projects / GitHub / mt8127 / android_kernel_alcatel_ttab.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
ext4: add punching hole support for non-extent-mapped files
[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);
0b8e58a1 963
ac27a0ec
DK		 964 return &ei->vfs_inode;
965}
966
7ff9c073
TT		 967static int ext4_drop_inode(struct inode *inode)
968{
969 int drop = generic_drop_inode(inode);
970
971 trace_ext4_drop_inode(inode, drop);
972 return drop;
973}
974
fa0d7e3d
NP		 975static void ext4_i_callback(struct rcu_head *head)
976{
977 struct inode *inode = container_of(head, struct inode, i_rcu);
fa0d7e3d
NP		 978 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
979}
980
617ba13b 981static void ext4_destroy_inode(struct inode *inode)
ac27a0ec 982{
9f7dd93d 983 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
b31e1552
ES		 984 ext4_msg(inode->i_sb, KERN_ERR,
985 "Inode %lu (%p): orphan list check failed!",
986 inode->i_ino, EXT4_I(inode));
9f7dd93d
VA		 987 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
988 EXT4_I(inode), sizeof(struct ext4_inode_info),
989 true);
990 dump_stack();
991 }
fa0d7e3d 992 call_rcu(&inode->i_rcu, ext4_i_callback);
ac27a0ec
DK		 993}
994
51cc5068 995static void init_once(void *foo)
ac27a0ec 996{
617ba13b 997 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
ac27a0ec 998
a35afb83 999 INIT_LIST_HEAD(&ei->i_orphan);
a35afb83 1000 init_rwsem(&ei->xattr_sem);
0e855ac8 1001 init_rwsem(&ei->i_data_sem);
a35afb83 1002 inode_init_once(&ei->vfs_inode);
ac27a0ec
DK		 1003}
1004
1005static int init_inodecache(void)
1006{
617ba13b
MC		 1007 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
1008 sizeof(struct ext4_inode_info),
ac27a0ec
DK		 1009 0, (SLAB_RECLAIM_ACCOUNT|
1010 SLAB_MEM_SPREAD),
20c2df83 1011 init_once);
617ba13b 1012 if (ext4_inode_cachep == NULL)
ac27a0ec
DK		 1013 return -ENOMEM;
1014 return 0;
1015}
1016
1017static void destroy_inodecache(void)
1018{
8c0a8537
KS		 1019 /*
1020 * Make sure all delayed rcu free inodes are flushed before we
1021 * destroy cache.
1022 */
1023 rcu_barrier();
617ba13b 1024 kmem_cache_destroy(ext4_inode_cachep);
ac27a0ec
DK		 1025}
1026
0930fcc1 1027void ext4_clear_inode(struct inode *inode)
ac27a0ec 1028{
0930fcc1 1029 invalidate_inode_buffers(inode);
dbd5768f 1030 clear_inode(inode);
9f754758 1031 dquot_drop(inode);
c2ea3fde 1032 ext4_discard_preallocations(inode);
51865fda 1033 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
8aefcd55
TT		 1034 if (EXT4_I(inode)->jinode) {
1035 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1036 EXT4_I(inode)->jinode);
1037 jbd2_free_inode(EXT4_I(inode)->jinode);
1038 EXT4_I(inode)->jinode = NULL;
1039 }
ac27a0ec
DK		 1040}
1041
1b961ac0 1042static struct inode *ext4_nfs_get_inode(struct super_block *sb,
0b8e58a1 1043 u64 ino, u32 generation)
ac27a0ec 1044{
ac27a0ec 1045 struct inode *inode;
ac27a0ec 1046
617ba13b 1047 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
ac27a0ec 1048 return ERR_PTR(-ESTALE);
617ba13b 1049 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
ac27a0ec
DK		 1050 return ERR_PTR(-ESTALE);
1051
1052 /* iget isn't really right if the inode is currently unallocated!!
1053 *
617ba13b 1054 * ext4_read_inode will return a bad_inode if the inode had been
ac27a0ec
DK		 1055 * deleted, so we should be safe.
1056 *
1057 * Currently we don't know the generation for parent directory, so
1058 * a generation of 0 means "accept any"
1059 */
1d1fe1ee
DH		 1060 inode = ext4_iget(sb, ino);
1061 if (IS_ERR(inode))
1062 return ERR_CAST(inode);
1063 if (generation && inode->i_generation != generation) {
ac27a0ec
DK		 1064 iput(inode);
1065 return ERR_PTR(-ESTALE);
1066 }
1b961ac0
CH		 1067
1068 return inode;
1069}
1070
1071static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
0b8e58a1 1072 int fh_len, int fh_type)
1b961ac0
CH		 1073{
1074 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1075 ext4_nfs_get_inode);
1076}
1077
1078static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
0b8e58a1 1079 int fh_len, int fh_type)
1b961ac0
CH		 1080{
1081 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1082 ext4_nfs_get_inode);
ac27a0ec
DK		 1083}
1084
c39a7f84
TO		 1085/*
1086 * Try to release metadata pages (indirect blocks, directories) which are
1087 * mapped via the block device. Since these pages could have journal heads
1088 * which would prevent try_to_free_buffers() from freeing them, we must use
1089 * jbd2 layer's try_to_free_buffers() function to release them.
1090 */
0b8e58a1
AD		 1091static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1092 gfp_t wait)
c39a7f84
TO		 1093{
1094 journal_t *journal = EXT4_SB(sb)->s_journal;
1095
1096 WARN_ON(PageChecked(page));
1097 if (!page_has_buffers(page))
1098 return 0;
1099 if (journal)
1100 return jbd2_journal_try_to_free_buffers(journal, page,
1101 wait & ~__GFP_WAIT);
1102 return try_to_free_buffers(page);
1103}
1104
ac27a0ec 1105#ifdef CONFIG_QUOTA
af5bc92d 1106#define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
2b2d6d01 1107#define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
ac27a0ec 1108
617ba13b
MC		 1109static int ext4_write_dquot(struct dquot *dquot);
1110static int ext4_acquire_dquot(struct dquot *dquot);
1111static int ext4_release_dquot(struct dquot *dquot);
1112static int ext4_mark_dquot_dirty(struct dquot *dquot);
1113static int ext4_write_info(struct super_block *sb, int type);
6f28e087 1114static int ext4_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 1115 struct path *path);
7c319d32
AK		 1116static int ext4_quota_on_sysfile(struct super_block *sb, int type,
1117 int format_id);
ca0e05e4 1118static int ext4_quota_off(struct super_block *sb, int type);
7c319d32 1119static int ext4_quota_off_sysfile(struct super_block *sb, int type);
617ba13b
MC		 1120static int ext4_quota_on_mount(struct super_block *sb, int type);
1121static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 1122 size_t len, loff_t off);
617ba13b 1123static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec 1124 const char *data, size_t len, loff_t off);
7c319d32
AK		 1125static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1126 unsigned int flags);
1127static int ext4_enable_quotas(struct super_block *sb);
ac27a0ec 1128
61e225dc 1129static const struct dquot_operations ext4_quota_operations = {
60e58e0f 1130 .get_reserved_space = ext4_get_reserved_space,
617ba13b
MC		 1131 .write_dquot = ext4_write_dquot,
1132 .acquire_dquot = ext4_acquire_dquot,
1133 .release_dquot = ext4_release_dquot,
1134 .mark_dirty = ext4_mark_dquot_dirty,
a5b5ee32
JK		 1135 .write_info = ext4_write_info,
1136 .alloc_dquot = dquot_alloc,
1137 .destroy_dquot = dquot_destroy,
ac27a0ec
DK		 1138};
1139
0d54b217 1140static const struct quotactl_ops ext4_qctl_operations = {
617ba13b 1141 .quota_on = ext4_quota_on,
ca0e05e4 1142 .quota_off = ext4_quota_off,
287a8095
CH		 1143 .quota_sync = dquot_quota_sync,
1144 .get_info = dquot_get_dqinfo,
1145 .set_info = dquot_set_dqinfo,
1146 .get_dqblk = dquot_get_dqblk,
1147 .set_dqblk = dquot_set_dqblk
ac27a0ec 1148};
7c319d32
AK		 1149
1150static const struct quotactl_ops ext4_qctl_sysfile_operations = {
1151 .quota_on_meta = ext4_quota_on_sysfile,
1152 .quota_off = ext4_quota_off_sysfile,
1153 .quota_sync = dquot_quota_sync,
1154 .get_info = dquot_get_dqinfo,
1155 .set_info = dquot_set_dqinfo,
1156 .get_dqblk = dquot_get_dqblk,
1157 .set_dqblk = dquot_set_dqblk
1158};
ac27a0ec
DK		 1159#endif
1160
ee9b6d61 1161static const struct super_operations ext4_sops = {
617ba13b
MC		 1162 .alloc_inode = ext4_alloc_inode,
1163 .destroy_inode = ext4_destroy_inode,
617ba13b
MC		 1164 .write_inode = ext4_write_inode,
1165 .dirty_inode = ext4_dirty_inode,
7ff9c073 1166 .drop_inode = ext4_drop_inode,
0930fcc1 1167 .evict_inode = ext4_evict_inode,
617ba13b 1168 .put_super = ext4_put_super,
617ba13b 1169 .sync_fs = ext4_sync_fs,
c4be0c1d
TS		 1170 .freeze_fs = ext4_freeze,
1171 .unfreeze_fs = ext4_unfreeze,
617ba13b
MC		 1172 .statfs = ext4_statfs,
1173 .remount_fs = ext4_remount,
617ba13b 1174 .show_options = ext4_show_options,
ac27a0ec 1175#ifdef CONFIG_QUOTA
617ba13b
MC		 1176 .quota_read = ext4_quota_read,
1177 .quota_write = ext4_quota_write,
ac27a0ec 1178#endif
c39a7f84 1179 .bdev_try_to_free_page = bdev_try_to_free_page,
ac27a0ec
DK		 1180};
1181
9ca92389
TT		 1182static const struct super_operations ext4_nojournal_sops = {
1183 .alloc_inode = ext4_alloc_inode,
1184 .destroy_inode = ext4_destroy_inode,
1185 .write_inode = ext4_write_inode,
1186 .dirty_inode = ext4_dirty_inode,
7ff9c073 1187 .drop_inode = ext4_drop_inode,
0930fcc1 1188 .evict_inode = ext4_evict_inode,
9ca92389
TT		 1189 .put_super = ext4_put_super,
1190 .statfs = ext4_statfs,
1191 .remount_fs = ext4_remount,
9ca92389
TT		 1192 .show_options = ext4_show_options,
1193#ifdef CONFIG_QUOTA
1194 .quota_read = ext4_quota_read,
1195 .quota_write = ext4_quota_write,
1196#endif
1197 .bdev_try_to_free_page = bdev_try_to_free_page,
1198};
1199
39655164 1200static const struct export_operations ext4_export_ops = {
1b961ac0
CH		 1201 .fh_to_dentry = ext4_fh_to_dentry,
1202 .fh_to_parent = ext4_fh_to_parent,
617ba13b 1203 .get_parent = ext4_get_parent,
ac27a0ec
DK		 1204};
1205
1206enum {
1207 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1208 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
72578c33 1209 Opt_nouid32, Opt_debug, Opt_removed,
ac27a0ec 1210 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
72578c33 1211 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
30773840 1212 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
43e625d8 1213 Opt_journal_dev, Opt_journal_checksum, Opt_journal_async_commit,
ac27a0ec 1214 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
296c355c 1215 Opt_data_err_abort, Opt_data_err_ignore,
ac27a0ec 1216 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
5a20bdfc 1217 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
ee4a3fcd 1218 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
661aa520 1219 Opt_usrquota, Opt_grpquota, Opt_i_version,
1449032b
TT		 1220 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1221 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
5328e635 1222 Opt_inode_readahead_blks, Opt_journal_ioprio,
744692dc 1223 Opt_dioread_nolock, Opt_dioread_lock,
fc6cb1cd 1224 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
df981d03 1225 Opt_max_dir_size_kb,
ac27a0ec
DK		 1226};
1227
a447c093 1228static const match_table_t tokens = {
ac27a0ec
DK		 1229 {Opt_bsd_df, "bsddf"},
1230 {Opt_minix_df, "minixdf"},
1231 {Opt_grpid, "grpid"},
1232 {Opt_grpid, "bsdgroups"},
1233 {Opt_nogrpid, "nogrpid"},
1234 {Opt_nogrpid, "sysvgroups"},
1235 {Opt_resgid, "resgid=%u"},
1236 {Opt_resuid, "resuid=%u"},
1237 {Opt_sb, "sb=%u"},
1238 {Opt_err_cont, "errors=continue"},
1239 {Opt_err_panic, "errors=panic"},
1240 {Opt_err_ro, "errors=remount-ro"},
1241 {Opt_nouid32, "nouid32"},
ac27a0ec 1242 {Opt_debug, "debug"},
72578c33
TT		 1243 {Opt_removed, "oldalloc"},
1244 {Opt_removed, "orlov"},
ac27a0ec
DK		 1245 {Opt_user_xattr, "user_xattr"},
1246 {Opt_nouser_xattr, "nouser_xattr"},
1247 {Opt_acl, "acl"},
1248 {Opt_noacl, "noacl"},
e3bb52ae 1249 {Opt_noload, "norecovery"},
5a916be1 1250 {Opt_noload, "noload"},
72578c33
TT		 1251 {Opt_removed, "nobh"},
1252 {Opt_removed, "bh"},
ac27a0ec 1253 {Opt_commit, "commit=%u"},
30773840
TT		 1254 {Opt_min_batch_time, "min_batch_time=%u"},
1255 {Opt_max_batch_time, "max_batch_time=%u"},
ac27a0ec 1256 {Opt_journal_dev, "journal_dev=%u"},
818d276c
GS		 1257 {Opt_journal_checksum, "journal_checksum"},
1258 {Opt_journal_async_commit, "journal_async_commit"},
ac27a0ec
DK		 1259 {Opt_abort, "abort"},
1260 {Opt_data_journal, "data=journal"},
1261 {Opt_data_ordered, "data=ordered"},
1262 {Opt_data_writeback, "data=writeback"},
5bf5683a
HK		 1263 {Opt_data_err_abort, "data_err=abort"},
1264 {Opt_data_err_ignore, "data_err=ignore"},
ac27a0ec
DK		 1265 {Opt_offusrjquota, "usrjquota="},
1266 {Opt_usrjquota, "usrjquota=%s"},
1267 {Opt_offgrpjquota, "grpjquota="},
1268 {Opt_grpjquota, "grpjquota=%s"},
1269 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1270 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
5a20bdfc 1271 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
ac27a0ec
DK		 1272 {Opt_grpquota, "grpquota"},
1273 {Opt_noquota, "noquota"},
1274 {Opt_quota, "quota"},
1275 {Opt_usrquota, "usrquota"},
1276 {Opt_barrier, "barrier=%u"},
06705bff
TT		 1277 {Opt_barrier, "barrier"},
1278 {Opt_nobarrier, "nobarrier"},
25ec56b5 1279 {Opt_i_version, "i_version"},
c9de560d 1280 {Opt_stripe, "stripe=%u"},
64769240 1281 {Opt_delalloc, "delalloc"},
dd919b98 1282 {Opt_nodelalloc, "nodelalloc"},
1449032b
TT		 1283 {Opt_mblk_io_submit, "mblk_io_submit"},
1284 {Opt_nomblk_io_submit, "nomblk_io_submit"},
6fd058f7
TT		 1285 {Opt_block_validity, "block_validity"},
1286 {Opt_noblock_validity, "noblock_validity"},
240799cd 1287 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
b3881f74 1288 {Opt_journal_ioprio, "journal_ioprio=%u"},
afd4672d 1289 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
06705bff
TT		 1290 {Opt_auto_da_alloc, "auto_da_alloc"},
1291 {Opt_noauto_da_alloc, "noauto_da_alloc"},
744692dc
JZ		 1292 {Opt_dioread_nolock, "dioread_nolock"},
1293 {Opt_dioread_lock, "dioread_lock"},
5328e635
ES		 1294 {Opt_discard, "discard"},
1295 {Opt_nodiscard, "nodiscard"},
fc6cb1cd
TT		 1296 {Opt_init_itable, "init_itable=%u"},
1297 {Opt_init_itable, "init_itable"},
1298 {Opt_noinit_itable, "noinit_itable"},
df981d03 1299 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
c7198b9c
TT		 1300 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1301 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1302 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1303 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1304 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
f3f12faa 1305 {Opt_err, NULL},
ac27a0ec
DK		 1306};
1307
617ba13b 1308static ext4_fsblk_t get_sb_block(void **data)
ac27a0ec 1309{
617ba13b 1310 ext4_fsblk_t sb_block;
ac27a0ec
DK		 1311 char *options = (char *) *data;
1312
1313 if (!options || strncmp(options, "sb=", 3) != 0)
1314 return 1; /* Default location */
0b8e58a1 1315
ac27a0ec 1316 options += 3;
0b8e58a1 1317 /* TODO: use simple_strtoll with >32bit ext4 */
ac27a0ec
DK		 1318 sb_block = simple_strtoul(options, &options, 0);
1319 if (*options && *options != ',') {
4776004f 1320 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
ac27a0ec
DK		 1321 (char *) *data);
1322 return 1;
1323 }
1324 if (*options == ',')
1325 options++;
1326 *data = (void *) options;
0b8e58a1 1327
ac27a0ec
DK		 1328 return sb_block;
1329}
1330
b3881f74 1331#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
437ca0fd
DM		 1332static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1333 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
b3881f74 1334
56c50f11
DM		 1335#ifdef CONFIG_QUOTA
1336static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1337{
1338 struct ext4_sb_info *sbi = EXT4_SB(sb);
1339 char *qname;
03dafb5f 1340 int ret = -1;
56c50f11
DM		 1341
1342 if (sb_any_quota_loaded(sb) &&
1343 !sbi->s_qf_names[qtype]) {
1344 ext4_msg(sb, KERN_ERR,
1345 "Cannot change journaled "
1346 "quota options when quota turned on");
57f73c2c 1347 return -1;
56c50f11
DM		 1348 }
1349 qname = match_strdup(args);
1350 if (!qname) {
1351 ext4_msg(sb, KERN_ERR,
1352 "Not enough memory for storing quotafile name");
57f73c2c 1353 return -1;
56c50f11 1354 }
03dafb5f
CG		 1355 if (sbi->s_qf_names[qtype]) {
1356 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1357 ret = 1;
1358 else
1359 ext4_msg(sb, KERN_ERR,
1360 "%s quota file already specified",
1361 QTYPE2NAME(qtype));
1362 goto errout;
56c50f11 1363 }
03dafb5f 1364 if (strchr(qname, '/')) {
56c50f11
DM		 1365 ext4_msg(sb, KERN_ERR,
1366 "quotafile must be on filesystem root");
03dafb5f 1367 goto errout;
56c50f11 1368 }
03dafb5f 1369 sbi->s_qf_names[qtype] = qname;
fd8c37ec 1370 set_opt(sb, QUOTA);
56c50f11 1371 return 1;
03dafb5f
CG		 1372errout:
1373 kfree(qname);
1374 return ret;
56c50f11
DM		 1375}
1376
1377static int clear_qf_name(struct super_block *sb, int qtype)
1378{
1379
1380 struct ext4_sb_info *sbi = EXT4_SB(sb);
1381
1382 if (sb_any_quota_loaded(sb) &&
1383 sbi->s_qf_names[qtype]) {
1384 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1385 " when quota turned on");
57f73c2c 1386 return -1;
56c50f11 1387 }
03dafb5f 1388 kfree(sbi->s_qf_names[qtype]);
56c50f11
DM		 1389 sbi->s_qf_names[qtype] = NULL;
1390 return 1;
1391}
1392#endif
1393
26092bf5
TT		 1394#define MOPT_SET 0x0001
1395#define MOPT_CLEAR 0x0002
1396#define MOPT_NOSUPPORT 0x0004
1397#define MOPT_EXPLICIT 0x0008
1398#define MOPT_CLEAR_ERR 0x0010
1399#define MOPT_GTE0 0x0020
ac27a0ec 1400#ifdef CONFIG_QUOTA
26092bf5
TT		 1401#define MOPT_Q 0
1402#define MOPT_QFMT 0x0040
1403#else
1404#define MOPT_Q MOPT_NOSUPPORT
1405#define MOPT_QFMT MOPT_NOSUPPORT
ac27a0ec 1406#endif
26092bf5
TT		 1407#define MOPT_DATAJ 0x0080
1408
1409static const struct mount_opts {
1410 int token;
1411 int mount_opt;
1412 int flags;
1413} ext4_mount_opts[] = {
1414 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1415 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1416 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1417 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1418 {Opt_mblk_io_submit, EXT4_MOUNT_MBLK_IO_SUBMIT, MOPT_SET},
1419 {Opt_nomblk_io_submit, EXT4_MOUNT_MBLK_IO_SUBMIT, MOPT_CLEAR},
1420 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1421 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1422 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK, MOPT_SET},
1423 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK, MOPT_CLEAR},
1424 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1425 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1426 {Opt_delalloc, EXT4_MOUNT_DELALLOC, MOPT_SET | MOPT_EXPLICIT},
1427 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC, MOPT_CLEAR | MOPT_EXPLICIT},
1428 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM, MOPT_SET},
1429 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1430 EXT4_MOUNT_JOURNAL_CHECKSUM), MOPT_SET},
1431 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_SET},
1432 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1433 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1434 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1435 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT, MOPT_SET},
1436 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT, MOPT_CLEAR},
1437 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1438 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1439 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1440 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1441 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1442 {Opt_commit, 0, MOPT_GTE0},
1443 {Opt_max_batch_time, 0, MOPT_GTE0},
1444 {Opt_min_batch_time, 0, MOPT_GTE0},
1445 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1446 {Opt_init_itable, 0, MOPT_GTE0},
1447 {Opt_stripe, 0, MOPT_GTE0},
1448 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_DATAJ},
1449 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_DATAJ},
1450 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA, MOPT_DATAJ},
26092bf5
TT		 1451 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1452 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
03010a33 1453#ifdef CONFIG_EXT4_FS_POSIX_ACL
26092bf5
TT		 1454 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1455 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
ac27a0ec 1456#else
26092bf5
TT		 1457 {Opt_acl, 0, MOPT_NOSUPPORT},
1458 {Opt_noacl, 0, MOPT_NOSUPPORT},
ac27a0ec 1459#endif
26092bf5
TT		 1460 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1461 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1462 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1463 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1464 MOPT_SET | MOPT_Q},
1465 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1466 MOPT_SET | MOPT_Q},
1467 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1468 EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1469 {Opt_usrjquota, 0, MOPT_Q},
1470 {Opt_grpjquota, 0, MOPT_Q},
1471 {Opt_offusrjquota, 0, MOPT_Q},
1472 {Opt_offgrpjquota, 0, MOPT_Q},
1473 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1474 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1475 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
df981d03 1476 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
26092bf5
TT		 1477 {Opt_err, 0, 0}
1478};
1479
1480static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1481 substring_t *args, unsigned long *journal_devnum,
1482 unsigned int *journal_ioprio, int is_remount)
1483{
1484 struct ext4_sb_info *sbi = EXT4_SB(sb);
1485 const struct mount_opts *m;
08cefc7a
EB		 1486 kuid_t uid;
1487 kgid_t gid;
26092bf5
TT		 1488 int arg = 0;
1489
57f73c2c
TT		 1490#ifdef CONFIG_QUOTA
1491 if (token == Opt_usrjquota)
1492 return set_qf_name(sb, USRQUOTA, &args[0]);
1493 else if (token == Opt_grpjquota)
1494 return set_qf_name(sb, GRPQUOTA, &args[0]);
1495 else if (token == Opt_offusrjquota)
1496 return clear_qf_name(sb, USRQUOTA);
1497 else if (token == Opt_offgrpjquota)
1498 return clear_qf_name(sb, GRPQUOTA);
1499#endif
26092bf5
TT		 1500 if (args->from && match_int(args, &arg))
1501 return -1;
1502 switch (token) {
f7048605
TT		 1503 case Opt_noacl:
1504 case Opt_nouser_xattr:
1505 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1506 break;
26092bf5
TT		 1507 case Opt_sb:
1508 return 1; /* handled by get_sb_block() */
1509 case Opt_removed:
1510 ext4_msg(sb, KERN_WARNING,
1511 "Ignoring removed %s option", opt);
1512 return 1;
1513 case Opt_resuid:
08cefc7a
EB		 1514 uid = make_kuid(current_user_ns(), arg);
1515 if (!uid_valid(uid)) {
1516 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1517 return -1;
1518 }
1519 sbi->s_resuid = uid;
26092bf5
TT		 1520 return 1;
1521 case Opt_resgid:
08cefc7a
EB		 1522 gid = make_kgid(current_user_ns(), arg);
1523 if (!gid_valid(gid)) {
1524 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1525 return -1;
1526 }
1527 sbi->s_resgid = gid;
26092bf5
TT		 1528 return 1;
1529 case Opt_abort:
1530 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1531 return 1;
1532 case Opt_i_version:
1533 sb->s_flags |= MS_I_VERSION;
1534 return 1;
1535 case Opt_journal_dev:
1536 if (is_remount) {
1537 ext4_msg(sb, KERN_ERR,
1538 "Cannot specify journal on remount");
1539 return -1;
1540 }
1541 *journal_devnum = arg;
1542 return 1;
1543 case Opt_journal_ioprio:
1544 if (arg < 0 || arg > 7)
1545 return -1;
1546 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1547 return 1;
1548 }
1549
1550 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1551 if (token != m->token)
1552 continue;
1553 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1554 return -1;
1555 if (m->flags & MOPT_EXPLICIT)
1556 set_opt2(sb, EXPLICIT_DELALLOC);
1557 if (m->flags & MOPT_CLEAR_ERR)
1558 clear_opt(sb, ERRORS_MASK);
1559 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1560 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1561 "options when quota turned on");
1562 return -1;
1563 }
1564
1565 if (m->flags & MOPT_NOSUPPORT) {
1566 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1567 } else if (token == Opt_commit) {
1568 if (arg == 0)
1569 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1570 sbi->s_commit_interval = HZ * arg;
1571 } else if (token == Opt_max_batch_time) {
1572 if (arg == 0)
1573 arg = EXT4_DEF_MAX_BATCH_TIME;
1574 sbi->s_max_batch_time = arg;
1575 } else if (token == Opt_min_batch_time) {
1576 sbi->s_min_batch_time = arg;
1577 } else if (token == Opt_inode_readahead_blks) {
1578 if (arg > (1 << 30))
1579 return -1;
1580 if (arg && !is_power_of_2(arg)) {
b31e1552 1581 ext4_msg(sb, KERN_ERR,
26092bf5
TT		 1582 "EXT4-fs: inode_readahead_blks"
1583 " must be a power of 2");
1584 return -1;
ac27a0ec 1585 }
26092bf5
TT		 1586 sbi->s_inode_readahead_blks = arg;
1587 } else if (token == Opt_init_itable) {
1588 set_opt(sb, INIT_INODE_TABLE);
1589 if (!args->from)
1590 arg = EXT4_DEF_LI_WAIT_MULT;
1591 sbi->s_li_wait_mult = arg;
df981d03
TT		 1592 } else if (token == Opt_max_dir_size_kb) {
1593 sbi->s_max_dir_size_kb = arg;
26092bf5
TT		 1594 } else if (token == Opt_stripe) {
1595 sbi->s_stripe = arg;
1596 } else if (m->flags & MOPT_DATAJ) {
ac27a0ec 1597 if (is_remount) {
eb513689
TT		 1598 if (!sbi->s_journal)
1599 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
26092bf5
TT		 1600 else if (test_opt(sb, DATA_FLAGS) !=
1601 m->mount_opt) {
b31e1552 1602 ext4_msg(sb, KERN_ERR,
26092bf5
TT		 1603 "Cannot change data mode on remount");
1604 return -1;
ac27a0ec
DK		 1605 }
1606 } else {
fd8c37ec 1607 clear_opt(sb, DATA_FLAGS);
26092bf5 1608 sbi->s_mount_opt |= m->mount_opt;
ac27a0ec 1609 }
ac27a0ec 1610#ifdef CONFIG_QUOTA
26092bf5 1611 } else if (m->flags & MOPT_QFMT) {
17bd13b3 1612 if (sb_any_quota_loaded(sb) &&
26092bf5
TT		 1613 sbi->s_jquota_fmt != m->mount_opt) {
1614 ext4_msg(sb, KERN_ERR, "Cannot "
1615 "change journaled quota options "
1616 "when quota turned on");
1617 return -1;
ac27a0ec 1618 }
26092bf5 1619 sbi->s_jquota_fmt = m->mount_opt;
ac27a0ec 1620#endif
26092bf5
TT		 1621 } else {
1622 if (!args->from)
1623 arg = 1;
1624 if (m->flags & MOPT_CLEAR)
1625 arg = !arg;
1626 else if (unlikely(!(m->flags & MOPT_SET))) {
1627 ext4_msg(sb, KERN_WARNING,
1628 "buggy handling of option %s", opt);
1629 WARN_ON(1);
1630 return -1;
3197ebdb 1631 }
26092bf5
TT		 1632 if (arg != 0)
1633 sbi->s_mount_opt |= m->mount_opt;
afd4672d 1634 else
26092bf5 1635 sbi->s_mount_opt &= ~m->mount_opt;
ac27a0ec 1636 }
26092bf5
TT		 1637 return 1;
1638 }
1639 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1640 "or missing value", opt);
1641 return -1;
1642}
1643
1644static int parse_options(char *options, struct super_block *sb,
1645 unsigned long *journal_devnum,
1646 unsigned int *journal_ioprio,
1647 int is_remount)
1648{
1649 struct ext4_sb_info *sbi = EXT4_SB(sb);
1650 char *p;
1651 substring_t args[MAX_OPT_ARGS];
1652 int token;
1653
1654 if (!options)
1655 return 1;
1656
1657 while ((p = strsep(&options, ",")) != NULL) {
1658 if (!*p)
1659 continue;
1660 /*
1661 * Initialize args struct so we know whether arg was
1662 * found; some options take optional arguments.
1663 */
caecd0af 1664 args[0].to = args[0].from = NULL;
26092bf5
TT		 1665 token = match_token(p, tokens, args);
1666 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1667 journal_ioprio, is_remount) < 0)
1668 return 0;
ac27a0ec
DK		 1669 }
1670#ifdef CONFIG_QUOTA
1671 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
482a7425 1672 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
fd8c37ec 1673 clear_opt(sb, USRQUOTA);
ac27a0ec 1674
482a7425 1675 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
fd8c37ec 1676 clear_opt(sb, GRPQUOTA);
ac27a0ec 1677
56c50f11 1678 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
b31e1552
ES		 1679 ext4_msg(sb, KERN_ERR, "old and new quota "
1680 "format mixing");
ac27a0ec
DK		 1681 return 0;
1682 }
1683
1684 if (!sbi->s_jquota_fmt) {
b31e1552
ES		 1685 ext4_msg(sb, KERN_ERR, "journaled quota format "
1686 "not specified");
ac27a0ec
DK		 1687 return 0;
1688 }
1689 } else {
1690 if (sbi->s_jquota_fmt) {
b31e1552 1691 ext4_msg(sb, KERN_ERR, "journaled quota format "
2c8be6b2 1692 "specified with no journaling "
b31e1552 1693 "enabled");
ac27a0ec
DK		 1694 return 0;
1695 }
1696 }
1697#endif
261cb20c
JK		 1698 if (test_opt(sb, DIOREAD_NOLOCK)) {
1699 int blocksize =
1700 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1701
1702 if (blocksize < PAGE_CACHE_SIZE) {
1703 ext4_msg(sb, KERN_ERR, "can't mount with "
1704 "dioread_nolock if block size != PAGE_SIZE");
1705 return 0;
1706 }
1707 }
ac27a0ec
DK		 1708 return 1;
1709}
1710
2adf6da8
TT		 1711static inline void ext4_show_quota_options(struct seq_file *seq,
1712 struct super_block *sb)
1713{
1714#if defined(CONFIG_QUOTA)
1715 struct ext4_sb_info *sbi = EXT4_SB(sb);
1716
1717 if (sbi->s_jquota_fmt) {
1718 char *fmtname = "";
1719
1720 switch (sbi->s_jquota_fmt) {
1721 case QFMT_VFS_OLD:
1722 fmtname = "vfsold";
1723 break;
1724 case QFMT_VFS_V0:
1725 fmtname = "vfsv0";
1726 break;
1727 case QFMT_VFS_V1:
1728 fmtname = "vfsv1";
1729 break;
1730 }
1731 seq_printf(seq, ",jqfmt=%s", fmtname);
1732 }
1733
1734 if (sbi->s_qf_names[USRQUOTA])
1735 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1736
1737 if (sbi->s_qf_names[GRPQUOTA])
1738 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
1739
1740 if (test_opt(sb, USRQUOTA))
1741 seq_puts(seq, ",usrquota");
1742
1743 if (test_opt(sb, GRPQUOTA))
1744 seq_puts(seq, ",grpquota");
1745#endif
1746}
1747
5a916be1
TT		 1748static const char *token2str(int token)
1749{
50df9fd5 1750 const struct match_token *t;
5a916be1
TT		 1751
1752 for (t = tokens; t->token != Opt_err; t++)
1753 if (t->token == token && !strchr(t->pattern, '='))
1754 break;
1755 return t->pattern;
1756}
1757
2adf6da8
TT		 1758/*
1759 * Show an option if
1760 * - it's set to a non-default value OR
1761 * - if the per-sb default is different from the global default
1762 */
66acdcf4
TT		 1763static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1764 int nodefs)
2adf6da8 1765{
2adf6da8
TT		 1766 struct ext4_sb_info *sbi = EXT4_SB(sb);
1767 struct ext4_super_block *es = sbi->s_es;
66acdcf4 1768 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
5a916be1 1769 const struct mount_opts *m;
66acdcf4 1770 char sep = nodefs ? '\n' : ',';
2adf6da8 1771
66acdcf4
TT		 1772#define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1773#define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2adf6da8
TT		 1774
1775 if (sbi->s_sb_block != 1)
5a916be1
TT		 1776 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1777
1778 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1779 int want_set = m->flags & MOPT_SET;
1780 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1781 (m->flags & MOPT_CLEAR_ERR))
1782 continue;
66acdcf4 1783 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
5a916be1
TT		 1784 continue; /* skip if same as the default */
1785 if ((want_set &&
1786 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1787 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1788 continue; /* select Opt_noFoo vs Opt_Foo */
1789 SEQ_OPTS_PRINT("%s", token2str(m->token));
2adf6da8 1790 }
5a916be1 1791
08cefc7a 1792 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
5a916be1 1793 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
08cefc7a
EB		 1794 SEQ_OPTS_PRINT("resuid=%u",
1795 from_kuid_munged(&init_user_ns, sbi->s_resuid));
1796 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
5a916be1 1797 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
08cefc7a
EB		 1798 SEQ_OPTS_PRINT("resgid=%u",
1799 from_kgid_munged(&init_user_ns, sbi->s_resgid));
66acdcf4 1800 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
5a916be1
TT		 1801 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1802 SEQ_OPTS_PUTS("errors=remount-ro");
2adf6da8 1803 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
5a916be1 1804 SEQ_OPTS_PUTS("errors=continue");
2adf6da8 1805 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
5a916be1 1806 SEQ_OPTS_PUTS("errors=panic");
66acdcf4 1807 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
5a916be1 1808 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
66acdcf4 1809 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
5a916be1 1810 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
66acdcf4 1811 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
5a916be1 1812 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2adf6da8 1813 if (sb->s_flags & MS_I_VERSION)
5a916be1 1814 SEQ_OPTS_PUTS("i_version");
66acdcf4 1815 if (nodefs || sbi->s_stripe)
5a916be1 1816 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
66acdcf4 1817 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
5a916be1
TT		 1818 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1819 SEQ_OPTS_PUTS("data=journal");
1820 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1821 SEQ_OPTS_PUTS("data=ordered");
1822 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1823 SEQ_OPTS_PUTS("data=writeback");
1824 }
66acdcf4
TT		 1825 if (nodefs ||
1826 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
5a916be1
TT		 1827 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1828 sbi->s_inode_readahead_blks);
2adf6da8 1829
66acdcf4
TT		 1830 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1831 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
5a916be1 1832 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
df981d03
TT		 1833 if (nodefs || sbi->s_max_dir_size_kb)
1834 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
2adf6da8
TT		 1835
1836 ext4_show_quota_options(seq, sb);
2adf6da8
TT		 1837 return 0;
1838}
1839
66acdcf4
TT		 1840static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1841{
1842 return _ext4_show_options(seq, root->d_sb, 0);
1843}
1844
1845static int options_seq_show(struct seq_file *seq, void *offset)
1846{
1847 struct super_block *sb = seq->private;
1848 int rc;
1849
1850 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1851 rc = _ext4_show_options(seq, sb, 1);
1852 seq_puts(seq, "\n");
1853 return rc;
1854}
1855
1856static int options_open_fs(struct inode *inode, struct file *file)
1857{
1858 return single_open(file, options_seq_show, PDE(inode)->data);
1859}
1860
1861static const struct file_operations ext4_seq_options_fops = {
1862 .owner = THIS_MODULE,
1863 .open = options_open_fs,
1864 .read = seq_read,
1865 .llseek = seq_lseek,
1866 .release = single_release,
1867};
1868
617ba13b 1869static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 1870 int read_only)
1871{
617ba13b 1872 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 1873 int res = 0;
1874
617ba13b 1875 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
b31e1552
ES		 1876 ext4_msg(sb, KERN_ERR, "revision level too high, "
1877 "forcing read-only mode");
ac27a0ec
DK		 1878 res = MS_RDONLY;
1879 }
1880 if (read_only)
281b5995 1881 goto done;
617ba13b 1882 if (!(sbi->s_mount_state & EXT4_VALID_FS))
b31e1552
ES		 1883 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1884 "running e2fsck is recommended");
617ba13b 1885 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
b31e1552
ES		 1886 ext4_msg(sb, KERN_WARNING,
1887 "warning: mounting fs with errors, "
1888 "running e2fsck is recommended");
ed3ce80a 1889 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
ac27a0ec
DK		 1890 le16_to_cpu(es->s_mnt_count) >=
1891 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
b31e1552
ES		 1892 ext4_msg(sb, KERN_WARNING,
1893 "warning: maximal mount count reached, "
1894 "running e2fsck is recommended");
ac27a0ec
DK		 1895 else if (le32_to_cpu(es->s_checkinterval) &&
1896 (le32_to_cpu(es->s_lastcheck) +
1897 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
b31e1552
ES		 1898 ext4_msg(sb, KERN_WARNING,
1899 "warning: checktime reached, "
1900 "running e2fsck is recommended");
0b8e58a1 1901 if (!sbi->s_journal)
0390131b 1902 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec 1903 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 1904 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 1905 le16_add_cpu(&es->s_mnt_count, 1);
ac27a0ec 1906 es->s_mtime = cpu_to_le32(get_seconds());
617ba13b 1907 ext4_update_dynamic_rev(sb);
0390131b
FM		 1908 if (sbi->s_journal)
1909 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 1910
e2d67052 1911 ext4_commit_super(sb, 1);
281b5995 1912done:
ac27a0ec 1913 if (test_opt(sb, DEBUG))
a9df9a49 1914 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
a2595b8a 1915 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
ac27a0ec
DK		 1916 sb->s_blocksize,
1917 sbi->s_groups_count,
617ba13b
MC		 1918 EXT4_BLOCKS_PER_GROUP(sb),
1919 EXT4_INODES_PER_GROUP(sb),
a2595b8a 1920 sbi->s_mount_opt, sbi->s_mount_opt2);
ac27a0ec 1921
7abc52c2 1922 cleancache_init_fs(sb);
ac27a0ec
DK		 1923 return res;
1924}
1925
117fff10
TT		 1926int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
1927{
1928 struct ext4_sb_info *sbi = EXT4_SB(sb);
1929 struct flex_groups *new_groups;
1930 int size;
1931
1932 if (!sbi->s_log_groups_per_flex)
1933 return 0;
1934
1935 size = ext4_flex_group(sbi, ngroup - 1) + 1;
1936 if (size <= sbi->s_flex_groups_allocated)
1937 return 0;
1938
1939 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
1940 new_groups = ext4_kvzalloc(size, GFP_KERNEL);
1941 if (!new_groups) {
1942 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
1943 size / (int) sizeof(struct flex_groups));
1944 return -ENOMEM;
1945 }
1946
1947 if (sbi->s_flex_groups) {
1948 memcpy(new_groups, sbi->s_flex_groups,
1949 (sbi->s_flex_groups_allocated *
1950 sizeof(struct flex_groups)));
1951 ext4_kvfree(sbi->s_flex_groups);
1952 }
1953 sbi->s_flex_groups = new_groups;
1954 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
1955 return 0;
1956}
1957
772cb7c8
JS		 1958static int ext4_fill_flex_info(struct super_block *sb)
1959{
1960 struct ext4_sb_info *sbi = EXT4_SB(sb);
1961 struct ext4_group_desc *gdp = NULL;
772cb7c8 1962 ext4_group_t flex_group;
d50f2ab6 1963 unsigned int groups_per_flex = 0;
117fff10 1964 int i, err;
772cb7c8 1965
503358ae 1966 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
d50f2ab6 1967 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
772cb7c8
JS		 1968 sbi->s_log_groups_per_flex = 0;
1969 return 1;
1970 }
76495ec1 1971 groups_per_flex = 1U << sbi->s_log_groups_per_flex;
772cb7c8 1972
117fff10
TT		 1973 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
1974 if (err)
9933fc0a 1975 goto failed;
772cb7c8 1976
772cb7c8 1977 for (i = 0; i < sbi->s_groups_count; i++) {
88b6edd1 1978 gdp = ext4_get_group_desc(sb, i, NULL);
772cb7c8
JS		 1979
1980 flex_group = ext4_flex_group(sbi, i);
7ad9bb65
TT		 1981 atomic_add(ext4_free_inodes_count(sb, gdp),
1982 &sbi->s_flex_groups[flex_group].free_inodes);
021b65bb 1983 atomic_add(ext4_free_group_clusters(sb, gdp),
24aaa8ef 1984 &sbi->s_flex_groups[flex_group].free_clusters);
7ad9bb65
TT		 1985 atomic_add(ext4_used_dirs_count(sb, gdp),
1986 &sbi->s_flex_groups[flex_group].used_dirs);
772cb7c8
JS		 1987 }
1988
1989 return 1;
1990failed:
1991 return 0;
1992}
1993
feb0ab32
DW		 1994static __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1995 struct ext4_group_desc *gdp)
717d50e4 1996{
feb0ab32 1997 int offset;
717d50e4 1998 __u16 crc = 0;
feb0ab32 1999 __le32 le_group = cpu_to_le32(block_group);
717d50e4 2000
feb0ab32
DW		 2001 if ((sbi->s_es->s_feature_ro_compat &
2002 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))) {
2003 /* Use new metadata_csum algorithm */
2004 __u16 old_csum;
2005 __u32 csum32;
2006
2007 old_csum = gdp->bg_checksum;
2008 gdp->bg_checksum = 0;
2009 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2010 sizeof(le_group));
2011 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp,
2012 sbi->s_desc_size);
2013 gdp->bg_checksum = old_csum;
2014
2015 crc = csum32 & 0xFFFF;
2016 goto out;
717d50e4
AD		 2017 }
2018
feb0ab32
DW		 2019 /* old crc16 code */
2020 offset = offsetof(struct ext4_group_desc, bg_checksum);
2021
2022 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2023 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2024 crc = crc16(crc, (__u8 *)gdp, offset);
2025 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2026 /* for checksum of struct ext4_group_desc do the rest...*/
2027 if ((sbi->s_es->s_feature_incompat &
2028 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
2029 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2030 crc = crc16(crc, (__u8 *)gdp + offset,
2031 le16_to_cpu(sbi->s_es->s_desc_size) -
2032 offset);
2033
2034out:
717d50e4
AD		 2035 return cpu_to_le16(crc);
2036}
2037
feb0ab32 2038int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
717d50e4
AD		 2039 struct ext4_group_desc *gdp)
2040{
feb0ab32
DW		 2041 if (ext4_has_group_desc_csum(sb) &&
2042 (gdp->bg_checksum != ext4_group_desc_csum(EXT4_SB(sb),
2043 block_group, gdp)))
717d50e4
AD		 2044 return 0;
2045
2046 return 1;
2047}
2048
feb0ab32
DW		 2049void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2050 struct ext4_group_desc *gdp)
2051{
2052 if (!ext4_has_group_desc_csum(sb))
2053 return;
2054 gdp->bg_checksum = ext4_group_desc_csum(EXT4_SB(sb), block_group, gdp);
2055}
2056
ac27a0ec 2057/* Called at mount-time, super-block is locked */
bfff6873
LC		 2058static int ext4_check_descriptors(struct super_block *sb,
2059 ext4_group_t *first_not_zeroed)
ac27a0ec 2060{
617ba13b
MC		 2061 struct ext4_sb_info *sbi = EXT4_SB(sb);
2062 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2063 ext4_fsblk_t last_block;
bd81d8ee
LV		 2064 ext4_fsblk_t block_bitmap;
2065 ext4_fsblk_t inode_bitmap;
2066 ext4_fsblk_t inode_table;
ce421581 2067 int flexbg_flag = 0;
bfff6873 2068 ext4_group_t i, grp = sbi->s_groups_count;
ac27a0ec 2069
ce421581
JS		 2070 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2071 flexbg_flag = 1;
2072
af5bc92d 2073 ext4_debug("Checking group descriptors");
ac27a0ec 2074
197cd65a
AM		 2075 for (i = 0; i < sbi->s_groups_count; i++) {
2076 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2077
ce421581 2078 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 2079 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 2080 else
2081 last_block = first_block +
617ba13b 2082 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 2083
bfff6873
LC		 2084 if ((grp == sbi->s_groups_count) &&
2085 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2086 grp = i;
2087
8fadc143 2088 block_bitmap = ext4_block_bitmap(sb, gdp);
2b2d6d01 2089 if (block_bitmap < first_block || block_bitmap > last_block) {
b31e1552 2090 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2091 "Block bitmap for group %u not in group "
b31e1552 2092 "(block %llu)!", i, block_bitmap);
ac27a0ec
DK		 2093 return 0;
2094 }
8fadc143 2095 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2b2d6d01 2096 if (inode_bitmap < first_block || inode_bitmap > last_block) {
b31e1552 2097 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2098 "Inode bitmap for group %u not in group "
b31e1552 2099 "(block %llu)!", i, inode_bitmap);
ac27a0ec
DK		 2100 return 0;
2101 }
8fadc143 2102 inode_table = ext4_inode_table(sb, gdp);
bd81d8ee 2103 if (inode_table < first_block ||
2b2d6d01 2104 inode_table + sbi->s_itb_per_group - 1 > last_block) {
b31e1552 2105 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2106 "Inode table for group %u not in group "
b31e1552 2107 "(block %llu)!", i, inode_table);
ac27a0ec
DK		 2108 return 0;
2109 }
955ce5f5 2110 ext4_lock_group(sb, i);
feb0ab32 2111 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
b31e1552
ES		 2112 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2113 "Checksum for group %u failed (%u!=%u)",
2114 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2115 gdp)), le16_to_cpu(gdp->bg_checksum));
7ee1ec4c 2116 if (!(sb->s_flags & MS_RDONLY)) {
955ce5f5 2117 ext4_unlock_group(sb, i);
8a266467 2118 return 0;
7ee1ec4c 2119 }
717d50e4 2120 }
955ce5f5 2121 ext4_unlock_group(sb, i);
ce421581
JS		 2122 if (!flexbg_flag)
2123 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec 2124 }
bfff6873
LC		 2125 if (NULL != first_not_zeroed)
2126 *first_not_zeroed = grp;
ac27a0ec 2127
5dee5437
TT		 2128 ext4_free_blocks_count_set(sbi->s_es,
2129 EXT4_C2B(sbi, ext4_count_free_clusters(sb)));
0b8e58a1 2130 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
ac27a0ec
DK		 2131 return 1;
2132}
2133
617ba13b 2134/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 2135 * the superblock) which were deleted from all directories, but held open by
2136 * a process at the time of a crash. We walk the list and try to delete these
2137 * inodes at recovery time (only with a read-write filesystem).
2138 *
2139 * In order to keep the orphan inode chain consistent during traversal (in
2140 * case of crash during recovery), we link each inode into the superblock
2141 * orphan list_head and handle it the same way as an inode deletion during
2142 * normal operation (which journals the operations for us).
2143 *
2144 * We only do an iget() and an iput() on each inode, which is very safe if we
2145 * accidentally point at an in-use or already deleted inode. The worst that
2146 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 2147 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 2148 * e2fsck was run on this filesystem, and it must have already done the orphan
2149 * inode cleanup for us, so we can safely abort without any further action.
2150 */
2b2d6d01
TT		 2151static void ext4_orphan_cleanup(struct super_block *sb,
2152 struct ext4_super_block *es)
ac27a0ec
DK		 2153{
2154 unsigned int s_flags = sb->s_flags;
2155 int nr_orphans = 0, nr_truncates = 0;
2156#ifdef CONFIG_QUOTA
2157 int i;
2158#endif
2159 if (!es->s_last_orphan) {
2160 jbd_debug(4, "no orphan inodes to clean up\n");
2161 return;
2162 }
2163
a8f48a95 2164 if (bdev_read_only(sb->s_bdev)) {
b31e1552
ES		 2165 ext4_msg(sb, KERN_ERR, "write access "
2166 "unavailable, skipping orphan cleanup");
a8f48a95
ES		 2167 return;
2168 }
2169
d39195c3
AG		 2170 /* Check if feature set would not allow a r/w mount */
2171 if (!ext4_feature_set_ok(sb, 0)) {
2172 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2173 "unknown ROCOMPAT features");
2174 return;
2175 }
2176
617ba13b 2177 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
c25f9bc6
ES		 2178 /* don't clear list on RO mount w/ errors */
2179 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
ac27a0ec
DK		 2180 jbd_debug(1, "Errors on filesystem, "
2181 "clearing orphan list.\n");
c25f9bc6
ES		 2182 es->s_last_orphan = 0;
2183 }
ac27a0ec
DK		 2184 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2185 return;
2186 }
2187
2188 if (s_flags & MS_RDONLY) {
b31e1552 2189 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
ac27a0ec
DK		 2190 sb->s_flags &= ~MS_RDONLY;
2191 }
2192#ifdef CONFIG_QUOTA
2193 /* Needed for iput() to work correctly and not trash data */
2194 sb->s_flags |= MS_ACTIVE;
2195 /* Turn on quotas so that they are updated correctly */
2196 for (i = 0; i < MAXQUOTAS; i++) {
617ba13b
MC		 2197 if (EXT4_SB(sb)->s_qf_names[i]) {
2198 int ret = ext4_quota_on_mount(sb, i);
ac27a0ec 2199 if (ret < 0)
b31e1552
ES		 2200 ext4_msg(sb, KERN_ERR,
2201 "Cannot turn on journaled "
2202 "quota: error %d", ret);
ac27a0ec
DK		 2203 }
2204 }
2205#endif
2206
2207 while (es->s_last_orphan) {
2208 struct inode *inode;
2209
97bd42b9
JB		 2210 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2211 if (IS_ERR(inode)) {
ac27a0ec
DK		 2212 es->s_last_orphan = 0;
2213 break;
2214 }
2215
617ba13b 2216 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
871a2931 2217 dquot_initialize(inode);
ac27a0ec 2218 if (inode->i_nlink) {
b31e1552
ES		 2219 ext4_msg(sb, KERN_DEBUG,
2220 "%s: truncating inode %lu to %lld bytes",
46e665e9 2221 __func__, inode->i_ino, inode->i_size);
e5f8eab8 2222 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
ac27a0ec 2223 inode->i_ino, inode->i_size);
721e3eba 2224 mutex_lock(&inode->i_mutex);
617ba13b 2225 ext4_truncate(inode);
721e3eba 2226 mutex_unlock(&inode->i_mutex);
ac27a0ec
DK		 2227 nr_truncates++;
2228 } else {
b31e1552
ES		 2229 ext4_msg(sb, KERN_DEBUG,
2230 "%s: deleting unreferenced inode %lu",
46e665e9 2231 __func__, inode->i_ino);
ac27a0ec
DK		 2232 jbd_debug(2, "deleting unreferenced inode %lu\n",
2233 inode->i_ino);
2234 nr_orphans++;
2235 }
2236 iput(inode); /* The delete magic happens here! */
2237 }
2238
2b2d6d01 2239#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
ac27a0ec
DK		 2240
2241 if (nr_orphans)
b31e1552
ES		 2242 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2243 PLURAL(nr_orphans));
ac27a0ec 2244 if (nr_truncates)
b31e1552
ES		 2245 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2246 PLURAL(nr_truncates));
ac27a0ec
DK		 2247#ifdef CONFIG_QUOTA
2248 /* Turn quotas off */
2249 for (i = 0; i < MAXQUOTAS; i++) {
2250 if (sb_dqopt(sb)->files[i])
287a8095 2251 dquot_quota_off(sb, i);
ac27a0ec
DK		 2252 }
2253#endif
2254 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2255}
0b8e58a1 2256
cd2291a4
ES		 2257/*
2258 * Maximal extent format file size.
2259 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2260 * extent format containers, within a sector_t, and within i_blocks
2261 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2262 * so that won't be a limiting factor.
2263 *
f17722f9
LC		 2264 * However there is other limiting factor. We do store extents in the form
2265 * of starting block and length, hence the resulting length of the extent
2266 * covering maximum file size must fit into on-disk format containers as
2267 * well. Given that length is always by 1 unit bigger than max unit (because
2268 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2269 *
cd2291a4
ES		 2270 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2271 */
f287a1a5 2272static loff_t ext4_max_size(int blkbits, int has_huge_files)
cd2291a4
ES		 2273{
2274 loff_t res;
2275 loff_t upper_limit = MAX_LFS_FILESIZE;
2276
2277 /* small i_blocks in vfs inode? */
f287a1a5 2278 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
cd2291a4 2279 /*
90c699a9 2280 * CONFIG_LBDAF is not enabled implies the inode
cd2291a4
ES		 2281 * i_block represent total blocks in 512 bytes
2282 * 32 == size of vfs inode i_blocks * 8
2283 */
2284 upper_limit = (1LL << 32) - 1;
2285
2286 /* total blocks in file system block size */
2287 upper_limit >>= (blkbits - 9);
2288 upper_limit <<= blkbits;
2289 }
2290
f17722f9
LC		 2291 /*
2292 * 32-bit extent-start container, ee_block. We lower the maxbytes
2293 * by one fs block, so ee_len can cover the extent of maximum file
2294 * size
2295 */
2296 res = (1LL << 32) - 1;
cd2291a4 2297 res <<= blkbits;
cd2291a4
ES		 2298
2299 /* Sanity check against vm- & vfs- imposed limits */
2300 if (res > upper_limit)
2301 res = upper_limit;
2302
2303 return res;
2304}
ac27a0ec 2305
ac27a0ec 2306/*
cd2291a4 2307 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 2308 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2309 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 2310 */
f287a1a5 2311static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
ac27a0ec 2312{
617ba13b 2313 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 2314 int meta_blocks;
2315 loff_t upper_limit;
0b8e58a1
AD		 2316 /* This is calculated to be the largest file size for a dense, block
2317 * mapped file such that the file's total number of 512-byte sectors,
2318 * including data and all indirect blocks, does not exceed (2^48 - 1).
2319 *
2320 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2321 * number of 512-byte sectors of the file.
0fc1b451
AK		 2322 */
2323
f287a1a5 2324 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
0fc1b451 2325 /*
90c699a9 2326 * !has_huge_files or CONFIG_LBDAF not enabled implies that
0b8e58a1
AD		 2327 * the inode i_block field represents total file blocks in
2328 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
0fc1b451
AK		 2329 */
2330 upper_limit = (1LL << 32) - 1;
2331
2332 /* total blocks in file system block size */
2333 upper_limit >>= (bits - 9);
2334
2335 } else {
8180a562
AK		 2336 /*
2337 * We use 48 bit ext4_inode i_blocks
2338 * With EXT4_HUGE_FILE_FL set the i_blocks
2339 * represent total number of blocks in
2340 * file system block size
2341 */
0fc1b451
AK		 2342 upper_limit = (1LL << 48) - 1;
2343
0fc1b451
AK		 2344 }
2345
2346 /* indirect blocks */
2347 meta_blocks = 1;
2348 /* double indirect blocks */
2349 meta_blocks += 1 + (1LL << (bits-2));
2350 /* tripple indirect blocks */
2351 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2352
2353 upper_limit -= meta_blocks;
2354 upper_limit <<= bits;
ac27a0ec
DK		 2355
2356 res += 1LL << (bits-2);
2357 res += 1LL << (2*(bits-2));
2358 res += 1LL << (3*(bits-2));
2359 res <<= bits;
2360 if (res > upper_limit)
2361 res = upper_limit;
0fc1b451
AK		 2362
2363 if (res > MAX_LFS_FILESIZE)
2364 res = MAX_LFS_FILESIZE;
2365
ac27a0ec
DK		 2366 return res;
2367}
2368
617ba13b 2369static ext4_fsblk_t descriptor_loc(struct super_block *sb,
0b8e58a1 2370 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 2371{
617ba13b 2372 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 2373 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 2374 int has_super = 0;
2375
2376 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2377
617ba13b 2378 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
ac27a0ec 2379 nr < first_meta_bg)
70bbb3e0 2380 return logical_sb_block + nr + 1;
ac27a0ec 2381 bg = sbi->s_desc_per_block * nr;
617ba13b 2382 if (ext4_bg_has_super(sb, bg))
ac27a0ec 2383 has_super = 1;
0b8e58a1 2384
617ba13b 2385 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 2386}
2387
c9de560d
AT		 2388/**
2389 * ext4_get_stripe_size: Get the stripe size.
2390 * @sbi: In memory super block info
2391 *
2392 * If we have specified it via mount option, then
2393 * use the mount option value. If the value specified at mount time is
2394 * greater than the blocks per group use the super block value.
2395 * If the super block value is greater than blocks per group return 0.
2396 * Allocator needs it be less than blocks per group.
2397 *
2398 */
2399static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2400{
2401 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2402 unsigned long stripe_width =
2403 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3eb08658 2404 int ret;
c9de560d
AT		 2405
2406 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3eb08658
DE		 2407 ret = sbi->s_stripe;
2408 else if (stripe_width <= sbi->s_blocks_per_group)
2409 ret = stripe_width;
2410 else if (stride <= sbi->s_blocks_per_group)
2411 ret = stride;
2412 else
2413 ret = 0;
c9de560d 2414
3eb08658
DE		 2415 /*
2416 * If the stripe width is 1, this makes no sense and
2417 * we set it to 0 to turn off stripe handling code.
2418 */
2419 if (ret <= 1)
2420 ret = 0;
c9de560d 2421
3eb08658 2422 return ret;
c9de560d 2423}
ac27a0ec 2424
3197ebdb
TT		 2425/* sysfs supprt */
2426
2427struct ext4_attr {
2428 struct attribute attr;
2429 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
60e6679e 2430 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
3197ebdb
TT		 2431 const char *, size_t);
2432 int offset;
2433};
2434
2435static int parse_strtoul(const char *buf,
2436 unsigned long max, unsigned long *value)
2437{
2438 char *endp;
2439
e7d2860b
AGR		 2440 *value = simple_strtoul(skip_spaces(buf), &endp, 0);
2441 endp = skip_spaces(endp);
3197ebdb
TT		 2442 if (*endp || *value > max)
2443 return -EINVAL;
2444
2445 return 0;
2446}
2447
2448static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2449 struct ext4_sb_info *sbi,
2450 char *buf)
2451{
2452 return snprintf(buf, PAGE_SIZE, "%llu\n",
7b415bf6
AK		 2453 (s64) EXT4_C2B(sbi,
2454 percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
3197ebdb
TT		 2455}
2456
2457static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2458 struct ext4_sb_info *sbi, char *buf)
2459{
2460 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2461
f613dfcb
TT		 2462 if (!sb->s_bdev->bd_part)
2463 return snprintf(buf, PAGE_SIZE, "0\n");
3197ebdb
TT		 2464 return snprintf(buf, PAGE_SIZE, "%lu\n",
2465 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2466 sbi->s_sectors_written_start) >> 1);
2467}
2468
2469static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2470 struct ext4_sb_info *sbi, char *buf)
2471{
2472 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2473
f613dfcb
TT		 2474 if (!sb->s_bdev->bd_part)
2475 return snprintf(buf, PAGE_SIZE, "0\n");
3197ebdb 2476 return snprintf(buf, PAGE_SIZE, "%llu\n",
a6b43e38 2477 (unsigned long long)(sbi->s_kbytes_written +
3197ebdb 2478 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
a6b43e38 2479 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
3197ebdb
TT		 2480}
2481
2482static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2483 struct ext4_sb_info *sbi,
2484 const char *buf, size_t count)
2485{
2486 unsigned long t;
2487
2488 if (parse_strtoul(buf, 0x40000000, &t))
2489 return -EINVAL;
2490
5dbd571d 2491 if (t && !is_power_of_2(t))
3197ebdb
TT		 2492 return -EINVAL;
2493
2494 sbi->s_inode_readahead_blks = t;
2495 return count;
2496}
2497
2498static ssize_t sbi_ui_show(struct ext4_attr *a,
0b8e58a1 2499 struct ext4_sb_info *sbi, char *buf)
3197ebdb
TT		 2500{
2501 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2502
2503 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2504}
2505
2506static ssize_t sbi_ui_store(struct ext4_attr *a,
2507 struct ext4_sb_info *sbi,
2508 const char *buf, size_t count)
2509{
2510 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2511 unsigned long t;
2512
2513 if (parse_strtoul(buf, 0xffffffff, &t))
2514 return -EINVAL;
2515 *ui = t;
2516 return count;
2517}
2518
2c0544b2
TT		 2519static ssize_t trigger_test_error(struct ext4_attr *a,
2520 struct ext4_sb_info *sbi,
2521 const char *buf, size_t count)
2522{
2523 int len = count;
2524
2525 if (!capable(CAP_SYS_ADMIN))
2526 return -EPERM;
2527
2528 if (len && buf[len-1] == '\n')
2529 len--;
2530
2531 if (len)
2532 ext4_error(sbi->s_sb, "%.*s", len, buf);
2533 return count;
2534}
2535
3197ebdb
TT		 2536#define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2537static struct ext4_attr ext4_attr_##_name = { \
2538 .attr = {.name = __stringify(_name), .mode = _mode }, \
2539 .show = _show, \
2540 .store = _store, \
2541 .offset = offsetof(struct ext4_sb_info, _elname), \
2542}
2543#define EXT4_ATTR(name, mode, show, store) \
2544static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2545
857ac889 2546#define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
3197ebdb
TT		 2547#define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2548#define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2549#define EXT4_RW_ATTR_SBI_UI(name, elname) \
2550 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2551#define ATTR_LIST(name) &ext4_attr_##name.attr
2552
2553EXT4_RO_ATTR(delayed_allocation_blocks);
2554EXT4_RO_ATTR(session_write_kbytes);
2555EXT4_RO_ATTR(lifetime_write_kbytes);
2556EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2557 inode_readahead_blks_store, s_inode_readahead_blks);
11013911 2558EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
3197ebdb
TT		 2559EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2560EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2561EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2562EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2563EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2564EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
55138e0b 2565EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
67a5da56 2566EXT4_RW_ATTR_SBI_UI(extent_max_zeroout_kb, s_extent_max_zeroout_kb);
2c0544b2 2567EXT4_ATTR(trigger_fs_error, 0200, NULL, trigger_test_error);
3197ebdb
TT		 2568
2569static struct attribute *ext4_attrs[] = {
2570 ATTR_LIST(delayed_allocation_blocks),
2571 ATTR_LIST(session_write_kbytes),
2572 ATTR_LIST(lifetime_write_kbytes),
2573 ATTR_LIST(inode_readahead_blks),
11013911 2574 ATTR_LIST(inode_goal),
3197ebdb
TT		 2575 ATTR_LIST(mb_stats),
2576 ATTR_LIST(mb_max_to_scan),
2577 ATTR_LIST(mb_min_to_scan),
2578 ATTR_LIST(mb_order2_req),
2579 ATTR_LIST(mb_stream_req),
2580 ATTR_LIST(mb_group_prealloc),
55138e0b 2581 ATTR_LIST(max_writeback_mb_bump),
67a5da56 2582 ATTR_LIST(extent_max_zeroout_kb),
2c0544b2 2583 ATTR_LIST(trigger_fs_error),
3197ebdb
TT		 2584 NULL,
2585};
2586
857ac889
LC		 2587/* Features this copy of ext4 supports */
2588EXT4_INFO_ATTR(lazy_itable_init);
27ee40df 2589EXT4_INFO_ATTR(batched_discard);
5e7bbef1 2590EXT4_INFO_ATTR(meta_bg_resize);
857ac889
LC		 2591
2592static struct attribute *ext4_feat_attrs[] = {
2593 ATTR_LIST(lazy_itable_init),
27ee40df 2594 ATTR_LIST(batched_discard),
5e7bbef1 2595 ATTR_LIST(meta_bg_resize),
857ac889
LC		 2596 NULL,
2597};
2598
3197ebdb
TT		 2599static ssize_t ext4_attr_show(struct kobject *kobj,
2600 struct attribute *attr, char *buf)
2601{
2602 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2603 s_kobj);
2604 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2605
2606 return a->show ? a->show(a, sbi, buf) : 0;
2607}
2608
2609static ssize_t ext4_attr_store(struct kobject *kobj,
2610 struct attribute *attr,
2611 const char *buf, size_t len)
2612{
2613 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2614 s_kobj);
2615 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2616
2617 return a->store ? a->store(a, sbi, buf, len) : 0;
2618}
2619
2620static void ext4_sb_release(struct kobject *kobj)
2621{
2622 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2623 s_kobj);
2624 complete(&sbi->s_kobj_unregister);
2625}
2626
52cf25d0 2627static const struct sysfs_ops ext4_attr_ops = {
3197ebdb
TT		 2628 .show = ext4_attr_show,
2629 .store = ext4_attr_store,
2630};
2631
2632static struct kobj_type ext4_ktype = {
2633 .default_attrs = ext4_attrs,
2634 .sysfs_ops = &ext4_attr_ops,
2635 .release = ext4_sb_release,
2636};
2637
857ac889
LC		 2638static void ext4_feat_release(struct kobject *kobj)
2639{
2640 complete(&ext4_feat->f_kobj_unregister);
2641}
2642
2643static struct kobj_type ext4_feat_ktype = {
2644 .default_attrs = ext4_feat_attrs,
2645 .sysfs_ops = &ext4_attr_ops,
2646 .release = ext4_feat_release,
2647};
2648
a13fb1a4
ES		 2649/*
2650 * Check whether this filesystem can be mounted based on
2651 * the features present and the RDONLY/RDWR mount requested.
2652 * Returns 1 if this filesystem can be mounted as requested,
2653 * 0 if it cannot be.
2654 */
2655static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2656{
2657 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2658 ext4_msg(sb, KERN_ERR,
2659 "Couldn't mount because of "
2660 "unsupported optional features (%x)",
2661 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2662 ~EXT4_FEATURE_INCOMPAT_SUPP));
2663 return 0;
2664 }
2665
2666 if (readonly)
2667 return 1;
2668
2669 /* Check that feature set is OK for a read-write mount */
2670 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2671 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2672 "unsupported optional features (%x)",
2673 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2674 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2675 return 0;
2676 }
2677 /*
2678 * Large file size enabled file system can only be mounted
2679 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2680 */
2681 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2682 if (sizeof(blkcnt_t) < sizeof(u64)) {
2683 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2684 "cannot be mounted RDWR without "
2685 "CONFIG_LBDAF");
2686 return 0;
2687 }
2688 }
bab08ab9
TT		 2689 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2690 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2691 ext4_msg(sb, KERN_ERR,
2692 "Can't support bigalloc feature without "
2693 "extents feature\n");
2694 return 0;
2695 }
7c319d32
AK		 2696
2697#ifndef CONFIG_QUOTA
2698 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
2699 !readonly) {
2700 ext4_msg(sb, KERN_ERR,
2701 "Filesystem with quota feature cannot be mounted RDWR "
2702 "without CONFIG_QUOTA");
2703 return 0;
2704 }
2705#endif /* CONFIG_QUOTA */
a13fb1a4
ES		 2706 return 1;
2707}
2708
66e61a9e
TT		 2709/*
2710 * This function is called once a day if we have errors logged
2711 * on the file system
2712 */
2713static void print_daily_error_info(unsigned long arg)
2714{
2715 struct super_block *sb = (struct super_block *) arg;
2716 struct ext4_sb_info *sbi;
2717 struct ext4_super_block *es;
2718
2719 sbi = EXT4_SB(sb);
2720 es = sbi->s_es;
2721
2722 if (es->s_error_count)
2723 ext4_msg(sb, KERN_NOTICE, "error count: %u",
2724 le32_to_cpu(es->s_error_count));
2725 if (es->s_first_error_time) {
2726 printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d",
2727 sb->s_id, le32_to_cpu(es->s_first_error_time),
2728 (int) sizeof(es->s_first_error_func),
2729 es->s_first_error_func,
2730 le32_to_cpu(es->s_first_error_line));
2731 if (es->s_first_error_ino)
2732 printk(": inode %u",
2733 le32_to_cpu(es->s_first_error_ino));
2734 if (es->s_first_error_block)
2735 printk(": block %llu", (unsigned long long)
2736 le64_to_cpu(es->s_first_error_block));
2737 printk("\n");
2738 }
2739 if (es->s_last_error_time) {
2740 printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d",
2741 sb->s_id, le32_to_cpu(es->s_last_error_time),
2742 (int) sizeof(es->s_last_error_func),
2743 es->s_last_error_func,
2744 le32_to_cpu(es->s_last_error_line));
2745 if (es->s_last_error_ino)
2746 printk(": inode %u",
2747 le32_to_cpu(es->s_last_error_ino));
2748 if (es->s_last_error_block)
2749 printk(": block %llu", (unsigned long long)
2750 le64_to_cpu(es->s_last_error_block));
2751 printk("\n");
2752 }
2753 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2754}
2755
bfff6873
LC		 2756/* Find next suitable group and run ext4_init_inode_table */
2757static int ext4_run_li_request(struct ext4_li_request *elr)
2758{
2759 struct ext4_group_desc *gdp = NULL;
2760 ext4_group_t group, ngroups;
2761 struct super_block *sb;
2762 unsigned long timeout = 0;
2763 int ret = 0;
2764
2765 sb = elr->lr_super;
2766 ngroups = EXT4_SB(sb)->s_groups_count;
2767
8e8ad8a5 2768 sb_start_write(sb);
bfff6873
LC		 2769 for (group = elr->lr_next_group; group < ngroups; group++) {
2770 gdp = ext4_get_group_desc(sb, group, NULL);
2771 if (!gdp) {
2772 ret = 1;
2773 break;
2774 }
2775
2776 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2777 break;
2778 }
2779
7f511862 2780 if (group >= ngroups)
bfff6873
LC		 2781 ret = 1;
2782
2783 if (!ret) {
2784 timeout = jiffies;
2785 ret = ext4_init_inode_table(sb, group,
2786 elr->lr_timeout ? 0 : 1);
2787 if (elr->lr_timeout == 0) {
51ce6511
LC		 2788 timeout = (jiffies - timeout) *
2789 elr->lr_sbi->s_li_wait_mult;
bfff6873
LC		 2790 elr->lr_timeout = timeout;
2791 }
2792 elr->lr_next_sched = jiffies + elr->lr_timeout;
2793 elr->lr_next_group = group + 1;
2794 }
8e8ad8a5 2795 sb_end_write(sb);
bfff6873
LC		 2796
2797 return ret;
2798}
2799
2800/*
2801 * Remove lr_request from the list_request and free the
4ed5c033 2802 * request structure. Should be called with li_list_mtx held
bfff6873
LC		 2803 */
2804static void ext4_remove_li_request(struct ext4_li_request *elr)
2805{
2806 struct ext4_sb_info *sbi;
2807
2808 if (!elr)
2809 return;
2810
2811 sbi = elr->lr_sbi;
2812
2813 list_del(&elr->lr_request);
2814 sbi->s_li_request = NULL;
2815 kfree(elr);
2816}
2817
2818static void ext4_unregister_li_request(struct super_block *sb)
2819{
1bb933fb
LC		 2820 mutex_lock(&ext4_li_mtx);
2821 if (!ext4_li_info) {
2822 mutex_unlock(&ext4_li_mtx);
bfff6873 2823 return;
1bb933fb 2824 }
bfff6873
LC		 2825
2826 mutex_lock(&ext4_li_info->li_list_mtx);
1bb933fb 2827 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
bfff6873 2828 mutex_unlock(&ext4_li_info->li_list_mtx);
1bb933fb 2829 mutex_unlock(&ext4_li_mtx);
bfff6873
LC		 2830}
2831
8f1f7453
ES		 2832static struct task_struct *ext4_lazyinit_task;
2833
bfff6873
LC		 2834/*
2835 * This is the function where ext4lazyinit thread lives. It walks
2836 * through the request list searching for next scheduled filesystem.
2837 * When such a fs is found, run the lazy initialization request
2838 * (ext4_rn_li_request) and keep track of the time spend in this
2839 * function. Based on that time we compute next schedule time of
2840 * the request. When walking through the list is complete, compute
2841 * next waking time and put itself into sleep.
2842 */
2843static int ext4_lazyinit_thread(void *arg)
2844{
2845 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2846 struct list_head *pos, *n;
2847 struct ext4_li_request *elr;
4ed5c033 2848 unsigned long next_wakeup, cur;
bfff6873
LC		 2849
2850 BUG_ON(NULL == eli);
2851
bfff6873
LC		 2852cont_thread:
2853 while (true) {
2854 next_wakeup = MAX_JIFFY_OFFSET;
2855
2856 mutex_lock(&eli->li_list_mtx);
2857 if (list_empty(&eli->li_request_list)) {
2858 mutex_unlock(&eli->li_list_mtx);
2859 goto exit_thread;
2860 }
2861
2862 list_for_each_safe(pos, n, &eli->li_request_list) {
2863 elr = list_entry(pos, struct ext4_li_request,
2864 lr_request);
2865
b2c78cd0
TT		 2866 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2867 if (ext4_run_li_request(elr) != 0) {
2868 /* error, remove the lazy_init job */
2869 ext4_remove_li_request(elr);
2870 continue;
2871 }
bfff6873
LC		 2872 }
2873
2874 if (time_before(elr->lr_next_sched, next_wakeup))
2875 next_wakeup = elr->lr_next_sched;
2876 }
2877 mutex_unlock(&eli->li_list_mtx);
2878
a0acae0e 2879 try_to_freeze();
bfff6873 2880
4ed5c033
LC		 2881 cur = jiffies;
2882 if ((time_after_eq(cur, next_wakeup)) ||
f4245bd4 2883 (MAX_JIFFY_OFFSET == next_wakeup)) {
bfff6873
LC		 2884 cond_resched();
2885 continue;
2886 }
2887
4ed5c033
LC		 2888 schedule_timeout_interruptible(next_wakeup - cur);
2889
8f1f7453
ES		 2890 if (kthread_should_stop()) {
2891 ext4_clear_request_list();
2892 goto exit_thread;
2893 }
bfff6873
LC		 2894 }
2895
2896exit_thread:
2897 /*
2898 * It looks like the request list is empty, but we need
2899 * to check it under the li_list_mtx lock, to prevent any
2900 * additions into it, and of course we should lock ext4_li_mtx
2901 * to atomically free the list and ext4_li_info, because at
2902 * this point another ext4 filesystem could be registering
2903 * new one.
2904 */
2905 mutex_lock(&ext4_li_mtx);
2906 mutex_lock(&eli->li_list_mtx);
2907 if (!list_empty(&eli->li_request_list)) {
2908 mutex_unlock(&eli->li_list_mtx);
2909 mutex_unlock(&ext4_li_mtx);
2910 goto cont_thread;
2911 }
2912 mutex_unlock(&eli->li_list_mtx);
bfff6873
LC		 2913 kfree(ext4_li_info);
2914 ext4_li_info = NULL;
2915 mutex_unlock(&ext4_li_mtx);
2916
2917 return 0;
2918}
2919
2920static void ext4_clear_request_list(void)
2921{
2922 struct list_head *pos, *n;
2923 struct ext4_li_request *elr;
2924
2925 mutex_lock(&ext4_li_info->li_list_mtx);
bfff6873
LC		 2926 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2927 elr = list_entry(pos, struct ext4_li_request,
2928 lr_request);
2929 ext4_remove_li_request(elr);
2930 }
2931 mutex_unlock(&ext4_li_info->li_list_mtx);
2932}
2933
2934static int ext4_run_lazyinit_thread(void)
2935{
8f1f7453
ES		 2936 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2937 ext4_li_info, "ext4lazyinit");
2938 if (IS_ERR(ext4_lazyinit_task)) {
2939 int err = PTR_ERR(ext4_lazyinit_task);
bfff6873 2940 ext4_clear_request_list();
bfff6873
LC		 2941 kfree(ext4_li_info);
2942 ext4_li_info = NULL;
92b97816 2943 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
bfff6873
LC		 2944 "initialization thread\n",
2945 err);
2946 return err;
2947 }
2948 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
bfff6873
LC		 2949 return 0;
2950}
2951
2952/*
2953 * Check whether it make sense to run itable init. thread or not.
2954 * If there is at least one uninitialized inode table, return
2955 * corresponding group number, else the loop goes through all
2956 * groups and return total number of groups.
2957 */
2958static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2959{
2960 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2961 struct ext4_group_desc *gdp = NULL;
2962
2963 for (group = 0; group < ngroups; group++) {
2964 gdp = ext4_get_group_desc(sb, group, NULL);
2965 if (!gdp)
2966 continue;
2967
2968 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2969 break;
2970 }
2971
2972 return group;
2973}
2974
2975static int ext4_li_info_new(void)
2976{
2977 struct ext4_lazy_init *eli = NULL;
2978
2979 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2980 if (!eli)
2981 return -ENOMEM;
2982
bfff6873
LC		 2983 INIT_LIST_HEAD(&eli->li_request_list);
2984 mutex_init(&eli->li_list_mtx);
2985
bfff6873
LC		 2986 eli->li_state |= EXT4_LAZYINIT_QUIT;
2987
2988 ext4_li_info = eli;
2989
2990 return 0;
2991}
2992
2993static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
2994 ext4_group_t start)
2995{
2996 struct ext4_sb_info *sbi = EXT4_SB(sb);
2997 struct ext4_li_request *elr;
2998 unsigned long rnd;
2999
3000 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3001 if (!elr)
3002 return NULL;
3003
3004 elr->lr_super = sb;
3005 elr->lr_sbi = sbi;
3006 elr->lr_next_group = start;
3007
3008 /*
3009 * Randomize first schedule time of the request to
3010 * spread the inode table initialization requests
3011 * better.
3012 */
3013 get_random_bytes(&rnd, sizeof(rnd));
3014 elr->lr_next_sched = jiffies + (unsigned long)rnd %
3015 (EXT4_DEF_LI_MAX_START_DELAY * HZ);
3016
3017 return elr;
3018}
3019
7f511862
TT		 3020int ext4_register_li_request(struct super_block *sb,
3021 ext4_group_t first_not_zeroed)
bfff6873
LC		 3022{
3023 struct ext4_sb_info *sbi = EXT4_SB(sb);
7f511862 3024 struct ext4_li_request *elr = NULL;
bfff6873 3025 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
6c5a6cb9 3026 int ret = 0;
bfff6873 3027
7f511862 3028 mutex_lock(&ext4_li_mtx);
51ce6511
LC		 3029 if (sbi->s_li_request != NULL) {
3030 /*
3031 * Reset timeout so it can be computed again, because
3032 * s_li_wait_mult might have changed.
3033 */
3034 sbi->s_li_request->lr_timeout = 0;
7f511862 3035 goto out;
51ce6511 3036 }
bfff6873
LC		 3037
3038 if (first_not_zeroed == ngroups ||
3039 (sb->s_flags & MS_RDONLY) ||
55ff3840 3040 !test_opt(sb, INIT_INODE_TABLE))
7f511862 3041 goto out;
bfff6873
LC		 3042
3043 elr = ext4_li_request_new(sb, first_not_zeroed);
7f511862
TT		 3044 if (!elr) {
3045 ret = -ENOMEM;
3046 goto out;
3047 }
bfff6873
LC		 3048
3049 if (NULL == ext4_li_info) {
3050 ret = ext4_li_info_new();
3051 if (ret)
3052 goto out;
3053 }
3054
3055 mutex_lock(&ext4_li_info->li_list_mtx);
3056 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3057 mutex_unlock(&ext4_li_info->li_list_mtx);
3058
3059 sbi->s_li_request = elr;
46e4690b
TM		 3060 /*
3061 * set elr to NULL here since it has been inserted to
3062 * the request_list and the removal and free of it is
3063 * handled by ext4_clear_request_list from now on.
3064 */
3065 elr = NULL;
bfff6873
LC		 3066
3067 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3068 ret = ext4_run_lazyinit_thread();
3069 if (ret)
3070 goto out;
3071 }
bfff6873 3072out:
beed5ecb
NK		 3073 mutex_unlock(&ext4_li_mtx);
3074 if (ret)
bfff6873 3075 kfree(elr);
bfff6873
LC		 3076 return ret;
3077}
3078
3079/*
3080 * We do not need to lock anything since this is called on
3081 * module unload.
3082 */
3083static void ext4_destroy_lazyinit_thread(void)
3084{
3085 /*
3086 * If thread exited earlier
3087 * there's nothing to be done.
3088 */
8f1f7453 3089 if (!ext4_li_info || !ext4_lazyinit_task)
bfff6873
LC		 3090 return;
3091
8f1f7453 3092 kthread_stop(ext4_lazyinit_task);
bfff6873
LC		 3093}
3094
25ed6e8a
DW		 3095static int set_journal_csum_feature_set(struct super_block *sb)
3096{
3097 int ret = 1;
3098 int compat, incompat;
3099 struct ext4_sb_info *sbi = EXT4_SB(sb);
3100
3101 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3102 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3103 /* journal checksum v2 */
3104 compat = 0;
3105 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V2;
3106 } else {
3107 /* journal checksum v1 */
3108 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3109 incompat = 0;
3110 }
3111
3112 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3113 ret = jbd2_journal_set_features(sbi->s_journal,
3114 compat, 0,
3115 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3116 incompat);
3117 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3118 ret = jbd2_journal_set_features(sbi->s_journal,
3119 compat, 0,
3120 incompat);
3121 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3122 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3123 } else {
3124 jbd2_journal_clear_features(sbi->s_journal,
3125 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3126 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3127 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3128 }
3129
3130 return ret;
3131}
3132
952fc18e
TT		 3133/*
3134 * Note: calculating the overhead so we can be compatible with
3135 * historical BSD practice is quite difficult in the face of
3136 * clusters/bigalloc. This is because multiple metadata blocks from
3137 * different block group can end up in the same allocation cluster.
3138 * Calculating the exact overhead in the face of clustered allocation
3139 * requires either O(all block bitmaps) in memory or O(number of block
3140 * groups**2) in time. We will still calculate the superblock for
3141 * older file systems --- and if we come across with a bigalloc file
3142 * system with zero in s_overhead_clusters the estimate will be close to
3143 * correct especially for very large cluster sizes --- but for newer
3144 * file systems, it's better to calculate this figure once at mkfs
3145 * time, and store it in the superblock. If the superblock value is
3146 * present (even for non-bigalloc file systems), we will use it.
3147 */
3148static int count_overhead(struct super_block *sb, ext4_group_t grp,
3149 char *buf)
3150{
3151 struct ext4_sb_info *sbi = EXT4_SB(sb);
3152 struct ext4_group_desc *gdp;
3153 ext4_fsblk_t first_block, last_block, b;
3154 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3155 int s, j, count = 0;
3156
0548bbb8
TT		 3157 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC))
3158 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3159 sbi->s_itb_per_group + 2);
3160
952fc18e
TT		 3161 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3162 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3163 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3164 for (i = 0; i < ngroups; i++) {
3165 gdp = ext4_get_group_desc(sb, i, NULL);
3166 b = ext4_block_bitmap(sb, gdp);
3167 if (b >= first_block && b <= last_block) {
3168 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3169 count++;
3170 }
3171 b = ext4_inode_bitmap(sb, gdp);
3172 if (b >= first_block && b <= last_block) {
3173 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3174 count++;
3175 }
3176 b = ext4_inode_table(sb, gdp);
3177 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3178 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3179 int c = EXT4_B2C(sbi, b - first_block);
3180 ext4_set_bit(c, buf);
3181 count++;
3182 }
3183 if (i != grp)
3184 continue;
3185 s = 0;
3186 if (ext4_bg_has_super(sb, grp)) {
3187 ext4_set_bit(s++, buf);
3188 count++;
3189 }
3190 for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
3191 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3192 count++;
3193 }
3194 }
3195 if (!count)
3196 return 0;
3197 return EXT4_CLUSTERS_PER_GROUP(sb) -
3198 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3199}
3200
3201/*
3202 * Compute the overhead and stash it in sbi->s_overhead
3203 */
3204int ext4_calculate_overhead(struct super_block *sb)
3205{
3206 struct ext4_sb_info *sbi = EXT4_SB(sb);
3207 struct ext4_super_block *es = sbi->s_es;
3208 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3209 ext4_fsblk_t overhead = 0;
3210 char *buf = (char *) get_zeroed_page(GFP_KERNEL);
3211
952fc18e
TT		 3212 if (!buf)
3213 return -ENOMEM;
3214
3215 /*
3216 * Compute the overhead (FS structures). This is constant
3217 * for a given filesystem unless the number of block groups
3218 * changes so we cache the previous value until it does.
3219 */
3220
3221 /*
3222 * All of the blocks before first_data_block are overhead
3223 */
3224 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3225
3226 /*
3227 * Add the overhead found in each block group
3228 */
3229 for (i = 0; i < ngroups; i++) {
3230 int blks;
3231
3232 blks = count_overhead(sb, i, buf);
3233 overhead += blks;
3234 if (blks)
3235 memset(buf, 0, PAGE_SIZE);
3236 cond_resched();
3237 }
0875a2b4
ES		 3238 /* Add the journal blocks as well */
3239 if (sbi->s_journal)
3240 overhead += EXT4_B2C(sbi, sbi->s_journal->j_maxlen);
3241
952fc18e
TT		 3242 sbi->s_overhead = overhead;
3243 smp_wmb();
3244 free_page((unsigned long) buf);
3245 return 0;
3246}
3247
2b2d6d01 3248static int ext4_fill_super(struct super_block *sb, void *data, int silent)
ac27a0ec 3249{
d4c402d9 3250 char *orig_data = kstrdup(data, GFP_KERNEL);
2b2d6d01 3251 struct buffer_head *bh;
617ba13b
MC		 3252 struct ext4_super_block *es = NULL;
3253 struct ext4_sb_info *sbi;
3254 ext4_fsblk_t block;
3255 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 3256 ext4_fsblk_t logical_sb_block;
ac27a0ec 3257 unsigned long offset = 0;
ac27a0ec
DK		 3258 unsigned long journal_devnum = 0;
3259 unsigned long def_mount_opts;
3260 struct inode *root;
9f6200bb 3261 char *cp;
0390131b 3262 const char *descr;
dcc7dae3 3263 int ret = -ENOMEM;
281b5995 3264 int blocksize, clustersize;
4ec11028
TT		 3265 unsigned int db_count;
3266 unsigned int i;
281b5995 3267 int needs_recovery, has_huge_files, has_bigalloc;
bd81d8ee 3268 __u64 blocks_count;
07aa2ea1 3269 int err = 0;
b3881f74 3270 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
bfff6873 3271 ext4_group_t first_not_zeroed;
ac27a0ec
DK		 3272
3273 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3274 if (!sbi)
dcc7dae3 3275 goto out_free_orig;
705895b6
PE		 3276
3277 sbi->s_blockgroup_lock =
3278 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3279 if (!sbi->s_blockgroup_lock) {
3280 kfree(sbi);
dcc7dae3 3281 goto out_free_orig;
705895b6 3282 }
ac27a0ec 3283 sb->s_fs_info = sbi;
2c0544b2 3284 sbi->s_sb = sb;
240799cd 3285 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
d9c9bef1 3286 sbi->s_sb_block = sb_block;
f613dfcb
TT		 3287 if (sb->s_bdev->bd_part)
3288 sbi->s_sectors_written_start =
3289 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
ac27a0ec 3290
9f6200bb
TT		 3291 /* Cleanup superblock name */
3292 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3293 *cp = '!';
3294
07aa2ea1 3295 /* -EINVAL is default */
dcc7dae3 3296 ret = -EINVAL;
617ba13b 3297 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 3298 if (!blocksize) {
b31e1552 3299 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
ac27a0ec
DK		 3300 goto out_fail;
3301 }
3302
3303 /*
617ba13b 3304 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 3305 * block sizes. We need to calculate the offset from buffer start.
3306 */
617ba13b 3307 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 3308 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3309 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 3310 } else {
70bbb3e0 3311 logical_sb_block = sb_block;
ac27a0ec
DK		 3312 }
3313
70bbb3e0 3314 if (!(bh = sb_bread(sb, logical_sb_block))) {
b31e1552 3315 ext4_msg(sb, KERN_ERR, "unable to read superblock");
ac27a0ec
DK		 3316 goto out_fail;
3317 }
3318 /*
3319 * Note: s_es must be initialized as soon as possible because
617ba13b 3320 * some ext4 macro-instructions depend on its value
ac27a0ec 3321 */
2716b802 3322 es = (struct ext4_super_block *) (bh->b_data + offset);
ac27a0ec
DK		 3323 sbi->s_es = es;
3324 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 3325 if (sb->s_magic != EXT4_SUPER_MAGIC)
3326 goto cantfind_ext4;
afc32f7e 3327 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
ac27a0ec 3328
feb0ab32
DW		 3329 /* Warn if metadata_csum and gdt_csum are both set. */
3330 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3331 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
3332 EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_GDT_CSUM))
3333 ext4_warning(sb, KERN_INFO "metadata_csum and uninit_bg are "
3334 "redundant flags; please run fsck.");
3335
d25425f8
DW		 3336 /* Check for a known checksum algorithm */
3337 if (!ext4_verify_csum_type(sb, es)) {
3338 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3339 "unknown checksum algorithm.");
3340 silent = 1;
3341 goto cantfind_ext4;
3342 }
3343
0441984a
DW		 3344 /* Load the checksum driver */
3345 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3346 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3347 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3348 if (IS_ERR(sbi->s_chksum_driver)) {
3349 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3350 ret = PTR_ERR(sbi->s_chksum_driver);
3351 sbi->s_chksum_driver = NULL;
3352 goto failed_mount;
3353 }
3354 }
3355
a9c47317
DW		 3356 /* Check superblock checksum */
3357 if (!ext4_superblock_csum_verify(sb, es)) {
3358 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3359 "invalid superblock checksum. Run e2fsck?");
3360 silent = 1;
3361 goto cantfind_ext4;
3362 }
3363
3364 /* Precompute checksum seed for all metadata */
3365 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3366 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
3367 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3368 sizeof(es->s_uuid));
3369
ac27a0ec
DK		 3370 /* Set defaults before we parse the mount options */
3371 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
fd8c37ec 3372 set_opt(sb, INIT_INODE_TABLE);
617ba13b 3373 if (def_mount_opts & EXT4_DEFM_DEBUG)
fd8c37ec 3374 set_opt(sb, DEBUG);
87f26807 3375 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
fd8c37ec 3376 set_opt(sb, GRPID);
617ba13b 3377 if (def_mount_opts & EXT4_DEFM_UID16)
fd8c37ec 3378 set_opt(sb, NO_UID32);
ea663336 3379 /* xattr user namespace & acls are now defaulted on */
ea663336 3380 set_opt(sb, XATTR_USER);
03010a33 3381#ifdef CONFIG_EXT4_FS_POSIX_ACL
ea663336 3382 set_opt(sb, POSIX_ACL);
2e7842b8 3383#endif
6fd7a467 3384 set_opt(sb, MBLK_IO_SUBMIT);
617ba13b 3385 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
fd8c37ec 3386 set_opt(sb, JOURNAL_DATA);
617ba13b 3387 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
fd8c37ec 3388 set_opt(sb, ORDERED_DATA);
617ba13b 3389 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
fd8c37ec 3390 set_opt(sb, WRITEBACK_DATA);
617ba13b
MC		 3391
3392 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
fd8c37ec 3393 set_opt(sb, ERRORS_PANIC);
bb4f397a 3394 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
fd8c37ec 3395 set_opt(sb, ERRORS_CONT);
bb4f397a 3396 else
fd8c37ec 3397 set_opt(sb, ERRORS_RO);
8b67f04a 3398 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
fd8c37ec 3399 set_opt(sb, BLOCK_VALIDITY);
8b67f04a 3400 if (def_mount_opts & EXT4_DEFM_DISCARD)
fd8c37ec 3401 set_opt(sb, DISCARD);
ac27a0ec 3402
08cefc7a
EB		 3403 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3404 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
30773840
TT		 3405 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3406 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3407 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
ac27a0ec 3408
8b67f04a 3409 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
fd8c37ec 3410 set_opt(sb, BARRIER);
ac27a0ec 3411
dd919b98
AK		 3412 /*
3413 * enable delayed allocation by default
3414 * Use -o nodelalloc to turn it off
3415 */
bc0b75f7 3416 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
8b67f04a 3417 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
fd8c37ec 3418 set_opt(sb, DELALLOC);
dd919b98 3419
51ce6511
LC		 3420 /*
3421 * set default s_li_wait_mult for lazyinit, for the case there is
3422 * no mount option specified.
3423 */
3424 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3425
8b67f04a 3426 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
661aa520 3427 &journal_devnum, &journal_ioprio, 0)) {
8b67f04a
TT		 3428 ext4_msg(sb, KERN_WARNING,
3429 "failed to parse options in superblock: %s",
3430 sbi->s_es->s_mount_opts);
3431 }
5a916be1 3432 sbi->s_def_mount_opt = sbi->s_mount_opt;
b3881f74 3433 if (!parse_options((char *) data, sb, &journal_devnum,
661aa520 3434 &journal_ioprio, 0))
ac27a0ec
DK		 3435 goto failed_mount;
3436
56889787
TT		 3437 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3438 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3439 "with data=journal disables delayed "
3440 "allocation and O_DIRECT support!\n");
3441 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3442 ext4_msg(sb, KERN_ERR, "can't mount with "
3443 "both data=journal and delalloc");
3444 goto failed_mount;
3445 }
3446 if (test_opt(sb, DIOREAD_NOLOCK)) {
3447 ext4_msg(sb, KERN_ERR, "can't mount with "
3448 "both data=journal and delalloc");
3449 goto failed_mount;
3450 }
3451 if (test_opt(sb, DELALLOC))
3452 clear_opt(sb, DELALLOC);
3453 }
3454
ac27a0ec 3455 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 3456 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec 3457
617ba13b
MC		 3458 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3459 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3460 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3461 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
b31e1552
ES		 3462 ext4_msg(sb, KERN_WARNING,
3463 "feature flags set on rev 0 fs, "
3464 "running e2fsck is recommended");
469108ff 3465
2035e776
TT		 3466 if (IS_EXT2_SB(sb)) {
3467 if (ext2_feature_set_ok(sb))
3468 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3469 "using the ext4 subsystem");
3470 else {
3471 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3472 "to feature incompatibilities");
3473 goto failed_mount;
3474 }
3475 }
3476
3477 if (IS_EXT3_SB(sb)) {
3478 if (ext3_feature_set_ok(sb))
3479 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3480 "using the ext4 subsystem");
3481 else {
3482 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3483 "to feature incompatibilities");
3484 goto failed_mount;
3485 }
3486 }
3487
ac27a0ec
DK		 3488 /*
3489 * Check feature flags regardless of the revision level, since we
3490 * previously didn't change the revision level when setting the flags,
3491 * so there is a chance incompat flags are set on a rev 0 filesystem.
3492 */
a13fb1a4 3493 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
ac27a0ec 3494 goto failed_mount;
a13fb1a4 3495
261cb20c 3496 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
617ba13b
MC		 3497 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3498 blocksize > EXT4_MAX_BLOCK_SIZE) {
b31e1552
ES		 3499 ext4_msg(sb, KERN_ERR,
3500 "Unsupported filesystem blocksize %d", blocksize);
ac27a0ec
DK		 3501 goto failed_mount;
3502 }
3503
ac27a0ec 3504 if (sb->s_blocksize != blocksize) {
ce40733c
AK		 3505 /* Validate the filesystem blocksize */
3506 if (!sb_set_blocksize(sb, blocksize)) {
b31e1552 3507 ext4_msg(sb, KERN_ERR, "bad block size %d",
ce40733c 3508 blocksize);
ac27a0ec
DK		 3509 goto failed_mount;
3510 }
3511
2b2d6d01 3512 brelse(bh);
70bbb3e0
AM		 3513 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3514 offset = do_div(logical_sb_block, blocksize);
3515 bh = sb_bread(sb, logical_sb_block);
ac27a0ec 3516 if (!bh) {
b31e1552
ES		 3517 ext4_msg(sb, KERN_ERR,
3518 "Can't read superblock on 2nd try");
ac27a0ec
DK		 3519 goto failed_mount;
3520 }
2716b802 3521 es = (struct ext4_super_block *)(bh->b_data + offset);
ac27a0ec 3522 sbi->s_es = es;
617ba13b 3523 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
b31e1552
ES		 3524 ext4_msg(sb, KERN_ERR,
3525 "Magic mismatch, very weird!");
ac27a0ec
DK		 3526 goto failed_mount;
3527 }
3528 }
3529
a13fb1a4
ES		 3530 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3531 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
f287a1a5
TT		 3532 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3533 has_huge_files);
3534 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
ac27a0ec 3535
617ba13b
MC		 3536 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3537 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3538 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
ac27a0ec
DK		 3539 } else {
3540 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3541 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
617ba13b 3542 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1330593e 3543 (!is_power_of_2(sbi->s_inode_size)) ||
ac27a0ec 3544 (sbi->s_inode_size > blocksize)) {
b31e1552
ES		 3545 ext4_msg(sb, KERN_ERR,
3546 "unsupported inode size: %d",
2b2d6d01 3547 sbi->s_inode_size);
ac27a0ec
DK		 3548 goto failed_mount;
3549 }
ef7f3835
KS		 3550 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3551 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
ac27a0ec 3552 }
0b8e58a1 3553
0d1ee42f
AR		 3554 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3555 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
8fadc143 3556 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 3557 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 3558 !is_power_of_2(sbi->s_desc_size)) {
b31e1552
ES		 3559 ext4_msg(sb, KERN_ERR,
3560 "unsupported descriptor size %lu",
0d1ee42f
AR		 3561 sbi->s_desc_size);
3562 goto failed_mount;
3563 }
3564 } else
3565 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
0b8e58a1 3566
ac27a0ec 3567 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 3568 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
b47b6f38 3569 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
617ba13b 3570 goto cantfind_ext4;
0b8e58a1 3571
617ba13b 3572 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 3573 if (sbi->s_inodes_per_block == 0)
617ba13b 3574 goto cantfind_ext4;
ac27a0ec
DK		 3575 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3576 sbi->s_inodes_per_block;
0d1ee42f 3577 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 3578 sbi->s_sbh = bh;
3579 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 3580 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3581 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
0b8e58a1 3582
2b2d6d01 3583 for (i = 0; i < 4; i++)
ac27a0ec
DK		 3584 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3585 sbi->s_def_hash_version = es->s_def_hash_version;
f99b2589
TT		 3586 i = le32_to_cpu(es->s_flags);
3587 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3588 sbi->s_hash_unsigned = 3;
3589 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3590#ifdef __CHAR_UNSIGNED__
3591 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3592 sbi->s_hash_unsigned = 3;
3593#else
3594 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3595#endif
f99b2589 3596 }
ac27a0ec 3597
281b5995
TT		 3598 /* Handle clustersize */
3599 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3600 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3601 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3602 if (has_bigalloc) {
3603 if (clustersize < blocksize) {
3604 ext4_msg(sb, KERN_ERR,
3605 "cluster size (%d) smaller than "
3606 "block size (%d)", clustersize, blocksize);
3607 goto failed_mount;
3608 }
3609 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3610 le32_to_cpu(es->s_log_block_size);
3611 sbi->s_clusters_per_group =
3612 le32_to_cpu(es->s_clusters_per_group);
3613 if (sbi->s_clusters_per_group > blocksize * 8) {
3614 ext4_msg(sb, KERN_ERR,
3615 "#clusters per group too big: %lu",
3616 sbi->s_clusters_per_group);
3617 goto failed_mount;
3618 }
3619 if (sbi->s_blocks_per_group !=
3620 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3621 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3622 "clusters per group (%lu) inconsistent",
3623 sbi->s_blocks_per_group,
3624 sbi->s_clusters_per_group);
3625 goto failed_mount;
3626 }
3627 } else {
3628 if (clustersize != blocksize) {
3629 ext4_warning(sb, "fragment/cluster size (%d) != "
3630 "block size (%d)", clustersize,
3631 blocksize);
3632 clustersize = blocksize;
3633 }
3634 if (sbi->s_blocks_per_group > blocksize * 8) {
3635 ext4_msg(sb, KERN_ERR,
3636 "#blocks per group too big: %lu",
3637 sbi->s_blocks_per_group);
3638 goto failed_mount;
3639 }
3640 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3641 sbi->s_cluster_bits = 0;
ac27a0ec 3642 }
281b5995
TT		 3643 sbi->s_cluster_ratio = clustersize / blocksize;
3644
ac27a0ec 3645 if (sbi->s_inodes_per_group > blocksize * 8) {
b31e1552
ES		 3646 ext4_msg(sb, KERN_ERR,
3647 "#inodes per group too big: %lu",
2b2d6d01 3648 sbi->s_inodes_per_group);
ac27a0ec
DK		 3649 goto failed_mount;
3650 }
3651
bf43d84b
ES		 3652 /*
3653 * Test whether we have more sectors than will fit in sector_t,
3654 * and whether the max offset is addressable by the page cache.
3655 */
5a9ae68a 3656 err = generic_check_addressable(sb->s_blocksize_bits,
30ca22c7 3657 ext4_blocks_count(es));
5a9ae68a 3658 if (err) {
b31e1552 3659 ext4_msg(sb, KERN_ERR, "filesystem"
bf43d84b 3660 " too large to mount safely on this system");
ac27a0ec 3661 if (sizeof(sector_t) < 8)
90c699a9 3662 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
ac27a0ec
DK		 3663 goto failed_mount;
3664 }
3665
617ba13b
MC		 3666 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3667 goto cantfind_ext4;
e7c95593 3668
0f2ddca6
FTN		 3669 /* check blocks count against device size */
3670 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3671 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
b31e1552
ES		 3672 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3673 "exceeds size of device (%llu blocks)",
0f2ddca6
FTN		 3674 ext4_blocks_count(es), blocks_count);
3675 goto failed_mount;
3676 }
3677
0b8e58a1
AD		 3678 /*
3679 * It makes no sense for the first data block to be beyond the end
3680 * of the filesystem.
3681 */
3682 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
5635a62b 3683 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
b31e1552
ES		 3684 "block %u is beyond end of filesystem (%llu)",
3685 le32_to_cpu(es->s_first_data_block),
3686 ext4_blocks_count(es));
e7c95593
ES		 3687 goto failed_mount;
3688 }
bd81d8ee
LV		 3689 blocks_count = (ext4_blocks_count(es) -
3690 le32_to_cpu(es->s_first_data_block) +
3691 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3692 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4ec11028 3693 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
b31e1552 3694 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
4ec11028 3695 "(block count %llu, first data block %u, "
b31e1552 3696 "blocks per group %lu)", sbi->s_groups_count,
4ec11028
TT		 3697 ext4_blocks_count(es),
3698 le32_to_cpu(es->s_first_data_block),
3699 EXT4_BLOCKS_PER_GROUP(sb));
3700 goto failed_mount;
3701 }
bd81d8ee 3702 sbi->s_groups_count = blocks_count;
fb0a387d
ES		 3703 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3704 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
617ba13b
MC		 3705 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3706 EXT4_DESC_PER_BLOCK(sb);
f18a5f21
TT		 3707 sbi->s_group_desc = ext4_kvmalloc(db_count *
3708 sizeof(struct buffer_head *),
3709 GFP_KERNEL);
ac27a0ec 3710 if (sbi->s_group_desc == NULL) {
b31e1552 3711 ext4_msg(sb, KERN_ERR, "not enough memory");
2cde417d 3712 ret = -ENOMEM;
ac27a0ec
DK		 3713 goto failed_mount;
3714 }
3715
9f6200bb
TT		 3716 if (ext4_proc_root)
3717 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
240799cd 3718
66acdcf4
TT		 3719 if (sbi->s_proc)
3720 proc_create_data("options", S_IRUGO, sbi->s_proc,
3721 &ext4_seq_options_fops, sb);
3722
705895b6 3723 bgl_lock_init(sbi->s_blockgroup_lock);
ac27a0ec
DK		 3724
3725 for (i = 0; i < db_count; i++) {
70bbb3e0 3726 block = descriptor_loc(sb, logical_sb_block, i);
ac27a0ec
DK		 3727 sbi->s_group_desc[i] = sb_bread(sb, block);
3728 if (!sbi->s_group_desc[i]) {
b31e1552
ES		 3729 ext4_msg(sb, KERN_ERR,
3730 "can't read group descriptor %d", i);
ac27a0ec
DK		 3731 db_count = i;
3732 goto failed_mount2;
3733 }
3734 }
bfff6873 3735 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
b31e1552 3736 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
ac27a0ec
DK		 3737 goto failed_mount2;
3738 }
772cb7c8
JS		 3739 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3740 if (!ext4_fill_flex_info(sb)) {
b31e1552
ES		 3741 ext4_msg(sb, KERN_ERR,
3742 "unable to initialize "
3743 "flex_bg meta info!");
772cb7c8
JS		 3744 goto failed_mount2;
3745 }
3746
ac27a0ec
DK		 3747 sbi->s_gdb_count = db_count;
3748 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3749 spin_lock_init(&sbi->s_next_gen_lock);
3750
04496411
TM		 3751 init_timer(&sbi->s_err_report);
3752 sbi->s_err_report.function = print_daily_error_info;
3753 sbi->s_err_report.data = (unsigned long) sb;
3754
57042651 3755 err = percpu_counter_init(&sbi->s_freeclusters_counter,
5dee5437 3756 ext4_count_free_clusters(sb));
ce7e010a
TT		 3757 if (!err) {
3758 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3759 ext4_count_free_inodes(sb));
3760 }
3761 if (!err) {
3762 err = percpu_counter_init(&sbi->s_dirs_counter,
3763 ext4_count_dirs(sb));
3764 }
3765 if (!err) {
57042651 3766 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0);
ce7e010a
TT		 3767 }
3768 if (err) {
3769 ext4_msg(sb, KERN_ERR, "insufficient memory");
3770 goto failed_mount3;
3771 }
3772
c9de560d 3773 sbi->s_stripe = ext4_get_stripe_size(sbi);
55138e0b 3774 sbi->s_max_writeback_mb_bump = 128;
67a5da56 3775 sbi->s_extent_max_zeroout_kb = 32;
c9de560d 3776
ac27a0ec
DK		 3777 /*
3778 * set up enough so that it can read an inode
3779 */
9ca92389
TT		 3780 if (!test_opt(sb, NOLOAD) &&
3781 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3782 sb->s_op = &ext4_sops;
3783 else
3784 sb->s_op = &ext4_nojournal_sops;
617ba13b
MC		 3785 sb->s_export_op = &ext4_export_ops;
3786 sb->s_xattr = ext4_xattr_handlers;
ac27a0ec 3787#ifdef CONFIG_QUOTA
617ba13b
MC		 3788 sb->s_qcop = &ext4_qctl_operations;
3789 sb->dq_op = &ext4_quota_operations;
7c319d32
AK		 3790
3791 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) {
3792 /* Use qctl operations for hidden quota files. */
3793 sb->s_qcop = &ext4_qctl_sysfile_operations;
3794 }
ac27a0ec 3795#endif
f2fa2ffc
AK		 3796 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3797
ac27a0ec 3798 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3b9d4ed2 3799 mutex_init(&sbi->s_orphan_lock);
ac27a0ec
DK		 3800
3801 sb->s_root = NULL;
3802
3803 needs_recovery = (es->s_last_orphan != 0 ||
617ba13b
MC		 3804 EXT4_HAS_INCOMPAT_FEATURE(sb,
3805 EXT4_FEATURE_INCOMPAT_RECOVER));
ac27a0ec 3806
c5e06d10
JL		 3807 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3808 !(sb->s_flags & MS_RDONLY))
3809 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3810 goto failed_mount3;
3811
ac27a0ec
DK		 3812 /*
3813 * The first inode we look at is the journal inode. Don't try
3814 * root first: it may be modified in the journal!
3815 */
3816 if (!test_opt(sb, NOLOAD) &&
617ba13b
MC		 3817 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3818 if (ext4_load_journal(sb, es, journal_devnum))
ac27a0ec 3819 goto failed_mount3;
0390131b
FM		 3820 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3821 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
b31e1552
ES		 3822 ext4_msg(sb, KERN_ERR, "required journal recovery "
3823 "suppressed and not mounted read-only");
744692dc 3824 goto failed_mount_wq;
ac27a0ec 3825 } else {
fd8c37ec 3826 clear_opt(sb, DATA_FLAGS);
0390131b
FM		 3827 sbi->s_journal = NULL;
3828 needs_recovery = 0;
3829 goto no_journal;
ac27a0ec
DK		 3830 }
3831
f32aaf2d 3832 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT) &&
eb40a09c
JS		 3833 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3834 JBD2_FEATURE_INCOMPAT_64BIT)) {
b31e1552 3835 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
744692dc 3836 goto failed_mount_wq;
eb40a09c
JS		 3837 }
3838
25ed6e8a
DW		 3839 if (!set_journal_csum_feature_set(sb)) {
3840 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
3841 "feature set");
3842 goto failed_mount_wq;
d4da6c9c 3843 }
818d276c 3844
ac27a0ec
DK		 3845 /* We have now updated the journal if required, so we can
3846 * validate the data journaling mode. */
3847 switch (test_opt(sb, DATA_FLAGS)) {
3848 case 0:
3849 /* No mode set, assume a default based on the journal
63f57933
AM		 3850 * capabilities: ORDERED_DATA if the journal can
3851 * cope, else JOURNAL_DATA
3852 */
dab291af
MC		 3853 if (jbd2_journal_check_available_features
3854 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
fd8c37ec 3855 set_opt(sb, ORDERED_DATA);
ac27a0ec 3856 else
fd8c37ec 3857 set_opt(sb, JOURNAL_DATA);
ac27a0ec
DK		 3858 break;
3859
617ba13b
MC		 3860 case EXT4_MOUNT_ORDERED_DATA:
3861 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 3862 if (!jbd2_journal_check_available_features
3863 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
b31e1552
ES		 3864 ext4_msg(sb, KERN_ERR, "Journal does not support "
3865 "requested data journaling mode");
744692dc 3866 goto failed_mount_wq;
ac27a0ec
DK		 3867 }
3868 default:
3869 break;
3870 }
b3881f74 3871 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
ac27a0ec 3872
18aadd47
BJ		 3873 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
3874
ce7e010a
TT		 3875 /*
3876 * The journal may have updated the bg summary counts, so we
3877 * need to update the global counters.
3878 */
57042651 3879 percpu_counter_set(&sbi->s_freeclusters_counter,
5dee5437 3880 ext4_count_free_clusters(sb));
ce7e010a
TT		 3881 percpu_counter_set(&sbi->s_freeinodes_counter,
3882 ext4_count_free_inodes(sb));
3883 percpu_counter_set(&sbi->s_dirs_counter,
3884 ext4_count_dirs(sb));
57042651 3885 percpu_counter_set(&sbi->s_dirtyclusters_counter, 0);
206f7ab4 3886
ce7e010a 3887no_journal:
952fc18e
TT		 3888 /*
3889 * Get the # of file system overhead blocks from the
3890 * superblock if present.
3891 */
3892 if (es->s_overhead_clusters)
3893 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
3894 else {
07aa2ea1
LC		 3895 err = ext4_calculate_overhead(sb);
3896 if (err)
952fc18e
TT		 3897 goto failed_mount_wq;
3898 }
3899
fd89d5f2
TH		 3900 /*
3901 * The maximum number of concurrent works can be high and
3902 * concurrency isn't really necessary. Limit it to 1.
3903 */
3904 EXT4_SB(sb)->dio_unwritten_wq =
ae005cbe 3905 alloc_workqueue("ext4-dio-unwritten", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4c0425ff
MC		 3906 if (!EXT4_SB(sb)->dio_unwritten_wq) {
3907 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
07aa2ea1 3908 ret = -ENOMEM;
4c0425ff
MC		 3909 goto failed_mount_wq;
3910 }
3911
ac27a0ec 3912 /*
dab291af 3913 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 3914 * so we can safely mount the rest of the filesystem now.
3915 */
3916
1d1fe1ee
DH		 3917 root = ext4_iget(sb, EXT4_ROOT_INO);
3918 if (IS_ERR(root)) {
b31e1552 3919 ext4_msg(sb, KERN_ERR, "get root inode failed");
1d1fe1ee 3920 ret = PTR_ERR(root);
32a9bb57 3921 root = NULL;
ac27a0ec
DK		 3922 goto failed_mount4;
3923 }
3924 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
b31e1552 3925 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
94bf608a 3926 iput(root);
ac27a0ec
DK		 3927 goto failed_mount4;
3928 }
48fde701 3929 sb->s_root = d_make_root(root);
1d1fe1ee 3930 if (!sb->s_root) {
b31e1552 3931 ext4_msg(sb, KERN_ERR, "get root dentry failed");
1d1fe1ee
DH		 3932 ret = -ENOMEM;
3933 goto failed_mount4;
3934 }
ac27a0ec 3935
7e84b621
ES		 3936 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
3937 sb->s_flags |= MS_RDONLY;
ef7f3835
KS		 3938
3939 /* determine the minimum size of new large inodes, if present */
3940 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3941 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3942 EXT4_GOOD_OLD_INODE_SIZE;
3943 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3944 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
3945 if (sbi->s_want_extra_isize <
3946 le16_to_cpu(es->s_want_extra_isize))
3947 sbi->s_want_extra_isize =
3948 le16_to_cpu(es->s_want_extra_isize);
3949 if (sbi->s_want_extra_isize <
3950 le16_to_cpu(es->s_min_extra_isize))
3951 sbi->s_want_extra_isize =
3952 le16_to_cpu(es->s_min_extra_isize);
3953 }
3954 }
3955 /* Check if enough inode space is available */
3956 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3957 sbi->s_inode_size) {
3958 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3959 EXT4_GOOD_OLD_INODE_SIZE;
b31e1552
ES		 3960 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3961 "available");
ef7f3835
KS		 3962 }
3963
6fd058f7
TT		 3964 err = ext4_setup_system_zone(sb);
3965 if (err) {
b31e1552 3966 ext4_msg(sb, KERN_ERR, "failed to initialize system "
fbe845dd 3967 "zone (%d)", err);
94bf608a 3968 goto failed_mount4a;
6fd058f7
TT		 3969 }
3970
c2774d84 3971 ext4_ext_init(sb);
9d99012f 3972 err = ext4_mb_init(sb);
c2774d84 3973 if (err) {
421f91d2 3974 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
b31e1552 3975 err);
dcf2d804 3976 goto failed_mount5;
c2774d84
AK		 3977 }
3978
bfff6873
LC		 3979 err = ext4_register_li_request(sb, first_not_zeroed);
3980 if (err)
dcf2d804 3981 goto failed_mount6;
bfff6873 3982
3197ebdb
TT		 3983 sbi->s_kobj.kset = ext4_kset;
3984 init_completion(&sbi->s_kobj_unregister);
3985 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
3986 "%s", sb->s_id);
dcf2d804
TM		 3987 if (err)
3988 goto failed_mount7;
3197ebdb 3989
617ba13b
MC		 3990 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3991 ext4_orphan_cleanup(sb, es);
3992 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
0390131b 3993 if (needs_recovery) {
b31e1552 3994 ext4_msg(sb, KERN_INFO, "recovery complete");
0390131b
FM		 3995 ext4_mark_recovery_complete(sb, es);
3996 }
3997 if (EXT4_SB(sb)->s_journal) {
3998 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3999 descr = " journalled data mode";
4000 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4001 descr = " ordered data mode";
4002 else
4003 descr = " writeback data mode";
4004 } else
4005 descr = "out journal";
4006
7c319d32
AK		 4007#ifdef CONFIG_QUOTA
4008 /* Enable quota usage during mount. */
4009 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
4010 !(sb->s_flags & MS_RDONLY)) {
07aa2ea1
LC		 4011 err = ext4_enable_quotas(sb);
4012 if (err)
72ba7450 4013 goto failed_mount8;
7c319d32
AK		 4014 }
4015#endif /* CONFIG_QUOTA */
4016
79add3a3
LC		 4017 if (test_opt(sb, DISCARD)) {
4018 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4019 if (!blk_queue_discard(q))
4020 ext4_msg(sb, KERN_WARNING,
4021 "mounting with \"discard\" option, but "
4022 "the device does not support discard");
4023 }
4024
d4c402d9 4025 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
8b67f04a
TT		 4026 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
4027 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
ac27a0ec 4028
66e61a9e
TT		 4029 if (es->s_error_count)
4030 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
ac27a0ec 4031
d4c402d9 4032 kfree(orig_data);
ac27a0ec
DK		 4033 return 0;
4034
617ba13b 4035cantfind_ext4:
ac27a0ec 4036 if (!silent)
b31e1552 4037 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
ac27a0ec
DK		 4038 goto failed_mount;
4039
72ba7450
TT		 4040#ifdef CONFIG_QUOTA
4041failed_mount8:
4042 kobject_del(&sbi->s_kobj);
4043#endif
dcf2d804
TM		 4044failed_mount7:
4045 ext4_unregister_li_request(sb);
4046failed_mount6:
dcf2d804 4047 ext4_mb_release(sb);
94bf608a
AV		 4048failed_mount5:
4049 ext4_ext_release(sb);
dcf2d804 4050 ext4_release_system_zone(sb);
94bf608a
AV		 4051failed_mount4a:
4052 dput(sb->s_root);
32a9bb57 4053 sb->s_root = NULL;
94bf608a 4054failed_mount4:
b31e1552 4055 ext4_msg(sb, KERN_ERR, "mount failed");
4c0425ff
MC		 4056 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
4057failed_mount_wq:
0390131b
FM		 4058 if (sbi->s_journal) {
4059 jbd2_journal_destroy(sbi->s_journal);
4060 sbi->s_journal = NULL;
4061 }
ac27a0ec 4062failed_mount3:
04496411 4063 del_timer(&sbi->s_err_report);
9933fc0a
TT		 4064 if (sbi->s_flex_groups)
4065 ext4_kvfree(sbi->s_flex_groups);
57042651 4066 percpu_counter_destroy(&sbi->s_freeclusters_counter);
ce7e010a
TT		 4067 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4068 percpu_counter_destroy(&sbi->s_dirs_counter);
57042651 4069 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
c5e06d10
JL		 4070 if (sbi->s_mmp_tsk)
4071 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec
DK		 4072failed_mount2:
4073 for (i = 0; i < db_count; i++)
4074 brelse(sbi->s_group_desc[i]);
f18a5f21 4075 ext4_kvfree(sbi->s_group_desc);
ac27a0ec 4076failed_mount:
0441984a
DW		 4077 if (sbi->s_chksum_driver)
4078 crypto_free_shash(sbi->s_chksum_driver);
240799cd 4079 if (sbi->s_proc) {
66acdcf4 4080 remove_proc_entry("options", sbi->s_proc);
9f6200bb 4081 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 4082 }
ac27a0ec
DK		 4083#ifdef CONFIG_QUOTA
4084 for (i = 0; i < MAXQUOTAS; i++)
4085 kfree(sbi->s_qf_names[i]);
4086#endif
617ba13b 4087 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 4088 brelse(bh);
4089out_fail:
4090 sb->s_fs_info = NULL;
f6830165 4091 kfree(sbi->s_blockgroup_lock);
ac27a0ec 4092 kfree(sbi);
dcc7dae3 4093out_free_orig:
d4c402d9 4094 kfree(orig_data);
07aa2ea1 4095 return err ? err : ret;
ac27a0ec
DK		 4096}
4097
4098/*
4099 * Setup any per-fs journal parameters now. We'll do this both on
4100 * initial mount, once the journal has been initialised but before we've
4101 * done any recovery; and again on any subsequent remount.
4102 */
617ba13b 4103static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 4104{
617ba13b 4105 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4106
30773840
TT		 4107 journal->j_commit_interval = sbi->s_commit_interval;
4108 journal->j_min_batch_time = sbi->s_min_batch_time;
4109 journal->j_max_batch_time = sbi->s_max_batch_time;
ac27a0ec 4110
a931da6a 4111 write_lock(&journal->j_state_lock);
ac27a0ec 4112 if (test_opt(sb, BARRIER))
dab291af 4113 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 4114 else
dab291af 4115 journal->j_flags &= ~JBD2_BARRIER;
5bf5683a
HK		 4116 if (test_opt(sb, DATA_ERR_ABORT))
4117 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4118 else
4119 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
a931da6a 4120 write_unlock(&journal->j_state_lock);
ac27a0ec
DK		 4121}
4122
617ba13b 4123static journal_t *ext4_get_journal(struct super_block *sb,
ac27a0ec
DK		 4124 unsigned int journal_inum)
4125{
4126 struct inode *journal_inode;
4127 journal_t *journal;
4128
0390131b
FM		 4129 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4130
ac27a0ec
DK		 4131 /* First, test for the existence of a valid inode on disk. Bad
4132 * things happen if we iget() an unused inode, as the subsequent
4133 * iput() will try to delete it. */
4134
1d1fe1ee
DH		 4135 journal_inode = ext4_iget(sb, journal_inum);
4136 if (IS_ERR(journal_inode)) {
b31e1552 4137 ext4_msg(sb, KERN_ERR, "no journal found");
ac27a0ec
DK		 4138 return NULL;
4139 }
4140 if (!journal_inode->i_nlink) {
4141 make_bad_inode(journal_inode);
4142 iput(journal_inode);
b31e1552 4143 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
ac27a0ec
DK		 4144 return NULL;
4145 }
4146
e5f8eab8 4147 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
ac27a0ec 4148 journal_inode, journal_inode->i_size);
1d1fe1ee 4149 if (!S_ISREG(journal_inode->i_mode)) {
b31e1552 4150 ext4_msg(sb, KERN_ERR, "invalid journal inode");
ac27a0ec
DK		 4151 iput(journal_inode);
4152 return NULL;
4153 }
4154
dab291af 4155 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 4156 if (!journal) {
b31e1552 4157 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
ac27a0ec
DK		 4158 iput(journal_inode);
4159 return NULL;
4160 }
4161 journal->j_private = sb;
617ba13b 4162 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 4163 return journal;
4164}
4165
617ba13b 4166static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 4167 dev_t j_dev)
4168{
2b2d6d01 4169 struct buffer_head *bh;
ac27a0ec 4170 journal_t *journal;
617ba13b
MC		 4171 ext4_fsblk_t start;
4172 ext4_fsblk_t len;
ac27a0ec 4173 int hblock, blocksize;
617ba13b 4174 ext4_fsblk_t sb_block;
ac27a0ec 4175 unsigned long offset;
2b2d6d01 4176 struct ext4_super_block *es;
ac27a0ec
DK		 4177 struct block_device *bdev;
4178
0390131b
FM		 4179 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4180
b31e1552 4181 bdev = ext4_blkdev_get(j_dev, sb);
ac27a0ec
DK		 4182 if (bdev == NULL)
4183 return NULL;
4184
ac27a0ec 4185 blocksize = sb->s_blocksize;
e1defc4f 4186 hblock = bdev_logical_block_size(bdev);
ac27a0ec 4187 if (blocksize < hblock) {
b31e1552
ES		 4188 ext4_msg(sb, KERN_ERR,
4189 "blocksize too small for journal device");
ac27a0ec
DK		 4190 goto out_bdev;
4191 }
4192
617ba13b
MC		 4193 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4194 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 4195 set_blocksize(bdev, blocksize);
4196 if (!(bh = __bread(bdev, sb_block, blocksize))) {
b31e1552
ES		 4197 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4198 "external journal");
ac27a0ec
DK		 4199 goto out_bdev;
4200 }
4201
2716b802 4202 es = (struct ext4_super_block *) (bh->b_data + offset);
617ba13b 4203 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 4204 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b 4205 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
b31e1552
ES		 4206 ext4_msg(sb, KERN_ERR, "external journal has "
4207 "bad superblock");
ac27a0ec
DK		 4208 brelse(bh);
4209 goto out_bdev;
4210 }
4211
617ba13b 4212 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
b31e1552 4213 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
ac27a0ec
DK		 4214 brelse(bh);
4215 goto out_bdev;
4216 }
4217
bd81d8ee 4218 len = ext4_blocks_count(es);
ac27a0ec
DK		 4219 start = sb_block + 1;
4220 brelse(bh); /* we're done with the superblock */
4221
dab291af 4222 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 4223 start, len, blocksize);
4224 if (!journal) {
b31e1552 4225 ext4_msg(sb, KERN_ERR, "failed to create device journal");
ac27a0ec
DK		 4226 goto out_bdev;
4227 }
4228 journal->j_private = sb;
4229 ll_rw_block(READ, 1, &journal->j_sb_buffer);
4230 wait_on_buffer(journal->j_sb_buffer);
4231 if (!buffer_uptodate(journal->j_sb_buffer)) {
b31e1552 4232 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
ac27a0ec
DK		 4233 goto out_journal;
4234 }
4235 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
b31e1552
ES		 4236 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4237 "user (unsupported) - %d",
ac27a0ec
DK		 4238 be32_to_cpu(journal->j_superblock->s_nr_users));
4239 goto out_journal;
4240 }
617ba13b
MC		 4241 EXT4_SB(sb)->journal_bdev = bdev;
4242 ext4_init_journal_params(sb, journal);
ac27a0ec 4243 return journal;
0b8e58a1 4244
ac27a0ec 4245out_journal:
dab291af 4246 jbd2_journal_destroy(journal);
ac27a0ec 4247out_bdev:
617ba13b 4248 ext4_blkdev_put(bdev);
ac27a0ec
DK		 4249 return NULL;
4250}
4251
617ba13b
MC		 4252static int ext4_load_journal(struct super_block *sb,
4253 struct ext4_super_block *es,
ac27a0ec
DK		 4254 unsigned long journal_devnum)
4255{
4256 journal_t *journal;
4257 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4258 dev_t journal_dev;
4259 int err = 0;
4260 int really_read_only;
4261
0390131b
FM		 4262 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4263
ac27a0ec
DK		 4264 if (journal_devnum &&
4265 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
b31e1552
ES		 4266 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4267 "numbers have changed");
ac27a0ec
DK		 4268 journal_dev = new_decode_dev(journal_devnum);
4269 } else
4270 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4271
4272 really_read_only = bdev_read_only(sb->s_bdev);
4273
4274 /*
4275 * Are we loading a blank journal or performing recovery after a
4276 * crash? For recovery, we need to check in advance whether we
4277 * can get read-write access to the device.
4278 */
617ba13b 4279 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
ac27a0ec 4280 if (sb->s_flags & MS_RDONLY) {
b31e1552
ES		 4281 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4282 "required on readonly filesystem");
ac27a0ec 4283 if (really_read_only) {
b31e1552
ES		 4284 ext4_msg(sb, KERN_ERR, "write access "
4285 "unavailable, cannot proceed");
ac27a0ec
DK		 4286 return -EROFS;
4287 }
b31e1552
ES		 4288 ext4_msg(sb, KERN_INFO, "write access will "
4289 "be enabled during recovery");
ac27a0ec
DK		 4290 }
4291 }
4292
4293 if (journal_inum && journal_dev) {
b31e1552
ES		 4294 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4295 "and inode journals!");
ac27a0ec
DK		 4296 return -EINVAL;
4297 }
4298
4299 if (journal_inum) {
617ba13b 4300 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 4301 return -EINVAL;
4302 } else {
617ba13b 4303 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 4304 return -EINVAL;
4305 }
4306
90576c0b 4307 if (!(journal->j_flags & JBD2_BARRIER))
b31e1552 4308 ext4_msg(sb, KERN_INFO, "barriers disabled");
4776004f 4309
617ba13b 4310 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
dab291af 4311 err = jbd2_journal_wipe(journal, !really_read_only);
1c13d5c0
TT		 4312 if (!err) {
4313 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4314 if (save)
4315 memcpy(save, ((char *) es) +
4316 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
dab291af 4317 err = jbd2_journal_load(journal);
1c13d5c0
TT		 4318 if (save)
4319 memcpy(((char *) es) + EXT4_S_ERR_START,
4320 save, EXT4_S_ERR_LEN);
4321 kfree(save);
4322 }
ac27a0ec
DK		 4323
4324 if (err) {
b31e1552 4325 ext4_msg(sb, KERN_ERR, "error loading journal");
dab291af 4326 jbd2_journal_destroy(journal);
ac27a0ec
DK		 4327 return err;
4328 }
4329
617ba13b
MC		 4330 EXT4_SB(sb)->s_journal = journal;
4331 ext4_clear_journal_err(sb, es);
ac27a0ec 4332
c41303ce 4333 if (!really_read_only && journal_devnum &&
ac27a0ec
DK		 4334 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4335 es->s_journal_dev = cpu_to_le32(journal_devnum);
ac27a0ec
DK		 4336
4337 /* Make sure we flush the recovery flag to disk. */
e2d67052 4338 ext4_commit_super(sb, 1);
ac27a0ec
DK		 4339 }
4340
4341 return 0;
4342}
4343
e2d67052 4344static int ext4_commit_super(struct super_block *sb, int sync)
ac27a0ec 4345{
e2d67052 4346 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
617ba13b 4347 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
c4be0c1d 4348 int error = 0;
ac27a0ec 4349
7c2e7087 4350 if (!sbh || block_device_ejected(sb))
c4be0c1d 4351 return error;
914258bf
TT		 4352 if (buffer_write_io_error(sbh)) {
4353 /*
4354 * Oh, dear. A previous attempt to write the
4355 * superblock failed. This could happen because the
4356 * USB device was yanked out. Or it could happen to
4357 * be a transient write error and maybe the block will
4358 * be remapped. Nothing we can do but to retry the
4359 * write and hope for the best.
4360 */
b31e1552
ES		 4361 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4362 "superblock detected");
914258bf
TT		 4363 clear_buffer_write_io_error(sbh);
4364 set_buffer_uptodate(sbh);
4365 }
71290b36
TT		 4366 /*
4367 * If the file system is mounted read-only, don't update the
4368 * superblock write time. This avoids updating the superblock
4369 * write time when we are mounting the root file system
4370 * read/only but we need to replay the journal; at that point,
4371 * for people who are east of GMT and who make their clock
4372 * tick in localtime for Windows bug-for-bug compatibility,
4373 * the clock is set in the future, and this will cause e2fsck
4374 * to complain and force a full file system check.
4375 */
4376 if (!(sb->s_flags & MS_RDONLY))
4377 es->s_wtime = cpu_to_le32(get_seconds());
f613dfcb
TT		 4378 if (sb->s_bdev->bd_part)
4379 es->s_kbytes_written =
4380 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
afc32f7e
TT		 4381 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4382 EXT4_SB(sb)->s_sectors_written_start) >> 1));
f613dfcb
TT		 4383 else
4384 es->s_kbytes_written =
4385 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
57042651
TT		 4386 ext4_free_blocks_count_set(es,
4387 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4388 &EXT4_SB(sb)->s_freeclusters_counter)));
ce7e010a
TT		 4389 es->s_free_inodes_count =
4390 cpu_to_le32(percpu_counter_sum_positive(
4391 &EXT4_SB(sb)->s_freeinodes_counter));
ac27a0ec 4392 BUFFER_TRACE(sbh, "marking dirty");
06db49e6 4393 ext4_superblock_csum_set(sb);
ac27a0ec 4394 mark_buffer_dirty(sbh);
914258bf 4395 if (sync) {
c4be0c1d
TS		 4396 error = sync_dirty_buffer(sbh);
4397 if (error)
4398 return error;
4399
4400 error = buffer_write_io_error(sbh);
4401 if (error) {
b31e1552
ES		 4402 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4403 "superblock");
914258bf
TT		 4404 clear_buffer_write_io_error(sbh);
4405 set_buffer_uptodate(sbh);
4406 }
4407 }
c4be0c1d 4408 return error;
ac27a0ec
DK		 4409}
4410
ac27a0ec
DK		 4411/*
4412 * Have we just finished recovery? If so, and if we are mounting (or
4413 * remounting) the filesystem readonly, then we will end up with a
4414 * consistent fs on disk. Record that fact.
4415 */
2b2d6d01
TT		 4416static void ext4_mark_recovery_complete(struct super_block *sb,
4417 struct ext4_super_block *es)
ac27a0ec 4418{
617ba13b 4419 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 4420
0390131b
FM		 4421 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4422 BUG_ON(journal != NULL);
4423 return;
4424 }
dab291af 4425 jbd2_journal_lock_updates(journal);
7ffe1ea8
HK		 4426 if (jbd2_journal_flush(journal) < 0)
4427 goto out;
4428
617ba13b 4429 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
ac27a0ec 4430 sb->s_flags & MS_RDONLY) {
617ba13b 4431 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
e2d67052 4432 ext4_commit_super(sb, 1);
ac27a0ec 4433 }
7ffe1ea8
HK		 4434
4435out:
dab291af 4436 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 4437}
4438
4439/*
4440 * If we are mounting (or read-write remounting) a filesystem whose journal
4441 * has recorded an error from a previous lifetime, move that error to the
4442 * main filesystem now.
4443 */
2b2d6d01
TT		 4444static void ext4_clear_journal_err(struct super_block *sb,
4445 struct ext4_super_block *es)
ac27a0ec
DK		 4446{
4447 journal_t *journal;
4448 int j_errno;
4449 const char *errstr;
4450
0390131b
FM		 4451 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4452
617ba13b 4453 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 4454
4455 /*
4456 * Now check for any error status which may have been recorded in the
617ba13b 4457 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 4458 */
4459
dab291af 4460 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 4461 if (j_errno) {
4462 char nbuf[16];
4463
617ba13b 4464 errstr = ext4_decode_error(sb, j_errno, nbuf);
12062ddd 4465 ext4_warning(sb, "Filesystem error recorded "
ac27a0ec 4466 "from previous mount: %s", errstr);
12062ddd 4467 ext4_warning(sb, "Marking fs in need of filesystem check.");
ac27a0ec 4468
617ba13b
MC		 4469 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4470 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
e2d67052 4471 ext4_commit_super(sb, 1);
ac27a0ec 4472
dab291af 4473 jbd2_journal_clear_err(journal);
d796c52e 4474 jbd2_journal_update_sb_errno(journal);
ac27a0ec
DK		 4475 }
4476}
4477
4478/*
4479 * Force the running and committing transactions to commit,
4480 * and wait on the commit.
4481 */
617ba13b 4482int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 4483{
4484 journal_t *journal;
0390131b 4485 int ret = 0;
ac27a0ec
DK		 4486
4487 if (sb->s_flags & MS_RDONLY)
4488 return 0;
4489
617ba13b 4490 journal = EXT4_SB(sb)->s_journal;
8e8ad8a5 4491 if (journal)
0390131b 4492 ret = ext4_journal_force_commit(journal);
0390131b 4493
ac27a0ec
DK		 4494 return ret;
4495}
4496
617ba13b 4497static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec 4498{
14ce0cb4 4499 int ret = 0;
9eddacf9 4500 tid_t target;
8d5d02e6 4501 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4502
9bffad1e 4503 trace_ext4_sync_fs(sb, wait);
8d5d02e6 4504 flush_workqueue(sbi->dio_unwritten_wq);
a1177825
JK		 4505 /*
4506 * Writeback quota in non-journalled quota case - journalled quota has
4507 * no dirty dquots
4508 */
4509 dquot_writeback_dquots(sb, -1);
8d5d02e6 4510 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
9ca92389 4511 if (wait)
8d5d02e6 4512 jbd2_log_wait_commit(sbi->s_journal, target);
0390131b 4513 }
14ce0cb4 4514 return ret;
ac27a0ec
DK		 4515}
4516
4517/*
4518 * LVM calls this function before a (read-only) snapshot is created. This
4519 * gives us a chance to flush the journal completely and mark the fs clean.
be4f27d3
YY		 4520 *
4521 * Note that only this function cannot bring a filesystem to be in a clean
8e8ad8a5
JK		 4522 * state independently. It relies on upper layer to stop all data & metadata
4523 * modifications.
ac27a0ec 4524 */
c4be0c1d 4525static int ext4_freeze(struct super_block *sb)
ac27a0ec 4526{
c4be0c1d
TS		 4527 int error = 0;
4528 journal_t *journal;
ac27a0ec 4529
9ca92389
TT		 4530 if (sb->s_flags & MS_RDONLY)
4531 return 0;
ac27a0ec 4532
9ca92389 4533 journal = EXT4_SB(sb)->s_journal;
7ffe1ea8 4534
9ca92389
TT		 4535 /* Now we set up the journal barrier. */
4536 jbd2_journal_lock_updates(journal);
ac27a0ec 4537
9ca92389
TT		 4538 /*
4539 * Don't clear the needs_recovery flag if we failed to flush
4540 * the journal.
4541 */
4542 error = jbd2_journal_flush(journal);
6b0310fb
ES		 4543 if (error < 0)
4544 goto out;
9ca92389
TT		 4545
4546 /* Journal blocked and flushed, clear needs_recovery flag. */
4547 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4548 error = ext4_commit_super(sb, 1);
6b0310fb 4549out:
8e8ad8a5 4550 /* we rely on upper layer to stop further updates */
6b0310fb
ES		 4551 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4552 return error;
ac27a0ec
DK		 4553}
4554
4555/*
4556 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4557 * flag here, even though the filesystem is not technically dirty yet.
4558 */
c4be0c1d 4559static int ext4_unfreeze(struct super_block *sb)
ac27a0ec 4560{
9ca92389
TT		 4561 if (sb->s_flags & MS_RDONLY)
4562 return 0;
4563
9ca92389
TT		 4564 /* Reset the needs_recovery flag before the fs is unlocked. */
4565 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4566 ext4_commit_super(sb, 1);
c4be0c1d 4567 return 0;
ac27a0ec
DK		 4568}
4569
673c6100
TT		 4570/*
4571 * Structure to save mount options for ext4_remount's benefit
4572 */
4573struct ext4_mount_options {
4574 unsigned long s_mount_opt;
a2595b8a 4575 unsigned long s_mount_opt2;
08cefc7a
EB		 4576 kuid_t s_resuid;
4577 kgid_t s_resgid;
673c6100
TT		 4578 unsigned long s_commit_interval;
4579 u32 s_min_batch_time, s_max_batch_time;
4580#ifdef CONFIG_QUOTA
4581 int s_jquota_fmt;
4582 char *s_qf_names[MAXQUOTAS];
4583#endif
4584};
4585
2b2d6d01 4586static int ext4_remount(struct super_block *sb, int *flags, char *data)
ac27a0ec 4587{
2b2d6d01 4588 struct ext4_super_block *es;
617ba13b 4589 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4590 unsigned long old_sb_flags;
617ba13b 4591 struct ext4_mount_options old_opts;
c79d967d 4592 int enable_quota = 0;
8a266467 4593 ext4_group_t g;
b3881f74 4594 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
c5e06d10 4595 int err = 0;
ac27a0ec 4596#ifdef CONFIG_QUOTA
03dafb5f 4597 int i, j;
ac27a0ec 4598#endif
d4c402d9 4599 char *orig_data = kstrdup(data, GFP_KERNEL);
ac27a0ec
DK		 4600
4601 /* Store the original options */
4602 old_sb_flags = sb->s_flags;
4603 old_opts.s_mount_opt = sbi->s_mount_opt;
a2595b8a 4604 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
ac27a0ec
DK		 4605 old_opts.s_resuid = sbi->s_resuid;
4606 old_opts.s_resgid = sbi->s_resgid;
4607 old_opts.s_commit_interval = sbi->s_commit_interval;
30773840
TT		 4608 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4609 old_opts.s_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 4610#ifdef CONFIG_QUOTA
4611 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4612 for (i = 0; i < MAXQUOTAS; i++)
03dafb5f
CG		 4613 if (sbi->s_qf_names[i]) {
4614 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4615 GFP_KERNEL);
4616 if (!old_opts.s_qf_names[i]) {
4617 for (j = 0; j < i; j++)
4618 kfree(old_opts.s_qf_names[j]);
4619 return -ENOMEM;
4620 }
4621 } else
4622 old_opts.s_qf_names[i] = NULL;
ac27a0ec 4623#endif
b3881f74
TT		 4624 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4625 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
ac27a0ec
DK		 4626
4627 /*
4628 * Allow the "check" option to be passed as a remount option.
4629 */
661aa520 4630 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
ac27a0ec
DK		 4631 err = -EINVAL;
4632 goto restore_opts;
4633 }
4634
4ab2f15b 4635 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
c67d859e 4636 ext4_abort(sb, "Abort forced by user");
ac27a0ec
DK		 4637
4638 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 4639 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec
DK		 4640
4641 es = sbi->s_es;
4642
b3881f74 4643 if (sbi->s_journal) {
0390131b 4644 ext4_init_journal_params(sb, sbi->s_journal);
b3881f74
TT		 4645 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4646 }
ac27a0ec 4647
661aa520 4648 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4ab2f15b 4649 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
ac27a0ec
DK		 4650 err = -EROFS;
4651 goto restore_opts;
4652 }
4653
4654 if (*flags & MS_RDONLY) {
0f0dd62f
CH		 4655 err = dquot_suspend(sb, -1);
4656 if (err < 0)
c79d967d 4657 goto restore_opts;
c79d967d 4658
ac27a0ec
DK		 4659 /*
4660 * First of all, the unconditional stuff we have to do
4661 * to disable replay of the journal when we next remount
4662 */
4663 sb->s_flags |= MS_RDONLY;
4664
4665 /*
4666 * OK, test if we are remounting a valid rw partition
4667 * readonly, and if so set the rdonly flag and then
4668 * mark the partition as valid again.
4669 */
617ba13b
MC		 4670 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4671 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 4672 es->s_state = cpu_to_le16(sbi->s_mount_state);
4673
a63c9eb2 4674 if (sbi->s_journal)
0390131b 4675 ext4_mark_recovery_complete(sb, es);
ac27a0ec 4676 } else {
a13fb1a4
ES		 4677 /* Make sure we can mount this feature set readwrite */
4678 if (!ext4_feature_set_ok(sb, 0)) {
ac27a0ec
DK		 4679 err = -EROFS;
4680 goto restore_opts;
4681 }
8a266467
TT		 4682 /*
4683 * Make sure the group descriptor checksums
0b8e58a1 4684 * are sane. If they aren't, refuse to remount r/w.
8a266467
TT		 4685 */
4686 for (g = 0; g < sbi->s_groups_count; g++) {
4687 struct ext4_group_desc *gdp =
4688 ext4_get_group_desc(sb, g, NULL);
4689
feb0ab32 4690 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
b31e1552
ES		 4691 ext4_msg(sb, KERN_ERR,
4692 "ext4_remount: Checksum for group %u failed (%u!=%u)",
8a266467
TT		 4693 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4694 le16_to_cpu(gdp->bg_checksum));
4695 err = -EINVAL;
4696 goto restore_opts;
4697 }
4698 }
4699
ead6596b
ES		 4700 /*
4701 * If we have an unprocessed orphan list hanging
4702 * around from a previously readonly bdev mount,
4703 * require a full umount/remount for now.
4704 */
4705 if (es->s_last_orphan) {
b31e1552 4706 ext4_msg(sb, KERN_WARNING, "Couldn't "
ead6596b
ES		 4707 "remount RDWR because of unprocessed "
4708 "orphan inode list. Please "
b31e1552 4709 "umount/remount instead");
ead6596b
ES		 4710 err = -EINVAL;
4711 goto restore_opts;
4712 }
4713
ac27a0ec
DK		 4714 /*
4715 * Mounting a RDONLY partition read-write, so reread
4716 * and store the current valid flag. (It may have
4717 * been changed by e2fsck since we originally mounted
4718 * the partition.)
4719 */
0390131b
FM		 4720 if (sbi->s_journal)
4721 ext4_clear_journal_err(sb, es);
ac27a0ec 4722 sbi->s_mount_state = le16_to_cpu(es->s_state);
2b2d6d01 4723 if (!ext4_setup_super(sb, es, 0))
ac27a0ec 4724 sb->s_flags &= ~MS_RDONLY;
c5e06d10
JL		 4725 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
4726 EXT4_FEATURE_INCOMPAT_MMP))
4727 if (ext4_multi_mount_protect(sb,
4728 le64_to_cpu(es->s_mmp_block))) {
4729 err = -EROFS;
4730 goto restore_opts;
4731 }
c79d967d 4732 enable_quota = 1;
ac27a0ec
DK		 4733 }
4734 }
bfff6873
LC		 4735
4736 /*
4737 * Reinitialize lazy itable initialization thread based on
4738 * current settings
4739 */
4740 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4741 ext4_unregister_li_request(sb);
4742 else {
4743 ext4_group_t first_not_zeroed;
4744 first_not_zeroed = ext4_has_uninit_itable(sb);
4745 ext4_register_li_request(sb, first_not_zeroed);
4746 }
4747
6fd058f7 4748 ext4_setup_system_zone(sb);
d096ad0f 4749 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
e2d67052 4750 ext4_commit_super(sb, 1);
0390131b 4751
ac27a0ec
DK		 4752#ifdef CONFIG_QUOTA
4753 /* Release old quota file names */
4754 for (i = 0; i < MAXQUOTAS; i++)
03dafb5f 4755 kfree(old_opts.s_qf_names[i]);
7c319d32
AK		 4756 if (enable_quota) {
4757 if (sb_any_quota_suspended(sb))
4758 dquot_resume(sb, -1);
4759 else if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4760 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
4761 err = ext4_enable_quotas(sb);
07724f98 4762 if (err)
7c319d32 4763 goto restore_opts;
7c319d32
AK		 4764 }
4765 }
ac27a0ec 4766#endif
d4c402d9
CW		 4767
4768 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4769 kfree(orig_data);
ac27a0ec 4770 return 0;
0b8e58a1 4771
ac27a0ec
DK		 4772restore_opts:
4773 sb->s_flags = old_sb_flags;
4774 sbi->s_mount_opt = old_opts.s_mount_opt;
a2595b8a 4775 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
ac27a0ec
DK		 4776 sbi->s_resuid = old_opts.s_resuid;
4777 sbi->s_resgid = old_opts.s_resgid;
4778 sbi->s_commit_interval = old_opts.s_commit_interval;
30773840
TT		 4779 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4780 sbi->s_max_batch_time = old_opts.s_max_batch_time;
ac27a0ec
DK		 4781#ifdef CONFIG_QUOTA
4782 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4783 for (i = 0; i < MAXQUOTAS; i++) {
03dafb5f 4784 kfree(sbi->s_qf_names[i]);
ac27a0ec
DK		 4785 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4786 }
4787#endif
d4c402d9 4788 kfree(orig_data);
ac27a0ec
DK		 4789 return err;
4790}
4791
2b2d6d01 4792static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
ac27a0ec
DK		 4793{
4794 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 4795 struct ext4_sb_info *sbi = EXT4_SB(sb);
4796 struct ext4_super_block *es = sbi->s_es;
952fc18e 4797 ext4_fsblk_t overhead = 0;
960cc398 4798 u64 fsid;
d02a9391 4799 s64 bfree;
ac27a0ec 4800
952fc18e
TT		 4801 if (!test_opt(sb, MINIX_DF))
4802 overhead = sbi->s_overhead;
ac27a0ec 4803
617ba13b 4804 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 4805 buf->f_bsize = sb->s_blocksize;
b72f78cb 4806 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
57042651
TT		 4807 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
4808 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
d02a9391 4809 /* prevent underflow in case that few free space is available */
57042651 4810 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
bd81d8ee
LV		 4811 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
4812 if (buf->f_bfree < ext4_r_blocks_count(es))
ac27a0ec
DK		 4813 buf->f_bavail = 0;
4814 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 4815 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
617ba13b 4816 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 4817 fsid = le64_to_cpup((void *)es->s_uuid) ^
4818 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4819 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4820 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
0b8e58a1 4821
ac27a0ec
DK		 4822 return 0;
4823}
4824
0b8e58a1
AD		 4825/* Helper function for writing quotas on sync - we need to start transaction
4826 * before quota file is locked for write. Otherwise the are possible deadlocks:
ac27a0ec 4827 * Process 1 Process 2
617ba13b 4828 * ext4_create() quota_sync()
a269eb18 4829 * jbd2_journal_start() write_dquot()
871a2931 4830 * dquot_initialize() down(dqio_mutex)
dab291af 4831 * down(dqio_mutex) jbd2_journal_start()
ac27a0ec
DK		 4832 *
4833 */
4834
4835#ifdef CONFIG_QUOTA
4836
4837static inline struct inode *dquot_to_inode(struct dquot *dquot)
4838{
4c376dca 4839 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
ac27a0ec
DK		 4840}
4841
617ba13b 4842static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 4843{
4844 int ret, err;
4845 handle_t *handle;
4846 struct inode *inode;
4847
4848 inode = dquot_to_inode(dquot);
617ba13b 4849 handle = ext4_journal_start(inode,
0b8e58a1 4850 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 4851 if (IS_ERR(handle))
4852 return PTR_ERR(handle);
4853 ret = dquot_commit(dquot);
617ba13b 4854 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4855 if (!ret)
4856 ret = err;
4857 return ret;
4858}
4859
617ba13b 4860static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 4861{
4862 int ret, err;
4863 handle_t *handle;
4864
617ba13b 4865 handle = ext4_journal_start(dquot_to_inode(dquot),
0b8e58a1 4866 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 4867 if (IS_ERR(handle))
4868 return PTR_ERR(handle);
4869 ret = dquot_acquire(dquot);
617ba13b 4870 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4871 if (!ret)
4872 ret = err;
4873 return ret;
4874}
4875
617ba13b 4876static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 4877{
4878 int ret, err;
4879 handle_t *handle;
4880
617ba13b 4881 handle = ext4_journal_start(dquot_to_inode(dquot),
0b8e58a1 4882 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 4883 if (IS_ERR(handle)) {
4884 /* Release dquot anyway to avoid endless cycle in dqput() */
4885 dquot_release(dquot);
ac27a0ec 4886 return PTR_ERR(handle);
9c3013e9 4887 }
ac27a0ec 4888 ret = dquot_release(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_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec 4896{
2c8be6b2 4897 /* Are we journaling quotas? */
617ba13b
MC		 4898 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
4899 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
ac27a0ec 4900 dquot_mark_dquot_dirty(dquot);
617ba13b 4901 return ext4_write_dquot(dquot);
ac27a0ec
DK		 4902 } else {
4903 return dquot_mark_dquot_dirty(dquot);
4904 }
4905}
4906
617ba13b 4907static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 4908{
4909 int ret, err;
4910 handle_t *handle;
4911
4912 /* Data block + inode block */
617ba13b 4913 handle = ext4_journal_start(sb->s_root->d_inode, 2);
ac27a0ec
DK		 4914 if (IS_ERR(handle))
4915 return PTR_ERR(handle);
4916 ret = dquot_commit_info(sb, type);
617ba13b 4917 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4918 if (!ret)
4919 ret = err;
4920 return ret;
4921}
4922
4923/*
4924 * Turn on quotas during mount time - we need to find
4925 * the quota file and such...
4926 */
617ba13b 4927static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 4928{
287a8095
CH		 4929 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4930 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 4931}
4932
4933/*
4934 * Standard function to be called on quota_on
4935 */
617ba13b 4936static int ext4_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 4937 struct path *path)
ac27a0ec
DK		 4938{
4939 int err;
ac27a0ec
DK		 4940
4941 if (!test_opt(sb, QUOTA))
4942 return -EINVAL;
0623543b 4943
ac27a0ec 4944 /* Quotafile not on the same filesystem? */
d8c9584e 4945 if (path->dentry->d_sb != sb)
ac27a0ec 4946 return -EXDEV;
0623543b
JK		 4947 /* Journaling quota? */
4948 if (EXT4_SB(sb)->s_qf_names[type]) {
2b2d6d01 4949 /* Quotafile not in fs root? */
f00c9e44 4950 if (path->dentry->d_parent != sb->s_root)
b31e1552
ES		 4951 ext4_msg(sb, KERN_WARNING,
4952 "Quota file not on filesystem root. "
4953 "Journaled quota will not work");
2b2d6d01 4954 }
0623543b
JK		 4955
4956 /*
4957 * When we journal data on quota file, we have to flush journal to see
4958 * all updates to the file when we bypass pagecache...
4959 */
0390131b 4960 if (EXT4_SB(sb)->s_journal &&
f00c9e44 4961 ext4_should_journal_data(path->dentry->d_inode)) {
0623543b
JK		 4962 /*
4963 * We don't need to lock updates but journal_flush() could
4964 * otherwise be livelocked...
4965 */
4966 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 4967 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
0623543b 4968 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
f00c9e44 4969 if (err)
7ffe1ea8 4970 return err;
0623543b
JK		 4971 }
4972
f00c9e44 4973 return dquot_quota_on(sb, type, format_id, path);
ac27a0ec
DK		 4974}
4975
7c319d32
AK		 4976static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
4977 unsigned int flags)
4978{
4979 int err;
4980 struct inode *qf_inode;
4981 unsigned long qf_inums[MAXQUOTAS] = {
4982 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
4983 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
4984 };
4985
4986 BUG_ON(!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA));
4987
4988 if (!qf_inums[type])
4989 return -EPERM;
4990
4991 qf_inode = ext4_iget(sb, qf_inums[type]);
4992 if (IS_ERR(qf_inode)) {
4993 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
4994 return PTR_ERR(qf_inode);
4995 }
4996
4997 err = dquot_enable(qf_inode, type, format_id, flags);
4998 iput(qf_inode);
4999
5000 return err;
5001}
5002
5003/* Enable usage tracking for all quota types. */
5004static int ext4_enable_quotas(struct super_block *sb)
5005{
5006 int type, err = 0;
5007 unsigned long qf_inums[MAXQUOTAS] = {
5008 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5009 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5010 };
5011
5012 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
5013 for (type = 0; type < MAXQUOTAS; type++) {
5014 if (qf_inums[type]) {
5015 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5016 DQUOT_USAGE_ENABLED);
5017 if (err) {
5018 ext4_warning(sb,
72ba7450
TT		 5019 "Failed to enable quota tracking "
5020 "(type=%d, err=%d). Please run "
5021 "e2fsck to fix.", type, err);
7c319d32
AK		 5022 return err;
5023 }
5024 }
5025 }
5026 return 0;
5027}
5028
5029/*
5030 * quota_on function that is used when QUOTA feature is set.
5031 */
5032static int ext4_quota_on_sysfile(struct super_block *sb, int type,
5033 int format_id)
5034{
5035 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5036 return -EINVAL;
5037
5038 /*
5039 * USAGE was enabled at mount time. Only need to enable LIMITS now.
5040 */
5041 return ext4_quota_enable(sb, type, format_id, DQUOT_LIMITS_ENABLED);
5042}
5043
ca0e05e4
DM		 5044static int ext4_quota_off(struct super_block *sb, int type)
5045{
21f97697
JK		 5046 struct inode *inode = sb_dqopt(sb)->files[type];
5047 handle_t *handle;
5048
87009d86
DM		 5049 /* Force all delayed allocation blocks to be allocated.
5050 * Caller already holds s_umount sem */
5051 if (test_opt(sb, DELALLOC))
ca0e05e4 5052 sync_filesystem(sb);
ca0e05e4 5053
0b268590
AG		 5054 if (!inode)
5055 goto out;
5056
21f97697
JK		 5057 /* Update modification times of quota files when userspace can
5058 * start looking at them */
5059 handle = ext4_journal_start(inode, 1);
5060 if (IS_ERR(handle))
5061 goto out;
5062 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
5063 ext4_mark_inode_dirty(handle, inode);
5064 ext4_journal_stop(handle);
5065
5066out:
ca0e05e4
DM		 5067 return dquot_quota_off(sb, type);
5068}
5069
7c319d32
AK		 5070/*
5071 * quota_off function that is used when QUOTA feature is set.
5072 */
5073static int ext4_quota_off_sysfile(struct super_block *sb, int type)
5074{
5075 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5076 return -EINVAL;
5077
5078 /* Disable only the limits. */
5079 return dquot_disable(sb, type, DQUOT_LIMITS_ENABLED);
5080}
5081
ac27a0ec
DK		 5082/* Read data from quotafile - avoid pagecache and such because we cannot afford
5083 * acquiring the locks... As quota files are never truncated and quota code
25985edc 5084 * itself serializes the operations (and no one else should touch the files)
ac27a0ec 5085 * we don't have to be afraid of races */
617ba13b 5086static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 5087 size_t len, loff_t off)
5088{
5089 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 5090 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 5091 int err = 0;
5092 int offset = off & (sb->s_blocksize - 1);
5093 int tocopy;
5094 size_t toread;
5095 struct buffer_head *bh;
5096 loff_t i_size = i_size_read(inode);
5097
5098 if (off > i_size)
5099 return 0;
5100 if (off+len > i_size)
5101 len = i_size-off;
5102 toread = len;
5103 while (toread > 0) {
5104 tocopy = sb->s_blocksize - offset < toread ?
5105 sb->s_blocksize - offset : toread;
617ba13b 5106 bh = ext4_bread(NULL, inode, blk, 0, &err);
ac27a0ec
DK		 5107 if (err)
5108 return err;
5109 if (!bh) /* A hole? */
5110 memset(data, 0, tocopy);
5111 else
5112 memcpy(data, bh->b_data+offset, tocopy);
5113 brelse(bh);
5114 offset = 0;
5115 toread -= tocopy;
5116 data += tocopy;
5117 blk++;
5118 }
5119 return len;
5120}
5121
5122/* Write to quotafile (we know the transaction is already started and has
5123 * enough credits) */
617ba13b 5124static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 5125 const char *data, size_t len, loff_t off)
5126{
5127 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 5128 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 5129 int err = 0;
5130 int offset = off & (sb->s_blocksize - 1);
ac27a0ec
DK		 5131 struct buffer_head *bh;
5132 handle_t *handle = journal_current_handle();
5133
0390131b 5134 if (EXT4_SB(sb)->s_journal && !handle) {
b31e1552
ES		 5135 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5136 " cancelled because transaction is not started",
9c3013e9
JK		 5137 (unsigned long long)off, (unsigned long long)len);
5138 return -EIO;
5139 }
67eeb568
DM		 5140 /*
5141 * Since we account only one data block in transaction credits,
5142 * then it is impossible to cross a block boundary.
5143 */
5144 if (sb->s_blocksize - offset < len) {
5145 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5146 " cancelled because not block aligned",
5147 (unsigned long long)off, (unsigned long long)len);
5148 return -EIO;
5149 }
5150
67eeb568
DM		 5151 bh = ext4_bread(handle, inode, blk, 1, &err);
5152 if (!bh)
5153 goto out;
62d2b5f2
JK		 5154 err = ext4_journal_get_write_access(handle, bh);
5155 if (err) {
5156 brelse(bh);
5157 goto out;
ac27a0ec 5158 }
67eeb568
DM		 5159 lock_buffer(bh);
5160 memcpy(bh->b_data+offset, data, len);
5161 flush_dcache_page(bh->b_page);
5162 unlock_buffer(bh);
62d2b5f2 5163 err = ext4_handle_dirty_metadata(handle, NULL, bh);
67eeb568 5164 brelse(bh);
ac27a0ec 5165out:
0b7f7cef 5166 if (err)
ac27a0ec 5167 return err;
67eeb568
DM		 5168 if (inode->i_size < off + len) {
5169 i_size_write(inode, off + len);
617ba13b 5170 EXT4_I(inode)->i_disksize = inode->i_size;
21f97697 5171 ext4_mark_inode_dirty(handle, inode);
ac27a0ec 5172 }
67eeb568 5173 return len;
ac27a0ec
DK		 5174}
5175
5176#endif
5177
152a0836
AV		 5178static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5179 const char *dev_name, void *data)
ac27a0ec 5180{
152a0836 5181 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
ac27a0ec
DK		 5182}
5183
37f328eb 5184#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
24b58424
TT		 5185static inline void register_as_ext2(void)
5186{
5187 int err = register_filesystem(&ext2_fs_type);
5188 if (err)
5189 printk(KERN_WARNING
5190 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5191}
5192
5193static inline void unregister_as_ext2(void)
5194{
5195 unregister_filesystem(&ext2_fs_type);
5196}
2035e776
TT		 5197
5198static inline int ext2_feature_set_ok(struct super_block *sb)
5199{
5200 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
5201 return 0;
5202 if (sb->s_flags & MS_RDONLY)
5203 return 1;
5204 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
5205 return 0;
5206 return 1;
5207}
51b7e3c9 5208MODULE_ALIAS("ext2");
24b58424
TT		 5209#else
5210static inline void register_as_ext2(void) { }
5211static inline void unregister_as_ext2(void) { }
2035e776 5212static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 5213#endif
5214
37f328eb 5215#if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
24b58424
TT		 5216static inline void register_as_ext3(void)
5217{
5218 int err = register_filesystem(&ext3_fs_type);
5219 if (err)
5220 printk(KERN_WARNING
5221 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5222}
5223
5224static inline void unregister_as_ext3(void)
5225{
5226 unregister_filesystem(&ext3_fs_type);
5227}
2035e776
TT		 5228
5229static inline int ext3_feature_set_ok(struct super_block *sb)
5230{
5231 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
5232 return 0;
5233 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
5234 return 0;
5235 if (sb->s_flags & MS_RDONLY)
5236 return 1;
5237 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
5238 return 0;
5239 return 1;
5240}
51b7e3c9 5241MODULE_ALIAS("ext3");
24b58424
TT		 5242#else
5243static inline void register_as_ext3(void) { }
5244static inline void unregister_as_ext3(void) { }
2035e776 5245static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 5246#endif
5247
03010a33
TT		 5248static struct file_system_type ext4_fs_type = {
5249 .owner = THIS_MODULE,
5250 .name = "ext4",
152a0836 5251 .mount = ext4_mount,
03010a33
TT		 5252 .kill_sb = kill_block_super,
5253 .fs_flags = FS_REQUIRES_DEV,
5254};
5255
8f021222 5256static int __init ext4_init_feat_adverts(void)
857ac889
LC		 5257{
5258 struct ext4_features *ef;
5259 int ret = -ENOMEM;
5260
5261 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
5262 if (!ef)
5263 goto out;
5264
5265 ef->f_kobj.kset = ext4_kset;
5266 init_completion(&ef->f_kobj_unregister);
5267 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
5268 "features");
5269 if (ret) {
5270 kfree(ef);
5271 goto out;
5272 }
5273
5274 ext4_feat = ef;
5275 ret = 0;
5276out:
5277 return ret;
5278}
5279
8f021222
LC		 5280static void ext4_exit_feat_adverts(void)
5281{
5282 kobject_put(&ext4_feat->f_kobj);
5283 wait_for_completion(&ext4_feat->f_kobj_unregister);
5284 kfree(ext4_feat);
5285}
5286
e9e3bcec
ES		 5287/* Shared across all ext4 file systems */
5288wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5289struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5290
5dabfc78 5291static int __init ext4_init_fs(void)
ac27a0ec 5292{
e9e3bcec 5293 int i, err;
c9de560d 5294
07c0c5d8
AV		 5295 ext4_li_info = NULL;
5296 mutex_init(&ext4_li_mtx);
5297
9a4c8019 5298 /* Build-time check for flags consistency */
12e9b892 5299 ext4_check_flag_values();
e9e3bcec
ES		 5300
5301 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5302 mutex_init(&ext4__aio_mutex[i]);
5303 init_waitqueue_head(&ext4__ioend_wq[i]);
5304 }
5305
51865fda 5306 err = ext4_init_es();
6fd058f7
TT		 5307 if (err)
5308 return err;
51865fda
ZL		 5309
5310 err = ext4_init_pageio();
5311 if (err)
5312 goto out7;
5313
5dabfc78 5314 err = ext4_init_system_zone();
bd2d0210 5315 if (err)
d44651d0 5316 goto out6;
3197ebdb 5317 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
0e376b1e
TT		 5318 if (!ext4_kset) {
5319 err = -ENOMEM;
dd68314c 5320 goto out5;
0e376b1e 5321 }
d44651d0 5322 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
857ac889
LC		 5323
5324 err = ext4_init_feat_adverts();
dd68314c
TT		 5325 if (err)
5326 goto out4;
857ac889 5327
5dabfc78 5328 err = ext4_init_mballoc();
ac27a0ec 5329 if (err)
6fd058f7 5330 goto out3;
c9de560d 5331
5dabfc78 5332 err = ext4_init_xattr();
c9de560d
AT		 5333 if (err)
5334 goto out2;
ac27a0ec
DK		 5335 err = init_inodecache();
5336 if (err)
5337 goto out1;
24b58424 5338 register_as_ext3();
2035e776 5339 register_as_ext2();
03010a33 5340 err = register_filesystem(&ext4_fs_type);
ac27a0ec
DK		 5341 if (err)
5342 goto out;
bfff6873 5343
ac27a0ec
DK		 5344 return 0;
5345out:
24b58424
TT		 5346 unregister_as_ext2();
5347 unregister_as_ext3();
ac27a0ec
DK		 5348 destroy_inodecache();
5349out1:
5dabfc78 5350 ext4_exit_xattr();
c9de560d 5351out2:
5dabfc78 5352 ext4_exit_mballoc();
6fd058f7 5353out3:
8f021222 5354 ext4_exit_feat_adverts();
dd68314c 5355out4:
d44651d0
FJ		 5356 if (ext4_proc_root)
5357 remove_proc_entry("fs/ext4", NULL);
6fd058f7 5358 kset_unregister(ext4_kset);
d44651d0 5359out5:
5dabfc78 5360 ext4_exit_system_zone();
d44651d0 5361out6:
5dabfc78 5362 ext4_exit_pageio();
51865fda
ZL		 5363out7:
5364 ext4_exit_es();
5365
ac27a0ec
DK		 5366 return err;
5367}
5368
5dabfc78 5369static void __exit ext4_exit_fs(void)
ac27a0ec 5370{
bfff6873 5371 ext4_destroy_lazyinit_thread();
24b58424
TT		 5372 unregister_as_ext2();
5373 unregister_as_ext3();
03010a33 5374 unregister_filesystem(&ext4_fs_type);
ac27a0ec 5375 destroy_inodecache();
5dabfc78
TT		 5376 ext4_exit_xattr();
5377 ext4_exit_mballoc();
8f021222 5378 ext4_exit_feat_adverts();
9f6200bb 5379 remove_proc_entry("fs/ext4", NULL);
3197ebdb 5380 kset_unregister(ext4_kset);
5dabfc78
TT		 5381 ext4_exit_system_zone();
5382 ext4_exit_pageio();
ac27a0ec
DK		 5383}
5384
5385MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
83982b6f 5386MODULE_DESCRIPTION("Fourth Extended Filesystem");
ac27a0ec 5387MODULE_LICENSE("GPL");
5dabfc78
TT		 5388module_init(ext4_init_fs)
5389module_exit(ext4_exit_fs)
kernel source for telekom puls (alcatel/mediatek)