aboutsummaryrefslogtreecommitdiffstats
path: root/fs/ext4
diff options
context:
space:
mode:
Diffstat (limited to 'fs/ext4')
-rw-r--r--fs/ext4/Makefile2
-rw-r--r--fs/ext4/acl.c1
-rw-r--r--fs/ext4/balloc.c11
-rw-r--r--fs/ext4/block_validity.c15
-rw-r--r--fs/ext4/dir.c25
-rw-r--r--fs/ext4/ext4.h265
-rw-r--r--fs/ext4/ext4_extents.h65
-rw-r--r--fs/ext4/ext4_jbd2.c71
-rw-r--r--fs/ext4/ext4_jbd2.h56
-rw-r--r--fs/ext4/extents.c386
-rw-r--r--fs/ext4/file.c49
-rw-r--r--fs/ext4/fsync.c88
-rw-r--r--fs/ext4/ialloc.c141
-rw-r--r--fs/ext4/inode.c838
-rw-r--r--fs/ext4/mballoc.c698
-rw-r--r--fs/ext4/migrate.c4
-rw-r--r--fs/ext4/move_extent.c32
-rw-r--r--fs/ext4/namei.c103
-rw-r--r--fs/ext4/page-io.c430
-rw-r--r--fs/ext4/resize.c60
-rw-r--r--fs/ext4/super.c896
-rw-r--r--fs/ext4/xattr.c19
-rw-r--r--fs/ext4/xattr.h10
23 files changed, 2836 insertions, 1429 deletions
diff --git a/fs/ext4/Makefile b/fs/ext4/Makefile
index 8867b2a1e5fe..c947e36eda6c 100644
--- a/fs/ext4/Makefile
+++ b/fs/ext4/Makefile
@@ -4,7 +4,7 @@
4 4
5obj-$(CONFIG_EXT4_FS) += ext4.o 5obj-$(CONFIG_EXT4_FS) += ext4.o
6 6
7ext4-y := balloc.o bitmap.o dir.o file.o fsync.o ialloc.o inode.o \ 7ext4-y := balloc.o bitmap.o dir.o file.o fsync.o ialloc.o inode.o page-io.o \
8 ioctl.o namei.o super.o symlink.o hash.o resize.o extents.o \ 8 ioctl.o namei.o super.o symlink.o hash.o resize.o extents.o \
9 ext4_jbd2.o migrate.o mballoc.o block_validity.o move_extent.o 9 ext4_jbd2.o migrate.o mballoc.o block_validity.o move_extent.o
10 10
diff --git a/fs/ext4/acl.c b/fs/ext4/acl.c
index feaf498feaa6..5e2ed4504ead 100644
--- a/fs/ext4/acl.c
+++ b/fs/ext4/acl.c
@@ -204,6 +204,7 @@ ext4_set_acl(handle_t *handle, struct inode *inode, int type,
204 return error; 204 return error;
205 else { 205 else {
206 inode->i_mode = mode; 206 inode->i_mode = mode;
207 inode->i_ctime = ext4_current_time(inode);
207 ext4_mark_inode_dirty(handle, inode); 208 ext4_mark_inode_dirty(handle, inode);
208 if (error == 0) 209 if (error == 0)
209 acl = NULL; 210 acl = NULL;
diff --git a/fs/ext4/balloc.c b/fs/ext4/balloc.c
index 95b7594c76f9..14c3af26c671 100644
--- a/fs/ext4/balloc.c
+++ b/fs/ext4/balloc.c
@@ -171,7 +171,8 @@ unsigned ext4_init_block_bitmap(struct super_block *sb, struct buffer_head *bh,
171 * less than the blocksize * 8 ( which is the size 171 * less than the blocksize * 8 ( which is the size
172 * of bitmap ), set rest of the block bitmap to 1 172 * of bitmap ), set rest of the block bitmap to 1
173 */ 173 */
174 mark_bitmap_end(group_blocks, sb->s_blocksize * 8, bh->b_data); 174 ext4_mark_bitmap_end(group_blocks, sb->s_blocksize * 8,
175 bh->b_data);
175 } 176 }
176 return free_blocks - ext4_group_used_meta_blocks(sb, block_group, gdp); 177 return free_blocks - ext4_group_used_meta_blocks(sb, block_group, gdp);
177} 178}
@@ -377,14 +378,11 @@ void ext4_add_groupblocks(handle_t *handle, struct super_block *sb,
377 ext4_grpblk_t bit; 378 ext4_grpblk_t bit;
378 unsigned int i; 379 unsigned int i;
379 struct ext4_group_desc *desc; 380 struct ext4_group_desc *desc;
380 struct ext4_super_block *es; 381 struct ext4_sb_info *sbi = EXT4_SB(sb);
381 struct ext4_sb_info *sbi;
382 int err = 0, ret, blk_free_count; 382 int err = 0, ret, blk_free_count;
383 ext4_grpblk_t blocks_freed; 383 ext4_grpblk_t blocks_freed;
384 struct ext4_group_info *grp; 384 struct ext4_group_info *grp;
385 385
386 sbi = EXT4_SB(sb);
387 es = sbi->s_es;
388 ext4_debug("Adding block(s) %llu-%llu\n", block, block + count - 1); 386 ext4_debug("Adding block(s) %llu-%llu\n", block, block + count - 1);
389 387
390 ext4_get_group_no_and_offset(sb, block, &block_group, &bit); 388 ext4_get_group_no_and_offset(sb, block, &block_group, &bit);
@@ -477,7 +475,6 @@ void ext4_add_groupblocks(handle_t *handle, struct super_block *sb,
477 ret = ext4_handle_dirty_metadata(handle, NULL, gd_bh); 475 ret = ext4_handle_dirty_metadata(handle, NULL, gd_bh);
478 if (!err) 476 if (!err)
479 err = ret; 477 err = ret;
480 sb->s_dirt = 1;
481 478
482error_return: 479error_return:
483 brelse(bitmap_bh); 480 brelse(bitmap_bh);
@@ -493,7 +490,7 @@ error_return:
493 * Check if filesystem has nblocks free & available for allocation. 490 * Check if filesystem has nblocks free & available for allocation.
494 * On success return 1, return 0 on failure. 491 * On success return 1, return 0 on failure.
495 */ 492 */
496int ext4_has_free_blocks(struct ext4_sb_info *sbi, s64 nblocks) 493static int ext4_has_free_blocks(struct ext4_sb_info *sbi, s64 nblocks)
497{ 494{
498 s64 free_blocks, dirty_blocks, root_blocks; 495 s64 free_blocks, dirty_blocks, root_blocks;
499 struct percpu_counter *fbc = &sbi->s_freeblocks_counter; 496 struct percpu_counter *fbc = &sbi->s_freeblocks_counter;
diff --git a/fs/ext4/block_validity.c b/fs/ext4/block_validity.c
index 5b6973fbf1bd..fac90f3fba80 100644
--- a/fs/ext4/block_validity.c
+++ b/fs/ext4/block_validity.c
@@ -29,16 +29,15 @@ struct ext4_system_zone {
29 29
30static struct kmem_cache *ext4_system_zone_cachep; 30static struct kmem_cache *ext4_system_zone_cachep;
31 31
32int __init init_ext4_system_zone(void) 32int __init ext4_init_system_zone(void)
33{ 33{
34 ext4_system_zone_cachep = KMEM_CACHE(ext4_system_zone, 34 ext4_system_zone_cachep = KMEM_CACHE(ext4_system_zone, 0);
35 SLAB_RECLAIM_ACCOUNT);
36 if (ext4_system_zone_cachep == NULL) 35 if (ext4_system_zone_cachep == NULL)
37 return -ENOMEM; 36 return -ENOMEM;
38 return 0; 37 return 0;
39} 38}
40 39
41void exit_ext4_system_zone(void) 40void ext4_exit_system_zone(void)
42{ 41{
43 kmem_cache_destroy(ext4_system_zone_cachep); 42 kmem_cache_destroy(ext4_system_zone_cachep);
44} 43}
@@ -229,16 +228,20 @@ int ext4_data_block_valid(struct ext4_sb_info *sbi, ext4_fsblk_t start_blk,
229 228
230 if ((start_blk <= le32_to_cpu(sbi->s_es->s_first_data_block)) || 229 if ((start_blk <= le32_to_cpu(sbi->s_es->s_first_data_block)) ||
231 (start_blk + count < start_blk) || 230 (start_blk + count < start_blk) ||
232 (start_blk + count > ext4_blocks_count(sbi->s_es))) 231 (start_blk + count > ext4_blocks_count(sbi->s_es))) {
232 sbi->s_es->s_last_error_block = cpu_to_le64(start_blk);
233 return 0; 233 return 0;
234 }
234 while (n) { 235 while (n) {
235 entry = rb_entry(n, struct ext4_system_zone, node); 236 entry = rb_entry(n, struct ext4_system_zone, node);
236 if (start_blk + count - 1 < entry->start_blk) 237 if (start_blk + count - 1 < entry->start_blk)
237 n = n->rb_left; 238 n = n->rb_left;
238 else if (start_blk >= (entry->start_blk + entry->count)) 239 else if (start_blk >= (entry->start_blk + entry->count))
239 n = n->rb_right; 240 n = n->rb_right;
240 else 241 else {
242 sbi->s_es->s_last_error_block = cpu_to_le64(start_blk);
241 return 0; 243 return 0;
244 }
242 } 245 }
243 return 1; 246 return 1;
244} 247}
diff --git a/fs/ext4/dir.c b/fs/ext4/dir.c
index ea5e6cb7e2a5..ece76fb6a40c 100644
--- a/fs/ext4/dir.c
+++ b/fs/ext4/dir.c
@@ -39,7 +39,7 @@ static int ext4_release_dir(struct inode *inode,
39 struct file *filp); 39 struct file *filp);
40 40
41const struct file_operations ext4_dir_operations = { 41const struct file_operations ext4_dir_operations = {
42 .llseek = generic_file_llseek, 42 .llseek = ext4_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 .unlocked_ioctl = ext4_ioctl, 45 .unlocked_ioctl = ext4_ioctl,
@@ -61,10 +61,11 @@ static unsigned char get_dtype(struct super_block *sb, int filetype)
61} 61}
62 62
63 63
64int ext4_check_dir_entry(const char *function, struct inode *dir, 64int __ext4_check_dir_entry(const char *function, unsigned int line,
65 struct ext4_dir_entry_2 *de, 65 struct inode *dir,
66 struct buffer_head *bh, 66 struct ext4_dir_entry_2 *de,
67 unsigned int offset) 67 struct buffer_head *bh,
68 unsigned int offset)
68{ 69{
69 const char *error_msg = NULL; 70 const char *error_msg = NULL;
70 const int rlen = ext4_rec_len_from_disk(de->rec_len, 71 const int rlen = ext4_rec_len_from_disk(de->rec_len,
@@ -83,11 +84,10 @@ int ext4_check_dir_entry(const char *function, struct inode *dir,
83 error_msg = "inode out of bounds"; 84 error_msg = "inode out of bounds";
84 85
85 if (error_msg != NULL) 86 if (error_msg != NULL)
86 ext4_error_inode(function, dir, 87 ext4_error_inode(dir, function, line, bh->b_blocknr,
87 "bad entry in directory: %s - block=%llu" 88 "bad entry in directory: %s - "
88 "offset=%u(%u), inode=%u, rec_len=%d, name_len=%d", 89 "offset=%u(%u), inode=%u, rec_len=%d, name_len=%d",
89 error_msg, (unsigned long long) bh->b_blocknr, 90 error_msg, (unsigned) (offset%bh->b_size), offset,
90 (unsigned) (offset%bh->b_size), offset,
91 le32_to_cpu(de->inode), 91 le32_to_cpu(de->inode),
92 rlen, de->name_len); 92 rlen, de->name_len);
93 return error_msg == NULL ? 1 : 0; 93 return error_msg == NULL ? 1 : 0;
@@ -121,7 +121,8 @@ static int ext4_readdir(struct file *filp,
121 * We don't set the inode dirty flag since it's not 121 * We don't set the inode dirty flag since it's not
122 * critical that it get flushed back to the disk. 122 * critical that it get flushed back to the disk.
123 */ 123 */
124 ext4_clear_inode_flag(filp->f_path.dentry->d_inode, EXT4_INODE_INDEX); 124 ext4_clear_inode_flag(filp->f_path.dentry->d_inode,
125 EXT4_INODE_INDEX);
125 } 126 }
126 stored = 0; 127 stored = 0;
127 offset = filp->f_pos & (sb->s_blocksize - 1); 128 offset = filp->f_pos & (sb->s_blocksize - 1);
@@ -193,7 +194,7 @@ revalidate:
193 while (!error && filp->f_pos < inode->i_size 194 while (!error && filp->f_pos < inode->i_size
194 && offset < sb->s_blocksize) { 195 && offset < sb->s_blocksize) {
195 de = (struct ext4_dir_entry_2 *) (bh->b_data + offset); 196 de = (struct ext4_dir_entry_2 *) (bh->b_data + offset);
196 if (!ext4_check_dir_entry("ext4_readdir", inode, de, 197 if (!ext4_check_dir_entry(inode, de,
197 bh, offset)) { 198 bh, offset)) {
198 /* 199 /*
199 * On error, skip the f_pos to the next block 200 * On error, skip the f_pos to the next block
@@ -343,7 +344,7 @@ int ext4_htree_store_dirent(struct file *dir_file, __u32 hash,
343 struct dir_private_info *info; 344 struct dir_private_info *info;
344 int len; 345 int len;
345 346
346 info = (struct dir_private_info *) dir_file->private_data; 347 info = dir_file->private_data;
347 p = &info->root.rb_node; 348 p = &info->root.rb_node;
348 349
349 /* Create and allocate the fname structure */ 350 /* Create and allocate the fname structure */
diff --git a/fs/ext4/ext4.h b/fs/ext4/ext4.h
index 19a4de57128a..8b5dd6369f82 100644
--- a/fs/ext4/ext4.h
+++ b/fs/ext4/ext4.h
@@ -57,10 +57,13 @@
57#endif 57#endif
58 58
59#define EXT4_ERROR_INODE(inode, fmt, a...) \ 59#define EXT4_ERROR_INODE(inode, fmt, a...) \
60 ext4_error_inode(__func__, (inode), (fmt), ## a) 60 ext4_error_inode((inode), __func__, __LINE__, 0, (fmt), ## a)
61
62#define EXT4_ERROR_INODE_BLOCK(inode, block, fmt, a...) \
63 ext4_error_inode((inode), __func__, __LINE__, (block), (fmt), ## a)
61 64
62#define EXT4_ERROR_FILE(file, fmt, a...) \ 65#define EXT4_ERROR_FILE(file, fmt, a...) \
63 ext4_error_file(__func__, (file), (fmt), ## a) 66 ext4_error_file(__func__, __LINE__, (file), (fmt), ## a)
64 67
65/* data type for block offset of block group */ 68/* data type for block offset of block group */
66typedef int ext4_grpblk_t; 69typedef int ext4_grpblk_t;
@@ -165,17 +168,42 @@ struct mpage_da_data {
165 int pages_written; 168 int pages_written;
166 int retval; 169 int retval;
167}; 170};
168#define EXT4_IO_UNWRITTEN 0x1 171
172/*
173 * Flags for ext4_io_end->flags
174 */
175#define EXT4_IO_END_UNWRITTEN 0x0001
176#define EXT4_IO_END_ERROR 0x0002
177
178struct ext4_io_page {
179 struct page *p_page;
180 int p_count;
181};
182
183#define MAX_IO_PAGES 128
184
169typedef struct ext4_io_end { 185typedef struct ext4_io_end {
170 struct list_head list; /* per-file finished AIO list */ 186 struct list_head list; /* per-file finished IO list */
171 struct inode *inode; /* file being written to */ 187 struct inode *inode; /* file being written to */
172 unsigned int flag; /* unwritten or not */ 188 unsigned int flag; /* unwritten or not */
173 struct page *page; /* page struct for buffer write */ 189 struct page *page; /* page struct for buffer write */
174 loff_t offset; /* offset in the file */ 190 loff_t offset; /* offset in the file */
175 ssize_t size; /* size of the extent */ 191 ssize_t size; /* size of the extent */
176 struct work_struct work; /* data work queue */ 192 struct work_struct work; /* data work queue */
193 struct kiocb *iocb; /* iocb struct for AIO */
194 int result; /* error value for AIO */
195 int num_io_pages;
196 struct ext4_io_page *pages[MAX_IO_PAGES];
177} ext4_io_end_t; 197} ext4_io_end_t;
178 198
199struct ext4_io_submit {
200 int io_op;
201 struct bio *io_bio;
202 ext4_io_end_t *io_end;
203 struct ext4_io_page *io_page;
204 sector_t io_next_block;
205};
206
179/* 207/*
180 * Special inodes numbers 208 * Special inodes numbers
181 */ 209 */
@@ -200,6 +228,7 @@ typedef struct ext4_io_end {
200#define EXT4_MIN_BLOCK_SIZE 1024 228#define EXT4_MIN_BLOCK_SIZE 1024
201#define EXT4_MAX_BLOCK_SIZE 65536 229#define EXT4_MAX_BLOCK_SIZE 65536
202#define EXT4_MIN_BLOCK_LOG_SIZE 10 230#define EXT4_MIN_BLOCK_LOG_SIZE 10
231#define EXT4_MAX_BLOCK_LOG_SIZE 16
203#ifdef __KERNEL__ 232#ifdef __KERNEL__
204# define EXT4_BLOCK_SIZE(s) ((s)->s_blocksize) 233# define EXT4_BLOCK_SIZE(s) ((s)->s_blocksize)
205#else 234#else
@@ -460,7 +489,7 @@ struct ext4_new_group_data {
460}; 489};
461 490
462/* 491/*
463 * Flags used by ext4_get_blocks() 492 * Flags used by ext4_map_blocks()
464 */ 493 */
465 /* Allocate any needed blocks and/or convert an unitialized 494 /* Allocate any needed blocks and/or convert an unitialized
466 extent to be an initialized ext4 */ 495 extent to be an initialized ext4 */
@@ -873,7 +902,6 @@ struct ext4_inode_info {
873#define EXT4_MOUNT_POSIX_ACL 0x08000 /* POSIX Access Control Lists */ 902#define EXT4_MOUNT_POSIX_ACL 0x08000 /* POSIX Access Control Lists */
874#define EXT4_MOUNT_NO_AUTO_DA_ALLOC 0x10000 /* No auto delalloc mapping */ 903#define EXT4_MOUNT_NO_AUTO_DA_ALLOC 0x10000 /* No auto delalloc mapping */
875#define EXT4_MOUNT_BARRIER 0x20000 /* Use block barriers */ 904#define EXT4_MOUNT_BARRIER 0x20000 /* Use block barriers */
876#define EXT4_MOUNT_NOBH 0x40000 /* No bufferheads */
877#define EXT4_MOUNT_QUOTA 0x80000 /* Some quota option set */ 905#define EXT4_MOUNT_QUOTA 0x80000 /* Some quota option set */
878#define EXT4_MOUNT_USRQUOTA 0x100000 /* "old" user quota */ 906#define EXT4_MOUNT_USRQUOTA 0x100000 /* "old" user quota */
879#define EXT4_MOUNT_GRPQUOTA 0x200000 /* "old" group quota */ 907#define EXT4_MOUNT_GRPQUOTA 0x200000 /* "old" group quota */
@@ -885,6 +913,7 @@ struct ext4_inode_info {
885#define EXT4_MOUNT_DATA_ERR_ABORT 0x10000000 /* Abort on file data write */ 913#define EXT4_MOUNT_DATA_ERR_ABORT 0x10000000 /* Abort on file data write */
886#define EXT4_MOUNT_BLOCK_VALIDITY 0x20000000 /* Block validity checking */ 914#define EXT4_MOUNT_BLOCK_VALIDITY 0x20000000 /* Block validity checking */
887#define EXT4_MOUNT_DISCARD 0x40000000 /* Issue DISCARD requests */ 915#define EXT4_MOUNT_DISCARD 0x40000000 /* Issue DISCARD requests */
916#define EXT4_MOUNT_INIT_INODE_TABLE 0x80000000 /* Initialize uninitialized itables */
888 917
889#define clear_opt(o, opt) o &= ~EXT4_MOUNT_##opt 918#define clear_opt(o, opt) o &= ~EXT4_MOUNT_##opt
890#define set_opt(o, opt) o |= EXT4_MOUNT_##opt 919#define set_opt(o, opt) o |= EXT4_MOUNT_##opt
@@ -982,7 +1011,7 @@ struct ext4_super_block {
982 __le32 s_last_orphan; /* start of list of inodes to delete */ 1011 __le32 s_last_orphan; /* start of list of inodes to delete */
983 __le32 s_hash_seed[4]; /* HTREE hash seed */ 1012 __le32 s_hash_seed[4]; /* HTREE hash seed */
984 __u8 s_def_hash_version; /* Default hash version to use */ 1013 __u8 s_def_hash_version; /* Default hash version to use */
985 __u8 s_reserved_char_pad; 1014 __u8 s_jnl_backup_type;
986 __le16 s_desc_size; /* size of group descriptor */ 1015 __le16 s_desc_size; /* size of group descriptor */
987/*100*/ __le32 s_default_mount_opts; 1016/*100*/ __le32 s_default_mount_opts;
988 __le32 s_first_meta_bg; /* First metablock block group */ 1017 __le32 s_first_meta_bg; /* First metablock block group */
@@ -1000,12 +1029,34 @@ struct ext4_super_block {
1000 __le64 s_mmp_block; /* Block for multi-mount protection */ 1029 __le64 s_mmp_block; /* Block for multi-mount protection */
1001 __le32 s_raid_stripe_width; /* blocks on all data disks (N*stride)*/ 1030 __le32 s_raid_stripe_width; /* blocks on all data disks (N*stride)*/
1002 __u8 s_log_groups_per_flex; /* FLEX_BG group size */ 1031 __u8 s_log_groups_per_flex; /* FLEX_BG group size */
1003 __u8 s_reserved_char_pad2; 1032 __u8 s_reserved_char_pad;
1004 __le16 s_reserved_pad; 1033 __le16 s_reserved_pad;
1005 __le64 s_kbytes_written; /* nr of lifetime kilobytes written */ 1034 __le64 s_kbytes_written; /* nr of lifetime kilobytes written */
1006 __u32 s_reserved[160]; /* Padding to the end of the block */ 1035 __le32 s_snapshot_inum; /* Inode number of active snapshot */
1036 __le32 s_snapshot_id; /* sequential ID of active snapshot */
1037 __le64 s_snapshot_r_blocks_count; /* reserved blocks for active
1038 snapshot's future use */
1039 __le32 s_snapshot_list; /* inode number of the head of the
1040 on-disk snapshot list */
1041#define EXT4_S_ERR_START offsetof(struct ext4_super_block, s_error_count)
1042 __le32 s_error_count; /* number of fs errors */
1043 __le32 s_first_error_time; /* first time an error happened */
1044 __le32 s_first_error_ino; /* inode involved in first error */
1045 __le64 s_first_error_block; /* block involved of first error */
1046 __u8 s_first_error_func[32]; /* function where the error happened */
1047 __le32 s_first_error_line; /* line number where error happened */
1048 __le32 s_last_error_time; /* most recent time of an error */
1049 __le32 s_last_error_ino; /* inode involved in last error */
1050 __le32 s_last_error_line; /* line number where error happened */
1051 __le64 s_last_error_block; /* block involved of last error */
1052 __u8 s_last_error_func[32]; /* function where the error happened */
1053#define EXT4_S_ERR_END offsetof(struct ext4_super_block, s_mount_opts)
1054 __u8 s_mount_opts[64];
1055 __le32 s_reserved[112]; /* Padding to the end of the block */
1007}; 1056};
1008 1057
1058#define EXT4_S_ERR_LEN (EXT4_S_ERR_END - EXT4_S_ERR_START)
1059
1009#ifdef __KERNEL__ 1060#ifdef __KERNEL__
1010 1061
1011/* 1062/*
@@ -1061,7 +1112,6 @@ struct ext4_sb_info {
1061 struct completion s_kobj_unregister; 1112 struct completion s_kobj_unregister;
1062 1113
1063 /* Journaling */ 1114 /* Journaling */
1064 struct inode *s_journal_inode;
1065 struct journal_s *s_journal; 1115 struct journal_s *s_journal;
1066 struct list_head s_orphan; 1116 struct list_head s_orphan;
1067 struct mutex s_orphan_lock; 1117 struct mutex s_orphan_lock;
@@ -1094,10 +1144,7 @@ struct ext4_sb_info {
1094 /* for buddy allocator */ 1144 /* for buddy allocator */
1095 struct ext4_group_info ***s_group_info; 1145 struct ext4_group_info ***s_group_info;
1096 struct inode *s_buddy_cache; 1146 struct inode *s_buddy_cache;
1097 long s_blocks_reserved;
1098 spinlock_t s_reserve_lock;
1099 spinlock_t s_md_lock; 1147 spinlock_t s_md_lock;
1100 tid_t s_last_transaction;
1101 unsigned short *s_mb_offsets; 1148 unsigned short *s_mb_offsets;
1102 unsigned int *s_mb_maxs; 1149 unsigned int *s_mb_maxs;
1103 1150
@@ -1115,7 +1162,6 @@ struct ext4_sb_info {
1115 unsigned long s_mb_last_start; 1162 unsigned long s_mb_last_start;
1116 1163
1117 /* stats for buddy allocator */ 1164 /* stats for buddy allocator */
1118 spinlock_t s_mb_pa_lock;
1119 atomic_t s_bal_reqs; /* number of reqs with len > 1 */ 1165 atomic_t s_bal_reqs; /* number of reqs with len > 1 */
1120 atomic_t s_bal_success; /* we found long enough chunks */ 1166 atomic_t s_bal_success; /* we found long enough chunks */
1121 atomic_t s_bal_allocated; /* in blocks */ 1167 atomic_t s_bal_allocated; /* in blocks */
@@ -1143,6 +1189,14 @@ struct ext4_sb_info {
1143 1189
1144 /* workqueue for dio unwritten */ 1190 /* workqueue for dio unwritten */
1145 struct workqueue_struct *dio_unwritten_wq; 1191 struct workqueue_struct *dio_unwritten_wq;
1192
1193 /* timer for periodic error stats printing */
1194 struct timer_list s_err_report;
1195
1196 /* Lazy inode table initialization info */
1197 struct ext4_li_request *s_li_request;
1198 /* Wait multiplier for lazy initialization thread */
1199 unsigned int s_li_wait_mult;
1146}; 1200};
1147 1201
1148static inline struct ext4_sb_info *EXT4_SB(struct super_block *sb) 1202static inline struct ext4_sb_info *EXT4_SB(struct super_block *sb)
@@ -1313,6 +1367,10 @@ EXT4_INODE_BIT_FNS(state, state_flags)
1313#define EXT4_DEFM_JMODE_DATA 0x0020 1367#define EXT4_DEFM_JMODE_DATA 0x0020
1314#define EXT4_DEFM_JMODE_ORDERED 0x0040 1368#define EXT4_DEFM_JMODE_ORDERED 0x0040
1315#define EXT4_DEFM_JMODE_WBACK 0x0060 1369#define EXT4_DEFM_JMODE_WBACK 0x0060
1370#define EXT4_DEFM_NOBARRIER 0x0100
1371#define EXT4_DEFM_BLOCK_VALIDITY 0x0200
1372#define EXT4_DEFM_DISCARD 0x0400
1373#define EXT4_DEFM_NODELALLOC 0x0800
1316 1374
1317/* 1375/*
1318 * Default journal batch times 1376 * Default journal batch times
@@ -1379,6 +1437,43 @@ struct ext4_dir_entry_2 {
1379#define EXT4_MAX_REC_LEN ((1<<16)-1) 1437#define EXT4_MAX_REC_LEN ((1<<16)-1)
1380 1438
1381/* 1439/*
1440 * If we ever get support for fs block sizes > page_size, we'll need
1441 * to remove the #if statements in the next two functions...
1442 */
1443static inline unsigned int
1444ext4_rec_len_from_disk(__le16 dlen, unsigned blocksize)
1445{
1446 unsigned len = le16_to_cpu(dlen);
1447
1448#if (PAGE_CACHE_SIZE >= 65536)
1449 if (len == EXT4_MAX_REC_LEN || len == 0)
1450 return blocksize;
1451 return (len & 65532) | ((len & 3) << 16);
1452#else
1453 return len;
1454#endif
1455}
1456
1457static inline __le16 ext4_rec_len_to_disk(unsigned len, unsigned blocksize)
1458{
1459 if ((len > blocksize) || (blocksize > (1 << 18)) || (len & 3))
1460 BUG();
1461#if (PAGE_CACHE_SIZE >= 65536)
1462 if (len < 65536)
1463 return cpu_to_le16(len);
1464 if (len == blocksize) {
1465 if (blocksize == 65536)
1466 return cpu_to_le16(EXT4_MAX_REC_LEN);
1467 else
1468 return cpu_to_le16(0);
1469 }
1470 return cpu_to_le16((len & 65532) | ((len >> 16) & 3));
1471#else
1472 return cpu_to_le16(len);
1473#endif
1474}
1475
1476/*
1382 * Hash Tree Directory indexing 1477 * Hash Tree Directory indexing
1383 * (c) Daniel Phillips, 2001 1478 * (c) Daniel Phillips, 2001
1384 */ 1479 */
@@ -1463,7 +1558,42 @@ ext4_group_first_block_no(struct super_block *sb, ext4_group_t group_no)
1463void ext4_get_group_no_and_offset(struct super_block *sb, ext4_fsblk_t blocknr, 1558void ext4_get_group_no_and_offset(struct super_block *sb, ext4_fsblk_t blocknr,
1464 ext4_group_t *blockgrpp, ext4_grpblk_t *offsetp); 1559 ext4_group_t *blockgrpp, ext4_grpblk_t *offsetp);
1465 1560
1466extern struct proc_dir_entry *ext4_proc_root; 1561/*
1562 * Timeout and state flag for lazy initialization inode thread.
1563 */
1564#define EXT4_DEF_LI_WAIT_MULT 10
1565#define EXT4_DEF_LI_MAX_START_DELAY 5
1566#define EXT4_LAZYINIT_QUIT 0x0001
1567#define EXT4_LAZYINIT_RUNNING 0x0002
1568
1569/*
1570 * Lazy inode table initialization info
1571 */
1572struct ext4_lazy_init {
1573 unsigned long li_state;
1574
1575 wait_queue_head_t li_wait_daemon;
1576 wait_queue_head_t li_wait_task;
1577 struct timer_list li_timer;
1578 struct task_struct *li_task;
1579
1580 struct list_head li_request_list;
1581 struct mutex li_list_mtx;
1582};
1583
1584struct ext4_li_request {
1585 struct super_block *lr_super;
1586 struct ext4_sb_info *lr_sbi;
1587 ext4_group_t lr_next_group;
1588 struct list_head lr_request;
1589 unsigned long lr_next_sched;
1590 unsigned long lr_timeout;
1591};
1592
1593struct ext4_features {
1594 struct kobject f_kobj;
1595 struct completion f_kobj_unregister;
1596};
1467 1597
1468/* 1598/*
1469 * Function prototypes 1599 * Function prototypes
@@ -1491,7 +1621,6 @@ extern unsigned long ext4_bg_num_gdb(struct super_block *sb,
1491extern ext4_fsblk_t ext4_new_meta_blocks(handle_t *handle, struct inode *inode, 1621extern ext4_fsblk_t ext4_new_meta_blocks(handle_t *handle, struct inode *inode,
1492 ext4_fsblk_t goal, unsigned long *count, int *errp); 1622 ext4_fsblk_t goal, unsigned long *count, int *errp);
1493extern int ext4_claim_free_blocks(struct ext4_sb_info *sbi, s64 nblocks); 1623extern int ext4_claim_free_blocks(struct ext4_sb_info *sbi, s64 nblocks);
1494extern int ext4_has_free_blocks(struct ext4_sb_info *sbi, s64 nblocks);
1495extern void ext4_add_groupblocks(handle_t *handle, struct super_block *sb, 1624extern void ext4_add_groupblocks(handle_t *handle, struct super_block *sb,
1496 ext4_fsblk_t block, unsigned long count); 1625 ext4_fsblk_t block, unsigned long count);
1497extern ext4_fsblk_t ext4_count_free_blocks(struct super_block *); 1626extern ext4_fsblk_t ext4_count_free_blocks(struct super_block *);
@@ -1510,9 +1639,11 @@ extern unsigned ext4_init_block_bitmap(struct super_block *sb,
1510 ext4_init_block_bitmap(sb, NULL, group, desc) 1639 ext4_init_block_bitmap(sb, NULL, group, desc)
1511 1640
1512/* dir.c */ 1641/* dir.c */
1513extern int ext4_check_dir_entry(const char *, struct inode *, 1642extern int __ext4_check_dir_entry(const char *, unsigned int, struct inode *,
1514 struct ext4_dir_entry_2 *, 1643 struct ext4_dir_entry_2 *,
1515 struct buffer_head *, unsigned int); 1644 struct buffer_head *, unsigned int);
1645#define ext4_check_dir_entry(dir, de, bh, offset) \
1646 __ext4_check_dir_entry(__func__, __LINE__, (dir), (de), (bh), (offset))
1516extern int ext4_htree_store_dirent(struct file *dir_file, __u32 hash, 1647extern int ext4_htree_store_dirent(struct file *dir_file, __u32 hash,
1517 __u32 minor_hash, 1648 __u32 minor_hash,
1518 struct ext4_dir_entry_2 *dirent); 1649 struct ext4_dir_entry_2 *dirent);
@@ -1533,11 +1664,9 @@ extern struct inode * ext4_orphan_get(struct super_block *, unsigned long);
1533extern unsigned long ext4_count_free_inodes(struct super_block *); 1664extern unsigned long ext4_count_free_inodes(struct super_block *);
1534extern unsigned long ext4_count_dirs(struct super_block *); 1665extern unsigned long ext4_count_dirs(struct super_block *);
1535extern void ext4_check_inodes_bitmap(struct super_block *); 1666extern void ext4_check_inodes_bitmap(struct super_block *);
1536extern unsigned ext4_init_inode_bitmap(struct super_block *sb, 1667extern void ext4_mark_bitmap_end(int start_bit, int end_bit, char *bitmap);
1537 struct buffer_head *bh, 1668extern int ext4_init_inode_table(struct super_block *sb,
1538 ext4_group_t group, 1669 ext4_group_t group, int barrier);
1539 struct ext4_group_desc *desc);
1540extern void mark_bitmap_end(int start_bit, int end_bit, char *bitmap);
1541 1670
1542/* mballoc.c */ 1671/* mballoc.c */
1543extern long ext4_mb_stats; 1672extern long ext4_mb_stats;
@@ -1548,16 +1677,15 @@ extern ext4_fsblk_t ext4_mb_new_blocks(handle_t *,
1548 struct ext4_allocation_request *, int *); 1677 struct ext4_allocation_request *, int *);
1549extern int ext4_mb_reserve_blocks(struct super_block *, int); 1678extern int ext4_mb_reserve_blocks(struct super_block *, int);
1550extern void ext4_discard_preallocations(struct inode *); 1679extern void ext4_discard_preallocations(struct inode *);
1551extern int __init init_ext4_mballoc(void); 1680extern int __init ext4_init_mballoc(void);
1552extern void exit_ext4_mballoc(void); 1681extern void ext4_exit_mballoc(void);
1553extern void ext4_free_blocks(handle_t *handle, struct inode *inode, 1682extern void ext4_free_blocks(handle_t *handle, struct inode *inode,
1554 struct buffer_head *bh, ext4_fsblk_t block, 1683 struct buffer_head *bh, ext4_fsblk_t block,
1555 unsigned long count, int flags); 1684 unsigned long count, int flags);
1556extern int ext4_mb_add_groupinfo(struct super_block *sb, 1685extern int ext4_mb_add_groupinfo(struct super_block *sb,
1557 ext4_group_t i, struct ext4_group_desc *desc); 1686 ext4_group_t i, struct ext4_group_desc *desc);
1558extern int ext4_mb_get_buddy_cache_lock(struct super_block *, ext4_group_t); 1687extern int ext4_trim_fs(struct super_block *, struct fstrim_range *);
1559extern void ext4_mb_put_buddy_cache_lock(struct super_block *, 1688
1560 ext4_group_t, int);
1561/* inode.c */ 1689/* inode.c */
1562struct buffer_head *ext4_getblk(handle_t *, struct inode *, 1690struct buffer_head *ext4_getblk(handle_t *, struct inode *,
1563 ext4_lblk_t, int, int *); 1691 ext4_lblk_t, int, int *);
@@ -1571,7 +1699,8 @@ extern int ext4_write_inode(struct inode *, struct writeback_control *);
1571extern int ext4_setattr(struct dentry *, struct iattr *); 1699extern int ext4_setattr(struct dentry *, struct iattr *);
1572extern int ext4_getattr(struct vfsmount *mnt, struct dentry *dentry, 1700extern int ext4_getattr(struct vfsmount *mnt, struct dentry *dentry,
1573 struct kstat *stat); 1701 struct kstat *stat);
1574extern void ext4_delete_inode(struct inode *); 1702extern void ext4_evict_inode(struct inode *);
1703extern void ext4_clear_inode(struct inode *);
1575extern int ext4_sync_inode(handle_t *, struct inode *); 1704extern int ext4_sync_inode(handle_t *, struct inode *);
1576extern void ext4_dirty_inode(struct inode *); 1705extern void ext4_dirty_inode(struct inode *);
1577extern int ext4_change_inode_journal_flag(struct inode *, int); 1706extern int ext4_change_inode_journal_flag(struct inode *, int);
@@ -1584,13 +1713,11 @@ extern void ext4_get_inode_flags(struct ext4_inode_info *);
1584extern int ext4_alloc_da_blocks(struct inode *inode); 1713extern int ext4_alloc_da_blocks(struct inode *inode);
1585extern void ext4_set_aops(struct inode *inode); 1714extern void ext4_set_aops(struct inode *inode);
1586extern int ext4_writepage_trans_blocks(struct inode *); 1715extern int ext4_writepage_trans_blocks(struct inode *);
1587extern int ext4_meta_trans_blocks(struct inode *, int nrblocks, int idxblocks);
1588extern int ext4_chunk_trans_blocks(struct inode *, int nrblocks); 1716extern int ext4_chunk_trans_blocks(struct inode *, int nrblocks);
1589extern int ext4_block_truncate_page(handle_t *handle, 1717extern int ext4_block_truncate_page(handle_t *handle,
1590 struct address_space *mapping, loff_t from); 1718 struct address_space *mapping, loff_t from);
1591extern int ext4_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf); 1719extern int ext4_page_mkwrite(struct vm_area_struct *vma, struct vm_fault *vmf);
1592extern qsize_t *ext4_get_reserved_space(struct inode *inode); 1720extern qsize_t *ext4_get_reserved_space(struct inode *inode);
1593extern int flush_completed_IO(struct inode *inode);
1594extern void ext4_da_update_reserve_space(struct inode *inode, 1721extern void ext4_da_update_reserve_space(struct inode *inode,
1595 int used, int quota_claim); 1722 int used, int quota_claim);
1596/* ioctl.c */ 1723/* ioctl.c */
@@ -1601,8 +1728,6 @@ extern long ext4_compat_ioctl(struct file *, unsigned int, unsigned long);
1601extern int ext4_ext_migrate(struct inode *); 1728extern int ext4_ext_migrate(struct inode *);
1602 1729
1603/* namei.c */ 1730/* namei.c */
1604extern unsigned int ext4_rec_len_from_disk(__le16 dlen, unsigned blocksize);
1605extern __le16 ext4_rec_len_to_disk(unsigned len, unsigned blocksize);
1606extern int ext4_orphan_add(handle_t *, struct inode *); 1731extern int ext4_orphan_add(handle_t *, struct inode *);
1607extern int ext4_orphan_del(handle_t *, struct inode *); 1732extern int ext4_orphan_del(handle_t *, struct inode *);
1608extern int ext4_htree_fill_tree(struct file *dir_file, __u32 start_hash, 1733extern int ext4_htree_fill_tree(struct file *dir_file, __u32 start_hash,
@@ -1616,25 +1741,38 @@ extern int ext4_group_extend(struct super_block *sb,
1616 ext4_fsblk_t n_blocks_count); 1741 ext4_fsblk_t n_blocks_count);
1617 1742
1618/* super.c */ 1743/* super.c */
1619extern void __ext4_error(struct super_block *, const char *, const char *, ...) 1744extern void __ext4_error(struct super_block *, const char *, unsigned int,
1620 __attribute__ ((format (printf, 3, 4))); 1745 const char *, ...)
1621#define ext4_error(sb, message...) __ext4_error(sb, __func__, ## message) 1746 __attribute__ ((format (printf, 4, 5)));
1622extern void ext4_error_inode(const char *, struct inode *, const char *, ...) 1747#define ext4_error(sb, message...) __ext4_error(sb, __func__, \
1623 __attribute__ ((format (printf, 3, 4))); 1748 __LINE__, ## message)
1624extern void ext4_error_file(const char *, struct file *, const char *, ...) 1749extern void ext4_error_inode(struct inode *, const char *, unsigned int,
1625 __attribute__ ((format (printf, 3, 4))); 1750 ext4_fsblk_t, const char *, ...)
1626extern void __ext4_std_error(struct super_block *, const char *, int); 1751 __attribute__ ((format (printf, 5, 6)));
1627extern void ext4_abort(struct super_block *, const char *, const char *, ...) 1752extern void ext4_error_file(struct file *, const char *, unsigned int,
1628 __attribute__ ((format (printf, 3, 4))); 1753 const char *, ...)
1629extern void __ext4_warning(struct super_block *, const char *, 1754 __attribute__ ((format (printf, 4, 5)));
1755extern void __ext4_std_error(struct super_block *, const char *,
1756 unsigned int, int);
1757extern void __ext4_abort(struct super_block *, const char *, unsigned int,
1758 const char *, ...)
1759 __attribute__ ((format (printf, 4, 5)));
1760#define ext4_abort(sb, message...) __ext4_abort(sb, __func__, \
1761 __LINE__, ## message)
1762extern void __ext4_warning(struct super_block *, const char *, unsigned int,
1630 const char *, ...) 1763 const char *, ...)
1631 __attribute__ ((format (printf, 3, 4))); 1764 __attribute__ ((format (printf, 4, 5)));
1632#define ext4_warning(sb, message...) __ext4_warning(sb, __func__, ## message) 1765#define ext4_warning(sb, message...) __ext4_warning(sb, __func__, \
1766 __LINE__, ## message)
1633extern void ext4_msg(struct super_block *, const char *, const char *, ...) 1767extern void ext4_msg(struct super_block *, const char *, const char *, ...)
1634 __attribute__ ((format (printf, 3, 4))); 1768 __attribute__ ((format (printf, 3, 4)));
1635extern void ext4_grp_locked_error(struct super_block *, ext4_group_t, 1769extern void __ext4_grp_locked_error(const char *, unsigned int, \
1636 const char *, const char *, ...) 1770 struct super_block *, ext4_group_t, \
1637 __attribute__ ((format (printf, 4, 5))); 1771 unsigned long, ext4_fsblk_t, \
1772 const char *, ...)
1773 __attribute__ ((format (printf, 7, 8)));
1774#define ext4_grp_locked_error(sb, grp, message...) \
1775 __ext4_grp_locked_error(__func__, __LINE__, (sb), (grp), ## message)
1638extern void ext4_update_dynamic_rev(struct super_block *sb); 1776extern void ext4_update_dynamic_rev(struct super_block *sb);
1639extern int ext4_update_compat_feature(handle_t *handle, struct super_block *sb, 1777extern int ext4_update_compat_feature(handle_t *handle, struct super_block *sb,
1640 __u32 compat); 1778 __u32 compat);
@@ -1768,7 +1906,7 @@ static inline unsigned int ext4_flex_bg_size(struct ext4_sb_info *sbi)
1768#define ext4_std_error(sb, errno) \ 1906#define ext4_std_error(sb, errno) \
1769do { \ 1907do { \
1770 if ((errno)) \ 1908 if ((errno)) \
1771 __ext4_std_error((sb), __func__, (errno)); \ 1909 __ext4_std_error((sb), __func__, __LINE__, (errno)); \
1772} while (0) 1910} while (0)
1773 1911
1774#ifdef CONFIG_SMP 1912#ifdef CONFIG_SMP
@@ -1860,6 +1998,12 @@ static inline void ext4_unlock_group(struct super_block *sb,
1860 spin_unlock(ext4_group_lock_ptr(sb, group)); 1998 spin_unlock(ext4_group_lock_ptr(sb, group));
1861} 1999}
1862 2000
2001static inline void ext4_mark_super_dirty(struct super_block *sb)
2002{
2003 if (EXT4_SB(sb)->s_journal == NULL)
2004 sb->s_dirt =1;
2005}
2006
1863/* 2007/*
1864 * Inodes and files operations 2008 * Inodes and files operations
1865 */ 2009 */
@@ -1870,6 +2014,7 @@ extern const struct file_operations ext4_dir_operations;
1870/* file.c */ 2014/* file.c */
1871extern const struct inode_operations ext4_file_inode_operations; 2015extern const struct inode_operations ext4_file_inode_operations;
1872extern const struct file_operations ext4_file_operations; 2016extern const struct file_operations ext4_file_operations;
2017extern loff_t ext4_llseek(struct file *file, loff_t offset, int origin);
1873 2018
1874/* namei.c */ 2019/* namei.c */
1875extern const struct inode_operations ext4_dir_inode_operations; 2020extern const struct inode_operations ext4_dir_inode_operations;
@@ -1883,8 +2028,8 @@ extern const struct inode_operations ext4_fast_symlink_inode_operations;
1883/* block_validity */ 2028/* block_validity */
1884extern void ext4_release_system_zone(struct super_block *sb); 2029extern void ext4_release_system_zone(struct super_block *sb);
1885extern int ext4_setup_system_zone(struct super_block *sb); 2030extern int ext4_setup_system_zone(struct super_block *sb);
1886extern int __init init_ext4_system_zone(void); 2031extern int __init ext4_init_system_zone(void);
1887extern void exit_ext4_system_zone(void); 2032extern void ext4_exit_system_zone(void);
1888extern int ext4_data_block_valid(struct ext4_sb_info *sbi, 2033extern int ext4_data_block_valid(struct ext4_sb_info *sbi,
1889 ext4_fsblk_t start_blk, 2034 ext4_fsblk_t start_blk,
1890 unsigned int count); 2035 unsigned int count);
@@ -1905,9 +2050,6 @@ extern int ext4_convert_unwritten_extents(struct inode *inode, loff_t offset,
1905 ssize_t len); 2050 ssize_t len);
1906extern int ext4_map_blocks(handle_t *handle, struct inode *inode, 2051extern int ext4_map_blocks(handle_t *handle, struct inode *inode,
1907 struct ext4_map_blocks *map, int flags); 2052 struct ext4_map_blocks *map, int flags);
1908extern int ext4_get_blocks(handle_t *handle, struct inode *inode,
1909 sector_t block, unsigned int max_blocks,
1910 struct buffer_head *bh, int flags);
1911extern int ext4_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo, 2053extern int ext4_fiemap(struct inode *inode, struct fiemap_extent_info *fieinfo,
1912 __u64 start, __u64 len); 2054 __u64 start, __u64 len);
1913/* move_extent.c */ 2055/* move_extent.c */
@@ -1915,6 +2057,17 @@ extern int ext4_move_extents(struct file *o_filp, struct file *d_filp,
1915 __u64 start_orig, __u64 start_donor, 2057 __u64 start_orig, __u64 start_donor,
1916 __u64 len, __u64 *moved_len); 2058 __u64 len, __u64 *moved_len);
1917 2059
2060/* page-io.c */
2061extern int __init ext4_init_pageio(void);
2062extern void ext4_exit_pageio(void);
2063extern void ext4_free_io_end(ext4_io_end_t *io);
2064extern ext4_io_end_t *ext4_init_io_end(struct inode *inode, gfp_t flags);
2065extern int ext4_end_io_nolock(ext4_io_end_t *io);
2066extern void ext4_io_submit(struct ext4_io_submit *io);
2067extern int ext4_bio_write_page(struct ext4_io_submit *io,
2068 struct page *page,
2069 int len,
2070 struct writeback_control *wbc);
1918 2071
1919/* BH_Uninit flag: blocks are allocated but uninitialized on disk */ 2072/* BH_Uninit flag: blocks are allocated but uninitialized on disk */
1920enum ext4_state_bits { 2073enum ext4_state_bits {
diff --git a/fs/ext4/ext4_extents.h b/fs/ext4/ext4_extents.h
index bdb6ce7e2eb4..28ce70fd9cd0 100644
--- a/fs/ext4/ext4_extents.h
+++ b/fs/ext4/ext4_extents.h
@@ -225,11 +225,60 @@ static inline void ext4_ext_mark_initialized(struct ext4_extent *ext)
225 ext->ee_len = cpu_to_le16(ext4_ext_get_actual_len(ext)); 225 ext->ee_len = cpu_to_le16(ext4_ext_get_actual_len(ext));
226} 226}
227 227
228/*
229 * ext4_ext_pblock:
230 * combine low and high parts of physical block number into ext4_fsblk_t
231 */
232static inline ext4_fsblk_t ext4_ext_pblock(struct ext4_extent *ex)
233{
234 ext4_fsblk_t block;
235
236 block = le32_to_cpu(ex->ee_start_lo);
237 block |= ((ext4_fsblk_t) le16_to_cpu(ex->ee_start_hi) << 31) << 1;
238 return block;
239}
240
241/*
242 * ext4_idx_pblock:
243 * combine low and high parts of a leaf physical block number into ext4_fsblk_t
244 */
245static inline ext4_fsblk_t ext4_idx_pblock(struct ext4_extent_idx *ix)
246{
247 ext4_fsblk_t block;
248
249 block = le32_to_cpu(ix->ei_leaf_lo);
250 block |= ((ext4_fsblk_t) le16_to_cpu(ix->ei_leaf_hi) << 31) << 1;
251 return block;
252}
253
254/*
255 * ext4_ext_store_pblock:
256 * stores a large physical block number into an extent struct,
257 * breaking it into parts
258 */
259static inline void ext4_ext_store_pblock(struct ext4_extent *ex,
260 ext4_fsblk_t pb)
261{
262 ex->ee_start_lo = cpu_to_le32((unsigned long) (pb & 0xffffffff));
263 ex->ee_start_hi = cpu_to_le16((unsigned long) ((pb >> 31) >> 1) &
264 0xffff);
265}
266
267/*
268 * ext4_idx_store_pblock:
269 * stores a large physical block number into an index struct,
270 * breaking it into parts
271 */
272static inline void ext4_idx_store_pblock(struct ext4_extent_idx *ix,
273 ext4_fsblk_t pb)
274{
275 ix->ei_leaf_lo = cpu_to_le32((unsigned long) (pb & 0xffffffff));
276 ix->ei_leaf_hi = cpu_to_le16((unsigned long) ((pb >> 31) >> 1) &
277 0xffff);
278}
279
228extern int ext4_ext_calc_metadata_amount(struct inode *inode, 280extern int ext4_ext_calc_metadata_amount(struct inode *inode,
229 sector_t lblocks); 281 sector_t lblocks);
230extern ext4_fsblk_t ext_pblock(struct ext4_extent *ex);
231extern ext4_fsblk_t idx_pblock(struct ext4_extent_idx *);
232extern void ext4_ext_store_pblock(struct ext4_extent *, ext4_fsblk_t);
233extern int ext4_extent_tree_init(handle_t *, struct inode *); 282extern int ext4_extent_tree_init(handle_t *, struct inode *);
234extern int ext4_ext_calc_credits_for_single_extent(struct inode *inode, 283extern int ext4_ext_calc_credits_for_single_extent(struct inode *inode,
235 int num, 284 int num,
@@ -237,19 +286,9 @@ extern int ext4_ext_calc_credits_for_single_extent(struct inode *inode,
237extern int ext4_can_extents_be_merged(struct inode *inode, 286extern int ext4_can_extents_be_merged(struct inode *inode,
238 struct ext4_extent *ex1, 287 struct ext4_extent *ex1,
239 struct ext4_extent *ex2); 288 struct ext4_extent *ex2);
240extern int ext4_ext_try_to_merge(struct inode *inode,
241 struct ext4_ext_path *path,
242 struct ext4_extent *);
243extern unsigned int ext4_ext_check_overlap(struct inode *, struct ext4_extent *, struct ext4_ext_path *);
244extern int ext4_ext_insert_extent(handle_t *, struct inode *, struct ext4_ext_path *, struct ext4_extent *, int); 289extern int ext4_ext_insert_extent(handle_t *, struct inode *, struct ext4_ext_path *, struct ext4_extent *, int);
245extern int ext4_ext_walk_space(struct inode *, ext4_lblk_t, ext4_lblk_t,
246 ext_prepare_callback, void *);
247extern struct ext4_ext_path *ext4_ext_find_extent(struct inode *, ext4_lblk_t, 290extern struct ext4_ext_path *ext4_ext_find_extent(struct inode *, ext4_lblk_t,
248 struct ext4_ext_path *); 291 struct ext4_ext_path *);
249extern int ext4_ext_search_left(struct inode *, struct ext4_ext_path *,
250 ext4_lblk_t *, ext4_fsblk_t *);
251extern int ext4_ext_search_right(struct inode *, struct ext4_ext_path *,
252 ext4_lblk_t *, ext4_fsblk_t *);
253extern void ext4_ext_drop_refs(struct ext4_ext_path *); 292extern void ext4_ext_drop_refs(struct ext4_ext_path *);
254extern int ext4_ext_check_inode(struct inode *inode); 293extern int ext4_ext_check_inode(struct inode *inode);
255#endif /* _EXT4_EXTENTS */ 294#endif /* _EXT4_EXTENTS */
diff --git a/fs/ext4/ext4_jbd2.c b/fs/ext4/ext4_jbd2.c
index 53d2764d71ca..6e272ef6ba96 100644
--- a/fs/ext4/ext4_jbd2.c
+++ b/fs/ext4/ext4_jbd2.c
@@ -6,29 +6,29 @@
6 6
7#include <trace/events/ext4.h> 7#include <trace/events/ext4.h>
8 8
9int __ext4_journal_get_undo_access(const char *where, handle_t *handle, 9int __ext4_journal_get_undo_access(const char *where, unsigned int line,
10 struct buffer_head *bh) 10 handle_t *handle, struct buffer_head *bh)
11{ 11{
12 int err = 0; 12 int err = 0;
13 13
14 if (ext4_handle_valid(handle)) { 14 if (ext4_handle_valid(handle)) {
15 err = jbd2_journal_get_undo_access(handle, bh); 15 err = jbd2_journal_get_undo_access(handle, bh);
16 if (err) 16 if (err)
17 ext4_journal_abort_handle(where, __func__, bh, 17 ext4_journal_abort_handle(where, line, __func__, bh,
18 handle, err); 18 handle, err);
19 } 19 }
20 return err; 20 return err;
21} 21}
22 22
23int __ext4_journal_get_write_access(const char *where, handle_t *handle, 23int __ext4_journal_get_write_access(const char *where, unsigned int line,
24 struct buffer_head *bh) 24 handle_t *handle, struct buffer_head *bh)
25{ 25{
26 int err = 0; 26 int err = 0;
27 27
28 if (ext4_handle_valid(handle)) { 28 if (ext4_handle_valid(handle)) {
29 err = jbd2_journal_get_write_access(handle, bh); 29 err = jbd2_journal_get_write_access(handle, bh);
30 if (err) 30 if (err)
31 ext4_journal_abort_handle(where, __func__, bh, 31 ext4_journal_abort_handle(where, line, __func__, bh,
32 handle, err); 32 handle, err);
33 } 33 }
34 return err; 34 return err;
@@ -46,9 +46,9 @@ int __ext4_journal_get_write_access(const char *where, handle_t *handle,
46 * If the handle isn't valid we're not journaling, but we still need to 46 * If the handle isn't valid we're not journaling, but we still need to
47 * call into ext4_journal_revoke() to put the buffer head. 47 * call into ext4_journal_revoke() to put the buffer head.
48 */ 48 */
49int __ext4_forget(const char *where, handle_t *handle, int is_metadata, 49int __ext4_forget(const char *where, unsigned int line, handle_t *handle,
50 struct inode *inode, struct buffer_head *bh, 50 int is_metadata, struct inode *inode,
51 ext4_fsblk_t blocknr) 51 struct buffer_head *bh, ext4_fsblk_t blocknr)
52{ 52{
53 int err; 53 int err;
54 54
@@ -79,8 +79,8 @@ int __ext4_forget(const char *where, handle_t *handle, int is_metadata,
79 BUFFER_TRACE(bh, "call jbd2_journal_forget"); 79 BUFFER_TRACE(bh, "call jbd2_journal_forget");
80 err = jbd2_journal_forget(handle, bh); 80 err = jbd2_journal_forget(handle, bh);
81 if (err) 81 if (err)
82 ext4_journal_abort_handle(where, __func__, bh, 82 ext4_journal_abort_handle(where, line, __func__,
83 handle, err); 83 bh, handle, err);
84 return err; 84 return err;
85 } 85 }
86 return 0; 86 return 0;
@@ -92,15 +92,16 @@ int __ext4_forget(const char *where, handle_t *handle, int is_metadata,
92 BUFFER_TRACE(bh, "call jbd2_journal_revoke"); 92 BUFFER_TRACE(bh, "call jbd2_journal_revoke");
93 err = jbd2_journal_revoke(handle, blocknr, bh); 93 err = jbd2_journal_revoke(handle, blocknr, bh);
94 if (err) { 94 if (err) {
95 ext4_journal_abort_handle(where, __func__, bh, handle, err); 95 ext4_journal_abort_handle(where, line, __func__,
96 ext4_abort(inode->i_sb, __func__, 96 bh, handle, err);
97 __ext4_abort(inode->i_sb, where, line,
97 "error %d when attempting revoke", err); 98 "error %d when attempting revoke", err);
98 } 99 }
99 BUFFER_TRACE(bh, "exit"); 100 BUFFER_TRACE(bh, "exit");
100 return err; 101 return err;
101} 102}
102 103
103int __ext4_journal_get_create_access(const char *where, 104int __ext4_journal_get_create_access(const char *where, unsigned int line,
104 handle_t *handle, struct buffer_head *bh) 105 handle_t *handle, struct buffer_head *bh)
105{ 106{
106 int err = 0; 107 int err = 0;
@@ -108,22 +109,23 @@ int __ext4_journal_get_create_access(const char *where,
108 if (ext4_handle_valid(handle)) { 109 if (ext4_handle_valid(handle)) {
109 err = jbd2_journal_get_create_access(handle, bh); 110 err = jbd2_journal_get_create_access(handle, bh);
110 if (err) 111 if (err)
111 ext4_journal_abort_handle(where, __func__, bh, 112 ext4_journal_abort_handle(where, line, __func__,
112 handle, err); 113 bh, handle, err);
113 } 114 }
114 return err; 115 return err;
115} 116}
116 117
117int __ext4_handle_dirty_metadata(const char *where, handle_t *handle, 118int __ext4_handle_dirty_metadata(const char *where, unsigned int line,
118 struct inode *inode, struct buffer_head *bh) 119 handle_t *handle, struct inode *inode,
120 struct buffer_head *bh)
119{ 121{
120 int err = 0; 122 int err = 0;
121 123
122 if (ext4_handle_valid(handle)) { 124 if (ext4_handle_valid(handle)) {
123 err = jbd2_journal_dirty_metadata(handle, bh); 125 err = jbd2_journal_dirty_metadata(handle, bh);
124 if (err) 126 if (err)
125 ext4_journal_abort_handle(where, __func__, bh, 127 ext4_journal_abort_handle(where, line, __func__,
126 handle, err); 128 bh, handle, err);
127 } else { 129 } else {
128 if (inode) 130 if (inode)
129 mark_buffer_dirty_inode(bh, inode); 131 mark_buffer_dirty_inode(bh, inode);
@@ -132,14 +134,33 @@ int __ext4_handle_dirty_metadata(const char *where, handle_t *handle,
132 if (inode && inode_needs_sync(inode)) { 134 if (inode && inode_needs_sync(inode)) {
133 sync_dirty_buffer(bh); 135 sync_dirty_buffer(bh);
134 if (buffer_req(bh) && !buffer_uptodate(bh)) { 136 if (buffer_req(bh) && !buffer_uptodate(bh)) {
135 ext4_error(inode->i_sb, 137 struct ext4_super_block *es;
136 "IO error syncing inode, " 138
137 "inode=%lu, block=%llu", 139 es = EXT4_SB(inode->i_sb)->s_es;
138 inode->i_ino, 140 es->s_last_error_block =
139 (unsigned long long) bh->b_blocknr); 141 cpu_to_le64(bh->b_blocknr);
142 ext4_error_inode(inode, where, line,
143 bh->b_blocknr,
144 "IO error syncing itable block");
140 err = -EIO; 145 err = -EIO;
141 } 146 }
142 } 147 }
143 } 148 }
144 return err; 149 return err;
145} 150}
151
152int __ext4_handle_dirty_super(const char *where, unsigned int line,
153 handle_t *handle, struct super_block *sb)
154{
155 struct buffer_head *bh = EXT4_SB(sb)->s_sbh;
156 int err = 0;
157
158 if (ext4_handle_valid(handle)) {
159 err = jbd2_journal_dirty_metadata(handle, bh);
160 if (err)
161 ext4_journal_abort_handle(where, line, __func__,
162 bh, handle, err);
163 } else
164 sb->s_dirt = 1;
165 return err;
166}
diff --git a/fs/ext4/ext4_jbd2.h b/fs/ext4/ext4_jbd2.h
index dade0c024797..b0bd792c58c5 100644
--- a/fs/ext4/ext4_jbd2.h
+++ b/fs/ext4/ext4_jbd2.h
@@ -122,39 +122,47 @@ int ext4_mark_inode_dirty(handle_t *handle, struct inode *inode);
122/* 122/*
123 * Wrapper functions with which ext4 calls into JBD. 123 * Wrapper functions with which ext4 calls into JBD.
124 */ 124 */
125void ext4_journal_abort_handle(const char *caller, const char *err_fn, 125void ext4_journal_abort_handle(const char *caller, unsigned int line,
126 const char *err_fn,
126 struct buffer_head *bh, handle_t *handle, int err); 127 struct buffer_head *bh, handle_t *handle, int err);
127 128
128int __ext4_journal_get_undo_access(const char *where, handle_t *handle, 129int __ext4_journal_get_undo_access(const char *where, unsigned int line,
129 struct buffer_head *bh); 130 handle_t *handle, struct buffer_head *bh);
130 131
131int __ext4_journal_get_write_access(const char *where, handle_t *handle, 132int __ext4_journal_get_write_access(const char *where, unsigned int line,
132 struct buffer_head *bh); 133 handle_t *handle, struct buffer_head *bh);
133 134
134int __ext4_forget(const char *where, handle_t *handle, int is_metadata, 135int __ext4_forget(const char *where, unsigned int line, handle_t *handle,
135 struct inode *inode, struct buffer_head *bh, 136 int is_metadata, struct inode *inode,
136 ext4_fsblk_t blocknr); 137 struct buffer_head *bh, ext4_fsblk_t blocknr);
137 138
138int __ext4_journal_get_create_access(const char *where, 139int __ext4_journal_get_create_access(const char *where, unsigned int line,
139 handle_t *handle, struct buffer_head *bh); 140 handle_t *handle, struct buffer_head *bh);
140 141
141int __ext4_handle_dirty_metadata(const char *where, handle_t *handle, 142int __ext4_handle_dirty_metadata(const char *where, unsigned int line,
142 struct inode *inode, struct buffer_head *bh); 143 handle_t *handle, struct inode *inode,
144 struct buffer_head *bh);
145
146int __ext4_handle_dirty_super(const char *where, unsigned int line,
147 handle_t *handle, struct super_block *sb);
143 148
144#define ext4_journal_get_undo_access(handle, bh) \ 149#define ext4_journal_get_undo_access(handle, bh) \
145 __ext4_journal_get_undo_access(__func__, (handle), (bh)) 150 __ext4_journal_get_undo_access(__func__, __LINE__, (handle), (bh))
146#define ext4_journal_get_write_access(handle, bh) \ 151#define ext4_journal_get_write_access(handle, bh) \
147 __ext4_journal_get_write_access(__func__, (handle), (bh)) 152 __ext4_journal_get_write_access(__func__, __LINE__, (handle), (bh))
148#define ext4_forget(handle, is_metadata, inode, bh, block_nr) \ 153#define ext4_forget(handle, is_metadata, inode, bh, block_nr) \
149 __ext4_forget(__func__, (handle), (is_metadata), (inode), (bh),\ 154 __ext4_forget(__func__, __LINE__, (handle), (is_metadata), (inode), \
150 (block_nr)) 155 (bh), (block_nr))
151#define ext4_journal_get_create_access(handle, bh) \ 156#define ext4_journal_get_create_access(handle, bh) \
152 __ext4_journal_get_create_access(__func__, (handle), (bh)) 157 __ext4_journal_get_create_access(__func__, __LINE__, (handle), (bh))
153#define ext4_handle_dirty_metadata(handle, inode, bh) \ 158#define ext4_handle_dirty_metadata(handle, inode, bh) \
154 __ext4_handle_dirty_metadata(__func__, (handle), (inode), (bh)) 159 __ext4_handle_dirty_metadata(__func__, __LINE__, (handle), (inode), \
160 (bh))
161#define ext4_handle_dirty_super(handle, sb) \
162 __ext4_handle_dirty_super(__func__, __LINE__, (handle), (sb))
155 163
156handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks); 164handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks);
157int __ext4_journal_stop(const char *where, handle_t *handle); 165int __ext4_journal_stop(const char *where, unsigned int line, handle_t *handle);
158 166
159#define EXT4_NOJOURNAL_MAX_REF_COUNT ((unsigned long) 4096) 167#define EXT4_NOJOURNAL_MAX_REF_COUNT ((unsigned long) 4096)
160 168
@@ -207,7 +215,7 @@ static inline handle_t *ext4_journal_start(struct inode *inode, int nblocks)
207} 215}
208 216
209#define ext4_journal_stop(handle) \ 217#define ext4_journal_stop(handle) \
210 __ext4_journal_stop(__func__, (handle)) 218 __ext4_journal_stop(__func__, __LINE__, (handle))
211 219
212static inline handle_t *ext4_journal_current_handle(void) 220static inline handle_t *ext4_journal_current_handle(void)
213{ 221{
@@ -308,17 +316,15 @@ static inline int ext4_should_writeback_data(struct inode *inode)
308 * This function controls whether or not we should try to go down the 316 * This function controls whether or not we should try to go down the
309 * dioread_nolock code paths, which makes it safe to avoid taking 317 * dioread_nolock code paths, which makes it safe to avoid taking
310 * i_mutex for direct I/O reads. This only works for extent-based 318 * i_mutex for direct I/O reads. This only works for extent-based
311 * files, and it doesn't work for nobh or if data journaling is 319 * files, and it doesn't work if data journaling is enabled, since the
312 * enabled, since the dioread_nolock code uses b_private to pass 320 * dioread_nolock code uses b_private to pass information back to the
313 * information back to the I/O completion handler, and this conflicts 321 * I/O completion handler, and this conflicts with the jbd's use of
314 * with the jbd's use of b_private. 322 * b_private.
315 */ 323 */
316static inline int ext4_should_dioread_nolock(struct inode *inode) 324static inline int ext4_should_dioread_nolock(struct inode *inode)
317{ 325{
318 if (!test_opt(inode->i_sb, DIOREAD_NOLOCK)) 326 if (!test_opt(inode->i_sb, DIOREAD_NOLOCK))
319 return 0; 327 return 0;
320 if (test_opt(inode->i_sb, NOBH))
321 return 0;
322 if (!S_ISREG(inode->i_mode)) 328 if (!S_ISREG(inode->i_mode))
323 return 0; 329 return 0;
324 if (!(ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS))) 330 if (!(ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS)))
diff --git a/fs/ext4/extents.c b/fs/ext4/extents.c
index 377309c1af65..0554c48cb1fd 100644
--- a/fs/ext4/extents.c
+++ b/fs/ext4/extents.c
@@ -44,55 +44,6 @@
44#include "ext4_jbd2.h" 44#include "ext4_jbd2.h"
45#include "ext4_extents.h" 45#include "ext4_extents.h"
46 46
47
48/*
49 * ext_pblock:
50 * combine low and high parts of physical block number into ext4_fsblk_t
51 */
52ext4_fsblk_t ext_pblock(struct ext4_extent *ex)
53{
54 ext4_fsblk_t block;
55
56 block = le32_to_cpu(ex->ee_start_lo);
57 block |= ((ext4_fsblk_t) le16_to_cpu(ex->ee_start_hi) << 31) << 1;
58 return block;
59}
60
61/*
62 * idx_pblock:
63 * combine low and high parts of a leaf physical block number into ext4_fsblk_t
64 */
65ext4_fsblk_t idx_pblock(struct ext4_extent_idx *ix)
66{
67 ext4_fsblk_t block;
68
69 block = le32_to_cpu(ix->ei_leaf_lo);
70 block |= ((ext4_fsblk_t) le16_to_cpu(ix->ei_leaf_hi) << 31) << 1;
71 return block;
72}
73
74/*
75 * ext4_ext_store_pblock:
76 * stores a large physical block number into an extent struct,
77 * breaking it into parts
78 */
79void ext4_ext_store_pblock(struct ext4_extent *ex, ext4_fsblk_t pb)
80{
81 ex->ee_start_lo = cpu_to_le32((unsigned long) (pb & 0xffffffff));
82 ex->ee_start_hi = cpu_to_le16((unsigned long) ((pb >> 31) >> 1) & 0xffff);
83}
84
85/*
86 * ext4_idx_store_pblock:
87 * stores a large physical block number into an index struct,
88 * breaking it into parts
89 */
90static void ext4_idx_store_pblock(struct ext4_extent_idx *ix, ext4_fsblk_t pb)
91{
92 ix->ei_leaf_lo = cpu_to_le32((unsigned long) (pb & 0xffffffff));
93 ix->ei_leaf_hi = cpu_to_le16((unsigned long) ((pb >> 31) >> 1) & 0xffff);
94}
95
96static int ext4_ext_truncate_extend_restart(handle_t *handle, 47static int ext4_ext_truncate_extend_restart(handle_t *handle,
97 struct inode *inode, 48 struct inode *inode,
98 int needed) 49 int needed)
@@ -169,7 +120,8 @@ static ext4_fsblk_t ext4_ext_find_goal(struct inode *inode,
169 /* try to predict block placement */ 120 /* try to predict block placement */
170 ex = path[depth].p_ext; 121 ex = path[depth].p_ext;
171 if (ex) 122 if (ex)
172 return ext_pblock(ex)+(block-le32_to_cpu(ex->ee_block)); 123 return (ext4_ext_pblock(ex) +
124 (block - le32_to_cpu(ex->ee_block)));
173 125
174 /* it looks like index is empty; 126 /* it looks like index is empty;
175 * try to find starting block from index itself */ 127 * try to find starting block from index itself */
@@ -354,7 +306,7 @@ ext4_ext_max_entries(struct inode *inode, int depth)
354 306
355static int ext4_valid_extent(struct inode *inode, struct ext4_extent *ext) 307static int ext4_valid_extent(struct inode *inode, struct ext4_extent *ext)
356{ 308{
357 ext4_fsblk_t block = ext_pblock(ext); 309 ext4_fsblk_t block = ext4_ext_pblock(ext);
358 int len = ext4_ext_get_actual_len(ext); 310 int len = ext4_ext_get_actual_len(ext);
359 311
360 return ext4_data_block_valid(EXT4_SB(inode->i_sb), block, len); 312 return ext4_data_block_valid(EXT4_SB(inode->i_sb), block, len);
@@ -363,7 +315,7 @@ static int ext4_valid_extent(struct inode *inode, struct ext4_extent *ext)
363static int ext4_valid_extent_idx(struct inode *inode, 315static int ext4_valid_extent_idx(struct inode *inode,
364 struct ext4_extent_idx *ext_idx) 316 struct ext4_extent_idx *ext_idx)
365{ 317{
366 ext4_fsblk_t block = idx_pblock(ext_idx); 318 ext4_fsblk_t block = ext4_idx_pblock(ext_idx);
367 319
368 return ext4_data_block_valid(EXT4_SB(inode->i_sb), block, 1); 320 return ext4_data_block_valid(EXT4_SB(inode->i_sb), block, 1);
369} 321}
@@ -401,9 +353,9 @@ static int ext4_valid_extent_entries(struct inode *inode,
401 return 1; 353 return 1;
402} 354}
403 355
404static int __ext4_ext_check(const char *function, struct inode *inode, 356static int __ext4_ext_check(const char *function, unsigned int line,
405 struct ext4_extent_header *eh, 357 struct inode *inode, struct ext4_extent_header *eh,
406 int depth) 358 int depth)
407{ 359{
408 const char *error_msg; 360 const char *error_msg;
409 int max = 0; 361 int max = 0;
@@ -436,7 +388,7 @@ static int __ext4_ext_check(const char *function, struct inode *inode,
436 return 0; 388 return 0;
437 389
438corrupted: 390corrupted:
439 ext4_error_inode(function, inode, 391 ext4_error_inode(inode, function, line, 0,
440 "bad header/extent: %s - magic %x, " 392 "bad header/extent: %s - magic %x, "
441 "entries %u, max %u(%u), depth %u(%u)", 393 "entries %u, max %u(%u), depth %u(%u)",
442 error_msg, le16_to_cpu(eh->eh_magic), 394 error_msg, le16_to_cpu(eh->eh_magic),
@@ -447,7 +399,7 @@ corrupted:
447} 399}
448 400
449#define ext4_ext_check(inode, eh, depth) \ 401#define ext4_ext_check(inode, eh, depth) \
450 __ext4_ext_check(__func__, inode, eh, depth) 402 __ext4_ext_check(__func__, __LINE__, inode, eh, depth)
451 403
452int ext4_ext_check_inode(struct inode *inode) 404int ext4_ext_check_inode(struct inode *inode)
453{ 405{
@@ -463,13 +415,13 @@ static void ext4_ext_show_path(struct inode *inode, struct ext4_ext_path *path)
463 for (k = 0; k <= l; k++, path++) { 415 for (k = 0; k <= l; k++, path++) {
464 if (path->p_idx) { 416 if (path->p_idx) {
465 ext_debug(" %d->%llu", le32_to_cpu(path->p_idx->ei_block), 417 ext_debug(" %d->%llu", le32_to_cpu(path->p_idx->ei_block),
466 idx_pblock(path->p_idx)); 418 ext4_idx_pblock(path->p_idx));
467 } else if (path->p_ext) { 419 } else if (path->p_ext) {
468 ext_debug(" %d:[%d]%d:%llu ", 420 ext_debug(" %d:[%d]%d:%llu ",
469 le32_to_cpu(path->p_ext->ee_block), 421 le32_to_cpu(path->p_ext->ee_block),
470 ext4_ext_is_uninitialized(path->p_ext), 422 ext4_ext_is_uninitialized(path->p_ext),
471 ext4_ext_get_actual_len(path->p_ext), 423 ext4_ext_get_actual_len(path->p_ext),
472 ext_pblock(path->p_ext)); 424 ext4_ext_pblock(path->p_ext));
473 } else 425 } else
474 ext_debug(" []"); 426 ext_debug(" []");
475 } 427 }
@@ -494,7 +446,7 @@ static void ext4_ext_show_leaf(struct inode *inode, struct ext4_ext_path *path)
494 for (i = 0; i < le16_to_cpu(eh->eh_entries); i++, ex++) { 446 for (i = 0; i < le16_to_cpu(eh->eh_entries); i++, ex++) {
495 ext_debug("%d:[%d]%d:%llu ", le32_to_cpu(ex->ee_block), 447 ext_debug("%d:[%d]%d:%llu ", le32_to_cpu(ex->ee_block),
496 ext4_ext_is_uninitialized(ex), 448 ext4_ext_is_uninitialized(ex),
497 ext4_ext_get_actual_len(ex), ext_pblock(ex)); 449 ext4_ext_get_actual_len(ex), ext4_ext_pblock(ex));
498 } 450 }
499 ext_debug("\n"); 451 ext_debug("\n");
500} 452}
@@ -545,7 +497,7 @@ ext4_ext_binsearch_idx(struct inode *inode,
545 497
546 path->p_idx = l - 1; 498 path->p_idx = l - 1;
547 ext_debug(" -> %d->%lld ", le32_to_cpu(path->p_idx->ei_block), 499 ext_debug(" -> %d->%lld ", le32_to_cpu(path->p_idx->ei_block),
548 idx_pblock(path->p_idx)); 500 ext4_idx_pblock(path->p_idx));
549 501
550#ifdef CHECK_BINSEARCH 502#ifdef CHECK_BINSEARCH
551 { 503 {
@@ -614,7 +566,7 @@ ext4_ext_binsearch(struct inode *inode,
614 path->p_ext = l - 1; 566 path->p_ext = l - 1;
615 ext_debug(" -> %d:%llu:[%d]%d ", 567 ext_debug(" -> %d:%llu:[%d]%d ",
616 le32_to_cpu(path->p_ext->ee_block), 568 le32_to_cpu(path->p_ext->ee_block),
617 ext_pblock(path->p_ext), 569 ext4_ext_pblock(path->p_ext),
618 ext4_ext_is_uninitialized(path->p_ext), 570 ext4_ext_is_uninitialized(path->p_ext),
619 ext4_ext_get_actual_len(path->p_ext)); 571 ext4_ext_get_actual_len(path->p_ext));
620 572
@@ -682,7 +634,7 @@ ext4_ext_find_extent(struct inode *inode, ext4_lblk_t block,
682 ppos, le16_to_cpu(eh->eh_entries), le16_to_cpu(eh->eh_max)); 634 ppos, le16_to_cpu(eh->eh_entries), le16_to_cpu(eh->eh_max));
683 635
684 ext4_ext_binsearch_idx(inode, path + ppos, block); 636 ext4_ext_binsearch_idx(inode, path + ppos, block);
685 path[ppos].p_block = idx_pblock(path[ppos].p_idx); 637 path[ppos].p_block = ext4_idx_pblock(path[ppos].p_idx);
686 path[ppos].p_depth = i; 638 path[ppos].p_depth = i;
687 path[ppos].p_ext = NULL; 639 path[ppos].p_ext = NULL;
688 640
@@ -721,7 +673,7 @@ ext4_ext_find_extent(struct inode *inode, ext4_lblk_t block,
721 ext4_ext_binsearch(inode, path + ppos, block); 673 ext4_ext_binsearch(inode, path + ppos, block);
722 /* if not an empty leaf */ 674 /* if not an empty leaf */
723 if (path[ppos].p_ext) 675 if (path[ppos].p_ext)
724 path[ppos].p_block = ext_pblock(path[ppos].p_ext); 676 path[ppos].p_block = ext4_ext_pblock(path[ppos].p_ext);
725 677
726 ext4_ext_show_path(inode, path); 678 ext4_ext_show_path(inode, path);
727 679
@@ -739,9 +691,9 @@ err:
739 * insert new index [@logical;@ptr] into the block at @curp; 691 * insert new index [@logical;@ptr] into the block at @curp;
740 * check where to insert: before @curp or after @curp 692 * check where to insert: before @curp or after @curp
741 */ 693 */
742int ext4_ext_insert_index(handle_t *handle, struct inode *inode, 694static int ext4_ext_insert_index(handle_t *handle, struct inode *inode,
743 struct ext4_ext_path *curp, 695 struct ext4_ext_path *curp,
744 int logical, ext4_fsblk_t ptr) 696 int logical, ext4_fsblk_t ptr)
745{ 697{
746 struct ext4_extent_idx *ix; 698 struct ext4_extent_idx *ix;
747 int len, err; 699 int len, err;
@@ -917,7 +869,7 @@ static int ext4_ext_split(handle_t *handle, struct inode *inode,
917 EXT_MAX_EXTENT(path[depth].p_hdr)) { 869 EXT_MAX_EXTENT(path[depth].p_hdr)) {
918 ext_debug("move %d:%llu:[%d]%d in new leaf %llu\n", 870 ext_debug("move %d:%llu:[%d]%d in new leaf %llu\n",
919 le32_to_cpu(path[depth].p_ext->ee_block), 871 le32_to_cpu(path[depth].p_ext->ee_block),
920 ext_pblock(path[depth].p_ext), 872 ext4_ext_pblock(path[depth].p_ext),
921 ext4_ext_is_uninitialized(path[depth].p_ext), 873 ext4_ext_is_uninitialized(path[depth].p_ext),
922 ext4_ext_get_actual_len(path[depth].p_ext), 874 ext4_ext_get_actual_len(path[depth].p_ext),
923 newblock); 875 newblock);
@@ -1007,7 +959,7 @@ static int ext4_ext_split(handle_t *handle, struct inode *inode,
1007 while (path[i].p_idx <= EXT_MAX_INDEX(path[i].p_hdr)) { 959 while (path[i].p_idx <= EXT_MAX_INDEX(path[i].p_hdr)) {
1008 ext_debug("%d: move %d:%llu in new index %llu\n", i, 960 ext_debug("%d: move %d:%llu in new index %llu\n", i,
1009 le32_to_cpu(path[i].p_idx->ei_block), 961 le32_to_cpu(path[i].p_idx->ei_block),
1010 idx_pblock(path[i].p_idx), 962 ext4_idx_pblock(path[i].p_idx),
1011 newblock); 963 newblock);
1012 /*memmove(++fidx, path[i].p_idx++, 964 /*memmove(++fidx, path[i].p_idx++,
1013 sizeof(struct ext4_extent_idx)); 965 sizeof(struct ext4_extent_idx));
@@ -1083,7 +1035,6 @@ static int ext4_ext_grow_indepth(handle_t *handle, struct inode *inode,
1083{ 1035{
1084 struct ext4_ext_path *curp = path; 1036 struct ext4_ext_path *curp = path;
1085 struct ext4_extent_header *neh; 1037 struct ext4_extent_header *neh;
1086 struct ext4_extent_idx *fidx;
1087 struct buffer_head *bh; 1038 struct buffer_head *bh;
1088 ext4_fsblk_t newblock; 1039 ext4_fsblk_t newblock;
1089 int err = 0; 1040 int err = 0;
@@ -1144,10 +1095,10 @@ static int ext4_ext_grow_indepth(handle_t *handle, struct inode *inode,
1144 ext4_idx_store_pblock(curp->p_idx, newblock); 1095 ext4_idx_store_pblock(curp->p_idx, newblock);
1145 1096
1146 neh = ext_inode_hdr(inode); 1097 neh = ext_inode_hdr(inode);
1147 fidx = EXT_FIRST_INDEX(neh);
1148 ext_debug("new root: num %d(%d), lblock %d, ptr %llu\n", 1098 ext_debug("new root: num %d(%d), lblock %d, ptr %llu\n",
1149 le16_to_cpu(neh->eh_entries), le16_to_cpu(neh->eh_max), 1099 le16_to_cpu(neh->eh_entries), le16_to_cpu(neh->eh_max),
1150 le32_to_cpu(fidx->ei_block), idx_pblock(fidx)); 1100 le32_to_cpu(EXT_FIRST_INDEX(neh)->ei_block),
1101 ext4_idx_pblock(EXT_FIRST_INDEX(neh)));
1151 1102
1152 neh->eh_depth = cpu_to_le16(path->p_depth + 1); 1103 neh->eh_depth = cpu_to_le16(path->p_depth + 1);
1153 err = ext4_ext_dirty(handle, inode, curp); 1104 err = ext4_ext_dirty(handle, inode, curp);
@@ -1233,9 +1184,9 @@ out:
1233 * returns 0 at @phys 1184 * returns 0 at @phys
1234 * return value contains 0 (success) or error code 1185 * return value contains 0 (success) or error code
1235 */ 1186 */
1236int 1187static int ext4_ext_search_left(struct inode *inode,
1237ext4_ext_search_left(struct inode *inode, struct ext4_ext_path *path, 1188 struct ext4_ext_path *path,
1238 ext4_lblk_t *logical, ext4_fsblk_t *phys) 1189 ext4_lblk_t *logical, ext4_fsblk_t *phys)
1239{ 1190{
1240 struct ext4_extent_idx *ix; 1191 struct ext4_extent_idx *ix;
1241 struct ext4_extent *ex; 1192 struct ext4_extent *ex;
@@ -1287,7 +1238,7 @@ ext4_ext_search_left(struct inode *inode, struct ext4_ext_path *path,
1287 } 1238 }
1288 1239
1289 *logical = le32_to_cpu(ex->ee_block) + ee_len - 1; 1240 *logical = le32_to_cpu(ex->ee_block) + ee_len - 1;
1290 *phys = ext_pblock(ex) + ee_len - 1; 1241 *phys = ext4_ext_pblock(ex) + ee_len - 1;
1291 return 0; 1242 return 0;
1292} 1243}
1293 1244
@@ -1298,9 +1249,9 @@ ext4_ext_search_left(struct inode *inode, struct ext4_ext_path *path,
1298 * returns 0 at @phys 1249 * returns 0 at @phys
1299 * return value contains 0 (success) or error code 1250 * return value contains 0 (success) or error code
1300 */ 1251 */
1301int 1252static int ext4_ext_search_right(struct inode *inode,
1302ext4_ext_search_right(struct inode *inode, struct ext4_ext_path *path, 1253 struct ext4_ext_path *path,
1303 ext4_lblk_t *logical, ext4_fsblk_t *phys) 1254 ext4_lblk_t *logical, ext4_fsblk_t *phys)
1304{ 1255{
1305 struct buffer_head *bh = NULL; 1256 struct buffer_head *bh = NULL;
1306 struct ext4_extent_header *eh; 1257 struct ext4_extent_header *eh;
@@ -1343,7 +1294,7 @@ ext4_ext_search_right(struct inode *inode, struct ext4_ext_path *path,
1343 } 1294 }
1344 } 1295 }
1345 *logical = le32_to_cpu(ex->ee_block); 1296 *logical = le32_to_cpu(ex->ee_block);
1346 *phys = ext_pblock(ex); 1297 *phys = ext4_ext_pblock(ex);
1347 return 0; 1298 return 0;
1348 } 1299 }
1349 1300
@@ -1358,7 +1309,7 @@ ext4_ext_search_right(struct inode *inode, struct ext4_ext_path *path,
1358 /* next allocated block in this leaf */ 1309 /* next allocated block in this leaf */
1359 ex++; 1310 ex++;
1360 *logical = le32_to_cpu(ex->ee_block); 1311 *logical = le32_to_cpu(ex->ee_block);
1361 *phys = ext_pblock(ex); 1312 *phys = ext4_ext_pblock(ex);
1362 return 0; 1313 return 0;
1363 } 1314 }
1364 1315
@@ -1377,7 +1328,7 @@ got_index:
1377 * follow it and find the closest allocated 1328 * follow it and find the closest allocated
1378 * block to the right */ 1329 * block to the right */
1379 ix++; 1330 ix++;
1380 block = idx_pblock(ix); 1331 block = ext4_idx_pblock(ix);
1381 while (++depth < path->p_depth) { 1332 while (++depth < path->p_depth) {
1382 bh = sb_bread(inode->i_sb, block); 1333 bh = sb_bread(inode->i_sb, block);
1383 if (bh == NULL) 1334 if (bh == NULL)
@@ -1389,7 +1340,7 @@ got_index:
1389 return -EIO; 1340 return -EIO;
1390 } 1341 }
1391 ix = EXT_FIRST_INDEX(eh); 1342 ix = EXT_FIRST_INDEX(eh);
1392 block = idx_pblock(ix); 1343 block = ext4_idx_pblock(ix);
1393 put_bh(bh); 1344 put_bh(bh);
1394 } 1345 }
1395 1346
@@ -1403,7 +1354,7 @@ got_index:
1403 } 1354 }
1404 ex = EXT_FIRST_EXTENT(eh); 1355 ex = EXT_FIRST_EXTENT(eh);
1405 *logical = le32_to_cpu(ex->ee_block); 1356 *logical = le32_to_cpu(ex->ee_block);
1406 *phys = ext_pblock(ex); 1357 *phys = ext4_ext_pblock(ex);
1407 put_bh(bh); 1358 put_bh(bh);
1408 return 0; 1359 return 0;
1409} 1360}
@@ -1574,7 +1525,7 @@ ext4_can_extents_be_merged(struct inode *inode, struct ext4_extent *ex1,
1574 return 0; 1525 return 0;
1575#endif 1526#endif
1576 1527
1577 if (ext_pblock(ex1) + ext1_ee_len == ext_pblock(ex2)) 1528 if (ext4_ext_pblock(ex1) + ext1_ee_len == ext4_ext_pblock(ex2))
1578 return 1; 1529 return 1;
1579 return 0; 1530 return 0;
1580} 1531}
@@ -1586,9 +1537,9 @@ ext4_can_extents_be_merged(struct inode *inode, struct ext4_extent *ex1,
1586 * Returns 0 if the extents (ex and ex+1) were _not_ merged and returns 1537 * Returns 0 if the extents (ex and ex+1) were _not_ merged and returns
1587 * 1 if they got merged. 1538 * 1 if they got merged.
1588 */ 1539 */
1589int ext4_ext_try_to_merge(struct inode *inode, 1540static int ext4_ext_try_to_merge(struct inode *inode,
1590 struct ext4_ext_path *path, 1541 struct ext4_ext_path *path,
1591 struct ext4_extent *ex) 1542 struct ext4_extent *ex)
1592{ 1543{
1593 struct ext4_extent_header *eh; 1544 struct ext4_extent_header *eh;
1594 unsigned int depth, len; 1545 unsigned int depth, len;
@@ -1633,9 +1584,9 @@ int ext4_ext_try_to_merge(struct inode *inode,
1633 * such that there will be no overlap, and then returns 1. 1584 * such that there will be no overlap, and then returns 1.
1634 * If there is no overlap found, it returns 0. 1585 * If there is no overlap found, it returns 0.
1635 */ 1586 */
1636unsigned int ext4_ext_check_overlap(struct inode *inode, 1587static unsigned int ext4_ext_check_overlap(struct inode *inode,
1637 struct ext4_extent *newext, 1588 struct ext4_extent *newext,
1638 struct ext4_ext_path *path) 1589 struct ext4_ext_path *path)
1639{ 1590{
1640 ext4_lblk_t b1, b2; 1591 ext4_lblk_t b1, b2;
1641 unsigned int depth, len1; 1592 unsigned int depth, len1;
@@ -1707,11 +1658,12 @@ int ext4_ext_insert_extent(handle_t *handle, struct inode *inode,
1707 if (ex && !(flag & EXT4_GET_BLOCKS_PRE_IO) 1658 if (ex && !(flag & EXT4_GET_BLOCKS_PRE_IO)
1708 && ext4_can_extents_be_merged(inode, ex, newext)) { 1659 && ext4_can_extents_be_merged(inode, ex, newext)) {
1709 ext_debug("append [%d]%d block to %d:[%d]%d (from %llu)\n", 1660 ext_debug("append [%d]%d block to %d:[%d]%d (from %llu)\n",
1710 ext4_ext_is_uninitialized(newext), 1661 ext4_ext_is_uninitialized(newext),
1711 ext4_ext_get_actual_len(newext), 1662 ext4_ext_get_actual_len(newext),
1712 le32_to_cpu(ex->ee_block), 1663 le32_to_cpu(ex->ee_block),
1713 ext4_ext_is_uninitialized(ex), 1664 ext4_ext_is_uninitialized(ex),
1714 ext4_ext_get_actual_len(ex), ext_pblock(ex)); 1665 ext4_ext_get_actual_len(ex),
1666 ext4_ext_pblock(ex));
1715 err = ext4_ext_get_access(handle, inode, path + depth); 1667 err = ext4_ext_get_access(handle, inode, path + depth);
1716 if (err) 1668 if (err)
1717 return err; 1669 return err;
@@ -1781,7 +1733,7 @@ has_space:
1781 /* there is no extent in this leaf, create first one */ 1733 /* there is no extent in this leaf, create first one */
1782 ext_debug("first extent in the leaf: %d:%llu:[%d]%d\n", 1734 ext_debug("first extent in the leaf: %d:%llu:[%d]%d\n",
1783 le32_to_cpu(newext->ee_block), 1735 le32_to_cpu(newext->ee_block),
1784 ext_pblock(newext), 1736 ext4_ext_pblock(newext),
1785 ext4_ext_is_uninitialized(newext), 1737 ext4_ext_is_uninitialized(newext),
1786 ext4_ext_get_actual_len(newext)); 1738 ext4_ext_get_actual_len(newext));
1787 path[depth].p_ext = EXT_FIRST_EXTENT(eh); 1739 path[depth].p_ext = EXT_FIRST_EXTENT(eh);
@@ -1795,7 +1747,7 @@ has_space:
1795 ext_debug("insert %d:%llu:[%d]%d after: nearest 0x%p, " 1747 ext_debug("insert %d:%llu:[%d]%d after: nearest 0x%p, "
1796 "move %d from 0x%p to 0x%p\n", 1748 "move %d from 0x%p to 0x%p\n",
1797 le32_to_cpu(newext->ee_block), 1749 le32_to_cpu(newext->ee_block),
1798 ext_pblock(newext), 1750 ext4_ext_pblock(newext),
1799 ext4_ext_is_uninitialized(newext), 1751 ext4_ext_is_uninitialized(newext),
1800 ext4_ext_get_actual_len(newext), 1752 ext4_ext_get_actual_len(newext),
1801 nearex, len, nearex + 1, nearex + 2); 1753 nearex, len, nearex + 1, nearex + 2);
@@ -1809,7 +1761,7 @@ has_space:
1809 ext_debug("insert %d:%llu:[%d]%d before: nearest 0x%p, " 1761 ext_debug("insert %d:%llu:[%d]%d before: nearest 0x%p, "
1810 "move %d from 0x%p to 0x%p\n", 1762 "move %d from 0x%p to 0x%p\n",
1811 le32_to_cpu(newext->ee_block), 1763 le32_to_cpu(newext->ee_block),
1812 ext_pblock(newext), 1764 ext4_ext_pblock(newext),
1813 ext4_ext_is_uninitialized(newext), 1765 ext4_ext_is_uninitialized(newext),
1814 ext4_ext_get_actual_len(newext), 1766 ext4_ext_get_actual_len(newext),
1815 nearex, len, nearex + 1, nearex + 2); 1767 nearex, len, nearex + 1, nearex + 2);
@@ -1820,7 +1772,7 @@ has_space:
1820 le16_add_cpu(&eh->eh_entries, 1); 1772 le16_add_cpu(&eh->eh_entries, 1);
1821 nearex = path[depth].p_ext; 1773 nearex = path[depth].p_ext;
1822 nearex->ee_block = newext->ee_block; 1774 nearex->ee_block = newext->ee_block;
1823 ext4_ext_store_pblock(nearex, ext_pblock(newext)); 1775 ext4_ext_store_pblock(nearex, ext4_ext_pblock(newext));
1824 nearex->ee_len = newext->ee_len; 1776 nearex->ee_len = newext->ee_len;
1825 1777
1826merge: 1778merge:
@@ -1846,9 +1798,9 @@ cleanup:
1846 return err; 1798 return err;
1847} 1799}
1848 1800
1849int ext4_ext_walk_space(struct inode *inode, ext4_lblk_t block, 1801static int ext4_ext_walk_space(struct inode *inode, ext4_lblk_t block,
1850 ext4_lblk_t num, ext_prepare_callback func, 1802 ext4_lblk_t num, ext_prepare_callback func,
1851 void *cbdata) 1803 void *cbdata)
1852{ 1804{
1853 struct ext4_ext_path *path = NULL; 1805 struct ext4_ext_path *path = NULL;
1854 struct ext4_ext_cache cbex; 1806 struct ext4_ext_cache cbex;
@@ -1924,7 +1876,7 @@ int ext4_ext_walk_space(struct inode *inode, ext4_lblk_t block,
1924 } else { 1876 } else {
1925 cbex.ec_block = le32_to_cpu(ex->ee_block); 1877 cbex.ec_block = le32_to_cpu(ex->ee_block);
1926 cbex.ec_len = ext4_ext_get_actual_len(ex); 1878 cbex.ec_len = ext4_ext_get_actual_len(ex);
1927 cbex.ec_start = ext_pblock(ex); 1879 cbex.ec_start = ext4_ext_pblock(ex);
1928 cbex.ec_type = EXT4_EXT_CACHE_EXTENT; 1880 cbex.ec_type = EXT4_EXT_CACHE_EXTENT;
1929 } 1881 }
1930 1882
@@ -2074,7 +2026,7 @@ static int ext4_ext_rm_idx(handle_t *handle, struct inode *inode,
2074 2026
2075 /* free index block */ 2027 /* free index block */
2076 path--; 2028 path--;
2077 leaf = idx_pblock(path->p_idx); 2029 leaf = ext4_idx_pblock(path->p_idx);
2078 if (unlikely(path->p_hdr->eh_entries == 0)) { 2030 if (unlikely(path->p_hdr->eh_entries == 0)) {
2079 EXT4_ERROR_INODE(inode, "path->p_hdr->eh_entries == 0"); 2031 EXT4_ERROR_INODE(inode, "path->p_hdr->eh_entries == 0");
2080 return -EIO; 2032 return -EIO;
@@ -2182,7 +2134,7 @@ static int ext4_remove_blocks(handle_t *handle, struct inode *inode,
2182 ext4_fsblk_t start; 2134 ext4_fsblk_t start;
2183 2135
2184 num = le32_to_cpu(ex->ee_block) + ee_len - from; 2136 num = le32_to_cpu(ex->ee_block) + ee_len - from;
2185 start = ext_pblock(ex) + ee_len - num; 2137 start = ext4_ext_pblock(ex) + ee_len - num;
2186 ext_debug("free last %u blocks starting %llu\n", num, start); 2138 ext_debug("free last %u blocks starting %llu\n", num, start);
2187 ext4_free_blocks(handle, inode, 0, start, num, flags); 2139 ext4_free_blocks(handle, inode, 0, start, num, flags);
2188 } else if (from == le32_to_cpu(ex->ee_block) 2140 } else if (from == le32_to_cpu(ex->ee_block)
@@ -2311,7 +2263,7 @@ ext4_ext_rm_leaf(handle_t *handle, struct inode *inode,
2311 goto out; 2263 goto out;
2312 2264
2313 ext_debug("new extent: %u:%u:%llu\n", block, num, 2265 ext_debug("new extent: %u:%u:%llu\n", block, num,
2314 ext_pblock(ex)); 2266 ext4_ext_pblock(ex));
2315 ex--; 2267 ex--;
2316 ex_ee_block = le32_to_cpu(ex->ee_block); 2268 ex_ee_block = le32_to_cpu(ex->ee_block);
2317 ex_ee_len = ext4_ext_get_actual_len(ex); 2269 ex_ee_len = ext4_ext_get_actual_len(ex);
@@ -2422,9 +2374,9 @@ again:
2422 struct buffer_head *bh; 2374 struct buffer_head *bh;
2423 /* go to the next level */ 2375 /* go to the next level */
2424 ext_debug("move to level %d (block %llu)\n", 2376 ext_debug("move to level %d (block %llu)\n",
2425 i + 1, idx_pblock(path[i].p_idx)); 2377 i + 1, ext4_idx_pblock(path[i].p_idx));
2426 memset(path + i + 1, 0, sizeof(*path)); 2378 memset(path + i + 1, 0, sizeof(*path));
2427 bh = sb_bread(sb, idx_pblock(path[i].p_idx)); 2379 bh = sb_bread(sb, ext4_idx_pblock(path[i].p_idx));
2428 if (!bh) { 2380 if (!bh) {
2429 /* should we reset i_size? */ 2381 /* should we reset i_size? */
2430 err = -EIO; 2382 err = -EIO;
@@ -2536,77 +2488,21 @@ void ext4_ext_release(struct super_block *sb)
2536#endif 2488#endif
2537} 2489}
2538 2490
2539static void bi_complete(struct bio *bio, int error)
2540{
2541 complete((struct completion *)bio->bi_private);
2542}
2543
2544/* FIXME!! we need to try to merge to left or right after zero-out */ 2491/* FIXME!! we need to try to merge to left or right after zero-out */
2545static int ext4_ext_zeroout(struct inode *inode, struct ext4_extent *ex) 2492static int ext4_ext_zeroout(struct inode *inode, struct ext4_extent *ex)
2546{ 2493{
2494 ext4_fsblk_t ee_pblock;
2495 unsigned int ee_len;
2547 int ret; 2496 int ret;
2548 struct bio *bio;
2549 int blkbits, blocksize;
2550 sector_t ee_pblock;
2551 struct completion event;
2552 unsigned int ee_len, len, done, offset;
2553 2497
2554
2555 blkbits = inode->i_blkbits;
2556 blocksize = inode->i_sb->s_blocksize;
2557 ee_len = ext4_ext_get_actual_len(ex); 2498 ee_len = ext4_ext_get_actual_len(ex);
2558 ee_pblock = ext_pblock(ex); 2499 ee_pblock = ext4_ext_pblock(ex);
2559
2560 /* convert ee_pblock to 512 byte sectors */
2561 ee_pblock = ee_pblock << (blkbits - 9);
2562
2563 while (ee_len > 0) {
2564
2565 if (ee_len > BIO_MAX_PAGES)
2566 len = BIO_MAX_PAGES;
2567 else
2568 len = ee_len;
2569
2570 bio = bio_alloc(GFP_NOIO, len);
2571 if (!bio)
2572 return -ENOMEM;
2573
2574 bio->bi_sector = ee_pblock;
2575 bio->bi_bdev = inode->i_sb->s_bdev;
2576
2577 done = 0;
2578 offset = 0;
2579 while (done < len) {
2580 ret = bio_add_page(bio, ZERO_PAGE(0),
2581 blocksize, offset);
2582 if (ret != blocksize) {
2583 /*
2584 * We can't add any more pages because of
2585 * hardware limitations. Start a new bio.
2586 */
2587 break;
2588 }
2589 done++;
2590 offset += blocksize;
2591 if (offset >= PAGE_CACHE_SIZE)
2592 offset = 0;
2593 }
2594 2500
2595 init_completion(&event); 2501 ret = sb_issue_zeroout(inode->i_sb, ee_pblock, ee_len, GFP_NOFS);
2596 bio->bi_private = &event; 2502 if (ret > 0)
2597 bio->bi_end_io = bi_complete; 2503 ret = 0;
2598 submit_bio(WRITE, bio);
2599 wait_for_completion(&event);
2600 2504
2601 if (!test_bit(BIO_UPTODATE, &bio->bi_flags)) { 2505 return ret;
2602 bio_put(bio);
2603 return -EIO;
2604 }
2605 bio_put(bio);
2606 ee_len -= done;
2607 ee_pblock += done << (blkbits - 9);
2608 }
2609 return 0;
2610} 2506}
2611 2507
2612#define EXT4_EXT_ZERO_LEN 7 2508#define EXT4_EXT_ZERO_LEN 7
@@ -2652,12 +2548,12 @@ static int ext4_ext_convert_to_initialized(handle_t *handle,
2652 ee_block = le32_to_cpu(ex->ee_block); 2548 ee_block = le32_to_cpu(ex->ee_block);
2653 ee_len = ext4_ext_get_actual_len(ex); 2549 ee_len = ext4_ext_get_actual_len(ex);
2654 allocated = ee_len - (map->m_lblk - ee_block); 2550 allocated = ee_len - (map->m_lblk - ee_block);
2655 newblock = map->m_lblk - ee_block + ext_pblock(ex); 2551 newblock = map->m_lblk - ee_block + ext4_ext_pblock(ex);
2656 2552
2657 ex2 = ex; 2553 ex2 = ex;
2658 orig_ex.ee_block = ex->ee_block; 2554 orig_ex.ee_block = ex->ee_block;
2659 orig_ex.ee_len = cpu_to_le16(ee_len); 2555 orig_ex.ee_len = cpu_to_le16(ee_len);
2660 ext4_ext_store_pblock(&orig_ex, ext_pblock(ex)); 2556 ext4_ext_store_pblock(&orig_ex, ext4_ext_pblock(ex));
2661 2557
2662 /* 2558 /*
2663 * It is safe to convert extent to initialized via explicit 2559 * It is safe to convert extent to initialized via explicit
@@ -2676,7 +2572,7 @@ static int ext4_ext_convert_to_initialized(handle_t *handle,
2676 /* update the extent length and mark as initialized */ 2572 /* update the extent length and mark as initialized */
2677 ex->ee_block = orig_ex.ee_block; 2573 ex->ee_block = orig_ex.ee_block;
2678 ex->ee_len = orig_ex.ee_len; 2574 ex->ee_len = orig_ex.ee_len;
2679 ext4_ext_store_pblock(ex, ext_pblock(&orig_ex)); 2575 ext4_ext_store_pblock(ex, ext4_ext_pblock(&orig_ex));
2680 ext4_ext_dirty(handle, inode, path + depth); 2576 ext4_ext_dirty(handle, inode, path + depth);
2681 /* zeroed the full extent */ 2577 /* zeroed the full extent */
2682 return allocated; 2578 return allocated;
@@ -2711,7 +2607,7 @@ static int ext4_ext_convert_to_initialized(handle_t *handle,
2711 ex->ee_block = orig_ex.ee_block; 2607 ex->ee_block = orig_ex.ee_block;
2712 ex->ee_len = cpu_to_le16(ee_len - allocated); 2608 ex->ee_len = cpu_to_le16(ee_len - allocated);
2713 ext4_ext_mark_uninitialized(ex); 2609 ext4_ext_mark_uninitialized(ex);
2714 ext4_ext_store_pblock(ex, ext_pblock(&orig_ex)); 2610 ext4_ext_store_pblock(ex, ext4_ext_pblock(&orig_ex));
2715 ext4_ext_dirty(handle, inode, path + depth); 2611 ext4_ext_dirty(handle, inode, path + depth);
2716 2612
2717 ex3 = &newex; 2613 ex3 = &newex;
@@ -2726,7 +2622,8 @@ static int ext4_ext_convert_to_initialized(handle_t *handle,
2726 goto fix_extent_len; 2622 goto fix_extent_len;
2727 ex->ee_block = orig_ex.ee_block; 2623 ex->ee_block = orig_ex.ee_block;
2728 ex->ee_len = orig_ex.ee_len; 2624 ex->ee_len = orig_ex.ee_len;
2729 ext4_ext_store_pblock(ex, ext_pblock(&orig_ex)); 2625 ext4_ext_store_pblock(ex,
2626 ext4_ext_pblock(&orig_ex));
2730 ext4_ext_dirty(handle, inode, path + depth); 2627 ext4_ext_dirty(handle, inode, path + depth);
2731 /* blocks available from map->m_lblk */ 2628 /* blocks available from map->m_lblk */
2732 return allocated; 2629 return allocated;
@@ -2783,7 +2680,7 @@ static int ext4_ext_convert_to_initialized(handle_t *handle,
2783 /* update the extent length and mark as initialized */ 2680 /* update the extent length and mark as initialized */
2784 ex->ee_block = orig_ex.ee_block; 2681 ex->ee_block = orig_ex.ee_block;
2785 ex->ee_len = orig_ex.ee_len; 2682 ex->ee_len = orig_ex.ee_len;
2786 ext4_ext_store_pblock(ex, ext_pblock(&orig_ex)); 2683 ext4_ext_store_pblock(ex, ext4_ext_pblock(&orig_ex));
2787 ext4_ext_dirty(handle, inode, path + depth); 2684 ext4_ext_dirty(handle, inode, path + depth);
2788 /* zeroed the full extent */ 2685 /* zeroed the full extent */
2789 /* blocks available from map->m_lblk */ 2686 /* blocks available from map->m_lblk */
@@ -2834,7 +2731,7 @@ static int ext4_ext_convert_to_initialized(handle_t *handle,
2834 /* update the extent length and mark as initialized */ 2731 /* update the extent length and mark as initialized */
2835 ex->ee_block = orig_ex.ee_block; 2732 ex->ee_block = orig_ex.ee_block;
2836 ex->ee_len = orig_ex.ee_len; 2733 ex->ee_len = orig_ex.ee_len;
2837 ext4_ext_store_pblock(ex, ext_pblock(&orig_ex)); 2734 ext4_ext_store_pblock(ex, ext4_ext_pblock(&orig_ex));
2838 ext4_ext_dirty(handle, inode, path + depth); 2735 ext4_ext_dirty(handle, inode, path + depth);
2839 /* zero out the first half */ 2736 /* zero out the first half */
2840 /* blocks available from map->m_lblk */ 2737 /* blocks available from map->m_lblk */
@@ -2903,7 +2800,7 @@ insert:
2903 /* update the extent length and mark as initialized */ 2800 /* update the extent length and mark as initialized */
2904 ex->ee_block = orig_ex.ee_block; 2801 ex->ee_block = orig_ex.ee_block;
2905 ex->ee_len = orig_ex.ee_len; 2802 ex->ee_len = orig_ex.ee_len;
2906 ext4_ext_store_pblock(ex, ext_pblock(&orig_ex)); 2803 ext4_ext_store_pblock(ex, ext4_ext_pblock(&orig_ex));
2907 ext4_ext_dirty(handle, inode, path + depth); 2804 ext4_ext_dirty(handle, inode, path + depth);
2908 /* zero out the first half */ 2805 /* zero out the first half */
2909 return allocated; 2806 return allocated;
@@ -2916,7 +2813,7 @@ out:
2916fix_extent_len: 2813fix_extent_len:
2917 ex->ee_block = orig_ex.ee_block; 2814 ex->ee_block = orig_ex.ee_block;
2918 ex->ee_len = orig_ex.ee_len; 2815 ex->ee_len = orig_ex.ee_len;
2919 ext4_ext_store_pblock(ex, ext_pblock(&orig_ex)); 2816 ext4_ext_store_pblock(ex, ext4_ext_pblock(&orig_ex));
2920 ext4_ext_mark_uninitialized(ex); 2817 ext4_ext_mark_uninitialized(ex);
2921 ext4_ext_dirty(handle, inode, path + depth); 2818 ext4_ext_dirty(handle, inode, path + depth);
2922 return err; 2819 return err;
@@ -2937,7 +2834,7 @@ fix_extent_len:
2937 * One of more index blocks maybe needed if the extent tree grow after 2834 * One of more index blocks maybe needed if the extent tree grow after
2938 * the unintialized extent split. To prevent ENOSPC occur at the IO 2835 * the unintialized extent split. To prevent ENOSPC occur at the IO
2939 * complete, we need to split the uninitialized extent before DIO submit 2836 * complete, we need to split the uninitialized extent before DIO submit
2940 * the IO. The uninitilized extent called at this time will be split 2837 * the IO. The uninitialized extent called at this time will be split
2941 * into three uninitialized extent(at most). After IO complete, the part 2838 * into three uninitialized extent(at most). After IO complete, the part
2942 * being filled will be convert to initialized by the end_io callback function 2839 * being filled will be convert to initialized by the end_io callback function
2943 * via ext4_convert_unwritten_extents(). 2840 * via ext4_convert_unwritten_extents().
@@ -2954,7 +2851,6 @@ static int ext4_split_unwritten_extents(handle_t *handle,
2954 struct ext4_extent *ex1 = NULL; 2851 struct ext4_extent *ex1 = NULL;
2955 struct ext4_extent *ex2 = NULL; 2852 struct ext4_extent *ex2 = NULL;
2956 struct ext4_extent *ex3 = NULL; 2853 struct ext4_extent *ex3 = NULL;
2957 struct ext4_extent_header *eh;
2958 ext4_lblk_t ee_block, eof_block; 2854 ext4_lblk_t ee_block, eof_block;
2959 unsigned int allocated, ee_len, depth; 2855 unsigned int allocated, ee_len, depth;
2960 ext4_fsblk_t newblock; 2856 ext4_fsblk_t newblock;
@@ -2971,17 +2867,16 @@ static int ext4_split_unwritten_extents(handle_t *handle,
2971 eof_block = map->m_lblk + map->m_len; 2867 eof_block = map->m_lblk + map->m_len;
2972 2868
2973 depth = ext_depth(inode); 2869 depth = ext_depth(inode);
2974 eh = path[depth].p_hdr;
2975 ex = path[depth].p_ext; 2870 ex = path[depth].p_ext;
2976 ee_block = le32_to_cpu(ex->ee_block); 2871 ee_block = le32_to_cpu(ex->ee_block);
2977 ee_len = ext4_ext_get_actual_len(ex); 2872 ee_len = ext4_ext_get_actual_len(ex);
2978 allocated = ee_len - (map->m_lblk - ee_block); 2873 allocated = ee_len - (map->m_lblk - ee_block);
2979 newblock = map->m_lblk - ee_block + ext_pblock(ex); 2874 newblock = map->m_lblk - ee_block + ext4_ext_pblock(ex);
2980 2875
2981 ex2 = ex; 2876 ex2 = ex;
2982 orig_ex.ee_block = ex->ee_block; 2877 orig_ex.ee_block = ex->ee_block;
2983 orig_ex.ee_len = cpu_to_le16(ee_len); 2878 orig_ex.ee_len = cpu_to_le16(ee_len);
2984 ext4_ext_store_pblock(&orig_ex, ext_pblock(ex)); 2879 ext4_ext_store_pblock(&orig_ex, ext4_ext_pblock(ex));
2985 2880
2986 /* 2881 /*
2987 * It is safe to convert extent to initialized via explicit 2882 * It is safe to convert extent to initialized via explicit
@@ -3030,7 +2925,7 @@ static int ext4_split_unwritten_extents(handle_t *handle,
3030 /* update the extent length and mark as initialized */ 2925 /* update the extent length and mark as initialized */
3031 ex->ee_block = orig_ex.ee_block; 2926 ex->ee_block = orig_ex.ee_block;
3032 ex->ee_len = orig_ex.ee_len; 2927 ex->ee_len = orig_ex.ee_len;
3033 ext4_ext_store_pblock(ex, ext_pblock(&orig_ex)); 2928 ext4_ext_store_pblock(ex, ext4_ext_pblock(&orig_ex));
3034 ext4_ext_dirty(handle, inode, path + depth); 2929 ext4_ext_dirty(handle, inode, path + depth);
3035 /* zeroed the full extent */ 2930 /* zeroed the full extent */
3036 /* blocks available from map->m_lblk */ 2931 /* blocks available from map->m_lblk */
@@ -3058,7 +2953,6 @@ static int ext4_split_unwritten_extents(handle_t *handle,
3058 err = PTR_ERR(path); 2953 err = PTR_ERR(path);
3059 goto out; 2954 goto out;
3060 } 2955 }
3061 eh = path[depth].p_hdr;
3062 ex = path[depth].p_ext; 2956 ex = path[depth].p_ext;
3063 if (ex2 != &newex) 2957 if (ex2 != &newex)
3064 ex2 = ex; 2958 ex2 = ex;
@@ -3103,7 +2997,7 @@ insert:
3103 /* update the extent length and mark as initialized */ 2997 /* update the extent length and mark as initialized */
3104 ex->ee_block = orig_ex.ee_block; 2998 ex->ee_block = orig_ex.ee_block;
3105 ex->ee_len = orig_ex.ee_len; 2999 ex->ee_len = orig_ex.ee_len;
3106 ext4_ext_store_pblock(ex, ext_pblock(&orig_ex)); 3000 ext4_ext_store_pblock(ex, ext4_ext_pblock(&orig_ex));
3107 ext4_ext_dirty(handle, inode, path + depth); 3001 ext4_ext_dirty(handle, inode, path + depth);
3108 /* zero out the first half */ 3002 /* zero out the first half */
3109 return allocated; 3003 return allocated;
@@ -3116,7 +3010,7 @@ out:
3116fix_extent_len: 3010fix_extent_len:
3117 ex->ee_block = orig_ex.ee_block; 3011 ex->ee_block = orig_ex.ee_block;
3118 ex->ee_len = orig_ex.ee_len; 3012 ex->ee_len = orig_ex.ee_len;
3119 ext4_ext_store_pblock(ex, ext_pblock(&orig_ex)); 3013 ext4_ext_store_pblock(ex, ext4_ext_pblock(&orig_ex));
3120 ext4_ext_mark_uninitialized(ex); 3014 ext4_ext_mark_uninitialized(ex);
3121 ext4_ext_dirty(handle, inode, path + depth); 3015 ext4_ext_dirty(handle, inode, path + depth);
3122 return err; 3016 return err;
@@ -3184,6 +3078,57 @@ static void unmap_underlying_metadata_blocks(struct block_device *bdev,
3184 unmap_underlying_metadata(bdev, block + i); 3078 unmap_underlying_metadata(bdev, block + i);
3185} 3079}
3186 3080
3081/*
3082 * Handle EOFBLOCKS_FL flag, clearing it if necessary
3083 */
3084static int check_eofblocks_fl(handle_t *handle, struct inode *inode,
3085 struct ext4_map_blocks *map,
3086 struct ext4_ext_path *path,
3087 unsigned int len)
3088{
3089 int i, depth;
3090 struct ext4_extent_header *eh;
3091 struct ext4_extent *ex, *last_ex;
3092
3093 if (!ext4_test_inode_flag(inode, EXT4_INODE_EOFBLOCKS))
3094 return 0;
3095
3096 depth = ext_depth(inode);
3097 eh = path[depth].p_hdr;
3098 ex = path[depth].p_ext;
3099
3100 if (unlikely(!eh->eh_entries)) {
3101 EXT4_ERROR_INODE(inode, "eh->eh_entries == 0 and "
3102 "EOFBLOCKS_FL set");
3103 return -EIO;
3104 }
3105 last_ex = EXT_LAST_EXTENT(eh);
3106 /*
3107 * We should clear the EOFBLOCKS_FL flag if we are writing the
3108 * last block in the last extent in the file. We test this by
3109 * first checking to see if the caller to
3110 * ext4_ext_get_blocks() was interested in the last block (or
3111 * a block beyond the last block) in the current extent. If
3112 * this turns out to be false, we can bail out from this
3113 * function immediately.
3114 */
3115 if (map->m_lblk + len < le32_to_cpu(last_ex->ee_block) +
3116 ext4_ext_get_actual_len(last_ex))
3117 return 0;
3118 /*
3119 * If the caller does appear to be planning to write at or
3120 * beyond the end of the current extent, we then test to see
3121 * if the current extent is the last extent in the file, by
3122 * checking to make sure it was reached via the rightmost node
3123 * at each level of the tree.
3124 */
3125 for (i = depth-1; i >= 0; i--)
3126 if (path[i].p_idx != EXT_LAST_INDEX(path[i].p_hdr))
3127 return 0;
3128 ext4_clear_inode_flag(inode, EXT4_INODE_EOFBLOCKS);
3129 return ext4_mark_inode_dirty(handle, inode);
3130}
3131
3187static int 3132static int
3188ext4_ext_handle_uninitialized_extents(handle_t *handle, struct inode *inode, 3133ext4_ext_handle_uninitialized_extents(handle_t *handle, struct inode *inode,
3189 struct ext4_map_blocks *map, 3134 struct ext4_map_blocks *map,
@@ -3210,7 +3155,7 @@ ext4_ext_handle_uninitialized_extents(handle_t *handle, struct inode *inode,
3210 * completed 3155 * completed
3211 */ 3156 */
3212 if (io) 3157 if (io)
3213 io->flag = EXT4_IO_UNWRITTEN; 3158 io->flag = EXT4_IO_END_UNWRITTEN;
3214 else 3159 else
3215 ext4_set_inode_state(inode, EXT4_STATE_DIO_UNWRITTEN); 3160 ext4_set_inode_state(inode, EXT4_STATE_DIO_UNWRITTEN);
3216 if (ext4_should_dioread_nolock(inode)) 3161 if (ext4_should_dioread_nolock(inode))
@@ -3221,8 +3166,12 @@ ext4_ext_handle_uninitialized_extents(handle_t *handle, struct inode *inode,
3221 if ((flags & EXT4_GET_BLOCKS_CONVERT)) { 3166 if ((flags & EXT4_GET_BLOCKS_CONVERT)) {
3222 ret = ext4_convert_unwritten_extents_endio(handle, inode, 3167 ret = ext4_convert_unwritten_extents_endio(handle, inode,
3223 path); 3168 path);
3224 if (ret >= 0) 3169 if (ret >= 0) {
3225 ext4_update_inode_fsync_trans(handle, inode, 1); 3170 ext4_update_inode_fsync_trans(handle, inode, 1);
3171 err = check_eofblocks_fl(handle, inode, map, path,
3172 map->m_len);
3173 } else
3174 err = ret;
3226 goto out2; 3175 goto out2;
3227 } 3176 }
3228 /* buffered IO case */ 3177 /* buffered IO case */
@@ -3248,8 +3197,13 @@ ext4_ext_handle_uninitialized_extents(handle_t *handle, struct inode *inode,
3248 3197
3249 /* buffered write, writepage time, convert*/ 3198 /* buffered write, writepage time, convert*/
3250 ret = ext4_ext_convert_to_initialized(handle, inode, map, path); 3199 ret = ext4_ext_convert_to_initialized(handle, inode, map, path);
3251 if (ret >= 0) 3200 if (ret >= 0) {
3252 ext4_update_inode_fsync_trans(handle, inode, 1); 3201 ext4_update_inode_fsync_trans(handle, inode, 1);
3202 err = check_eofblocks_fl(handle, inode, map, path, map->m_len);
3203 if (err < 0)
3204 goto out2;
3205 }
3206
3253out: 3207out:
3254 if (ret <= 0) { 3208 if (ret <= 0) {
3255 err = ret; 3209 err = ret;
@@ -3296,6 +3250,7 @@ out2:
3296 } 3250 }
3297 return err ? err : allocated; 3251 return err ? err : allocated;
3298} 3252}
3253
3299/* 3254/*
3300 * Block allocation/map/preallocation routine for extents based files 3255 * Block allocation/map/preallocation routine for extents based files
3301 * 3256 *
@@ -3319,9 +3274,9 @@ int ext4_ext_map_blocks(handle_t *handle, struct inode *inode,
3319{ 3274{
3320 struct ext4_ext_path *path = NULL; 3275 struct ext4_ext_path *path = NULL;
3321 struct ext4_extent_header *eh; 3276 struct ext4_extent_header *eh;
3322 struct ext4_extent newex, *ex, *last_ex; 3277 struct ext4_extent newex, *ex;
3323 ext4_fsblk_t newblock; 3278 ext4_fsblk_t newblock;
3324 int i, err = 0, depth, ret, cache_type; 3279 int err = 0, depth, ret, cache_type;
3325 unsigned int allocated = 0; 3280 unsigned int allocated = 0;
3326 struct ext4_allocation_request ar; 3281 struct ext4_allocation_request ar;
3327 ext4_io_end_t *io = EXT4_I(inode)->cur_aio_dio; 3282 ext4_io_end_t *io = EXT4_I(inode)->cur_aio_dio;
@@ -3345,7 +3300,7 @@ int ext4_ext_map_blocks(handle_t *handle, struct inode *inode,
3345 /* block is already allocated */ 3300 /* block is already allocated */
3346 newblock = map->m_lblk 3301 newblock = map->m_lblk
3347 - le32_to_cpu(newex.ee_block) 3302 - le32_to_cpu(newex.ee_block)
3348 + ext_pblock(&newex); 3303 + ext4_ext_pblock(&newex);
3349 /* number of remaining blocks in the extent */ 3304 /* number of remaining blocks in the extent */
3350 allocated = ext4_ext_get_actual_len(&newex) - 3305 allocated = ext4_ext_get_actual_len(&newex) -
3351 (map->m_lblk - le32_to_cpu(newex.ee_block)); 3306 (map->m_lblk - le32_to_cpu(newex.ee_block));
@@ -3383,7 +3338,7 @@ int ext4_ext_map_blocks(handle_t *handle, struct inode *inode,
3383 ex = path[depth].p_ext; 3338 ex = path[depth].p_ext;
3384 if (ex) { 3339 if (ex) {
3385 ext4_lblk_t ee_block = le32_to_cpu(ex->ee_block); 3340 ext4_lblk_t ee_block = le32_to_cpu(ex->ee_block);
3386 ext4_fsblk_t ee_start = ext_pblock(ex); 3341 ext4_fsblk_t ee_start = ext4_ext_pblock(ex);
3387 unsigned short ee_len; 3342 unsigned short ee_len;
3388 3343
3389 /* 3344 /*
@@ -3492,7 +3447,7 @@ int ext4_ext_map_blocks(handle_t *handle, struct inode *inode,
3492 */ 3447 */
3493 if ((flags & EXT4_GET_BLOCKS_PRE_IO)) { 3448 if ((flags & EXT4_GET_BLOCKS_PRE_IO)) {
3494 if (io) 3449 if (io)
3495 io->flag = EXT4_IO_UNWRITTEN; 3450 io->flag = EXT4_IO_END_UNWRITTEN;
3496 else 3451 else
3497 ext4_set_inode_state(inode, 3452 ext4_set_inode_state(inode,
3498 EXT4_STATE_DIO_UNWRITTEN); 3453 EXT4_STATE_DIO_UNWRITTEN);
@@ -3501,44 +3456,23 @@ int ext4_ext_map_blocks(handle_t *handle, struct inode *inode,
3501 map->m_flags |= EXT4_MAP_UNINIT; 3456 map->m_flags |= EXT4_MAP_UNINIT;
3502 } 3457 }
3503 3458
3504 if (unlikely(ext4_test_inode_flag(inode, EXT4_INODE_EOFBLOCKS))) { 3459 err = check_eofblocks_fl(handle, inode, map, path, ar.len);
3505 if (unlikely(!eh->eh_entries)) { 3460 if (err)
3506 EXT4_ERROR_INODE(inode, 3461 goto out2;
3507 "eh->eh_entries == 0 and " 3462
3508 "EOFBLOCKS_FL set");
3509 err = -EIO;
3510 goto out2;
3511 }
3512 last_ex = EXT_LAST_EXTENT(eh);
3513 /*
3514 * If the current leaf block was reached by looking at
3515 * the last index block all the way down the tree, and
3516 * we are extending the inode beyond the last extent
3517 * in the current leaf block, then clear the
3518 * EOFBLOCKS_FL flag.
3519 */
3520 for (i = depth-1; i >= 0; i--) {
3521 if (path[i].p_idx != EXT_LAST_INDEX(path[i].p_hdr))
3522 break;
3523 }
3524 if ((i < 0) &&
3525 (map->m_lblk + ar.len > le32_to_cpu(last_ex->ee_block) +
3526 ext4_ext_get_actual_len(last_ex)))
3527 ext4_clear_inode_flag(inode, EXT4_INODE_EOFBLOCKS);
3528 }
3529 err = ext4_ext_insert_extent(handle, inode, path, &newex, flags); 3463 err = ext4_ext_insert_extent(handle, inode, path, &newex, flags);
3530 if (err) { 3464 if (err) {
3531 /* free data blocks we just allocated */ 3465 /* free data blocks we just allocated */
3532 /* not a good idea to call discard here directly, 3466 /* not a good idea to call discard here directly,
3533 * but otherwise we'd need to call it every free() */ 3467 * but otherwise we'd need to call it every free() */
3534 ext4_discard_preallocations(inode); 3468 ext4_discard_preallocations(inode);
3535 ext4_free_blocks(handle, inode, 0, ext_pblock(&newex), 3469 ext4_free_blocks(handle, inode, 0, ext4_ext_pblock(&newex),
3536 ext4_ext_get_actual_len(&newex), 0); 3470 ext4_ext_get_actual_len(&newex), 0);
3537 goto out2; 3471 goto out2;
3538 } 3472 }
3539 3473
3540 /* previous routine could use block we allocated */ 3474 /* previous routine could use block we allocated */
3541 newblock = ext_pblock(&newex); 3475 newblock = ext4_ext_pblock(&newex);
3542 allocated = ext4_ext_get_actual_len(&newex); 3476 allocated = ext4_ext_get_actual_len(&newex);
3543 if (allocated > map->m_len) 3477 if (allocated > map->m_len)
3544 allocated = map->m_len; 3478 allocated = map->m_len;
@@ -3733,7 +3667,7 @@ retry:
3733 printk(KERN_ERR "%s: ext4_ext_map_blocks " 3667 printk(KERN_ERR "%s: ext4_ext_map_blocks "
3734 "returned error inode#%lu, block=%u, " 3668 "returned error inode#%lu, block=%u, "
3735 "max_blocks=%u", __func__, 3669 "max_blocks=%u", __func__,
3736 inode->i_ino, block, max_blocks); 3670 inode->i_ino, map.m_lblk, max_blocks);
3737#endif 3671#endif
3738 ext4_mark_inode_dirty(handle, inode); 3672 ext4_mark_inode_dirty(handle, inode);
3739 ret2 = ext4_journal_stop(handle); 3673 ret2 = ext4_journal_stop(handle);
diff --git a/fs/ext4/file.c b/fs/ext4/file.c
index 5313ae4cda2d..5a5c55ddceef 100644
--- a/fs/ext4/file.c
+++ b/fs/ext4/file.c
@@ -70,7 +70,8 @@ ext4_file_write(struct kiocb *iocb, const struct iovec *iov,
70 struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb); 70 struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
71 size_t length = iov_length(iov, nr_segs); 71 size_t length = iov_length(iov, nr_segs);
72 72
73 if (pos > sbi->s_bitmap_maxbytes) 73 if ((pos > sbi->s_bitmap_maxbytes ||
74 (pos == sbi->s_bitmap_maxbytes && length > 0)))
74 return -EFBIG; 75 return -EFBIG;
75 76
76 if (pos + length > sbi->s_bitmap_maxbytes) { 77 if (pos + length > sbi->s_bitmap_maxbytes) {
@@ -123,14 +124,56 @@ static int ext4_file_open(struct inode * inode, struct file * filp)
123 if (!IS_ERR(cp)) { 124 if (!IS_ERR(cp)) {
124 memcpy(sbi->s_es->s_last_mounted, cp, 125 memcpy(sbi->s_es->s_last_mounted, cp,
125 sizeof(sbi->s_es->s_last_mounted)); 126 sizeof(sbi->s_es->s_last_mounted));
126 sb->s_dirt = 1; 127 ext4_mark_super_dirty(sb);
127 } 128 }
128 } 129 }
129 return dquot_file_open(inode, filp); 130 return dquot_file_open(inode, filp);
130} 131}
131 132
133/*
134 * ext4_llseek() copied from generic_file_llseek() to handle both
135 * block-mapped and extent-mapped maxbytes values. This should
136 * otherwise be identical with generic_file_llseek().
137 */
138loff_t ext4_llseek(struct file *file, loff_t offset, int origin)
139{
140 struct inode *inode = file->f_mapping->host;
141 loff_t maxbytes;
142
143 if (!(ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS)))
144 maxbytes = EXT4_SB(inode->i_sb)->s_bitmap_maxbytes;
145 else
146 maxbytes = inode->i_sb->s_maxbytes;
147 mutex_lock(&inode->i_mutex);
148 switch (origin) {
149 case SEEK_END:
150 offset += inode->i_size;
151 break;
152 case SEEK_CUR:
153 if (offset == 0) {
154 mutex_unlock(&inode->i_mutex);
155 return file->f_pos;
156 }
157 offset += file->f_pos;
158 break;
159 }
160
161 if (offset < 0 || offset > maxbytes) {
162 mutex_unlock(&inode->i_mutex);
163 return -EINVAL;
164 }
165
166 if (offset != file->f_pos) {
167 file->f_pos = offset;
168 file->f_version = 0;
169 }
170 mutex_unlock(&inode->i_mutex);
171
172 return offset;
173}
174
132const struct file_operations ext4_file_operations = { 175const struct file_operations ext4_file_operations = {
133 .llseek = generic_file_llseek, 176 .llseek = ext4_llseek,
134 .read = do_sync_read, 177 .read = do_sync_read,
135 .write = do_sync_write, 178 .write = do_sync_write,
136 .aio_read = generic_file_aio_read, 179 .aio_read = generic_file_aio_read,
diff --git a/fs/ext4/fsync.c b/fs/ext4/fsync.c
index 592adf2e546e..c1a7bc923cf6 100644
--- a/fs/ext4/fsync.c
+++ b/fs/ext4/fsync.c
@@ -34,6 +34,89 @@
34 34
35#include <trace/events/ext4.h> 35#include <trace/events/ext4.h>
36 36
37static void dump_completed_IO(struct inode * inode)
38{
39#ifdef EXT4_DEBUG
40 struct list_head *cur, *before, *after;
41 ext4_io_end_t *io, *io0, *io1;
42 unsigned long flags;
43
44 if (list_empty(&EXT4_I(inode)->i_completed_io_list)){
45 ext4_debug("inode %lu completed_io list is empty\n", inode->i_ino);
46 return;
47 }
48
49 ext4_debug("Dump inode %lu completed_io list \n", inode->i_ino);
50 spin_lock_irqsave(&EXT4_I(inode)->i_completed_io_lock, flags);
51 list_for_each_entry(io, &EXT4_I(inode)->i_completed_io_list, list){
52 cur = &io->list;
53 before = cur->prev;
54 io0 = container_of(before, ext4_io_end_t, list);
55 after = cur->next;
56 io1 = container_of(after, ext4_io_end_t, list);
57
58 ext4_debug("io 0x%p from inode %lu,prev 0x%p,next 0x%p\n",
59 io, inode->i_ino, io0, io1);
60 }
61 spin_unlock_irqrestore(&EXT4_I(inode)->i_completed_io_lock, flags);
62#endif
63}
64
65/*
66 * This function is called from ext4_sync_file().
67 *
68 * When IO is completed, the work to convert unwritten extents to
69 * written is queued on workqueue but may not get immediately
70 * scheduled. When fsync is called, we need to ensure the
71 * conversion is complete before fsync returns.
72 * The inode keeps track of a list of pending/completed IO that
73 * might needs to do the conversion. This function walks through
74 * the list and convert the related unwritten extents for completed IO
75 * to written.
76 * The function return the number of pending IOs on success.
77 */
78static int flush_completed_IO(struct inode *inode)
79{
80 ext4_io_end_t *io;
81 struct ext4_inode_info *ei = EXT4_I(inode);
82 unsigned long flags;
83 int ret = 0;
84 int ret2 = 0;
85
86 if (list_empty(&ei->i_completed_io_list))
87 return ret;
88
89 dump_completed_IO(inode);
90 spin_lock_irqsave(&ei->i_completed_io_lock, flags);
91 while (!list_empty(&ei->i_completed_io_list)){
92 io = list_entry(ei->i_completed_io_list.next,
93 ext4_io_end_t, list);
94 /*
95 * Calling ext4_end_io_nolock() to convert completed
96 * IO to written.
97 *
98 * When ext4_sync_file() is called, run_queue() may already
99 * about to flush the work corresponding to this io structure.
100 * It will be upset if it founds the io structure related
101 * to the work-to-be schedule is freed.
102 *
103 * Thus we need to keep the io structure still valid here after
104 * convertion finished. The io structure has a flag to
105 * avoid double converting from both fsync and background work
106 * queue work.
107 */
108 spin_unlock_irqrestore(&ei->i_completed_io_lock, flags);
109 ret = ext4_end_io_nolock(io);
110 spin_lock_irqsave(&ei->i_completed_io_lock, flags);
111 if (ret < 0)
112 ret2 = ret;
113 else
114 list_del_init(&io->list);
115 }
116 spin_unlock_irqrestore(&ei->i_completed_io_lock, flags);
117 return (ret2 < 0) ? ret2 : 0;
118}
119
37/* 120/*
38 * If we're not journaling and this is a just-created file, we have to 121 * If we're not journaling and this is a just-created file, we have to
39 * sync our parent directory (if it was freshly created) since 122 * sync our parent directory (if it was freshly created) since
@@ -128,10 +211,9 @@ int ext4_sync_file(struct file *file, int datasync)
128 (journal->j_fs_dev != journal->j_dev) && 211 (journal->j_fs_dev != journal->j_dev) &&
129 (journal->j_flags & JBD2_BARRIER)) 212 (journal->j_flags & JBD2_BARRIER))
130 blkdev_issue_flush(inode->i_sb->s_bdev, GFP_KERNEL, 213 blkdev_issue_flush(inode->i_sb->s_bdev, GFP_KERNEL,
131 NULL, BLKDEV_IFL_WAIT); 214 NULL);
132 ret = jbd2_log_wait_commit(journal, commit_tid); 215 ret = jbd2_log_wait_commit(journal, commit_tid);
133 } else if (journal->j_flags & JBD2_BARRIER) 216 } else if (journal->j_flags & JBD2_BARRIER)
134 blkdev_issue_flush(inode->i_sb->s_bdev, GFP_KERNEL, NULL, 217 blkdev_issue_flush(inode->i_sb->s_bdev, GFP_KERNEL, NULL);
135 BLKDEV_IFL_WAIT);
136 return ret; 218 return ret;
137} 219}
diff --git a/fs/ext4/ialloc.c b/fs/ext4/ialloc.c
index 25c4b3173fd9..1ce240a23ebb 100644
--- a/fs/ext4/ialloc.c
+++ b/fs/ext4/ialloc.c
@@ -50,7 +50,7 @@
50 * need to use it within a single byte (to ensure we get endianness right). 50 * need to use it within a single byte (to ensure we get endianness right).
51 * We can use memset for the rest of the bitmap as there are no other users. 51 * We can use memset for the rest of the bitmap as there are no other users.
52 */ 52 */
53void mark_bitmap_end(int start_bit, int end_bit, char *bitmap) 53void ext4_mark_bitmap_end(int start_bit, int end_bit, char *bitmap)
54{ 54{
55 int i; 55 int i;
56 56
@@ -65,9 +65,10 @@ void mark_bitmap_end(int start_bit, int end_bit, char *bitmap)
65} 65}
66 66
67/* Initializes an uninitialized inode bitmap */ 67/* Initializes an uninitialized inode bitmap */
68unsigned ext4_init_inode_bitmap(struct super_block *sb, struct buffer_head *bh, 68static unsigned ext4_init_inode_bitmap(struct super_block *sb,
69 ext4_group_t block_group, 69 struct buffer_head *bh,
70 struct ext4_group_desc *gdp) 70 ext4_group_t block_group,
71 struct ext4_group_desc *gdp)
71{ 72{
72 struct ext4_sb_info *sbi = EXT4_SB(sb); 73 struct ext4_sb_info *sbi = EXT4_SB(sb);
73 74
@@ -85,7 +86,7 @@ unsigned ext4_init_inode_bitmap(struct super_block *sb, struct buffer_head *bh,
85 } 86 }
86 87
87 memset(bh->b_data, 0, (EXT4_INODES_PER_GROUP(sb) + 7) / 8); 88 memset(bh->b_data, 0, (EXT4_INODES_PER_GROUP(sb) + 7) / 8);
88 mark_bitmap_end(EXT4_INODES_PER_GROUP(sb), sb->s_blocksize * 8, 89 ext4_mark_bitmap_end(EXT4_INODES_PER_GROUP(sb), sb->s_blocksize * 8,
89 bh->b_data); 90 bh->b_data);
90 91
91 return EXT4_INODES_PER_GROUP(sb); 92 return EXT4_INODES_PER_GROUP(sb);
@@ -107,6 +108,7 @@ ext4_read_inode_bitmap(struct super_block *sb, ext4_group_t block_group)
107 desc = ext4_get_group_desc(sb, block_group, NULL); 108 desc = ext4_get_group_desc(sb, block_group, NULL);
108 if (!desc) 109 if (!desc)
109 return NULL; 110 return NULL;
111
110 bitmap_blk = ext4_inode_bitmap(sb, desc); 112 bitmap_blk = ext4_inode_bitmap(sb, desc);
111 bh = sb_getblk(sb, bitmap_blk); 113 bh = sb_getblk(sb, bitmap_blk);
112 if (unlikely(!bh)) { 114 if (unlikely(!bh)) {
@@ -123,6 +125,7 @@ ext4_read_inode_bitmap(struct super_block *sb, ext4_group_t block_group)
123 unlock_buffer(bh); 125 unlock_buffer(bh);
124 return bh; 126 return bh;
125 } 127 }
128
126 ext4_lock_group(sb, block_group); 129 ext4_lock_group(sb, block_group);
127 if (desc->bg_flags & cpu_to_le16(EXT4_BG_INODE_UNINIT)) { 130 if (desc->bg_flags & cpu_to_le16(EXT4_BG_INODE_UNINIT)) {
128 ext4_init_inode_bitmap(sb, bh, block_group, desc); 131 ext4_init_inode_bitmap(sb, bh, block_group, desc);
@@ -133,6 +136,7 @@ ext4_read_inode_bitmap(struct super_block *sb, ext4_group_t block_group)
133 return bh; 136 return bh;
134 } 137 }
135 ext4_unlock_group(sb, block_group); 138 ext4_unlock_group(sb, block_group);
139
136 if (buffer_uptodate(bh)) { 140 if (buffer_uptodate(bh)) {
137 /* 141 /*
138 * if not uninit if bh is uptodate, 142 * if not uninit if bh is uptodate,
@@ -222,7 +226,7 @@ void ext4_free_inode(handle_t *handle, struct inode *inode)
222 is_directory = S_ISDIR(inode->i_mode); 226 is_directory = S_ISDIR(inode->i_mode);
223 227
224 /* Do this BEFORE marking the inode not in use or returning an error */ 228 /* Do this BEFORE marking the inode not in use or returning an error */
225 clear_inode(inode); 229 ext4_clear_inode(inode);
226 230
227 es = EXT4_SB(sb)->s_es; 231 es = EXT4_SB(sb)->s_es;
228 if (ino < EXT4_FIRST_INO(sb) || ino > le32_to_cpu(es->s_inodes_count)) { 232 if (ino < EXT4_FIRST_INO(sb) || ino > le32_to_cpu(es->s_inodes_count)) {
@@ -279,7 +283,7 @@ out:
279 err = ext4_handle_dirty_metadata(handle, NULL, bitmap_bh); 283 err = ext4_handle_dirty_metadata(handle, NULL, bitmap_bh);
280 if (!fatal) 284 if (!fatal)
281 fatal = err; 285 fatal = err;
282 sb->s_dirt = 1; 286 ext4_mark_super_dirty(sb);
283 } else 287 } else
284 ext4_error(sb, "bit already cleared for inode %lu", ino); 288 ext4_error(sb, "bit already cleared for inode %lu", ino);
285 289
@@ -411,8 +415,8 @@ struct orlov_stats {
411 * for a particular block group or flex_bg. If flex_size is 1, then g 415 * for a particular block group or flex_bg. If flex_size is 1, then g
412 * is a block group number; otherwise it is flex_bg number. 416 * is a block group number; otherwise it is flex_bg number.
413 */ 417 */
414void get_orlov_stats(struct super_block *sb, ext4_group_t g, 418static void get_orlov_stats(struct super_block *sb, ext4_group_t g,
415 int flex_size, struct orlov_stats *stats) 419 int flex_size, struct orlov_stats *stats)
416{ 420{
417 struct ext4_group_desc *desc; 421 struct ext4_group_desc *desc;
418 struct flex_groups *flex_group = EXT4_SB(sb)->s_flex_groups; 422 struct flex_groups *flex_group = EXT4_SB(sb)->s_flex_groups;
@@ -712,8 +716,17 @@ static int ext4_claim_inode(struct super_block *sb,
712{ 716{
713 int free = 0, retval = 0, count; 717 int free = 0, retval = 0, count;
714 struct ext4_sb_info *sbi = EXT4_SB(sb); 718 struct ext4_sb_info *sbi = EXT4_SB(sb);
719 struct ext4_group_info *grp = ext4_get_group_info(sb, group);
715 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, group, NULL); 720 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, group, NULL);
716 721
722 /*
723 * We have to be sure that new inode allocation does not race with
724 * inode table initialization, because otherwise we may end up
725 * allocating and writing new inode right before sb_issue_zeroout
726 * takes place and overwriting our new inode with zeroes. So we
727 * take alloc_sem to prevent it.
728 */
729 down_read(&grp->alloc_sem);
717 ext4_lock_group(sb, group); 730 ext4_lock_group(sb, group);
718 if (ext4_set_bit(ino, inode_bitmap_bh->b_data)) { 731 if (ext4_set_bit(ino, inode_bitmap_bh->b_data)) {
719 /* not a free inode */ 732 /* not a free inode */
@@ -724,6 +737,7 @@ static int ext4_claim_inode(struct super_block *sb,
724 if ((group == 0 && ino < EXT4_FIRST_INO(sb)) || 737 if ((group == 0 && ino < EXT4_FIRST_INO(sb)) ||
725 ino > EXT4_INODES_PER_GROUP(sb)) { 738 ino > EXT4_INODES_PER_GROUP(sb)) {
726 ext4_unlock_group(sb, group); 739 ext4_unlock_group(sb, group);
740 up_read(&grp->alloc_sem);
727 ext4_error(sb, "reserved inode or inode > inodes count - " 741 ext4_error(sb, "reserved inode or inode > inodes count - "
728 "block_group = %u, inode=%lu", group, 742 "block_group = %u, inode=%lu", group,
729 ino + group * EXT4_INODES_PER_GROUP(sb)); 743 ino + group * EXT4_INODES_PER_GROUP(sb));
@@ -772,6 +786,7 @@ static int ext4_claim_inode(struct super_block *sb,
772 gdp->bg_checksum = ext4_group_desc_csum(sbi, group, gdp); 786 gdp->bg_checksum = ext4_group_desc_csum(sbi, group, gdp);
773err_ret: 787err_ret:
774 ext4_unlock_group(sb, group); 788 ext4_unlock_group(sb, group);
789 up_read(&grp->alloc_sem);
775 return retval; 790 return retval;
776} 791}
777 792
@@ -965,7 +980,7 @@ got:
965 percpu_counter_dec(&sbi->s_freeinodes_counter); 980 percpu_counter_dec(&sbi->s_freeinodes_counter);
966 if (S_ISDIR(mode)) 981 if (S_ISDIR(mode))
967 percpu_counter_inc(&sbi->s_dirs_counter); 982 percpu_counter_inc(&sbi->s_dirs_counter);
968 sb->s_dirt = 1; 983 ext4_mark_super_dirty(sb);
969 984
970 if (sbi->s_log_groups_per_flex) { 985 if (sbi->s_log_groups_per_flex) {
971 flex_group = ext4_flex_group(sbi, group); 986 flex_group = ext4_flex_group(sbi, group);
@@ -1205,3 +1220,109 @@ unsigned long ext4_count_dirs(struct super_block * sb)
1205 } 1220 }
1206 return count; 1221 return count;
1207} 1222}
1223
1224/*
1225 * Zeroes not yet zeroed inode table - just write zeroes through the whole
1226 * inode table. Must be called without any spinlock held. The only place
1227 * where it is called from on active part of filesystem is ext4lazyinit
1228 * thread, so we do not need any special locks, however we have to prevent
1229 * inode allocation from the current group, so we take alloc_sem lock, to
1230 * block ext4_claim_inode until we are finished.
1231 */
1232extern int ext4_init_inode_table(struct super_block *sb, ext4_group_t group,
1233 int barrier)
1234{
1235 struct ext4_group_info *grp = ext4_get_group_info(sb, group);
1236 struct ext4_sb_info *sbi = EXT4_SB(sb);
1237 struct ext4_group_desc *gdp = NULL;
1238 struct buffer_head *group_desc_bh;
1239 handle_t *handle;
1240 ext4_fsblk_t blk;
1241 int num, ret = 0, used_blks = 0;
1242
1243 /* This should not happen, but just to be sure check this */
1244 if (sb->s_flags & MS_RDONLY) {
1245 ret = 1;
1246 goto out;
1247 }
1248
1249 gdp = ext4_get_group_desc(sb, group, &group_desc_bh);
1250 if (!gdp)
1251 goto out;
1252
1253 /*
1254 * We do not need to lock this, because we are the only one
1255 * handling this flag.
1256 */
1257 if (gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED))
1258 goto out;
1259
1260 handle = ext4_journal_start_sb(sb, 1);
1261 if (IS_ERR(handle)) {
1262 ret = PTR_ERR(handle);
1263 goto out;
1264 }
1265
1266 down_write(&grp->alloc_sem);
1267 /*
1268 * If inode bitmap was already initialized there may be some
1269 * used inodes so we need to skip blocks with used inodes in
1270 * inode table.
1271 */
1272 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_UNINIT)))
1273 used_blks = DIV_ROUND_UP((EXT4_INODES_PER_GROUP(sb) -
1274 ext4_itable_unused_count(sb, gdp)),
1275 sbi->s_inodes_per_block);
1276
1277 if ((used_blks < 0) || (used_blks > sbi->s_itb_per_group)) {
1278 ext4_error(sb, "Something is wrong with group %u\n"
1279 "Used itable blocks: %d"
1280 "itable unused count: %u\n",
1281 group, used_blks,
1282 ext4_itable_unused_count(sb, gdp));
1283 ret = 1;
1284 goto out;
1285 }
1286
1287 blk = ext4_inode_table(sb, gdp) + used_blks;
1288 num = sbi->s_itb_per_group - used_blks;
1289
1290 BUFFER_TRACE(group_desc_bh, "get_write_access");
1291 ret = ext4_journal_get_write_access(handle,
1292 group_desc_bh);
1293 if (ret)
1294 goto err_out;
1295
1296 /*
1297 * Skip zeroout if the inode table is full. But we set the ZEROED
1298 * flag anyway, because obviously, when it is full it does not need
1299 * further zeroing.
1300 */
1301 if (unlikely(num == 0))
1302 goto skip_zeroout;
1303
1304 ext4_debug("going to zero out inode table in group %d\n",
1305 group);
1306 ret = sb_issue_zeroout(sb, blk, num, GFP_NOFS);
1307 if (ret < 0)
1308 goto err_out;
1309 if (barrier)
1310 blkdev_issue_flush(sb->s_bdev, GFP_NOFS, NULL);
1311
1312skip_zeroout:
1313 ext4_lock_group(sb, group);
1314 gdp->bg_flags |= cpu_to_le16(EXT4_BG_INODE_ZEROED);
1315 gdp->bg_checksum = ext4_group_desc_csum(sbi, group, gdp);
1316 ext4_unlock_group(sb, group);
1317
1318 BUFFER_TRACE(group_desc_bh,
1319 "call ext4_handle_dirty_metadata");
1320 ret = ext4_handle_dirty_metadata(handle, NULL,
1321 group_desc_bh);
1322
1323err_out:
1324 up_write(&grp->alloc_sem);
1325 ext4_journal_stop(handle);
1326out:
1327 return ret;
1328}
diff --git a/fs/ext4/inode.c b/fs/ext4/inode.c
index 42272d67955a..2d6c6c8c036d 100644
--- a/fs/ext4/inode.c
+++ b/fs/ext4/inode.c
@@ -60,6 +60,12 @@ static inline int ext4_begin_ordered_truncate(struct inode *inode,
60} 60}
61 61
62static void ext4_invalidatepage(struct page *page, unsigned long offset); 62static void ext4_invalidatepage(struct page *page, unsigned long offset);
63static int noalloc_get_block_write(struct inode *inode, sector_t iblock,
64 struct buffer_head *bh_result, int create);
65static int ext4_set_bh_endio(struct buffer_head *bh, struct inode *inode);
66static void ext4_end_io_buffer_write(struct buffer_head *bh, int uptodate);
67static int __ext4_journalled_writepage(struct page *page, unsigned int len);
68static int ext4_bh_delay_or_unwritten(handle_t *handle, struct buffer_head *bh);
63 69
64/* 70/*
65 * Test whether an inode is a fast symlink. 71 * Test whether an inode is a fast symlink.
@@ -167,11 +173,16 @@ int ext4_truncate_restart_trans(handle_t *handle, struct inode *inode,
167/* 173/*
168 * Called at the last iput() if i_nlink is zero. 174 * Called at the last iput() if i_nlink is zero.
169 */ 175 */
170void ext4_delete_inode(struct inode *inode) 176void ext4_evict_inode(struct inode *inode)
171{ 177{
172 handle_t *handle; 178 handle_t *handle;
173 int err; 179 int err;
174 180
181 if (inode->i_nlink) {
182 truncate_inode_pages(&inode->i_data, 0);
183 goto no_delete;
184 }
185
175 if (!is_bad_inode(inode)) 186 if (!is_bad_inode(inode))
176 dquot_initialize(inode); 187 dquot_initialize(inode);
177 188
@@ -221,6 +232,7 @@ void ext4_delete_inode(struct inode *inode)
221 "couldn't extend journal (err %d)", err); 232 "couldn't extend journal (err %d)", err);
222 stop_handle: 233 stop_handle:
223 ext4_journal_stop(handle); 234 ext4_journal_stop(handle);
235 ext4_orphan_del(NULL, inode);
224 goto no_delete; 236 goto no_delete;
225 } 237 }
226 } 238 }
@@ -245,13 +257,13 @@ void ext4_delete_inode(struct inode *inode)
245 */ 257 */
246 if (ext4_mark_inode_dirty(handle, inode)) 258 if (ext4_mark_inode_dirty(handle, inode))
247 /* If that failed, just do the required in-core inode clear. */ 259 /* If that failed, just do the required in-core inode clear. */
248 clear_inode(inode); 260 ext4_clear_inode(inode);
249 else 261 else
250 ext4_free_inode(handle, inode); 262 ext4_free_inode(handle, inode);
251 ext4_journal_stop(handle); 263 ext4_journal_stop(handle);
252 return; 264 return;
253no_delete: 265no_delete:
254 clear_inode(inode); /* We must guarantee clearing of inode... */ 266 ext4_clear_inode(inode); /* We must guarantee clearing of inode... */
255} 267}
256 268
257typedef struct { 269typedef struct {
@@ -337,9 +349,11 @@ static int ext4_block_to_path(struct inode *inode,
337 return n; 349 return n;
338} 350}
339 351
340static int __ext4_check_blockref(const char *function, struct inode *inode, 352static int __ext4_check_blockref(const char *function, unsigned int line,
353 struct inode *inode,
341 __le32 *p, unsigned int max) 354 __le32 *p, unsigned int max)
342{ 355{
356 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
343 __le32 *bref = p; 357 __le32 *bref = p;
344 unsigned int blk; 358 unsigned int blk;
345 359
@@ -348,8 +362,9 @@ static int __ext4_check_blockref(const char *function, struct inode *inode,
348 if (blk && 362 if (blk &&
349 unlikely(!ext4_data_block_valid(EXT4_SB(inode->i_sb), 363 unlikely(!ext4_data_block_valid(EXT4_SB(inode->i_sb),
350 blk, 1))) { 364 blk, 1))) {
351 ext4_error_inode(function, inode, 365 es->s_last_error_block = cpu_to_le64(blk);
352 "invalid block reference %u", blk); 366 ext4_error_inode(inode, function, line, blk,
367 "invalid block");
353 return -EIO; 368 return -EIO;
354 } 369 }
355 } 370 }
@@ -358,11 +373,13 @@ static int __ext4_check_blockref(const char *function, struct inode *inode,
358 373
359 374
360#define ext4_check_indirect_blockref(inode, bh) \ 375#define ext4_check_indirect_blockref(inode, bh) \
361 __ext4_check_blockref(__func__, inode, (__le32 *)(bh)->b_data, \ 376 __ext4_check_blockref(__func__, __LINE__, inode, \
377 (__le32 *)(bh)->b_data, \
362 EXT4_ADDR_PER_BLOCK((inode)->i_sb)) 378 EXT4_ADDR_PER_BLOCK((inode)->i_sb))
363 379
364#define ext4_check_inode_blockref(inode) \ 380#define ext4_check_inode_blockref(inode) \
365 __ext4_check_blockref(__func__, inode, EXT4_I(inode)->i_data, \ 381 __ext4_check_blockref(__func__, __LINE__, inode, \
382 EXT4_I(inode)->i_data, \
366 EXT4_NDIR_BLOCKS) 383 EXT4_NDIR_BLOCKS)
367 384
368/** 385/**
@@ -744,6 +761,11 @@ static int ext4_alloc_branch(handle_t *handle, struct inode *inode,
744 * parent to disk. 761 * parent to disk.
745 */ 762 */
746 bh = sb_getblk(inode->i_sb, new_blocks[n-1]); 763 bh = sb_getblk(inode->i_sb, new_blocks[n-1]);
764 if (unlikely(!bh)) {
765 err = -EIO;
766 goto failed;
767 }
768
747 branch[n].bh = bh; 769 branch[n].bh = bh;
748 lock_buffer(bh); 770 lock_buffer(bh);
749 BUFFER_TRACE(bh, "call get_create_access"); 771 BUFFER_TRACE(bh, "call get_create_access");
@@ -1128,20 +1150,24 @@ void ext4_da_update_reserve_space(struct inode *inode,
1128 ext4_discard_preallocations(inode); 1150 ext4_discard_preallocations(inode);
1129} 1151}
1130 1152
1131static int check_block_validity(struct inode *inode, const char *func, 1153static int __check_block_validity(struct inode *inode, const char *func,
1154 unsigned int line,
1132 struct ext4_map_blocks *map) 1155 struct ext4_map_blocks *map)
1133{ 1156{
1134 if (!ext4_data_block_valid(EXT4_SB(inode->i_sb), map->m_pblk, 1157 if (!ext4_data_block_valid(EXT4_SB(inode->i_sb), map->m_pblk,
1135 map->m_len)) { 1158 map->m_len)) {
1136 ext4_error_inode(func, inode, 1159 ext4_error_inode(inode, func, line, map->m_pblk,
1137 "lblock %lu mapped to illegal pblock %llu " 1160 "lblock %lu mapped to illegal pblock "
1138 "(length %d)", (unsigned long) map->m_lblk, 1161 "(length %d)", (unsigned long) map->m_lblk,
1139 map->m_pblk, map->m_len); 1162 map->m_len);
1140 return -EIO; 1163 return -EIO;
1141 } 1164 }
1142 return 0; 1165 return 0;
1143} 1166}
1144 1167
1168#define check_block_validity(inode, map) \
1169 __check_block_validity((inode), __func__, __LINE__, (map))
1170
1145/* 1171/*
1146 * Return the number of contiguous dirty pages in a given inode 1172 * Return the number of contiguous dirty pages in a given inode
1147 * starting at page frame idx. 1173 * starting at page frame idx.
@@ -1192,8 +1218,10 @@ static pgoff_t ext4_num_dirty_pages(struct inode *inode, pgoff_t idx,
1192 break; 1218 break;
1193 idx++; 1219 idx++;
1194 num++; 1220 num++;
1195 if (num >= max_pages) 1221 if (num >= max_pages) {
1222 done = 1;
1196 break; 1223 break;
1224 }
1197 } 1225 }
1198 pagevec_release(&pvec); 1226 pagevec_release(&pvec);
1199 } 1227 }
@@ -1244,7 +1272,7 @@ int ext4_map_blocks(handle_t *handle, struct inode *inode,
1244 up_read((&EXT4_I(inode)->i_data_sem)); 1272 up_read((&EXT4_I(inode)->i_data_sem));
1245 1273
1246 if (retval > 0 && map->m_flags & EXT4_MAP_MAPPED) { 1274 if (retval > 0 && map->m_flags & EXT4_MAP_MAPPED) {
1247 int ret = check_block_validity(inode, __func__, map); 1275 int ret = check_block_validity(inode, map);
1248 if (ret != 0) 1276 if (ret != 0)
1249 return ret; 1277 return ret;
1250 } 1278 }
@@ -1324,9 +1352,7 @@ int ext4_map_blocks(handle_t *handle, struct inode *inode,
1324 1352
1325 up_write((&EXT4_I(inode)->i_data_sem)); 1353 up_write((&EXT4_I(inode)->i_data_sem));
1326 if (retval > 0 && map->m_flags & EXT4_MAP_MAPPED) { 1354 if (retval > 0 && map->m_flags & EXT4_MAP_MAPPED) {
1327 int ret = check_block_validity(inode, 1355 int ret = check_block_validity(inode, map);
1328 "ext4_map_blocks_after_alloc",
1329 map);
1330 if (ret != 0) 1356 if (ret != 0)
1331 return ret; 1357 return ret;
1332 } 1358 }
@@ -1519,9 +1545,25 @@ static int walk_page_buffers(handle_t *handle,
1519static int do_journal_get_write_access(handle_t *handle, 1545static int do_journal_get_write_access(handle_t *handle,
1520 struct buffer_head *bh) 1546 struct buffer_head *bh)
1521{ 1547{
1548 int dirty = buffer_dirty(bh);
1549 int ret;
1550
1522 if (!buffer_mapped(bh) || buffer_freed(bh)) 1551 if (!buffer_mapped(bh) || buffer_freed(bh))
1523 return 0; 1552 return 0;
1524 return ext4_journal_get_write_access(handle, bh); 1553 /*
1554 * __block_write_begin() could have dirtied some buffers. Clean
1555 * the dirty bit as jbd2_journal_get_write_access() could complain
1556 * otherwise about fs integrity issues. Setting of the dirty bit
1557 * by __block_write_begin() isn't a real problem here as we clear
1558 * the bit before releasing a page lock and thus writeback cannot
1559 * ever write the buffer.
1560 */
1561 if (dirty)
1562 clear_buffer_dirty(bh);
1563 ret = ext4_journal_get_write_access(handle, bh);
1564 if (!ret && dirty)
1565 ret = ext4_handle_dirty_metadata(handle, NULL, bh);
1566 return ret;
1525} 1567}
1526 1568
1527/* 1569/*
@@ -1578,11 +1620,9 @@ retry:
1578 *pagep = page; 1620 *pagep = page;
1579 1621
1580 if (ext4_should_dioread_nolock(inode)) 1622 if (ext4_should_dioread_nolock(inode))
1581 ret = block_write_begin(file, mapping, pos, len, flags, pagep, 1623 ret = __block_write_begin(page, pos, len, ext4_get_block_write);
1582 fsdata, ext4_get_block_write);
1583 else 1624 else
1584 ret = block_write_begin(file, mapping, pos, len, flags, pagep, 1625 ret = __block_write_begin(page, pos, len, ext4_get_block);
1585 fsdata, ext4_get_block);
1586 1626
1587 if (!ret && ext4_should_journal_data(inode)) { 1627 if (!ret && ext4_should_journal_data(inode)) {
1588 ret = walk_page_buffers(handle, page_buffers(page), 1628 ret = walk_page_buffers(handle, page_buffers(page),
@@ -1593,7 +1633,7 @@ retry:
1593 unlock_page(page); 1633 unlock_page(page);
1594 page_cache_release(page); 1634 page_cache_release(page);
1595 /* 1635 /*
1596 * block_write_begin may have instantiated a few blocks 1636 * __block_write_begin may have instantiated a few blocks
1597 * outside i_size. Trim these off again. Don't need 1637 * outside i_size. Trim these off again. Don't need
1598 * i_size_read because we hold i_mutex. 1638 * i_size_read because we hold i_mutex.
1599 * 1639 *
@@ -1968,16 +2008,23 @@ static void ext4_da_page_release_reservation(struct page *page,
1968 * 2008 *
1969 * As pages are already locked by write_cache_pages(), we can't use it 2009 * As pages are already locked by write_cache_pages(), we can't use it
1970 */ 2010 */
1971static int mpage_da_submit_io(struct mpage_da_data *mpd) 2011static int mpage_da_submit_io(struct mpage_da_data *mpd,
2012 struct ext4_map_blocks *map)
1972{ 2013{
1973 long pages_skipped;
1974 struct pagevec pvec; 2014 struct pagevec pvec;
1975 unsigned long index, end; 2015 unsigned long index, end;
1976 int ret = 0, err, nr_pages, i; 2016 int ret = 0, err, nr_pages, i;
1977 struct inode *inode = mpd->inode; 2017 struct inode *inode = mpd->inode;
1978 struct address_space *mapping = inode->i_mapping; 2018 struct address_space *mapping = inode->i_mapping;
2019 loff_t size = i_size_read(inode);
2020 unsigned int len, block_start;
2021 struct buffer_head *bh, *page_bufs = NULL;
2022 int journal_data = ext4_should_journal_data(inode);
2023 sector_t pblock = 0, cur_logical = 0;
2024 struct ext4_io_submit io_submit;
1979 2025
1980 BUG_ON(mpd->next_page <= mpd->first_page); 2026 BUG_ON(mpd->next_page <= mpd->first_page);
2027 memset(&io_submit, 0, sizeof(io_submit));
1981 /* 2028 /*
1982 * We need to start from the first_page to the next_page - 1 2029 * We need to start from the first_page to the next_page - 1
1983 * to make sure we also write the mapped dirty buffer_heads. 2030 * to make sure we also write the mapped dirty buffer_heads.
@@ -1993,122 +2040,108 @@ static int mpage_da_submit_io(struct mpage_da_data *mpd)
1993 if (nr_pages == 0) 2040 if (nr_pages == 0)
1994 break; 2041 break;
1995 for (i = 0; i < nr_pages; i++) { 2042 for (i = 0; i < nr_pages; i++) {
2043 int commit_write = 0, redirty_page = 0;
1996 struct page *page = pvec.pages[i]; 2044 struct page *page = pvec.pages[i];
1997 2045
1998 index = page->index; 2046 index = page->index;
1999 if (index > end) 2047 if (index > end)
2000 break; 2048 break;
2049
2050 if (index == size >> PAGE_CACHE_SHIFT)
2051 len = size & ~PAGE_CACHE_MASK;
2052 else
2053 len = PAGE_CACHE_SIZE;
2054 if (map) {
2055 cur_logical = index << (PAGE_CACHE_SHIFT -
2056 inode->i_blkbits);
2057 pblock = map->m_pblk + (cur_logical -
2058 map->m_lblk);
2059 }
2001 index++; 2060 index++;
2002 2061
2003 BUG_ON(!PageLocked(page)); 2062 BUG_ON(!PageLocked(page));
2004 BUG_ON(PageWriteback(page)); 2063 BUG_ON(PageWriteback(page));
2005 2064
2006 pages_skipped = mpd->wbc->pages_skipped;
2007 err = mapping->a_ops->writepage(page, mpd->wbc);
2008 if (!err && (pages_skipped == mpd->wbc->pages_skipped))
2009 /*
2010 * have successfully written the page
2011 * without skipping the same
2012 */
2013 mpd->pages_written++;
2014 /* 2065 /*
2015 * In error case, we have to continue because 2066 * If the page does not have buffers (for
2016 * remaining pages are still locked 2067 * whatever reason), try to create them using
2017 * XXX: unlock and re-dirty them? 2068 * __block_write_begin. If this fails,
2069 * redirty the page and move on.
2018 */ 2070 */
2019 if (ret == 0) 2071 if (!page_has_buffers(page)) {
2020 ret = err; 2072 if (__block_write_begin(page, 0, len,
2021 } 2073 noalloc_get_block_write)) {
2022 pagevec_release(&pvec); 2074 redirty_page:
2023 } 2075 redirty_page_for_writepage(mpd->wbc,
2024 return ret; 2076 page);
2025} 2077 unlock_page(page);
2026 2078 continue;
2027/* 2079 }
2028 * mpage_put_bnr_to_bhs - walk blocks and assign them actual numbers 2080 commit_write = 1;
2029 * 2081 }
2030 * the function goes through all passed space and put actual disk
2031 * block numbers into buffer heads, dropping BH_Delay and BH_Unwritten
2032 */
2033static void mpage_put_bnr_to_bhs(struct mpage_da_data *mpd,
2034 struct ext4_map_blocks *map)
2035{
2036 struct inode *inode = mpd->inode;
2037 struct address_space *mapping = inode->i_mapping;
2038 int blocks = map->m_len;
2039 sector_t pblock = map->m_pblk, cur_logical;
2040 struct buffer_head *head, *bh;
2041 pgoff_t index, end;
2042 struct pagevec pvec;
2043 int nr_pages, i;
2044
2045 index = map->m_lblk >> (PAGE_CACHE_SHIFT - inode->i_blkbits);
2046 end = (map->m_lblk + blocks - 1) >> (PAGE_CACHE_SHIFT - inode->i_blkbits);
2047 cur_logical = index << (PAGE_CACHE_SHIFT - inode->i_blkbits);
2048
2049 pagevec_init(&pvec, 0);
2050
2051 while (index <= end) {
2052 /* XXX: optimize tail */
2053 nr_pages = pagevec_lookup(&pvec, mapping, index, PAGEVEC_SIZE);
2054 if (nr_pages == 0)
2055 break;
2056 for (i = 0; i < nr_pages; i++) {
2057 struct page *page = pvec.pages[i];
2058
2059 index = page->index;
2060 if (index > end)
2061 break;
2062 index++;
2063
2064 BUG_ON(!PageLocked(page));
2065 BUG_ON(PageWriteback(page));
2066 BUG_ON(!page_has_buffers(page));
2067
2068 bh = page_buffers(page);
2069 head = bh;
2070
2071 /* skip blocks out of the range */
2072 do {
2073 if (cur_logical >= map->m_lblk)
2074 break;
2075 cur_logical++;
2076 } while ((bh = bh->b_this_page) != head);
2077 2082
2083 bh = page_bufs = page_buffers(page);
2084 block_start = 0;
2078 do { 2085 do {
2079 if (cur_logical >= map->m_lblk + blocks) 2086 if (!bh)
2080 break; 2087 goto redirty_page;
2081 2088 if (map && (cur_logical >= map->m_lblk) &&
2082 if (buffer_delay(bh) || buffer_unwritten(bh)) { 2089 (cur_logical <= (map->m_lblk +
2083 2090 (map->m_len - 1)))) {
2084 BUG_ON(bh->b_bdev != inode->i_sb->s_bdev);
2085
2086 if (buffer_delay(bh)) { 2091 if (buffer_delay(bh)) {
2087 clear_buffer_delay(bh); 2092 clear_buffer_delay(bh);
2088 bh->b_blocknr = pblock; 2093 bh->b_blocknr = pblock;
2089 } else {
2090 /*
2091 * unwritten already should have
2092 * blocknr assigned. Verify that
2093 */
2094 clear_buffer_unwritten(bh);
2095 BUG_ON(bh->b_blocknr != pblock);
2096 } 2094 }
2095 if (buffer_unwritten(bh) ||
2096 buffer_mapped(bh))
2097 BUG_ON(bh->b_blocknr != pblock);
2098 if (map->m_flags & EXT4_MAP_UNINIT)
2099 set_buffer_uninit(bh);
2100 clear_buffer_unwritten(bh);
2101 }
2097 2102
2098 } else if (buffer_mapped(bh)) 2103 /* redirty page if block allocation undone */
2099 BUG_ON(bh->b_blocknr != pblock); 2104 if (buffer_delay(bh) || buffer_unwritten(bh))
2100 2105 redirty_page = 1;
2101 if (map->m_flags & EXT4_MAP_UNINIT) 2106 bh = bh->b_this_page;
2102 set_buffer_uninit(bh); 2107 block_start += bh->b_size;
2103 cur_logical++; 2108 cur_logical++;
2104 pblock++; 2109 pblock++;
2105 } while ((bh = bh->b_this_page) != head); 2110 } while (bh != page_bufs);
2111
2112 if (redirty_page)
2113 goto redirty_page;
2114
2115 if (commit_write)
2116 /* mark the buffer_heads as dirty & uptodate */
2117 block_commit_write(page, 0, len);
2118
2119 /*
2120 * Delalloc doesn't support data journalling,
2121 * but eventually maybe we'll lift this
2122 * restriction.
2123 */
2124 if (unlikely(journal_data && PageChecked(page)))
2125 err = __ext4_journalled_writepage(page, len);
2126 else
2127 err = ext4_bio_write_page(&io_submit, page,
2128 len, mpd->wbc);
2129
2130 if (!err)
2131 mpd->pages_written++;
2132 /*
2133 * In error case, we have to continue because
2134 * remaining pages are still locked
2135 */
2136 if (ret == 0)
2137 ret = err;
2106 } 2138 }
2107 pagevec_release(&pvec); 2139 pagevec_release(&pvec);
2108 } 2140 }
2141 ext4_io_submit(&io_submit);
2142 return ret;
2109} 2143}
2110 2144
2111
2112static void ext4_da_block_invalidatepages(struct mpage_da_data *mpd, 2145static void ext4_da_block_invalidatepages(struct mpage_da_data *mpd,
2113 sector_t logical, long blk_cnt) 2146 sector_t logical, long blk_cnt)
2114{ 2147{
@@ -2160,41 +2193,38 @@ static void ext4_print_free_blocks(struct inode *inode)
2160} 2193}
2161 2194
2162/* 2195/*
2163 * mpage_da_map_blocks - go through given space 2196 * mpage_da_map_and_submit - go through given space, map them
2197 * if necessary, and then submit them for I/O
2164 * 2198 *
2165 * @mpd - bh describing space 2199 * @mpd - bh describing space
2166 * 2200 *
2167 * The function skips space we know is already mapped to disk blocks. 2201 * The function skips space we know is already mapped to disk blocks.
2168 * 2202 *
2169 */ 2203 */
2170static int mpage_da_map_blocks(struct mpage_da_data *mpd) 2204static void mpage_da_map_and_submit(struct mpage_da_data *mpd)
2171{ 2205{
2172 int err, blks, get_blocks_flags; 2206 int err, blks, get_blocks_flags;
2173 struct ext4_map_blocks map; 2207 struct ext4_map_blocks map, *mapp = NULL;
2174 sector_t next = mpd->b_blocknr; 2208 sector_t next = mpd->b_blocknr;
2175 unsigned max_blocks = mpd->b_size >> mpd->inode->i_blkbits; 2209 unsigned max_blocks = mpd->b_size >> mpd->inode->i_blkbits;
2176 loff_t disksize = EXT4_I(mpd->inode)->i_disksize; 2210 loff_t disksize = EXT4_I(mpd->inode)->i_disksize;
2177 handle_t *handle = NULL; 2211 handle_t *handle = NULL;
2178 2212
2179 /* 2213 /*
2180 * We consider only non-mapped and non-allocated blocks 2214 * If the blocks are mapped already, or we couldn't accumulate
2181 */ 2215 * any blocks, then proceed immediately to the submission stage.
2182 if ((mpd->b_state & (1 << BH_Mapped)) &&
2183 !(mpd->b_state & (1 << BH_Delay)) &&
2184 !(mpd->b_state & (1 << BH_Unwritten)))
2185 return 0;
2186
2187 /*
2188 * If we didn't accumulate anything to write simply return
2189 */ 2216 */
2190 if (!mpd->b_size) 2217 if ((mpd->b_size == 0) ||
2191 return 0; 2218 ((mpd->b_state & (1 << BH_Mapped)) &&
2219 !(mpd->b_state & (1 << BH_Delay)) &&
2220 !(mpd->b_state & (1 << BH_Unwritten))))
2221 goto submit_io;
2192 2222
2193 handle = ext4_journal_current_handle(); 2223 handle = ext4_journal_current_handle();
2194 BUG_ON(!handle); 2224 BUG_ON(!handle);
2195 2225
2196 /* 2226 /*
2197 * Call ext4_get_blocks() to allocate any delayed allocation 2227 * Call ext4_map_blocks() to allocate any delayed allocation
2198 * blocks, or to convert an uninitialized extent to be 2228 * blocks, or to convert an uninitialized extent to be
2199 * initialized (in the case where we have written into 2229 * initialized (in the case where we have written into
2200 * one or more preallocated blocks). 2230 * one or more preallocated blocks).
@@ -2203,7 +2233,7 @@ static int mpage_da_map_blocks(struct mpage_da_data *mpd)
2203 * indicate that we are on the delayed allocation path. This 2233 * indicate that we are on the delayed allocation path. This
2204 * affects functions in many different parts of the allocation 2234 * affects functions in many different parts of the allocation
2205 * call path. This flag exists primarily because we don't 2235 * call path. This flag exists primarily because we don't
2206 * want to change *many* call functions, so ext4_get_blocks() 2236 * want to change *many* call functions, so ext4_map_blocks()
2207 * will set the magic i_delalloc_reserved_flag once the 2237 * will set the magic i_delalloc_reserved_flag once the
2208 * inode's allocation semaphore is taken. 2238 * inode's allocation semaphore is taken.
2209 * 2239 *
@@ -2221,19 +2251,22 @@ static int mpage_da_map_blocks(struct mpage_da_data *mpd)
2221 2251
2222 blks = ext4_map_blocks(handle, mpd->inode, &map, get_blocks_flags); 2252 blks = ext4_map_blocks(handle, mpd->inode, &map, get_blocks_flags);
2223 if (blks < 0) { 2253 if (blks < 0) {
2254 struct super_block *sb = mpd->inode->i_sb;
2255
2224 err = blks; 2256 err = blks;
2225 /* 2257 /*
2226 * If get block returns with error we simply 2258 * If get block returns EAGAIN or ENOSPC and there
2227 * return. Later writepage will redirty the page and 2259 * appears to be free blocks we will call
2228 * writepages will find the dirty page again 2260 * ext4_writepage() for all of the pages which will
2261 * just redirty the pages.
2229 */ 2262 */
2230 if (err == -EAGAIN) 2263 if (err == -EAGAIN)
2231 return 0; 2264 goto submit_io;
2232 2265
2233 if (err == -ENOSPC && 2266 if (err == -ENOSPC &&
2234 ext4_count_free_blocks(mpd->inode->i_sb)) { 2267 ext4_count_free_blocks(sb)) {
2235 mpd->retval = err; 2268 mpd->retval = err;
2236 return 0; 2269 goto submit_io;
2237 } 2270 }
2238 2271
2239 /* 2272 /*
@@ -2243,24 +2276,26 @@ static int mpage_da_map_blocks(struct mpage_da_data *mpd)
2243 * writepage and writepages will again try to write 2276 * writepage and writepages will again try to write
2244 * the same. 2277 * the same.
2245 */ 2278 */
2246 ext4_msg(mpd->inode->i_sb, KERN_CRIT, 2279 if (!(EXT4_SB(sb)->s_mount_flags & EXT4_MF_FS_ABORTED)) {
2247 "delayed block allocation failed for inode %lu at " 2280 ext4_msg(sb, KERN_CRIT,
2248 "logical offset %llu with max blocks %zd with " 2281 "delayed block allocation failed for inode %lu "
2249 "error %d", mpd->inode->i_ino, 2282 "at logical offset %llu with max blocks %zd "
2250 (unsigned long long) next, 2283 "with error %d", mpd->inode->i_ino,
2251 mpd->b_size >> mpd->inode->i_blkbits, err); 2284 (unsigned long long) next,
2252 printk(KERN_CRIT "This should not happen!! " 2285 mpd->b_size >> mpd->inode->i_blkbits, err);
2253 "Data will be lost\n"); 2286 ext4_msg(sb, KERN_CRIT,
2254 if (err == -ENOSPC) { 2287 "This should not happen!! Data will be lost\n");
2255 ext4_print_free_blocks(mpd->inode); 2288 if (err == -ENOSPC)
2289 ext4_print_free_blocks(mpd->inode);
2256 } 2290 }
2257 /* invalidate all the pages */ 2291 /* invalidate all the pages */
2258 ext4_da_block_invalidatepages(mpd, next, 2292 ext4_da_block_invalidatepages(mpd, next,
2259 mpd->b_size >> mpd->inode->i_blkbits); 2293 mpd->b_size >> mpd->inode->i_blkbits);
2260 return err; 2294 return;
2261 } 2295 }
2262 BUG_ON(blks == 0); 2296 BUG_ON(blks == 0);
2263 2297
2298 mapp = &map;
2264 if (map.m_flags & EXT4_MAP_NEW) { 2299 if (map.m_flags & EXT4_MAP_NEW) {
2265 struct block_device *bdev = mpd->inode->i_sb->s_bdev; 2300 struct block_device *bdev = mpd->inode->i_sb->s_bdev;
2266 int i; 2301 int i;
@@ -2269,18 +2304,11 @@ static int mpage_da_map_blocks(struct mpage_da_data *mpd)
2269 unmap_underlying_metadata(bdev, map.m_pblk + i); 2304 unmap_underlying_metadata(bdev, map.m_pblk + i);
2270 } 2305 }
2271 2306
2272 /*
2273 * If blocks are delayed marked, we need to
2274 * put actual blocknr and drop delayed bit
2275 */
2276 if ((mpd->b_state & (1 << BH_Delay)) ||
2277 (mpd->b_state & (1 << BH_Unwritten)))
2278 mpage_put_bnr_to_bhs(mpd, &map);
2279
2280 if (ext4_should_order_data(mpd->inode)) { 2307 if (ext4_should_order_data(mpd->inode)) {
2281 err = ext4_jbd2_file_inode(handle, mpd->inode); 2308 err = ext4_jbd2_file_inode(handle, mpd->inode);
2282 if (err) 2309 if (err)
2283 return err; 2310 /* This only happens if the journal is aborted */
2311 return;
2284 } 2312 }
2285 2313
2286 /* 2314 /*
@@ -2291,10 +2319,16 @@ static int mpage_da_map_blocks(struct mpage_da_data *mpd)
2291 disksize = i_size_read(mpd->inode); 2319 disksize = i_size_read(mpd->inode);
2292 if (disksize > EXT4_I(mpd->inode)->i_disksize) { 2320 if (disksize > EXT4_I(mpd->inode)->i_disksize) {
2293 ext4_update_i_disksize(mpd->inode, disksize); 2321 ext4_update_i_disksize(mpd->inode, disksize);
2294 return ext4_mark_inode_dirty(handle, mpd->inode); 2322 err = ext4_mark_inode_dirty(handle, mpd->inode);
2323 if (err)
2324 ext4_error(mpd->inode->i_sb,
2325 "Failed to mark inode %lu dirty",
2326 mpd->inode->i_ino);
2295 } 2327 }
2296 2328
2297 return 0; 2329submit_io:
2330 mpage_da_submit_io(mpd, mapp);
2331 mpd->io_done = 1;
2298} 2332}
2299 2333
2300#define BH_FLAGS ((1 << BH_Uptodate) | (1 << BH_Mapped) | \ 2334#define BH_FLAGS ((1 << BH_Uptodate) | (1 << BH_Mapped) | \
@@ -2320,7 +2354,7 @@ static void mpage_add_bh_to_extent(struct mpage_da_data *mpd,
2320 * XXX Don't go larger than mballoc is willing to allocate 2354 * XXX Don't go larger than mballoc is willing to allocate
2321 * This is a stopgap solution. We eventually need to fold 2355 * This is a stopgap solution. We eventually need to fold
2322 * mpage_da_submit_io() into this function and then call 2356 * mpage_da_submit_io() into this function and then call
2323 * ext4_get_blocks() multiple times in a loop 2357 * ext4_map_blocks() multiple times in a loop
2324 */ 2358 */
2325 if (nrblocks >= 8*1024*1024/mpd->inode->i_sb->s_blocksize) 2359 if (nrblocks >= 8*1024*1024/mpd->inode->i_sb->s_blocksize)
2326 goto flush_it; 2360 goto flush_it;
@@ -2371,9 +2405,7 @@ flush_it:
2371 * We couldn't merge the block to our extent, so we 2405 * We couldn't merge the block to our extent, so we
2372 * need to flush current extent and start new one 2406 * need to flush current extent and start new one
2373 */ 2407 */
2374 if (mpage_da_map_blocks(mpd) == 0) 2408 mpage_da_map_and_submit(mpd);
2375 mpage_da_submit_io(mpd);
2376 mpd->io_done = 1;
2377 return; 2409 return;
2378} 2410}
2379 2411
@@ -2392,9 +2424,9 @@ static int ext4_bh_delay_or_unwritten(handle_t *handle, struct buffer_head *bh)
2392 * The function finds extents of pages and scan them for all blocks. 2424 * The function finds extents of pages and scan them for all blocks.
2393 */ 2425 */
2394static int __mpage_da_writepage(struct page *page, 2426static int __mpage_da_writepage(struct page *page,
2395 struct writeback_control *wbc, void *data) 2427 struct writeback_control *wbc,
2428 struct mpage_da_data *mpd)
2396{ 2429{
2397 struct mpage_da_data *mpd = data;
2398 struct inode *inode = mpd->inode; 2430 struct inode *inode = mpd->inode;
2399 struct buffer_head *bh, *head; 2431 struct buffer_head *bh, *head;
2400 sector_t logical; 2432 sector_t logical;
@@ -2405,15 +2437,13 @@ static int __mpage_da_writepage(struct page *page,
2405 if (mpd->next_page != page->index) { 2437 if (mpd->next_page != page->index) {
2406 /* 2438 /*
2407 * Nope, we can't. So, we map non-allocated blocks 2439 * Nope, we can't. So, we map non-allocated blocks
2408 * and start IO on them using writepage() 2440 * and start IO on them
2409 */ 2441 */
2410 if (mpd->next_page != mpd->first_page) { 2442 if (mpd->next_page != mpd->first_page) {
2411 if (mpage_da_map_blocks(mpd) == 0) 2443 mpage_da_map_and_submit(mpd);
2412 mpage_da_submit_io(mpd);
2413 /* 2444 /*
2414 * skip rest of the page in the page_vec 2445 * skip rest of the page in the page_vec
2415 */ 2446 */
2416 mpd->io_done = 1;
2417 redirty_page_for_writepage(wbc, page); 2447 redirty_page_for_writepage(wbc, page);
2418 unlock_page(page); 2448 unlock_page(page);
2419 return MPAGE_DA_EXTENT_TAIL; 2449 return MPAGE_DA_EXTENT_TAIL;
@@ -2520,8 +2550,7 @@ static int ext4_da_get_block_prep(struct inode *inode, sector_t iblock,
2520 if (buffer_delay(bh)) 2550 if (buffer_delay(bh))
2521 return 0; /* Not sure this could or should happen */ 2551 return 0; /* Not sure this could or should happen */
2522 /* 2552 /*
2523 * XXX: __block_prepare_write() unmaps passed block, 2553 * XXX: __block_write_begin() unmaps passed block, is it OK?
2524 * is it OK?
2525 */ 2554 */
2526 ret = ext4_da_reserve_space(inode, iblock); 2555 ret = ext4_da_reserve_space(inode, iblock);
2527 if (ret) 2556 if (ret)
@@ -2553,18 +2582,16 @@ static int ext4_da_get_block_prep(struct inode *inode, sector_t iblock,
2553/* 2582/*
2554 * This function is used as a standard get_block_t calback function 2583 * This function is used as a standard get_block_t calback function
2555 * when there is no desire to allocate any blocks. It is used as a 2584 * when there is no desire to allocate any blocks. It is used as a
2556 * callback function for block_prepare_write(), nobh_writepage(), and 2585 * callback function for block_write_begin() and block_write_full_page().
2557 * block_write_full_page(). These functions should only try to map a 2586 * These functions should only try to map a single block at a time.
2558 * single block at a time.
2559 * 2587 *
2560 * Since this function doesn't do block allocations even if the caller 2588 * Since this function doesn't do block allocations even if the caller
2561 * requests it by passing in create=1, it is critically important that 2589 * requests it by passing in create=1, it is critically important that
2562 * any caller checks to make sure that any buffer heads are returned 2590 * any caller checks to make sure that any buffer heads are returned
2563 * by this function are either all already mapped or marked for 2591 * by this function are either all already mapped or marked for
2564 * delayed allocation before calling nobh_writepage() or 2592 * delayed allocation before calling block_write_full_page(). Otherwise,
2565 * block_write_full_page(). Otherwise, b_blocknr could be left 2593 * b_blocknr could be left unitialized, and the page write functions will
2566 * unitialized, and the page write functions will be taken by 2594 * be taken by surprise.
2567 * surprise.
2568 */ 2595 */
2569static int noalloc_get_block_write(struct inode *inode, sector_t iblock, 2596static int noalloc_get_block_write(struct inode *inode, sector_t iblock,
2570 struct buffer_head *bh_result, int create) 2597 struct buffer_head *bh_result, int create)
@@ -2595,6 +2622,7 @@ static int __ext4_journalled_writepage(struct page *page,
2595 int ret = 0; 2622 int ret = 0;
2596 int err; 2623 int err;
2597 2624
2625 ClearPageChecked(page);
2598 page_bufs = page_buffers(page); 2626 page_bufs = page_buffers(page);
2599 BUG_ON(!page_bufs); 2627 BUG_ON(!page_bufs);
2600 walk_page_buffers(handle, page_bufs, 0, len, NULL, bget_one); 2628 walk_page_buffers(handle, page_bufs, 0, len, NULL, bget_one);
@@ -2672,7 +2700,7 @@ static void ext4_end_io_buffer_write(struct buffer_head *bh, int uptodate);
2672static int ext4_writepage(struct page *page, 2700static int ext4_writepage(struct page *page,
2673 struct writeback_control *wbc) 2701 struct writeback_control *wbc)
2674{ 2702{
2675 int ret = 0; 2703 int ret = 0, commit_write = 0;
2676 loff_t size; 2704 loff_t size;
2677 unsigned int len; 2705 unsigned int len;
2678 struct buffer_head *page_bufs = NULL; 2706 struct buffer_head *page_bufs = NULL;
@@ -2685,73 +2713,46 @@ static int ext4_writepage(struct page *page,
2685 else 2713 else
2686 len = PAGE_CACHE_SIZE; 2714 len = PAGE_CACHE_SIZE;
2687 2715
2688 if (page_has_buffers(page)) { 2716 /*
2689 page_bufs = page_buffers(page); 2717 * If the page does not have buffers (for whatever reason),
2690 if (walk_page_buffers(NULL, page_bufs, 0, len, NULL, 2718 * try to create them using __block_write_begin. If this
2691 ext4_bh_delay_or_unwritten)) { 2719 * fails, redirty the page and move on.
2692 /* 2720 */
2693 * We don't want to do block allocation 2721 if (!page_buffers(page)) {
2694 * So redirty the page and return 2722 if (__block_write_begin(page, 0, len,
2695 * We may reach here when we do a journal commit 2723 noalloc_get_block_write)) {
2696 * via journal_submit_inode_data_buffers. 2724 redirty_page:
2697 * If we don't have mapping block we just ignore
2698 * them. We can also reach here via shrink_page_list
2699 */
2700 redirty_page_for_writepage(wbc, page); 2725 redirty_page_for_writepage(wbc, page);
2701 unlock_page(page); 2726 unlock_page(page);
2702 return 0; 2727 return 0;
2703 } 2728 }
2704 } else { 2729 commit_write = 1;
2730 }
2731 page_bufs = page_buffers(page);
2732 if (walk_page_buffers(NULL, page_bufs, 0, len, NULL,
2733 ext4_bh_delay_or_unwritten)) {
2705 /* 2734 /*
2706 * The test for page_has_buffers() is subtle: 2735 * We don't want to do block allocation So redirty the
2707 * We know the page is dirty but it lost buffers. That means 2736 * page and return We may reach here when we do a
2708 * that at some moment in time after write_begin()/write_end() 2737 * journal commit via
2709 * has been called all buffers have been clean and thus they 2738 * journal_submit_inode_data_buffers. If we don't
2710 * must have been written at least once. So they are all 2739 * have mapping block we just ignore them. We can also
2711 * mapped and we can happily proceed with mapping them 2740 * reach here via shrink_page_list
2712 * and writing the page.
2713 *
2714 * Try to initialize the buffer_heads and check whether
2715 * all are mapped and non delay. We don't want to
2716 * do block allocation here.
2717 */ 2741 */
2718 ret = block_prepare_write(page, 0, len, 2742 goto redirty_page;
2719 noalloc_get_block_write); 2743 }
2720 if (!ret) { 2744 if (commit_write)
2721 page_bufs = page_buffers(page);
2722 /* check whether all are mapped and non delay */
2723 if (walk_page_buffers(NULL, page_bufs, 0, len, NULL,
2724 ext4_bh_delay_or_unwritten)) {
2725 redirty_page_for_writepage(wbc, page);
2726 unlock_page(page);
2727 return 0;
2728 }
2729 } else {
2730 /*
2731 * We can't do block allocation here
2732 * so just redity the page and unlock
2733 * and return
2734 */
2735 redirty_page_for_writepage(wbc, page);
2736 unlock_page(page);
2737 return 0;
2738 }
2739 /* now mark the buffer_heads as dirty and uptodate */ 2745 /* now mark the buffer_heads as dirty and uptodate */
2740 block_commit_write(page, 0, len); 2746 block_commit_write(page, 0, len);
2741 }
2742 2747
2743 if (PageChecked(page) && ext4_should_journal_data(inode)) { 2748 if (PageChecked(page) && ext4_should_journal_data(inode))
2744 /* 2749 /*
2745 * It's mmapped pagecache. Add buffers and journal it. There 2750 * It's mmapped pagecache. Add buffers and journal it. There
2746 * doesn't seem much point in redirtying the page here. 2751 * doesn't seem much point in redirtying the page here.
2747 */ 2752 */
2748 ClearPageChecked(page);
2749 return __ext4_journalled_writepage(page, len); 2753 return __ext4_journalled_writepage(page, len);
2750 }
2751 2754
2752 if (test_opt(inode->i_sb, NOBH) && ext4_should_writeback_data(inode)) 2755 if (buffer_uninit(page_bufs)) {
2753 ret = nobh_writepage(page, noalloc_get_block_write, wbc);
2754 else if (page_bufs && buffer_uninit(page_bufs)) {
2755 ext4_set_bh_endio(page_bufs, inode); 2756 ext4_set_bh_endio(page_bufs, inode);
2756 ret = block_write_full_page_endio(page, noalloc_get_block_write, 2757 ret = block_write_full_page_endio(page, noalloc_get_block_write,
2757 wbc, ext4_end_io_buffer_write); 2758 wbc, ext4_end_io_buffer_write);
@@ -2798,25 +2799,32 @@ static int ext4_da_writepages_trans_blocks(struct inode *inode)
2798 */ 2799 */
2799static int write_cache_pages_da(struct address_space *mapping, 2800static int write_cache_pages_da(struct address_space *mapping,
2800 struct writeback_control *wbc, 2801 struct writeback_control *wbc,
2801 struct mpage_da_data *mpd) 2802 struct mpage_da_data *mpd,
2803 pgoff_t *done_index)
2802{ 2804{
2803 int ret = 0; 2805 int ret = 0;
2804 int done = 0; 2806 int done = 0;
2805 struct pagevec pvec; 2807 struct pagevec pvec;
2806 int nr_pages; 2808 unsigned nr_pages;
2807 pgoff_t index; 2809 pgoff_t index;
2808 pgoff_t end; /* Inclusive */ 2810 pgoff_t end; /* Inclusive */
2809 long nr_to_write = wbc->nr_to_write; 2811 long nr_to_write = wbc->nr_to_write;
2812 int tag;
2810 2813
2811 pagevec_init(&pvec, 0); 2814 pagevec_init(&pvec, 0);
2812 index = wbc->range_start >> PAGE_CACHE_SHIFT; 2815 index = wbc->range_start >> PAGE_CACHE_SHIFT;
2813 end = wbc->range_end >> PAGE_CACHE_SHIFT; 2816 end = wbc->range_end >> PAGE_CACHE_SHIFT;
2814 2817
2818 if (wbc->sync_mode == WB_SYNC_ALL)
2819 tag = PAGECACHE_TAG_TOWRITE;
2820 else
2821 tag = PAGECACHE_TAG_DIRTY;
2822
2823 *done_index = index;
2815 while (!done && (index <= end)) { 2824 while (!done && (index <= end)) {
2816 int i; 2825 int i;
2817 2826
2818 nr_pages = pagevec_lookup_tag(&pvec, mapping, &index, 2827 nr_pages = pagevec_lookup_tag(&pvec, mapping, &index, tag,
2819 PAGECACHE_TAG_DIRTY,
2820 min(end - index, (pgoff_t)PAGEVEC_SIZE-1) + 1); 2828 min(end - index, (pgoff_t)PAGEVEC_SIZE-1) + 1);
2821 if (nr_pages == 0) 2829 if (nr_pages == 0)
2822 break; 2830 break;
@@ -2836,6 +2844,8 @@ static int write_cache_pages_da(struct address_space *mapping,
2836 break; 2844 break;
2837 } 2845 }
2838 2846
2847 *done_index = page->index + 1;
2848
2839 lock_page(page); 2849 lock_page(page);
2840 2850
2841 /* 2851 /*
@@ -2921,6 +2931,8 @@ static int ext4_da_writepages(struct address_space *mapping,
2921 long desired_nr_to_write, nr_to_writebump = 0; 2931 long desired_nr_to_write, nr_to_writebump = 0;
2922 loff_t range_start = wbc->range_start; 2932 loff_t range_start = wbc->range_start;
2923 struct ext4_sb_info *sbi = EXT4_SB(mapping->host->i_sb); 2933 struct ext4_sb_info *sbi = EXT4_SB(mapping->host->i_sb);
2934 pgoff_t done_index = 0;
2935 pgoff_t end;
2924 2936
2925 trace_ext4_da_writepages(inode, wbc); 2937 trace_ext4_da_writepages(inode, wbc);
2926 2938
@@ -2956,8 +2968,11 @@ static int ext4_da_writepages(struct address_space *mapping,
2956 wbc->range_start = index << PAGE_CACHE_SHIFT; 2968 wbc->range_start = index << PAGE_CACHE_SHIFT;
2957 wbc->range_end = LLONG_MAX; 2969 wbc->range_end = LLONG_MAX;
2958 wbc->range_cyclic = 0; 2970 wbc->range_cyclic = 0;
2959 } else 2971 end = -1;
2972 } else {
2960 index = wbc->range_start >> PAGE_CACHE_SHIFT; 2973 index = wbc->range_start >> PAGE_CACHE_SHIFT;
2974 end = wbc->range_end >> PAGE_CACHE_SHIFT;
2975 }
2961 2976
2962 /* 2977 /*
2963 * This works around two forms of stupidity. The first is in 2978 * This works around two forms of stupidity. The first is in
@@ -2976,9 +2991,12 @@ static int ext4_da_writepages(struct address_space *mapping,
2976 * sbi->max_writeback_mb_bump whichever is smaller. 2991 * sbi->max_writeback_mb_bump whichever is smaller.
2977 */ 2992 */
2978 max_pages = sbi->s_max_writeback_mb_bump << (20 - PAGE_CACHE_SHIFT); 2993 max_pages = sbi->s_max_writeback_mb_bump << (20 - PAGE_CACHE_SHIFT);
2979 if (!range_cyclic && range_whole) 2994 if (!range_cyclic && range_whole) {
2980 desired_nr_to_write = wbc->nr_to_write * 8; 2995 if (wbc->nr_to_write == LONG_MAX)
2981 else 2996 desired_nr_to_write = wbc->nr_to_write;
2997 else
2998 desired_nr_to_write = wbc->nr_to_write * 8;
2999 } else
2982 desired_nr_to_write = ext4_num_dirty_pages(inode, index, 3000 desired_nr_to_write = ext4_num_dirty_pages(inode, index,
2983 max_pages); 3001 max_pages);
2984 if (desired_nr_to_write > max_pages) 3002 if (desired_nr_to_write > max_pages)
@@ -2995,6 +3013,9 @@ static int ext4_da_writepages(struct address_space *mapping,
2995 pages_skipped = wbc->pages_skipped; 3013 pages_skipped = wbc->pages_skipped;
2996 3014
2997retry: 3015retry:
3016 if (wbc->sync_mode == WB_SYNC_ALL)
3017 tag_pages_for_writeback(mapping, index, end);
3018
2998 while (!ret && wbc->nr_to_write > 0) { 3019 while (!ret && wbc->nr_to_write > 0) {
2999 3020
3000 /* 3021 /*
@@ -3033,16 +3054,14 @@ retry:
3033 mpd.io_done = 0; 3054 mpd.io_done = 0;
3034 mpd.pages_written = 0; 3055 mpd.pages_written = 0;
3035 mpd.retval = 0; 3056 mpd.retval = 0;
3036 ret = write_cache_pages_da(mapping, wbc, &mpd); 3057 ret = write_cache_pages_da(mapping, wbc, &mpd, &done_index);
3037 /* 3058 /*
3038 * If we have a contiguous extent of pages and we 3059 * If we have a contiguous extent of pages and we
3039 * haven't done the I/O yet, map the blocks and submit 3060 * haven't done the I/O yet, map the blocks and submit
3040 * them for I/O. 3061 * them for I/O.
3041 */ 3062 */
3042 if (!mpd.io_done && mpd.next_page != mpd.first_page) { 3063 if (!mpd.io_done && mpd.next_page != mpd.first_page) {
3043 if (mpage_da_map_blocks(&mpd) == 0) 3064 mpage_da_map_and_submit(&mpd);
3044 mpage_da_submit_io(&mpd);
3045 mpd.io_done = 1;
3046 ret = MPAGE_DA_EXTENT_TAIL; 3065 ret = MPAGE_DA_EXTENT_TAIL;
3047 } 3066 }
3048 trace_ext4_da_write_pages(inode, &mpd); 3067 trace_ext4_da_write_pages(inode, &mpd);
@@ -3089,14 +3108,13 @@ retry:
3089 __func__, wbc->nr_to_write, ret); 3108 __func__, wbc->nr_to_write, ret);
3090 3109
3091 /* Update index */ 3110 /* Update index */
3092 index += pages_written;
3093 wbc->range_cyclic = range_cyclic; 3111 wbc->range_cyclic = range_cyclic;
3094 if (wbc->range_cyclic || (range_whole && wbc->nr_to_write > 0)) 3112 if (wbc->range_cyclic || (range_whole && wbc->nr_to_write > 0))
3095 /* 3113 /*
3096 * set the writeback_index so that range_cyclic 3114 * set the writeback_index so that range_cyclic
3097 * mode will write it back later 3115 * mode will write it back later
3098 */ 3116 */
3099 mapping->writeback_index = index; 3117 mapping->writeback_index = done_index;
3100 3118
3101out_writepages: 3119out_writepages:
3102 wbc->nr_to_write -= nr_to_writebump; 3120 wbc->nr_to_write -= nr_to_writebump;
@@ -3146,13 +3164,10 @@ static int ext4_da_write_begin(struct file *file, struct address_space *mapping,
3146 int ret, retries = 0; 3164 int ret, retries = 0;
3147 struct page *page; 3165 struct page *page;
3148 pgoff_t index; 3166 pgoff_t index;
3149 unsigned from, to;
3150 struct inode *inode = mapping->host; 3167 struct inode *inode = mapping->host;
3151 handle_t *handle; 3168 handle_t *handle;
3152 3169
3153 index = pos >> PAGE_CACHE_SHIFT; 3170 index = pos >> PAGE_CACHE_SHIFT;
3154 from = pos & (PAGE_CACHE_SIZE - 1);
3155 to = from + len;
3156 3171
3157 if (ext4_nonda_switch(inode->i_sb)) { 3172 if (ext4_nonda_switch(inode->i_sb)) {
3158 *fsdata = (void *)FALL_BACK_TO_NONDELALLOC; 3173 *fsdata = (void *)FALL_BACK_TO_NONDELALLOC;
@@ -3185,8 +3200,7 @@ retry:
3185 } 3200 }
3186 *pagep = page; 3201 *pagep = page;
3187 3202
3188 ret = block_write_begin(file, mapping, pos, len, flags, pagep, fsdata, 3203 ret = __block_write_begin(page, pos, len, ext4_da_get_block_prep);
3189 ext4_da_get_block_prep);
3190 if (ret < 0) { 3204 if (ret < 0) {
3191 unlock_page(page); 3205 unlock_page(page);
3192 ext4_journal_stop(handle); 3206 ext4_journal_stop(handle);
@@ -3435,15 +3449,6 @@ ext4_readpages(struct file *file, struct address_space *mapping,
3435 return mpage_readpages(mapping, pages, nr_pages, ext4_get_block); 3449 return mpage_readpages(mapping, pages, nr_pages, ext4_get_block);
3436} 3450}
3437 3451
3438static void ext4_free_io_end(ext4_io_end_t *io)
3439{
3440 BUG_ON(!io);
3441 if (io->page)
3442 put_page(io->page);
3443 iput(io->inode);
3444 kfree(io);
3445}
3446
3447static void ext4_invalidatepage_free_endio(struct page *page, unsigned long offset) 3452static void ext4_invalidatepage_free_endio(struct page *page, unsigned long offset)
3448{ 3453{
3449 struct buffer_head *head, *bh; 3454 struct buffer_head *head, *bh;
@@ -3545,15 +3550,24 @@ static ssize_t ext4_ind_direct_IO(int rw, struct kiocb *iocb,
3545 3550
3546retry: 3551retry:
3547 if (rw == READ && ext4_should_dioread_nolock(inode)) 3552 if (rw == READ && ext4_should_dioread_nolock(inode))
3548 ret = blockdev_direct_IO_no_locking(rw, iocb, inode, 3553 ret = __blockdev_direct_IO(rw, iocb, inode,
3549 inode->i_sb->s_bdev, iov, 3554 inode->i_sb->s_bdev, iov,
3550 offset, nr_segs, 3555 offset, nr_segs,
3551 ext4_get_block, NULL); 3556 ext4_get_block, NULL, NULL, 0);
3552 else 3557 else {
3553 ret = blockdev_direct_IO(rw, iocb, inode, 3558 ret = blockdev_direct_IO(rw, iocb, inode,
3554 inode->i_sb->s_bdev, iov, 3559 inode->i_sb->s_bdev, iov,
3555 offset, nr_segs, 3560 offset, nr_segs,
3556 ext4_get_block, NULL); 3561 ext4_get_block, NULL);
3562
3563 if (unlikely((rw & WRITE) && ret < 0)) {
3564 loff_t isize = i_size_read(inode);
3565 loff_t end = offset + iov_length(iov, nr_segs);
3566
3567 if (end > isize)
3568 vmtruncate(inode, isize);
3569 }
3570 }
3557 if (ret == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries)) 3571 if (ret == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries))
3558 goto retry; 3572 goto retry;
3559 3573
@@ -3611,171 +3625,9 @@ static int ext4_get_block_write(struct inode *inode, sector_t iblock,
3611 EXT4_GET_BLOCKS_IO_CREATE_EXT); 3625 EXT4_GET_BLOCKS_IO_CREATE_EXT);
3612} 3626}
3613 3627
3614static void dump_completed_IO(struct inode * inode)
3615{
3616#ifdef EXT4_DEBUG
3617 struct list_head *cur, *before, *after;
3618 ext4_io_end_t *io, *io0, *io1;
3619 unsigned long flags;
3620
3621 if (list_empty(&EXT4_I(inode)->i_completed_io_list)){
3622 ext4_debug("inode %lu completed_io list is empty\n", inode->i_ino);
3623 return;
3624 }
3625
3626 ext4_debug("Dump inode %lu completed_io list \n", inode->i_ino);
3627 spin_lock_irqsave(&EXT4_I(inode)->i_completed_io_lock, flags);
3628 list_for_each_entry(io, &EXT4_I(inode)->i_completed_io_list, list){
3629 cur = &io->list;
3630 before = cur->prev;
3631 io0 = container_of(before, ext4_io_end_t, list);
3632 after = cur->next;
3633 io1 = container_of(after, ext4_io_end_t, list);
3634
3635 ext4_debug("io 0x%p from inode %lu,prev 0x%p,next 0x%p\n",
3636 io, inode->i_ino, io0, io1);
3637 }
3638 spin_unlock_irqrestore(&EXT4_I(inode)->i_completed_io_lock, flags);
3639#endif
3640}
3641
3642/*
3643 * check a range of space and convert unwritten extents to written.
3644 */
3645static int ext4_end_io_nolock(ext4_io_end_t *io)
3646{
3647 struct inode *inode = io->inode;
3648 loff_t offset = io->offset;
3649 ssize_t size = io->size;
3650 int ret = 0;
3651
3652 ext4_debug("ext4_end_io_nolock: io 0x%p from inode %lu,list->next 0x%p,"
3653 "list->prev 0x%p\n",
3654 io, inode->i_ino, io->list.next, io->list.prev);
3655
3656 if (list_empty(&io->list))
3657 return ret;
3658
3659 if (io->flag != EXT4_IO_UNWRITTEN)
3660 return ret;
3661
3662 ret = ext4_convert_unwritten_extents(inode, offset, size);
3663 if (ret < 0) {
3664 printk(KERN_EMERG "%s: failed to convert unwritten"
3665 "extents to written extents, error is %d"
3666 " io is still on inode %lu aio dio list\n",
3667 __func__, ret, inode->i_ino);
3668 return ret;
3669 }
3670
3671 /* clear the DIO AIO unwritten flag */
3672 io->flag = 0;
3673 return ret;
3674}
3675
3676/*
3677 * work on completed aio dio IO, to convert unwritten extents to extents
3678 */
3679static void ext4_end_io_work(struct work_struct *work)
3680{
3681 ext4_io_end_t *io = container_of(work, ext4_io_end_t, work);
3682 struct inode *inode = io->inode;
3683 struct ext4_inode_info *ei = EXT4_I(inode);
3684 unsigned long flags;
3685 int ret;
3686
3687 mutex_lock(&inode->i_mutex);
3688 ret = ext4_end_io_nolock(io);
3689 if (ret < 0) {
3690 mutex_unlock(&inode->i_mutex);
3691 return;
3692 }
3693
3694 spin_lock_irqsave(&ei->i_completed_io_lock, flags);
3695 if (!list_empty(&io->list))
3696 list_del_init(&io->list);
3697 spin_unlock_irqrestore(&ei->i_completed_io_lock, flags);
3698 mutex_unlock(&inode->i_mutex);
3699 ext4_free_io_end(io);
3700}
3701
3702/*
3703 * This function is called from ext4_sync_file().
3704 *
3705 * When IO is completed, the work to convert unwritten extents to
3706 * written is queued on workqueue but may not get immediately
3707 * scheduled. When fsync is called, we need to ensure the
3708 * conversion is complete before fsync returns.
3709 * The inode keeps track of a list of pending/completed IO that
3710 * might needs to do the conversion. This function walks through
3711 * the list and convert the related unwritten extents for completed IO
3712 * to written.
3713 * The function return the number of pending IOs on success.
3714 */
3715int flush_completed_IO(struct inode *inode)
3716{
3717 ext4_io_end_t *io;
3718 struct ext4_inode_info *ei = EXT4_I(inode);
3719 unsigned long flags;
3720 int ret = 0;
3721 int ret2 = 0;
3722
3723 if (list_empty(&ei->i_completed_io_list))
3724 return ret;
3725
3726 dump_completed_IO(inode);
3727 spin_lock_irqsave(&ei->i_completed_io_lock, flags);
3728 while (!list_empty(&ei->i_completed_io_list)){
3729 io = list_entry(ei->i_completed_io_list.next,
3730 ext4_io_end_t, list);
3731 /*
3732 * Calling ext4_end_io_nolock() to convert completed
3733 * IO to written.
3734 *
3735 * When ext4_sync_file() is called, run_queue() may already
3736 * about to flush the work corresponding to this io structure.
3737 * It will be upset if it founds the io structure related
3738 * to the work-to-be schedule is freed.
3739 *
3740 * Thus we need to keep the io structure still valid here after
3741 * convertion finished. The io structure has a flag to
3742 * avoid double converting from both fsync and background work
3743 * queue work.
3744 */
3745 spin_unlock_irqrestore(&ei->i_completed_io_lock, flags);
3746 ret = ext4_end_io_nolock(io);
3747 spin_lock_irqsave(&ei->i_completed_io_lock, flags);
3748 if (ret < 0)
3749 ret2 = ret;
3750 else
3751 list_del_init(&io->list);
3752 }
3753 spin_unlock_irqrestore(&ei->i_completed_io_lock, flags);
3754 return (ret2 < 0) ? ret2 : 0;
3755}
3756
3757static ext4_io_end_t *ext4_init_io_end (struct inode *inode, gfp_t flags)
3758{
3759 ext4_io_end_t *io = NULL;
3760
3761 io = kmalloc(sizeof(*io), flags);
3762
3763 if (io) {
3764 igrab(inode);
3765 io->inode = inode;
3766 io->flag = 0;
3767 io->offset = 0;
3768 io->size = 0;
3769 io->page = NULL;
3770 INIT_WORK(&io->work, ext4_end_io_work);
3771 INIT_LIST_HEAD(&io->list);
3772 }
3773
3774 return io;
3775}
3776
3777static void ext4_end_io_dio(struct kiocb *iocb, loff_t offset, 3628static void ext4_end_io_dio(struct kiocb *iocb, loff_t offset,
3778 ssize_t size, void *private) 3629 ssize_t size, void *private, int ret,
3630 bool is_async)
3779{ 3631{
3780 ext4_io_end_t *io_end = iocb->private; 3632 ext4_io_end_t *io_end = iocb->private;
3781 struct workqueue_struct *wq; 3633 struct workqueue_struct *wq;
@@ -3784,7 +3636,7 @@ static void ext4_end_io_dio(struct kiocb *iocb, loff_t offset,
3784 3636
3785 /* if not async direct IO or dio with 0 bytes write, just return */ 3637 /* if not async direct IO or dio with 0 bytes write, just return */
3786 if (!io_end || !size) 3638 if (!io_end || !size)
3787 return; 3639 goto out;
3788 3640
3789 ext_debug("ext4_end_io_dio(): io_end 0x%p" 3641 ext_debug("ext4_end_io_dio(): io_end 0x%p"
3790 "for inode %lu, iocb 0x%p, offset %llu, size %llu\n", 3642 "for inode %lu, iocb 0x%p, offset %llu, size %llu\n",
@@ -3792,25 +3644,31 @@ static void ext4_end_io_dio(struct kiocb *iocb, loff_t offset,
3792 size); 3644 size);
3793 3645
3794 /* if not aio dio with unwritten extents, just free io and return */ 3646 /* if not aio dio with unwritten extents, just free io and return */
3795 if (io_end->flag != EXT4_IO_UNWRITTEN){ 3647 if (!(io_end->flag & EXT4_IO_END_UNWRITTEN)) {
3796 ext4_free_io_end(io_end); 3648 ext4_free_io_end(io_end);
3797 iocb->private = NULL; 3649 iocb->private = NULL;
3650out:
3651 if (is_async)
3652 aio_complete(iocb, ret, 0);
3798 return; 3653 return;
3799 } 3654 }
3800 3655
3801 io_end->offset = offset; 3656 io_end->offset = offset;
3802 io_end->size = size; 3657 io_end->size = size;
3803 io_end->flag = EXT4_IO_UNWRITTEN; 3658 if (is_async) {
3659 io_end->iocb = iocb;
3660 io_end->result = ret;
3661 }
3804 wq = EXT4_SB(io_end->inode->i_sb)->dio_unwritten_wq; 3662 wq = EXT4_SB(io_end->inode->i_sb)->dio_unwritten_wq;
3805 3663
3806 /* queue the work to convert unwritten extents to written */
3807 queue_work(wq, &io_end->work);
3808
3809 /* Add the io_end to per-inode completed aio dio list*/ 3664 /* Add the io_end to per-inode completed aio dio list*/
3810 ei = EXT4_I(io_end->inode); 3665 ei = EXT4_I(io_end->inode);
3811 spin_lock_irqsave(&ei->i_completed_io_lock, flags); 3666 spin_lock_irqsave(&ei->i_completed_io_lock, flags);
3812 list_add_tail(&io_end->list, &ei->i_completed_io_list); 3667 list_add_tail(&io_end->list, &ei->i_completed_io_list);
3813 spin_unlock_irqrestore(&ei->i_completed_io_lock, flags); 3668 spin_unlock_irqrestore(&ei->i_completed_io_lock, flags);
3669
3670 /* queue the work to convert unwritten extents to written */
3671 queue_work(wq, &io_end->work);
3814 iocb->private = NULL; 3672 iocb->private = NULL;
3815} 3673}
3816 3674
@@ -3831,7 +3689,7 @@ static void ext4_end_io_buffer_write(struct buffer_head *bh, int uptodate)
3831 goto out; 3689 goto out;
3832 } 3690 }
3833 3691
3834 io_end->flag = EXT4_IO_UNWRITTEN; 3692 io_end->flag = EXT4_IO_END_UNWRITTEN;
3835 inode = io_end->inode; 3693 inode = io_end->inode;
3836 3694
3837 /* Add the io_end to per-inode completed io list*/ 3695 /* Add the io_end to per-inode completed io list*/
@@ -3937,7 +3795,7 @@ static ssize_t ext4_ext_direct_IO(int rw, struct kiocb *iocb,
3937 return -ENOMEM; 3795 return -ENOMEM;
3938 /* 3796 /*
3939 * we save the io structure for current async 3797 * we save the io structure for current async
3940 * direct IO, so that later ext4_get_blocks() 3798 * direct IO, so that later ext4_map_blocks()
3941 * could flag the io structure whether there 3799 * could flag the io structure whether there
3942 * is a unwritten extents needs to be converted 3800 * is a unwritten extents needs to be converted
3943 * when IO is completed. 3801 * when IO is completed.
@@ -4128,17 +3986,6 @@ int ext4_block_truncate_page(handle_t *handle,
4128 length = blocksize - (offset & (blocksize - 1)); 3986 length = blocksize - (offset & (blocksize - 1));
4129 iblock = index << (PAGE_CACHE_SHIFT - inode->i_sb->s_blocksize_bits); 3987 iblock = index << (PAGE_CACHE_SHIFT - inode->i_sb->s_blocksize_bits);
4130 3988
4131 /*
4132 * For "nobh" option, we can only work if we don't need to
4133 * read-in the page - otherwise we create buffers to do the IO.
4134 */
4135 if (!page_has_buffers(page) && test_opt(inode->i_sb, NOBH) &&
4136 ext4_should_writeback_data(inode) && PageUptodate(page)) {
4137 zero_user(page, offset, length);
4138 set_page_dirty(page);
4139 goto unlock;
4140 }
4141
4142 if (!page_has_buffers(page)) 3989 if (!page_has_buffers(page))
4143 create_empty_buffers(page, blocksize, 0); 3990 create_empty_buffers(page, blocksize, 0);
4144 3991
@@ -4488,9 +4335,8 @@ static void ext4_free_branches(handle_t *handle, struct inode *inode,
4488 * (should be rare). 4335 * (should be rare).
4489 */ 4336 */
4490 if (!bh) { 4337 if (!bh) {
4491 EXT4_ERROR_INODE(inode, 4338 EXT4_ERROR_INODE_BLOCK(inode, nr,
4492 "Read failure block=%llu", 4339 "Read failure");
4493 (unsigned long long) nr);
4494 continue; 4340 continue;
4495 } 4341 }
4496 4342
@@ -4502,27 +4348,6 @@ static void ext4_free_branches(handle_t *handle, struct inode *inode,
4502 depth); 4348 depth);
4503 4349
4504 /* 4350 /*
4505 * We've probably journalled the indirect block several
4506 * times during the truncate. But it's no longer
4507 * needed and we now drop it from the transaction via
4508 * jbd2_journal_revoke().
4509 *
4510 * That's easy if it's exclusively part of this
4511 * transaction. But if it's part of the committing
4512 * transaction then jbd2_journal_forget() will simply
4513 * brelse() it. That means that if the underlying
4514 * block is reallocated in ext4_get_block(),
4515 * unmap_underlying_metadata() will find this block
4516 * and will try to get rid of it. damn, damn.
4517 *
4518 * If this block has already been committed to the
4519 * journal, a revoke record will be written. And
4520 * revoke records must be emitted *before* clearing
4521 * this block's bit in the bitmaps.
4522 */
4523 ext4_forget(handle, 1, inode, bh, bh->b_blocknr);
4524
4525 /*
4526 * Everything below this this pointer has been 4351 * Everything below this this pointer has been
4527 * released. Now let this top-of-subtree go. 4352 * released. Now let this top-of-subtree go.
4528 * 4353 *
@@ -4546,8 +4371,20 @@ static void ext4_free_branches(handle_t *handle, struct inode *inode,
4546 blocks_for_truncate(inode)); 4371 blocks_for_truncate(inode));
4547 } 4372 }
4548 4373
4374 /*
4375 * The forget flag here is critical because if
4376 * we are journaling (and not doing data
4377 * journaling), we have to make sure a revoke
4378 * record is written to prevent the journal
4379 * replay from overwriting the (former)
4380 * indirect block if it gets reallocated as a
4381 * data block. This must happen in the same
4382 * transaction where the data blocks are
4383 * actually freed.
4384 */
4549 ext4_free_blocks(handle, inode, 0, nr, 1, 4385 ext4_free_blocks(handle, inode, 0, nr, 1,
4550 EXT4_FREE_BLOCKS_METADATA); 4386 EXT4_FREE_BLOCKS_METADATA|
4387 EXT4_FREE_BLOCKS_FORGET);
4551 4388
4552 if (parent_bh) { 4389 if (parent_bh) {
4553 /* 4390 /*
@@ -4805,8 +4642,8 @@ static int __ext4_get_inode_loc(struct inode *inode,
4805 4642
4806 bh = sb_getblk(sb, block); 4643 bh = sb_getblk(sb, block);
4807 if (!bh) { 4644 if (!bh) {
4808 EXT4_ERROR_INODE(inode, "unable to read inode block - " 4645 EXT4_ERROR_INODE_BLOCK(inode, block,
4809 "block %llu", block); 4646 "unable to read itable block");
4810 return -EIO; 4647 return -EIO;
4811 } 4648 }
4812 if (!buffer_uptodate(bh)) { 4649 if (!buffer_uptodate(bh)) {
@@ -4904,8 +4741,8 @@ make_io:
4904 submit_bh(READ_META, bh); 4741 submit_bh(READ_META, bh);
4905 wait_on_buffer(bh); 4742 wait_on_buffer(bh);
4906 if (!buffer_uptodate(bh)) { 4743 if (!buffer_uptodate(bh)) {
4907 EXT4_ERROR_INODE(inode, "unable to read inode " 4744 EXT4_ERROR_INODE_BLOCK(inode, block,
4908 "block %llu", block); 4745 "unable to read itable block");
4909 brelse(bh); 4746 brelse(bh);
4910 return -EIO; 4747 return -EIO;
4911 } 4748 }
@@ -4976,7 +4813,7 @@ static blkcnt_t ext4_inode_blocks(struct ext4_inode *raw_inode,
4976 /* we are using combined 48 bit field */ 4813 /* we are using combined 48 bit field */
4977 i_blocks = ((u64)le16_to_cpu(raw_inode->i_blocks_high)) << 32 | 4814 i_blocks = ((u64)le16_to_cpu(raw_inode->i_blocks_high)) << 32 |
4978 le32_to_cpu(raw_inode->i_blocks_lo); 4815 le32_to_cpu(raw_inode->i_blocks_lo);
4979 if (ei->i_flags & EXT4_HUGE_FILE_FL) { 4816 if (ext4_test_inode_flag(inode, EXT4_INODE_HUGE_FILE)) {
4980 /* i_blocks represent file system block size */ 4817 /* i_blocks represent file system block size */
4981 return i_blocks << (inode->i_blkbits - 9); 4818 return i_blocks << (inode->i_blkbits - 9);
4982 } else { 4819 } else {
@@ -5072,7 +4909,7 @@ struct inode *ext4_iget(struct super_block *sb, unsigned long ino)
5072 transaction_t *transaction; 4909 transaction_t *transaction;
5073 tid_t tid; 4910 tid_t tid;
5074 4911
5075 spin_lock(&journal->j_state_lock); 4912 read_lock(&journal->j_state_lock);
5076 if (journal->j_running_transaction) 4913 if (journal->j_running_transaction)
5077 transaction = journal->j_running_transaction; 4914 transaction = journal->j_running_transaction;
5078 else 4915 else
@@ -5081,7 +4918,7 @@ struct inode *ext4_iget(struct super_block *sb, unsigned long ino)
5081 tid = transaction->t_tid; 4918 tid = transaction->t_tid;
5082 else 4919 else
5083 tid = journal->j_commit_sequence; 4920 tid = journal->j_commit_sequence;
5084 spin_unlock(&journal->j_state_lock); 4921 read_unlock(&journal->j_state_lock);
5085 ei->i_sync_tid = tid; 4922 ei->i_sync_tid = tid;
5086 ei->i_datasync_tid = tid; 4923 ei->i_datasync_tid = tid;
5087 } 4924 }
@@ -5126,7 +4963,7 @@ struct inode *ext4_iget(struct super_block *sb, unsigned long ino)
5126 ei->i_file_acl); 4963 ei->i_file_acl);
5127 ret = -EIO; 4964 ret = -EIO;
5128 goto bad_inode; 4965 goto bad_inode;
5129 } else if (ei->i_flags & EXT4_EXTENTS_FL) { 4966 } else if (ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS)) {
5130 if (S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) || 4967 if (S_ISREG(inode->i_mode) || S_ISDIR(inode->i_mode) ||
5131 (S_ISLNK(inode->i_mode) && 4968 (S_ISLNK(inode->i_mode) &&
5132 !ext4_inode_is_fast_symlink(inode))) 4969 !ext4_inode_is_fast_symlink(inode)))
@@ -5406,9 +5243,8 @@ int ext4_write_inode(struct inode *inode, struct writeback_control *wbc)
5406 if (wbc->sync_mode == WB_SYNC_ALL) 5243 if (wbc->sync_mode == WB_SYNC_ALL)
5407 sync_dirty_buffer(iloc.bh); 5244 sync_dirty_buffer(iloc.bh);
5408 if (buffer_req(iloc.bh) && !buffer_uptodate(iloc.bh)) { 5245 if (buffer_req(iloc.bh) && !buffer_uptodate(iloc.bh)) {
5409 EXT4_ERROR_INODE(inode, 5246 EXT4_ERROR_INODE_BLOCK(inode, iloc.bh->b_blocknr,
5410 "IO error syncing inode (block=%llu)", 5247 "IO error syncing inode");
5411 (unsigned long long) iloc.bh->b_blocknr);
5412 err = -EIO; 5248 err = -EIO;
5413 } 5249 }
5414 brelse(iloc.bh); 5250 brelse(iloc.bh);
@@ -5444,6 +5280,7 @@ int ext4_setattr(struct dentry *dentry, struct iattr *attr)
5444{ 5280{
5445 struct inode *inode = dentry->d_inode; 5281 struct inode *inode = dentry->d_inode;
5446 int error, rc = 0; 5282 int error, rc = 0;
5283 int orphan = 0;
5447 const unsigned int ia_valid = attr->ia_valid; 5284 const unsigned int ia_valid = attr->ia_valid;
5448 5285
5449 error = inode_change_ok(inode, attr); 5286 error = inode_change_ok(inode, attr);
@@ -5483,10 +5320,8 @@ int ext4_setattr(struct dentry *dentry, struct iattr *attr)
5483 if (!(ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS))) { 5320 if (!(ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS))) {
5484 struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb); 5321 struct ext4_sb_info *sbi = EXT4_SB(inode->i_sb);
5485 5322
5486 if (attr->ia_size > sbi->s_bitmap_maxbytes) { 5323 if (attr->ia_size > sbi->s_bitmap_maxbytes)
5487 error = -EFBIG; 5324 return -EFBIG;
5488 goto err_out;
5489 }
5490 } 5325 }
5491 } 5326 }
5492 5327
@@ -5501,8 +5336,10 @@ int ext4_setattr(struct dentry *dentry, struct iattr *attr)
5501 error = PTR_ERR(handle); 5336 error = PTR_ERR(handle);
5502 goto err_out; 5337 goto err_out;
5503 } 5338 }
5504 5339 if (ext4_handle_valid(handle)) {
5505 error = ext4_orphan_add(handle, inode); 5340 error = ext4_orphan_add(handle, inode);
5341 orphan = 1;
5342 }
5506 EXT4_I(inode)->i_disksize = attr->ia_size; 5343 EXT4_I(inode)->i_disksize = attr->ia_size;
5507 rc = ext4_mark_inode_dirty(handle, inode); 5344 rc = ext4_mark_inode_dirty(handle, inode);
5508 if (!error) 5345 if (!error)
@@ -5520,6 +5357,7 @@ int ext4_setattr(struct dentry *dentry, struct iattr *attr)
5520 goto err_out; 5357 goto err_out;
5521 } 5358 }
5522 ext4_orphan_del(handle, inode); 5359 ext4_orphan_del(handle, inode);
5360 orphan = 0;
5523 ext4_journal_stop(handle); 5361 ext4_journal_stop(handle);
5524 goto err_out; 5362 goto err_out;
5525 } 5363 }
@@ -5529,12 +5367,20 @@ int ext4_setattr(struct dentry *dentry, struct iattr *attr)
5529 ext4_truncate(inode); 5367 ext4_truncate(inode);
5530 } 5368 }
5531 5369
5532 rc = inode_setattr(inode, attr); 5370 if ((attr->ia_valid & ATTR_SIZE) &&
5371 attr->ia_size != i_size_read(inode))
5372 rc = vmtruncate(inode, attr->ia_size);
5533 5373
5534 /* If inode_setattr's call to ext4_truncate failed to get a 5374 if (!rc) {
5535 * transaction handle at all, we need to clean up the in-core 5375 setattr_copy(inode, attr);
5536 * orphan list manually. */ 5376 mark_inode_dirty(inode);
5537 if (inode->i_nlink) 5377 }
5378
5379 /*
5380 * If the call to ext4_truncate failed to get a transaction handle at
5381 * all, we need to clean up the in-core orphan list manually.
5382 */
5383 if (orphan && inode->i_nlink)
5538 ext4_orphan_del(NULL, inode); 5384 ext4_orphan_del(NULL, inode);
5539 5385
5540 if (!rc && (ia_valid & ATTR_MODE)) 5386 if (!rc && (ia_valid & ATTR_MODE))
@@ -5617,7 +5463,7 @@ static int ext4_index_trans_blocks(struct inode *inode, int nrblocks, int chunk)
5617 * 5463 *
5618 * Also account for superblock, inode, quota and xattr blocks 5464 * Also account for superblock, inode, quota and xattr blocks
5619 */ 5465 */
5620int ext4_meta_trans_blocks(struct inode *inode, int nrblocks, int chunk) 5466static int ext4_meta_trans_blocks(struct inode *inode, int nrblocks, int chunk)
5621{ 5467{
5622 ext4_group_t groups, ngroups = ext4_get_groups_count(inode->i_sb); 5468 ext4_group_t groups, ngroups = ext4_get_groups_count(inode->i_sb);
5623 int gdpblocks; 5469 int gdpblocks;
@@ -5688,7 +5534,7 @@ int ext4_writepage_trans_blocks(struct inode *inode)
5688 * Calculate the journal credits for a chunk of data modification. 5534 * Calculate the journal credits for a chunk of data modification.
5689 * 5535 *
5690 * This is called from DIO, fallocate or whoever calling 5536 * This is called from DIO, fallocate or whoever calling
5691 * ext4_get_blocks() to map/allocate a chunk of contiguous disk blocks. 5537 * ext4_map_blocks() to map/allocate a chunk of contiguous disk blocks.
5692 * 5538 *
5693 * journal buffers for data blocks are not included here, as DIO 5539 * journal buffers for data blocks are not included here, as DIO
5694 * and fallocate do no need to journal data buffers. 5540 * and fallocate do no need to journal data buffers.
@@ -5754,7 +5600,6 @@ static int ext4_expand_extra_isize(struct inode *inode,
5754{ 5600{
5755 struct ext4_inode *raw_inode; 5601 struct ext4_inode *raw_inode;
5756 struct ext4_xattr_ibody_header *header; 5602 struct ext4_xattr_ibody_header *header;
5757 struct ext4_xattr_entry *entry;
5758 5603
5759 if (EXT4_I(inode)->i_extra_isize >= new_extra_isize) 5604 if (EXT4_I(inode)->i_extra_isize >= new_extra_isize)
5760 return 0; 5605 return 0;
@@ -5762,7 +5607,6 @@ static int ext4_expand_extra_isize(struct inode *inode,
5762 raw_inode = ext4_raw_inode(&iloc); 5607 raw_inode = ext4_raw_inode(&iloc);
5763 5608
5764 header = IHDR(inode, raw_inode); 5609 header = IHDR(inode, raw_inode);
5765 entry = IFIRST(header);
5766 5610
5767 /* No extended attributes present */ 5611 /* No extended attributes present */
5768 if (!ext4_test_inode_state(inode, EXT4_STATE_XATTR) || 5612 if (!ext4_test_inode_state(inode, EXT4_STATE_XATTR) ||
diff --git a/fs/ext4/mballoc.c b/fs/ext4/mballoc.c
index 12b3bc026a68..c58eba34724a 100644
--- a/fs/ext4/mballoc.c
+++ b/fs/ext4/mballoc.c
@@ -338,6 +338,14 @@
338static struct kmem_cache *ext4_pspace_cachep; 338static struct kmem_cache *ext4_pspace_cachep;
339static struct kmem_cache *ext4_ac_cachep; 339static struct kmem_cache *ext4_ac_cachep;
340static struct kmem_cache *ext4_free_ext_cachep; 340static struct kmem_cache *ext4_free_ext_cachep;
341
342/* We create slab caches for groupinfo data structures based on the
343 * superblock block size. There will be one per mounted filesystem for
344 * each unique s_blocksize_bits */
345#define NR_GRPINFO_CACHES \
346 (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE + 1)
347static struct kmem_cache *ext4_groupinfo_caches[NR_GRPINFO_CACHES];
348
341static void ext4_mb_generate_from_pa(struct super_block *sb, void *bitmap, 349static void ext4_mb_generate_from_pa(struct super_block *sb, void *bitmap,
342 ext4_group_t group); 350 ext4_group_t group);
343static void ext4_mb_generate_from_freelist(struct super_block *sb, void *bitmap, 351static void ext4_mb_generate_from_freelist(struct super_block *sb, void *bitmap,
@@ -446,10 +454,11 @@ static void mb_free_blocks_double(struct inode *inode, struct ext4_buddy *e4b,
446 blocknr = ext4_group_first_block_no(sb, e4b->bd_group); 454 blocknr = ext4_group_first_block_no(sb, e4b->bd_group);
447 blocknr += first + i; 455 blocknr += first + i;
448 ext4_grp_locked_error(sb, e4b->bd_group, 456 ext4_grp_locked_error(sb, e4b->bd_group,
449 __func__, "double-free of inode" 457 inode ? inode->i_ino : 0,
450 " %lu's block %llu(bit %u in group %u)", 458 blocknr,
451 inode ? inode->i_ino : 0, blocknr, 459 "freeing block already freed "
452 first + i, e4b->bd_group); 460 "(bit %u)",
461 first + i);
453 } 462 }
454 mb_clear_bit(first + i, e4b->bd_info->bb_bitmap); 463 mb_clear_bit(first + i, e4b->bd_info->bb_bitmap);
455 } 464 }
@@ -712,9 +721,9 @@ void ext4_mb_generate_buddy(struct super_block *sb,
712 grp->bb_fragments = fragments; 721 grp->bb_fragments = fragments;
713 722
714 if (free != grp->bb_free) { 723 if (free != grp->bb_free) {
715 ext4_grp_locked_error(sb, group, __func__, 724 ext4_grp_locked_error(sb, group, 0, 0,
716 "EXT4-fs: group %u: %u blocks in bitmap, %u in gd", 725 "%u blocks in bitmap, %u in gd",
717 group, free, grp->bb_free); 726 free, grp->bb_free);
718 /* 727 /*
719 * If we intent to continue, we consider group descritor 728 * If we intent to continue, we consider group descritor
720 * corrupt and update bb_free using bitmap value 729 * corrupt and update bb_free using bitmap value
@@ -938,6 +947,85 @@ out:
938} 947}
939 948
940/* 949/*
950 * lock the group_info alloc_sem of all the groups
951 * belonging to the same buddy cache page. This
952 * make sure other parallel operation on the buddy
953 * cache doesn't happen whild holding the buddy cache
954 * lock
955 */
956static int ext4_mb_get_buddy_cache_lock(struct super_block *sb,
957 ext4_group_t group)
958{
959 int i;
960 int block, pnum;
961 int blocks_per_page;
962 int groups_per_page;
963 ext4_group_t ngroups = ext4_get_groups_count(sb);
964 ext4_group_t first_group;
965 struct ext4_group_info *grp;
966
967 blocks_per_page = PAGE_CACHE_SIZE / sb->s_blocksize;
968 /*
969 * the buddy cache inode stores the block bitmap
970 * and buddy information in consecutive blocks.
971 * So for each group we need two blocks.
972 */
973 block = group * 2;
974 pnum = block / blocks_per_page;
975 first_group = pnum * blocks_per_page / 2;
976
977 groups_per_page = blocks_per_page >> 1;
978 if (groups_per_page == 0)
979 groups_per_page = 1;
980 /* read all groups the page covers into the cache */
981 for (i = 0; i < groups_per_page; i++) {
982
983 if ((first_group + i) >= ngroups)
984 break;
985 grp = ext4_get_group_info(sb, first_group + i);
986 /* take all groups write allocation
987 * semaphore. This make sure there is
988 * no block allocation going on in any
989 * of that groups
990 */
991 down_write_nested(&grp->alloc_sem, i);
992 }
993 return i;
994}
995
996static void ext4_mb_put_buddy_cache_lock(struct super_block *sb,
997 ext4_group_t group, int locked_group)
998{
999 int i;
1000 int block, pnum;
1001 int blocks_per_page;
1002 ext4_group_t first_group;
1003 struct ext4_group_info *grp;
1004
1005 blocks_per_page = PAGE_CACHE_SIZE / sb->s_blocksize;
1006 /*
1007 * the buddy cache inode stores the block bitmap
1008 * and buddy information in consecutive blocks.
1009 * So for each group we need two blocks.
1010 */
1011 block = group * 2;
1012 pnum = block / blocks_per_page;
1013 first_group = pnum * blocks_per_page / 2;
1014 /* release locks on all the groups */
1015 for (i = 0; i < locked_group; i++) {
1016
1017 grp = ext4_get_group_info(sb, first_group + i);
1018 /* take all groups write allocation
1019 * semaphore. This make sure there is
1020 * no block allocation going on in any
1021 * of that groups
1022 */
1023 up_write(&grp->alloc_sem);
1024 }
1025
1026}
1027
1028/*
941 * Locking note: This routine calls ext4_mb_init_cache(), which takes the 1029 * Locking note: This routine calls ext4_mb_init_cache(), which takes the
942 * block group lock of all groups for this page; do not hold the BG lock when 1030 * block group lock of all groups for this page; do not hold the BG lock when
943 * calling this routine! 1031 * calling this routine!
@@ -1296,10 +1384,10 @@ static void mb_free_blocks(struct inode *inode, struct ext4_buddy *e4b,
1296 blocknr = ext4_group_first_block_no(sb, e4b->bd_group); 1384 blocknr = ext4_group_first_block_no(sb, e4b->bd_group);
1297 blocknr += block; 1385 blocknr += block;
1298 ext4_grp_locked_error(sb, e4b->bd_group, 1386 ext4_grp_locked_error(sb, e4b->bd_group,
1299 __func__, "double-free of inode" 1387 inode ? inode->i_ino : 0,
1300 " %lu's block %llu(bit %u in group %u)", 1388 blocknr,
1301 inode ? inode->i_ino : 0, blocknr, block, 1389 "freeing already freed block "
1302 e4b->bd_group); 1390 "(bit %u)", block);
1303 } 1391 }
1304 mb_clear_bit(block, EXT4_MB_BITMAP(e4b)); 1392 mb_clear_bit(block, EXT4_MB_BITMAP(e4b));
1305 e4b->bd_info->bb_counters[order]++; 1393 e4b->bd_info->bb_counters[order]++;
@@ -1788,8 +1876,8 @@ void ext4_mb_complex_scan_group(struct ext4_allocation_context *ac,
1788 * free blocks even though group info says we 1876 * free blocks even though group info says we
1789 * we have free blocks 1877 * we have free blocks
1790 */ 1878 */
1791 ext4_grp_locked_error(sb, e4b->bd_group, 1879 ext4_grp_locked_error(sb, e4b->bd_group, 0, 0,
1792 __func__, "%d free blocks as per " 1880 "%d free blocks as per "
1793 "group info. But bitmap says 0", 1881 "group info. But bitmap says 0",
1794 free); 1882 free);
1795 break; 1883 break;
@@ -1798,8 +1886,8 @@ void ext4_mb_complex_scan_group(struct ext4_allocation_context *ac,
1798 mb_find_extent(e4b, 0, i, ac->ac_g_ex.fe_len, &ex); 1886 mb_find_extent(e4b, 0, i, ac->ac_g_ex.fe_len, &ex);
1799 BUG_ON(ex.fe_len <= 0); 1887 BUG_ON(ex.fe_len <= 0);
1800 if (free < ex.fe_len) { 1888 if (free < ex.fe_len) {
1801 ext4_grp_locked_error(sb, e4b->bd_group, 1889 ext4_grp_locked_error(sb, e4b->bd_group, 0, 0,
1802 __func__, "%d free blocks as per " 1890 "%d free blocks as per "
1803 "group info. But got %d blocks", 1891 "group info. But got %d blocks",
1804 free, ex.fe_len); 1892 free, ex.fe_len);
1805 /* 1893 /*
@@ -1821,8 +1909,7 @@ void ext4_mb_complex_scan_group(struct ext4_allocation_context *ac,
1821 1909
1822/* 1910/*
1823 * This is a special case for storages like raid5 1911 * This is a special case for storages like raid5
1824 * we try to find stripe-aligned chunks for stripe-size requests 1912 * we try to find stripe-aligned chunks for stripe-size-multiple requests
1825 * XXX should do so at least for multiples of stripe size as well
1826 */ 1913 */
1827static noinline_for_stack 1914static noinline_for_stack
1828void ext4_mb_scan_aligned(struct ext4_allocation_context *ac, 1915void ext4_mb_scan_aligned(struct ext4_allocation_context *ac,
@@ -1915,91 +2002,12 @@ static int ext4_mb_good_group(struct ext4_allocation_context *ac,
1915 return 0; 2002 return 0;
1916} 2003}
1917 2004
1918/*
1919 * lock the group_info alloc_sem of all the groups
1920 * belonging to the same buddy cache page. This
1921 * make sure other parallel operation on the buddy
1922 * cache doesn't happen whild holding the buddy cache
1923 * lock
1924 */
1925int ext4_mb_get_buddy_cache_lock(struct super_block *sb, ext4_group_t group)
1926{
1927 int i;
1928 int block, pnum;
1929 int blocks_per_page;
1930 int groups_per_page;
1931 ext4_group_t ngroups = ext4_get_groups_count(sb);
1932 ext4_group_t first_group;
1933 struct ext4_group_info *grp;
1934
1935 blocks_per_page = PAGE_CACHE_SIZE / sb->s_blocksize;
1936 /*
1937 * the buddy cache inode stores the block bitmap
1938 * and buddy information in consecutive blocks.
1939 * So for each group we need two blocks.
1940 */
1941 block = group * 2;
1942 pnum = block / blocks_per_page;
1943 first_group = pnum * blocks_per_page / 2;
1944
1945 groups_per_page = blocks_per_page >> 1;
1946 if (groups_per_page == 0)
1947 groups_per_page = 1;
1948 /* read all groups the page covers into the cache */
1949 for (i = 0; i < groups_per_page; i++) {
1950
1951 if ((first_group + i) >= ngroups)
1952 break;
1953 grp = ext4_get_group_info(sb, first_group + i);
1954 /* take all groups write allocation
1955 * semaphore. This make sure there is
1956 * no block allocation going on in any
1957 * of that groups
1958 */
1959 down_write_nested(&grp->alloc_sem, i);
1960 }
1961 return i;
1962}
1963
1964void ext4_mb_put_buddy_cache_lock(struct super_block *sb,
1965 ext4_group_t group, int locked_group)
1966{
1967 int i;
1968 int block, pnum;
1969 int blocks_per_page;
1970 ext4_group_t first_group;
1971 struct ext4_group_info *grp;
1972
1973 blocks_per_page = PAGE_CACHE_SIZE / sb->s_blocksize;
1974 /*
1975 * the buddy cache inode stores the block bitmap
1976 * and buddy information in consecutive blocks.
1977 * So for each group we need two blocks.
1978 */
1979 block = group * 2;
1980 pnum = block / blocks_per_page;
1981 first_group = pnum * blocks_per_page / 2;
1982 /* release locks on all the groups */
1983 for (i = 0; i < locked_group; i++) {
1984
1985 grp = ext4_get_group_info(sb, first_group + i);
1986 /* take all groups write allocation
1987 * semaphore. This make sure there is
1988 * no block allocation going on in any
1989 * of that groups
1990 */
1991 up_write(&grp->alloc_sem);
1992 }
1993
1994}
1995
1996static noinline_for_stack int 2005static noinline_for_stack int
1997ext4_mb_regular_allocator(struct ext4_allocation_context *ac) 2006ext4_mb_regular_allocator(struct ext4_allocation_context *ac)
1998{ 2007{
1999 ext4_group_t ngroups, group, i; 2008 ext4_group_t ngroups, group, i;
2000 int cr; 2009 int cr;
2001 int err = 0; 2010 int err = 0;
2002 int bsbits;
2003 struct ext4_sb_info *sbi; 2011 struct ext4_sb_info *sbi;
2004 struct super_block *sb; 2012 struct super_block *sb;
2005 struct ext4_buddy e4b; 2013 struct ext4_buddy e4b;
@@ -2041,8 +2049,6 @@ ext4_mb_regular_allocator(struct ext4_allocation_context *ac)
2041 ac->ac_2order = i - 1; 2049 ac->ac_2order = i - 1;
2042 } 2050 }
2043 2051
2044 bsbits = ac->ac_sb->s_blocksize_bits;
2045
2046 /* if stream allocation is enabled, use global goal */ 2052 /* if stream allocation is enabled, use global goal */
2047 if (ac->ac_flags & EXT4_MB_STREAM_ALLOC) { 2053 if (ac->ac_flags & EXT4_MB_STREAM_ALLOC) {
2048 /* TBD: may be hot point */ 2054 /* TBD: may be hot point */
@@ -2094,8 +2100,8 @@ repeat:
2094 ac->ac_groups_scanned++; 2100 ac->ac_groups_scanned++;
2095 if (cr == 0) 2101 if (cr == 0)
2096 ext4_mb_simple_scan_group(ac, &e4b); 2102 ext4_mb_simple_scan_group(ac, &e4b);
2097 else if (cr == 1 && 2103 else if (cr == 1 && sbi->s_stripe &&
2098 ac->ac_g_ex.fe_len == sbi->s_stripe) 2104 !(ac->ac_g_ex.fe_len % sbi->s_stripe))
2099 ext4_mb_scan_aligned(ac, &e4b); 2105 ext4_mb_scan_aligned(ac, &e4b);
2100 else 2106 else
2101 ext4_mb_complex_scan_group(ac, &e4b); 2107 ext4_mb_complex_scan_group(ac, &e4b);
@@ -2221,7 +2227,7 @@ static int ext4_mb_seq_groups_open(struct inode *inode, struct file *file)
2221 2227
2222 rc = seq_open(file, &ext4_mb_seq_groups_ops); 2228 rc = seq_open(file, &ext4_mb_seq_groups_ops);
2223 if (rc == 0) { 2229 if (rc == 0) {
2224 struct seq_file *m = (struct seq_file *)file->private_data; 2230 struct seq_file *m = file->private_data;
2225 m->private = sb; 2231 m->private = sb;
2226 } 2232 }
2227 return rc; 2233 return rc;
@@ -2236,15 +2242,24 @@ static const struct file_operations ext4_mb_seq_groups_fops = {
2236 .release = seq_release, 2242 .release = seq_release,
2237}; 2243};
2238 2244
2245static struct kmem_cache *get_groupinfo_cache(int blocksize_bits)
2246{
2247 int cache_index = blocksize_bits - EXT4_MIN_BLOCK_LOG_SIZE;
2248 struct kmem_cache *cachep = ext4_groupinfo_caches[cache_index];
2249
2250 BUG_ON(!cachep);
2251 return cachep;
2252}
2239 2253
2240/* Create and initialize ext4_group_info data for the given group. */ 2254/* Create and initialize ext4_group_info data for the given group. */
2241int ext4_mb_add_groupinfo(struct super_block *sb, ext4_group_t group, 2255int ext4_mb_add_groupinfo(struct super_block *sb, ext4_group_t group,
2242 struct ext4_group_desc *desc) 2256 struct ext4_group_desc *desc)
2243{ 2257{
2244 int i, len; 2258 int i;
2245 int metalen = 0; 2259 int metalen = 0;
2246 struct ext4_sb_info *sbi = EXT4_SB(sb); 2260 struct ext4_sb_info *sbi = EXT4_SB(sb);
2247 struct ext4_group_info **meta_group_info; 2261 struct ext4_group_info **meta_group_info;
2262 struct kmem_cache *cachep = get_groupinfo_cache(sb->s_blocksize_bits);
2248 2263
2249 /* 2264 /*
2250 * First check if this group is the first of a reserved block. 2265 * First check if this group is the first of a reserved block.
@@ -2264,22 +2279,16 @@ int ext4_mb_add_groupinfo(struct super_block *sb, ext4_group_t group,
2264 meta_group_info; 2279 meta_group_info;
2265 } 2280 }
2266 2281
2267 /*
2268 * calculate needed size. if change bb_counters size,
2269 * don't forget about ext4_mb_generate_buddy()
2270 */
2271 len = offsetof(typeof(**meta_group_info),
2272 bb_counters[sb->s_blocksize_bits + 2]);
2273
2274 meta_group_info = 2282 meta_group_info =
2275 sbi->s_group_info[group >> EXT4_DESC_PER_BLOCK_BITS(sb)]; 2283 sbi->s_group_info[group >> EXT4_DESC_PER_BLOCK_BITS(sb)];
2276 i = group & (EXT4_DESC_PER_BLOCK(sb) - 1); 2284 i = group & (EXT4_DESC_PER_BLOCK(sb) - 1);
2277 2285
2278 meta_group_info[i] = kzalloc(len, GFP_KERNEL); 2286 meta_group_info[i] = kmem_cache_alloc(cachep, GFP_KERNEL);
2279 if (meta_group_info[i] == NULL) { 2287 if (meta_group_info[i] == NULL) {
2280 printk(KERN_ERR "EXT4-fs: can't allocate buddy mem\n"); 2288 printk(KERN_ERR "EXT4-fs: can't allocate buddy mem\n");
2281 goto exit_group_info; 2289 goto exit_group_info;
2282 } 2290 }
2291 memset(meta_group_info[i], 0, kmem_cache_size(cachep));
2283 set_bit(EXT4_GROUP_INFO_NEED_INIT_BIT, 2292 set_bit(EXT4_GROUP_INFO_NEED_INIT_BIT,
2284 &(meta_group_info[i]->bb_state)); 2293 &(meta_group_info[i]->bb_state));
2285 2294
@@ -2334,6 +2343,7 @@ static int ext4_mb_init_backend(struct super_block *sb)
2334 int num_meta_group_infos_max; 2343 int num_meta_group_infos_max;
2335 int array_size; 2344 int array_size;
2336 struct ext4_group_desc *desc; 2345 struct ext4_group_desc *desc;
2346 struct kmem_cache *cachep;
2337 2347
2338 /* This is the number of blocks used by GDT */ 2348 /* This is the number of blocks used by GDT */
2339 num_meta_group_infos = (ngroups + EXT4_DESC_PER_BLOCK(sb) - 2349 num_meta_group_infos = (ngroups + EXT4_DESC_PER_BLOCK(sb) -
@@ -2376,6 +2386,7 @@ static int ext4_mb_init_backend(struct super_block *sb)
2376 printk(KERN_ERR "EXT4-fs: can't get new inode\n"); 2386 printk(KERN_ERR "EXT4-fs: can't get new inode\n");
2377 goto err_freesgi; 2387 goto err_freesgi;
2378 } 2388 }
2389 sbi->s_buddy_cache->i_ino = get_next_ino();
2379 EXT4_I(sbi->s_buddy_cache)->i_disksize = 0; 2390 EXT4_I(sbi->s_buddy_cache)->i_disksize = 0;
2380 for (i = 0; i < ngroups; i++) { 2391 for (i = 0; i < ngroups; i++) {
2381 desc = ext4_get_group_desc(sb, i, NULL); 2392 desc = ext4_get_group_desc(sb, i, NULL);
@@ -2391,8 +2402,9 @@ static int ext4_mb_init_backend(struct super_block *sb)
2391 return 0; 2402 return 0;
2392 2403
2393err_freebuddy: 2404err_freebuddy:
2405 cachep = get_groupinfo_cache(sb->s_blocksize_bits);
2394 while (i-- > 0) 2406 while (i-- > 0)
2395 kfree(ext4_get_group_info(sb, i)); 2407 kmem_cache_free(cachep, ext4_get_group_info(sb, i));
2396 i = num_meta_group_infos; 2408 i = num_meta_group_infos;
2397 while (i-- > 0) 2409 while (i-- > 0)
2398 kfree(sbi->s_group_info[i]); 2410 kfree(sbi->s_group_info[i]);
@@ -2409,19 +2421,48 @@ int ext4_mb_init(struct super_block *sb, int needs_recovery)
2409 unsigned offset; 2421 unsigned offset;
2410 unsigned max; 2422 unsigned max;
2411 int ret; 2423 int ret;
2424 int cache_index;
2425 struct kmem_cache *cachep;
2426 char *namep = NULL;
2412 2427
2413 i = (sb->s_blocksize_bits + 2) * sizeof(*sbi->s_mb_offsets); 2428 i = (sb->s_blocksize_bits + 2) * sizeof(*sbi->s_mb_offsets);
2414 2429
2415 sbi->s_mb_offsets = kmalloc(i, GFP_KERNEL); 2430 sbi->s_mb_offsets = kmalloc(i, GFP_KERNEL);
2416 if (sbi->s_mb_offsets == NULL) { 2431 if (sbi->s_mb_offsets == NULL) {
2417 return -ENOMEM; 2432 ret = -ENOMEM;
2433 goto out;
2418 } 2434 }
2419 2435
2420 i = (sb->s_blocksize_bits + 2) * sizeof(*sbi->s_mb_maxs); 2436 i = (sb->s_blocksize_bits + 2) * sizeof(*sbi->s_mb_maxs);
2421 sbi->s_mb_maxs = kmalloc(i, GFP_KERNEL); 2437 sbi->s_mb_maxs = kmalloc(i, GFP_KERNEL);
2422 if (sbi->s_mb_maxs == NULL) { 2438 if (sbi->s_mb_maxs == NULL) {
2423 kfree(sbi->s_mb_offsets); 2439 ret = -ENOMEM;
2424 return -ENOMEM; 2440 goto out;
2441 }
2442
2443 cache_index = sb->s_blocksize_bits - EXT4_MIN_BLOCK_LOG_SIZE;
2444 cachep = ext4_groupinfo_caches[cache_index];
2445 if (!cachep) {
2446 char name[32];
2447 int len = offsetof(struct ext4_group_info,
2448 bb_counters[sb->s_blocksize_bits + 2]);
2449
2450 sprintf(name, "ext4_groupinfo_%d", sb->s_blocksize_bits);
2451 namep = kstrdup(name, GFP_KERNEL);
2452 if (!namep) {
2453 ret = -ENOMEM;
2454 goto out;
2455 }
2456
2457 /* Need to free the kmem_cache_name() when we
2458 * destroy the slab */
2459 cachep = kmem_cache_create(namep, len, 0,
2460 SLAB_RECLAIM_ACCOUNT, NULL);
2461 if (!cachep) {
2462 ret = -ENOMEM;
2463 goto out;
2464 }
2465 ext4_groupinfo_caches[cache_index] = cachep;
2425 } 2466 }
2426 2467
2427 /* order 0 is regular bitmap */ 2468 /* order 0 is regular bitmap */
@@ -2442,9 +2483,7 @@ int ext4_mb_init(struct super_block *sb, int needs_recovery)
2442 /* init file for buddy data */ 2483 /* init file for buddy data */
2443 ret = ext4_mb_init_backend(sb); 2484 ret = ext4_mb_init_backend(sb);
2444 if (ret != 0) { 2485 if (ret != 0) {
2445 kfree(sbi->s_mb_offsets); 2486 goto out;
2446 kfree(sbi->s_mb_maxs);
2447 return ret;
2448 } 2487 }
2449 2488
2450 spin_lock_init(&sbi->s_md_lock); 2489 spin_lock_init(&sbi->s_md_lock);
@@ -2459,9 +2498,8 @@ int ext4_mb_init(struct super_block *sb, int needs_recovery)
2459 2498
2460 sbi->s_locality_groups = alloc_percpu(struct ext4_locality_group); 2499 sbi->s_locality_groups = alloc_percpu(struct ext4_locality_group);
2461 if (sbi->s_locality_groups == NULL) { 2500 if (sbi->s_locality_groups == NULL) {
2462 kfree(sbi->s_mb_offsets); 2501 ret = -ENOMEM;
2463 kfree(sbi->s_mb_maxs); 2502 goto out;
2464 return -ENOMEM;
2465 } 2503 }
2466 for_each_possible_cpu(i) { 2504 for_each_possible_cpu(i) {
2467 struct ext4_locality_group *lg; 2505 struct ext4_locality_group *lg;
@@ -2478,7 +2516,13 @@ int ext4_mb_init(struct super_block *sb, int needs_recovery)
2478 2516
2479 if (sbi->s_journal) 2517 if (sbi->s_journal)
2480 sbi->s_journal->j_commit_callback = release_blocks_on_commit; 2518 sbi->s_journal->j_commit_callback = release_blocks_on_commit;
2481 return 0; 2519out:
2520 if (ret) {
2521 kfree(sbi->s_mb_offsets);
2522 kfree(sbi->s_mb_maxs);
2523 kfree(namep);
2524 }
2525 return ret;
2482} 2526}
2483 2527
2484/* need to called with the ext4 group lock held */ 2528/* need to called with the ext4 group lock held */
@@ -2506,6 +2550,7 @@ int ext4_mb_release(struct super_block *sb)
2506 int num_meta_group_infos; 2550 int num_meta_group_infos;
2507 struct ext4_group_info *grinfo; 2551 struct ext4_group_info *grinfo;
2508 struct ext4_sb_info *sbi = EXT4_SB(sb); 2552 struct ext4_sb_info *sbi = EXT4_SB(sb);
2553 struct kmem_cache *cachep = get_groupinfo_cache(sb->s_blocksize_bits);
2509 2554
2510 if (sbi->s_group_info) { 2555 if (sbi->s_group_info) {
2511 for (i = 0; i < ngroups; i++) { 2556 for (i = 0; i < ngroups; i++) {
@@ -2516,7 +2561,7 @@ int ext4_mb_release(struct super_block *sb)
2516 ext4_lock_group(sb, i); 2561 ext4_lock_group(sb, i);
2517 ext4_mb_cleanup_pa(grinfo); 2562 ext4_mb_cleanup_pa(grinfo);
2518 ext4_unlock_group(sb, i); 2563 ext4_unlock_group(sb, i);
2519 kfree(grinfo); 2564 kmem_cache_free(cachep, grinfo);
2520 } 2565 }
2521 num_meta_group_infos = (ngroups + 2566 num_meta_group_infos = (ngroups +
2522 EXT4_DESC_PER_BLOCK(sb) - 1) >> 2567 EXT4_DESC_PER_BLOCK(sb) - 1) >>
@@ -2560,6 +2605,23 @@ int ext4_mb_release(struct super_block *sb)
2560 return 0; 2605 return 0;
2561} 2606}
2562 2607
2608static inline int ext4_issue_discard(struct super_block *sb,
2609 ext4_group_t block_group, ext4_grpblk_t block, int count)
2610{
2611 int ret;
2612 ext4_fsblk_t discard_block;
2613
2614 discard_block = block + ext4_group_first_block_no(sb, block_group);
2615 trace_ext4_discard_blocks(sb,
2616 (unsigned long long) discard_block, count);
2617 ret = sb_issue_discard(sb, discard_block, count, GFP_NOFS, 0);
2618 if (ret == -EOPNOTSUPP) {
2619 ext4_warning(sb, "discard not supported, disabling");
2620 clear_opt(EXT4_SB(sb)->s_mount_opt, DISCARD);
2621 }
2622 return ret;
2623}
2624
2563/* 2625/*
2564 * This function is called by the jbd2 layer once the commit has finished, 2626 * This function is called by the jbd2 layer once the commit has finished,
2565 * so we know we can free the blocks that were released with that commit. 2627 * so we know we can free the blocks that were released with that commit.
@@ -2579,22 +2641,9 @@ static void release_blocks_on_commit(journal_t *journal, transaction_t *txn)
2579 mb_debug(1, "gonna free %u blocks in group %u (0x%p):", 2641 mb_debug(1, "gonna free %u blocks in group %u (0x%p):",
2580 entry->count, entry->group, entry); 2642 entry->count, entry->group, entry);
2581 2643
2582 if (test_opt(sb, DISCARD)) { 2644 if (test_opt(sb, DISCARD))
2583 int ret; 2645 ext4_issue_discard(sb, entry->group,
2584 ext4_fsblk_t discard_block; 2646 entry->start_blk, entry->count);
2585
2586 discard_block = entry->start_blk +
2587 ext4_group_first_block_no(sb, entry->group);
2588 trace_ext4_discard_blocks(sb,
2589 (unsigned long long)discard_block,
2590 entry->count);
2591 ret = sb_issue_discard(sb, discard_block, entry->count);
2592 if (ret == EOPNOTSUPP) {
2593 ext4_warning(sb,
2594 "discard not supported, disabling");
2595 clear_opt(EXT4_SB(sb)->s_mount_opt, DISCARD);
2596 }
2597 }
2598 2647
2599 err = ext4_mb_load_buddy(sb, entry->group, &e4b); 2648 err = ext4_mb_load_buddy(sb, entry->group, &e4b);
2600 /* we expect to find existing buddy because it's pinned */ 2649 /* we expect to find existing buddy because it's pinned */
@@ -2658,28 +2707,22 @@ static void ext4_remove_debugfs_entry(void)
2658 2707
2659#endif 2708#endif
2660 2709
2661int __init init_ext4_mballoc(void) 2710int __init ext4_init_mballoc(void)
2662{ 2711{
2663 ext4_pspace_cachep = 2712 ext4_pspace_cachep = KMEM_CACHE(ext4_prealloc_space,
2664 kmem_cache_create("ext4_prealloc_space", 2713 SLAB_RECLAIM_ACCOUNT);
2665 sizeof(struct ext4_prealloc_space),
2666 0, SLAB_RECLAIM_ACCOUNT, NULL);
2667 if (ext4_pspace_cachep == NULL) 2714 if (ext4_pspace_cachep == NULL)
2668 return -ENOMEM; 2715 return -ENOMEM;
2669 2716
2670 ext4_ac_cachep = 2717 ext4_ac_cachep = KMEM_CACHE(ext4_allocation_context,
2671 kmem_cache_create("ext4_alloc_context", 2718 SLAB_RECLAIM_ACCOUNT);
2672 sizeof(struct ext4_allocation_context),
2673 0, SLAB_RECLAIM_ACCOUNT, NULL);
2674 if (ext4_ac_cachep == NULL) { 2719 if (ext4_ac_cachep == NULL) {
2675 kmem_cache_destroy(ext4_pspace_cachep); 2720 kmem_cache_destroy(ext4_pspace_cachep);
2676 return -ENOMEM; 2721 return -ENOMEM;
2677 } 2722 }
2678 2723
2679 ext4_free_ext_cachep = 2724 ext4_free_ext_cachep = KMEM_CACHE(ext4_free_data,
2680 kmem_cache_create("ext4_free_block_extents", 2725 SLAB_RECLAIM_ACCOUNT);
2681 sizeof(struct ext4_free_data),
2682 0, SLAB_RECLAIM_ACCOUNT, NULL);
2683 if (ext4_free_ext_cachep == NULL) { 2726 if (ext4_free_ext_cachep == NULL) {
2684 kmem_cache_destroy(ext4_pspace_cachep); 2727 kmem_cache_destroy(ext4_pspace_cachep);
2685 kmem_cache_destroy(ext4_ac_cachep); 2728 kmem_cache_destroy(ext4_ac_cachep);
@@ -2689,8 +2732,9 @@ int __init init_ext4_mballoc(void)
2689 return 0; 2732 return 0;
2690} 2733}
2691 2734
2692void exit_ext4_mballoc(void) 2735void ext4_exit_mballoc(void)
2693{ 2736{
2737 int i;
2694 /* 2738 /*
2695 * Wait for completion of call_rcu()'s on ext4_pspace_cachep 2739 * Wait for completion of call_rcu()'s on ext4_pspace_cachep
2696 * before destroying the slab cache. 2740 * before destroying the slab cache.
@@ -2699,12 +2743,21 @@ void exit_ext4_mballoc(void)
2699 kmem_cache_destroy(ext4_pspace_cachep); 2743 kmem_cache_destroy(ext4_pspace_cachep);
2700 kmem_cache_destroy(ext4_ac_cachep); 2744 kmem_cache_destroy(ext4_ac_cachep);
2701 kmem_cache_destroy(ext4_free_ext_cachep); 2745 kmem_cache_destroy(ext4_free_ext_cachep);
2746
2747 for (i = 0; i < NR_GRPINFO_CACHES; i++) {
2748 struct kmem_cache *cachep = ext4_groupinfo_caches[i];
2749 if (cachep) {
2750 char *name = (char *)kmem_cache_name(cachep);
2751 kmem_cache_destroy(cachep);
2752 kfree(name);
2753 }
2754 }
2702 ext4_remove_debugfs_entry(); 2755 ext4_remove_debugfs_entry();
2703} 2756}
2704 2757
2705 2758
2706/* 2759/*
2707 * Check quota and mark choosed space (ac->ac_b_ex) non-free in bitmaps 2760 * Check quota and mark chosen space (ac->ac_b_ex) non-free in bitmaps
2708 * Returns 0 if success or error code 2761 * Returns 0 if success or error code
2709 */ 2762 */
2710static noinline_for_stack int 2763static noinline_for_stack int
@@ -2712,7 +2765,6 @@ ext4_mb_mark_diskspace_used(struct ext4_allocation_context *ac,
2712 handle_t *handle, unsigned int reserv_blks) 2765 handle_t *handle, unsigned int reserv_blks)
2713{ 2766{
2714 struct buffer_head *bitmap_bh = NULL; 2767 struct buffer_head *bitmap_bh = NULL;
2715 struct ext4_super_block *es;
2716 struct ext4_group_desc *gdp; 2768 struct ext4_group_desc *gdp;
2717 struct buffer_head *gdp_bh; 2769 struct buffer_head *gdp_bh;
2718 struct ext4_sb_info *sbi; 2770 struct ext4_sb_info *sbi;
@@ -2725,8 +2777,6 @@ ext4_mb_mark_diskspace_used(struct ext4_allocation_context *ac,
2725 2777
2726 sb = ac->ac_sb; 2778 sb = ac->ac_sb;
2727 sbi = EXT4_SB(sb); 2779 sbi = EXT4_SB(sb);
2728 es = sbi->s_es;
2729
2730 2780
2731 err = -EIO; 2781 err = -EIO;
2732 bitmap_bh = ext4_read_block_bitmap(sb, ac->ac_b_ex.fe_group); 2782 bitmap_bh = ext4_read_block_bitmap(sb, ac->ac_b_ex.fe_group);
@@ -2812,7 +2862,7 @@ ext4_mb_mark_diskspace_used(struct ext4_allocation_context *ac,
2812 err = ext4_handle_dirty_metadata(handle, NULL, gdp_bh); 2862 err = ext4_handle_dirty_metadata(handle, NULL, gdp_bh);
2813 2863
2814out_err: 2864out_err:
2815 sb->s_dirt = 1; 2865 ext4_mark_super_dirty(sb);
2816 brelse(bitmap_bh); 2866 brelse(bitmap_bh);
2817 return err; 2867 return err;
2818} 2868}
@@ -2850,7 +2900,7 @@ ext4_mb_normalize_request(struct ext4_allocation_context *ac,
2850 int bsbits, max; 2900 int bsbits, max;
2851 ext4_lblk_t end; 2901 ext4_lblk_t end;
2852 loff_t size, orig_size, start_off; 2902 loff_t size, orig_size, start_off;
2853 ext4_lblk_t start, orig_start; 2903 ext4_lblk_t start;
2854 struct ext4_inode_info *ei = EXT4_I(ac->ac_inode); 2904 struct ext4_inode_info *ei = EXT4_I(ac->ac_inode);
2855 struct ext4_prealloc_space *pa; 2905 struct ext4_prealloc_space *pa;
2856 2906
@@ -2881,6 +2931,7 @@ ext4_mb_normalize_request(struct ext4_allocation_context *ac,
2881 size = size << bsbits; 2931 size = size << bsbits;
2882 if (size < i_size_read(ac->ac_inode)) 2932 if (size < i_size_read(ac->ac_inode))
2883 size = i_size_read(ac->ac_inode); 2933 size = i_size_read(ac->ac_inode);
2934 orig_size = size;
2884 2935
2885 /* max size of free chunks */ 2936 /* max size of free chunks */
2886 max = 2 << bsbits; 2937 max = 2 << bsbits;
@@ -2922,8 +2973,8 @@ ext4_mb_normalize_request(struct ext4_allocation_context *ac,
2922 start_off = (loff_t)ac->ac_o_ex.fe_logical << bsbits; 2973 start_off = (loff_t)ac->ac_o_ex.fe_logical << bsbits;
2923 size = ac->ac_o_ex.fe_len << bsbits; 2974 size = ac->ac_o_ex.fe_len << bsbits;
2924 } 2975 }
2925 orig_size = size = size >> bsbits; 2976 size = size >> bsbits;
2926 orig_start = start = start_off >> bsbits; 2977 start = start_off >> bsbits;
2927 2978
2928 /* don't cover already allocated blocks in selected range */ 2979 /* don't cover already allocated blocks in selected range */
2929 if (ar->pleft && start <= ar->lleft) { 2980 if (ar->pleft && start <= ar->lleft) {
@@ -3537,8 +3588,7 @@ static int ext4_mb_new_preallocation(struct ext4_allocation_context *ac)
3537 */ 3588 */
3538static noinline_for_stack int 3589static noinline_for_stack int
3539ext4_mb_release_inode_pa(struct ext4_buddy *e4b, struct buffer_head *bitmap_bh, 3590ext4_mb_release_inode_pa(struct ext4_buddy *e4b, struct buffer_head *bitmap_bh,
3540 struct ext4_prealloc_space *pa, 3591 struct ext4_prealloc_space *pa)
3541 struct ext4_allocation_context *ac)
3542{ 3592{
3543 struct super_block *sb = e4b->bd_sb; 3593 struct super_block *sb = e4b->bd_sb;
3544 struct ext4_sb_info *sbi = EXT4_SB(sb); 3594 struct ext4_sb_info *sbi = EXT4_SB(sb);
@@ -3547,7 +3597,6 @@ ext4_mb_release_inode_pa(struct ext4_buddy *e4b, struct buffer_head *bitmap_bh,
3547 ext4_group_t group; 3597 ext4_group_t group;
3548 ext4_grpblk_t bit; 3598 ext4_grpblk_t bit;
3549 unsigned long long grp_blk_start; 3599 unsigned long long grp_blk_start;
3550 sector_t start;
3551 int err = 0; 3600 int err = 0;
3552 int free = 0; 3601 int free = 0;
3553 3602
@@ -3557,32 +3606,19 @@ ext4_mb_release_inode_pa(struct ext4_buddy *e4b, struct buffer_head *bitmap_bh,
3557 BUG_ON(group != e4b->bd_group && pa->pa_len != 0); 3606 BUG_ON(group != e4b->bd_group && pa->pa_len != 0);
3558 end = bit + pa->pa_len; 3607 end = bit + pa->pa_len;
3559 3608
3560 if (ac) {
3561 ac->ac_sb = sb;
3562 ac->ac_inode = pa->pa_inode;
3563 }
3564
3565 while (bit < end) { 3609 while (bit < end) {
3566 bit = mb_find_next_zero_bit(bitmap_bh->b_data, end, bit); 3610 bit = mb_find_next_zero_bit(bitmap_bh->b_data, end, bit);
3567 if (bit >= end) 3611 if (bit >= end)
3568 break; 3612 break;
3569 next = mb_find_next_bit(bitmap_bh->b_data, end, bit); 3613 next = mb_find_next_bit(bitmap_bh->b_data, end, bit);
3570 start = ext4_group_first_block_no(sb, group) + bit;
3571 mb_debug(1, " free preallocated %u/%u in group %u\n", 3614 mb_debug(1, " free preallocated %u/%u in group %u\n",
3572 (unsigned) start, (unsigned) next - bit, 3615 (unsigned) ext4_group_first_block_no(sb, group) + bit,
3573 (unsigned) group); 3616 (unsigned) next - bit, (unsigned) group);
3574 free += next - bit; 3617 free += next - bit;
3575 3618
3576 if (ac) { 3619 trace_ext4_mballoc_discard(sb, NULL, group, bit, next - bit);
3577 ac->ac_b_ex.fe_group = group; 3620 trace_ext4_mb_release_inode_pa(sb, pa->pa_inode, pa,
3578 ac->ac_b_ex.fe_start = bit; 3621 grp_blk_start + bit, next - bit);
3579 ac->ac_b_ex.fe_len = next - bit;
3580 ac->ac_b_ex.fe_logical = 0;
3581 trace_ext4_mballoc_discard(ac);
3582 }
3583
3584 trace_ext4_mb_release_inode_pa(ac, pa, grp_blk_start + bit,
3585 next - bit);
3586 mb_free_blocks(pa->pa_inode, e4b, bit, next - bit); 3622 mb_free_blocks(pa->pa_inode, e4b, bit, next - bit);
3587 bit = next + 1; 3623 bit = next + 1;
3588 } 3624 }
@@ -3591,8 +3627,7 @@ ext4_mb_release_inode_pa(struct ext4_buddy *e4b, struct buffer_head *bitmap_bh,
3591 pa, (unsigned long) pa->pa_lstart, 3627 pa, (unsigned long) pa->pa_lstart,
3592 (unsigned long) pa->pa_pstart, 3628 (unsigned long) pa->pa_pstart,
3593 (unsigned long) pa->pa_len); 3629 (unsigned long) pa->pa_len);
3594 ext4_grp_locked_error(sb, group, 3630 ext4_grp_locked_error(sb, group, 0, 0, "free %u, pa_free %u",
3595 __func__, "free %u, pa_free %u",
3596 free, pa->pa_free); 3631 free, pa->pa_free);
3597 /* 3632 /*
3598 * pa is already deleted so we use the value obtained 3633 * pa is already deleted so we use the value obtained
@@ -3606,29 +3641,19 @@ ext4_mb_release_inode_pa(struct ext4_buddy *e4b, struct buffer_head *bitmap_bh,
3606 3641
3607static noinline_for_stack int 3642static noinline_for_stack int
3608ext4_mb_release_group_pa(struct ext4_buddy *e4b, 3643ext4_mb_release_group_pa(struct ext4_buddy *e4b,
3609 struct ext4_prealloc_space *pa, 3644 struct ext4_prealloc_space *pa)
3610 struct ext4_allocation_context *ac)
3611{ 3645{
3612 struct super_block *sb = e4b->bd_sb; 3646 struct super_block *sb = e4b->bd_sb;
3613 ext4_group_t group; 3647 ext4_group_t group;
3614 ext4_grpblk_t bit; 3648 ext4_grpblk_t bit;
3615 3649
3616 trace_ext4_mb_release_group_pa(ac, pa); 3650 trace_ext4_mb_release_group_pa(sb, pa);
3617 BUG_ON(pa->pa_deleted == 0); 3651 BUG_ON(pa->pa_deleted == 0);
3618 ext4_get_group_no_and_offset(sb, pa->pa_pstart, &group, &bit); 3652 ext4_get_group_no_and_offset(sb, pa->pa_pstart, &group, &bit);
3619 BUG_ON(group != e4b->bd_group && pa->pa_len != 0); 3653 BUG_ON(group != e4b->bd_group && pa->pa_len != 0);
3620 mb_free_blocks(pa->pa_inode, e4b, bit, pa->pa_len); 3654 mb_free_blocks(pa->pa_inode, e4b, bit, pa->pa_len);
3621 atomic_add(pa->pa_len, &EXT4_SB(sb)->s_mb_discarded); 3655 atomic_add(pa->pa_len, &EXT4_SB(sb)->s_mb_discarded);
3622 3656 trace_ext4_mballoc_discard(sb, NULL, group, bit, pa->pa_len);
3623 if (ac) {
3624 ac->ac_sb = sb;
3625 ac->ac_inode = NULL;
3626 ac->ac_b_ex.fe_group = group;
3627 ac->ac_b_ex.fe_start = bit;
3628 ac->ac_b_ex.fe_len = pa->pa_len;
3629 ac->ac_b_ex.fe_logical = 0;
3630 trace_ext4_mballoc_discard(ac);
3631 }
3632 3657
3633 return 0; 3658 return 0;
3634} 3659}
@@ -3649,7 +3674,6 @@ ext4_mb_discard_group_preallocations(struct super_block *sb,
3649 struct ext4_group_info *grp = ext4_get_group_info(sb, group); 3674 struct ext4_group_info *grp = ext4_get_group_info(sb, group);
3650 struct buffer_head *bitmap_bh = NULL; 3675 struct buffer_head *bitmap_bh = NULL;
3651 struct ext4_prealloc_space *pa, *tmp; 3676 struct ext4_prealloc_space *pa, *tmp;
3652 struct ext4_allocation_context *ac;
3653 struct list_head list; 3677 struct list_head list;
3654 struct ext4_buddy e4b; 3678 struct ext4_buddy e4b;
3655 int err; 3679 int err;
@@ -3678,9 +3702,6 @@ ext4_mb_discard_group_preallocations(struct super_block *sb,
3678 needed = EXT4_BLOCKS_PER_GROUP(sb) + 1; 3702 needed = EXT4_BLOCKS_PER_GROUP(sb) + 1;
3679 3703
3680 INIT_LIST_HEAD(&list); 3704 INIT_LIST_HEAD(&list);
3681 ac = kmem_cache_alloc(ext4_ac_cachep, GFP_NOFS);
3682 if (ac)
3683 ac->ac_sb = sb;
3684repeat: 3705repeat:
3685 ext4_lock_group(sb, group); 3706 ext4_lock_group(sb, group);
3686 list_for_each_entry_safe(pa, tmp, 3707 list_for_each_entry_safe(pa, tmp,
@@ -3735,9 +3756,9 @@ repeat:
3735 spin_unlock(pa->pa_obj_lock); 3756 spin_unlock(pa->pa_obj_lock);
3736 3757
3737 if (pa->pa_type == MB_GROUP_PA) 3758 if (pa->pa_type == MB_GROUP_PA)
3738 ext4_mb_release_group_pa(&e4b, pa, ac); 3759 ext4_mb_release_group_pa(&e4b, pa);
3739 else 3760 else
3740 ext4_mb_release_inode_pa(&e4b, bitmap_bh, pa, ac); 3761 ext4_mb_release_inode_pa(&e4b, bitmap_bh, pa);
3741 3762
3742 list_del(&pa->u.pa_tmp_list); 3763 list_del(&pa->u.pa_tmp_list);
3743 call_rcu(&(pa)->u.pa_rcu, ext4_mb_pa_callback); 3764 call_rcu(&(pa)->u.pa_rcu, ext4_mb_pa_callback);
@@ -3745,8 +3766,6 @@ repeat:
3745 3766
3746out: 3767out:
3747 ext4_unlock_group(sb, group); 3768 ext4_unlock_group(sb, group);
3748 if (ac)
3749 kmem_cache_free(ext4_ac_cachep, ac);
3750 ext4_mb_unload_buddy(&e4b); 3769 ext4_mb_unload_buddy(&e4b);
3751 put_bh(bitmap_bh); 3770 put_bh(bitmap_bh);
3752 return free; 3771 return free;
@@ -3767,7 +3786,6 @@ void ext4_discard_preallocations(struct inode *inode)
3767 struct super_block *sb = inode->i_sb; 3786 struct super_block *sb = inode->i_sb;
3768 struct buffer_head *bitmap_bh = NULL; 3787 struct buffer_head *bitmap_bh = NULL;
3769 struct ext4_prealloc_space *pa, *tmp; 3788 struct ext4_prealloc_space *pa, *tmp;
3770 struct ext4_allocation_context *ac;
3771 ext4_group_t group = 0; 3789 ext4_group_t group = 0;
3772 struct list_head list; 3790 struct list_head list;
3773 struct ext4_buddy e4b; 3791 struct ext4_buddy e4b;
@@ -3783,11 +3801,6 @@ void ext4_discard_preallocations(struct inode *inode)
3783 3801
3784 INIT_LIST_HEAD(&list); 3802 INIT_LIST_HEAD(&list);
3785 3803
3786 ac = kmem_cache_alloc(ext4_ac_cachep, GFP_NOFS);
3787 if (ac) {
3788 ac->ac_sb = sb;
3789 ac->ac_inode = inode;
3790 }
3791repeat: 3804repeat:
3792 /* first, collect all pa's in the inode */ 3805 /* first, collect all pa's in the inode */
3793 spin_lock(&ei->i_prealloc_lock); 3806 spin_lock(&ei->i_prealloc_lock);
@@ -3857,7 +3870,7 @@ repeat:
3857 3870
3858 ext4_lock_group(sb, group); 3871 ext4_lock_group(sb, group);
3859 list_del(&pa->pa_group_list); 3872 list_del(&pa->pa_group_list);
3860 ext4_mb_release_inode_pa(&e4b, bitmap_bh, pa, ac); 3873 ext4_mb_release_inode_pa(&e4b, bitmap_bh, pa);
3861 ext4_unlock_group(sb, group); 3874 ext4_unlock_group(sb, group);
3862 3875
3863 ext4_mb_unload_buddy(&e4b); 3876 ext4_mb_unload_buddy(&e4b);
@@ -3866,8 +3879,6 @@ repeat:
3866 list_del(&pa->u.pa_tmp_list); 3879 list_del(&pa->u.pa_tmp_list);
3867 call_rcu(&(pa)->u.pa_rcu, ext4_mb_pa_callback); 3880 call_rcu(&(pa)->u.pa_rcu, ext4_mb_pa_callback);
3868 } 3881 }
3869 if (ac)
3870 kmem_cache_free(ext4_ac_cachep, ac);
3871} 3882}
3872 3883
3873/* 3884/*
@@ -3889,6 +3900,9 @@ static void ext4_mb_show_ac(struct ext4_allocation_context *ac)
3889 struct super_block *sb = ac->ac_sb; 3900 struct super_block *sb = ac->ac_sb;
3890 ext4_group_t ngroups, i; 3901 ext4_group_t ngroups, i;
3891 3902
3903 if (EXT4_SB(sb)->s_mount_flags & EXT4_MF_FS_ABORTED)
3904 return;
3905
3892 printk(KERN_ERR "EXT4-fs: Can't allocate:" 3906 printk(KERN_ERR "EXT4-fs: Can't allocate:"
3893 " Allocation context details:\n"); 3907 " Allocation context details:\n");
3894 printk(KERN_ERR "EXT4-fs: status %d flags %d\n", 3908 printk(KERN_ERR "EXT4-fs: status %d flags %d\n",
@@ -4062,14 +4076,10 @@ ext4_mb_discard_lg_preallocations(struct super_block *sb,
4062 struct ext4_buddy e4b; 4076 struct ext4_buddy e4b;
4063 struct list_head discard_list; 4077 struct list_head discard_list;
4064 struct ext4_prealloc_space *pa, *tmp; 4078 struct ext4_prealloc_space *pa, *tmp;
4065 struct ext4_allocation_context *ac;
4066 4079
4067 mb_debug(1, "discard locality group preallocation\n"); 4080 mb_debug(1, "discard locality group preallocation\n");
4068 4081
4069 INIT_LIST_HEAD(&discard_list); 4082 INIT_LIST_HEAD(&discard_list);
4070 ac = kmem_cache_alloc(ext4_ac_cachep, GFP_NOFS);
4071 if (ac)
4072 ac->ac_sb = sb;
4073 4083
4074 spin_lock(&lg->lg_prealloc_lock); 4084 spin_lock(&lg->lg_prealloc_lock);
4075 list_for_each_entry_rcu(pa, &lg->lg_prealloc_list[order], 4085 list_for_each_entry_rcu(pa, &lg->lg_prealloc_list[order],
@@ -4121,15 +4131,13 @@ ext4_mb_discard_lg_preallocations(struct super_block *sb,
4121 } 4131 }
4122 ext4_lock_group(sb, group); 4132 ext4_lock_group(sb, group);
4123 list_del(&pa->pa_group_list); 4133 list_del(&pa->pa_group_list);
4124 ext4_mb_release_group_pa(&e4b, pa, ac); 4134 ext4_mb_release_group_pa(&e4b, pa);
4125 ext4_unlock_group(sb, group); 4135 ext4_unlock_group(sb, group);
4126 4136
4127 ext4_mb_unload_buddy(&e4b); 4137 ext4_mb_unload_buddy(&e4b);
4128 list_del(&pa->u.pa_tmp_list); 4138 list_del(&pa->u.pa_tmp_list);
4129 call_rcu(&(pa)->u.pa_rcu, ext4_mb_pa_callback); 4139 call_rcu(&(pa)->u.pa_rcu, ext4_mb_pa_callback);
4130 } 4140 }
4131 if (ac)
4132 kmem_cache_free(ext4_ac_cachep, ac);
4133} 4141}
4134 4142
4135/* 4143/*
@@ -4255,7 +4263,7 @@ static int ext4_mb_discard_preallocations(struct super_block *sb, int needed)
4255 * to usual allocation 4263 * to usual allocation
4256 */ 4264 */
4257ext4_fsblk_t ext4_mb_new_blocks(handle_t *handle, 4265ext4_fsblk_t ext4_mb_new_blocks(handle_t *handle,
4258 struct ext4_allocation_request *ar, int *errp) 4266 struct ext4_allocation_request *ar, int *errp)
4259{ 4267{
4260 int freed; 4268 int freed;
4261 struct ext4_allocation_context *ac = NULL; 4269 struct ext4_allocation_context *ac = NULL;
@@ -4299,7 +4307,7 @@ ext4_fsblk_t ext4_mb_new_blocks(handle_t *handle,
4299 inquota = ar->len; 4307 inquota = ar->len;
4300 if (ar->len == 0) { 4308 if (ar->len == 0) {
4301 *errp = -EDQUOT; 4309 *errp = -EDQUOT;
4302 goto out3; 4310 goto out;
4303 } 4311 }
4304 } 4312 }
4305 4313
@@ -4307,13 +4315,13 @@ ext4_fsblk_t ext4_mb_new_blocks(handle_t *handle,
4307 if (!ac) { 4315 if (!ac) {
4308 ar->len = 0; 4316 ar->len = 0;
4309 *errp = -ENOMEM; 4317 *errp = -ENOMEM;
4310 goto out1; 4318 goto out;
4311 } 4319 }
4312 4320
4313 *errp = ext4_mb_initialize_context(ac, ar); 4321 *errp = ext4_mb_initialize_context(ac, ar);
4314 if (*errp) { 4322 if (*errp) {
4315 ar->len = 0; 4323 ar->len = 0;
4316 goto out2; 4324 goto out;
4317 } 4325 }
4318 4326
4319 ac->ac_op = EXT4_MB_HISTORY_PREALLOC; 4327 ac->ac_op = EXT4_MB_HISTORY_PREALLOC;
@@ -4322,7 +4330,9 @@ ext4_fsblk_t ext4_mb_new_blocks(handle_t *handle,
4322 ext4_mb_normalize_request(ac, ar); 4330 ext4_mb_normalize_request(ac, ar);
4323repeat: 4331repeat:
4324 /* allocate space in core */ 4332 /* allocate space in core */
4325 ext4_mb_regular_allocator(ac); 4333 *errp = ext4_mb_regular_allocator(ac);
4334 if (*errp)
4335 goto errout;
4326 4336
4327 /* as we've just preallocated more space than 4337 /* as we've just preallocated more space than
4328 * user requested orinally, we store allocated 4338 * user requested orinally, we store allocated
@@ -4333,7 +4343,7 @@ repeat:
4333 } 4343 }
4334 if (likely(ac->ac_status == AC_STATUS_FOUND)) { 4344 if (likely(ac->ac_status == AC_STATUS_FOUND)) {
4335 *errp = ext4_mb_mark_diskspace_used(ac, handle, reserv_blks); 4345 *errp = ext4_mb_mark_diskspace_used(ac, handle, reserv_blks);
4336 if (*errp == -EAGAIN) { 4346 if (*errp == -EAGAIN) {
4337 /* 4347 /*
4338 * drop the reference that we took 4348 * drop the reference that we took
4339 * in ext4_mb_use_best_found 4349 * in ext4_mb_use_best_found
@@ -4344,12 +4354,10 @@ repeat:
4344 ac->ac_b_ex.fe_len = 0; 4354 ac->ac_b_ex.fe_len = 0;
4345 ac->ac_status = AC_STATUS_CONTINUE; 4355 ac->ac_status = AC_STATUS_CONTINUE;
4346 goto repeat; 4356 goto repeat;
4347 } else if (*errp) { 4357 } else if (*errp)
4358 errout:
4348 ext4_discard_allocated_blocks(ac); 4359 ext4_discard_allocated_blocks(ac);
4349 ac->ac_b_ex.fe_len = 0; 4360 else {
4350 ar->len = 0;
4351 ext4_mb_show_ac(ac);
4352 } else {
4353 block = ext4_grp_offs_to_block(sb, &ac->ac_b_ex); 4361 block = ext4_grp_offs_to_block(sb, &ac->ac_b_ex);
4354 ar->len = ac->ac_b_ex.fe_len; 4362 ar->len = ac->ac_b_ex.fe_len;
4355 } 4363 }
@@ -4358,19 +4366,19 @@ repeat:
4358 if (freed) 4366 if (freed)
4359 goto repeat; 4367 goto repeat;
4360 *errp = -ENOSPC; 4368 *errp = -ENOSPC;
4369 }
4370
4371 if (*errp) {
4361 ac->ac_b_ex.fe_len = 0; 4372 ac->ac_b_ex.fe_len = 0;
4362 ar->len = 0; 4373 ar->len = 0;
4363 ext4_mb_show_ac(ac); 4374 ext4_mb_show_ac(ac);
4364 } 4375 }
4365
4366 ext4_mb_release_context(ac); 4376 ext4_mb_release_context(ac);
4367 4377out:
4368out2: 4378 if (ac)
4369 kmem_cache_free(ext4_ac_cachep, ac); 4379 kmem_cache_free(ext4_ac_cachep, ac);
4370out1:
4371 if (inquota && ar->len < inquota) 4380 if (inquota && ar->len < inquota)
4372 dquot_free_block(ar->inode, inquota - ar->len); 4381 dquot_free_block(ar->inode, inquota - ar->len);
4373out3:
4374 if (!ar->len) { 4382 if (!ar->len) {
4375 if (!EXT4_I(ar->inode)->i_delalloc_reserved_flag) 4383 if (!EXT4_I(ar->inode)->i_delalloc_reserved_flag)
4376 /* release all the reserved blocks if non delalloc */ 4384 /* release all the reserved blocks if non delalloc */
@@ -4402,6 +4410,7 @@ static noinline_for_stack int
4402ext4_mb_free_metadata(handle_t *handle, struct ext4_buddy *e4b, 4410ext4_mb_free_metadata(handle_t *handle, struct ext4_buddy *e4b,
4403 struct ext4_free_data *new_entry) 4411 struct ext4_free_data *new_entry)
4404{ 4412{
4413 ext4_group_t group = e4b->bd_group;
4405 ext4_grpblk_t block; 4414 ext4_grpblk_t block;
4406 struct ext4_free_data *entry; 4415 struct ext4_free_data *entry;
4407 struct ext4_group_info *db = e4b->bd_info; 4416 struct ext4_group_info *db = e4b->bd_info;
@@ -4434,9 +4443,9 @@ ext4_mb_free_metadata(handle_t *handle, struct ext4_buddy *e4b,
4434 else if (block >= (entry->start_blk + entry->count)) 4443 else if (block >= (entry->start_blk + entry->count))
4435 n = &(*n)->rb_right; 4444 n = &(*n)->rb_right;
4436 else { 4445 else {
4437 ext4_grp_locked_error(sb, e4b->bd_group, __func__, 4446 ext4_grp_locked_error(sb, group, 0,
4438 "Double free of blocks %d (%d %d)", 4447 ext4_group_first_block_no(sb, group) + block,
4439 block, entry->start_blk, entry->count); 4448 "Block already on to-be-freed list");
4440 return 0; 4449 return 0;
4441 } 4450 }
4442 } 4451 }
@@ -4492,9 +4501,7 @@ void ext4_free_blocks(handle_t *handle, struct inode *inode,
4492{ 4501{
4493 struct buffer_head *bitmap_bh = NULL; 4502 struct buffer_head *bitmap_bh = NULL;
4494 struct super_block *sb = inode->i_sb; 4503 struct super_block *sb = inode->i_sb;
4495 struct ext4_allocation_context *ac = NULL;
4496 struct ext4_group_desc *gdp; 4504 struct ext4_group_desc *gdp;
4497 struct ext4_super_block *es;
4498 unsigned long freed = 0; 4505 unsigned long freed = 0;
4499 unsigned int overflow; 4506 unsigned int overflow;
4500 ext4_grpblk_t bit; 4507 ext4_grpblk_t bit;
@@ -4513,7 +4520,6 @@ void ext4_free_blocks(handle_t *handle, struct inode *inode,
4513 } 4520 }
4514 4521
4515 sbi = EXT4_SB(sb); 4522 sbi = EXT4_SB(sb);
4516 es = EXT4_SB(sb)->s_es;
4517 if (!(flags & EXT4_FREE_BLOCKS_VALIDATED) && 4523 if (!(flags & EXT4_FREE_BLOCKS_VALIDATED) &&
4518 !ext4_data_block_valid(sbi, block, count)) { 4524 !ext4_data_block_valid(sbi, block, count)) {
4519 ext4_error(sb, "Freeing blocks not in datazone - " 4525 ext4_error(sb, "Freeing blocks not in datazone - "
@@ -4534,6 +4540,8 @@ void ext4_free_blocks(handle_t *handle, struct inode *inode,
4534 if (!bh) 4540 if (!bh)
4535 tbh = sb_find_get_block(inode->i_sb, 4541 tbh = sb_find_get_block(inode->i_sb,
4536 block + i); 4542 block + i);
4543 if (unlikely(!tbh))
4544 continue;
4537 ext4_forget(handle, flags & EXT4_FREE_BLOCKS_METADATA, 4545 ext4_forget(handle, flags & EXT4_FREE_BLOCKS_METADATA,
4538 inode, tbh, block + i); 4546 inode, tbh, block + i);
4539 } 4547 }
@@ -4549,12 +4557,6 @@ void ext4_free_blocks(handle_t *handle, struct inode *inode,
4549 if (!ext4_should_writeback_data(inode)) 4557 if (!ext4_should_writeback_data(inode))
4550 flags |= EXT4_FREE_BLOCKS_METADATA; 4558 flags |= EXT4_FREE_BLOCKS_METADATA;
4551 4559
4552 ac = kmem_cache_alloc(ext4_ac_cachep, GFP_NOFS);
4553 if (ac) {
4554 ac->ac_inode = inode;
4555 ac->ac_sb = sb;
4556 }
4557
4558do_more: 4560do_more:
4559 overflow = 0; 4561 overflow = 0;
4560 ext4_get_group_no_and_offset(sb, block, &block_group, &bit); 4562 ext4_get_group_no_and_offset(sb, block, &block_group, &bit);
@@ -4612,12 +4614,7 @@ do_more:
4612 BUG_ON(!mb_test_bit(bit + i, bitmap_bh->b_data)); 4614 BUG_ON(!mb_test_bit(bit + i, bitmap_bh->b_data));
4613 } 4615 }
4614#endif 4616#endif
4615 if (ac) { 4617 trace_ext4_mballoc_free(sb, inode, block_group, bit, count);
4616 ac->ac_b_ex.fe_group = block_group;
4617 ac->ac_b_ex.fe_start = bit;
4618 ac->ac_b_ex.fe_len = count;
4619 trace_ext4_mballoc_free(ac);
4620 }
4621 4618
4622 err = ext4_mb_load_buddy(sb, block_group, &e4b); 4619 err = ext4_mb_load_buddy(sb, block_group, &e4b);
4623 if (err) 4620 if (err)
@@ -4643,6 +4640,8 @@ do_more:
4643 * with group lock held. generate_buddy look at 4640 * with group lock held. generate_buddy look at
4644 * them with group lock_held 4641 * them with group lock_held
4645 */ 4642 */
4643 if (test_opt(sb, DISCARD))
4644 ext4_issue_discard(sb, block_group, bit, count);
4646 ext4_lock_group(sb, block_group); 4645 ext4_lock_group(sb, block_group);
4647 mb_clear_bits(bitmap_bh->b_data, bit, count); 4646 mb_clear_bits(bitmap_bh->b_data, bit, count);
4648 mb_free_blocks(inode, &e4b, bit, count); 4647 mb_free_blocks(inode, &e4b, bit, count);
@@ -4680,13 +4679,196 @@ do_more:
4680 put_bh(bitmap_bh); 4679 put_bh(bitmap_bh);
4681 goto do_more; 4680 goto do_more;
4682 } 4681 }
4683 sb->s_dirt = 1; 4682 ext4_mark_super_dirty(sb);
4684error_return: 4683error_return:
4685 if (freed) 4684 if (freed)
4686 dquot_free_block(inode, freed); 4685 dquot_free_block(inode, freed);
4687 brelse(bitmap_bh); 4686 brelse(bitmap_bh);
4688 ext4_std_error(sb, err); 4687 ext4_std_error(sb, err);
4689 if (ac)
4690 kmem_cache_free(ext4_ac_cachep, ac);
4691 return; 4688 return;
4692} 4689}
4690
4691/**
4692 * ext4_trim_extent -- function to TRIM one single free extent in the group
4693 * @sb: super block for the file system
4694 * @start: starting block of the free extent in the alloc. group
4695 * @count: number of blocks to TRIM
4696 * @group: alloc. group we are working with
4697 * @e4b: ext4 buddy for the group
4698 *
4699 * Trim "count" blocks starting at "start" in the "group". To assure that no
4700 * one will allocate those blocks, mark it as used in buddy bitmap. This must
4701 * be called with under the group lock.
4702 */
4703static int ext4_trim_extent(struct super_block *sb, int start, int count,
4704 ext4_group_t group, struct ext4_buddy *e4b)
4705{
4706 struct ext4_free_extent ex;
4707 int ret = 0;
4708
4709 assert_spin_locked(ext4_group_lock_ptr(sb, group));
4710
4711 ex.fe_start = start;
4712 ex.fe_group = group;
4713 ex.fe_len = count;
4714
4715 /*
4716 * Mark blocks used, so no one can reuse them while
4717 * being trimmed.
4718 */
4719 mb_mark_used(e4b, &ex);
4720 ext4_unlock_group(sb, group);
4721
4722 ret = ext4_issue_discard(sb, group, start, count);
4723 if (ret)
4724 ext4_std_error(sb, ret);
4725
4726 ext4_lock_group(sb, group);
4727 mb_free_blocks(NULL, e4b, start, ex.fe_len);
4728 return ret;
4729}
4730
4731/**
4732 * ext4_trim_all_free -- function to trim all free space in alloc. group
4733 * @sb: super block for file system
4734 * @e4b: ext4 buddy
4735 * @start: first group block to examine
4736 * @max: last group block to examine
4737 * @minblocks: minimum extent block count
4738 *
4739 * ext4_trim_all_free walks through group's buddy bitmap searching for free
4740 * extents. When the free block is found, ext4_trim_extent is called to TRIM
4741 * the extent.
4742 *
4743 *
4744 * ext4_trim_all_free walks through group's block bitmap searching for free
4745 * extents. When the free extent is found, mark it as used in group buddy
4746 * bitmap. Then issue a TRIM command on this extent and free the extent in
4747 * the group buddy bitmap. This is done until whole group is scanned.
4748 */
4749ext4_grpblk_t ext4_trim_all_free(struct super_block *sb, struct ext4_buddy *e4b,
4750 ext4_grpblk_t start, ext4_grpblk_t max, ext4_grpblk_t minblocks)
4751{
4752 void *bitmap;
4753 ext4_grpblk_t next, count = 0;
4754 ext4_group_t group;
4755 int ret = 0;
4756
4757 BUG_ON(e4b == NULL);
4758
4759 bitmap = e4b->bd_bitmap;
4760 group = e4b->bd_group;
4761 start = (e4b->bd_info->bb_first_free > start) ?
4762 e4b->bd_info->bb_first_free : start;
4763 ext4_lock_group(sb, group);
4764
4765 while (start < max) {
4766 start = mb_find_next_zero_bit(bitmap, max, start);
4767 if (start >= max)
4768 break;
4769 next = mb_find_next_bit(bitmap, max, start);
4770
4771 if ((next - start) >= minblocks) {
4772 ret = ext4_trim_extent(sb, start,
4773 next - start, group, e4b);
4774 if (ret < 0)
4775 break;
4776 count += next - start;
4777 }
4778 start = next + 1;
4779
4780 if (fatal_signal_pending(current)) {
4781 count = -ERESTARTSYS;
4782 break;
4783 }
4784
4785 if (need_resched()) {
4786 ext4_unlock_group(sb, group);
4787 cond_resched();
4788 ext4_lock_group(sb, group);
4789 }
4790
4791 if ((e4b->bd_info->bb_free - count) < minblocks)
4792 break;
4793 }
4794 ext4_unlock_group(sb, group);
4795
4796 ext4_debug("trimmed %d blocks in the group %d\n",
4797 count, group);
4798
4799 if (ret < 0)
4800 count = ret;
4801
4802 return count;
4803}
4804
4805/**
4806 * ext4_trim_fs() -- trim ioctl handle function
4807 * @sb: superblock for filesystem
4808 * @range: fstrim_range structure
4809 *
4810 * start: First Byte to trim
4811 * len: number of Bytes to trim from start
4812 * minlen: minimum extent length in Bytes
4813 * ext4_trim_fs goes through all allocation groups containing Bytes from
4814 * start to start+len. For each such a group ext4_trim_all_free function
4815 * is invoked to trim all free space.
4816 */
4817int ext4_trim_fs(struct super_block *sb, struct fstrim_range *range)
4818{
4819 struct ext4_buddy e4b;
4820 ext4_group_t first_group, last_group;
4821 ext4_group_t group, ngroups = ext4_get_groups_count(sb);
4822 ext4_grpblk_t cnt = 0, first_block, last_block;
4823 uint64_t start, len, minlen, trimmed;
4824 int ret = 0;
4825
4826 start = range->start >> sb->s_blocksize_bits;
4827 len = range->len >> sb->s_blocksize_bits;
4828 minlen = range->minlen >> sb->s_blocksize_bits;
4829 trimmed = 0;
4830
4831 if (unlikely(minlen > EXT4_BLOCKS_PER_GROUP(sb)))
4832 return -EINVAL;
4833
4834 /* Determine first and last group to examine based on start and len */
4835 ext4_get_group_no_and_offset(sb, (ext4_fsblk_t) start,
4836 &first_group, &first_block);
4837 ext4_get_group_no_and_offset(sb, (ext4_fsblk_t) (start + len),
4838 &last_group, &last_block);
4839 last_group = (last_group > ngroups - 1) ? ngroups - 1 : last_group;
4840 last_block = EXT4_BLOCKS_PER_GROUP(sb);
4841
4842 if (first_group > last_group)
4843 return -EINVAL;
4844
4845 for (group = first_group; group <= last_group; group++) {
4846 ret = ext4_mb_load_buddy(sb, group, &e4b);
4847 if (ret) {
4848 ext4_error(sb, "Error in loading buddy "
4849 "information for %u", group);
4850 break;
4851 }
4852
4853 if (len >= EXT4_BLOCKS_PER_GROUP(sb))
4854 len -= (EXT4_BLOCKS_PER_GROUP(sb) - first_block);
4855 else
4856 last_block = len;
4857
4858 if (e4b.bd_info->bb_free >= minlen) {
4859 cnt = ext4_trim_all_free(sb, &e4b, first_block,
4860 last_block, minlen);
4861 if (cnt < 0) {
4862 ret = cnt;
4863 ext4_mb_unload_buddy(&e4b);
4864 break;
4865 }
4866 }
4867 ext4_mb_unload_buddy(&e4b);
4868 trimmed += cnt;
4869 first_block = 0;
4870 }
4871 range->len = trimmed * sb->s_blocksize;
4872
4873 return ret;
4874}
diff --git a/fs/ext4/migrate.c b/fs/ext4/migrate.c
index 6f3a27ec30bf..25f3a974b725 100644
--- a/fs/ext4/migrate.c
+++ b/fs/ext4/migrate.c
@@ -376,7 +376,7 @@ static int ext4_ext_swap_inode_data(handle_t *handle, struct inode *inode,
376 * We have the extent map build with the tmp inode. 376 * We have the extent map build with the tmp inode.
377 * Now copy the i_data across 377 * Now copy the i_data across
378 */ 378 */
379 ei->i_flags |= EXT4_EXTENTS_FL; 379 ext4_test_inode_flag(inode, EXT4_INODE_EXTENTS);
380 memcpy(ei->i_data, tmp_ei->i_data, sizeof(ei->i_data)); 380 memcpy(ei->i_data, tmp_ei->i_data, sizeof(ei->i_data));
381 381
382 /* 382 /*
@@ -412,7 +412,7 @@ static int free_ext_idx(handle_t *handle, struct inode *inode,
412 struct buffer_head *bh; 412 struct buffer_head *bh;
413 struct ext4_extent_header *eh; 413 struct ext4_extent_header *eh;
414 414
415 block = idx_pblock(ix); 415 block = ext4_idx_pblock(ix);
416 bh = sb_bread(inode->i_sb, block); 416 bh = sb_bread(inode->i_sb, block);
417 if (!bh) 417 if (!bh)
418 return -EIO; 418 return -EIO;
diff --git a/fs/ext4/move_extent.c b/fs/ext4/move_extent.c
index 52abfa12762a..b9f3e7862f13 100644
--- a/fs/ext4/move_extent.c
+++ b/fs/ext4/move_extent.c
@@ -85,7 +85,7 @@ mext_next_extent(struct inode *inode, struct ext4_ext_path *path,
85 if (EXT_LAST_EXTENT(path[ppos].p_hdr) > path[ppos].p_ext) { 85 if (EXT_LAST_EXTENT(path[ppos].p_hdr) > path[ppos].p_ext) {
86 /* leaf block */ 86 /* leaf block */
87 *extent = ++path[ppos].p_ext; 87 *extent = ++path[ppos].p_ext;
88 path[ppos].p_block = ext_pblock(path[ppos].p_ext); 88 path[ppos].p_block = ext4_ext_pblock(path[ppos].p_ext);
89 return 0; 89 return 0;
90 } 90 }
91 91
@@ -96,7 +96,7 @@ mext_next_extent(struct inode *inode, struct ext4_ext_path *path,
96 96
97 /* index block */ 97 /* index block */
98 path[ppos].p_idx++; 98 path[ppos].p_idx++;
99 path[ppos].p_block = idx_pblock(path[ppos].p_idx); 99 path[ppos].p_block = ext4_idx_pblock(path[ppos].p_idx);
100 if (path[ppos+1].p_bh) 100 if (path[ppos+1].p_bh)
101 brelse(path[ppos+1].p_bh); 101 brelse(path[ppos+1].p_bh);
102 path[ppos+1].p_bh = 102 path[ppos+1].p_bh =
@@ -111,7 +111,7 @@ mext_next_extent(struct inode *inode, struct ext4_ext_path *path,
111 path[cur_ppos].p_idx = 111 path[cur_ppos].p_idx =
112 EXT_FIRST_INDEX(path[cur_ppos].p_hdr); 112 EXT_FIRST_INDEX(path[cur_ppos].p_hdr);
113 path[cur_ppos].p_block = 113 path[cur_ppos].p_block =
114 idx_pblock(path[cur_ppos].p_idx); 114 ext4_idx_pblock(path[cur_ppos].p_idx);
115 if (path[cur_ppos+1].p_bh) 115 if (path[cur_ppos+1].p_bh)
116 brelse(path[cur_ppos+1].p_bh); 116 brelse(path[cur_ppos+1].p_bh);
117 path[cur_ppos+1].p_bh = sb_bread(inode->i_sb, 117 path[cur_ppos+1].p_bh = sb_bread(inode->i_sb,
@@ -133,7 +133,7 @@ mext_next_extent(struct inode *inode, struct ext4_ext_path *path,
133 path[leaf_ppos].p_ext = *extent = 133 path[leaf_ppos].p_ext = *extent =
134 EXT_FIRST_EXTENT(path[leaf_ppos].p_hdr); 134 EXT_FIRST_EXTENT(path[leaf_ppos].p_hdr);
135 path[leaf_ppos].p_block = 135 path[leaf_ppos].p_block =
136 ext_pblock(path[leaf_ppos].p_ext); 136 ext4_ext_pblock(path[leaf_ppos].p_ext);
137 return 0; 137 return 0;
138 } 138 }
139 } 139 }
@@ -148,17 +148,17 @@ mext_next_extent(struct inode *inode, struct ext4_ext_path *path,
148 */ 148 */
149static int 149static int
150mext_check_null_inode(struct inode *inode1, struct inode *inode2, 150mext_check_null_inode(struct inode *inode1, struct inode *inode2,
151 const char *function) 151 const char *function, unsigned int line)
152{ 152{
153 int ret = 0; 153 int ret = 0;
154 154
155 if (inode1 == NULL) { 155 if (inode1 == NULL) {
156 __ext4_error(inode2->i_sb, function, 156 __ext4_error(inode2->i_sb, function, line,
157 "Both inodes should not be NULL: " 157 "Both inodes should not be NULL: "
158 "inode1 NULL inode2 %lu", inode2->i_ino); 158 "inode1 NULL inode2 %lu", inode2->i_ino);
159 ret = -EIO; 159 ret = -EIO;
160 } else if (inode2 == NULL) { 160 } else if (inode2 == NULL) {
161 __ext4_error(inode1->i_sb, function, 161 __ext4_error(inode1->i_sb, function, line,
162 "Both inodes should not be NULL: " 162 "Both inodes should not be NULL: "
163 "inode1 %lu inode2 NULL", inode1->i_ino); 163 "inode1 %lu inode2 NULL", inode1->i_ino);
164 ret = -EIO; 164 ret = -EIO;
@@ -249,7 +249,7 @@ mext_insert_across_blocks(handle_t *handle, struct inode *orig_inode,
249 */ 249 */
250 o_end->ee_block = end_ext->ee_block; 250 o_end->ee_block = end_ext->ee_block;
251 o_end->ee_len = end_ext->ee_len; 251 o_end->ee_len = end_ext->ee_len;
252 ext4_ext_store_pblock(o_end, ext_pblock(end_ext)); 252 ext4_ext_store_pblock(o_end, ext4_ext_pblock(end_ext));
253 } 253 }
254 254
255 o_start->ee_len = start_ext->ee_len; 255 o_start->ee_len = start_ext->ee_len;
@@ -276,7 +276,7 @@ mext_insert_across_blocks(handle_t *handle, struct inode *orig_inode,
276 */ 276 */
277 o_end->ee_block = end_ext->ee_block; 277 o_end->ee_block = end_ext->ee_block;
278 o_end->ee_len = end_ext->ee_len; 278 o_end->ee_len = end_ext->ee_len;
279 ext4_ext_store_pblock(o_end, ext_pblock(end_ext)); 279 ext4_ext_store_pblock(o_end, ext4_ext_pblock(end_ext));
280 280
281 /* 281 /*
282 * Set 0 to the extent block if new_ext was 282 * Set 0 to the extent block if new_ext was
@@ -361,7 +361,7 @@ mext_insert_inside_block(struct ext4_extent *o_start,
361 /* Insert new entry */ 361 /* Insert new entry */
362 if (new_ext->ee_len) { 362 if (new_ext->ee_len) {
363 o_start[i] = *new_ext; 363 o_start[i] = *new_ext;
364 ext4_ext_store_pblock(&o_start[i++], ext_pblock(new_ext)); 364 ext4_ext_store_pblock(&o_start[i++], ext4_ext_pblock(new_ext));
365 } 365 }
366 366
367 /* Insert end entry */ 367 /* Insert end entry */
@@ -488,7 +488,7 @@ mext_leaf_block(handle_t *handle, struct inode *orig_inode,
488 start_ext.ee_len = end_ext.ee_len = 0; 488 start_ext.ee_len = end_ext.ee_len = 0;
489 489
490 new_ext.ee_block = cpu_to_le32(*from); 490 new_ext.ee_block = cpu_to_le32(*from);
491 ext4_ext_store_pblock(&new_ext, ext_pblock(dext)); 491 ext4_ext_store_pblock(&new_ext, ext4_ext_pblock(dext));
492 new_ext.ee_len = dext->ee_len; 492 new_ext.ee_len = dext->ee_len;
493 new_ext_alen = ext4_ext_get_actual_len(&new_ext); 493 new_ext_alen = ext4_ext_get_actual_len(&new_ext);
494 new_ext_end = le32_to_cpu(new_ext.ee_block) + new_ext_alen - 1; 494 new_ext_end = le32_to_cpu(new_ext.ee_block) + new_ext_alen - 1;
@@ -553,7 +553,7 @@ mext_leaf_block(handle_t *handle, struct inode *orig_inode,
553 copy_extent_status(oext, &end_ext); 553 copy_extent_status(oext, &end_ext);
554 end_ext_alen = ext4_ext_get_actual_len(&end_ext); 554 end_ext_alen = ext4_ext_get_actual_len(&end_ext);
555 ext4_ext_store_pblock(&end_ext, 555 ext4_ext_store_pblock(&end_ext,
556 (ext_pblock(o_end) + oext_alen - end_ext_alen)); 556 (ext4_ext_pblock(o_end) + oext_alen - end_ext_alen));
557 end_ext.ee_block = 557 end_ext.ee_block =
558 cpu_to_le32(le32_to_cpu(o_end->ee_block) + 558 cpu_to_le32(le32_to_cpu(o_end->ee_block) +
559 oext_alen - end_ext_alen); 559 oext_alen - end_ext_alen);
@@ -604,7 +604,7 @@ mext_calc_swap_extents(struct ext4_extent *tmp_dext,
604 /* When tmp_dext is too large, pick up the target range. */ 604 /* When tmp_dext is too large, pick up the target range. */
605 diff = donor_off - le32_to_cpu(tmp_dext->ee_block); 605 diff = donor_off - le32_to_cpu(tmp_dext->ee_block);
606 606
607 ext4_ext_store_pblock(tmp_dext, ext_pblock(tmp_dext) + diff); 607 ext4_ext_store_pblock(tmp_dext, ext4_ext_pblock(tmp_dext) + diff);
608 tmp_dext->ee_block = 608 tmp_dext->ee_block =
609 cpu_to_le32(le32_to_cpu(tmp_dext->ee_block) + diff); 609 cpu_to_le32(le32_to_cpu(tmp_dext->ee_block) + diff);
610 tmp_dext->ee_len = cpu_to_le16(le16_to_cpu(tmp_dext->ee_len) - diff); 610 tmp_dext->ee_len = cpu_to_le16(le16_to_cpu(tmp_dext->ee_len) - diff);
@@ -613,7 +613,7 @@ mext_calc_swap_extents(struct ext4_extent *tmp_dext,
613 tmp_dext->ee_len = cpu_to_le16(max_count); 613 tmp_dext->ee_len = cpu_to_le16(max_count);
614 614
615 orig_diff = orig_off - le32_to_cpu(tmp_oext->ee_block); 615 orig_diff = orig_off - le32_to_cpu(tmp_oext->ee_block);
616 ext4_ext_store_pblock(tmp_oext, ext_pblock(tmp_oext) + orig_diff); 616 ext4_ext_store_pblock(tmp_oext, ext4_ext_pblock(tmp_oext) + orig_diff);
617 617
618 /* Adjust extent length if donor extent is larger than orig */ 618 /* Adjust extent length if donor extent is larger than orig */
619 if (ext4_ext_get_actual_len(tmp_dext) > 619 if (ext4_ext_get_actual_len(tmp_dext) >
@@ -1084,7 +1084,7 @@ mext_inode_double_lock(struct inode *inode1, struct inode *inode2)
1084 1084
1085 BUG_ON(inode1 == NULL && inode2 == NULL); 1085 BUG_ON(inode1 == NULL && inode2 == NULL);
1086 1086
1087 ret = mext_check_null_inode(inode1, inode2, __func__); 1087 ret = mext_check_null_inode(inode1, inode2, __func__, __LINE__);
1088 if (ret < 0) 1088 if (ret < 0)
1089 goto out; 1089 goto out;
1090 1090
@@ -1121,7 +1121,7 @@ mext_inode_double_unlock(struct inode *inode1, struct inode *inode2)
1121 1121
1122 BUG_ON(inode1 == NULL && inode2 == NULL); 1122 BUG_ON(inode1 == NULL && inode2 == NULL);
1123 1123
1124 ret = mext_check_null_inode(inode1, inode2, __func__); 1124 ret = mext_check_null_inode(inode1, inode2, __func__, __LINE__);
1125 if (ret < 0) 1125 if (ret < 0)
1126 goto out; 1126 goto out;
1127 1127
diff --git a/fs/ext4/namei.c b/fs/ext4/namei.c
index a43e6617b351..92203b8a099f 100644
--- a/fs/ext4/namei.c
+++ b/fs/ext4/namei.c
@@ -179,30 +179,6 @@ static struct buffer_head * ext4_dx_find_entry(struct inode *dir,
179static int ext4_dx_add_entry(handle_t *handle, struct dentry *dentry, 179static int ext4_dx_add_entry(handle_t *handle, struct dentry *dentry,
180 struct inode *inode); 180 struct inode *inode);
181 181
182unsigned int ext4_rec_len_from_disk(__le16 dlen, unsigned blocksize)
183{
184 unsigned len = le16_to_cpu(dlen);
185
186 if (len == EXT4_MAX_REC_LEN || len == 0)
187 return blocksize;
188 return (len & 65532) | ((len & 3) << 16);
189}
190
191__le16 ext4_rec_len_to_disk(unsigned len, unsigned blocksize)
192{
193 if ((len > blocksize) || (blocksize > (1 << 18)) || (len & 3))
194 BUG();
195 if (len < 65536)
196 return cpu_to_le16(len);
197 if (len == blocksize) {
198 if (blocksize == 65536)
199 return cpu_to_le16(EXT4_MAX_REC_LEN);
200 else
201 return cpu_to_le16(0);
202 }
203 return cpu_to_le16((len & 65532) | ((len >> 16) & 3));
204}
205
206/* 182/*
207 * p is at least 6 bytes before the end of page 183 * p is at least 6 bytes before the end of page
208 */ 184 */
@@ -605,7 +581,7 @@ static int htree_dirblock_to_tree(struct file *dir_file,
605 dir->i_sb->s_blocksize - 581 dir->i_sb->s_blocksize -
606 EXT4_DIR_REC_LEN(0)); 582 EXT4_DIR_REC_LEN(0));
607 for (; de < top; de = ext4_next_entry(de, dir->i_sb->s_blocksize)) { 583 for (; de < top; de = ext4_next_entry(de, dir->i_sb->s_blocksize)) {
608 if (!ext4_check_dir_entry("htree_dirblock_to_tree", dir, de, bh, 584 if (!ext4_check_dir_entry(dir, de, bh,
609 (block<<EXT4_BLOCK_SIZE_BITS(dir->i_sb)) 585 (block<<EXT4_BLOCK_SIZE_BITS(dir->i_sb))
610 +((char *)de - bh->b_data))) { 586 +((char *)de - bh->b_data))) {
611 /* On error, skip the f_pos to the next block. */ 587 /* On error, skip the f_pos to the next block. */
@@ -844,8 +820,7 @@ static inline int search_dirblock(struct buffer_head *bh,
844 if ((char *) de + namelen <= dlimit && 820 if ((char *) de + namelen <= dlimit &&
845 ext4_match (namelen, name, de)) { 821 ext4_match (namelen, name, de)) {
846 /* found a match - just to be sure, do a full check */ 822 /* found a match - just to be sure, do a full check */
847 if (!ext4_check_dir_entry("ext4_find_entry", 823 if (!ext4_check_dir_entry(dir, de, bh, offset))
848 dir, de, bh, offset))
849 return -1; 824 return -1;
850 *res_dir = de; 825 *res_dir = de;
851 return 1; 826 return 1;
@@ -881,6 +856,7 @@ static struct buffer_head * ext4_find_entry (struct inode *dir,
881 struct buffer_head *bh_use[NAMEI_RA_SIZE]; 856 struct buffer_head *bh_use[NAMEI_RA_SIZE];
882 struct buffer_head *bh, *ret = NULL; 857 struct buffer_head *bh, *ret = NULL;
883 ext4_lblk_t start, block, b; 858 ext4_lblk_t start, block, b;
859 const u8 *name = d_name->name;
884 int ra_max = 0; /* Number of bh's in the readahead 860 int ra_max = 0; /* Number of bh's in the readahead
885 buffer, bh_use[] */ 861 buffer, bh_use[] */
886 int ra_ptr = 0; /* Current index into readahead 862 int ra_ptr = 0; /* Current index into readahead
@@ -895,6 +871,16 @@ static struct buffer_head * ext4_find_entry (struct inode *dir,
895 namelen = d_name->len; 871 namelen = d_name->len;
896 if (namelen > EXT4_NAME_LEN) 872 if (namelen > EXT4_NAME_LEN)
897 return NULL; 873 return NULL;
874 if ((namelen <= 2) && (name[0] == '.') &&
875 (name[1] == '.' || name[1] == '0')) {
876 /*
877 * "." or ".." will only be in the first block
878 * NFS may look up ".."; "." should be handled by the VFS
879 */
880 block = start = 0;
881 nblocks = 1;
882 goto restart;
883 }
898 if (is_dx(dir)) { 884 if (is_dx(dir)) {
899 bh = ext4_dx_find_entry(dir, d_name, res_dir, &err); 885 bh = ext4_dx_find_entry(dir, d_name, res_dir, &err);
900 /* 886 /*
@@ -985,55 +971,35 @@ cleanup_and_exit:
985static struct buffer_head * ext4_dx_find_entry(struct inode *dir, const struct qstr *d_name, 971static struct buffer_head * ext4_dx_find_entry(struct inode *dir, const struct qstr *d_name,
986 struct ext4_dir_entry_2 **res_dir, int *err) 972 struct ext4_dir_entry_2 **res_dir, int *err)
987{ 973{
988 struct super_block * sb; 974 struct super_block * sb = dir->i_sb;
989 struct dx_hash_info hinfo; 975 struct dx_hash_info hinfo;
990 u32 hash;
991 struct dx_frame frames[2], *frame; 976 struct dx_frame frames[2], *frame;
992 struct ext4_dir_entry_2 *de, *top;
993 struct buffer_head *bh; 977 struct buffer_head *bh;
994 ext4_lblk_t block; 978 ext4_lblk_t block;
995 int retval; 979 int retval;
996 int namelen = d_name->len;
997 const u8 *name = d_name->name;
998 980
999 sb = dir->i_sb; 981 if (!(frame = dx_probe(d_name, dir, &hinfo, frames, err)))
1000 /* NFS may look up ".." - look at dx_root directory block */ 982 return NULL;
1001 if (namelen > 2 || name[0] != '.'||(name[1] != '.' && name[1] != '\0')){
1002 if (!(frame = dx_probe(d_name, dir, &hinfo, frames, err)))
1003 return NULL;
1004 } else {
1005 frame = frames;
1006 frame->bh = NULL; /* for dx_release() */
1007 frame->at = (struct dx_entry *)frames; /* hack for zero entry*/
1008 dx_set_block(frame->at, 0); /* dx_root block is 0 */
1009 }
1010 hash = hinfo.hash;
1011 do { 983 do {
1012 block = dx_get_block(frame->at); 984 block = dx_get_block(frame->at);
1013 if (!(bh = ext4_bread (NULL,dir, block, 0, err))) 985 if (!(bh = ext4_bread(NULL, dir, block, 0, err)))
1014 goto errout; 986 goto errout;
1015 de = (struct ext4_dir_entry_2 *) bh->b_data;
1016 top = (struct ext4_dir_entry_2 *) ((char *) de + sb->s_blocksize -
1017 EXT4_DIR_REC_LEN(0));
1018 for (; de < top; de = ext4_next_entry(de, sb->s_blocksize)) {
1019 int off = (block << EXT4_BLOCK_SIZE_BITS(sb))
1020 + ((char *) de - bh->b_data);
1021
1022 if (!ext4_check_dir_entry(__func__, dir, de, bh, off)) {
1023 brelse(bh);
1024 *err = ERR_BAD_DX_DIR;
1025 goto errout;
1026 }
1027 987
1028 if (ext4_match(namelen, name, de)) { 988 retval = search_dirblock(bh, dir, d_name,
1029 *res_dir = de; 989 block << EXT4_BLOCK_SIZE_BITS(sb),
1030 dx_release(frames); 990 res_dir);
1031 return bh; 991 if (retval == 1) { /* Success! */
1032 } 992 dx_release(frames);
993 return bh;
1033 } 994 }
1034 brelse(bh); 995 brelse(bh);
996 if (retval == -1) {
997 *err = ERR_BAD_DX_DIR;
998 goto errout;
999 }
1000
1035 /* Check to see if we should continue to search */ 1001 /* Check to see if we should continue to search */
1036 retval = ext4_htree_next_block(dir, hash, frame, 1002 retval = ext4_htree_next_block(dir, hinfo.hash, frame,
1037 frames, NULL); 1003 frames, NULL);
1038 if (retval < 0) { 1004 if (retval < 0) {
1039 ext4_warning(sb, 1005 ext4_warning(sb,
@@ -1088,7 +1054,6 @@ static struct dentry *ext4_lookup(struct inode *dir, struct dentry *dentry, stru
1088struct dentry *ext4_get_parent(struct dentry *child) 1054struct dentry *ext4_get_parent(struct dentry *child)
1089{ 1055{
1090 __u32 ino; 1056 __u32 ino;
1091 struct inode *inode;
1092 static const struct qstr dotdot = { 1057 static const struct qstr dotdot = {
1093 .name = "..", 1058 .name = "..",
1094 .len = 2, 1059 .len = 2,
@@ -1097,7 +1062,6 @@ struct dentry *ext4_get_parent(struct dentry *child)
1097 struct buffer_head *bh; 1062 struct buffer_head *bh;
1098 1063
1099 bh = ext4_find_entry(child->d_inode, &dotdot, &de); 1064 bh = ext4_find_entry(child->d_inode, &dotdot, &de);
1100 inode = NULL;
1101 if (!bh) 1065 if (!bh)
1102 return ERR_PTR(-ENOENT); 1066 return ERR_PTR(-ENOENT);
1103 ino = le32_to_cpu(de->inode); 1067 ino = le32_to_cpu(de->inode);
@@ -1305,8 +1269,7 @@ static int add_dirent_to_buf(handle_t *handle, struct dentry *dentry,
1305 de = (struct ext4_dir_entry_2 *)bh->b_data; 1269 de = (struct ext4_dir_entry_2 *)bh->b_data;
1306 top = bh->b_data + blocksize - reclen; 1270 top = bh->b_data + blocksize - reclen;
1307 while ((char *) de <= top) { 1271 while ((char *) de <= top) {
1308 if (!ext4_check_dir_entry("ext4_add_entry", dir, de, 1272 if (!ext4_check_dir_entry(dir, de, bh, offset))
1309 bh, offset))
1310 return -EIO; 1273 return -EIO;
1311 if (ext4_match(namelen, name, de)) 1274 if (ext4_match(namelen, name, de))
1312 return -EEXIST; 1275 return -EEXIST;
@@ -1673,7 +1636,7 @@ static int ext4_delete_entry(handle_t *handle,
1673 pde = NULL; 1636 pde = NULL;
1674 de = (struct ext4_dir_entry_2 *) bh->b_data; 1637 de = (struct ext4_dir_entry_2 *) bh->b_data;
1675 while (i < bh->b_size) { 1638 while (i < bh->b_size) {
1676 if (!ext4_check_dir_entry("ext4_delete_entry", dir, de, bh, i)) 1639 if (!ext4_check_dir_entry(dir, de, bh, i))
1677 return -EIO; 1640 return -EIO;
1678 if (de == de_del) { 1641 if (de == de_del) {
1679 BUFFER_TRACE(bh, "get_write_access"); 1642 BUFFER_TRACE(bh, "get_write_access");
@@ -1956,7 +1919,7 @@ static int empty_dir(struct inode *inode)
1956 } 1919 }
1957 de = (struct ext4_dir_entry_2 *) bh->b_data; 1920 de = (struct ext4_dir_entry_2 *) bh->b_data;
1958 } 1921 }
1959 if (!ext4_check_dir_entry("empty_dir", inode, de, bh, offset)) { 1922 if (!ext4_check_dir_entry(inode, de, bh, offset)) {
1960 de = (struct ext4_dir_entry_2 *)(bh->b_data + 1923 de = (struct ext4_dir_entry_2 *)(bh->b_data +
1961 sb->s_blocksize); 1924 sb->s_blocksize);
1962 offset = (offset | (sb->s_blocksize - 1)) + 1; 1925 offset = (offset | (sb->s_blocksize - 1)) + 1;
@@ -2340,7 +2303,7 @@ retry:
2340 2303
2341 inode->i_ctime = ext4_current_time(inode); 2304 inode->i_ctime = ext4_current_time(inode);
2342 ext4_inc_count(handle, inode); 2305 ext4_inc_count(handle, inode);
2343 atomic_inc(&inode->i_count); 2306 ihold(inode);
2344 2307
2345 err = ext4_add_entry(handle, dentry, inode); 2308 err = ext4_add_entry(handle, dentry, inode);
2346 if (!err) { 2309 if (!err) {
diff --git a/fs/ext4/page-io.c b/fs/ext4/page-io.c
new file mode 100644
index 000000000000..46a7d6a9d976
--- /dev/null
+++ b/fs/ext4/page-io.c
@@ -0,0 +1,430 @@
1/*
2 * linux/fs/ext4/page-io.c
3 *
4 * This contains the new page_io functions for ext4
5 *
6 * Written by Theodore Ts'o, 2010.
7 */
8
9#include <linux/module.h>
10#include <linux/fs.h>
11#include <linux/time.h>
12#include <linux/jbd2.h>
13#include <linux/highuid.h>
14#include <linux/pagemap.h>
15#include <linux/quotaops.h>
16#include <linux/string.h>
17#include <linux/buffer_head.h>
18#include <linux/writeback.h>
19#include <linux/pagevec.h>
20#include <linux/mpage.h>
21#include <linux/namei.h>
22#include <linux/uio.h>
23#include <linux/bio.h>
24#include <linux/workqueue.h>
25#include <linux/kernel.h>
26#include <linux/slab.h>
27
28#include "ext4_jbd2.h"
29#include "xattr.h"
30#include "acl.h"
31#include "ext4_extents.h"
32
33static struct kmem_cache *io_page_cachep, *io_end_cachep;
34
35int __init ext4_init_pageio(void)
36{
37 io_page_cachep = KMEM_CACHE(ext4_io_page, SLAB_RECLAIM_ACCOUNT);
38 if (io_page_cachep == NULL)
39 return -ENOMEM;
40 io_end_cachep = KMEM_CACHE(ext4_io_end, SLAB_RECLAIM_ACCOUNT);
41 if (io_page_cachep == NULL) {
42 kmem_cache_destroy(io_page_cachep);
43 return -ENOMEM;
44 }
45
46 return 0;
47}
48
49void ext4_exit_pageio(void)
50{
51 kmem_cache_destroy(io_end_cachep);
52 kmem_cache_destroy(io_page_cachep);
53}
54
55void ext4_free_io_end(ext4_io_end_t *io)
56{
57 int i;
58
59 BUG_ON(!io);
60 if (io->page)
61 put_page(io->page);
62 for (i = 0; i < io->num_io_pages; i++) {
63 if (--io->pages[i]->p_count == 0) {
64 struct page *page = io->pages[i]->p_page;
65
66 end_page_writeback(page);
67 put_page(page);
68 kmem_cache_free(io_page_cachep, io->pages[i]);
69 }
70 }
71 io->num_io_pages = 0;
72 iput(io->inode);
73 kmem_cache_free(io_end_cachep, io);
74}
75
76/*
77 * check a range of space and convert unwritten extents to written.
78 */
79int ext4_end_io_nolock(ext4_io_end_t *io)
80{
81 struct inode *inode = io->inode;
82 loff_t offset = io->offset;
83 ssize_t size = io->size;
84 int ret = 0;
85
86 ext4_debug("ext4_end_io_nolock: io 0x%p from inode %lu,list->next 0x%p,"
87 "list->prev 0x%p\n",
88 io, inode->i_ino, io->list.next, io->list.prev);
89
90 if (list_empty(&io->list))
91 return ret;
92
93 if (!(io->flag & EXT4_IO_END_UNWRITTEN))
94 return ret;
95
96 ret = ext4_convert_unwritten_extents(inode, offset, size);
97 if (ret < 0) {
98 printk(KERN_EMERG "%s: failed to convert unwritten "
99 "extents to written extents, error is %d "
100 "io is still on inode %lu aio dio list\n",
101 __func__, ret, inode->i_ino);
102 return ret;
103 }
104
105 if (io->iocb)
106 aio_complete(io->iocb, io->result, 0);
107 /* clear the DIO AIO unwritten flag */
108 io->flag &= ~EXT4_IO_END_UNWRITTEN;
109 return ret;
110}
111
112/*
113 * work on completed aio dio IO, to convert unwritten extents to extents
114 */
115static void ext4_end_io_work(struct work_struct *work)
116{
117 ext4_io_end_t *io = container_of(work, ext4_io_end_t, work);
118 struct inode *inode = io->inode;
119 struct ext4_inode_info *ei = EXT4_I(inode);
120 unsigned long flags;
121 int ret;
122
123 mutex_lock(&inode->i_mutex);
124 ret = ext4_end_io_nolock(io);
125 if (ret < 0) {
126 mutex_unlock(&inode->i_mutex);
127 return;
128 }
129
130 spin_lock_irqsave(&ei->i_completed_io_lock, flags);
131 if (!list_empty(&io->list))
132 list_del_init(&io->list);
133 spin_unlock_irqrestore(&ei->i_completed_io_lock, flags);
134 mutex_unlock(&inode->i_mutex);
135 ext4_free_io_end(io);
136}
137
138ext4_io_end_t *ext4_init_io_end(struct inode *inode, gfp_t flags)
139{
140 ext4_io_end_t *io = NULL;
141
142 io = kmem_cache_alloc(io_end_cachep, flags);
143 if (io) {
144 memset(io, 0, sizeof(*io));
145 io->inode = igrab(inode);
146 BUG_ON(!io->inode);
147 INIT_WORK(&io->work, ext4_end_io_work);
148 INIT_LIST_HEAD(&io->list);
149 }
150 return io;
151}
152
153/*
154 * Print an buffer I/O error compatible with the fs/buffer.c. This
155 * provides compatibility with dmesg scrapers that look for a specific
156 * buffer I/O error message. We really need a unified error reporting
157 * structure to userspace ala Digital Unix's uerf system, but it's
158 * probably not going to happen in my lifetime, due to LKML politics...
159 */
160static void buffer_io_error(struct buffer_head *bh)
161{
162 char b[BDEVNAME_SIZE];
163 printk(KERN_ERR "Buffer I/O error on device %s, logical block %llu\n",
164 bdevname(bh->b_bdev, b),
165 (unsigned long long)bh->b_blocknr);
166}
167
168static void ext4_end_bio(struct bio *bio, int error)
169{
170 ext4_io_end_t *io_end = bio->bi_private;
171 struct workqueue_struct *wq;
172 struct inode *inode;
173 unsigned long flags;
174 ext4_fsblk_t err_block;
175 int i;
176
177 BUG_ON(!io_end);
178 inode = io_end->inode;
179 bio->bi_private = NULL;
180 bio->bi_end_io = NULL;
181 if (test_bit(BIO_UPTODATE, &bio->bi_flags))
182 error = 0;
183 err_block = bio->bi_sector >> (inode->i_blkbits - 9);
184 bio_put(bio);
185
186 if (!(inode->i_sb->s_flags & MS_ACTIVE)) {
187 pr_err("sb umounted, discard end_io request for inode %lu\n",
188 io_end->inode->i_ino);
189 ext4_free_io_end(io_end);
190 return;
191 }
192
193 if (error) {
194 io_end->flag |= EXT4_IO_END_ERROR;
195 ext4_warning(inode->i_sb, "I/O error writing to inode %lu "
196 "(offset %llu size %ld starting block %llu)",
197 inode->i_ino,
198 (unsigned long long) io_end->offset,
199 (long) io_end->size,
200 (unsigned long long) err_block);
201 }
202
203 for (i = 0; i < io_end->num_io_pages; i++) {
204 struct page *page = io_end->pages[i]->p_page;
205 struct buffer_head *bh, *head;
206 int partial_write = 0;
207
208 head = page_buffers(page);
209 if (error)
210 SetPageError(page);
211 BUG_ON(!head);
212 if (head->b_size == PAGE_CACHE_SIZE)
213 clear_buffer_dirty(head);
214 else {
215 loff_t offset;
216 loff_t io_end_offset = io_end->offset + io_end->size;
217
218 offset = (sector_t) page->index << PAGE_CACHE_SHIFT;
219 bh = head;
220 do {
221 if ((offset >= io_end->offset) &&
222 (offset+bh->b_size <= io_end_offset)) {
223 if (error)
224 buffer_io_error(bh);
225
226 clear_buffer_dirty(bh);
227 }
228 if (buffer_delay(bh))
229 partial_write = 1;
230 else if (!buffer_mapped(bh))
231 clear_buffer_dirty(bh);
232 else if (buffer_dirty(bh))
233 partial_write = 1;
234 offset += bh->b_size;
235 bh = bh->b_this_page;
236 } while (bh != head);
237 }
238
239 if (--io_end->pages[i]->p_count == 0) {
240 struct page *page = io_end->pages[i]->p_page;
241
242 end_page_writeback(page);
243 put_page(page);
244 kmem_cache_free(io_page_cachep, io_end->pages[i]);
245 }
246
247 /*
248 * If this is a partial write which happened to make
249 * all buffers uptodate then we can optimize away a
250 * bogus readpage() for the next read(). Here we
251 * 'discover' whether the page went uptodate as a
252 * result of this (potentially partial) write.
253 */
254 if (!partial_write)
255 SetPageUptodate(page);
256 }
257
258 io_end->num_io_pages = 0;
259
260 /* Add the io_end to per-inode completed io list*/
261 spin_lock_irqsave(&EXT4_I(inode)->i_completed_io_lock, flags);
262 list_add_tail(&io_end->list, &EXT4_I(inode)->i_completed_io_list);
263 spin_unlock_irqrestore(&EXT4_I(inode)->i_completed_io_lock, flags);
264
265 wq = EXT4_SB(inode->i_sb)->dio_unwritten_wq;
266 /* queue the work to convert unwritten extents to written */
267 queue_work(wq, &io_end->work);
268}
269
270void ext4_io_submit(struct ext4_io_submit *io)
271{
272 struct bio *bio = io->io_bio;
273
274 if (bio) {
275 bio_get(io->io_bio);
276 submit_bio(io->io_op, io->io_bio);
277 BUG_ON(bio_flagged(io->io_bio, BIO_EOPNOTSUPP));
278 bio_put(io->io_bio);
279 }
280 io->io_bio = 0;
281 io->io_op = 0;
282 io->io_end = 0;
283}
284
285static int io_submit_init(struct ext4_io_submit *io,
286 struct inode *inode,
287 struct writeback_control *wbc,
288 struct buffer_head *bh)
289{
290 ext4_io_end_t *io_end;
291 struct page *page = bh->b_page;
292 int nvecs = bio_get_nr_vecs(bh->b_bdev);
293 struct bio *bio;
294
295 io_end = ext4_init_io_end(inode, GFP_NOFS);
296 if (!io_end)
297 return -ENOMEM;
298 do {
299 bio = bio_alloc(GFP_NOIO, nvecs);
300 nvecs >>= 1;
301 } while (bio == NULL);
302
303 bio->bi_sector = bh->b_blocknr * (bh->b_size >> 9);
304 bio->bi_bdev = bh->b_bdev;
305 bio->bi_private = io->io_end = io_end;
306 bio->bi_end_io = ext4_end_bio;
307
308 io_end->inode = inode;
309 io_end->offset = (page->index << PAGE_CACHE_SHIFT) + bh_offset(bh);
310
311 io->io_bio = bio;
312 io->io_op = (wbc->sync_mode == WB_SYNC_ALL ?
313 WRITE_SYNC_PLUG : WRITE);
314 io->io_next_block = bh->b_blocknr;
315 return 0;
316}
317
318static int io_submit_add_bh(struct ext4_io_submit *io,
319 struct ext4_io_page *io_page,
320 struct inode *inode,
321 struct writeback_control *wbc,
322 struct buffer_head *bh)
323{
324 ext4_io_end_t *io_end;
325 int ret;
326
327 if (buffer_new(bh)) {
328 clear_buffer_new(bh);
329 unmap_underlying_metadata(bh->b_bdev, bh->b_blocknr);
330 }
331
332 if (!buffer_mapped(bh) || buffer_delay(bh)) {
333 if (!buffer_mapped(bh))
334 clear_buffer_dirty(bh);
335 if (io->io_bio)
336 ext4_io_submit(io);
337 return 0;
338 }
339
340 if (io->io_bio && bh->b_blocknr != io->io_next_block) {
341submit_and_retry:
342 ext4_io_submit(io);
343 }
344 if (io->io_bio == NULL) {
345 ret = io_submit_init(io, inode, wbc, bh);
346 if (ret)
347 return ret;
348 }
349 io_end = io->io_end;
350 if ((io_end->num_io_pages >= MAX_IO_PAGES) &&
351 (io_end->pages[io_end->num_io_pages-1] != io_page))
352 goto submit_and_retry;
353 if (buffer_uninit(bh))
354 io->io_end->flag |= EXT4_IO_END_UNWRITTEN;
355 io->io_end->size += bh->b_size;
356 io->io_next_block++;
357 ret = bio_add_page(io->io_bio, bh->b_page, bh->b_size, bh_offset(bh));
358 if (ret != bh->b_size)
359 goto submit_and_retry;
360 if ((io_end->num_io_pages == 0) ||
361 (io_end->pages[io_end->num_io_pages-1] != io_page)) {
362 io_end->pages[io_end->num_io_pages++] = io_page;
363 io_page->p_count++;
364 }
365 return 0;
366}
367
368int ext4_bio_write_page(struct ext4_io_submit *io,
369 struct page *page,
370 int len,
371 struct writeback_control *wbc)
372{
373 struct inode *inode = page->mapping->host;
374 unsigned block_start, block_end, blocksize;
375 struct ext4_io_page *io_page;
376 struct buffer_head *bh, *head;
377 int ret = 0;
378
379 blocksize = 1 << inode->i_blkbits;
380
381 BUG_ON(PageWriteback(page));
382 set_page_writeback(page);
383 ClearPageError(page);
384
385 io_page = kmem_cache_alloc(io_page_cachep, GFP_NOFS);
386 if (!io_page) {
387 set_page_dirty(page);
388 unlock_page(page);
389 return -ENOMEM;
390 }
391 io_page->p_page = page;
392 io_page->p_count = 0;
393 get_page(page);
394
395 for (bh = head = page_buffers(page), block_start = 0;
396 bh != head || !block_start;
397 block_start = block_end, bh = bh->b_this_page) {
398 block_end = block_start + blocksize;
399 if (block_start >= len) {
400 clear_buffer_dirty(bh);
401 set_buffer_uptodate(bh);
402 continue;
403 }
404 ret = io_submit_add_bh(io, io_page, inode, wbc, bh);
405 if (ret) {
406 /*
407 * We only get here on ENOMEM. Not much else
408 * we can do but mark the page as dirty, and
409 * better luck next time.
410 */
411 set_page_dirty(page);
412 break;
413 }
414 }
415 unlock_page(page);
416 /*
417 * If the page was truncated before we could do the writeback,
418 * or we had a memory allocation error while trying to write
419 * the first buffer head, we won't have submitted any pages for
420 * I/O. In that case we need to make sure we've cleared the
421 * PageWriteback bit from the page to prevent the system from
422 * wedging later on.
423 */
424 if (io_page->p_count == 0) {
425 put_page(page);
426 end_page_writeback(page);
427 kmem_cache_free(io_page_cachep, io_page);
428 }
429 return ret;
430}
diff --git a/fs/ext4/resize.c b/fs/ext4/resize.c
index 6df797eb9aeb..dc963929de65 100644
--- a/fs/ext4/resize.c
+++ b/fs/ext4/resize.c
@@ -226,23 +226,13 @@ static int setup_new_group_blocks(struct super_block *sb,
226 } 226 }
227 227
228 /* Zero out all of the reserved backup group descriptor table blocks */ 228 /* Zero out all of the reserved backup group descriptor table blocks */
229 for (i = 0, bit = gdblocks + 1, block = start + bit; 229 ext4_debug("clear inode table blocks %#04llx -> %#04llx\n",
230 i < reserved_gdb; i++, block++, bit++) { 230 block, sbi->s_itb_per_group);
231 struct buffer_head *gdb; 231 err = sb_issue_zeroout(sb, gdblocks + start + 1, reserved_gdb,
232 232 GFP_NOFS);
233 ext4_debug("clear reserved block %#04llx (+%d)\n", block, bit); 233 if (err)
234 234 goto exit_bh;
235 if ((err = extend_or_restart_transaction(handle, 1, bh)))
236 goto exit_bh;
237 235
238 if (IS_ERR(gdb = bclean(handle, sb, block))) {
239 err = PTR_ERR(gdb);
240 goto exit_bh;
241 }
242 ext4_handle_dirty_metadata(handle, NULL, gdb);
243 ext4_set_bit(bit, bh->b_data);
244 brelse(gdb);
245 }
246 ext4_debug("mark block bitmap %#04llx (+%llu)\n", input->block_bitmap, 236 ext4_debug("mark block bitmap %#04llx (+%llu)\n", input->block_bitmap,
247 input->block_bitmap - start); 237 input->block_bitmap - start);
248 ext4_set_bit(input->block_bitmap - start, bh->b_data); 238 ext4_set_bit(input->block_bitmap - start, bh->b_data);
@@ -251,28 +241,18 @@ static int setup_new_group_blocks(struct super_block *sb,
251 ext4_set_bit(input->inode_bitmap - start, bh->b_data); 241 ext4_set_bit(input->inode_bitmap - start, bh->b_data);
252 242
253 /* Zero out all of the inode table blocks */ 243 /* Zero out all of the inode table blocks */
254 for (i = 0, block = input->inode_table, bit = block - start; 244 block = input->inode_table;
255 i < sbi->s_itb_per_group; i++, bit++, block++) { 245 ext4_debug("clear inode table blocks %#04llx -> %#04llx\n",
256 struct buffer_head *it; 246 block, sbi->s_itb_per_group);
257 247 err = sb_issue_zeroout(sb, block, sbi->s_itb_per_group, GFP_NOFS);
258 ext4_debug("clear inode block %#04llx (+%d)\n", block, bit); 248 if (err)
259 249 goto exit_bh;
260 if ((err = extend_or_restart_transaction(handle, 1, bh)))
261 goto exit_bh;
262
263 if (IS_ERR(it = bclean(handle, sb, block))) {
264 err = PTR_ERR(it);
265 goto exit_bh;
266 }
267 ext4_handle_dirty_metadata(handle, NULL, it);
268 brelse(it);
269 ext4_set_bit(bit, bh->b_data);
270 }
271 250
272 if ((err = extend_or_restart_transaction(handle, 2, bh))) 251 if ((err = extend_or_restart_transaction(handle, 2, bh)))
273 goto exit_bh; 252 goto exit_bh;
274 253
275 mark_bitmap_end(input->blocks_count, sb->s_blocksize * 8, bh->b_data); 254 ext4_mark_bitmap_end(input->blocks_count, sb->s_blocksize * 8,
255 bh->b_data);
276 ext4_handle_dirty_metadata(handle, NULL, bh); 256 ext4_handle_dirty_metadata(handle, NULL, bh);
277 brelse(bh); 257 brelse(bh);
278 /* Mark unused entries in inode bitmap used */ 258 /* Mark unused entries in inode bitmap used */
@@ -283,8 +263,8 @@ static int setup_new_group_blocks(struct super_block *sb,
283 goto exit_journal; 263 goto exit_journal;
284 } 264 }
285 265
286 mark_bitmap_end(EXT4_INODES_PER_GROUP(sb), sb->s_blocksize * 8, 266 ext4_mark_bitmap_end(EXT4_INODES_PER_GROUP(sb), sb->s_blocksize * 8,
287 bh->b_data); 267 bh->b_data);
288 ext4_handle_dirty_metadata(handle, NULL, bh); 268 ext4_handle_dirty_metadata(handle, NULL, bh);
289exit_bh: 269exit_bh:
290 brelse(bh); 270 brelse(bh);
@@ -921,8 +901,7 @@ int ext4_group_add(struct super_block *sb, struct ext4_new_group_data *input)
921 &sbi->s_flex_groups[flex_group].free_inodes); 901 &sbi->s_flex_groups[flex_group].free_inodes);
922 } 902 }
923 903
924 ext4_handle_dirty_metadata(handle, NULL, sbi->s_sbh); 904 ext4_handle_dirty_super(handle, sb);
925 sb->s_dirt = 1;
926 905
927exit_journal: 906exit_journal:
928 mutex_unlock(&sbi->s_resize_lock); 907 mutex_unlock(&sbi->s_resize_lock);
@@ -953,7 +932,6 @@ int ext4_group_extend(struct super_block *sb, struct ext4_super_block *es,
953 ext4_fsblk_t n_blocks_count) 932 ext4_fsblk_t n_blocks_count)
954{ 933{
955 ext4_fsblk_t o_blocks_count; 934 ext4_fsblk_t o_blocks_count;
956 ext4_group_t o_groups_count;
957 ext4_grpblk_t last; 935 ext4_grpblk_t last;
958 ext4_grpblk_t add; 936 ext4_grpblk_t add;
959 struct buffer_head *bh; 937 struct buffer_head *bh;
@@ -965,7 +943,6 @@ int ext4_group_extend(struct super_block *sb, struct ext4_super_block *es,
965 * yet: we're going to revalidate es->s_blocks_count after 943 * yet: we're going to revalidate es->s_blocks_count after
966 * taking the s_resize_lock below. */ 944 * taking the s_resize_lock below. */
967 o_blocks_count = ext4_blocks_count(es); 945 o_blocks_count = ext4_blocks_count(es);
968 o_groups_count = EXT4_SB(sb)->s_groups_count;
969 946
970 if (test_opt(sb, DEBUG)) 947 if (test_opt(sb, DEBUG))
971 printk(KERN_DEBUG "EXT4-fs: extending last group from %llu uto %llu blocks\n", 948 printk(KERN_DEBUG "EXT4-fs: extending last group from %llu uto %llu blocks\n",
@@ -1045,13 +1022,12 @@ int ext4_group_extend(struct super_block *sb, struct ext4_super_block *es,
1045 goto exit_put; 1022 goto exit_put;
1046 } 1023 }
1047 ext4_blocks_count_set(es, o_blocks_count + add); 1024 ext4_blocks_count_set(es, o_blocks_count + add);
1048 ext4_handle_dirty_metadata(handle, NULL, EXT4_SB(sb)->s_sbh);
1049 sb->s_dirt = 1;
1050 mutex_unlock(&EXT4_SB(sb)->s_resize_lock); 1025 mutex_unlock(&EXT4_SB(sb)->s_resize_lock);
1051 ext4_debug("freeing blocks %llu through %llu\n", o_blocks_count, 1026 ext4_debug("freeing blocks %llu through %llu\n", o_blocks_count,
1052 o_blocks_count + add); 1027 o_blocks_count + add);
1053 /* We add the blocks to the bitmap and set the group need init bit */ 1028 /* We add the blocks to the bitmap and set the group need init bit */
1054 ext4_add_groupblocks(handle, sb, o_blocks_count, add); 1029 ext4_add_groupblocks(handle, sb, o_blocks_count, add);
1030 ext4_handle_dirty_super(handle, sb);
1055 ext4_debug("freed blocks %llu through %llu\n", o_blocks_count, 1031 ext4_debug("freed blocks %llu through %llu\n", o_blocks_count,
1056 o_blocks_count + add); 1032 o_blocks_count + add);
1057 if ((err = ext4_journal_stop(handle))) 1033 if ((err = ext4_journal_stop(handle)))
diff --git a/fs/ext4/super.c b/fs/ext4/super.c
index 4e8983a9811b..0348ce066592 100644
--- a/fs/ext4/super.c
+++ b/fs/ext4/super.c
@@ -26,7 +26,6 @@
26#include <linux/init.h> 26#include <linux/init.h>
27#include <linux/blkdev.h> 27#include <linux/blkdev.h>
28#include <linux/parser.h> 28#include <linux/parser.h>
29#include <linux/smp_lock.h>
30#include <linux/buffer_head.h> 29#include <linux/buffer_head.h>
31#include <linux/exportfs.h> 30#include <linux/exportfs.h>
32#include <linux/vfs.h> 31#include <linux/vfs.h>
@@ -41,6 +40,9 @@
41#include <linux/crc16.h> 40#include <linux/crc16.h>
42#include <asm/uaccess.h> 41#include <asm/uaccess.h>
43 42
43#include <linux/kthread.h>
44#include <linux/freezer.h>
45
44#include "ext4.h" 46#include "ext4.h"
45#include "ext4_jbd2.h" 47#include "ext4_jbd2.h"
46#include "xattr.h" 48#include "xattr.h"
@@ -50,8 +52,11 @@
50#define CREATE_TRACE_POINTS 52#define CREATE_TRACE_POINTS
51#include <trace/events/ext4.h> 53#include <trace/events/ext4.h>
52 54
53struct proc_dir_entry *ext4_proc_root; 55static struct proc_dir_entry *ext4_proc_root;
54static struct kset *ext4_kset; 56static struct kset *ext4_kset;
57struct ext4_lazy_init *ext4_li_info;
58struct mutex ext4_li_mtx;
59struct ext4_features *ext4_feat;
55 60
56static int ext4_load_journal(struct super_block *, struct ext4_super_block *, 61static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
57 unsigned long journal_devnum); 62 unsigned long journal_devnum);
@@ -70,6 +75,8 @@ static void ext4_write_super(struct super_block *sb);
70static int ext4_freeze(struct super_block *sb); 75static int ext4_freeze(struct super_block *sb);
71static int ext4_get_sb(struct file_system_type *fs_type, int flags, 76static int ext4_get_sb(struct file_system_type *fs_type, int flags,
72 const char *dev_name, void *data, struct vfsmount *mnt); 77 const char *dev_name, void *data, struct vfsmount *mnt);
78static void ext4_destroy_lazyinit_thread(void);
79static void ext4_unregister_li_request(struct super_block *sb);
73 80
74#if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23) 81#if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
75static struct file_system_type ext3_fs_type = { 82static struct file_system_type ext3_fs_type = {
@@ -241,14 +248,14 @@ handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
241 if (sb->s_flags & MS_RDONLY) 248 if (sb->s_flags & MS_RDONLY)
242 return ERR_PTR(-EROFS); 249 return ERR_PTR(-EROFS);
243 250
244 vfs_check_frozen(sb, SB_FREEZE_WRITE); 251 vfs_check_frozen(sb, SB_FREEZE_TRANS);
245 /* Special case here: if the journal has aborted behind our 252 /* Special case here: if the journal has aborted behind our
246 * backs (eg. EIO in the commit thread), then we still need to 253 * backs (eg. EIO in the commit thread), then we still need to
247 * take the FS itself readonly cleanly. */ 254 * take the FS itself readonly cleanly. */
248 journal = EXT4_SB(sb)->s_journal; 255 journal = EXT4_SB(sb)->s_journal;
249 if (journal) { 256 if (journal) {
250 if (is_journal_aborted(journal)) { 257 if (is_journal_aborted(journal)) {
251 ext4_abort(sb, __func__, "Detected aborted journal"); 258 ext4_abort(sb, "Detected aborted journal");
252 return ERR_PTR(-EROFS); 259 return ERR_PTR(-EROFS);
253 } 260 }
254 return jbd2_journal_start(journal, nblocks); 261 return jbd2_journal_start(journal, nblocks);
@@ -262,7 +269,7 @@ handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
262 * that sync() will call the filesystem's write_super callback if 269 * that sync() will call the filesystem's write_super callback if
263 * appropriate. 270 * appropriate.
264 */ 271 */
265int __ext4_journal_stop(const char *where, handle_t *handle) 272int __ext4_journal_stop(const char *where, unsigned int line, handle_t *handle)
266{ 273{
267 struct super_block *sb; 274 struct super_block *sb;
268 int err; 275 int err;
@@ -279,12 +286,13 @@ int __ext4_journal_stop(const char *where, handle_t *handle)
279 if (!err) 286 if (!err)
280 err = rc; 287 err = rc;
281 if (err) 288 if (err)
282 __ext4_std_error(sb, where, err); 289 __ext4_std_error(sb, where, line, err);
283 return err; 290 return err;
284} 291}
285 292
286void ext4_journal_abort_handle(const char *caller, const char *err_fn, 293void ext4_journal_abort_handle(const char *caller, unsigned int line,
287 struct buffer_head *bh, handle_t *handle, int err) 294 const char *err_fn, struct buffer_head *bh,
295 handle_t *handle, int err)
288{ 296{
289 char nbuf[16]; 297 char nbuf[16];
290 const char *errstr = ext4_decode_error(NULL, err, nbuf); 298 const char *errstr = ext4_decode_error(NULL, err, nbuf);
@@ -300,12 +308,47 @@ void ext4_journal_abort_handle(const char *caller, const char *err_fn,
300 if (is_handle_aborted(handle)) 308 if (is_handle_aborted(handle))
301 return; 309 return;
302 310
303 printk(KERN_ERR "%s: aborting transaction: %s in %s\n", 311 printk(KERN_ERR "%s:%d: aborting transaction: %s in %s\n",
304 caller, errstr, err_fn); 312 caller, line, errstr, err_fn);
305 313
306 jbd2_journal_abort_handle(handle); 314 jbd2_journal_abort_handle(handle);
307} 315}
308 316
317static void __save_error_info(struct super_block *sb, const char *func,
318 unsigned int line)
319{
320 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
321
322 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
323 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
324 es->s_last_error_time = cpu_to_le32(get_seconds());
325 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
326 es->s_last_error_line = cpu_to_le32(line);
327 if (!es->s_first_error_time) {
328 es->s_first_error_time = es->s_last_error_time;
329 strncpy(es->s_first_error_func, func,
330 sizeof(es->s_first_error_func));
331 es->s_first_error_line = cpu_to_le32(line);
332 es->s_first_error_ino = es->s_last_error_ino;
333 es->s_first_error_block = es->s_last_error_block;
334 }
335 /*
336 * Start the daily error reporting function if it hasn't been
337 * started already
338 */
339 if (!es->s_error_count)
340 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
341 es->s_error_count = cpu_to_le32(le32_to_cpu(es->s_error_count) + 1);
342}
343
344static void save_error_info(struct super_block *sb, const char *func,
345 unsigned int line)
346{
347 __save_error_info(sb, func, line);
348 ext4_commit_super(sb, 1);
349}
350
351
309/* Deal with the reporting of failure conditions on a filesystem such as 352/* Deal with the reporting of failure conditions on a filesystem such as
310 * inconsistencies detected or read IO failures. 353 * inconsistencies detected or read IO failures.
311 * 354 *
@@ -323,11 +366,6 @@ void ext4_journal_abort_handle(const char *caller, const char *err_fn,
323 366
324static void ext4_handle_error(struct super_block *sb) 367static void ext4_handle_error(struct super_block *sb)
325{ 368{
326 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
327
328 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
329 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
330
331 if (sb->s_flags & MS_RDONLY) 369 if (sb->s_flags & MS_RDONLY)
332 return; 370 return;
333 371
@@ -342,19 +380,19 @@ static void ext4_handle_error(struct super_block *sb)
342 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only"); 380 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
343 sb->s_flags |= MS_RDONLY; 381 sb->s_flags |= MS_RDONLY;
344 } 382 }
345 ext4_commit_super(sb, 1);
346 if (test_opt(sb, ERRORS_PANIC)) 383 if (test_opt(sb, ERRORS_PANIC))
347 panic("EXT4-fs (device %s): panic forced after error\n", 384 panic("EXT4-fs (device %s): panic forced after error\n",
348 sb->s_id); 385 sb->s_id);
349} 386}
350 387
351void __ext4_error(struct super_block *sb, const char *function, 388void __ext4_error(struct super_block *sb, const char *function,
352 const char *fmt, ...) 389 unsigned int line, const char *fmt, ...)
353{ 390{
354 va_list args; 391 va_list args;
355 392
356 va_start(args, fmt); 393 va_start(args, fmt);
357 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function); 394 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: comm %s: ",
395 sb->s_id, function, line, current->comm);
358 vprintk(fmt, args); 396 vprintk(fmt, args);
359 printk("\n"); 397 printk("\n");
360 va_end(args); 398 va_end(args);
@@ -362,14 +400,22 @@ void __ext4_error(struct super_block *sb, const char *function,
362 ext4_handle_error(sb); 400 ext4_handle_error(sb);
363} 401}
364 402
365void ext4_error_inode(const char *function, struct inode *inode, 403void ext4_error_inode(struct inode *inode, const char *function,
404 unsigned int line, ext4_fsblk_t block,
366 const char *fmt, ...) 405 const char *fmt, ...)
367{ 406{
368 va_list args; 407 va_list args;
408 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
369 409
410 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
411 es->s_last_error_block = cpu_to_le64(block);
412 save_error_info(inode->i_sb, function, line);
370 va_start(args, fmt); 413 va_start(args, fmt);
371 printk(KERN_CRIT "EXT4-fs error (device %s): %s: inode #%lu: (comm %s) ", 414 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: inode #%lu: ",
372 inode->i_sb->s_id, function, inode->i_ino, current->comm); 415 inode->i_sb->s_id, function, line, inode->i_ino);
416 if (block)
417 printk("block %llu: ", block);
418 printk("comm %s: ", current->comm);
373 vprintk(fmt, args); 419 vprintk(fmt, args);
374 printk("\n"); 420 printk("\n");
375 va_end(args); 421 va_end(args);
@@ -377,20 +423,26 @@ void ext4_error_inode(const char *function, struct inode *inode,
377 ext4_handle_error(inode->i_sb); 423 ext4_handle_error(inode->i_sb);
378} 424}
379 425
380void ext4_error_file(const char *function, struct file *file, 426void ext4_error_file(struct file *file, const char *function,
381 const char *fmt, ...) 427 unsigned int line, const char *fmt, ...)
382{ 428{
383 va_list args; 429 va_list args;
430 struct ext4_super_block *es;
384 struct inode *inode = file->f_dentry->d_inode; 431 struct inode *inode = file->f_dentry->d_inode;
385 char pathname[80], *path; 432 char pathname[80], *path;
386 433
434 es = EXT4_SB(inode->i_sb)->s_es;
435 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
436 save_error_info(inode->i_sb, function, line);
387 va_start(args, fmt); 437 va_start(args, fmt);
388 path = d_path(&(file->f_path), pathname, sizeof(pathname)); 438 path = d_path(&(file->f_path), pathname, sizeof(pathname));
389 if (!path) 439 if (!path)
390 path = "(unknown)"; 440 path = "(unknown)";
391 printk(KERN_CRIT 441 printk(KERN_CRIT
392 "EXT4-fs error (device %s): %s: inode #%lu (comm %s path %s): ", 442 "EXT4-fs error (device %s): %s:%d: inode #%lu "
393 inode->i_sb->s_id, function, inode->i_ino, current->comm, path); 443 "(comm %s path %s): ",
444 inode->i_sb->s_id, function, line, inode->i_ino,
445 current->comm, path);
394 vprintk(fmt, args); 446 vprintk(fmt, args);
395 printk("\n"); 447 printk("\n");
396 va_end(args); 448 va_end(args);
@@ -435,7 +487,8 @@ static const char *ext4_decode_error(struct super_block *sb, int errno,
435/* __ext4_std_error decodes expected errors from journaling functions 487/* __ext4_std_error decodes expected errors from journaling functions
436 * automatically and invokes the appropriate error response. */ 488 * automatically and invokes the appropriate error response. */
437 489
438void __ext4_std_error(struct super_block *sb, const char *function, int errno) 490void __ext4_std_error(struct super_block *sb, const char *function,
491 unsigned int line, int errno)
439{ 492{
440 char nbuf[16]; 493 char nbuf[16];
441 const char *errstr; 494 const char *errstr;
@@ -448,8 +501,9 @@ void __ext4_std_error(struct super_block *sb, const char *function, int errno)
448 return; 501 return;
449 502
450 errstr = ext4_decode_error(sb, errno, nbuf); 503 errstr = ext4_decode_error(sb, errno, nbuf);
451 printk(KERN_CRIT "EXT4-fs error (device %s) in %s: %s\n", 504 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
452 sb->s_id, function, errstr); 505 sb->s_id, function, line, errstr);
506 save_error_info(sb, function, line);
453 507
454 ext4_handle_error(sb); 508 ext4_handle_error(sb);
455} 509}
@@ -464,29 +518,29 @@ void __ext4_std_error(struct super_block *sb, const char *function, int errno)
464 * case we take the easy way out and panic immediately. 518 * case we take the easy way out and panic immediately.
465 */ 519 */
466 520
467void ext4_abort(struct super_block *sb, const char *function, 521void __ext4_abort(struct super_block *sb, const char *function,
468 const char *fmt, ...) 522 unsigned int line, const char *fmt, ...)
469{ 523{
470 va_list args; 524 va_list args;
471 525
526 save_error_info(sb, function, line);
472 va_start(args, fmt); 527 va_start(args, fmt);
473 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function); 528 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
529 function, line);
474 vprintk(fmt, args); 530 vprintk(fmt, args);
475 printk("\n"); 531 printk("\n");
476 va_end(args); 532 va_end(args);
477 533
534 if ((sb->s_flags & MS_RDONLY) == 0) {
535 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
536 sb->s_flags |= MS_RDONLY;
537 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
538 if (EXT4_SB(sb)->s_journal)
539 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
540 save_error_info(sb, function, line);
541 }
478 if (test_opt(sb, ERRORS_PANIC)) 542 if (test_opt(sb, ERRORS_PANIC))
479 panic("EXT4-fs panic from previous error\n"); 543 panic("EXT4-fs panic from previous error\n");
480
481 if (sb->s_flags & MS_RDONLY)
482 return;
483
484 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
485 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
486 sb->s_flags |= MS_RDONLY;
487 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
488 if (EXT4_SB(sb)->s_journal)
489 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
490} 544}
491 545
492void ext4_msg (struct super_block * sb, const char *prefix, 546void ext4_msg (struct super_block * sb, const char *prefix,
@@ -502,38 +556,47 @@ void ext4_msg (struct super_block * sb, const char *prefix,
502} 556}
503 557
504void __ext4_warning(struct super_block *sb, const char *function, 558void __ext4_warning(struct super_block *sb, const char *function,
505 const char *fmt, ...) 559 unsigned int line, const char *fmt, ...)
506{ 560{
507 va_list args; 561 va_list args;
508 562
509 va_start(args, fmt); 563 va_start(args, fmt);
510 printk(KERN_WARNING "EXT4-fs warning (device %s): %s: ", 564 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: ",
511 sb->s_id, function); 565 sb->s_id, function, line);
512 vprintk(fmt, args); 566 vprintk(fmt, args);
513 printk("\n"); 567 printk("\n");
514 va_end(args); 568 va_end(args);
515} 569}
516 570
517void ext4_grp_locked_error(struct super_block *sb, ext4_group_t grp, 571void __ext4_grp_locked_error(const char *function, unsigned int line,
518 const char *function, const char *fmt, ...) 572 struct super_block *sb, ext4_group_t grp,
573 unsigned long ino, ext4_fsblk_t block,
574 const char *fmt, ...)
519__releases(bitlock) 575__releases(bitlock)
520__acquires(bitlock) 576__acquires(bitlock)
521{ 577{
522 va_list args; 578 va_list args;
523 struct ext4_super_block *es = EXT4_SB(sb)->s_es; 579 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
524 580
581 es->s_last_error_ino = cpu_to_le32(ino);
582 es->s_last_error_block = cpu_to_le64(block);
583 __save_error_info(sb, function, line);
525 va_start(args, fmt); 584 va_start(args, fmt);
526 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function); 585 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u",
586 sb->s_id, function, line, grp);
587 if (ino)
588 printk("inode %lu: ", ino);
589 if (block)
590 printk("block %llu:", (unsigned long long) block);
527 vprintk(fmt, args); 591 vprintk(fmt, args);
528 printk("\n"); 592 printk("\n");
529 va_end(args); 593 va_end(args);
530 594
531 if (test_opt(sb, ERRORS_CONT)) { 595 if (test_opt(sb, ERRORS_CONT)) {
532 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
533 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
534 ext4_commit_super(sb, 0); 596 ext4_commit_super(sb, 0);
535 return; 597 return;
536 } 598 }
599
537 ext4_unlock_group(sb, grp); 600 ext4_unlock_group(sb, grp);
538 ext4_handle_error(sb); 601 ext4_handle_error(sb);
539 /* 602 /*
@@ -646,13 +709,13 @@ static void ext4_put_super(struct super_block *sb)
646 struct ext4_super_block *es = sbi->s_es; 709 struct ext4_super_block *es = sbi->s_es;
647 int i, err; 710 int i, err;
648 711
712 ext4_unregister_li_request(sb);
649 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED); 713 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
650 714
651 flush_workqueue(sbi->dio_unwritten_wq); 715 flush_workqueue(sbi->dio_unwritten_wq);
652 destroy_workqueue(sbi->dio_unwritten_wq); 716 destroy_workqueue(sbi->dio_unwritten_wq);
653 717
654 lock_super(sb); 718 lock_super(sb);
655 lock_kernel();
656 if (sb->s_dirt) 719 if (sb->s_dirt)
657 ext4_commit_super(sb, 1); 720 ext4_commit_super(sb, 1);
658 721
@@ -660,10 +723,10 @@ static void ext4_put_super(struct super_block *sb)
660 err = jbd2_journal_destroy(sbi->s_journal); 723 err = jbd2_journal_destroy(sbi->s_journal);
661 sbi->s_journal = NULL; 724 sbi->s_journal = NULL;
662 if (err < 0) 725 if (err < 0)
663 ext4_abort(sb, __func__, 726 ext4_abort(sb, "Couldn't clean up the journal");
664 "Couldn't clean up the journal");
665 } 727 }
666 728
729 del_timer(&sbi->s_err_report);
667 ext4_release_system_zone(sb); 730 ext4_release_system_zone(sb);
668 ext4_mb_release(sb); 731 ext4_mb_release(sb);
669 ext4_ext_release(sb); 732 ext4_ext_release(sb);
@@ -720,7 +783,6 @@ static void ext4_put_super(struct super_block *sb)
720 * Now that we are completely done shutting down the 783 * Now that we are completely done shutting down the
721 * superblock, we need to actually destroy the kobject. 784 * superblock, we need to actually destroy the kobject.
722 */ 785 */
723 unlock_kernel();
724 unlock_super(sb); 786 unlock_super(sb);
725 kobject_put(&sbi->s_kobj); 787 kobject_put(&sbi->s_kobj);
726 wait_for_completion(&sbi->s_kobj_unregister); 788 wait_for_completion(&sbi->s_kobj_unregister);
@@ -813,8 +875,10 @@ static void destroy_inodecache(void)
813 kmem_cache_destroy(ext4_inode_cachep); 875 kmem_cache_destroy(ext4_inode_cachep);
814} 876}
815 877
816static void ext4_clear_inode(struct inode *inode) 878void ext4_clear_inode(struct inode *inode)
817{ 879{
880 invalidate_inode_buffers(inode);
881 end_writeback(inode);
818 dquot_drop(inode); 882 dquot_drop(inode);
819 ext4_discard_preallocations(inode); 883 ext4_discard_preallocations(inode);
820 if (EXT4_JOURNAL(inode)) 884 if (EXT4_JOURNAL(inode))
@@ -946,14 +1010,12 @@ static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
946 seq_puts(seq, ",journal_async_commit"); 1010 seq_puts(seq, ",journal_async_commit");
947 else if (test_opt(sb, JOURNAL_CHECKSUM)) 1011 else if (test_opt(sb, JOURNAL_CHECKSUM))
948 seq_puts(seq, ",journal_checksum"); 1012 seq_puts(seq, ",journal_checksum");
949 if (test_opt(sb, NOBH))
950 seq_puts(seq, ",nobh");
951 if (test_opt(sb, I_VERSION)) 1013 if (test_opt(sb, I_VERSION))
952 seq_puts(seq, ",i_version"); 1014 seq_puts(seq, ",i_version");
953 if (!test_opt(sb, DELALLOC)) 1015 if (!test_opt(sb, DELALLOC) &&
1016 !(def_mount_opts & EXT4_DEFM_NODELALLOC))
954 seq_puts(seq, ",nodelalloc"); 1017 seq_puts(seq, ",nodelalloc");
955 1018
956
957 if (sbi->s_stripe) 1019 if (sbi->s_stripe)
958 seq_printf(seq, ",stripe=%lu", sbi->s_stripe); 1020 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
959 /* 1021 /*
@@ -977,7 +1039,7 @@ static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
977 if (test_opt(sb, NO_AUTO_DA_ALLOC)) 1039 if (test_opt(sb, NO_AUTO_DA_ALLOC))
978 seq_puts(seq, ",noauto_da_alloc"); 1040 seq_puts(seq, ",noauto_da_alloc");
979 1041
980 if (test_opt(sb, DISCARD)) 1042 if (test_opt(sb, DISCARD) && !(def_mount_opts & EXT4_DEFM_DISCARD))
981 seq_puts(seq, ",discard"); 1043 seq_puts(seq, ",discard");
982 1044
983 if (test_opt(sb, NOLOAD)) 1045 if (test_opt(sb, NOLOAD))
@@ -986,6 +1048,16 @@ static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
986 if (test_opt(sb, DIOREAD_NOLOCK)) 1048 if (test_opt(sb, DIOREAD_NOLOCK))
987 seq_puts(seq, ",dioread_nolock"); 1049 seq_puts(seq, ",dioread_nolock");
988 1050
1051 if (test_opt(sb, BLOCK_VALIDITY) &&
1052 !(def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY))
1053 seq_puts(seq, ",block_validity");
1054
1055 if (!test_opt(sb, INIT_INODE_TABLE))
1056 seq_puts(seq, ",noinit_inode_table");
1057 else if (sbi->s_li_wait_mult)
1058 seq_printf(seq, ",init_inode_table=%u",
1059 (unsigned) sbi->s_li_wait_mult);
1060
989 ext4_show_quota_options(seq, sb); 1061 ext4_show_quota_options(seq, sb);
990 1062
991 return 0; 1063 return 0;
@@ -1065,6 +1137,7 @@ static int ext4_mark_dquot_dirty(struct dquot *dquot);
1065static int ext4_write_info(struct super_block *sb, int type); 1137static int ext4_write_info(struct super_block *sb, int type);
1066static int ext4_quota_on(struct super_block *sb, int type, int format_id, 1138static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1067 char *path); 1139 char *path);
1140static int ext4_quota_off(struct super_block *sb, int type);
1068static int ext4_quota_on_mount(struct super_block *sb, int type); 1141static int ext4_quota_on_mount(struct super_block *sb, int type);
1069static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data, 1142static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1070 size_t len, loff_t off); 1143 size_t len, loff_t off);
@@ -1086,7 +1159,7 @@ static const struct dquot_operations ext4_quota_operations = {
1086 1159
1087static const struct quotactl_ops ext4_qctl_operations = { 1160static const struct quotactl_ops ext4_qctl_operations = {
1088 .quota_on = ext4_quota_on, 1161 .quota_on = ext4_quota_on,
1089 .quota_off = dquot_quota_off, 1162 .quota_off = ext4_quota_off,
1090 .quota_sync = dquot_quota_sync, 1163 .quota_sync = dquot_quota_sync,
1091 .get_info = dquot_get_dqinfo, 1164 .get_info = dquot_get_dqinfo,
1092 .set_info = dquot_set_dqinfo, 1165 .set_info = dquot_set_dqinfo,
@@ -1100,20 +1173,20 @@ static const struct super_operations ext4_sops = {
1100 .destroy_inode = ext4_destroy_inode, 1173 .destroy_inode = ext4_destroy_inode,
1101 .write_inode = ext4_write_inode, 1174 .write_inode = ext4_write_inode,
1102 .dirty_inode = ext4_dirty_inode, 1175 .dirty_inode = ext4_dirty_inode,
1103 .delete_inode = ext4_delete_inode, 1176 .evict_inode = ext4_evict_inode,
1104 .put_super = ext4_put_super, 1177 .put_super = ext4_put_super,
1105 .sync_fs = ext4_sync_fs, 1178 .sync_fs = ext4_sync_fs,
1106 .freeze_fs = ext4_freeze, 1179 .freeze_fs = ext4_freeze,
1107 .unfreeze_fs = ext4_unfreeze, 1180 .unfreeze_fs = ext4_unfreeze,
1108 .statfs = ext4_statfs, 1181 .statfs = ext4_statfs,
1109 .remount_fs = ext4_remount, 1182 .remount_fs = ext4_remount,
1110 .clear_inode = ext4_clear_inode,
1111 .show_options = ext4_show_options, 1183 .show_options = ext4_show_options,
1112#ifdef CONFIG_QUOTA 1184#ifdef CONFIG_QUOTA
1113 .quota_read = ext4_quota_read, 1185 .quota_read = ext4_quota_read,
1114 .quota_write = ext4_quota_write, 1186 .quota_write = ext4_quota_write,
1115#endif 1187#endif
1116 .bdev_try_to_free_page = bdev_try_to_free_page, 1188 .bdev_try_to_free_page = bdev_try_to_free_page,
1189 .trim_fs = ext4_trim_fs
1117}; 1190};
1118 1191
1119static const struct super_operations ext4_nojournal_sops = { 1192static const struct super_operations ext4_nojournal_sops = {
@@ -1121,12 +1194,11 @@ static const struct super_operations ext4_nojournal_sops = {
1121 .destroy_inode = ext4_destroy_inode, 1194 .destroy_inode = ext4_destroy_inode,
1122 .write_inode = ext4_write_inode, 1195 .write_inode = ext4_write_inode,
1123 .dirty_inode = ext4_dirty_inode, 1196 .dirty_inode = ext4_dirty_inode,
1124 .delete_inode = ext4_delete_inode, 1197 .evict_inode = ext4_evict_inode,
1125 .write_super = ext4_write_super, 1198 .write_super = ext4_write_super,
1126 .put_super = ext4_put_super, 1199 .put_super = ext4_put_super,
1127 .statfs = ext4_statfs, 1200 .statfs = ext4_statfs,
1128 .remount_fs = ext4_remount, 1201 .remount_fs = ext4_remount,
1129 .clear_inode = ext4_clear_inode,
1130 .show_options = ext4_show_options, 1202 .show_options = ext4_show_options,
1131#ifdef CONFIG_QUOTA 1203#ifdef CONFIG_QUOTA
1132 .quota_read = ext4_quota_read, 1204 .quota_read = ext4_quota_read,
@@ -1161,6 +1233,7 @@ enum {
1161 Opt_inode_readahead_blks, Opt_journal_ioprio, 1233 Opt_inode_readahead_blks, Opt_journal_ioprio,
1162 Opt_dioread_nolock, Opt_dioread_lock, 1234 Opt_dioread_nolock, Opt_dioread_lock,
1163 Opt_discard, Opt_nodiscard, 1235 Opt_discard, Opt_nodiscard,
1236 Opt_init_inode_table, Opt_noinit_inode_table,
1164}; 1237};
1165 1238
1166static const match_table_t tokens = { 1239static const match_table_t tokens = {
@@ -1231,6 +1304,9 @@ static const match_table_t tokens = {
1231 {Opt_dioread_lock, "dioread_lock"}, 1304 {Opt_dioread_lock, "dioread_lock"},
1232 {Opt_discard, "discard"}, 1305 {Opt_discard, "discard"},
1233 {Opt_nodiscard, "nodiscard"}, 1306 {Opt_nodiscard, "nodiscard"},
1307 {Opt_init_inode_table, "init_itable=%u"},
1308 {Opt_init_inode_table, "init_itable"},
1309 {Opt_noinit_inode_table, "noinit_itable"},
1234 {Opt_err, NULL}, 1310 {Opt_err, NULL},
1235}; 1311};
1236 1312
@@ -1624,10 +1700,12 @@ set_qf_format:
1624 *n_blocks_count = option; 1700 *n_blocks_count = option;
1625 break; 1701 break;
1626 case Opt_nobh: 1702 case Opt_nobh:
1627 set_opt(sbi->s_mount_opt, NOBH); 1703 ext4_msg(sb, KERN_WARNING,
1704 "Ignoring deprecated nobh option");
1628 break; 1705 break;
1629 case Opt_bh: 1706 case Opt_bh:
1630 clear_opt(sbi->s_mount_opt, NOBH); 1707 ext4_msg(sb, KERN_WARNING,
1708 "Ignoring deprecated bh option");
1631 break; 1709 break;
1632 case Opt_i_version: 1710 case Opt_i_version:
1633 set_opt(sbi->s_mount_opt, I_VERSION); 1711 set_opt(sbi->s_mount_opt, I_VERSION);
@@ -1699,6 +1777,20 @@ set_qf_format:
1699 case Opt_dioread_lock: 1777 case Opt_dioread_lock:
1700 clear_opt(sbi->s_mount_opt, DIOREAD_NOLOCK); 1778 clear_opt(sbi->s_mount_opt, DIOREAD_NOLOCK);
1701 break; 1779 break;
1780 case Opt_init_inode_table:
1781 set_opt(sbi->s_mount_opt, INIT_INODE_TABLE);
1782 if (args[0].from) {
1783 if (match_int(&args[0], &option))
1784 return 0;
1785 } else
1786 option = EXT4_DEF_LI_WAIT_MULT;
1787 if (option < 0)
1788 return 0;
1789 sbi->s_li_wait_mult = option;
1790 break;
1791 case Opt_noinit_inode_table:
1792 clear_opt(sbi->s_mount_opt, INIT_INODE_TABLE);
1793 break;
1702 default: 1794 default:
1703 ext4_msg(sb, KERN_ERR, 1795 ext4_msg(sb, KERN_ERR,
1704 "Unrecognized mount option \"%s\" " 1796 "Unrecognized mount option \"%s\" "
@@ -1882,7 +1974,8 @@ int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1882} 1974}
1883 1975
1884/* Called at mount-time, super-block is locked */ 1976/* Called at mount-time, super-block is locked */
1885static int ext4_check_descriptors(struct super_block *sb) 1977static int ext4_check_descriptors(struct super_block *sb,
1978 ext4_group_t *first_not_zeroed)
1886{ 1979{
1887 struct ext4_sb_info *sbi = EXT4_SB(sb); 1980 struct ext4_sb_info *sbi = EXT4_SB(sb);
1888 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block); 1981 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
@@ -1891,7 +1984,7 @@ static int ext4_check_descriptors(struct super_block *sb)
1891 ext4_fsblk_t inode_bitmap; 1984 ext4_fsblk_t inode_bitmap;
1892 ext4_fsblk_t inode_table; 1985 ext4_fsblk_t inode_table;
1893 int flexbg_flag = 0; 1986 int flexbg_flag = 0;
1894 ext4_group_t i; 1987 ext4_group_t i, grp = sbi->s_groups_count;
1895 1988
1896 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG)) 1989 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1897 flexbg_flag = 1; 1990 flexbg_flag = 1;
@@ -1907,6 +2000,10 @@ static int ext4_check_descriptors(struct super_block *sb)
1907 last_block = first_block + 2000 last_block = first_block +
1908 (EXT4_BLOCKS_PER_GROUP(sb) - 1); 2001 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
1909 2002
2003 if ((grp == sbi->s_groups_count) &&
2004 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2005 grp = i;
2006
1910 block_bitmap = ext4_block_bitmap(sb, gdp); 2007 block_bitmap = ext4_block_bitmap(sb, gdp);
1911 if (block_bitmap < first_block || block_bitmap > last_block) { 2008 if (block_bitmap < first_block || block_bitmap > last_block) {
1912 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: " 2009 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
@@ -1944,6 +2041,8 @@ static int ext4_check_descriptors(struct super_block *sb)
1944 if (!flexbg_flag) 2041 if (!flexbg_flag)
1945 first_block += EXT4_BLOCKS_PER_GROUP(sb); 2042 first_block += EXT4_BLOCKS_PER_GROUP(sb);
1946 } 2043 }
2044 if (NULL != first_not_zeroed)
2045 *first_not_zeroed = grp;
1947 2046
1948 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb)); 2047 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
1949 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb)); 2048 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
@@ -2249,6 +2348,8 @@ static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2249{ 2348{
2250 struct super_block *sb = sbi->s_buddy_cache->i_sb; 2349 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2251 2350
2351 if (!sb->s_bdev->bd_part)
2352 return snprintf(buf, PAGE_SIZE, "0\n");
2252 return snprintf(buf, PAGE_SIZE, "%lu\n", 2353 return snprintf(buf, PAGE_SIZE, "%lu\n",
2253 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) - 2354 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2254 sbi->s_sectors_written_start) >> 1); 2355 sbi->s_sectors_written_start) >> 1);
@@ -2259,6 +2360,8 @@ static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2259{ 2360{
2260 struct super_block *sb = sbi->s_buddy_cache->i_sb; 2361 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2261 2362
2363 if (!sb->s_bdev->bd_part)
2364 return snprintf(buf, PAGE_SIZE, "0\n");
2262 return snprintf(buf, PAGE_SIZE, "%llu\n", 2365 return snprintf(buf, PAGE_SIZE, "%llu\n",
2263 (unsigned long long)(sbi->s_kbytes_written + 2366 (unsigned long long)(sbi->s_kbytes_written +
2264 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) - 2367 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
@@ -2312,6 +2415,7 @@ static struct ext4_attr ext4_attr_##_name = { \
2312#define EXT4_ATTR(name, mode, show, store) \ 2415#define EXT4_ATTR(name, mode, show, store) \
2313static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store) 2416static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2314 2417
2418#define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
2315#define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL) 2419#define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2316#define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store) 2420#define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2317#define EXT4_RW_ATTR_SBI_UI(name, elname) \ 2421#define EXT4_RW_ATTR_SBI_UI(name, elname) \
@@ -2348,6 +2452,16 @@ static struct attribute *ext4_attrs[] = {
2348 NULL, 2452 NULL,
2349}; 2453};
2350 2454
2455/* Features this copy of ext4 supports */
2456EXT4_INFO_ATTR(lazy_itable_init);
2457EXT4_INFO_ATTR(batched_discard);
2458
2459static struct attribute *ext4_feat_attrs[] = {
2460 ATTR_LIST(lazy_itable_init),
2461 ATTR_LIST(batched_discard),
2462 NULL,
2463};
2464
2351static ssize_t ext4_attr_show(struct kobject *kobj, 2465static ssize_t ext4_attr_show(struct kobject *kobj,
2352 struct attribute *attr, char *buf) 2466 struct attribute *attr, char *buf)
2353{ 2467{
@@ -2376,7 +2490,6 @@ static void ext4_sb_release(struct kobject *kobj)
2376 complete(&sbi->s_kobj_unregister); 2490 complete(&sbi->s_kobj_unregister);
2377} 2491}
2378 2492
2379
2380static const struct sysfs_ops ext4_attr_ops = { 2493static const struct sysfs_ops ext4_attr_ops = {
2381 .show = ext4_attr_show, 2494 .show = ext4_attr_show,
2382 .store = ext4_attr_store, 2495 .store = ext4_attr_store,
@@ -2388,6 +2501,17 @@ static struct kobj_type ext4_ktype = {
2388 .release = ext4_sb_release, 2501 .release = ext4_sb_release,
2389}; 2502};
2390 2503
2504static void ext4_feat_release(struct kobject *kobj)
2505{
2506 complete(&ext4_feat->f_kobj_unregister);
2507}
2508
2509static struct kobj_type ext4_feat_ktype = {
2510 .default_attrs = ext4_feat_attrs,
2511 .sysfs_ops = &ext4_attr_ops,
2512 .release = ext4_feat_release,
2513};
2514
2391/* 2515/*
2392 * Check whether this filesystem can be mounted based on 2516 * Check whether this filesystem can be mounted based on
2393 * the features present and the RDONLY/RDWR mount requested. 2517 * the features present and the RDONLY/RDWR mount requested.
@@ -2431,6 +2555,419 @@ static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2431 return 1; 2555 return 1;
2432} 2556}
2433 2557
2558/*
2559 * This function is called once a day if we have errors logged
2560 * on the file system
2561 */
2562static void print_daily_error_info(unsigned long arg)
2563{
2564 struct super_block *sb = (struct super_block *) arg;
2565 struct ext4_sb_info *sbi;
2566 struct ext4_super_block *es;
2567
2568 sbi = EXT4_SB(sb);
2569 es = sbi->s_es;
2570
2571 if (es->s_error_count)
2572 ext4_msg(sb, KERN_NOTICE, "error count: %u",
2573 le32_to_cpu(es->s_error_count));
2574 if (es->s_first_error_time) {
2575 printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d",
2576 sb->s_id, le32_to_cpu(es->s_first_error_time),
2577 (int) sizeof(es->s_first_error_func),
2578 es->s_first_error_func,
2579 le32_to_cpu(es->s_first_error_line));
2580 if (es->s_first_error_ino)
2581 printk(": inode %u",
2582 le32_to_cpu(es->s_first_error_ino));
2583 if (es->s_first_error_block)
2584 printk(": block %llu", (unsigned long long)
2585 le64_to_cpu(es->s_first_error_block));
2586 printk("\n");
2587 }
2588 if (es->s_last_error_time) {
2589 printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d",
2590 sb->s_id, le32_to_cpu(es->s_last_error_time),
2591 (int) sizeof(es->s_last_error_func),
2592 es->s_last_error_func,
2593 le32_to_cpu(es->s_last_error_line));
2594 if (es->s_last_error_ino)
2595 printk(": inode %u",
2596 le32_to_cpu(es->s_last_error_ino));
2597 if (es->s_last_error_block)
2598 printk(": block %llu", (unsigned long long)
2599 le64_to_cpu(es->s_last_error_block));
2600 printk("\n");
2601 }
2602 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2603}
2604
2605static void ext4_lazyinode_timeout(unsigned long data)
2606{
2607 struct task_struct *p = (struct task_struct *)data;
2608 wake_up_process(p);
2609}
2610
2611/* Find next suitable group and run ext4_init_inode_table */
2612static int ext4_run_li_request(struct ext4_li_request *elr)
2613{
2614 struct ext4_group_desc *gdp = NULL;
2615 ext4_group_t group, ngroups;
2616 struct super_block *sb;
2617 unsigned long timeout = 0;
2618 int ret = 0;
2619
2620 sb = elr->lr_super;
2621 ngroups = EXT4_SB(sb)->s_groups_count;
2622
2623 for (group = elr->lr_next_group; group < ngroups; group++) {
2624 gdp = ext4_get_group_desc(sb, group, NULL);
2625 if (!gdp) {
2626 ret = 1;
2627 break;
2628 }
2629
2630 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2631 break;
2632 }
2633
2634 if (group == ngroups)
2635 ret = 1;
2636
2637 if (!ret) {
2638 timeout = jiffies;
2639 ret = ext4_init_inode_table(sb, group,
2640 elr->lr_timeout ? 0 : 1);
2641 if (elr->lr_timeout == 0) {
2642 timeout = jiffies - timeout;
2643 if (elr->lr_sbi->s_li_wait_mult)
2644 timeout *= elr->lr_sbi->s_li_wait_mult;
2645 else
2646 timeout *= 20;
2647 elr->lr_timeout = timeout;
2648 }
2649 elr->lr_next_sched = jiffies + elr->lr_timeout;
2650 elr->lr_next_group = group + 1;
2651 }
2652
2653 return ret;
2654}
2655
2656/*
2657 * Remove lr_request from the list_request and free the
2658 * request tructure. Should be called with li_list_mtx held
2659 */
2660static void ext4_remove_li_request(struct ext4_li_request *elr)
2661{
2662 struct ext4_sb_info *sbi;
2663
2664 if (!elr)
2665 return;
2666
2667 sbi = elr->lr_sbi;
2668
2669 list_del(&elr->lr_request);
2670 sbi->s_li_request = NULL;
2671 kfree(elr);
2672}
2673
2674static void ext4_unregister_li_request(struct super_block *sb)
2675{
2676 struct ext4_li_request *elr = EXT4_SB(sb)->s_li_request;
2677
2678 if (!ext4_li_info)
2679 return;
2680
2681 mutex_lock(&ext4_li_info->li_list_mtx);
2682 ext4_remove_li_request(elr);
2683 mutex_unlock(&ext4_li_info->li_list_mtx);
2684}
2685
2686/*
2687 * This is the function where ext4lazyinit thread lives. It walks
2688 * through the request list searching for next scheduled filesystem.
2689 * When such a fs is found, run the lazy initialization request
2690 * (ext4_rn_li_request) and keep track of the time spend in this
2691 * function. Based on that time we compute next schedule time of
2692 * the request. When walking through the list is complete, compute
2693 * next waking time and put itself into sleep.
2694 */
2695static int ext4_lazyinit_thread(void *arg)
2696{
2697 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2698 struct list_head *pos, *n;
2699 struct ext4_li_request *elr;
2700 unsigned long next_wakeup;
2701 DEFINE_WAIT(wait);
2702 int ret;
2703
2704 BUG_ON(NULL == eli);
2705
2706 eli->li_timer.data = (unsigned long)current;
2707 eli->li_timer.function = ext4_lazyinode_timeout;
2708
2709 eli->li_task = current;
2710 wake_up(&eli->li_wait_task);
2711
2712cont_thread:
2713 while (true) {
2714 next_wakeup = MAX_JIFFY_OFFSET;
2715
2716 mutex_lock(&eli->li_list_mtx);
2717 if (list_empty(&eli->li_request_list)) {
2718 mutex_unlock(&eli->li_list_mtx);
2719 goto exit_thread;
2720 }
2721
2722 list_for_each_safe(pos, n, &eli->li_request_list) {
2723 elr = list_entry(pos, struct ext4_li_request,
2724 lr_request);
2725
2726 if (time_after_eq(jiffies, elr->lr_next_sched))
2727 ret = ext4_run_li_request(elr);
2728
2729 if (ret) {
2730 ret = 0;
2731 ext4_remove_li_request(elr);
2732 continue;
2733 }
2734
2735 if (time_before(elr->lr_next_sched, next_wakeup))
2736 next_wakeup = elr->lr_next_sched;
2737 }
2738 mutex_unlock(&eli->li_list_mtx);
2739
2740 if (freezing(current))
2741 refrigerator();
2742
2743 if (time_after_eq(jiffies, next_wakeup)) {
2744 cond_resched();
2745 continue;
2746 }
2747
2748 eli->li_timer.expires = next_wakeup;
2749 add_timer(&eli->li_timer);
2750 prepare_to_wait(&eli->li_wait_daemon, &wait,
2751 TASK_INTERRUPTIBLE);
2752 if (time_before(jiffies, next_wakeup))
2753 schedule();
2754 finish_wait(&eli->li_wait_daemon, &wait);
2755 }
2756
2757exit_thread:
2758 /*
2759 * It looks like the request list is empty, but we need
2760 * to check it under the li_list_mtx lock, to prevent any
2761 * additions into it, and of course we should lock ext4_li_mtx
2762 * to atomically free the list and ext4_li_info, because at
2763 * this point another ext4 filesystem could be registering
2764 * new one.
2765 */
2766 mutex_lock(&ext4_li_mtx);
2767 mutex_lock(&eli->li_list_mtx);
2768 if (!list_empty(&eli->li_request_list)) {
2769 mutex_unlock(&eli->li_list_mtx);
2770 mutex_unlock(&ext4_li_mtx);
2771 goto cont_thread;
2772 }
2773 mutex_unlock(&eli->li_list_mtx);
2774 del_timer_sync(&ext4_li_info->li_timer);
2775 eli->li_task = NULL;
2776 wake_up(&eli->li_wait_task);
2777
2778 kfree(ext4_li_info);
2779 ext4_li_info = NULL;
2780 mutex_unlock(&ext4_li_mtx);
2781
2782 return 0;
2783}
2784
2785static void ext4_clear_request_list(void)
2786{
2787 struct list_head *pos, *n;
2788 struct ext4_li_request *elr;
2789
2790 mutex_lock(&ext4_li_info->li_list_mtx);
2791 if (list_empty(&ext4_li_info->li_request_list))
2792 return;
2793
2794 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2795 elr = list_entry(pos, struct ext4_li_request,
2796 lr_request);
2797 ext4_remove_li_request(elr);
2798 }
2799 mutex_unlock(&ext4_li_info->li_list_mtx);
2800}
2801
2802static int ext4_run_lazyinit_thread(void)
2803{
2804 struct task_struct *t;
2805
2806 t = kthread_run(ext4_lazyinit_thread, ext4_li_info, "ext4lazyinit");
2807 if (IS_ERR(t)) {
2808 int err = PTR_ERR(t);
2809 ext4_clear_request_list();
2810 del_timer_sync(&ext4_li_info->li_timer);
2811 kfree(ext4_li_info);
2812 ext4_li_info = NULL;
2813 printk(KERN_CRIT "EXT4: error %d creating inode table "
2814 "initialization thread\n",
2815 err);
2816 return err;
2817 }
2818 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
2819
2820 wait_event(ext4_li_info->li_wait_task, ext4_li_info->li_task != NULL);
2821 return 0;
2822}
2823
2824/*
2825 * Check whether it make sense to run itable init. thread or not.
2826 * If there is at least one uninitialized inode table, return
2827 * corresponding group number, else the loop goes through all
2828 * groups and return total number of groups.
2829 */
2830static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2831{
2832 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2833 struct ext4_group_desc *gdp = NULL;
2834
2835 for (group = 0; group < ngroups; group++) {
2836 gdp = ext4_get_group_desc(sb, group, NULL);
2837 if (!gdp)
2838 continue;
2839
2840 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2841 break;
2842 }
2843
2844 return group;
2845}
2846
2847static int ext4_li_info_new(void)
2848{
2849 struct ext4_lazy_init *eli = NULL;
2850
2851 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2852 if (!eli)
2853 return -ENOMEM;
2854
2855 eli->li_task = NULL;
2856 INIT_LIST_HEAD(&eli->li_request_list);
2857 mutex_init(&eli->li_list_mtx);
2858
2859 init_waitqueue_head(&eli->li_wait_daemon);
2860 init_waitqueue_head(&eli->li_wait_task);
2861 init_timer(&eli->li_timer);
2862 eli->li_state |= EXT4_LAZYINIT_QUIT;
2863
2864 ext4_li_info = eli;
2865
2866 return 0;
2867}
2868
2869static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
2870 ext4_group_t start)
2871{
2872 struct ext4_sb_info *sbi = EXT4_SB(sb);
2873 struct ext4_li_request *elr;
2874 unsigned long rnd;
2875
2876 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
2877 if (!elr)
2878 return NULL;
2879
2880 elr->lr_super = sb;
2881 elr->lr_sbi = sbi;
2882 elr->lr_next_group = start;
2883
2884 /*
2885 * Randomize first schedule time of the request to
2886 * spread the inode table initialization requests
2887 * better.
2888 */
2889 get_random_bytes(&rnd, sizeof(rnd));
2890 elr->lr_next_sched = jiffies + (unsigned long)rnd %
2891 (EXT4_DEF_LI_MAX_START_DELAY * HZ);
2892
2893 return elr;
2894}
2895
2896static int ext4_register_li_request(struct super_block *sb,
2897 ext4_group_t first_not_zeroed)
2898{
2899 struct ext4_sb_info *sbi = EXT4_SB(sb);
2900 struct ext4_li_request *elr;
2901 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
2902 int ret;
2903
2904 if (sbi->s_li_request != NULL)
2905 return 0;
2906
2907 if (first_not_zeroed == ngroups ||
2908 (sb->s_flags & MS_RDONLY) ||
2909 !test_opt(sb, INIT_INODE_TABLE)) {
2910 sbi->s_li_request = NULL;
2911 return 0;
2912 }
2913
2914 if (first_not_zeroed == ngroups) {
2915 sbi->s_li_request = NULL;
2916 return 0;
2917 }
2918
2919 elr = ext4_li_request_new(sb, first_not_zeroed);
2920 if (!elr)
2921 return -ENOMEM;
2922
2923 mutex_lock(&ext4_li_mtx);
2924
2925 if (NULL == ext4_li_info) {
2926 ret = ext4_li_info_new();
2927 if (ret)
2928 goto out;
2929 }
2930
2931 mutex_lock(&ext4_li_info->li_list_mtx);
2932 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
2933 mutex_unlock(&ext4_li_info->li_list_mtx);
2934
2935 sbi->s_li_request = elr;
2936
2937 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
2938 ret = ext4_run_lazyinit_thread();
2939 if (ret)
2940 goto out;
2941 }
2942out:
2943 mutex_unlock(&ext4_li_mtx);
2944 if (ret)
2945 kfree(elr);
2946 return ret;
2947}
2948
2949/*
2950 * We do not need to lock anything since this is called on
2951 * module unload.
2952 */
2953static void ext4_destroy_lazyinit_thread(void)
2954{
2955 /*
2956 * If thread exited earlier
2957 * there's nothing to be done.
2958 */
2959 if (!ext4_li_info)
2960 return;
2961
2962 ext4_clear_request_list();
2963
2964 while (ext4_li_info->li_task) {
2965 wake_up(&ext4_li_info->li_wait_daemon);
2966 wait_event(ext4_li_info->li_wait_task,
2967 ext4_li_info->li_task == NULL);
2968 }
2969}
2970
2434static int ext4_fill_super(struct super_block *sb, void *data, int silent) 2971static int ext4_fill_super(struct super_block *sb, void *data, int silent)
2435 __releases(kernel_lock) 2972 __releases(kernel_lock)
2436 __acquires(kernel_lock) 2973 __acquires(kernel_lock)
@@ -2448,7 +2985,7 @@ static int ext4_fill_super(struct super_block *sb, void *data, int silent)
2448 struct inode *root; 2985 struct inode *root;
2449 char *cp; 2986 char *cp;
2450 const char *descr; 2987 const char *descr;
2451 int ret = -EINVAL; 2988 int ret = -ENOMEM;
2452 int blocksize; 2989 int blocksize;
2453 unsigned int db_count; 2990 unsigned int db_count;
2454 unsigned int i; 2991 unsigned int i;
@@ -2456,16 +2993,17 @@ static int ext4_fill_super(struct super_block *sb, void *data, int silent)
2456 __u64 blocks_count; 2993 __u64 blocks_count;
2457 int err; 2994 int err;
2458 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO; 2995 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
2996 ext4_group_t first_not_zeroed;
2459 2997
2460 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL); 2998 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
2461 if (!sbi) 2999 if (!sbi)
2462 return -ENOMEM; 3000 goto out_free_orig;
2463 3001
2464 sbi->s_blockgroup_lock = 3002 sbi->s_blockgroup_lock =
2465 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL); 3003 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
2466 if (!sbi->s_blockgroup_lock) { 3004 if (!sbi->s_blockgroup_lock) {
2467 kfree(sbi); 3005 kfree(sbi);
2468 return -ENOMEM; 3006 goto out_free_orig;
2469 } 3007 }
2470 sb->s_fs_info = sbi; 3008 sb->s_fs_info = sbi;
2471 sbi->s_mount_opt = 0; 3009 sbi->s_mount_opt = 0;
@@ -2473,15 +3011,15 @@ static int ext4_fill_super(struct super_block *sb, void *data, int silent)
2473 sbi->s_resgid = EXT4_DEF_RESGID; 3011 sbi->s_resgid = EXT4_DEF_RESGID;
2474 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS; 3012 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
2475 sbi->s_sb_block = sb_block; 3013 sbi->s_sb_block = sb_block;
2476 sbi->s_sectors_written_start = part_stat_read(sb->s_bdev->bd_part, 3014 if (sb->s_bdev->bd_part)
2477 sectors[1]); 3015 sbi->s_sectors_written_start =
2478 3016 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
2479 unlock_kernel();
2480 3017
2481 /* Cleanup superblock name */ 3018 /* Cleanup superblock name */
2482 for (cp = sb->s_id; (cp = strchr(cp, '/'));) 3019 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
2483 *cp = '!'; 3020 *cp = '!';
2484 3021
3022 ret = -EINVAL;
2485 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE); 3023 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
2486 if (!blocksize) { 3024 if (!blocksize) {
2487 ext4_msg(sb, KERN_ERR, "unable to set blocksize"); 3025 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
@@ -2516,6 +3054,7 @@ static int ext4_fill_super(struct super_block *sb, void *data, int silent)
2516 3054
2517 /* Set defaults before we parse the mount options */ 3055 /* Set defaults before we parse the mount options */
2518 def_mount_opts = le32_to_cpu(es->s_default_mount_opts); 3056 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3057 set_opt(sbi->s_mount_opt, INIT_INODE_TABLE);
2519 if (def_mount_opts & EXT4_DEFM_DEBUG) 3058 if (def_mount_opts & EXT4_DEFM_DEBUG)
2520 set_opt(sbi->s_mount_opt, DEBUG); 3059 set_opt(sbi->s_mount_opt, DEBUG);
2521 if (def_mount_opts & EXT4_DEFM_BSDGROUPS) { 3060 if (def_mount_opts & EXT4_DEFM_BSDGROUPS) {
@@ -2546,6 +3085,10 @@ static int ext4_fill_super(struct super_block *sb, void *data, int silent)
2546 set_opt(sbi->s_mount_opt, ERRORS_CONT); 3085 set_opt(sbi->s_mount_opt, ERRORS_CONT);
2547 else 3086 else
2548 set_opt(sbi->s_mount_opt, ERRORS_RO); 3087 set_opt(sbi->s_mount_opt, ERRORS_RO);
3088 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
3089 set_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
3090 if (def_mount_opts & EXT4_DEFM_DISCARD)
3091 set_opt(sbi->s_mount_opt, DISCARD);
2549 3092
2550 sbi->s_resuid = le16_to_cpu(es->s_def_resuid); 3093 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
2551 sbi->s_resgid = le16_to_cpu(es->s_def_resgid); 3094 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
@@ -2553,15 +3096,23 @@ static int ext4_fill_super(struct super_block *sb, void *data, int silent)
2553 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME; 3096 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
2554 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME; 3097 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
2555 3098
2556 set_opt(sbi->s_mount_opt, BARRIER); 3099 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3100 set_opt(sbi->s_mount_opt, BARRIER);
2557 3101
2558 /* 3102 /*
2559 * enable delayed allocation by default 3103 * enable delayed allocation by default
2560 * Use -o nodelalloc to turn it off 3104 * Use -o nodelalloc to turn it off
2561 */ 3105 */
2562 if (!IS_EXT3_SB(sb)) 3106 if (!IS_EXT3_SB(sb) &&
3107 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
2563 set_opt(sbi->s_mount_opt, DELALLOC); 3108 set_opt(sbi->s_mount_opt, DELALLOC);
2564 3109
3110 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3111 &journal_devnum, &journal_ioprio, NULL, 0)) {
3112 ext4_msg(sb, KERN_WARNING,
3113 "failed to parse options in superblock: %s",
3114 sbi->s_es->s_mount_opts);
3115 }
2565 if (!parse_options((char *) data, sb, &journal_devnum, 3116 if (!parse_options((char *) data, sb, &journal_devnum,
2566 &journal_ioprio, NULL, 0)) 3117 &journal_ioprio, NULL, 0))
2567 goto failed_mount; 3118 goto failed_mount;
@@ -2706,15 +3257,13 @@ static int ext4_fill_super(struct super_block *sb, void *data, int silent)
2706 * Test whether we have more sectors than will fit in sector_t, 3257 * Test whether we have more sectors than will fit in sector_t,
2707 * and whether the max offset is addressable by the page cache. 3258 * and whether the max offset is addressable by the page cache.
2708 */ 3259 */
2709 if ((ext4_blocks_count(es) > 3260 ret = generic_check_addressable(sb->s_blocksize_bits,
2710 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) || 3261 ext4_blocks_count(es));
2711 (ext4_blocks_count(es) > 3262 if (ret) {
2712 (pgoff_t)(~0ULL) >> (PAGE_CACHE_SHIFT - sb->s_blocksize_bits))) {
2713 ext4_msg(sb, KERN_ERR, "filesystem" 3263 ext4_msg(sb, KERN_ERR, "filesystem"
2714 " too large to mount safely on this system"); 3264 " too large to mount safely on this system");
2715 if (sizeof(sector_t) < 8) 3265 if (sizeof(sector_t) < 8)
2716 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled"); 3266 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
2717 ret = -EFBIG;
2718 goto failed_mount; 3267 goto failed_mount;
2719 } 3268 }
2720 3269
@@ -2783,7 +3332,7 @@ static int ext4_fill_super(struct super_block *sb, void *data, int silent)
2783 goto failed_mount2; 3332 goto failed_mount2;
2784 } 3333 }
2785 } 3334 }
2786 if (!ext4_check_descriptors(sb)) { 3335 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
2787 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!"); 3336 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
2788 goto failed_mount2; 3337 goto failed_mount2;
2789 } 3338 }
@@ -2912,18 +3461,7 @@ no_journal:
2912 ext4_msg(sb, KERN_ERR, "insufficient memory"); 3461 ext4_msg(sb, KERN_ERR, "insufficient memory");
2913 goto failed_mount_wq; 3462 goto failed_mount_wq;
2914 } 3463 }
2915 if (test_opt(sb, NOBH)) { 3464
2916 if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
2917 ext4_msg(sb, KERN_WARNING, "Ignoring nobh option - "
2918 "its supported only with writeback mode");
2919 clear_opt(sbi->s_mount_opt, NOBH);
2920 }
2921 if (test_opt(sb, DIOREAD_NOLOCK)) {
2922 ext4_msg(sb, KERN_WARNING, "dioread_nolock option is "
2923 "not supported with nobh mode");
2924 goto failed_mount_wq;
2925 }
2926 }
2927 EXT4_SB(sb)->dio_unwritten_wq = create_workqueue("ext4-dio-unwritten"); 3465 EXT4_SB(sb)->dio_unwritten_wq = create_workqueue("ext4-dio-unwritten");
2928 if (!EXT4_SB(sb)->dio_unwritten_wq) { 3466 if (!EXT4_SB(sb)->dio_unwritten_wq) {
2929 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n"); 3467 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
@@ -3010,11 +3548,15 @@ no_journal:
3010 ext4_ext_init(sb); 3548 ext4_ext_init(sb);
3011 err = ext4_mb_init(sb, needs_recovery); 3549 err = ext4_mb_init(sb, needs_recovery);
3012 if (err) { 3550 if (err) {
3013 ext4_msg(sb, KERN_ERR, "failed to initalize mballoc (%d)", 3551 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
3014 err); 3552 err);
3015 goto failed_mount4; 3553 goto failed_mount4;
3016 } 3554 }
3017 3555
3556 err = ext4_register_li_request(sb, first_not_zeroed);
3557 if (err)
3558 goto failed_mount4;
3559
3018 sbi->s_kobj.kset = ext4_kset; 3560 sbi->s_kobj.kset = ext4_kset;
3019 init_completion(&sbi->s_kobj_unregister); 3561 init_completion(&sbi->s_kobj_unregister);
3020 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL, 3562 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
@@ -3043,9 +3585,15 @@ no_journal:
3043 descr = "out journal"; 3585 descr = "out journal";
3044 3586
3045 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. " 3587 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
3046 "Opts: %s", descr, orig_data); 3588 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
3589 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
3590
3591 init_timer(&sbi->s_err_report);
3592 sbi->s_err_report.function = print_daily_error_info;
3593 sbi->s_err_report.data = (unsigned long) sb;
3594 if (es->s_error_count)
3595 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
3047 3596
3048 lock_kernel();
3049 kfree(orig_data); 3597 kfree(orig_data);
3050 return 0; 3598 return 0;
3051 3599
@@ -3092,7 +3640,7 @@ out_fail:
3092 sb->s_fs_info = NULL; 3640 sb->s_fs_info = NULL;
3093 kfree(sbi->s_blockgroup_lock); 3641 kfree(sbi->s_blockgroup_lock);
3094 kfree(sbi); 3642 kfree(sbi);
3095 lock_kernel(); 3643out_free_orig:
3096 kfree(orig_data); 3644 kfree(orig_data);
3097 return ret; 3645 return ret;
3098} 3646}
@@ -3110,7 +3658,7 @@ static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
3110 journal->j_min_batch_time = sbi->s_min_batch_time; 3658 journal->j_min_batch_time = sbi->s_min_batch_time;
3111 journal->j_max_batch_time = sbi->s_max_batch_time; 3659 journal->j_max_batch_time = sbi->s_max_batch_time;
3112 3660
3113 spin_lock(&journal->j_state_lock); 3661 write_lock(&journal->j_state_lock);
3114 if (test_opt(sb, BARRIER)) 3662 if (test_opt(sb, BARRIER))
3115 journal->j_flags |= JBD2_BARRIER; 3663 journal->j_flags |= JBD2_BARRIER;
3116 else 3664 else
@@ -3119,7 +3667,7 @@ static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
3119 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR; 3667 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
3120 else 3668 else
3121 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR; 3669 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
3122 spin_unlock(&journal->j_state_lock); 3670 write_unlock(&journal->j_state_lock);
3123} 3671}
3124 3672
3125static journal_t *ext4_get_journal(struct super_block *sb, 3673static journal_t *ext4_get_journal(struct super_block *sb,
@@ -3327,8 +3875,17 @@ static int ext4_load_journal(struct super_block *sb,
3327 3875
3328 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) 3876 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
3329 err = jbd2_journal_wipe(journal, !really_read_only); 3877 err = jbd2_journal_wipe(journal, !really_read_only);
3330 if (!err) 3878 if (!err) {
3879 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
3880 if (save)
3881 memcpy(save, ((char *) es) +
3882 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
3331 err = jbd2_journal_load(journal); 3883 err = jbd2_journal_load(journal);
3884 if (save)
3885 memcpy(((char *) es) + EXT4_S_ERR_START,
3886 save, EXT4_S_ERR_LEN);
3887 kfree(save);
3888 }
3332 3889
3333 if (err) { 3890 if (err) {
3334 ext4_msg(sb, KERN_ERR, "error loading journal"); 3891 ext4_msg(sb, KERN_ERR, "error loading journal");
@@ -3339,7 +3896,7 @@ static int ext4_load_journal(struct super_block *sb,
3339 EXT4_SB(sb)->s_journal = journal; 3896 EXT4_SB(sb)->s_journal = journal;
3340 ext4_clear_journal_err(sb, es); 3897 ext4_clear_journal_err(sb, es);
3341 3898
3342 if (journal_devnum && 3899 if (!really_read_only && journal_devnum &&
3343 journal_devnum != le32_to_cpu(es->s_journal_dev)) { 3900 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3344 es->s_journal_dev = cpu_to_le32(journal_devnum); 3901 es->s_journal_dev = cpu_to_le32(journal_devnum);
3345 3902
@@ -3384,13 +3941,20 @@ static int ext4_commit_super(struct super_block *sb, int sync)
3384 */ 3941 */
3385 if (!(sb->s_flags & MS_RDONLY)) 3942 if (!(sb->s_flags & MS_RDONLY))
3386 es->s_wtime = cpu_to_le32(get_seconds()); 3943 es->s_wtime = cpu_to_le32(get_seconds());
3387 es->s_kbytes_written = 3944 if (sb->s_bdev->bd_part)
3388 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written + 3945 es->s_kbytes_written =
3946 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
3389 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) - 3947 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
3390 EXT4_SB(sb)->s_sectors_written_start) >> 1)); 3948 EXT4_SB(sb)->s_sectors_written_start) >> 1));
3391 ext4_free_blocks_count_set(es, percpu_counter_sum_positive( 3949 else
3950 es->s_kbytes_written =
3951 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
3952 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeblocks_counter))
3953 ext4_free_blocks_count_set(es, percpu_counter_sum_positive(
3392 &EXT4_SB(sb)->s_freeblocks_counter)); 3954 &EXT4_SB(sb)->s_freeblocks_counter));
3393 es->s_free_inodes_count = cpu_to_le32(percpu_counter_sum_positive( 3955 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
3956 es->s_free_inodes_count =
3957 cpu_to_le32(percpu_counter_sum_positive(
3394 &EXT4_SB(sb)->s_freeinodes_counter)); 3958 &EXT4_SB(sb)->s_freeinodes_counter));
3395 sb->s_dirt = 0; 3959 sb->s_dirt = 0;
3396 BUFFER_TRACE(sbh, "marking dirty"); 3960 BUFFER_TRACE(sbh, "marking dirty");
@@ -3491,7 +4055,7 @@ int ext4_force_commit(struct super_block *sb)
3491 4055
3492 journal = EXT4_SB(sb)->s_journal; 4056 journal = EXT4_SB(sb)->s_journal;
3493 if (journal) { 4057 if (journal) {
3494 vfs_check_frozen(sb, SB_FREEZE_WRITE); 4058 vfs_check_frozen(sb, SB_FREEZE_TRANS);
3495 ret = ext4_journal_force_commit(journal); 4059 ret = ext4_journal_force_commit(journal);
3496 } 4060 }
3497 4061
@@ -3587,8 +4151,6 @@ static int ext4_remount(struct super_block *sb, int *flags, char *data)
3587#endif 4151#endif
3588 char *orig_data = kstrdup(data, GFP_KERNEL); 4152 char *orig_data = kstrdup(data, GFP_KERNEL);
3589 4153
3590 lock_kernel();
3591
3592 /* Store the original options */ 4154 /* Store the original options */
3593 lock_super(sb); 4155 lock_super(sb);
3594 old_sb_flags = sb->s_flags; 4156 old_sb_flags = sb->s_flags;
@@ -3616,7 +4178,7 @@ static int ext4_remount(struct super_block *sb, int *flags, char *data)
3616 } 4178 }
3617 4179
3618 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) 4180 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
3619 ext4_abort(sb, __func__, "Abort forced by user"); 4181 ext4_abort(sb, "Abort forced by user");
3620 4182
3621 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) | 4183 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3622 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0); 4184 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
@@ -3711,6 +4273,19 @@ static int ext4_remount(struct super_block *sb, int *flags, char *data)
3711 enable_quota = 1; 4273 enable_quota = 1;
3712 } 4274 }
3713 } 4275 }
4276
4277 /*
4278 * Reinitialize lazy itable initialization thread based on
4279 * current settings
4280 */
4281 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4282 ext4_unregister_li_request(sb);
4283 else {
4284 ext4_group_t first_not_zeroed;
4285 first_not_zeroed = ext4_has_uninit_itable(sb);
4286 ext4_register_li_request(sb, first_not_zeroed);
4287 }
4288
3714 ext4_setup_system_zone(sb); 4289 ext4_setup_system_zone(sb);
3715 if (sbi->s_journal == NULL) 4290 if (sbi->s_journal == NULL)
3716 ext4_commit_super(sb, 1); 4291 ext4_commit_super(sb, 1);
@@ -3723,7 +4298,6 @@ static int ext4_remount(struct super_block *sb, int *flags, char *data)
3723 kfree(old_opts.s_qf_names[i]); 4298 kfree(old_opts.s_qf_names[i]);
3724#endif 4299#endif
3725 unlock_super(sb); 4300 unlock_super(sb);
3726 unlock_kernel();
3727 if (enable_quota) 4301 if (enable_quota)
3728 dquot_resume(sb, -1); 4302 dquot_resume(sb, -1);
3729 4303
@@ -3749,7 +4323,6 @@ restore_opts:
3749 } 4323 }
3750#endif 4324#endif
3751 unlock_super(sb); 4325 unlock_super(sb);
3752 unlock_kernel();
3753 kfree(orig_data); 4326 kfree(orig_data);
3754 return err; 4327 return err;
3755} 4328}
@@ -3981,6 +4554,18 @@ static int ext4_quota_on(struct super_block *sb, int type, int format_id,
3981 return err; 4554 return err;
3982} 4555}
3983 4556
4557static int ext4_quota_off(struct super_block *sb, int type)
4558{
4559 /* Force all delayed allocation blocks to be allocated */
4560 if (test_opt(sb, DELALLOC)) {
4561 down_read(&sb->s_umount);
4562 sync_filesystem(sb);
4563 up_read(&sb->s_umount);
4564 }
4565
4566 return dquot_quota_off(sb, type);
4567}
4568
3984/* Read data from quotafile - avoid pagecache and such because we cannot afford 4569/* Read data from quotafile - avoid pagecache and such because we cannot afford
3985 * acquiring the locks... As quota files are never truncated and quota code 4570 * acquiring the locks... As quota files are never truncated and quota code
3986 * itself serializes the operations (and noone else should touch the files) 4571 * itself serializes the operations (and noone else should touch the files)
@@ -4030,7 +4615,6 @@ static ssize_t ext4_quota_write(struct super_block *sb, int type,
4030 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb); 4615 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
4031 int err = 0; 4616 int err = 0;
4032 int offset = off & (sb->s_blocksize - 1); 4617 int offset = off & (sb->s_blocksize - 1);
4033 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
4034 struct buffer_head *bh; 4618 struct buffer_head *bh;
4035 handle_t *handle = journal_current_handle(); 4619 handle_t *handle = journal_current_handle();
4036 4620
@@ -4055,24 +4639,16 @@ static ssize_t ext4_quota_write(struct super_block *sb, int type,
4055 bh = ext4_bread(handle, inode, blk, 1, &err); 4639 bh = ext4_bread(handle, inode, blk, 1, &err);
4056 if (!bh) 4640 if (!bh)
4057 goto out; 4641 goto out;
4058 if (journal_quota) { 4642 err = ext4_journal_get_write_access(handle, bh);
4059 err = ext4_journal_get_write_access(handle, bh); 4643 if (err) {
4060 if (err) { 4644 brelse(bh);
4061 brelse(bh); 4645 goto out;
4062 goto out;
4063 }
4064 } 4646 }
4065 lock_buffer(bh); 4647 lock_buffer(bh);
4066 memcpy(bh->b_data+offset, data, len); 4648 memcpy(bh->b_data+offset, data, len);
4067 flush_dcache_page(bh->b_page); 4649 flush_dcache_page(bh->b_page);
4068 unlock_buffer(bh); 4650 unlock_buffer(bh);
4069 if (journal_quota) 4651 err = ext4_handle_dirty_metadata(handle, NULL, bh);
4070 err = ext4_handle_dirty_metadata(handle, NULL, bh);
4071 else {
4072 /* Always do at least ordered writes for quotas */
4073 err = ext4_jbd2_file_inode(handle, inode);
4074 mark_buffer_dirty(bh);
4075 }
4076 brelse(bh); 4652 brelse(bh);
4077out: 4653out:
4078 if (err) { 4654 if (err) {
@@ -4151,23 +4727,53 @@ static struct file_system_type ext4_fs_type = {
4151 .fs_flags = FS_REQUIRES_DEV, 4727 .fs_flags = FS_REQUIRES_DEV,
4152}; 4728};
4153 4729
4154static int __init init_ext4_fs(void) 4730int __init ext4_init_feat_adverts(void)
4731{
4732 struct ext4_features *ef;
4733 int ret = -ENOMEM;
4734
4735 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
4736 if (!ef)
4737 goto out;
4738
4739 ef->f_kobj.kset = ext4_kset;
4740 init_completion(&ef->f_kobj_unregister);
4741 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
4742 "features");
4743 if (ret) {
4744 kfree(ef);
4745 goto out;
4746 }
4747
4748 ext4_feat = ef;
4749 ret = 0;
4750out:
4751 return ret;
4752}
4753
4754static int __init ext4_init_fs(void)
4155{ 4755{
4156 int err; 4756 int err;
4157 4757
4158 ext4_check_flag_values(); 4758 ext4_check_flag_values();
4159 err = init_ext4_system_zone(); 4759 err = ext4_init_pageio();
4160 if (err) 4760 if (err)
4161 return err; 4761 return err;
4762 err = ext4_init_system_zone();
4763 if (err)
4764 goto out5;
4162 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj); 4765 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
4163 if (!ext4_kset) 4766 if (!ext4_kset)
4164 goto out4; 4767 goto out4;
4165 ext4_proc_root = proc_mkdir("fs/ext4", NULL); 4768 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
4166 err = init_ext4_mballoc(); 4769
4770 err = ext4_init_feat_adverts();
4771
4772 err = ext4_init_mballoc();
4167 if (err) 4773 if (err)
4168 goto out3; 4774 goto out3;
4169 4775
4170 err = init_ext4_xattr(); 4776 err = ext4_init_xattr();
4171 if (err) 4777 if (err)
4172 goto out2; 4778 goto out2;
4173 err = init_inodecache(); 4779 err = init_inodecache();
@@ -4178,38 +4784,46 @@ static int __init init_ext4_fs(void)
4178 err = register_filesystem(&ext4_fs_type); 4784 err = register_filesystem(&ext4_fs_type);
4179 if (err) 4785 if (err)
4180 goto out; 4786 goto out;
4787
4788 ext4_li_info = NULL;
4789 mutex_init(&ext4_li_mtx);
4181 return 0; 4790 return 0;
4182out: 4791out:
4183 unregister_as_ext2(); 4792 unregister_as_ext2();
4184 unregister_as_ext3(); 4793 unregister_as_ext3();
4185 destroy_inodecache(); 4794 destroy_inodecache();
4186out1: 4795out1:
4187 exit_ext4_xattr(); 4796 ext4_exit_xattr();
4188out2: 4797out2:
4189 exit_ext4_mballoc(); 4798 ext4_exit_mballoc();
4190out3: 4799out3:
4800 kfree(ext4_feat);
4191 remove_proc_entry("fs/ext4", NULL); 4801 remove_proc_entry("fs/ext4", NULL);
4192 kset_unregister(ext4_kset); 4802 kset_unregister(ext4_kset);
4193out4: 4803out4:
4194 exit_ext4_system_zone(); 4804 ext4_exit_system_zone();
4805out5:
4806 ext4_exit_pageio();
4195 return err; 4807 return err;
4196} 4808}
4197 4809
4198static void __exit exit_ext4_fs(void) 4810static void __exit ext4_exit_fs(void)
4199{ 4811{
4812 ext4_destroy_lazyinit_thread();
4200 unregister_as_ext2(); 4813 unregister_as_ext2();
4201 unregister_as_ext3(); 4814 unregister_as_ext3();
4202 unregister_filesystem(&ext4_fs_type); 4815 unregister_filesystem(&ext4_fs_type);
4203 destroy_inodecache(); 4816 destroy_inodecache();
4204 exit_ext4_xattr(); 4817 ext4_exit_xattr();
4205 exit_ext4_mballoc(); 4818 ext4_exit_mballoc();
4206 remove_proc_entry("fs/ext4", NULL); 4819 remove_proc_entry("fs/ext4", NULL);
4207 kset_unregister(ext4_kset); 4820 kset_unregister(ext4_kset);
4208 exit_ext4_system_zone(); 4821 ext4_exit_system_zone();
4822 ext4_exit_pageio();
4209} 4823}
4210 4824
4211MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others"); 4825MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
4212MODULE_DESCRIPTION("Fourth Extended Filesystem"); 4826MODULE_DESCRIPTION("Fourth Extended Filesystem");
4213MODULE_LICENSE("GPL"); 4827MODULE_LICENSE("GPL");
4214module_init(init_ext4_fs) 4828module_init(ext4_init_fs)
4215module_exit(exit_ext4_fs) 4829module_exit(ext4_exit_fs)
diff --git a/fs/ext4/xattr.c b/fs/ext4/xattr.c
index 04338009793a..fa4b899da4b3 100644
--- a/fs/ext4/xattr.c
+++ b/fs/ext4/xattr.c
@@ -458,8 +458,7 @@ static void ext4_xattr_update_super_block(handle_t *handle,
458 458
459 if (ext4_journal_get_write_access(handle, EXT4_SB(sb)->s_sbh) == 0) { 459 if (ext4_journal_get_write_access(handle, EXT4_SB(sb)->s_sbh) == 0) {
460 EXT4_SET_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_EXT_ATTR); 460 EXT4_SET_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_EXT_ATTR);
461 sb->s_dirt = 1; 461 ext4_handle_dirty_super(handle, sb);
462 ext4_handle_dirty_metadata(handle, NULL, EXT4_SB(sb)->s_sbh);
463 } 462 }
464} 463}
465 464
@@ -1418,7 +1417,7 @@ ext4_xattr_cache_insert(struct buffer_head *bh)
1418 ea_bdebug(bh, "out of memory"); 1417 ea_bdebug(bh, "out of memory");
1419 return; 1418 return;
1420 } 1419 }
1421 error = mb_cache_entry_insert(ce, bh->b_bdev, bh->b_blocknr, &hash); 1420 error = mb_cache_entry_insert(ce, bh->b_bdev, bh->b_blocknr, hash);
1422 if (error) { 1421 if (error) {
1423 mb_cache_entry_free(ce); 1422 mb_cache_entry_free(ce);
1424 if (error == -EBUSY) { 1423 if (error == -EBUSY) {
@@ -1490,8 +1489,8 @@ ext4_xattr_cache_find(struct inode *inode, struct ext4_xattr_header *header,
1490 return NULL; /* never share */ 1489 return NULL; /* never share */
1491 ea_idebug(inode, "looking for cached blocks [%x]", (int)hash); 1490 ea_idebug(inode, "looking for cached blocks [%x]", (int)hash);
1492again: 1491again:
1493 ce = mb_cache_entry_find_first(ext4_xattr_cache, 0, 1492 ce = mb_cache_entry_find_first(ext4_xattr_cache, inode->i_sb->s_bdev,
1494 inode->i_sb->s_bdev, hash); 1493 hash);
1495 while (ce) { 1494 while (ce) {
1496 struct buffer_head *bh; 1495 struct buffer_head *bh;
1497 1496
@@ -1515,7 +1514,7 @@ again:
1515 return bh; 1514 return bh;
1516 } 1515 }
1517 brelse(bh); 1516 brelse(bh);
1518 ce = mb_cache_entry_find_next(ce, 0, inode->i_sb->s_bdev, hash); 1517 ce = mb_cache_entry_find_next(ce, inode->i_sb->s_bdev, hash);
1519 } 1518 }
1520 return NULL; 1519 return NULL;
1521} 1520}
@@ -1589,18 +1588,16 @@ static void ext4_xattr_rehash(struct ext4_xattr_header *header,
1589#undef BLOCK_HASH_SHIFT 1588#undef BLOCK_HASH_SHIFT
1590 1589
1591int __init 1590int __init
1592init_ext4_xattr(void) 1591ext4_init_xattr(void)
1593{ 1592{
1594 ext4_xattr_cache = mb_cache_create("ext4_xattr", NULL, 1593 ext4_xattr_cache = mb_cache_create("ext4_xattr", 6);
1595 sizeof(struct mb_cache_entry) +
1596 sizeof(((struct mb_cache_entry *) 0)->e_indexes[0]), 1, 6);
1597 if (!ext4_xattr_cache) 1594 if (!ext4_xattr_cache)
1598 return -ENOMEM; 1595 return -ENOMEM;
1599 return 0; 1596 return 0;
1600} 1597}
1601 1598
1602void 1599void
1603exit_ext4_xattr(void) 1600ext4_exit_xattr(void)
1604{ 1601{
1605 if (ext4_xattr_cache) 1602 if (ext4_xattr_cache)
1606 mb_cache_destroy(ext4_xattr_cache); 1603 mb_cache_destroy(ext4_xattr_cache);
diff --git a/fs/ext4/xattr.h b/fs/ext4/xattr.h
index 518e96e43905..1ef16520b950 100644
--- a/fs/ext4/xattr.h
+++ b/fs/ext4/xattr.h
@@ -83,8 +83,8 @@ extern void ext4_xattr_put_super(struct super_block *);
83extern int ext4_expand_extra_isize_ea(struct inode *inode, int new_extra_isize, 83extern int ext4_expand_extra_isize_ea(struct inode *inode, int new_extra_isize,
84 struct ext4_inode *raw_inode, handle_t *handle); 84 struct ext4_inode *raw_inode, handle_t *handle);
85 85
86extern int init_ext4_xattr(void); 86extern int __init ext4_init_xattr(void);
87extern void exit_ext4_xattr(void); 87extern void ext4_exit_xattr(void);
88 88
89extern const struct xattr_handler *ext4_xattr_handlers[]; 89extern const struct xattr_handler *ext4_xattr_handlers[];
90 90
@@ -121,14 +121,14 @@ ext4_xattr_put_super(struct super_block *sb)
121{ 121{
122} 122}
123 123
124static inline int 124static __init inline int
125init_ext4_xattr(void) 125ext4_init_xattr(void)
126{ 126{
127 return 0; 127 return 0;
128} 128}
129 129
130static inline void 130static inline void
131exit_ext4_xattr(void) 131ext4_exit_xattr(void)
132{ 132{
133} 133}
134 134
generated by cgit v1.2.2 (git 2.25.0) at 2025-04-25 18:53:38 -0400