aboutsummaryrefslogtreecommitdiffstats
path: root/fs
diff options
context:
space:
mode:
authorLinus Torvalds <torvalds@linux-foundation.org>2008-04-29 23:34:49 -0400
committerLinus Torvalds <torvalds@linux-foundation.org>2008-04-29 23:34:49 -0400
commitc4755d16fce9a4cdbb316c17657444856821bd4b (patch)
tree56060ddcdb8e1e3ddcc7970a2324cadb3e8e4e27 /fs
parentc15a2434ed4868cad99278ac4d4ae4de9de62e02 (diff)
parentf1fa3342e271029f93d323ca664809b94594fe04 (diff)
Merge branch 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tytso/ext4
* 'for_linus' of git://git.kernel.org/pub/scm/linux/kernel/git/tytso/ext4: (48 commits) ext4: fix hot spins in mballoc after err_freebuddy and err_freemeta ext4: fix test ext_generic_write_end() copied return value ext3: fix test ext_generic_write_end() copied return value ext4: Move mballoc headers/structures to a seperate header file mballoc.h ext4: cleanup for compiling mballoc with verification and debugging #defines ext4: don't use ext4_error in ext4_check_descriptors ext4: mark inode dirty after initializing the extent tree ext4: update ctime and mtime for truncate with extents. ext4: Don't do GFP_NOFS allocations after taking ext4_lock_group ext4: move headers out of include/linux ext4: fix wrong gfp type under transaction ext4: Fix hang on umount with quotas when journal is aborted ext4: Fix update of mtime and ctime on rename jdb2: replace remaining __FUNCTION__ occurrences ext4: replace remaining __FUNCTION__ occurrences jbd2: only create debugfs and stats entries if init is successful jbd2: fix kernel-doc notation jbd2: replace potentially false assertion with if block jbd2: eliminate duplicated code in revocation table init/destroy functions jbd2: tidy up revoke cache initialisation and destruction ...
Diffstat (limited to 'fs')
-rw-r--r--fs/ext3/inode.c14
-rw-r--r--fs/ext4/acl.c12
-rw-r--r--fs/ext4/balloc.c33
-rw-r--r--fs/ext4/bitmap.c2
-rw-r--r--fs/ext4/dir.c4
-rw-r--r--fs/ext4/ext4.h1205
-rw-r--r--fs/ext4/ext4_extents.h232
-rw-r--r--fs/ext4/ext4_i.h167
-rw-r--r--fs/ext4/ext4_jbd2.c14
-rw-r--r--fs/ext4/ext4_jbd2.h231
-rw-r--r--fs/ext4/ext4_sb.h148
-rw-r--r--fs/ext4/extents.c354
-rw-r--r--fs/ext4/file.c6
-rw-r--r--fs/ext4/fsync.c7
-rw-r--r--fs/ext4/hash.c2
-rw-r--r--fs/ext4/ialloc.c44
-rw-r--r--fs/ext4/inode.c57
-rw-r--r--fs/ext4/ioctl.c16
-rw-r--r--fs/ext4/mballoc.c437
-rw-r--r--fs/ext4/mballoc.h304
-rw-r--r--fs/ext4/migrate.c43
-rw-r--r--fs/ext4/namei.c44
-rw-r--r--fs/ext4/resize.c83
-rw-r--r--fs/ext4/super.c66
-rw-r--r--fs/ext4/symlink.c2
-rw-r--r--fs/ext4/xattr.c40
-rw-r--r--fs/ext4/xattr.h7
-rw-r--r--fs/ext4/xattr_security.c4
-rw-r--r--fs/ext4/xattr_trusted.c4
-rw-r--r--fs/ext4/xattr_user.c4
-rw-r--r--fs/jbd2/commit.c19
-rw-r--r--fs/jbd2/journal.c38
-rw-r--r--fs/jbd2/revoke.c165
-rw-r--r--fs/jbd2/transaction.c41
34 files changed, 3039 insertions, 810 deletions
diff --git a/fs/ext3/inode.c b/fs/ext3/inode.c
index cc47b76091bf..6ae4ecf3ce40 100644
--- a/fs/ext3/inode.c
+++ b/fs/ext3/inode.c
@@ -1261,10 +1261,11 @@ static int ext3_ordered_write_end(struct file *file,
1261 new_i_size = pos + copied; 1261 new_i_size = pos + copied;
1262 if (new_i_size > EXT3_I(inode)->i_disksize) 1262 if (new_i_size > EXT3_I(inode)->i_disksize)
1263 EXT3_I(inode)->i_disksize = new_i_size; 1263 EXT3_I(inode)->i_disksize = new_i_size;
1264 copied = ext3_generic_write_end(file, mapping, pos, len, copied, 1264 ret2 = ext3_generic_write_end(file, mapping, pos, len, copied,
1265 page, fsdata); 1265 page, fsdata);
1266 if (copied < 0) 1266 copied = ret2;
1267 ret = copied; 1267 if (ret2 < 0)
1268 ret = ret2;
1268 } 1269 }
1269 ret2 = ext3_journal_stop(handle); 1270 ret2 = ext3_journal_stop(handle);
1270 if (!ret) 1271 if (!ret)
@@ -1289,10 +1290,11 @@ static int ext3_writeback_write_end(struct file *file,
1289 if (new_i_size > EXT3_I(inode)->i_disksize) 1290 if (new_i_size > EXT3_I(inode)->i_disksize)
1290 EXT3_I(inode)->i_disksize = new_i_size; 1291 EXT3_I(inode)->i_disksize = new_i_size;
1291 1292
1292 copied = ext3_generic_write_end(file, mapping, pos, len, copied, 1293 ret2 = ext3_generic_write_end(file, mapping, pos, len, copied,
1293 page, fsdata); 1294 page, fsdata);
1294 if (copied < 0) 1295 copied = ret2;
1295 ret = copied; 1296 if (ret2 < 0)
1297 ret = ret2;
1296 1298
1297 ret2 = ext3_journal_stop(handle); 1299 ret2 = ext3_journal_stop(handle);
1298 if (!ret) 1300 if (!ret)
diff --git a/fs/ext4/acl.c b/fs/ext4/acl.c
index a8bae8cd1d5d..3c8dab880d91 100644
--- a/fs/ext4/acl.c
+++ b/fs/ext4/acl.c
@@ -9,8 +9,8 @@
9#include <linux/slab.h> 9#include <linux/slab.h>
10#include <linux/capability.h> 10#include <linux/capability.h>
11#include <linux/fs.h> 11#include <linux/fs.h>
12#include <linux/ext4_jbd2.h> 12#include "ext4_jbd2.h"
13#include <linux/ext4_fs.h> 13#include "ext4.h"
14#include "xattr.h" 14#include "xattr.h"
15#include "acl.h" 15#include "acl.h"
16 16
@@ -37,7 +37,7 @@ ext4_acl_from_disk(const void *value, size_t size)
37 return ERR_PTR(-EINVAL); 37 return ERR_PTR(-EINVAL);
38 if (count == 0) 38 if (count == 0)
39 return NULL; 39 return NULL;
40 acl = posix_acl_alloc(count, GFP_KERNEL); 40 acl = posix_acl_alloc(count, GFP_NOFS);
41 if (!acl) 41 if (!acl)
42 return ERR_PTR(-ENOMEM); 42 return ERR_PTR(-ENOMEM);
43 for (n=0; n < count; n++) { 43 for (n=0; n < count; n++) {
@@ -91,7 +91,7 @@ ext4_acl_to_disk(const struct posix_acl *acl, size_t *size)
91 91
92 *size = ext4_acl_size(acl->a_count); 92 *size = ext4_acl_size(acl->a_count);
93 ext_acl = kmalloc(sizeof(ext4_acl_header) + acl->a_count * 93 ext_acl = kmalloc(sizeof(ext4_acl_header) + acl->a_count *
94 sizeof(ext4_acl_entry), GFP_KERNEL); 94 sizeof(ext4_acl_entry), GFP_NOFS);
95 if (!ext_acl) 95 if (!ext_acl)
96 return ERR_PTR(-ENOMEM); 96 return ERR_PTR(-ENOMEM);
97 ext_acl->a_version = cpu_to_le32(EXT4_ACL_VERSION); 97 ext_acl->a_version = cpu_to_le32(EXT4_ACL_VERSION);
@@ -187,7 +187,7 @@ ext4_get_acl(struct inode *inode, int type)
187 } 187 }
188 retval = ext4_xattr_get(inode, name_index, "", NULL, 0); 188 retval = ext4_xattr_get(inode, name_index, "", NULL, 0);
189 if (retval > 0) { 189 if (retval > 0) {
190 value = kmalloc(retval, GFP_KERNEL); 190 value = kmalloc(retval, GFP_NOFS);
191 if (!value) 191 if (!value)
192 return ERR_PTR(-ENOMEM); 192 return ERR_PTR(-ENOMEM);
193 retval = ext4_xattr_get(inode, name_index, "", value, retval); 193 retval = ext4_xattr_get(inode, name_index, "", value, retval);
@@ -335,7 +335,7 @@ ext4_init_acl(handle_t *handle, struct inode *inode, struct inode *dir)
335 if (error) 335 if (error)
336 goto cleanup; 336 goto cleanup;
337 } 337 }
338 clone = posix_acl_clone(acl, GFP_KERNEL); 338 clone = posix_acl_clone(acl, GFP_NOFS);
339 error = -ENOMEM; 339 error = -ENOMEM;
340 if (!clone) 340 if (!clone)
341 goto cleanup; 341 goto cleanup;
diff --git a/fs/ext4/balloc.c b/fs/ext4/balloc.c
index 0737e05ba3dd..da994374ec3b 100644
--- a/fs/ext4/balloc.c
+++ b/fs/ext4/balloc.c
@@ -15,12 +15,12 @@
15#include <linux/capability.h> 15#include <linux/capability.h>
16#include <linux/fs.h> 16#include <linux/fs.h>
17#include <linux/jbd2.h> 17#include <linux/jbd2.h>
18#include <linux/ext4_fs.h>
19#include <linux/ext4_jbd2.h>
20#include <linux/quotaops.h> 18#include <linux/quotaops.h>
21#include <linux/buffer_head.h> 19#include <linux/buffer_head.h>
22 20#include "ext4.h"
21#include "ext4_jbd2.h"
23#include "group.h" 22#include "group.h"
23
24/* 24/*
25 * balloc.c contains the blocks allocation and deallocation routines 25 * balloc.c contains the blocks allocation and deallocation routines
26 */ 26 */
@@ -48,7 +48,6 @@ void ext4_get_group_no_and_offset(struct super_block *sb, ext4_fsblk_t blocknr,
48unsigned ext4_init_block_bitmap(struct super_block *sb, struct buffer_head *bh, 48unsigned ext4_init_block_bitmap(struct super_block *sb, struct buffer_head *bh,
49 ext4_group_t block_group, struct ext4_group_desc *gdp) 49 ext4_group_t block_group, struct ext4_group_desc *gdp)
50{ 50{
51 unsigned long start;
52 int bit, bit_max; 51 int bit, bit_max;
53 unsigned free_blocks, group_blocks; 52 unsigned free_blocks, group_blocks;
54 struct ext4_sb_info *sbi = EXT4_SB(sb); 53 struct ext4_sb_info *sbi = EXT4_SB(sb);
@@ -59,7 +58,7 @@ unsigned ext4_init_block_bitmap(struct super_block *sb, struct buffer_head *bh,
59 /* If checksum is bad mark all blocks used to prevent allocation 58 /* If checksum is bad mark all blocks used to prevent allocation
60 * essentially implementing a per-group read-only flag. */ 59 * essentially implementing a per-group read-only flag. */
61 if (!ext4_group_desc_csum_verify(sbi, block_group, gdp)) { 60 if (!ext4_group_desc_csum_verify(sbi, block_group, gdp)) {
62 ext4_error(sb, __FUNCTION__, 61 ext4_error(sb, __func__,
63 "Checksum bad for group %lu\n", block_group); 62 "Checksum bad for group %lu\n", block_group);
64 gdp->bg_free_blocks_count = 0; 63 gdp->bg_free_blocks_count = 0;
65 gdp->bg_free_inodes_count = 0; 64 gdp->bg_free_inodes_count = 0;
@@ -106,11 +105,12 @@ unsigned ext4_init_block_bitmap(struct super_block *sb, struct buffer_head *bh,
106 free_blocks = group_blocks - bit_max; 105 free_blocks = group_blocks - bit_max;
107 106
108 if (bh) { 107 if (bh) {
108 ext4_fsblk_t start;
109
109 for (bit = 0; bit < bit_max; bit++) 110 for (bit = 0; bit < bit_max; bit++)
110 ext4_set_bit(bit, bh->b_data); 111 ext4_set_bit(bit, bh->b_data);
111 112
112 start = block_group * EXT4_BLOCKS_PER_GROUP(sb) + 113 start = ext4_group_first_block_no(sb, block_group);
113 le32_to_cpu(sbi->s_es->s_first_data_block);
114 114
115 /* Set bits for block and inode bitmaps, and inode table */ 115 /* Set bits for block and inode bitmaps, and inode table */
116 ext4_set_bit(ext4_block_bitmap(sb, gdp) - start, bh->b_data); 116 ext4_set_bit(ext4_block_bitmap(sb, gdp) - start, bh->b_data);
@@ -235,7 +235,7 @@ static int ext4_valid_block_bitmap(struct super_block *sb,
235 return 1; 235 return 1;
236 236
237err_out: 237err_out:
238 ext4_error(sb, __FUNCTION__, 238 ext4_error(sb, __func__,
239 "Invalid block bitmap - " 239 "Invalid block bitmap - "
240 "block_group = %d, block = %llu", 240 "block_group = %d, block = %llu",
241 block_group, bitmap_blk); 241 block_group, bitmap_blk);
@@ -264,7 +264,7 @@ read_block_bitmap(struct super_block *sb, ext4_group_t block_group)
264 bitmap_blk = ext4_block_bitmap(sb, desc); 264 bitmap_blk = ext4_block_bitmap(sb, desc);
265 bh = sb_getblk(sb, bitmap_blk); 265 bh = sb_getblk(sb, bitmap_blk);
266 if (unlikely(!bh)) { 266 if (unlikely(!bh)) {
267 ext4_error(sb, __FUNCTION__, 267 ext4_error(sb, __func__,
268 "Cannot read block bitmap - " 268 "Cannot read block bitmap - "
269 "block_group = %d, block_bitmap = %llu", 269 "block_group = %d, block_bitmap = %llu",
270 (int)block_group, (unsigned long long)bitmap_blk); 270 (int)block_group, (unsigned long long)bitmap_blk);
@@ -281,7 +281,7 @@ read_block_bitmap(struct super_block *sb, ext4_group_t block_group)
281 } 281 }
282 if (bh_submit_read(bh) < 0) { 282 if (bh_submit_read(bh) < 0) {
283 put_bh(bh); 283 put_bh(bh);
284 ext4_error(sb, __FUNCTION__, 284 ext4_error(sb, __func__,
285 "Cannot read block bitmap - " 285 "Cannot read block bitmap - "
286 "block_group = %d, block_bitmap = %llu", 286 "block_group = %d, block_bitmap = %llu",
287 (int)block_group, (unsigned long long)bitmap_blk); 287 (int)block_group, (unsigned long long)bitmap_blk);
@@ -360,7 +360,7 @@ restart:
360 BUG(); 360 BUG();
361} 361}
362#define rsv_window_dump(root, verbose) \ 362#define rsv_window_dump(root, verbose) \
363 __rsv_window_dump((root), (verbose), __FUNCTION__) 363 __rsv_window_dump((root), (verbose), __func__)
364#else 364#else
365#define rsv_window_dump(root, verbose) do {} while (0) 365#define rsv_window_dump(root, verbose) do {} while (0)
366#endif 366#endif
@@ -740,7 +740,7 @@ do_more:
740 if (!ext4_clear_bit_atomic(sb_bgl_lock(sbi, block_group), 740 if (!ext4_clear_bit_atomic(sb_bgl_lock(sbi, block_group),
741 bit + i, bitmap_bh->b_data)) { 741 bit + i, bitmap_bh->b_data)) {
742 jbd_unlock_bh_state(bitmap_bh); 742 jbd_unlock_bh_state(bitmap_bh);
743 ext4_error(sb, __FUNCTION__, 743 ext4_error(sb, __func__,
744 "bit already cleared for block %llu", 744 "bit already cleared for block %llu",
745 (ext4_fsblk_t)(block + i)); 745 (ext4_fsblk_t)(block + i));
746 jbd_lock_bh_state(bitmap_bh); 746 jbd_lock_bh_state(bitmap_bh);
@@ -752,9 +752,7 @@ do_more:
752 jbd_unlock_bh_state(bitmap_bh); 752 jbd_unlock_bh_state(bitmap_bh);
753 753
754 spin_lock(sb_bgl_lock(sbi, block_group)); 754 spin_lock(sb_bgl_lock(sbi, block_group));
755 desc->bg_free_blocks_count = 755 le16_add_cpu(&desc->bg_free_blocks_count, group_freed);
756 cpu_to_le16(le16_to_cpu(desc->bg_free_blocks_count) +
757 group_freed);
758 desc->bg_checksum = ext4_group_desc_csum(sbi, block_group, desc); 756 desc->bg_checksum = ext4_group_desc_csum(sbi, block_group, desc);
759 spin_unlock(sb_bgl_lock(sbi, block_group)); 757 spin_unlock(sb_bgl_lock(sbi, block_group));
760 percpu_counter_add(&sbi->s_freeblocks_counter, count); 758 percpu_counter_add(&sbi->s_freeblocks_counter, count);
@@ -1798,7 +1796,7 @@ allocated:
1798 if (ext4_test_bit(grp_alloc_blk+i, 1796 if (ext4_test_bit(grp_alloc_blk+i,
1799 bh2jh(bitmap_bh)->b_committed_data)) { 1797 bh2jh(bitmap_bh)->b_committed_data)) {
1800 printk("%s: block was unexpectedly set in " 1798 printk("%s: block was unexpectedly set in "
1801 "b_committed_data\n", __FUNCTION__); 1799 "b_committed_data\n", __func__);
1802 } 1800 }
1803 } 1801 }
1804 } 1802 }
@@ -1823,8 +1821,7 @@ allocated:
1823 spin_lock(sb_bgl_lock(sbi, group_no)); 1821 spin_lock(sb_bgl_lock(sbi, group_no));
1824 if (gdp->bg_flags & cpu_to_le16(EXT4_BG_BLOCK_UNINIT)) 1822 if (gdp->bg_flags & cpu_to_le16(EXT4_BG_BLOCK_UNINIT))
1825 gdp->bg_flags &= cpu_to_le16(~EXT4_BG_BLOCK_UNINIT); 1823 gdp->bg_flags &= cpu_to_le16(~EXT4_BG_BLOCK_UNINIT);
1826 gdp->bg_free_blocks_count = 1824 le16_add_cpu(&gdp->bg_free_blocks_count, -num);
1827 cpu_to_le16(le16_to_cpu(gdp->bg_free_blocks_count)-num);
1828 gdp->bg_checksum = ext4_group_desc_csum(sbi, group_no, gdp); 1825 gdp->bg_checksum = ext4_group_desc_csum(sbi, group_no, gdp);
1829 spin_unlock(sb_bgl_lock(sbi, group_no)); 1826 spin_unlock(sb_bgl_lock(sbi, group_no));
1830 percpu_counter_sub(&sbi->s_freeblocks_counter, num); 1827 percpu_counter_sub(&sbi->s_freeblocks_counter, num);
diff --git a/fs/ext4/bitmap.c b/fs/ext4/bitmap.c
index 420554f8f79d..d37ea6750454 100644
--- a/fs/ext4/bitmap.c
+++ b/fs/ext4/bitmap.c
@@ -9,7 +9,7 @@
9 9
10#include <linux/buffer_head.h> 10#include <linux/buffer_head.h>
11#include <linux/jbd2.h> 11#include <linux/jbd2.h>
12#include <linux/ext4_fs.h> 12#include "ext4.h"
13 13
14#ifdef EXT4FS_DEBUG 14#ifdef EXT4FS_DEBUG
15 15
diff --git a/fs/ext4/dir.c b/fs/ext4/dir.c
index 2c23bade9aa6..2bf0331ea194 100644
--- a/fs/ext4/dir.c
+++ b/fs/ext4/dir.c
@@ -23,10 +23,10 @@
23 23
24#include <linux/fs.h> 24#include <linux/fs.h>
25#include <linux/jbd2.h> 25#include <linux/jbd2.h>
26#include <linux/ext4_fs.h>
27#include <linux/buffer_head.h> 26#include <linux/buffer_head.h>
28#include <linux/slab.h> 27#include <linux/slab.h>
29#include <linux/rbtree.h> 28#include <linux/rbtree.h>
29#include "ext4.h"
30 30
31static unsigned char ext4_filetype_table[] = { 31static unsigned char ext4_filetype_table[] = {
32 DT_UNKNOWN, DT_REG, DT_DIR, DT_CHR, DT_BLK, DT_FIFO, DT_SOCK, DT_LNK 32 DT_UNKNOWN, DT_REG, DT_DIR, DT_CHR, DT_BLK, DT_FIFO, DT_SOCK, DT_LNK
@@ -42,7 +42,7 @@ const struct file_operations ext4_dir_operations = {
42 .llseek = generic_file_llseek, 42 .llseek = generic_file_llseek,
43 .read = generic_read_dir, 43 .read = generic_read_dir,
44 .readdir = ext4_readdir, /* we take BKL. needed?*/ 44 .readdir = ext4_readdir, /* we take BKL. needed?*/
45 .ioctl = ext4_ioctl, /* BKL held */ 45 .unlocked_ioctl = ext4_ioctl,
46#ifdef CONFIG_COMPAT 46#ifdef CONFIG_COMPAT
47 .compat_ioctl = ext4_compat_ioctl, 47 .compat_ioctl = ext4_compat_ioctl,
48#endif 48#endif
diff --git a/fs/ext4/ext4.h b/fs/ext4/ext4.h
new file mode 100644
index 000000000000..8158083f7ac0
--- /dev/null
+++ b/fs/ext4/ext4.h
@@ -0,0 +1,1205 @@
1/*
2 * ext4.h
3 *
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
8 *
9 * from
10 *
11 * linux/include/linux/minix_fs.h
12 *
13 * Copyright (C) 1991, 1992 Linus Torvalds
14 */
15
16#ifndef _EXT4_H
17#define _EXT4_H
18
19#include <linux/types.h>
20#include <linux/blkdev.h>
21#include <linux/magic.h>
22#include "ext4_i.h"
23
24/*
25 * The second extended filesystem constants/structures
26 */
27
28/*
29 * Define EXT4FS_DEBUG to produce debug messages
30 */
31#undef EXT4FS_DEBUG
32
33/*
34 * Define EXT4_RESERVATION to reserve data blocks for expanding files
35 */
36#define EXT4_DEFAULT_RESERVE_BLOCKS 8
37/*max window size: 1024(direct blocks) + 3([t,d]indirect blocks) */
38#define EXT4_MAX_RESERVE_BLOCKS 1027
39#define EXT4_RESERVE_WINDOW_NOT_ALLOCATED 0
40
41/*
42 * Debug code
43 */
44#ifdef EXT4FS_DEBUG
45#define ext4_debug(f, a...) \
46 do { \
47 printk (KERN_DEBUG "EXT4-fs DEBUG (%s, %d): %s:", \
48 __FILE__, __LINE__, __FUNCTION__); \
49 printk (KERN_DEBUG f, ## a); \
50 } while (0)
51#else
52#define ext4_debug(f, a...) do {} while (0)
53#endif
54
55#define EXT4_MULTIBLOCK_ALLOCATOR 1
56
57/* prefer goal again. length */
58#define EXT4_MB_HINT_MERGE 1
59/* blocks already reserved */
60#define EXT4_MB_HINT_RESERVED 2
61/* metadata is being allocated */
62#define EXT4_MB_HINT_METADATA 4
63/* first blocks in the file */
64#define EXT4_MB_HINT_FIRST 8
65/* search for the best chunk */
66#define EXT4_MB_HINT_BEST 16
67/* data is being allocated */
68#define EXT4_MB_HINT_DATA 32
69/* don't preallocate (for tails) */
70#define EXT4_MB_HINT_NOPREALLOC 64
71/* allocate for locality group */
72#define EXT4_MB_HINT_GROUP_ALLOC 128
73/* allocate goal blocks or none */
74#define EXT4_MB_HINT_GOAL_ONLY 256
75/* goal is meaningful */
76#define EXT4_MB_HINT_TRY_GOAL 512
77
78struct ext4_allocation_request {
79 /* target inode for block we're allocating */
80 struct inode *inode;
81 /* logical block in target inode */
82 ext4_lblk_t logical;
83 /* phys. target (a hint) */
84 ext4_fsblk_t goal;
85 /* the closest logical allocated block to the left */
86 ext4_lblk_t lleft;
87 /* phys. block for ^^^ */
88 ext4_fsblk_t pleft;
89 /* the closest logical allocated block to the right */
90 ext4_lblk_t lright;
91 /* phys. block for ^^^ */
92 ext4_fsblk_t pright;
93 /* how many blocks we want to allocate */
94 unsigned long len;
95 /* flags. see above EXT4_MB_HINT_* */
96 unsigned long flags;
97};
98
99/*
100 * Special inodes numbers
101 */
102#define EXT4_BAD_INO 1 /* Bad blocks inode */
103#define EXT4_ROOT_INO 2 /* Root inode */
104#define EXT4_BOOT_LOADER_INO 5 /* Boot loader inode */
105#define EXT4_UNDEL_DIR_INO 6 /* Undelete directory inode */
106#define EXT4_RESIZE_INO 7 /* Reserved group descriptors inode */
107#define EXT4_JOURNAL_INO 8 /* Journal inode */
108
109/* First non-reserved inode for old ext4 filesystems */
110#define EXT4_GOOD_OLD_FIRST_INO 11
111
112/*
113 * Maximal count of links to a file
114 */
115#define EXT4_LINK_MAX 65000
116
117/*
118 * Macro-instructions used to manage several block sizes
119 */
120#define EXT4_MIN_BLOCK_SIZE 1024
121#define EXT4_MAX_BLOCK_SIZE 65536
122#define EXT4_MIN_BLOCK_LOG_SIZE 10
123#ifdef __KERNEL__
124# define EXT4_BLOCK_SIZE(s) ((s)->s_blocksize)
125#else
126# define EXT4_BLOCK_SIZE(s) (EXT4_MIN_BLOCK_SIZE << (s)->s_log_block_size)
127#endif
128#define EXT4_ADDR_PER_BLOCK(s) (EXT4_BLOCK_SIZE(s) / sizeof (__u32))
129#ifdef __KERNEL__
130# define EXT4_BLOCK_SIZE_BITS(s) ((s)->s_blocksize_bits)
131#else
132# define EXT4_BLOCK_SIZE_BITS(s) ((s)->s_log_block_size + 10)
133#endif
134#ifdef __KERNEL__
135#define EXT4_ADDR_PER_BLOCK_BITS(s) (EXT4_SB(s)->s_addr_per_block_bits)
136#define EXT4_INODE_SIZE(s) (EXT4_SB(s)->s_inode_size)
137#define EXT4_FIRST_INO(s) (EXT4_SB(s)->s_first_ino)
138#else
139#define EXT4_INODE_SIZE(s) (((s)->s_rev_level == EXT4_GOOD_OLD_REV) ? \
140 EXT4_GOOD_OLD_INODE_SIZE : \
141 (s)->s_inode_size)
142#define EXT4_FIRST_INO(s) (((s)->s_rev_level == EXT4_GOOD_OLD_REV) ? \
143 EXT4_GOOD_OLD_FIRST_INO : \
144 (s)->s_first_ino)
145#endif
146#define EXT4_BLOCK_ALIGN(size, blkbits) ALIGN((size), (1 << (blkbits)))
147
148/*
149 * Structure of a blocks group descriptor
150 */
151struct ext4_group_desc
152{
153 __le32 bg_block_bitmap_lo; /* Blocks bitmap block */
154 __le32 bg_inode_bitmap_lo; /* Inodes bitmap block */
155 __le32 bg_inode_table_lo; /* Inodes table block */
156 __le16 bg_free_blocks_count; /* Free blocks count */
157 __le16 bg_free_inodes_count; /* Free inodes count */
158 __le16 bg_used_dirs_count; /* Directories count */
159 __le16 bg_flags; /* EXT4_BG_flags (INODE_UNINIT, etc) */
160 __u32 bg_reserved[2]; /* Likely block/inode bitmap checksum */
161 __le16 bg_itable_unused; /* Unused inodes count */
162 __le16 bg_checksum; /* crc16(sb_uuid+group+desc) */
163 __le32 bg_block_bitmap_hi; /* Blocks bitmap block MSB */
164 __le32 bg_inode_bitmap_hi; /* Inodes bitmap block MSB */
165 __le32 bg_inode_table_hi; /* Inodes table block MSB */
166 __le16 bg_free_blocks_count_hi;/* Free blocks count MSB */
167 __le16 bg_free_inodes_count_hi;/* Free inodes count MSB */
168 __le16 bg_used_dirs_count_hi; /* Directories count MSB */
169 __le16 bg_itable_unused_hi; /* Unused inodes count MSB */
170 __u32 bg_reserved2[3];
171};
172
173#define EXT4_BG_INODE_UNINIT 0x0001 /* Inode table/bitmap not in use */
174#define EXT4_BG_BLOCK_UNINIT 0x0002 /* Block bitmap not in use */
175#define EXT4_BG_INODE_ZEROED 0x0004 /* On-disk itable initialized to zero */
176
177#ifdef __KERNEL__
178#include "ext4_sb.h"
179#endif
180/*
181 * Macro-instructions used to manage group descriptors
182 */
183#define EXT4_MIN_DESC_SIZE 32
184#define EXT4_MIN_DESC_SIZE_64BIT 64
185#define EXT4_MAX_DESC_SIZE EXT4_MIN_BLOCK_SIZE
186#define EXT4_DESC_SIZE(s) (EXT4_SB(s)->s_desc_size)
187#ifdef __KERNEL__
188# define EXT4_BLOCKS_PER_GROUP(s) (EXT4_SB(s)->s_blocks_per_group)
189# define EXT4_DESC_PER_BLOCK(s) (EXT4_SB(s)->s_desc_per_block)
190# define EXT4_INODES_PER_GROUP(s) (EXT4_SB(s)->s_inodes_per_group)
191# define EXT4_DESC_PER_BLOCK_BITS(s) (EXT4_SB(s)->s_desc_per_block_bits)
192#else
193# define EXT4_BLOCKS_PER_GROUP(s) ((s)->s_blocks_per_group)
194# define EXT4_DESC_PER_BLOCK(s) (EXT4_BLOCK_SIZE(s) / EXT4_DESC_SIZE(s))
195# define EXT4_INODES_PER_GROUP(s) ((s)->s_inodes_per_group)
196#endif
197
198/*
199 * Constants relative to the data blocks
200 */
201#define EXT4_NDIR_BLOCKS 12
202#define EXT4_IND_BLOCK EXT4_NDIR_BLOCKS
203#define EXT4_DIND_BLOCK (EXT4_IND_BLOCK + 1)
204#define EXT4_TIND_BLOCK (EXT4_DIND_BLOCK + 1)
205#define EXT4_N_BLOCKS (EXT4_TIND_BLOCK + 1)
206
207/*
208 * Inode flags
209 */
210#define EXT4_SECRM_FL 0x00000001 /* Secure deletion */
211#define EXT4_UNRM_FL 0x00000002 /* Undelete */
212#define EXT4_COMPR_FL 0x00000004 /* Compress file */
213#define EXT4_SYNC_FL 0x00000008 /* Synchronous updates */
214#define EXT4_IMMUTABLE_FL 0x00000010 /* Immutable file */
215#define EXT4_APPEND_FL 0x00000020 /* writes to file may only append */
216#define EXT4_NODUMP_FL 0x00000040 /* do not dump file */
217#define EXT4_NOATIME_FL 0x00000080 /* do not update atime */
218/* Reserved for compression usage... */
219#define EXT4_DIRTY_FL 0x00000100
220#define EXT4_COMPRBLK_FL 0x00000200 /* One or more compressed clusters */
221#define EXT4_NOCOMPR_FL 0x00000400 /* Don't compress */
222#define EXT4_ECOMPR_FL 0x00000800 /* Compression error */
223/* End compression flags --- maybe not all used */
224#define EXT4_INDEX_FL 0x00001000 /* hash-indexed directory */
225#define EXT4_IMAGIC_FL 0x00002000 /* AFS directory */
226#define EXT4_JOURNAL_DATA_FL 0x00004000 /* file data should be journaled */
227#define EXT4_NOTAIL_FL 0x00008000 /* file tail should not be merged */
228#define EXT4_DIRSYNC_FL 0x00010000 /* dirsync behaviour (directories only) */
229#define EXT4_TOPDIR_FL 0x00020000 /* Top of directory hierarchies*/
230#define EXT4_HUGE_FILE_FL 0x00040000 /* Set to each huge file */
231#define EXT4_EXTENTS_FL 0x00080000 /* Inode uses extents */
232#define EXT4_EXT_MIGRATE 0x00100000 /* Inode is migrating */
233#define EXT4_RESERVED_FL 0x80000000 /* reserved for ext4 lib */
234
235#define EXT4_FL_USER_VISIBLE 0x000BDFFF /* User visible flags */
236#define EXT4_FL_USER_MODIFIABLE 0x000380FF /* User modifiable flags */
237
238/*
239 * Inode dynamic state flags
240 */
241#define EXT4_STATE_JDATA 0x00000001 /* journaled data exists */
242#define EXT4_STATE_NEW 0x00000002 /* inode is newly created */
243#define EXT4_STATE_XATTR 0x00000004 /* has in-inode xattrs */
244#define EXT4_STATE_NO_EXPAND 0x00000008 /* No space for expansion */
245
246/* Used to pass group descriptor data when online resize is done */
247struct ext4_new_group_input {
248 __u32 group; /* Group number for this data */
249 __u64 block_bitmap; /* Absolute block number of block bitmap */
250 __u64 inode_bitmap; /* Absolute block number of inode bitmap */
251 __u64 inode_table; /* Absolute block number of inode table start */
252 __u32 blocks_count; /* Total number of blocks in this group */
253 __u16 reserved_blocks; /* Number of reserved blocks in this group */
254 __u16 unused;
255};
256
257/* The struct ext4_new_group_input in kernel space, with free_blocks_count */
258struct ext4_new_group_data {
259 __u32 group;
260 __u64 block_bitmap;
261 __u64 inode_bitmap;
262 __u64 inode_table;
263 __u32 blocks_count;
264 __u16 reserved_blocks;
265 __u16 unused;
266 __u32 free_blocks_count;
267};
268
269/*
270 * Following is used by preallocation code to tell get_blocks() that we
271 * want uninitialzed extents.
272 */
273#define EXT4_CREATE_UNINITIALIZED_EXT 2
274
275/*
276 * ioctl commands
277 */
278#define EXT4_IOC_GETFLAGS FS_IOC_GETFLAGS
279#define EXT4_IOC_SETFLAGS FS_IOC_SETFLAGS
280#define EXT4_IOC_GETVERSION _IOR('f', 3, long)
281#define EXT4_IOC_SETVERSION _IOW('f', 4, long)
282#define EXT4_IOC_GROUP_EXTEND _IOW('f', 7, unsigned long)
283#define EXT4_IOC_GROUP_ADD _IOW('f', 8,struct ext4_new_group_input)
284#define EXT4_IOC_GETVERSION_OLD FS_IOC_GETVERSION
285#define EXT4_IOC_SETVERSION_OLD FS_IOC_SETVERSION
286#ifdef CONFIG_JBD2_DEBUG
287#define EXT4_IOC_WAIT_FOR_READONLY _IOR('f', 99, long)
288#endif
289#define EXT4_IOC_GETRSVSZ _IOR('f', 5, long)
290#define EXT4_IOC_SETRSVSZ _IOW('f', 6, long)
291#define EXT4_IOC_MIGRATE _IO('f', 7)
292
293/*
294 * ioctl commands in 32 bit emulation
295 */
296#define EXT4_IOC32_GETFLAGS FS_IOC32_GETFLAGS
297#define EXT4_IOC32_SETFLAGS FS_IOC32_SETFLAGS
298#define EXT4_IOC32_GETVERSION _IOR('f', 3, int)
299#define EXT4_IOC32_SETVERSION _IOW('f', 4, int)
300#define EXT4_IOC32_GETRSVSZ _IOR('f', 5, int)
301#define EXT4_IOC32_SETRSVSZ _IOW('f', 6, int)
302#define EXT4_IOC32_GROUP_EXTEND _IOW('f', 7, unsigned int)
303#ifdef CONFIG_JBD2_DEBUG
304#define EXT4_IOC32_WAIT_FOR_READONLY _IOR('f', 99, int)
305#endif
306#define EXT4_IOC32_GETVERSION_OLD FS_IOC32_GETVERSION
307#define EXT4_IOC32_SETVERSION_OLD FS_IOC32_SETVERSION
308
309
310/*
311 * Mount options
312 */
313struct ext4_mount_options {
314 unsigned long s_mount_opt;
315 uid_t s_resuid;
316 gid_t s_resgid;
317 unsigned long s_commit_interval;
318#ifdef CONFIG_QUOTA
319 int s_jquota_fmt;
320 char *s_qf_names[MAXQUOTAS];
321#endif
322};
323
324/*
325 * Structure of an inode on the disk
326 */
327struct ext4_inode {
328 __le16 i_mode; /* File mode */
329 __le16 i_uid; /* Low 16 bits of Owner Uid */
330 __le32 i_size_lo; /* Size in bytes */
331 __le32 i_atime; /* Access time */
332 __le32 i_ctime; /* Inode Change time */
333 __le32 i_mtime; /* Modification time */
334 __le32 i_dtime; /* Deletion Time */
335 __le16 i_gid; /* Low 16 bits of Group Id */
336 __le16 i_links_count; /* Links count */
337 __le32 i_blocks_lo; /* Blocks count */
338 __le32 i_flags; /* File flags */
339 union {
340 struct {
341 __le32 l_i_version;
342 } linux1;
343 struct {
344 __u32 h_i_translator;
345 } hurd1;
346 struct {
347 __u32 m_i_reserved1;
348 } masix1;
349 } osd1; /* OS dependent 1 */
350 __le32 i_block[EXT4_N_BLOCKS];/* Pointers to blocks */
351 __le32 i_generation; /* File version (for NFS) */
352 __le32 i_file_acl_lo; /* File ACL */
353 __le32 i_size_high;
354 __le32 i_obso_faddr; /* Obsoleted fragment address */
355 union {
356 struct {
357 __le16 l_i_blocks_high; /* were l_i_reserved1 */
358 __le16 l_i_file_acl_high;
359 __le16 l_i_uid_high; /* these 2 fields */
360 __le16 l_i_gid_high; /* were reserved2[0] */
361 __u32 l_i_reserved2;
362 } linux2;
363 struct {
364 __le16 h_i_reserved1; /* Obsoleted fragment number/size which are removed in ext4 */
365 __u16 h_i_mode_high;
366 __u16 h_i_uid_high;
367 __u16 h_i_gid_high;
368 __u32 h_i_author;
369 } hurd2;
370 struct {
371 __le16 h_i_reserved1; /* Obsoleted fragment number/size which are removed in ext4 */
372 __le16 m_i_file_acl_high;
373 __u32 m_i_reserved2[2];
374 } masix2;
375 } osd2; /* OS dependent 2 */
376 __le16 i_extra_isize;
377 __le16 i_pad1;
378 __le32 i_ctime_extra; /* extra Change time (nsec << 2 | epoch) */
379 __le32 i_mtime_extra; /* extra Modification time(nsec << 2 | epoch) */
380 __le32 i_atime_extra; /* extra Access time (nsec << 2 | epoch) */
381 __le32 i_crtime; /* File Creation time */
382 __le32 i_crtime_extra; /* extra FileCreationtime (nsec << 2 | epoch) */
383 __le32 i_version_hi; /* high 32 bits for 64-bit version */
384};
385
386
387#define EXT4_EPOCH_BITS 2
388#define EXT4_EPOCH_MASK ((1 << EXT4_EPOCH_BITS) - 1)
389#define EXT4_NSEC_MASK (~0UL << EXT4_EPOCH_BITS)
390
391/*
392 * Extended fields will fit into an inode if the filesystem was formatted
393 * with large inodes (-I 256 or larger) and there are not currently any EAs
394 * consuming all of the available space. For new inodes we always reserve
395 * enough space for the kernel's known extended fields, but for inodes
396 * created with an old kernel this might not have been the case. None of
397 * the extended inode fields is critical for correct filesystem operation.
398 * This macro checks if a certain field fits in the inode. Note that
399 * inode-size = GOOD_OLD_INODE_SIZE + i_extra_isize
400 */
401#define EXT4_FITS_IN_INODE(ext4_inode, einode, field) \
402 ((offsetof(typeof(*ext4_inode), field) + \
403 sizeof((ext4_inode)->field)) \
404 <= (EXT4_GOOD_OLD_INODE_SIZE + \
405 (einode)->i_extra_isize)) \
406
407static inline __le32 ext4_encode_extra_time(struct timespec *time)
408{
409 return cpu_to_le32((sizeof(time->tv_sec) > 4 ?
410 time->tv_sec >> 32 : 0) |
411 ((time->tv_nsec << 2) & EXT4_NSEC_MASK));
412}
413
414static inline void ext4_decode_extra_time(struct timespec *time, __le32 extra)
415{
416 if (sizeof(time->tv_sec) > 4)
417 time->tv_sec |= (__u64)(le32_to_cpu(extra) & EXT4_EPOCH_MASK)
418 << 32;
419 time->tv_nsec = (le32_to_cpu(extra) & EXT4_NSEC_MASK) >> 2;
420}
421
422#define EXT4_INODE_SET_XTIME(xtime, inode, raw_inode) \
423do { \
424 (raw_inode)->xtime = cpu_to_le32((inode)->xtime.tv_sec); \
425 if (EXT4_FITS_IN_INODE(raw_inode, EXT4_I(inode), xtime ## _extra)) \
426 (raw_inode)->xtime ## _extra = \
427 ext4_encode_extra_time(&(inode)->xtime); \
428} while (0)
429
430#define EXT4_EINODE_SET_XTIME(xtime, einode, raw_inode) \
431do { \
432 if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime)) \
433 (raw_inode)->xtime = cpu_to_le32((einode)->xtime.tv_sec); \
434 if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime ## _extra)) \
435 (raw_inode)->xtime ## _extra = \
436 ext4_encode_extra_time(&(einode)->xtime); \
437} while (0)
438
439#define EXT4_INODE_GET_XTIME(xtime, inode, raw_inode) \
440do { \
441 (inode)->xtime.tv_sec = (signed)le32_to_cpu((raw_inode)->xtime); \
442 if (EXT4_FITS_IN_INODE(raw_inode, EXT4_I(inode), xtime ## _extra)) \
443 ext4_decode_extra_time(&(inode)->xtime, \
444 raw_inode->xtime ## _extra); \
445} while (0)
446
447#define EXT4_EINODE_GET_XTIME(xtime, einode, raw_inode) \
448do { \
449 if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime)) \
450 (einode)->xtime.tv_sec = \
451 (signed)le32_to_cpu((raw_inode)->xtime); \
452 if (EXT4_FITS_IN_INODE(raw_inode, einode, xtime ## _extra)) \
453 ext4_decode_extra_time(&(einode)->xtime, \
454 raw_inode->xtime ## _extra); \
455} while (0)
456
457#define i_disk_version osd1.linux1.l_i_version
458
459#if defined(__KERNEL__) || defined(__linux__)
460#define i_reserved1 osd1.linux1.l_i_reserved1
461#define i_file_acl_high osd2.linux2.l_i_file_acl_high
462#define i_blocks_high osd2.linux2.l_i_blocks_high
463#define i_uid_low i_uid
464#define i_gid_low i_gid
465#define i_uid_high osd2.linux2.l_i_uid_high
466#define i_gid_high osd2.linux2.l_i_gid_high
467#define i_reserved2 osd2.linux2.l_i_reserved2
468
469#elif defined(__GNU__)
470
471#define i_translator osd1.hurd1.h_i_translator
472#define i_uid_high osd2.hurd2.h_i_uid_high
473#define i_gid_high osd2.hurd2.h_i_gid_high
474#define i_author osd2.hurd2.h_i_author
475
476#elif defined(__masix__)
477
478#define i_reserved1 osd1.masix1.m_i_reserved1
479#define i_file_acl_high osd2.masix2.m_i_file_acl_high
480#define i_reserved2 osd2.masix2.m_i_reserved2
481
482#endif /* defined(__KERNEL__) || defined(__linux__) */
483
484/*
485 * File system states
486 */
487#define EXT4_VALID_FS 0x0001 /* Unmounted cleanly */
488#define EXT4_ERROR_FS 0x0002 /* Errors detected */
489#define EXT4_ORPHAN_FS 0x0004 /* Orphans being recovered */
490
491/*
492 * Misc. filesystem flags
493 */
494#define EXT2_FLAGS_SIGNED_HASH 0x0001 /* Signed dirhash in use */
495#define EXT2_FLAGS_UNSIGNED_HASH 0x0002 /* Unsigned dirhash in use */
496#define EXT2_FLAGS_TEST_FILESYS 0x0004 /* to test development code */
497
498/*
499 * Mount flags
500 */
501#define EXT4_MOUNT_CHECK 0x00001 /* Do mount-time checks */
502#define EXT4_MOUNT_OLDALLOC 0x00002 /* Don't use the new Orlov allocator */
503#define EXT4_MOUNT_GRPID 0x00004 /* Create files with directory's group */
504#define EXT4_MOUNT_DEBUG 0x00008 /* Some debugging messages */
505#define EXT4_MOUNT_ERRORS_CONT 0x00010 /* Continue on errors */
506#define EXT4_MOUNT_ERRORS_RO 0x00020 /* Remount fs ro on errors */
507#define EXT4_MOUNT_ERRORS_PANIC 0x00040 /* Panic on errors */
508#define EXT4_MOUNT_MINIX_DF 0x00080 /* Mimics the Minix statfs */
509#define EXT4_MOUNT_NOLOAD 0x00100 /* Don't use existing journal*/
510#define EXT4_MOUNT_ABORT 0x00200 /* Fatal error detected */
511#define EXT4_MOUNT_DATA_FLAGS 0x00C00 /* Mode for data writes: */
512#define EXT4_MOUNT_JOURNAL_DATA 0x00400 /* Write data to journal */
513#define EXT4_MOUNT_ORDERED_DATA 0x00800 /* Flush data before commit */
514#define EXT4_MOUNT_WRITEBACK_DATA 0x00C00 /* No data ordering */
515#define EXT4_MOUNT_UPDATE_JOURNAL 0x01000 /* Update the journal format */
516#define EXT4_MOUNT_NO_UID32 0x02000 /* Disable 32-bit UIDs */
517#define EXT4_MOUNT_XATTR_USER 0x04000 /* Extended user attributes */
518#define EXT4_MOUNT_POSIX_ACL 0x08000 /* POSIX Access Control Lists */
519#define EXT4_MOUNT_RESERVATION 0x10000 /* Preallocation */
520#define EXT4_MOUNT_BARRIER 0x20000 /* Use block barriers */
521#define EXT4_MOUNT_NOBH 0x40000 /* No bufferheads */
522#define EXT4_MOUNT_QUOTA 0x80000 /* Some quota option set */
523#define EXT4_MOUNT_USRQUOTA 0x100000 /* "old" user quota */
524#define EXT4_MOUNT_GRPQUOTA 0x200000 /* "old" group quota */
525#define EXT4_MOUNT_EXTENTS 0x400000 /* Extents support */
526#define EXT4_MOUNT_JOURNAL_CHECKSUM 0x800000 /* Journal checksums */
527#define EXT4_MOUNT_JOURNAL_ASYNC_COMMIT 0x1000000 /* Journal Async Commit */
528#define EXT4_MOUNT_I_VERSION 0x2000000 /* i_version support */
529#define EXT4_MOUNT_MBALLOC 0x4000000 /* Buddy allocation support */
530/* Compatibility, for having both ext2_fs.h and ext4_fs.h included at once */
531#ifndef _LINUX_EXT2_FS_H
532#define clear_opt(o, opt) o &= ~EXT4_MOUNT_##opt
533#define set_opt(o, opt) o |= EXT4_MOUNT_##opt
534#define test_opt(sb, opt) (EXT4_SB(sb)->s_mount_opt & \
535 EXT4_MOUNT_##opt)
536#else
537#define EXT2_MOUNT_NOLOAD EXT4_MOUNT_NOLOAD
538#define EXT2_MOUNT_ABORT EXT4_MOUNT_ABORT
539#define EXT2_MOUNT_DATA_FLAGS EXT4_MOUNT_DATA_FLAGS
540#endif
541
542#define ext4_set_bit ext2_set_bit
543#define ext4_set_bit_atomic ext2_set_bit_atomic
544#define ext4_clear_bit ext2_clear_bit
545#define ext4_clear_bit_atomic ext2_clear_bit_atomic
546#define ext4_test_bit ext2_test_bit
547#define ext4_find_first_zero_bit ext2_find_first_zero_bit
548#define ext4_find_next_zero_bit ext2_find_next_zero_bit
549#define ext4_find_next_bit ext2_find_next_bit
550
551/*
552 * Maximal mount counts between two filesystem checks
553 */
554#define EXT4_DFL_MAX_MNT_COUNT 20 /* Allow 20 mounts */
555#define EXT4_DFL_CHECKINTERVAL 0 /* Don't use interval check */
556
557/*
558 * Behaviour when detecting errors
559 */
560#define EXT4_ERRORS_CONTINUE 1 /* Continue execution */
561#define EXT4_ERRORS_RO 2 /* Remount fs read-only */
562#define EXT4_ERRORS_PANIC 3 /* Panic */
563#define EXT4_ERRORS_DEFAULT EXT4_ERRORS_CONTINUE
564
565/*
566 * Structure of the super block
567 */
568struct ext4_super_block {
569/*00*/ __le32 s_inodes_count; /* Inodes count */
570 __le32 s_blocks_count_lo; /* Blocks count */
571 __le32 s_r_blocks_count_lo; /* Reserved blocks count */
572 __le32 s_free_blocks_count_lo; /* Free blocks count */
573/*10*/ __le32 s_free_inodes_count; /* Free inodes count */
574 __le32 s_first_data_block; /* First Data Block */
575 __le32 s_log_block_size; /* Block size */
576 __le32 s_obso_log_frag_size; /* Obsoleted fragment size */
577/*20*/ __le32 s_blocks_per_group; /* # Blocks per group */
578 __le32 s_obso_frags_per_group; /* Obsoleted fragments per group */
579 __le32 s_inodes_per_group; /* # Inodes per group */
580 __le32 s_mtime; /* Mount time */
581/*30*/ __le32 s_wtime; /* Write time */
582 __le16 s_mnt_count; /* Mount count */
583 __le16 s_max_mnt_count; /* Maximal mount count */
584 __le16 s_magic; /* Magic signature */
585 __le16 s_state; /* File system state */
586 __le16 s_errors; /* Behaviour when detecting errors */
587 __le16 s_minor_rev_level; /* minor revision level */
588/*40*/ __le32 s_lastcheck; /* time of last check */
589 __le32 s_checkinterval; /* max. time between checks */
590 __le32 s_creator_os; /* OS */
591 __le32 s_rev_level; /* Revision level */
592/*50*/ __le16 s_def_resuid; /* Default uid for reserved blocks */
593 __le16 s_def_resgid; /* Default gid for reserved blocks */
594 /*
595 * These fields are for EXT4_DYNAMIC_REV superblocks only.
596 *
597 * Note: the difference between the compatible feature set and
598 * the incompatible feature set is that if there is a bit set
599 * in the incompatible feature set that the kernel doesn't
600 * know about, it should refuse to mount the filesystem.
601 *
602 * e2fsck's requirements are more strict; if it doesn't know
603 * about a feature in either the compatible or incompatible
604 * feature set, it must abort and not try to meddle with
605 * things it doesn't understand...
606 */
607 __le32 s_first_ino; /* First non-reserved inode */
608 __le16 s_inode_size; /* size of inode structure */
609 __le16 s_block_group_nr; /* block group # of this superblock */
610 __le32 s_feature_compat; /* compatible feature set */
611/*60*/ __le32 s_feature_incompat; /* incompatible feature set */
612 __le32 s_feature_ro_compat; /* readonly-compatible feature set */
613/*68*/ __u8 s_uuid[16]; /* 128-bit uuid for volume */
614/*78*/ char s_volume_name[16]; /* volume name */
615/*88*/ char s_last_mounted[64]; /* directory where last mounted */
616/*C8*/ __le32 s_algorithm_usage_bitmap; /* For compression */
617 /*
618 * Performance hints. Directory preallocation should only
619 * happen if the EXT4_FEATURE_COMPAT_DIR_PREALLOC flag is on.
620 */
621 __u8 s_prealloc_blocks; /* Nr of blocks to try to preallocate*/
622 __u8 s_prealloc_dir_blocks; /* Nr to preallocate for dirs */
623 __le16 s_reserved_gdt_blocks; /* Per group desc for online growth */
624 /*
625 * Journaling support valid if EXT4_FEATURE_COMPAT_HAS_JOURNAL set.
626 */
627/*D0*/ __u8 s_journal_uuid[16]; /* uuid of journal superblock */
628/*E0*/ __le32 s_journal_inum; /* inode number of journal file */
629 __le32 s_journal_dev; /* device number of journal file */
630 __le32 s_last_orphan; /* start of list of inodes to delete */
631 __le32 s_hash_seed[4]; /* HTREE hash seed */
632 __u8 s_def_hash_version; /* Default hash version to use */
633 __u8 s_reserved_char_pad;
634 __le16 s_desc_size; /* size of group descriptor */
635/*100*/ __le32 s_default_mount_opts;
636 __le32 s_first_meta_bg; /* First metablock block group */
637 __le32 s_mkfs_time; /* When the filesystem was created */
638 __le32 s_jnl_blocks[17]; /* Backup of the journal inode */
639 /* 64bit support valid if EXT4_FEATURE_COMPAT_64BIT */
640/*150*/ __le32 s_blocks_count_hi; /* Blocks count */
641 __le32 s_r_blocks_count_hi; /* Reserved blocks count */
642 __le32 s_free_blocks_count_hi; /* Free blocks count */
643 __le16 s_min_extra_isize; /* All inodes have at least # bytes */
644 __le16 s_want_extra_isize; /* New inodes should reserve # bytes */
645 __le32 s_flags; /* Miscellaneous flags */
646 __le16 s_raid_stride; /* RAID stride */
647 __le16 s_mmp_interval; /* # seconds to wait in MMP checking */
648 __le64 s_mmp_block; /* Block for multi-mount protection */
649 __le32 s_raid_stripe_width; /* blocks on all data disks (N*stride)*/
650 __u32 s_reserved[163]; /* Padding to the end of the block */
651};
652
653#ifdef __KERNEL__
654static inline struct ext4_sb_info * EXT4_SB(struct super_block *sb)
655{
656 return sb->s_fs_info;
657}
658static inline struct ext4_inode_info *EXT4_I(struct inode *inode)
659{
660 return container_of(inode, struct ext4_inode_info, vfs_inode);
661}
662
663static inline struct timespec ext4_current_time(struct inode *inode)
664{
665 return (inode->i_sb->s_time_gran < NSEC_PER_SEC) ?
666 current_fs_time(inode->i_sb) : CURRENT_TIME_SEC;
667}
668
669
670static inline int ext4_valid_inum(struct super_block *sb, unsigned long ino)
671{
672 return ino == EXT4_ROOT_INO ||
673 ino == EXT4_JOURNAL_INO ||
674 ino == EXT4_RESIZE_INO ||
675 (ino >= EXT4_FIRST_INO(sb) &&
676 ino <= le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count));
677}
678#else
679/* Assume that user mode programs are passing in an ext4fs superblock, not
680 * a kernel struct super_block. This will allow us to call the feature-test
681 * macros from user land. */
682#define EXT4_SB(sb) (sb)
683#endif
684
685#define NEXT_ORPHAN(inode) EXT4_I(inode)->i_dtime
686
687/*
688 * Codes for operating systems
689 */
690#define EXT4_OS_LINUX 0
691#define EXT4_OS_HURD 1
692#define EXT4_OS_MASIX 2
693#define EXT4_OS_FREEBSD 3
694#define EXT4_OS_LITES 4
695
696/*
697 * Revision levels
698 */
699#define EXT4_GOOD_OLD_REV 0 /* The good old (original) format */
700#define EXT4_DYNAMIC_REV 1 /* V2 format w/ dynamic inode sizes */
701
702#define EXT4_CURRENT_REV EXT4_GOOD_OLD_REV
703#define EXT4_MAX_SUPP_REV EXT4_DYNAMIC_REV
704
705#define EXT4_GOOD_OLD_INODE_SIZE 128
706
707/*
708 * Feature set definitions
709 */
710
711#define EXT4_HAS_COMPAT_FEATURE(sb,mask) \
712 ( EXT4_SB(sb)->s_es->s_feature_compat & cpu_to_le32(mask) )
713#define EXT4_HAS_RO_COMPAT_FEATURE(sb,mask) \
714 ( EXT4_SB(sb)->s_es->s_feature_ro_compat & cpu_to_le32(mask) )
715#define EXT4_HAS_INCOMPAT_FEATURE(sb,mask) \
716 ( EXT4_SB(sb)->s_es->s_feature_incompat & cpu_to_le32(mask) )
717#define EXT4_SET_COMPAT_FEATURE(sb,mask) \
718 EXT4_SB(sb)->s_es->s_feature_compat |= cpu_to_le32(mask)
719#define EXT4_SET_RO_COMPAT_FEATURE(sb,mask) \
720 EXT4_SB(sb)->s_es->s_feature_ro_compat |= cpu_to_le32(mask)
721#define EXT4_SET_INCOMPAT_FEATURE(sb,mask) \
722 EXT4_SB(sb)->s_es->s_feature_incompat |= cpu_to_le32(mask)
723#define EXT4_CLEAR_COMPAT_FEATURE(sb,mask) \
724 EXT4_SB(sb)->s_es->s_feature_compat &= ~cpu_to_le32(mask)
725#define EXT4_CLEAR_RO_COMPAT_FEATURE(sb,mask) \
726 EXT4_SB(sb)->s_es->s_feature_ro_compat &= ~cpu_to_le32(mask)
727#define EXT4_CLEAR_INCOMPAT_FEATURE(sb,mask) \
728 EXT4_SB(sb)->s_es->s_feature_incompat &= ~cpu_to_le32(mask)
729
730#define EXT4_FEATURE_COMPAT_DIR_PREALLOC 0x0001
731#define EXT4_FEATURE_COMPAT_IMAGIC_INODES 0x0002
732#define EXT4_FEATURE_COMPAT_HAS_JOURNAL 0x0004
733#define EXT4_FEATURE_COMPAT_EXT_ATTR 0x0008
734#define EXT4_FEATURE_COMPAT_RESIZE_INODE 0x0010
735#define EXT4_FEATURE_COMPAT_DIR_INDEX 0x0020
736
737#define EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER 0x0001
738#define EXT4_FEATURE_RO_COMPAT_LARGE_FILE 0x0002
739#define EXT4_FEATURE_RO_COMPAT_BTREE_DIR 0x0004
740#define EXT4_FEATURE_RO_COMPAT_HUGE_FILE 0x0008
741#define EXT4_FEATURE_RO_COMPAT_GDT_CSUM 0x0010
742#define EXT4_FEATURE_RO_COMPAT_DIR_NLINK 0x0020
743#define EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE 0x0040
744
745#define EXT4_FEATURE_INCOMPAT_COMPRESSION 0x0001
746#define EXT4_FEATURE_INCOMPAT_FILETYPE 0x0002
747#define EXT4_FEATURE_INCOMPAT_RECOVER 0x0004 /* Needs recovery */
748#define EXT4_FEATURE_INCOMPAT_JOURNAL_DEV 0x0008 /* Journal device */
749#define EXT4_FEATURE_INCOMPAT_META_BG 0x0010
750#define EXT4_FEATURE_INCOMPAT_EXTENTS 0x0040 /* extents support */
751#define EXT4_FEATURE_INCOMPAT_64BIT 0x0080
752#define EXT4_FEATURE_INCOMPAT_MMP 0x0100
753#define EXT4_FEATURE_INCOMPAT_FLEX_BG 0x0200
754
755#define EXT4_FEATURE_COMPAT_SUPP EXT2_FEATURE_COMPAT_EXT_ATTR
756#define EXT4_FEATURE_INCOMPAT_SUPP (EXT4_FEATURE_INCOMPAT_FILETYPE| \
757 EXT4_FEATURE_INCOMPAT_RECOVER| \
758 EXT4_FEATURE_INCOMPAT_META_BG| \
759 EXT4_FEATURE_INCOMPAT_EXTENTS| \
760 EXT4_FEATURE_INCOMPAT_64BIT| \
761 EXT4_FEATURE_INCOMPAT_FLEX_BG)
762#define EXT4_FEATURE_RO_COMPAT_SUPP (EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER| \
763 EXT4_FEATURE_RO_COMPAT_LARGE_FILE| \
764 EXT4_FEATURE_RO_COMPAT_GDT_CSUM| \
765 EXT4_FEATURE_RO_COMPAT_DIR_NLINK | \
766 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE | \
767 EXT4_FEATURE_RO_COMPAT_BTREE_DIR |\
768 EXT4_FEATURE_RO_COMPAT_HUGE_FILE)
769
770/*
771 * Default values for user and/or group using reserved blocks
772 */
773#define EXT4_DEF_RESUID 0
774#define EXT4_DEF_RESGID 0
775
776/*
777 * Default mount options
778 */
779#define EXT4_DEFM_DEBUG 0x0001
780#define EXT4_DEFM_BSDGROUPS 0x0002
781#define EXT4_DEFM_XATTR_USER 0x0004
782#define EXT4_DEFM_ACL 0x0008
783#define EXT4_DEFM_UID16 0x0010
784#define EXT4_DEFM_JMODE 0x0060
785#define EXT4_DEFM_JMODE_DATA 0x0020
786#define EXT4_DEFM_JMODE_ORDERED 0x0040
787#define EXT4_DEFM_JMODE_WBACK 0x0060
788
789/*
790 * Structure of a directory entry
791 */
792#define EXT4_NAME_LEN 255
793
794struct ext4_dir_entry {
795 __le32 inode; /* Inode number */
796 __le16 rec_len; /* Directory entry length */
797 __le16 name_len; /* Name length */
798 char name[EXT4_NAME_LEN]; /* File name */
799};
800
801/*
802 * The new version of the directory entry. Since EXT4 structures are
803 * stored in intel byte order, and the name_len field could never be
804 * bigger than 255 chars, it's safe to reclaim the extra byte for the
805 * file_type field.
806 */
807struct ext4_dir_entry_2 {
808 __le32 inode; /* Inode number */
809 __le16 rec_len; /* Directory entry length */
810 __u8 name_len; /* Name length */
811 __u8 file_type;
812 char name[EXT4_NAME_LEN]; /* File name */
813};
814
815/*
816 * Ext4 directory file types. Only the low 3 bits are used. The
817 * other bits are reserved for now.
818 */
819#define EXT4_FT_UNKNOWN 0
820#define EXT4_FT_REG_FILE 1
821#define EXT4_FT_DIR 2
822#define EXT4_FT_CHRDEV 3
823#define EXT4_FT_BLKDEV 4
824#define EXT4_FT_FIFO 5
825#define EXT4_FT_SOCK 6
826#define EXT4_FT_SYMLINK 7
827
828#define EXT4_FT_MAX 8
829
830/*
831 * EXT4_DIR_PAD defines the directory entries boundaries
832 *
833 * NOTE: It must be a multiple of 4
834 */
835#define EXT4_DIR_PAD 4
836#define EXT4_DIR_ROUND (EXT4_DIR_PAD - 1)
837#define EXT4_DIR_REC_LEN(name_len) (((name_len) + 8 + EXT4_DIR_ROUND) & \
838 ~EXT4_DIR_ROUND)
839#define EXT4_MAX_REC_LEN ((1<<16)-1)
840
841static inline unsigned ext4_rec_len_from_disk(__le16 dlen)
842{
843 unsigned len = le16_to_cpu(dlen);
844
845 if (len == EXT4_MAX_REC_LEN)
846 return 1 << 16;
847 return len;
848}
849
850static inline __le16 ext4_rec_len_to_disk(unsigned len)
851{
852 if (len == (1 << 16))
853 return cpu_to_le16(EXT4_MAX_REC_LEN);
854 else if (len > (1 << 16))
855 BUG();
856 return cpu_to_le16(len);
857}
858
859/*
860 * Hash Tree Directory indexing
861 * (c) Daniel Phillips, 2001
862 */
863
864#define is_dx(dir) (EXT4_HAS_COMPAT_FEATURE(dir->i_sb, \
865 EXT4_FEATURE_COMPAT_DIR_INDEX) && \
866 (EXT4_I(dir)->i_flags & EXT4_INDEX_FL))
867#define EXT4_DIR_LINK_MAX(dir) (!is_dx(dir) && (dir)->i_nlink >= EXT4_LINK_MAX)
868#define EXT4_DIR_LINK_EMPTY(dir) ((dir)->i_nlink == 2 || (dir)->i_nlink == 1)
869
870/* Legal values for the dx_root hash_version field: */
871
872#define DX_HASH_LEGACY 0
873#define DX_HASH_HALF_MD4 1
874#define DX_HASH_TEA 2
875
876#ifdef __KERNEL__
877
878/* hash info structure used by the directory hash */
879struct dx_hash_info
880{
881 u32 hash;
882 u32 minor_hash;
883 int hash_version;
884 u32 *seed;
885};
886
887#define EXT4_HTREE_EOF 0x7fffffff
888
889/*
890 * Control parameters used by ext4_htree_next_block
891 */
892#define HASH_NB_ALWAYS 1
893
894
895/*
896 * Describe an inode's exact location on disk and in memory
897 */
898struct ext4_iloc
899{
900 struct buffer_head *bh;
901 unsigned long offset;
902 ext4_group_t block_group;
903};
904
905static inline struct ext4_inode *ext4_raw_inode(struct ext4_iloc *iloc)
906{
907 return (struct ext4_inode *) (iloc->bh->b_data + iloc->offset);
908}
909
910/*
911 * This structure is stuffed into the struct file's private_data field
912 * for directories. It is where we put information so that we can do
913 * readdir operations in hash tree order.
914 */
915struct dir_private_info {
916 struct rb_root root;
917 struct rb_node *curr_node;
918 struct fname *extra_fname;
919 loff_t last_pos;
920 __u32 curr_hash;
921 __u32 curr_minor_hash;
922 __u32 next_hash;
923};
924
925/* calculate the first block number of the group */
926static inline ext4_fsblk_t
927ext4_group_first_block_no(struct super_block *sb, ext4_group_t group_no)
928{
929 return group_no * (ext4_fsblk_t)EXT4_BLOCKS_PER_GROUP(sb) +
930 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block);
931}
932
933/*
934 * Special error return code only used by dx_probe() and its callers.
935 */
936#define ERR_BAD_DX_DIR -75000
937
938void ext4_get_group_no_and_offset(struct super_block *sb, ext4_fsblk_t blocknr,
939 unsigned long *blockgrpp, ext4_grpblk_t *offsetp);
940
941/*
942 * Function prototypes
943 */
944
945/*
946 * Ok, these declarations are also in <linux/kernel.h> but none of the
947 * ext4 source programs needs to include it so they are duplicated here.
948 */
949# define NORET_TYPE /**/
950# define ATTRIB_NORET __attribute__((noreturn))
951# define NORET_AND noreturn,
952
953/* balloc.c */
954extern unsigned int ext4_block_group(struct super_block *sb,
955 ext4_fsblk_t blocknr);
956extern ext4_grpblk_t ext4_block_group_offset(struct super_block *sb,
957 ext4_fsblk_t blocknr);
958extern int ext4_bg_has_super(struct super_block *sb, ext4_group_t group);
959extern unsigned long ext4_bg_num_gdb(struct super_block *sb,
960 ext4_group_t group);
961extern ext4_fsblk_t ext4_new_block (handle_t *handle, struct inode *inode,
962 ext4_fsblk_t goal, int *errp);
963extern ext4_fsblk_t ext4_new_blocks (handle_t *handle, struct inode *inode,
964 ext4_fsblk_t goal, unsigned long *count, int *errp);
965extern ext4_fsblk_t ext4_new_blocks_old(handle_t *handle, struct inode *inode,
966 ext4_fsblk_t goal, unsigned long *count, int *errp);
967extern void ext4_free_blocks (handle_t *handle, struct inode *inode,
968 ext4_fsblk_t block, unsigned long count, int metadata);
969extern void ext4_free_blocks_sb (handle_t *handle, struct super_block *sb,
970 ext4_fsblk_t block, unsigned long count,
971 unsigned long *pdquot_freed_blocks);
972extern ext4_fsblk_t ext4_count_free_blocks (struct super_block *);
973extern void ext4_check_blocks_bitmap (struct super_block *);
974extern struct ext4_group_desc * ext4_get_group_desc(struct super_block * sb,
975 ext4_group_t block_group,
976 struct buffer_head ** bh);
977extern int ext4_should_retry_alloc(struct super_block *sb, int *retries);
978extern void ext4_init_block_alloc_info(struct inode *);
979extern void ext4_rsv_window_add(struct super_block *sb, struct ext4_reserve_window_node *rsv);
980
981/* dir.c */
982extern int ext4_check_dir_entry(const char *, struct inode *,
983 struct ext4_dir_entry_2 *,
984 struct buffer_head *, unsigned long);
985extern int ext4_htree_store_dirent(struct file *dir_file, __u32 hash,
986 __u32 minor_hash,
987 struct ext4_dir_entry_2 *dirent);
988extern void ext4_htree_free_dir_info(struct dir_private_info *p);
989
990/* fsync.c */
991extern int ext4_sync_file (struct file *, struct dentry *, int);
992
993/* hash.c */
994extern int ext4fs_dirhash(const char *name, int len, struct
995 dx_hash_info *hinfo);
996
997/* ialloc.c */
998extern struct inode * ext4_new_inode (handle_t *, struct inode *, int);
999extern void ext4_free_inode (handle_t *, struct inode *);
1000extern struct inode * ext4_orphan_get (struct super_block *, unsigned long);
1001extern unsigned long ext4_count_free_inodes (struct super_block *);
1002extern unsigned long ext4_count_dirs (struct super_block *);
1003extern void ext4_check_inodes_bitmap (struct super_block *);
1004extern unsigned long ext4_count_free (struct buffer_head *, unsigned);
1005
1006/* mballoc.c */
1007extern long ext4_mb_stats;
1008extern long ext4_mb_max_to_scan;
1009extern int ext4_mb_init(struct super_block *, int);
1010extern int ext4_mb_release(struct super_block *);
1011extern ext4_fsblk_t ext4_mb_new_blocks(handle_t *,
1012 struct ext4_allocation_request *, int *);
1013extern int ext4_mb_reserve_blocks(struct super_block *, int);
1014extern void ext4_mb_discard_inode_preallocations(struct inode *);
1015extern int __init init_ext4_mballoc(void);
1016extern void exit_ext4_mballoc(void);
1017extern void ext4_mb_free_blocks(handle_t *, struct inode *,
1018 unsigned long, unsigned long, int, unsigned long *);
1019
1020
1021/* inode.c */
1022int ext4_forget(handle_t *handle, int is_metadata, struct inode *inode,
1023 struct buffer_head *bh, ext4_fsblk_t blocknr);
1024struct buffer_head *ext4_getblk(handle_t *, struct inode *,
1025 ext4_lblk_t, int, int *);
1026struct buffer_head *ext4_bread(handle_t *, struct inode *,
1027 ext4_lblk_t, int, int *);
1028int ext4_get_blocks_handle(handle_t *handle, struct inode *inode,
1029 ext4_lblk_t iblock, unsigned long maxblocks,
1030 struct buffer_head *bh_result,
1031 int create, int extend_disksize);
1032
1033extern struct inode *ext4_iget(struct super_block *, unsigned long);
1034extern int ext4_write_inode (struct inode *, int);
1035extern int ext4_setattr (struct dentry *, struct iattr *);
1036extern void ext4_delete_inode (struct inode *);
1037extern int ext4_sync_inode (handle_t *, struct inode *);
1038extern void ext4_discard_reservation (struct inode *);
1039extern void ext4_dirty_inode(struct inode *);
1040extern int ext4_change_inode_journal_flag(struct inode *, int);
1041extern int ext4_get_inode_loc(struct inode *, struct ext4_iloc *);
1042extern void ext4_truncate (struct inode *);
1043extern void ext4_set_inode_flags(struct inode *);
1044extern void ext4_get_inode_flags(struct ext4_inode_info *);
1045extern void ext4_set_aops(struct inode *inode);
1046extern int ext4_writepage_trans_blocks(struct inode *);
1047extern int ext4_block_truncate_page(handle_t *handle, struct page *page,
1048 struct address_space *mapping, loff_t from);
1049
1050/* ioctl.c */
1051extern long ext4_ioctl(struct file *, unsigned int, unsigned long);
1052extern long ext4_compat_ioctl (struct file *, unsigned int, unsigned long);
1053
1054/* migrate.c */
1055extern int ext4_ext_migrate(struct inode *, struct file *, unsigned int,
1056 unsigned long);
1057/* namei.c */
1058extern int ext4_orphan_add(handle_t *, struct inode *);
1059extern int ext4_orphan_del(handle_t *, struct inode *);
1060extern int ext4_htree_fill_tree(struct file *dir_file, __u32 start_hash,
1061 __u32 start_minor_hash, __u32 *next_hash);
1062
1063/* resize.c */
1064extern int ext4_group_add(struct super_block *sb,
1065 struct ext4_new_group_data *input);
1066extern int ext4_group_extend(struct super_block *sb,
1067 struct ext4_super_block *es,
1068 ext4_fsblk_t n_blocks_count);
1069
1070/* super.c */
1071extern void ext4_error (struct super_block *, const char *, const char *, ...)
1072 __attribute__ ((format (printf, 3, 4)));
1073extern void __ext4_std_error (struct super_block *, const char *, int);
1074extern void ext4_abort (struct super_block *, const char *, const char *, ...)
1075 __attribute__ ((format (printf, 3, 4)));
1076extern void ext4_warning (struct super_block *, const char *, const char *, ...)
1077 __attribute__ ((format (printf, 3, 4)));
1078extern void ext4_update_dynamic_rev (struct super_block *sb);
1079extern int ext4_update_compat_feature(handle_t *handle, struct super_block *sb,
1080 __u32 compat);
1081extern int ext4_update_rocompat_feature(handle_t *handle,
1082 struct super_block *sb, __u32 rocompat);
1083extern int ext4_update_incompat_feature(handle_t *handle,
1084 struct super_block *sb, __u32 incompat);
1085extern ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
1086 struct ext4_group_desc *bg);
1087extern ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
1088 struct ext4_group_desc *bg);
1089extern ext4_fsblk_t ext4_inode_table(struct super_block *sb,
1090 struct ext4_group_desc *bg);
1091extern void ext4_block_bitmap_set(struct super_block *sb,
1092 struct ext4_group_desc *bg, ext4_fsblk_t blk);
1093extern void ext4_inode_bitmap_set(struct super_block *sb,
1094 struct ext4_group_desc *bg, ext4_fsblk_t blk);
1095extern void ext4_inode_table_set(struct super_block *sb,
1096 struct ext4_group_desc *bg, ext4_fsblk_t blk);
1097
1098static inline ext4_fsblk_t ext4_blocks_count(struct ext4_super_block *es)
1099{
1100 return ((ext4_fsblk_t)le32_to_cpu(es->s_blocks_count_hi) << 32) |
1101 le32_to_cpu(es->s_blocks_count_lo);
1102}
1103
1104static inline ext4_fsblk_t ext4_r_blocks_count(struct ext4_super_block *es)
1105{
1106 return ((ext4_fsblk_t)le32_to_cpu(es->s_r_blocks_count_hi) << 32) |
1107 le32_to_cpu(es->s_r_blocks_count_lo);
1108}
1109
1110static inline ext4_fsblk_t ext4_free_blocks_count(struct ext4_super_block *es)
1111{
1112 return ((ext4_fsblk_t)le32_to_cpu(es->s_free_blocks_count_hi) << 32) |
1113 le32_to_cpu(es->s_free_blocks_count_lo);
1114}
1115
1116static inline void ext4_blocks_count_set(struct ext4_super_block *es,
1117 ext4_fsblk_t blk)
1118{
1119 es->s_blocks_count_lo = cpu_to_le32((u32)blk);
1120 es->s_blocks_count_hi = cpu_to_le32(blk >> 32);
1121}
1122
1123static inline void ext4_free_blocks_count_set(struct ext4_super_block *es,
1124 ext4_fsblk_t blk)
1125{
1126 es->s_free_blocks_count_lo = cpu_to_le32((u32)blk);
1127 es->s_free_blocks_count_hi = cpu_to_le32(blk >> 32);
1128}
1129
1130static inline void ext4_r_blocks_count_set(struct ext4_super_block *es,
1131 ext4_fsblk_t blk)
1132{
1133 es->s_r_blocks_count_lo = cpu_to_le32((u32)blk);
1134 es->s_r_blocks_count_hi = cpu_to_le32(blk >> 32);
1135}
1136
1137static inline loff_t ext4_isize(struct ext4_inode *raw_inode)
1138{
1139 return ((loff_t)le32_to_cpu(raw_inode->i_size_high) << 32) |
1140 le32_to_cpu(raw_inode->i_size_lo);
1141}
1142
1143static inline void ext4_isize_set(struct ext4_inode *raw_inode, loff_t i_size)
1144{
1145 raw_inode->i_size_lo = cpu_to_le32(i_size);
1146 raw_inode->i_size_high = cpu_to_le32(i_size >> 32);
1147}
1148
1149static inline
1150struct ext4_group_info *ext4_get_group_info(struct super_block *sb,
1151 ext4_group_t group)
1152{
1153 struct ext4_group_info ***grp_info;
1154 long indexv, indexh;
1155 grp_info = EXT4_SB(sb)->s_group_info;
1156 indexv = group >> (EXT4_DESC_PER_BLOCK_BITS(sb));
1157 indexh = group & ((EXT4_DESC_PER_BLOCK(sb)) - 1);
1158 return grp_info[indexv][indexh];
1159}
1160
1161
1162#define ext4_std_error(sb, errno) \
1163do { \
1164 if ((errno)) \
1165 __ext4_std_error((sb), __FUNCTION__, (errno)); \
1166} while (0)
1167
1168/*
1169 * Inodes and files operations
1170 */
1171
1172/* dir.c */
1173extern const struct file_operations ext4_dir_operations;
1174
1175/* file.c */
1176extern const struct inode_operations ext4_file_inode_operations;
1177extern const struct file_operations ext4_file_operations;
1178
1179/* namei.c */
1180extern const struct inode_operations ext4_dir_inode_operations;
1181extern const struct inode_operations ext4_special_inode_operations;
1182
1183/* symlink.c */
1184extern const struct inode_operations ext4_symlink_inode_operations;
1185extern const struct inode_operations ext4_fast_symlink_inode_operations;
1186
1187/* extents.c */
1188extern int ext4_ext_tree_init(handle_t *handle, struct inode *);
1189extern int ext4_ext_writepage_trans_blocks(struct inode *, int);
1190extern int ext4_ext_get_blocks(handle_t *handle, struct inode *inode,
1191 ext4_lblk_t iblock,
1192 unsigned long max_blocks, struct buffer_head *bh_result,
1193 int create, int extend_disksize);
1194extern void ext4_ext_truncate(struct inode *, struct page *);
1195extern void ext4_ext_init(struct super_block *);
1196extern void ext4_ext_release(struct super_block *);
1197extern long ext4_fallocate(struct inode *inode, int mode, loff_t offset,
1198 loff_t len);
1199extern int ext4_get_blocks_wrap(handle_t *handle, struct inode *inode,
1200 sector_t block, unsigned long max_blocks,
1201 struct buffer_head *bh, int create,
1202 int extend_disksize);
1203#endif /* __KERNEL__ */
1204
1205#endif /* _EXT4_H */
diff --git a/fs/ext4/ext4_extents.h b/fs/ext4/ext4_extents.h
new file mode 100644
index 000000000000..75333b595fab
--- /dev/null
+++ b/fs/ext4/ext4_extents.h
@@ -0,0 +1,232 @@
1/*
2 * Copyright (c) 2003-2006, Cluster File Systems, Inc, info@clusterfs.com
3 * Written by Alex Tomas <alex@clusterfs.com>
4 *
5 * This program is free software; you can redistribute it and/or modify
6 * it under the terms of the GNU General Public License version 2 as
7 * published by the Free Software Foundation.
8 *
9 * This program is distributed in the hope that it will be useful,
10 * but WITHOUT ANY WARRANTY; without even the implied warranty of
11 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
12 * GNU General Public License for more details.
13 *
14 * You should have received a copy of the GNU General Public Licens
15 * along with this program; if not, write to the Free Software
16 * Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA 02111-
17 */
18
19#ifndef _EXT4_EXTENTS
20#define _EXT4_EXTENTS
21
22#include "ext4.h"
23
24/*
25 * With AGGRESSIVE_TEST defined, the capacity of index/leaf blocks
26 * becomes very small, so index split, in-depth growing and
27 * other hard changes happen much more often.
28 * This is for debug purposes only.
29 */
30#define AGGRESSIVE_TEST_
31
32/*
33 * With EXTENTS_STATS defined, the number of blocks and extents
34 * are collected in the truncate path. They'll be shown at
35 * umount time.
36 */
37#define EXTENTS_STATS__
38
39/*
40 * If CHECK_BINSEARCH is defined, then the results of the binary search
41 * will also be checked by linear search.
42 */
43#define CHECK_BINSEARCH__
44
45/*
46 * If EXT_DEBUG is defined you can use the 'extdebug' mount option
47 * to get lots of info about what's going on.
48 */
49#define EXT_DEBUG__
50#ifdef EXT_DEBUG
51#define ext_debug(a...) printk(a)
52#else
53#define ext_debug(a...)
54#endif
55
56/*
57 * If EXT_STATS is defined then stats numbers are collected.
58 * These number will be displayed at umount time.
59 */
60#define EXT_STATS_
61
62
63/*
64 * ext4_inode has i_block array (60 bytes total).
65 * The first 12 bytes store ext4_extent_header;
66 * the remainder stores an array of ext4_extent.
67 */
68
69/*
70 * This is the extent on-disk structure.
71 * It's used at the bottom of the tree.
72 */
73struct ext4_extent {
74 __le32 ee_block; /* first logical block extent covers */
75 __le16 ee_len; /* number of blocks covered by extent */
76 __le16 ee_start_hi; /* high 16 bits of physical block */
77 __le32 ee_start_lo; /* low 32 bits of physical block */
78};
79
80/*
81 * This is index on-disk structure.
82 * It's used at all the levels except the bottom.
83 */
84struct ext4_extent_idx {
85 __le32 ei_block; /* index covers logical blocks from 'block' */
86 __le32 ei_leaf_lo; /* pointer to the physical block of the next *
87 * level. leaf or next index could be there */
88 __le16 ei_leaf_hi; /* high 16 bits of physical block */
89 __u16 ei_unused;
90};
91
92/*
93 * Each block (leaves and indexes), even inode-stored has header.
94 */
95struct ext4_extent_header {
96 __le16 eh_magic; /* probably will support different formats */
97 __le16 eh_entries; /* number of valid entries */
98 __le16 eh_max; /* capacity of store in entries */
99 __le16 eh_depth; /* has tree real underlying blocks? */
100 __le32 eh_generation; /* generation of the tree */
101};
102
103#define EXT4_EXT_MAGIC cpu_to_le16(0xf30a)
104
105/*
106 * Array of ext4_ext_path contains path to some extent.
107 * Creation/lookup routines use it for traversal/splitting/etc.
108 * Truncate uses it to simulate recursive walking.
109 */
110struct ext4_ext_path {
111 ext4_fsblk_t p_block;
112 __u16 p_depth;
113 struct ext4_extent *p_ext;
114 struct ext4_extent_idx *p_idx;
115 struct ext4_extent_header *p_hdr;
116 struct buffer_head *p_bh;
117};
118
119/*
120 * structure for external API
121 */
122
123#define EXT4_EXT_CACHE_NO 0
124#define EXT4_EXT_CACHE_GAP 1
125#define EXT4_EXT_CACHE_EXTENT 2
126
127
128#define EXT_MAX_BLOCK 0xffffffff
129
130/*
131 * EXT_INIT_MAX_LEN is the maximum number of blocks we can have in an
132 * initialized extent. This is 2^15 and not (2^16 - 1), since we use the
133 * MSB of ee_len field in the extent datastructure to signify if this
134 * particular extent is an initialized extent or an uninitialized (i.e.
135 * preallocated).
136 * EXT_UNINIT_MAX_LEN is the maximum number of blocks we can have in an
137 * uninitialized extent.
138 * If ee_len is <= 0x8000, it is an initialized extent. Otherwise, it is an
139 * uninitialized one. In other words, if MSB of ee_len is set, it is an
140 * uninitialized extent with only one special scenario when ee_len = 0x8000.
141 * In this case we can not have an uninitialized extent of zero length and
142 * thus we make it as a special case of initialized extent with 0x8000 length.
143 * This way we get better extent-to-group alignment for initialized extents.
144 * Hence, the maximum number of blocks we can have in an *initialized*
145 * extent is 2^15 (32768) and in an *uninitialized* extent is 2^15-1 (32767).
146 */
147#define EXT_INIT_MAX_LEN (1UL << 15)
148#define EXT_UNINIT_MAX_LEN (EXT_INIT_MAX_LEN - 1)
149
150
151#define EXT_FIRST_EXTENT(__hdr__) \
152 ((struct ext4_extent *) (((char *) (__hdr__)) + \
153 sizeof(struct ext4_extent_header)))
154#define EXT_FIRST_INDEX(__hdr__) \
155 ((struct ext4_extent_idx *) (((char *) (__hdr__)) + \
156 sizeof(struct ext4_extent_header)))
157#define EXT_HAS_FREE_INDEX(__path__) \
158 (le16_to_cpu((__path__)->p_hdr->eh_entries) \
159 < le16_to_cpu((__path__)->p_hdr->eh_max))
160#define EXT_LAST_EXTENT(__hdr__) \
161 (EXT_FIRST_EXTENT((__hdr__)) + le16_to_cpu((__hdr__)->eh_entries) - 1)
162#define EXT_LAST_INDEX(__hdr__) \
163 (EXT_FIRST_INDEX((__hdr__)) + le16_to_cpu((__hdr__)->eh_entries) - 1)
164#define EXT_MAX_EXTENT(__hdr__) \
165 (EXT_FIRST_EXTENT((__hdr__)) + le16_to_cpu((__hdr__)->eh_max) - 1)
166#define EXT_MAX_INDEX(__hdr__) \
167 (EXT_FIRST_INDEX((__hdr__)) + le16_to_cpu((__hdr__)->eh_max) - 1)
168
169static inline struct ext4_extent_header *ext_inode_hdr(struct inode *inode)
170{
171 return (struct ext4_extent_header *) EXT4_I(inode)->i_data;
172}
173
174static inline struct ext4_extent_header *ext_block_hdr(struct buffer_head *bh)
175{
176 return (struct ext4_extent_header *) bh->b_data;
177}
178
179static inline unsigned short ext_depth(struct inode *inode)
180{
181 return le16_to_cpu(ext_inode_hdr(inode)->eh_depth);
182}
183
184static inline void ext4_ext_tree_changed(struct inode *inode)
185{
186 EXT4_I(inode)->i_ext_generation++;
187}
188
189static inline void
190ext4_ext_invalidate_cache(struct inode *inode)
191{
192 EXT4_I(inode)->i_cached_extent.ec_type = EXT4_EXT_CACHE_NO;
193}
194
195static inline void ext4_ext_mark_uninitialized(struct ext4_extent *ext)
196{
197 /* We can not have an uninitialized extent of zero length! */
198 BUG_ON((le16_to_cpu(ext->ee_len) & ~EXT_INIT_MAX_LEN) == 0);
199 ext->ee_len |= cpu_to_le16(EXT_INIT_MAX_LEN);
200}
201
202static inline int ext4_ext_is_uninitialized(struct ext4_extent *ext)
203{
204 /* Extent with ee_len of 0x8000 is treated as an initialized extent */
205 return (le16_to_cpu(ext->ee_len) > EXT_INIT_MAX_LEN);
206}
207
208static inline int ext4_ext_get_actual_len(struct ext4_extent *ext)
209{
210 return (le16_to_cpu(ext->ee_len) <= EXT_INIT_MAX_LEN ?
211 le16_to_cpu(ext->ee_len) :
212 (le16_to_cpu(ext->ee_len) - EXT_INIT_MAX_LEN));
213}
214
215extern ext4_fsblk_t idx_pblock(struct ext4_extent_idx *);
216extern void ext4_ext_store_pblock(struct ext4_extent *, ext4_fsblk_t);
217extern int ext4_extent_tree_init(handle_t *, struct inode *);
218extern int ext4_ext_calc_credits_for_insert(struct inode *, struct ext4_ext_path *);
219extern int ext4_ext_try_to_merge(struct inode *inode,
220 struct ext4_ext_path *path,
221 struct ext4_extent *);
222extern unsigned int ext4_ext_check_overlap(struct inode *, struct ext4_extent *, struct ext4_ext_path *);
223extern int ext4_ext_insert_extent(handle_t *, struct inode *, struct ext4_ext_path *, struct ext4_extent *);
224extern struct ext4_ext_path *ext4_ext_find_extent(struct inode *, ext4_lblk_t,
225 struct ext4_ext_path *);
226extern int ext4_ext_search_left(struct inode *, struct ext4_ext_path *,
227 ext4_lblk_t *, ext4_fsblk_t *);
228extern int ext4_ext_search_right(struct inode *, struct ext4_ext_path *,
229 ext4_lblk_t *, ext4_fsblk_t *);
230extern void ext4_ext_drop_refs(struct ext4_ext_path *);
231#endif /* _EXT4_EXTENTS */
232
diff --git a/fs/ext4/ext4_i.h b/fs/ext4/ext4_i.h
new file mode 100644
index 000000000000..26a4ae255d79
--- /dev/null
+++ b/fs/ext4/ext4_i.h
@@ -0,0 +1,167 @@
1/*
2 * ext4_i.h
3 *
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
8 *
9 * from
10 *
11 * linux/include/linux/minix_fs_i.h
12 *
13 * Copyright (C) 1991, 1992 Linus Torvalds
14 */
15
16#ifndef _EXT4_I
17#define _EXT4_I
18
19#include <linux/rwsem.h>
20#include <linux/rbtree.h>
21#include <linux/seqlock.h>
22#include <linux/mutex.h>
23
24/* data type for block offset of block group */
25typedef int ext4_grpblk_t;
26
27/* data type for filesystem-wide blocks number */
28typedef unsigned long long ext4_fsblk_t;
29
30/* data type for file logical block number */
31typedef __u32 ext4_lblk_t;
32
33/* data type for block group number */
34typedef unsigned long ext4_group_t;
35
36struct ext4_reserve_window {
37 ext4_fsblk_t _rsv_start; /* First byte reserved */
38 ext4_fsblk_t _rsv_end; /* Last byte reserved or 0 */
39};
40
41struct ext4_reserve_window_node {
42 struct rb_node rsv_node;
43 __u32 rsv_goal_size;
44 __u32 rsv_alloc_hit;
45 struct ext4_reserve_window rsv_window;
46};
47
48struct ext4_block_alloc_info {
49 /* information about reservation window */
50 struct ext4_reserve_window_node rsv_window_node;
51 /*
52 * was i_next_alloc_block in ext4_inode_info
53 * is the logical (file-relative) number of the
54 * most-recently-allocated block in this file.
55 * We use this for detecting linearly ascending allocation requests.
56 */
57 ext4_lblk_t last_alloc_logical_block;
58 /*
59 * Was i_next_alloc_goal in ext4_inode_info
60 * is the *physical* companion to i_next_alloc_block.
61 * it the physical block number of the block which was most-recentl
62 * allocated to this file. This give us the goal (target) for the next
63 * allocation when we detect linearly ascending requests.
64 */
65 ext4_fsblk_t last_alloc_physical_block;
66};
67
68#define rsv_start rsv_window._rsv_start
69#define rsv_end rsv_window._rsv_end
70
71/*
72 * storage for cached extent
73 */
74struct ext4_ext_cache {
75 ext4_fsblk_t ec_start;
76 ext4_lblk_t ec_block;
77 __u32 ec_len; /* must be 32bit to return holes */
78 __u32 ec_type;
79};
80
81/*
82 * third extended file system inode data in memory
83 */
84struct ext4_inode_info {
85 __le32 i_data[15]; /* unconverted */
86 __u32 i_flags;
87 ext4_fsblk_t i_file_acl;
88 __u32 i_dtime;
89
90 /*
91 * i_block_group is the number of the block group which contains
92 * this file's inode. Constant across the lifetime of the inode,
93 * it is ued for making block allocation decisions - we try to
94 * place a file's data blocks near its inode block, and new inodes
95 * near to their parent directory's inode.
96 */
97 ext4_group_t i_block_group;
98 __u32 i_state; /* Dynamic state flags for ext4 */
99
100 /* block reservation info */
101 struct ext4_block_alloc_info *i_block_alloc_info;
102
103 ext4_lblk_t i_dir_start_lookup;
104#ifdef CONFIG_EXT4DEV_FS_XATTR
105 /*
106 * Extended attributes can be read independently of the main file
107 * data. Taking i_mutex even when reading would cause contention
108 * between readers of EAs and writers of regular file data, so
109 * instead we synchronize on xattr_sem when reading or changing
110 * EAs.
111 */
112 struct rw_semaphore xattr_sem;
113#endif
114#ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
115 struct posix_acl *i_acl;
116 struct posix_acl *i_default_acl;
117#endif
118
119 struct list_head i_orphan; /* unlinked but open inodes */
120
121 /*
122 * i_disksize keeps track of what the inode size is ON DISK, not
123 * in memory. During truncate, i_size is set to the new size by
124 * the VFS prior to calling ext4_truncate(), but the filesystem won't
125 * set i_disksize to 0 until the truncate is actually under way.
126 *
127 * The intent is that i_disksize always represents the blocks which
128 * are used by this file. This allows recovery to restart truncate
129 * on orphans if we crash during truncate. We actually write i_disksize
130 * into the on-disk inode when writing inodes out, instead of i_size.
131 *
132 * The only time when i_disksize and i_size may be different is when
133 * a truncate is in progress. The only things which change i_disksize
134 * are ext4_get_block (growth) and ext4_truncate (shrinkth).
135 */
136 loff_t i_disksize;
137
138 /* on-disk additional length */
139 __u16 i_extra_isize;
140
141 /*
142 * i_data_sem is for serialising ext4_truncate() against
143 * ext4_getblock(). In the 2.4 ext2 design, great chunks of inode's
144 * data tree are chopped off during truncate. We can't do that in
145 * ext4 because whenever we perform intermediate commits during
146 * truncate, the inode and all the metadata blocks *must* be in a
147 * consistent state which allows truncation of the orphans to restart
148 * during recovery. Hence we must fix the get_block-vs-truncate race
149 * by other means, so we have i_data_sem.
150 */
151 struct rw_semaphore i_data_sem;
152 struct inode vfs_inode;
153
154 unsigned long i_ext_generation;
155 struct ext4_ext_cache i_cached_extent;
156 /*
157 * File creation time. Its function is same as that of
158 * struct timespec i_{a,c,m}time in the generic inode.
159 */
160 struct timespec i_crtime;
161
162 /* mballoc */
163 struct list_head i_prealloc_list;
164 spinlock_t i_prealloc_lock;
165};
166
167#endif /* _EXT4_I */
diff --git a/fs/ext4/ext4_jbd2.c b/fs/ext4/ext4_jbd2.c
index d6afe4e27340..c75384b34f2c 100644
--- a/fs/ext4/ext4_jbd2.c
+++ b/fs/ext4/ext4_jbd2.c
@@ -2,14 +2,14 @@
2 * Interface between ext4 and JBD 2 * Interface between ext4 and JBD
3 */ 3 */
4 4
5#include <linux/ext4_jbd2.h> 5#include "ext4_jbd2.h"
6 6
7int __ext4_journal_get_undo_access(const char *where, handle_t *handle, 7int __ext4_journal_get_undo_access(const char *where, handle_t *handle,
8 struct buffer_head *bh) 8 struct buffer_head *bh)
9{ 9{
10 int err = jbd2_journal_get_undo_access(handle, bh); 10 int err = jbd2_journal_get_undo_access(handle, bh);
11 if (err) 11 if (err)
12 ext4_journal_abort_handle(where, __FUNCTION__, bh, handle,err); 12 ext4_journal_abort_handle(where, __func__, bh, handle, err);
13 return err; 13 return err;
14} 14}
15 15
@@ -18,7 +18,7 @@ int __ext4_journal_get_write_access(const char *where, handle_t *handle,
18{ 18{
19 int err = jbd2_journal_get_write_access(handle, bh); 19 int err = jbd2_journal_get_write_access(handle, bh);
20 if (err) 20 if (err)
21 ext4_journal_abort_handle(where, __FUNCTION__, bh, handle,err); 21 ext4_journal_abort_handle(where, __func__, bh, handle, err);
22 return err; 22 return err;
23} 23}
24 24
@@ -27,7 +27,7 @@ int __ext4_journal_forget(const char *where, handle_t *handle,
27{ 27{
28 int err = jbd2_journal_forget(handle, bh); 28 int err = jbd2_journal_forget(handle, bh);
29 if (err) 29 if (err)
30 ext4_journal_abort_handle(where, __FUNCTION__, bh, handle,err); 30 ext4_journal_abort_handle(where, __func__, bh, handle, err);
31 return err; 31 return err;
32} 32}
33 33
@@ -36,7 +36,7 @@ int __ext4_journal_revoke(const char *where, handle_t *handle,
36{ 36{
37 int err = jbd2_journal_revoke(handle, blocknr, bh); 37 int err = jbd2_journal_revoke(handle, blocknr, bh);
38 if (err) 38 if (err)
39 ext4_journal_abort_handle(where, __FUNCTION__, bh, handle,err); 39 ext4_journal_abort_handle(where, __func__, bh, handle, err);
40 return err; 40 return err;
41} 41}
42 42
@@ -45,7 +45,7 @@ int __ext4_journal_get_create_access(const char *where,
45{ 45{
46 int err = jbd2_journal_get_create_access(handle, bh); 46 int err = jbd2_journal_get_create_access(handle, bh);
47 if (err) 47 if (err)
48 ext4_journal_abort_handle(where, __FUNCTION__, bh, handle,err); 48 ext4_journal_abort_handle(where, __func__, bh, handle, err);
49 return err; 49 return err;
50} 50}
51 51
@@ -54,6 +54,6 @@ int __ext4_journal_dirty_metadata(const char *where,
54{ 54{
55 int err = jbd2_journal_dirty_metadata(handle, bh); 55 int err = jbd2_journal_dirty_metadata(handle, bh);
56 if (err) 56 if (err)
57 ext4_journal_abort_handle(where, __FUNCTION__, bh, handle,err); 57 ext4_journal_abort_handle(where, __func__, bh, handle, err);
58 return err; 58 return err;
59} 59}
diff --git a/fs/ext4/ext4_jbd2.h b/fs/ext4/ext4_jbd2.h
new file mode 100644
index 000000000000..9255a7d28b24
--- /dev/null
+++ b/fs/ext4/ext4_jbd2.h
@@ -0,0 +1,231 @@
1/*
2 * ext4_jbd2.h
3 *
4 * Written by Stephen C. Tweedie <sct@redhat.com>, 1999
5 *
6 * Copyright 1998--1999 Red Hat corp --- All Rights Reserved
7 *
8 * This file is part of the Linux kernel and is made available under
9 * the terms of the GNU General Public License, version 2, or at your
10 * option, any later version, incorporated herein by reference.
11 *
12 * Ext4-specific journaling extensions.
13 */
14
15#ifndef _EXT4_JBD2_H
16#define _EXT4_JBD2_H
17
18#include <linux/fs.h>
19#include <linux/jbd2.h>
20#include "ext4.h"
21
22#define EXT4_JOURNAL(inode) (EXT4_SB((inode)->i_sb)->s_journal)
23
24/* Define the number of blocks we need to account to a transaction to
25 * modify one block of data.
26 *
27 * We may have to touch one inode, one bitmap buffer, up to three
28 * indirection blocks, the group and superblock summaries, and the data
29 * block to complete the transaction.
30 *
31 * For extents-enabled fs we may have to allocate and modify up to
32 * 5 levels of tree + root which are stored in the inode. */
33
34#define EXT4_SINGLEDATA_TRANS_BLOCKS(sb) \
35 (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS) \
36 || test_opt(sb, EXTENTS) ? 27U : 8U)
37
38/* Extended attribute operations touch at most two data buffers,
39 * two bitmap buffers, and two group summaries, in addition to the inode
40 * and the superblock, which are already accounted for. */
41
42#define EXT4_XATTR_TRANS_BLOCKS 6U
43
44/* Define the minimum size for a transaction which modifies data. This
45 * needs to take into account the fact that we may end up modifying two
46 * quota files too (one for the group, one for the user quota). The
47 * superblock only gets updated once, of course, so don't bother
48 * counting that again for the quota updates. */
49
50#define EXT4_DATA_TRANS_BLOCKS(sb) (EXT4_SINGLEDATA_TRANS_BLOCKS(sb) + \
51 EXT4_XATTR_TRANS_BLOCKS - 2 + \
52 2*EXT4_QUOTA_TRANS_BLOCKS(sb))
53
54/* Delete operations potentially hit one directory's namespace plus an
55 * entire inode, plus arbitrary amounts of bitmap/indirection data. Be
56 * generous. We can grow the delete transaction later if necessary. */
57
58#define EXT4_DELETE_TRANS_BLOCKS(sb) (2 * EXT4_DATA_TRANS_BLOCKS(sb) + 64)
59
60/* Define an arbitrary limit for the amount of data we will anticipate
61 * writing to any given transaction. For unbounded transactions such as
62 * write(2) and truncate(2) we can write more than this, but we always
63 * start off at the maximum transaction size and grow the transaction
64 * optimistically as we go. */
65
66#define EXT4_MAX_TRANS_DATA 64U
67
68/* We break up a large truncate or write transaction once the handle's
69 * buffer credits gets this low, we need either to extend the
70 * transaction or to start a new one. Reserve enough space here for
71 * inode, bitmap, superblock, group and indirection updates for at least
72 * one block, plus two quota updates. Quota allocations are not
73 * needed. */
74
75#define EXT4_RESERVE_TRANS_BLOCKS 12U
76
77#define EXT4_INDEX_EXTRA_TRANS_BLOCKS 8
78
79#ifdef CONFIG_QUOTA
80/* Amount of blocks needed for quota update - we know that the structure was
81 * allocated so we need to update only inode+data */
82#define EXT4_QUOTA_TRANS_BLOCKS(sb) (test_opt(sb, QUOTA) ? 2 : 0)
83/* Amount of blocks needed for quota insert/delete - we do some block writes
84 * but inode, sb and group updates are done only once */
85#define EXT4_QUOTA_INIT_BLOCKS(sb) (test_opt(sb, QUOTA) ? (DQUOT_INIT_ALLOC*\
86 (EXT4_SINGLEDATA_TRANS_BLOCKS(sb)-3)+3+DQUOT_INIT_REWRITE) : 0)
87#define EXT4_QUOTA_DEL_BLOCKS(sb) (test_opt(sb, QUOTA) ? (DQUOT_DEL_ALLOC*\
88 (EXT4_SINGLEDATA_TRANS_BLOCKS(sb)-3)+3+DQUOT_DEL_REWRITE) : 0)
89#else
90#define EXT4_QUOTA_TRANS_BLOCKS(sb) 0
91#define EXT4_QUOTA_INIT_BLOCKS(sb) 0
92#define EXT4_QUOTA_DEL_BLOCKS(sb) 0
93#endif
94
95int
96ext4_mark_iloc_dirty(handle_t *handle,
97 struct inode *inode,
98 struct ext4_iloc *iloc);
99
100/*
101 * On success, We end up with an outstanding reference count against
102 * iloc->bh. This _must_ be cleaned up later.
103 */
104
105int ext4_reserve_inode_write(handle_t *handle, struct inode *inode,
106 struct ext4_iloc *iloc);
107
108int ext4_mark_inode_dirty(handle_t *handle, struct inode *inode);
109
110/*
111 * Wrapper functions with which ext4 calls into JBD. The intent here is
112 * to allow these to be turned into appropriate stubs so ext4 can control
113 * ext2 filesystems, so ext2+ext4 systems only nee one fs. This work hasn't
114 * been done yet.
115 */
116
117static inline void ext4_journal_release_buffer(handle_t *handle,
118 struct buffer_head *bh)
119{
120 jbd2_journal_release_buffer(handle, bh);
121}
122
123void ext4_journal_abort_handle(const char *caller, const char *err_fn,
124 struct buffer_head *bh, handle_t *handle, int err);
125
126int __ext4_journal_get_undo_access(const char *where, handle_t *handle,
127 struct buffer_head *bh);
128
129int __ext4_journal_get_write_access(const char *where, handle_t *handle,
130 struct buffer_head *bh);
131
132int __ext4_journal_forget(const char *where, handle_t *handle,
133 struct buffer_head *bh);
134
135int __ext4_journal_revoke(const char *where, handle_t *handle,
136 ext4_fsblk_t blocknr, struct buffer_head *bh);
137
138int __ext4_journal_get_create_access(const char *where,
139 handle_t *handle, struct buffer_head *bh);
140
141int __ext4_journal_dirty_metadata(const char *where,
142 handle_t *handle, struct buffer_head *bh);
143
144#define ext4_journal_get_undo_access(handle, bh) \
145 __ext4_journal_get_undo_access(__FUNCTION__, (handle), (bh))
146#define ext4_journal_get_write_access(handle, bh) \
147 __ext4_journal_get_write_access(__FUNCTION__, (handle), (bh))
148#define ext4_journal_revoke(handle, blocknr, bh) \
149 __ext4_journal_revoke(__FUNCTION__, (handle), (blocknr), (bh))
150#define ext4_journal_get_create_access(handle, bh) \
151 __ext4_journal_get_create_access(__FUNCTION__, (handle), (bh))
152#define ext4_journal_dirty_metadata(handle, bh) \
153 __ext4_journal_dirty_metadata(__FUNCTION__, (handle), (bh))
154#define ext4_journal_forget(handle, bh) \
155 __ext4_journal_forget(__FUNCTION__, (handle), (bh))
156
157int ext4_journal_dirty_data(handle_t *handle, struct buffer_head *bh);
158
159handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks);
160int __ext4_journal_stop(const char *where, handle_t *handle);
161
162static inline handle_t *ext4_journal_start(struct inode *inode, int nblocks)
163{
164 return ext4_journal_start_sb(inode->i_sb, nblocks);
165}
166
167#define ext4_journal_stop(handle) \
168 __ext4_journal_stop(__FUNCTION__, (handle))
169
170static inline handle_t *ext4_journal_current_handle(void)
171{
172 return journal_current_handle();
173}
174
175static inline int ext4_journal_extend(handle_t *handle, int nblocks)
176{
177 return jbd2_journal_extend(handle, nblocks);
178}
179
180static inline int ext4_journal_restart(handle_t *handle, int nblocks)
181{
182 return jbd2_journal_restart(handle, nblocks);
183}
184
185static inline int ext4_journal_blocks_per_page(struct inode *inode)
186{
187 return jbd2_journal_blocks_per_page(inode);
188}
189
190static inline int ext4_journal_force_commit(journal_t *journal)
191{
192 return jbd2_journal_force_commit(journal);
193}
194
195/* super.c */
196int ext4_force_commit(struct super_block *sb);
197
198static inline int ext4_should_journal_data(struct inode *inode)
199{
200 if (!S_ISREG(inode->i_mode))
201 return 1;
202 if (test_opt(inode->i_sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
203 return 1;
204 if (EXT4_I(inode)->i_flags & EXT4_JOURNAL_DATA_FL)
205 return 1;
206 return 0;
207}
208
209static inline int ext4_should_order_data(struct inode *inode)
210{
211 if (!S_ISREG(inode->i_mode))
212 return 0;
213 if (EXT4_I(inode)->i_flags & EXT4_JOURNAL_DATA_FL)
214 return 0;
215 if (test_opt(inode->i_sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
216 return 1;
217 return 0;
218}
219
220static inline int ext4_should_writeback_data(struct inode *inode)
221{
222 if (!S_ISREG(inode->i_mode))
223 return 0;
224 if (EXT4_I(inode)->i_flags & EXT4_JOURNAL_DATA_FL)
225 return 0;
226 if (test_opt(inode->i_sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
227 return 1;
228 return 0;
229}
230
231#endif /* _EXT4_JBD2_H */
diff --git a/fs/ext4/ext4_sb.h b/fs/ext4/ext4_sb.h
new file mode 100644
index 000000000000..5802e69f2191
--- /dev/null
+++ b/fs/ext4/ext4_sb.h
@@ -0,0 +1,148 @@
1/*
2 * ext4_sb.h
3 *
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
8 *
9 * from
10 *
11 * linux/include/linux/minix_fs_sb.h
12 *
13 * Copyright (C) 1991, 1992 Linus Torvalds
14 */
15
16#ifndef _EXT4_SB
17#define _EXT4_SB
18
19#ifdef __KERNEL__
20#include <linux/timer.h>
21#include <linux/wait.h>
22#include <linux/blockgroup_lock.h>
23#include <linux/percpu_counter.h>
24#endif
25#include <linux/rbtree.h>
26
27/*
28 * third extended-fs super-block data in memory
29 */
30struct ext4_sb_info {
31 unsigned long s_desc_size; /* Size of a group descriptor in bytes */
32 unsigned long s_inodes_per_block;/* Number of inodes per block */
33 unsigned long s_blocks_per_group;/* Number of blocks in a group */
34 unsigned long s_inodes_per_group;/* Number of inodes in a group */
35 unsigned long s_itb_per_group; /* Number of inode table blocks per group */
36 unsigned long s_gdb_count; /* Number of group descriptor blocks */
37 unsigned long s_desc_per_block; /* Number of group descriptors per block */
38 ext4_group_t s_groups_count; /* Number of groups in the fs */
39 unsigned long s_overhead_last; /* Last calculated overhead */
40 unsigned long s_blocks_last; /* Last seen block count */
41 loff_t s_bitmap_maxbytes; /* max bytes for bitmap files */
42 struct buffer_head * s_sbh; /* Buffer containing the super block */
43 struct ext4_super_block * s_es; /* Pointer to the super block in the buffer */
44 struct buffer_head ** s_group_desc;
45 unsigned long s_mount_opt;
46 ext4_fsblk_t s_sb_block;
47 uid_t s_resuid;
48 gid_t s_resgid;
49 unsigned short s_mount_state;
50 unsigned short s_pad;
51 int s_addr_per_block_bits;
52 int s_desc_per_block_bits;
53 int s_inode_size;
54 int s_first_ino;
55 spinlock_t s_next_gen_lock;
56 u32 s_next_generation;
57 u32 s_hash_seed[4];
58 int s_def_hash_version;
59 struct percpu_counter s_freeblocks_counter;
60 struct percpu_counter s_freeinodes_counter;
61 struct percpu_counter s_dirs_counter;
62 struct blockgroup_lock s_blockgroup_lock;
63
64 /* root of the per fs reservation window tree */
65 spinlock_t s_rsv_window_lock;
66 struct rb_root s_rsv_window_root;
67 struct ext4_reserve_window_node s_rsv_window_head;
68
69 /* Journaling */
70 struct inode * s_journal_inode;
71 struct journal_s * s_journal;
72 struct list_head s_orphan;
73 unsigned long s_commit_interval;
74 struct block_device *journal_bdev;
75#ifdef CONFIG_JBD2_DEBUG
76 struct timer_list turn_ro_timer; /* For turning read-only (crash simulation) */
77 wait_queue_head_t ro_wait_queue; /* For people waiting for the fs to go read-only */
78#endif
79#ifdef CONFIG_QUOTA
80 char *s_qf_names[MAXQUOTAS]; /* Names of quota files with journalled quota */
81 int s_jquota_fmt; /* Format of quota to use */
82#endif
83 unsigned int s_want_extra_isize; /* New inodes should reserve # bytes */
84
85#ifdef EXTENTS_STATS
86 /* ext4 extents stats */
87 unsigned long s_ext_min;
88 unsigned long s_ext_max;
89 unsigned long s_depth_max;
90 spinlock_t s_ext_stats_lock;
91 unsigned long s_ext_blocks;
92 unsigned long s_ext_extents;
93#endif
94
95 /* for buddy allocator */
96 struct ext4_group_info ***s_group_info;
97 struct inode *s_buddy_cache;
98 long s_blocks_reserved;
99 spinlock_t s_reserve_lock;
100 struct list_head s_active_transaction;
101 struct list_head s_closed_transaction;
102 struct list_head s_committed_transaction;
103 spinlock_t s_md_lock;
104 tid_t s_last_transaction;
105 unsigned short *s_mb_offsets, *s_mb_maxs;
106
107 /* tunables */
108 unsigned long s_stripe;
109 unsigned long s_mb_stream_request;
110 unsigned long s_mb_max_to_scan;
111 unsigned long s_mb_min_to_scan;
112 unsigned long s_mb_stats;
113 unsigned long s_mb_order2_reqs;
114 unsigned long s_mb_group_prealloc;
115 /* where last allocation was done - for stream allocation */
116 unsigned long s_mb_last_group;
117 unsigned long s_mb_last_start;
118
119 /* history to debug policy */
120 struct ext4_mb_history *s_mb_history;
121 int s_mb_history_cur;
122 int s_mb_history_max;
123 int s_mb_history_num;
124 struct proc_dir_entry *s_mb_proc;
125 spinlock_t s_mb_history_lock;
126 int s_mb_history_filter;
127
128 /* stats for buddy allocator */
129 spinlock_t s_mb_pa_lock;
130 atomic_t s_bal_reqs; /* number of reqs with len > 1 */
131 atomic_t s_bal_success; /* we found long enough chunks */
132 atomic_t s_bal_allocated; /* in blocks */
133 atomic_t s_bal_ex_scanned; /* total extents scanned */
134 atomic_t s_bal_goals; /* goal hits */
135 atomic_t s_bal_breaks; /* too long searches */
136 atomic_t s_bal_2orders; /* 2^order hits */
137 spinlock_t s_bal_lock;
138 unsigned long s_mb_buddies_generated;
139 unsigned long long s_mb_generation_time;
140 atomic_t s_mb_lost_chunks;
141 atomic_t s_mb_preallocated;
142 atomic_t s_mb_discarded;
143
144 /* locality groups */
145 struct ext4_locality_group *s_locality_groups;
146};
147
148#endif /* _EXT4_SB */
diff --git a/fs/ext4/extents.c b/fs/ext4/extents.c
index 9ae6e67090cd..47929c4e3dae 100644
--- a/fs/ext4/extents.c
+++ b/fs/ext4/extents.c
@@ -32,7 +32,6 @@
32#include <linux/module.h> 32#include <linux/module.h>
33#include <linux/fs.h> 33#include <linux/fs.h>
34#include <linux/time.h> 34#include <linux/time.h>
35#include <linux/ext4_jbd2.h>
36#include <linux/jbd2.h> 35#include <linux/jbd2.h>
37#include <linux/highuid.h> 36#include <linux/highuid.h>
38#include <linux/pagemap.h> 37#include <linux/pagemap.h>
@@ -40,8 +39,9 @@
40#include <linux/string.h> 39#include <linux/string.h>
41#include <linux/slab.h> 40#include <linux/slab.h>
42#include <linux/falloc.h> 41#include <linux/falloc.h>
43#include <linux/ext4_fs_extents.h>
44#include <asm/uaccess.h> 42#include <asm/uaccess.h>
43#include "ext4_jbd2.h"
44#include "ext4_extents.h"
45 45
46 46
47/* 47/*
@@ -308,7 +308,7 @@ corrupted:
308} 308}
309 309
310#define ext4_ext_check_header(inode, eh, depth) \ 310#define ext4_ext_check_header(inode, eh, depth) \
311 __ext4_ext_check_header(__FUNCTION__, inode, eh, depth) 311 __ext4_ext_check_header(__func__, inode, eh, depth)
312 312
313#ifdef EXT_DEBUG 313#ifdef EXT_DEBUG
314static void ext4_ext_show_path(struct inode *inode, struct ext4_ext_path *path) 314static void ext4_ext_show_path(struct inode *inode, struct ext4_ext_path *path)
@@ -614,7 +614,7 @@ static int ext4_ext_insert_index(handle_t *handle, struct inode *inode,
614 614
615 ix->ei_block = cpu_to_le32(logical); 615 ix->ei_block = cpu_to_le32(logical);
616 ext4_idx_store_pblock(ix, ptr); 616 ext4_idx_store_pblock(ix, ptr);
617 curp->p_hdr->eh_entries = cpu_to_le16(le16_to_cpu(curp->p_hdr->eh_entries)+1); 617 le16_add_cpu(&curp->p_hdr->eh_entries, 1);
618 618
619 BUG_ON(le16_to_cpu(curp->p_hdr->eh_entries) 619 BUG_ON(le16_to_cpu(curp->p_hdr->eh_entries)
620 > le16_to_cpu(curp->p_hdr->eh_max)); 620 > le16_to_cpu(curp->p_hdr->eh_max));
@@ -736,7 +736,7 @@ static int ext4_ext_split(handle_t *handle, struct inode *inode,
736 } 736 }
737 if (m) { 737 if (m) {
738 memmove(ex, path[depth].p_ext-m, sizeof(struct ext4_extent)*m); 738 memmove(ex, path[depth].p_ext-m, sizeof(struct ext4_extent)*m);
739 neh->eh_entries = cpu_to_le16(le16_to_cpu(neh->eh_entries)+m); 739 le16_add_cpu(&neh->eh_entries, m);
740 } 740 }
741 741
742 set_buffer_uptodate(bh); 742 set_buffer_uptodate(bh);
@@ -753,8 +753,7 @@ static int ext4_ext_split(handle_t *handle, struct inode *inode,
753 err = ext4_ext_get_access(handle, inode, path + depth); 753 err = ext4_ext_get_access(handle, inode, path + depth);
754 if (err) 754 if (err)
755 goto cleanup; 755 goto cleanup;
756 path[depth].p_hdr->eh_entries = 756 le16_add_cpu(&path[depth].p_hdr->eh_entries, -m);
757 cpu_to_le16(le16_to_cpu(path[depth].p_hdr->eh_entries)-m);
758 err = ext4_ext_dirty(handle, inode, path + depth); 757 err = ext4_ext_dirty(handle, inode, path + depth);
759 if (err) 758 if (err)
760 goto cleanup; 759 goto cleanup;
@@ -817,8 +816,7 @@ static int ext4_ext_split(handle_t *handle, struct inode *inode,
817 if (m) { 816 if (m) {
818 memmove(++fidx, path[i].p_idx - m, 817 memmove(++fidx, path[i].p_idx - m,
819 sizeof(struct ext4_extent_idx) * m); 818 sizeof(struct ext4_extent_idx) * m);
820 neh->eh_entries = 819 le16_add_cpu(&neh->eh_entries, m);
821 cpu_to_le16(le16_to_cpu(neh->eh_entries) + m);
822 } 820 }
823 set_buffer_uptodate(bh); 821 set_buffer_uptodate(bh);
824 unlock_buffer(bh); 822 unlock_buffer(bh);
@@ -834,7 +832,7 @@ static int ext4_ext_split(handle_t *handle, struct inode *inode,
834 err = ext4_ext_get_access(handle, inode, path + i); 832 err = ext4_ext_get_access(handle, inode, path + i);
835 if (err) 833 if (err)
836 goto cleanup; 834 goto cleanup;
837 path[i].p_hdr->eh_entries = cpu_to_le16(le16_to_cpu(path[i].p_hdr->eh_entries)-m); 835 le16_add_cpu(&path[i].p_hdr->eh_entries, -m);
838 err = ext4_ext_dirty(handle, inode, path + i); 836 err = ext4_ext_dirty(handle, inode, path + i);
839 if (err) 837 if (err)
840 goto cleanup; 838 goto cleanup;
@@ -1369,7 +1367,7 @@ int ext4_ext_try_to_merge(struct inode *inode,
1369 * sizeof(struct ext4_extent); 1367 * sizeof(struct ext4_extent);
1370 memmove(ex + 1, ex + 2, len); 1368 memmove(ex + 1, ex + 2, len);
1371 } 1369 }
1372 eh->eh_entries = cpu_to_le16(le16_to_cpu(eh->eh_entries) - 1); 1370 le16_add_cpu(&eh->eh_entries, -1);
1373 merge_done = 1; 1371 merge_done = 1;
1374 WARN_ON(eh->eh_entries == 0); 1372 WARN_ON(eh->eh_entries == 0);
1375 if (!eh->eh_entries) 1373 if (!eh->eh_entries)
@@ -1560,7 +1558,7 @@ has_space:
1560 path[depth].p_ext = nearex; 1558 path[depth].p_ext = nearex;
1561 } 1559 }
1562 1560
1563 eh->eh_entries = cpu_to_le16(le16_to_cpu(eh->eh_entries)+1); 1561 le16_add_cpu(&eh->eh_entries, 1);
1564 nearex = path[depth].p_ext; 1562 nearex = path[depth].p_ext;
1565 nearex->ee_block = newext->ee_block; 1563 nearex->ee_block = newext->ee_block;
1566 ext4_ext_store_pblock(nearex, ext_pblock(newext)); 1564 ext4_ext_store_pblock(nearex, ext_pblock(newext));
@@ -1699,7 +1697,7 @@ static int ext4_ext_rm_idx(handle_t *handle, struct inode *inode,
1699 err = ext4_ext_get_access(handle, inode, path); 1697 err = ext4_ext_get_access(handle, inode, path);
1700 if (err) 1698 if (err)
1701 return err; 1699 return err;
1702 path->p_hdr->eh_entries = cpu_to_le16(le16_to_cpu(path->p_hdr->eh_entries)-1); 1700 le16_add_cpu(&path->p_hdr->eh_entries, -1);
1703 err = ext4_ext_dirty(handle, inode, path); 1701 err = ext4_ext_dirty(handle, inode, path);
1704 if (err) 1702 if (err)
1705 return err; 1703 return err;
@@ -1902,7 +1900,7 @@ ext4_ext_rm_leaf(handle_t *handle, struct inode *inode,
1902 if (num == 0) { 1900 if (num == 0) {
1903 /* this extent is removed; mark slot entirely unused */ 1901 /* this extent is removed; mark slot entirely unused */
1904 ext4_ext_store_pblock(ex, 0); 1902 ext4_ext_store_pblock(ex, 0);
1905 eh->eh_entries = cpu_to_le16(le16_to_cpu(eh->eh_entries)-1); 1903 le16_add_cpu(&eh->eh_entries, -1);
1906 } 1904 }
1907 1905
1908 ex->ee_block = cpu_to_le32(block); 1906 ex->ee_block = cpu_to_le32(block);
@@ -1979,7 +1977,7 @@ static int ext4_ext_remove_space(struct inode *inode, ext4_lblk_t start)
1979 * We start scanning from right side, freeing all the blocks 1977 * We start scanning from right side, freeing all the blocks
1980 * after i_size and walking into the tree depth-wise. 1978 * after i_size and walking into the tree depth-wise.
1981 */ 1979 */
1982 path = kzalloc(sizeof(struct ext4_ext_path) * (depth + 1), GFP_KERNEL); 1980 path = kzalloc(sizeof(struct ext4_ext_path) * (depth + 1), GFP_NOFS);
1983 if (path == NULL) { 1981 if (path == NULL) {
1984 ext4_journal_stop(handle); 1982 ext4_journal_stop(handle);
1985 return -ENOMEM; 1983 return -ENOMEM;
@@ -2138,6 +2136,82 @@ void ext4_ext_release(struct super_block *sb)
2138#endif 2136#endif
2139} 2137}
2140 2138
2139static void bi_complete(struct bio *bio, int error)
2140{
2141 complete((struct completion *)bio->bi_private);
2142}
2143
2144/* FIXME!! we need to try to merge to left or right after zero-out */
2145static int ext4_ext_zeroout(struct inode *inode, struct ext4_extent *ex)
2146{
2147 int ret = -EIO;
2148 struct bio *bio;
2149 int blkbits, blocksize;
2150 sector_t ee_pblock;
2151 struct completion event;
2152 unsigned int ee_len, len, done, offset;
2153
2154
2155 blkbits = inode->i_blkbits;
2156 blocksize = inode->i_sb->s_blocksize;
2157 ee_len = ext4_ext_get_actual_len(ex);
2158 ee_pblock = ext_pblock(ex);
2159
2160 /* convert ee_pblock to 512 byte sectors */
2161 ee_pblock = ee_pblock << (blkbits - 9);
2162
2163 while (ee_len > 0) {
2164
2165 if (ee_len > BIO_MAX_PAGES)
2166 len = BIO_MAX_PAGES;
2167 else
2168 len = ee_len;
2169
2170 bio = bio_alloc(GFP_NOIO, len);
2171 if (!bio)
2172 return -ENOMEM;
2173 bio->bi_sector = ee_pblock;
2174 bio->bi_bdev = inode->i_sb->s_bdev;
2175
2176 done = 0;
2177 offset = 0;
2178 while (done < len) {
2179 ret = bio_add_page(bio, ZERO_PAGE(0),
2180 blocksize, offset);
2181 if (ret != blocksize) {
2182 /*
2183 * We can't add any more pages because of
2184 * hardware limitations. Start a new bio.
2185 */
2186 break;
2187 }
2188 done++;
2189 offset += blocksize;
2190 if (offset >= PAGE_CACHE_SIZE)
2191 offset = 0;
2192 }
2193
2194 init_completion(&event);
2195 bio->bi_private = &event;
2196 bio->bi_end_io = bi_complete;
2197 submit_bio(WRITE, bio);
2198 wait_for_completion(&event);
2199
2200 if (test_bit(BIO_UPTODATE, &bio->bi_flags))
2201 ret = 0;
2202 else {
2203 ret = -EIO;
2204 break;
2205 }
2206 bio_put(bio);
2207 ee_len -= done;
2208 ee_pblock += done << (blkbits - 9);
2209 }
2210 return ret;
2211}
2212
2213#define EXT4_EXT_ZERO_LEN 7
2214
2141/* 2215/*
2142 * This function is called by ext4_ext_get_blocks() if someone tries to write 2216 * This function is called by ext4_ext_get_blocks() if someone tries to write
2143 * to an uninitialized extent. It may result in splitting the uninitialized 2217 * to an uninitialized extent. It may result in splitting the uninitialized
@@ -2154,7 +2228,7 @@ static int ext4_ext_convert_to_initialized(handle_t *handle,
2154 ext4_lblk_t iblock, 2228 ext4_lblk_t iblock,
2155 unsigned long max_blocks) 2229 unsigned long max_blocks)
2156{ 2230{
2157 struct ext4_extent *ex, newex; 2231 struct ext4_extent *ex, newex, orig_ex;
2158 struct ext4_extent *ex1 = NULL; 2232 struct ext4_extent *ex1 = NULL;
2159 struct ext4_extent *ex2 = NULL; 2233 struct ext4_extent *ex2 = NULL;
2160 struct ext4_extent *ex3 = NULL; 2234 struct ext4_extent *ex3 = NULL;
@@ -2173,10 +2247,26 @@ static int ext4_ext_convert_to_initialized(handle_t *handle,
2173 allocated = ee_len - (iblock - ee_block); 2247 allocated = ee_len - (iblock - ee_block);
2174 newblock = iblock - ee_block + ext_pblock(ex); 2248 newblock = iblock - ee_block + ext_pblock(ex);
2175 ex2 = ex; 2249 ex2 = ex;
2250 orig_ex.ee_block = ex->ee_block;
2251 orig_ex.ee_len = cpu_to_le16(ee_len);
2252 ext4_ext_store_pblock(&orig_ex, ext_pblock(ex));
2176 2253
2177 err = ext4_ext_get_access(handle, inode, path + depth); 2254 err = ext4_ext_get_access(handle, inode, path + depth);
2178 if (err) 2255 if (err)
2179 goto out; 2256 goto out;
2257 /* If extent has less than 2*EXT4_EXT_ZERO_LEN zerout directly */
2258 if (ee_len <= 2*EXT4_EXT_ZERO_LEN) {
2259 err = ext4_ext_zeroout(inode, &orig_ex);
2260 if (err)
2261 goto fix_extent_len;
2262 /* update the extent length and mark as initialized */
2263 ex->ee_block = orig_ex.ee_block;
2264 ex->ee_len = orig_ex.ee_len;
2265 ext4_ext_store_pblock(ex, ext_pblock(&orig_ex));
2266 ext4_ext_dirty(handle, inode, path + depth);
2267 /* zeroed the full extent */
2268 return allocated;
2269 }
2180 2270
2181 /* ex1: ee_block to iblock - 1 : uninitialized */ 2271 /* ex1: ee_block to iblock - 1 : uninitialized */
2182 if (iblock > ee_block) { 2272 if (iblock > ee_block) {
@@ -2195,19 +2285,103 @@ static int ext4_ext_convert_to_initialized(handle_t *handle,
2195 /* ex3: to ee_block + ee_len : uninitialised */ 2285 /* ex3: to ee_block + ee_len : uninitialised */
2196 if (allocated > max_blocks) { 2286 if (allocated > max_blocks) {
2197 unsigned int newdepth; 2287 unsigned int newdepth;
2288 /* If extent has less than EXT4_EXT_ZERO_LEN zerout directly */
2289 if (allocated <= EXT4_EXT_ZERO_LEN) {
2290 /* Mark first half uninitialized.
2291 * Mark second half initialized and zero out the
2292 * initialized extent
2293 */
2294 ex->ee_block = orig_ex.ee_block;
2295 ex->ee_len = cpu_to_le16(ee_len - allocated);
2296 ext4_ext_mark_uninitialized(ex);
2297 ext4_ext_store_pblock(ex, ext_pblock(&orig_ex));
2298 ext4_ext_dirty(handle, inode, path + depth);
2299
2300 ex3 = &newex;
2301 ex3->ee_block = cpu_to_le32(iblock);
2302 ext4_ext_store_pblock(ex3, newblock);
2303 ex3->ee_len = cpu_to_le16(allocated);
2304 err = ext4_ext_insert_extent(handle, inode, path, ex3);
2305 if (err == -ENOSPC) {
2306 err = ext4_ext_zeroout(inode, &orig_ex);
2307 if (err)
2308 goto fix_extent_len;
2309 ex->ee_block = orig_ex.ee_block;
2310 ex->ee_len = orig_ex.ee_len;
2311 ext4_ext_store_pblock(ex, ext_pblock(&orig_ex));
2312 ext4_ext_dirty(handle, inode, path + depth);
2313 /* zeroed the full extent */
2314 return allocated;
2315
2316 } else if (err)
2317 goto fix_extent_len;
2318
2319 /*
2320 * We need to zero out the second half because
2321 * an fallocate request can update file size and
2322 * converting the second half to initialized extent
2323 * implies that we can leak some junk data to user
2324 * space.
2325 */
2326 err = ext4_ext_zeroout(inode, ex3);
2327 if (err) {
2328 /*
2329 * We should actually mark the
2330 * second half as uninit and return error
2331 * Insert would have changed the extent
2332 */
2333 depth = ext_depth(inode);
2334 ext4_ext_drop_refs(path);
2335 path = ext4_ext_find_extent(inode,
2336 iblock, path);
2337 if (IS_ERR(path)) {
2338 err = PTR_ERR(path);
2339 return err;
2340 }
2341 ex = path[depth].p_ext;
2342 err = ext4_ext_get_access(handle, inode,
2343 path + depth);
2344 if (err)
2345 return err;
2346 ext4_ext_mark_uninitialized(ex);
2347 ext4_ext_dirty(handle, inode, path + depth);
2348 return err;
2349 }
2350
2351 /* zeroed the second half */
2352 return allocated;
2353 }
2198 ex3 = &newex; 2354 ex3 = &newex;
2199 ex3->ee_block = cpu_to_le32(iblock + max_blocks); 2355 ex3->ee_block = cpu_to_le32(iblock + max_blocks);
2200 ext4_ext_store_pblock(ex3, newblock + max_blocks); 2356 ext4_ext_store_pblock(ex3, newblock + max_blocks);
2201 ex3->ee_len = cpu_to_le16(allocated - max_blocks); 2357 ex3->ee_len = cpu_to_le16(allocated - max_blocks);
2202 ext4_ext_mark_uninitialized(ex3); 2358 ext4_ext_mark_uninitialized(ex3);
2203 err = ext4_ext_insert_extent(handle, inode, path, ex3); 2359 err = ext4_ext_insert_extent(handle, inode, path, ex3);
2204 if (err) 2360 if (err == -ENOSPC) {
2205 goto out; 2361 err = ext4_ext_zeroout(inode, &orig_ex);
2362 if (err)
2363 goto fix_extent_len;
2364 /* update the extent length and mark as initialized */
2365 ex->ee_block = orig_ex.ee_block;
2366 ex->ee_len = orig_ex.ee_len;
2367 ext4_ext_store_pblock(ex, ext_pblock(&orig_ex));
2368 ext4_ext_dirty(handle, inode, path + depth);
2369 /* zeroed the full extent */
2370 return allocated;
2371
2372 } else if (err)
2373 goto fix_extent_len;
2206 /* 2374 /*
2207 * The depth, and hence eh & ex might change 2375 * The depth, and hence eh & ex might change
2208 * as part of the insert above. 2376 * as part of the insert above.
2209 */ 2377 */
2210 newdepth = ext_depth(inode); 2378 newdepth = ext_depth(inode);
2379 /*
2380 * update the extent length after successfull insert of the
2381 * split extent
2382 */
2383 orig_ex.ee_len = cpu_to_le16(ee_len -
2384 ext4_ext_get_actual_len(ex3));
2211 if (newdepth != depth) { 2385 if (newdepth != depth) {
2212 depth = newdepth; 2386 depth = newdepth;
2213 ext4_ext_drop_refs(path); 2387 ext4_ext_drop_refs(path);
@@ -2226,6 +2400,24 @@ static int ext4_ext_convert_to_initialized(handle_t *handle,
2226 goto out; 2400 goto out;
2227 } 2401 }
2228 allocated = max_blocks; 2402 allocated = max_blocks;
2403
2404 /* If extent has less than EXT4_EXT_ZERO_LEN and we are trying
2405 * to insert a extent in the middle zerout directly
2406 * otherwise give the extent a chance to merge to left
2407 */
2408 if (le16_to_cpu(orig_ex.ee_len) <= EXT4_EXT_ZERO_LEN &&
2409 iblock != ee_block) {
2410 err = ext4_ext_zeroout(inode, &orig_ex);
2411 if (err)
2412 goto fix_extent_len;
2413 /* update the extent length and mark as initialized */
2414 ex->ee_block = orig_ex.ee_block;
2415 ex->ee_len = orig_ex.ee_len;
2416 ext4_ext_store_pblock(ex, ext_pblock(&orig_ex));
2417 ext4_ext_dirty(handle, inode, path + depth);
2418 /* zero out the first half */
2419 return allocated;
2420 }
2229 } 2421 }
2230 /* 2422 /*
2231 * If there was a change of depth as part of the 2423 * If there was a change of depth as part of the
@@ -2282,8 +2474,29 @@ static int ext4_ext_convert_to_initialized(handle_t *handle,
2282 goto out; 2474 goto out;
2283insert: 2475insert:
2284 err = ext4_ext_insert_extent(handle, inode, path, &newex); 2476 err = ext4_ext_insert_extent(handle, inode, path, &newex);
2477 if (err == -ENOSPC) {
2478 err = ext4_ext_zeroout(inode, &orig_ex);
2479 if (err)
2480 goto fix_extent_len;
2481 /* update the extent length and mark as initialized */
2482 ex->ee_block = orig_ex.ee_block;
2483 ex->ee_len = orig_ex.ee_len;
2484 ext4_ext_store_pblock(ex, ext_pblock(&orig_ex));
2485 ext4_ext_dirty(handle, inode, path + depth);
2486 /* zero out the first half */
2487 return allocated;
2488 } else if (err)
2489 goto fix_extent_len;
2285out: 2490out:
2286 return err ? err : allocated; 2491 return err ? err : allocated;
2492
2493fix_extent_len:
2494 ex->ee_block = orig_ex.ee_block;
2495 ex->ee_len = orig_ex.ee_len;
2496 ext4_ext_store_pblock(ex, ext_pblock(&orig_ex));
2497 ext4_ext_mark_uninitialized(ex);
2498 ext4_ext_dirty(handle, inode, path + depth);
2499 return err;
2287} 2500}
2288 2501
2289/* 2502/*
@@ -2393,8 +2606,20 @@ int ext4_ext_get_blocks(handle_t *handle, struct inode *inode,
2393 } 2606 }
2394 if (create == EXT4_CREATE_UNINITIALIZED_EXT) 2607 if (create == EXT4_CREATE_UNINITIALIZED_EXT)
2395 goto out; 2608 goto out;
2396 if (!create) 2609 if (!create) {
2610 /*
2611 * We have blocks reserved already. We
2612 * return allocated blocks so that delalloc
2613 * won't do block reservation for us. But
2614 * the buffer head will be unmapped so that
2615 * a read from the block returns 0s.
2616 */
2617 if (allocated > max_blocks)
2618 allocated = max_blocks;
2619 /* mark the buffer unwritten */
2620 __set_bit(BH_Unwritten, &bh_result->b_state);
2397 goto out2; 2621 goto out2;
2622 }
2398 2623
2399 ret = ext4_ext_convert_to_initialized(handle, inode, 2624 ret = ext4_ext_convert_to_initialized(handle, inode,
2400 path, iblock, 2625 path, iblock,
@@ -2584,6 +2809,8 @@ out_stop:
2584 ext4_orphan_del(handle, inode); 2809 ext4_orphan_del(handle, inode);
2585 2810
2586 up_write(&EXT4_I(inode)->i_data_sem); 2811 up_write(&EXT4_I(inode)->i_data_sem);
2812 inode->i_mtime = inode->i_ctime = ext4_current_time(inode);
2813 ext4_mark_inode_dirty(handle, inode);
2587 ext4_journal_stop(handle); 2814 ext4_journal_stop(handle);
2588} 2815}
2589 2816
@@ -2608,6 +2835,28 @@ int ext4_ext_writepage_trans_blocks(struct inode *inode, int num)
2608 return needed; 2835 return needed;
2609} 2836}
2610 2837
2838static void ext4_falloc_update_inode(struct inode *inode,
2839 int mode, loff_t new_size, int update_ctime)
2840{
2841 struct timespec now;
2842
2843 if (update_ctime) {
2844 now = current_fs_time(inode->i_sb);
2845 if (!timespec_equal(&inode->i_ctime, &now))
2846 inode->i_ctime = now;
2847 }
2848 /*
2849 * Update only when preallocation was requested beyond
2850 * the file size.
2851 */
2852 if (!(mode & FALLOC_FL_KEEP_SIZE) &&
2853 new_size > i_size_read(inode)) {
2854 i_size_write(inode, new_size);
2855 EXT4_I(inode)->i_disksize = new_size;
2856 }
2857
2858}
2859
2611/* 2860/*
2612 * preallocate space for a file. This implements ext4's fallocate inode 2861 * preallocate space for a file. This implements ext4's fallocate inode
2613 * operation, which gets called from sys_fallocate system call. 2862 * operation, which gets called from sys_fallocate system call.
@@ -2619,8 +2868,8 @@ long ext4_fallocate(struct inode *inode, int mode, loff_t offset, loff_t len)
2619{ 2868{
2620 handle_t *handle; 2869 handle_t *handle;
2621 ext4_lblk_t block; 2870 ext4_lblk_t block;
2871 loff_t new_size;
2622 unsigned long max_blocks; 2872 unsigned long max_blocks;
2623 ext4_fsblk_t nblocks = 0;
2624 int ret = 0; 2873 int ret = 0;
2625 int ret2 = 0; 2874 int ret2 = 0;
2626 int retries = 0; 2875 int retries = 0;
@@ -2639,9 +2888,12 @@ long ext4_fallocate(struct inode *inode, int mode, loff_t offset, loff_t len)
2639 return -ENODEV; 2888 return -ENODEV;
2640 2889
2641 block = offset >> blkbits; 2890 block = offset >> blkbits;
2891 /*
2892 * We can't just convert len to max_blocks because
2893 * If blocksize = 4096 offset = 3072 and len = 2048
2894 */
2642 max_blocks = (EXT4_BLOCK_ALIGN(len + offset, blkbits) >> blkbits) 2895 max_blocks = (EXT4_BLOCK_ALIGN(len + offset, blkbits) >> blkbits)
2643 - block; 2896 - block;
2644
2645 /* 2897 /*
2646 * credits to insert 1 extent into extent tree + buffers to be able to 2898 * credits to insert 1 extent into extent tree + buffers to be able to
2647 * modify 1 super block, 1 block bitmap and 1 group descriptor. 2899 * modify 1 super block, 1 block bitmap and 1 group descriptor.
@@ -2657,7 +2909,6 @@ retry:
2657 ret = PTR_ERR(handle); 2909 ret = PTR_ERR(handle);
2658 break; 2910 break;
2659 } 2911 }
2660
2661 ret = ext4_get_blocks_wrap(handle, inode, block, 2912 ret = ext4_get_blocks_wrap(handle, inode, block,
2662 max_blocks, &map_bh, 2913 max_blocks, &map_bh,
2663 EXT4_CREATE_UNINITIALIZED_EXT, 0); 2914 EXT4_CREATE_UNINITIALIZED_EXT, 0);
@@ -2673,61 +2924,24 @@ retry:
2673 ret2 = ext4_journal_stop(handle); 2924 ret2 = ext4_journal_stop(handle);
2674 break; 2925 break;
2675 } 2926 }
2676 if (ret > 0) { 2927 if ((block + ret) >= (EXT4_BLOCK_ALIGN(offset + len,
2677 /* check wrap through sign-bit/zero here */ 2928 blkbits) >> blkbits))
2678 if ((block + ret) < 0 || (block + ret) < block) { 2929 new_size = offset + len;
2679 ret = -EIO; 2930 else
2680 ext4_mark_inode_dirty(handle, inode); 2931 new_size = (block + ret) << blkbits;
2681 ret2 = ext4_journal_stop(handle);
2682 break;
2683 }
2684 if (buffer_new(&map_bh) && ((block + ret) >
2685 (EXT4_BLOCK_ALIGN(i_size_read(inode), blkbits)
2686 >> blkbits)))
2687 nblocks = nblocks + ret;
2688 }
2689
2690 /* Update ctime if new blocks get allocated */
2691 if (nblocks) {
2692 struct timespec now;
2693
2694 now = current_fs_time(inode->i_sb);
2695 if (!timespec_equal(&inode->i_ctime, &now))
2696 inode->i_ctime = now;
2697 }
2698 2932
2933 ext4_falloc_update_inode(inode, mode, new_size,
2934 buffer_new(&map_bh));
2699 ext4_mark_inode_dirty(handle, inode); 2935 ext4_mark_inode_dirty(handle, inode);
2700 ret2 = ext4_journal_stop(handle); 2936 ret2 = ext4_journal_stop(handle);
2701 if (ret2) 2937 if (ret2)
2702 break; 2938 break;
2703 } 2939 }
2704 2940 if (ret == -ENOSPC &&
2705 if (ret == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries)) 2941 ext4_should_retry_alloc(inode->i_sb, &retries)) {
2942 ret = 0;
2706 goto retry; 2943 goto retry;
2707
2708 /*
2709 * Time to update the file size.
2710 * Update only when preallocation was requested beyond the file size.
2711 */
2712 if (!(mode & FALLOC_FL_KEEP_SIZE) &&
2713 (offset + len) > i_size_read(inode)) {
2714 if (ret > 0) {
2715 /*
2716 * if no error, we assume preallocation succeeded
2717 * completely
2718 */
2719 i_size_write(inode, offset + len);
2720 EXT4_I(inode)->i_disksize = i_size_read(inode);
2721 } else if (ret < 0 && nblocks) {
2722 /* Handle partial allocation scenario */
2723 loff_t newsize;
2724
2725 newsize = (nblocks << blkbits) + i_size_read(inode);
2726 i_size_write(inode, EXT4_BLOCK_ALIGN(newsize, blkbits));
2727 EXT4_I(inode)->i_disksize = i_size_read(inode);
2728 }
2729 } 2944 }
2730
2731 mutex_unlock(&inode->i_mutex); 2945 mutex_unlock(&inode->i_mutex);
2732 return ret > 0 ? ret2 : ret; 2946 return ret > 0 ? ret2 : ret;
2733} 2947}
diff --git a/fs/ext4/file.c b/fs/ext4/file.c
index ac35ec58db55..4159be6366ab 100644
--- a/fs/ext4/file.c
+++ b/fs/ext4/file.c
@@ -21,8 +21,8 @@
21#include <linux/time.h> 21#include <linux/time.h>
22#include <linux/fs.h> 22#include <linux/fs.h>
23#include <linux/jbd2.h> 23#include <linux/jbd2.h>
24#include <linux/ext4_fs.h> 24#include "ext4.h"
25#include <linux/ext4_jbd2.h> 25#include "ext4_jbd2.h"
26#include "xattr.h" 26#include "xattr.h"
27#include "acl.h" 27#include "acl.h"
28 28
@@ -129,7 +129,7 @@ const struct file_operations ext4_file_operations = {
129 .write = do_sync_write, 129 .write = do_sync_write,
130 .aio_read = generic_file_aio_read, 130 .aio_read = generic_file_aio_read,
131 .aio_write = ext4_file_write, 131 .aio_write = ext4_file_write,
132 .ioctl = ext4_ioctl, 132 .unlocked_ioctl = ext4_ioctl,
133#ifdef CONFIG_COMPAT 133#ifdef CONFIG_COMPAT
134 .compat_ioctl = ext4_compat_ioctl, 134 .compat_ioctl = ext4_compat_ioctl,
135#endif 135#endif
diff --git a/fs/ext4/fsync.c b/fs/ext4/fsync.c
index 8d50879d1c2c..1c8ba48d4f8d 100644
--- a/fs/ext4/fsync.c
+++ b/fs/ext4/fsync.c
@@ -27,8 +27,8 @@
27#include <linux/sched.h> 27#include <linux/sched.h>
28#include <linux/writeback.h> 28#include <linux/writeback.h>
29#include <linux/jbd2.h> 29#include <linux/jbd2.h>
30#include <linux/ext4_fs.h> 30#include "ext4.h"
31#include <linux/ext4_jbd2.h> 31#include "ext4_jbd2.h"
32 32
33/* 33/*
34 * akpm: A new design for ext4_sync_file(). 34 * akpm: A new design for ext4_sync_file().
@@ -72,6 +72,9 @@ int ext4_sync_file(struct file * file, struct dentry *dentry, int datasync)
72 goto out; 72 goto out;
73 } 73 }
74 74
75 if (datasync && !(inode->i_state & I_DIRTY_DATASYNC))
76 goto out;
77
75 /* 78 /*
76 * The VFS has written the file data. If the inode is unaltered 79 * The VFS has written the file data. If the inode is unaltered
77 * then we need not start a commit. 80 * then we need not start a commit.
diff --git a/fs/ext4/hash.c b/fs/ext4/hash.c
index 1555024e3b36..1d6329dbe390 100644
--- a/fs/ext4/hash.c
+++ b/fs/ext4/hash.c
@@ -11,8 +11,8 @@
11 11
12#include <linux/fs.h> 12#include <linux/fs.h>
13#include <linux/jbd2.h> 13#include <linux/jbd2.h>
14#include <linux/ext4_fs.h>
15#include <linux/cryptohash.h> 14#include <linux/cryptohash.h>
15#include "ext4.h"
16 16
17#define DELTA 0x9E3779B9 17#define DELTA 0x9E3779B9
18 18
diff --git a/fs/ext4/ialloc.c b/fs/ext4/ialloc.c
index 486e46a3918d..c6efbab0c801 100644
--- a/fs/ext4/ialloc.c
+++ b/fs/ext4/ialloc.c
@@ -15,8 +15,6 @@
15#include <linux/time.h> 15#include <linux/time.h>
16#include <linux/fs.h> 16#include <linux/fs.h>
17#include <linux/jbd2.h> 17#include <linux/jbd2.h>
18#include <linux/ext4_fs.h>
19#include <linux/ext4_jbd2.h>
20#include <linux/stat.h> 18#include <linux/stat.h>
21#include <linux/string.h> 19#include <linux/string.h>
22#include <linux/quotaops.h> 20#include <linux/quotaops.h>
@@ -25,7 +23,8 @@
25#include <linux/bitops.h> 23#include <linux/bitops.h>
26#include <linux/blkdev.h> 24#include <linux/blkdev.h>
27#include <asm/byteorder.h> 25#include <asm/byteorder.h>
28 26#include "ext4.h"
27#include "ext4_jbd2.h"
29#include "xattr.h" 28#include "xattr.h"
30#include "acl.h" 29#include "acl.h"
31#include "group.h" 30#include "group.h"
@@ -75,7 +74,7 @@ unsigned ext4_init_inode_bitmap(struct super_block *sb, struct buffer_head *bh,
75 /* If checksum is bad mark all blocks and inodes use to prevent 74 /* If checksum is bad mark all blocks and inodes use to prevent
76 * allocation, essentially implementing a per-group read-only flag. */ 75 * allocation, essentially implementing a per-group read-only flag. */
77 if (!ext4_group_desc_csum_verify(sbi, block_group, gdp)) { 76 if (!ext4_group_desc_csum_verify(sbi, block_group, gdp)) {
78 ext4_error(sb, __FUNCTION__, "Checksum bad for group %lu\n", 77 ext4_error(sb, __func__, "Checksum bad for group %lu\n",
79 block_group); 78 block_group);
80 gdp->bg_free_blocks_count = 0; 79 gdp->bg_free_blocks_count = 0;
81 gdp->bg_free_inodes_count = 0; 80 gdp->bg_free_inodes_count = 0;
@@ -223,11 +222,9 @@ void ext4_free_inode (handle_t *handle, struct inode * inode)
223 222
224 if (gdp) { 223 if (gdp) {
225 spin_lock(sb_bgl_lock(sbi, block_group)); 224 spin_lock(sb_bgl_lock(sbi, block_group));
226 gdp->bg_free_inodes_count = cpu_to_le16( 225 le16_add_cpu(&gdp->bg_free_inodes_count, 1);
227 le16_to_cpu(gdp->bg_free_inodes_count) + 1);
228 if (is_directory) 226 if (is_directory)
229 gdp->bg_used_dirs_count = cpu_to_le16( 227 le16_add_cpu(&gdp->bg_used_dirs_count, -1);
230 le16_to_cpu(gdp->bg_used_dirs_count) - 1);
231 gdp->bg_checksum = ext4_group_desc_csum(sbi, 228 gdp->bg_checksum = ext4_group_desc_csum(sbi,
232 block_group, gdp); 229 block_group, gdp);
233 spin_unlock(sb_bgl_lock(sbi, block_group)); 230 spin_unlock(sb_bgl_lock(sbi, block_group));
@@ -588,7 +585,7 @@ got:
588 ino++; 585 ino++;
589 if ((group == 0 && ino < EXT4_FIRST_INO(sb)) || 586 if ((group == 0 && ino < EXT4_FIRST_INO(sb)) ||
590 ino > EXT4_INODES_PER_GROUP(sb)) { 587 ino > EXT4_INODES_PER_GROUP(sb)) {
591 ext4_error(sb, __FUNCTION__, 588 ext4_error(sb, __func__,
592 "reserved inode or inode > inodes count - " 589 "reserved inode or inode > inodes count - "
593 "block_group = %lu, inode=%lu", group, 590 "block_group = %lu, inode=%lu", group,
594 ino + group * EXT4_INODES_PER_GROUP(sb)); 591 ino + group * EXT4_INODES_PER_GROUP(sb));
@@ -664,11 +661,9 @@ got:
664 cpu_to_le16(EXT4_INODES_PER_GROUP(sb) - ino); 661 cpu_to_le16(EXT4_INODES_PER_GROUP(sb) - ino);
665 } 662 }
666 663
667 gdp->bg_free_inodes_count = 664 le16_add_cpu(&gdp->bg_free_inodes_count, -1);
668 cpu_to_le16(le16_to_cpu(gdp->bg_free_inodes_count) - 1);
669 if (S_ISDIR(mode)) { 665 if (S_ISDIR(mode)) {
670 gdp->bg_used_dirs_count = 666 le16_add_cpu(&gdp->bg_used_dirs_count, 1);
671 cpu_to_le16(le16_to_cpu(gdp->bg_used_dirs_count) + 1);
672 } 667 }
673 gdp->bg_checksum = ext4_group_desc_csum(sbi, group, gdp); 668 gdp->bg_checksum = ext4_group_desc_csum(sbi, group, gdp);
674 spin_unlock(sb_bgl_lock(sbi, group)); 669 spin_unlock(sb_bgl_lock(sbi, group));
@@ -744,23 +739,24 @@ got:
744 if (err) 739 if (err)
745 goto fail_free_drop; 740 goto fail_free_drop;
746 741
747 err = ext4_mark_inode_dirty(handle, inode);
748 if (err) {
749 ext4_std_error(sb, err);
750 goto fail_free_drop;
751 }
752 if (test_opt(sb, EXTENTS)) { 742 if (test_opt(sb, EXTENTS)) {
753 /* set extent flag only for directory and file */ 743 /* set extent flag only for diretory, file and normal symlink*/
754 if (S_ISDIR(mode) || S_ISREG(mode)) { 744 if (S_ISDIR(mode) || S_ISREG(mode) || S_ISLNK(mode)) {
755 EXT4_I(inode)->i_flags |= EXT4_EXTENTS_FL; 745 EXT4_I(inode)->i_flags |= EXT4_EXTENTS_FL;
756 ext4_ext_tree_init(handle, inode); 746 ext4_ext_tree_init(handle, inode);
757 err = ext4_update_incompat_feature(handle, sb, 747 err = ext4_update_incompat_feature(handle, sb,
758 EXT4_FEATURE_INCOMPAT_EXTENTS); 748 EXT4_FEATURE_INCOMPAT_EXTENTS);
759 if (err) 749 if (err)
760 goto fail; 750 goto fail_free_drop;
761 } 751 }
762 } 752 }
763 753
754 err = ext4_mark_inode_dirty(handle, inode);
755 if (err) {
756 ext4_std_error(sb, err);
757 goto fail_free_drop;
758 }
759
764 ext4_debug("allocating inode %lu\n", inode->i_ino); 760 ext4_debug("allocating inode %lu\n", inode->i_ino);
765 goto really_out; 761 goto really_out;
766fail: 762fail:
@@ -796,7 +792,7 @@ struct inode *ext4_orphan_get(struct super_block *sb, unsigned long ino)
796 792
797 /* Error cases - e2fsck has already cleaned up for us */ 793 /* Error cases - e2fsck has already cleaned up for us */
798 if (ino > max_ino) { 794 if (ino > max_ino) {
799 ext4_warning(sb, __FUNCTION__, 795 ext4_warning(sb, __func__,
800 "bad orphan ino %lu! e2fsck was run?", ino); 796 "bad orphan ino %lu! e2fsck was run?", ino);
801 goto error; 797 goto error;
802 } 798 }
@@ -805,7 +801,7 @@ struct inode *ext4_orphan_get(struct super_block *sb, unsigned long ino)
805 bit = (ino - 1) % EXT4_INODES_PER_GROUP(sb); 801 bit = (ino - 1) % EXT4_INODES_PER_GROUP(sb);
806 bitmap_bh = read_inode_bitmap(sb, block_group); 802 bitmap_bh = read_inode_bitmap(sb, block_group);
807 if (!bitmap_bh) { 803 if (!bitmap_bh) {
808 ext4_warning(sb, __FUNCTION__, 804 ext4_warning(sb, __func__,
809 "inode bitmap error for orphan %lu", ino); 805 "inode bitmap error for orphan %lu", ino);
810 goto error; 806 goto error;
811 } 807 }
@@ -830,7 +826,7 @@ iget_failed:
830 err = PTR_ERR(inode); 826 err = PTR_ERR(inode);
831 inode = NULL; 827 inode = NULL;
832bad_orphan: 828bad_orphan:
833 ext4_warning(sb, __FUNCTION__, 829 ext4_warning(sb, __func__,
834 "bad orphan inode %lu! e2fsck was run?", ino); 830 "bad orphan inode %lu! e2fsck was run?", ino);
835 printk(KERN_NOTICE "ext4_test_bit(bit=%d, block=%llu) = %d\n", 831 printk(KERN_NOTICE "ext4_test_bit(bit=%d, block=%llu) = %d\n",
836 bit, (unsigned long long)bitmap_bh->b_blocknr, 832 bit, (unsigned long long)bitmap_bh->b_blocknr,
diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c
index 8fab233cb05f..8d9707746413 100644
--- a/fs/ext4/inode.c
+++ b/fs/ext4/inode.c
@@ -25,7 +25,6 @@
25#include <linux/module.h> 25#include <linux/module.h>
26#include <linux/fs.h> 26#include <linux/fs.h>
27#include <linux/time.h> 27#include <linux/time.h>
28#include <linux/ext4_jbd2.h>
29#include <linux/jbd2.h> 28#include <linux/jbd2.h>
30#include <linux/highuid.h> 29#include <linux/highuid.h>
31#include <linux/pagemap.h> 30#include <linux/pagemap.h>
@@ -36,6 +35,7 @@
36#include <linux/mpage.h> 35#include <linux/mpage.h>
37#include <linux/uio.h> 36#include <linux/uio.h>
38#include <linux/bio.h> 37#include <linux/bio.h>
38#include "ext4_jbd2.h"
39#include "xattr.h" 39#include "xattr.h"
40#include "acl.h" 40#include "acl.h"
41 41
@@ -93,7 +93,7 @@ int ext4_forget(handle_t *handle, int is_metadata, struct inode *inode,
93 BUFFER_TRACE(bh, "call ext4_journal_revoke"); 93 BUFFER_TRACE(bh, "call ext4_journal_revoke");
94 err = ext4_journal_revoke(handle, blocknr, bh); 94 err = ext4_journal_revoke(handle, blocknr, bh);
95 if (err) 95 if (err)
96 ext4_abort(inode->i_sb, __FUNCTION__, 96 ext4_abort(inode->i_sb, __func__,
97 "error %d when attempting revoke", err); 97 "error %d when attempting revoke", err);
98 BUFFER_TRACE(bh, "exit"); 98 BUFFER_TRACE(bh, "exit");
99 return err; 99 return err;
@@ -985,6 +985,16 @@ int ext4_get_blocks_wrap(handle_t *handle, struct inode *inode, sector_t block,
985 } else { 985 } else {
986 retval = ext4_get_blocks_handle(handle, inode, block, 986 retval = ext4_get_blocks_handle(handle, inode, block,
987 max_blocks, bh, create, extend_disksize); 987 max_blocks, bh, create, extend_disksize);
988
989 if (retval > 0 && buffer_new(bh)) {
990 /*
991 * We allocated new blocks which will result in
992 * i_data's format changing. Force the migrate
993 * to fail by clearing migrate flags
994 */
995 EXT4_I(inode)->i_flags = EXT4_I(inode)->i_flags &
996 ~EXT4_EXT_MIGRATE;
997 }
988 } 998 }
989 up_write((&EXT4_I(inode)->i_data_sem)); 999 up_write((&EXT4_I(inode)->i_data_sem));
990 return retval; 1000 return retval;
@@ -1230,7 +1240,7 @@ int ext4_journal_dirty_data(handle_t *handle, struct buffer_head *bh)
1230{ 1240{
1231 int err = jbd2_journal_dirty_data(handle, bh); 1241 int err = jbd2_journal_dirty_data(handle, bh);
1232 if (err) 1242 if (err)
1233 ext4_journal_abort_handle(__FUNCTION__, __FUNCTION__, 1243 ext4_journal_abort_handle(__func__, __func__,
1234 bh, handle, err); 1244 bh, handle, err);
1235 return err; 1245 return err;
1236} 1246}
@@ -1301,10 +1311,11 @@ static int ext4_ordered_write_end(struct file *file,
1301 new_i_size = pos + copied; 1311 new_i_size = pos + copied;
1302 if (new_i_size > EXT4_I(inode)->i_disksize) 1312 if (new_i_size > EXT4_I(inode)->i_disksize)
1303 EXT4_I(inode)->i_disksize = new_i_size; 1313 EXT4_I(inode)->i_disksize = new_i_size;
1304 copied = ext4_generic_write_end(file, mapping, pos, len, copied, 1314 ret2 = ext4_generic_write_end(file, mapping, pos, len, copied,
1305 page, fsdata); 1315 page, fsdata);
1306 if (copied < 0) 1316 copied = ret2;
1307 ret = copied; 1317 if (ret2 < 0)
1318 ret = ret2;
1308 } 1319 }
1309 ret2 = ext4_journal_stop(handle); 1320 ret2 = ext4_journal_stop(handle);
1310 if (!ret) 1321 if (!ret)
@@ -1329,10 +1340,11 @@ static int ext4_writeback_write_end(struct file *file,
1329 if (new_i_size > EXT4_I(inode)->i_disksize) 1340 if (new_i_size > EXT4_I(inode)->i_disksize)
1330 EXT4_I(inode)->i_disksize = new_i_size; 1341 EXT4_I(inode)->i_disksize = new_i_size;
1331 1342
1332 copied = ext4_generic_write_end(file, mapping, pos, len, copied, 1343 ret2 = ext4_generic_write_end(file, mapping, pos, len, copied,
1333 page, fsdata); 1344 page, fsdata);
1334 if (copied < 0) 1345 copied = ret2;
1335 ret = copied; 1346 if (ret2 < 0)
1347 ret = ret2;
1336 1348
1337 ret2 = ext4_journal_stop(handle); 1349 ret2 = ext4_journal_stop(handle);
1338 if (!ret) 1350 if (!ret)
@@ -2501,12 +2513,10 @@ out_stop:
2501static ext4_fsblk_t ext4_get_inode_block(struct super_block *sb, 2513static ext4_fsblk_t ext4_get_inode_block(struct super_block *sb,
2502 unsigned long ino, struct ext4_iloc *iloc) 2514 unsigned long ino, struct ext4_iloc *iloc)
2503{ 2515{
2504 unsigned long desc, group_desc;
2505 ext4_group_t block_group; 2516 ext4_group_t block_group;
2506 unsigned long offset; 2517 unsigned long offset;
2507 ext4_fsblk_t block; 2518 ext4_fsblk_t block;
2508 struct buffer_head *bh; 2519 struct ext4_group_desc *gdp;
2509 struct ext4_group_desc * gdp;
2510 2520
2511 if (!ext4_valid_inum(sb, ino)) { 2521 if (!ext4_valid_inum(sb, ino)) {
2512 /* 2522 /*
@@ -2518,22 +2528,10 @@ static ext4_fsblk_t ext4_get_inode_block(struct super_block *sb,
2518 } 2528 }
2519 2529
2520 block_group = (ino - 1) / EXT4_INODES_PER_GROUP(sb); 2530 block_group = (ino - 1) / EXT4_INODES_PER_GROUP(sb);
2521 if (block_group >= EXT4_SB(sb)->s_groups_count) { 2531 gdp = ext4_get_group_desc(sb, block_group, NULL);
2522 ext4_error(sb,"ext4_get_inode_block","group >= groups count"); 2532 if (!gdp)
2523 return 0; 2533 return 0;
2524 }
2525 smp_rmb();
2526 group_desc = block_group >> EXT4_DESC_PER_BLOCK_BITS(sb);
2527 desc = block_group & (EXT4_DESC_PER_BLOCK(sb) - 1);
2528 bh = EXT4_SB(sb)->s_group_desc[group_desc];
2529 if (!bh) {
2530 ext4_error (sb, "ext4_get_inode_block",
2531 "Descriptor not loaded");
2532 return 0;
2533 }
2534 2534
2535 gdp = (struct ext4_group_desc *)((__u8 *)bh->b_data +
2536 desc * EXT4_DESC_SIZE(sb));
2537 /* 2535 /*
2538 * Figure out the offset within the block group inode table 2536 * Figure out the offset within the block group inode table
2539 */ 2537 */
@@ -2976,7 +2974,8 @@ static int ext4_do_update_inode(handle_t *handle,
2976 if (ext4_inode_blocks_set(handle, raw_inode, ei)) 2974 if (ext4_inode_blocks_set(handle, raw_inode, ei))
2977 goto out_brelse; 2975 goto out_brelse;
2978 raw_inode->i_dtime = cpu_to_le32(ei->i_dtime); 2976 raw_inode->i_dtime = cpu_to_le32(ei->i_dtime);
2979 raw_inode->i_flags = cpu_to_le32(ei->i_flags); 2977 /* clear the migrate flag in the raw_inode */
2978 raw_inode->i_flags = cpu_to_le32(ei->i_flags & ~EXT4_EXT_MIGRATE);
2980 if (EXT4_SB(inode->i_sb)->s_es->s_creator_os != 2979 if (EXT4_SB(inode->i_sb)->s_es->s_creator_os !=
2981 cpu_to_le32(EXT4_OS_HURD)) 2980 cpu_to_le32(EXT4_OS_HURD))
2982 raw_inode->i_file_acl_high = 2981 raw_inode->i_file_acl_high =
@@ -3374,7 +3373,7 @@ int ext4_mark_inode_dirty(handle_t *handle, struct inode *inode)
3374 EXT4_I(inode)->i_state |= EXT4_STATE_NO_EXPAND; 3373 EXT4_I(inode)->i_state |= EXT4_STATE_NO_EXPAND;
3375 if (mnt_count != 3374 if (mnt_count !=
3376 le16_to_cpu(sbi->s_es->s_mnt_count)) { 3375 le16_to_cpu(sbi->s_es->s_mnt_count)) {
3377 ext4_warning(inode->i_sb, __FUNCTION__, 3376 ext4_warning(inode->i_sb, __func__,
3378 "Unable to expand inode %lu. Delete" 3377 "Unable to expand inode %lu. Delete"
3379 " some EAs or run e2fsck.", 3378 " some EAs or run e2fsck.",
3380 inode->i_ino); 3379 inode->i_ino);
@@ -3415,7 +3414,7 @@ void ext4_dirty_inode(struct inode *inode)
3415 current_handle->h_transaction != handle->h_transaction) { 3414 current_handle->h_transaction != handle->h_transaction) {
3416 /* This task has a transaction open against a different fs */ 3415 /* This task has a transaction open against a different fs */
3417 printk(KERN_EMERG "%s: transactions do not match!\n", 3416 printk(KERN_EMERG "%s: transactions do not match!\n",
3418 __FUNCTION__); 3417 __func__);
3419 } else { 3418 } else {
3420 jbd_debug(5, "marking dirty. outer handle=%p\n", 3419 jbd_debug(5, "marking dirty. outer handle=%p\n",
3421 current_handle); 3420 current_handle);
diff --git a/fs/ext4/ioctl.c b/fs/ext4/ioctl.c
index 25b13ede8086..7a6c2f1faba6 100644
--- a/fs/ext4/ioctl.c
+++ b/fs/ext4/ioctl.c
@@ -10,17 +10,17 @@
10#include <linux/fs.h> 10#include <linux/fs.h>
11#include <linux/jbd2.h> 11#include <linux/jbd2.h>
12#include <linux/capability.h> 12#include <linux/capability.h>
13#include <linux/ext4_fs.h>
14#include <linux/ext4_jbd2.h>
15#include <linux/time.h> 13#include <linux/time.h>
16#include <linux/compat.h> 14#include <linux/compat.h>
17#include <linux/smp_lock.h> 15#include <linux/smp_lock.h>
18#include <linux/mount.h> 16#include <linux/mount.h>
19#include <asm/uaccess.h> 17#include <asm/uaccess.h>
18#include "ext4_jbd2.h"
19#include "ext4.h"
20 20
21int ext4_ioctl (struct inode * inode, struct file * filp, unsigned int cmd, 21long ext4_ioctl(struct file *filp, unsigned int cmd, unsigned long arg)
22 unsigned long arg)
23{ 22{
23 struct inode *inode = filp->f_dentry->d_inode;
24 struct ext4_inode_info *ei = EXT4_I(inode); 24 struct ext4_inode_info *ei = EXT4_I(inode);
25 unsigned int flags; 25 unsigned int flags;
26 unsigned short rsv_window_size; 26 unsigned short rsv_window_size;
@@ -277,9 +277,6 @@ setversion_out:
277#ifdef CONFIG_COMPAT 277#ifdef CONFIG_COMPAT
278long ext4_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg) 278long ext4_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
279{ 279{
280 struct inode *inode = file->f_path.dentry->d_inode;
281 int ret;
282
283 /* These are just misnamed, they actually get/put from/to user an int */ 280 /* These are just misnamed, they actually get/put from/to user an int */
284 switch (cmd) { 281 switch (cmd) {
285 case EXT4_IOC32_GETFLAGS: 282 case EXT4_IOC32_GETFLAGS:
@@ -319,9 +316,6 @@ long ext4_compat_ioctl(struct file *file, unsigned int cmd, unsigned long arg)
319 default: 316 default:
320 return -ENOIOCTLCMD; 317 return -ENOIOCTLCMD;
321 } 318 }
322 lock_kernel(); 319 return ext4_ioctl(file, cmd, (unsigned long) compat_ptr(arg));
323 ret = ext4_ioctl(inode, file, cmd, (unsigned long) compat_ptr(arg));
324 unlock_kernel();
325 return ret;
326} 320}
327#endif 321#endif
diff --git a/fs/ext4/mballoc.c b/fs/ext4/mballoc.c
index 9d57695de746..fbec2ef93797 100644
--- a/fs/ext4/mballoc.c
+++ b/fs/ext4/mballoc.c
@@ -21,21 +21,7 @@
21 * mballoc.c contains the multiblocks allocation routines 21 * mballoc.c contains the multiblocks allocation routines
22 */ 22 */
23 23
24#include <linux/time.h> 24#include "mballoc.h"
25#include <linux/fs.h>
26#include <linux/namei.h>
27#include <linux/ext4_jbd2.h>
28#include <linux/ext4_fs.h>
29#include <linux/quotaops.h>
30#include <linux/buffer_head.h>
31#include <linux/module.h>
32#include <linux/swap.h>
33#include <linux/proc_fs.h>
34#include <linux/pagemap.h>
35#include <linux/seq_file.h>
36#include <linux/version.h>
37#include "group.h"
38
39/* 25/*
40 * MUSTDO: 26 * MUSTDO:
41 * - test ext4_ext_search_left() and ext4_ext_search_right() 27 * - test ext4_ext_search_left() and ext4_ext_search_right()
@@ -345,288 +331,6 @@
345 * 331 *
346 */ 332 */
347 333
348/*
349 * with AGGRESSIVE_CHECK allocator runs consistency checks over
350 * structures. these checks slow things down a lot
351 */
352#define AGGRESSIVE_CHECK__
353
354/*
355 * with DOUBLE_CHECK defined mballoc creates persistent in-core
356 * bitmaps, maintains and uses them to check for double allocations
357 */
358#define DOUBLE_CHECK__
359
360/*
361 */
362#define MB_DEBUG__
363#ifdef MB_DEBUG
364#define mb_debug(fmt, a...) printk(fmt, ##a)
365#else
366#define mb_debug(fmt, a...)
367#endif
368
369/*
370 * with EXT4_MB_HISTORY mballoc stores last N allocations in memory
371 * and you can monitor it in /proc/fs/ext4/<dev>/mb_history
372 */
373#define EXT4_MB_HISTORY
374#define EXT4_MB_HISTORY_ALLOC 1 /* allocation */
375#define EXT4_MB_HISTORY_PREALLOC 2 /* preallocated blocks used */
376#define EXT4_MB_HISTORY_DISCARD 4 /* preallocation discarded */
377#define EXT4_MB_HISTORY_FREE 8 /* free */
378
379#define EXT4_MB_HISTORY_DEFAULT (EXT4_MB_HISTORY_ALLOC | \
380 EXT4_MB_HISTORY_PREALLOC)
381
382/*
383 * How long mballoc can look for a best extent (in found extents)
384 */
385#define MB_DEFAULT_MAX_TO_SCAN 200
386
387/*
388 * How long mballoc must look for a best extent
389 */
390#define MB_DEFAULT_MIN_TO_SCAN 10
391
392/*
393 * How many groups mballoc will scan looking for the best chunk
394 */
395#define MB_DEFAULT_MAX_GROUPS_TO_SCAN 5
396
397/*
398 * with 'ext4_mb_stats' allocator will collect stats that will be
399 * shown at umount. The collecting costs though!
400 */
401#define MB_DEFAULT_STATS 1
402
403/*
404 * files smaller than MB_DEFAULT_STREAM_THRESHOLD are served
405 * by the stream allocator, which purpose is to pack requests
406 * as close each to other as possible to produce smooth I/O traffic
407 * We use locality group prealloc space for stream request.
408 * We can tune the same via /proc/fs/ext4/<parition>/stream_req
409 */
410#define MB_DEFAULT_STREAM_THRESHOLD 16 /* 64K */
411
412/*
413 * for which requests use 2^N search using buddies
414 */
415#define MB_DEFAULT_ORDER2_REQS 2
416
417/*
418 * default group prealloc size 512 blocks
419 */
420#define MB_DEFAULT_GROUP_PREALLOC 512
421
422static struct kmem_cache *ext4_pspace_cachep;
423static struct kmem_cache *ext4_ac_cachep;
424
425#ifdef EXT4_BB_MAX_BLOCKS
426#undef EXT4_BB_MAX_BLOCKS
427#endif
428#define EXT4_BB_MAX_BLOCKS 30
429
430struct ext4_free_metadata {
431 ext4_group_t group;
432 unsigned short num;
433 ext4_grpblk_t blocks[EXT4_BB_MAX_BLOCKS];
434 struct list_head list;
435};
436
437struct ext4_group_info {
438 unsigned long bb_state;
439 unsigned long bb_tid;
440 struct ext4_free_metadata *bb_md_cur;
441 unsigned short bb_first_free;
442 unsigned short bb_free;
443 unsigned short bb_fragments;
444 struct list_head bb_prealloc_list;
445#ifdef DOUBLE_CHECK
446 void *bb_bitmap;
447#endif
448 unsigned short bb_counters[];
449};
450
451#define EXT4_GROUP_INFO_NEED_INIT_BIT 0
452#define EXT4_GROUP_INFO_LOCKED_BIT 1
453
454#define EXT4_MB_GRP_NEED_INIT(grp) \
455 (test_bit(EXT4_GROUP_INFO_NEED_INIT_BIT, &((grp)->bb_state)))
456
457
458struct ext4_prealloc_space {
459 struct list_head pa_inode_list;
460 struct list_head pa_group_list;
461 union {
462 struct list_head pa_tmp_list;
463 struct rcu_head pa_rcu;
464 } u;
465 spinlock_t pa_lock;
466 atomic_t pa_count;
467 unsigned pa_deleted;
468 ext4_fsblk_t pa_pstart; /* phys. block */
469 ext4_lblk_t pa_lstart; /* log. block */
470 unsigned short pa_len; /* len of preallocated chunk */
471 unsigned short pa_free; /* how many blocks are free */
472 unsigned short pa_linear; /* consumed in one direction
473 * strictly, for grp prealloc */
474 spinlock_t *pa_obj_lock;
475 struct inode *pa_inode; /* hack, for history only */
476};
477
478
479struct ext4_free_extent {
480 ext4_lblk_t fe_logical;
481 ext4_grpblk_t fe_start;
482 ext4_group_t fe_group;
483 int fe_len;
484};
485
486/*
487 * Locality group:
488 * we try to group all related changes together
489 * so that writeback can flush/allocate them together as well
490 */
491struct ext4_locality_group {
492 /* for allocator */
493 struct mutex lg_mutex; /* to serialize allocates */
494 struct list_head lg_prealloc_list;/* list of preallocations */
495 spinlock_t lg_prealloc_lock;
496};
497
498struct ext4_allocation_context {
499 struct inode *ac_inode;
500 struct super_block *ac_sb;
501
502 /* original request */
503 struct ext4_free_extent ac_o_ex;
504
505 /* goal request (after normalization) */
506 struct ext4_free_extent ac_g_ex;
507
508 /* the best found extent */
509 struct ext4_free_extent ac_b_ex;
510
511 /* copy of the bext found extent taken before preallocation efforts */
512 struct ext4_free_extent ac_f_ex;
513
514 /* number of iterations done. we have to track to limit searching */
515 unsigned long ac_ex_scanned;
516 __u16 ac_groups_scanned;
517 __u16 ac_found;
518 __u16 ac_tail;
519 __u16 ac_buddy;
520 __u16 ac_flags; /* allocation hints */
521 __u8 ac_status;
522 __u8 ac_criteria;
523 __u8 ac_repeats;
524 __u8 ac_2order; /* if request is to allocate 2^N blocks and
525 * N > 0, the field stores N, otherwise 0 */
526 __u8 ac_op; /* operation, for history only */
527 struct page *ac_bitmap_page;
528 struct page *ac_buddy_page;
529 struct ext4_prealloc_space *ac_pa;
530 struct ext4_locality_group *ac_lg;
531};
532
533#define AC_STATUS_CONTINUE 1
534#define AC_STATUS_FOUND 2
535#define AC_STATUS_BREAK 3
536
537struct ext4_mb_history {
538 struct ext4_free_extent orig; /* orig allocation */
539 struct ext4_free_extent goal; /* goal allocation */
540 struct ext4_free_extent result; /* result allocation */
541 unsigned pid;
542 unsigned ino;
543 __u16 found; /* how many extents have been found */
544 __u16 groups; /* how many groups have been scanned */
545 __u16 tail; /* what tail broke some buddy */
546 __u16 buddy; /* buddy the tail ^^^ broke */
547 __u16 flags;
548 __u8 cr:3; /* which phase the result extent was found at */
549 __u8 op:4;
550 __u8 merged:1;
551};
552
553struct ext4_buddy {
554 struct page *bd_buddy_page;
555 void *bd_buddy;
556 struct page *bd_bitmap_page;
557 void *bd_bitmap;
558 struct ext4_group_info *bd_info;
559 struct super_block *bd_sb;
560 __u16 bd_blkbits;
561 ext4_group_t bd_group;
562};
563#define EXT4_MB_BITMAP(e4b) ((e4b)->bd_bitmap)
564#define EXT4_MB_BUDDY(e4b) ((e4b)->bd_buddy)
565
566#ifndef EXT4_MB_HISTORY
567static inline void ext4_mb_store_history(struct ext4_allocation_context *ac)
568{
569 return;
570}
571#else
572static void ext4_mb_store_history(struct ext4_allocation_context *ac);
573#endif
574
575#define in_range(b, first, len) ((b) >= (first) && (b) <= (first) + (len) - 1)
576
577static struct proc_dir_entry *proc_root_ext4;
578struct buffer_head *read_block_bitmap(struct super_block *, ext4_group_t);
579ext4_fsblk_t ext4_new_blocks_old(handle_t *handle, struct inode *inode,
580 ext4_fsblk_t goal, unsigned long *count, int *errp);
581
582static void ext4_mb_generate_from_pa(struct super_block *sb, void *bitmap,
583 ext4_group_t group);
584static void ext4_mb_poll_new_transaction(struct super_block *, handle_t *);
585static void ext4_mb_free_committed_blocks(struct super_block *);
586static void ext4_mb_return_to_preallocation(struct inode *inode,
587 struct ext4_buddy *e4b, sector_t block,
588 int count);
589static void ext4_mb_put_pa(struct ext4_allocation_context *,
590 struct super_block *, struct ext4_prealloc_space *pa);
591static int ext4_mb_init_per_dev_proc(struct super_block *sb);
592static int ext4_mb_destroy_per_dev_proc(struct super_block *sb);
593
594
595static inline void ext4_lock_group(struct super_block *sb, ext4_group_t group)
596{
597 struct ext4_group_info *grinfo = ext4_get_group_info(sb, group);
598
599 bit_spin_lock(EXT4_GROUP_INFO_LOCKED_BIT, &(grinfo->bb_state));
600}
601
602static inline void ext4_unlock_group(struct super_block *sb,
603 ext4_group_t group)
604{
605 struct ext4_group_info *grinfo = ext4_get_group_info(sb, group);
606
607 bit_spin_unlock(EXT4_GROUP_INFO_LOCKED_BIT, &(grinfo->bb_state));
608}
609
610static inline int ext4_is_group_locked(struct super_block *sb,
611 ext4_group_t group)
612{
613 struct ext4_group_info *grinfo = ext4_get_group_info(sb, group);
614
615 return bit_spin_is_locked(EXT4_GROUP_INFO_LOCKED_BIT,
616 &(grinfo->bb_state));
617}
618
619static ext4_fsblk_t ext4_grp_offs_to_block(struct super_block *sb,
620 struct ext4_free_extent *fex)
621{
622 ext4_fsblk_t block;
623
624 block = (ext4_fsblk_t) fex->fe_group * EXT4_BLOCKS_PER_GROUP(sb)
625 + fex->fe_start
626 + le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block);
627 return block;
628}
629
630static inline void *mb_correct_addr_and_bit(int *bit, void *addr) 334static inline void *mb_correct_addr_and_bit(int *bit, void *addr)
631{ 335{
632#if BITS_PER_LONG == 64 336#if BITS_PER_LONG == 64
@@ -736,7 +440,7 @@ static void mb_free_blocks_double(struct inode *inode, struct ext4_buddy *e4b,
736 blocknr += 440 blocknr +=
737 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block); 441 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block);
738 442
739 ext4_error(sb, __FUNCTION__, "double-free of inode" 443 ext4_error(sb, __func__, "double-free of inode"
740 " %lu's block %llu(bit %u in group %lu)\n", 444 " %lu's block %llu(bit %u in group %lu)\n",
741 inode ? inode->i_ino : 0, blocknr, 445 inode ? inode->i_ino : 0, blocknr,
742 first + i, e4b->bd_group); 446 first + i, e4b->bd_group);
@@ -898,17 +602,17 @@ static int __mb_check_buddy(struct ext4_buddy *e4b, char *file,
898 list_for_each(cur, &grp->bb_prealloc_list) { 602 list_for_each(cur, &grp->bb_prealloc_list) {
899 ext4_group_t groupnr; 603 ext4_group_t groupnr;
900 struct ext4_prealloc_space *pa; 604 struct ext4_prealloc_space *pa;
901 pa = list_entry(cur, struct ext4_prealloc_space, group_list); 605 pa = list_entry(cur, struct ext4_prealloc_space, pa_group_list);
902 ext4_get_group_no_and_offset(sb, pa->pstart, &groupnr, &k); 606 ext4_get_group_no_and_offset(sb, pa->pa_pstart, &groupnr, &k);
903 MB_CHECK_ASSERT(groupnr == e4b->bd_group); 607 MB_CHECK_ASSERT(groupnr == e4b->bd_group);
904 for (i = 0; i < pa->len; i++) 608 for (i = 0; i < pa->pa_len; i++)
905 MB_CHECK_ASSERT(mb_test_bit(k + i, buddy)); 609 MB_CHECK_ASSERT(mb_test_bit(k + i, buddy));
906 } 610 }
907 return 0; 611 return 0;
908} 612}
909#undef MB_CHECK_ASSERT 613#undef MB_CHECK_ASSERT
910#define mb_check_buddy(e4b) __mb_check_buddy(e4b, \ 614#define mb_check_buddy(e4b) __mb_check_buddy(e4b, \
911 __FILE__, __FUNCTION__, __LINE__) 615 __FILE__, __func__, __LINE__)
912#else 616#else
913#define mb_check_buddy(e4b) 617#define mb_check_buddy(e4b)
914#endif 618#endif
@@ -982,7 +686,7 @@ static void ext4_mb_generate_buddy(struct super_block *sb,
982 grp->bb_fragments = fragments; 686 grp->bb_fragments = fragments;
983 687
984 if (free != grp->bb_free) { 688 if (free != grp->bb_free) {
985 ext4_error(sb, __FUNCTION__, 689 ext4_error(sb, __func__,
986 "EXT4-fs: group %lu: %u blocks in bitmap, %u in gd\n", 690 "EXT4-fs: group %lu: %u blocks in bitmap, %u in gd\n",
987 group, free, grp->bb_free); 691 group, free, grp->bb_free);
988 /* 692 /*
@@ -1168,8 +872,9 @@ out:
1168 return err; 872 return err;
1169} 873}
1170 874
1171static int ext4_mb_load_buddy(struct super_block *sb, ext4_group_t group, 875static noinline_for_stack int
1172 struct ext4_buddy *e4b) 876ext4_mb_load_buddy(struct super_block *sb, ext4_group_t group,
877 struct ext4_buddy *e4b)
1173{ 878{
1174 struct ext4_sb_info *sbi = EXT4_SB(sb); 879 struct ext4_sb_info *sbi = EXT4_SB(sb);
1175 struct inode *inode = sbi->s_buddy_cache; 880 struct inode *inode = sbi->s_buddy_cache;
@@ -1367,7 +1072,7 @@ static int mb_free_blocks(struct inode *inode, struct ext4_buddy *e4b,
1367 blocknr += 1072 blocknr +=
1368 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block); 1073 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block);
1369 1074
1370 ext4_error(sb, __FUNCTION__, "double-free of inode" 1075 ext4_error(sb, __func__, "double-free of inode"
1371 " %lu's block %llu(bit %u in group %lu)\n", 1076 " %lu's block %llu(bit %u in group %lu)\n",
1372 inode ? inode->i_ino : 0, blocknr, block, 1077 inode ? inode->i_ino : 0, blocknr, block,
1373 e4b->bd_group); 1078 e4b->bd_group);
@@ -1848,7 +1553,7 @@ static void ext4_mb_complex_scan_group(struct ext4_allocation_context *ac,
1848 * free blocks even though group info says we 1553 * free blocks even though group info says we
1849 * we have free blocks 1554 * we have free blocks
1850 */ 1555 */
1851 ext4_error(sb, __FUNCTION__, "%d free blocks as per " 1556 ext4_error(sb, __func__, "%d free blocks as per "
1852 "group info. But bitmap says 0\n", 1557 "group info. But bitmap says 0\n",
1853 free); 1558 free);
1854 break; 1559 break;
@@ -1857,7 +1562,7 @@ static void ext4_mb_complex_scan_group(struct ext4_allocation_context *ac,
1857 mb_find_extent(e4b, 0, i, ac->ac_g_ex.fe_len, &ex); 1562 mb_find_extent(e4b, 0, i, ac->ac_g_ex.fe_len, &ex);
1858 BUG_ON(ex.fe_len <= 0); 1563 BUG_ON(ex.fe_len <= 0);
1859 if (free < ex.fe_len) { 1564 if (free < ex.fe_len) {
1860 ext4_error(sb, __FUNCTION__, "%d free blocks as per " 1565 ext4_error(sb, __func__, "%d free blocks as per "
1861 "group info. But got %d blocks\n", 1566 "group info. But got %d blocks\n",
1862 free, ex.fe_len); 1567 free, ex.fe_len);
1863 /* 1568 /*
@@ -1965,7 +1670,8 @@ static int ext4_mb_good_group(struct ext4_allocation_context *ac,
1965 return 0; 1670 return 0;
1966} 1671}
1967 1672
1968static int ext4_mb_regular_allocator(struct ext4_allocation_context *ac) 1673static noinline_for_stack int
1674ext4_mb_regular_allocator(struct ext4_allocation_context *ac)
1969{ 1675{
1970 ext4_group_t group; 1676 ext4_group_t group;
1971 ext4_group_t i; 1677 ext4_group_t i;
@@ -2465,7 +2171,8 @@ static void ext4_mb_history_init(struct super_block *sb)
2465 /* if we can't allocate history, then we simple won't use it */ 2171 /* if we can't allocate history, then we simple won't use it */
2466} 2172}
2467 2173
2468static void ext4_mb_store_history(struct ext4_allocation_context *ac) 2174static noinline_for_stack void
2175ext4_mb_store_history(struct ext4_allocation_context *ac)
2469{ 2176{
2470 struct ext4_sb_info *sbi = EXT4_SB(ac->ac_sb); 2177 struct ext4_sb_info *sbi = EXT4_SB(ac->ac_sb);
2471 struct ext4_mb_history h; 2178 struct ext4_mb_history h;
@@ -2565,13 +2272,13 @@ static int ext4_mb_init_backend(struct super_block *sb)
2565 meta_group_info[j] = kzalloc(len, GFP_KERNEL); 2272 meta_group_info[j] = kzalloc(len, GFP_KERNEL);
2566 if (meta_group_info[j] == NULL) { 2273 if (meta_group_info[j] == NULL) {
2567 printk(KERN_ERR "EXT4-fs: can't allocate buddy mem\n"); 2274 printk(KERN_ERR "EXT4-fs: can't allocate buddy mem\n");
2568 i--;
2569 goto err_freebuddy; 2275 goto err_freebuddy;
2570 } 2276 }
2571 desc = ext4_get_group_desc(sb, i, NULL); 2277 desc = ext4_get_group_desc(sb, i, NULL);
2572 if (desc == NULL) { 2278 if (desc == NULL) {
2573 printk(KERN_ERR 2279 printk(KERN_ERR
2574 "EXT4-fs: can't read descriptor %lu\n", i); 2280 "EXT4-fs: can't read descriptor %lu\n", i);
2281 i++;
2575 goto err_freebuddy; 2282 goto err_freebuddy;
2576 } 2283 }
2577 memset(meta_group_info[j], 0, len); 2284 memset(meta_group_info[j], 0, len);
@@ -2611,13 +2318,11 @@ static int ext4_mb_init_backend(struct super_block *sb)
2611 return 0; 2318 return 0;
2612 2319
2613err_freebuddy: 2320err_freebuddy:
2614 while (i >= 0) { 2321 while (i-- > 0)
2615 kfree(ext4_get_group_info(sb, i)); 2322 kfree(ext4_get_group_info(sb, i));
2616 i--;
2617 }
2618 i = num_meta_group_infos; 2323 i = num_meta_group_infos;
2619err_freemeta: 2324err_freemeta:
2620 while (--i >= 0) 2325 while (i-- > 0)
2621 kfree(sbi->s_group_info[i]); 2326 kfree(sbi->s_group_info[i]);
2622 iput(sbi->s_buddy_cache); 2327 iput(sbi->s_buddy_cache);
2623err_freesgi: 2328err_freesgi:
@@ -2801,7 +2506,8 @@ int ext4_mb_release(struct super_block *sb)
2801 return 0; 2506 return 0;
2802} 2507}
2803 2508
2804static void ext4_mb_free_committed_blocks(struct super_block *sb) 2509static noinline_for_stack void
2510ext4_mb_free_committed_blocks(struct super_block *sb)
2805{ 2511{
2806 struct ext4_sb_info *sbi = EXT4_SB(sb); 2512 struct ext4_sb_info *sbi = EXT4_SB(sb);
2807 int err; 2513 int err;
@@ -3021,7 +2727,8 @@ void exit_ext4_mballoc(void)
3021 * Check quota and mark choosed space (ac->ac_b_ex) non-free in bitmaps 2727 * Check quota and mark choosed space (ac->ac_b_ex) non-free in bitmaps
3022 * Returns 0 if success or error code 2728 * Returns 0 if success or error code
3023 */ 2729 */
3024static int ext4_mb_mark_diskspace_used(struct ext4_allocation_context *ac, 2730static noinline_for_stack int
2731ext4_mb_mark_diskspace_used(struct ext4_allocation_context *ac,
3025 handle_t *handle) 2732 handle_t *handle)
3026{ 2733{
3027 struct buffer_head *bitmap_bh = NULL; 2734 struct buffer_head *bitmap_bh = NULL;
@@ -3070,7 +2777,7 @@ static int ext4_mb_mark_diskspace_used(struct ext4_allocation_context *ac,
3070 in_range(block, ext4_inode_table(sb, gdp), 2777 in_range(block, ext4_inode_table(sb, gdp),
3071 EXT4_SB(sb)->s_itb_per_group)) { 2778 EXT4_SB(sb)->s_itb_per_group)) {
3072 2779
3073 ext4_error(sb, __FUNCTION__, 2780 ext4_error(sb, __func__,
3074 "Allocating block in system zone - block = %llu", 2781 "Allocating block in system zone - block = %llu",
3075 block); 2782 block);
3076 } 2783 }
@@ -3094,9 +2801,7 @@ static int ext4_mb_mark_diskspace_used(struct ext4_allocation_context *ac,
3094 ac->ac_b_ex.fe_group, 2801 ac->ac_b_ex.fe_group,
3095 gdp)); 2802 gdp));
3096 } 2803 }
3097 gdp->bg_free_blocks_count = 2804 le16_add_cpu(&gdp->bg_free_blocks_count, -ac->ac_b_ex.fe_len);
3098 cpu_to_le16(le16_to_cpu(gdp->bg_free_blocks_count)
3099 - ac->ac_b_ex.fe_len);
3100 gdp->bg_checksum = ext4_group_desc_csum(sbi, ac->ac_b_ex.fe_group, gdp); 2805 gdp->bg_checksum = ext4_group_desc_csum(sbi, ac->ac_b_ex.fe_group, gdp);
3101 spin_unlock(sb_bgl_lock(sbi, ac->ac_b_ex.fe_group)); 2806 spin_unlock(sb_bgl_lock(sbi, ac->ac_b_ex.fe_group));
3102 percpu_counter_sub(&sbi->s_freeblocks_counter, ac->ac_b_ex.fe_len); 2807 percpu_counter_sub(&sbi->s_freeblocks_counter, ac->ac_b_ex.fe_len);
@@ -3130,7 +2835,7 @@ static void ext4_mb_normalize_group_request(struct ext4_allocation_context *ac)
3130 ac->ac_g_ex.fe_len = EXT4_SB(sb)->s_stripe; 2835 ac->ac_g_ex.fe_len = EXT4_SB(sb)->s_stripe;
3131 else 2836 else
3132 ac->ac_g_ex.fe_len = EXT4_SB(sb)->s_mb_group_prealloc; 2837 ac->ac_g_ex.fe_len = EXT4_SB(sb)->s_mb_group_prealloc;
3133 mb_debug("#%u: goal %lu blocks for locality group\n", 2838 mb_debug("#%u: goal %u blocks for locality group\n",
3134 current->pid, ac->ac_g_ex.fe_len); 2839 current->pid, ac->ac_g_ex.fe_len);
3135} 2840}
3136 2841
@@ -3138,15 +2843,16 @@ static void ext4_mb_normalize_group_request(struct ext4_allocation_context *ac)
3138 * Normalization means making request better in terms of 2843 * Normalization means making request better in terms of
3139 * size and alignment 2844 * size and alignment
3140 */ 2845 */
3141static void ext4_mb_normalize_request(struct ext4_allocation_context *ac, 2846static noinline_for_stack void
2847ext4_mb_normalize_request(struct ext4_allocation_context *ac,
3142 struct ext4_allocation_request *ar) 2848 struct ext4_allocation_request *ar)
3143{ 2849{
3144 int bsbits, max; 2850 int bsbits, max;
3145 ext4_lblk_t end; 2851 ext4_lblk_t end;
3146 struct list_head *cur;
3147 loff_t size, orig_size, start_off; 2852 loff_t size, orig_size, start_off;
3148 ext4_lblk_t start, orig_start; 2853 ext4_lblk_t start, orig_start;
3149 struct ext4_inode_info *ei = EXT4_I(ac->ac_inode); 2854 struct ext4_inode_info *ei = EXT4_I(ac->ac_inode);
2855 struct ext4_prealloc_space *pa;
3150 2856
3151 /* do normalize only data requests, metadata requests 2857 /* do normalize only data requests, metadata requests
3152 do not need preallocation */ 2858 do not need preallocation */
@@ -3232,12 +2938,9 @@ static void ext4_mb_normalize_request(struct ext4_allocation_context *ac,
3232 2938
3233 /* check we don't cross already preallocated blocks */ 2939 /* check we don't cross already preallocated blocks */
3234 rcu_read_lock(); 2940 rcu_read_lock();
3235 list_for_each_rcu(cur, &ei->i_prealloc_list) { 2941 list_for_each_entry_rcu(pa, &ei->i_prealloc_list, pa_inode_list) {
3236 struct ext4_prealloc_space *pa;
3237 unsigned long pa_end; 2942 unsigned long pa_end;
3238 2943
3239 pa = list_entry(cur, struct ext4_prealloc_space, pa_inode_list);
3240
3241 if (pa->pa_deleted) 2944 if (pa->pa_deleted)
3242 continue; 2945 continue;
3243 spin_lock(&pa->pa_lock); 2946 spin_lock(&pa->pa_lock);
@@ -3279,10 +2982,8 @@ static void ext4_mb_normalize_request(struct ext4_allocation_context *ac,
3279 2982
3280 /* XXX: extra loop to check we really don't overlap preallocations */ 2983 /* XXX: extra loop to check we really don't overlap preallocations */
3281 rcu_read_lock(); 2984 rcu_read_lock();
3282 list_for_each_rcu(cur, &ei->i_prealloc_list) { 2985 list_for_each_entry_rcu(pa, &ei->i_prealloc_list, pa_inode_list) {
3283 struct ext4_prealloc_space *pa;
3284 unsigned long pa_end; 2986 unsigned long pa_end;
3285 pa = list_entry(cur, struct ext4_prealloc_space, pa_inode_list);
3286 spin_lock(&pa->pa_lock); 2987 spin_lock(&pa->pa_lock);
3287 if (pa->pa_deleted == 0) { 2988 if (pa->pa_deleted == 0) {
3288 pa_end = pa->pa_lstart + pa->pa_len; 2989 pa_end = pa->pa_lstart + pa->pa_len;
@@ -3374,7 +3075,7 @@ static void ext4_mb_use_inode_pa(struct ext4_allocation_context *ac,
3374 BUG_ON(pa->pa_free < len); 3075 BUG_ON(pa->pa_free < len);
3375 pa->pa_free -= len; 3076 pa->pa_free -= len;
3376 3077
3377 mb_debug("use %llu/%lu from inode pa %p\n", start, len, pa); 3078 mb_debug("use %llu/%u from inode pa %p\n", start, len, pa);
3378} 3079}
3379 3080
3380/* 3081/*
@@ -3404,12 +3105,12 @@ static void ext4_mb_use_group_pa(struct ext4_allocation_context *ac,
3404/* 3105/*
3405 * search goal blocks in preallocated space 3106 * search goal blocks in preallocated space
3406 */ 3107 */
3407static int ext4_mb_use_preallocated(struct ext4_allocation_context *ac) 3108static noinline_for_stack int
3109ext4_mb_use_preallocated(struct ext4_allocation_context *ac)
3408{ 3110{
3409 struct ext4_inode_info *ei = EXT4_I(ac->ac_inode); 3111 struct ext4_inode_info *ei = EXT4_I(ac->ac_inode);
3410 struct ext4_locality_group *lg; 3112 struct ext4_locality_group *lg;
3411 struct ext4_prealloc_space *pa; 3113 struct ext4_prealloc_space *pa;
3412 struct list_head *cur;
3413 3114
3414 /* only data can be preallocated */ 3115 /* only data can be preallocated */
3415 if (!(ac->ac_flags & EXT4_MB_HINT_DATA)) 3116 if (!(ac->ac_flags & EXT4_MB_HINT_DATA))
@@ -3417,8 +3118,7 @@ static int ext4_mb_use_preallocated(struct ext4_allocation_context *ac)
3417 3118
3418 /* first, try per-file preallocation */ 3119 /* first, try per-file preallocation */
3419 rcu_read_lock(); 3120 rcu_read_lock();
3420 list_for_each_rcu(cur, &ei->i_prealloc_list) { 3121 list_for_each_entry_rcu(pa, &ei->i_prealloc_list, pa_inode_list) {
3421 pa = list_entry(cur, struct ext4_prealloc_space, pa_inode_list);
3422 3122
3423 /* all fields in this condition don't change, 3123 /* all fields in this condition don't change,
3424 * so we can skip locking for them */ 3124 * so we can skip locking for them */
@@ -3450,8 +3150,7 @@ static int ext4_mb_use_preallocated(struct ext4_allocation_context *ac)
3450 return 0; 3150 return 0;
3451 3151
3452 rcu_read_lock(); 3152 rcu_read_lock();
3453 list_for_each_rcu(cur, &lg->lg_prealloc_list) { 3153 list_for_each_entry_rcu(pa, &lg->lg_prealloc_list, pa_inode_list) {
3454 pa = list_entry(cur, struct ext4_prealloc_space, pa_inode_list);
3455 spin_lock(&pa->pa_lock); 3154 spin_lock(&pa->pa_lock);
3456 if (pa->pa_deleted == 0 && pa->pa_free >= ac->ac_o_ex.fe_len) { 3155 if (pa->pa_deleted == 0 && pa->pa_free >= ac->ac_o_ex.fe_len) {
3457 atomic_inc(&pa->pa_count); 3156 atomic_inc(&pa->pa_count);
@@ -3571,7 +3270,8 @@ static void ext4_mb_put_pa(struct ext4_allocation_context *ac,
3571/* 3270/*
3572 * creates new preallocated space for given inode 3271 * creates new preallocated space for given inode
3573 */ 3272 */
3574static int ext4_mb_new_inode_pa(struct ext4_allocation_context *ac) 3273static noinline_for_stack int
3274ext4_mb_new_inode_pa(struct ext4_allocation_context *ac)
3575{ 3275{
3576 struct super_block *sb = ac->ac_sb; 3276 struct super_block *sb = ac->ac_sb;
3577 struct ext4_prealloc_space *pa; 3277 struct ext4_prealloc_space *pa;
@@ -3658,7 +3358,8 @@ static int ext4_mb_new_inode_pa(struct ext4_allocation_context *ac)
3658/* 3358/*
3659 * creates new preallocated space for locality group inodes belongs to 3359 * creates new preallocated space for locality group inodes belongs to
3660 */ 3360 */
3661static int ext4_mb_new_group_pa(struct ext4_allocation_context *ac) 3361static noinline_for_stack int
3362ext4_mb_new_group_pa(struct ext4_allocation_context *ac)
3662{ 3363{
3663 struct super_block *sb = ac->ac_sb; 3364 struct super_block *sb = ac->ac_sb;
3664 struct ext4_locality_group *lg; 3365 struct ext4_locality_group *lg;
@@ -3731,11 +3432,11 @@ static int ext4_mb_new_preallocation(struct ext4_allocation_context *ac)
3731 * the caller MUST hold group/inode locks. 3432 * the caller MUST hold group/inode locks.
3732 * TODO: optimize the case when there are no in-core structures yet 3433 * TODO: optimize the case when there are no in-core structures yet
3733 */ 3434 */
3734static int ext4_mb_release_inode_pa(struct ext4_buddy *e4b, 3435static noinline_for_stack int
3735 struct buffer_head *bitmap_bh, 3436ext4_mb_release_inode_pa(struct ext4_buddy *e4b, struct buffer_head *bitmap_bh,
3736 struct ext4_prealloc_space *pa) 3437 struct ext4_prealloc_space *pa,
3438 struct ext4_allocation_context *ac)
3737{ 3439{
3738 struct ext4_allocation_context *ac;
3739 struct super_block *sb = e4b->bd_sb; 3440 struct super_block *sb = e4b->bd_sb;
3740 struct ext4_sb_info *sbi = EXT4_SB(sb); 3441 struct ext4_sb_info *sbi = EXT4_SB(sb);
3741 unsigned long end; 3442 unsigned long end;
@@ -3751,8 +3452,6 @@ static int ext4_mb_release_inode_pa(struct ext4_buddy *e4b,
3751 BUG_ON(group != e4b->bd_group && pa->pa_len != 0); 3452 BUG_ON(group != e4b->bd_group && pa->pa_len != 0);
3752 end = bit + pa->pa_len; 3453 end = bit + pa->pa_len;
3753 3454
3754 ac = kmem_cache_alloc(ext4_ac_cachep, GFP_NOFS);
3755
3756 if (ac) { 3455 if (ac) {
3757 ac->ac_sb = sb; 3456 ac->ac_sb = sb;
3758 ac->ac_inode = pa->pa_inode; 3457 ac->ac_inode = pa->pa_inode;
@@ -3789,7 +3488,7 @@ static int ext4_mb_release_inode_pa(struct ext4_buddy *e4b,
3789 pa, (unsigned long) pa->pa_lstart, 3488 pa, (unsigned long) pa->pa_lstart,
3790 (unsigned long) pa->pa_pstart, 3489 (unsigned long) pa->pa_pstart,
3791 (unsigned long) pa->pa_len); 3490 (unsigned long) pa->pa_len);
3792 ext4_error(sb, __FUNCTION__, "free %u, pa_free %u\n", 3491 ext4_error(sb, __func__, "free %u, pa_free %u\n",
3793 free, pa->pa_free); 3492 free, pa->pa_free);
3794 /* 3493 /*
3795 * pa is already deleted so we use the value obtained 3494 * pa is already deleted so we use the value obtained
@@ -3797,22 +3496,19 @@ static int ext4_mb_release_inode_pa(struct ext4_buddy *e4b,
3797 */ 3496 */
3798 } 3497 }
3799 atomic_add(free, &sbi->s_mb_discarded); 3498 atomic_add(free, &sbi->s_mb_discarded);
3800 if (ac)
3801 kmem_cache_free(ext4_ac_cachep, ac);
3802 3499
3803 return err; 3500 return err;
3804} 3501}
3805 3502
3806static int ext4_mb_release_group_pa(struct ext4_buddy *e4b, 3503static noinline_for_stack int
3807 struct ext4_prealloc_space *pa) 3504ext4_mb_release_group_pa(struct ext4_buddy *e4b,
3505 struct ext4_prealloc_space *pa,
3506 struct ext4_allocation_context *ac)
3808{ 3507{
3809 struct ext4_allocation_context *ac;
3810 struct super_block *sb = e4b->bd_sb; 3508 struct super_block *sb = e4b->bd_sb;
3811 ext4_group_t group; 3509 ext4_group_t group;
3812 ext4_grpblk_t bit; 3510 ext4_grpblk_t bit;
3813 3511
3814 ac = kmem_cache_alloc(ext4_ac_cachep, GFP_NOFS);
3815
3816 if (ac) 3512 if (ac)
3817 ac->ac_op = EXT4_MB_HISTORY_DISCARD; 3513 ac->ac_op = EXT4_MB_HISTORY_DISCARD;
3818 3514
@@ -3830,7 +3526,6 @@ static int ext4_mb_release_group_pa(struct ext4_buddy *e4b,
3830 ac->ac_b_ex.fe_len = pa->pa_len; 3526 ac->ac_b_ex.fe_len = pa->pa_len;
3831 ac->ac_b_ex.fe_logical = 0; 3527 ac->ac_b_ex.fe_logical = 0;
3832 ext4_mb_store_history(ac); 3528 ext4_mb_store_history(ac);
3833 kmem_cache_free(ext4_ac_cachep, ac);
3834 } 3529 }
3835 3530
3836 return 0; 3531 return 0;
@@ -3845,12 +3540,14 @@ static int ext4_mb_release_group_pa(struct ext4_buddy *e4b,
3845 * - how many do we discard 3540 * - how many do we discard
3846 * 1) how many requested 3541 * 1) how many requested
3847 */ 3542 */
3848static int ext4_mb_discard_group_preallocations(struct super_block *sb, 3543static noinline_for_stack int
3544ext4_mb_discard_group_preallocations(struct super_block *sb,
3849 ext4_group_t group, int needed) 3545 ext4_group_t group, int needed)
3850{ 3546{
3851 struct ext4_group_info *grp = ext4_get_group_info(sb, group); 3547 struct ext4_group_info *grp = ext4_get_group_info(sb, group);
3852 struct buffer_head *bitmap_bh = NULL; 3548 struct buffer_head *bitmap_bh = NULL;
3853 struct ext4_prealloc_space *pa, *tmp; 3549 struct ext4_prealloc_space *pa, *tmp;
3550 struct ext4_allocation_context *ac;
3854 struct list_head list; 3551 struct list_head list;
3855 struct ext4_buddy e4b; 3552 struct ext4_buddy e4b;
3856 int err; 3553 int err;
@@ -3878,6 +3575,7 @@ static int ext4_mb_discard_group_preallocations(struct super_block *sb,
3878 grp = ext4_get_group_info(sb, group); 3575 grp = ext4_get_group_info(sb, group);
3879 INIT_LIST_HEAD(&list); 3576 INIT_LIST_HEAD(&list);
3880 3577
3578 ac = kmem_cache_alloc(ext4_ac_cachep, GFP_NOFS);
3881repeat: 3579repeat:
3882 ext4_lock_group(sb, group); 3580 ext4_lock_group(sb, group);
3883 list_for_each_entry_safe(pa, tmp, 3581 list_for_each_entry_safe(pa, tmp,
@@ -3932,9 +3630,9 @@ repeat:
3932 spin_unlock(pa->pa_obj_lock); 3630 spin_unlock(pa->pa_obj_lock);
3933 3631
3934 if (pa->pa_linear) 3632 if (pa->pa_linear)
3935 ext4_mb_release_group_pa(&e4b, pa); 3633 ext4_mb_release_group_pa(&e4b, pa, ac);
3936 else 3634 else
3937 ext4_mb_release_inode_pa(&e4b, bitmap_bh, pa); 3635 ext4_mb_release_inode_pa(&e4b, bitmap_bh, pa, ac);
3938 3636
3939 list_del(&pa->u.pa_tmp_list); 3637 list_del(&pa->u.pa_tmp_list);
3940 call_rcu(&(pa)->u.pa_rcu, ext4_mb_pa_callback); 3638 call_rcu(&(pa)->u.pa_rcu, ext4_mb_pa_callback);
@@ -3942,6 +3640,8 @@ repeat:
3942 3640
3943out: 3641out:
3944 ext4_unlock_group(sb, group); 3642 ext4_unlock_group(sb, group);
3643 if (ac)
3644 kmem_cache_free(ext4_ac_cachep, ac);
3945 ext4_mb_release_desc(&e4b); 3645 ext4_mb_release_desc(&e4b);
3946 put_bh(bitmap_bh); 3646 put_bh(bitmap_bh);
3947 return free; 3647 return free;
@@ -3962,6 +3662,7 @@ void ext4_mb_discard_inode_preallocations(struct inode *inode)
3962 struct super_block *sb = inode->i_sb; 3662 struct super_block *sb = inode->i_sb;
3963 struct buffer_head *bitmap_bh = NULL; 3663 struct buffer_head *bitmap_bh = NULL;
3964 struct ext4_prealloc_space *pa, *tmp; 3664 struct ext4_prealloc_space *pa, *tmp;
3665 struct ext4_allocation_context *ac;
3965 ext4_group_t group = 0; 3666 ext4_group_t group = 0;
3966 struct list_head list; 3667 struct list_head list;
3967 struct ext4_buddy e4b; 3668 struct ext4_buddy e4b;
@@ -3976,6 +3677,7 @@ void ext4_mb_discard_inode_preallocations(struct inode *inode)
3976 3677
3977 INIT_LIST_HEAD(&list); 3678 INIT_LIST_HEAD(&list);
3978 3679
3680 ac = kmem_cache_alloc(ext4_ac_cachep, GFP_NOFS);
3979repeat: 3681repeat:
3980 /* first, collect all pa's in the inode */ 3682 /* first, collect all pa's in the inode */
3981 spin_lock(&ei->i_prealloc_lock); 3683 spin_lock(&ei->i_prealloc_lock);
@@ -4040,7 +3742,7 @@ repeat:
4040 3742
4041 ext4_lock_group(sb, group); 3743 ext4_lock_group(sb, group);
4042 list_del(&pa->pa_group_list); 3744 list_del(&pa->pa_group_list);
4043 ext4_mb_release_inode_pa(&e4b, bitmap_bh, pa); 3745 ext4_mb_release_inode_pa(&e4b, bitmap_bh, pa, ac);
4044 ext4_unlock_group(sb, group); 3746 ext4_unlock_group(sb, group);
4045 3747
4046 ext4_mb_release_desc(&e4b); 3748 ext4_mb_release_desc(&e4b);
@@ -4049,6 +3751,8 @@ repeat:
4049 list_del(&pa->u.pa_tmp_list); 3751 list_del(&pa->u.pa_tmp_list);
4050 call_rcu(&(pa)->u.pa_rcu, ext4_mb_pa_callback); 3752 call_rcu(&(pa)->u.pa_rcu, ext4_mb_pa_callback);
4051 } 3753 }
3754 if (ac)
3755 kmem_cache_free(ext4_ac_cachep, ac);
4052} 3756}
4053 3757
4054/* 3758/*
@@ -4108,7 +3812,7 @@ static void ext4_mb_show_ac(struct ext4_allocation_context *ac)
4108 printk(KERN_ERR "PA:%lu:%d:%u \n", i, 3812 printk(KERN_ERR "PA:%lu:%d:%u \n", i,
4109 start, pa->pa_len); 3813 start, pa->pa_len);
4110 } 3814 }
4111 ext4_lock_group(sb, i); 3815 ext4_unlock_group(sb, i);
4112 3816
4113 if (grp->bb_free == 0) 3817 if (grp->bb_free == 0)
4114 continue; 3818 continue;
@@ -4167,7 +3871,8 @@ static void ext4_mb_group_or_file(struct ext4_allocation_context *ac)
4167 mutex_lock(&ac->ac_lg->lg_mutex); 3871 mutex_lock(&ac->ac_lg->lg_mutex);
4168} 3872}
4169 3873
4170static int ext4_mb_initialize_context(struct ext4_allocation_context *ac, 3874static noinline_for_stack int
3875ext4_mb_initialize_context(struct ext4_allocation_context *ac,
4171 struct ext4_allocation_request *ar) 3876 struct ext4_allocation_request *ar)
4172{ 3877{
4173 struct super_block *sb = ar->inode->i_sb; 3878 struct super_block *sb = ar->inode->i_sb;
@@ -4398,7 +4103,8 @@ static void ext4_mb_poll_new_transaction(struct super_block *sb,
4398 ext4_mb_free_committed_blocks(sb); 4103 ext4_mb_free_committed_blocks(sb);
4399} 4104}
4400 4105
4401static int ext4_mb_free_metadata(handle_t *handle, struct ext4_buddy *e4b, 4106static noinline_for_stack int
4107ext4_mb_free_metadata(handle_t *handle, struct ext4_buddy *e4b,
4402 ext4_group_t group, ext4_grpblk_t block, int count) 4108 ext4_group_t group, ext4_grpblk_t block, int count)
4403{ 4109{
4404 struct ext4_group_info *db = e4b->bd_info; 4110 struct ext4_group_info *db = e4b->bd_info;
@@ -4489,7 +4195,7 @@ void ext4_mb_free_blocks(handle_t *handle, struct inode *inode,
4489 if (block < le32_to_cpu(es->s_first_data_block) || 4195 if (block < le32_to_cpu(es->s_first_data_block) ||
4490 block + count < block || 4196 block + count < block ||
4491 block + count > ext4_blocks_count(es)) { 4197 block + count > ext4_blocks_count(es)) {
4492 ext4_error(sb, __FUNCTION__, 4198 ext4_error(sb, __func__,
4493 "Freeing blocks not in datazone - " 4199 "Freeing blocks not in datazone - "
4494 "block = %lu, count = %lu", block, count); 4200 "block = %lu, count = %lu", block, count);
4495 goto error_return; 4201 goto error_return;
@@ -4530,7 +4236,7 @@ do_more:
4530 in_range(block + count - 1, ext4_inode_table(sb, gdp), 4236 in_range(block + count - 1, ext4_inode_table(sb, gdp),
4531 EXT4_SB(sb)->s_itb_per_group)) { 4237 EXT4_SB(sb)->s_itb_per_group)) {
4532 4238
4533 ext4_error(sb, __FUNCTION__, 4239 ext4_error(sb, __func__,
4534 "Freeing blocks in system zone - " 4240 "Freeing blocks in system zone - "
4535 "Block = %lu, count = %lu", block, count); 4241 "Block = %lu, count = %lu", block, count);
4536 } 4242 }
@@ -4588,8 +4294,7 @@ do_more:
4588 } 4294 }
4589 4295
4590 spin_lock(sb_bgl_lock(sbi, block_group)); 4296 spin_lock(sb_bgl_lock(sbi, block_group));
4591 gdp->bg_free_blocks_count = 4297 le16_add_cpu(&gdp->bg_free_blocks_count, count);
4592 cpu_to_le16(le16_to_cpu(gdp->bg_free_blocks_count) + count);
4593 gdp->bg_checksum = ext4_group_desc_csum(sbi, block_group, gdp); 4298 gdp->bg_checksum = ext4_group_desc_csum(sbi, block_group, gdp);
4594 spin_unlock(sb_bgl_lock(sbi, block_group)); 4299 spin_unlock(sb_bgl_lock(sbi, block_group));
4595 percpu_counter_add(&sbi->s_freeblocks_counter, count); 4300 percpu_counter_add(&sbi->s_freeblocks_counter, count);
diff --git a/fs/ext4/mballoc.h b/fs/ext4/mballoc.h
new file mode 100644
index 000000000000..bfe6add46bcf
--- /dev/null
+++ b/fs/ext4/mballoc.h
@@ -0,0 +1,304 @@
1/*
2 * fs/ext4/mballoc.h
3 *
4 * Written by: Alex Tomas <alex@clusterfs.com>
5 *
6 */
7#ifndef _EXT4_MBALLOC_H
8#define _EXT4_MBALLOC_H
9
10#include <linux/time.h>
11#include <linux/fs.h>
12#include <linux/namei.h>
13#include <linux/quotaops.h>
14#include <linux/buffer_head.h>
15#include <linux/module.h>
16#include <linux/swap.h>
17#include <linux/proc_fs.h>
18#include <linux/pagemap.h>
19#include <linux/seq_file.h>
20#include <linux/version.h>
21#include "ext4_jbd2.h"
22#include "ext4.h"
23#include "group.h"
24
25/*
26 * with AGGRESSIVE_CHECK allocator runs consistency checks over
27 * structures. these checks slow things down a lot
28 */
29#define AGGRESSIVE_CHECK__
30
31/*
32 * with DOUBLE_CHECK defined mballoc creates persistent in-core
33 * bitmaps, maintains and uses them to check for double allocations
34 */
35#define DOUBLE_CHECK__
36
37/*
38 */
39#define MB_DEBUG__
40#ifdef MB_DEBUG
41#define mb_debug(fmt, a...) printk(fmt, ##a)
42#else
43#define mb_debug(fmt, a...)
44#endif
45
46/*
47 * with EXT4_MB_HISTORY mballoc stores last N allocations in memory
48 * and you can monitor it in /proc/fs/ext4/<dev>/mb_history
49 */
50#define EXT4_MB_HISTORY
51#define EXT4_MB_HISTORY_ALLOC 1 /* allocation */
52#define EXT4_MB_HISTORY_PREALLOC 2 /* preallocated blocks used */
53#define EXT4_MB_HISTORY_DISCARD 4 /* preallocation discarded */
54#define EXT4_MB_HISTORY_FREE 8 /* free */
55
56#define EXT4_MB_HISTORY_DEFAULT (EXT4_MB_HISTORY_ALLOC | \
57 EXT4_MB_HISTORY_PREALLOC)
58
59/*
60 * How long mballoc can look for a best extent (in found extents)
61 */
62#define MB_DEFAULT_MAX_TO_SCAN 200
63
64/*
65 * How long mballoc must look for a best extent
66 */
67#define MB_DEFAULT_MIN_TO_SCAN 10
68
69/*
70 * How many groups mballoc will scan looking for the best chunk
71 */
72#define MB_DEFAULT_MAX_GROUPS_TO_SCAN 5
73
74/*
75 * with 'ext4_mb_stats' allocator will collect stats that will be
76 * shown at umount. The collecting costs though!
77 */
78#define MB_DEFAULT_STATS 1
79
80/*
81 * files smaller than MB_DEFAULT_STREAM_THRESHOLD are served
82 * by the stream allocator, which purpose is to pack requests
83 * as close each to other as possible to produce smooth I/O traffic
84 * We use locality group prealloc space for stream request.
85 * We can tune the same via /proc/fs/ext4/<parition>/stream_req
86 */
87#define MB_DEFAULT_STREAM_THRESHOLD 16 /* 64K */
88
89/*
90 * for which requests use 2^N search using buddies
91 */
92#define MB_DEFAULT_ORDER2_REQS 2
93
94/*
95 * default group prealloc size 512 blocks
96 */
97#define MB_DEFAULT_GROUP_PREALLOC 512
98
99static struct kmem_cache *ext4_pspace_cachep;
100static struct kmem_cache *ext4_ac_cachep;
101
102#ifdef EXT4_BB_MAX_BLOCKS
103#undef EXT4_BB_MAX_BLOCKS
104#endif
105#define EXT4_BB_MAX_BLOCKS 30
106
107struct ext4_free_metadata {
108 ext4_group_t group;
109 unsigned short num;
110 ext4_grpblk_t blocks[EXT4_BB_MAX_BLOCKS];
111 struct list_head list;
112};
113
114struct ext4_group_info {
115 unsigned long bb_state;
116 unsigned long bb_tid;
117 struct ext4_free_metadata *bb_md_cur;
118 unsigned short bb_first_free;
119 unsigned short bb_free;
120 unsigned short bb_fragments;
121 struct list_head bb_prealloc_list;
122#ifdef DOUBLE_CHECK
123 void *bb_bitmap;
124#endif
125 unsigned short bb_counters[];
126};
127
128#define EXT4_GROUP_INFO_NEED_INIT_BIT 0
129#define EXT4_GROUP_INFO_LOCKED_BIT 1
130
131#define EXT4_MB_GRP_NEED_INIT(grp) \
132 (test_bit(EXT4_GROUP_INFO_NEED_INIT_BIT, &((grp)->bb_state)))
133
134
135struct ext4_prealloc_space {
136 struct list_head pa_inode_list;
137 struct list_head pa_group_list;
138 union {
139 struct list_head pa_tmp_list;
140 struct rcu_head pa_rcu;
141 } u;
142 spinlock_t pa_lock;
143 atomic_t pa_count;
144 unsigned pa_deleted;
145 ext4_fsblk_t pa_pstart; /* phys. block */
146 ext4_lblk_t pa_lstart; /* log. block */
147 unsigned short pa_len; /* len of preallocated chunk */
148 unsigned short pa_free; /* how many blocks are free */
149 unsigned short pa_linear; /* consumed in one direction
150 * strictly, for grp prealloc */
151 spinlock_t *pa_obj_lock;
152 struct inode *pa_inode; /* hack, for history only */
153};
154
155
156struct ext4_free_extent {
157 ext4_lblk_t fe_logical;
158 ext4_grpblk_t fe_start;
159 ext4_group_t fe_group;
160 int fe_len;
161};
162
163/*
164 * Locality group:
165 * we try to group all related changes together
166 * so that writeback can flush/allocate them together as well
167 */
168struct ext4_locality_group {
169 /* for allocator */
170 struct mutex lg_mutex; /* to serialize allocates */
171 struct list_head lg_prealloc_list;/* list of preallocations */
172 spinlock_t lg_prealloc_lock;
173};
174
175struct ext4_allocation_context {
176 struct inode *ac_inode;
177 struct super_block *ac_sb;
178
179 /* original request */
180 struct ext4_free_extent ac_o_ex;
181
182 /* goal request (after normalization) */
183 struct ext4_free_extent ac_g_ex;
184
185 /* the best found extent */
186 struct ext4_free_extent ac_b_ex;
187
188 /* copy of the bext found extent taken before preallocation efforts */
189 struct ext4_free_extent ac_f_ex;
190
191 /* number of iterations done. we have to track to limit searching */
192 unsigned long ac_ex_scanned;
193 __u16 ac_groups_scanned;
194 __u16 ac_found;
195 __u16 ac_tail;
196 __u16 ac_buddy;
197 __u16 ac_flags; /* allocation hints */
198 __u8 ac_status;
199 __u8 ac_criteria;
200 __u8 ac_repeats;
201 __u8 ac_2order; /* if request is to allocate 2^N blocks and
202 * N > 0, the field stores N, otherwise 0 */
203 __u8 ac_op; /* operation, for history only */
204 struct page *ac_bitmap_page;
205 struct page *ac_buddy_page;
206 struct ext4_prealloc_space *ac_pa;
207 struct ext4_locality_group *ac_lg;
208};
209
210#define AC_STATUS_CONTINUE 1
211#define AC_STATUS_FOUND 2
212#define AC_STATUS_BREAK 3
213
214struct ext4_mb_history {
215 struct ext4_free_extent orig; /* orig allocation */
216 struct ext4_free_extent goal; /* goal allocation */
217 struct ext4_free_extent result; /* result allocation */
218 unsigned pid;
219 unsigned ino;
220 __u16 found; /* how many extents have been found */
221 __u16 groups; /* how many groups have been scanned */
222 __u16 tail; /* what tail broke some buddy */
223 __u16 buddy; /* buddy the tail ^^^ broke */
224 __u16 flags;
225 __u8 cr:3; /* which phase the result extent was found at */
226 __u8 op:4;
227 __u8 merged:1;
228};
229
230struct ext4_buddy {
231 struct page *bd_buddy_page;
232 void *bd_buddy;
233 struct page *bd_bitmap_page;
234 void *bd_bitmap;
235 struct ext4_group_info *bd_info;
236 struct super_block *bd_sb;
237 __u16 bd_blkbits;
238 ext4_group_t bd_group;
239};
240#define EXT4_MB_BITMAP(e4b) ((e4b)->bd_bitmap)
241#define EXT4_MB_BUDDY(e4b) ((e4b)->bd_buddy)
242
243#ifndef EXT4_MB_HISTORY
244static inline void ext4_mb_store_history(struct ext4_allocation_context *ac)
245{
246 return;
247}
248#else
249static void ext4_mb_store_history(struct ext4_allocation_context *ac);
250#endif
251
252#define in_range(b, first, len) ((b) >= (first) && (b) <= (first) + (len) - 1)
253
254static struct proc_dir_entry *proc_root_ext4;
255struct buffer_head *read_block_bitmap(struct super_block *, ext4_group_t);
256
257static void ext4_mb_generate_from_pa(struct super_block *sb, void *bitmap,
258 ext4_group_t group);
259static void ext4_mb_poll_new_transaction(struct super_block *, handle_t *);
260static void ext4_mb_free_committed_blocks(struct super_block *);
261static void ext4_mb_return_to_preallocation(struct inode *inode,
262 struct ext4_buddy *e4b, sector_t block,
263 int count);
264static void ext4_mb_put_pa(struct ext4_allocation_context *,
265 struct super_block *, struct ext4_prealloc_space *pa);
266static int ext4_mb_init_per_dev_proc(struct super_block *sb);
267static int ext4_mb_destroy_per_dev_proc(struct super_block *sb);
268
269
270static inline void ext4_lock_group(struct super_block *sb, ext4_group_t group)
271{
272 struct ext4_group_info *grinfo = ext4_get_group_info(sb, group);
273
274 bit_spin_lock(EXT4_GROUP_INFO_LOCKED_BIT, &(grinfo->bb_state));
275}
276
277static inline void ext4_unlock_group(struct super_block *sb,
278 ext4_group_t group)
279{
280 struct ext4_group_info *grinfo = ext4_get_group_info(sb, group);
281
282 bit_spin_unlock(EXT4_GROUP_INFO_LOCKED_BIT, &(grinfo->bb_state));
283}
284
285static inline int ext4_is_group_locked(struct super_block *sb,
286 ext4_group_t group)
287{
288 struct ext4_group_info *grinfo = ext4_get_group_info(sb, group);
289
290 return bit_spin_is_locked(EXT4_GROUP_INFO_LOCKED_BIT,
291 &(grinfo->bb_state));
292}
293
294static ext4_fsblk_t ext4_grp_offs_to_block(struct super_block *sb,
295 struct ext4_free_extent *fex)
296{
297 ext4_fsblk_t block;
298
299 block = (ext4_fsblk_t) fex->fe_group * EXT4_BLOCKS_PER_GROUP(sb)
300 + fex->fe_start
301 + le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block);
302 return block;
303}
304#endif
diff --git a/fs/ext4/migrate.c b/fs/ext4/migrate.c
index 5c1e27de7755..b9e077ba07e9 100644
--- a/fs/ext4/migrate.c
+++ b/fs/ext4/migrate.c
@@ -13,8 +13,8 @@
13 */ 13 */
14 14
15#include <linux/module.h> 15#include <linux/module.h>
16#include <linux/ext4_jbd2.h> 16#include "ext4_jbd2.h"
17#include <linux/ext4_fs_extents.h> 17#include "ext4_extents.h"
18 18
19/* 19/*
20 * The contiguous blocks details which can be 20 * The contiguous blocks details which can be
@@ -327,7 +327,7 @@ static int free_ind_block(handle_t *handle, struct inode *inode, __le32 *i_data)
327} 327}
328 328
329static int ext4_ext_swap_inode_data(handle_t *handle, struct inode *inode, 329static int ext4_ext_swap_inode_data(handle_t *handle, struct inode *inode,
330 struct inode *tmp_inode) 330 struct inode *tmp_inode)
331{ 331{
332 int retval; 332 int retval;
333 __le32 i_data[3]; 333 __le32 i_data[3];
@@ -339,7 +339,7 @@ static int ext4_ext_swap_inode_data(handle_t *handle, struct inode *inode,
339 * i_data field of the original inode 339 * i_data field of the original inode
340 */ 340 */
341 retval = ext4_journal_extend(handle, 1); 341 retval = ext4_journal_extend(handle, 1);
342 if (retval != 0) { 342 if (retval) {
343 retval = ext4_journal_restart(handle, 1); 343 retval = ext4_journal_restart(handle, 1);
344 if (retval) 344 if (retval)
345 goto err_out; 345 goto err_out;
@@ -351,6 +351,18 @@ static int ext4_ext_swap_inode_data(handle_t *handle, struct inode *inode,
351 351
352 down_write(&EXT4_I(inode)->i_data_sem); 352 down_write(&EXT4_I(inode)->i_data_sem);
353 /* 353 /*
354 * if EXT4_EXT_MIGRATE is cleared a block allocation
355 * happened after we started the migrate. We need to
356 * fail the migrate
357 */
358 if (!(EXT4_I(inode)->i_flags & EXT4_EXT_MIGRATE)) {
359 retval = -EAGAIN;
360 up_write(&EXT4_I(inode)->i_data_sem);
361 goto err_out;
362 } else
363 EXT4_I(inode)->i_flags = EXT4_I(inode)->i_flags &
364 ~EXT4_EXT_MIGRATE;
365 /*
354 * We have the extent map build with the tmp inode. 366 * We have the extent map build with the tmp inode.
355 * Now copy the i_data across 367 * Now copy the i_data across
356 */ 368 */
@@ -508,6 +520,17 @@ int ext4_ext_migrate(struct inode *inode, struct file *filp,
508 * switch the inode format to prevent read. 520 * switch the inode format to prevent read.
509 */ 521 */
510 mutex_lock(&(inode->i_mutex)); 522 mutex_lock(&(inode->i_mutex));
523 /*
524 * Even though we take i_mutex we can still cause block allocation
525 * via mmap write to holes. If we have allocated new blocks we fail
526 * migrate. New block allocation will clear EXT4_EXT_MIGRATE flag.
527 * The flag is updated with i_data_sem held to prevent racing with
528 * block allocation.
529 */
530 down_read((&EXT4_I(inode)->i_data_sem));
531 EXT4_I(inode)->i_flags = EXT4_I(inode)->i_flags | EXT4_EXT_MIGRATE;
532 up_read((&EXT4_I(inode)->i_data_sem));
533
511 handle = ext4_journal_start(inode, 1); 534 handle = ext4_journal_start(inode, 1);
512 535
513 ei = EXT4_I(inode); 536 ei = EXT4_I(inode);
@@ -559,9 +582,15 @@ err_out:
559 * tmp_inode 582 * tmp_inode
560 */ 583 */
561 free_ext_block(handle, tmp_inode); 584 free_ext_block(handle, tmp_inode);
562 else 585 else {
563 retval = ext4_ext_swap_inode_data(handle, inode, 586 retval = ext4_ext_swap_inode_data(handle, inode, tmp_inode);
564 tmp_inode); 587 if (retval)
588 /*
589 * if we fail to swap inode data free the extent
590 * details of the tmp inode
591 */
592 free_ext_block(handle, tmp_inode);
593 }
565 594
566 /* We mark the tmp_inode dirty via ext4_ext_tree_init. */ 595 /* We mark the tmp_inode dirty via ext4_ext_tree_init. */
567 if (ext4_journal_extend(handle, 1) != 0) 596 if (ext4_journal_extend(handle, 1) != 0)
diff --git a/fs/ext4/namei.c b/fs/ext4/namei.c
index 28aa2ed4297e..ab16beaa830d 100644
--- a/fs/ext4/namei.c
+++ b/fs/ext4/namei.c
@@ -28,14 +28,14 @@
28#include <linux/pagemap.h> 28#include <linux/pagemap.h>
29#include <linux/jbd2.h> 29#include <linux/jbd2.h>
30#include <linux/time.h> 30#include <linux/time.h>
31#include <linux/ext4_fs.h>
32#include <linux/ext4_jbd2.h>
33#include <linux/fcntl.h> 31#include <linux/fcntl.h>
34#include <linux/stat.h> 32#include <linux/stat.h>
35#include <linux/string.h> 33#include <linux/string.h>
36#include <linux/quotaops.h> 34#include <linux/quotaops.h>
37#include <linux/buffer_head.h> 35#include <linux/buffer_head.h>
38#include <linux/bio.h> 36#include <linux/bio.h>
37#include "ext4.h"
38#include "ext4_jbd2.h"
39 39
40#include "namei.h" 40#include "namei.h"
41#include "xattr.h" 41#include "xattr.h"
@@ -57,10 +57,15 @@ static struct buffer_head *ext4_append(handle_t *handle,
57 57
58 *block = inode->i_size >> inode->i_sb->s_blocksize_bits; 58 *block = inode->i_size >> inode->i_sb->s_blocksize_bits;
59 59
60 if ((bh = ext4_bread(handle, inode, *block, 1, err))) { 60 bh = ext4_bread(handle, inode, *block, 1, err);
61 if (bh) {
61 inode->i_size += inode->i_sb->s_blocksize; 62 inode->i_size += inode->i_sb->s_blocksize;
62 EXT4_I(inode)->i_disksize = inode->i_size; 63 EXT4_I(inode)->i_disksize = inode->i_size;
63 ext4_journal_get_write_access(handle,bh); 64 *err = ext4_journal_get_write_access(handle, bh);
65 if (*err) {
66 brelse(bh);
67 bh = NULL;
68 }
64 } 69 }
65 return bh; 70 return bh;
66} 71}
@@ -348,7 +353,7 @@ dx_probe(struct dentry *dentry, struct inode *dir,
348 if (root->info.hash_version != DX_HASH_TEA && 353 if (root->info.hash_version != DX_HASH_TEA &&
349 root->info.hash_version != DX_HASH_HALF_MD4 && 354 root->info.hash_version != DX_HASH_HALF_MD4 &&
350 root->info.hash_version != DX_HASH_LEGACY) { 355 root->info.hash_version != DX_HASH_LEGACY) {
351 ext4_warning(dir->i_sb, __FUNCTION__, 356 ext4_warning(dir->i_sb, __func__,
352 "Unrecognised inode hash code %d", 357 "Unrecognised inode hash code %d",
353 root->info.hash_version); 358 root->info.hash_version);
354 brelse(bh); 359 brelse(bh);
@@ -362,7 +367,7 @@ dx_probe(struct dentry *dentry, struct inode *dir,
362 hash = hinfo->hash; 367 hash = hinfo->hash;
363 368
364 if (root->info.unused_flags & 1) { 369 if (root->info.unused_flags & 1) {
365 ext4_warning(dir->i_sb, __FUNCTION__, 370 ext4_warning(dir->i_sb, __func__,
366 "Unimplemented inode hash flags: %#06x", 371 "Unimplemented inode hash flags: %#06x",
367 root->info.unused_flags); 372 root->info.unused_flags);
368 brelse(bh); 373 brelse(bh);
@@ -371,7 +376,7 @@ dx_probe(struct dentry *dentry, struct inode *dir,
371 } 376 }
372 377
373 if ((indirect = root->info.indirect_levels) > 1) { 378 if ((indirect = root->info.indirect_levels) > 1) {
374 ext4_warning(dir->i_sb, __FUNCTION__, 379 ext4_warning(dir->i_sb, __func__,
375 "Unimplemented inode hash depth: %#06x", 380 "Unimplemented inode hash depth: %#06x",
376 root->info.indirect_levels); 381 root->info.indirect_levels);
377 brelse(bh); 382 brelse(bh);
@@ -384,7 +389,7 @@ dx_probe(struct dentry *dentry, struct inode *dir,
384 389
385 if (dx_get_limit(entries) != dx_root_limit(dir, 390 if (dx_get_limit(entries) != dx_root_limit(dir,
386 root->info.info_length)) { 391 root->info.info_length)) {
387 ext4_warning(dir->i_sb, __FUNCTION__, 392 ext4_warning(dir->i_sb, __func__,
388 "dx entry: limit != root limit"); 393 "dx entry: limit != root limit");
389 brelse(bh); 394 brelse(bh);
390 *err = ERR_BAD_DX_DIR; 395 *err = ERR_BAD_DX_DIR;
@@ -396,7 +401,7 @@ dx_probe(struct dentry *dentry, struct inode *dir,
396 { 401 {
397 count = dx_get_count(entries); 402 count = dx_get_count(entries);
398 if (!count || count > dx_get_limit(entries)) { 403 if (!count || count > dx_get_limit(entries)) {
399 ext4_warning(dir->i_sb, __FUNCTION__, 404 ext4_warning(dir->i_sb, __func__,
400 "dx entry: no count or count > limit"); 405 "dx entry: no count or count > limit");
401 brelse(bh); 406 brelse(bh);
402 *err = ERR_BAD_DX_DIR; 407 *err = ERR_BAD_DX_DIR;
@@ -441,7 +446,7 @@ dx_probe(struct dentry *dentry, struct inode *dir,
441 goto fail2; 446 goto fail2;
442 at = entries = ((struct dx_node *) bh->b_data)->entries; 447 at = entries = ((struct dx_node *) bh->b_data)->entries;
443 if (dx_get_limit(entries) != dx_node_limit (dir)) { 448 if (dx_get_limit(entries) != dx_node_limit (dir)) {
444 ext4_warning(dir->i_sb, __FUNCTION__, 449 ext4_warning(dir->i_sb, __func__,
445 "dx entry: limit != node limit"); 450 "dx entry: limit != node limit");
446 brelse(bh); 451 brelse(bh);
447 *err = ERR_BAD_DX_DIR; 452 *err = ERR_BAD_DX_DIR;
@@ -457,7 +462,7 @@ fail2:
457 } 462 }
458fail: 463fail:
459 if (*err == ERR_BAD_DX_DIR) 464 if (*err == ERR_BAD_DX_DIR)
460 ext4_warning(dir->i_sb, __FUNCTION__, 465 ext4_warning(dir->i_sb, __func__,
461 "Corrupt dir inode %ld, running e2fsck is " 466 "Corrupt dir inode %ld, running e2fsck is "
462 "recommended.", dir->i_ino); 467 "recommended.", dir->i_ino);
463 return NULL; 468 return NULL;
@@ -914,7 +919,7 @@ restart:
914 wait_on_buffer(bh); 919 wait_on_buffer(bh);
915 if (!buffer_uptodate(bh)) { 920 if (!buffer_uptodate(bh)) {
916 /* read error, skip block & hope for the best */ 921 /* read error, skip block & hope for the best */
917 ext4_error(sb, __FUNCTION__, "reading directory #%lu " 922 ext4_error(sb, __func__, "reading directory #%lu "
918 "offset %lu", dir->i_ino, 923 "offset %lu", dir->i_ino,
919 (unsigned long)block); 924 (unsigned long)block);
920 brelse(bh); 925 brelse(bh);
@@ -1007,7 +1012,7 @@ static struct buffer_head * ext4_dx_find_entry(struct dentry *dentry,
1007 retval = ext4_htree_next_block(dir, hash, frame, 1012 retval = ext4_htree_next_block(dir, hash, frame,
1008 frames, NULL); 1013 frames, NULL);
1009 if (retval < 0) { 1014 if (retval < 0) {
1010 ext4_warning(sb, __FUNCTION__, 1015 ext4_warning(sb, __func__,
1011 "error reading index page in directory #%lu", 1016 "error reading index page in directory #%lu",
1012 dir->i_ino); 1017 dir->i_ino);
1013 *err = retval; 1018 *err = retval;
@@ -1532,7 +1537,7 @@ static int ext4_dx_add_entry(handle_t *handle, struct dentry *dentry,
1532 1537
1533 if (levels && (dx_get_count(frames->entries) == 1538 if (levels && (dx_get_count(frames->entries) ==
1534 dx_get_limit(frames->entries))) { 1539 dx_get_limit(frames->entries))) {
1535 ext4_warning(sb, __FUNCTION__, 1540 ext4_warning(sb, __func__,
1536 "Directory index full!"); 1541 "Directory index full!");
1537 err = -ENOSPC; 1542 err = -ENOSPC;
1538 goto cleanup; 1543 goto cleanup;
@@ -1860,11 +1865,11 @@ static int empty_dir (struct inode * inode)
1860 if (inode->i_size < EXT4_DIR_REC_LEN(1) + EXT4_DIR_REC_LEN(2) || 1865 if (inode->i_size < EXT4_DIR_REC_LEN(1) + EXT4_DIR_REC_LEN(2) ||
1861 !(bh = ext4_bread (NULL, inode, 0, 0, &err))) { 1866 !(bh = ext4_bread (NULL, inode, 0, 0, &err))) {
1862 if (err) 1867 if (err)
1863 ext4_error(inode->i_sb, __FUNCTION__, 1868 ext4_error(inode->i_sb, __func__,
1864 "error %d reading directory #%lu offset 0", 1869 "error %d reading directory #%lu offset 0",
1865 err, inode->i_ino); 1870 err, inode->i_ino);
1866 else 1871 else
1867 ext4_warning(inode->i_sb, __FUNCTION__, 1872 ext4_warning(inode->i_sb, __func__,
1868 "bad directory (dir #%lu) - no data block", 1873 "bad directory (dir #%lu) - no data block",
1869 inode->i_ino); 1874 inode->i_ino);
1870 return 1; 1875 return 1;
@@ -1893,7 +1898,7 @@ static int empty_dir (struct inode * inode)
1893 offset >> EXT4_BLOCK_SIZE_BITS(sb), 0, &err); 1898 offset >> EXT4_BLOCK_SIZE_BITS(sb), 0, &err);
1894 if (!bh) { 1899 if (!bh) {
1895 if (err) 1900 if (err)
1896 ext4_error(sb, __FUNCTION__, 1901 ext4_error(sb, __func__,
1897 "error %d reading directory" 1902 "error %d reading directory"
1898 " #%lu offset %lu", 1903 " #%lu offset %lu",
1899 err, inode->i_ino, offset); 1904 err, inode->i_ino, offset);
@@ -2217,6 +2222,8 @@ retry:
2217 goto out_stop; 2222 goto out_stop;
2218 } 2223 }
2219 } else { 2224 } else {
2225 /* clear the extent format for fast symlink */
2226 EXT4_I(inode)->i_flags &= ~EXT4_EXTENTS_FL;
2220 inode->i_op = &ext4_fast_symlink_inode_operations; 2227 inode->i_op = &ext4_fast_symlink_inode_operations;
2221 memcpy((char*)&EXT4_I(inode)->i_data,symname,l); 2228 memcpy((char*)&EXT4_I(inode)->i_data,symname,l);
2222 inode->i_size = l-1; 2229 inode->i_size = l-1;
@@ -2347,6 +2354,9 @@ static int ext4_rename (struct inode * old_dir, struct dentry *old_dentry,
2347 EXT4_FEATURE_INCOMPAT_FILETYPE)) 2354 EXT4_FEATURE_INCOMPAT_FILETYPE))
2348 new_de->file_type = old_de->file_type; 2355 new_de->file_type = old_de->file_type;
2349 new_dir->i_version++; 2356 new_dir->i_version++;
2357 new_dir->i_ctime = new_dir->i_mtime =
2358 ext4_current_time(new_dir);
2359 ext4_mark_inode_dirty(handle, new_dir);
2350 BUFFER_TRACE(new_bh, "call ext4_journal_dirty_metadata"); 2360 BUFFER_TRACE(new_bh, "call ext4_journal_dirty_metadata");
2351 ext4_journal_dirty_metadata(handle, new_bh); 2361 ext4_journal_dirty_metadata(handle, new_bh);
2352 brelse(new_bh); 2362 brelse(new_bh);
diff --git a/fs/ext4/resize.c b/fs/ext4/resize.c
index e29efa0f9d62..9f086a6a472b 100644
--- a/fs/ext4/resize.c
+++ b/fs/ext4/resize.c
@@ -11,11 +11,10 @@
11 11
12#define EXT4FS_DEBUG 12#define EXT4FS_DEBUG
13 13
14#include <linux/ext4_jbd2.h>
15
16#include <linux/errno.h> 14#include <linux/errno.h>
17#include <linux/slab.h> 15#include <linux/slab.h>
18 16
17#include "ext4_jbd2.h"
19#include "group.h" 18#include "group.h"
20 19
21#define outside(b, first, last) ((b) < (first) || (b) >= (last)) 20#define outside(b, first, last) ((b) < (first) || (b) >= (last))
@@ -50,63 +49,63 @@ static int verify_group_input(struct super_block *sb,
50 49
51 ext4_get_group_no_and_offset(sb, start, NULL, &offset); 50 ext4_get_group_no_and_offset(sb, start, NULL, &offset);
52 if (group != sbi->s_groups_count) 51 if (group != sbi->s_groups_count)
53 ext4_warning(sb, __FUNCTION__, 52 ext4_warning(sb, __func__,
54 "Cannot add at group %u (only %lu groups)", 53 "Cannot add at group %u (only %lu groups)",
55 input->group, sbi->s_groups_count); 54 input->group, sbi->s_groups_count);
56 else if (offset != 0) 55 else if (offset != 0)
57 ext4_warning(sb, __FUNCTION__, "Last group not full"); 56 ext4_warning(sb, __func__, "Last group not full");
58 else if (input->reserved_blocks > input->blocks_count / 5) 57 else if (input->reserved_blocks > input->blocks_count / 5)
59 ext4_warning(sb, __FUNCTION__, "Reserved blocks too high (%u)", 58 ext4_warning(sb, __func__, "Reserved blocks too high (%u)",
60 input->reserved_blocks); 59 input->reserved_blocks);
61 else if (free_blocks_count < 0) 60 else if (free_blocks_count < 0)
62 ext4_warning(sb, __FUNCTION__, "Bad blocks count %u", 61 ext4_warning(sb, __func__, "Bad blocks count %u",
63 input->blocks_count); 62 input->blocks_count);
64 else if (!(bh = sb_bread(sb, end - 1))) 63 else if (!(bh = sb_bread(sb, end - 1)))
65 ext4_warning(sb, __FUNCTION__, 64 ext4_warning(sb, __func__,
66 "Cannot read last block (%llu)", 65 "Cannot read last block (%llu)",
67 end - 1); 66 end - 1);
68 else if (outside(input->block_bitmap, start, end)) 67 else if (outside(input->block_bitmap, start, end))
69 ext4_warning(sb, __FUNCTION__, 68 ext4_warning(sb, __func__,
70 "Block bitmap not in group (block %llu)", 69 "Block bitmap not in group (block %llu)",
71 (unsigned long long)input->block_bitmap); 70 (unsigned long long)input->block_bitmap);
72 else if (outside(input->inode_bitmap, start, end)) 71 else if (outside(input->inode_bitmap, start, end))
73 ext4_warning(sb, __FUNCTION__, 72 ext4_warning(sb, __func__,
74 "Inode bitmap not in group (block %llu)", 73 "Inode bitmap not in group (block %llu)",
75 (unsigned long long)input->inode_bitmap); 74 (unsigned long long)input->inode_bitmap);
76 else if (outside(input->inode_table, start, end) || 75 else if (outside(input->inode_table, start, end) ||
77 outside(itend - 1, start, end)) 76 outside(itend - 1, start, end))
78 ext4_warning(sb, __FUNCTION__, 77 ext4_warning(sb, __func__,
79 "Inode table not in group (blocks %llu-%llu)", 78 "Inode table not in group (blocks %llu-%llu)",
80 (unsigned long long)input->inode_table, itend - 1); 79 (unsigned long long)input->inode_table, itend - 1);
81 else if (input->inode_bitmap == input->block_bitmap) 80 else if (input->inode_bitmap == input->block_bitmap)
82 ext4_warning(sb, __FUNCTION__, 81 ext4_warning(sb, __func__,
83 "Block bitmap same as inode bitmap (%llu)", 82 "Block bitmap same as inode bitmap (%llu)",
84 (unsigned long long)input->block_bitmap); 83 (unsigned long long)input->block_bitmap);
85 else if (inside(input->block_bitmap, input->inode_table, itend)) 84 else if (inside(input->block_bitmap, input->inode_table, itend))
86 ext4_warning(sb, __FUNCTION__, 85 ext4_warning(sb, __func__,
87 "Block bitmap (%llu) in inode table (%llu-%llu)", 86 "Block bitmap (%llu) in inode table (%llu-%llu)",
88 (unsigned long long)input->block_bitmap, 87 (unsigned long long)input->block_bitmap,
89 (unsigned long long)input->inode_table, itend - 1); 88 (unsigned long long)input->inode_table, itend - 1);
90 else if (inside(input->inode_bitmap, input->inode_table, itend)) 89 else if (inside(input->inode_bitmap, input->inode_table, itend))
91 ext4_warning(sb, __FUNCTION__, 90 ext4_warning(sb, __func__,
92 "Inode bitmap (%llu) in inode table (%llu-%llu)", 91 "Inode bitmap (%llu) in inode table (%llu-%llu)",
93 (unsigned long long)input->inode_bitmap, 92 (unsigned long long)input->inode_bitmap,
94 (unsigned long long)input->inode_table, itend - 1); 93 (unsigned long long)input->inode_table, itend - 1);
95 else if (inside(input->block_bitmap, start, metaend)) 94 else if (inside(input->block_bitmap, start, metaend))
96 ext4_warning(sb, __FUNCTION__, 95 ext4_warning(sb, __func__,
97 "Block bitmap (%llu) in GDT table" 96 "Block bitmap (%llu) in GDT table"
98 " (%llu-%llu)", 97 " (%llu-%llu)",
99 (unsigned long long)input->block_bitmap, 98 (unsigned long long)input->block_bitmap,
100 start, metaend - 1); 99 start, metaend - 1);
101 else if (inside(input->inode_bitmap, start, metaend)) 100 else if (inside(input->inode_bitmap, start, metaend))
102 ext4_warning(sb, __FUNCTION__, 101 ext4_warning(sb, __func__,
103 "Inode bitmap (%llu) in GDT table" 102 "Inode bitmap (%llu) in GDT table"
104 " (%llu-%llu)", 103 " (%llu-%llu)",
105 (unsigned long long)input->inode_bitmap, 104 (unsigned long long)input->inode_bitmap,
106 start, metaend - 1); 105 start, metaend - 1);
107 else if (inside(input->inode_table, start, metaend) || 106 else if (inside(input->inode_table, start, metaend) ||
108 inside(itend - 1, start, metaend)) 107 inside(itend - 1, start, metaend))
109 ext4_warning(sb, __FUNCTION__, 108 ext4_warning(sb, __func__,
110 "Inode table (%llu-%llu) overlaps" 109 "Inode table (%llu-%llu) overlaps"
111 "GDT table (%llu-%llu)", 110 "GDT table (%llu-%llu)",
112 (unsigned long long)input->inode_table, 111 (unsigned long long)input->inode_table,
@@ -368,7 +367,7 @@ static int verify_reserved_gdb(struct super_block *sb,
368 while ((grp = ext4_list_backups(sb, &three, &five, &seven)) < end) { 367 while ((grp = ext4_list_backups(sb, &three, &five, &seven)) < end) {
369 if (le32_to_cpu(*p++) != 368 if (le32_to_cpu(*p++) !=
370 grp * EXT4_BLOCKS_PER_GROUP(sb) + blk){ 369 grp * EXT4_BLOCKS_PER_GROUP(sb) + blk){
371 ext4_warning(sb, __FUNCTION__, 370 ext4_warning(sb, __func__,
372 "reserved GDT %llu" 371 "reserved GDT %llu"
373 " missing grp %d (%llu)", 372 " missing grp %d (%llu)",
374 blk, grp, 373 blk, grp,
@@ -424,7 +423,7 @@ static int add_new_gdb(handle_t *handle, struct inode *inode,
424 */ 423 */
425 if (EXT4_SB(sb)->s_sbh->b_blocknr != 424 if (EXT4_SB(sb)->s_sbh->b_blocknr !=
426 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block)) { 425 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block)) {
427 ext4_warning(sb, __FUNCTION__, 426 ext4_warning(sb, __func__,
428 "won't resize using backup superblock at %llu", 427 "won't resize using backup superblock at %llu",
429 (unsigned long long)EXT4_SB(sb)->s_sbh->b_blocknr); 428 (unsigned long long)EXT4_SB(sb)->s_sbh->b_blocknr);
430 return -EPERM; 429 return -EPERM;
@@ -448,7 +447,7 @@ static int add_new_gdb(handle_t *handle, struct inode *inode,
448 447
449 data = (__le32 *)dind->b_data; 448 data = (__le32 *)dind->b_data;
450 if (le32_to_cpu(data[gdb_num % EXT4_ADDR_PER_BLOCK(sb)]) != gdblock) { 449 if (le32_to_cpu(data[gdb_num % EXT4_ADDR_PER_BLOCK(sb)]) != gdblock) {
451 ext4_warning(sb, __FUNCTION__, 450 ext4_warning(sb, __func__,
452 "new group %u GDT block %llu not reserved", 451 "new group %u GDT block %llu not reserved",
453 input->group, gdblock); 452 input->group, gdblock);
454 err = -EINVAL; 453 err = -EINVAL;
@@ -469,10 +468,10 @@ static int add_new_gdb(handle_t *handle, struct inode *inode,
469 goto exit_dindj; 468 goto exit_dindj;
470 469
471 n_group_desc = kmalloc((gdb_num + 1) * sizeof(struct buffer_head *), 470 n_group_desc = kmalloc((gdb_num + 1) * sizeof(struct buffer_head *),
472 GFP_KERNEL); 471 GFP_NOFS);
473 if (!n_group_desc) { 472 if (!n_group_desc) {
474 err = -ENOMEM; 473 err = -ENOMEM;
475 ext4_warning (sb, __FUNCTION__, 474 ext4_warning(sb, __func__,
476 "not enough memory for %lu groups", gdb_num + 1); 475 "not enough memory for %lu groups", gdb_num + 1);
477 goto exit_inode; 476 goto exit_inode;
478 } 477 }
@@ -502,8 +501,7 @@ static int add_new_gdb(handle_t *handle, struct inode *inode,
502 EXT4_SB(sb)->s_gdb_count++; 501 EXT4_SB(sb)->s_gdb_count++;
503 kfree(o_group_desc); 502 kfree(o_group_desc);
504 503
505 es->s_reserved_gdt_blocks = 504 le16_add_cpu(&es->s_reserved_gdt_blocks, -1);
506 cpu_to_le16(le16_to_cpu(es->s_reserved_gdt_blocks) - 1);
507 ext4_journal_dirty_metadata(handle, EXT4_SB(sb)->s_sbh); 505 ext4_journal_dirty_metadata(handle, EXT4_SB(sb)->s_sbh);
508 506
509 return 0; 507 return 0;
@@ -553,7 +551,7 @@ static int reserve_backup_gdb(handle_t *handle, struct inode *inode,
553 int res, i; 551 int res, i;
554 int err; 552 int err;
555 553
556 primary = kmalloc(reserved_gdb * sizeof(*primary), GFP_KERNEL); 554 primary = kmalloc(reserved_gdb * sizeof(*primary), GFP_NOFS);
557 if (!primary) 555 if (!primary)
558 return -ENOMEM; 556 return -ENOMEM;
559 557
@@ -571,7 +569,7 @@ static int reserve_backup_gdb(handle_t *handle, struct inode *inode,
571 /* Get each reserved primary GDT block and verify it holds backups */ 569 /* Get each reserved primary GDT block and verify it holds backups */
572 for (res = 0; res < reserved_gdb; res++, blk++) { 570 for (res = 0; res < reserved_gdb; res++, blk++) {
573 if (le32_to_cpu(*data) != blk) { 571 if (le32_to_cpu(*data) != blk) {
574 ext4_warning(sb, __FUNCTION__, 572 ext4_warning(sb, __func__,
575 "reserved block %llu" 573 "reserved block %llu"
576 " not at offset %ld", 574 " not at offset %ld",
577 blk, 575 blk,
@@ -715,7 +713,7 @@ static void update_backups(struct super_block *sb,
715 */ 713 */
716exit_err: 714exit_err:
717 if (err) { 715 if (err) {
718 ext4_warning(sb, __FUNCTION__, 716 ext4_warning(sb, __func__,
719 "can't update backup for group %lu (err %d), " 717 "can't update backup for group %lu (err %d), "
720 "forcing fsck on next reboot", group, err); 718 "forcing fsck on next reboot", group, err);
721 sbi->s_mount_state &= ~EXT4_VALID_FS; 719 sbi->s_mount_state &= ~EXT4_VALID_FS;
@@ -755,33 +753,33 @@ int ext4_group_add(struct super_block *sb, struct ext4_new_group_data *input)
755 753
756 if (gdb_off == 0 && !EXT4_HAS_RO_COMPAT_FEATURE(sb, 754 if (gdb_off == 0 && !EXT4_HAS_RO_COMPAT_FEATURE(sb,
757 EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER)) { 755 EXT4_FEATURE_RO_COMPAT_SPARSE_SUPER)) {
758 ext4_warning(sb, __FUNCTION__, 756 ext4_warning(sb, __func__,
759 "Can't resize non-sparse filesystem further"); 757 "Can't resize non-sparse filesystem further");
760 return -EPERM; 758 return -EPERM;
761 } 759 }
762 760
763 if (ext4_blocks_count(es) + input->blocks_count < 761 if (ext4_blocks_count(es) + input->blocks_count <
764 ext4_blocks_count(es)) { 762 ext4_blocks_count(es)) {
765 ext4_warning(sb, __FUNCTION__, "blocks_count overflow\n"); 763 ext4_warning(sb, __func__, "blocks_count overflow\n");
766 return -EINVAL; 764 return -EINVAL;
767 } 765 }
768 766
769 if (le32_to_cpu(es->s_inodes_count) + EXT4_INODES_PER_GROUP(sb) < 767 if (le32_to_cpu(es->s_inodes_count) + EXT4_INODES_PER_GROUP(sb) <
770 le32_to_cpu(es->s_inodes_count)) { 768 le32_to_cpu(es->s_inodes_count)) {
771 ext4_warning(sb, __FUNCTION__, "inodes_count overflow\n"); 769 ext4_warning(sb, __func__, "inodes_count overflow\n");
772 return -EINVAL; 770 return -EINVAL;
773 } 771 }
774 772
775 if (reserved_gdb || gdb_off == 0) { 773 if (reserved_gdb || gdb_off == 0) {
776 if (!EXT4_HAS_COMPAT_FEATURE(sb, 774 if (!EXT4_HAS_COMPAT_FEATURE(sb,
777 EXT4_FEATURE_COMPAT_RESIZE_INODE)){ 775 EXT4_FEATURE_COMPAT_RESIZE_INODE)){
778 ext4_warning(sb, __FUNCTION__, 776 ext4_warning(sb, __func__,
779 "No reserved GDT blocks, can't resize"); 777 "No reserved GDT blocks, can't resize");
780 return -EPERM; 778 return -EPERM;
781 } 779 }
782 inode = ext4_iget(sb, EXT4_RESIZE_INO); 780 inode = ext4_iget(sb, EXT4_RESIZE_INO);
783 if (IS_ERR(inode)) { 781 if (IS_ERR(inode)) {
784 ext4_warning(sb, __FUNCTION__, 782 ext4_warning(sb, __func__,
785 "Error opening resize inode"); 783 "Error opening resize inode");
786 return PTR_ERR(inode); 784 return PTR_ERR(inode);
787 } 785 }
@@ -810,7 +808,7 @@ int ext4_group_add(struct super_block *sb, struct ext4_new_group_data *input)
810 808
811 lock_super(sb); 809 lock_super(sb);
812 if (input->group != sbi->s_groups_count) { 810 if (input->group != sbi->s_groups_count) {
813 ext4_warning(sb, __FUNCTION__, 811 ext4_warning(sb, __func__,
814 "multiple resizers run on filesystem!"); 812 "multiple resizers run on filesystem!");
815 err = -EBUSY; 813 err = -EBUSY;
816 goto exit_journal; 814 goto exit_journal;
@@ -877,8 +875,7 @@ int ext4_group_add(struct super_block *sb, struct ext4_new_group_data *input)
877 */ 875 */
878 ext4_blocks_count_set(es, ext4_blocks_count(es) + 876 ext4_blocks_count_set(es, ext4_blocks_count(es) +
879 input->blocks_count); 877 input->blocks_count);
880 es->s_inodes_count = cpu_to_le32(le32_to_cpu(es->s_inodes_count) + 878 le32_add_cpu(&es->s_inodes_count, EXT4_INODES_PER_GROUP(sb));
881 EXT4_INODES_PER_GROUP(sb));
882 879
883 /* 880 /*
884 * We need to protect s_groups_count against other CPUs seeing 881 * We need to protect s_groups_count against other CPUs seeing
@@ -977,13 +974,13 @@ int ext4_group_extend(struct super_block *sb, struct ext4_super_block *es,
977 " too large to resize to %llu blocks safely\n", 974 " too large to resize to %llu blocks safely\n",
978 sb->s_id, n_blocks_count); 975 sb->s_id, n_blocks_count);
979 if (sizeof(sector_t) < 8) 976 if (sizeof(sector_t) < 8)
980 ext4_warning(sb, __FUNCTION__, 977 ext4_warning(sb, __func__,
981 "CONFIG_LBD not enabled\n"); 978 "CONFIG_LBD not enabled\n");
982 return -EINVAL; 979 return -EINVAL;
983 } 980 }
984 981
985 if (n_blocks_count < o_blocks_count) { 982 if (n_blocks_count < o_blocks_count) {
986 ext4_warning(sb, __FUNCTION__, 983 ext4_warning(sb, __func__,
987 "can't shrink FS - resize aborted"); 984 "can't shrink FS - resize aborted");
988 return -EBUSY; 985 return -EBUSY;
989 } 986 }
@@ -992,7 +989,7 @@ int ext4_group_extend(struct super_block *sb, struct ext4_super_block *es,
992 ext4_get_group_no_and_offset(sb, o_blocks_count, NULL, &last); 989 ext4_get_group_no_and_offset(sb, o_blocks_count, NULL, &last);
993 990
994 if (last == 0) { 991 if (last == 0) {
995 ext4_warning(sb, __FUNCTION__, 992 ext4_warning(sb, __func__,
996 "need to use ext2online to resize further"); 993 "need to use ext2online to resize further");
997 return -EPERM; 994 return -EPERM;
998 } 995 }
@@ -1000,7 +997,7 @@ int ext4_group_extend(struct super_block *sb, struct ext4_super_block *es,
1000 add = EXT4_BLOCKS_PER_GROUP(sb) - last; 997 add = EXT4_BLOCKS_PER_GROUP(sb) - last;
1001 998
1002 if (o_blocks_count + add < o_blocks_count) { 999 if (o_blocks_count + add < o_blocks_count) {
1003 ext4_warning(sb, __FUNCTION__, "blocks_count overflow"); 1000 ext4_warning(sb, __func__, "blocks_count overflow");
1004 return -EINVAL; 1001 return -EINVAL;
1005 } 1002 }
1006 1003
@@ -1008,7 +1005,7 @@ int ext4_group_extend(struct super_block *sb, struct ext4_super_block *es,
1008 add = n_blocks_count - o_blocks_count; 1005 add = n_blocks_count - o_blocks_count;
1009 1006
1010 if (o_blocks_count + add < n_blocks_count) 1007 if (o_blocks_count + add < n_blocks_count)
1011 ext4_warning(sb, __FUNCTION__, 1008 ext4_warning(sb, __func__,
1012 "will only finish group (%llu" 1009 "will only finish group (%llu"
1013 " blocks, %u new)", 1010 " blocks, %u new)",
1014 o_blocks_count + add, add); 1011 o_blocks_count + add, add);
@@ -1016,7 +1013,7 @@ int ext4_group_extend(struct super_block *sb, struct ext4_super_block *es,
1016 /* See if the device is actually as big as what was requested */ 1013 /* See if the device is actually as big as what was requested */
1017 bh = sb_bread(sb, o_blocks_count + add -1); 1014 bh = sb_bread(sb, o_blocks_count + add -1);
1018 if (!bh) { 1015 if (!bh) {
1019 ext4_warning(sb, __FUNCTION__, 1016 ext4_warning(sb, __func__,
1020 "can't read last block, resize aborted"); 1017 "can't read last block, resize aborted");
1021 return -ENOSPC; 1018 return -ENOSPC;
1022 } 1019 }
@@ -1028,13 +1025,13 @@ int ext4_group_extend(struct super_block *sb, struct ext4_super_block *es,
1028 handle = ext4_journal_start_sb(sb, 3); 1025 handle = ext4_journal_start_sb(sb, 3);
1029 if (IS_ERR(handle)) { 1026 if (IS_ERR(handle)) {
1030 err = PTR_ERR(handle); 1027 err = PTR_ERR(handle);
1031 ext4_warning(sb, __FUNCTION__, "error %d on journal start",err); 1028 ext4_warning(sb, __func__, "error %d on journal start", err);
1032 goto exit_put; 1029 goto exit_put;
1033 } 1030 }
1034 1031
1035 lock_super(sb); 1032 lock_super(sb);
1036 if (o_blocks_count != ext4_blocks_count(es)) { 1033 if (o_blocks_count != ext4_blocks_count(es)) {
1037 ext4_warning(sb, __FUNCTION__, 1034 ext4_warning(sb, __func__,
1038 "multiple resizers run on filesystem!"); 1035 "multiple resizers run on filesystem!");
1039 unlock_super(sb); 1036 unlock_super(sb);
1040 ext4_journal_stop(handle); 1037 ext4_journal_stop(handle);
@@ -1044,7 +1041,7 @@ int ext4_group_extend(struct super_block *sb, struct ext4_super_block *es,
1044 1041
1045 if ((err = ext4_journal_get_write_access(handle, 1042 if ((err = ext4_journal_get_write_access(handle,
1046 EXT4_SB(sb)->s_sbh))) { 1043 EXT4_SB(sb)->s_sbh))) {
1047 ext4_warning(sb, __FUNCTION__, 1044 ext4_warning(sb, __func__,
1048 "error %d on journal write access", err); 1045 "error %d on journal write access", err);
1049 unlock_super(sb); 1046 unlock_super(sb);
1050 ext4_journal_stop(handle); 1047 ext4_journal_stop(handle);
diff --git a/fs/ext4/super.c b/fs/ext4/super.c
index c81a8e759bad..52dd0679a4e2 100644
--- a/fs/ext4/super.c
+++ b/fs/ext4/super.c
@@ -21,8 +21,6 @@
21#include <linux/fs.h> 21#include <linux/fs.h>
22#include <linux/time.h> 22#include <linux/time.h>
23#include <linux/jbd2.h> 23#include <linux/jbd2.h>
24#include <linux/ext4_fs.h>
25#include <linux/ext4_jbd2.h>
26#include <linux/slab.h> 24#include <linux/slab.h>
27#include <linux/init.h> 25#include <linux/init.h>
28#include <linux/blkdev.h> 26#include <linux/blkdev.h>
@@ -38,9 +36,10 @@
38#include <linux/seq_file.h> 36#include <linux/seq_file.h>
39#include <linux/log2.h> 37#include <linux/log2.h>
40#include <linux/crc16.h> 38#include <linux/crc16.h>
41
42#include <asm/uaccess.h> 39#include <asm/uaccess.h>
43 40
41#include "ext4.h"
42#include "ext4_jbd2.h"
44#include "xattr.h" 43#include "xattr.h"
45#include "acl.h" 44#include "acl.h"
46#include "namei.h" 45#include "namei.h"
@@ -135,7 +134,7 @@ handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
135 * take the FS itself readonly cleanly. */ 134 * take the FS itself readonly cleanly. */
136 journal = EXT4_SB(sb)->s_journal; 135 journal = EXT4_SB(sb)->s_journal;
137 if (is_journal_aborted(journal)) { 136 if (is_journal_aborted(journal)) {
138 ext4_abort(sb, __FUNCTION__, 137 ext4_abort(sb, __func__,
139 "Detected aborted journal"); 138 "Detected aborted journal");
140 return ERR_PTR(-EROFS); 139 return ERR_PTR(-EROFS);
141 } 140 }
@@ -355,7 +354,7 @@ void ext4_update_dynamic_rev(struct super_block *sb)
355 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV) 354 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
356 return; 355 return;
357 356
358 ext4_warning(sb, __FUNCTION__, 357 ext4_warning(sb, __func__,
359 "updating to rev %d because of new feature flag, " 358 "updating to rev %d because of new feature flag, "
360 "running e2fsck is recommended", 359 "running e2fsck is recommended",
361 EXT4_DYNAMIC_REV); 360 EXT4_DYNAMIC_REV);
@@ -945,8 +944,8 @@ static match_table_t tokens = {
945 {Opt_mballoc, "mballoc"}, 944 {Opt_mballoc, "mballoc"},
946 {Opt_nomballoc, "nomballoc"}, 945 {Opt_nomballoc, "nomballoc"},
947 {Opt_stripe, "stripe=%u"}, 946 {Opt_stripe, "stripe=%u"},
948 {Opt_err, NULL},
949 {Opt_resize, "resize"}, 947 {Opt_resize, "resize"},
948 {Opt_err, NULL},
950}; 949};
951 950
952static ext4_fsblk_t get_sb_block(void **data) 951static ext4_fsblk_t get_sb_block(void **data)
@@ -1388,11 +1387,11 @@ static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1388 * a plain journaled filesystem we can keep it set as 1387 * a plain journaled filesystem we can keep it set as
1389 * valid forever! :) 1388 * valid forever! :)
1390 */ 1389 */
1391 es->s_state = cpu_to_le16(le16_to_cpu(es->s_state) & ~EXT4_VALID_FS); 1390 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1392#endif 1391#endif
1393 if (!(__s16) le16_to_cpu(es->s_max_mnt_count)) 1392 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1394 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT); 1393 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1395 es->s_mnt_count=cpu_to_le16(le16_to_cpu(es->s_mnt_count) + 1); 1394 le16_add_cpu(&es->s_mnt_count, 1);
1396 es->s_mtime = cpu_to_le32(get_seconds()); 1395 es->s_mtime = cpu_to_le32(get_seconds());
1397 ext4_update_dynamic_rev(sb); 1396 ext4_update_dynamic_rev(sb);
1398 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER); 1397 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
@@ -1485,36 +1484,33 @@ static int ext4_check_descriptors(struct super_block *sb)
1485 block_bitmap = ext4_block_bitmap(sb, gdp); 1484 block_bitmap = ext4_block_bitmap(sb, gdp);
1486 if (block_bitmap < first_block || block_bitmap > last_block) 1485 if (block_bitmap < first_block || block_bitmap > last_block)
1487 { 1486 {
1488 ext4_error (sb, "ext4_check_descriptors", 1487 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1489 "Block bitmap for group %lu" 1488 "Block bitmap for group %lu not in group "
1490 " not in group (block %llu)!", 1489 "(block %llu)!", i, block_bitmap);
1491 i, block_bitmap);
1492 return 0; 1490 return 0;
1493 } 1491 }
1494 inode_bitmap = ext4_inode_bitmap(sb, gdp); 1492 inode_bitmap = ext4_inode_bitmap(sb, gdp);
1495 if (inode_bitmap < first_block || inode_bitmap > last_block) 1493 if (inode_bitmap < first_block || inode_bitmap > last_block)
1496 { 1494 {
1497 ext4_error (sb, "ext4_check_descriptors", 1495 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1498 "Inode bitmap for group %lu" 1496 "Inode bitmap for group %lu not in group "
1499 " not in group (block %llu)!", 1497 "(block %llu)!", i, inode_bitmap);
1500 i, inode_bitmap);
1501 return 0; 1498 return 0;
1502 } 1499 }
1503 inode_table = ext4_inode_table(sb, gdp); 1500 inode_table = ext4_inode_table(sb, gdp);
1504 if (inode_table < first_block || 1501 if (inode_table < first_block ||
1505 inode_table + sbi->s_itb_per_group - 1 > last_block) 1502 inode_table + sbi->s_itb_per_group - 1 > last_block)
1506 { 1503 {
1507 ext4_error (sb, "ext4_check_descriptors", 1504 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1508 "Inode table for group %lu" 1505 "Inode table for group %lu not in group "
1509 " not in group (block %llu)!", 1506 "(block %llu)!", i, inode_table);
1510 i, inode_table);
1511 return 0; 1507 return 0;
1512 } 1508 }
1513 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) { 1509 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
1514 ext4_error(sb, __FUNCTION__, 1510 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1515 "Checksum for group %lu failed (%u!=%u)\n", 1511 "Checksum for group %lu failed (%u!=%u)\n",
1516 i, le16_to_cpu(ext4_group_desc_csum(sbi, i, 1512 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1517 gdp)), le16_to_cpu(gdp->bg_checksum)); 1513 gdp)), le16_to_cpu(gdp->bg_checksum));
1518 return 0; 1514 return 0;
1519 } 1515 }
1520 if (!flexbg_flag) 1516 if (!flexbg_flag)
@@ -1594,8 +1590,8 @@ static void ext4_orphan_cleanup (struct super_block * sb,
1594 while (es->s_last_orphan) { 1590 while (es->s_last_orphan) {
1595 struct inode *inode; 1591 struct inode *inode;
1596 1592
1597 if (!(inode = 1593 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1598 ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan)))) { 1594 if (IS_ERR(inode)) {
1599 es->s_last_orphan = 0; 1595 es->s_last_orphan = 0;
1600 break; 1596 break;
1601 } 1597 }
@@ -1605,7 +1601,7 @@ static void ext4_orphan_cleanup (struct super_block * sb,
1605 if (inode->i_nlink) { 1601 if (inode->i_nlink) {
1606 printk(KERN_DEBUG 1602 printk(KERN_DEBUG
1607 "%s: truncating inode %lu to %Ld bytes\n", 1603 "%s: truncating inode %lu to %Ld bytes\n",
1608 __FUNCTION__, inode->i_ino, inode->i_size); 1604 __func__, inode->i_ino, inode->i_size);
1609 jbd_debug(2, "truncating inode %lu to %Ld bytes\n", 1605 jbd_debug(2, "truncating inode %lu to %Ld bytes\n",
1610 inode->i_ino, inode->i_size); 1606 inode->i_ino, inode->i_size);
1611 ext4_truncate(inode); 1607 ext4_truncate(inode);
@@ -1613,7 +1609,7 @@ static void ext4_orphan_cleanup (struct super_block * sb,
1613 } else { 1609 } else {
1614 printk(KERN_DEBUG 1610 printk(KERN_DEBUG
1615 "%s: deleting unreferenced inode %lu\n", 1611 "%s: deleting unreferenced inode %lu\n",
1616 __FUNCTION__, inode->i_ino); 1612 __func__, inode->i_ino);
1617 jbd_debug(2, "deleting unreferenced inode %lu\n", 1613 jbd_debug(2, "deleting unreferenced inode %lu\n",
1618 inode->i_ino); 1614 inode->i_ino);
1619 nr_orphans++; 1615 nr_orphans++;
@@ -2699,9 +2695,9 @@ static void ext4_clear_journal_err(struct super_block * sb,
2699 char nbuf[16]; 2695 char nbuf[16];
2700 2696
2701 errstr = ext4_decode_error(sb, j_errno, nbuf); 2697 errstr = ext4_decode_error(sb, j_errno, nbuf);
2702 ext4_warning(sb, __FUNCTION__, "Filesystem error recorded " 2698 ext4_warning(sb, __func__, "Filesystem error recorded "
2703 "from previous mount: %s", errstr); 2699 "from previous mount: %s", errstr);
2704 ext4_warning(sb, __FUNCTION__, "Marking fs in need of " 2700 ext4_warning(sb, __func__, "Marking fs in need of "
2705 "filesystem check."); 2701 "filesystem check.");
2706 2702
2707 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS; 2703 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
@@ -2828,7 +2824,7 @@ static int ext4_remount (struct super_block * sb, int * flags, char * data)
2828 } 2824 }
2829 2825
2830 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT) 2826 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT)
2831 ext4_abort(sb, __FUNCTION__, "Abort forced by user"); 2827 ext4_abort(sb, __func__, "Abort forced by user");
2832 2828
2833 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) | 2829 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
2834 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0); 2830 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
@@ -3040,8 +3036,14 @@ static int ext4_dquot_drop(struct inode *inode)
3040 3036
3041 /* We may delete quota structure so we need to reserve enough blocks */ 3037 /* We may delete quota structure so we need to reserve enough blocks */
3042 handle = ext4_journal_start(inode, 2*EXT4_QUOTA_DEL_BLOCKS(inode->i_sb)); 3038 handle = ext4_journal_start(inode, 2*EXT4_QUOTA_DEL_BLOCKS(inode->i_sb));
3043 if (IS_ERR(handle)) 3039 if (IS_ERR(handle)) {
3040 /*
3041 * We call dquot_drop() anyway to at least release references
3042 * to quota structures so that umount does not hang.
3043 */
3044 dquot_drop(inode);
3044 return PTR_ERR(handle); 3045 return PTR_ERR(handle);
3046 }
3045 ret = dquot_drop(inode); 3047 ret = dquot_drop(inode);
3046 err = ext4_journal_stop(handle); 3048 err = ext4_journal_stop(handle);
3047 if (!ret) 3049 if (!ret)
diff --git a/fs/ext4/symlink.c b/fs/ext4/symlink.c
index e6f9da4287c4..e9178643dc01 100644
--- a/fs/ext4/symlink.c
+++ b/fs/ext4/symlink.c
@@ -19,8 +19,8 @@
19 19
20#include <linux/fs.h> 20#include <linux/fs.h>
21#include <linux/jbd2.h> 21#include <linux/jbd2.h>
22#include <linux/ext4_fs.h>
23#include <linux/namei.h> 22#include <linux/namei.h>
23#include "ext4.h"
24#include "xattr.h" 24#include "xattr.h"
25 25
26static void * ext4_follow_link(struct dentry *dentry, struct nameidata *nd) 26static void * ext4_follow_link(struct dentry *dentry, struct nameidata *nd)
diff --git a/fs/ext4/xattr.c b/fs/ext4/xattr.c
index e9054c1c7d93..3fbc2c6c3d0e 100644
--- a/fs/ext4/xattr.c
+++ b/fs/ext4/xattr.c
@@ -53,11 +53,11 @@
53#include <linux/init.h> 53#include <linux/init.h>
54#include <linux/fs.h> 54#include <linux/fs.h>
55#include <linux/slab.h> 55#include <linux/slab.h>
56#include <linux/ext4_jbd2.h>
57#include <linux/ext4_fs.h>
58#include <linux/mbcache.h> 56#include <linux/mbcache.h>
59#include <linux/quotaops.h> 57#include <linux/quotaops.h>
60#include <linux/rwsem.h> 58#include <linux/rwsem.h>
59#include "ext4_jbd2.h"
60#include "ext4.h"
61#include "xattr.h" 61#include "xattr.h"
62#include "acl.h" 62#include "acl.h"
63 63
@@ -92,6 +92,8 @@ static struct buffer_head *ext4_xattr_cache_find(struct inode *,
92 struct mb_cache_entry **); 92 struct mb_cache_entry **);
93static void ext4_xattr_rehash(struct ext4_xattr_header *, 93static void ext4_xattr_rehash(struct ext4_xattr_header *,
94 struct ext4_xattr_entry *); 94 struct ext4_xattr_entry *);
95static int ext4_xattr_list(struct inode *inode, char *buffer,
96 size_t buffer_size);
95 97
96static struct mb_cache *ext4_xattr_cache; 98static struct mb_cache *ext4_xattr_cache;
97 99
@@ -225,7 +227,7 @@ ext4_xattr_block_get(struct inode *inode, int name_index, const char *name,
225 ea_bdebug(bh, "b_count=%d, refcount=%d", 227 ea_bdebug(bh, "b_count=%d, refcount=%d",
226 atomic_read(&(bh->b_count)), le32_to_cpu(BHDR(bh)->h_refcount)); 228 atomic_read(&(bh->b_count)), le32_to_cpu(BHDR(bh)->h_refcount));
227 if (ext4_xattr_check_block(bh)) { 229 if (ext4_xattr_check_block(bh)) {
228bad_block: ext4_error(inode->i_sb, __FUNCTION__, 230bad_block: ext4_error(inode->i_sb, __func__,
229 "inode %lu: bad block %llu", inode->i_ino, 231 "inode %lu: bad block %llu", inode->i_ino,
230 EXT4_I(inode)->i_file_acl); 232 EXT4_I(inode)->i_file_acl);
231 error = -EIO; 233 error = -EIO;
@@ -367,7 +369,7 @@ ext4_xattr_block_list(struct inode *inode, char *buffer, size_t buffer_size)
367 ea_bdebug(bh, "b_count=%d, refcount=%d", 369 ea_bdebug(bh, "b_count=%d, refcount=%d",
368 atomic_read(&(bh->b_count)), le32_to_cpu(BHDR(bh)->h_refcount)); 370 atomic_read(&(bh->b_count)), le32_to_cpu(BHDR(bh)->h_refcount));
369 if (ext4_xattr_check_block(bh)) { 371 if (ext4_xattr_check_block(bh)) {
370 ext4_error(inode->i_sb, __FUNCTION__, 372 ext4_error(inode->i_sb, __func__,
371 "inode %lu: bad block %llu", inode->i_ino, 373 "inode %lu: bad block %llu", inode->i_ino,
372 EXT4_I(inode)->i_file_acl); 374 EXT4_I(inode)->i_file_acl);
373 error = -EIO; 375 error = -EIO;
@@ -420,7 +422,7 @@ cleanup:
420 * Returns a negative error number on failure, or the number of bytes 422 * Returns a negative error number on failure, or the number of bytes
421 * used / required on success. 423 * used / required on success.
422 */ 424 */
423int 425static int
424ext4_xattr_list(struct inode *inode, char *buffer, size_t buffer_size) 426ext4_xattr_list(struct inode *inode, char *buffer, size_t buffer_size)
425{ 427{
426 int i_error, b_error; 428 int i_error, b_error;
@@ -484,8 +486,7 @@ ext4_xattr_release_block(handle_t *handle, struct inode *inode,
484 get_bh(bh); 486 get_bh(bh);
485 ext4_forget(handle, 1, inode, bh, bh->b_blocknr); 487 ext4_forget(handle, 1, inode, bh, bh->b_blocknr);
486 } else { 488 } else {
487 BHDR(bh)->h_refcount = cpu_to_le32( 489 le32_add_cpu(&BHDR(bh)->h_refcount, -1);
488 le32_to_cpu(BHDR(bh)->h_refcount) - 1);
489 error = ext4_journal_dirty_metadata(handle, bh); 490 error = ext4_journal_dirty_metadata(handle, bh);
490 if (IS_SYNC(inode)) 491 if (IS_SYNC(inode))
491 handle->h_sync = 1; 492 handle->h_sync = 1;
@@ -660,7 +661,7 @@ ext4_xattr_block_find(struct inode *inode, struct ext4_xattr_info *i,
660 atomic_read(&(bs->bh->b_count)), 661 atomic_read(&(bs->bh->b_count)),
661 le32_to_cpu(BHDR(bs->bh)->h_refcount)); 662 le32_to_cpu(BHDR(bs->bh)->h_refcount));
662 if (ext4_xattr_check_block(bs->bh)) { 663 if (ext4_xattr_check_block(bs->bh)) {
663 ext4_error(sb, __FUNCTION__, 664 ext4_error(sb, __func__,
664 "inode %lu: bad block %llu", inode->i_ino, 665 "inode %lu: bad block %llu", inode->i_ino,
665 EXT4_I(inode)->i_file_acl); 666 EXT4_I(inode)->i_file_acl);
666 error = -EIO; 667 error = -EIO;
@@ -738,7 +739,7 @@ ext4_xattr_block_set(handle_t *handle, struct inode *inode,
738 ce = NULL; 739 ce = NULL;
739 } 740 }
740 ea_bdebug(bs->bh, "cloning"); 741 ea_bdebug(bs->bh, "cloning");
741 s->base = kmalloc(bs->bh->b_size, GFP_KERNEL); 742 s->base = kmalloc(bs->bh->b_size, GFP_NOFS);
742 error = -ENOMEM; 743 error = -ENOMEM;
743 if (s->base == NULL) 744 if (s->base == NULL)
744 goto cleanup; 745 goto cleanup;
@@ -750,7 +751,7 @@ ext4_xattr_block_set(handle_t *handle, struct inode *inode,
750 } 751 }
751 } else { 752 } else {
752 /* Allocate a buffer where we construct the new block. */ 753 /* Allocate a buffer where we construct the new block. */
753 s->base = kzalloc(sb->s_blocksize, GFP_KERNEL); 754 s->base = kzalloc(sb->s_blocksize, GFP_NOFS);
754 /* assert(header == s->base) */ 755 /* assert(header == s->base) */
755 error = -ENOMEM; 756 error = -ENOMEM;
756 if (s->base == NULL) 757 if (s->base == NULL)
@@ -789,8 +790,7 @@ inserted:
789 if (error) 790 if (error)
790 goto cleanup_dquot; 791 goto cleanup_dquot;
791 lock_buffer(new_bh); 792 lock_buffer(new_bh);
792 BHDR(new_bh)->h_refcount = cpu_to_le32(1 + 793 le32_add_cpu(&BHDR(new_bh)->h_refcount, 1);
793 le32_to_cpu(BHDR(new_bh)->h_refcount));
794 ea_bdebug(new_bh, "reusing; refcount now=%d", 794 ea_bdebug(new_bh, "reusing; refcount now=%d",
795 le32_to_cpu(BHDR(new_bh)->h_refcount)); 795 le32_to_cpu(BHDR(new_bh)->h_refcount));
796 unlock_buffer(new_bh); 796 unlock_buffer(new_bh);
@@ -808,10 +808,8 @@ inserted:
808 get_bh(new_bh); 808 get_bh(new_bh);
809 } else { 809 } else {
810 /* We need to allocate a new block */ 810 /* We need to allocate a new block */
811 ext4_fsblk_t goal = le32_to_cpu( 811 ext4_fsblk_t goal = ext4_group_first_block_no(sb,
812 EXT4_SB(sb)->s_es->s_first_data_block) + 812 EXT4_I(inode)->i_block_group);
813 (ext4_fsblk_t)EXT4_I(inode)->i_block_group *
814 EXT4_BLOCKS_PER_GROUP(sb);
815 ext4_fsblk_t block = ext4_new_block(handle, inode, 813 ext4_fsblk_t block = ext4_new_block(handle, inode,
816 goal, &error); 814 goal, &error);
817 if (error) 815 if (error)
@@ -863,7 +861,7 @@ cleanup_dquot:
863 goto cleanup; 861 goto cleanup;
864 862
865bad_block: 863bad_block:
866 ext4_error(inode->i_sb, __FUNCTION__, 864 ext4_error(inode->i_sb, __func__,
867 "inode %lu: bad block %llu", inode->i_ino, 865 "inode %lu: bad block %llu", inode->i_ino,
868 EXT4_I(inode)->i_file_acl); 866 EXT4_I(inode)->i_file_acl);
869 goto cleanup; 867 goto cleanup;
@@ -1166,7 +1164,7 @@ retry:
1166 if (!bh) 1164 if (!bh)
1167 goto cleanup; 1165 goto cleanup;
1168 if (ext4_xattr_check_block(bh)) { 1166 if (ext4_xattr_check_block(bh)) {
1169 ext4_error(inode->i_sb, __FUNCTION__, 1167 ext4_error(inode->i_sb, __func__,
1170 "inode %lu: bad block %llu", inode->i_ino, 1168 "inode %lu: bad block %llu", inode->i_ino,
1171 EXT4_I(inode)->i_file_acl); 1169 EXT4_I(inode)->i_file_acl);
1172 error = -EIO; 1170 error = -EIO;
@@ -1341,14 +1339,14 @@ ext4_xattr_delete_inode(handle_t *handle, struct inode *inode)
1341 goto cleanup; 1339 goto cleanup;
1342 bh = sb_bread(inode->i_sb, EXT4_I(inode)->i_file_acl); 1340 bh = sb_bread(inode->i_sb, EXT4_I(inode)->i_file_acl);
1343 if (!bh) { 1341 if (!bh) {
1344 ext4_error(inode->i_sb, __FUNCTION__, 1342 ext4_error(inode->i_sb, __func__,
1345 "inode %lu: block %llu read error", inode->i_ino, 1343 "inode %lu: block %llu read error", inode->i_ino,
1346 EXT4_I(inode)->i_file_acl); 1344 EXT4_I(inode)->i_file_acl);
1347 goto cleanup; 1345 goto cleanup;
1348 } 1346 }
1349 if (BHDR(bh)->h_magic != cpu_to_le32(EXT4_XATTR_MAGIC) || 1347 if (BHDR(bh)->h_magic != cpu_to_le32(EXT4_XATTR_MAGIC) ||
1350 BHDR(bh)->h_blocks != cpu_to_le32(1)) { 1348 BHDR(bh)->h_blocks != cpu_to_le32(1)) {
1351 ext4_error(inode->i_sb, __FUNCTION__, 1349 ext4_error(inode->i_sb, __func__,
1352 "inode %lu: bad block %llu", inode->i_ino, 1350 "inode %lu: bad block %llu", inode->i_ino,
1353 EXT4_I(inode)->i_file_acl); 1351 EXT4_I(inode)->i_file_acl);
1354 goto cleanup; 1352 goto cleanup;
@@ -1475,7 +1473,7 @@ again:
1475 } 1473 }
1476 bh = sb_bread(inode->i_sb, ce->e_block); 1474 bh = sb_bread(inode->i_sb, ce->e_block);
1477 if (!bh) { 1475 if (!bh) {
1478 ext4_error(inode->i_sb, __FUNCTION__, 1476 ext4_error(inode->i_sb, __func__,
1479 "inode %lu: block %lu read error", 1477 "inode %lu: block %lu read error",
1480 inode->i_ino, (unsigned long) ce->e_block); 1478 inode->i_ino, (unsigned long) ce->e_block);
1481 } else if (le32_to_cpu(BHDR(bh)->h_refcount) >= 1479 } else if (le32_to_cpu(BHDR(bh)->h_refcount) >=
diff --git a/fs/ext4/xattr.h b/fs/ext4/xattr.h
index d7f5d6a12651..5992fe979bb9 100644
--- a/fs/ext4/xattr.h
+++ b/fs/ext4/xattr.h
@@ -74,7 +74,6 @@ extern struct xattr_handler ext4_xattr_security_handler;
74extern ssize_t ext4_listxattr(struct dentry *, char *, size_t); 74extern ssize_t ext4_listxattr(struct dentry *, char *, size_t);
75 75
76extern int ext4_xattr_get(struct inode *, int, const char *, void *, size_t); 76extern int ext4_xattr_get(struct inode *, int, const char *, void *, size_t);
77extern int ext4_xattr_list(struct inode *, char *, size_t);
78extern int ext4_xattr_set(struct inode *, int, const char *, const void *, size_t, int); 77extern int ext4_xattr_set(struct inode *, int, const char *, const void *, size_t, int);
79extern int ext4_xattr_set_handle(handle_t *, struct inode *, int, const char *, const void *, size_t, int); 78extern int ext4_xattr_set_handle(handle_t *, struct inode *, int, const char *, const void *, size_t, int);
80 79
@@ -99,12 +98,6 @@ ext4_xattr_get(struct inode *inode, int name_index, const char *name,
99} 98}
100 99
101static inline int 100static inline int
102ext4_xattr_list(struct inode *inode, void *buffer, size_t size)
103{
104 return -EOPNOTSUPP;
105}
106
107static inline int
108ext4_xattr_set(struct inode *inode, int name_index, const char *name, 101ext4_xattr_set(struct inode *inode, int name_index, const char *name,
109 const void *value, size_t size, int flags) 102 const void *value, size_t size, int flags)
110{ 103{
diff --git a/fs/ext4/xattr_security.c b/fs/ext4/xattr_security.c
index f17eaf2321b9..ca5f89fc6cae 100644
--- a/fs/ext4/xattr_security.c
+++ b/fs/ext4/xattr_security.c
@@ -6,9 +6,9 @@
6#include <linux/module.h> 6#include <linux/module.h>
7#include <linux/string.h> 7#include <linux/string.h>
8#include <linux/fs.h> 8#include <linux/fs.h>
9#include <linux/ext4_jbd2.h>
10#include <linux/ext4_fs.h>
11#include <linux/security.h> 9#include <linux/security.h>
10#include "ext4_jbd2.h"
11#include "ext4.h"
12#include "xattr.h" 12#include "xattr.h"
13 13
14static size_t 14static size_t
diff --git a/fs/ext4/xattr_trusted.c b/fs/ext4/xattr_trusted.c
index e0f05acdafec..fff33382cadc 100644
--- a/fs/ext4/xattr_trusted.c
+++ b/fs/ext4/xattr_trusted.c
@@ -9,8 +9,8 @@
9#include <linux/string.h> 9#include <linux/string.h>
10#include <linux/capability.h> 10#include <linux/capability.h>
11#include <linux/fs.h> 11#include <linux/fs.h>
12#include <linux/ext4_jbd2.h> 12#include "ext4_jbd2.h"
13#include <linux/ext4_fs.h> 13#include "ext4.h"
14#include "xattr.h" 14#include "xattr.h"
15 15
16#define XATTR_TRUSTED_PREFIX "trusted." 16#define XATTR_TRUSTED_PREFIX "trusted."
diff --git a/fs/ext4/xattr_user.c b/fs/ext4/xattr_user.c
index 7ed3d8ebf096..67be723fcc4e 100644
--- a/fs/ext4/xattr_user.c
+++ b/fs/ext4/xattr_user.c
@@ -8,8 +8,8 @@
8#include <linux/module.h> 8#include <linux/module.h>
9#include <linux/string.h> 9#include <linux/string.h>
10#include <linux/fs.h> 10#include <linux/fs.h>
11#include <linux/ext4_jbd2.h> 11#include "ext4_jbd2.h"
12#include <linux/ext4_fs.h> 12#include "ext4.h"
13#include "xattr.h" 13#include "xattr.h"
14 14
15#define XATTR_USER_PREFIX "user." 15#define XATTR_USER_PREFIX "user."
diff --git a/fs/jbd2/commit.c b/fs/jbd2/commit.c
index a8173081f831..e0139786f717 100644
--- a/fs/jbd2/commit.c
+++ b/fs/jbd2/commit.c
@@ -520,22 +520,6 @@ void jbd2_journal_commit_transaction(journal_t *journal)
520 jbd_debug (3, "JBD: commit phase 2\n"); 520 jbd_debug (3, "JBD: commit phase 2\n");
521 521
522 /* 522 /*
523 * First, drop modified flag: all accesses to the buffers
524 * will be tracked for a new trasaction only -bzzz
525 */
526 spin_lock(&journal->j_list_lock);
527 if (commit_transaction->t_buffers) {
528 new_jh = jh = commit_transaction->t_buffers->b_tnext;
529 do {
530 J_ASSERT_JH(new_jh, new_jh->b_modified == 1 ||
531 new_jh->b_modified == 0);
532 new_jh->b_modified = 0;
533 new_jh = new_jh->b_tnext;
534 } while (new_jh != jh);
535 }
536 spin_unlock(&journal->j_list_lock);
537
538 /*
539 * Now start flushing things to disk, in the order they appear 523 * Now start flushing things to disk, in the order they appear
540 * on the transaction lists. Data blocks go first. 524 * on the transaction lists. Data blocks go first.
541 */ 525 */
@@ -584,6 +568,9 @@ void jbd2_journal_commit_transaction(journal_t *journal)
584 stats.u.run.rs_blocks = commit_transaction->t_outstanding_credits; 568 stats.u.run.rs_blocks = commit_transaction->t_outstanding_credits;
585 stats.u.run.rs_blocks_logged = 0; 569 stats.u.run.rs_blocks_logged = 0;
586 570
571 J_ASSERT(commit_transaction->t_nr_buffers <=
572 commit_transaction->t_outstanding_credits);
573
587 descriptor = NULL; 574 descriptor = NULL;
588 bufs = 0; 575 bufs = 0;
589 while (commit_transaction->t_buffers) { 576 while (commit_transaction->t_buffers) {
diff --git a/fs/jbd2/journal.c b/fs/jbd2/journal.c
index eb7eb6c27bcb..53632e3e8457 100644
--- a/fs/jbd2/journal.c
+++ b/fs/jbd2/journal.c
@@ -534,7 +534,7 @@ int jbd2_log_wait_commit(journal_t *journal, tid_t tid)
534 if (!tid_geq(journal->j_commit_request, tid)) { 534 if (!tid_geq(journal->j_commit_request, tid)) {
535 printk(KERN_EMERG 535 printk(KERN_EMERG
536 "%s: error: j_commit_request=%d, tid=%d\n", 536 "%s: error: j_commit_request=%d, tid=%d\n",
537 __FUNCTION__, journal->j_commit_request, tid); 537 __func__, journal->j_commit_request, tid);
538 } 538 }
539 spin_unlock(&journal->j_state_lock); 539 spin_unlock(&journal->j_state_lock);
540#endif 540#endif
@@ -599,7 +599,7 @@ int jbd2_journal_bmap(journal_t *journal, unsigned long blocknr,
599 599
600 printk(KERN_ALERT "%s: journal block not found " 600 printk(KERN_ALERT "%s: journal block not found "
601 "at offset %lu on %s\n", 601 "at offset %lu on %s\n",
602 __FUNCTION__, 602 __func__,
603 blocknr, 603 blocknr,
604 bdevname(journal->j_dev, b)); 604 bdevname(journal->j_dev, b));
605 err = -EIO; 605 err = -EIO;
@@ -997,13 +997,14 @@ fail:
997 */ 997 */
998 998
999/** 999/**
1000 * journal_t * jbd2_journal_init_dev() - creates an initialises a journal structure 1000 * journal_t * jbd2_journal_init_dev() - creates and initialises a journal structure
1001 * @bdev: Block device on which to create the journal 1001 * @bdev: Block device on which to create the journal
1002 * @fs_dev: Device which hold journalled filesystem for this journal. 1002 * @fs_dev: Device which hold journalled filesystem for this journal.
1003 * @start: Block nr Start of journal. 1003 * @start: Block nr Start of journal.
1004 * @len: Length of the journal in blocks. 1004 * @len: Length of the journal in blocks.
1005 * @blocksize: blocksize of journalling device 1005 * @blocksize: blocksize of journalling device
1006 * @returns: a newly created journal_t * 1006 *
1007 * Returns: a newly created journal_t *
1007 * 1008 *
1008 * jbd2_journal_init_dev creates a journal which maps a fixed contiguous 1009 * jbd2_journal_init_dev creates a journal which maps a fixed contiguous
1009 * range of blocks on an arbitrary block device. 1010 * range of blocks on an arbitrary block device.
@@ -1027,7 +1028,7 @@ journal_t * jbd2_journal_init_dev(struct block_device *bdev,
1027 journal->j_wbuf = kmalloc(n * sizeof(struct buffer_head*), GFP_KERNEL); 1028 journal->j_wbuf = kmalloc(n * sizeof(struct buffer_head*), GFP_KERNEL);
1028 if (!journal->j_wbuf) { 1029 if (!journal->j_wbuf) {
1029 printk(KERN_ERR "%s: Cant allocate bhs for commit thread\n", 1030 printk(KERN_ERR "%s: Cant allocate bhs for commit thread\n",
1030 __FUNCTION__); 1031 __func__);
1031 kfree(journal); 1032 kfree(journal);
1032 journal = NULL; 1033 journal = NULL;
1033 goto out; 1034 goto out;
@@ -1083,7 +1084,7 @@ journal_t * jbd2_journal_init_inode (struct inode *inode)
1083 journal->j_wbuf = kmalloc(n * sizeof(struct buffer_head*), GFP_KERNEL); 1084 journal->j_wbuf = kmalloc(n * sizeof(struct buffer_head*), GFP_KERNEL);
1084 if (!journal->j_wbuf) { 1085 if (!journal->j_wbuf) {
1085 printk(KERN_ERR "%s: Cant allocate bhs for commit thread\n", 1086 printk(KERN_ERR "%s: Cant allocate bhs for commit thread\n",
1086 __FUNCTION__); 1087 __func__);
1087 kfree(journal); 1088 kfree(journal);
1088 return NULL; 1089 return NULL;
1089 } 1090 }
@@ -1092,7 +1093,7 @@ journal_t * jbd2_journal_init_inode (struct inode *inode)
1092 /* If that failed, give up */ 1093 /* If that failed, give up */
1093 if (err) { 1094 if (err) {
1094 printk(KERN_ERR "%s: Cannnot locate journal superblock\n", 1095 printk(KERN_ERR "%s: Cannnot locate journal superblock\n",
1095 __FUNCTION__); 1096 __func__);
1096 kfree(journal); 1097 kfree(journal);
1097 return NULL; 1098 return NULL;
1098 } 1099 }
@@ -1178,7 +1179,7 @@ int jbd2_journal_create(journal_t *journal)
1178 */ 1179 */
1179 printk(KERN_EMERG 1180 printk(KERN_EMERG
1180 "%s: creation of journal on external device!\n", 1181 "%s: creation of journal on external device!\n",
1181 __FUNCTION__); 1182 __func__);
1182 BUG(); 1183 BUG();
1183 } 1184 }
1184 1185
@@ -1976,9 +1977,10 @@ static int journal_init_jbd2_journal_head_cache(void)
1976 1977
1977static void jbd2_journal_destroy_jbd2_journal_head_cache(void) 1978static void jbd2_journal_destroy_jbd2_journal_head_cache(void)
1978{ 1979{
1979 J_ASSERT(jbd2_journal_head_cache != NULL); 1980 if (jbd2_journal_head_cache) {
1980 kmem_cache_destroy(jbd2_journal_head_cache); 1981 kmem_cache_destroy(jbd2_journal_head_cache);
1981 jbd2_journal_head_cache = NULL; 1982 jbd2_journal_head_cache = NULL;
1983 }
1982} 1984}
1983 1985
1984/* 1986/*
@@ -1997,7 +1999,7 @@ static struct journal_head *journal_alloc_journal_head(void)
1997 jbd_debug(1, "out of memory for journal_head\n"); 1999 jbd_debug(1, "out of memory for journal_head\n");
1998 if (time_after(jiffies, last_warning + 5*HZ)) { 2000 if (time_after(jiffies, last_warning + 5*HZ)) {
1999 printk(KERN_NOTICE "ENOMEM in %s, retrying.\n", 2001 printk(KERN_NOTICE "ENOMEM in %s, retrying.\n",
2000 __FUNCTION__); 2002 __func__);
2001 last_warning = jiffies; 2003 last_warning = jiffies;
2002 } 2004 }
2003 while (!ret) { 2005 while (!ret) {
@@ -2134,13 +2136,13 @@ static void __journal_remove_journal_head(struct buffer_head *bh)
2134 if (jh->b_frozen_data) { 2136 if (jh->b_frozen_data) {
2135 printk(KERN_WARNING "%s: freeing " 2137 printk(KERN_WARNING "%s: freeing "
2136 "b_frozen_data\n", 2138 "b_frozen_data\n",
2137 __FUNCTION__); 2139 __func__);
2138 jbd2_free(jh->b_frozen_data, bh->b_size); 2140 jbd2_free(jh->b_frozen_data, bh->b_size);
2139 } 2141 }
2140 if (jh->b_committed_data) { 2142 if (jh->b_committed_data) {
2141 printk(KERN_WARNING "%s: freeing " 2143 printk(KERN_WARNING "%s: freeing "
2142 "b_committed_data\n", 2144 "b_committed_data\n",
2143 __FUNCTION__); 2145 __func__);
2144 jbd2_free(jh->b_committed_data, bh->b_size); 2146 jbd2_free(jh->b_committed_data, bh->b_size);
2145 } 2147 }
2146 bh->b_private = NULL; 2148 bh->b_private = NULL;
@@ -2305,10 +2307,12 @@ static int __init journal_init(void)
2305 BUILD_BUG_ON(sizeof(struct journal_superblock_s) != 1024); 2307 BUILD_BUG_ON(sizeof(struct journal_superblock_s) != 1024);
2306 2308
2307 ret = journal_init_caches(); 2309 ret = journal_init_caches();
2308 if (ret != 0) 2310 if (ret == 0) {
2311 jbd2_create_debugfs_entry();
2312 jbd2_create_jbd_stats_proc_entry();
2313 } else {
2309 jbd2_journal_destroy_caches(); 2314 jbd2_journal_destroy_caches();
2310 jbd2_create_debugfs_entry(); 2315 }
2311 jbd2_create_jbd_stats_proc_entry();
2312 return ret; 2316 return ret;
2313} 2317}
2314 2318
diff --git a/fs/jbd2/revoke.c b/fs/jbd2/revoke.c
index 2e1453a5e998..257ff2625765 100644
--- a/fs/jbd2/revoke.c
+++ b/fs/jbd2/revoke.c
@@ -139,7 +139,7 @@ repeat:
139oom: 139oom:
140 if (!journal_oom_retry) 140 if (!journal_oom_retry)
141 return -ENOMEM; 141 return -ENOMEM;
142 jbd_debug(1, "ENOMEM in %s, retrying\n", __FUNCTION__); 142 jbd_debug(1, "ENOMEM in %s, retrying\n", __func__);
143 yield(); 143 yield();
144 goto repeat; 144 goto repeat;
145} 145}
@@ -167,138 +167,121 @@ static struct jbd2_revoke_record_s *find_revoke_record(journal_t *journal,
167 return NULL; 167 return NULL;
168} 168}
169 169
170void jbd2_journal_destroy_revoke_caches(void)
171{
172 if (jbd2_revoke_record_cache) {
173 kmem_cache_destroy(jbd2_revoke_record_cache);
174 jbd2_revoke_record_cache = NULL;
175 }
176 if (jbd2_revoke_table_cache) {
177 kmem_cache_destroy(jbd2_revoke_table_cache);
178 jbd2_revoke_table_cache = NULL;
179 }
180}
181
170int __init jbd2_journal_init_revoke_caches(void) 182int __init jbd2_journal_init_revoke_caches(void)
171{ 183{
184 J_ASSERT(!jbd2_revoke_record_cache);
185 J_ASSERT(!jbd2_revoke_table_cache);
186
172 jbd2_revoke_record_cache = kmem_cache_create("jbd2_revoke_record", 187 jbd2_revoke_record_cache = kmem_cache_create("jbd2_revoke_record",
173 sizeof(struct jbd2_revoke_record_s), 188 sizeof(struct jbd2_revoke_record_s),
174 0, 189 0,
175 SLAB_HWCACHE_ALIGN|SLAB_TEMPORARY, 190 SLAB_HWCACHE_ALIGN|SLAB_TEMPORARY,
176 NULL); 191 NULL);
177 if (!jbd2_revoke_record_cache) 192 if (!jbd2_revoke_record_cache)
178 return -ENOMEM; 193 goto record_cache_failure;
179 194
180 jbd2_revoke_table_cache = kmem_cache_create("jbd2_revoke_table", 195 jbd2_revoke_table_cache = kmem_cache_create("jbd2_revoke_table",
181 sizeof(struct jbd2_revoke_table_s), 196 sizeof(struct jbd2_revoke_table_s),
182 0, SLAB_TEMPORARY, NULL); 197 0, SLAB_TEMPORARY, NULL);
183 if (!jbd2_revoke_table_cache) { 198 if (!jbd2_revoke_table_cache)
184 kmem_cache_destroy(jbd2_revoke_record_cache); 199 goto table_cache_failure;
185 jbd2_revoke_record_cache = NULL;
186 return -ENOMEM;
187 }
188 return 0; 200 return 0;
201table_cache_failure:
202 jbd2_journal_destroy_revoke_caches();
203record_cache_failure:
204 return -ENOMEM;
189} 205}
190 206
191void jbd2_journal_destroy_revoke_caches(void) 207static struct jbd2_revoke_table_s *jbd2_journal_init_revoke_table(int hash_size)
192{ 208{
193 kmem_cache_destroy(jbd2_revoke_record_cache); 209 int shift = 0;
194 jbd2_revoke_record_cache = NULL; 210 int tmp = hash_size;
195 kmem_cache_destroy(jbd2_revoke_table_cache); 211 struct jbd2_revoke_table_s *table;
196 jbd2_revoke_table_cache = NULL;
197}
198
199/* Initialise the revoke table for a given journal to a given size. */
200
201int jbd2_journal_init_revoke(journal_t *journal, int hash_size)
202{
203 int shift, tmp;
204 212
205 J_ASSERT (journal->j_revoke_table[0] == NULL); 213 table = kmem_cache_alloc(jbd2_revoke_table_cache, GFP_KERNEL);
214 if (!table)
215 goto out;
206 216
207 shift = 0;
208 tmp = hash_size;
209 while((tmp >>= 1UL) != 0UL) 217 while((tmp >>= 1UL) != 0UL)
210 shift++; 218 shift++;
211 219
212 journal->j_revoke_table[0] = kmem_cache_alloc(jbd2_revoke_table_cache, GFP_KERNEL); 220 table->hash_size = hash_size;
213 if (!journal->j_revoke_table[0]) 221 table->hash_shift = shift;
214 return -ENOMEM; 222 table->hash_table =
215 journal->j_revoke = journal->j_revoke_table[0];
216
217 /* Check that the hash_size is a power of two */
218 J_ASSERT(is_power_of_2(hash_size));
219
220 journal->j_revoke->hash_size = hash_size;
221
222 journal->j_revoke->hash_shift = shift;
223
224 journal->j_revoke->hash_table =
225 kmalloc(hash_size * sizeof(struct list_head), GFP_KERNEL); 223 kmalloc(hash_size * sizeof(struct list_head), GFP_KERNEL);
226 if (!journal->j_revoke->hash_table) { 224 if (!table->hash_table) {
227 kmem_cache_free(jbd2_revoke_table_cache, journal->j_revoke_table[0]); 225 kmem_cache_free(jbd2_revoke_table_cache, table);
228 journal->j_revoke = NULL; 226 table = NULL;
229 return -ENOMEM; 227 goto out;
230 } 228 }
231 229
232 for (tmp = 0; tmp < hash_size; tmp++) 230 for (tmp = 0; tmp < hash_size; tmp++)
233 INIT_LIST_HEAD(&journal->j_revoke->hash_table[tmp]); 231 INIT_LIST_HEAD(&table->hash_table[tmp]);
234 232
235 journal->j_revoke_table[1] = kmem_cache_alloc(jbd2_revoke_table_cache, GFP_KERNEL); 233out:
236 if (!journal->j_revoke_table[1]) { 234 return table;
237 kfree(journal->j_revoke_table[0]->hash_table); 235}
238 kmem_cache_free(jbd2_revoke_table_cache, journal->j_revoke_table[0]); 236
239 return -ENOMEM; 237static void jbd2_journal_destroy_revoke_table(struct jbd2_revoke_table_s *table)
238{
239 int i;
240 struct list_head *hash_list;
241
242 for (i = 0; i < table->hash_size; i++) {
243 hash_list = &table->hash_table[i];
244 J_ASSERT(list_empty(hash_list));
240 } 245 }
241 246
242 journal->j_revoke = journal->j_revoke_table[1]; 247 kfree(table->hash_table);
248 kmem_cache_free(jbd2_revoke_table_cache, table);
249}
243 250
244 /* Check that the hash_size is a power of two */ 251/* Initialise the revoke table for a given journal to a given size. */
252int jbd2_journal_init_revoke(journal_t *journal, int hash_size)
253{
254 J_ASSERT(journal->j_revoke_table[0] == NULL);
245 J_ASSERT(is_power_of_2(hash_size)); 255 J_ASSERT(is_power_of_2(hash_size));
246 256
247 journal->j_revoke->hash_size = hash_size; 257 journal->j_revoke_table[0] = jbd2_journal_init_revoke_table(hash_size);
248 258 if (!journal->j_revoke_table[0])
249 journal->j_revoke->hash_shift = shift; 259 goto fail0;
250 260
251 journal->j_revoke->hash_table = 261 journal->j_revoke_table[1] = jbd2_journal_init_revoke_table(hash_size);
252 kmalloc(hash_size * sizeof(struct list_head), GFP_KERNEL); 262 if (!journal->j_revoke_table[1])
253 if (!journal->j_revoke->hash_table) { 263 goto fail1;
254 kfree(journal->j_revoke_table[0]->hash_table);
255 kmem_cache_free(jbd2_revoke_table_cache, journal->j_revoke_table[0]);
256 kmem_cache_free(jbd2_revoke_table_cache, journal->j_revoke_table[1]);
257 journal->j_revoke = NULL;
258 return -ENOMEM;
259 }
260 264
261 for (tmp = 0; tmp < hash_size; tmp++) 265 journal->j_revoke = journal->j_revoke_table[1];
262 INIT_LIST_HEAD(&journal->j_revoke->hash_table[tmp]);
263 266
264 spin_lock_init(&journal->j_revoke_lock); 267 spin_lock_init(&journal->j_revoke_lock);
265 268
266 return 0; 269 return 0;
267}
268 270
269/* Destoy a journal's revoke table. The table must already be empty! */ 271fail1:
272 jbd2_journal_destroy_revoke_table(journal->j_revoke_table[0]);
273fail0:
274 return -ENOMEM;
275}
270 276
277/* Destroy a journal's revoke table. The table must already be empty! */
271void jbd2_journal_destroy_revoke(journal_t *journal) 278void jbd2_journal_destroy_revoke(journal_t *journal)
272{ 279{
273 struct jbd2_revoke_table_s *table;
274 struct list_head *hash_list;
275 int i;
276
277 table = journal->j_revoke_table[0];
278 if (!table)
279 return;
280
281 for (i=0; i<table->hash_size; i++) {
282 hash_list = &table->hash_table[i];
283 J_ASSERT (list_empty(hash_list));
284 }
285
286 kfree(table->hash_table);
287 kmem_cache_free(jbd2_revoke_table_cache, table);
288 journal->j_revoke = NULL;
289
290 table = journal->j_revoke_table[1];
291 if (!table)
292 return;
293
294 for (i=0; i<table->hash_size; i++) {
295 hash_list = &table->hash_table[i];
296 J_ASSERT (list_empty(hash_list));
297 }
298
299 kfree(table->hash_table);
300 kmem_cache_free(jbd2_revoke_table_cache, table);
301 journal->j_revoke = NULL; 280 journal->j_revoke = NULL;
281 if (journal->j_revoke_table[0])
282 jbd2_journal_destroy_revoke_table(journal->j_revoke_table[0]);
283 if (journal->j_revoke_table[1])
284 jbd2_journal_destroy_revoke_table(journal->j_revoke_table[1]);
302} 285}
303 286
304 287
diff --git a/fs/jbd2/transaction.c b/fs/jbd2/transaction.c
index b9b0b6f899b9..d6e006e67804 100644
--- a/fs/jbd2/transaction.c
+++ b/fs/jbd2/transaction.c
@@ -618,6 +618,12 @@ repeat:
618 goto done; 618 goto done;
619 619
620 /* 620 /*
621 * this is the first time this transaction is touching this buffer,
622 * reset the modified flag
623 */
624 jh->b_modified = 0;
625
626 /*
621 * If there is already a copy-out version of this buffer, then we don't 627 * If there is already a copy-out version of this buffer, then we don't
622 * need to make another one 628 * need to make another one
623 */ 629 */
@@ -690,7 +696,7 @@ repeat:
690 if (!frozen_buffer) { 696 if (!frozen_buffer) {
691 printk(KERN_EMERG 697 printk(KERN_EMERG
692 "%s: OOM for frozen_buffer\n", 698 "%s: OOM for frozen_buffer\n",
693 __FUNCTION__); 699 __func__);
694 JBUFFER_TRACE(jh, "oom!"); 700 JBUFFER_TRACE(jh, "oom!");
695 error = -ENOMEM; 701 error = -ENOMEM;
696 jbd_lock_bh_state(bh); 702 jbd_lock_bh_state(bh);
@@ -829,9 +835,16 @@ int jbd2_journal_get_create_access(handle_t *handle, struct buffer_head *bh)
829 835
830 if (jh->b_transaction == NULL) { 836 if (jh->b_transaction == NULL) {
831 jh->b_transaction = transaction; 837 jh->b_transaction = transaction;
838
839 /* first access by this transaction */
840 jh->b_modified = 0;
841
832 JBUFFER_TRACE(jh, "file as BJ_Reserved"); 842 JBUFFER_TRACE(jh, "file as BJ_Reserved");
833 __jbd2_journal_file_buffer(jh, transaction, BJ_Reserved); 843 __jbd2_journal_file_buffer(jh, transaction, BJ_Reserved);
834 } else if (jh->b_transaction == journal->j_committing_transaction) { 844 } else if (jh->b_transaction == journal->j_committing_transaction) {
845 /* first access by this transaction */
846 jh->b_modified = 0;
847
835 JBUFFER_TRACE(jh, "set next transaction"); 848 JBUFFER_TRACE(jh, "set next transaction");
836 jh->b_next_transaction = transaction; 849 jh->b_next_transaction = transaction;
837 } 850 }
@@ -901,7 +914,7 @@ repeat:
901 committed_data = jbd2_alloc(jh2bh(jh)->b_size, GFP_NOFS); 914 committed_data = jbd2_alloc(jh2bh(jh)->b_size, GFP_NOFS);
902 if (!committed_data) { 915 if (!committed_data) {
903 printk(KERN_EMERG "%s: No memory for committed data\n", 916 printk(KERN_EMERG "%s: No memory for committed data\n",
904 __FUNCTION__); 917 __func__);
905 err = -ENOMEM; 918 err = -ENOMEM;
906 goto out; 919 goto out;
907 } 920 }
@@ -1230,6 +1243,7 @@ int jbd2_journal_forget (handle_t *handle, struct buffer_head *bh)
1230 struct journal_head *jh; 1243 struct journal_head *jh;
1231 int drop_reserve = 0; 1244 int drop_reserve = 0;
1232 int err = 0; 1245 int err = 0;
1246 int was_modified = 0;
1233 1247
1234 BUFFER_TRACE(bh, "entry"); 1248 BUFFER_TRACE(bh, "entry");
1235 1249
@@ -1248,6 +1262,9 @@ int jbd2_journal_forget (handle_t *handle, struct buffer_head *bh)
1248 goto not_jbd; 1262 goto not_jbd;
1249 } 1263 }
1250 1264
1265 /* keep track of wether or not this transaction modified us */
1266 was_modified = jh->b_modified;
1267
1251 /* 1268 /*
1252 * The buffer's going from the transaction, we must drop 1269 * The buffer's going from the transaction, we must drop
1253 * all references -bzzz 1270 * all references -bzzz
@@ -1265,7 +1282,12 @@ int jbd2_journal_forget (handle_t *handle, struct buffer_head *bh)
1265 1282
1266 JBUFFER_TRACE(jh, "belongs to current transaction: unfile"); 1283 JBUFFER_TRACE(jh, "belongs to current transaction: unfile");
1267 1284
1268 drop_reserve = 1; 1285 /*
1286 * we only want to drop a reference if this transaction
1287 * modified the buffer
1288 */
1289 if (was_modified)
1290 drop_reserve = 1;
1269 1291
1270 /* 1292 /*
1271 * We are no longer going to journal this buffer. 1293 * We are no longer going to journal this buffer.
@@ -1305,7 +1327,13 @@ int jbd2_journal_forget (handle_t *handle, struct buffer_head *bh)
1305 if (jh->b_next_transaction) { 1327 if (jh->b_next_transaction) {
1306 J_ASSERT(jh->b_next_transaction == transaction); 1328 J_ASSERT(jh->b_next_transaction == transaction);
1307 jh->b_next_transaction = NULL; 1329 jh->b_next_transaction = NULL;
1308 drop_reserve = 1; 1330
1331 /*
1332 * only drop a reference if this transaction modified
1333 * the buffer
1334 */
1335 if (was_modified)
1336 drop_reserve = 1;
1309 } 1337 }
1310 } 1338 }
1311 1339
@@ -1434,7 +1462,8 @@ int jbd2_journal_stop(handle_t *handle)
1434 return err; 1462 return err;
1435} 1463}
1436 1464
1437/**int jbd2_journal_force_commit() - force any uncommitted transactions 1465/**
1466 * int jbd2_journal_force_commit() - force any uncommitted transactions
1438 * @journal: journal to force 1467 * @journal: journal to force
1439 * 1468 *
1440 * For synchronous operations: force any uncommitted transactions 1469 * For synchronous operations: force any uncommitted transactions
@@ -2077,7 +2106,7 @@ void __jbd2_journal_refile_buffer(struct journal_head *jh)
2077 jh->b_transaction = jh->b_next_transaction; 2106 jh->b_transaction = jh->b_next_transaction;
2078 jh->b_next_transaction = NULL; 2107 jh->b_next_transaction = NULL;
2079 __jbd2_journal_file_buffer(jh, jh->b_transaction, 2108 __jbd2_journal_file_buffer(jh, jh->b_transaction,
2080 was_dirty ? BJ_Metadata : BJ_Reserved); 2109 jh->b_modified ? BJ_Metadata : BJ_Reserved);
2081 J_ASSERT_JH(jh, jh->b_transaction->t_state == T_RUNNING); 2110 J_ASSERT_JH(jh, jh->b_transaction->t_state == T_RUNNING);
2082 2111
2083 if (was_dirty) 2112 if (was_dirty)
generated by cgit v1.2.2 (git 2.25.0) at 2024-09-25 12:23:57 -0400