projects / GitHub / mt8127 / android_kernel_alcatel_ttab.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
ext4: initialize extent status tree
[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"
6f91bc5f 48#include "ext4_extents.h"
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;
91246c00 942 ei->vfs_inode.i_data.writeback_index = 0;
a86c6181 943 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
c9de560d
AT		 944 INIT_LIST_HEAD(&ei->i_prealloc_list);
945 spin_lock_init(&ei->i_prealloc_lock);
9a26b661
ZL		 946 ext4_es_init_tree(&ei->i_es_tree);
947 rwlock_init(&ei->i_es_lock);
d2a17637
MC		 948 ei->i_reserved_data_blocks = 0;
949 ei->i_reserved_meta_blocks = 0;
950 ei->i_allocated_meta_blocks = 0;
9d0be502 951 ei->i_da_metadata_calc_len = 0;
7e731bc9 952 ei->i_da_metadata_calc_last_lblock = 0;
d2a17637 953 spin_lock_init(&(ei->i_block_reservation_lock));
a9e7f447
DM		 954#ifdef CONFIG_QUOTA
955 ei->i_reserved_quota = 0;
956#endif
8aefcd55 957 ei->jinode = NULL;
c7064ef1 958 INIT_LIST_HEAD(&ei->i_completed_io_list);
744692dc 959 spin_lock_init(&ei->i_completed_io_lock);
b436b9be
JK		 960 ei->i_sync_tid = 0;
961 ei->i_datasync_tid = 0;
f7ad6d2e 962 atomic_set(&ei->i_ioend_count, 0);
e27f41e1 963 atomic_set(&ei->i_unwritten, 0);
0b8e58a1 964
ac27a0ec
DK		 965 return &ei->vfs_inode;
966}
967
7ff9c073
TT		 968static int ext4_drop_inode(struct inode *inode)
969{
970 int drop = generic_drop_inode(inode);
971
972 trace_ext4_drop_inode(inode, drop);
973 return drop;
974}
975
fa0d7e3d
NP		 976static void ext4_i_callback(struct rcu_head *head)
977{
978 struct inode *inode = container_of(head, struct inode, i_rcu);
fa0d7e3d
NP		 979 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
980}
981
617ba13b 982static void ext4_destroy_inode(struct inode *inode)
ac27a0ec 983{
9f7dd93d 984 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
b31e1552
ES		 985 ext4_msg(inode->i_sb, KERN_ERR,
986 "Inode %lu (%p): orphan list check failed!",
987 inode->i_ino, EXT4_I(inode));
9f7dd93d
VA		 988 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
989 EXT4_I(inode), sizeof(struct ext4_inode_info),
990 true);
991 dump_stack();
992 }
fa0d7e3d 993 call_rcu(&inode->i_rcu, ext4_i_callback);
ac27a0ec
DK		 994}
995
51cc5068 996static void init_once(void *foo)
ac27a0ec 997{
617ba13b 998 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
ac27a0ec 999
a35afb83 1000 INIT_LIST_HEAD(&ei->i_orphan);
03010a33 1001#ifdef CONFIG_EXT4_FS_XATTR
a35afb83 1002 init_rwsem(&ei->xattr_sem);
ac27a0ec 1003#endif
0e855ac8 1004 init_rwsem(&ei->i_data_sem);
a35afb83 1005 inode_init_once(&ei->vfs_inode);
ac27a0ec
DK		 1006}
1007
1008static int init_inodecache(void)
1009{
617ba13b
MC		 1010 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
1011 sizeof(struct ext4_inode_info),
ac27a0ec
DK		 1012 0, (SLAB_RECLAIM_ACCOUNT|
1013 SLAB_MEM_SPREAD),
20c2df83 1014 init_once);
617ba13b 1015 if (ext4_inode_cachep == NULL)
ac27a0ec
DK		 1016 return -ENOMEM;
1017 return 0;
1018}
1019
1020static void destroy_inodecache(void)
1021{
8c0a8537
KS		 1022 /*
1023 * Make sure all delayed rcu free inodes are flushed before we
1024 * destroy cache.
1025 */
1026 rcu_barrier();
617ba13b 1027 kmem_cache_destroy(ext4_inode_cachep);
ac27a0ec
DK		 1028}
1029
0930fcc1 1030void ext4_clear_inode(struct inode *inode)
ac27a0ec 1031{
0930fcc1 1032 invalidate_inode_buffers(inode);
dbd5768f 1033 clear_inode(inode);
9f754758 1034 dquot_drop(inode);
c2ea3fde 1035 ext4_discard_preallocations(inode);
8aefcd55
TT		 1036 if (EXT4_I(inode)->jinode) {
1037 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1038 EXT4_I(inode)->jinode);
1039 jbd2_free_inode(EXT4_I(inode)->jinode);
1040 EXT4_I(inode)->jinode = NULL;
1041 }
ac27a0ec
DK		 1042}
1043
1b961ac0 1044static struct inode *ext4_nfs_get_inode(struct super_block *sb,
0b8e58a1 1045 u64 ino, u32 generation)
ac27a0ec 1046{
ac27a0ec 1047 struct inode *inode;
ac27a0ec 1048
617ba13b 1049 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
ac27a0ec 1050 return ERR_PTR(-ESTALE);
617ba13b 1051 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
ac27a0ec
DK		 1052 return ERR_PTR(-ESTALE);
1053
1054 /* iget isn't really right if the inode is currently unallocated!!
1055 *
617ba13b 1056 * ext4_read_inode will return a bad_inode if the inode had been
ac27a0ec
DK		 1057 * deleted, so we should be safe.
1058 *
1059 * Currently we don't know the generation for parent directory, so
1060 * a generation of 0 means "accept any"
1061 */
1d1fe1ee
DH		 1062 inode = ext4_iget(sb, ino);
1063 if (IS_ERR(inode))
1064 return ERR_CAST(inode);
1065 if (generation && inode->i_generation != generation) {
ac27a0ec
DK		 1066 iput(inode);
1067 return ERR_PTR(-ESTALE);
1068 }
1b961ac0
CH		 1069
1070 return inode;
1071}
1072
1073static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
0b8e58a1 1074 int fh_len, int fh_type)
1b961ac0
CH		 1075{
1076 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1077 ext4_nfs_get_inode);
1078}
1079
1080static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
0b8e58a1 1081 int fh_len, int fh_type)
1b961ac0
CH		 1082{
1083 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1084 ext4_nfs_get_inode);
ac27a0ec
DK		 1085}
1086
c39a7f84
TO		 1087/*
1088 * Try to release metadata pages (indirect blocks, directories) which are
1089 * mapped via the block device. Since these pages could have journal heads
1090 * which would prevent try_to_free_buffers() from freeing them, we must use
1091 * jbd2 layer's try_to_free_buffers() function to release them.
1092 */
0b8e58a1
AD		 1093static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1094 gfp_t wait)
c39a7f84
TO		 1095{
1096 journal_t *journal = EXT4_SB(sb)->s_journal;
1097
1098 WARN_ON(PageChecked(page));
1099 if (!page_has_buffers(page))
1100 return 0;
1101 if (journal)
1102 return jbd2_journal_try_to_free_buffers(journal, page,
1103 wait & ~__GFP_WAIT);
1104 return try_to_free_buffers(page);
1105}
1106
ac27a0ec 1107#ifdef CONFIG_QUOTA
af5bc92d 1108#define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
2b2d6d01 1109#define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
ac27a0ec 1110
617ba13b
MC		 1111static int ext4_write_dquot(struct dquot *dquot);
1112static int ext4_acquire_dquot(struct dquot *dquot);
1113static int ext4_release_dquot(struct dquot *dquot);
1114static int ext4_mark_dquot_dirty(struct dquot *dquot);
1115static int ext4_write_info(struct super_block *sb, int type);
6f28e087 1116static int ext4_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 1117 struct path *path);
7c319d32
AK		 1118static int ext4_quota_on_sysfile(struct super_block *sb, int type,
1119 int format_id);
ca0e05e4 1120static int ext4_quota_off(struct super_block *sb, int type);
7c319d32 1121static int ext4_quota_off_sysfile(struct super_block *sb, int type);
617ba13b
MC		 1122static int ext4_quota_on_mount(struct super_block *sb, int type);
1123static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 1124 size_t len, loff_t off);
617ba13b 1125static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec 1126 const char *data, size_t len, loff_t off);
7c319d32
AK		 1127static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1128 unsigned int flags);
1129static int ext4_enable_quotas(struct super_block *sb);
ac27a0ec 1130
61e225dc 1131static const struct dquot_operations ext4_quota_operations = {
60e58e0f 1132 .get_reserved_space = ext4_get_reserved_space,
617ba13b
MC		 1133 .write_dquot = ext4_write_dquot,
1134 .acquire_dquot = ext4_acquire_dquot,
1135 .release_dquot = ext4_release_dquot,
1136 .mark_dirty = ext4_mark_dquot_dirty,
a5b5ee32
JK		 1137 .write_info = ext4_write_info,
1138 .alloc_dquot = dquot_alloc,
1139 .destroy_dquot = dquot_destroy,
ac27a0ec
DK		 1140};
1141
0d54b217 1142static const struct quotactl_ops ext4_qctl_operations = {
617ba13b 1143 .quota_on = ext4_quota_on,
ca0e05e4 1144 .quota_off = ext4_quota_off,
287a8095
CH		 1145 .quota_sync = dquot_quota_sync,
1146 .get_info = dquot_get_dqinfo,
1147 .set_info = dquot_set_dqinfo,
1148 .get_dqblk = dquot_get_dqblk,
1149 .set_dqblk = dquot_set_dqblk
ac27a0ec 1150};
7c319d32
AK		 1151
1152static const struct quotactl_ops ext4_qctl_sysfile_operations = {
1153 .quota_on_meta = ext4_quota_on_sysfile,
1154 .quota_off = ext4_quota_off_sysfile,
1155 .quota_sync = dquot_quota_sync,
1156 .get_info = dquot_get_dqinfo,
1157 .set_info = dquot_set_dqinfo,
1158 .get_dqblk = dquot_get_dqblk,
1159 .set_dqblk = dquot_set_dqblk
1160};
ac27a0ec
DK		 1161#endif
1162
ee9b6d61 1163static const struct super_operations ext4_sops = {
617ba13b
MC		 1164 .alloc_inode = ext4_alloc_inode,
1165 .destroy_inode = ext4_destroy_inode,
617ba13b
MC		 1166 .write_inode = ext4_write_inode,
1167 .dirty_inode = ext4_dirty_inode,
7ff9c073 1168 .drop_inode = ext4_drop_inode,
0930fcc1 1169 .evict_inode = ext4_evict_inode,
617ba13b 1170 .put_super = ext4_put_super,
617ba13b 1171 .sync_fs = ext4_sync_fs,
c4be0c1d
TS		 1172 .freeze_fs = ext4_freeze,
1173 .unfreeze_fs = ext4_unfreeze,
617ba13b
MC		 1174 .statfs = ext4_statfs,
1175 .remount_fs = ext4_remount,
617ba13b 1176 .show_options = ext4_show_options,
ac27a0ec 1177#ifdef CONFIG_QUOTA
617ba13b
MC		 1178 .quota_read = ext4_quota_read,
1179 .quota_write = ext4_quota_write,
ac27a0ec 1180#endif
c39a7f84 1181 .bdev_try_to_free_page = bdev_try_to_free_page,
ac27a0ec
DK		 1182};
1183
9ca92389
TT		 1184static const struct super_operations ext4_nojournal_sops = {
1185 .alloc_inode = ext4_alloc_inode,
1186 .destroy_inode = ext4_destroy_inode,
1187 .write_inode = ext4_write_inode,
1188 .dirty_inode = ext4_dirty_inode,
7ff9c073 1189 .drop_inode = ext4_drop_inode,
0930fcc1 1190 .evict_inode = ext4_evict_inode,
9ca92389
TT		 1191 .put_super = ext4_put_super,
1192 .statfs = ext4_statfs,
1193 .remount_fs = ext4_remount,
9ca92389
TT		 1194 .show_options = ext4_show_options,
1195#ifdef CONFIG_QUOTA
1196 .quota_read = ext4_quota_read,
1197 .quota_write = ext4_quota_write,
1198#endif
1199 .bdev_try_to_free_page = bdev_try_to_free_page,
1200};
1201
39655164 1202static const struct export_operations ext4_export_ops = {
1b961ac0
CH		 1203 .fh_to_dentry = ext4_fh_to_dentry,
1204 .fh_to_parent = ext4_fh_to_parent,
617ba13b 1205 .get_parent = ext4_get_parent,
ac27a0ec
DK		 1206};
1207
1208enum {
1209 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1210 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
72578c33 1211 Opt_nouid32, Opt_debug, Opt_removed,
ac27a0ec 1212 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
72578c33 1213 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
30773840 1214 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
43e625d8 1215 Opt_journal_dev, Opt_journal_checksum, Opt_journal_async_commit,
ac27a0ec 1216 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
296c355c 1217 Opt_data_err_abort, Opt_data_err_ignore,
ac27a0ec 1218 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
5a20bdfc 1219 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
ee4a3fcd 1220 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
661aa520 1221 Opt_usrquota, Opt_grpquota, Opt_i_version,
1449032b
TT		 1222 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1223 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
5328e635 1224 Opt_inode_readahead_blks, Opt_journal_ioprio,
744692dc 1225 Opt_dioread_nolock, Opt_dioread_lock,
fc6cb1cd 1226 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
df981d03 1227 Opt_max_dir_size_kb,
ac27a0ec
DK		 1228};
1229
a447c093 1230static const match_table_t tokens = {
ac27a0ec
DK		 1231 {Opt_bsd_df, "bsddf"},
1232 {Opt_minix_df, "minixdf"},
1233 {Opt_grpid, "grpid"},
1234 {Opt_grpid, "bsdgroups"},
1235 {Opt_nogrpid, "nogrpid"},
1236 {Opt_nogrpid, "sysvgroups"},
1237 {Opt_resgid, "resgid=%u"},
1238 {Opt_resuid, "resuid=%u"},
1239 {Opt_sb, "sb=%u"},
1240 {Opt_err_cont, "errors=continue"},
1241 {Opt_err_panic, "errors=panic"},
1242 {Opt_err_ro, "errors=remount-ro"},
1243 {Opt_nouid32, "nouid32"},
ac27a0ec 1244 {Opt_debug, "debug"},
72578c33
TT		 1245 {Opt_removed, "oldalloc"},
1246 {Opt_removed, "orlov"},
ac27a0ec
DK		 1247 {Opt_user_xattr, "user_xattr"},
1248 {Opt_nouser_xattr, "nouser_xattr"},
1249 {Opt_acl, "acl"},
1250 {Opt_noacl, "noacl"},
e3bb52ae 1251 {Opt_noload, "norecovery"},
5a916be1 1252 {Opt_noload, "noload"},
72578c33
TT		 1253 {Opt_removed, "nobh"},
1254 {Opt_removed, "bh"},
ac27a0ec 1255 {Opt_commit, "commit=%u"},
30773840
TT		 1256 {Opt_min_batch_time, "min_batch_time=%u"},
1257 {Opt_max_batch_time, "max_batch_time=%u"},
ac27a0ec 1258 {Opt_journal_dev, "journal_dev=%u"},
818d276c
GS		 1259 {Opt_journal_checksum, "journal_checksum"},
1260 {Opt_journal_async_commit, "journal_async_commit"},
ac27a0ec
DK		 1261 {Opt_abort, "abort"},
1262 {Opt_data_journal, "data=journal"},
1263 {Opt_data_ordered, "data=ordered"},
1264 {Opt_data_writeback, "data=writeback"},
5bf5683a
HK		 1265 {Opt_data_err_abort, "data_err=abort"},
1266 {Opt_data_err_ignore, "data_err=ignore"},
ac27a0ec
DK		 1267 {Opt_offusrjquota, "usrjquota="},
1268 {Opt_usrjquota, "usrjquota=%s"},
1269 {Opt_offgrpjquota, "grpjquota="},
1270 {Opt_grpjquota, "grpjquota=%s"},
1271 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1272 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
5a20bdfc 1273 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
ac27a0ec
DK		 1274 {Opt_grpquota, "grpquota"},
1275 {Opt_noquota, "noquota"},
1276 {Opt_quota, "quota"},
1277 {Opt_usrquota, "usrquota"},
1278 {Opt_barrier, "barrier=%u"},
06705bff
TT		 1279 {Opt_barrier, "barrier"},
1280 {Opt_nobarrier, "nobarrier"},
25ec56b5 1281 {Opt_i_version, "i_version"},
c9de560d 1282 {Opt_stripe, "stripe=%u"},
64769240 1283 {Opt_delalloc, "delalloc"},
dd919b98 1284 {Opt_nodelalloc, "nodelalloc"},
1449032b
TT		 1285 {Opt_mblk_io_submit, "mblk_io_submit"},
1286 {Opt_nomblk_io_submit, "nomblk_io_submit"},
6fd058f7
TT		 1287 {Opt_block_validity, "block_validity"},
1288 {Opt_noblock_validity, "noblock_validity"},
240799cd 1289 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
b3881f74 1290 {Opt_journal_ioprio, "journal_ioprio=%u"},
afd4672d 1291 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
06705bff
TT		 1292 {Opt_auto_da_alloc, "auto_da_alloc"},
1293 {Opt_noauto_da_alloc, "noauto_da_alloc"},
744692dc
JZ		 1294 {Opt_dioread_nolock, "dioread_nolock"},
1295 {Opt_dioread_lock, "dioread_lock"},
5328e635
ES		 1296 {Opt_discard, "discard"},
1297 {Opt_nodiscard, "nodiscard"},
fc6cb1cd
TT		 1298 {Opt_init_itable, "init_itable=%u"},
1299 {Opt_init_itable, "init_itable"},
1300 {Opt_noinit_itable, "noinit_itable"},
df981d03 1301 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
c7198b9c
TT		 1302 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1303 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1304 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1305 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1306 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
f3f12faa 1307 {Opt_err, NULL},
ac27a0ec
DK		 1308};
1309
617ba13b 1310static ext4_fsblk_t get_sb_block(void **data)
ac27a0ec 1311{
617ba13b 1312 ext4_fsblk_t sb_block;
ac27a0ec
DK		 1313 char *options = (char *) *data;
1314
1315 if (!options || strncmp(options, "sb=", 3) != 0)
1316 return 1; /* Default location */
0b8e58a1 1317
ac27a0ec 1318 options += 3;
0b8e58a1 1319 /* TODO: use simple_strtoll with >32bit ext4 */
ac27a0ec
DK		 1320 sb_block = simple_strtoul(options, &options, 0);
1321 if (*options && *options != ',') {
4776004f 1322 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
ac27a0ec
DK		 1323 (char *) *data);
1324 return 1;
1325 }
1326 if (*options == ',')
1327 options++;
1328 *data = (void *) options;
0b8e58a1 1329
ac27a0ec
DK		 1330 return sb_block;
1331}
1332
b3881f74 1333#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
437ca0fd
DM		 1334static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1335 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
b3881f74 1336
56c50f11
DM		 1337#ifdef CONFIG_QUOTA
1338static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1339{
1340 struct ext4_sb_info *sbi = EXT4_SB(sb);
1341 char *qname;
1342
1343 if (sb_any_quota_loaded(sb) &&
1344 !sbi->s_qf_names[qtype]) {
1345 ext4_msg(sb, KERN_ERR,
1346 "Cannot change journaled "
1347 "quota options when quota turned on");
57f73c2c 1348 return -1;
56c50f11
DM		 1349 }
1350 qname = match_strdup(args);
1351 if (!qname) {
1352 ext4_msg(sb, KERN_ERR,
1353 "Not enough memory for storing quotafile name");
57f73c2c 1354 return -1;
56c50f11
DM		 1355 }
1356 if (sbi->s_qf_names[qtype] &&
1357 strcmp(sbi->s_qf_names[qtype], qname)) {
1358 ext4_msg(sb, KERN_ERR,
1359 "%s quota file already specified", QTYPE2NAME(qtype));
1360 kfree(qname);
57f73c2c 1361 return -1;
56c50f11
DM		 1362 }
1363 sbi->s_qf_names[qtype] = qname;
1364 if (strchr(sbi->s_qf_names[qtype], '/')) {
1365 ext4_msg(sb, KERN_ERR,
1366 "quotafile must be on filesystem root");
1367 kfree(sbi->s_qf_names[qtype]);
1368 sbi->s_qf_names[qtype] = NULL;
57f73c2c 1369 return -1;
56c50f11 1370 }
fd8c37ec 1371 set_opt(sb, QUOTA);
56c50f11
DM		 1372 return 1;
1373}
1374
1375static int clear_qf_name(struct super_block *sb, int qtype)
1376{
1377
1378 struct ext4_sb_info *sbi = EXT4_SB(sb);
1379
1380 if (sb_any_quota_loaded(sb) &&
1381 sbi->s_qf_names[qtype]) {
1382 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1383 " when quota turned on");
57f73c2c 1384 return -1;
56c50f11
DM		 1385 }
1386 /*
1387 * The space will be released later when all options are confirmed
1388 * to be correct
1389 */
1390 sbi->s_qf_names[qtype] = NULL;
1391 return 1;
1392}
1393#endif
1394
26092bf5
TT		 1395#define MOPT_SET 0x0001
1396#define MOPT_CLEAR 0x0002
1397#define MOPT_NOSUPPORT 0x0004
1398#define MOPT_EXPLICIT 0x0008
1399#define MOPT_CLEAR_ERR 0x0010
1400#define MOPT_GTE0 0x0020
ac27a0ec 1401#ifdef CONFIG_QUOTA
26092bf5
TT		 1402#define MOPT_Q 0
1403#define MOPT_QFMT 0x0040
1404#else
1405#define MOPT_Q MOPT_NOSUPPORT
1406#define MOPT_QFMT MOPT_NOSUPPORT
ac27a0ec 1407#endif
26092bf5
TT		 1408#define MOPT_DATAJ 0x0080
1409
1410static const struct mount_opts {
1411 int token;
1412 int mount_opt;
1413 int flags;
1414} ext4_mount_opts[] = {
1415 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1416 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1417 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1418 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1419 {Opt_mblk_io_submit, EXT4_MOUNT_MBLK_IO_SUBMIT, MOPT_SET},
1420 {Opt_nomblk_io_submit, EXT4_MOUNT_MBLK_IO_SUBMIT, MOPT_CLEAR},
1421 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1422 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1423 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK, MOPT_SET},
1424 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK, MOPT_CLEAR},
1425 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1426 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1427 {Opt_delalloc, EXT4_MOUNT_DELALLOC, MOPT_SET | MOPT_EXPLICIT},
1428 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC, MOPT_CLEAR | MOPT_EXPLICIT},
1429 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM, MOPT_SET},
1430 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1431 EXT4_MOUNT_JOURNAL_CHECKSUM), MOPT_SET},
1432 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_SET},
1433 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1434 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1435 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1436 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT, MOPT_SET},
1437 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT, MOPT_CLEAR},
1438 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1439 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1440 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1441 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1442 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1443 {Opt_commit, 0, MOPT_GTE0},
1444 {Opt_max_batch_time, 0, MOPT_GTE0},
1445 {Opt_min_batch_time, 0, MOPT_GTE0},
1446 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1447 {Opt_init_itable, 0, MOPT_GTE0},
1448 {Opt_stripe, 0, MOPT_GTE0},
1449 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_DATAJ},
1450 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_DATAJ},
1451 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA, MOPT_DATAJ},
03010a33 1452#ifdef CONFIG_EXT4_FS_XATTR
26092bf5
TT		 1453 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1454 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
ac27a0ec 1455#else
26092bf5
TT		 1456 {Opt_user_xattr, 0, MOPT_NOSUPPORT},
1457 {Opt_nouser_xattr, 0, MOPT_NOSUPPORT},
ac27a0ec 1458#endif
03010a33 1459#ifdef CONFIG_EXT4_FS_POSIX_ACL
26092bf5
TT		 1460 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1461 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
ac27a0ec 1462#else
26092bf5
TT		 1463 {Opt_acl, 0, MOPT_NOSUPPORT},
1464 {Opt_noacl, 0, MOPT_NOSUPPORT},
ac27a0ec 1465#endif
26092bf5
TT		 1466 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1467 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1468 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1469 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1470 MOPT_SET | MOPT_Q},
1471 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1472 MOPT_SET | MOPT_Q},
1473 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1474 EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1475 {Opt_usrjquota, 0, MOPT_Q},
1476 {Opt_grpjquota, 0, MOPT_Q},
1477 {Opt_offusrjquota, 0, MOPT_Q},
1478 {Opt_offgrpjquota, 0, MOPT_Q},
1479 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1480 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1481 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
df981d03 1482 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
26092bf5
TT		 1483 {Opt_err, 0, 0}
1484};
1485
1486static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1487 substring_t *args, unsigned long *journal_devnum,
1488 unsigned int *journal_ioprio, int is_remount)
1489{
1490 struct ext4_sb_info *sbi = EXT4_SB(sb);
1491 const struct mount_opts *m;
08cefc7a
EB		 1492 kuid_t uid;
1493 kgid_t gid;
26092bf5
TT		 1494 int arg = 0;
1495
57f73c2c
TT		 1496#ifdef CONFIG_QUOTA
1497 if (token == Opt_usrjquota)
1498 return set_qf_name(sb, USRQUOTA, &args[0]);
1499 else if (token == Opt_grpjquota)
1500 return set_qf_name(sb, GRPQUOTA, &args[0]);
1501 else if (token == Opt_offusrjquota)
1502 return clear_qf_name(sb, USRQUOTA);
1503 else if (token == Opt_offgrpjquota)
1504 return clear_qf_name(sb, GRPQUOTA);
1505#endif
26092bf5
TT		 1506 if (args->from && match_int(args, &arg))
1507 return -1;
1508 switch (token) {
f7048605
TT		 1509 case Opt_noacl:
1510 case Opt_nouser_xattr:
1511 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1512 break;
26092bf5
TT		 1513 case Opt_sb:
1514 return 1; /* handled by get_sb_block() */
1515 case Opt_removed:
1516 ext4_msg(sb, KERN_WARNING,
1517 "Ignoring removed %s option", opt);
1518 return 1;
1519 case Opt_resuid:
08cefc7a
EB		 1520 uid = make_kuid(current_user_ns(), arg);
1521 if (!uid_valid(uid)) {
1522 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1523 return -1;
1524 }
1525 sbi->s_resuid = uid;
26092bf5
TT		 1526 return 1;
1527 case Opt_resgid:
08cefc7a
EB		 1528 gid = make_kgid(current_user_ns(), arg);
1529 if (!gid_valid(gid)) {
1530 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1531 return -1;
1532 }
1533 sbi->s_resgid = gid;
26092bf5
TT		 1534 return 1;
1535 case Opt_abort:
1536 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1537 return 1;
1538 case Opt_i_version:
1539 sb->s_flags |= MS_I_VERSION;
1540 return 1;
1541 case Opt_journal_dev:
1542 if (is_remount) {
1543 ext4_msg(sb, KERN_ERR,
1544 "Cannot specify journal on remount");
1545 return -1;
1546 }
1547 *journal_devnum = arg;
1548 return 1;
1549 case Opt_journal_ioprio:
1550 if (arg < 0 || arg > 7)
1551 return -1;
1552 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1553 return 1;
1554 }
1555
1556 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1557 if (token != m->token)
1558 continue;
1559 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1560 return -1;
1561 if (m->flags & MOPT_EXPLICIT)
1562 set_opt2(sb, EXPLICIT_DELALLOC);
1563 if (m->flags & MOPT_CLEAR_ERR)
1564 clear_opt(sb, ERRORS_MASK);
1565 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1566 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1567 "options when quota turned on");
1568 return -1;
1569 }
1570
1571 if (m->flags & MOPT_NOSUPPORT) {
1572 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1573 } else if (token == Opt_commit) {
1574 if (arg == 0)
1575 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1576 sbi->s_commit_interval = HZ * arg;
1577 } else if (token == Opt_max_batch_time) {
1578 if (arg == 0)
1579 arg = EXT4_DEF_MAX_BATCH_TIME;
1580 sbi->s_max_batch_time = arg;
1581 } else if (token == Opt_min_batch_time) {
1582 sbi->s_min_batch_time = arg;
1583 } else if (token == Opt_inode_readahead_blks) {
1584 if (arg > (1 << 30))
1585 return -1;
1586 if (arg && !is_power_of_2(arg)) {
b31e1552 1587 ext4_msg(sb, KERN_ERR,
26092bf5
TT		 1588 "EXT4-fs: inode_readahead_blks"
1589 " must be a power of 2");
1590 return -1;
ac27a0ec 1591 }
26092bf5
TT		 1592 sbi->s_inode_readahead_blks = arg;
1593 } else if (token == Opt_init_itable) {
1594 set_opt(sb, INIT_INODE_TABLE);
1595 if (!args->from)
1596 arg = EXT4_DEF_LI_WAIT_MULT;
1597 sbi->s_li_wait_mult = arg;
df981d03
TT		 1598 } else if (token == Opt_max_dir_size_kb) {
1599 sbi->s_max_dir_size_kb = arg;
26092bf5
TT		 1600 } else if (token == Opt_stripe) {
1601 sbi->s_stripe = arg;
1602 } else if (m->flags & MOPT_DATAJ) {
ac27a0ec 1603 if (is_remount) {
eb513689
TT		 1604 if (!sbi->s_journal)
1605 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
26092bf5
TT		 1606 else if (test_opt(sb, DATA_FLAGS) !=
1607 m->mount_opt) {
b31e1552 1608 ext4_msg(sb, KERN_ERR,
26092bf5
TT		 1609 "Cannot change data mode on remount");
1610 return -1;
ac27a0ec
DK		 1611 }
1612 } else {
fd8c37ec 1613 clear_opt(sb, DATA_FLAGS);
26092bf5 1614 sbi->s_mount_opt |= m->mount_opt;
ac27a0ec 1615 }
ac27a0ec 1616#ifdef CONFIG_QUOTA
26092bf5 1617 } else if (m->flags & MOPT_QFMT) {
17bd13b3 1618 if (sb_any_quota_loaded(sb) &&
26092bf5
TT		 1619 sbi->s_jquota_fmt != m->mount_opt) {
1620 ext4_msg(sb, KERN_ERR, "Cannot "
1621 "change journaled quota options "
1622 "when quota turned on");
1623 return -1;
ac27a0ec 1624 }
26092bf5 1625 sbi->s_jquota_fmt = m->mount_opt;
ac27a0ec 1626#endif
26092bf5
TT		 1627 } else {
1628 if (!args->from)
1629 arg = 1;
1630 if (m->flags & MOPT_CLEAR)
1631 arg = !arg;
1632 else if (unlikely(!(m->flags & MOPT_SET))) {
1633 ext4_msg(sb, KERN_WARNING,
1634 "buggy handling of option %s", opt);
1635 WARN_ON(1);
1636 return -1;
3197ebdb 1637 }
26092bf5
TT		 1638 if (arg != 0)
1639 sbi->s_mount_opt |= m->mount_opt;
afd4672d 1640 else
26092bf5 1641 sbi->s_mount_opt &= ~m->mount_opt;
ac27a0ec 1642 }
26092bf5
TT		 1643 return 1;
1644 }
1645 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1646 "or missing value", opt);
1647 return -1;
1648}
1649
1650static int parse_options(char *options, struct super_block *sb,
1651 unsigned long *journal_devnum,
1652 unsigned int *journal_ioprio,
1653 int is_remount)
1654{
db7e5c66 1655#ifdef CONFIG_QUOTA
26092bf5 1656 struct ext4_sb_info *sbi = EXT4_SB(sb);
db7e5c66 1657#endif
26092bf5
TT		 1658 char *p;
1659 substring_t args[MAX_OPT_ARGS];
1660 int token;
1661
1662 if (!options)
1663 return 1;
1664
1665 while ((p = strsep(&options, ",")) != NULL) {
1666 if (!*p)
1667 continue;
1668 /*
1669 * Initialize args struct so we know whether arg was
1670 * found; some options take optional arguments.
1671 */
caecd0af 1672 args[0].to = args[0].from = NULL;
26092bf5
TT		 1673 token = match_token(p, tokens, args);
1674 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1675 journal_ioprio, is_remount) < 0)
1676 return 0;
ac27a0ec
DK		 1677 }
1678#ifdef CONFIG_QUOTA
1679 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
482a7425 1680 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
fd8c37ec 1681 clear_opt(sb, USRQUOTA);
ac27a0ec 1682
482a7425 1683 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
fd8c37ec 1684 clear_opt(sb, GRPQUOTA);
ac27a0ec 1685
56c50f11 1686 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
b31e1552
ES		 1687 ext4_msg(sb, KERN_ERR, "old and new quota "
1688 "format mixing");
ac27a0ec
DK		 1689 return 0;
1690 }
1691
1692 if (!sbi->s_jquota_fmt) {
b31e1552
ES		 1693 ext4_msg(sb, KERN_ERR, "journaled quota format "
1694 "not specified");
ac27a0ec
DK		 1695 return 0;
1696 }
1697 } else {
1698 if (sbi->s_jquota_fmt) {
b31e1552 1699 ext4_msg(sb, KERN_ERR, "journaled quota format "
2c8be6b2 1700 "specified with no journaling "
b31e1552 1701 "enabled");
ac27a0ec
DK		 1702 return 0;
1703 }
1704 }
1705#endif
1706 return 1;
1707}
1708
2adf6da8
TT		 1709static inline void ext4_show_quota_options(struct seq_file *seq,
1710 struct super_block *sb)
1711{
1712#if defined(CONFIG_QUOTA)
1713 struct ext4_sb_info *sbi = EXT4_SB(sb);
1714
1715 if (sbi->s_jquota_fmt) {
1716 char *fmtname = "";
1717
1718 switch (sbi->s_jquota_fmt) {
1719 case QFMT_VFS_OLD:
1720 fmtname = "vfsold";
1721 break;
1722 case QFMT_VFS_V0:
1723 fmtname = "vfsv0";
1724 break;
1725 case QFMT_VFS_V1:
1726 fmtname = "vfsv1";
1727 break;
1728 }
1729 seq_printf(seq, ",jqfmt=%s", fmtname);
1730 }
1731
1732 if (sbi->s_qf_names[USRQUOTA])
1733 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1734
1735 if (sbi->s_qf_names[GRPQUOTA])
1736 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
1737
1738 if (test_opt(sb, USRQUOTA))
1739 seq_puts(seq, ",usrquota");
1740
1741 if (test_opt(sb, GRPQUOTA))
1742 seq_puts(seq, ",grpquota");
1743#endif
1744}
1745
5a916be1
TT		 1746static const char *token2str(int token)
1747{
50df9fd5 1748 const struct match_token *t;
5a916be1
TT		 1749
1750 for (t = tokens; t->token != Opt_err; t++)
1751 if (t->token == token && !strchr(t->pattern, '='))
1752 break;
1753 return t->pattern;
1754}
1755
2adf6da8
TT		 1756/*
1757 * Show an option if
1758 * - it's set to a non-default value OR
1759 * - if the per-sb default is different from the global default
1760 */
66acdcf4
TT		 1761static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1762 int nodefs)
2adf6da8 1763{
2adf6da8
TT		 1764 struct ext4_sb_info *sbi = EXT4_SB(sb);
1765 struct ext4_super_block *es = sbi->s_es;
66acdcf4 1766 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
5a916be1 1767 const struct mount_opts *m;
66acdcf4 1768 char sep = nodefs ? '\n' : ',';
2adf6da8 1769
66acdcf4
TT		 1770#define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1771#define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2adf6da8
TT		 1772
1773 if (sbi->s_sb_block != 1)
5a916be1
TT		 1774 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1775
1776 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1777 int want_set = m->flags & MOPT_SET;
1778 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1779 (m->flags & MOPT_CLEAR_ERR))
1780 continue;
66acdcf4 1781 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
5a916be1
TT		 1782 continue; /* skip if same as the default */
1783 if ((want_set &&
1784 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1785 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1786 continue; /* select Opt_noFoo vs Opt_Foo */
1787 SEQ_OPTS_PRINT("%s", token2str(m->token));
2adf6da8 1788 }
5a916be1 1789
08cefc7a 1790 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
5a916be1 1791 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
08cefc7a
EB		 1792 SEQ_OPTS_PRINT("resuid=%u",
1793 from_kuid_munged(&init_user_ns, sbi->s_resuid));
1794 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
5a916be1 1795 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
08cefc7a
EB		 1796 SEQ_OPTS_PRINT("resgid=%u",
1797 from_kgid_munged(&init_user_ns, sbi->s_resgid));
66acdcf4 1798 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
5a916be1
TT		 1799 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1800 SEQ_OPTS_PUTS("errors=remount-ro");
2adf6da8 1801 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
5a916be1 1802 SEQ_OPTS_PUTS("errors=continue");
2adf6da8 1803 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
5a916be1 1804 SEQ_OPTS_PUTS("errors=panic");
66acdcf4 1805 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
5a916be1 1806 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
66acdcf4 1807 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
5a916be1 1808 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
66acdcf4 1809 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
5a916be1 1810 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2adf6da8 1811 if (sb->s_flags & MS_I_VERSION)
5a916be1 1812 SEQ_OPTS_PUTS("i_version");
66acdcf4 1813 if (nodefs || sbi->s_stripe)
5a916be1 1814 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
66acdcf4 1815 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
5a916be1
TT		 1816 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1817 SEQ_OPTS_PUTS("data=journal");
1818 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1819 SEQ_OPTS_PUTS("data=ordered");
1820 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1821 SEQ_OPTS_PUTS("data=writeback");
1822 }
66acdcf4
TT		 1823 if (nodefs ||
1824 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
5a916be1
TT		 1825 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1826 sbi->s_inode_readahead_blks);
2adf6da8 1827
66acdcf4
TT		 1828 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1829 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
5a916be1 1830 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
df981d03
TT		 1831 if (nodefs || sbi->s_max_dir_size_kb)
1832 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
2adf6da8
TT		 1833
1834 ext4_show_quota_options(seq, sb);
2adf6da8
TT		 1835 return 0;
1836}
1837
66acdcf4
TT		 1838static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1839{
1840 return _ext4_show_options(seq, root->d_sb, 0);
1841}
1842
1843static int options_seq_show(struct seq_file *seq, void *offset)
1844{
1845 struct super_block *sb = seq->private;
1846 int rc;
1847
1848 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1849 rc = _ext4_show_options(seq, sb, 1);
1850 seq_puts(seq, "\n");
1851 return rc;
1852}
1853
1854static int options_open_fs(struct inode *inode, struct file *file)
1855{
1856 return single_open(file, options_seq_show, PDE(inode)->data);
1857}
1858
1859static const struct file_operations ext4_seq_options_fops = {
1860 .owner = THIS_MODULE,
1861 .open = options_open_fs,
1862 .read = seq_read,
1863 .llseek = seq_lseek,
1864 .release = single_release,
1865};
1866
617ba13b 1867static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 1868 int read_only)
1869{
617ba13b 1870 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 1871 int res = 0;
1872
617ba13b 1873 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
b31e1552
ES		 1874 ext4_msg(sb, KERN_ERR, "revision level too high, "
1875 "forcing read-only mode");
ac27a0ec
DK		 1876 res = MS_RDONLY;
1877 }
1878 if (read_only)
281b5995 1879 goto done;
617ba13b 1880 if (!(sbi->s_mount_state & EXT4_VALID_FS))
b31e1552
ES		 1881 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1882 "running e2fsck is recommended");
617ba13b 1883 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
b31e1552
ES		 1884 ext4_msg(sb, KERN_WARNING,
1885 "warning: mounting fs with errors, "
1886 "running e2fsck is recommended");
ed3ce80a 1887 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
ac27a0ec
DK		 1888 le16_to_cpu(es->s_mnt_count) >=
1889 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
b31e1552
ES		 1890 ext4_msg(sb, KERN_WARNING,
1891 "warning: maximal mount count reached, "
1892 "running e2fsck is recommended");
ac27a0ec
DK		 1893 else if (le32_to_cpu(es->s_checkinterval) &&
1894 (le32_to_cpu(es->s_lastcheck) +
1895 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
b31e1552
ES		 1896 ext4_msg(sb, KERN_WARNING,
1897 "warning: checktime reached, "
1898 "running e2fsck is recommended");
0b8e58a1 1899 if (!sbi->s_journal)
0390131b 1900 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec 1901 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 1902 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 1903 le16_add_cpu(&es->s_mnt_count, 1);
ac27a0ec 1904 es->s_mtime = cpu_to_le32(get_seconds());
617ba13b 1905 ext4_update_dynamic_rev(sb);
0390131b
FM		 1906 if (sbi->s_journal)
1907 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 1908
e2d67052 1909 ext4_commit_super(sb, 1);
281b5995 1910done:
ac27a0ec 1911 if (test_opt(sb, DEBUG))
a9df9a49 1912 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
a2595b8a 1913 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
ac27a0ec
DK		 1914 sb->s_blocksize,
1915 sbi->s_groups_count,
617ba13b
MC		 1916 EXT4_BLOCKS_PER_GROUP(sb),
1917 EXT4_INODES_PER_GROUP(sb),
a2595b8a 1918 sbi->s_mount_opt, sbi->s_mount_opt2);
ac27a0ec 1919
7abc52c2 1920 cleancache_init_fs(sb);
ac27a0ec
DK		 1921 return res;
1922}
1923
117fff10
TT		 1924int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
1925{
1926 struct ext4_sb_info *sbi = EXT4_SB(sb);
1927 struct flex_groups *new_groups;
1928 int size;
1929
1930 if (!sbi->s_log_groups_per_flex)
1931 return 0;
1932
1933 size = ext4_flex_group(sbi, ngroup - 1) + 1;
1934 if (size <= sbi->s_flex_groups_allocated)
1935 return 0;
1936
1937 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
1938 new_groups = ext4_kvzalloc(size, GFP_KERNEL);
1939 if (!new_groups) {
1940 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
1941 size / (int) sizeof(struct flex_groups));
1942 return -ENOMEM;
1943 }
1944
1945 if (sbi->s_flex_groups) {
1946 memcpy(new_groups, sbi->s_flex_groups,
1947 (sbi->s_flex_groups_allocated *
1948 sizeof(struct flex_groups)));
1949 ext4_kvfree(sbi->s_flex_groups);
1950 }
1951 sbi->s_flex_groups = new_groups;
1952 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
1953 return 0;
1954}
1955
772cb7c8
JS		 1956static int ext4_fill_flex_info(struct super_block *sb)
1957{
1958 struct ext4_sb_info *sbi = EXT4_SB(sb);
1959 struct ext4_group_desc *gdp = NULL;
772cb7c8 1960 ext4_group_t flex_group;
d50f2ab6 1961 unsigned int groups_per_flex = 0;
117fff10 1962 int i, err;
772cb7c8 1963
503358ae 1964 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
d50f2ab6 1965 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
772cb7c8
JS		 1966 sbi->s_log_groups_per_flex = 0;
1967 return 1;
1968 }
76495ec1 1969 groups_per_flex = 1U << sbi->s_log_groups_per_flex;
772cb7c8 1970
117fff10
TT		 1971 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
1972 if (err)
9933fc0a 1973 goto failed;
772cb7c8 1974
772cb7c8 1975 for (i = 0; i < sbi->s_groups_count; i++) {
88b6edd1 1976 gdp = ext4_get_group_desc(sb, i, NULL);
772cb7c8
JS		 1977
1978 flex_group = ext4_flex_group(sbi, i);
7ad9bb65
TT		 1979 atomic_add(ext4_free_inodes_count(sb, gdp),
1980 &sbi->s_flex_groups[flex_group].free_inodes);
021b65bb 1981 atomic_add(ext4_free_group_clusters(sb, gdp),
24aaa8ef 1982 &sbi->s_flex_groups[flex_group].free_clusters);
7ad9bb65
TT		 1983 atomic_add(ext4_used_dirs_count(sb, gdp),
1984 &sbi->s_flex_groups[flex_group].used_dirs);
772cb7c8
JS		 1985 }
1986
1987 return 1;
1988failed:
1989 return 0;
1990}
1991
feb0ab32
DW		 1992static __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1993 struct ext4_group_desc *gdp)
717d50e4 1994{
feb0ab32 1995 int offset;
717d50e4 1996 __u16 crc = 0;
feb0ab32 1997 __le32 le_group = cpu_to_le32(block_group);
717d50e4 1998
feb0ab32
DW		 1999 if ((sbi->s_es->s_feature_ro_compat &
2000 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))) {
2001 /* Use new metadata_csum algorithm */
2002 __u16 old_csum;
2003 __u32 csum32;
2004
2005 old_csum = gdp->bg_checksum;
2006 gdp->bg_checksum = 0;
2007 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2008 sizeof(le_group));
2009 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp,
2010 sbi->s_desc_size);
2011 gdp->bg_checksum = old_csum;
2012
2013 crc = csum32 & 0xFFFF;
2014 goto out;
717d50e4
AD		 2015 }
2016
feb0ab32
DW		 2017 /* old crc16 code */
2018 offset = offsetof(struct ext4_group_desc, bg_checksum);
2019
2020 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2021 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2022 crc = crc16(crc, (__u8 *)gdp, offset);
2023 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2024 /* for checksum of struct ext4_group_desc do the rest...*/
2025 if ((sbi->s_es->s_feature_incompat &
2026 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
2027 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2028 crc = crc16(crc, (__u8 *)gdp + offset,
2029 le16_to_cpu(sbi->s_es->s_desc_size) -
2030 offset);
2031
2032out:
717d50e4
AD		 2033 return cpu_to_le16(crc);
2034}
2035
feb0ab32 2036int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
717d50e4
AD		 2037 struct ext4_group_desc *gdp)
2038{
feb0ab32
DW		 2039 if (ext4_has_group_desc_csum(sb) &&
2040 (gdp->bg_checksum != ext4_group_desc_csum(EXT4_SB(sb),
2041 block_group, gdp)))
717d50e4
AD		 2042 return 0;
2043
2044 return 1;
2045}
2046
feb0ab32
DW		 2047void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2048 struct ext4_group_desc *gdp)
2049{
2050 if (!ext4_has_group_desc_csum(sb))
2051 return;
2052 gdp->bg_checksum = ext4_group_desc_csum(EXT4_SB(sb), block_group, gdp);
2053}
2054
ac27a0ec 2055/* Called at mount-time, super-block is locked */
bfff6873
LC		 2056static int ext4_check_descriptors(struct super_block *sb,
2057 ext4_group_t *first_not_zeroed)
ac27a0ec 2058{
617ba13b
MC		 2059 struct ext4_sb_info *sbi = EXT4_SB(sb);
2060 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2061 ext4_fsblk_t last_block;
bd81d8ee
LV		 2062 ext4_fsblk_t block_bitmap;
2063 ext4_fsblk_t inode_bitmap;
2064 ext4_fsblk_t inode_table;
ce421581 2065 int flexbg_flag = 0;
bfff6873 2066 ext4_group_t i, grp = sbi->s_groups_count;
ac27a0ec 2067
ce421581
JS		 2068 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2069 flexbg_flag = 1;
2070
af5bc92d 2071 ext4_debug("Checking group descriptors");
ac27a0ec 2072
197cd65a
AM		 2073 for (i = 0; i < sbi->s_groups_count; i++) {
2074 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2075
ce421581 2076 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 2077 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 2078 else
2079 last_block = first_block +
617ba13b 2080 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 2081
bfff6873
LC		 2082 if ((grp == sbi->s_groups_count) &&
2083 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2084 grp = i;
2085
8fadc143 2086 block_bitmap = ext4_block_bitmap(sb, gdp);
2b2d6d01 2087 if (block_bitmap < first_block || block_bitmap > last_block) {
b31e1552 2088 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2089 "Block bitmap for group %u not in group "
b31e1552 2090 "(block %llu)!", i, block_bitmap);
ac27a0ec
DK		 2091 return 0;
2092 }
8fadc143 2093 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2b2d6d01 2094 if (inode_bitmap < first_block || inode_bitmap > last_block) {
b31e1552 2095 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2096 "Inode bitmap for group %u not in group "
b31e1552 2097 "(block %llu)!", i, inode_bitmap);
ac27a0ec
DK		 2098 return 0;
2099 }
8fadc143 2100 inode_table = ext4_inode_table(sb, gdp);
bd81d8ee 2101 if (inode_table < first_block ||
2b2d6d01 2102 inode_table + sbi->s_itb_per_group - 1 > last_block) {
b31e1552 2103 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2104 "Inode table for group %u not in group "
b31e1552 2105 "(block %llu)!", i, inode_table);
ac27a0ec
DK		 2106 return 0;
2107 }
955ce5f5 2108 ext4_lock_group(sb, i);
feb0ab32 2109 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
b31e1552
ES		 2110 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2111 "Checksum for group %u failed (%u!=%u)",
2112 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2113 gdp)), le16_to_cpu(gdp->bg_checksum));
7ee1ec4c 2114 if (!(sb->s_flags & MS_RDONLY)) {
955ce5f5 2115 ext4_unlock_group(sb, i);
8a266467 2116 return 0;
7ee1ec4c 2117 }
717d50e4 2118 }
955ce5f5 2119 ext4_unlock_group(sb, i);
ce421581
JS		 2120 if (!flexbg_flag)
2121 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec 2122 }
bfff6873
LC		 2123 if (NULL != first_not_zeroed)
2124 *first_not_zeroed = grp;
ac27a0ec 2125
5dee5437
TT		 2126 ext4_free_blocks_count_set(sbi->s_es,
2127 EXT4_C2B(sbi, ext4_count_free_clusters(sb)));
0b8e58a1 2128 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
ac27a0ec
DK		 2129 return 1;
2130}
2131
617ba13b 2132/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 2133 * the superblock) which were deleted from all directories, but held open by
2134 * a process at the time of a crash. We walk the list and try to delete these
2135 * inodes at recovery time (only with a read-write filesystem).
2136 *
2137 * In order to keep the orphan inode chain consistent during traversal (in
2138 * case of crash during recovery), we link each inode into the superblock
2139 * orphan list_head and handle it the same way as an inode deletion during
2140 * normal operation (which journals the operations for us).
2141 *
2142 * We only do an iget() and an iput() on each inode, which is very safe if we
2143 * accidentally point at an in-use or already deleted inode. The worst that
2144 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 2145 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 2146 * e2fsck was run on this filesystem, and it must have already done the orphan
2147 * inode cleanup for us, so we can safely abort without any further action.
2148 */
2b2d6d01
TT		 2149static void ext4_orphan_cleanup(struct super_block *sb,
2150 struct ext4_super_block *es)
ac27a0ec
DK		 2151{
2152 unsigned int s_flags = sb->s_flags;
2153 int nr_orphans = 0, nr_truncates = 0;
2154#ifdef CONFIG_QUOTA
2155 int i;
2156#endif
2157 if (!es->s_last_orphan) {
2158 jbd_debug(4, "no orphan inodes to clean up\n");
2159 return;
2160 }
2161
a8f48a95 2162 if (bdev_read_only(sb->s_bdev)) {
b31e1552
ES		 2163 ext4_msg(sb, KERN_ERR, "write access "
2164 "unavailable, skipping orphan cleanup");
a8f48a95
ES		 2165 return;
2166 }
2167
d39195c3
AG		 2168 /* Check if feature set would not allow a r/w mount */
2169 if (!ext4_feature_set_ok(sb, 0)) {
2170 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2171 "unknown ROCOMPAT features");
2172 return;
2173 }
2174
617ba13b 2175 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
c25f9bc6
ES		 2176 /* don't clear list on RO mount w/ errors */
2177 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
ac27a0ec
DK		 2178 jbd_debug(1, "Errors on filesystem, "
2179 "clearing orphan list.\n");
c25f9bc6
ES		 2180 es->s_last_orphan = 0;
2181 }
ac27a0ec
DK		 2182 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2183 return;
2184 }
2185
2186 if (s_flags & MS_RDONLY) {
b31e1552 2187 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
ac27a0ec
DK		 2188 sb->s_flags &= ~MS_RDONLY;
2189 }
2190#ifdef CONFIG_QUOTA
2191 /* Needed for iput() to work correctly and not trash data */
2192 sb->s_flags |= MS_ACTIVE;
2193 /* Turn on quotas so that they are updated correctly */
2194 for (i = 0; i < MAXQUOTAS; i++) {
617ba13b
MC		 2195 if (EXT4_SB(sb)->s_qf_names[i]) {
2196 int ret = ext4_quota_on_mount(sb, i);
ac27a0ec 2197 if (ret < 0)
b31e1552
ES		 2198 ext4_msg(sb, KERN_ERR,
2199 "Cannot turn on journaled "
2200 "quota: error %d", ret);
ac27a0ec
DK		 2201 }
2202 }
2203#endif
2204
2205 while (es->s_last_orphan) {
2206 struct inode *inode;
2207
97bd42b9
JB		 2208 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2209 if (IS_ERR(inode)) {
ac27a0ec
DK		 2210 es->s_last_orphan = 0;
2211 break;
2212 }
2213
617ba13b 2214 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
871a2931 2215 dquot_initialize(inode);
ac27a0ec 2216 if (inode->i_nlink) {
b31e1552
ES		 2217 ext4_msg(sb, KERN_DEBUG,
2218 "%s: truncating inode %lu to %lld bytes",
46e665e9 2219 __func__, inode->i_ino, inode->i_size);
e5f8eab8 2220 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
ac27a0ec 2221 inode->i_ino, inode->i_size);
617ba13b 2222 ext4_truncate(inode);
ac27a0ec
DK		 2223 nr_truncates++;
2224 } else {
b31e1552
ES		 2225 ext4_msg(sb, KERN_DEBUG,
2226 "%s: deleting unreferenced inode %lu",
46e665e9 2227 __func__, inode->i_ino);
ac27a0ec
DK		 2228 jbd_debug(2, "deleting unreferenced inode %lu\n",
2229 inode->i_ino);
2230 nr_orphans++;
2231 }
2232 iput(inode); /* The delete magic happens here! */
2233 }
2234
2b2d6d01 2235#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
ac27a0ec
DK		 2236
2237 if (nr_orphans)
b31e1552
ES		 2238 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2239 PLURAL(nr_orphans));
ac27a0ec 2240 if (nr_truncates)
b31e1552
ES		 2241 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2242 PLURAL(nr_truncates));
ac27a0ec
DK		 2243#ifdef CONFIG_QUOTA
2244 /* Turn quotas off */
2245 for (i = 0; i < MAXQUOTAS; i++) {
2246 if (sb_dqopt(sb)->files[i])
287a8095 2247 dquot_quota_off(sb, i);
ac27a0ec
DK		 2248 }
2249#endif
2250 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2251}
0b8e58a1 2252
cd2291a4
ES		 2253/*
2254 * Maximal extent format file size.
2255 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2256 * extent format containers, within a sector_t, and within i_blocks
2257 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2258 * so that won't be a limiting factor.
2259 *
f17722f9
LC		 2260 * However there is other limiting factor. We do store extents in the form
2261 * of starting block and length, hence the resulting length of the extent
2262 * covering maximum file size must fit into on-disk format containers as
2263 * well. Given that length is always by 1 unit bigger than max unit (because
2264 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2265 *
cd2291a4
ES		 2266 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2267 */
f287a1a5 2268static loff_t ext4_max_size(int blkbits, int has_huge_files)
cd2291a4
ES		 2269{
2270 loff_t res;
2271 loff_t upper_limit = MAX_LFS_FILESIZE;
2272
2273 /* small i_blocks in vfs inode? */
f287a1a5 2274 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
cd2291a4 2275 /*
90c699a9 2276 * CONFIG_LBDAF is not enabled implies the inode
cd2291a4
ES		 2277 * i_block represent total blocks in 512 bytes
2278 * 32 == size of vfs inode i_blocks * 8
2279 */
2280 upper_limit = (1LL << 32) - 1;
2281
2282 /* total blocks in file system block size */
2283 upper_limit >>= (blkbits - 9);
2284 upper_limit <<= blkbits;
2285 }
2286
f17722f9
LC		 2287 /*
2288 * 32-bit extent-start container, ee_block. We lower the maxbytes
2289 * by one fs block, so ee_len can cover the extent of maximum file
2290 * size
2291 */
2292 res = (1LL << 32) - 1;
cd2291a4 2293 res <<= blkbits;
cd2291a4
ES		 2294
2295 /* Sanity check against vm- & vfs- imposed limits */
2296 if (res > upper_limit)
2297 res = upper_limit;
2298
2299 return res;
2300}
ac27a0ec 2301
ac27a0ec 2302/*
cd2291a4 2303 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 2304 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2305 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 2306 */
f287a1a5 2307static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
ac27a0ec 2308{
617ba13b 2309 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 2310 int meta_blocks;
2311 loff_t upper_limit;
0b8e58a1
AD		 2312 /* This is calculated to be the largest file size for a dense, block
2313 * mapped file such that the file's total number of 512-byte sectors,
2314 * including data and all indirect blocks, does not exceed (2^48 - 1).
2315 *
2316 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2317 * number of 512-byte sectors of the file.
0fc1b451
AK		 2318 */
2319
f287a1a5 2320 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
0fc1b451 2321 /*
90c699a9 2322 * !has_huge_files or CONFIG_LBDAF not enabled implies that
0b8e58a1
AD		 2323 * the inode i_block field represents total file blocks in
2324 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
0fc1b451
AK		 2325 */
2326 upper_limit = (1LL << 32) - 1;
2327
2328 /* total blocks in file system block size */
2329 upper_limit >>= (bits - 9);
2330
2331 } else {
8180a562
AK		 2332 /*
2333 * We use 48 bit ext4_inode i_blocks
2334 * With EXT4_HUGE_FILE_FL set the i_blocks
2335 * represent total number of blocks in
2336 * file system block size
2337 */
0fc1b451
AK		 2338 upper_limit = (1LL << 48) - 1;
2339
0fc1b451
AK		 2340 }
2341
2342 /* indirect blocks */
2343 meta_blocks = 1;
2344 /* double indirect blocks */
2345 meta_blocks += 1 + (1LL << (bits-2));
2346 /* tripple indirect blocks */
2347 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2348
2349 upper_limit -= meta_blocks;
2350 upper_limit <<= bits;
ac27a0ec
DK		 2351
2352 res += 1LL << (bits-2);
2353 res += 1LL << (2*(bits-2));
2354 res += 1LL << (3*(bits-2));
2355 res <<= bits;
2356 if (res > upper_limit)
2357 res = upper_limit;
0fc1b451
AK		 2358
2359 if (res > MAX_LFS_FILESIZE)
2360 res = MAX_LFS_FILESIZE;
2361
ac27a0ec
DK		 2362 return res;
2363}
2364
617ba13b 2365static ext4_fsblk_t descriptor_loc(struct super_block *sb,
0b8e58a1 2366 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 2367{
617ba13b 2368 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 2369 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 2370 int has_super = 0;
2371
2372 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2373
617ba13b 2374 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
ac27a0ec 2375 nr < first_meta_bg)
70bbb3e0 2376 return logical_sb_block + nr + 1;
ac27a0ec 2377 bg = sbi->s_desc_per_block * nr;
617ba13b 2378 if (ext4_bg_has_super(sb, bg))
ac27a0ec 2379 has_super = 1;
0b8e58a1 2380
617ba13b 2381 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 2382}
2383
c9de560d
AT		 2384/**
2385 * ext4_get_stripe_size: Get the stripe size.
2386 * @sbi: In memory super block info
2387 *
2388 * If we have specified it via mount option, then
2389 * use the mount option value. If the value specified at mount time is
2390 * greater than the blocks per group use the super block value.
2391 * If the super block value is greater than blocks per group return 0.
2392 * Allocator needs it be less than blocks per group.
2393 *
2394 */
2395static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2396{
2397 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2398 unsigned long stripe_width =
2399 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3eb08658 2400 int ret;
c9de560d
AT		 2401
2402 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3eb08658
DE		 2403 ret = sbi->s_stripe;
2404 else if (stripe_width <= sbi->s_blocks_per_group)
2405 ret = stripe_width;
2406 else if (stride <= sbi->s_blocks_per_group)
2407 ret = stride;
2408 else
2409 ret = 0;
c9de560d 2410
3eb08658
DE		 2411 /*
2412 * If the stripe width is 1, this makes no sense and
2413 * we set it to 0 to turn off stripe handling code.
2414 */
2415 if (ret <= 1)
2416 ret = 0;
c9de560d 2417
3eb08658 2418 return ret;
c9de560d 2419}
ac27a0ec 2420
3197ebdb
TT		 2421/* sysfs supprt */
2422
2423struct ext4_attr {
2424 struct attribute attr;
2425 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
60e6679e 2426 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
3197ebdb
TT		 2427 const char *, size_t);
2428 int offset;
2429};
2430
2431static int parse_strtoul(const char *buf,
2432 unsigned long max, unsigned long *value)
2433{
2434 char *endp;
2435
e7d2860b
AGR		 2436 *value = simple_strtoul(skip_spaces(buf), &endp, 0);
2437 endp = skip_spaces(endp);
3197ebdb
TT		 2438 if (*endp || *value > max)
2439 return -EINVAL;
2440
2441 return 0;
2442}
2443
2444static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2445 struct ext4_sb_info *sbi,
2446 char *buf)
2447{
2448 return snprintf(buf, PAGE_SIZE, "%llu\n",
7b415bf6
AK		 2449 (s64) EXT4_C2B(sbi,
2450 percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
3197ebdb
TT		 2451}
2452
2453static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2454 struct ext4_sb_info *sbi, char *buf)
2455{
2456 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2457
f613dfcb
TT		 2458 if (!sb->s_bdev->bd_part)
2459 return snprintf(buf, PAGE_SIZE, "0\n");
3197ebdb
TT		 2460 return snprintf(buf, PAGE_SIZE, "%lu\n",
2461 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2462 sbi->s_sectors_written_start) >> 1);
2463}
2464
2465static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2466 struct ext4_sb_info *sbi, char *buf)
2467{
2468 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2469
f613dfcb
TT		 2470 if (!sb->s_bdev->bd_part)
2471 return snprintf(buf, PAGE_SIZE, "0\n");
3197ebdb 2472 return snprintf(buf, PAGE_SIZE, "%llu\n",
a6b43e38 2473 (unsigned long long)(sbi->s_kbytes_written +
3197ebdb 2474 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
a6b43e38 2475 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
3197ebdb
TT		 2476}
2477
2478static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2479 struct ext4_sb_info *sbi,
2480 const char *buf, size_t count)
2481{
2482 unsigned long t;
2483
2484 if (parse_strtoul(buf, 0x40000000, &t))
2485 return -EINVAL;
2486
5dbd571d 2487 if (t && !is_power_of_2(t))
3197ebdb
TT		 2488 return -EINVAL;
2489
2490 sbi->s_inode_readahead_blks = t;
2491 return count;
2492}
2493
2494static ssize_t sbi_ui_show(struct ext4_attr *a,
0b8e58a1 2495 struct ext4_sb_info *sbi, char *buf)
3197ebdb
TT		 2496{
2497 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2498
2499 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2500}
2501
2502static ssize_t sbi_ui_store(struct ext4_attr *a,
2503 struct ext4_sb_info *sbi,
2504 const char *buf, size_t count)
2505{
2506 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2507 unsigned long t;
2508
2509 if (parse_strtoul(buf, 0xffffffff, &t))
2510 return -EINVAL;
2511 *ui = t;
2512 return count;
2513}
2514
2c0544b2
TT		 2515static ssize_t trigger_test_error(struct ext4_attr *a,
2516 struct ext4_sb_info *sbi,
2517 const char *buf, size_t count)
2518{
2519 int len = count;
2520
2521 if (!capable(CAP_SYS_ADMIN))
2522 return -EPERM;
2523
2524 if (len && buf[len-1] == '\n')
2525 len--;
2526
2527 if (len)
2528 ext4_error(sbi->s_sb, "%.*s", len, buf);
2529 return count;
2530}
2531
3197ebdb
TT		 2532#define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2533static struct ext4_attr ext4_attr_##_name = { \
2534 .attr = {.name = __stringify(_name), .mode = _mode }, \
2535 .show = _show, \
2536 .store = _store, \
2537 .offset = offsetof(struct ext4_sb_info, _elname), \
2538}
2539#define EXT4_ATTR(name, mode, show, store) \
2540static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2541
857ac889 2542#define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
3197ebdb
TT		 2543#define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2544#define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2545#define EXT4_RW_ATTR_SBI_UI(name, elname) \
2546 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2547#define ATTR_LIST(name) &ext4_attr_##name.attr
2548
2549EXT4_RO_ATTR(delayed_allocation_blocks);
2550EXT4_RO_ATTR(session_write_kbytes);
2551EXT4_RO_ATTR(lifetime_write_kbytes);
2552EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2553 inode_readahead_blks_store, s_inode_readahead_blks);
11013911 2554EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
3197ebdb
TT		 2555EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2556EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2557EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2558EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2559EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2560EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
55138e0b 2561EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
67a5da56 2562EXT4_RW_ATTR_SBI_UI(extent_max_zeroout_kb, s_extent_max_zeroout_kb);
2c0544b2 2563EXT4_ATTR(trigger_fs_error, 0200, NULL, trigger_test_error);
3197ebdb
TT		 2564
2565static struct attribute *ext4_attrs[] = {
2566 ATTR_LIST(delayed_allocation_blocks),
2567 ATTR_LIST(session_write_kbytes),
2568 ATTR_LIST(lifetime_write_kbytes),
2569 ATTR_LIST(inode_readahead_blks),
11013911 2570 ATTR_LIST(inode_goal),
3197ebdb
TT		 2571 ATTR_LIST(mb_stats),
2572 ATTR_LIST(mb_max_to_scan),
2573 ATTR_LIST(mb_min_to_scan),
2574 ATTR_LIST(mb_order2_req),
2575 ATTR_LIST(mb_stream_req),
2576 ATTR_LIST(mb_group_prealloc),
55138e0b 2577 ATTR_LIST(max_writeback_mb_bump),
67a5da56 2578 ATTR_LIST(extent_max_zeroout_kb),
2c0544b2 2579 ATTR_LIST(trigger_fs_error),
3197ebdb
TT		 2580 NULL,
2581};
2582
857ac889
LC		 2583/* Features this copy of ext4 supports */
2584EXT4_INFO_ATTR(lazy_itable_init);
27ee40df 2585EXT4_INFO_ATTR(batched_discard);
5e7bbef1 2586EXT4_INFO_ATTR(meta_bg_resize);
857ac889
LC		 2587
2588static struct attribute *ext4_feat_attrs[] = {
2589 ATTR_LIST(lazy_itable_init),
27ee40df 2590 ATTR_LIST(batched_discard),
5e7bbef1 2591 ATTR_LIST(meta_bg_resize),
857ac889
LC		 2592 NULL,
2593};
2594
3197ebdb
TT		 2595static ssize_t ext4_attr_show(struct kobject *kobj,
2596 struct attribute *attr, char *buf)
2597{
2598 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2599 s_kobj);
2600 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2601
2602 return a->show ? a->show(a, sbi, buf) : 0;
2603}
2604
2605static ssize_t ext4_attr_store(struct kobject *kobj,
2606 struct attribute *attr,
2607 const char *buf, size_t len)
2608{
2609 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2610 s_kobj);
2611 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2612
2613 return a->store ? a->store(a, sbi, buf, len) : 0;
2614}
2615
2616static void ext4_sb_release(struct kobject *kobj)
2617{
2618 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2619 s_kobj);
2620 complete(&sbi->s_kobj_unregister);
2621}
2622
52cf25d0 2623static const struct sysfs_ops ext4_attr_ops = {
3197ebdb
TT		 2624 .show = ext4_attr_show,
2625 .store = ext4_attr_store,
2626};
2627
2628static struct kobj_type ext4_ktype = {
2629 .default_attrs = ext4_attrs,
2630 .sysfs_ops = &ext4_attr_ops,
2631 .release = ext4_sb_release,
2632};
2633
857ac889
LC		 2634static void ext4_feat_release(struct kobject *kobj)
2635{
2636 complete(&ext4_feat->f_kobj_unregister);
2637}
2638
2639static struct kobj_type ext4_feat_ktype = {
2640 .default_attrs = ext4_feat_attrs,
2641 .sysfs_ops = &ext4_attr_ops,
2642 .release = ext4_feat_release,
2643};
2644
a13fb1a4
ES		 2645/*
2646 * Check whether this filesystem can be mounted based on
2647 * the features present and the RDONLY/RDWR mount requested.
2648 * Returns 1 if this filesystem can be mounted as requested,
2649 * 0 if it cannot be.
2650 */
2651static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2652{
2653 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2654 ext4_msg(sb, KERN_ERR,
2655 "Couldn't mount because of "
2656 "unsupported optional features (%x)",
2657 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2658 ~EXT4_FEATURE_INCOMPAT_SUPP));
2659 return 0;
2660 }
2661
2662 if (readonly)
2663 return 1;
2664
2665 /* Check that feature set is OK for a read-write mount */
2666 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2667 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2668 "unsupported optional features (%x)",
2669 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2670 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2671 return 0;
2672 }
2673 /*
2674 * Large file size enabled file system can only be mounted
2675 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2676 */
2677 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2678 if (sizeof(blkcnt_t) < sizeof(u64)) {
2679 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2680 "cannot be mounted RDWR without "
2681 "CONFIG_LBDAF");
2682 return 0;
2683 }
2684 }
bab08ab9
TT		 2685 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2686 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2687 ext4_msg(sb, KERN_ERR,
2688 "Can't support bigalloc feature without "
2689 "extents feature\n");
2690 return 0;
2691 }
7c319d32
AK		 2692
2693#ifndef CONFIG_QUOTA
2694 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
2695 !readonly) {
2696 ext4_msg(sb, KERN_ERR,
2697 "Filesystem with quota feature cannot be mounted RDWR "
2698 "without CONFIG_QUOTA");
2699 return 0;
2700 }
2701#endif /* CONFIG_QUOTA */
a13fb1a4
ES		 2702 return 1;
2703}
2704
66e61a9e
TT		 2705/*
2706 * This function is called once a day if we have errors logged
2707 * on the file system
2708 */
2709static void print_daily_error_info(unsigned long arg)
2710{
2711 struct super_block *sb = (struct super_block *) arg;
2712 struct ext4_sb_info *sbi;
2713 struct ext4_super_block *es;
2714
2715 sbi = EXT4_SB(sb);
2716 es = sbi->s_es;
2717
2718 if (es->s_error_count)
2719 ext4_msg(sb, KERN_NOTICE, "error count: %u",
2720 le32_to_cpu(es->s_error_count));
2721 if (es->s_first_error_time) {
2722 printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d",
2723 sb->s_id, le32_to_cpu(es->s_first_error_time),
2724 (int) sizeof(es->s_first_error_func),
2725 es->s_first_error_func,
2726 le32_to_cpu(es->s_first_error_line));
2727 if (es->s_first_error_ino)
2728 printk(": inode %u",
2729 le32_to_cpu(es->s_first_error_ino));
2730 if (es->s_first_error_block)
2731 printk(": block %llu", (unsigned long long)
2732 le64_to_cpu(es->s_first_error_block));
2733 printk("\n");
2734 }
2735 if (es->s_last_error_time) {
2736 printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d",
2737 sb->s_id, le32_to_cpu(es->s_last_error_time),
2738 (int) sizeof(es->s_last_error_func),
2739 es->s_last_error_func,
2740 le32_to_cpu(es->s_last_error_line));
2741 if (es->s_last_error_ino)
2742 printk(": inode %u",
2743 le32_to_cpu(es->s_last_error_ino));
2744 if (es->s_last_error_block)
2745 printk(": block %llu", (unsigned long long)
2746 le64_to_cpu(es->s_last_error_block));
2747 printk("\n");
2748 }
2749 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2750}
2751
bfff6873
LC		 2752/* Find next suitable group and run ext4_init_inode_table */
2753static int ext4_run_li_request(struct ext4_li_request *elr)
2754{
2755 struct ext4_group_desc *gdp = NULL;
2756 ext4_group_t group, ngroups;
2757 struct super_block *sb;
2758 unsigned long timeout = 0;
2759 int ret = 0;
2760
2761 sb = elr->lr_super;
2762 ngroups = EXT4_SB(sb)->s_groups_count;
2763
8e8ad8a5 2764 sb_start_write(sb);
bfff6873
LC		 2765 for (group = elr->lr_next_group; group < ngroups; group++) {
2766 gdp = ext4_get_group_desc(sb, group, NULL);
2767 if (!gdp) {
2768 ret = 1;
2769 break;
2770 }
2771
2772 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2773 break;
2774 }
2775
2776 if (group == ngroups)
2777 ret = 1;
2778
2779 if (!ret) {
2780 timeout = jiffies;
2781 ret = ext4_init_inode_table(sb, group,
2782 elr->lr_timeout ? 0 : 1);
2783 if (elr->lr_timeout == 0) {
51ce6511
LC		 2784 timeout = (jiffies - timeout) *
2785 elr->lr_sbi->s_li_wait_mult;
bfff6873
LC		 2786 elr->lr_timeout = timeout;
2787 }
2788 elr->lr_next_sched = jiffies + elr->lr_timeout;
2789 elr->lr_next_group = group + 1;
2790 }
8e8ad8a5 2791 sb_end_write(sb);
bfff6873
LC		 2792
2793 return ret;
2794}
2795
2796/*
2797 * Remove lr_request from the list_request and free the
4ed5c033 2798 * request structure. Should be called with li_list_mtx held
bfff6873
LC		 2799 */
2800static void ext4_remove_li_request(struct ext4_li_request *elr)
2801{
2802 struct ext4_sb_info *sbi;
2803
2804 if (!elr)
2805 return;
2806
2807 sbi = elr->lr_sbi;
2808
2809 list_del(&elr->lr_request);
2810 sbi->s_li_request = NULL;
2811 kfree(elr);
2812}
2813
2814static void ext4_unregister_li_request(struct super_block *sb)
2815{
1bb933fb
LC		 2816 mutex_lock(&ext4_li_mtx);
2817 if (!ext4_li_info) {
2818 mutex_unlock(&ext4_li_mtx);
bfff6873 2819 return;
1bb933fb 2820 }
bfff6873
LC		 2821
2822 mutex_lock(&ext4_li_info->li_list_mtx);
1bb933fb 2823 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
bfff6873 2824 mutex_unlock(&ext4_li_info->li_list_mtx);
1bb933fb 2825 mutex_unlock(&ext4_li_mtx);
bfff6873
LC		 2826}
2827
8f1f7453
ES		 2828static struct task_struct *ext4_lazyinit_task;
2829
bfff6873
LC		 2830/*
2831 * This is the function where ext4lazyinit thread lives. It walks
2832 * through the request list searching for next scheduled filesystem.
2833 * When such a fs is found, run the lazy initialization request
2834 * (ext4_rn_li_request) and keep track of the time spend in this
2835 * function. Based on that time we compute next schedule time of
2836 * the request. When walking through the list is complete, compute
2837 * next waking time and put itself into sleep.
2838 */
2839static int ext4_lazyinit_thread(void *arg)
2840{
2841 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2842 struct list_head *pos, *n;
2843 struct ext4_li_request *elr;
4ed5c033 2844 unsigned long next_wakeup, cur;
bfff6873
LC		 2845
2846 BUG_ON(NULL == eli);
2847
bfff6873
LC		 2848cont_thread:
2849 while (true) {
2850 next_wakeup = MAX_JIFFY_OFFSET;
2851
2852 mutex_lock(&eli->li_list_mtx);
2853 if (list_empty(&eli->li_request_list)) {
2854 mutex_unlock(&eli->li_list_mtx);
2855 goto exit_thread;
2856 }
2857
2858 list_for_each_safe(pos, n, &eli->li_request_list) {
2859 elr = list_entry(pos, struct ext4_li_request,
2860 lr_request);
2861
b2c78cd0
TT		 2862 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2863 if (ext4_run_li_request(elr) != 0) {
2864 /* error, remove the lazy_init job */
2865 ext4_remove_li_request(elr);
2866 continue;
2867 }
bfff6873
LC		 2868 }
2869
2870 if (time_before(elr->lr_next_sched, next_wakeup))
2871 next_wakeup = elr->lr_next_sched;
2872 }
2873 mutex_unlock(&eli->li_list_mtx);
2874
a0acae0e 2875 try_to_freeze();
bfff6873 2876
4ed5c033
LC		 2877 cur = jiffies;
2878 if ((time_after_eq(cur, next_wakeup)) ||
f4245bd4 2879 (MAX_JIFFY_OFFSET == next_wakeup)) {
bfff6873
LC		 2880 cond_resched();
2881 continue;
2882 }
2883
4ed5c033
LC		 2884 schedule_timeout_interruptible(next_wakeup - cur);
2885
8f1f7453
ES		 2886 if (kthread_should_stop()) {
2887 ext4_clear_request_list();
2888 goto exit_thread;
2889 }
bfff6873
LC		 2890 }
2891
2892exit_thread:
2893 /*
2894 * It looks like the request list is empty, but we need
2895 * to check it under the li_list_mtx lock, to prevent any
2896 * additions into it, and of course we should lock ext4_li_mtx
2897 * to atomically free the list and ext4_li_info, because at
2898 * this point another ext4 filesystem could be registering
2899 * new one.
2900 */
2901 mutex_lock(&ext4_li_mtx);
2902 mutex_lock(&eli->li_list_mtx);
2903 if (!list_empty(&eli->li_request_list)) {
2904 mutex_unlock(&eli->li_list_mtx);
2905 mutex_unlock(&ext4_li_mtx);
2906 goto cont_thread;
2907 }
2908 mutex_unlock(&eli->li_list_mtx);
bfff6873
LC		 2909 kfree(ext4_li_info);
2910 ext4_li_info = NULL;
2911 mutex_unlock(&ext4_li_mtx);
2912
2913 return 0;
2914}
2915
2916static void ext4_clear_request_list(void)
2917{
2918 struct list_head *pos, *n;
2919 struct ext4_li_request *elr;
2920
2921 mutex_lock(&ext4_li_info->li_list_mtx);
bfff6873
LC		 2922 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2923 elr = list_entry(pos, struct ext4_li_request,
2924 lr_request);
2925 ext4_remove_li_request(elr);
2926 }
2927 mutex_unlock(&ext4_li_info->li_list_mtx);
2928}
2929
2930static int ext4_run_lazyinit_thread(void)
2931{
8f1f7453
ES		 2932 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2933 ext4_li_info, "ext4lazyinit");
2934 if (IS_ERR(ext4_lazyinit_task)) {
2935 int err = PTR_ERR(ext4_lazyinit_task);
bfff6873 2936 ext4_clear_request_list();
bfff6873
LC		 2937 kfree(ext4_li_info);
2938 ext4_li_info = NULL;
92b97816 2939 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
bfff6873
LC		 2940 "initialization thread\n",
2941 err);
2942 return err;
2943 }
2944 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
bfff6873
LC		 2945 return 0;
2946}
2947
2948/*
2949 * Check whether it make sense to run itable init. thread or not.
2950 * If there is at least one uninitialized inode table, return
2951 * corresponding group number, else the loop goes through all
2952 * groups and return total number of groups.
2953 */
2954static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2955{
2956 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2957 struct ext4_group_desc *gdp = NULL;
2958
2959 for (group = 0; group < ngroups; group++) {
2960 gdp = ext4_get_group_desc(sb, group, NULL);
2961 if (!gdp)
2962 continue;
2963
2964 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2965 break;
2966 }
2967
2968 return group;
2969}
2970
2971static int ext4_li_info_new(void)
2972{
2973 struct ext4_lazy_init *eli = NULL;
2974
2975 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2976 if (!eli)
2977 return -ENOMEM;
2978
bfff6873
LC		 2979 INIT_LIST_HEAD(&eli->li_request_list);
2980 mutex_init(&eli->li_list_mtx);
2981
bfff6873
LC		 2982 eli->li_state |= EXT4_LAZYINIT_QUIT;
2983
2984 ext4_li_info = eli;
2985
2986 return 0;
2987}
2988
2989static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
2990 ext4_group_t start)
2991{
2992 struct ext4_sb_info *sbi = EXT4_SB(sb);
2993 struct ext4_li_request *elr;
2994 unsigned long rnd;
2995
2996 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
2997 if (!elr)
2998 return NULL;
2999
3000 elr->lr_super = sb;
3001 elr->lr_sbi = sbi;
3002 elr->lr_next_group = start;
3003
3004 /*
3005 * Randomize first schedule time of the request to
3006 * spread the inode table initialization requests
3007 * better.
3008 */
3009 get_random_bytes(&rnd, sizeof(rnd));
3010 elr->lr_next_sched = jiffies + (unsigned long)rnd %
3011 (EXT4_DEF_LI_MAX_START_DELAY * HZ);
3012
3013 return elr;
3014}
3015
3016static int ext4_register_li_request(struct super_block *sb,
3017 ext4_group_t first_not_zeroed)
3018{
3019 struct ext4_sb_info *sbi = EXT4_SB(sb);
3020 struct ext4_li_request *elr;
3021 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
6c5a6cb9 3022 int ret = 0;
bfff6873 3023
51ce6511
LC		 3024 if (sbi->s_li_request != NULL) {
3025 /*
3026 * Reset timeout so it can be computed again, because
3027 * s_li_wait_mult might have changed.
3028 */
3029 sbi->s_li_request->lr_timeout = 0;
beed5ecb 3030 return 0;
51ce6511 3031 }
bfff6873
LC		 3032
3033 if (first_not_zeroed == ngroups ||
3034 (sb->s_flags & MS_RDONLY) ||
55ff3840 3035 !test_opt(sb, INIT_INODE_TABLE))
beed5ecb 3036 return 0;
bfff6873
LC		 3037
3038 elr = ext4_li_request_new(sb, first_not_zeroed);
beed5ecb
NK		 3039 if (!elr)
3040 return -ENOMEM;
bfff6873
LC		 3041
3042 mutex_lock(&ext4_li_mtx);
3043
3044 if (NULL == ext4_li_info) {
3045 ret = ext4_li_info_new();
3046 if (ret)
3047 goto out;
3048 }
3049
3050 mutex_lock(&ext4_li_info->li_list_mtx);
3051 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3052 mutex_unlock(&ext4_li_info->li_list_mtx);
3053
3054 sbi->s_li_request = elr;
46e4690b
TM		 3055 /*
3056 * set elr to NULL here since it has been inserted to
3057 * the request_list and the removal and free of it is
3058 * handled by ext4_clear_request_list from now on.
3059 */
3060 elr = NULL;
bfff6873
LC		 3061
3062 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3063 ret = ext4_run_lazyinit_thread();
3064 if (ret)
3065 goto out;
3066 }
bfff6873 3067out:
beed5ecb
NK		 3068 mutex_unlock(&ext4_li_mtx);
3069 if (ret)
bfff6873 3070 kfree(elr);
bfff6873
LC		 3071 return ret;
3072}
3073
3074/*
3075 * We do not need to lock anything since this is called on
3076 * module unload.
3077 */
3078static void ext4_destroy_lazyinit_thread(void)
3079{
3080 /*
3081 * If thread exited earlier
3082 * there's nothing to be done.
3083 */
8f1f7453 3084 if (!ext4_li_info || !ext4_lazyinit_task)
bfff6873
LC		 3085 return;
3086
8f1f7453 3087 kthread_stop(ext4_lazyinit_task);
bfff6873
LC		 3088}
3089
25ed6e8a
DW		 3090static int set_journal_csum_feature_set(struct super_block *sb)
3091{
3092 int ret = 1;
3093 int compat, incompat;
3094 struct ext4_sb_info *sbi = EXT4_SB(sb);
3095
3096 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3097 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3098 /* journal checksum v2 */
3099 compat = 0;
3100 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V2;
3101 } else {
3102 /* journal checksum v1 */
3103 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3104 incompat = 0;
3105 }
3106
3107 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3108 ret = jbd2_journal_set_features(sbi->s_journal,
3109 compat, 0,
3110 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3111 incompat);
3112 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3113 ret = jbd2_journal_set_features(sbi->s_journal,
3114 compat, 0,
3115 incompat);
3116 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3117 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3118 } else {
3119 jbd2_journal_clear_features(sbi->s_journal,
3120 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3121 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3122 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3123 }
3124
3125 return ret;
3126}
3127
952fc18e
TT		 3128/*
3129 * Note: calculating the overhead so we can be compatible with
3130 * historical BSD practice is quite difficult in the face of
3131 * clusters/bigalloc. This is because multiple metadata blocks from
3132 * different block group can end up in the same allocation cluster.
3133 * Calculating the exact overhead in the face of clustered allocation
3134 * requires either O(all block bitmaps) in memory or O(number of block
3135 * groups**2) in time. We will still calculate the superblock for
3136 * older file systems --- and if we come across with a bigalloc file
3137 * system with zero in s_overhead_clusters the estimate will be close to
3138 * correct especially for very large cluster sizes --- but for newer
3139 * file systems, it's better to calculate this figure once at mkfs
3140 * time, and store it in the superblock. If the superblock value is
3141 * present (even for non-bigalloc file systems), we will use it.
3142 */
3143static int count_overhead(struct super_block *sb, ext4_group_t grp,
3144 char *buf)
3145{
3146 struct ext4_sb_info *sbi = EXT4_SB(sb);
3147 struct ext4_group_desc *gdp;
3148 ext4_fsblk_t first_block, last_block, b;
3149 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3150 int s, j, count = 0;
3151
0548bbb8
TT		 3152 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC))
3153 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3154 sbi->s_itb_per_group + 2);
3155
952fc18e
TT		 3156 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3157 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3158 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3159 for (i = 0; i < ngroups; i++) {
3160 gdp = ext4_get_group_desc(sb, i, NULL);
3161 b = ext4_block_bitmap(sb, gdp);
3162 if (b >= first_block && b <= last_block) {
3163 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3164 count++;
3165 }
3166 b = ext4_inode_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_table(sb, gdp);
3172 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3173 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3174 int c = EXT4_B2C(sbi, b - first_block);
3175 ext4_set_bit(c, buf);
3176 count++;
3177 }
3178 if (i != grp)
3179 continue;
3180 s = 0;
3181 if (ext4_bg_has_super(sb, grp)) {
3182 ext4_set_bit(s++, buf);
3183 count++;
3184 }
3185 for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
3186 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3187 count++;
3188 }
3189 }
3190 if (!count)
3191 return 0;
3192 return EXT4_CLUSTERS_PER_GROUP(sb) -
3193 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3194}
3195
3196/*
3197 * Compute the overhead and stash it in sbi->s_overhead
3198 */
3199int ext4_calculate_overhead(struct super_block *sb)
3200{
3201 struct ext4_sb_info *sbi = EXT4_SB(sb);
3202 struct ext4_super_block *es = sbi->s_es;
3203 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3204 ext4_fsblk_t overhead = 0;
3205 char *buf = (char *) get_zeroed_page(GFP_KERNEL);
3206
3207 memset(buf, 0, PAGE_SIZE);
3208 if (!buf)
3209 return -ENOMEM;
3210
3211 /*
3212 * Compute the overhead (FS structures). This is constant
3213 * for a given filesystem unless the number of block groups
3214 * changes so we cache the previous value until it does.
3215 */
3216
3217 /*
3218 * All of the blocks before first_data_block are overhead
3219 */
3220 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3221
3222 /*
3223 * Add the overhead found in each block group
3224 */
3225 for (i = 0; i < ngroups; i++) {
3226 int blks;
3227
3228 blks = count_overhead(sb, i, buf);
3229 overhead += blks;
3230 if (blks)
3231 memset(buf, 0, PAGE_SIZE);
3232 cond_resched();
3233 }
3234 sbi->s_overhead = overhead;
3235 smp_wmb();
3236 free_page((unsigned long) buf);
3237 return 0;
3238}
3239
2b2d6d01 3240static int ext4_fill_super(struct super_block *sb, void *data, int silent)
ac27a0ec 3241{
d4c402d9 3242 char *orig_data = kstrdup(data, GFP_KERNEL);
2b2d6d01 3243 struct buffer_head *bh;
617ba13b
MC		 3244 struct ext4_super_block *es = NULL;
3245 struct ext4_sb_info *sbi;
3246 ext4_fsblk_t block;
3247 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 3248 ext4_fsblk_t logical_sb_block;
ac27a0ec 3249 unsigned long offset = 0;
ac27a0ec
DK		 3250 unsigned long journal_devnum = 0;
3251 unsigned long def_mount_opts;
3252 struct inode *root;
9f6200bb 3253 char *cp;
0390131b 3254 const char *descr;
dcc7dae3 3255 int ret = -ENOMEM;
281b5995 3256 int blocksize, clustersize;
4ec11028
TT		 3257 unsigned int db_count;
3258 unsigned int i;
281b5995 3259 int needs_recovery, has_huge_files, has_bigalloc;
bd81d8ee 3260 __u64 blocks_count;
07aa2ea1 3261 int err = 0;
b3881f74 3262 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
bfff6873 3263 ext4_group_t first_not_zeroed;
ac27a0ec
DK		 3264
3265 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3266 if (!sbi)
dcc7dae3 3267 goto out_free_orig;
705895b6
PE		 3268
3269 sbi->s_blockgroup_lock =
3270 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3271 if (!sbi->s_blockgroup_lock) {
3272 kfree(sbi);
dcc7dae3 3273 goto out_free_orig;
705895b6 3274 }
ac27a0ec 3275 sb->s_fs_info = sbi;
2c0544b2 3276 sbi->s_sb = sb;
240799cd 3277 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
d9c9bef1 3278 sbi->s_sb_block = sb_block;
f613dfcb
TT		 3279 if (sb->s_bdev->bd_part)
3280 sbi->s_sectors_written_start =
3281 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
ac27a0ec 3282
9f6200bb
TT		 3283 /* Cleanup superblock name */
3284 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3285 *cp = '!';
3286
07aa2ea1 3287 /* -EINVAL is default */
dcc7dae3 3288 ret = -EINVAL;
617ba13b 3289 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 3290 if (!blocksize) {
b31e1552 3291 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
ac27a0ec
DK		 3292 goto out_fail;
3293 }
3294
3295 /*
617ba13b 3296 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 3297 * block sizes. We need to calculate the offset from buffer start.
3298 */
617ba13b 3299 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 3300 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3301 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 3302 } else {
70bbb3e0 3303 logical_sb_block = sb_block;
ac27a0ec
DK		 3304 }
3305
70bbb3e0 3306 if (!(bh = sb_bread(sb, logical_sb_block))) {
b31e1552 3307 ext4_msg(sb, KERN_ERR, "unable to read superblock");
ac27a0ec
DK		 3308 goto out_fail;
3309 }
3310 /*
3311 * Note: s_es must be initialized as soon as possible because
617ba13b 3312 * some ext4 macro-instructions depend on its value
ac27a0ec 3313 */
2716b802 3314 es = (struct ext4_super_block *) (bh->b_data + offset);
ac27a0ec
DK		 3315 sbi->s_es = es;
3316 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 3317 if (sb->s_magic != EXT4_SUPER_MAGIC)
3318 goto cantfind_ext4;
afc32f7e 3319 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
ac27a0ec 3320
feb0ab32
DW		 3321 /* Warn if metadata_csum and gdt_csum are both set. */
3322 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3323 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
3324 EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_GDT_CSUM))
3325 ext4_warning(sb, KERN_INFO "metadata_csum and uninit_bg are "
3326 "redundant flags; please run fsck.");
3327
d25425f8
DW		 3328 /* Check for a known checksum algorithm */
3329 if (!ext4_verify_csum_type(sb, es)) {
3330 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3331 "unknown checksum algorithm.");
3332 silent = 1;
3333 goto cantfind_ext4;
3334 }
3335
0441984a
DW		 3336 /* Load the checksum driver */
3337 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3338 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3339 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3340 if (IS_ERR(sbi->s_chksum_driver)) {
3341 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3342 ret = PTR_ERR(sbi->s_chksum_driver);
3343 sbi->s_chksum_driver = NULL;
3344 goto failed_mount;
3345 }
3346 }
3347
a9c47317
DW		 3348 /* Check superblock checksum */
3349 if (!ext4_superblock_csum_verify(sb, es)) {
3350 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3351 "invalid superblock checksum. Run e2fsck?");
3352 silent = 1;
3353 goto cantfind_ext4;
3354 }
3355
3356 /* Precompute checksum seed for all metadata */
3357 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3358 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
3359 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3360 sizeof(es->s_uuid));
3361
ac27a0ec
DK		 3362 /* Set defaults before we parse the mount options */
3363 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
fd8c37ec 3364 set_opt(sb, INIT_INODE_TABLE);
617ba13b 3365 if (def_mount_opts & EXT4_DEFM_DEBUG)
fd8c37ec 3366 set_opt(sb, DEBUG);
87f26807 3367 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
fd8c37ec 3368 set_opt(sb, GRPID);
617ba13b 3369 if (def_mount_opts & EXT4_DEFM_UID16)
fd8c37ec 3370 set_opt(sb, NO_UID32);
ea663336 3371 /* xattr user namespace & acls are now defaulted on */
03010a33 3372#ifdef CONFIG_EXT4_FS_XATTR
ea663336 3373 set_opt(sb, XATTR_USER);
2e7842b8 3374#endif
03010a33 3375#ifdef CONFIG_EXT4_FS_POSIX_ACL
ea663336 3376 set_opt(sb, POSIX_ACL);
2e7842b8 3377#endif
6fd7a467 3378 set_opt(sb, MBLK_IO_SUBMIT);
617ba13b 3379 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
fd8c37ec 3380 set_opt(sb, JOURNAL_DATA);
617ba13b 3381 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
fd8c37ec 3382 set_opt(sb, ORDERED_DATA);
617ba13b 3383 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
fd8c37ec 3384 set_opt(sb, WRITEBACK_DATA);
617ba13b
MC		 3385
3386 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
fd8c37ec 3387 set_opt(sb, ERRORS_PANIC);
bb4f397a 3388 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
fd8c37ec 3389 set_opt(sb, ERRORS_CONT);
bb4f397a 3390 else
fd8c37ec 3391 set_opt(sb, ERRORS_RO);
8b67f04a 3392 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
fd8c37ec 3393 set_opt(sb, BLOCK_VALIDITY);
8b67f04a 3394 if (def_mount_opts & EXT4_DEFM_DISCARD)
fd8c37ec 3395 set_opt(sb, DISCARD);
ac27a0ec 3396
08cefc7a
EB		 3397 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3398 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
30773840
TT		 3399 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3400 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3401 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
ac27a0ec 3402
8b67f04a 3403 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
fd8c37ec 3404 set_opt(sb, BARRIER);
ac27a0ec 3405
dd919b98
AK		 3406 /*
3407 * enable delayed allocation by default
3408 * Use -o nodelalloc to turn it off
3409 */
bc0b75f7 3410 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
8b67f04a 3411 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
fd8c37ec 3412 set_opt(sb, DELALLOC);
dd919b98 3413
51ce6511
LC		 3414 /*
3415 * set default s_li_wait_mult for lazyinit, for the case there is
3416 * no mount option specified.
3417 */
3418 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3419
8b67f04a 3420 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
661aa520 3421 &journal_devnum, &journal_ioprio, 0)) {
8b67f04a
TT		 3422 ext4_msg(sb, KERN_WARNING,
3423 "failed to parse options in superblock: %s",
3424 sbi->s_es->s_mount_opts);
3425 }
5a916be1 3426 sbi->s_def_mount_opt = sbi->s_mount_opt;
b3881f74 3427 if (!parse_options((char *) data, sb, &journal_devnum,
661aa520 3428 &journal_ioprio, 0))
ac27a0ec
DK		 3429 goto failed_mount;
3430
56889787
TT		 3431 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3432 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3433 "with data=journal disables delayed "
3434 "allocation and O_DIRECT support!\n");
3435 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3436 ext4_msg(sb, KERN_ERR, "can't mount with "
3437 "both data=journal and delalloc");
3438 goto failed_mount;
3439 }
3440 if (test_opt(sb, DIOREAD_NOLOCK)) {
3441 ext4_msg(sb, KERN_ERR, "can't mount with "
3442 "both data=journal and delalloc");
3443 goto failed_mount;
3444 }
3445 if (test_opt(sb, DELALLOC))
3446 clear_opt(sb, DELALLOC);
3447 }
3448
3449 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3450 if (test_opt(sb, DIOREAD_NOLOCK)) {
3451 if (blocksize < PAGE_SIZE) {
3452 ext4_msg(sb, KERN_ERR, "can't mount with "
3453 "dioread_nolock if block size != PAGE_SIZE");
3454 goto failed_mount;
3455 }
3456 }
3457
ac27a0ec 3458 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 3459 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec 3460
617ba13b
MC		 3461 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3462 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3463 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3464 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
b31e1552
ES		 3465 ext4_msg(sb, KERN_WARNING,
3466 "feature flags set on rev 0 fs, "
3467 "running e2fsck is recommended");
469108ff 3468
2035e776
TT		 3469 if (IS_EXT2_SB(sb)) {
3470 if (ext2_feature_set_ok(sb))
3471 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3472 "using the ext4 subsystem");
3473 else {
3474 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3475 "to feature incompatibilities");
3476 goto failed_mount;
3477 }
3478 }
3479
3480 if (IS_EXT3_SB(sb)) {
3481 if (ext3_feature_set_ok(sb))
3482 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3483 "using the ext4 subsystem");
3484 else {
3485 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3486 "to feature incompatibilities");
3487 goto failed_mount;
3488 }
3489 }
3490
ac27a0ec
DK		 3491 /*
3492 * Check feature flags regardless of the revision level, since we
3493 * previously didn't change the revision level when setting the flags,
3494 * so there is a chance incompat flags are set on a rev 0 filesystem.
3495 */
a13fb1a4 3496 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
ac27a0ec 3497 goto failed_mount;
a13fb1a4 3498
617ba13b
MC		 3499 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3500 blocksize > EXT4_MAX_BLOCK_SIZE) {
b31e1552
ES		 3501 ext4_msg(sb, KERN_ERR,
3502 "Unsupported filesystem blocksize %d", blocksize);
ac27a0ec
DK		 3503 goto failed_mount;
3504 }
3505
ac27a0ec 3506 if (sb->s_blocksize != blocksize) {
ce40733c
AK		 3507 /* Validate the filesystem blocksize */
3508 if (!sb_set_blocksize(sb, blocksize)) {
b31e1552 3509 ext4_msg(sb, KERN_ERR, "bad block size %d",
ce40733c 3510 blocksize);
ac27a0ec
DK		 3511 goto failed_mount;
3512 }
3513
2b2d6d01 3514 brelse(bh);
70bbb3e0
AM		 3515 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3516 offset = do_div(logical_sb_block, blocksize);
3517 bh = sb_bread(sb, logical_sb_block);
ac27a0ec 3518 if (!bh) {
b31e1552
ES		 3519 ext4_msg(sb, KERN_ERR,
3520 "Can't read superblock on 2nd try");
ac27a0ec
DK		 3521 goto failed_mount;
3522 }
2716b802 3523 es = (struct ext4_super_block *)(bh->b_data + offset);
ac27a0ec 3524 sbi->s_es = es;
617ba13b 3525 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
b31e1552
ES		 3526 ext4_msg(sb, KERN_ERR,
3527 "Magic mismatch, very weird!");
ac27a0ec
DK		 3528 goto failed_mount;
3529 }
3530 }
3531
a13fb1a4
ES		 3532 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3533 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
f287a1a5
TT		 3534 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3535 has_huge_files);
3536 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
ac27a0ec 3537
617ba13b
MC		 3538 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3539 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3540 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
ac27a0ec
DK		 3541 } else {
3542 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3543 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
617ba13b 3544 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1330593e 3545 (!is_power_of_2(sbi->s_inode_size)) ||
ac27a0ec 3546 (sbi->s_inode_size > blocksize)) {
b31e1552
ES		 3547 ext4_msg(sb, KERN_ERR,
3548 "unsupported inode size: %d",
2b2d6d01 3549 sbi->s_inode_size);
ac27a0ec
DK		 3550 goto failed_mount;
3551 }
ef7f3835
KS		 3552 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3553 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
ac27a0ec 3554 }
0b8e58a1 3555
0d1ee42f
AR		 3556 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3557 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
8fadc143 3558 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 3559 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 3560 !is_power_of_2(sbi->s_desc_size)) {
b31e1552
ES		 3561 ext4_msg(sb, KERN_ERR,
3562 "unsupported descriptor size %lu",
0d1ee42f
AR		 3563 sbi->s_desc_size);
3564 goto failed_mount;
3565 }
3566 } else
3567 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
0b8e58a1 3568
ac27a0ec 3569 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 3570 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
b47b6f38 3571 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
617ba13b 3572 goto cantfind_ext4;
0b8e58a1 3573
617ba13b 3574 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 3575 if (sbi->s_inodes_per_block == 0)
617ba13b 3576 goto cantfind_ext4;
ac27a0ec
DK		 3577 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3578 sbi->s_inodes_per_block;
0d1ee42f 3579 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 3580 sbi->s_sbh = bh;
3581 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 3582 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3583 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
0b8e58a1 3584
2b2d6d01 3585 for (i = 0; i < 4; i++)
ac27a0ec
DK		 3586 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3587 sbi->s_def_hash_version = es->s_def_hash_version;
f99b2589
TT		 3588 i = le32_to_cpu(es->s_flags);
3589 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3590 sbi->s_hash_unsigned = 3;
3591 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3592#ifdef __CHAR_UNSIGNED__
3593 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3594 sbi->s_hash_unsigned = 3;
3595#else
3596 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3597#endif
f99b2589 3598 }
ac27a0ec 3599
281b5995
TT		 3600 /* Handle clustersize */
3601 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3602 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3603 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3604 if (has_bigalloc) {
3605 if (clustersize < blocksize) {
3606 ext4_msg(sb, KERN_ERR,
3607 "cluster size (%d) smaller than "
3608 "block size (%d)", clustersize, blocksize);
3609 goto failed_mount;
3610 }
3611 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3612 le32_to_cpu(es->s_log_block_size);
3613 sbi->s_clusters_per_group =
3614 le32_to_cpu(es->s_clusters_per_group);
3615 if (sbi->s_clusters_per_group > blocksize * 8) {
3616 ext4_msg(sb, KERN_ERR,
3617 "#clusters per group too big: %lu",
3618 sbi->s_clusters_per_group);
3619 goto failed_mount;
3620 }
3621 if (sbi->s_blocks_per_group !=
3622 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3623 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3624 "clusters per group (%lu) inconsistent",
3625 sbi->s_blocks_per_group,
3626 sbi->s_clusters_per_group);
3627 goto failed_mount;
3628 }
3629 } else {
3630 if (clustersize != blocksize) {
3631 ext4_warning(sb, "fragment/cluster size (%d) != "
3632 "block size (%d)", clustersize,
3633 blocksize);
3634 clustersize = blocksize;
3635 }
3636 if (sbi->s_blocks_per_group > blocksize * 8) {
3637 ext4_msg(sb, KERN_ERR,
3638 "#blocks per group too big: %lu",
3639 sbi->s_blocks_per_group);
3640 goto failed_mount;
3641 }
3642 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3643 sbi->s_cluster_bits = 0;
ac27a0ec 3644 }
281b5995
TT		 3645 sbi->s_cluster_ratio = clustersize / blocksize;
3646
ac27a0ec 3647 if (sbi->s_inodes_per_group > blocksize * 8) {
b31e1552
ES		 3648 ext4_msg(sb, KERN_ERR,
3649 "#inodes per group too big: %lu",
2b2d6d01 3650 sbi->s_inodes_per_group);
ac27a0ec
DK		 3651 goto failed_mount;
3652 }
3653
bf43d84b
ES		 3654 /*
3655 * Test whether we have more sectors than will fit in sector_t,
3656 * and whether the max offset is addressable by the page cache.
3657 */
5a9ae68a 3658 err = generic_check_addressable(sb->s_blocksize_bits,
30ca22c7 3659 ext4_blocks_count(es));
5a9ae68a 3660 if (err) {
b31e1552 3661 ext4_msg(sb, KERN_ERR, "filesystem"
bf43d84b 3662 " too large to mount safely on this system");
ac27a0ec 3663 if (sizeof(sector_t) < 8)
90c699a9 3664 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
ac27a0ec
DK		 3665 goto failed_mount;
3666 }
3667
617ba13b
MC		 3668 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3669 goto cantfind_ext4;
e7c95593 3670
0f2ddca6
FTN		 3671 /* check blocks count against device size */
3672 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3673 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
b31e1552
ES		 3674 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3675 "exceeds size of device (%llu blocks)",
0f2ddca6
FTN		 3676 ext4_blocks_count(es), blocks_count);
3677 goto failed_mount;
3678 }
3679
0b8e58a1
AD		 3680 /*
3681 * It makes no sense for the first data block to be beyond the end
3682 * of the filesystem.
3683 */
3684 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
5635a62b 3685 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
b31e1552
ES		 3686 "block %u is beyond end of filesystem (%llu)",
3687 le32_to_cpu(es->s_first_data_block),
3688 ext4_blocks_count(es));
e7c95593
ES		 3689 goto failed_mount;
3690 }
bd81d8ee
LV		 3691 blocks_count = (ext4_blocks_count(es) -
3692 le32_to_cpu(es->s_first_data_block) +
3693 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3694 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4ec11028 3695 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
b31e1552 3696 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
4ec11028 3697 "(block count %llu, first data block %u, "
b31e1552 3698 "blocks per group %lu)", sbi->s_groups_count,
4ec11028
TT		 3699 ext4_blocks_count(es),
3700 le32_to_cpu(es->s_first_data_block),
3701 EXT4_BLOCKS_PER_GROUP(sb));
3702 goto failed_mount;
3703 }
bd81d8ee 3704 sbi->s_groups_count = blocks_count;
fb0a387d
ES		 3705 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3706 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
617ba13b
MC		 3707 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3708 EXT4_DESC_PER_BLOCK(sb);
f18a5f21
TT		 3709 sbi->s_group_desc = ext4_kvmalloc(db_count *
3710 sizeof(struct buffer_head *),
3711 GFP_KERNEL);
ac27a0ec 3712 if (sbi->s_group_desc == NULL) {
b31e1552 3713 ext4_msg(sb, KERN_ERR, "not enough memory");
2cde417d 3714 ret = -ENOMEM;
ac27a0ec
DK		 3715 goto failed_mount;
3716 }
3717
9f6200bb
TT		 3718 if (ext4_proc_root)
3719 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
240799cd 3720
66acdcf4
TT		 3721 if (sbi->s_proc)
3722 proc_create_data("options", S_IRUGO, sbi->s_proc,
3723 &ext4_seq_options_fops, sb);
3724
705895b6 3725 bgl_lock_init(sbi->s_blockgroup_lock);
ac27a0ec
DK		 3726
3727 for (i = 0; i < db_count; i++) {
70bbb3e0 3728 block = descriptor_loc(sb, logical_sb_block, i);
ac27a0ec
DK		 3729 sbi->s_group_desc[i] = sb_bread(sb, block);
3730 if (!sbi->s_group_desc[i]) {
b31e1552
ES		 3731 ext4_msg(sb, KERN_ERR,
3732 "can't read group descriptor %d", i);
ac27a0ec
DK		 3733 db_count = i;
3734 goto failed_mount2;
3735 }
3736 }
bfff6873 3737 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
b31e1552 3738 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
ac27a0ec
DK		 3739 goto failed_mount2;
3740 }
772cb7c8
JS		 3741 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3742 if (!ext4_fill_flex_info(sb)) {
b31e1552
ES		 3743 ext4_msg(sb, KERN_ERR,
3744 "unable to initialize "
3745 "flex_bg meta info!");
772cb7c8
JS		 3746 goto failed_mount2;
3747 }
3748
ac27a0ec
DK		 3749 sbi->s_gdb_count = db_count;
3750 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3751 spin_lock_init(&sbi->s_next_gen_lock);
3752
04496411
TM		 3753 init_timer(&sbi->s_err_report);
3754 sbi->s_err_report.function = print_daily_error_info;
3755 sbi->s_err_report.data = (unsigned long) sb;
3756
57042651 3757 err = percpu_counter_init(&sbi->s_freeclusters_counter,
5dee5437 3758 ext4_count_free_clusters(sb));
ce7e010a
TT		 3759 if (!err) {
3760 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3761 ext4_count_free_inodes(sb));
3762 }
3763 if (!err) {
3764 err = percpu_counter_init(&sbi->s_dirs_counter,
3765 ext4_count_dirs(sb));
3766 }
3767 if (!err) {
57042651 3768 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0);
ce7e010a
TT		 3769 }
3770 if (err) {
3771 ext4_msg(sb, KERN_ERR, "insufficient memory");
3772 goto failed_mount3;
3773 }
3774
c9de560d 3775 sbi->s_stripe = ext4_get_stripe_size(sbi);
55138e0b 3776 sbi->s_max_writeback_mb_bump = 128;
67a5da56 3777 sbi->s_extent_max_zeroout_kb = 32;
c9de560d 3778
ac27a0ec
DK		 3779 /*
3780 * set up enough so that it can read an inode
3781 */
9ca92389
TT		 3782 if (!test_opt(sb, NOLOAD) &&
3783 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3784 sb->s_op = &ext4_sops;
3785 else
3786 sb->s_op = &ext4_nojournal_sops;
617ba13b
MC		 3787 sb->s_export_op = &ext4_export_ops;
3788 sb->s_xattr = ext4_xattr_handlers;
ac27a0ec 3789#ifdef CONFIG_QUOTA
617ba13b
MC		 3790 sb->s_qcop = &ext4_qctl_operations;
3791 sb->dq_op = &ext4_quota_operations;
7c319d32
AK		 3792
3793 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) {
3794 /* Use qctl operations for hidden quota files. */
3795 sb->s_qcop = &ext4_qctl_sysfile_operations;
3796 }
ac27a0ec 3797#endif
f2fa2ffc
AK		 3798 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3799
ac27a0ec 3800 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3b9d4ed2 3801 mutex_init(&sbi->s_orphan_lock);
8f82f840 3802 sbi->s_resize_flags = 0;
ac27a0ec
DK		 3803
3804 sb->s_root = NULL;
3805
3806 needs_recovery = (es->s_last_orphan != 0 ||
617ba13b
MC		 3807 EXT4_HAS_INCOMPAT_FEATURE(sb,
3808 EXT4_FEATURE_INCOMPAT_RECOVER));
ac27a0ec 3809
c5e06d10
JL		 3810 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3811 !(sb->s_flags & MS_RDONLY))
3812 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3813 goto failed_mount3;
3814
ac27a0ec
DK		 3815 /*
3816 * The first inode we look at is the journal inode. Don't try
3817 * root first: it may be modified in the journal!
3818 */
3819 if (!test_opt(sb, NOLOAD) &&
617ba13b
MC		 3820 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3821 if (ext4_load_journal(sb, es, journal_devnum))
ac27a0ec 3822 goto failed_mount3;
0390131b
FM		 3823 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3824 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
b31e1552
ES		 3825 ext4_msg(sb, KERN_ERR, "required journal recovery "
3826 "suppressed and not mounted read-only");
744692dc 3827 goto failed_mount_wq;
ac27a0ec 3828 } else {
fd8c37ec 3829 clear_opt(sb, DATA_FLAGS);
0390131b
FM		 3830 sbi->s_journal = NULL;
3831 needs_recovery = 0;
3832 goto no_journal;
ac27a0ec
DK		 3833 }
3834
f32aaf2d 3835 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT) &&
eb40a09c
JS		 3836 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3837 JBD2_FEATURE_INCOMPAT_64BIT)) {
b31e1552 3838 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
744692dc 3839 goto failed_mount_wq;
eb40a09c
JS		 3840 }
3841
25ed6e8a
DW		 3842 if (!set_journal_csum_feature_set(sb)) {
3843 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
3844 "feature set");
3845 goto failed_mount_wq;
d4da6c9c 3846 }
818d276c 3847
ac27a0ec
DK		 3848 /* We have now updated the journal if required, so we can
3849 * validate the data journaling mode. */
3850 switch (test_opt(sb, DATA_FLAGS)) {
3851 case 0:
3852 /* No mode set, assume a default based on the journal
63f57933
AM		 3853 * capabilities: ORDERED_DATA if the journal can
3854 * cope, else JOURNAL_DATA
3855 */
dab291af
MC		 3856 if (jbd2_journal_check_available_features
3857 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
fd8c37ec 3858 set_opt(sb, ORDERED_DATA);
ac27a0ec 3859 else
fd8c37ec 3860 set_opt(sb, JOURNAL_DATA);
ac27a0ec
DK		 3861 break;
3862
617ba13b
MC		 3863 case EXT4_MOUNT_ORDERED_DATA:
3864 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 3865 if (!jbd2_journal_check_available_features
3866 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
b31e1552
ES		 3867 ext4_msg(sb, KERN_ERR, "Journal does not support "
3868 "requested data journaling mode");
744692dc 3869 goto failed_mount_wq;
ac27a0ec
DK		 3870 }
3871 default:
3872 break;
3873 }
b3881f74 3874 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
ac27a0ec 3875
18aadd47
BJ		 3876 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
3877
ce7e010a
TT		 3878 /*
3879 * The journal may have updated the bg summary counts, so we
3880 * need to update the global counters.
3881 */
57042651 3882 percpu_counter_set(&sbi->s_freeclusters_counter,
5dee5437 3883 ext4_count_free_clusters(sb));
ce7e010a
TT		 3884 percpu_counter_set(&sbi->s_freeinodes_counter,
3885 ext4_count_free_inodes(sb));
3886 percpu_counter_set(&sbi->s_dirs_counter,
3887 ext4_count_dirs(sb));
57042651 3888 percpu_counter_set(&sbi->s_dirtyclusters_counter, 0);
206f7ab4 3889
ce7e010a 3890no_journal:
952fc18e
TT		 3891 /*
3892 * Get the # of file system overhead blocks from the
3893 * superblock if present.
3894 */
3895 if (es->s_overhead_clusters)
3896 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
3897 else {
07aa2ea1
LC		 3898 err = ext4_calculate_overhead(sb);
3899 if (err)
952fc18e
TT		 3900 goto failed_mount_wq;
3901 }
3902
fd89d5f2
TH		 3903 /*
3904 * The maximum number of concurrent works can be high and
3905 * concurrency isn't really necessary. Limit it to 1.
3906 */
3907 EXT4_SB(sb)->dio_unwritten_wq =
ae005cbe 3908 alloc_workqueue("ext4-dio-unwritten", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4c0425ff
MC		 3909 if (!EXT4_SB(sb)->dio_unwritten_wq) {
3910 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
07aa2ea1 3911 ret = -ENOMEM;
4c0425ff
MC		 3912 goto failed_mount_wq;
3913 }
3914
ac27a0ec 3915 /*
dab291af 3916 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 3917 * so we can safely mount the rest of the filesystem now.
3918 */
3919
1d1fe1ee
DH		 3920 root = ext4_iget(sb, EXT4_ROOT_INO);
3921 if (IS_ERR(root)) {
b31e1552 3922 ext4_msg(sb, KERN_ERR, "get root inode failed");
1d1fe1ee 3923 ret = PTR_ERR(root);
32a9bb57 3924 root = NULL;
ac27a0ec
DK		 3925 goto failed_mount4;
3926 }
3927 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
b31e1552 3928 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
94bf608a 3929 iput(root);
ac27a0ec
DK		 3930 goto failed_mount4;
3931 }
48fde701 3932 sb->s_root = d_make_root(root);
1d1fe1ee 3933 if (!sb->s_root) {
b31e1552 3934 ext4_msg(sb, KERN_ERR, "get root dentry failed");
1d1fe1ee
DH		 3935 ret = -ENOMEM;
3936 goto failed_mount4;
3937 }
ac27a0ec 3938
7e84b621
ES		 3939 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
3940 sb->s_flags |= MS_RDONLY;
ef7f3835
KS		 3941
3942 /* determine the minimum size of new large inodes, if present */
3943 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3944 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3945 EXT4_GOOD_OLD_INODE_SIZE;
3946 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3947 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
3948 if (sbi->s_want_extra_isize <
3949 le16_to_cpu(es->s_want_extra_isize))
3950 sbi->s_want_extra_isize =
3951 le16_to_cpu(es->s_want_extra_isize);
3952 if (sbi->s_want_extra_isize <
3953 le16_to_cpu(es->s_min_extra_isize))
3954 sbi->s_want_extra_isize =
3955 le16_to_cpu(es->s_min_extra_isize);
3956 }
3957 }
3958 /* Check if enough inode space is available */
3959 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3960 sbi->s_inode_size) {
3961 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3962 EXT4_GOOD_OLD_INODE_SIZE;
b31e1552
ES		 3963 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3964 "available");
ef7f3835
KS		 3965 }
3966
6fd058f7
TT		 3967 err = ext4_setup_system_zone(sb);
3968 if (err) {
b31e1552 3969 ext4_msg(sb, KERN_ERR, "failed to initialize system "
fbe845dd 3970 "zone (%d)", err);
94bf608a 3971 goto failed_mount4a;
6fd058f7
TT		 3972 }
3973
c2774d84 3974 ext4_ext_init(sb);
9d99012f 3975 err = ext4_mb_init(sb);
c2774d84 3976 if (err) {
421f91d2 3977 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
b31e1552 3978 err);
dcf2d804 3979 goto failed_mount5;
c2774d84
AK		 3980 }
3981
bfff6873
LC		 3982 err = ext4_register_li_request(sb, first_not_zeroed);
3983 if (err)
dcf2d804 3984 goto failed_mount6;
bfff6873 3985
3197ebdb
TT		 3986 sbi->s_kobj.kset = ext4_kset;
3987 init_completion(&sbi->s_kobj_unregister);
3988 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
3989 "%s", sb->s_id);
dcf2d804
TM		 3990 if (err)
3991 goto failed_mount7;
3197ebdb 3992
617ba13b
MC		 3993 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3994 ext4_orphan_cleanup(sb, es);
3995 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
0390131b 3996 if (needs_recovery) {
b31e1552 3997 ext4_msg(sb, KERN_INFO, "recovery complete");
0390131b
FM		 3998 ext4_mark_recovery_complete(sb, es);
3999 }
4000 if (EXT4_SB(sb)->s_journal) {
4001 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4002 descr = " journalled data mode";
4003 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4004 descr = " ordered data mode";
4005 else
4006 descr = " writeback data mode";
4007 } else
4008 descr = "out journal";
4009
7c319d32
AK		 4010#ifdef CONFIG_QUOTA
4011 /* Enable quota usage during mount. */
4012 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
4013 !(sb->s_flags & MS_RDONLY)) {
07aa2ea1
LC		 4014 err = ext4_enable_quotas(sb);
4015 if (err)
7c319d32
AK		 4016 goto failed_mount7;
4017 }
4018#endif /* CONFIG_QUOTA */
4019
79add3a3
LC		 4020 if (test_opt(sb, DISCARD)) {
4021 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4022 if (!blk_queue_discard(q))
4023 ext4_msg(sb, KERN_WARNING,
4024 "mounting with \"discard\" option, but "
4025 "the device does not support discard");
4026 }
4027
d4c402d9 4028 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
8b67f04a
TT		 4029 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
4030 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
ac27a0ec 4031
66e61a9e
TT		 4032 if (es->s_error_count)
4033 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
ac27a0ec 4034
d4c402d9 4035 kfree(orig_data);
ac27a0ec
DK		 4036 return 0;
4037
617ba13b 4038cantfind_ext4:
ac27a0ec 4039 if (!silent)
b31e1552 4040 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
ac27a0ec
DK		 4041 goto failed_mount;
4042
dcf2d804
TM		 4043failed_mount7:
4044 ext4_unregister_li_request(sb);
4045failed_mount6:
dcf2d804 4046 ext4_mb_release(sb);
94bf608a
AV		 4047failed_mount5:
4048 ext4_ext_release(sb);
dcf2d804 4049 ext4_release_system_zone(sb);
94bf608a
AV		 4050failed_mount4a:
4051 dput(sb->s_root);
32a9bb57 4052 sb->s_root = NULL;
94bf608a 4053failed_mount4:
b31e1552 4054 ext4_msg(sb, KERN_ERR, "mount failed");
4c0425ff
MC		 4055 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
4056failed_mount_wq:
0390131b
FM		 4057 if (sbi->s_journal) {
4058 jbd2_journal_destroy(sbi->s_journal);
4059 sbi->s_journal = NULL;
4060 }
ac27a0ec 4061failed_mount3:
04496411 4062 del_timer(&sbi->s_err_report);
9933fc0a
TT		 4063 if (sbi->s_flex_groups)
4064 ext4_kvfree(sbi->s_flex_groups);
57042651 4065 percpu_counter_destroy(&sbi->s_freeclusters_counter);
ce7e010a
TT		 4066 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4067 percpu_counter_destroy(&sbi->s_dirs_counter);
57042651 4068 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
c5e06d10
JL		 4069 if (sbi->s_mmp_tsk)
4070 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec
DK		 4071failed_mount2:
4072 for (i = 0; i < db_count; i++)
4073 brelse(sbi->s_group_desc[i]);
f18a5f21 4074 ext4_kvfree(sbi->s_group_desc);
ac27a0ec 4075failed_mount:
0441984a
DW		 4076 if (sbi->s_chksum_driver)
4077 crypto_free_shash(sbi->s_chksum_driver);
240799cd 4078 if (sbi->s_proc) {
66acdcf4 4079 remove_proc_entry("options", sbi->s_proc);
9f6200bb 4080 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 4081 }
ac27a0ec
DK		 4082#ifdef CONFIG_QUOTA
4083 for (i = 0; i < MAXQUOTAS; i++)
4084 kfree(sbi->s_qf_names[i]);
4085#endif
617ba13b 4086 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 4087 brelse(bh);
4088out_fail:
4089 sb->s_fs_info = NULL;
f6830165 4090 kfree(sbi->s_blockgroup_lock);
ac27a0ec 4091 kfree(sbi);
dcc7dae3 4092out_free_orig:
d4c402d9 4093 kfree(orig_data);
07aa2ea1 4094 return err ? err : ret;
ac27a0ec
DK		 4095}
4096
4097/*
4098 * Setup any per-fs journal parameters now. We'll do this both on
4099 * initial mount, once the journal has been initialised but before we've
4100 * done any recovery; and again on any subsequent remount.
4101 */
617ba13b 4102static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 4103{
617ba13b 4104 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4105
30773840
TT		 4106 journal->j_commit_interval = sbi->s_commit_interval;
4107 journal->j_min_batch_time = sbi->s_min_batch_time;
4108 journal->j_max_batch_time = sbi->s_max_batch_time;
ac27a0ec 4109
a931da6a 4110 write_lock(&journal->j_state_lock);
ac27a0ec 4111 if (test_opt(sb, BARRIER))
dab291af 4112 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 4113 else
dab291af 4114 journal->j_flags &= ~JBD2_BARRIER;
5bf5683a
HK		 4115 if (test_opt(sb, DATA_ERR_ABORT))
4116 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4117 else
4118 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
a931da6a 4119 write_unlock(&journal->j_state_lock);
ac27a0ec
DK		 4120}
4121
617ba13b 4122static journal_t *ext4_get_journal(struct super_block *sb,
ac27a0ec
DK		 4123 unsigned int journal_inum)
4124{
4125 struct inode *journal_inode;
4126 journal_t *journal;
4127
0390131b
FM		 4128 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4129
ac27a0ec
DK		 4130 /* First, test for the existence of a valid inode on disk. Bad
4131 * things happen if we iget() an unused inode, as the subsequent
4132 * iput() will try to delete it. */
4133
1d1fe1ee
DH		 4134 journal_inode = ext4_iget(sb, journal_inum);
4135 if (IS_ERR(journal_inode)) {
b31e1552 4136 ext4_msg(sb, KERN_ERR, "no journal found");
ac27a0ec
DK		 4137 return NULL;
4138 }
4139 if (!journal_inode->i_nlink) {
4140 make_bad_inode(journal_inode);
4141 iput(journal_inode);
b31e1552 4142 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
ac27a0ec
DK		 4143 return NULL;
4144 }
4145
e5f8eab8 4146 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
ac27a0ec 4147 journal_inode, journal_inode->i_size);
1d1fe1ee 4148 if (!S_ISREG(journal_inode->i_mode)) {
b31e1552 4149 ext4_msg(sb, KERN_ERR, "invalid journal inode");
ac27a0ec
DK		 4150 iput(journal_inode);
4151 return NULL;
4152 }
4153
dab291af 4154 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 4155 if (!journal) {
b31e1552 4156 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
ac27a0ec
DK		 4157 iput(journal_inode);
4158 return NULL;
4159 }
4160 journal->j_private = sb;
617ba13b 4161 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 4162 return journal;
4163}
4164
617ba13b 4165static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 4166 dev_t j_dev)
4167{
2b2d6d01 4168 struct buffer_head *bh;
ac27a0ec 4169 journal_t *journal;
617ba13b
MC		 4170 ext4_fsblk_t start;
4171 ext4_fsblk_t len;
ac27a0ec 4172 int hblock, blocksize;
617ba13b 4173 ext4_fsblk_t sb_block;
ac27a0ec 4174 unsigned long offset;
2b2d6d01 4175 struct ext4_super_block *es;
ac27a0ec
DK		 4176 struct block_device *bdev;
4177
0390131b
FM		 4178 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4179
b31e1552 4180 bdev = ext4_blkdev_get(j_dev, sb);
ac27a0ec
DK		 4181 if (bdev == NULL)
4182 return NULL;
4183
ac27a0ec 4184 blocksize = sb->s_blocksize;
e1defc4f 4185 hblock = bdev_logical_block_size(bdev);
ac27a0ec 4186 if (blocksize < hblock) {
b31e1552
ES		 4187 ext4_msg(sb, KERN_ERR,
4188 "blocksize too small for journal device");
ac27a0ec
DK		 4189 goto out_bdev;
4190 }
4191
617ba13b
MC		 4192 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4193 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 4194 set_blocksize(bdev, blocksize);
4195 if (!(bh = __bread(bdev, sb_block, blocksize))) {
b31e1552
ES		 4196 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4197 "external journal");
ac27a0ec
DK		 4198 goto out_bdev;
4199 }
4200
2716b802 4201 es = (struct ext4_super_block *) (bh->b_data + offset);
617ba13b 4202 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 4203 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b 4204 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
b31e1552
ES		 4205 ext4_msg(sb, KERN_ERR, "external journal has "
4206 "bad superblock");
ac27a0ec
DK		 4207 brelse(bh);
4208 goto out_bdev;
4209 }
4210
617ba13b 4211 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
b31e1552 4212 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
ac27a0ec
DK		 4213 brelse(bh);
4214 goto out_bdev;
4215 }
4216
bd81d8ee 4217 len = ext4_blocks_count(es);
ac27a0ec
DK		 4218 start = sb_block + 1;
4219 brelse(bh); /* we're done with the superblock */
4220
dab291af 4221 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 4222 start, len, blocksize);
4223 if (!journal) {
b31e1552 4224 ext4_msg(sb, KERN_ERR, "failed to create device journal");
ac27a0ec
DK		 4225 goto out_bdev;
4226 }
4227 journal->j_private = sb;
4228 ll_rw_block(READ, 1, &journal->j_sb_buffer);
4229 wait_on_buffer(journal->j_sb_buffer);
4230 if (!buffer_uptodate(journal->j_sb_buffer)) {
b31e1552 4231 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
ac27a0ec
DK		 4232 goto out_journal;
4233 }
4234 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
b31e1552
ES		 4235 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4236 "user (unsupported) - %d",
ac27a0ec
DK		 4237 be32_to_cpu(journal->j_superblock->s_nr_users));
4238 goto out_journal;
4239 }
617ba13b
MC		 4240 EXT4_SB(sb)->journal_bdev = bdev;
4241 ext4_init_journal_params(sb, journal);
ac27a0ec 4242 return journal;
0b8e58a1 4243
ac27a0ec 4244out_journal:
dab291af 4245 jbd2_journal_destroy(journal);
ac27a0ec 4246out_bdev:
617ba13b 4247 ext4_blkdev_put(bdev);
ac27a0ec
DK		 4248 return NULL;
4249}
4250
617ba13b
MC		 4251static int ext4_load_journal(struct super_block *sb,
4252 struct ext4_super_block *es,
ac27a0ec
DK		 4253 unsigned long journal_devnum)
4254{
4255 journal_t *journal;
4256 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4257 dev_t journal_dev;
4258 int err = 0;
4259 int really_read_only;
4260
0390131b
FM		 4261 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4262
ac27a0ec
DK		 4263 if (journal_devnum &&
4264 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
b31e1552
ES		 4265 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4266 "numbers have changed");
ac27a0ec
DK		 4267 journal_dev = new_decode_dev(journal_devnum);
4268 } else
4269 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4270
4271 really_read_only = bdev_read_only(sb->s_bdev);
4272
4273 /*
4274 * Are we loading a blank journal or performing recovery after a
4275 * crash? For recovery, we need to check in advance whether we
4276 * can get read-write access to the device.
4277 */
617ba13b 4278 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
ac27a0ec 4279 if (sb->s_flags & MS_RDONLY) {
b31e1552
ES		 4280 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4281 "required on readonly filesystem");
ac27a0ec 4282 if (really_read_only) {
b31e1552
ES		 4283 ext4_msg(sb, KERN_ERR, "write access "
4284 "unavailable, cannot proceed");
ac27a0ec
DK		 4285 return -EROFS;
4286 }
b31e1552
ES		 4287 ext4_msg(sb, KERN_INFO, "write access will "
4288 "be enabled during recovery");
ac27a0ec
DK		 4289 }
4290 }
4291
4292 if (journal_inum && journal_dev) {
b31e1552
ES		 4293 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4294 "and inode journals!");
ac27a0ec
DK		 4295 return -EINVAL;
4296 }
4297
4298 if (journal_inum) {
617ba13b 4299 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 4300 return -EINVAL;
4301 } else {
617ba13b 4302 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 4303 return -EINVAL;
4304 }
4305
90576c0b 4306 if (!(journal->j_flags & JBD2_BARRIER))
b31e1552 4307 ext4_msg(sb, KERN_INFO, "barriers disabled");
4776004f 4308
617ba13b 4309 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
dab291af 4310 err = jbd2_journal_wipe(journal, !really_read_only);
1c13d5c0
TT		 4311 if (!err) {
4312 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4313 if (save)
4314 memcpy(save, ((char *) es) +
4315 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
dab291af 4316 err = jbd2_journal_load(journal);
1c13d5c0
TT		 4317 if (save)
4318 memcpy(((char *) es) + EXT4_S_ERR_START,
4319 save, EXT4_S_ERR_LEN);
4320 kfree(save);
4321 }
ac27a0ec
DK		 4322
4323 if (err) {
b31e1552 4324 ext4_msg(sb, KERN_ERR, "error loading journal");
dab291af 4325 jbd2_journal_destroy(journal);
ac27a0ec
DK		 4326 return err;
4327 }
4328
617ba13b
MC		 4329 EXT4_SB(sb)->s_journal = journal;
4330 ext4_clear_journal_err(sb, es);
ac27a0ec 4331
c41303ce 4332 if (!really_read_only && journal_devnum &&
ac27a0ec
DK		 4333 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4334 es->s_journal_dev = cpu_to_le32(journal_devnum);
ac27a0ec
DK		 4335
4336 /* Make sure we flush the recovery flag to disk. */
e2d67052 4337 ext4_commit_super(sb, 1);
ac27a0ec
DK		 4338 }
4339
4340 return 0;
4341}
4342
e2d67052 4343static int ext4_commit_super(struct super_block *sb, int sync)
ac27a0ec 4344{
e2d67052 4345 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
617ba13b 4346 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
c4be0c1d 4347 int error = 0;
ac27a0ec 4348
7c2e7087 4349 if (!sbh || block_device_ejected(sb))
c4be0c1d 4350 return error;
914258bf
TT		 4351 if (buffer_write_io_error(sbh)) {
4352 /*
4353 * Oh, dear. A previous attempt to write the
4354 * superblock failed. This could happen because the
4355 * USB device was yanked out. Or it could happen to
4356 * be a transient write error and maybe the block will
4357 * be remapped. Nothing we can do but to retry the
4358 * write and hope for the best.
4359 */
b31e1552
ES		 4360 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4361 "superblock detected");
914258bf
TT		 4362 clear_buffer_write_io_error(sbh);
4363 set_buffer_uptodate(sbh);
4364 }
71290b36
TT		 4365 /*
4366 * If the file system is mounted read-only, don't update the
4367 * superblock write time. This avoids updating the superblock
4368 * write time when we are mounting the root file system
4369 * read/only but we need to replay the journal; at that point,
4370 * for people who are east of GMT and who make their clock
4371 * tick in localtime for Windows bug-for-bug compatibility,
4372 * the clock is set in the future, and this will cause e2fsck
4373 * to complain and force a full file system check.
4374 */
4375 if (!(sb->s_flags & MS_RDONLY))
4376 es->s_wtime = cpu_to_le32(get_seconds());
f613dfcb
TT		 4377 if (sb->s_bdev->bd_part)
4378 es->s_kbytes_written =
4379 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
afc32f7e
TT		 4380 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4381 EXT4_SB(sb)->s_sectors_written_start) >> 1));
f613dfcb
TT		 4382 else
4383 es->s_kbytes_written =
4384 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
57042651
TT		 4385 ext4_free_blocks_count_set(es,
4386 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4387 &EXT4_SB(sb)->s_freeclusters_counter)));
ce7e010a
TT		 4388 es->s_free_inodes_count =
4389 cpu_to_le32(percpu_counter_sum_positive(
4390 &EXT4_SB(sb)->s_freeinodes_counter));
ac27a0ec 4391 BUFFER_TRACE(sbh, "marking dirty");
06db49e6 4392 ext4_superblock_csum_set(sb);
ac27a0ec 4393 mark_buffer_dirty(sbh);
914258bf 4394 if (sync) {
c4be0c1d
TS		 4395 error = sync_dirty_buffer(sbh);
4396 if (error)
4397 return error;
4398
4399 error = buffer_write_io_error(sbh);
4400 if (error) {
b31e1552
ES		 4401 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4402 "superblock");
914258bf
TT		 4403 clear_buffer_write_io_error(sbh);
4404 set_buffer_uptodate(sbh);
4405 }
4406 }
c4be0c1d 4407 return error;
ac27a0ec
DK		 4408}
4409
ac27a0ec
DK		 4410/*
4411 * Have we just finished recovery? If so, and if we are mounting (or
4412 * remounting) the filesystem readonly, then we will end up with a
4413 * consistent fs on disk. Record that fact.
4414 */
2b2d6d01
TT		 4415static void ext4_mark_recovery_complete(struct super_block *sb,
4416 struct ext4_super_block *es)
ac27a0ec 4417{
617ba13b 4418 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 4419
0390131b
FM		 4420 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4421 BUG_ON(journal != NULL);
4422 return;
4423 }
dab291af 4424 jbd2_journal_lock_updates(journal);
7ffe1ea8
HK		 4425 if (jbd2_journal_flush(journal) < 0)
4426 goto out;
4427
617ba13b 4428 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
ac27a0ec 4429 sb->s_flags & MS_RDONLY) {
617ba13b 4430 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
e2d67052 4431 ext4_commit_super(sb, 1);
ac27a0ec 4432 }
7ffe1ea8
HK		 4433
4434out:
dab291af 4435 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 4436}
4437
4438/*
4439 * If we are mounting (or read-write remounting) a filesystem whose journal
4440 * has recorded an error from a previous lifetime, move that error to the
4441 * main filesystem now.
4442 */
2b2d6d01
TT		 4443static void ext4_clear_journal_err(struct super_block *sb,
4444 struct ext4_super_block *es)
ac27a0ec
DK		 4445{
4446 journal_t *journal;
4447 int j_errno;
4448 const char *errstr;
4449
0390131b
FM		 4450 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4451
617ba13b 4452 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 4453
4454 /*
4455 * Now check for any error status which may have been recorded in the
617ba13b 4456 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 4457 */
4458
dab291af 4459 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 4460 if (j_errno) {
4461 char nbuf[16];
4462
617ba13b 4463 errstr = ext4_decode_error(sb, j_errno, nbuf);
12062ddd 4464 ext4_warning(sb, "Filesystem error recorded "
ac27a0ec 4465 "from previous mount: %s", errstr);
12062ddd 4466 ext4_warning(sb, "Marking fs in need of filesystem check.");
ac27a0ec 4467
617ba13b
MC		 4468 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4469 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
e2d67052 4470 ext4_commit_super(sb, 1);
ac27a0ec 4471
dab291af 4472 jbd2_journal_clear_err(journal);
d796c52e 4473 jbd2_journal_update_sb_errno(journal);
ac27a0ec
DK		 4474 }
4475}
4476
4477/*
4478 * Force the running and committing transactions to commit,
4479 * and wait on the commit.
4480 */
617ba13b 4481int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 4482{
4483 journal_t *journal;
0390131b 4484 int ret = 0;
ac27a0ec
DK		 4485
4486 if (sb->s_flags & MS_RDONLY)
4487 return 0;
4488
617ba13b 4489 journal = EXT4_SB(sb)->s_journal;
8e8ad8a5 4490 if (journal)
0390131b 4491 ret = ext4_journal_force_commit(journal);
0390131b 4492
ac27a0ec
DK		 4493 return ret;
4494}
4495
617ba13b 4496static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec 4497{
14ce0cb4 4498 int ret = 0;
9eddacf9 4499 tid_t target;
8d5d02e6 4500 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4501
9bffad1e 4502 trace_ext4_sync_fs(sb, wait);
8d5d02e6 4503 flush_workqueue(sbi->dio_unwritten_wq);
a1177825
JK		 4504 /*
4505 * Writeback quota in non-journalled quota case - journalled quota has
4506 * no dirty dquots
4507 */
4508 dquot_writeback_dquots(sb, -1);
8d5d02e6 4509 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
9ca92389 4510 if (wait)
8d5d02e6 4511 jbd2_log_wait_commit(sbi->s_journal, target);
0390131b 4512 }
14ce0cb4 4513 return ret;
ac27a0ec
DK		 4514}
4515
4516/*
4517 * LVM calls this function before a (read-only) snapshot is created. This
4518 * gives us a chance to flush the journal completely and mark the fs clean.
be4f27d3
YY		 4519 *
4520 * Note that only this function cannot bring a filesystem to be in a clean
8e8ad8a5
JK		 4521 * state independently. It relies on upper layer to stop all data & metadata
4522 * modifications.
ac27a0ec 4523 */
c4be0c1d 4524static int ext4_freeze(struct super_block *sb)
ac27a0ec 4525{
c4be0c1d
TS		 4526 int error = 0;
4527 journal_t *journal;
ac27a0ec 4528
9ca92389
TT		 4529 if (sb->s_flags & MS_RDONLY)
4530 return 0;
ac27a0ec 4531
9ca92389 4532 journal = EXT4_SB(sb)->s_journal;
7ffe1ea8 4533
9ca92389
TT		 4534 /* Now we set up the journal barrier. */
4535 jbd2_journal_lock_updates(journal);
ac27a0ec 4536
9ca92389
TT		 4537 /*
4538 * Don't clear the needs_recovery flag if we failed to flush
4539 * the journal.
4540 */
4541 error = jbd2_journal_flush(journal);
6b0310fb
ES		 4542 if (error < 0)
4543 goto out;
9ca92389
TT		 4544
4545 /* Journal blocked and flushed, clear needs_recovery flag. */
4546 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4547 error = ext4_commit_super(sb, 1);
6b0310fb 4548out:
8e8ad8a5 4549 /* we rely on upper layer to stop further updates */
6b0310fb
ES		 4550 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4551 return error;
ac27a0ec
DK		 4552}
4553
4554/*
4555 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4556 * flag here, even though the filesystem is not technically dirty yet.
4557 */
c4be0c1d 4558static int ext4_unfreeze(struct super_block *sb)
ac27a0ec 4559{
9ca92389
TT		 4560 if (sb->s_flags & MS_RDONLY)
4561 return 0;
4562
9ca92389
TT		 4563 /* Reset the needs_recovery flag before the fs is unlocked. */
4564 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4565 ext4_commit_super(sb, 1);
c4be0c1d 4566 return 0;
ac27a0ec
DK		 4567}
4568
673c6100
TT		 4569/*
4570 * Structure to save mount options for ext4_remount's benefit
4571 */
4572struct ext4_mount_options {
4573 unsigned long s_mount_opt;
a2595b8a 4574 unsigned long s_mount_opt2;
08cefc7a
EB		 4575 kuid_t s_resuid;
4576 kgid_t s_resgid;
673c6100
TT		 4577 unsigned long s_commit_interval;
4578 u32 s_min_batch_time, s_max_batch_time;
4579#ifdef CONFIG_QUOTA
4580 int s_jquota_fmt;
4581 char *s_qf_names[MAXQUOTAS];
4582#endif
4583};
4584
2b2d6d01 4585static int ext4_remount(struct super_block *sb, int *flags, char *data)
ac27a0ec 4586{
2b2d6d01 4587 struct ext4_super_block *es;
617ba13b 4588 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4589 unsigned long old_sb_flags;
617ba13b 4590 struct ext4_mount_options old_opts;
c79d967d 4591 int enable_quota = 0;
8a266467 4592 ext4_group_t g;
b3881f74 4593 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
c5e06d10 4594 int err = 0;
ac27a0ec
DK		 4595#ifdef CONFIG_QUOTA
4596 int i;
4597#endif
d4c402d9 4598 char *orig_data = kstrdup(data, GFP_KERNEL);
ac27a0ec
DK		 4599
4600 /* Store the original options */
4601 old_sb_flags = sb->s_flags;
4602 old_opts.s_mount_opt = sbi->s_mount_opt;
a2595b8a 4603 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
ac27a0ec
DK		 4604 old_opts.s_resuid = sbi->s_resuid;
4605 old_opts.s_resgid = sbi->s_resgid;
4606 old_opts.s_commit_interval = sbi->s_commit_interval;
30773840
TT		 4607 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4608 old_opts.s_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 4609#ifdef CONFIG_QUOTA
4610 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4611 for (i = 0; i < MAXQUOTAS; i++)
4612 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
4613#endif
b3881f74
TT		 4614 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4615 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
ac27a0ec
DK		 4616
4617 /*
4618 * Allow the "check" option to be passed as a remount option.
4619 */
661aa520 4620 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
ac27a0ec
DK		 4621 err = -EINVAL;
4622 goto restore_opts;
4623 }
4624
4ab2f15b 4625 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
c67d859e 4626 ext4_abort(sb, "Abort forced by user");
ac27a0ec
DK		 4627
4628 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 4629 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec
DK		 4630
4631 es = sbi->s_es;
4632
b3881f74 4633 if (sbi->s_journal) {
0390131b 4634 ext4_init_journal_params(sb, sbi->s_journal);
b3881f74
TT		 4635 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4636 }
ac27a0ec 4637
661aa520 4638 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4ab2f15b 4639 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
ac27a0ec
DK		 4640 err = -EROFS;
4641 goto restore_opts;
4642 }
4643
4644 if (*flags & MS_RDONLY) {
0f0dd62f
CH		 4645 err = dquot_suspend(sb, -1);
4646 if (err < 0)
c79d967d 4647 goto restore_opts;
c79d967d 4648
ac27a0ec
DK		 4649 /*
4650 * First of all, the unconditional stuff we have to do
4651 * to disable replay of the journal when we next remount
4652 */
4653 sb->s_flags |= MS_RDONLY;
4654
4655 /*
4656 * OK, test if we are remounting a valid rw partition
4657 * readonly, and if so set the rdonly flag and then
4658 * mark the partition as valid again.
4659 */
617ba13b
MC		 4660 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4661 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 4662 es->s_state = cpu_to_le16(sbi->s_mount_state);
4663
a63c9eb2 4664 if (sbi->s_journal)
0390131b 4665 ext4_mark_recovery_complete(sb, es);
ac27a0ec 4666 } else {
a13fb1a4
ES		 4667 /* Make sure we can mount this feature set readwrite */
4668 if (!ext4_feature_set_ok(sb, 0)) {
ac27a0ec
DK		 4669 err = -EROFS;
4670 goto restore_opts;
4671 }
8a266467
TT		 4672 /*
4673 * Make sure the group descriptor checksums
0b8e58a1 4674 * are sane. If they aren't, refuse to remount r/w.
8a266467
TT		 4675 */
4676 for (g = 0; g < sbi->s_groups_count; g++) {
4677 struct ext4_group_desc *gdp =
4678 ext4_get_group_desc(sb, g, NULL);
4679
feb0ab32 4680 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
b31e1552
ES		 4681 ext4_msg(sb, KERN_ERR,
4682 "ext4_remount: Checksum for group %u failed (%u!=%u)",
8a266467
TT		 4683 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4684 le16_to_cpu(gdp->bg_checksum));
4685 err = -EINVAL;
4686 goto restore_opts;
4687 }
4688 }
4689
ead6596b
ES		 4690 /*
4691 * If we have an unprocessed orphan list hanging
4692 * around from a previously readonly bdev mount,
4693 * require a full umount/remount for now.
4694 */
4695 if (es->s_last_orphan) {
b31e1552 4696 ext4_msg(sb, KERN_WARNING, "Couldn't "
ead6596b
ES		 4697 "remount RDWR because of unprocessed "
4698 "orphan inode list. Please "
b31e1552 4699 "umount/remount instead");
ead6596b
ES		 4700 err = -EINVAL;
4701 goto restore_opts;
4702 }
4703
ac27a0ec
DK		 4704 /*
4705 * Mounting a RDONLY partition read-write, so reread
4706 * and store the current valid flag. (It may have
4707 * been changed by e2fsck since we originally mounted
4708 * the partition.)
4709 */
0390131b
FM		 4710 if (sbi->s_journal)
4711 ext4_clear_journal_err(sb, es);
ac27a0ec 4712 sbi->s_mount_state = le16_to_cpu(es->s_state);
2b2d6d01 4713 if (!ext4_setup_super(sb, es, 0))
ac27a0ec 4714 sb->s_flags &= ~MS_RDONLY;
c5e06d10
JL		 4715 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
4716 EXT4_FEATURE_INCOMPAT_MMP))
4717 if (ext4_multi_mount_protect(sb,
4718 le64_to_cpu(es->s_mmp_block))) {
4719 err = -EROFS;
4720 goto restore_opts;
4721 }
c79d967d 4722 enable_quota = 1;
ac27a0ec
DK		 4723 }
4724 }
bfff6873
LC		 4725
4726 /*
4727 * Reinitialize lazy itable initialization thread based on
4728 * current settings
4729 */
4730 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4731 ext4_unregister_li_request(sb);
4732 else {
4733 ext4_group_t first_not_zeroed;
4734 first_not_zeroed = ext4_has_uninit_itable(sb);
4735 ext4_register_li_request(sb, first_not_zeroed);
4736 }
4737
6fd058f7 4738 ext4_setup_system_zone(sb);
0390131b 4739 if (sbi->s_journal == NULL)
e2d67052 4740 ext4_commit_super(sb, 1);
0390131b 4741
ac27a0ec
DK		 4742#ifdef CONFIG_QUOTA
4743 /* Release old quota file names */
4744 for (i = 0; i < MAXQUOTAS; i++)
4745 if (old_opts.s_qf_names[i] &&
4746 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4747 kfree(old_opts.s_qf_names[i]);
7c319d32
AK		 4748 if (enable_quota) {
4749 if (sb_any_quota_suspended(sb))
4750 dquot_resume(sb, -1);
4751 else if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4752 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
4753 err = ext4_enable_quotas(sb);
07724f98 4754 if (err)
7c319d32 4755 goto restore_opts;
7c319d32
AK		 4756 }
4757 }
ac27a0ec 4758#endif
d4c402d9
CW		 4759
4760 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4761 kfree(orig_data);
ac27a0ec 4762 return 0;
0b8e58a1 4763
ac27a0ec
DK		 4764restore_opts:
4765 sb->s_flags = old_sb_flags;
4766 sbi->s_mount_opt = old_opts.s_mount_opt;
a2595b8a 4767 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
ac27a0ec
DK		 4768 sbi->s_resuid = old_opts.s_resuid;
4769 sbi->s_resgid = old_opts.s_resgid;
4770 sbi->s_commit_interval = old_opts.s_commit_interval;
30773840
TT		 4771 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4772 sbi->s_max_batch_time = old_opts.s_max_batch_time;
ac27a0ec
DK		 4773#ifdef CONFIG_QUOTA
4774 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4775 for (i = 0; i < MAXQUOTAS; i++) {
4776 if (sbi->s_qf_names[i] &&
4777 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4778 kfree(sbi->s_qf_names[i]);
4779 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4780 }
4781#endif
d4c402d9 4782 kfree(orig_data);
ac27a0ec
DK		 4783 return err;
4784}
4785
2b2d6d01 4786static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
ac27a0ec
DK		 4787{
4788 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 4789 struct ext4_sb_info *sbi = EXT4_SB(sb);
4790 struct ext4_super_block *es = sbi->s_es;
952fc18e 4791 ext4_fsblk_t overhead = 0;
960cc398 4792 u64 fsid;
d02a9391 4793 s64 bfree;
ac27a0ec 4794
952fc18e
TT		 4795 if (!test_opt(sb, MINIX_DF))
4796 overhead = sbi->s_overhead;
ac27a0ec 4797
617ba13b 4798 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 4799 buf->f_bsize = sb->s_blocksize;
b72f78cb 4800 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
57042651
TT		 4801 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
4802 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
d02a9391 4803 /* prevent underflow in case that few free space is available */
57042651 4804 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
bd81d8ee
LV		 4805 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
4806 if (buf->f_bfree < ext4_r_blocks_count(es))
ac27a0ec
DK		 4807 buf->f_bavail = 0;
4808 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 4809 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
617ba13b 4810 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 4811 fsid = le64_to_cpup((void *)es->s_uuid) ^
4812 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4813 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4814 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
0b8e58a1 4815
ac27a0ec
DK		 4816 return 0;
4817}
4818
0b8e58a1
AD		 4819/* Helper function for writing quotas on sync - we need to start transaction
4820 * before quota file is locked for write. Otherwise the are possible deadlocks:
ac27a0ec 4821 * Process 1 Process 2
617ba13b 4822 * ext4_create() quota_sync()
a269eb18 4823 * jbd2_journal_start() write_dquot()
871a2931 4824 * dquot_initialize() down(dqio_mutex)
dab291af 4825 * down(dqio_mutex) jbd2_journal_start()
ac27a0ec
DK		 4826 *
4827 */
4828
4829#ifdef CONFIG_QUOTA
4830
4831static inline struct inode *dquot_to_inode(struct dquot *dquot)
4832{
4c376dca 4833 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
ac27a0ec
DK		 4834}
4835
617ba13b 4836static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 4837{
4838 int ret, err;
4839 handle_t *handle;
4840 struct inode *inode;
4841
4842 inode = dquot_to_inode(dquot);
617ba13b 4843 handle = ext4_journal_start(inode,
0b8e58a1 4844 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 4845 if (IS_ERR(handle))
4846 return PTR_ERR(handle);
4847 ret = dquot_commit(dquot);
617ba13b 4848 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4849 if (!ret)
4850 ret = err;
4851 return ret;
4852}
4853
617ba13b 4854static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 4855{
4856 int ret, err;
4857 handle_t *handle;
4858
617ba13b 4859 handle = ext4_journal_start(dquot_to_inode(dquot),
0b8e58a1 4860 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 4861 if (IS_ERR(handle))
4862 return PTR_ERR(handle);
4863 ret = dquot_acquire(dquot);
617ba13b 4864 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4865 if (!ret)
4866 ret = err;
4867 return ret;
4868}
4869
617ba13b 4870static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 4871{
4872 int ret, err;
4873 handle_t *handle;
4874
617ba13b 4875 handle = ext4_journal_start(dquot_to_inode(dquot),
0b8e58a1 4876 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 4877 if (IS_ERR(handle)) {
4878 /* Release dquot anyway to avoid endless cycle in dqput() */
4879 dquot_release(dquot);
ac27a0ec 4880 return PTR_ERR(handle);
9c3013e9 4881 }
ac27a0ec 4882 ret = dquot_release(dquot);
617ba13b 4883 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4884 if (!ret)
4885 ret = err;
4886 return ret;
4887}
4888
617ba13b 4889static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec 4890{
2c8be6b2 4891 /* Are we journaling quotas? */
617ba13b
MC		 4892 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
4893 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
ac27a0ec 4894 dquot_mark_dquot_dirty(dquot);
617ba13b 4895 return ext4_write_dquot(dquot);
ac27a0ec
DK		 4896 } else {
4897 return dquot_mark_dquot_dirty(dquot);
4898 }
4899}
4900
617ba13b 4901static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 4902{
4903 int ret, err;
4904 handle_t *handle;
4905
4906 /* Data block + inode block */
617ba13b 4907 handle = ext4_journal_start(sb->s_root->d_inode, 2);
ac27a0ec
DK		 4908 if (IS_ERR(handle))
4909 return PTR_ERR(handle);
4910 ret = dquot_commit_info(sb, type);
617ba13b 4911 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4912 if (!ret)
4913 ret = err;
4914 return ret;
4915}
4916
4917/*
4918 * Turn on quotas during mount time - we need to find
4919 * the quota file and such...
4920 */
617ba13b 4921static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 4922{
287a8095
CH		 4923 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4924 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 4925}
4926
4927/*
4928 * Standard function to be called on quota_on
4929 */
617ba13b 4930static int ext4_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 4931 struct path *path)
ac27a0ec
DK		 4932{
4933 int err;
ac27a0ec
DK		 4934
4935 if (!test_opt(sb, QUOTA))
4936 return -EINVAL;
0623543b 4937
ac27a0ec 4938 /* Quotafile not on the same filesystem? */
d8c9584e 4939 if (path->dentry->d_sb != sb)
ac27a0ec 4940 return -EXDEV;
0623543b
JK		 4941 /* Journaling quota? */
4942 if (EXT4_SB(sb)->s_qf_names[type]) {
2b2d6d01 4943 /* Quotafile not in fs root? */
f00c9e44 4944 if (path->dentry->d_parent != sb->s_root)
b31e1552
ES		 4945 ext4_msg(sb, KERN_WARNING,
4946 "Quota file not on filesystem root. "
4947 "Journaled quota will not work");
2b2d6d01 4948 }
0623543b
JK		 4949
4950 /*
4951 * When we journal data on quota file, we have to flush journal to see
4952 * all updates to the file when we bypass pagecache...
4953 */
0390131b 4954 if (EXT4_SB(sb)->s_journal &&
f00c9e44 4955 ext4_should_journal_data(path->dentry->d_inode)) {
0623543b
JK		 4956 /*
4957 * We don't need to lock updates but journal_flush() could
4958 * otherwise be livelocked...
4959 */
4960 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 4961 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
0623543b 4962 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
f00c9e44 4963 if (err)
7ffe1ea8 4964 return err;
0623543b
JK		 4965 }
4966
f00c9e44 4967 return dquot_quota_on(sb, type, format_id, path);
ac27a0ec
DK		 4968}
4969
7c319d32
AK		 4970static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
4971 unsigned int flags)
4972{
4973 int err;
4974 struct inode *qf_inode;
4975 unsigned long qf_inums[MAXQUOTAS] = {
4976 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
4977 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
4978 };
4979
4980 BUG_ON(!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA));
4981
4982 if (!qf_inums[type])
4983 return -EPERM;
4984
4985 qf_inode = ext4_iget(sb, qf_inums[type]);
4986 if (IS_ERR(qf_inode)) {
4987 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
4988 return PTR_ERR(qf_inode);
4989 }
4990
4991 err = dquot_enable(qf_inode, type, format_id, flags);
4992 iput(qf_inode);
4993
4994 return err;
4995}
4996
4997/* Enable usage tracking for all quota types. */
4998static int ext4_enable_quotas(struct super_block *sb)
4999{
5000 int type, err = 0;
5001 unsigned long qf_inums[MAXQUOTAS] = {
5002 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5003 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5004 };
5005
5006 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
5007 for (type = 0; type < MAXQUOTAS; type++) {
5008 if (qf_inums[type]) {
5009 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5010 DQUOT_USAGE_ENABLED);
5011 if (err) {
5012 ext4_warning(sb,
5013 "Failed to enable quota (type=%d) "
5014 "tracking. Please run e2fsck to fix.",
5015 type);
5016 return err;
5017 }
5018 }
5019 }
5020 return 0;
5021}
5022
5023/*
5024 * quota_on function that is used when QUOTA feature is set.
5025 */
5026static int ext4_quota_on_sysfile(struct super_block *sb, int type,
5027 int format_id)
5028{
5029 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5030 return -EINVAL;
5031
5032 /*
5033 * USAGE was enabled at mount time. Only need to enable LIMITS now.
5034 */
5035 return ext4_quota_enable(sb, type, format_id, DQUOT_LIMITS_ENABLED);
5036}
5037
ca0e05e4
DM		 5038static int ext4_quota_off(struct super_block *sb, int type)
5039{
21f97697
JK		 5040 struct inode *inode = sb_dqopt(sb)->files[type];
5041 handle_t *handle;
5042
87009d86
DM		 5043 /* Force all delayed allocation blocks to be allocated.
5044 * Caller already holds s_umount sem */
5045 if (test_opt(sb, DELALLOC))
ca0e05e4 5046 sync_filesystem(sb);
ca0e05e4 5047
0b268590
AG		 5048 if (!inode)
5049 goto out;
5050
21f97697
JK		 5051 /* Update modification times of quota files when userspace can
5052 * start looking at them */
5053 handle = ext4_journal_start(inode, 1);
5054 if (IS_ERR(handle))
5055 goto out;
5056 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
5057 ext4_mark_inode_dirty(handle, inode);
5058 ext4_journal_stop(handle);
5059
5060out:
ca0e05e4
DM		 5061 return dquot_quota_off(sb, type);
5062}
5063
7c319d32
AK		 5064/*
5065 * quota_off function that is used when QUOTA feature is set.
5066 */
5067static int ext4_quota_off_sysfile(struct super_block *sb, int type)
5068{
5069 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5070 return -EINVAL;
5071
5072 /* Disable only the limits. */
5073 return dquot_disable(sb, type, DQUOT_LIMITS_ENABLED);
5074}
5075
ac27a0ec
DK		 5076/* Read data from quotafile - avoid pagecache and such because we cannot afford
5077 * acquiring the locks... As quota files are never truncated and quota code
25985edc 5078 * itself serializes the operations (and no one else should touch the files)
ac27a0ec 5079 * we don't have to be afraid of races */
617ba13b 5080static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 5081 size_t len, loff_t off)
5082{
5083 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 5084 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 5085 int err = 0;
5086 int offset = off & (sb->s_blocksize - 1);
5087 int tocopy;
5088 size_t toread;
5089 struct buffer_head *bh;
5090 loff_t i_size = i_size_read(inode);
5091
5092 if (off > i_size)
5093 return 0;
5094 if (off+len > i_size)
5095 len = i_size-off;
5096 toread = len;
5097 while (toread > 0) {
5098 tocopy = sb->s_blocksize - offset < toread ?
5099 sb->s_blocksize - offset : toread;
617ba13b 5100 bh = ext4_bread(NULL, inode, blk, 0, &err);
ac27a0ec
DK		 5101 if (err)
5102 return err;
5103 if (!bh) /* A hole? */
5104 memset(data, 0, tocopy);
5105 else
5106 memcpy(data, bh->b_data+offset, tocopy);
5107 brelse(bh);
5108 offset = 0;
5109 toread -= tocopy;
5110 data += tocopy;
5111 blk++;
5112 }
5113 return len;
5114}
5115
5116/* Write to quotafile (we know the transaction is already started and has
5117 * enough credits) */
617ba13b 5118static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 5119 const char *data, size_t len, loff_t off)
5120{
5121 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 5122 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 5123 int err = 0;
5124 int offset = off & (sb->s_blocksize - 1);
ac27a0ec
DK		 5125 struct buffer_head *bh;
5126 handle_t *handle = journal_current_handle();
5127
0390131b 5128 if (EXT4_SB(sb)->s_journal && !handle) {
b31e1552
ES		 5129 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5130 " cancelled because transaction is not started",
9c3013e9
JK		 5131 (unsigned long long)off, (unsigned long long)len);
5132 return -EIO;
5133 }
67eeb568
DM		 5134 /*
5135 * Since we account only one data block in transaction credits,
5136 * then it is impossible to cross a block boundary.
5137 */
5138 if (sb->s_blocksize - offset < len) {
5139 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5140 " cancelled because not block aligned",
5141 (unsigned long long)off, (unsigned long long)len);
5142 return -EIO;
5143 }
5144
67eeb568
DM		 5145 bh = ext4_bread(handle, inode, blk, 1, &err);
5146 if (!bh)
5147 goto out;
62d2b5f2
JK		 5148 err = ext4_journal_get_write_access(handle, bh);
5149 if (err) {
5150 brelse(bh);
5151 goto out;
ac27a0ec 5152 }
67eeb568
DM		 5153 lock_buffer(bh);
5154 memcpy(bh->b_data+offset, data, len);
5155 flush_dcache_page(bh->b_page);
5156 unlock_buffer(bh);
62d2b5f2 5157 err = ext4_handle_dirty_metadata(handle, NULL, bh);
67eeb568 5158 brelse(bh);
ac27a0ec 5159out:
0b7f7cef 5160 if (err)
ac27a0ec 5161 return err;
67eeb568
DM		 5162 if (inode->i_size < off + len) {
5163 i_size_write(inode, off + len);
617ba13b 5164 EXT4_I(inode)->i_disksize = inode->i_size;
21f97697 5165 ext4_mark_inode_dirty(handle, inode);
ac27a0ec 5166 }
67eeb568 5167 return len;
ac27a0ec
DK		 5168}
5169
5170#endif
5171
152a0836
AV		 5172static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5173 const char *dev_name, void *data)
ac27a0ec 5174{
152a0836 5175 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
ac27a0ec
DK		 5176}
5177
37f328eb 5178#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
24b58424
TT		 5179static inline void register_as_ext2(void)
5180{
5181 int err = register_filesystem(&ext2_fs_type);
5182 if (err)
5183 printk(KERN_WARNING
5184 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5185}
5186
5187static inline void unregister_as_ext2(void)
5188{
5189 unregister_filesystem(&ext2_fs_type);
5190}
2035e776
TT		 5191
5192static inline int ext2_feature_set_ok(struct super_block *sb)
5193{
5194 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
5195 return 0;
5196 if (sb->s_flags & MS_RDONLY)
5197 return 1;
5198 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
5199 return 0;
5200 return 1;
5201}
51b7e3c9 5202MODULE_ALIAS("ext2");
24b58424
TT		 5203#else
5204static inline void register_as_ext2(void) { }
5205static inline void unregister_as_ext2(void) { }
2035e776 5206static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 5207#endif
5208
37f328eb 5209#if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
24b58424
TT		 5210static inline void register_as_ext3(void)
5211{
5212 int err = register_filesystem(&ext3_fs_type);
5213 if (err)
5214 printk(KERN_WARNING
5215 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5216}
5217
5218static inline void unregister_as_ext3(void)
5219{
5220 unregister_filesystem(&ext3_fs_type);
5221}
2035e776
TT		 5222
5223static inline int ext3_feature_set_ok(struct super_block *sb)
5224{
5225 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
5226 return 0;
5227 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
5228 return 0;
5229 if (sb->s_flags & MS_RDONLY)
5230 return 1;
5231 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
5232 return 0;
5233 return 1;
5234}
51b7e3c9 5235MODULE_ALIAS("ext3");
24b58424
TT		 5236#else
5237static inline void register_as_ext3(void) { }
5238static inline void unregister_as_ext3(void) { }
2035e776 5239static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 5240#endif
5241
03010a33
TT		 5242static struct file_system_type ext4_fs_type = {
5243 .owner = THIS_MODULE,
5244 .name = "ext4",
152a0836 5245 .mount = ext4_mount,
03010a33
TT		 5246 .kill_sb = kill_block_super,
5247 .fs_flags = FS_REQUIRES_DEV,
5248};
5249
8f021222 5250static int __init ext4_init_feat_adverts(void)
857ac889
LC		 5251{
5252 struct ext4_features *ef;
5253 int ret = -ENOMEM;
5254
5255 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
5256 if (!ef)
5257 goto out;
5258
5259 ef->f_kobj.kset = ext4_kset;
5260 init_completion(&ef->f_kobj_unregister);
5261 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
5262 "features");
5263 if (ret) {
5264 kfree(ef);
5265 goto out;
5266 }
5267
5268 ext4_feat = ef;
5269 ret = 0;
5270out:
5271 return ret;
5272}
5273
8f021222
LC		 5274static void ext4_exit_feat_adverts(void)
5275{
5276 kobject_put(&ext4_feat->f_kobj);
5277 wait_for_completion(&ext4_feat->f_kobj_unregister);
5278 kfree(ext4_feat);
5279}
5280
e9e3bcec
ES		 5281/* Shared across all ext4 file systems */
5282wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5283struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5284
5dabfc78 5285static int __init ext4_init_fs(void)
ac27a0ec 5286{
e9e3bcec 5287 int i, err;
c9de560d 5288
07c0c5d8
AV		 5289 ext4_li_info = NULL;
5290 mutex_init(&ext4_li_mtx);
5291
12e9b892 5292 ext4_check_flag_values();
e9e3bcec
ES		 5293
5294 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5295 mutex_init(&ext4__aio_mutex[i]);
5296 init_waitqueue_head(&ext4__ioend_wq[i]);
5297 }
5298
5dabfc78 5299 err = ext4_init_pageio();
6fd058f7
TT		 5300 if (err)
5301 return err;
5dabfc78 5302 err = ext4_init_system_zone();
bd2d0210 5303 if (err)
d44651d0 5304 goto out6;
3197ebdb 5305 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
0e376b1e
TT		 5306 if (!ext4_kset) {
5307 err = -ENOMEM;
dd68314c 5308 goto out5;
0e376b1e 5309 }
d44651d0 5310 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
857ac889
LC		 5311
5312 err = ext4_init_feat_adverts();
dd68314c
TT		 5313 if (err)
5314 goto out4;
857ac889 5315
5dabfc78 5316 err = ext4_init_mballoc();
ac27a0ec 5317 if (err)
6fd058f7 5318 goto out3;
c9de560d 5319
5dabfc78 5320 err = ext4_init_xattr();
c9de560d
AT		 5321 if (err)
5322 goto out2;
ac27a0ec
DK		 5323 err = init_inodecache();
5324 if (err)
5325 goto out1;
24b58424 5326 register_as_ext3();
2035e776 5327 register_as_ext2();
03010a33 5328 err = register_filesystem(&ext4_fs_type);
ac27a0ec
DK		 5329 if (err)
5330 goto out;
bfff6873 5331
ac27a0ec
DK		 5332 return 0;
5333out:
24b58424
TT		 5334 unregister_as_ext2();
5335 unregister_as_ext3();
ac27a0ec
DK		 5336 destroy_inodecache();
5337out1:
5dabfc78 5338 ext4_exit_xattr();
c9de560d 5339out2:
5dabfc78 5340 ext4_exit_mballoc();
6fd058f7 5341out3:
8f021222 5342 ext4_exit_feat_adverts();
dd68314c 5343out4:
d44651d0
FJ		 5344 if (ext4_proc_root)
5345 remove_proc_entry("fs/ext4", NULL);
6fd058f7 5346 kset_unregister(ext4_kset);
d44651d0 5347out5:
5dabfc78 5348 ext4_exit_system_zone();
d44651d0 5349out6:
5dabfc78 5350 ext4_exit_pageio();
ac27a0ec
DK		 5351 return err;
5352}
5353
5dabfc78 5354static void __exit ext4_exit_fs(void)
ac27a0ec 5355{
bfff6873 5356 ext4_destroy_lazyinit_thread();
24b58424
TT		 5357 unregister_as_ext2();
5358 unregister_as_ext3();
03010a33 5359 unregister_filesystem(&ext4_fs_type);
ac27a0ec 5360 destroy_inodecache();
5dabfc78
TT		 5361 ext4_exit_xattr();
5362 ext4_exit_mballoc();
8f021222 5363 ext4_exit_feat_adverts();
9f6200bb 5364 remove_proc_entry("fs/ext4", NULL);
3197ebdb 5365 kset_unregister(ext4_kset);
5dabfc78
TT		 5366 ext4_exit_system_zone();
5367 ext4_exit_pageio();
ac27a0ec
DK		 5368}
5369
5370MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
83982b6f 5371MODULE_DESCRIPTION("Fourth Extended Filesystem");
ac27a0ec 5372MODULE_LICENSE("GPL");
5dabfc78
TT		 5373module_init(ext4_init_fs)
5374module_exit(ext4_exit_fs)
kernel source for telekom puls (alcatel/mediatek)