projects / GitHub / mt8127 / android_kernel_alcatel_ttab.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
jbd2: add tracepoints which provide per-handle statistics
[GitHub/mt8127/android_kernel_alcatel_ttab.git] / fs / ext4 / super.c
CommitLineData
ac27a0ec 1/*
617ba13b 2 * linux/fs/ext4/super.c
ac27a0ec
DK		 3 *
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
8 *
9 * from
10 *
11 * linux/fs/minix/inode.c
12 *
13 * Copyright (C) 1991, 1992 Linus Torvalds
14 *
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
17 */
18
19#include <linux/module.h>
20#include <linux/string.h>
21#include <linux/fs.h>
22#include <linux/time.h>
c5ca7c76 23#include <linux/vmalloc.h>
dab291af 24#include <linux/jbd2.h>
ac27a0ec
DK		 25#include <linux/slab.h>
26#include <linux/init.h>
27#include <linux/blkdev.h>
28#include <linux/parser.h>
ac27a0ec 29#include <linux/buffer_head.h>
a5694255 30#include <linux/exportfs.h>
ac27a0ec
DK		 31#include <linux/vfs.h>
32#include <linux/random.h>
33#include <linux/mount.h>
34#include <linux/namei.h>
35#include <linux/quotaops.h>
36#include <linux/seq_file.h>
9f6200bb 37#include <linux/proc_fs.h>
3197ebdb 38#include <linux/ctype.h>
1330593e 39#include <linux/log2.h>
717d50e4 40#include <linux/crc16.h>
7abc52c2 41#include <linux/cleancache.h>
ac27a0ec
DK		 42#include <asm/uaccess.h>
43
bfff6873
LC		 44#include <linux/kthread.h>
45#include <linux/freezer.h>
46
3dcf5451 47#include "ext4.h"
4a092d73 48#include "ext4_extents.h" /* Needed for trace points definition */
3dcf5451 49#include "ext4_jbd2.h"
ac27a0ec
DK		 50#include "xattr.h"
51#include "acl.h"
3661d286 52#include "mballoc.h"
ac27a0ec 53
9bffad1e
TT		 54#define CREATE_TRACE_POINTS
55#include <trace/events/ext4.h>
56
1f109d5a 57static struct proc_dir_entry *ext4_proc_root;
3197ebdb 58static struct kset *ext4_kset;
0b75a840
LC		 59static struct ext4_lazy_init *ext4_li_info;
60static struct mutex ext4_li_mtx;
61static struct ext4_features *ext4_feat;
9f6200bb 62
617ba13b 63static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
ac27a0ec 64 unsigned long journal_devnum);
2adf6da8 65static int ext4_show_options(struct seq_file *seq, struct dentry *root);
e2d67052 66static int ext4_commit_super(struct super_block *sb, int sync);
2b2d6d01
TT		 67static void ext4_mark_recovery_complete(struct super_block *sb,
68 struct ext4_super_block *es);
69static void ext4_clear_journal_err(struct super_block *sb,
70 struct ext4_super_block *es);
617ba13b 71static int ext4_sync_fs(struct super_block *sb, int wait);
2b2d6d01 72static const char *ext4_decode_error(struct super_block *sb, int errno,
ac27a0ec 73 char nbuf[16]);
2b2d6d01
TT		 74static int ext4_remount(struct super_block *sb, int *flags, char *data);
75static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
c4be0c1d 76static int ext4_unfreeze(struct super_block *sb);
c4be0c1d 77static int ext4_freeze(struct super_block *sb);
152a0836
AV		 78static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
79 const char *dev_name, void *data);
2035e776
TT		 80static inline int ext2_feature_set_ok(struct super_block *sb);
81static inline int ext3_feature_set_ok(struct super_block *sb);
d39195c3 82static int ext4_feature_set_ok(struct super_block *sb, int readonly);
bfff6873
LC		 83static void ext4_destroy_lazyinit_thread(void);
84static void ext4_unregister_li_request(struct super_block *sb);
8f1f7453 85static void ext4_clear_request_list(void);
ac27a0ec 86
2035e776
TT		 87#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
88static struct file_system_type ext2_fs_type = {
89 .owner = THIS_MODULE,
90 .name = "ext2",
91 .mount = ext4_mount,
92 .kill_sb = kill_block_super,
93 .fs_flags = FS_REQUIRES_DEV,
94};
95#define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
96#else
97#define IS_EXT2_SB(sb) (0)
98#endif
99
100
ba69f9ab
JK		 101#if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
102static struct file_system_type ext3_fs_type = {
103 .owner = THIS_MODULE,
104 .name = "ext3",
152a0836 105 .mount = ext4_mount,
ba69f9ab
JK		 106 .kill_sb = kill_block_super,
107 .fs_flags = FS_REQUIRES_DEV,
108};
109#define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
110#else
111#define IS_EXT3_SB(sb) (0)
112#endif
bd81d8ee 113
d25425f8
DW		 114static int ext4_verify_csum_type(struct super_block *sb,
115 struct ext4_super_block *es)
116{
117 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
118 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
119 return 1;
120
121 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
122}
123
a9c47317
DW		 124static __le32 ext4_superblock_csum(struct super_block *sb,
125 struct ext4_super_block *es)
126{
127 struct ext4_sb_info *sbi = EXT4_SB(sb);
128 int offset = offsetof(struct ext4_super_block, s_checksum);
129 __u32 csum;
130
131 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
132
133 return cpu_to_le32(csum);
134}
135
136int ext4_superblock_csum_verify(struct super_block *sb,
137 struct ext4_super_block *es)
138{
139 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
140 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
141 return 1;
142
143 return es->s_checksum == ext4_superblock_csum(sb, es);
144}
145
06db49e6 146void ext4_superblock_csum_set(struct super_block *sb)
a9c47317 147{
06db49e6
TT		 148 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
149
a9c47317
DW		 150 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
151 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
152 return;
153
154 es->s_checksum = ext4_superblock_csum(sb, es);
155}
156
9933fc0a
TT		 157void *ext4_kvmalloc(size_t size, gfp_t flags)
158{
159 void *ret;
160
161 ret = kmalloc(size, flags);
162 if (!ret)
163 ret = __vmalloc(size, flags, PAGE_KERNEL);
164 return ret;
165}
166
167void *ext4_kvzalloc(size_t size, gfp_t flags)
168{
169 void *ret;
170
db9481c0 171 ret = kzalloc(size, flags);
9933fc0a
TT		 172 if (!ret)
173 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
174 return ret;
175}
176
177void ext4_kvfree(void *ptr)
178{
179 if (is_vmalloc_addr(ptr))
180 vfree(ptr);
181 else
182 kfree(ptr);
183
184}
185
8fadc143
AR		 186ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
187 struct ext4_group_desc *bg)
bd81d8ee 188{
3a14589c 189 return le32_to_cpu(bg->bg_block_bitmap_lo) |
8fadc143 190 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 191 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 192}
193
8fadc143
AR		 194ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
195 struct ext4_group_desc *bg)
bd81d8ee 196{
5272f837 197 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
8fadc143 198 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 199 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 200}
201
8fadc143
AR		 202ext4_fsblk_t ext4_inode_table(struct super_block *sb,
203 struct ext4_group_desc *bg)
bd81d8ee 204{
5272f837 205 return le32_to_cpu(bg->bg_inode_table_lo) |
8fadc143 206 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 207 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
bd81d8ee
LV		 208}
209
021b65bb
TT		 210__u32 ext4_free_group_clusters(struct super_block *sb,
211 struct ext4_group_desc *bg)
560671a0
AK		 212{
213 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
214 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 215 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
560671a0
AK		 216}
217
218__u32 ext4_free_inodes_count(struct super_block *sb,
219 struct ext4_group_desc *bg)
220{
221 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
222 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 223 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
560671a0
AK		 224}
225
226__u32 ext4_used_dirs_count(struct super_block *sb,
227 struct ext4_group_desc *bg)
228{
229 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
230 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 231 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
560671a0
AK		 232}
233
234__u32 ext4_itable_unused_count(struct super_block *sb,
235 struct ext4_group_desc *bg)
236{
237 return le16_to_cpu(bg->bg_itable_unused_lo) |
238 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 239 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
560671a0
AK		 240}
241
8fadc143
AR		 242void ext4_block_bitmap_set(struct super_block *sb,
243 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 244{
3a14589c 245 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 246 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
247 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 248}
249
8fadc143
AR		 250void ext4_inode_bitmap_set(struct super_block *sb,
251 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 252{
5272f837 253 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 254 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
255 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 256}
257
8fadc143
AR		 258void ext4_inode_table_set(struct super_block *sb,
259 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 260{
5272f837 261 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 262 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
263 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 264}
265
021b65bb
TT		 266void ext4_free_group_clusters_set(struct super_block *sb,
267 struct ext4_group_desc *bg, __u32 count)
560671a0
AK		 268{
269 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
270 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
271 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
272}
273
274void ext4_free_inodes_set(struct super_block *sb,
275 struct ext4_group_desc *bg, __u32 count)
276{
277 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
278 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
279 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
280}
281
282void ext4_used_dirs_set(struct super_block *sb,
283 struct ext4_group_desc *bg, __u32 count)
284{
285 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
286 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
287 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
288}
289
290void ext4_itable_unused_set(struct super_block *sb,
291 struct ext4_group_desc *bg, __u32 count)
292{
293 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
294 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
295 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
296}
297
d3d1faf6
CW		 298
299/* Just increment the non-pointer handle value */
300static handle_t *ext4_get_nojournal(void)
301{
302 handle_t *handle = current->journal_info;
303 unsigned long ref_cnt = (unsigned long)handle;
304
305 BUG_ON(ref_cnt >= EXT4_NOJOURNAL_MAX_REF_COUNT);
306
307 ref_cnt++;
308 handle = (handle_t *)ref_cnt;
309
310 current->journal_info = handle;
311 return handle;
312}
313
314
315/* Decrement the non-pointer handle value */
316static void ext4_put_nojournal(handle_t *handle)
317{
318 unsigned long ref_cnt = (unsigned long)handle;
319
320 BUG_ON(ref_cnt == 0);
321
322 ref_cnt--;
323 handle = (handle_t *)ref_cnt;
324
325 current->journal_info = handle;
326}
327
ac27a0ec 328/*
dab291af 329 * Wrappers for jbd2_journal_start/end.
ac27a0ec 330 */
617ba13b 331handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
ac27a0ec
DK		 332{
333 journal_t *journal;
334
12706394 335 trace_ext4_journal_start(sb, nblocks, _RET_IP_);
ac27a0ec
DK		 336 if (sb->s_flags & MS_RDONLY)
337 return ERR_PTR(-EROFS);
338
8e8ad8a5 339 WARN_ON(sb->s_writers.frozen == SB_FREEZE_COMPLETE);
617ba13b 340 journal = EXT4_SB(sb)->s_journal;
be4f27d3
YY		 341 if (!journal)
342 return ext4_get_nojournal();
343 /*
344 * Special case here: if the journal has aborted behind our
345 * backs (eg. EIO in the commit thread), then we still need to
346 * take the FS itself readonly cleanly.
347 */
348 if (is_journal_aborted(journal)) {
349 ext4_abort(sb, "Detected aborted journal");
350 return ERR_PTR(-EROFS);
ac27a0ec 351 }
be4f27d3 352 return jbd2_journal_start(journal, nblocks);
ac27a0ec
DK		 353}
354
c398eda0 355int __ext4_journal_stop(const char *where, unsigned int line, handle_t *handle)
ac27a0ec
DK		 356{
357 struct super_block *sb;
358 int err;
359 int rc;
360
0390131b 361 if (!ext4_handle_valid(handle)) {
d3d1faf6 362 ext4_put_nojournal(handle);
0390131b
FM		 363 return 0;
364 }
ac27a0ec
DK		 365 sb = handle->h_transaction->t_journal->j_private;
366 err = handle->h_err;
dab291af 367 rc = jbd2_journal_stop(handle);
ac27a0ec
DK		 368
369 if (!err)
370 err = rc;
371 if (err)
c398eda0 372 __ext4_std_error(sb, where, line, err);
ac27a0ec
DK		 373 return err;
374}
375
90c7201b
TT		 376void ext4_journal_abort_handle(const char *caller, unsigned int line,
377 const char *err_fn, struct buffer_head *bh,
378 handle_t *handle, int err)
ac27a0ec
DK		 379{
380 char nbuf[16];
617ba13b 381 const char *errstr = ext4_decode_error(NULL, err, nbuf);
ac27a0ec 382
0390131b
FM		 383 BUG_ON(!ext4_handle_valid(handle));
384
ac27a0ec
DK		 385 if (bh)
386 BUFFER_TRACE(bh, "abort");
387
388 if (!handle->h_err)
389 handle->h_err = err;
390
391 if (is_handle_aborted(handle))
392 return;
393
92b97816 394 printk(KERN_ERR "EXT4-fs: %s:%d: aborting transaction: %s in %s\n",
90c7201b 395 caller, line, errstr, err_fn);
ac27a0ec 396
dab291af 397 jbd2_journal_abort_handle(handle);
ac27a0ec
DK		 398}
399
1c13d5c0
TT		 400static void __save_error_info(struct super_block *sb, const char *func,
401 unsigned int line)
402{
403 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
404
405 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
406 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
407 es->s_last_error_time = cpu_to_le32(get_seconds());
408 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
409 es->s_last_error_line = cpu_to_le32(line);
410 if (!es->s_first_error_time) {
411 es->s_first_error_time = es->s_last_error_time;
412 strncpy(es->s_first_error_func, func,
413 sizeof(es->s_first_error_func));
414 es->s_first_error_line = cpu_to_le32(line);
415 es->s_first_error_ino = es->s_last_error_ino;
416 es->s_first_error_block = es->s_last_error_block;
417 }
66e61a9e
TT		 418 /*
419 * Start the daily error reporting function if it hasn't been
420 * started already
421 */
422 if (!es->s_error_count)
423 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
ba39ebb6 424 le32_add_cpu(&es->s_error_count, 1);
1c13d5c0
TT		 425}
426
427static void save_error_info(struct super_block *sb, const char *func,
428 unsigned int line)
429{
430 __save_error_info(sb, func, line);
431 ext4_commit_super(sb, 1);
432}
433
7c2e7087
TT		 434/*
435 * The del_gendisk() function uninitializes the disk-specific data
436 * structures, including the bdi structure, without telling anyone
437 * else. Once this happens, any attempt to call mark_buffer_dirty()
438 * (for example, by ext4_commit_super), will cause a kernel OOPS.
439 * This is a kludge to prevent these oops until we can put in a proper
440 * hook in del_gendisk() to inform the VFS and file system layers.
441 */
442static int block_device_ejected(struct super_block *sb)
443{
444 struct inode *bd_inode = sb->s_bdev->bd_inode;
445 struct backing_dev_info *bdi = bd_inode->i_mapping->backing_dev_info;
446
447 return bdi->dev == NULL;
448}
449
18aadd47
BJ		 450static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
451{
452 struct super_block *sb = journal->j_private;
453 struct ext4_sb_info *sbi = EXT4_SB(sb);
454 int error = is_journal_aborted(journal);
455 struct ext4_journal_cb_entry *jce, *tmp;
456
457 spin_lock(&sbi->s_md_lock);
458 list_for_each_entry_safe(jce, tmp, &txn->t_private_list, jce_list) {
459 list_del_init(&jce->jce_list);
460 spin_unlock(&sbi->s_md_lock);
461 jce->jce_func(sb, jce, error);
462 spin_lock(&sbi->s_md_lock);
463 }
464 spin_unlock(&sbi->s_md_lock);
465}
1c13d5c0 466
ac27a0ec
DK		 467/* Deal with the reporting of failure conditions on a filesystem such as
468 * inconsistencies detected or read IO failures.
469 *
470 * On ext2, we can store the error state of the filesystem in the
617ba13b 471 * superblock. That is not possible on ext4, because we may have other
ac27a0ec
DK		 472 * write ordering constraints on the superblock which prevent us from
473 * writing it out straight away; and given that the journal is about to
474 * be aborted, we can't rely on the current, or future, transactions to
475 * write out the superblock safely.
476 *
dab291af 477 * We'll just use the jbd2_journal_abort() error code to record an error in
d6b198bc 478 * the journal instead. On recovery, the journal will complain about
ac27a0ec
DK		 479 * that error until we've noted it down and cleared it.
480 */
481
617ba13b 482static void ext4_handle_error(struct super_block *sb)
ac27a0ec 483{
ac27a0ec
DK		 484 if (sb->s_flags & MS_RDONLY)
485 return;
486
2b2d6d01 487 if (!test_opt(sb, ERRORS_CONT)) {
617ba13b 488 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 489
4ab2f15b 490 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
ac27a0ec 491 if (journal)
dab291af 492 jbd2_journal_abort(journal, -EIO);
ac27a0ec 493 }
2b2d6d01 494 if (test_opt(sb, ERRORS_RO)) {
b31e1552 495 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
ac27a0ec
DK		 496 sb->s_flags |= MS_RDONLY;
497 }
ac27a0ec 498 if (test_opt(sb, ERRORS_PANIC))
617ba13b 499 panic("EXT4-fs (device %s): panic forced after error\n",
ac27a0ec
DK		 500 sb->s_id);
501}
502
12062ddd 503void __ext4_error(struct super_block *sb, const char *function,
c398eda0 504 unsigned int line, const char *fmt, ...)
ac27a0ec 505{
0ff2ea7d 506 struct va_format vaf;
ac27a0ec
DK		 507 va_list args;
508
509 va_start(args, fmt);
0ff2ea7d
JP		 510 vaf.fmt = fmt;
511 vaf.va = &args;
512 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
513 sb->s_id, function, line, current->comm, &vaf);
ac27a0ec 514 va_end(args);
f3fc0210 515 save_error_info(sb, function, line);
ac27a0ec 516
617ba13b 517 ext4_handle_error(sb);
ac27a0ec
DK		 518}
519
c398eda0
TT		 520void ext4_error_inode(struct inode *inode, const char *function,
521 unsigned int line, ext4_fsblk_t block,
273df556
FM		 522 const char *fmt, ...)
523{
524 va_list args;
f7c21177 525 struct va_format vaf;
1c13d5c0 526 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
273df556 527
1c13d5c0
TT		 528 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
529 es->s_last_error_block = cpu_to_le64(block);
530 save_error_info(inode->i_sb, function, line);
273df556 531 va_start(args, fmt);
f7c21177
TT		 532 vaf.fmt = fmt;
533 vaf.va = &args;
c398eda0 534 if (block)
d9ee81da
JP		 535 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
536 "inode #%lu: block %llu: comm %s: %pV\n",
537 inode->i_sb->s_id, function, line, inode->i_ino,
538 block, current->comm, &vaf);
539 else
540 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
541 "inode #%lu: comm %s: %pV\n",
542 inode->i_sb->s_id, function, line, inode->i_ino,
543 current->comm, &vaf);
273df556
FM		 544 va_end(args);
545
546 ext4_handle_error(inode->i_sb);
547}
548
c398eda0 549void ext4_error_file(struct file *file, const char *function,
f7c21177
TT		 550 unsigned int line, ext4_fsblk_t block,
551 const char *fmt, ...)
273df556
FM		 552{
553 va_list args;
f7c21177 554 struct va_format vaf;
1c13d5c0 555 struct ext4_super_block *es;
273df556
FM		 556 struct inode *inode = file->f_dentry->d_inode;
557 char pathname[80], *path;
558
1c13d5c0
TT		 559 es = EXT4_SB(inode->i_sb)->s_es;
560 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
561 save_error_info(inode->i_sb, function, line);
273df556 562 path = d_path(&(file->f_path), pathname, sizeof(pathname));
f9a62d09 563 if (IS_ERR(path))
273df556 564 path = "(unknown)";
f7c21177
TT		 565 va_start(args, fmt);
566 vaf.fmt = fmt;
567 vaf.va = &args;
d9ee81da
JP		 568 if (block)
569 printk(KERN_CRIT
570 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
571 "block %llu: comm %s: path %s: %pV\n",
572 inode->i_sb->s_id, function, line, inode->i_ino,
573 block, current->comm, path, &vaf);
574 else
575 printk(KERN_CRIT
576 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
577 "comm %s: path %s: %pV\n",
578 inode->i_sb->s_id, function, line, inode->i_ino,
579 current->comm, path, &vaf);
273df556
FM		 580 va_end(args);
581
582 ext4_handle_error(inode->i_sb);
583}
584
2b2d6d01 585static const char *ext4_decode_error(struct super_block *sb, int errno,
ac27a0ec
DK		 586 char nbuf[16])
587{
588 char *errstr = NULL;
589
590 switch (errno) {
591 case -EIO:
592 errstr = "IO failure";
593 break;
594 case -ENOMEM:
595 errstr = "Out of memory";
596 break;
597 case -EROFS:
78f1ddbb
TT		 598 if (!sb || (EXT4_SB(sb)->s_journal &&
599 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
ac27a0ec
DK		 600 errstr = "Journal has aborted";
601 else
602 errstr = "Readonly filesystem";
603 break;
604 default:
605 /* If the caller passed in an extra buffer for unknown
606 * errors, textualise them now. Else we just return
607 * NULL. */
608 if (nbuf) {
609 /* Check for truncated error codes... */
610 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
611 errstr = nbuf;
612 }
613 break;
614 }
615
616 return errstr;
617}
618
617ba13b 619/* __ext4_std_error decodes expected errors from journaling functions
ac27a0ec
DK		 620 * automatically and invokes the appropriate error response. */
621
c398eda0
TT		 622void __ext4_std_error(struct super_block *sb, const char *function,
623 unsigned int line, int errno)
ac27a0ec
DK		 624{
625 char nbuf[16];
626 const char *errstr;
627
628 /* Special case: if the error is EROFS, and we're not already
629 * inside a transaction, then there's really no point in logging
630 * an error. */
631 if (errno == -EROFS && journal_current_handle() == NULL &&
632 (sb->s_flags & MS_RDONLY))
633 return;
634
617ba13b 635 errstr = ext4_decode_error(sb, errno, nbuf);
c398eda0
TT		 636 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
637 sb->s_id, function, line, errstr);
1c13d5c0 638 save_error_info(sb, function, line);
ac27a0ec 639
617ba13b 640 ext4_handle_error(sb);
ac27a0ec
DK		 641}
642
643/*
617ba13b 644 * ext4_abort is a much stronger failure handler than ext4_error. The
ac27a0ec
DK		 645 * abort function may be used to deal with unrecoverable failures such
646 * as journal IO errors or ENOMEM at a critical moment in log management.
647 *
648 * We unconditionally force the filesystem into an ABORT|READONLY state,
649 * unless the error response on the fs has been set to panic in which
650 * case we take the easy way out and panic immediately.
651 */
652
c67d859e 653void __ext4_abort(struct super_block *sb, const char *function,
c398eda0 654 unsigned int line, const char *fmt, ...)
ac27a0ec
DK		 655{
656 va_list args;
657
1c13d5c0 658 save_error_info(sb, function, line);
ac27a0ec 659 va_start(args, fmt);
c398eda0
TT		 660 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
661 function, line);
ac27a0ec
DK		 662 vprintk(fmt, args);
663 printk("\n");
664 va_end(args);
665
1c13d5c0
TT		 666 if ((sb->s_flags & MS_RDONLY) == 0) {
667 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
668 sb->s_flags |= MS_RDONLY;
669 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
670 if (EXT4_SB(sb)->s_journal)
671 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
672 save_error_info(sb, function, line);
673 }
ac27a0ec 674 if (test_opt(sb, ERRORS_PANIC))
617ba13b 675 panic("EXT4-fs panic from previous error\n");
ac27a0ec
DK		 676}
677
0ff2ea7d 678void ext4_msg(struct super_block *sb, const char *prefix, const char *fmt, ...)
b31e1552 679{
0ff2ea7d 680 struct va_format vaf;
b31e1552
ES		 681 va_list args;
682
683 va_start(args, fmt);
0ff2ea7d
JP		 684 vaf.fmt = fmt;
685 vaf.va = &args;
686 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
b31e1552
ES		 687 va_end(args);
688}
689
12062ddd 690void __ext4_warning(struct super_block *sb, const char *function,
c398eda0 691 unsigned int line, const char *fmt, ...)
ac27a0ec 692{
0ff2ea7d 693 struct va_format vaf;
ac27a0ec
DK		 694 va_list args;
695
696 va_start(args, fmt);
0ff2ea7d
JP		 697 vaf.fmt = fmt;
698 vaf.va = &args;
699 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
700 sb->s_id, function, line, &vaf);
ac27a0ec
DK		 701 va_end(args);
702}
703
e29136f8
TT		 704void __ext4_grp_locked_error(const char *function, unsigned int line,
705 struct super_block *sb, ext4_group_t grp,
706 unsigned long ino, ext4_fsblk_t block,
707 const char *fmt, ...)
5d1b1b3f
AK		 708__releases(bitlock)
709__acquires(bitlock)
710{
0ff2ea7d 711 struct va_format vaf;
5d1b1b3f
AK		 712 va_list args;
713 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
714
1c13d5c0
TT		 715 es->s_last_error_ino = cpu_to_le32(ino);
716 es->s_last_error_block = cpu_to_le64(block);
717 __save_error_info(sb, function, line);
0ff2ea7d 718
5d1b1b3f 719 va_start(args, fmt);
0ff2ea7d
JP		 720
721 vaf.fmt = fmt;
722 vaf.va = &args;
21149d61 723 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
e29136f8
TT		 724 sb->s_id, function, line, grp);
725 if (ino)
0ff2ea7d 726 printk(KERN_CONT "inode %lu: ", ino);
e29136f8 727 if (block)
0ff2ea7d
JP		 728 printk(KERN_CONT "block %llu:", (unsigned long long) block);
729 printk(KERN_CONT "%pV\n", &vaf);
5d1b1b3f
AK		 730 va_end(args);
731
732 if (test_opt(sb, ERRORS_CONT)) {
e2d67052 733 ext4_commit_super(sb, 0);
5d1b1b3f
AK		 734 return;
735 }
1c13d5c0 736
5d1b1b3f
AK		 737 ext4_unlock_group(sb, grp);
738 ext4_handle_error(sb);
739 /*
740 * We only get here in the ERRORS_RO case; relocking the group
741 * may be dangerous, but nothing bad will happen since the
742 * filesystem will have already been marked read/only and the
743 * journal has been aborted. We return 1 as a hint to callers
744 * who might what to use the return value from
25985edc 745 * ext4_grp_locked_error() to distinguish between the
5d1b1b3f
AK		 746 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
747 * aggressively from the ext4 function in question, with a
748 * more appropriate error code.
749 */
750 ext4_lock_group(sb, grp);
751 return;
752}
753
617ba13b 754void ext4_update_dynamic_rev(struct super_block *sb)
ac27a0ec 755{
617ba13b 756 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 757
617ba13b 758 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
ac27a0ec
DK		 759 return;
760
12062ddd 761 ext4_warning(sb,
ac27a0ec
DK		 762 "updating to rev %d because of new feature flag, "
763 "running e2fsck is recommended",
617ba13b 764 EXT4_DYNAMIC_REV);
ac27a0ec 765
617ba13b
MC		 766 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
767 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
768 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
ac27a0ec
DK		 769 /* leave es->s_feature_*compat flags alone */
770 /* es->s_uuid will be set by e2fsck if empty */
771
772 /*
773 * The rest of the superblock fields should be zero, and if not it
774 * means they are likely already in use, so leave them alone. We
775 * can leave it up to e2fsck to clean up any inconsistencies there.
776 */
777}
778
779/*
780 * Open the external journal device
781 */
b31e1552 782static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
ac27a0ec
DK		 783{
784 struct block_device *bdev;
785 char b[BDEVNAME_SIZE];
786
d4d77629 787 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
ac27a0ec
DK		 788 if (IS_ERR(bdev))
789 goto fail;
790 return bdev;
791
792fail:
b31e1552 793 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
ac27a0ec
DK		 794 __bdevname(dev, b), PTR_ERR(bdev));
795 return NULL;
796}
797
798/*
799 * Release the journal device
800 */
617ba13b 801static int ext4_blkdev_put(struct block_device *bdev)
ac27a0ec 802{
e525fd89 803 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
ac27a0ec
DK		 804}
805
617ba13b 806static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
ac27a0ec
DK		 807{
808 struct block_device *bdev;
809 int ret = -ENODEV;
810
811 bdev = sbi->journal_bdev;
812 if (bdev) {
617ba13b 813 ret = ext4_blkdev_put(bdev);
ac27a0ec
DK		 814 sbi->journal_bdev = NULL;
815 }
816 return ret;
817}
818
819static inline struct inode *orphan_list_entry(struct list_head *l)
820{
617ba13b 821 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
ac27a0ec
DK		 822}
823
617ba13b 824static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
ac27a0ec
DK		 825{
826 struct list_head *l;
827
b31e1552
ES		 828 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
829 le32_to_cpu(sbi->s_es->s_last_orphan));
ac27a0ec
DK		 830
831 printk(KERN_ERR "sb_info orphan list:\n");
832 list_for_each(l, &sbi->s_orphan) {
833 struct inode *inode = orphan_list_entry(l);
834 printk(KERN_ERR " "
835 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
836 inode->i_sb->s_id, inode->i_ino, inode,
837 inode->i_mode, inode->i_nlink,
838 NEXT_ORPHAN(inode));
839 }
840}
841
2b2d6d01 842static void ext4_put_super(struct super_block *sb)
ac27a0ec 843{
617ba13b
MC		 844 struct ext4_sb_info *sbi = EXT4_SB(sb);
845 struct ext4_super_block *es = sbi->s_es;
ef2cabf7 846 int i, err;
ac27a0ec 847
857ac889 848 ext4_unregister_li_request(sb);
e0ccfd95
CH		 849 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
850
4c0425ff
MC		 851 flush_workqueue(sbi->dio_unwritten_wq);
852 destroy_workqueue(sbi->dio_unwritten_wq);
853
0390131b
FM		 854 if (sbi->s_journal) {
855 err = jbd2_journal_destroy(sbi->s_journal);
856 sbi->s_journal = NULL;
857 if (err < 0)
c67d859e 858 ext4_abort(sb, "Couldn't clean up the journal");
0390131b 859 }
d4edac31 860
a1c6c569 861 del_timer(&sbi->s_err_report);
d4edac31
JB		 862 ext4_release_system_zone(sb);
863 ext4_mb_release(sb);
864 ext4_ext_release(sb);
865 ext4_xattr_put_super(sb);
866
ac27a0ec 867 if (!(sb->s_flags & MS_RDONLY)) {
617ba13b 868 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 869 es->s_state = cpu_to_le16(sbi->s_mount_state);
ac27a0ec 870 }
58c5873a 871 if (!(sb->s_flags & MS_RDONLY))
a8e25a83
AB		 872 ext4_commit_super(sb, 1);
873
240799cd 874 if (sbi->s_proc) {
66acdcf4 875 remove_proc_entry("options", sbi->s_proc);
9f6200bb 876 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 877 }
3197ebdb 878 kobject_del(&sbi->s_kobj);
ac27a0ec
DK		 879
880 for (i = 0; i < sbi->s_gdb_count; i++)
881 brelse(sbi->s_group_desc[i]);
f18a5f21 882 ext4_kvfree(sbi->s_group_desc);
9933fc0a 883 ext4_kvfree(sbi->s_flex_groups);
57042651 884 percpu_counter_destroy(&sbi->s_freeclusters_counter);
ac27a0ec
DK		 885 percpu_counter_destroy(&sbi->s_freeinodes_counter);
886 percpu_counter_destroy(&sbi->s_dirs_counter);
57042651 887 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
ac27a0ec
DK		 888 brelse(sbi->s_sbh);
889#ifdef CONFIG_QUOTA
890 for (i = 0; i < MAXQUOTAS; i++)
891 kfree(sbi->s_qf_names[i]);
892#endif
893
894 /* Debugging code just in case the in-memory inode orphan list
895 * isn't empty. The on-disk one can be non-empty if we've
896 * detected an error and taken the fs readonly, but the
897 * in-memory list had better be clean by this point. */
898 if (!list_empty(&sbi->s_orphan))
899 dump_orphan_list(sb, sbi);
900 J_ASSERT(list_empty(&sbi->s_orphan));
901
f98393a6 902 invalidate_bdev(sb->s_bdev);
ac27a0ec
DK		 903 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
904 /*
905 * Invalidate the journal device's buffers. We don't want them
906 * floating about in memory - the physical journal device may
907 * hotswapped, and it breaks the `ro-after' testing code.
908 */
909 sync_blockdev(sbi->journal_bdev);
f98393a6 910 invalidate_bdev(sbi->journal_bdev);
617ba13b 911 ext4_blkdev_remove(sbi);
ac27a0ec 912 }
c5e06d10
JL		 913 if (sbi->s_mmp_tsk)
914 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec 915 sb->s_fs_info = NULL;
3197ebdb
TT		 916 /*
917 * Now that we are completely done shutting down the
918 * superblock, we need to actually destroy the kobject.
919 */
3197ebdb
TT		 920 kobject_put(&sbi->s_kobj);
921 wait_for_completion(&sbi->s_kobj_unregister);
0441984a
DW		 922 if (sbi->s_chksum_driver)
923 crypto_free_shash(sbi->s_chksum_driver);
705895b6 924 kfree(sbi->s_blockgroup_lock);
ac27a0ec 925 kfree(sbi);
ac27a0ec
DK		 926}
927
e18b890b 928static struct kmem_cache *ext4_inode_cachep;
ac27a0ec
DK		 929
930/*
931 * Called inside transaction, so use GFP_NOFS
932 */
617ba13b 933static struct inode *ext4_alloc_inode(struct super_block *sb)
ac27a0ec 934{
617ba13b 935 struct ext4_inode_info *ei;
ac27a0ec 936
e6b4f8da 937 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
ac27a0ec
DK		 938 if (!ei)
939 return NULL;
0b8e58a1 940
ac27a0ec 941 ei->vfs_inode.i_version = 1;
a86c6181 942 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
c9de560d
AT		 943 INIT_LIST_HEAD(&ei->i_prealloc_list);
944 spin_lock_init(&ei->i_prealloc_lock);
9a26b661
ZL		 945 ext4_es_init_tree(&ei->i_es_tree);
946 rwlock_init(&ei->i_es_lock);
d2a17637
MC		 947 ei->i_reserved_data_blocks = 0;
948 ei->i_reserved_meta_blocks = 0;
949 ei->i_allocated_meta_blocks = 0;
9d0be502 950 ei->i_da_metadata_calc_len = 0;
7e731bc9 951 ei->i_da_metadata_calc_last_lblock = 0;
d2a17637 952 spin_lock_init(&(ei->i_block_reservation_lock));
a9e7f447
DM		 953#ifdef CONFIG_QUOTA
954 ei->i_reserved_quota = 0;
955#endif
8aefcd55 956 ei->jinode = NULL;
c7064ef1 957 INIT_LIST_HEAD(&ei->i_completed_io_list);
744692dc 958 spin_lock_init(&ei->i_completed_io_lock);
b436b9be
JK		 959 ei->i_sync_tid = 0;
960 ei->i_datasync_tid = 0;
f7ad6d2e 961 atomic_set(&ei->i_ioend_count, 0);
e27f41e1 962 atomic_set(&ei->i_unwritten, 0);
84c17543 963 INIT_WORK(&ei->i_unwritten_work, ext4_end_io_work);
0b8e58a1 964
ac27a0ec
DK		 965 return &ei->vfs_inode;
966}
967
7ff9c073
TT		 968static int ext4_drop_inode(struct inode *inode)
969{
970 int drop = generic_drop_inode(inode);
971
972 trace_ext4_drop_inode(inode, drop);
973 return drop;
974}
975
fa0d7e3d
NP		 976static void ext4_i_callback(struct rcu_head *head)
977{
978 struct inode *inode = container_of(head, struct inode, i_rcu);
fa0d7e3d
NP		 979 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
980}
981
617ba13b 982static void ext4_destroy_inode(struct inode *inode)
ac27a0ec 983{
9f7dd93d 984 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
b31e1552
ES		 985 ext4_msg(inode->i_sb, KERN_ERR,
986 "Inode %lu (%p): orphan list check failed!",
987 inode->i_ino, EXT4_I(inode));
9f7dd93d
VA		 988 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
989 EXT4_I(inode), sizeof(struct ext4_inode_info),
990 true);
991 dump_stack();
992 }
fa0d7e3d 993 call_rcu(&inode->i_rcu, ext4_i_callback);
ac27a0ec
DK		 994}
995
51cc5068 996static void init_once(void *foo)
ac27a0ec 997{
617ba13b 998 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
ac27a0ec 999
a35afb83 1000 INIT_LIST_HEAD(&ei->i_orphan);
a35afb83 1001 init_rwsem(&ei->xattr_sem);
0e855ac8 1002 init_rwsem(&ei->i_data_sem);
a35afb83 1003 inode_init_once(&ei->vfs_inode);
ac27a0ec
DK		 1004}
1005
1006static int init_inodecache(void)
1007{
617ba13b
MC		 1008 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
1009 sizeof(struct ext4_inode_info),
ac27a0ec
DK		 1010 0, (SLAB_RECLAIM_ACCOUNT|
1011 SLAB_MEM_SPREAD),
20c2df83 1012 init_once);
617ba13b 1013 if (ext4_inode_cachep == NULL)
ac27a0ec
DK		 1014 return -ENOMEM;
1015 return 0;
1016}
1017
1018static void destroy_inodecache(void)
1019{
8c0a8537
KS		 1020 /*
1021 * Make sure all delayed rcu free inodes are flushed before we
1022 * destroy cache.
1023 */
1024 rcu_barrier();
617ba13b 1025 kmem_cache_destroy(ext4_inode_cachep);
ac27a0ec
DK		 1026}
1027
0930fcc1 1028void ext4_clear_inode(struct inode *inode)
ac27a0ec 1029{
0930fcc1 1030 invalidate_inode_buffers(inode);
dbd5768f 1031 clear_inode(inode);
9f754758 1032 dquot_drop(inode);
c2ea3fde 1033 ext4_discard_preallocations(inode);
51865fda 1034 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
8aefcd55
TT		 1035 if (EXT4_I(inode)->jinode) {
1036 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1037 EXT4_I(inode)->jinode);
1038 jbd2_free_inode(EXT4_I(inode)->jinode);
1039 EXT4_I(inode)->jinode = NULL;
1040 }
ac27a0ec
DK		 1041}
1042
1b961ac0 1043static struct inode *ext4_nfs_get_inode(struct super_block *sb,
0b8e58a1 1044 u64 ino, u32 generation)
ac27a0ec 1045{
ac27a0ec 1046 struct inode *inode;
ac27a0ec 1047
617ba13b 1048 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
ac27a0ec 1049 return ERR_PTR(-ESTALE);
617ba13b 1050 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
ac27a0ec
DK		 1051 return ERR_PTR(-ESTALE);
1052
1053 /* iget isn't really right if the inode is currently unallocated!!
1054 *
617ba13b 1055 * ext4_read_inode will return a bad_inode if the inode had been
ac27a0ec
DK		 1056 * deleted, so we should be safe.
1057 *
1058 * Currently we don't know the generation for parent directory, so
1059 * a generation of 0 means "accept any"
1060 */
1d1fe1ee
DH		 1061 inode = ext4_iget(sb, ino);
1062 if (IS_ERR(inode))
1063 return ERR_CAST(inode);
1064 if (generation && inode->i_generation != generation) {
ac27a0ec
DK		 1065 iput(inode);
1066 return ERR_PTR(-ESTALE);
1067 }
1b961ac0
CH		 1068
1069 return inode;
1070}
1071
1072static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
0b8e58a1 1073 int fh_len, int fh_type)
1b961ac0
CH		 1074{
1075 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1076 ext4_nfs_get_inode);
1077}
1078
1079static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
0b8e58a1 1080 int fh_len, int fh_type)
1b961ac0
CH		 1081{
1082 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1083 ext4_nfs_get_inode);
ac27a0ec
DK		 1084}
1085
c39a7f84
TO		 1086/*
1087 * Try to release metadata pages (indirect blocks, directories) which are
1088 * mapped via the block device. Since these pages could have journal heads
1089 * which would prevent try_to_free_buffers() from freeing them, we must use
1090 * jbd2 layer's try_to_free_buffers() function to release them.
1091 */
0b8e58a1
AD		 1092static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1093 gfp_t wait)
c39a7f84
TO		 1094{
1095 journal_t *journal = EXT4_SB(sb)->s_journal;
1096
1097 WARN_ON(PageChecked(page));
1098 if (!page_has_buffers(page))
1099 return 0;
1100 if (journal)
1101 return jbd2_journal_try_to_free_buffers(journal, page,
1102 wait & ~__GFP_WAIT);
1103 return try_to_free_buffers(page);
1104}
1105
ac27a0ec 1106#ifdef CONFIG_QUOTA
af5bc92d 1107#define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
2b2d6d01 1108#define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
ac27a0ec 1109
617ba13b
MC		 1110static int ext4_write_dquot(struct dquot *dquot);
1111static int ext4_acquire_dquot(struct dquot *dquot);
1112static int ext4_release_dquot(struct dquot *dquot);
1113static int ext4_mark_dquot_dirty(struct dquot *dquot);
1114static int ext4_write_info(struct super_block *sb, int type);
6f28e087 1115static int ext4_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 1116 struct path *path);
7c319d32
AK		 1117static int ext4_quota_on_sysfile(struct super_block *sb, int type,
1118 int format_id);
ca0e05e4 1119static int ext4_quota_off(struct super_block *sb, int type);
7c319d32 1120static int ext4_quota_off_sysfile(struct super_block *sb, int type);
617ba13b
MC		 1121static int ext4_quota_on_mount(struct super_block *sb, int type);
1122static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 1123 size_t len, loff_t off);
617ba13b 1124static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec 1125 const char *data, size_t len, loff_t off);
7c319d32
AK		 1126static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1127 unsigned int flags);
1128static int ext4_enable_quotas(struct super_block *sb);
ac27a0ec 1129
61e225dc 1130static const struct dquot_operations ext4_quota_operations = {
60e58e0f 1131 .get_reserved_space = ext4_get_reserved_space,
617ba13b
MC		 1132 .write_dquot = ext4_write_dquot,
1133 .acquire_dquot = ext4_acquire_dquot,
1134 .release_dquot = ext4_release_dquot,
1135 .mark_dirty = ext4_mark_dquot_dirty,
a5b5ee32
JK		 1136 .write_info = ext4_write_info,
1137 .alloc_dquot = dquot_alloc,
1138 .destroy_dquot = dquot_destroy,
ac27a0ec
DK		 1139};
1140
0d54b217 1141static const struct quotactl_ops ext4_qctl_operations = {
617ba13b 1142 .quota_on = ext4_quota_on,
ca0e05e4 1143 .quota_off = ext4_quota_off,
287a8095
CH		 1144 .quota_sync = dquot_quota_sync,
1145 .get_info = dquot_get_dqinfo,
1146 .set_info = dquot_set_dqinfo,
1147 .get_dqblk = dquot_get_dqblk,
1148 .set_dqblk = dquot_set_dqblk
ac27a0ec 1149};
7c319d32
AK		 1150
1151static const struct quotactl_ops ext4_qctl_sysfile_operations = {
1152 .quota_on_meta = ext4_quota_on_sysfile,
1153 .quota_off = ext4_quota_off_sysfile,
1154 .quota_sync = dquot_quota_sync,
1155 .get_info = dquot_get_dqinfo,
1156 .set_info = dquot_set_dqinfo,
1157 .get_dqblk = dquot_get_dqblk,
1158 .set_dqblk = dquot_set_dqblk
1159};
ac27a0ec
DK		 1160#endif
1161
ee9b6d61 1162static const struct super_operations ext4_sops = {
617ba13b
MC		 1163 .alloc_inode = ext4_alloc_inode,
1164 .destroy_inode = ext4_destroy_inode,
617ba13b
MC		 1165 .write_inode = ext4_write_inode,
1166 .dirty_inode = ext4_dirty_inode,
7ff9c073 1167 .drop_inode = ext4_drop_inode,
0930fcc1 1168 .evict_inode = ext4_evict_inode,
617ba13b 1169 .put_super = ext4_put_super,
617ba13b 1170 .sync_fs = ext4_sync_fs,
c4be0c1d
TS		 1171 .freeze_fs = ext4_freeze,
1172 .unfreeze_fs = ext4_unfreeze,
617ba13b
MC		 1173 .statfs = ext4_statfs,
1174 .remount_fs = ext4_remount,
617ba13b 1175 .show_options = ext4_show_options,
ac27a0ec 1176#ifdef CONFIG_QUOTA
617ba13b
MC		 1177 .quota_read = ext4_quota_read,
1178 .quota_write = ext4_quota_write,
ac27a0ec 1179#endif
c39a7f84 1180 .bdev_try_to_free_page = bdev_try_to_free_page,
ac27a0ec
DK		 1181};
1182
9ca92389
TT		 1183static const struct super_operations ext4_nojournal_sops = {
1184 .alloc_inode = ext4_alloc_inode,
1185 .destroy_inode = ext4_destroy_inode,
1186 .write_inode = ext4_write_inode,
1187 .dirty_inode = ext4_dirty_inode,
7ff9c073 1188 .drop_inode = ext4_drop_inode,
0930fcc1 1189 .evict_inode = ext4_evict_inode,
9ca92389
TT		 1190 .put_super = ext4_put_super,
1191 .statfs = ext4_statfs,
1192 .remount_fs = ext4_remount,
9ca92389
TT		 1193 .show_options = ext4_show_options,
1194#ifdef CONFIG_QUOTA
1195 .quota_read = ext4_quota_read,
1196 .quota_write = ext4_quota_write,
1197#endif
1198 .bdev_try_to_free_page = bdev_try_to_free_page,
1199};
1200
39655164 1201static const struct export_operations ext4_export_ops = {
1b961ac0
CH		 1202 .fh_to_dentry = ext4_fh_to_dentry,
1203 .fh_to_parent = ext4_fh_to_parent,
617ba13b 1204 .get_parent = ext4_get_parent,
ac27a0ec
DK		 1205};
1206
1207enum {
1208 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1209 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
72578c33 1210 Opt_nouid32, Opt_debug, Opt_removed,
ac27a0ec 1211 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
72578c33 1212 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
30773840 1213 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
43e625d8 1214 Opt_journal_dev, Opt_journal_checksum, Opt_journal_async_commit,
ac27a0ec 1215 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
296c355c 1216 Opt_data_err_abort, Opt_data_err_ignore,
ac27a0ec 1217 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
5a20bdfc 1218 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
ee4a3fcd 1219 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
661aa520 1220 Opt_usrquota, Opt_grpquota, Opt_i_version,
1449032b
TT		 1221 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1222 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
5328e635 1223 Opt_inode_readahead_blks, Opt_journal_ioprio,
744692dc 1224 Opt_dioread_nolock, Opt_dioread_lock,
fc6cb1cd 1225 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
df981d03 1226 Opt_max_dir_size_kb,
ac27a0ec
DK		 1227};
1228
a447c093 1229static const match_table_t tokens = {
ac27a0ec
DK		 1230 {Opt_bsd_df, "bsddf"},
1231 {Opt_minix_df, "minixdf"},
1232 {Opt_grpid, "grpid"},
1233 {Opt_grpid, "bsdgroups"},
1234 {Opt_nogrpid, "nogrpid"},
1235 {Opt_nogrpid, "sysvgroups"},
1236 {Opt_resgid, "resgid=%u"},
1237 {Opt_resuid, "resuid=%u"},
1238 {Opt_sb, "sb=%u"},
1239 {Opt_err_cont, "errors=continue"},
1240 {Opt_err_panic, "errors=panic"},
1241 {Opt_err_ro, "errors=remount-ro"},
1242 {Opt_nouid32, "nouid32"},
ac27a0ec 1243 {Opt_debug, "debug"},
72578c33
TT		 1244 {Opt_removed, "oldalloc"},
1245 {Opt_removed, "orlov"},
ac27a0ec
DK		 1246 {Opt_user_xattr, "user_xattr"},
1247 {Opt_nouser_xattr, "nouser_xattr"},
1248 {Opt_acl, "acl"},
1249 {Opt_noacl, "noacl"},
e3bb52ae 1250 {Opt_noload, "norecovery"},
5a916be1 1251 {Opt_noload, "noload"},
72578c33
TT		 1252 {Opt_removed, "nobh"},
1253 {Opt_removed, "bh"},
ac27a0ec 1254 {Opt_commit, "commit=%u"},
30773840
TT		 1255 {Opt_min_batch_time, "min_batch_time=%u"},
1256 {Opt_max_batch_time, "max_batch_time=%u"},
ac27a0ec 1257 {Opt_journal_dev, "journal_dev=%u"},
818d276c
GS		 1258 {Opt_journal_checksum, "journal_checksum"},
1259 {Opt_journal_async_commit, "journal_async_commit"},
ac27a0ec
DK		 1260 {Opt_abort, "abort"},
1261 {Opt_data_journal, "data=journal"},
1262 {Opt_data_ordered, "data=ordered"},
1263 {Opt_data_writeback, "data=writeback"},
5bf5683a
HK		 1264 {Opt_data_err_abort, "data_err=abort"},
1265 {Opt_data_err_ignore, "data_err=ignore"},
ac27a0ec
DK		 1266 {Opt_offusrjquota, "usrjquota="},
1267 {Opt_usrjquota, "usrjquota=%s"},
1268 {Opt_offgrpjquota, "grpjquota="},
1269 {Opt_grpjquota, "grpjquota=%s"},
1270 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1271 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
5a20bdfc 1272 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
ac27a0ec
DK		 1273 {Opt_grpquota, "grpquota"},
1274 {Opt_noquota, "noquota"},
1275 {Opt_quota, "quota"},
1276 {Opt_usrquota, "usrquota"},
1277 {Opt_barrier, "barrier=%u"},
06705bff
TT		 1278 {Opt_barrier, "barrier"},
1279 {Opt_nobarrier, "nobarrier"},
25ec56b5 1280 {Opt_i_version, "i_version"},
c9de560d 1281 {Opt_stripe, "stripe=%u"},
64769240 1282 {Opt_delalloc, "delalloc"},
dd919b98 1283 {Opt_nodelalloc, "nodelalloc"},
36ade451
JK		 1284 {Opt_removed, "mblk_io_submit"},
1285 {Opt_removed, "nomblk_io_submit"},
6fd058f7
TT		 1286 {Opt_block_validity, "block_validity"},
1287 {Opt_noblock_validity, "noblock_validity"},
240799cd 1288 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
b3881f74 1289 {Opt_journal_ioprio, "journal_ioprio=%u"},
afd4672d 1290 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
06705bff
TT		 1291 {Opt_auto_da_alloc, "auto_da_alloc"},
1292 {Opt_noauto_da_alloc, "noauto_da_alloc"},
744692dc
JZ		 1293 {Opt_dioread_nolock, "dioread_nolock"},
1294 {Opt_dioread_lock, "dioread_lock"},
5328e635
ES		 1295 {Opt_discard, "discard"},
1296 {Opt_nodiscard, "nodiscard"},
fc6cb1cd
TT		 1297 {Opt_init_itable, "init_itable=%u"},
1298 {Opt_init_itable, "init_itable"},
1299 {Opt_noinit_itable, "noinit_itable"},
df981d03 1300 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
c7198b9c
TT		 1301 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1302 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1303 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1304 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1305 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
f3f12faa 1306 {Opt_err, NULL},
ac27a0ec
DK		 1307};
1308
617ba13b 1309static ext4_fsblk_t get_sb_block(void **data)
ac27a0ec 1310{
617ba13b 1311 ext4_fsblk_t sb_block;
ac27a0ec
DK		 1312 char *options = (char *) *data;
1313
1314 if (!options || strncmp(options, "sb=", 3) != 0)
1315 return 1; /* Default location */
0b8e58a1 1316
ac27a0ec 1317 options += 3;
0b8e58a1 1318 /* TODO: use simple_strtoll with >32bit ext4 */
ac27a0ec
DK		 1319 sb_block = simple_strtoul(options, &options, 0);
1320 if (*options && *options != ',') {
4776004f 1321 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
ac27a0ec
DK		 1322 (char *) *data);
1323 return 1;
1324 }
1325 if (*options == ',')
1326 options++;
1327 *data = (void *) options;
0b8e58a1 1328
ac27a0ec
DK		 1329 return sb_block;
1330}
1331
b3881f74 1332#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
437ca0fd
DM		 1333static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1334 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
b3881f74 1335
56c50f11
DM		 1336#ifdef CONFIG_QUOTA
1337static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1338{
1339 struct ext4_sb_info *sbi = EXT4_SB(sb);
1340 char *qname;
03dafb5f 1341 int ret = -1;
56c50f11
DM		 1342
1343 if (sb_any_quota_loaded(sb) &&
1344 !sbi->s_qf_names[qtype]) {
1345 ext4_msg(sb, KERN_ERR,
1346 "Cannot change journaled "
1347 "quota options when quota turned on");
57f73c2c 1348 return -1;
56c50f11
DM		 1349 }
1350 qname = match_strdup(args);
1351 if (!qname) {
1352 ext4_msg(sb, KERN_ERR,
1353 "Not enough memory for storing quotafile name");
57f73c2c 1354 return -1;
56c50f11 1355 }
03dafb5f
CG		 1356 if (sbi->s_qf_names[qtype]) {
1357 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1358 ret = 1;
1359 else
1360 ext4_msg(sb, KERN_ERR,
1361 "%s quota file already specified",
1362 QTYPE2NAME(qtype));
1363 goto errout;
56c50f11 1364 }
03dafb5f 1365 if (strchr(qname, '/')) {
56c50f11
DM		 1366 ext4_msg(sb, KERN_ERR,
1367 "quotafile must be on filesystem root");
03dafb5f 1368 goto errout;
56c50f11 1369 }
03dafb5f 1370 sbi->s_qf_names[qtype] = qname;
fd8c37ec 1371 set_opt(sb, QUOTA);
56c50f11 1372 return 1;
03dafb5f
CG		 1373errout:
1374 kfree(qname);
1375 return ret;
56c50f11
DM		 1376}
1377
1378static int clear_qf_name(struct super_block *sb, int qtype)
1379{
1380
1381 struct ext4_sb_info *sbi = EXT4_SB(sb);
1382
1383 if (sb_any_quota_loaded(sb) &&
1384 sbi->s_qf_names[qtype]) {
1385 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1386 " when quota turned on");
57f73c2c 1387 return -1;
56c50f11 1388 }
03dafb5f 1389 kfree(sbi->s_qf_names[qtype]);
56c50f11
DM		 1390 sbi->s_qf_names[qtype] = NULL;
1391 return 1;
1392}
1393#endif
1394
26092bf5
TT		 1395#define MOPT_SET 0x0001
1396#define MOPT_CLEAR 0x0002
1397#define MOPT_NOSUPPORT 0x0004
1398#define MOPT_EXPLICIT 0x0008
1399#define MOPT_CLEAR_ERR 0x0010
1400#define MOPT_GTE0 0x0020
ac27a0ec 1401#ifdef CONFIG_QUOTA
26092bf5
TT		 1402#define MOPT_Q 0
1403#define MOPT_QFMT 0x0040
1404#else
1405#define MOPT_Q MOPT_NOSUPPORT
1406#define MOPT_QFMT MOPT_NOSUPPORT
ac27a0ec 1407#endif
26092bf5 1408#define MOPT_DATAJ 0x0080
8dc0aa8c
TT		 1409#define MOPT_NO_EXT2 0x0100
1410#define MOPT_NO_EXT3 0x0200
1411#define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
26092bf5
TT		 1412
1413static const struct mount_opts {
1414 int token;
1415 int mount_opt;
1416 int flags;
1417} ext4_mount_opts[] = {
1418 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1419 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1420 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1421 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
26092bf5
TT		 1422 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1423 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
8dc0aa8c
TT		 1424 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1425 MOPT_EXT4_ONLY | MOPT_SET},
1426 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1427 MOPT_EXT4_ONLY | MOPT_CLEAR},
26092bf5
TT		 1428 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1429 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
8dc0aa8c
TT		 1430 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1431 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1432 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1433 MOPT_EXT4_ONLY | MOPT_CLEAR | MOPT_EXPLICIT},
1434 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1435 MOPT_EXT4_ONLY | MOPT_SET},
26092bf5 1436 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
8dc0aa8c
TT		 1437 EXT4_MOUNT_JOURNAL_CHECKSUM),
1438 MOPT_EXT4_ONLY | MOPT_SET},
1439 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
26092bf5
TT		 1440 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1441 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1442 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
8dc0aa8c
TT		 1443 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1444 MOPT_NO_EXT2 | MOPT_SET},
1445 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1446 MOPT_NO_EXT2 | MOPT_CLEAR},
26092bf5
TT		 1447 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1448 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1449 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1450 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1451 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1452 {Opt_commit, 0, MOPT_GTE0},
1453 {Opt_max_batch_time, 0, MOPT_GTE0},
1454 {Opt_min_batch_time, 0, MOPT_GTE0},
1455 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1456 {Opt_init_itable, 0, MOPT_GTE0},
1457 {Opt_stripe, 0, MOPT_GTE0},
0efb3b23
JK		 1458 {Opt_resuid, 0, MOPT_GTE0},
1459 {Opt_resgid, 0, MOPT_GTE0},
1460 {Opt_journal_dev, 0, MOPT_GTE0},
1461 {Opt_journal_ioprio, 0, MOPT_GTE0},
8dc0aa8c
TT		 1462 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1463 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1464 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1465 MOPT_NO_EXT2 | MOPT_DATAJ},
26092bf5
TT		 1466 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1467 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
03010a33 1468#ifdef CONFIG_EXT4_FS_POSIX_ACL
26092bf5
TT		 1469 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1470 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
ac27a0ec 1471#else
26092bf5
TT		 1472 {Opt_acl, 0, MOPT_NOSUPPORT},
1473 {Opt_noacl, 0, MOPT_NOSUPPORT},
ac27a0ec 1474#endif
26092bf5
TT		 1475 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1476 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1477 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1478 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1479 MOPT_SET | MOPT_Q},
1480 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1481 MOPT_SET | MOPT_Q},
1482 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1483 EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1484 {Opt_usrjquota, 0, MOPT_Q},
1485 {Opt_grpjquota, 0, MOPT_Q},
1486 {Opt_offusrjquota, 0, MOPT_Q},
1487 {Opt_offgrpjquota, 0, MOPT_Q},
1488 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1489 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1490 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
df981d03 1491 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
26092bf5
TT		 1492 {Opt_err, 0, 0}
1493};
1494
1495static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1496 substring_t *args, unsigned long *journal_devnum,
1497 unsigned int *journal_ioprio, int is_remount)
1498{
1499 struct ext4_sb_info *sbi = EXT4_SB(sb);
1500 const struct mount_opts *m;
08cefc7a
EB		 1501 kuid_t uid;
1502 kgid_t gid;
26092bf5
TT		 1503 int arg = 0;
1504
57f73c2c
TT		 1505#ifdef CONFIG_QUOTA
1506 if (token == Opt_usrjquota)
1507 return set_qf_name(sb, USRQUOTA, &args[0]);
1508 else if (token == Opt_grpjquota)
1509 return set_qf_name(sb, GRPQUOTA, &args[0]);
1510 else if (token == Opt_offusrjquota)
1511 return clear_qf_name(sb, USRQUOTA);
1512 else if (token == Opt_offgrpjquota)
1513 return clear_qf_name(sb, GRPQUOTA);
1514#endif
26092bf5 1515 switch (token) {
f7048605
TT		 1516 case Opt_noacl:
1517 case Opt_nouser_xattr:
1518 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1519 break;
26092bf5
TT		 1520 case Opt_sb:
1521 return 1; /* handled by get_sb_block() */
1522 case Opt_removed:
5f3633e3 1523 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
26092bf5 1524 return 1;
26092bf5
TT		 1525 case Opt_abort:
1526 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1527 return 1;
1528 case Opt_i_version:
1529 sb->s_flags |= MS_I_VERSION;
1530 return 1;
26092bf5
TT		 1531 }
1532
5f3633e3
JK		 1533 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1534 if (token == m->token)
1535 break;
1536
1537 if (m->token == Opt_err) {
1538 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1539 "or missing value", opt);
1540 return -1;
1541 }
1542
8dc0aa8c
TT		 1543 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1544 ext4_msg(sb, KERN_ERR,
1545 "Mount option \"%s\" incompatible with ext2", opt);
1546 return -1;
1547 }
1548 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1549 ext4_msg(sb, KERN_ERR,
1550 "Mount option \"%s\" incompatible with ext3", opt);
1551 return -1;
1552 }
1553
5f3633e3
JK		 1554 if (args->from && match_int(args, &arg))
1555 return -1;
1556 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1557 return -1;
1558 if (m->flags & MOPT_EXPLICIT)
1559 set_opt2(sb, EXPLICIT_DELALLOC);
1560 if (m->flags & MOPT_CLEAR_ERR)
1561 clear_opt(sb, ERRORS_MASK);
1562 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1563 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1564 "options when quota turned on");
1565 return -1;
1566 }
1567
1568 if (m->flags & MOPT_NOSUPPORT) {
1569 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1570 } else if (token == Opt_commit) {
1571 if (arg == 0)
1572 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1573 sbi->s_commit_interval = HZ * arg;
1574 } else if (token == Opt_max_batch_time) {
1575 if (arg == 0)
1576 arg = EXT4_DEF_MAX_BATCH_TIME;
1577 sbi->s_max_batch_time = arg;
1578 } else if (token == Opt_min_batch_time) {
1579 sbi->s_min_batch_time = arg;
1580 } else if (token == Opt_inode_readahead_blks) {
e33e60ea
JK		 1581 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1582 ext4_msg(sb, KERN_ERR,
1583 "EXT4-fs: inode_readahead_blks must be "
1584 "0 or a power of 2 smaller than 2^31");
26092bf5 1585 return -1;
5f3633e3
JK		 1586 }
1587 sbi->s_inode_readahead_blks = arg;
1588 } else if (token == Opt_init_itable) {
1589 set_opt(sb, INIT_INODE_TABLE);
1590 if (!args->from)
1591 arg = EXT4_DEF_LI_WAIT_MULT;
1592 sbi->s_li_wait_mult = arg;
1593 } else if (token == Opt_max_dir_size_kb) {
1594 sbi->s_max_dir_size_kb = arg;
1595 } else if (token == Opt_stripe) {
1596 sbi->s_stripe = arg;
1597 } else if (token == Opt_resuid) {
1598 uid = make_kuid(current_user_ns(), arg);
1599 if (!uid_valid(uid)) {
1600 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
26092bf5
TT		 1601 return -1;
1602 }
5f3633e3
JK		 1603 sbi->s_resuid = uid;
1604 } else if (token == Opt_resgid) {
1605 gid = make_kgid(current_user_ns(), arg);
1606 if (!gid_valid(gid)) {
1607 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1608 return -1;
1609 }
1610 sbi->s_resgid = gid;
1611 } else if (token == Opt_journal_dev) {
1612 if (is_remount) {
1613 ext4_msg(sb, KERN_ERR,
1614 "Cannot specify journal on remount");
1615 return -1;
1616 }
1617 *journal_devnum = arg;
1618 } else if (token == Opt_journal_ioprio) {
1619 if (arg > 7) {
1620 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1621 " (must be 0-7)");
1622 return -1;
1623 }
1624 *journal_ioprio =
1625 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1626 } else if (m->flags & MOPT_DATAJ) {
1627 if (is_remount) {
1628 if (!sbi->s_journal)
1629 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1630 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
0efb3b23 1631 ext4_msg(sb, KERN_ERR,
26092bf5 1632 "Cannot change data mode on remount");
26092bf5 1633 return -1;
ac27a0ec 1634 }
26092bf5 1635 } else {
5f3633e3
JK		 1636 clear_opt(sb, DATA_FLAGS);
1637 sbi->s_mount_opt |= m->mount_opt;
ac27a0ec 1638 }
5f3633e3
JK		 1639#ifdef CONFIG_QUOTA
1640 } else if (m->flags & MOPT_QFMT) {
1641 if (sb_any_quota_loaded(sb) &&
1642 sbi->s_jquota_fmt != m->mount_opt) {
1643 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1644 "quota options when quota turned on");
1645 return -1;
1646 }
1647 sbi->s_jquota_fmt = m->mount_opt;
1648#endif
1649 } else {
1650 if (!args->from)
1651 arg = 1;
1652 if (m->flags & MOPT_CLEAR)
1653 arg = !arg;
1654 else if (unlikely(!(m->flags & MOPT_SET))) {
1655 ext4_msg(sb, KERN_WARNING,
1656 "buggy handling of option %s", opt);
1657 WARN_ON(1);
1658 return -1;
1659 }
1660 if (arg != 0)
1661 sbi->s_mount_opt |= m->mount_opt;
1662 else
1663 sbi->s_mount_opt &= ~m->mount_opt;
26092bf5 1664 }
5f3633e3 1665 return 1;
26092bf5
TT		 1666}
1667
1668static int parse_options(char *options, struct super_block *sb,
1669 unsigned long *journal_devnum,
1670 unsigned int *journal_ioprio,
1671 int is_remount)
1672{
1673 struct ext4_sb_info *sbi = EXT4_SB(sb);
1674 char *p;
1675 substring_t args[MAX_OPT_ARGS];
1676 int token;
1677
1678 if (!options)
1679 return 1;
1680
1681 while ((p = strsep(&options, ",")) != NULL) {
1682 if (!*p)
1683 continue;
1684 /*
1685 * Initialize args struct so we know whether arg was
1686 * found; some options take optional arguments.
1687 */
caecd0af 1688 args[0].to = args[0].from = NULL;
26092bf5
TT		 1689 token = match_token(p, tokens, args);
1690 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1691 journal_ioprio, is_remount) < 0)
1692 return 0;
ac27a0ec
DK		 1693 }
1694#ifdef CONFIG_QUOTA
1695 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
482a7425 1696 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
fd8c37ec 1697 clear_opt(sb, USRQUOTA);
ac27a0ec 1698
482a7425 1699 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
fd8c37ec 1700 clear_opt(sb, GRPQUOTA);
ac27a0ec 1701
56c50f11 1702 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
b31e1552
ES		 1703 ext4_msg(sb, KERN_ERR, "old and new quota "
1704 "format mixing");
ac27a0ec
DK		 1705 return 0;
1706 }
1707
1708 if (!sbi->s_jquota_fmt) {
b31e1552
ES		 1709 ext4_msg(sb, KERN_ERR, "journaled quota format "
1710 "not specified");
ac27a0ec
DK		 1711 return 0;
1712 }
1713 } else {
1714 if (sbi->s_jquota_fmt) {
b31e1552 1715 ext4_msg(sb, KERN_ERR, "journaled quota format "
2c8be6b2 1716 "specified with no journaling "
b31e1552 1717 "enabled");
ac27a0ec
DK		 1718 return 0;
1719 }
1720 }
1721#endif
261cb20c
JK		 1722 if (test_opt(sb, DIOREAD_NOLOCK)) {
1723 int blocksize =
1724 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1725
1726 if (blocksize < PAGE_CACHE_SIZE) {
1727 ext4_msg(sb, KERN_ERR, "can't mount with "
1728 "dioread_nolock if block size != PAGE_SIZE");
1729 return 0;
1730 }
1731 }
ac27a0ec
DK		 1732 return 1;
1733}
1734
2adf6da8
TT		 1735static inline void ext4_show_quota_options(struct seq_file *seq,
1736 struct super_block *sb)
1737{
1738#if defined(CONFIG_QUOTA)
1739 struct ext4_sb_info *sbi = EXT4_SB(sb);
1740
1741 if (sbi->s_jquota_fmt) {
1742 char *fmtname = "";
1743
1744 switch (sbi->s_jquota_fmt) {
1745 case QFMT_VFS_OLD:
1746 fmtname = "vfsold";
1747 break;
1748 case QFMT_VFS_V0:
1749 fmtname = "vfsv0";
1750 break;
1751 case QFMT_VFS_V1:
1752 fmtname = "vfsv1";
1753 break;
1754 }
1755 seq_printf(seq, ",jqfmt=%s", fmtname);
1756 }
1757
1758 if (sbi->s_qf_names[USRQUOTA])
1759 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1760
1761 if (sbi->s_qf_names[GRPQUOTA])
1762 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
1763
1764 if (test_opt(sb, USRQUOTA))
1765 seq_puts(seq, ",usrquota");
1766
1767 if (test_opt(sb, GRPQUOTA))
1768 seq_puts(seq, ",grpquota");
1769#endif
1770}
1771
5a916be1
TT		 1772static const char *token2str(int token)
1773{
50df9fd5 1774 const struct match_token *t;
5a916be1
TT		 1775
1776 for (t = tokens; t->token != Opt_err; t++)
1777 if (t->token == token && !strchr(t->pattern, '='))
1778 break;
1779 return t->pattern;
1780}
1781
2adf6da8
TT		 1782/*
1783 * Show an option if
1784 * - it's set to a non-default value OR
1785 * - if the per-sb default is different from the global default
1786 */
66acdcf4
TT		 1787static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1788 int nodefs)
2adf6da8 1789{
2adf6da8
TT		 1790 struct ext4_sb_info *sbi = EXT4_SB(sb);
1791 struct ext4_super_block *es = sbi->s_es;
66acdcf4 1792 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
5a916be1 1793 const struct mount_opts *m;
66acdcf4 1794 char sep = nodefs ? '\n' : ',';
2adf6da8 1795
66acdcf4
TT		 1796#define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1797#define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2adf6da8
TT		 1798
1799 if (sbi->s_sb_block != 1)
5a916be1
TT		 1800 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1801
1802 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1803 int want_set = m->flags & MOPT_SET;
1804 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1805 (m->flags & MOPT_CLEAR_ERR))
1806 continue;
66acdcf4 1807 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
5a916be1
TT		 1808 continue; /* skip if same as the default */
1809 if ((want_set &&
1810 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1811 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1812 continue; /* select Opt_noFoo vs Opt_Foo */
1813 SEQ_OPTS_PRINT("%s", token2str(m->token));
2adf6da8 1814 }
5a916be1 1815
08cefc7a 1816 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
5a916be1 1817 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
08cefc7a
EB		 1818 SEQ_OPTS_PRINT("resuid=%u",
1819 from_kuid_munged(&init_user_ns, sbi->s_resuid));
1820 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
5a916be1 1821 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
08cefc7a
EB		 1822 SEQ_OPTS_PRINT("resgid=%u",
1823 from_kgid_munged(&init_user_ns, sbi->s_resgid));
66acdcf4 1824 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
5a916be1
TT		 1825 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1826 SEQ_OPTS_PUTS("errors=remount-ro");
2adf6da8 1827 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
5a916be1 1828 SEQ_OPTS_PUTS("errors=continue");
2adf6da8 1829 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
5a916be1 1830 SEQ_OPTS_PUTS("errors=panic");
66acdcf4 1831 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
5a916be1 1832 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
66acdcf4 1833 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
5a916be1 1834 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
66acdcf4 1835 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
5a916be1 1836 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2adf6da8 1837 if (sb->s_flags & MS_I_VERSION)
5a916be1 1838 SEQ_OPTS_PUTS("i_version");
66acdcf4 1839 if (nodefs || sbi->s_stripe)
5a916be1 1840 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
66acdcf4 1841 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
5a916be1
TT		 1842 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1843 SEQ_OPTS_PUTS("data=journal");
1844 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1845 SEQ_OPTS_PUTS("data=ordered");
1846 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1847 SEQ_OPTS_PUTS("data=writeback");
1848 }
66acdcf4
TT		 1849 if (nodefs ||
1850 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
5a916be1
TT		 1851 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1852 sbi->s_inode_readahead_blks);
2adf6da8 1853
66acdcf4
TT		 1854 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1855 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
5a916be1 1856 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
df981d03
TT		 1857 if (nodefs || sbi->s_max_dir_size_kb)
1858 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
2adf6da8
TT		 1859
1860 ext4_show_quota_options(seq, sb);
2adf6da8
TT		 1861 return 0;
1862}
1863
66acdcf4
TT		 1864static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1865{
1866 return _ext4_show_options(seq, root->d_sb, 0);
1867}
1868
1869static int options_seq_show(struct seq_file *seq, void *offset)
1870{
1871 struct super_block *sb = seq->private;
1872 int rc;
1873
1874 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1875 rc = _ext4_show_options(seq, sb, 1);
1876 seq_puts(seq, "\n");
1877 return rc;
1878}
1879
1880static int options_open_fs(struct inode *inode, struct file *file)
1881{
1882 return single_open(file, options_seq_show, PDE(inode)->data);
1883}
1884
1885static const struct file_operations ext4_seq_options_fops = {
1886 .owner = THIS_MODULE,
1887 .open = options_open_fs,
1888 .read = seq_read,
1889 .llseek = seq_lseek,
1890 .release = single_release,
1891};
1892
617ba13b 1893static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 1894 int read_only)
1895{
617ba13b 1896 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 1897 int res = 0;
1898
617ba13b 1899 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
b31e1552
ES		 1900 ext4_msg(sb, KERN_ERR, "revision level too high, "
1901 "forcing read-only mode");
ac27a0ec
DK		 1902 res = MS_RDONLY;
1903 }
1904 if (read_only)
281b5995 1905 goto done;
617ba13b 1906 if (!(sbi->s_mount_state & EXT4_VALID_FS))
b31e1552
ES		 1907 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1908 "running e2fsck is recommended");
617ba13b 1909 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
b31e1552
ES		 1910 ext4_msg(sb, KERN_WARNING,
1911 "warning: mounting fs with errors, "
1912 "running e2fsck is recommended");
ed3ce80a 1913 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
ac27a0ec
DK		 1914 le16_to_cpu(es->s_mnt_count) >=
1915 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
b31e1552
ES		 1916 ext4_msg(sb, KERN_WARNING,
1917 "warning: maximal mount count reached, "
1918 "running e2fsck is recommended");
ac27a0ec
DK		 1919 else if (le32_to_cpu(es->s_checkinterval) &&
1920 (le32_to_cpu(es->s_lastcheck) +
1921 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
b31e1552
ES		 1922 ext4_msg(sb, KERN_WARNING,
1923 "warning: checktime reached, "
1924 "running e2fsck is recommended");
0b8e58a1 1925 if (!sbi->s_journal)
0390131b 1926 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec 1927 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 1928 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 1929 le16_add_cpu(&es->s_mnt_count, 1);
ac27a0ec 1930 es->s_mtime = cpu_to_le32(get_seconds());
617ba13b 1931 ext4_update_dynamic_rev(sb);
0390131b
FM		 1932 if (sbi->s_journal)
1933 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 1934
e2d67052 1935 ext4_commit_super(sb, 1);
281b5995 1936done:
ac27a0ec 1937 if (test_opt(sb, DEBUG))
a9df9a49 1938 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
a2595b8a 1939 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
ac27a0ec
DK		 1940 sb->s_blocksize,
1941 sbi->s_groups_count,
617ba13b
MC		 1942 EXT4_BLOCKS_PER_GROUP(sb),
1943 EXT4_INODES_PER_GROUP(sb),
a2595b8a 1944 sbi->s_mount_opt, sbi->s_mount_opt2);
ac27a0ec 1945
7abc52c2 1946 cleancache_init_fs(sb);
ac27a0ec
DK		 1947 return res;
1948}
1949
117fff10
TT		 1950int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
1951{
1952 struct ext4_sb_info *sbi = EXT4_SB(sb);
1953 struct flex_groups *new_groups;
1954 int size;
1955
1956 if (!sbi->s_log_groups_per_flex)
1957 return 0;
1958
1959 size = ext4_flex_group(sbi, ngroup - 1) + 1;
1960 if (size <= sbi->s_flex_groups_allocated)
1961 return 0;
1962
1963 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
1964 new_groups = ext4_kvzalloc(size, GFP_KERNEL);
1965 if (!new_groups) {
1966 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
1967 size / (int) sizeof(struct flex_groups));
1968 return -ENOMEM;
1969 }
1970
1971 if (sbi->s_flex_groups) {
1972 memcpy(new_groups, sbi->s_flex_groups,
1973 (sbi->s_flex_groups_allocated *
1974 sizeof(struct flex_groups)));
1975 ext4_kvfree(sbi->s_flex_groups);
1976 }
1977 sbi->s_flex_groups = new_groups;
1978 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
1979 return 0;
1980}
1981
772cb7c8
JS		 1982static int ext4_fill_flex_info(struct super_block *sb)
1983{
1984 struct ext4_sb_info *sbi = EXT4_SB(sb);
1985 struct ext4_group_desc *gdp = NULL;
772cb7c8 1986 ext4_group_t flex_group;
d50f2ab6 1987 unsigned int groups_per_flex = 0;
117fff10 1988 int i, err;
772cb7c8 1989
503358ae 1990 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
d50f2ab6 1991 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
772cb7c8
JS		 1992 sbi->s_log_groups_per_flex = 0;
1993 return 1;
1994 }
76495ec1 1995 groups_per_flex = 1U << sbi->s_log_groups_per_flex;
772cb7c8 1996
117fff10
TT		 1997 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
1998 if (err)
9933fc0a 1999 goto failed;
772cb7c8 2000
772cb7c8 2001 for (i = 0; i < sbi->s_groups_count; i++) {
88b6edd1 2002 gdp = ext4_get_group_desc(sb, i, NULL);
772cb7c8
JS		 2003
2004 flex_group = ext4_flex_group(sbi, i);
7ad9bb65
TT		 2005 atomic_add(ext4_free_inodes_count(sb, gdp),
2006 &sbi->s_flex_groups[flex_group].free_inodes);
021b65bb 2007 atomic_add(ext4_free_group_clusters(sb, gdp),
24aaa8ef 2008 &sbi->s_flex_groups[flex_group].free_clusters);
7ad9bb65
TT		 2009 atomic_add(ext4_used_dirs_count(sb, gdp),
2010 &sbi->s_flex_groups[flex_group].used_dirs);
772cb7c8
JS		 2011 }
2012
2013 return 1;
2014failed:
2015 return 0;
2016}
2017
feb0ab32
DW		 2018static __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
2019 struct ext4_group_desc *gdp)
717d50e4 2020{
feb0ab32 2021 int offset;
717d50e4 2022 __u16 crc = 0;
feb0ab32 2023 __le32 le_group = cpu_to_le32(block_group);
717d50e4 2024
feb0ab32
DW		 2025 if ((sbi->s_es->s_feature_ro_compat &
2026 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))) {
2027 /* Use new metadata_csum algorithm */
2028 __u16 old_csum;
2029 __u32 csum32;
2030
2031 old_csum = gdp->bg_checksum;
2032 gdp->bg_checksum = 0;
2033 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2034 sizeof(le_group));
2035 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp,
2036 sbi->s_desc_size);
2037 gdp->bg_checksum = old_csum;
2038
2039 crc = csum32 & 0xFFFF;
2040 goto out;
717d50e4
AD		 2041 }
2042
feb0ab32
DW		 2043 /* old crc16 code */
2044 offset = offsetof(struct ext4_group_desc, bg_checksum);
2045
2046 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2047 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2048 crc = crc16(crc, (__u8 *)gdp, offset);
2049 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2050 /* for checksum of struct ext4_group_desc do the rest...*/
2051 if ((sbi->s_es->s_feature_incompat &
2052 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
2053 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2054 crc = crc16(crc, (__u8 *)gdp + offset,
2055 le16_to_cpu(sbi->s_es->s_desc_size) -
2056 offset);
2057
2058out:
717d50e4
AD		 2059 return cpu_to_le16(crc);
2060}
2061
feb0ab32 2062int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
717d50e4
AD		 2063 struct ext4_group_desc *gdp)
2064{
feb0ab32
DW		 2065 if (ext4_has_group_desc_csum(sb) &&
2066 (gdp->bg_checksum != ext4_group_desc_csum(EXT4_SB(sb),
2067 block_group, gdp)))
717d50e4
AD		 2068 return 0;
2069
2070 return 1;
2071}
2072
feb0ab32
DW		 2073void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2074 struct ext4_group_desc *gdp)
2075{
2076 if (!ext4_has_group_desc_csum(sb))
2077 return;
2078 gdp->bg_checksum = ext4_group_desc_csum(EXT4_SB(sb), block_group, gdp);
2079}
2080
ac27a0ec 2081/* Called at mount-time, super-block is locked */
bfff6873
LC		 2082static int ext4_check_descriptors(struct super_block *sb,
2083 ext4_group_t *first_not_zeroed)
ac27a0ec 2084{
617ba13b
MC		 2085 struct ext4_sb_info *sbi = EXT4_SB(sb);
2086 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2087 ext4_fsblk_t last_block;
bd81d8ee
LV		 2088 ext4_fsblk_t block_bitmap;
2089 ext4_fsblk_t inode_bitmap;
2090 ext4_fsblk_t inode_table;
ce421581 2091 int flexbg_flag = 0;
bfff6873 2092 ext4_group_t i, grp = sbi->s_groups_count;
ac27a0ec 2093
ce421581
JS		 2094 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2095 flexbg_flag = 1;
2096
af5bc92d 2097 ext4_debug("Checking group descriptors");
ac27a0ec 2098
197cd65a
AM		 2099 for (i = 0; i < sbi->s_groups_count; i++) {
2100 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2101
ce421581 2102 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 2103 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 2104 else
2105 last_block = first_block +
617ba13b 2106 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 2107
bfff6873
LC		 2108 if ((grp == sbi->s_groups_count) &&
2109 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2110 grp = i;
2111
8fadc143 2112 block_bitmap = ext4_block_bitmap(sb, gdp);
2b2d6d01 2113 if (block_bitmap < first_block || block_bitmap > last_block) {
b31e1552 2114 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2115 "Block bitmap for group %u not in group "
b31e1552 2116 "(block %llu)!", i, block_bitmap);
ac27a0ec
DK		 2117 return 0;
2118 }
8fadc143 2119 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2b2d6d01 2120 if (inode_bitmap < first_block || inode_bitmap > last_block) {
b31e1552 2121 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2122 "Inode bitmap for group %u not in group "
b31e1552 2123 "(block %llu)!", i, inode_bitmap);
ac27a0ec
DK		 2124 return 0;
2125 }
8fadc143 2126 inode_table = ext4_inode_table(sb, gdp);
bd81d8ee 2127 if (inode_table < first_block ||
2b2d6d01 2128 inode_table + sbi->s_itb_per_group - 1 > last_block) {
b31e1552 2129 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2130 "Inode table for group %u not in group "
b31e1552 2131 "(block %llu)!", i, inode_table);
ac27a0ec
DK		 2132 return 0;
2133 }
955ce5f5 2134 ext4_lock_group(sb, i);
feb0ab32 2135 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
b31e1552
ES		 2136 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2137 "Checksum for group %u failed (%u!=%u)",
2138 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2139 gdp)), le16_to_cpu(gdp->bg_checksum));
7ee1ec4c 2140 if (!(sb->s_flags & MS_RDONLY)) {
955ce5f5 2141 ext4_unlock_group(sb, i);
8a266467 2142 return 0;
7ee1ec4c 2143 }
717d50e4 2144 }
955ce5f5 2145 ext4_unlock_group(sb, i);
ce421581
JS		 2146 if (!flexbg_flag)
2147 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec 2148 }
bfff6873
LC		 2149 if (NULL != first_not_zeroed)
2150 *first_not_zeroed = grp;
ac27a0ec 2151
5dee5437
TT		 2152 ext4_free_blocks_count_set(sbi->s_es,
2153 EXT4_C2B(sbi, ext4_count_free_clusters(sb)));
0b8e58a1 2154 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
ac27a0ec
DK		 2155 return 1;
2156}
2157
617ba13b 2158/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 2159 * the superblock) which were deleted from all directories, but held open by
2160 * a process at the time of a crash. We walk the list and try to delete these
2161 * inodes at recovery time (only with a read-write filesystem).
2162 *
2163 * In order to keep the orphan inode chain consistent during traversal (in
2164 * case of crash during recovery), we link each inode into the superblock
2165 * orphan list_head and handle it the same way as an inode deletion during
2166 * normal operation (which journals the operations for us).
2167 *
2168 * We only do an iget() and an iput() on each inode, which is very safe if we
2169 * accidentally point at an in-use or already deleted inode. The worst that
2170 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 2171 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 2172 * e2fsck was run on this filesystem, and it must have already done the orphan
2173 * inode cleanup for us, so we can safely abort without any further action.
2174 */
2b2d6d01
TT		 2175static void ext4_orphan_cleanup(struct super_block *sb,
2176 struct ext4_super_block *es)
ac27a0ec
DK		 2177{
2178 unsigned int s_flags = sb->s_flags;
2179 int nr_orphans = 0, nr_truncates = 0;
2180#ifdef CONFIG_QUOTA
2181 int i;
2182#endif
2183 if (!es->s_last_orphan) {
2184 jbd_debug(4, "no orphan inodes to clean up\n");
2185 return;
2186 }
2187
a8f48a95 2188 if (bdev_read_only(sb->s_bdev)) {
b31e1552
ES		 2189 ext4_msg(sb, KERN_ERR, "write access "
2190 "unavailable, skipping orphan cleanup");
a8f48a95
ES		 2191 return;
2192 }
2193
d39195c3
AG		 2194 /* Check if feature set would not allow a r/w mount */
2195 if (!ext4_feature_set_ok(sb, 0)) {
2196 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2197 "unknown ROCOMPAT features");
2198 return;
2199 }
2200
617ba13b 2201 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
c25f9bc6
ES		 2202 /* don't clear list on RO mount w/ errors */
2203 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
ac27a0ec
DK		 2204 jbd_debug(1, "Errors on filesystem, "
2205 "clearing orphan list.\n");
c25f9bc6
ES		 2206 es->s_last_orphan = 0;
2207 }
ac27a0ec
DK		 2208 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2209 return;
2210 }
2211
2212 if (s_flags & MS_RDONLY) {
b31e1552 2213 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
ac27a0ec
DK		 2214 sb->s_flags &= ~MS_RDONLY;
2215 }
2216#ifdef CONFIG_QUOTA
2217 /* Needed for iput() to work correctly and not trash data */
2218 sb->s_flags |= MS_ACTIVE;
2219 /* Turn on quotas so that they are updated correctly */
2220 for (i = 0; i < MAXQUOTAS; i++) {
617ba13b
MC		 2221 if (EXT4_SB(sb)->s_qf_names[i]) {
2222 int ret = ext4_quota_on_mount(sb, i);
ac27a0ec 2223 if (ret < 0)
b31e1552
ES		 2224 ext4_msg(sb, KERN_ERR,
2225 "Cannot turn on journaled "
2226 "quota: error %d", ret);
ac27a0ec
DK		 2227 }
2228 }
2229#endif
2230
2231 while (es->s_last_orphan) {
2232 struct inode *inode;
2233
97bd42b9
JB		 2234 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2235 if (IS_ERR(inode)) {
ac27a0ec
DK		 2236 es->s_last_orphan = 0;
2237 break;
2238 }
2239
617ba13b 2240 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
871a2931 2241 dquot_initialize(inode);
ac27a0ec 2242 if (inode->i_nlink) {
b31e1552
ES		 2243 ext4_msg(sb, KERN_DEBUG,
2244 "%s: truncating inode %lu to %lld bytes",
46e665e9 2245 __func__, inode->i_ino, inode->i_size);
e5f8eab8 2246 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
ac27a0ec 2247 inode->i_ino, inode->i_size);
721e3eba 2248 mutex_lock(&inode->i_mutex);
617ba13b 2249 ext4_truncate(inode);
721e3eba 2250 mutex_unlock(&inode->i_mutex);
ac27a0ec
DK		 2251 nr_truncates++;
2252 } else {
b31e1552
ES		 2253 ext4_msg(sb, KERN_DEBUG,
2254 "%s: deleting unreferenced inode %lu",
46e665e9 2255 __func__, inode->i_ino);
ac27a0ec
DK		 2256 jbd_debug(2, "deleting unreferenced inode %lu\n",
2257 inode->i_ino);
2258 nr_orphans++;
2259 }
2260 iput(inode); /* The delete magic happens here! */
2261 }
2262
2b2d6d01 2263#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
ac27a0ec
DK		 2264
2265 if (nr_orphans)
b31e1552
ES		 2266 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2267 PLURAL(nr_orphans));
ac27a0ec 2268 if (nr_truncates)
b31e1552
ES		 2269 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2270 PLURAL(nr_truncates));
ac27a0ec
DK		 2271#ifdef CONFIG_QUOTA
2272 /* Turn quotas off */
2273 for (i = 0; i < MAXQUOTAS; i++) {
2274 if (sb_dqopt(sb)->files[i])
287a8095 2275 dquot_quota_off(sb, i);
ac27a0ec
DK		 2276 }
2277#endif
2278 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2279}
0b8e58a1 2280
cd2291a4
ES		 2281/*
2282 * Maximal extent format file size.
2283 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2284 * extent format containers, within a sector_t, and within i_blocks
2285 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2286 * so that won't be a limiting factor.
2287 *
f17722f9
LC		 2288 * However there is other limiting factor. We do store extents in the form
2289 * of starting block and length, hence the resulting length of the extent
2290 * covering maximum file size must fit into on-disk format containers as
2291 * well. Given that length is always by 1 unit bigger than max unit (because
2292 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2293 *
cd2291a4
ES		 2294 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2295 */
f287a1a5 2296static loff_t ext4_max_size(int blkbits, int has_huge_files)
cd2291a4
ES		 2297{
2298 loff_t res;
2299 loff_t upper_limit = MAX_LFS_FILESIZE;
2300
2301 /* small i_blocks in vfs inode? */
f287a1a5 2302 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
cd2291a4 2303 /*
90c699a9 2304 * CONFIG_LBDAF is not enabled implies the inode
cd2291a4
ES		 2305 * i_block represent total blocks in 512 bytes
2306 * 32 == size of vfs inode i_blocks * 8
2307 */
2308 upper_limit = (1LL << 32) - 1;
2309
2310 /* total blocks in file system block size */
2311 upper_limit >>= (blkbits - 9);
2312 upper_limit <<= blkbits;
2313 }
2314
f17722f9
LC		 2315 /*
2316 * 32-bit extent-start container, ee_block. We lower the maxbytes
2317 * by one fs block, so ee_len can cover the extent of maximum file
2318 * size
2319 */
2320 res = (1LL << 32) - 1;
cd2291a4 2321 res <<= blkbits;
cd2291a4
ES		 2322
2323 /* Sanity check against vm- & vfs- imposed limits */
2324 if (res > upper_limit)
2325 res = upper_limit;
2326
2327 return res;
2328}
ac27a0ec 2329
ac27a0ec 2330/*
cd2291a4 2331 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 2332 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2333 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 2334 */
f287a1a5 2335static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
ac27a0ec 2336{
617ba13b 2337 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 2338 int meta_blocks;
2339 loff_t upper_limit;
0b8e58a1
AD		 2340 /* This is calculated to be the largest file size for a dense, block
2341 * mapped file such that the file's total number of 512-byte sectors,
2342 * including data and all indirect blocks, does not exceed (2^48 - 1).
2343 *
2344 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2345 * number of 512-byte sectors of the file.
0fc1b451
AK		 2346 */
2347
f287a1a5 2348 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
0fc1b451 2349 /*
90c699a9 2350 * !has_huge_files or CONFIG_LBDAF not enabled implies that
0b8e58a1
AD		 2351 * the inode i_block field represents total file blocks in
2352 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
0fc1b451
AK		 2353 */
2354 upper_limit = (1LL << 32) - 1;
2355
2356 /* total blocks in file system block size */
2357 upper_limit >>= (bits - 9);
2358
2359 } else {
8180a562
AK		 2360 /*
2361 * We use 48 bit ext4_inode i_blocks
2362 * With EXT4_HUGE_FILE_FL set the i_blocks
2363 * represent total number of blocks in
2364 * file system block size
2365 */
0fc1b451
AK		 2366 upper_limit = (1LL << 48) - 1;
2367
0fc1b451
AK		 2368 }
2369
2370 /* indirect blocks */
2371 meta_blocks = 1;
2372 /* double indirect blocks */
2373 meta_blocks += 1 + (1LL << (bits-2));
2374 /* tripple indirect blocks */
2375 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2376
2377 upper_limit -= meta_blocks;
2378 upper_limit <<= bits;
ac27a0ec
DK		 2379
2380 res += 1LL << (bits-2);
2381 res += 1LL << (2*(bits-2));
2382 res += 1LL << (3*(bits-2));
2383 res <<= bits;
2384 if (res > upper_limit)
2385 res = upper_limit;
0fc1b451
AK		 2386
2387 if (res > MAX_LFS_FILESIZE)
2388 res = MAX_LFS_FILESIZE;
2389
ac27a0ec
DK		 2390 return res;
2391}
2392
617ba13b 2393static ext4_fsblk_t descriptor_loc(struct super_block *sb,
0b8e58a1 2394 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 2395{
617ba13b 2396 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 2397 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 2398 int has_super = 0;
2399
2400 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2401
617ba13b 2402 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
ac27a0ec 2403 nr < first_meta_bg)
70bbb3e0 2404 return logical_sb_block + nr + 1;
ac27a0ec 2405 bg = sbi->s_desc_per_block * nr;
617ba13b 2406 if (ext4_bg_has_super(sb, bg))
ac27a0ec 2407 has_super = 1;
0b8e58a1 2408
617ba13b 2409 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 2410}
2411
c9de560d
AT		 2412/**
2413 * ext4_get_stripe_size: Get the stripe size.
2414 * @sbi: In memory super block info
2415 *
2416 * If we have specified it via mount option, then
2417 * use the mount option value. If the value specified at mount time is
2418 * greater than the blocks per group use the super block value.
2419 * If the super block value is greater than blocks per group return 0.
2420 * Allocator needs it be less than blocks per group.
2421 *
2422 */
2423static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2424{
2425 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2426 unsigned long stripe_width =
2427 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3eb08658 2428 int ret;
c9de560d
AT		 2429
2430 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3eb08658
DE		 2431 ret = sbi->s_stripe;
2432 else if (stripe_width <= sbi->s_blocks_per_group)
2433 ret = stripe_width;
2434 else if (stride <= sbi->s_blocks_per_group)
2435 ret = stride;
2436 else
2437 ret = 0;
c9de560d 2438
3eb08658
DE		 2439 /*
2440 * If the stripe width is 1, this makes no sense and
2441 * we set it to 0 to turn off stripe handling code.
2442 */
2443 if (ret <= 1)
2444 ret = 0;
c9de560d 2445
3eb08658 2446 return ret;
c9de560d 2447}
ac27a0ec 2448
3197ebdb
TT		 2449/* sysfs supprt */
2450
2451struct ext4_attr {
2452 struct attribute attr;
2453 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
60e6679e 2454 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
3197ebdb
TT		 2455 const char *, size_t);
2456 int offset;
2457};
2458
2459static int parse_strtoul(const char *buf,
2460 unsigned long max, unsigned long *value)
2461{
2462 char *endp;
2463
e7d2860b
AGR		 2464 *value = simple_strtoul(skip_spaces(buf), &endp, 0);
2465 endp = skip_spaces(endp);
3197ebdb
TT		 2466 if (*endp || *value > max)
2467 return -EINVAL;
2468
2469 return 0;
2470}
2471
2472static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2473 struct ext4_sb_info *sbi,
2474 char *buf)
2475{
2476 return snprintf(buf, PAGE_SIZE, "%llu\n",
7b415bf6
AK		 2477 (s64) EXT4_C2B(sbi,
2478 percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
3197ebdb
TT		 2479}
2480
2481static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2482 struct ext4_sb_info *sbi, char *buf)
2483{
2484 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2485
f613dfcb
TT		 2486 if (!sb->s_bdev->bd_part)
2487 return snprintf(buf, PAGE_SIZE, "0\n");
3197ebdb
TT		 2488 return snprintf(buf, PAGE_SIZE, "%lu\n",
2489 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2490 sbi->s_sectors_written_start) >> 1);
2491}
2492
2493static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2494 struct ext4_sb_info *sbi, char *buf)
2495{
2496 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2497
f613dfcb
TT		 2498 if (!sb->s_bdev->bd_part)
2499 return snprintf(buf, PAGE_SIZE, "0\n");
3197ebdb 2500 return snprintf(buf, PAGE_SIZE, "%llu\n",
a6b43e38 2501 (unsigned long long)(sbi->s_kbytes_written +
3197ebdb 2502 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
a6b43e38 2503 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
3197ebdb
TT		 2504}
2505
2506static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2507 struct ext4_sb_info *sbi,
2508 const char *buf, size_t count)
2509{
2510 unsigned long t;
2511
2512 if (parse_strtoul(buf, 0x40000000, &t))
2513 return -EINVAL;
2514
5dbd571d 2515 if (t && !is_power_of_2(t))
3197ebdb
TT		 2516 return -EINVAL;
2517
2518 sbi->s_inode_readahead_blks = t;
2519 return count;
2520}
2521
2522static ssize_t sbi_ui_show(struct ext4_attr *a,
0b8e58a1 2523 struct ext4_sb_info *sbi, char *buf)
3197ebdb
TT		 2524{
2525 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2526
2527 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2528}
2529
2530static ssize_t sbi_ui_store(struct ext4_attr *a,
2531 struct ext4_sb_info *sbi,
2532 const char *buf, size_t count)
2533{
2534 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2535 unsigned long t;
2536
2537 if (parse_strtoul(buf, 0xffffffff, &t))
2538 return -EINVAL;
2539 *ui = t;
2540 return count;
2541}
2542
2c0544b2
TT		 2543static ssize_t trigger_test_error(struct ext4_attr *a,
2544 struct ext4_sb_info *sbi,
2545 const char *buf, size_t count)
2546{
2547 int len = count;
2548
2549 if (!capable(CAP_SYS_ADMIN))
2550 return -EPERM;
2551
2552 if (len && buf[len-1] == '\n')
2553 len--;
2554
2555 if (len)
2556 ext4_error(sbi->s_sb, "%.*s", len, buf);
2557 return count;
2558}
2559
3197ebdb
TT		 2560#define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2561static struct ext4_attr ext4_attr_##_name = { \
2562 .attr = {.name = __stringify(_name), .mode = _mode }, \
2563 .show = _show, \
2564 .store = _store, \
2565 .offset = offsetof(struct ext4_sb_info, _elname), \
2566}
2567#define EXT4_ATTR(name, mode, show, store) \
2568static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2569
857ac889 2570#define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
3197ebdb
TT		 2571#define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2572#define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2573#define EXT4_RW_ATTR_SBI_UI(name, elname) \
2574 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2575#define ATTR_LIST(name) &ext4_attr_##name.attr
2576
2577EXT4_RO_ATTR(delayed_allocation_blocks);
2578EXT4_RO_ATTR(session_write_kbytes);
2579EXT4_RO_ATTR(lifetime_write_kbytes);
2580EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2581 inode_readahead_blks_store, s_inode_readahead_blks);
11013911 2582EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
3197ebdb
TT		 2583EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2584EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2585EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2586EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2587EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2588EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
55138e0b 2589EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
67a5da56 2590EXT4_RW_ATTR_SBI_UI(extent_max_zeroout_kb, s_extent_max_zeroout_kb);
2c0544b2 2591EXT4_ATTR(trigger_fs_error, 0200, NULL, trigger_test_error);
3197ebdb
TT		 2592
2593static struct attribute *ext4_attrs[] = {
2594 ATTR_LIST(delayed_allocation_blocks),
2595 ATTR_LIST(session_write_kbytes),
2596 ATTR_LIST(lifetime_write_kbytes),
2597 ATTR_LIST(inode_readahead_blks),
11013911 2598 ATTR_LIST(inode_goal),
3197ebdb
TT		 2599 ATTR_LIST(mb_stats),
2600 ATTR_LIST(mb_max_to_scan),
2601 ATTR_LIST(mb_min_to_scan),
2602 ATTR_LIST(mb_order2_req),
2603 ATTR_LIST(mb_stream_req),
2604 ATTR_LIST(mb_group_prealloc),
55138e0b 2605 ATTR_LIST(max_writeback_mb_bump),
67a5da56 2606 ATTR_LIST(extent_max_zeroout_kb),
2c0544b2 2607 ATTR_LIST(trigger_fs_error),
3197ebdb
TT		 2608 NULL,
2609};
2610
857ac889
LC		 2611/* Features this copy of ext4 supports */
2612EXT4_INFO_ATTR(lazy_itable_init);
27ee40df 2613EXT4_INFO_ATTR(batched_discard);
5e7bbef1 2614EXT4_INFO_ATTR(meta_bg_resize);
857ac889
LC		 2615
2616static struct attribute *ext4_feat_attrs[] = {
2617 ATTR_LIST(lazy_itable_init),
27ee40df 2618 ATTR_LIST(batched_discard),
5e7bbef1 2619 ATTR_LIST(meta_bg_resize),
857ac889
LC		 2620 NULL,
2621};
2622
3197ebdb
TT		 2623static ssize_t ext4_attr_show(struct kobject *kobj,
2624 struct attribute *attr, char *buf)
2625{
2626 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2627 s_kobj);
2628 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2629
2630 return a->show ? a->show(a, sbi, buf) : 0;
2631}
2632
2633static ssize_t ext4_attr_store(struct kobject *kobj,
2634 struct attribute *attr,
2635 const char *buf, size_t len)
2636{
2637 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2638 s_kobj);
2639 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2640
2641 return a->store ? a->store(a, sbi, buf, len) : 0;
2642}
2643
2644static void ext4_sb_release(struct kobject *kobj)
2645{
2646 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2647 s_kobj);
2648 complete(&sbi->s_kobj_unregister);
2649}
2650
52cf25d0 2651static const struct sysfs_ops ext4_attr_ops = {
3197ebdb
TT		 2652 .show = ext4_attr_show,
2653 .store = ext4_attr_store,
2654};
2655
2656static struct kobj_type ext4_ktype = {
2657 .default_attrs = ext4_attrs,
2658 .sysfs_ops = &ext4_attr_ops,
2659 .release = ext4_sb_release,
2660};
2661
857ac889
LC		 2662static void ext4_feat_release(struct kobject *kobj)
2663{
2664 complete(&ext4_feat->f_kobj_unregister);
2665}
2666
2667static struct kobj_type ext4_feat_ktype = {
2668 .default_attrs = ext4_feat_attrs,
2669 .sysfs_ops = &ext4_attr_ops,
2670 .release = ext4_feat_release,
2671};
2672
a13fb1a4
ES		 2673/*
2674 * Check whether this filesystem can be mounted based on
2675 * the features present and the RDONLY/RDWR mount requested.
2676 * Returns 1 if this filesystem can be mounted as requested,
2677 * 0 if it cannot be.
2678 */
2679static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2680{
2681 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2682 ext4_msg(sb, KERN_ERR,
2683 "Couldn't mount because of "
2684 "unsupported optional features (%x)",
2685 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2686 ~EXT4_FEATURE_INCOMPAT_SUPP));
2687 return 0;
2688 }
2689
2690 if (readonly)
2691 return 1;
2692
2693 /* Check that feature set is OK for a read-write mount */
2694 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2695 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2696 "unsupported optional features (%x)",
2697 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2698 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2699 return 0;
2700 }
2701 /*
2702 * Large file size enabled file system can only be mounted
2703 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2704 */
2705 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2706 if (sizeof(blkcnt_t) < sizeof(u64)) {
2707 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2708 "cannot be mounted RDWR without "
2709 "CONFIG_LBDAF");
2710 return 0;
2711 }
2712 }
bab08ab9
TT		 2713 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2714 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2715 ext4_msg(sb, KERN_ERR,
2716 "Can't support bigalloc feature without "
2717 "extents feature\n");
2718 return 0;
2719 }
7c319d32
AK		 2720
2721#ifndef CONFIG_QUOTA
2722 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
2723 !readonly) {
2724 ext4_msg(sb, KERN_ERR,
2725 "Filesystem with quota feature cannot be mounted RDWR "
2726 "without CONFIG_QUOTA");
2727 return 0;
2728 }
2729#endif /* CONFIG_QUOTA */
a13fb1a4
ES		 2730 return 1;
2731}
2732
66e61a9e
TT		 2733/*
2734 * This function is called once a day if we have errors logged
2735 * on the file system
2736 */
2737static void print_daily_error_info(unsigned long arg)
2738{
2739 struct super_block *sb = (struct super_block *) arg;
2740 struct ext4_sb_info *sbi;
2741 struct ext4_super_block *es;
2742
2743 sbi = EXT4_SB(sb);
2744 es = sbi->s_es;
2745
2746 if (es->s_error_count)
2747 ext4_msg(sb, KERN_NOTICE, "error count: %u",
2748 le32_to_cpu(es->s_error_count));
2749 if (es->s_first_error_time) {
2750 printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d",
2751 sb->s_id, le32_to_cpu(es->s_first_error_time),
2752 (int) sizeof(es->s_first_error_func),
2753 es->s_first_error_func,
2754 le32_to_cpu(es->s_first_error_line));
2755 if (es->s_first_error_ino)
2756 printk(": inode %u",
2757 le32_to_cpu(es->s_first_error_ino));
2758 if (es->s_first_error_block)
2759 printk(": block %llu", (unsigned long long)
2760 le64_to_cpu(es->s_first_error_block));
2761 printk("\n");
2762 }
2763 if (es->s_last_error_time) {
2764 printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d",
2765 sb->s_id, le32_to_cpu(es->s_last_error_time),
2766 (int) sizeof(es->s_last_error_func),
2767 es->s_last_error_func,
2768 le32_to_cpu(es->s_last_error_line));
2769 if (es->s_last_error_ino)
2770 printk(": inode %u",
2771 le32_to_cpu(es->s_last_error_ino));
2772 if (es->s_last_error_block)
2773 printk(": block %llu", (unsigned long long)
2774 le64_to_cpu(es->s_last_error_block));
2775 printk("\n");
2776 }
2777 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2778}
2779
bfff6873
LC		 2780/* Find next suitable group and run ext4_init_inode_table */
2781static int ext4_run_li_request(struct ext4_li_request *elr)
2782{
2783 struct ext4_group_desc *gdp = NULL;
2784 ext4_group_t group, ngroups;
2785 struct super_block *sb;
2786 unsigned long timeout = 0;
2787 int ret = 0;
2788
2789 sb = elr->lr_super;
2790 ngroups = EXT4_SB(sb)->s_groups_count;
2791
8e8ad8a5 2792 sb_start_write(sb);
bfff6873
LC		 2793 for (group = elr->lr_next_group; group < ngroups; group++) {
2794 gdp = ext4_get_group_desc(sb, group, NULL);
2795 if (!gdp) {
2796 ret = 1;
2797 break;
2798 }
2799
2800 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2801 break;
2802 }
2803
7f511862 2804 if (group >= ngroups)
bfff6873
LC		 2805 ret = 1;
2806
2807 if (!ret) {
2808 timeout = jiffies;
2809 ret = ext4_init_inode_table(sb, group,
2810 elr->lr_timeout ? 0 : 1);
2811 if (elr->lr_timeout == 0) {
51ce6511
LC		 2812 timeout = (jiffies - timeout) *
2813 elr->lr_sbi->s_li_wait_mult;
bfff6873
LC		 2814 elr->lr_timeout = timeout;
2815 }
2816 elr->lr_next_sched = jiffies + elr->lr_timeout;
2817 elr->lr_next_group = group + 1;
2818 }
8e8ad8a5 2819 sb_end_write(sb);
bfff6873
LC		 2820
2821 return ret;
2822}
2823
2824/*
2825 * Remove lr_request from the list_request and free the
4ed5c033 2826 * request structure. Should be called with li_list_mtx held
bfff6873
LC		 2827 */
2828static void ext4_remove_li_request(struct ext4_li_request *elr)
2829{
2830 struct ext4_sb_info *sbi;
2831
2832 if (!elr)
2833 return;
2834
2835 sbi = elr->lr_sbi;
2836
2837 list_del(&elr->lr_request);
2838 sbi->s_li_request = NULL;
2839 kfree(elr);
2840}
2841
2842static void ext4_unregister_li_request(struct super_block *sb)
2843{
1bb933fb
LC		 2844 mutex_lock(&ext4_li_mtx);
2845 if (!ext4_li_info) {
2846 mutex_unlock(&ext4_li_mtx);
bfff6873 2847 return;
1bb933fb 2848 }
bfff6873
LC		 2849
2850 mutex_lock(&ext4_li_info->li_list_mtx);
1bb933fb 2851 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
bfff6873 2852 mutex_unlock(&ext4_li_info->li_list_mtx);
1bb933fb 2853 mutex_unlock(&ext4_li_mtx);
bfff6873
LC		 2854}
2855
8f1f7453
ES		 2856static struct task_struct *ext4_lazyinit_task;
2857
bfff6873
LC		 2858/*
2859 * This is the function where ext4lazyinit thread lives. It walks
2860 * through the request list searching for next scheduled filesystem.
2861 * When such a fs is found, run the lazy initialization request
2862 * (ext4_rn_li_request) and keep track of the time spend in this
2863 * function. Based on that time we compute next schedule time of
2864 * the request. When walking through the list is complete, compute
2865 * next waking time and put itself into sleep.
2866 */
2867static int ext4_lazyinit_thread(void *arg)
2868{
2869 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2870 struct list_head *pos, *n;
2871 struct ext4_li_request *elr;
4ed5c033 2872 unsigned long next_wakeup, cur;
bfff6873
LC		 2873
2874 BUG_ON(NULL == eli);
2875
bfff6873
LC		 2876cont_thread:
2877 while (true) {
2878 next_wakeup = MAX_JIFFY_OFFSET;
2879
2880 mutex_lock(&eli->li_list_mtx);
2881 if (list_empty(&eli->li_request_list)) {
2882 mutex_unlock(&eli->li_list_mtx);
2883 goto exit_thread;
2884 }
2885
2886 list_for_each_safe(pos, n, &eli->li_request_list) {
2887 elr = list_entry(pos, struct ext4_li_request,
2888 lr_request);
2889
b2c78cd0
TT		 2890 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2891 if (ext4_run_li_request(elr) != 0) {
2892 /* error, remove the lazy_init job */
2893 ext4_remove_li_request(elr);
2894 continue;
2895 }
bfff6873
LC		 2896 }
2897
2898 if (time_before(elr->lr_next_sched, next_wakeup))
2899 next_wakeup = elr->lr_next_sched;
2900 }
2901 mutex_unlock(&eli->li_list_mtx);
2902
a0acae0e 2903 try_to_freeze();
bfff6873 2904
4ed5c033
LC		 2905 cur = jiffies;
2906 if ((time_after_eq(cur, next_wakeup)) ||
f4245bd4 2907 (MAX_JIFFY_OFFSET == next_wakeup)) {
bfff6873
LC		 2908 cond_resched();
2909 continue;
2910 }
2911
4ed5c033
LC		 2912 schedule_timeout_interruptible(next_wakeup - cur);
2913
8f1f7453
ES		 2914 if (kthread_should_stop()) {
2915 ext4_clear_request_list();
2916 goto exit_thread;
2917 }
bfff6873
LC		 2918 }
2919
2920exit_thread:
2921 /*
2922 * It looks like the request list is empty, but we need
2923 * to check it under the li_list_mtx lock, to prevent any
2924 * additions into it, and of course we should lock ext4_li_mtx
2925 * to atomically free the list and ext4_li_info, because at
2926 * this point another ext4 filesystem could be registering
2927 * new one.
2928 */
2929 mutex_lock(&ext4_li_mtx);
2930 mutex_lock(&eli->li_list_mtx);
2931 if (!list_empty(&eli->li_request_list)) {
2932 mutex_unlock(&eli->li_list_mtx);
2933 mutex_unlock(&ext4_li_mtx);
2934 goto cont_thread;
2935 }
2936 mutex_unlock(&eli->li_list_mtx);
bfff6873
LC		 2937 kfree(ext4_li_info);
2938 ext4_li_info = NULL;
2939 mutex_unlock(&ext4_li_mtx);
2940
2941 return 0;
2942}
2943
2944static void ext4_clear_request_list(void)
2945{
2946 struct list_head *pos, *n;
2947 struct ext4_li_request *elr;
2948
2949 mutex_lock(&ext4_li_info->li_list_mtx);
bfff6873
LC		 2950 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2951 elr = list_entry(pos, struct ext4_li_request,
2952 lr_request);
2953 ext4_remove_li_request(elr);
2954 }
2955 mutex_unlock(&ext4_li_info->li_list_mtx);
2956}
2957
2958static int ext4_run_lazyinit_thread(void)
2959{
8f1f7453
ES		 2960 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2961 ext4_li_info, "ext4lazyinit");
2962 if (IS_ERR(ext4_lazyinit_task)) {
2963 int err = PTR_ERR(ext4_lazyinit_task);
bfff6873 2964 ext4_clear_request_list();
bfff6873
LC		 2965 kfree(ext4_li_info);
2966 ext4_li_info = NULL;
92b97816 2967 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
bfff6873
LC		 2968 "initialization thread\n",
2969 err);
2970 return err;
2971 }
2972 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
bfff6873
LC		 2973 return 0;
2974}
2975
2976/*
2977 * Check whether it make sense to run itable init. thread or not.
2978 * If there is at least one uninitialized inode table, return
2979 * corresponding group number, else the loop goes through all
2980 * groups and return total number of groups.
2981 */
2982static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2983{
2984 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2985 struct ext4_group_desc *gdp = NULL;
2986
2987 for (group = 0; group < ngroups; group++) {
2988 gdp = ext4_get_group_desc(sb, group, NULL);
2989 if (!gdp)
2990 continue;
2991
2992 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2993 break;
2994 }
2995
2996 return group;
2997}
2998
2999static int ext4_li_info_new(void)
3000{
3001 struct ext4_lazy_init *eli = NULL;
3002
3003 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3004 if (!eli)
3005 return -ENOMEM;
3006
bfff6873
LC		 3007 INIT_LIST_HEAD(&eli->li_request_list);
3008 mutex_init(&eli->li_list_mtx);
3009
bfff6873
LC		 3010 eli->li_state |= EXT4_LAZYINIT_QUIT;
3011
3012 ext4_li_info = eli;
3013
3014 return 0;
3015}
3016
3017static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3018 ext4_group_t start)
3019{
3020 struct ext4_sb_info *sbi = EXT4_SB(sb);
3021 struct ext4_li_request *elr;
3022 unsigned long rnd;
3023
3024 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3025 if (!elr)
3026 return NULL;
3027
3028 elr->lr_super = sb;
3029 elr->lr_sbi = sbi;
3030 elr->lr_next_group = start;
3031
3032 /*
3033 * Randomize first schedule time of the request to
3034 * spread the inode table initialization requests
3035 * better.
3036 */
3037 get_random_bytes(&rnd, sizeof(rnd));
3038 elr->lr_next_sched = jiffies + (unsigned long)rnd %
3039 (EXT4_DEF_LI_MAX_START_DELAY * HZ);
3040
3041 return elr;
3042}
3043
7f511862
TT		 3044int ext4_register_li_request(struct super_block *sb,
3045 ext4_group_t first_not_zeroed)
bfff6873
LC		 3046{
3047 struct ext4_sb_info *sbi = EXT4_SB(sb);
7f511862 3048 struct ext4_li_request *elr = NULL;
bfff6873 3049 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
6c5a6cb9 3050 int ret = 0;
bfff6873 3051
7f511862 3052 mutex_lock(&ext4_li_mtx);
51ce6511
LC		 3053 if (sbi->s_li_request != NULL) {
3054 /*
3055 * Reset timeout so it can be computed again, because
3056 * s_li_wait_mult might have changed.
3057 */
3058 sbi->s_li_request->lr_timeout = 0;
7f511862 3059 goto out;
51ce6511 3060 }
bfff6873
LC		 3061
3062 if (first_not_zeroed == ngroups ||
3063 (sb->s_flags & MS_RDONLY) ||
55ff3840 3064 !test_opt(sb, INIT_INODE_TABLE))
7f511862 3065 goto out;
bfff6873
LC		 3066
3067 elr = ext4_li_request_new(sb, first_not_zeroed);
7f511862
TT		 3068 if (!elr) {
3069 ret = -ENOMEM;
3070 goto out;
3071 }
bfff6873
LC		 3072
3073 if (NULL == ext4_li_info) {
3074 ret = ext4_li_info_new();
3075 if (ret)
3076 goto out;
3077 }
3078
3079 mutex_lock(&ext4_li_info->li_list_mtx);
3080 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3081 mutex_unlock(&ext4_li_info->li_list_mtx);
3082
3083 sbi->s_li_request = elr;
46e4690b
TM		 3084 /*
3085 * set elr to NULL here since it has been inserted to
3086 * the request_list and the removal and free of it is
3087 * handled by ext4_clear_request_list from now on.
3088 */
3089 elr = NULL;
bfff6873
LC		 3090
3091 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3092 ret = ext4_run_lazyinit_thread();
3093 if (ret)
3094 goto out;
3095 }
bfff6873 3096out:
beed5ecb
NK		 3097 mutex_unlock(&ext4_li_mtx);
3098 if (ret)
bfff6873 3099 kfree(elr);
bfff6873
LC		 3100 return ret;
3101}
3102
3103/*
3104 * We do not need to lock anything since this is called on
3105 * module unload.
3106 */
3107static void ext4_destroy_lazyinit_thread(void)
3108{
3109 /*
3110 * If thread exited earlier
3111 * there's nothing to be done.
3112 */
8f1f7453 3113 if (!ext4_li_info || !ext4_lazyinit_task)
bfff6873
LC		 3114 return;
3115
8f1f7453 3116 kthread_stop(ext4_lazyinit_task);
bfff6873
LC		 3117}
3118
25ed6e8a
DW		 3119static int set_journal_csum_feature_set(struct super_block *sb)
3120{
3121 int ret = 1;
3122 int compat, incompat;
3123 struct ext4_sb_info *sbi = EXT4_SB(sb);
3124
3125 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3126 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3127 /* journal checksum v2 */
3128 compat = 0;
3129 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V2;
3130 } else {
3131 /* journal checksum v1 */
3132 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3133 incompat = 0;
3134 }
3135
3136 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3137 ret = jbd2_journal_set_features(sbi->s_journal,
3138 compat, 0,
3139 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3140 incompat);
3141 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3142 ret = jbd2_journal_set_features(sbi->s_journal,
3143 compat, 0,
3144 incompat);
3145 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3146 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3147 } else {
3148 jbd2_journal_clear_features(sbi->s_journal,
3149 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3150 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3151 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3152 }
3153
3154 return ret;
3155}
3156
952fc18e
TT		 3157/*
3158 * Note: calculating the overhead so we can be compatible with
3159 * historical BSD practice is quite difficult in the face of
3160 * clusters/bigalloc. This is because multiple metadata blocks from
3161 * different block group can end up in the same allocation cluster.
3162 * Calculating the exact overhead in the face of clustered allocation
3163 * requires either O(all block bitmaps) in memory or O(number of block
3164 * groups**2) in time. We will still calculate the superblock for
3165 * older file systems --- and if we come across with a bigalloc file
3166 * system with zero in s_overhead_clusters the estimate will be close to
3167 * correct especially for very large cluster sizes --- but for newer
3168 * file systems, it's better to calculate this figure once at mkfs
3169 * time, and store it in the superblock. If the superblock value is
3170 * present (even for non-bigalloc file systems), we will use it.
3171 */
3172static int count_overhead(struct super_block *sb, ext4_group_t grp,
3173 char *buf)
3174{
3175 struct ext4_sb_info *sbi = EXT4_SB(sb);
3176 struct ext4_group_desc *gdp;
3177 ext4_fsblk_t first_block, last_block, b;
3178 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3179 int s, j, count = 0;
3180
0548bbb8
TT		 3181 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC))
3182 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3183 sbi->s_itb_per_group + 2);
3184
952fc18e
TT		 3185 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3186 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3187 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3188 for (i = 0; i < ngroups; i++) {
3189 gdp = ext4_get_group_desc(sb, i, NULL);
3190 b = ext4_block_bitmap(sb, gdp);
3191 if (b >= first_block && b <= last_block) {
3192 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3193 count++;
3194 }
3195 b = ext4_inode_bitmap(sb, gdp);
3196 if (b >= first_block && b <= last_block) {
3197 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3198 count++;
3199 }
3200 b = ext4_inode_table(sb, gdp);
3201 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3202 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3203 int c = EXT4_B2C(sbi, b - first_block);
3204 ext4_set_bit(c, buf);
3205 count++;
3206 }
3207 if (i != grp)
3208 continue;
3209 s = 0;
3210 if (ext4_bg_has_super(sb, grp)) {
3211 ext4_set_bit(s++, buf);
3212 count++;
3213 }
3214 for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
3215 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3216 count++;
3217 }
3218 }
3219 if (!count)
3220 return 0;
3221 return EXT4_CLUSTERS_PER_GROUP(sb) -
3222 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3223}
3224
3225/*
3226 * Compute the overhead and stash it in sbi->s_overhead
3227 */
3228int ext4_calculate_overhead(struct super_block *sb)
3229{
3230 struct ext4_sb_info *sbi = EXT4_SB(sb);
3231 struct ext4_super_block *es = sbi->s_es;
3232 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3233 ext4_fsblk_t overhead = 0;
3234 char *buf = (char *) get_zeroed_page(GFP_KERNEL);
3235
952fc18e
TT		 3236 if (!buf)
3237 return -ENOMEM;
3238
3239 /*
3240 * Compute the overhead (FS structures). This is constant
3241 * for a given filesystem unless the number of block groups
3242 * changes so we cache the previous value until it does.
3243 */
3244
3245 /*
3246 * All of the blocks before first_data_block are overhead
3247 */
3248 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3249
3250 /*
3251 * Add the overhead found in each block group
3252 */
3253 for (i = 0; i < ngroups; i++) {
3254 int blks;
3255
3256 blks = count_overhead(sb, i, buf);
3257 overhead += blks;
3258 if (blks)
3259 memset(buf, 0, PAGE_SIZE);
3260 cond_resched();
3261 }
0875a2b4
ES		 3262 /* Add the journal blocks as well */
3263 if (sbi->s_journal)
3264 overhead += EXT4_B2C(sbi, sbi->s_journal->j_maxlen);
3265
952fc18e
TT		 3266 sbi->s_overhead = overhead;
3267 smp_wmb();
3268 free_page((unsigned long) buf);
3269 return 0;
3270}
3271
2b2d6d01 3272static int ext4_fill_super(struct super_block *sb, void *data, int silent)
ac27a0ec 3273{
d4c402d9 3274 char *orig_data = kstrdup(data, GFP_KERNEL);
2b2d6d01 3275 struct buffer_head *bh;
617ba13b
MC		 3276 struct ext4_super_block *es = NULL;
3277 struct ext4_sb_info *sbi;
3278 ext4_fsblk_t block;
3279 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 3280 ext4_fsblk_t logical_sb_block;
ac27a0ec 3281 unsigned long offset = 0;
ac27a0ec
DK		 3282 unsigned long journal_devnum = 0;
3283 unsigned long def_mount_opts;
3284 struct inode *root;
9f6200bb 3285 char *cp;
0390131b 3286 const char *descr;
dcc7dae3 3287 int ret = -ENOMEM;
281b5995 3288 int blocksize, clustersize;
4ec11028
TT		 3289 unsigned int db_count;
3290 unsigned int i;
281b5995 3291 int needs_recovery, has_huge_files, has_bigalloc;
bd81d8ee 3292 __u64 blocks_count;
07aa2ea1 3293 int err = 0;
b3881f74 3294 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
bfff6873 3295 ext4_group_t first_not_zeroed;
ac27a0ec
DK		 3296
3297 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3298 if (!sbi)
dcc7dae3 3299 goto out_free_orig;
705895b6
PE		 3300
3301 sbi->s_blockgroup_lock =
3302 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3303 if (!sbi->s_blockgroup_lock) {
3304 kfree(sbi);
dcc7dae3 3305 goto out_free_orig;
705895b6 3306 }
ac27a0ec 3307 sb->s_fs_info = sbi;
2c0544b2 3308 sbi->s_sb = sb;
240799cd 3309 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
d9c9bef1 3310 sbi->s_sb_block = sb_block;
f613dfcb
TT		 3311 if (sb->s_bdev->bd_part)
3312 sbi->s_sectors_written_start =
3313 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
ac27a0ec 3314
9f6200bb
TT		 3315 /* Cleanup superblock name */
3316 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3317 *cp = '!';
3318
07aa2ea1 3319 /* -EINVAL is default */
dcc7dae3 3320 ret = -EINVAL;
617ba13b 3321 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 3322 if (!blocksize) {
b31e1552 3323 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
ac27a0ec
DK		 3324 goto out_fail;
3325 }
3326
3327 /*
617ba13b 3328 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 3329 * block sizes. We need to calculate the offset from buffer start.
3330 */
617ba13b 3331 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 3332 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3333 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 3334 } else {
70bbb3e0 3335 logical_sb_block = sb_block;
ac27a0ec
DK		 3336 }
3337
70bbb3e0 3338 if (!(bh = sb_bread(sb, logical_sb_block))) {
b31e1552 3339 ext4_msg(sb, KERN_ERR, "unable to read superblock");
ac27a0ec
DK		 3340 goto out_fail;
3341 }
3342 /*
3343 * Note: s_es must be initialized as soon as possible because
617ba13b 3344 * some ext4 macro-instructions depend on its value
ac27a0ec 3345 */
2716b802 3346 es = (struct ext4_super_block *) (bh->b_data + offset);
ac27a0ec
DK		 3347 sbi->s_es = es;
3348 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 3349 if (sb->s_magic != EXT4_SUPER_MAGIC)
3350 goto cantfind_ext4;
afc32f7e 3351 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
ac27a0ec 3352
feb0ab32
DW		 3353 /* Warn if metadata_csum and gdt_csum are both set. */
3354 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3355 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
3356 EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_GDT_CSUM))
3357 ext4_warning(sb, KERN_INFO "metadata_csum and uninit_bg are "
3358 "redundant flags; please run fsck.");
3359
d25425f8
DW		 3360 /* Check for a known checksum algorithm */
3361 if (!ext4_verify_csum_type(sb, es)) {
3362 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3363 "unknown checksum algorithm.");
3364 silent = 1;
3365 goto cantfind_ext4;
3366 }
3367
0441984a
DW		 3368 /* Load the checksum driver */
3369 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3370 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3371 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3372 if (IS_ERR(sbi->s_chksum_driver)) {
3373 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3374 ret = PTR_ERR(sbi->s_chksum_driver);
3375 sbi->s_chksum_driver = NULL;
3376 goto failed_mount;
3377 }
3378 }
3379
a9c47317
DW		 3380 /* Check superblock checksum */
3381 if (!ext4_superblock_csum_verify(sb, es)) {
3382 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3383 "invalid superblock checksum. Run e2fsck?");
3384 silent = 1;
3385 goto cantfind_ext4;
3386 }
3387
3388 /* Precompute checksum seed for all metadata */
3389 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3390 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
3391 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3392 sizeof(es->s_uuid));
3393
ac27a0ec
DK		 3394 /* Set defaults before we parse the mount options */
3395 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
fd8c37ec 3396 set_opt(sb, INIT_INODE_TABLE);
617ba13b 3397 if (def_mount_opts & EXT4_DEFM_DEBUG)
fd8c37ec 3398 set_opt(sb, DEBUG);
87f26807 3399 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
fd8c37ec 3400 set_opt(sb, GRPID);
617ba13b 3401 if (def_mount_opts & EXT4_DEFM_UID16)
fd8c37ec 3402 set_opt(sb, NO_UID32);
ea663336 3403 /* xattr user namespace & acls are now defaulted on */
ea663336 3404 set_opt(sb, XATTR_USER);
03010a33 3405#ifdef CONFIG_EXT4_FS_POSIX_ACL
ea663336 3406 set_opt(sb, POSIX_ACL);
2e7842b8 3407#endif
617ba13b 3408 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
fd8c37ec 3409 set_opt(sb, JOURNAL_DATA);
617ba13b 3410 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
fd8c37ec 3411 set_opt(sb, ORDERED_DATA);
617ba13b 3412 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
fd8c37ec 3413 set_opt(sb, WRITEBACK_DATA);
617ba13b
MC		 3414
3415 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
fd8c37ec 3416 set_opt(sb, ERRORS_PANIC);
bb4f397a 3417 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
fd8c37ec 3418 set_opt(sb, ERRORS_CONT);
bb4f397a 3419 else
fd8c37ec 3420 set_opt(sb, ERRORS_RO);
8b67f04a 3421 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
fd8c37ec 3422 set_opt(sb, BLOCK_VALIDITY);
8b67f04a 3423 if (def_mount_opts & EXT4_DEFM_DISCARD)
fd8c37ec 3424 set_opt(sb, DISCARD);
ac27a0ec 3425
08cefc7a
EB		 3426 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3427 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
30773840
TT		 3428 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3429 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3430 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
ac27a0ec 3431
8b67f04a 3432 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
fd8c37ec 3433 set_opt(sb, BARRIER);
ac27a0ec 3434
dd919b98
AK		 3435 /*
3436 * enable delayed allocation by default
3437 * Use -o nodelalloc to turn it off
3438 */
bc0b75f7 3439 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
8b67f04a 3440 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
fd8c37ec 3441 set_opt(sb, DELALLOC);
dd919b98 3442
51ce6511
LC		 3443 /*
3444 * set default s_li_wait_mult for lazyinit, for the case there is
3445 * no mount option specified.
3446 */
3447 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3448
8b67f04a 3449 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
661aa520 3450 &journal_devnum, &journal_ioprio, 0)) {
8b67f04a
TT		 3451 ext4_msg(sb, KERN_WARNING,
3452 "failed to parse options in superblock: %s",
3453 sbi->s_es->s_mount_opts);
3454 }
5a916be1 3455 sbi->s_def_mount_opt = sbi->s_mount_opt;
b3881f74 3456 if (!parse_options((char *) data, sb, &journal_devnum,
661aa520 3457 &journal_ioprio, 0))
ac27a0ec
DK		 3458 goto failed_mount;
3459
56889787
TT		 3460 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3461 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3462 "with data=journal disables delayed "
3463 "allocation and O_DIRECT support!\n");
3464 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3465 ext4_msg(sb, KERN_ERR, "can't mount with "
3466 "both data=journal and delalloc");
3467 goto failed_mount;
3468 }
3469 if (test_opt(sb, DIOREAD_NOLOCK)) {
3470 ext4_msg(sb, KERN_ERR, "can't mount with "
3471 "both data=journal and delalloc");
3472 goto failed_mount;
3473 }
3474 if (test_opt(sb, DELALLOC))
3475 clear_opt(sb, DELALLOC);
3476 }
3477
ac27a0ec 3478 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 3479 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec 3480
617ba13b
MC		 3481 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3482 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3483 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3484 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
b31e1552
ES		 3485 ext4_msg(sb, KERN_WARNING,
3486 "feature flags set on rev 0 fs, "
3487 "running e2fsck is recommended");
469108ff 3488
2035e776
TT		 3489 if (IS_EXT2_SB(sb)) {
3490 if (ext2_feature_set_ok(sb))
3491 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3492 "using the ext4 subsystem");
3493 else {
3494 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3495 "to feature incompatibilities");
3496 goto failed_mount;
3497 }
3498 }
3499
3500 if (IS_EXT3_SB(sb)) {
3501 if (ext3_feature_set_ok(sb))
3502 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3503 "using the ext4 subsystem");
3504 else {
3505 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3506 "to feature incompatibilities");
3507 goto failed_mount;
3508 }
3509 }
3510
ac27a0ec
DK		 3511 /*
3512 * Check feature flags regardless of the revision level, since we
3513 * previously didn't change the revision level when setting the flags,
3514 * so there is a chance incompat flags are set on a rev 0 filesystem.
3515 */
a13fb1a4 3516 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
ac27a0ec 3517 goto failed_mount;
a13fb1a4 3518
261cb20c 3519 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
617ba13b
MC		 3520 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3521 blocksize > EXT4_MAX_BLOCK_SIZE) {
b31e1552
ES		 3522 ext4_msg(sb, KERN_ERR,
3523 "Unsupported filesystem blocksize %d", blocksize);
ac27a0ec
DK		 3524 goto failed_mount;
3525 }
3526
ac27a0ec 3527 if (sb->s_blocksize != blocksize) {
ce40733c
AK		 3528 /* Validate the filesystem blocksize */
3529 if (!sb_set_blocksize(sb, blocksize)) {
b31e1552 3530 ext4_msg(sb, KERN_ERR, "bad block size %d",
ce40733c 3531 blocksize);
ac27a0ec
DK		 3532 goto failed_mount;
3533 }
3534
2b2d6d01 3535 brelse(bh);
70bbb3e0
AM		 3536 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3537 offset = do_div(logical_sb_block, blocksize);
3538 bh = sb_bread(sb, logical_sb_block);
ac27a0ec 3539 if (!bh) {
b31e1552
ES		 3540 ext4_msg(sb, KERN_ERR,
3541 "Can't read superblock on 2nd try");
ac27a0ec
DK		 3542 goto failed_mount;
3543 }
2716b802 3544 es = (struct ext4_super_block *)(bh->b_data + offset);
ac27a0ec 3545 sbi->s_es = es;
617ba13b 3546 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
b31e1552
ES		 3547 ext4_msg(sb, KERN_ERR,
3548 "Magic mismatch, very weird!");
ac27a0ec
DK		 3549 goto failed_mount;
3550 }
3551 }
3552
a13fb1a4
ES		 3553 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3554 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
f287a1a5
TT		 3555 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3556 has_huge_files);
3557 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
ac27a0ec 3558
617ba13b
MC		 3559 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3560 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3561 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
ac27a0ec
DK		 3562 } else {
3563 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3564 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
617ba13b 3565 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1330593e 3566 (!is_power_of_2(sbi->s_inode_size)) ||
ac27a0ec 3567 (sbi->s_inode_size > blocksize)) {
b31e1552
ES		 3568 ext4_msg(sb, KERN_ERR,
3569 "unsupported inode size: %d",
2b2d6d01 3570 sbi->s_inode_size);
ac27a0ec
DK		 3571 goto failed_mount;
3572 }
ef7f3835
KS		 3573 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3574 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
ac27a0ec 3575 }
0b8e58a1 3576
0d1ee42f
AR		 3577 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3578 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
8fadc143 3579 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 3580 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 3581 !is_power_of_2(sbi->s_desc_size)) {
b31e1552
ES		 3582 ext4_msg(sb, KERN_ERR,
3583 "unsupported descriptor size %lu",
0d1ee42f
AR		 3584 sbi->s_desc_size);
3585 goto failed_mount;
3586 }
3587 } else
3588 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
0b8e58a1 3589
ac27a0ec 3590 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 3591 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
b47b6f38 3592 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
617ba13b 3593 goto cantfind_ext4;
0b8e58a1 3594
617ba13b 3595 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 3596 if (sbi->s_inodes_per_block == 0)
617ba13b 3597 goto cantfind_ext4;
ac27a0ec
DK		 3598 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3599 sbi->s_inodes_per_block;
0d1ee42f 3600 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 3601 sbi->s_sbh = bh;
3602 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 3603 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3604 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
0b8e58a1 3605
2b2d6d01 3606 for (i = 0; i < 4; i++)
ac27a0ec
DK		 3607 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3608 sbi->s_def_hash_version = es->s_def_hash_version;
f99b2589
TT		 3609 i = le32_to_cpu(es->s_flags);
3610 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3611 sbi->s_hash_unsigned = 3;
3612 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3613#ifdef __CHAR_UNSIGNED__
3614 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3615 sbi->s_hash_unsigned = 3;
3616#else
3617 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3618#endif
f99b2589 3619 }
ac27a0ec 3620
281b5995
TT		 3621 /* Handle clustersize */
3622 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3623 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3624 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3625 if (has_bigalloc) {
3626 if (clustersize < blocksize) {
3627 ext4_msg(sb, KERN_ERR,
3628 "cluster size (%d) smaller than "
3629 "block size (%d)", clustersize, blocksize);
3630 goto failed_mount;
3631 }
3632 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3633 le32_to_cpu(es->s_log_block_size);
3634 sbi->s_clusters_per_group =
3635 le32_to_cpu(es->s_clusters_per_group);
3636 if (sbi->s_clusters_per_group > blocksize * 8) {
3637 ext4_msg(sb, KERN_ERR,
3638 "#clusters per group too big: %lu",
3639 sbi->s_clusters_per_group);
3640 goto failed_mount;
3641 }
3642 if (sbi->s_blocks_per_group !=
3643 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3644 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3645 "clusters per group (%lu) inconsistent",
3646 sbi->s_blocks_per_group,
3647 sbi->s_clusters_per_group);
3648 goto failed_mount;
3649 }
3650 } else {
3651 if (clustersize != blocksize) {
3652 ext4_warning(sb, "fragment/cluster size (%d) != "
3653 "block size (%d)", clustersize,
3654 blocksize);
3655 clustersize = blocksize;
3656 }
3657 if (sbi->s_blocks_per_group > blocksize * 8) {
3658 ext4_msg(sb, KERN_ERR,
3659 "#blocks per group too big: %lu",
3660 sbi->s_blocks_per_group);
3661 goto failed_mount;
3662 }
3663 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3664 sbi->s_cluster_bits = 0;
ac27a0ec 3665 }
281b5995
TT		 3666 sbi->s_cluster_ratio = clustersize / blocksize;
3667
ac27a0ec 3668 if (sbi->s_inodes_per_group > blocksize * 8) {
b31e1552
ES		 3669 ext4_msg(sb, KERN_ERR,
3670 "#inodes per group too big: %lu",
2b2d6d01 3671 sbi->s_inodes_per_group);
ac27a0ec
DK		 3672 goto failed_mount;
3673 }
3674
bf43d84b
ES		 3675 /*
3676 * Test whether we have more sectors than will fit in sector_t,
3677 * and whether the max offset is addressable by the page cache.
3678 */
5a9ae68a 3679 err = generic_check_addressable(sb->s_blocksize_bits,
30ca22c7 3680 ext4_blocks_count(es));
5a9ae68a 3681 if (err) {
b31e1552 3682 ext4_msg(sb, KERN_ERR, "filesystem"
bf43d84b 3683 " too large to mount safely on this system");
ac27a0ec 3684 if (sizeof(sector_t) < 8)
90c699a9 3685 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
ac27a0ec
DK		 3686 goto failed_mount;
3687 }
3688
617ba13b
MC		 3689 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3690 goto cantfind_ext4;
e7c95593 3691
0f2ddca6
FTN		 3692 /* check blocks count against device size */
3693 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3694 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
b31e1552
ES		 3695 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3696 "exceeds size of device (%llu blocks)",
0f2ddca6
FTN		 3697 ext4_blocks_count(es), blocks_count);
3698 goto failed_mount;
3699 }
3700
0b8e58a1
AD		 3701 /*
3702 * It makes no sense for the first data block to be beyond the end
3703 * of the filesystem.
3704 */
3705 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
5635a62b 3706 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
b31e1552
ES		 3707 "block %u is beyond end of filesystem (%llu)",
3708 le32_to_cpu(es->s_first_data_block),
3709 ext4_blocks_count(es));
e7c95593
ES		 3710 goto failed_mount;
3711 }
bd81d8ee
LV		 3712 blocks_count = (ext4_blocks_count(es) -
3713 le32_to_cpu(es->s_first_data_block) +
3714 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3715 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4ec11028 3716 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
b31e1552 3717 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
4ec11028 3718 "(block count %llu, first data block %u, "
b31e1552 3719 "blocks per group %lu)", sbi->s_groups_count,
4ec11028
TT		 3720 ext4_blocks_count(es),
3721 le32_to_cpu(es->s_first_data_block),
3722 EXT4_BLOCKS_PER_GROUP(sb));
3723 goto failed_mount;
3724 }
bd81d8ee 3725 sbi->s_groups_count = blocks_count;
fb0a387d
ES		 3726 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3727 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
617ba13b
MC		 3728 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3729 EXT4_DESC_PER_BLOCK(sb);
f18a5f21
TT		 3730 sbi->s_group_desc = ext4_kvmalloc(db_count *
3731 sizeof(struct buffer_head *),
3732 GFP_KERNEL);
ac27a0ec 3733 if (sbi->s_group_desc == NULL) {
b31e1552 3734 ext4_msg(sb, KERN_ERR, "not enough memory");
2cde417d 3735 ret = -ENOMEM;
ac27a0ec
DK		 3736 goto failed_mount;
3737 }
3738
9f6200bb
TT		 3739 if (ext4_proc_root)
3740 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
240799cd 3741
66acdcf4
TT		 3742 if (sbi->s_proc)
3743 proc_create_data("options", S_IRUGO, sbi->s_proc,
3744 &ext4_seq_options_fops, sb);
3745
705895b6 3746 bgl_lock_init(sbi->s_blockgroup_lock);
ac27a0ec
DK		 3747
3748 for (i = 0; i < db_count; i++) {
70bbb3e0 3749 block = descriptor_loc(sb, logical_sb_block, i);
ac27a0ec
DK		 3750 sbi->s_group_desc[i] = sb_bread(sb, block);
3751 if (!sbi->s_group_desc[i]) {
b31e1552
ES		 3752 ext4_msg(sb, KERN_ERR,
3753 "can't read group descriptor %d", i);
ac27a0ec
DK		 3754 db_count = i;
3755 goto failed_mount2;
3756 }
3757 }
bfff6873 3758 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
b31e1552 3759 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
ac27a0ec
DK		 3760 goto failed_mount2;
3761 }
772cb7c8
JS		 3762 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3763 if (!ext4_fill_flex_info(sb)) {
b31e1552
ES		 3764 ext4_msg(sb, KERN_ERR,
3765 "unable to initialize "
3766 "flex_bg meta info!");
772cb7c8
JS		 3767 goto failed_mount2;
3768 }
3769
ac27a0ec
DK		 3770 sbi->s_gdb_count = db_count;
3771 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3772 spin_lock_init(&sbi->s_next_gen_lock);
3773
04496411
TM		 3774 init_timer(&sbi->s_err_report);
3775 sbi->s_err_report.function = print_daily_error_info;
3776 sbi->s_err_report.data = (unsigned long) sb;
3777
57042651 3778 err = percpu_counter_init(&sbi->s_freeclusters_counter,
5dee5437 3779 ext4_count_free_clusters(sb));
ce7e010a
TT		 3780 if (!err) {
3781 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3782 ext4_count_free_inodes(sb));
3783 }
3784 if (!err) {
3785 err = percpu_counter_init(&sbi->s_dirs_counter,
3786 ext4_count_dirs(sb));
3787 }
3788 if (!err) {
57042651 3789 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0);
ce7e010a
TT		 3790 }
3791 if (err) {
3792 ext4_msg(sb, KERN_ERR, "insufficient memory");
3793 goto failed_mount3;
3794 }
3795
c9de560d 3796 sbi->s_stripe = ext4_get_stripe_size(sbi);
55138e0b 3797 sbi->s_max_writeback_mb_bump = 128;
67a5da56 3798 sbi->s_extent_max_zeroout_kb = 32;
c9de560d 3799
ac27a0ec
DK		 3800 /*
3801 * set up enough so that it can read an inode
3802 */
9ca92389
TT		 3803 if (!test_opt(sb, NOLOAD) &&
3804 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3805 sb->s_op = &ext4_sops;
3806 else
3807 sb->s_op = &ext4_nojournal_sops;
617ba13b
MC		 3808 sb->s_export_op = &ext4_export_ops;
3809 sb->s_xattr = ext4_xattr_handlers;
ac27a0ec 3810#ifdef CONFIG_QUOTA
617ba13b
MC		 3811 sb->s_qcop = &ext4_qctl_operations;
3812 sb->dq_op = &ext4_quota_operations;
7c319d32
AK		 3813
3814 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) {
3815 /* Use qctl operations for hidden quota files. */
3816 sb->s_qcop = &ext4_qctl_sysfile_operations;
3817 }
ac27a0ec 3818#endif
f2fa2ffc
AK		 3819 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3820
ac27a0ec 3821 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3b9d4ed2 3822 mutex_init(&sbi->s_orphan_lock);
ac27a0ec
DK		 3823
3824 sb->s_root = NULL;
3825
3826 needs_recovery = (es->s_last_orphan != 0 ||
617ba13b
MC		 3827 EXT4_HAS_INCOMPAT_FEATURE(sb,
3828 EXT4_FEATURE_INCOMPAT_RECOVER));
ac27a0ec 3829
c5e06d10
JL		 3830 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3831 !(sb->s_flags & MS_RDONLY))
3832 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3833 goto failed_mount3;
3834
ac27a0ec
DK		 3835 /*
3836 * The first inode we look at is the journal inode. Don't try
3837 * root first: it may be modified in the journal!
3838 */
3839 if (!test_opt(sb, NOLOAD) &&
617ba13b
MC		 3840 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3841 if (ext4_load_journal(sb, es, journal_devnum))
ac27a0ec 3842 goto failed_mount3;
0390131b
FM		 3843 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3844 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
b31e1552
ES		 3845 ext4_msg(sb, KERN_ERR, "required journal recovery "
3846 "suppressed and not mounted read-only");
744692dc 3847 goto failed_mount_wq;
ac27a0ec 3848 } else {
fd8c37ec 3849 clear_opt(sb, DATA_FLAGS);
0390131b
FM		 3850 sbi->s_journal = NULL;
3851 needs_recovery = 0;
3852 goto no_journal;
ac27a0ec
DK		 3853 }
3854
f32aaf2d 3855 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT) &&
eb40a09c
JS		 3856 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3857 JBD2_FEATURE_INCOMPAT_64BIT)) {
b31e1552 3858 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
744692dc 3859 goto failed_mount_wq;
eb40a09c
JS		 3860 }
3861
25ed6e8a
DW		 3862 if (!set_journal_csum_feature_set(sb)) {
3863 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
3864 "feature set");
3865 goto failed_mount_wq;
d4da6c9c 3866 }
818d276c 3867
ac27a0ec
DK		 3868 /* We have now updated the journal if required, so we can
3869 * validate the data journaling mode. */
3870 switch (test_opt(sb, DATA_FLAGS)) {
3871 case 0:
3872 /* No mode set, assume a default based on the journal
63f57933
AM		 3873 * capabilities: ORDERED_DATA if the journal can
3874 * cope, else JOURNAL_DATA
3875 */
dab291af
MC		 3876 if (jbd2_journal_check_available_features
3877 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
fd8c37ec 3878 set_opt(sb, ORDERED_DATA);
ac27a0ec 3879 else
fd8c37ec 3880 set_opt(sb, JOURNAL_DATA);
ac27a0ec
DK		 3881 break;
3882
617ba13b
MC		 3883 case EXT4_MOUNT_ORDERED_DATA:
3884 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 3885 if (!jbd2_journal_check_available_features
3886 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
b31e1552
ES		 3887 ext4_msg(sb, KERN_ERR, "Journal does not support "
3888 "requested data journaling mode");
744692dc 3889 goto failed_mount_wq;
ac27a0ec
DK		 3890 }
3891 default:
3892 break;
3893 }
b3881f74 3894 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
ac27a0ec 3895
18aadd47
BJ		 3896 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
3897
ce7e010a
TT		 3898 /*
3899 * The journal may have updated the bg summary counts, so we
3900 * need to update the global counters.
3901 */
57042651 3902 percpu_counter_set(&sbi->s_freeclusters_counter,
5dee5437 3903 ext4_count_free_clusters(sb));
ce7e010a
TT		 3904 percpu_counter_set(&sbi->s_freeinodes_counter,
3905 ext4_count_free_inodes(sb));
3906 percpu_counter_set(&sbi->s_dirs_counter,
3907 ext4_count_dirs(sb));
57042651 3908 percpu_counter_set(&sbi->s_dirtyclusters_counter, 0);
206f7ab4 3909
ce7e010a 3910no_journal:
952fc18e
TT		 3911 /*
3912 * Get the # of file system overhead blocks from the
3913 * superblock if present.
3914 */
3915 if (es->s_overhead_clusters)
3916 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
3917 else {
07aa2ea1
LC		 3918 err = ext4_calculate_overhead(sb);
3919 if (err)
952fc18e
TT		 3920 goto failed_mount_wq;
3921 }
3922
fd89d5f2
TH		 3923 /*
3924 * The maximum number of concurrent works can be high and
3925 * concurrency isn't really necessary. Limit it to 1.
3926 */
3927 EXT4_SB(sb)->dio_unwritten_wq =
ae005cbe 3928 alloc_workqueue("ext4-dio-unwritten", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4c0425ff
MC		 3929 if (!EXT4_SB(sb)->dio_unwritten_wq) {
3930 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
07aa2ea1 3931 ret = -ENOMEM;
4c0425ff
MC		 3932 goto failed_mount_wq;
3933 }
3934
ac27a0ec 3935 /*
dab291af 3936 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 3937 * so we can safely mount the rest of the filesystem now.
3938 */
3939
1d1fe1ee
DH		 3940 root = ext4_iget(sb, EXT4_ROOT_INO);
3941 if (IS_ERR(root)) {
b31e1552 3942 ext4_msg(sb, KERN_ERR, "get root inode failed");
1d1fe1ee 3943 ret = PTR_ERR(root);
32a9bb57 3944 root = NULL;
ac27a0ec
DK		 3945 goto failed_mount4;
3946 }
3947 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
b31e1552 3948 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
94bf608a 3949 iput(root);
ac27a0ec
DK		 3950 goto failed_mount4;
3951 }
48fde701 3952 sb->s_root = d_make_root(root);
1d1fe1ee 3953 if (!sb->s_root) {
b31e1552 3954 ext4_msg(sb, KERN_ERR, "get root dentry failed");
1d1fe1ee
DH		 3955 ret = -ENOMEM;
3956 goto failed_mount4;
3957 }
ac27a0ec 3958
7e84b621
ES		 3959 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
3960 sb->s_flags |= MS_RDONLY;
ef7f3835
KS		 3961
3962 /* determine the minimum size of new large inodes, if present */
3963 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3964 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3965 EXT4_GOOD_OLD_INODE_SIZE;
3966 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3967 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
3968 if (sbi->s_want_extra_isize <
3969 le16_to_cpu(es->s_want_extra_isize))
3970 sbi->s_want_extra_isize =
3971 le16_to_cpu(es->s_want_extra_isize);
3972 if (sbi->s_want_extra_isize <
3973 le16_to_cpu(es->s_min_extra_isize))
3974 sbi->s_want_extra_isize =
3975 le16_to_cpu(es->s_min_extra_isize);
3976 }
3977 }
3978 /* Check if enough inode space is available */
3979 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3980 sbi->s_inode_size) {
3981 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3982 EXT4_GOOD_OLD_INODE_SIZE;
b31e1552
ES		 3983 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3984 "available");
ef7f3835
KS		 3985 }
3986
6fd058f7
TT		 3987 err = ext4_setup_system_zone(sb);
3988 if (err) {
b31e1552 3989 ext4_msg(sb, KERN_ERR, "failed to initialize system "
fbe845dd 3990 "zone (%d)", err);
94bf608a 3991 goto failed_mount4a;
6fd058f7
TT		 3992 }
3993
c2774d84 3994 ext4_ext_init(sb);
9d99012f 3995 err = ext4_mb_init(sb);
c2774d84 3996 if (err) {
421f91d2 3997 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
b31e1552 3998 err);
dcf2d804 3999 goto failed_mount5;
c2774d84
AK		 4000 }
4001
bfff6873
LC		 4002 err = ext4_register_li_request(sb, first_not_zeroed);
4003 if (err)
dcf2d804 4004 goto failed_mount6;
bfff6873 4005
3197ebdb
TT		 4006 sbi->s_kobj.kset = ext4_kset;
4007 init_completion(&sbi->s_kobj_unregister);
4008 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
4009 "%s", sb->s_id);
dcf2d804
TM		 4010 if (err)
4011 goto failed_mount7;
3197ebdb 4012
617ba13b
MC		 4013 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4014 ext4_orphan_cleanup(sb, es);
4015 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
0390131b 4016 if (needs_recovery) {
b31e1552 4017 ext4_msg(sb, KERN_INFO, "recovery complete");
0390131b
FM		 4018 ext4_mark_recovery_complete(sb, es);
4019 }
4020 if (EXT4_SB(sb)->s_journal) {
4021 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4022 descr = " journalled data mode";
4023 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4024 descr = " ordered data mode";
4025 else
4026 descr = " writeback data mode";
4027 } else
4028 descr = "out journal";
4029
7c319d32
AK		 4030#ifdef CONFIG_QUOTA
4031 /* Enable quota usage during mount. */
4032 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
4033 !(sb->s_flags & MS_RDONLY)) {
07aa2ea1
LC		 4034 err = ext4_enable_quotas(sb);
4035 if (err)
72ba7450 4036 goto failed_mount8;
7c319d32
AK		 4037 }
4038#endif /* CONFIG_QUOTA */
4039
79add3a3
LC		 4040 if (test_opt(sb, DISCARD)) {
4041 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4042 if (!blk_queue_discard(q))
4043 ext4_msg(sb, KERN_WARNING,
4044 "mounting with \"discard\" option, but "
4045 "the device does not support discard");
4046 }
4047
d4c402d9 4048 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
8b67f04a
TT		 4049 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
4050 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
ac27a0ec 4051
66e61a9e
TT		 4052 if (es->s_error_count)
4053 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
ac27a0ec 4054
d4c402d9 4055 kfree(orig_data);
ac27a0ec
DK		 4056 return 0;
4057
617ba13b 4058cantfind_ext4:
ac27a0ec 4059 if (!silent)
b31e1552 4060 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
ac27a0ec
DK		 4061 goto failed_mount;
4062
72ba7450
TT		 4063#ifdef CONFIG_QUOTA
4064failed_mount8:
4065 kobject_del(&sbi->s_kobj);
4066#endif
dcf2d804
TM		 4067failed_mount7:
4068 ext4_unregister_li_request(sb);
4069failed_mount6:
dcf2d804 4070 ext4_mb_release(sb);
94bf608a
AV		 4071failed_mount5:
4072 ext4_ext_release(sb);
dcf2d804 4073 ext4_release_system_zone(sb);
94bf608a
AV		 4074failed_mount4a:
4075 dput(sb->s_root);
32a9bb57 4076 sb->s_root = NULL;
94bf608a 4077failed_mount4:
b31e1552 4078 ext4_msg(sb, KERN_ERR, "mount failed");
4c0425ff
MC		 4079 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
4080failed_mount_wq:
0390131b
FM		 4081 if (sbi->s_journal) {
4082 jbd2_journal_destroy(sbi->s_journal);
4083 sbi->s_journal = NULL;
4084 }
ac27a0ec 4085failed_mount3:
04496411 4086 del_timer(&sbi->s_err_report);
9933fc0a
TT		 4087 if (sbi->s_flex_groups)
4088 ext4_kvfree(sbi->s_flex_groups);
57042651 4089 percpu_counter_destroy(&sbi->s_freeclusters_counter);
ce7e010a
TT		 4090 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4091 percpu_counter_destroy(&sbi->s_dirs_counter);
57042651 4092 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
c5e06d10
JL		 4093 if (sbi->s_mmp_tsk)
4094 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec
DK		 4095failed_mount2:
4096 for (i = 0; i < db_count; i++)
4097 brelse(sbi->s_group_desc[i]);
f18a5f21 4098 ext4_kvfree(sbi->s_group_desc);
ac27a0ec 4099failed_mount:
0441984a
DW		 4100 if (sbi->s_chksum_driver)
4101 crypto_free_shash(sbi->s_chksum_driver);
240799cd 4102 if (sbi->s_proc) {
66acdcf4 4103 remove_proc_entry("options", sbi->s_proc);
9f6200bb 4104 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 4105 }
ac27a0ec
DK		 4106#ifdef CONFIG_QUOTA
4107 for (i = 0; i < MAXQUOTAS; i++)
4108 kfree(sbi->s_qf_names[i]);
4109#endif
617ba13b 4110 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 4111 brelse(bh);
4112out_fail:
4113 sb->s_fs_info = NULL;
f6830165 4114 kfree(sbi->s_blockgroup_lock);
ac27a0ec 4115 kfree(sbi);
dcc7dae3 4116out_free_orig:
d4c402d9 4117 kfree(orig_data);
07aa2ea1 4118 return err ? err : ret;
ac27a0ec
DK		 4119}
4120
4121/*
4122 * Setup any per-fs journal parameters now. We'll do this both on
4123 * initial mount, once the journal has been initialised but before we've
4124 * done any recovery; and again on any subsequent remount.
4125 */
617ba13b 4126static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 4127{
617ba13b 4128 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4129
30773840
TT		 4130 journal->j_commit_interval = sbi->s_commit_interval;
4131 journal->j_min_batch_time = sbi->s_min_batch_time;
4132 journal->j_max_batch_time = sbi->s_max_batch_time;
ac27a0ec 4133
a931da6a 4134 write_lock(&journal->j_state_lock);
ac27a0ec 4135 if (test_opt(sb, BARRIER))
dab291af 4136 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 4137 else
dab291af 4138 journal->j_flags &= ~JBD2_BARRIER;
5bf5683a
HK		 4139 if (test_opt(sb, DATA_ERR_ABORT))
4140 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4141 else
4142 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
a931da6a 4143 write_unlock(&journal->j_state_lock);
ac27a0ec
DK		 4144}
4145
617ba13b 4146static journal_t *ext4_get_journal(struct super_block *sb,
ac27a0ec
DK		 4147 unsigned int journal_inum)
4148{
4149 struct inode *journal_inode;
4150 journal_t *journal;
4151
0390131b
FM		 4152 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4153
ac27a0ec
DK		 4154 /* First, test for the existence of a valid inode on disk. Bad
4155 * things happen if we iget() an unused inode, as the subsequent
4156 * iput() will try to delete it. */
4157
1d1fe1ee
DH		 4158 journal_inode = ext4_iget(sb, journal_inum);
4159 if (IS_ERR(journal_inode)) {
b31e1552 4160 ext4_msg(sb, KERN_ERR, "no journal found");
ac27a0ec
DK		 4161 return NULL;
4162 }
4163 if (!journal_inode->i_nlink) {
4164 make_bad_inode(journal_inode);
4165 iput(journal_inode);
b31e1552 4166 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
ac27a0ec
DK		 4167 return NULL;
4168 }
4169
e5f8eab8 4170 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
ac27a0ec 4171 journal_inode, journal_inode->i_size);
1d1fe1ee 4172 if (!S_ISREG(journal_inode->i_mode)) {
b31e1552 4173 ext4_msg(sb, KERN_ERR, "invalid journal inode");
ac27a0ec
DK		 4174 iput(journal_inode);
4175 return NULL;
4176 }
4177
dab291af 4178 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 4179 if (!journal) {
b31e1552 4180 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
ac27a0ec
DK		 4181 iput(journal_inode);
4182 return NULL;
4183 }
4184 journal->j_private = sb;
617ba13b 4185 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 4186 return journal;
4187}
4188
617ba13b 4189static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 4190 dev_t j_dev)
4191{
2b2d6d01 4192 struct buffer_head *bh;
ac27a0ec 4193 journal_t *journal;
617ba13b
MC		 4194 ext4_fsblk_t start;
4195 ext4_fsblk_t len;
ac27a0ec 4196 int hblock, blocksize;
617ba13b 4197 ext4_fsblk_t sb_block;
ac27a0ec 4198 unsigned long offset;
2b2d6d01 4199 struct ext4_super_block *es;
ac27a0ec
DK		 4200 struct block_device *bdev;
4201
0390131b
FM		 4202 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4203
b31e1552 4204 bdev = ext4_blkdev_get(j_dev, sb);
ac27a0ec
DK		 4205 if (bdev == NULL)
4206 return NULL;
4207
ac27a0ec 4208 blocksize = sb->s_blocksize;
e1defc4f 4209 hblock = bdev_logical_block_size(bdev);
ac27a0ec 4210 if (blocksize < hblock) {
b31e1552
ES		 4211 ext4_msg(sb, KERN_ERR,
4212 "blocksize too small for journal device");
ac27a0ec
DK		 4213 goto out_bdev;
4214 }
4215
617ba13b
MC		 4216 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4217 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 4218 set_blocksize(bdev, blocksize);
4219 if (!(bh = __bread(bdev, sb_block, blocksize))) {
b31e1552
ES		 4220 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4221 "external journal");
ac27a0ec
DK		 4222 goto out_bdev;
4223 }
4224
2716b802 4225 es = (struct ext4_super_block *) (bh->b_data + offset);
617ba13b 4226 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 4227 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b 4228 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
b31e1552
ES		 4229 ext4_msg(sb, KERN_ERR, "external journal has "
4230 "bad superblock");
ac27a0ec
DK		 4231 brelse(bh);
4232 goto out_bdev;
4233 }
4234
617ba13b 4235 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
b31e1552 4236 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
ac27a0ec
DK		 4237 brelse(bh);
4238 goto out_bdev;
4239 }
4240
bd81d8ee 4241 len = ext4_blocks_count(es);
ac27a0ec
DK		 4242 start = sb_block + 1;
4243 brelse(bh); /* we're done with the superblock */
4244
dab291af 4245 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 4246 start, len, blocksize);
4247 if (!journal) {
b31e1552 4248 ext4_msg(sb, KERN_ERR, "failed to create device journal");
ac27a0ec
DK		 4249 goto out_bdev;
4250 }
4251 journal->j_private = sb;
4252 ll_rw_block(READ, 1, &journal->j_sb_buffer);
4253 wait_on_buffer(journal->j_sb_buffer);
4254 if (!buffer_uptodate(journal->j_sb_buffer)) {
b31e1552 4255 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
ac27a0ec
DK		 4256 goto out_journal;
4257 }
4258 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
b31e1552
ES		 4259 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4260 "user (unsupported) - %d",
ac27a0ec
DK		 4261 be32_to_cpu(journal->j_superblock->s_nr_users));
4262 goto out_journal;
4263 }
617ba13b
MC		 4264 EXT4_SB(sb)->journal_bdev = bdev;
4265 ext4_init_journal_params(sb, journal);
ac27a0ec 4266 return journal;
0b8e58a1 4267
ac27a0ec 4268out_journal:
dab291af 4269 jbd2_journal_destroy(journal);
ac27a0ec 4270out_bdev:
617ba13b 4271 ext4_blkdev_put(bdev);
ac27a0ec
DK		 4272 return NULL;
4273}
4274
617ba13b
MC		 4275static int ext4_load_journal(struct super_block *sb,
4276 struct ext4_super_block *es,
ac27a0ec
DK		 4277 unsigned long journal_devnum)
4278{
4279 journal_t *journal;
4280 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4281 dev_t journal_dev;
4282 int err = 0;
4283 int really_read_only;
4284
0390131b
FM		 4285 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4286
ac27a0ec
DK		 4287 if (journal_devnum &&
4288 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
b31e1552
ES		 4289 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4290 "numbers have changed");
ac27a0ec
DK		 4291 journal_dev = new_decode_dev(journal_devnum);
4292 } else
4293 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4294
4295 really_read_only = bdev_read_only(sb->s_bdev);
4296
4297 /*
4298 * Are we loading a blank journal or performing recovery after a
4299 * crash? For recovery, we need to check in advance whether we
4300 * can get read-write access to the device.
4301 */
617ba13b 4302 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
ac27a0ec 4303 if (sb->s_flags & MS_RDONLY) {
b31e1552
ES		 4304 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4305 "required on readonly filesystem");
ac27a0ec 4306 if (really_read_only) {
b31e1552
ES		 4307 ext4_msg(sb, KERN_ERR, "write access "
4308 "unavailable, cannot proceed");
ac27a0ec
DK		 4309 return -EROFS;
4310 }
b31e1552
ES		 4311 ext4_msg(sb, KERN_INFO, "write access will "
4312 "be enabled during recovery");
ac27a0ec
DK		 4313 }
4314 }
4315
4316 if (journal_inum && journal_dev) {
b31e1552
ES		 4317 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4318 "and inode journals!");
ac27a0ec
DK		 4319 return -EINVAL;
4320 }
4321
4322 if (journal_inum) {
617ba13b 4323 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 4324 return -EINVAL;
4325 } else {
617ba13b 4326 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 4327 return -EINVAL;
4328 }
4329
90576c0b 4330 if (!(journal->j_flags & JBD2_BARRIER))
b31e1552 4331 ext4_msg(sb, KERN_INFO, "barriers disabled");
4776004f 4332
617ba13b 4333 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
dab291af 4334 err = jbd2_journal_wipe(journal, !really_read_only);
1c13d5c0
TT		 4335 if (!err) {
4336 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4337 if (save)
4338 memcpy(save, ((char *) es) +
4339 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
dab291af 4340 err = jbd2_journal_load(journal);
1c13d5c0
TT		 4341 if (save)
4342 memcpy(((char *) es) + EXT4_S_ERR_START,
4343 save, EXT4_S_ERR_LEN);
4344 kfree(save);
4345 }
ac27a0ec
DK		 4346
4347 if (err) {
b31e1552 4348 ext4_msg(sb, KERN_ERR, "error loading journal");
dab291af 4349 jbd2_journal_destroy(journal);
ac27a0ec
DK		 4350 return err;
4351 }
4352
617ba13b
MC		 4353 EXT4_SB(sb)->s_journal = journal;
4354 ext4_clear_journal_err(sb, es);
ac27a0ec 4355
c41303ce 4356 if (!really_read_only && journal_devnum &&
ac27a0ec
DK		 4357 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4358 es->s_journal_dev = cpu_to_le32(journal_devnum);
ac27a0ec
DK		 4359
4360 /* Make sure we flush the recovery flag to disk. */
e2d67052 4361 ext4_commit_super(sb, 1);
ac27a0ec
DK		 4362 }
4363
4364 return 0;
4365}
4366
e2d67052 4367static int ext4_commit_super(struct super_block *sb, int sync)
ac27a0ec 4368{
e2d67052 4369 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
617ba13b 4370 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
c4be0c1d 4371 int error = 0;
ac27a0ec 4372
7c2e7087 4373 if (!sbh || block_device_ejected(sb))
c4be0c1d 4374 return error;
914258bf
TT		 4375 if (buffer_write_io_error(sbh)) {
4376 /*
4377 * Oh, dear. A previous attempt to write the
4378 * superblock failed. This could happen because the
4379 * USB device was yanked out. Or it could happen to
4380 * be a transient write error and maybe the block will
4381 * be remapped. Nothing we can do but to retry the
4382 * write and hope for the best.
4383 */
b31e1552
ES		 4384 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4385 "superblock detected");
914258bf
TT		 4386 clear_buffer_write_io_error(sbh);
4387 set_buffer_uptodate(sbh);
4388 }
71290b36
TT		 4389 /*
4390 * If the file system is mounted read-only, don't update the
4391 * superblock write time. This avoids updating the superblock
4392 * write time when we are mounting the root file system
4393 * read/only but we need to replay the journal; at that point,
4394 * for people who are east of GMT and who make their clock
4395 * tick in localtime for Windows bug-for-bug compatibility,
4396 * the clock is set in the future, and this will cause e2fsck
4397 * to complain and force a full file system check.
4398 */
4399 if (!(sb->s_flags & MS_RDONLY))
4400 es->s_wtime = cpu_to_le32(get_seconds());
f613dfcb
TT		 4401 if (sb->s_bdev->bd_part)
4402 es->s_kbytes_written =
4403 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
afc32f7e
TT		 4404 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4405 EXT4_SB(sb)->s_sectors_written_start) >> 1));
f613dfcb
TT		 4406 else
4407 es->s_kbytes_written =
4408 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
57042651
TT		 4409 ext4_free_blocks_count_set(es,
4410 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4411 &EXT4_SB(sb)->s_freeclusters_counter)));
ce7e010a
TT		 4412 es->s_free_inodes_count =
4413 cpu_to_le32(percpu_counter_sum_positive(
4414 &EXT4_SB(sb)->s_freeinodes_counter));
ac27a0ec 4415 BUFFER_TRACE(sbh, "marking dirty");
06db49e6 4416 ext4_superblock_csum_set(sb);
ac27a0ec 4417 mark_buffer_dirty(sbh);
914258bf 4418 if (sync) {
c4be0c1d
TS		 4419 error = sync_dirty_buffer(sbh);
4420 if (error)
4421 return error;
4422
4423 error = buffer_write_io_error(sbh);
4424 if (error) {
b31e1552
ES		 4425 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4426 "superblock");
914258bf
TT		 4427 clear_buffer_write_io_error(sbh);
4428 set_buffer_uptodate(sbh);
4429 }
4430 }
c4be0c1d 4431 return error;
ac27a0ec
DK		 4432}
4433
ac27a0ec
DK		 4434/*
4435 * Have we just finished recovery? If so, and if we are mounting (or
4436 * remounting) the filesystem readonly, then we will end up with a
4437 * consistent fs on disk. Record that fact.
4438 */
2b2d6d01
TT		 4439static void ext4_mark_recovery_complete(struct super_block *sb,
4440 struct ext4_super_block *es)
ac27a0ec 4441{
617ba13b 4442 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 4443
0390131b
FM		 4444 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4445 BUG_ON(journal != NULL);
4446 return;
4447 }
dab291af 4448 jbd2_journal_lock_updates(journal);
7ffe1ea8
HK		 4449 if (jbd2_journal_flush(journal) < 0)
4450 goto out;
4451
617ba13b 4452 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
ac27a0ec 4453 sb->s_flags & MS_RDONLY) {
617ba13b 4454 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
e2d67052 4455 ext4_commit_super(sb, 1);
ac27a0ec 4456 }
7ffe1ea8
HK		 4457
4458out:
dab291af 4459 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 4460}
4461
4462/*
4463 * If we are mounting (or read-write remounting) a filesystem whose journal
4464 * has recorded an error from a previous lifetime, move that error to the
4465 * main filesystem now.
4466 */
2b2d6d01
TT		 4467static void ext4_clear_journal_err(struct super_block *sb,
4468 struct ext4_super_block *es)
ac27a0ec
DK		 4469{
4470 journal_t *journal;
4471 int j_errno;
4472 const char *errstr;
4473
0390131b
FM		 4474 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4475
617ba13b 4476 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 4477
4478 /*
4479 * Now check for any error status which may have been recorded in the
617ba13b 4480 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 4481 */
4482
dab291af 4483 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 4484 if (j_errno) {
4485 char nbuf[16];
4486
617ba13b 4487 errstr = ext4_decode_error(sb, j_errno, nbuf);
12062ddd 4488 ext4_warning(sb, "Filesystem error recorded "
ac27a0ec 4489 "from previous mount: %s", errstr);
12062ddd 4490 ext4_warning(sb, "Marking fs in need of filesystem check.");
ac27a0ec 4491
617ba13b
MC		 4492 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4493 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
e2d67052 4494 ext4_commit_super(sb, 1);
ac27a0ec 4495
dab291af 4496 jbd2_journal_clear_err(journal);
d796c52e 4497 jbd2_journal_update_sb_errno(journal);
ac27a0ec
DK		 4498 }
4499}
4500
4501/*
4502 * Force the running and committing transactions to commit,
4503 * and wait on the commit.
4504 */
617ba13b 4505int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 4506{
4507 journal_t *journal;
ac27a0ec
DK		 4508
4509 if (sb->s_flags & MS_RDONLY)
4510 return 0;
4511
617ba13b 4512 journal = EXT4_SB(sb)->s_journal;
b1deefc9 4513 return ext4_journal_force_commit(journal);
ac27a0ec
DK		 4514}
4515
617ba13b 4516static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec 4517{
14ce0cb4 4518 int ret = 0;
9eddacf9 4519 tid_t target;
8d5d02e6 4520 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4521
9bffad1e 4522 trace_ext4_sync_fs(sb, wait);
8d5d02e6 4523 flush_workqueue(sbi->dio_unwritten_wq);
a1177825
JK		 4524 /*
4525 * Writeback quota in non-journalled quota case - journalled quota has
4526 * no dirty dquots
4527 */
4528 dquot_writeback_dquots(sb, -1);
8d5d02e6 4529 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
9ca92389 4530 if (wait)
8d5d02e6 4531 jbd2_log_wait_commit(sbi->s_journal, target);
0390131b 4532 }
14ce0cb4 4533 return ret;
ac27a0ec
DK		 4534}
4535
4536/*
4537 * LVM calls this function before a (read-only) snapshot is created. This
4538 * gives us a chance to flush the journal completely and mark the fs clean.
be4f27d3
YY		 4539 *
4540 * Note that only this function cannot bring a filesystem to be in a clean
8e8ad8a5
JK		 4541 * state independently. It relies on upper layer to stop all data & metadata
4542 * modifications.
ac27a0ec 4543 */
c4be0c1d 4544static int ext4_freeze(struct super_block *sb)
ac27a0ec 4545{
c4be0c1d
TS		 4546 int error = 0;
4547 journal_t *journal;
ac27a0ec 4548
9ca92389
TT		 4549 if (sb->s_flags & MS_RDONLY)
4550 return 0;
ac27a0ec 4551
9ca92389 4552 journal = EXT4_SB(sb)->s_journal;
7ffe1ea8 4553
9ca92389
TT		 4554 /* Now we set up the journal barrier. */
4555 jbd2_journal_lock_updates(journal);
ac27a0ec 4556
9ca92389
TT		 4557 /*
4558 * Don't clear the needs_recovery flag if we failed to flush
4559 * the journal.
4560 */
4561 error = jbd2_journal_flush(journal);
6b0310fb
ES		 4562 if (error < 0)
4563 goto out;
9ca92389
TT		 4564
4565 /* Journal blocked and flushed, clear needs_recovery flag. */
4566 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4567 error = ext4_commit_super(sb, 1);
6b0310fb 4568out:
8e8ad8a5 4569 /* we rely on upper layer to stop further updates */
6b0310fb
ES		 4570 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4571 return error;
ac27a0ec
DK		 4572}
4573
4574/*
4575 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4576 * flag here, even though the filesystem is not technically dirty yet.
4577 */
c4be0c1d 4578static int ext4_unfreeze(struct super_block *sb)
ac27a0ec 4579{
9ca92389
TT		 4580 if (sb->s_flags & MS_RDONLY)
4581 return 0;
4582
9ca92389
TT		 4583 /* Reset the needs_recovery flag before the fs is unlocked. */
4584 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4585 ext4_commit_super(sb, 1);
c4be0c1d 4586 return 0;
ac27a0ec
DK		 4587}
4588
673c6100
TT		 4589/*
4590 * Structure to save mount options for ext4_remount's benefit
4591 */
4592struct ext4_mount_options {
4593 unsigned long s_mount_opt;
a2595b8a 4594 unsigned long s_mount_opt2;
08cefc7a
EB		 4595 kuid_t s_resuid;
4596 kgid_t s_resgid;
673c6100
TT		 4597 unsigned long s_commit_interval;
4598 u32 s_min_batch_time, s_max_batch_time;
4599#ifdef CONFIG_QUOTA
4600 int s_jquota_fmt;
4601 char *s_qf_names[MAXQUOTAS];
4602#endif
4603};
4604
2b2d6d01 4605static int ext4_remount(struct super_block *sb, int *flags, char *data)
ac27a0ec 4606{
2b2d6d01 4607 struct ext4_super_block *es;
617ba13b 4608 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4609 unsigned long old_sb_flags;
617ba13b 4610 struct ext4_mount_options old_opts;
c79d967d 4611 int enable_quota = 0;
8a266467 4612 ext4_group_t g;
b3881f74 4613 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
c5e06d10 4614 int err = 0;
ac27a0ec 4615#ifdef CONFIG_QUOTA
03dafb5f 4616 int i, j;
ac27a0ec 4617#endif
d4c402d9 4618 char *orig_data = kstrdup(data, GFP_KERNEL);
ac27a0ec
DK		 4619
4620 /* Store the original options */
4621 old_sb_flags = sb->s_flags;
4622 old_opts.s_mount_opt = sbi->s_mount_opt;
a2595b8a 4623 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
ac27a0ec
DK		 4624 old_opts.s_resuid = sbi->s_resuid;
4625 old_opts.s_resgid = sbi->s_resgid;
4626 old_opts.s_commit_interval = sbi->s_commit_interval;
30773840
TT		 4627 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4628 old_opts.s_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 4629#ifdef CONFIG_QUOTA
4630 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4631 for (i = 0; i < MAXQUOTAS; i++)
03dafb5f
CG		 4632 if (sbi->s_qf_names[i]) {
4633 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4634 GFP_KERNEL);
4635 if (!old_opts.s_qf_names[i]) {
4636 for (j = 0; j < i; j++)
4637 kfree(old_opts.s_qf_names[j]);
4638 return -ENOMEM;
4639 }
4640 } else
4641 old_opts.s_qf_names[i] = NULL;
ac27a0ec 4642#endif
b3881f74
TT		 4643 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4644 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
ac27a0ec
DK		 4645
4646 /*
4647 * Allow the "check" option to be passed as a remount option.
4648 */
661aa520 4649 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
ac27a0ec
DK		 4650 err = -EINVAL;
4651 goto restore_opts;
4652 }
4653
4ab2f15b 4654 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
c67d859e 4655 ext4_abort(sb, "Abort forced by user");
ac27a0ec
DK		 4656
4657 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 4658 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec
DK		 4659
4660 es = sbi->s_es;
4661
b3881f74 4662 if (sbi->s_journal) {
0390131b 4663 ext4_init_journal_params(sb, sbi->s_journal);
b3881f74
TT		 4664 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4665 }
ac27a0ec 4666
661aa520 4667 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4ab2f15b 4668 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
ac27a0ec
DK		 4669 err = -EROFS;
4670 goto restore_opts;
4671 }
4672
4673 if (*flags & MS_RDONLY) {
0f0dd62f
CH		 4674 err = dquot_suspend(sb, -1);
4675 if (err < 0)
c79d967d 4676 goto restore_opts;
c79d967d 4677
ac27a0ec
DK		 4678 /*
4679 * First of all, the unconditional stuff we have to do
4680 * to disable replay of the journal when we next remount
4681 */
4682 sb->s_flags |= MS_RDONLY;
4683
4684 /*
4685 * OK, test if we are remounting a valid rw partition
4686 * readonly, and if so set the rdonly flag and then
4687 * mark the partition as valid again.
4688 */
617ba13b
MC		 4689 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4690 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 4691 es->s_state = cpu_to_le16(sbi->s_mount_state);
4692
a63c9eb2 4693 if (sbi->s_journal)
0390131b 4694 ext4_mark_recovery_complete(sb, es);
ac27a0ec 4695 } else {
a13fb1a4
ES		 4696 /* Make sure we can mount this feature set readwrite */
4697 if (!ext4_feature_set_ok(sb, 0)) {
ac27a0ec
DK		 4698 err = -EROFS;
4699 goto restore_opts;
4700 }
8a266467
TT		 4701 /*
4702 * Make sure the group descriptor checksums
0b8e58a1 4703 * are sane. If they aren't, refuse to remount r/w.
8a266467
TT		 4704 */
4705 for (g = 0; g < sbi->s_groups_count; g++) {
4706 struct ext4_group_desc *gdp =
4707 ext4_get_group_desc(sb, g, NULL);
4708
feb0ab32 4709 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
b31e1552
ES		 4710 ext4_msg(sb, KERN_ERR,
4711 "ext4_remount: Checksum for group %u failed (%u!=%u)",
8a266467
TT		 4712 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4713 le16_to_cpu(gdp->bg_checksum));
4714 err = -EINVAL;
4715 goto restore_opts;
4716 }
4717 }
4718
ead6596b
ES		 4719 /*
4720 * If we have an unprocessed orphan list hanging
4721 * around from a previously readonly bdev mount,
4722 * require a full umount/remount for now.
4723 */
4724 if (es->s_last_orphan) {
b31e1552 4725 ext4_msg(sb, KERN_WARNING, "Couldn't "
ead6596b
ES		 4726 "remount RDWR because of unprocessed "
4727 "orphan inode list. Please "
b31e1552 4728 "umount/remount instead");
ead6596b
ES		 4729 err = -EINVAL;
4730 goto restore_opts;
4731 }
4732
ac27a0ec
DK		 4733 /*
4734 * Mounting a RDONLY partition read-write, so reread
4735 * and store the current valid flag. (It may have
4736 * been changed by e2fsck since we originally mounted
4737 * the partition.)
4738 */
0390131b
FM		 4739 if (sbi->s_journal)
4740 ext4_clear_journal_err(sb, es);
ac27a0ec 4741 sbi->s_mount_state = le16_to_cpu(es->s_state);
2b2d6d01 4742 if (!ext4_setup_super(sb, es, 0))
ac27a0ec 4743 sb->s_flags &= ~MS_RDONLY;
c5e06d10
JL		 4744 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
4745 EXT4_FEATURE_INCOMPAT_MMP))
4746 if (ext4_multi_mount_protect(sb,
4747 le64_to_cpu(es->s_mmp_block))) {
4748 err = -EROFS;
4749 goto restore_opts;
4750 }
c79d967d 4751 enable_quota = 1;
ac27a0ec
DK		 4752 }
4753 }
bfff6873
LC		 4754
4755 /*
4756 * Reinitialize lazy itable initialization thread based on
4757 * current settings
4758 */
4759 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4760 ext4_unregister_li_request(sb);
4761 else {
4762 ext4_group_t first_not_zeroed;
4763 first_not_zeroed = ext4_has_uninit_itable(sb);
4764 ext4_register_li_request(sb, first_not_zeroed);
4765 }
4766
6fd058f7 4767 ext4_setup_system_zone(sb);
d096ad0f 4768 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
e2d67052 4769 ext4_commit_super(sb, 1);
0390131b 4770
ac27a0ec
DK		 4771#ifdef CONFIG_QUOTA
4772 /* Release old quota file names */
4773 for (i = 0; i < MAXQUOTAS; i++)
03dafb5f 4774 kfree(old_opts.s_qf_names[i]);
7c319d32
AK		 4775 if (enable_quota) {
4776 if (sb_any_quota_suspended(sb))
4777 dquot_resume(sb, -1);
4778 else if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4779 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
4780 err = ext4_enable_quotas(sb);
07724f98 4781 if (err)
7c319d32 4782 goto restore_opts;
7c319d32
AK		 4783 }
4784 }
ac27a0ec 4785#endif
d4c402d9
CW		 4786
4787 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4788 kfree(orig_data);
ac27a0ec 4789 return 0;
0b8e58a1 4790
ac27a0ec
DK		 4791restore_opts:
4792 sb->s_flags = old_sb_flags;
4793 sbi->s_mount_opt = old_opts.s_mount_opt;
a2595b8a 4794 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
ac27a0ec
DK		 4795 sbi->s_resuid = old_opts.s_resuid;
4796 sbi->s_resgid = old_opts.s_resgid;
4797 sbi->s_commit_interval = old_opts.s_commit_interval;
30773840
TT		 4798 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4799 sbi->s_max_batch_time = old_opts.s_max_batch_time;
ac27a0ec
DK		 4800#ifdef CONFIG_QUOTA
4801 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4802 for (i = 0; i < MAXQUOTAS; i++) {
03dafb5f 4803 kfree(sbi->s_qf_names[i]);
ac27a0ec
DK		 4804 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4805 }
4806#endif
d4c402d9 4807 kfree(orig_data);
ac27a0ec
DK		 4808 return err;
4809}
4810
2b2d6d01 4811static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
ac27a0ec
DK		 4812{
4813 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 4814 struct ext4_sb_info *sbi = EXT4_SB(sb);
4815 struct ext4_super_block *es = sbi->s_es;
952fc18e 4816 ext4_fsblk_t overhead = 0;
960cc398 4817 u64 fsid;
d02a9391 4818 s64 bfree;
ac27a0ec 4819
952fc18e
TT		 4820 if (!test_opt(sb, MINIX_DF))
4821 overhead = sbi->s_overhead;
ac27a0ec 4822
617ba13b 4823 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 4824 buf->f_bsize = sb->s_blocksize;
b72f78cb 4825 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
57042651
TT		 4826 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
4827 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
d02a9391 4828 /* prevent underflow in case that few free space is available */
57042651 4829 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
bd81d8ee
LV		 4830 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
4831 if (buf->f_bfree < ext4_r_blocks_count(es))
ac27a0ec
DK		 4832 buf->f_bavail = 0;
4833 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 4834 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
617ba13b 4835 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 4836 fsid = le64_to_cpup((void *)es->s_uuid) ^
4837 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4838 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4839 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
0b8e58a1 4840
ac27a0ec
DK		 4841 return 0;
4842}
4843
0b8e58a1
AD		 4844/* Helper function for writing quotas on sync - we need to start transaction
4845 * before quota file is locked for write. Otherwise the are possible deadlocks:
ac27a0ec 4846 * Process 1 Process 2
617ba13b 4847 * ext4_create() quota_sync()
a269eb18 4848 * jbd2_journal_start() write_dquot()
871a2931 4849 * dquot_initialize() down(dqio_mutex)
dab291af 4850 * down(dqio_mutex) jbd2_journal_start()
ac27a0ec
DK		 4851 *
4852 */
4853
4854#ifdef CONFIG_QUOTA
4855
4856static inline struct inode *dquot_to_inode(struct dquot *dquot)
4857{
4c376dca 4858 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
ac27a0ec
DK		 4859}
4860
617ba13b 4861static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 4862{
4863 int ret, err;
4864 handle_t *handle;
4865 struct inode *inode;
4866
4867 inode = dquot_to_inode(dquot);
617ba13b 4868 handle = ext4_journal_start(inode,
0b8e58a1 4869 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 4870 if (IS_ERR(handle))
4871 return PTR_ERR(handle);
4872 ret = dquot_commit(dquot);
617ba13b 4873 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4874 if (!ret)
4875 ret = err;
4876 return ret;
4877}
4878
617ba13b 4879static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 4880{
4881 int ret, err;
4882 handle_t *handle;
4883
617ba13b 4884 handle = ext4_journal_start(dquot_to_inode(dquot),
0b8e58a1 4885 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 4886 if (IS_ERR(handle))
4887 return PTR_ERR(handle);
4888 ret = dquot_acquire(dquot);
617ba13b 4889 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4890 if (!ret)
4891 ret = err;
4892 return ret;
4893}
4894
617ba13b 4895static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 4896{
4897 int ret, err;
4898 handle_t *handle;
4899
617ba13b 4900 handle = ext4_journal_start(dquot_to_inode(dquot),
0b8e58a1 4901 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 4902 if (IS_ERR(handle)) {
4903 /* Release dquot anyway to avoid endless cycle in dqput() */
4904 dquot_release(dquot);
ac27a0ec 4905 return PTR_ERR(handle);
9c3013e9 4906 }
ac27a0ec 4907 ret = dquot_release(dquot);
617ba13b 4908 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4909 if (!ret)
4910 ret = err;
4911 return ret;
4912}
4913
617ba13b 4914static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec 4915{
2c8be6b2 4916 /* Are we journaling quotas? */
617ba13b
MC		 4917 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
4918 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
ac27a0ec 4919 dquot_mark_dquot_dirty(dquot);
617ba13b 4920 return ext4_write_dquot(dquot);
ac27a0ec
DK		 4921 } else {
4922 return dquot_mark_dquot_dirty(dquot);
4923 }
4924}
4925
617ba13b 4926static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 4927{
4928 int ret, err;
4929 handle_t *handle;
4930
4931 /* Data block + inode block */
617ba13b 4932 handle = ext4_journal_start(sb->s_root->d_inode, 2);
ac27a0ec
DK		 4933 if (IS_ERR(handle))
4934 return PTR_ERR(handle);
4935 ret = dquot_commit_info(sb, type);
617ba13b 4936 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4937 if (!ret)
4938 ret = err;
4939 return ret;
4940}
4941
4942/*
4943 * Turn on quotas during mount time - we need to find
4944 * the quota file and such...
4945 */
617ba13b 4946static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 4947{
287a8095
CH		 4948 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4949 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 4950}
4951
4952/*
4953 * Standard function to be called on quota_on
4954 */
617ba13b 4955static int ext4_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 4956 struct path *path)
ac27a0ec
DK		 4957{
4958 int err;
ac27a0ec
DK		 4959
4960 if (!test_opt(sb, QUOTA))
4961 return -EINVAL;
0623543b 4962
ac27a0ec 4963 /* Quotafile not on the same filesystem? */
d8c9584e 4964 if (path->dentry->d_sb != sb)
ac27a0ec 4965 return -EXDEV;
0623543b
JK		 4966 /* Journaling quota? */
4967 if (EXT4_SB(sb)->s_qf_names[type]) {
2b2d6d01 4968 /* Quotafile not in fs root? */
f00c9e44 4969 if (path->dentry->d_parent != sb->s_root)
b31e1552
ES		 4970 ext4_msg(sb, KERN_WARNING,
4971 "Quota file not on filesystem root. "
4972 "Journaled quota will not work");
2b2d6d01 4973 }
0623543b
JK		 4974
4975 /*
4976 * When we journal data on quota file, we have to flush journal to see
4977 * all updates to the file when we bypass pagecache...
4978 */
0390131b 4979 if (EXT4_SB(sb)->s_journal &&
f00c9e44 4980 ext4_should_journal_data(path->dentry->d_inode)) {
0623543b
JK		 4981 /*
4982 * We don't need to lock updates but journal_flush() could
4983 * otherwise be livelocked...
4984 */
4985 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 4986 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
0623543b 4987 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
f00c9e44 4988 if (err)
7ffe1ea8 4989 return err;
0623543b
JK		 4990 }
4991
f00c9e44 4992 return dquot_quota_on(sb, type, format_id, path);
ac27a0ec
DK		 4993}
4994
7c319d32
AK		 4995static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
4996 unsigned int flags)
4997{
4998 int err;
4999 struct inode *qf_inode;
5000 unsigned long qf_inums[MAXQUOTAS] = {
5001 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5002 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5003 };
5004
5005 BUG_ON(!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA));
5006
5007 if (!qf_inums[type])
5008 return -EPERM;
5009
5010 qf_inode = ext4_iget(sb, qf_inums[type]);
5011 if (IS_ERR(qf_inode)) {
5012 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5013 return PTR_ERR(qf_inode);
5014 }
5015
5016 err = dquot_enable(qf_inode, type, format_id, flags);
5017 iput(qf_inode);
5018
5019 return err;
5020}
5021
5022/* Enable usage tracking for all quota types. */
5023static int ext4_enable_quotas(struct super_block *sb)
5024{
5025 int type, err = 0;
5026 unsigned long qf_inums[MAXQUOTAS] = {
5027 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5028 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5029 };
5030
5031 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
5032 for (type = 0; type < MAXQUOTAS; type++) {
5033 if (qf_inums[type]) {
5034 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5035 DQUOT_USAGE_ENABLED);
5036 if (err) {
5037 ext4_warning(sb,
72ba7450
TT		 5038 "Failed to enable quota tracking "
5039 "(type=%d, err=%d). Please run "
5040 "e2fsck to fix.", type, err);
7c319d32
AK		 5041 return err;
5042 }
5043 }
5044 }
5045 return 0;
5046}
5047
5048/*
5049 * quota_on function that is used when QUOTA feature is set.
5050 */
5051static int ext4_quota_on_sysfile(struct super_block *sb, int type,
5052 int format_id)
5053{
5054 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5055 return -EINVAL;
5056
5057 /*
5058 * USAGE was enabled at mount time. Only need to enable LIMITS now.
5059 */
5060 return ext4_quota_enable(sb, type, format_id, DQUOT_LIMITS_ENABLED);
5061}
5062
ca0e05e4
DM		 5063static int ext4_quota_off(struct super_block *sb, int type)
5064{
21f97697
JK		 5065 struct inode *inode = sb_dqopt(sb)->files[type];
5066 handle_t *handle;
5067
87009d86
DM		 5068 /* Force all delayed allocation blocks to be allocated.
5069 * Caller already holds s_umount sem */
5070 if (test_opt(sb, DELALLOC))
ca0e05e4 5071 sync_filesystem(sb);
ca0e05e4 5072
0b268590
AG		 5073 if (!inode)
5074 goto out;
5075
21f97697
JK		 5076 /* Update modification times of quota files when userspace can
5077 * start looking at them */
5078 handle = ext4_journal_start(inode, 1);
5079 if (IS_ERR(handle))
5080 goto out;
5081 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
5082 ext4_mark_inode_dirty(handle, inode);
5083 ext4_journal_stop(handle);
5084
5085out:
ca0e05e4
DM		 5086 return dquot_quota_off(sb, type);
5087}
5088
7c319d32
AK		 5089/*
5090 * quota_off function that is used when QUOTA feature is set.
5091 */
5092static int ext4_quota_off_sysfile(struct super_block *sb, int type)
5093{
5094 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5095 return -EINVAL;
5096
5097 /* Disable only the limits. */
5098 return dquot_disable(sb, type, DQUOT_LIMITS_ENABLED);
5099}
5100
ac27a0ec
DK		 5101/* Read data from quotafile - avoid pagecache and such because we cannot afford
5102 * acquiring the locks... As quota files are never truncated and quota code
25985edc 5103 * itself serializes the operations (and no one else should touch the files)
ac27a0ec 5104 * we don't have to be afraid of races */
617ba13b 5105static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 5106 size_t len, loff_t off)
5107{
5108 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 5109 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 5110 int err = 0;
5111 int offset = off & (sb->s_blocksize - 1);
5112 int tocopy;
5113 size_t toread;
5114 struct buffer_head *bh;
5115 loff_t i_size = i_size_read(inode);
5116
5117 if (off > i_size)
5118 return 0;
5119 if (off+len > i_size)
5120 len = i_size-off;
5121 toread = len;
5122 while (toread > 0) {
5123 tocopy = sb->s_blocksize - offset < toread ?
5124 sb->s_blocksize - offset : toread;
617ba13b 5125 bh = ext4_bread(NULL, inode, blk, 0, &err);
ac27a0ec
DK		 5126 if (err)
5127 return err;
5128 if (!bh) /* A hole? */
5129 memset(data, 0, tocopy);
5130 else
5131 memcpy(data, bh->b_data+offset, tocopy);
5132 brelse(bh);
5133 offset = 0;
5134 toread -= tocopy;
5135 data += tocopy;
5136 blk++;
5137 }
5138 return len;
5139}
5140
5141/* Write to quotafile (we know the transaction is already started and has
5142 * enough credits) */
617ba13b 5143static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 5144 const char *data, size_t len, loff_t off)
5145{
5146 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 5147 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 5148 int err = 0;
5149 int offset = off & (sb->s_blocksize - 1);
ac27a0ec
DK		 5150 struct buffer_head *bh;
5151 handle_t *handle = journal_current_handle();
5152
0390131b 5153 if (EXT4_SB(sb)->s_journal && !handle) {
b31e1552
ES		 5154 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5155 " cancelled because transaction is not started",
9c3013e9
JK		 5156 (unsigned long long)off, (unsigned long long)len);
5157 return -EIO;
5158 }
67eeb568
DM		 5159 /*
5160 * Since we account only one data block in transaction credits,
5161 * then it is impossible to cross a block boundary.
5162 */
5163 if (sb->s_blocksize - offset < len) {
5164 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5165 " cancelled because not block aligned",
5166 (unsigned long long)off, (unsigned long long)len);
5167 return -EIO;
5168 }
5169
67eeb568
DM		 5170 bh = ext4_bread(handle, inode, blk, 1, &err);
5171 if (!bh)
5172 goto out;
62d2b5f2
JK		 5173 err = ext4_journal_get_write_access(handle, bh);
5174 if (err) {
5175 brelse(bh);
5176 goto out;
ac27a0ec 5177 }
67eeb568
DM		 5178 lock_buffer(bh);
5179 memcpy(bh->b_data+offset, data, len);
5180 flush_dcache_page(bh->b_page);
5181 unlock_buffer(bh);
62d2b5f2 5182 err = ext4_handle_dirty_metadata(handle, NULL, bh);
67eeb568 5183 brelse(bh);
ac27a0ec 5184out:
0b7f7cef 5185 if (err)
ac27a0ec 5186 return err;
67eeb568
DM		 5187 if (inode->i_size < off + len) {
5188 i_size_write(inode, off + len);
617ba13b 5189 EXT4_I(inode)->i_disksize = inode->i_size;
21f97697 5190 ext4_mark_inode_dirty(handle, inode);
ac27a0ec 5191 }
67eeb568 5192 return len;
ac27a0ec
DK		 5193}
5194
5195#endif
5196
152a0836
AV		 5197static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5198 const char *dev_name, void *data)
ac27a0ec 5199{
152a0836 5200 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
ac27a0ec
DK		 5201}
5202
37f328eb 5203#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
24b58424
TT		 5204static inline void register_as_ext2(void)
5205{
5206 int err = register_filesystem(&ext2_fs_type);
5207 if (err)
5208 printk(KERN_WARNING
5209 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5210}
5211
5212static inline void unregister_as_ext2(void)
5213{
5214 unregister_filesystem(&ext2_fs_type);
5215}
2035e776
TT		 5216
5217static inline int ext2_feature_set_ok(struct super_block *sb)
5218{
5219 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
5220 return 0;
5221 if (sb->s_flags & MS_RDONLY)
5222 return 1;
5223 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
5224 return 0;
5225 return 1;
5226}
51b7e3c9 5227MODULE_ALIAS("ext2");
24b58424
TT		 5228#else
5229static inline void register_as_ext2(void) { }
5230static inline void unregister_as_ext2(void) { }
2035e776 5231static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 5232#endif
5233
37f328eb 5234#if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
24b58424
TT		 5235static inline void register_as_ext3(void)
5236{
5237 int err = register_filesystem(&ext3_fs_type);
5238 if (err)
5239 printk(KERN_WARNING
5240 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5241}
5242
5243static inline void unregister_as_ext3(void)
5244{
5245 unregister_filesystem(&ext3_fs_type);
5246}
2035e776
TT		 5247
5248static inline int ext3_feature_set_ok(struct super_block *sb)
5249{
5250 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
5251 return 0;
5252 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
5253 return 0;
5254 if (sb->s_flags & MS_RDONLY)
5255 return 1;
5256 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
5257 return 0;
5258 return 1;
5259}
51b7e3c9 5260MODULE_ALIAS("ext3");
24b58424
TT		 5261#else
5262static inline void register_as_ext3(void) { }
5263static inline void unregister_as_ext3(void) { }
2035e776 5264static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 5265#endif
5266
03010a33
TT		 5267static struct file_system_type ext4_fs_type = {
5268 .owner = THIS_MODULE,
5269 .name = "ext4",
152a0836 5270 .mount = ext4_mount,
03010a33
TT		 5271 .kill_sb = kill_block_super,
5272 .fs_flags = FS_REQUIRES_DEV,
5273};
5274
8f021222 5275static int __init ext4_init_feat_adverts(void)
857ac889
LC		 5276{
5277 struct ext4_features *ef;
5278 int ret = -ENOMEM;
5279
5280 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
5281 if (!ef)
5282 goto out;
5283
5284 ef->f_kobj.kset = ext4_kset;
5285 init_completion(&ef->f_kobj_unregister);
5286 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
5287 "features");
5288 if (ret) {
5289 kfree(ef);
5290 goto out;
5291 }
5292
5293 ext4_feat = ef;
5294 ret = 0;
5295out:
5296 return ret;
5297}
5298
8f021222
LC		 5299static void ext4_exit_feat_adverts(void)
5300{
5301 kobject_put(&ext4_feat->f_kobj);
5302 wait_for_completion(&ext4_feat->f_kobj_unregister);
5303 kfree(ext4_feat);
5304}
5305
e9e3bcec
ES		 5306/* Shared across all ext4 file systems */
5307wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5308struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5309
5dabfc78 5310static int __init ext4_init_fs(void)
ac27a0ec 5311{
e9e3bcec 5312 int i, err;
c9de560d 5313
07c0c5d8
AV		 5314 ext4_li_info = NULL;
5315 mutex_init(&ext4_li_mtx);
5316
9a4c8019 5317 /* Build-time check for flags consistency */
12e9b892 5318 ext4_check_flag_values();
e9e3bcec
ES		 5319
5320 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5321 mutex_init(&ext4__aio_mutex[i]);
5322 init_waitqueue_head(&ext4__ioend_wq[i]);
5323 }
5324
51865fda 5325 err = ext4_init_es();
6fd058f7
TT		 5326 if (err)
5327 return err;
51865fda
ZL		 5328
5329 err = ext4_init_pageio();
5330 if (err)
5331 goto out7;
5332
5dabfc78 5333 err = ext4_init_system_zone();
bd2d0210 5334 if (err)
d44651d0 5335 goto out6;
3197ebdb 5336 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
0e376b1e
TT		 5337 if (!ext4_kset) {
5338 err = -ENOMEM;
dd68314c 5339 goto out5;
0e376b1e 5340 }
d44651d0 5341 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
857ac889
LC		 5342
5343 err = ext4_init_feat_adverts();
dd68314c
TT		 5344 if (err)
5345 goto out4;
857ac889 5346
5dabfc78 5347 err = ext4_init_mballoc();
ac27a0ec 5348 if (err)
6fd058f7 5349 goto out3;
c9de560d 5350
5dabfc78 5351 err = ext4_init_xattr();
c9de560d
AT		 5352 if (err)
5353 goto out2;
ac27a0ec
DK		 5354 err = init_inodecache();
5355 if (err)
5356 goto out1;
24b58424 5357 register_as_ext3();
2035e776 5358 register_as_ext2();
03010a33 5359 err = register_filesystem(&ext4_fs_type);
ac27a0ec
DK		 5360 if (err)
5361 goto out;
bfff6873 5362
ac27a0ec
DK		 5363 return 0;
5364out:
24b58424
TT		 5365 unregister_as_ext2();
5366 unregister_as_ext3();
ac27a0ec
DK		 5367 destroy_inodecache();
5368out1:
5dabfc78 5369 ext4_exit_xattr();
c9de560d 5370out2:
5dabfc78 5371 ext4_exit_mballoc();
6fd058f7 5372out3:
8f021222 5373 ext4_exit_feat_adverts();
dd68314c 5374out4:
d44651d0
FJ		 5375 if (ext4_proc_root)
5376 remove_proc_entry("fs/ext4", NULL);
6fd058f7 5377 kset_unregister(ext4_kset);
d44651d0 5378out5:
5dabfc78 5379 ext4_exit_system_zone();
d44651d0 5380out6:
5dabfc78 5381 ext4_exit_pageio();
51865fda
ZL		 5382out7:
5383 ext4_exit_es();
5384
ac27a0ec
DK		 5385 return err;
5386}
5387
5dabfc78 5388static void __exit ext4_exit_fs(void)
ac27a0ec 5389{
bfff6873 5390 ext4_destroy_lazyinit_thread();
24b58424
TT		 5391 unregister_as_ext2();
5392 unregister_as_ext3();
03010a33 5393 unregister_filesystem(&ext4_fs_type);
ac27a0ec 5394 destroy_inodecache();
5dabfc78
TT		 5395 ext4_exit_xattr();
5396 ext4_exit_mballoc();
8f021222 5397 ext4_exit_feat_adverts();
9f6200bb 5398 remove_proc_entry("fs/ext4", NULL);
3197ebdb 5399 kset_unregister(ext4_kset);
5dabfc78
TT		 5400 ext4_exit_system_zone();
5401 ext4_exit_pageio();
ac27a0ec
DK		 5402}
5403
5404MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
83982b6f 5405MODULE_DESCRIPTION("Fourth Extended Filesystem");
ac27a0ec 5406MODULE_LICENSE("GPL");
5dabfc78
TT		 5407module_init(ext4_init_fs)
5408module_exit(ext4_exit_fs)
kernel source for telekom puls (alcatel/mediatek)