projects / linux-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
btrfs: reada_extent doesn't need kref for refcount
[linux-block.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
146void ext4_superblock_csum_set(struct super_block *sb,
147 struct ext4_super_block *es)
148{
149 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
150 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
151 return;
152
153 es->s_checksum = ext4_superblock_csum(sb, es);
154}
155
9933fc0a
TT		 156void *ext4_kvmalloc(size_t size, gfp_t flags)
157{
158 void *ret;
159
160 ret = kmalloc(size, flags);
161 if (!ret)
162 ret = __vmalloc(size, flags, PAGE_KERNEL);
163 return ret;
164}
165
166void *ext4_kvzalloc(size_t size, gfp_t flags)
167{
168 void *ret;
169
db9481c0 170 ret = kzalloc(size, flags);
9933fc0a
TT		 171 if (!ret)
172 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
173 return ret;
174}
175
176void ext4_kvfree(void *ptr)
177{
178 if (is_vmalloc_addr(ptr))
179 vfree(ptr);
180 else
181 kfree(ptr);
182
183}
184
8fadc143
AR		 185ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
186 struct ext4_group_desc *bg)
bd81d8ee 187{
3a14589c 188 return le32_to_cpu(bg->bg_block_bitmap_lo) |
8fadc143 189 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 190 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 191}
192
8fadc143
AR		 193ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
194 struct ext4_group_desc *bg)
bd81d8ee 195{
5272f837 196 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
8fadc143 197 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 198 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 199}
200
8fadc143
AR		 201ext4_fsblk_t ext4_inode_table(struct super_block *sb,
202 struct ext4_group_desc *bg)
bd81d8ee 203{
5272f837 204 return le32_to_cpu(bg->bg_inode_table_lo) |
8fadc143 205 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 206 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
bd81d8ee
LV		 207}
208
021b65bb
TT		 209__u32 ext4_free_group_clusters(struct super_block *sb,
210 struct ext4_group_desc *bg)
560671a0
AK		 211{
212 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
213 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 214 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
560671a0
AK		 215}
216
217__u32 ext4_free_inodes_count(struct super_block *sb,
218 struct ext4_group_desc *bg)
219{
220 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
221 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 222 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
560671a0
AK		 223}
224
225__u32 ext4_used_dirs_count(struct super_block *sb,
226 struct ext4_group_desc *bg)
227{
228 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
229 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 230 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
560671a0
AK		 231}
232
233__u32 ext4_itable_unused_count(struct super_block *sb,
234 struct ext4_group_desc *bg)
235{
236 return le16_to_cpu(bg->bg_itable_unused_lo) |
237 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 238 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
560671a0
AK		 239}
240
8fadc143
AR		 241void ext4_block_bitmap_set(struct super_block *sb,
242 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 243{
3a14589c 244 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 245 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
246 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 247}
248
8fadc143
AR		 249void ext4_inode_bitmap_set(struct super_block *sb,
250 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 251{
5272f837 252 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 253 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
254 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 255}
256
8fadc143
AR		 257void ext4_inode_table_set(struct super_block *sb,
258 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 259{
5272f837 260 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 261 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
262 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 263}
264
021b65bb
TT		 265void ext4_free_group_clusters_set(struct super_block *sb,
266 struct ext4_group_desc *bg, __u32 count)
560671a0
AK		 267{
268 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
269 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
270 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
271}
272
273void ext4_free_inodes_set(struct super_block *sb,
274 struct ext4_group_desc *bg, __u32 count)
275{
276 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
277 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
278 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
279}
280
281void ext4_used_dirs_set(struct super_block *sb,
282 struct ext4_group_desc *bg, __u32 count)
283{
284 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
285 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
286 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
287}
288
289void ext4_itable_unused_set(struct super_block *sb,
290 struct ext4_group_desc *bg, __u32 count)
291{
292 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
293 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
294 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
295}
296
d3d1faf6
CW		 297
298/* Just increment the non-pointer handle value */
299static handle_t *ext4_get_nojournal(void)
300{
301 handle_t *handle = current->journal_info;
302 unsigned long ref_cnt = (unsigned long)handle;
303
304 BUG_ON(ref_cnt >= EXT4_NOJOURNAL_MAX_REF_COUNT);
305
306 ref_cnt++;
307 handle = (handle_t *)ref_cnt;
308
309 current->journal_info = handle;
310 return handle;
311}
312
313
314/* Decrement the non-pointer handle value */
315static void ext4_put_nojournal(handle_t *handle)
316{
317 unsigned long ref_cnt = (unsigned long)handle;
318
319 BUG_ON(ref_cnt == 0);
320
321 ref_cnt--;
322 handle = (handle_t *)ref_cnt;
323
324 current->journal_info = handle;
325}
326
ac27a0ec 327/*
dab291af 328 * Wrappers for jbd2_journal_start/end.
ac27a0ec 329 */
617ba13b 330handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
ac27a0ec
DK		 331{
332 journal_t *journal;
333
12706394 334 trace_ext4_journal_start(sb, nblocks, _RET_IP_);
ac27a0ec
DK		 335 if (sb->s_flags & MS_RDONLY)
336 return ERR_PTR(-EROFS);
337
8e8ad8a5 338 WARN_ON(sb->s_writers.frozen == SB_FREEZE_COMPLETE);
617ba13b 339 journal = EXT4_SB(sb)->s_journal;
be4f27d3
YY		 340 if (!journal)
341 return ext4_get_nojournal();
342 /*
343 * Special case here: if the journal has aborted behind our
344 * backs (eg. EIO in the commit thread), then we still need to
345 * take the FS itself readonly cleanly.
346 */
347 if (is_journal_aborted(journal)) {
348 ext4_abort(sb, "Detected aborted journal");
349 return ERR_PTR(-EROFS);
ac27a0ec 350 }
be4f27d3 351 return jbd2_journal_start(journal, nblocks);
ac27a0ec
DK		 352}
353
c398eda0 354int __ext4_journal_stop(const char *where, unsigned int line, handle_t *handle)
ac27a0ec
DK		 355{
356 struct super_block *sb;
357 int err;
358 int rc;
359
0390131b 360 if (!ext4_handle_valid(handle)) {
d3d1faf6 361 ext4_put_nojournal(handle);
0390131b
FM		 362 return 0;
363 }
ac27a0ec
DK		 364 sb = handle->h_transaction->t_journal->j_private;
365 err = handle->h_err;
dab291af 366 rc = jbd2_journal_stop(handle);
ac27a0ec
DK		 367
368 if (!err)
369 err = rc;
370 if (err)
c398eda0 371 __ext4_std_error(sb, where, line, err);
ac27a0ec
DK		 372 return err;
373}
374
90c7201b
TT		 375void ext4_journal_abort_handle(const char *caller, unsigned int line,
376 const char *err_fn, struct buffer_head *bh,
377 handle_t *handle, int err)
ac27a0ec
DK		 378{
379 char nbuf[16];
617ba13b 380 const char *errstr = ext4_decode_error(NULL, err, nbuf);
ac27a0ec 381
0390131b
FM		 382 BUG_ON(!ext4_handle_valid(handle));
383
ac27a0ec
DK		 384 if (bh)
385 BUFFER_TRACE(bh, "abort");
386
387 if (!handle->h_err)
388 handle->h_err = err;
389
390 if (is_handle_aborted(handle))
391 return;
392
92b97816 393 printk(KERN_ERR "EXT4-fs: %s:%d: aborting transaction: %s in %s\n",
90c7201b 394 caller, line, errstr, err_fn);
ac27a0ec 395
dab291af 396 jbd2_journal_abort_handle(handle);
ac27a0ec
DK		 397}
398
1c13d5c0
TT		 399static void __save_error_info(struct super_block *sb, const char *func,
400 unsigned int line)
401{
402 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
403
404 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
405 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
406 es->s_last_error_time = cpu_to_le32(get_seconds());
407 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
408 es->s_last_error_line = cpu_to_le32(line);
409 if (!es->s_first_error_time) {
410 es->s_first_error_time = es->s_last_error_time;
411 strncpy(es->s_first_error_func, func,
412 sizeof(es->s_first_error_func));
413 es->s_first_error_line = cpu_to_le32(line);
414 es->s_first_error_ino = es->s_last_error_ino;
415 es->s_first_error_block = es->s_last_error_block;
416 }
66e61a9e
TT		 417 /*
418 * Start the daily error reporting function if it hasn't been
419 * started already
420 */
421 if (!es->s_error_count)
422 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
1c13d5c0
TT		 423 es->s_error_count = cpu_to_le32(le32_to_cpu(es->s_error_count) + 1);
424}
425
426static void save_error_info(struct super_block *sb, const char *func,
427 unsigned int line)
428{
429 __save_error_info(sb, func, line);
430 ext4_commit_super(sb, 1);
431}
432
7c2e7087
TT		 433/*
434 * The del_gendisk() function uninitializes the disk-specific data
435 * structures, including the bdi structure, without telling anyone
436 * else. Once this happens, any attempt to call mark_buffer_dirty()
437 * (for example, by ext4_commit_super), will cause a kernel OOPS.
438 * This is a kludge to prevent these oops until we can put in a proper
439 * hook in del_gendisk() to inform the VFS and file system layers.
440 */
441static int block_device_ejected(struct super_block *sb)
442{
443 struct inode *bd_inode = sb->s_bdev->bd_inode;
444 struct backing_dev_info *bdi = bd_inode->i_mapping->backing_dev_info;
445
446 return bdi->dev == NULL;
447}
448
18aadd47
BJ		 449static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
450{
451 struct super_block *sb = journal->j_private;
452 struct ext4_sb_info *sbi = EXT4_SB(sb);
453 int error = is_journal_aborted(journal);
454 struct ext4_journal_cb_entry *jce, *tmp;
455
456 spin_lock(&sbi->s_md_lock);
457 list_for_each_entry_safe(jce, tmp, &txn->t_private_list, jce_list) {
458 list_del_init(&jce->jce_list);
459 spin_unlock(&sbi->s_md_lock);
460 jce->jce_func(sb, jce, error);
461 spin_lock(&sbi->s_md_lock);
462 }
463 spin_unlock(&sbi->s_md_lock);
464}
1c13d5c0 465
ac27a0ec
DK		 466/* Deal with the reporting of failure conditions on a filesystem such as
467 * inconsistencies detected or read IO failures.
468 *
469 * On ext2, we can store the error state of the filesystem in the
617ba13b 470 * superblock. That is not possible on ext4, because we may have other
ac27a0ec
DK		 471 * write ordering constraints on the superblock which prevent us from
472 * writing it out straight away; and given that the journal is about to
473 * be aborted, we can't rely on the current, or future, transactions to
474 * write out the superblock safely.
475 *
dab291af 476 * We'll just use the jbd2_journal_abort() error code to record an error in
d6b198bc 477 * the journal instead. On recovery, the journal will complain about
ac27a0ec
DK		 478 * that error until we've noted it down and cleared it.
479 */
480
617ba13b 481static void ext4_handle_error(struct super_block *sb)
ac27a0ec 482{
ac27a0ec
DK		 483 if (sb->s_flags & MS_RDONLY)
484 return;
485
2b2d6d01 486 if (!test_opt(sb, ERRORS_CONT)) {
617ba13b 487 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 488
4ab2f15b 489 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
ac27a0ec 490 if (journal)
dab291af 491 jbd2_journal_abort(journal, -EIO);
ac27a0ec 492 }
2b2d6d01 493 if (test_opt(sb, ERRORS_RO)) {
b31e1552 494 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
ac27a0ec
DK		 495 sb->s_flags |= MS_RDONLY;
496 }
ac27a0ec 497 if (test_opt(sb, ERRORS_PANIC))
617ba13b 498 panic("EXT4-fs (device %s): panic forced after error\n",
ac27a0ec
DK		 499 sb->s_id);
500}
501
12062ddd 502void __ext4_error(struct super_block *sb, const char *function,
c398eda0 503 unsigned int line, const char *fmt, ...)
ac27a0ec 504{
0ff2ea7d 505 struct va_format vaf;
ac27a0ec
DK		 506 va_list args;
507
508 va_start(args, fmt);
0ff2ea7d
JP		 509 vaf.fmt = fmt;
510 vaf.va = &args;
511 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
512 sb->s_id, function, line, current->comm, &vaf);
ac27a0ec 513 va_end(args);
f3fc0210 514 save_error_info(sb, function, line);
ac27a0ec 515
617ba13b 516 ext4_handle_error(sb);
ac27a0ec
DK		 517}
518
c398eda0
TT		 519void ext4_error_inode(struct inode *inode, const char *function,
520 unsigned int line, ext4_fsblk_t block,
273df556
FM		 521 const char *fmt, ...)
522{
523 va_list args;
f7c21177 524 struct va_format vaf;
1c13d5c0 525 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
273df556 526
1c13d5c0
TT		 527 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
528 es->s_last_error_block = cpu_to_le64(block);
529 save_error_info(inode->i_sb, function, line);
273df556 530 va_start(args, fmt);
f7c21177
TT		 531 vaf.fmt = fmt;
532 vaf.va = &args;
c398eda0 533 if (block)
d9ee81da
JP		 534 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
535 "inode #%lu: block %llu: comm %s: %pV\n",
536 inode->i_sb->s_id, function, line, inode->i_ino,
537 block, current->comm, &vaf);
538 else
539 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
540 "inode #%lu: comm %s: %pV\n",
541 inode->i_sb->s_id, function, line, inode->i_ino,
542 current->comm, &vaf);
273df556
FM		 543 va_end(args);
544
545 ext4_handle_error(inode->i_sb);
546}
547
c398eda0 548void ext4_error_file(struct file *file, const char *function,
f7c21177
TT		 549 unsigned int line, ext4_fsblk_t block,
550 const char *fmt, ...)
273df556
FM		 551{
552 va_list args;
f7c21177 553 struct va_format vaf;
1c13d5c0 554 struct ext4_super_block *es;
273df556
FM		 555 struct inode *inode = file->f_dentry->d_inode;
556 char pathname[80], *path;
557
1c13d5c0
TT		 558 es = EXT4_SB(inode->i_sb)->s_es;
559 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
560 save_error_info(inode->i_sb, function, line);
273df556 561 path = d_path(&(file->f_path), pathname, sizeof(pathname));
f9a62d09 562 if (IS_ERR(path))
273df556 563 path = "(unknown)";
f7c21177
TT		 564 va_start(args, fmt);
565 vaf.fmt = fmt;
566 vaf.va = &args;
d9ee81da
JP		 567 if (block)
568 printk(KERN_CRIT
569 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
570 "block %llu: comm %s: path %s: %pV\n",
571 inode->i_sb->s_id, function, line, inode->i_ino,
572 block, current->comm, path, &vaf);
573 else
574 printk(KERN_CRIT
575 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
576 "comm %s: path %s: %pV\n",
577 inode->i_sb->s_id, function, line, inode->i_ino,
578 current->comm, path, &vaf);
273df556
FM		 579 va_end(args);
580
581 ext4_handle_error(inode->i_sb);
582}
583
2b2d6d01 584static const char *ext4_decode_error(struct super_block *sb, int errno,
ac27a0ec
DK		 585 char nbuf[16])
586{
587 char *errstr = NULL;
588
589 switch (errno) {
590 case -EIO:
591 errstr = "IO failure";
592 break;
593 case -ENOMEM:
594 errstr = "Out of memory";
595 break;
596 case -EROFS:
78f1ddbb
TT		 597 if (!sb || (EXT4_SB(sb)->s_journal &&
598 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
ac27a0ec
DK		 599 errstr = "Journal has aborted";
600 else
601 errstr = "Readonly filesystem";
602 break;
603 default:
604 /* If the caller passed in an extra buffer for unknown
605 * errors, textualise them now. Else we just return
606 * NULL. */
607 if (nbuf) {
608 /* Check for truncated error codes... */
609 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
610 errstr = nbuf;
611 }
612 break;
613 }
614
615 return errstr;
616}
617
617ba13b 618/* __ext4_std_error decodes expected errors from journaling functions
ac27a0ec
DK		 619 * automatically and invokes the appropriate error response. */
620
c398eda0
TT		 621void __ext4_std_error(struct super_block *sb, const char *function,
622 unsigned int line, int errno)
ac27a0ec
DK		 623{
624 char nbuf[16];
625 const char *errstr;
626
627 /* Special case: if the error is EROFS, and we're not already
628 * inside a transaction, then there's really no point in logging
629 * an error. */
630 if (errno == -EROFS && journal_current_handle() == NULL &&
631 (sb->s_flags & MS_RDONLY))
632 return;
633
617ba13b 634 errstr = ext4_decode_error(sb, errno, nbuf);
c398eda0
TT		 635 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
636 sb->s_id, function, line, errstr);
1c13d5c0 637 save_error_info(sb, function, line);
ac27a0ec 638
617ba13b 639 ext4_handle_error(sb);
ac27a0ec
DK		 640}
641
642/*
617ba13b 643 * ext4_abort is a much stronger failure handler than ext4_error. The
ac27a0ec
DK		 644 * abort function may be used to deal with unrecoverable failures such
645 * as journal IO errors or ENOMEM at a critical moment in log management.
646 *
647 * We unconditionally force the filesystem into an ABORT|READONLY state,
648 * unless the error response on the fs has been set to panic in which
649 * case we take the easy way out and panic immediately.
650 */
651
c67d859e 652void __ext4_abort(struct super_block *sb, const char *function,
c398eda0 653 unsigned int line, const char *fmt, ...)
ac27a0ec
DK		 654{
655 va_list args;
656
1c13d5c0 657 save_error_info(sb, function, line);
ac27a0ec 658 va_start(args, fmt);
c398eda0
TT		 659 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
660 function, line);
ac27a0ec
DK		 661 vprintk(fmt, args);
662 printk("\n");
663 va_end(args);
664
1c13d5c0
TT		 665 if ((sb->s_flags & MS_RDONLY) == 0) {
666 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
667 sb->s_flags |= MS_RDONLY;
668 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
669 if (EXT4_SB(sb)->s_journal)
670 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
671 save_error_info(sb, function, line);
672 }
ac27a0ec 673 if (test_opt(sb, ERRORS_PANIC))
617ba13b 674 panic("EXT4-fs panic from previous error\n");
ac27a0ec
DK		 675}
676
0ff2ea7d 677void ext4_msg(struct super_block *sb, const char *prefix, const char *fmt, ...)
b31e1552 678{
0ff2ea7d 679 struct va_format vaf;
b31e1552
ES		 680 va_list args;
681
682 va_start(args, fmt);
0ff2ea7d
JP		 683 vaf.fmt = fmt;
684 vaf.va = &args;
685 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
b31e1552
ES		 686 va_end(args);
687}
688
12062ddd 689void __ext4_warning(struct super_block *sb, const char *function,
c398eda0 690 unsigned int line, const char *fmt, ...)
ac27a0ec 691{
0ff2ea7d 692 struct va_format vaf;
ac27a0ec
DK		 693 va_list args;
694
695 va_start(args, fmt);
0ff2ea7d
JP		 696 vaf.fmt = fmt;
697 vaf.va = &args;
698 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
699 sb->s_id, function, line, &vaf);
ac27a0ec
DK		 700 va_end(args);
701}
702
e29136f8
TT		 703void __ext4_grp_locked_error(const char *function, unsigned int line,
704 struct super_block *sb, ext4_group_t grp,
705 unsigned long ino, ext4_fsblk_t block,
706 const char *fmt, ...)
5d1b1b3f
AK		 707__releases(bitlock)
708__acquires(bitlock)
709{
0ff2ea7d 710 struct va_format vaf;
5d1b1b3f
AK		 711 va_list args;
712 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
713
1c13d5c0
TT		 714 es->s_last_error_ino = cpu_to_le32(ino);
715 es->s_last_error_block = cpu_to_le64(block);
716 __save_error_info(sb, function, line);
0ff2ea7d 717
5d1b1b3f 718 va_start(args, fmt);
0ff2ea7d
JP		 719
720 vaf.fmt = fmt;
721 vaf.va = &args;
21149d61 722 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
e29136f8
TT		 723 sb->s_id, function, line, grp);
724 if (ino)
0ff2ea7d 725 printk(KERN_CONT "inode %lu: ", ino);
e29136f8 726 if (block)
0ff2ea7d
JP		 727 printk(KERN_CONT "block %llu:", (unsigned long long) block);
728 printk(KERN_CONT "%pV\n", &vaf);
5d1b1b3f
AK		 729 va_end(args);
730
731 if (test_opt(sb, ERRORS_CONT)) {
e2d67052 732 ext4_commit_super(sb, 0);
5d1b1b3f
AK		 733 return;
734 }
1c13d5c0 735
5d1b1b3f
AK		 736 ext4_unlock_group(sb, grp);
737 ext4_handle_error(sb);
738 /*
739 * We only get here in the ERRORS_RO case; relocking the group
740 * may be dangerous, but nothing bad will happen since the
741 * filesystem will have already been marked read/only and the
742 * journal has been aborted. We return 1 as a hint to callers
743 * who might what to use the return value from
25985edc 744 * ext4_grp_locked_error() to distinguish between the
5d1b1b3f
AK		 745 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
746 * aggressively from the ext4 function in question, with a
747 * more appropriate error code.
748 */
749 ext4_lock_group(sb, grp);
750 return;
751}
752
617ba13b 753void ext4_update_dynamic_rev(struct super_block *sb)
ac27a0ec 754{
617ba13b 755 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 756
617ba13b 757 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
ac27a0ec
DK		 758 return;
759
12062ddd 760 ext4_warning(sb,
ac27a0ec
DK		 761 "updating to rev %d because of new feature flag, "
762 "running e2fsck is recommended",
617ba13b 763 EXT4_DYNAMIC_REV);
ac27a0ec 764
617ba13b
MC		 765 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
766 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
767 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
ac27a0ec
DK		 768 /* leave es->s_feature_*compat flags alone */
769 /* es->s_uuid will be set by e2fsck if empty */
770
771 /*
772 * The rest of the superblock fields should be zero, and if not it
773 * means they are likely already in use, so leave them alone. We
774 * can leave it up to e2fsck to clean up any inconsistencies there.
775 */
776}
777
778/*
779 * Open the external journal device
780 */
b31e1552 781static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
ac27a0ec
DK		 782{
783 struct block_device *bdev;
784 char b[BDEVNAME_SIZE];
785
d4d77629 786 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
ac27a0ec
DK		 787 if (IS_ERR(bdev))
788 goto fail;
789 return bdev;
790
791fail:
b31e1552 792 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
ac27a0ec
DK		 793 __bdevname(dev, b), PTR_ERR(bdev));
794 return NULL;
795}
796
797/*
798 * Release the journal device
799 */
617ba13b 800static int ext4_blkdev_put(struct block_device *bdev)
ac27a0ec 801{
e525fd89 802 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
ac27a0ec
DK		 803}
804
617ba13b 805static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
ac27a0ec
DK		 806{
807 struct block_device *bdev;
808 int ret = -ENODEV;
809
810 bdev = sbi->journal_bdev;
811 if (bdev) {
617ba13b 812 ret = ext4_blkdev_put(bdev);
ac27a0ec
DK		 813 sbi->journal_bdev = NULL;
814 }
815 return ret;
816}
817
818static inline struct inode *orphan_list_entry(struct list_head *l)
819{
617ba13b 820 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
ac27a0ec
DK		 821}
822
617ba13b 823static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
ac27a0ec
DK		 824{
825 struct list_head *l;
826
b31e1552
ES		 827 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
828 le32_to_cpu(sbi->s_es->s_last_orphan));
ac27a0ec
DK		 829
830 printk(KERN_ERR "sb_info orphan list:\n");
831 list_for_each(l, &sbi->s_orphan) {
832 struct inode *inode = orphan_list_entry(l);
833 printk(KERN_ERR " "
834 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
835 inode->i_sb->s_id, inode->i_ino, inode,
836 inode->i_mode, inode->i_nlink,
837 NEXT_ORPHAN(inode));
838 }
839}
840
2b2d6d01 841static void ext4_put_super(struct super_block *sb)
ac27a0ec 842{
617ba13b
MC		 843 struct ext4_sb_info *sbi = EXT4_SB(sb);
844 struct ext4_super_block *es = sbi->s_es;
ef2cabf7 845 int i, err;
ac27a0ec 846
857ac889 847 ext4_unregister_li_request(sb);
e0ccfd95
CH		 848 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
849
4c0425ff
MC		 850 flush_workqueue(sbi->dio_unwritten_wq);
851 destroy_workqueue(sbi->dio_unwritten_wq);
852
a9e220f8 853 lock_super(sb);
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 unlock_super(sb);
921 kobject_put(&sbi->s_kobj);
922 wait_for_completion(&sbi->s_kobj_unregister);
0441984a
DW		 923 if (sbi->s_chksum_driver)
924 crypto_free_shash(sbi->s_chksum_driver);
705895b6 925 kfree(sbi->s_blockgroup_lock);
ac27a0ec 926 kfree(sbi);
ac27a0ec
DK		 927}
928
e18b890b 929static struct kmem_cache *ext4_inode_cachep;
ac27a0ec
DK		 930
931/*
932 * Called inside transaction, so use GFP_NOFS
933 */
617ba13b 934static struct inode *ext4_alloc_inode(struct super_block *sb)
ac27a0ec 935{
617ba13b 936 struct ext4_inode_info *ei;
ac27a0ec 937
e6b4f8da 938 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
ac27a0ec
DK		 939 if (!ei)
940 return NULL;
0b8e58a1 941
ac27a0ec 942 ei->vfs_inode.i_version = 1;
91246c00 943 ei->vfs_inode.i_data.writeback_index = 0;
a86c6181 944 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
c9de560d
AT		 945 INIT_LIST_HEAD(&ei->i_prealloc_list);
946 spin_lock_init(&ei->i_prealloc_lock);
d2a17637
MC		 947 ei->i_reserved_data_blocks = 0;
948 ei->i_reserved_meta_blocks = 0;
949 ei->i_allocated_meta_blocks = 0;
9d0be502 950 ei->i_da_metadata_calc_len = 0;
7e731bc9 951 ei->i_da_metadata_calc_last_lblock = 0;
d2a17637 952 spin_lock_init(&(ei->i_block_reservation_lock));
a9e7f447
DM		 953#ifdef CONFIG_QUOTA
954 ei->i_reserved_quota = 0;
955#endif
8aefcd55 956 ei->jinode = NULL;
c7064ef1 957 INIT_LIST_HEAD(&ei->i_completed_io_list);
744692dc 958 spin_lock_init(&ei->i_completed_io_lock);
8d5d02e6 959 ei->cur_aio_dio = NULL;
b436b9be
JK		 960 ei->i_sync_tid = 0;
961 ei->i_datasync_tid = 0;
f7ad6d2e 962 atomic_set(&ei->i_ioend_count, 0);
e9e3bcec 963 atomic_set(&ei->i_aiodio_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{
617ba13b 1022 kmem_cache_destroy(ext4_inode_cachep);
ac27a0ec
DK		 1023}
1024
0930fcc1 1025void ext4_clear_inode(struct inode *inode)
ac27a0ec 1026{
0930fcc1 1027 invalidate_inode_buffers(inode);
dbd5768f 1028 clear_inode(inode);
9f754758 1029 dquot_drop(inode);
c2ea3fde 1030 ext4_discard_preallocations(inode);
8aefcd55
TT		 1031 if (EXT4_I(inode)->jinode) {
1032 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1033 EXT4_I(inode)->jinode);
1034 jbd2_free_inode(EXT4_I(inode)->jinode);
1035 EXT4_I(inode)->jinode = NULL;
1036 }
ac27a0ec
DK		 1037}
1038
1b961ac0 1039static struct inode *ext4_nfs_get_inode(struct super_block *sb,
0b8e58a1 1040 u64 ino, u32 generation)
ac27a0ec 1041{
ac27a0ec 1042 struct inode *inode;
ac27a0ec 1043
617ba13b 1044 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
ac27a0ec 1045 return ERR_PTR(-ESTALE);
617ba13b 1046 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
ac27a0ec
DK		 1047 return ERR_PTR(-ESTALE);
1048
1049 /* iget isn't really right if the inode is currently unallocated!!
1050 *
617ba13b 1051 * ext4_read_inode will return a bad_inode if the inode had been
ac27a0ec
DK		 1052 * deleted, so we should be safe.
1053 *
1054 * Currently we don't know the generation for parent directory, so
1055 * a generation of 0 means "accept any"
1056 */
1d1fe1ee
DH		 1057 inode = ext4_iget(sb, ino);
1058 if (IS_ERR(inode))
1059 return ERR_CAST(inode);
1060 if (generation && inode->i_generation != generation) {
ac27a0ec
DK		 1061 iput(inode);
1062 return ERR_PTR(-ESTALE);
1063 }
1b961ac0
CH		 1064
1065 return inode;
1066}
1067
1068static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
0b8e58a1 1069 int fh_len, int fh_type)
1b961ac0
CH		 1070{
1071 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1072 ext4_nfs_get_inode);
1073}
1074
1075static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
0b8e58a1 1076 int fh_len, int fh_type)
1b961ac0
CH		 1077{
1078 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1079 ext4_nfs_get_inode);
ac27a0ec
DK		 1080}
1081
c39a7f84
TO		 1082/*
1083 * Try to release metadata pages (indirect blocks, directories) which are
1084 * mapped via the block device. Since these pages could have journal heads
1085 * which would prevent try_to_free_buffers() from freeing them, we must use
1086 * jbd2 layer's try_to_free_buffers() function to release them.
1087 */
0b8e58a1
AD		 1088static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1089 gfp_t wait)
c39a7f84
TO		 1090{
1091 journal_t *journal = EXT4_SB(sb)->s_journal;
1092
1093 WARN_ON(PageChecked(page));
1094 if (!page_has_buffers(page))
1095 return 0;
1096 if (journal)
1097 return jbd2_journal_try_to_free_buffers(journal, page,
1098 wait & ~__GFP_WAIT);
1099 return try_to_free_buffers(page);
1100}
1101
ac27a0ec 1102#ifdef CONFIG_QUOTA
af5bc92d 1103#define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
2b2d6d01 1104#define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
ac27a0ec 1105
617ba13b
MC		 1106static int ext4_write_dquot(struct dquot *dquot);
1107static int ext4_acquire_dquot(struct dquot *dquot);
1108static int ext4_release_dquot(struct dquot *dquot);
1109static int ext4_mark_dquot_dirty(struct dquot *dquot);
1110static int ext4_write_info(struct super_block *sb, int type);
6f28e087 1111static int ext4_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 1112 struct path *path);
7c319d32
AK		 1113static int ext4_quota_on_sysfile(struct super_block *sb, int type,
1114 int format_id);
ca0e05e4 1115static int ext4_quota_off(struct super_block *sb, int type);
7c319d32 1116static int ext4_quota_off_sysfile(struct super_block *sb, int type);
617ba13b
MC		 1117static int ext4_quota_on_mount(struct super_block *sb, int type);
1118static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 1119 size_t len, loff_t off);
617ba13b 1120static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec 1121 const char *data, size_t len, loff_t off);
7c319d32
AK		 1122static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1123 unsigned int flags);
1124static int ext4_enable_quotas(struct super_block *sb);
ac27a0ec 1125
61e225dc 1126static const struct dquot_operations ext4_quota_operations = {
60e58e0f 1127 .get_reserved_space = ext4_get_reserved_space,
617ba13b
MC		 1128 .write_dquot = ext4_write_dquot,
1129 .acquire_dquot = ext4_acquire_dquot,
1130 .release_dquot = ext4_release_dquot,
1131 .mark_dirty = ext4_mark_dquot_dirty,
a5b5ee32
JK		 1132 .write_info = ext4_write_info,
1133 .alloc_dquot = dquot_alloc,
1134 .destroy_dquot = dquot_destroy,
ac27a0ec
DK		 1135};
1136
0d54b217 1137static const struct quotactl_ops ext4_qctl_operations = {
617ba13b 1138 .quota_on = ext4_quota_on,
ca0e05e4 1139 .quota_off = ext4_quota_off,
287a8095
CH		 1140 .quota_sync = dquot_quota_sync,
1141 .get_info = dquot_get_dqinfo,
1142 .set_info = dquot_set_dqinfo,
1143 .get_dqblk = dquot_get_dqblk,
1144 .set_dqblk = dquot_set_dqblk
ac27a0ec 1145};
7c319d32
AK		 1146
1147static const struct quotactl_ops ext4_qctl_sysfile_operations = {
1148 .quota_on_meta = ext4_quota_on_sysfile,
1149 .quota_off = ext4_quota_off_sysfile,
1150 .quota_sync = dquot_quota_sync,
1151 .get_info = dquot_get_dqinfo,
1152 .set_info = dquot_set_dqinfo,
1153 .get_dqblk = dquot_get_dqblk,
1154 .set_dqblk = dquot_set_dqblk
1155};
ac27a0ec
DK		 1156#endif
1157
ee9b6d61 1158static const struct super_operations ext4_sops = {
617ba13b
MC		 1159 .alloc_inode = ext4_alloc_inode,
1160 .destroy_inode = ext4_destroy_inode,
617ba13b
MC		 1161 .write_inode = ext4_write_inode,
1162 .dirty_inode = ext4_dirty_inode,
7ff9c073 1163 .drop_inode = ext4_drop_inode,
0930fcc1 1164 .evict_inode = ext4_evict_inode,
617ba13b 1165 .put_super = ext4_put_super,
617ba13b 1166 .sync_fs = ext4_sync_fs,
c4be0c1d
TS		 1167 .freeze_fs = ext4_freeze,
1168 .unfreeze_fs = ext4_unfreeze,
617ba13b
MC		 1169 .statfs = ext4_statfs,
1170 .remount_fs = ext4_remount,
617ba13b 1171 .show_options = ext4_show_options,
ac27a0ec 1172#ifdef CONFIG_QUOTA
617ba13b
MC		 1173 .quota_read = ext4_quota_read,
1174 .quota_write = ext4_quota_write,
ac27a0ec 1175#endif
c39a7f84 1176 .bdev_try_to_free_page = bdev_try_to_free_page,
ac27a0ec
DK		 1177};
1178
9ca92389
TT		 1179static const struct super_operations ext4_nojournal_sops = {
1180 .alloc_inode = ext4_alloc_inode,
1181 .destroy_inode = ext4_destroy_inode,
1182 .write_inode = ext4_write_inode,
1183 .dirty_inode = ext4_dirty_inode,
7ff9c073 1184 .drop_inode = ext4_drop_inode,
0930fcc1 1185 .evict_inode = ext4_evict_inode,
9ca92389
TT		 1186 .put_super = ext4_put_super,
1187 .statfs = ext4_statfs,
1188 .remount_fs = ext4_remount,
9ca92389
TT		 1189 .show_options = ext4_show_options,
1190#ifdef CONFIG_QUOTA
1191 .quota_read = ext4_quota_read,
1192 .quota_write = ext4_quota_write,
1193#endif
1194 .bdev_try_to_free_page = bdev_try_to_free_page,
1195};
1196
39655164 1197static const struct export_operations ext4_export_ops = {
1b961ac0
CH		 1198 .fh_to_dentry = ext4_fh_to_dentry,
1199 .fh_to_parent = ext4_fh_to_parent,
617ba13b 1200 .get_parent = ext4_get_parent,
ac27a0ec
DK		 1201};
1202
1203enum {
1204 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1205 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
72578c33 1206 Opt_nouid32, Opt_debug, Opt_removed,
ac27a0ec 1207 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
72578c33 1208 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
30773840 1209 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
43e625d8 1210 Opt_journal_dev, Opt_journal_checksum, Opt_journal_async_commit,
ac27a0ec 1211 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
296c355c 1212 Opt_data_err_abort, Opt_data_err_ignore,
ac27a0ec 1213 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
5a20bdfc 1214 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
ee4a3fcd 1215 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
661aa520 1216 Opt_usrquota, Opt_grpquota, Opt_i_version,
1449032b
TT		 1217 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1218 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
5328e635 1219 Opt_inode_readahead_blks, Opt_journal_ioprio,
744692dc 1220 Opt_dioread_nolock, Opt_dioread_lock,
fc6cb1cd 1221 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
ac27a0ec
DK		 1222};
1223
a447c093 1224static const match_table_t tokens = {
ac27a0ec
DK		 1225 {Opt_bsd_df, "bsddf"},
1226 {Opt_minix_df, "minixdf"},
1227 {Opt_grpid, "grpid"},
1228 {Opt_grpid, "bsdgroups"},
1229 {Opt_nogrpid, "nogrpid"},
1230 {Opt_nogrpid, "sysvgroups"},
1231 {Opt_resgid, "resgid=%u"},
1232 {Opt_resuid, "resuid=%u"},
1233 {Opt_sb, "sb=%u"},
1234 {Opt_err_cont, "errors=continue"},
1235 {Opt_err_panic, "errors=panic"},
1236 {Opt_err_ro, "errors=remount-ro"},
1237 {Opt_nouid32, "nouid32"},
ac27a0ec 1238 {Opt_debug, "debug"},
72578c33
TT		 1239 {Opt_removed, "oldalloc"},
1240 {Opt_removed, "orlov"},
ac27a0ec
DK		 1241 {Opt_user_xattr, "user_xattr"},
1242 {Opt_nouser_xattr, "nouser_xattr"},
1243 {Opt_acl, "acl"},
1244 {Opt_noacl, "noacl"},
e3bb52ae 1245 {Opt_noload, "norecovery"},
5a916be1 1246 {Opt_noload, "noload"},
72578c33
TT		 1247 {Opt_removed, "nobh"},
1248 {Opt_removed, "bh"},
ac27a0ec 1249 {Opt_commit, "commit=%u"},
30773840
TT		 1250 {Opt_min_batch_time, "min_batch_time=%u"},
1251 {Opt_max_batch_time, "max_batch_time=%u"},
ac27a0ec 1252 {Opt_journal_dev, "journal_dev=%u"},
818d276c
GS		 1253 {Opt_journal_checksum, "journal_checksum"},
1254 {Opt_journal_async_commit, "journal_async_commit"},
ac27a0ec
DK		 1255 {Opt_abort, "abort"},
1256 {Opt_data_journal, "data=journal"},
1257 {Opt_data_ordered, "data=ordered"},
1258 {Opt_data_writeback, "data=writeback"},
5bf5683a
HK		 1259 {Opt_data_err_abort, "data_err=abort"},
1260 {Opt_data_err_ignore, "data_err=ignore"},
ac27a0ec
DK		 1261 {Opt_offusrjquota, "usrjquota="},
1262 {Opt_usrjquota, "usrjquota=%s"},
1263 {Opt_offgrpjquota, "grpjquota="},
1264 {Opt_grpjquota, "grpjquota=%s"},
1265 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1266 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
5a20bdfc 1267 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
ac27a0ec
DK		 1268 {Opt_grpquota, "grpquota"},
1269 {Opt_noquota, "noquota"},
1270 {Opt_quota, "quota"},
1271 {Opt_usrquota, "usrquota"},
1272 {Opt_barrier, "barrier=%u"},
06705bff
TT		 1273 {Opt_barrier, "barrier"},
1274 {Opt_nobarrier, "nobarrier"},
25ec56b5 1275 {Opt_i_version, "i_version"},
c9de560d 1276 {Opt_stripe, "stripe=%u"},
64769240 1277 {Opt_delalloc, "delalloc"},
dd919b98 1278 {Opt_nodelalloc, "nodelalloc"},
1449032b
TT		 1279 {Opt_mblk_io_submit, "mblk_io_submit"},
1280 {Opt_nomblk_io_submit, "nomblk_io_submit"},
6fd058f7
TT		 1281 {Opt_block_validity, "block_validity"},
1282 {Opt_noblock_validity, "noblock_validity"},
240799cd 1283 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
b3881f74 1284 {Opt_journal_ioprio, "journal_ioprio=%u"},
afd4672d 1285 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
06705bff
TT		 1286 {Opt_auto_da_alloc, "auto_da_alloc"},
1287 {Opt_noauto_da_alloc, "noauto_da_alloc"},
744692dc
JZ		 1288 {Opt_dioread_nolock, "dioread_nolock"},
1289 {Opt_dioread_lock, "dioread_lock"},
5328e635
ES		 1290 {Opt_discard, "discard"},
1291 {Opt_nodiscard, "nodiscard"},
fc6cb1cd
TT		 1292 {Opt_init_itable, "init_itable=%u"},
1293 {Opt_init_itable, "init_itable"},
1294 {Opt_noinit_itable, "noinit_itable"},
c7198b9c
TT		 1295 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1296 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1297 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1298 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1299 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
f3f12faa 1300 {Opt_err, NULL},
ac27a0ec
DK		 1301};
1302
617ba13b 1303static ext4_fsblk_t get_sb_block(void **data)
ac27a0ec 1304{
617ba13b 1305 ext4_fsblk_t sb_block;
ac27a0ec
DK		 1306 char *options = (char *) *data;
1307
1308 if (!options || strncmp(options, "sb=", 3) != 0)
1309 return 1; /* Default location */
0b8e58a1 1310
ac27a0ec 1311 options += 3;
0b8e58a1 1312 /* TODO: use simple_strtoll with >32bit ext4 */
ac27a0ec
DK		 1313 sb_block = simple_strtoul(options, &options, 0);
1314 if (*options && *options != ',') {
4776004f 1315 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
ac27a0ec
DK		 1316 (char *) *data);
1317 return 1;
1318 }
1319 if (*options == ',')
1320 options++;
1321 *data = (void *) options;
0b8e58a1 1322
ac27a0ec
DK		 1323 return sb_block;
1324}
1325
b3881f74 1326#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
437ca0fd
DM		 1327static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1328 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
b3881f74 1329
56c50f11
DM		 1330#ifdef CONFIG_QUOTA
1331static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1332{
1333 struct ext4_sb_info *sbi = EXT4_SB(sb);
1334 char *qname;
1335
1336 if (sb_any_quota_loaded(sb) &&
1337 !sbi->s_qf_names[qtype]) {
1338 ext4_msg(sb, KERN_ERR,
1339 "Cannot change journaled "
1340 "quota options when quota turned on");
57f73c2c 1341 return -1;
56c50f11
DM		 1342 }
1343 qname = match_strdup(args);
1344 if (!qname) {
1345 ext4_msg(sb, KERN_ERR,
1346 "Not enough memory for storing quotafile name");
57f73c2c 1347 return -1;
56c50f11
DM		 1348 }
1349 if (sbi->s_qf_names[qtype] &&
1350 strcmp(sbi->s_qf_names[qtype], qname)) {
1351 ext4_msg(sb, KERN_ERR,
1352 "%s quota file already specified", QTYPE2NAME(qtype));
1353 kfree(qname);
57f73c2c 1354 return -1;
56c50f11
DM		 1355 }
1356 sbi->s_qf_names[qtype] = qname;
1357 if (strchr(sbi->s_qf_names[qtype], '/')) {
1358 ext4_msg(sb, KERN_ERR,
1359 "quotafile must be on filesystem root");
1360 kfree(sbi->s_qf_names[qtype]);
1361 sbi->s_qf_names[qtype] = NULL;
57f73c2c 1362 return -1;
56c50f11 1363 }
fd8c37ec 1364 set_opt(sb, QUOTA);
56c50f11
DM		 1365 return 1;
1366}
1367
1368static int clear_qf_name(struct super_block *sb, int qtype)
1369{
1370
1371 struct ext4_sb_info *sbi = EXT4_SB(sb);
1372
1373 if (sb_any_quota_loaded(sb) &&
1374 sbi->s_qf_names[qtype]) {
1375 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1376 " when quota turned on");
57f73c2c 1377 return -1;
56c50f11
DM		 1378 }
1379 /*
1380 * The space will be released later when all options are confirmed
1381 * to be correct
1382 */
1383 sbi->s_qf_names[qtype] = NULL;
1384 return 1;
1385}
1386#endif
1387
26092bf5
TT		 1388#define MOPT_SET 0x0001
1389#define MOPT_CLEAR 0x0002
1390#define MOPT_NOSUPPORT 0x0004
1391#define MOPT_EXPLICIT 0x0008
1392#define MOPT_CLEAR_ERR 0x0010
1393#define MOPT_GTE0 0x0020
ac27a0ec 1394#ifdef CONFIG_QUOTA
26092bf5
TT		 1395#define MOPT_Q 0
1396#define MOPT_QFMT 0x0040
1397#else
1398#define MOPT_Q MOPT_NOSUPPORT
1399#define MOPT_QFMT MOPT_NOSUPPORT
ac27a0ec 1400#endif
26092bf5
TT		 1401#define MOPT_DATAJ 0x0080
1402
1403static const struct mount_opts {
1404 int token;
1405 int mount_opt;
1406 int flags;
1407} ext4_mount_opts[] = {
1408 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1409 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1410 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1411 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1412 {Opt_mblk_io_submit, EXT4_MOUNT_MBLK_IO_SUBMIT, MOPT_SET},
1413 {Opt_nomblk_io_submit, EXT4_MOUNT_MBLK_IO_SUBMIT, MOPT_CLEAR},
1414 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1415 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1416 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK, MOPT_SET},
1417 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK, MOPT_CLEAR},
1418 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1419 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1420 {Opt_delalloc, EXT4_MOUNT_DELALLOC, MOPT_SET | MOPT_EXPLICIT},
1421 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC, MOPT_CLEAR | MOPT_EXPLICIT},
1422 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM, MOPT_SET},
1423 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1424 EXT4_MOUNT_JOURNAL_CHECKSUM), MOPT_SET},
1425 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_SET},
1426 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1427 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1428 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1429 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT, MOPT_SET},
1430 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT, MOPT_CLEAR},
1431 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1432 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1433 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1434 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1435 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1436 {Opt_commit, 0, MOPT_GTE0},
1437 {Opt_max_batch_time, 0, MOPT_GTE0},
1438 {Opt_min_batch_time, 0, MOPT_GTE0},
1439 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1440 {Opt_init_itable, 0, MOPT_GTE0},
1441 {Opt_stripe, 0, MOPT_GTE0},
1442 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_DATAJ},
1443 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_DATAJ},
1444 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA, MOPT_DATAJ},
03010a33 1445#ifdef CONFIG_EXT4_FS_XATTR
26092bf5
TT		 1446 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1447 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
ac27a0ec 1448#else
26092bf5
TT		 1449 {Opt_user_xattr, 0, MOPT_NOSUPPORT},
1450 {Opt_nouser_xattr, 0, MOPT_NOSUPPORT},
ac27a0ec 1451#endif
03010a33 1452#ifdef CONFIG_EXT4_FS_POSIX_ACL
26092bf5
TT		 1453 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1454 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
ac27a0ec 1455#else
26092bf5
TT		 1456 {Opt_acl, 0, MOPT_NOSUPPORT},
1457 {Opt_noacl, 0, MOPT_NOSUPPORT},
ac27a0ec 1458#endif
26092bf5
TT		 1459 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1460 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1461 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1462 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1463 MOPT_SET | MOPT_Q},
1464 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1465 MOPT_SET | MOPT_Q},
1466 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1467 EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1468 {Opt_usrjquota, 0, MOPT_Q},
1469 {Opt_grpjquota, 0, MOPT_Q},
1470 {Opt_offusrjquota, 0, MOPT_Q},
1471 {Opt_offgrpjquota, 0, MOPT_Q},
1472 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1473 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1474 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1475 {Opt_err, 0, 0}
1476};
1477
1478static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1479 substring_t *args, unsigned long *journal_devnum,
1480 unsigned int *journal_ioprio, int is_remount)
1481{
1482 struct ext4_sb_info *sbi = EXT4_SB(sb);
1483 const struct mount_opts *m;
08cefc7a
EB		 1484 kuid_t uid;
1485 kgid_t gid;
26092bf5
TT		 1486 int arg = 0;
1487
57f73c2c
TT		 1488#ifdef CONFIG_QUOTA
1489 if (token == Opt_usrjquota)
1490 return set_qf_name(sb, USRQUOTA, &args[0]);
1491 else if (token == Opt_grpjquota)
1492 return set_qf_name(sb, GRPQUOTA, &args[0]);
1493 else if (token == Opt_offusrjquota)
1494 return clear_qf_name(sb, USRQUOTA);
1495 else if (token == Opt_offgrpjquota)
1496 return clear_qf_name(sb, GRPQUOTA);
1497#endif
26092bf5
TT		 1498 if (args->from && match_int(args, &arg))
1499 return -1;
1500 switch (token) {
f7048605
TT		 1501 case Opt_noacl:
1502 case Opt_nouser_xattr:
1503 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1504 break;
26092bf5
TT		 1505 case Opt_sb:
1506 return 1; /* handled by get_sb_block() */
1507 case Opt_removed:
1508 ext4_msg(sb, KERN_WARNING,
1509 "Ignoring removed %s option", opt);
1510 return 1;
1511 case Opt_resuid:
08cefc7a
EB		 1512 uid = make_kuid(current_user_ns(), arg);
1513 if (!uid_valid(uid)) {
1514 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1515 return -1;
1516 }
1517 sbi->s_resuid = uid;
26092bf5
TT		 1518 return 1;
1519 case Opt_resgid:
08cefc7a
EB		 1520 gid = make_kgid(current_user_ns(), arg);
1521 if (!gid_valid(gid)) {
1522 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1523 return -1;
1524 }
1525 sbi->s_resgid = gid;
26092bf5
TT		 1526 return 1;
1527 case Opt_abort:
1528 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1529 return 1;
1530 case Opt_i_version:
1531 sb->s_flags |= MS_I_VERSION;
1532 return 1;
1533 case Opt_journal_dev:
1534 if (is_remount) {
1535 ext4_msg(sb, KERN_ERR,
1536 "Cannot specify journal on remount");
1537 return -1;
1538 }
1539 *journal_devnum = arg;
1540 return 1;
1541 case Opt_journal_ioprio:
1542 if (arg < 0 || arg > 7)
1543 return -1;
1544 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1545 return 1;
1546 }
1547
1548 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1549 if (token != m->token)
1550 continue;
1551 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1552 return -1;
1553 if (m->flags & MOPT_EXPLICIT)
1554 set_opt2(sb, EXPLICIT_DELALLOC);
1555 if (m->flags & MOPT_CLEAR_ERR)
1556 clear_opt(sb, ERRORS_MASK);
1557 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1558 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1559 "options when quota turned on");
1560 return -1;
1561 }
1562
1563 if (m->flags & MOPT_NOSUPPORT) {
1564 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1565 } else if (token == Opt_commit) {
1566 if (arg == 0)
1567 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1568 sbi->s_commit_interval = HZ * arg;
1569 } else if (token == Opt_max_batch_time) {
1570 if (arg == 0)
1571 arg = EXT4_DEF_MAX_BATCH_TIME;
1572 sbi->s_max_batch_time = arg;
1573 } else if (token == Opt_min_batch_time) {
1574 sbi->s_min_batch_time = arg;
1575 } else if (token == Opt_inode_readahead_blks) {
1576 if (arg > (1 << 30))
1577 return -1;
1578 if (arg && !is_power_of_2(arg)) {
b31e1552 1579 ext4_msg(sb, KERN_ERR,
26092bf5
TT		 1580 "EXT4-fs: inode_readahead_blks"
1581 " must be a power of 2");
1582 return -1;
ac27a0ec 1583 }
26092bf5
TT		 1584 sbi->s_inode_readahead_blks = arg;
1585 } else if (token == Opt_init_itable) {
1586 set_opt(sb, INIT_INODE_TABLE);
1587 if (!args->from)
1588 arg = EXT4_DEF_LI_WAIT_MULT;
1589 sbi->s_li_wait_mult = arg;
1590 } else if (token == Opt_stripe) {
1591 sbi->s_stripe = arg;
1592 } else if (m->flags & MOPT_DATAJ) {
ac27a0ec 1593 if (is_remount) {
eb513689
TT		 1594 if (!sbi->s_journal)
1595 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
26092bf5
TT		 1596 else if (test_opt(sb, DATA_FLAGS) !=
1597 m->mount_opt) {
b31e1552 1598 ext4_msg(sb, KERN_ERR,
26092bf5
TT		 1599 "Cannot change data mode on remount");
1600 return -1;
ac27a0ec
DK		 1601 }
1602 } else {
fd8c37ec 1603 clear_opt(sb, DATA_FLAGS);
26092bf5 1604 sbi->s_mount_opt |= m->mount_opt;
ac27a0ec 1605 }
ac27a0ec 1606#ifdef CONFIG_QUOTA
26092bf5 1607 } else if (m->flags & MOPT_QFMT) {
17bd13b3 1608 if (sb_any_quota_loaded(sb) &&
26092bf5
TT		 1609 sbi->s_jquota_fmt != m->mount_opt) {
1610 ext4_msg(sb, KERN_ERR, "Cannot "
1611 "change journaled quota options "
1612 "when quota turned on");
1613 return -1;
ac27a0ec 1614 }
26092bf5 1615 sbi->s_jquota_fmt = m->mount_opt;
ac27a0ec 1616#endif
26092bf5
TT		 1617 } else {
1618 if (!args->from)
1619 arg = 1;
1620 if (m->flags & MOPT_CLEAR)
1621 arg = !arg;
1622 else if (unlikely(!(m->flags & MOPT_SET))) {
1623 ext4_msg(sb, KERN_WARNING,
1624 "buggy handling of option %s", opt);
1625 WARN_ON(1);
1626 return -1;
3197ebdb 1627 }
26092bf5
TT		 1628 if (arg != 0)
1629 sbi->s_mount_opt |= m->mount_opt;
afd4672d 1630 else
26092bf5 1631 sbi->s_mount_opt &= ~m->mount_opt;
ac27a0ec 1632 }
26092bf5
TT		 1633 return 1;
1634 }
1635 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1636 "or missing value", opt);
1637 return -1;
1638}
1639
1640static int parse_options(char *options, struct super_block *sb,
1641 unsigned long *journal_devnum,
1642 unsigned int *journal_ioprio,
1643 int is_remount)
1644{
db7e5c66 1645#ifdef CONFIG_QUOTA
26092bf5 1646 struct ext4_sb_info *sbi = EXT4_SB(sb);
db7e5c66 1647#endif
26092bf5
TT		 1648 char *p;
1649 substring_t args[MAX_OPT_ARGS];
1650 int token;
1651
1652 if (!options)
1653 return 1;
1654
1655 while ((p = strsep(&options, ",")) != NULL) {
1656 if (!*p)
1657 continue;
1658 /*
1659 * Initialize args struct so we know whether arg was
1660 * found; some options take optional arguments.
1661 */
1662 args[0].to = args[0].from = 0;
1663 token = match_token(p, tokens, args);
1664 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1665 journal_ioprio, is_remount) < 0)
1666 return 0;
ac27a0ec
DK		 1667 }
1668#ifdef CONFIG_QUOTA
1669 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
482a7425 1670 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
fd8c37ec 1671 clear_opt(sb, USRQUOTA);
ac27a0ec 1672
482a7425 1673 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
fd8c37ec 1674 clear_opt(sb, GRPQUOTA);
ac27a0ec 1675
56c50f11 1676 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
b31e1552
ES		 1677 ext4_msg(sb, KERN_ERR, "old and new quota "
1678 "format mixing");
ac27a0ec
DK		 1679 return 0;
1680 }
1681
1682 if (!sbi->s_jquota_fmt) {
b31e1552
ES		 1683 ext4_msg(sb, KERN_ERR, "journaled quota format "
1684 "not specified");
ac27a0ec
DK		 1685 return 0;
1686 }
1687 } else {
1688 if (sbi->s_jquota_fmt) {
b31e1552 1689 ext4_msg(sb, KERN_ERR, "journaled quota format "
2c8be6b2 1690 "specified with no journaling "
b31e1552 1691 "enabled");
ac27a0ec
DK		 1692 return 0;
1693 }
1694 }
1695#endif
1696 return 1;
1697}
1698
2adf6da8
TT		 1699static inline void ext4_show_quota_options(struct seq_file *seq,
1700 struct super_block *sb)
1701{
1702#if defined(CONFIG_QUOTA)
1703 struct ext4_sb_info *sbi = EXT4_SB(sb);
1704
1705 if (sbi->s_jquota_fmt) {
1706 char *fmtname = "";
1707
1708 switch (sbi->s_jquota_fmt) {
1709 case QFMT_VFS_OLD:
1710 fmtname = "vfsold";
1711 break;
1712 case QFMT_VFS_V0:
1713 fmtname = "vfsv0";
1714 break;
1715 case QFMT_VFS_V1:
1716 fmtname = "vfsv1";
1717 break;
1718 }
1719 seq_printf(seq, ",jqfmt=%s", fmtname);
1720 }
1721
1722 if (sbi->s_qf_names[USRQUOTA])
1723 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1724
1725 if (sbi->s_qf_names[GRPQUOTA])
1726 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
1727
1728 if (test_opt(sb, USRQUOTA))
1729 seq_puts(seq, ",usrquota");
1730
1731 if (test_opt(sb, GRPQUOTA))
1732 seq_puts(seq, ",grpquota");
1733#endif
1734}
1735
5a916be1
TT		 1736static const char *token2str(int token)
1737{
1738 static const struct match_token *t;
1739
1740 for (t = tokens; t->token != Opt_err; t++)
1741 if (t->token == token && !strchr(t->pattern, '='))
1742 break;
1743 return t->pattern;
1744}
1745
2adf6da8
TT		 1746/*
1747 * Show an option if
1748 * - it's set to a non-default value OR
1749 * - if the per-sb default is different from the global default
1750 */
66acdcf4
TT		 1751static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1752 int nodefs)
2adf6da8 1753{
2adf6da8
TT		 1754 struct ext4_sb_info *sbi = EXT4_SB(sb);
1755 struct ext4_super_block *es = sbi->s_es;
66acdcf4 1756 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
5a916be1 1757 const struct mount_opts *m;
66acdcf4 1758 char sep = nodefs ? '\n' : ',';
2adf6da8 1759
66acdcf4
TT		 1760#define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1761#define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2adf6da8
TT		 1762
1763 if (sbi->s_sb_block != 1)
5a916be1
TT		 1764 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1765
1766 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1767 int want_set = m->flags & MOPT_SET;
1768 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1769 (m->flags & MOPT_CLEAR_ERR))
1770 continue;
66acdcf4 1771 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
5a916be1
TT		 1772 continue; /* skip if same as the default */
1773 if ((want_set &&
1774 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1775 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1776 continue; /* select Opt_noFoo vs Opt_Foo */
1777 SEQ_OPTS_PRINT("%s", token2str(m->token));
2adf6da8 1778 }
5a916be1 1779
08cefc7a 1780 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
5a916be1 1781 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
08cefc7a
EB		 1782 SEQ_OPTS_PRINT("resuid=%u",
1783 from_kuid_munged(&init_user_ns, sbi->s_resuid));
1784 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
5a916be1 1785 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
08cefc7a
EB		 1786 SEQ_OPTS_PRINT("resgid=%u",
1787 from_kgid_munged(&init_user_ns, sbi->s_resgid));
66acdcf4 1788 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
5a916be1
TT		 1789 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1790 SEQ_OPTS_PUTS("errors=remount-ro");
2adf6da8 1791 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
5a916be1 1792 SEQ_OPTS_PUTS("errors=continue");
2adf6da8 1793 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
5a916be1 1794 SEQ_OPTS_PUTS("errors=panic");
66acdcf4 1795 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
5a916be1 1796 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
66acdcf4 1797 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
5a916be1 1798 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
66acdcf4 1799 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
5a916be1 1800 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2adf6da8 1801 if (sb->s_flags & MS_I_VERSION)
5a916be1 1802 SEQ_OPTS_PUTS("i_version");
66acdcf4 1803 if (nodefs || sbi->s_stripe)
5a916be1 1804 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
66acdcf4 1805 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
5a916be1
TT		 1806 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1807 SEQ_OPTS_PUTS("data=journal");
1808 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1809 SEQ_OPTS_PUTS("data=ordered");
1810 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1811 SEQ_OPTS_PUTS("data=writeback");
1812 }
66acdcf4
TT		 1813 if (nodefs ||
1814 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
5a916be1
TT		 1815 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1816 sbi->s_inode_readahead_blks);
2adf6da8 1817
66acdcf4
TT		 1818 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1819 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
5a916be1 1820 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
2adf6da8
TT		 1821
1822 ext4_show_quota_options(seq, sb);
2adf6da8
TT		 1823 return 0;
1824}
1825
66acdcf4
TT		 1826static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1827{
1828 return _ext4_show_options(seq, root->d_sb, 0);
1829}
1830
1831static int options_seq_show(struct seq_file *seq, void *offset)
1832{
1833 struct super_block *sb = seq->private;
1834 int rc;
1835
1836 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1837 rc = _ext4_show_options(seq, sb, 1);
1838 seq_puts(seq, "\n");
1839 return rc;
1840}
1841
1842static int options_open_fs(struct inode *inode, struct file *file)
1843{
1844 return single_open(file, options_seq_show, PDE(inode)->data);
1845}
1846
1847static const struct file_operations ext4_seq_options_fops = {
1848 .owner = THIS_MODULE,
1849 .open = options_open_fs,
1850 .read = seq_read,
1851 .llseek = seq_lseek,
1852 .release = single_release,
1853};
1854
617ba13b 1855static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 1856 int read_only)
1857{
617ba13b 1858 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 1859 int res = 0;
1860
617ba13b 1861 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
b31e1552
ES		 1862 ext4_msg(sb, KERN_ERR, "revision level too high, "
1863 "forcing read-only mode");
ac27a0ec
DK		 1864 res = MS_RDONLY;
1865 }
1866 if (read_only)
281b5995 1867 goto done;
617ba13b 1868 if (!(sbi->s_mount_state & EXT4_VALID_FS))
b31e1552
ES		 1869 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1870 "running e2fsck is recommended");
617ba13b 1871 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
b31e1552
ES		 1872 ext4_msg(sb, KERN_WARNING,
1873 "warning: mounting fs with errors, "
1874 "running e2fsck is recommended");
ed3ce80a 1875 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
ac27a0ec
DK		 1876 le16_to_cpu(es->s_mnt_count) >=
1877 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
b31e1552
ES		 1878 ext4_msg(sb, KERN_WARNING,
1879 "warning: maximal mount count reached, "
1880 "running e2fsck is recommended");
ac27a0ec
DK		 1881 else if (le32_to_cpu(es->s_checkinterval) &&
1882 (le32_to_cpu(es->s_lastcheck) +
1883 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
b31e1552
ES		 1884 ext4_msg(sb, KERN_WARNING,
1885 "warning: checktime reached, "
1886 "running e2fsck is recommended");
0b8e58a1 1887 if (!sbi->s_journal)
0390131b 1888 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec 1889 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 1890 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 1891 le16_add_cpu(&es->s_mnt_count, 1);
ac27a0ec 1892 es->s_mtime = cpu_to_le32(get_seconds());
617ba13b 1893 ext4_update_dynamic_rev(sb);
0390131b
FM		 1894 if (sbi->s_journal)
1895 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 1896
e2d67052 1897 ext4_commit_super(sb, 1);
281b5995 1898done:
ac27a0ec 1899 if (test_opt(sb, DEBUG))
a9df9a49 1900 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
a2595b8a 1901 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
ac27a0ec
DK		 1902 sb->s_blocksize,
1903 sbi->s_groups_count,
617ba13b
MC		 1904 EXT4_BLOCKS_PER_GROUP(sb),
1905 EXT4_INODES_PER_GROUP(sb),
a2595b8a 1906 sbi->s_mount_opt, sbi->s_mount_opt2);
ac27a0ec 1907
7abc52c2 1908 cleancache_init_fs(sb);
ac27a0ec
DK		 1909 return res;
1910}
1911
772cb7c8
JS		 1912static int ext4_fill_flex_info(struct super_block *sb)
1913{
1914 struct ext4_sb_info *sbi = EXT4_SB(sb);
1915 struct ext4_group_desc *gdp = NULL;
772cb7c8
JS		 1916 ext4_group_t flex_group_count;
1917 ext4_group_t flex_group;
d50f2ab6 1918 unsigned int groups_per_flex = 0;
c5ca7c76 1919 size_t size;
772cb7c8
JS		 1920 int i;
1921
503358ae 1922 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
d50f2ab6 1923 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
772cb7c8
JS		 1924 sbi->s_log_groups_per_flex = 0;
1925 return 1;
1926 }
d50f2ab6 1927 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
772cb7c8 1928
c62a11fd
FB		 1929 /* We allocate both existing and potentially added groups */
1930 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
d94e99a6
AK		 1931 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1932 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
c5ca7c76 1933 size = flex_group_count * sizeof(struct flex_groups);
9933fc0a 1934 sbi->s_flex_groups = ext4_kvzalloc(size, GFP_KERNEL);
c5ca7c76 1935 if (sbi->s_flex_groups == NULL) {
9933fc0a
TT		 1936 ext4_msg(sb, KERN_ERR, "not enough memory for %u flex groups",
1937 flex_group_count);
1938 goto failed;
772cb7c8 1939 }
772cb7c8 1940
772cb7c8 1941 for (i = 0; i < sbi->s_groups_count; i++) {
88b6edd1 1942 gdp = ext4_get_group_desc(sb, i, NULL);
772cb7c8
JS		 1943
1944 flex_group = ext4_flex_group(sbi, i);
7ad9bb65
TT		 1945 atomic_add(ext4_free_inodes_count(sb, gdp),
1946 &sbi->s_flex_groups[flex_group].free_inodes);
021b65bb 1947 atomic_add(ext4_free_group_clusters(sb, gdp),
24aaa8ef 1948 &sbi->s_flex_groups[flex_group].free_clusters);
7ad9bb65
TT		 1949 atomic_add(ext4_used_dirs_count(sb, gdp),
1950 &sbi->s_flex_groups[flex_group].used_dirs);
772cb7c8
JS		 1951 }
1952
1953 return 1;
1954failed:
1955 return 0;
1956}
1957
feb0ab32
DW		 1958static __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1959 struct ext4_group_desc *gdp)
717d50e4 1960{
feb0ab32 1961 int offset;
717d50e4 1962 __u16 crc = 0;
feb0ab32 1963 __le32 le_group = cpu_to_le32(block_group);
717d50e4 1964
feb0ab32
DW		 1965 if ((sbi->s_es->s_feature_ro_compat &
1966 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))) {
1967 /* Use new metadata_csum algorithm */
1968 __u16 old_csum;
1969 __u32 csum32;
1970
1971 old_csum = gdp->bg_checksum;
1972 gdp->bg_checksum = 0;
1973 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
1974 sizeof(le_group));
1975 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp,
1976 sbi->s_desc_size);
1977 gdp->bg_checksum = old_csum;
1978
1979 crc = csum32 & 0xFFFF;
1980 goto out;
717d50e4
AD		 1981 }
1982
feb0ab32
DW		 1983 /* old crc16 code */
1984 offset = offsetof(struct ext4_group_desc, bg_checksum);
1985
1986 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1987 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1988 crc = crc16(crc, (__u8 *)gdp, offset);
1989 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1990 /* for checksum of struct ext4_group_desc do the rest...*/
1991 if ((sbi->s_es->s_feature_incompat &
1992 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1993 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1994 crc = crc16(crc, (__u8 *)gdp + offset,
1995 le16_to_cpu(sbi->s_es->s_desc_size) -
1996 offset);
1997
1998out:
717d50e4
AD		 1999 return cpu_to_le16(crc);
2000}
2001
feb0ab32 2002int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
717d50e4
AD		 2003 struct ext4_group_desc *gdp)
2004{
feb0ab32
DW		 2005 if (ext4_has_group_desc_csum(sb) &&
2006 (gdp->bg_checksum != ext4_group_desc_csum(EXT4_SB(sb),
2007 block_group, gdp)))
717d50e4
AD		 2008 return 0;
2009
2010 return 1;
2011}
2012
feb0ab32
DW		 2013void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2014 struct ext4_group_desc *gdp)
2015{
2016 if (!ext4_has_group_desc_csum(sb))
2017 return;
2018 gdp->bg_checksum = ext4_group_desc_csum(EXT4_SB(sb), block_group, gdp);
2019}
2020
ac27a0ec 2021/* Called at mount-time, super-block is locked */
bfff6873
LC		 2022static int ext4_check_descriptors(struct super_block *sb,
2023 ext4_group_t *first_not_zeroed)
ac27a0ec 2024{
617ba13b
MC		 2025 struct ext4_sb_info *sbi = EXT4_SB(sb);
2026 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2027 ext4_fsblk_t last_block;
bd81d8ee
LV		 2028 ext4_fsblk_t block_bitmap;
2029 ext4_fsblk_t inode_bitmap;
2030 ext4_fsblk_t inode_table;
ce421581 2031 int flexbg_flag = 0;
bfff6873 2032 ext4_group_t i, grp = sbi->s_groups_count;
ac27a0ec 2033
ce421581
JS		 2034 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2035 flexbg_flag = 1;
2036
af5bc92d 2037 ext4_debug("Checking group descriptors");
ac27a0ec 2038
197cd65a
AM		 2039 for (i = 0; i < sbi->s_groups_count; i++) {
2040 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2041
ce421581 2042 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 2043 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 2044 else
2045 last_block = first_block +
617ba13b 2046 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 2047
bfff6873
LC		 2048 if ((grp == sbi->s_groups_count) &&
2049 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2050 grp = i;
2051
8fadc143 2052 block_bitmap = ext4_block_bitmap(sb, gdp);
2b2d6d01 2053 if (block_bitmap < first_block || block_bitmap > last_block) {
b31e1552 2054 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2055 "Block bitmap for group %u not in group "
b31e1552 2056 "(block %llu)!", i, block_bitmap);
ac27a0ec
DK		 2057 return 0;
2058 }
8fadc143 2059 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2b2d6d01 2060 if (inode_bitmap < first_block || inode_bitmap > last_block) {
b31e1552 2061 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2062 "Inode bitmap for group %u not in group "
b31e1552 2063 "(block %llu)!", i, inode_bitmap);
ac27a0ec
DK		 2064 return 0;
2065 }
8fadc143 2066 inode_table = ext4_inode_table(sb, gdp);
bd81d8ee 2067 if (inode_table < first_block ||
2b2d6d01 2068 inode_table + sbi->s_itb_per_group - 1 > last_block) {
b31e1552 2069 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2070 "Inode table for group %u not in group "
b31e1552 2071 "(block %llu)!", i, inode_table);
ac27a0ec
DK		 2072 return 0;
2073 }
955ce5f5 2074 ext4_lock_group(sb, i);
feb0ab32 2075 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
b31e1552
ES		 2076 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2077 "Checksum for group %u failed (%u!=%u)",
2078 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2079 gdp)), le16_to_cpu(gdp->bg_checksum));
7ee1ec4c 2080 if (!(sb->s_flags & MS_RDONLY)) {
955ce5f5 2081 ext4_unlock_group(sb, i);
8a266467 2082 return 0;
7ee1ec4c 2083 }
717d50e4 2084 }
955ce5f5 2085 ext4_unlock_group(sb, i);
ce421581
JS		 2086 if (!flexbg_flag)
2087 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec 2088 }
bfff6873
LC		 2089 if (NULL != first_not_zeroed)
2090 *first_not_zeroed = grp;
ac27a0ec 2091
5dee5437
TT		 2092 ext4_free_blocks_count_set(sbi->s_es,
2093 EXT4_C2B(sbi, ext4_count_free_clusters(sb)));
0b8e58a1 2094 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
ac27a0ec
DK		 2095 return 1;
2096}
2097
617ba13b 2098/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 2099 * the superblock) which were deleted from all directories, but held open by
2100 * a process at the time of a crash. We walk the list and try to delete these
2101 * inodes at recovery time (only with a read-write filesystem).
2102 *
2103 * In order to keep the orphan inode chain consistent during traversal (in
2104 * case of crash during recovery), we link each inode into the superblock
2105 * orphan list_head and handle it the same way as an inode deletion during
2106 * normal operation (which journals the operations for us).
2107 *
2108 * We only do an iget() and an iput() on each inode, which is very safe if we
2109 * accidentally point at an in-use or already deleted inode. The worst that
2110 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 2111 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 2112 * e2fsck was run on this filesystem, and it must have already done the orphan
2113 * inode cleanup for us, so we can safely abort without any further action.
2114 */
2b2d6d01
TT		 2115static void ext4_orphan_cleanup(struct super_block *sb,
2116 struct ext4_super_block *es)
ac27a0ec
DK		 2117{
2118 unsigned int s_flags = sb->s_flags;
2119 int nr_orphans = 0, nr_truncates = 0;
2120#ifdef CONFIG_QUOTA
2121 int i;
2122#endif
2123 if (!es->s_last_orphan) {
2124 jbd_debug(4, "no orphan inodes to clean up\n");
2125 return;
2126 }
2127
a8f48a95 2128 if (bdev_read_only(sb->s_bdev)) {
b31e1552
ES		 2129 ext4_msg(sb, KERN_ERR, "write access "
2130 "unavailable, skipping orphan cleanup");
a8f48a95
ES		 2131 return;
2132 }
2133
d39195c3
AG		 2134 /* Check if feature set would not allow a r/w mount */
2135 if (!ext4_feature_set_ok(sb, 0)) {
2136 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2137 "unknown ROCOMPAT features");
2138 return;
2139 }
2140
617ba13b 2141 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
ac27a0ec
DK		 2142 if (es->s_last_orphan)
2143 jbd_debug(1, "Errors on filesystem, "
2144 "clearing orphan list.\n");
2145 es->s_last_orphan = 0;
2146 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2147 return;
2148 }
2149
2150 if (s_flags & MS_RDONLY) {
b31e1552 2151 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
ac27a0ec
DK		 2152 sb->s_flags &= ~MS_RDONLY;
2153 }
2154#ifdef CONFIG_QUOTA
2155 /* Needed for iput() to work correctly and not trash data */
2156 sb->s_flags |= MS_ACTIVE;
2157 /* Turn on quotas so that they are updated correctly */
2158 for (i = 0; i < MAXQUOTAS; i++) {
617ba13b
MC		 2159 if (EXT4_SB(sb)->s_qf_names[i]) {
2160 int ret = ext4_quota_on_mount(sb, i);
ac27a0ec 2161 if (ret < 0)
b31e1552
ES		 2162 ext4_msg(sb, KERN_ERR,
2163 "Cannot turn on journaled "
2164 "quota: error %d", ret);
ac27a0ec
DK		 2165 }
2166 }
2167#endif
2168
2169 while (es->s_last_orphan) {
2170 struct inode *inode;
2171
97bd42b9
JB		 2172 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2173 if (IS_ERR(inode)) {
ac27a0ec
DK		 2174 es->s_last_orphan = 0;
2175 break;
2176 }
2177
617ba13b 2178 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
871a2931 2179 dquot_initialize(inode);
ac27a0ec 2180 if (inode->i_nlink) {
b31e1552
ES		 2181 ext4_msg(sb, KERN_DEBUG,
2182 "%s: truncating inode %lu to %lld bytes",
46e665e9 2183 __func__, inode->i_ino, inode->i_size);
e5f8eab8 2184 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
ac27a0ec 2185 inode->i_ino, inode->i_size);
617ba13b 2186 ext4_truncate(inode);
ac27a0ec
DK		 2187 nr_truncates++;
2188 } else {
b31e1552
ES		 2189 ext4_msg(sb, KERN_DEBUG,
2190 "%s: deleting unreferenced inode %lu",
46e665e9 2191 __func__, inode->i_ino);
ac27a0ec
DK		 2192 jbd_debug(2, "deleting unreferenced inode %lu\n",
2193 inode->i_ino);
2194 nr_orphans++;
2195 }
2196 iput(inode); /* The delete magic happens here! */
2197 }
2198
2b2d6d01 2199#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
ac27a0ec
DK		 2200
2201 if (nr_orphans)
b31e1552
ES		 2202 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2203 PLURAL(nr_orphans));
ac27a0ec 2204 if (nr_truncates)
b31e1552
ES		 2205 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2206 PLURAL(nr_truncates));
ac27a0ec
DK		 2207#ifdef CONFIG_QUOTA
2208 /* Turn quotas off */
2209 for (i = 0; i < MAXQUOTAS; i++) {
2210 if (sb_dqopt(sb)->files[i])
287a8095 2211 dquot_quota_off(sb, i);
ac27a0ec
DK		 2212 }
2213#endif
2214 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2215}
0b8e58a1 2216
cd2291a4
ES		 2217/*
2218 * Maximal extent format file size.
2219 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2220 * extent format containers, within a sector_t, and within i_blocks
2221 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2222 * so that won't be a limiting factor.
2223 *
f17722f9
LC		 2224 * However there is other limiting factor. We do store extents in the form
2225 * of starting block and length, hence the resulting length of the extent
2226 * covering maximum file size must fit into on-disk format containers as
2227 * well. Given that length is always by 1 unit bigger than max unit (because
2228 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2229 *
cd2291a4
ES		 2230 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2231 */
f287a1a5 2232static loff_t ext4_max_size(int blkbits, int has_huge_files)
cd2291a4
ES		 2233{
2234 loff_t res;
2235 loff_t upper_limit = MAX_LFS_FILESIZE;
2236
2237 /* small i_blocks in vfs inode? */
f287a1a5 2238 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
cd2291a4 2239 /*
90c699a9 2240 * CONFIG_LBDAF is not enabled implies the inode
cd2291a4
ES		 2241 * i_block represent total blocks in 512 bytes
2242 * 32 == size of vfs inode i_blocks * 8
2243 */
2244 upper_limit = (1LL << 32) - 1;
2245
2246 /* total blocks in file system block size */
2247 upper_limit >>= (blkbits - 9);
2248 upper_limit <<= blkbits;
2249 }
2250
f17722f9
LC		 2251 /*
2252 * 32-bit extent-start container, ee_block. We lower the maxbytes
2253 * by one fs block, so ee_len can cover the extent of maximum file
2254 * size
2255 */
2256 res = (1LL << 32) - 1;
cd2291a4 2257 res <<= blkbits;
cd2291a4
ES		 2258
2259 /* Sanity check against vm- & vfs- imposed limits */
2260 if (res > upper_limit)
2261 res = upper_limit;
2262
2263 return res;
2264}
ac27a0ec 2265
ac27a0ec 2266/*
cd2291a4 2267 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 2268 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2269 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 2270 */
f287a1a5 2271static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
ac27a0ec 2272{
617ba13b 2273 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 2274 int meta_blocks;
2275 loff_t upper_limit;
0b8e58a1
AD		 2276 /* This is calculated to be the largest file size for a dense, block
2277 * mapped file such that the file's total number of 512-byte sectors,
2278 * including data and all indirect blocks, does not exceed (2^48 - 1).
2279 *
2280 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2281 * number of 512-byte sectors of the file.
0fc1b451
AK		 2282 */
2283
f287a1a5 2284 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
0fc1b451 2285 /*
90c699a9 2286 * !has_huge_files or CONFIG_LBDAF not enabled implies that
0b8e58a1
AD		 2287 * the inode i_block field represents total file blocks in
2288 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
0fc1b451
AK		 2289 */
2290 upper_limit = (1LL << 32) - 1;
2291
2292 /* total blocks in file system block size */
2293 upper_limit >>= (bits - 9);
2294
2295 } else {
8180a562
AK		 2296 /*
2297 * We use 48 bit ext4_inode i_blocks
2298 * With EXT4_HUGE_FILE_FL set the i_blocks
2299 * represent total number of blocks in
2300 * file system block size
2301 */
0fc1b451
AK		 2302 upper_limit = (1LL << 48) - 1;
2303
0fc1b451
AK		 2304 }
2305
2306 /* indirect blocks */
2307 meta_blocks = 1;
2308 /* double indirect blocks */
2309 meta_blocks += 1 + (1LL << (bits-2));
2310 /* tripple indirect blocks */
2311 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2312
2313 upper_limit -= meta_blocks;
2314 upper_limit <<= bits;
ac27a0ec
DK		 2315
2316 res += 1LL << (bits-2);
2317 res += 1LL << (2*(bits-2));
2318 res += 1LL << (3*(bits-2));
2319 res <<= bits;
2320 if (res > upper_limit)
2321 res = upper_limit;
0fc1b451
AK		 2322
2323 if (res > MAX_LFS_FILESIZE)
2324 res = MAX_LFS_FILESIZE;
2325
ac27a0ec
DK		 2326 return res;
2327}
2328
617ba13b 2329static ext4_fsblk_t descriptor_loc(struct super_block *sb,
0b8e58a1 2330 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 2331{
617ba13b 2332 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 2333 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 2334 int has_super = 0;
2335
2336 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2337
617ba13b 2338 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
ac27a0ec 2339 nr < first_meta_bg)
70bbb3e0 2340 return logical_sb_block + nr + 1;
ac27a0ec 2341 bg = sbi->s_desc_per_block * nr;
617ba13b 2342 if (ext4_bg_has_super(sb, bg))
ac27a0ec 2343 has_super = 1;
0b8e58a1 2344
617ba13b 2345 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 2346}
2347
c9de560d
AT		 2348/**
2349 * ext4_get_stripe_size: Get the stripe size.
2350 * @sbi: In memory super block info
2351 *
2352 * If we have specified it via mount option, then
2353 * use the mount option value. If the value specified at mount time is
2354 * greater than the blocks per group use the super block value.
2355 * If the super block value is greater than blocks per group return 0.
2356 * Allocator needs it be less than blocks per group.
2357 *
2358 */
2359static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2360{
2361 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2362 unsigned long stripe_width =
2363 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3eb08658 2364 int ret;
c9de560d
AT		 2365
2366 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3eb08658
DE		 2367 ret = sbi->s_stripe;
2368 else if (stripe_width <= sbi->s_blocks_per_group)
2369 ret = stripe_width;
2370 else if (stride <= sbi->s_blocks_per_group)
2371 ret = stride;
2372 else
2373 ret = 0;
c9de560d 2374
3eb08658
DE		 2375 /*
2376 * If the stripe width is 1, this makes no sense and
2377 * we set it to 0 to turn off stripe handling code.
2378 */
2379 if (ret <= 1)
2380 ret = 0;
c9de560d 2381
3eb08658 2382 return ret;
c9de560d 2383}
ac27a0ec 2384
3197ebdb
TT		 2385/* sysfs supprt */
2386
2387struct ext4_attr {
2388 struct attribute attr;
2389 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
60e6679e 2390 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
3197ebdb
TT		 2391 const char *, size_t);
2392 int offset;
2393};
2394
2395static int parse_strtoul(const char *buf,
2396 unsigned long max, unsigned long *value)
2397{
2398 char *endp;
2399
e7d2860b
AGR		 2400 *value = simple_strtoul(skip_spaces(buf), &endp, 0);
2401 endp = skip_spaces(endp);
3197ebdb
TT		 2402 if (*endp || *value > max)
2403 return -EINVAL;
2404
2405 return 0;
2406}
2407
2408static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2409 struct ext4_sb_info *sbi,
2410 char *buf)
2411{
2412 return snprintf(buf, PAGE_SIZE, "%llu\n",
7b415bf6
AK		 2413 (s64) EXT4_C2B(sbi,
2414 percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
3197ebdb
TT		 2415}
2416
2417static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2418 struct ext4_sb_info *sbi, char *buf)
2419{
2420 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2421
f613dfcb
TT		 2422 if (!sb->s_bdev->bd_part)
2423 return snprintf(buf, PAGE_SIZE, "0\n");
3197ebdb
TT		 2424 return snprintf(buf, PAGE_SIZE, "%lu\n",
2425 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2426 sbi->s_sectors_written_start) >> 1);
2427}
2428
2429static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2430 struct ext4_sb_info *sbi, char *buf)
2431{
2432 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2433
f613dfcb
TT		 2434 if (!sb->s_bdev->bd_part)
2435 return snprintf(buf, PAGE_SIZE, "0\n");
3197ebdb 2436 return snprintf(buf, PAGE_SIZE, "%llu\n",
a6b43e38 2437 (unsigned long long)(sbi->s_kbytes_written +
3197ebdb 2438 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
a6b43e38 2439 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
3197ebdb
TT		 2440}
2441
2442static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2443 struct ext4_sb_info *sbi,
2444 const char *buf, size_t count)
2445{
2446 unsigned long t;
2447
2448 if (parse_strtoul(buf, 0x40000000, &t))
2449 return -EINVAL;
2450
5dbd571d 2451 if (t && !is_power_of_2(t))
3197ebdb
TT		 2452 return -EINVAL;
2453
2454 sbi->s_inode_readahead_blks = t;
2455 return count;
2456}
2457
2458static ssize_t sbi_ui_show(struct ext4_attr *a,
0b8e58a1 2459 struct ext4_sb_info *sbi, char *buf)
3197ebdb
TT		 2460{
2461 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2462
2463 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2464}
2465
2466static ssize_t sbi_ui_store(struct ext4_attr *a,
2467 struct ext4_sb_info *sbi,
2468 const char *buf, size_t count)
2469{
2470 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2471 unsigned long t;
2472
2473 if (parse_strtoul(buf, 0xffffffff, &t))
2474 return -EINVAL;
2475 *ui = t;
2476 return count;
2477}
2478
2c0544b2
TT		 2479static ssize_t trigger_test_error(struct ext4_attr *a,
2480 struct ext4_sb_info *sbi,
2481 const char *buf, size_t count)
2482{
2483 int len = count;
2484
2485 if (!capable(CAP_SYS_ADMIN))
2486 return -EPERM;
2487
2488 if (len && buf[len-1] == '\n')
2489 len--;
2490
2491 if (len)
2492 ext4_error(sbi->s_sb, "%.*s", len, buf);
2493 return count;
2494}
2495
3197ebdb
TT		 2496#define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2497static struct ext4_attr ext4_attr_##_name = { \
2498 .attr = {.name = __stringify(_name), .mode = _mode }, \
2499 .show = _show, \
2500 .store = _store, \
2501 .offset = offsetof(struct ext4_sb_info, _elname), \
2502}
2503#define EXT4_ATTR(name, mode, show, store) \
2504static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2505
857ac889 2506#define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
3197ebdb
TT		 2507#define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2508#define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2509#define EXT4_RW_ATTR_SBI_UI(name, elname) \
2510 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2511#define ATTR_LIST(name) &ext4_attr_##name.attr
2512
2513EXT4_RO_ATTR(delayed_allocation_blocks);
2514EXT4_RO_ATTR(session_write_kbytes);
2515EXT4_RO_ATTR(lifetime_write_kbytes);
2516EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2517 inode_readahead_blks_store, s_inode_readahead_blks);
11013911 2518EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
3197ebdb
TT		 2519EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2520EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2521EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2522EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2523EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2524EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
55138e0b 2525EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
2c0544b2 2526EXT4_ATTR(trigger_fs_error, 0200, NULL, trigger_test_error);
3197ebdb
TT		 2527
2528static struct attribute *ext4_attrs[] = {
2529 ATTR_LIST(delayed_allocation_blocks),
2530 ATTR_LIST(session_write_kbytes),
2531 ATTR_LIST(lifetime_write_kbytes),
2532 ATTR_LIST(inode_readahead_blks),
11013911 2533 ATTR_LIST(inode_goal),
3197ebdb
TT		 2534 ATTR_LIST(mb_stats),
2535 ATTR_LIST(mb_max_to_scan),
2536 ATTR_LIST(mb_min_to_scan),
2537 ATTR_LIST(mb_order2_req),
2538 ATTR_LIST(mb_stream_req),
2539 ATTR_LIST(mb_group_prealloc),
55138e0b 2540 ATTR_LIST(max_writeback_mb_bump),
2c0544b2 2541 ATTR_LIST(trigger_fs_error),
3197ebdb
TT		 2542 NULL,
2543};
2544
857ac889
LC		 2545/* Features this copy of ext4 supports */
2546EXT4_INFO_ATTR(lazy_itable_init);
27ee40df 2547EXT4_INFO_ATTR(batched_discard);
857ac889
LC		 2548
2549static struct attribute *ext4_feat_attrs[] = {
2550 ATTR_LIST(lazy_itable_init),
27ee40df 2551 ATTR_LIST(batched_discard),
857ac889
LC		 2552 NULL,
2553};
2554
3197ebdb
TT		 2555static ssize_t ext4_attr_show(struct kobject *kobj,
2556 struct attribute *attr, char *buf)
2557{
2558 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2559 s_kobj);
2560 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2561
2562 return a->show ? a->show(a, sbi, buf) : 0;
2563}
2564
2565static ssize_t ext4_attr_store(struct kobject *kobj,
2566 struct attribute *attr,
2567 const char *buf, size_t len)
2568{
2569 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2570 s_kobj);
2571 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2572
2573 return a->store ? a->store(a, sbi, buf, len) : 0;
2574}
2575
2576static void ext4_sb_release(struct kobject *kobj)
2577{
2578 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2579 s_kobj);
2580 complete(&sbi->s_kobj_unregister);
2581}
2582
52cf25d0 2583static const struct sysfs_ops ext4_attr_ops = {
3197ebdb
TT		 2584 .show = ext4_attr_show,
2585 .store = ext4_attr_store,
2586};
2587
2588static struct kobj_type ext4_ktype = {
2589 .default_attrs = ext4_attrs,
2590 .sysfs_ops = &ext4_attr_ops,
2591 .release = ext4_sb_release,
2592};
2593
857ac889
LC		 2594static void ext4_feat_release(struct kobject *kobj)
2595{
2596 complete(&ext4_feat->f_kobj_unregister);
2597}
2598
2599static struct kobj_type ext4_feat_ktype = {
2600 .default_attrs = ext4_feat_attrs,
2601 .sysfs_ops = &ext4_attr_ops,
2602 .release = ext4_feat_release,
2603};
2604
a13fb1a4
ES		 2605/*
2606 * Check whether this filesystem can be mounted based on
2607 * the features present and the RDONLY/RDWR mount requested.
2608 * Returns 1 if this filesystem can be mounted as requested,
2609 * 0 if it cannot be.
2610 */
2611static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2612{
2613 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2614 ext4_msg(sb, KERN_ERR,
2615 "Couldn't mount because of "
2616 "unsupported optional features (%x)",
2617 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2618 ~EXT4_FEATURE_INCOMPAT_SUPP));
2619 return 0;
2620 }
2621
2622 if (readonly)
2623 return 1;
2624
2625 /* Check that feature set is OK for a read-write mount */
2626 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2627 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2628 "unsupported optional features (%x)",
2629 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2630 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2631 return 0;
2632 }
2633 /*
2634 * Large file size enabled file system can only be mounted
2635 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2636 */
2637 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2638 if (sizeof(blkcnt_t) < sizeof(u64)) {
2639 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2640 "cannot be mounted RDWR without "
2641 "CONFIG_LBDAF");
2642 return 0;
2643 }
2644 }
bab08ab9
TT		 2645 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2646 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2647 ext4_msg(sb, KERN_ERR,
2648 "Can't support bigalloc feature without "
2649 "extents feature\n");
2650 return 0;
2651 }
7c319d32
AK		 2652
2653#ifndef CONFIG_QUOTA
2654 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
2655 !readonly) {
2656 ext4_msg(sb, KERN_ERR,
2657 "Filesystem with quota feature cannot be mounted RDWR "
2658 "without CONFIG_QUOTA");
2659 return 0;
2660 }
2661#endif /* CONFIG_QUOTA */
a13fb1a4
ES		 2662 return 1;
2663}
2664
66e61a9e
TT		 2665/*
2666 * This function is called once a day if we have errors logged
2667 * on the file system
2668 */
2669static void print_daily_error_info(unsigned long arg)
2670{
2671 struct super_block *sb = (struct super_block *) arg;
2672 struct ext4_sb_info *sbi;
2673 struct ext4_super_block *es;
2674
2675 sbi = EXT4_SB(sb);
2676 es = sbi->s_es;
2677
2678 if (es->s_error_count)
2679 ext4_msg(sb, KERN_NOTICE, "error count: %u",
2680 le32_to_cpu(es->s_error_count));
2681 if (es->s_first_error_time) {
2682 printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d",
2683 sb->s_id, le32_to_cpu(es->s_first_error_time),
2684 (int) sizeof(es->s_first_error_func),
2685 es->s_first_error_func,
2686 le32_to_cpu(es->s_first_error_line));
2687 if (es->s_first_error_ino)
2688 printk(": inode %u",
2689 le32_to_cpu(es->s_first_error_ino));
2690 if (es->s_first_error_block)
2691 printk(": block %llu", (unsigned long long)
2692 le64_to_cpu(es->s_first_error_block));
2693 printk("\n");
2694 }
2695 if (es->s_last_error_time) {
2696 printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d",
2697 sb->s_id, le32_to_cpu(es->s_last_error_time),
2698 (int) sizeof(es->s_last_error_func),
2699 es->s_last_error_func,
2700 le32_to_cpu(es->s_last_error_line));
2701 if (es->s_last_error_ino)
2702 printk(": inode %u",
2703 le32_to_cpu(es->s_last_error_ino));
2704 if (es->s_last_error_block)
2705 printk(": block %llu", (unsigned long long)
2706 le64_to_cpu(es->s_last_error_block));
2707 printk("\n");
2708 }
2709 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2710}
2711
bfff6873
LC		 2712/* Find next suitable group and run ext4_init_inode_table */
2713static int ext4_run_li_request(struct ext4_li_request *elr)
2714{
2715 struct ext4_group_desc *gdp = NULL;
2716 ext4_group_t group, ngroups;
2717 struct super_block *sb;
2718 unsigned long timeout = 0;
2719 int ret = 0;
2720
2721 sb = elr->lr_super;
2722 ngroups = EXT4_SB(sb)->s_groups_count;
2723
8e8ad8a5 2724 sb_start_write(sb);
bfff6873
LC		 2725 for (group = elr->lr_next_group; group < ngroups; group++) {
2726 gdp = ext4_get_group_desc(sb, group, NULL);
2727 if (!gdp) {
2728 ret = 1;
2729 break;
2730 }
2731
2732 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2733 break;
2734 }
2735
2736 if (group == ngroups)
2737 ret = 1;
2738
2739 if (!ret) {
2740 timeout = jiffies;
2741 ret = ext4_init_inode_table(sb, group,
2742 elr->lr_timeout ? 0 : 1);
2743 if (elr->lr_timeout == 0) {
51ce6511
LC		 2744 timeout = (jiffies - timeout) *
2745 elr->lr_sbi->s_li_wait_mult;
bfff6873
LC		 2746 elr->lr_timeout = timeout;
2747 }
2748 elr->lr_next_sched = jiffies + elr->lr_timeout;
2749 elr->lr_next_group = group + 1;
2750 }
8e8ad8a5 2751 sb_end_write(sb);
bfff6873
LC		 2752
2753 return ret;
2754}
2755
2756/*
2757 * Remove lr_request from the list_request and free the
4ed5c033 2758 * request structure. Should be called with li_list_mtx held
bfff6873
LC		 2759 */
2760static void ext4_remove_li_request(struct ext4_li_request *elr)
2761{
2762 struct ext4_sb_info *sbi;
2763
2764 if (!elr)
2765 return;
2766
2767 sbi = elr->lr_sbi;
2768
2769 list_del(&elr->lr_request);
2770 sbi->s_li_request = NULL;
2771 kfree(elr);
2772}
2773
2774static void ext4_unregister_li_request(struct super_block *sb)
2775{
1bb933fb
LC		 2776 mutex_lock(&ext4_li_mtx);
2777 if (!ext4_li_info) {
2778 mutex_unlock(&ext4_li_mtx);
bfff6873 2779 return;
1bb933fb 2780 }
bfff6873
LC		 2781
2782 mutex_lock(&ext4_li_info->li_list_mtx);
1bb933fb 2783 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
bfff6873 2784 mutex_unlock(&ext4_li_info->li_list_mtx);
1bb933fb 2785 mutex_unlock(&ext4_li_mtx);
bfff6873
LC		 2786}
2787
8f1f7453
ES		 2788static struct task_struct *ext4_lazyinit_task;
2789
bfff6873
LC		 2790/*
2791 * This is the function where ext4lazyinit thread lives. It walks
2792 * through the request list searching for next scheduled filesystem.
2793 * When such a fs is found, run the lazy initialization request
2794 * (ext4_rn_li_request) and keep track of the time spend in this
2795 * function. Based on that time we compute next schedule time of
2796 * the request. When walking through the list is complete, compute
2797 * next waking time and put itself into sleep.
2798 */
2799static int ext4_lazyinit_thread(void *arg)
2800{
2801 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2802 struct list_head *pos, *n;
2803 struct ext4_li_request *elr;
4ed5c033 2804 unsigned long next_wakeup, cur;
bfff6873
LC		 2805
2806 BUG_ON(NULL == eli);
2807
bfff6873
LC		 2808cont_thread:
2809 while (true) {
2810 next_wakeup = MAX_JIFFY_OFFSET;
2811
2812 mutex_lock(&eli->li_list_mtx);
2813 if (list_empty(&eli->li_request_list)) {
2814 mutex_unlock(&eli->li_list_mtx);
2815 goto exit_thread;
2816 }
2817
2818 list_for_each_safe(pos, n, &eli->li_request_list) {
2819 elr = list_entry(pos, struct ext4_li_request,
2820 lr_request);
2821
b2c78cd0
TT		 2822 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2823 if (ext4_run_li_request(elr) != 0) {
2824 /* error, remove the lazy_init job */
2825 ext4_remove_li_request(elr);
2826 continue;
2827 }
bfff6873
LC		 2828 }
2829
2830 if (time_before(elr->lr_next_sched, next_wakeup))
2831 next_wakeup = elr->lr_next_sched;
2832 }
2833 mutex_unlock(&eli->li_list_mtx);
2834
a0acae0e 2835 try_to_freeze();
bfff6873 2836
4ed5c033
LC		 2837 cur = jiffies;
2838 if ((time_after_eq(cur, next_wakeup)) ||
f4245bd4 2839 (MAX_JIFFY_OFFSET == next_wakeup)) {
bfff6873
LC		 2840 cond_resched();
2841 continue;
2842 }
2843
4ed5c033
LC		 2844 schedule_timeout_interruptible(next_wakeup - cur);
2845
8f1f7453
ES		 2846 if (kthread_should_stop()) {
2847 ext4_clear_request_list();
2848 goto exit_thread;
2849 }
bfff6873
LC		 2850 }
2851
2852exit_thread:
2853 /*
2854 * It looks like the request list is empty, but we need
2855 * to check it under the li_list_mtx lock, to prevent any
2856 * additions into it, and of course we should lock ext4_li_mtx
2857 * to atomically free the list and ext4_li_info, because at
2858 * this point another ext4 filesystem could be registering
2859 * new one.
2860 */
2861 mutex_lock(&ext4_li_mtx);
2862 mutex_lock(&eli->li_list_mtx);
2863 if (!list_empty(&eli->li_request_list)) {
2864 mutex_unlock(&eli->li_list_mtx);
2865 mutex_unlock(&ext4_li_mtx);
2866 goto cont_thread;
2867 }
2868 mutex_unlock(&eli->li_list_mtx);
bfff6873
LC		 2869 kfree(ext4_li_info);
2870 ext4_li_info = NULL;
2871 mutex_unlock(&ext4_li_mtx);
2872
2873 return 0;
2874}
2875
2876static void ext4_clear_request_list(void)
2877{
2878 struct list_head *pos, *n;
2879 struct ext4_li_request *elr;
2880
2881 mutex_lock(&ext4_li_info->li_list_mtx);
bfff6873
LC		 2882 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2883 elr = list_entry(pos, struct ext4_li_request,
2884 lr_request);
2885 ext4_remove_li_request(elr);
2886 }
2887 mutex_unlock(&ext4_li_info->li_list_mtx);
2888}
2889
2890static int ext4_run_lazyinit_thread(void)
2891{
8f1f7453
ES		 2892 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2893 ext4_li_info, "ext4lazyinit");
2894 if (IS_ERR(ext4_lazyinit_task)) {
2895 int err = PTR_ERR(ext4_lazyinit_task);
bfff6873 2896 ext4_clear_request_list();
bfff6873
LC		 2897 kfree(ext4_li_info);
2898 ext4_li_info = NULL;
92b97816 2899 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
bfff6873
LC		 2900 "initialization thread\n",
2901 err);
2902 return err;
2903 }
2904 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
bfff6873
LC		 2905 return 0;
2906}
2907
2908/*
2909 * Check whether it make sense to run itable init. thread or not.
2910 * If there is at least one uninitialized inode table, return
2911 * corresponding group number, else the loop goes through all
2912 * groups and return total number of groups.
2913 */
2914static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2915{
2916 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2917 struct ext4_group_desc *gdp = NULL;
2918
2919 for (group = 0; group < ngroups; group++) {
2920 gdp = ext4_get_group_desc(sb, group, NULL);
2921 if (!gdp)
2922 continue;
2923
2924 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2925 break;
2926 }
2927
2928 return group;
2929}
2930
2931static int ext4_li_info_new(void)
2932{
2933 struct ext4_lazy_init *eli = NULL;
2934
2935 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2936 if (!eli)
2937 return -ENOMEM;
2938
bfff6873
LC		 2939 INIT_LIST_HEAD(&eli->li_request_list);
2940 mutex_init(&eli->li_list_mtx);
2941
bfff6873
LC		 2942 eli->li_state |= EXT4_LAZYINIT_QUIT;
2943
2944 ext4_li_info = eli;
2945
2946 return 0;
2947}
2948
2949static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
2950 ext4_group_t start)
2951{
2952 struct ext4_sb_info *sbi = EXT4_SB(sb);
2953 struct ext4_li_request *elr;
2954 unsigned long rnd;
2955
2956 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
2957 if (!elr)
2958 return NULL;
2959
2960 elr->lr_super = sb;
2961 elr->lr_sbi = sbi;
2962 elr->lr_next_group = start;
2963
2964 /*
2965 * Randomize first schedule time of the request to
2966 * spread the inode table initialization requests
2967 * better.
2968 */
2969 get_random_bytes(&rnd, sizeof(rnd));
2970 elr->lr_next_sched = jiffies + (unsigned long)rnd %
2971 (EXT4_DEF_LI_MAX_START_DELAY * HZ);
2972
2973 return elr;
2974}
2975
2976static int ext4_register_li_request(struct super_block *sb,
2977 ext4_group_t first_not_zeroed)
2978{
2979 struct ext4_sb_info *sbi = EXT4_SB(sb);
2980 struct ext4_li_request *elr;
2981 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
6c5a6cb9 2982 int ret = 0;
bfff6873 2983
51ce6511
LC		 2984 if (sbi->s_li_request != NULL) {
2985 /*
2986 * Reset timeout so it can be computed again, because
2987 * s_li_wait_mult might have changed.
2988 */
2989 sbi->s_li_request->lr_timeout = 0;
beed5ecb 2990 return 0;
51ce6511 2991 }
bfff6873
LC		 2992
2993 if (first_not_zeroed == ngroups ||
2994 (sb->s_flags & MS_RDONLY) ||
55ff3840 2995 !test_opt(sb, INIT_INODE_TABLE))
beed5ecb 2996 return 0;
bfff6873
LC		 2997
2998 elr = ext4_li_request_new(sb, first_not_zeroed);
beed5ecb
NK		 2999 if (!elr)
3000 return -ENOMEM;
bfff6873
LC		 3001
3002 mutex_lock(&ext4_li_mtx);
3003
3004 if (NULL == ext4_li_info) {
3005 ret = ext4_li_info_new();
3006 if (ret)
3007 goto out;
3008 }
3009
3010 mutex_lock(&ext4_li_info->li_list_mtx);
3011 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3012 mutex_unlock(&ext4_li_info->li_list_mtx);
3013
3014 sbi->s_li_request = elr;
46e4690b
TM		 3015 /*
3016 * set elr to NULL here since it has been inserted to
3017 * the request_list and the removal and free of it is
3018 * handled by ext4_clear_request_list from now on.
3019 */
3020 elr = NULL;
bfff6873
LC		 3021
3022 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3023 ret = ext4_run_lazyinit_thread();
3024 if (ret)
3025 goto out;
3026 }
bfff6873 3027out:
beed5ecb
NK		 3028 mutex_unlock(&ext4_li_mtx);
3029 if (ret)
bfff6873 3030 kfree(elr);
bfff6873
LC		 3031 return ret;
3032}
3033
3034/*
3035 * We do not need to lock anything since this is called on
3036 * module unload.
3037 */
3038static void ext4_destroy_lazyinit_thread(void)
3039{
3040 /*
3041 * If thread exited earlier
3042 * there's nothing to be done.
3043 */
8f1f7453 3044 if (!ext4_li_info || !ext4_lazyinit_task)
bfff6873
LC		 3045 return;
3046
8f1f7453 3047 kthread_stop(ext4_lazyinit_task);
bfff6873
LC		 3048}
3049
25ed6e8a
DW		 3050static int set_journal_csum_feature_set(struct super_block *sb)
3051{
3052 int ret = 1;
3053 int compat, incompat;
3054 struct ext4_sb_info *sbi = EXT4_SB(sb);
3055
3056 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3057 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3058 /* journal checksum v2 */
3059 compat = 0;
3060 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V2;
3061 } else {
3062 /* journal checksum v1 */
3063 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3064 incompat = 0;
3065 }
3066
3067 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3068 ret = jbd2_journal_set_features(sbi->s_journal,
3069 compat, 0,
3070 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3071 incompat);
3072 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3073 ret = jbd2_journal_set_features(sbi->s_journal,
3074 compat, 0,
3075 incompat);
3076 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3077 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3078 } else {
3079 jbd2_journal_clear_features(sbi->s_journal,
3080 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3081 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3082 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3083 }
3084
3085 return ret;
3086}
3087
952fc18e
TT		 3088/*
3089 * Note: calculating the overhead so we can be compatible with
3090 * historical BSD practice is quite difficult in the face of
3091 * clusters/bigalloc. This is because multiple metadata blocks from
3092 * different block group can end up in the same allocation cluster.
3093 * Calculating the exact overhead in the face of clustered allocation
3094 * requires either O(all block bitmaps) in memory or O(number of block
3095 * groups**2) in time. We will still calculate the superblock for
3096 * older file systems --- and if we come across with a bigalloc file
3097 * system with zero in s_overhead_clusters the estimate will be close to
3098 * correct especially for very large cluster sizes --- but for newer
3099 * file systems, it's better to calculate this figure once at mkfs
3100 * time, and store it in the superblock. If the superblock value is
3101 * present (even for non-bigalloc file systems), we will use it.
3102 */
3103static int count_overhead(struct super_block *sb, ext4_group_t grp,
3104 char *buf)
3105{
3106 struct ext4_sb_info *sbi = EXT4_SB(sb);
3107 struct ext4_group_desc *gdp;
3108 ext4_fsblk_t first_block, last_block, b;
3109 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3110 int s, j, count = 0;
3111
0548bbb8
TT		 3112 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC))
3113 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3114 sbi->s_itb_per_group + 2);
3115
952fc18e
TT		 3116 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3117 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3118 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3119 for (i = 0; i < ngroups; i++) {
3120 gdp = ext4_get_group_desc(sb, i, NULL);
3121 b = ext4_block_bitmap(sb, gdp);
3122 if (b >= first_block && b <= last_block) {
3123 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3124 count++;
3125 }
3126 b = ext4_inode_bitmap(sb, gdp);
3127 if (b >= first_block && b <= last_block) {
3128 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3129 count++;
3130 }
3131 b = ext4_inode_table(sb, gdp);
3132 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3133 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3134 int c = EXT4_B2C(sbi, b - first_block);
3135 ext4_set_bit(c, buf);
3136 count++;
3137 }
3138 if (i != grp)
3139 continue;
3140 s = 0;
3141 if (ext4_bg_has_super(sb, grp)) {
3142 ext4_set_bit(s++, buf);
3143 count++;
3144 }
3145 for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
3146 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3147 count++;
3148 }
3149 }
3150 if (!count)
3151 return 0;
3152 return EXT4_CLUSTERS_PER_GROUP(sb) -
3153 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3154}
3155
3156/*
3157 * Compute the overhead and stash it in sbi->s_overhead
3158 */
3159int ext4_calculate_overhead(struct super_block *sb)
3160{
3161 struct ext4_sb_info *sbi = EXT4_SB(sb);
3162 struct ext4_super_block *es = sbi->s_es;
3163 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3164 ext4_fsblk_t overhead = 0;
3165 char *buf = (char *) get_zeroed_page(GFP_KERNEL);
3166
3167 memset(buf, 0, PAGE_SIZE);
3168 if (!buf)
3169 return -ENOMEM;
3170
3171 /*
3172 * Compute the overhead (FS structures). This is constant
3173 * for a given filesystem unless the number of block groups
3174 * changes so we cache the previous value until it does.
3175 */
3176
3177 /*
3178 * All of the blocks before first_data_block are overhead
3179 */
3180 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3181
3182 /*
3183 * Add the overhead found in each block group
3184 */
3185 for (i = 0; i < ngroups; i++) {
3186 int blks;
3187
3188 blks = count_overhead(sb, i, buf);
3189 overhead += blks;
3190 if (blks)
3191 memset(buf, 0, PAGE_SIZE);
3192 cond_resched();
3193 }
3194 sbi->s_overhead = overhead;
3195 smp_wmb();
3196 free_page((unsigned long) buf);
3197 return 0;
3198}
3199
2b2d6d01 3200static int ext4_fill_super(struct super_block *sb, void *data, int silent)
ac27a0ec 3201{
d4c402d9 3202 char *orig_data = kstrdup(data, GFP_KERNEL);
2b2d6d01 3203 struct buffer_head *bh;
617ba13b
MC		 3204 struct ext4_super_block *es = NULL;
3205 struct ext4_sb_info *sbi;
3206 ext4_fsblk_t block;
3207 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 3208 ext4_fsblk_t logical_sb_block;
ac27a0ec 3209 unsigned long offset = 0;
ac27a0ec
DK		 3210 unsigned long journal_devnum = 0;
3211 unsigned long def_mount_opts;
3212 struct inode *root;
9f6200bb 3213 char *cp;
0390131b 3214 const char *descr;
dcc7dae3 3215 int ret = -ENOMEM;
281b5995 3216 int blocksize, clustersize;
4ec11028
TT		 3217 unsigned int db_count;
3218 unsigned int i;
281b5995 3219 int needs_recovery, has_huge_files, has_bigalloc;
bd81d8ee 3220 __u64 blocks_count;
833f4077 3221 int err;
b3881f74 3222 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
bfff6873 3223 ext4_group_t first_not_zeroed;
ac27a0ec
DK		 3224
3225 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3226 if (!sbi)
dcc7dae3 3227 goto out_free_orig;
705895b6
PE		 3228
3229 sbi->s_blockgroup_lock =
3230 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3231 if (!sbi->s_blockgroup_lock) {
3232 kfree(sbi);
dcc7dae3 3233 goto out_free_orig;
705895b6 3234 }
ac27a0ec 3235 sb->s_fs_info = sbi;
2c0544b2 3236 sbi->s_sb = sb;
ac27a0ec 3237 sbi->s_mount_opt = 0;
08cefc7a
EB		 3238 sbi->s_resuid = make_kuid(&init_user_ns, EXT4_DEF_RESUID);
3239 sbi->s_resgid = make_kgid(&init_user_ns, EXT4_DEF_RESGID);
240799cd 3240 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
d9c9bef1 3241 sbi->s_sb_block = sb_block;
f613dfcb
TT		 3242 if (sb->s_bdev->bd_part)
3243 sbi->s_sectors_written_start =
3244 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
ac27a0ec 3245
9f6200bb
TT		 3246 /* Cleanup superblock name */
3247 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3248 *cp = '!';
3249
dcc7dae3 3250 ret = -EINVAL;
617ba13b 3251 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 3252 if (!blocksize) {
b31e1552 3253 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
ac27a0ec
DK		 3254 goto out_fail;
3255 }
3256
3257 /*
617ba13b 3258 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 3259 * block sizes. We need to calculate the offset from buffer start.
3260 */
617ba13b 3261 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 3262 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3263 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 3264 } else {
70bbb3e0 3265 logical_sb_block = sb_block;
ac27a0ec
DK		 3266 }
3267
70bbb3e0 3268 if (!(bh = sb_bread(sb, logical_sb_block))) {
b31e1552 3269 ext4_msg(sb, KERN_ERR, "unable to read superblock");
ac27a0ec
DK		 3270 goto out_fail;
3271 }
3272 /*
3273 * Note: s_es must be initialized as soon as possible because
617ba13b 3274 * some ext4 macro-instructions depend on its value
ac27a0ec 3275 */
2716b802 3276 es = (struct ext4_super_block *) (bh->b_data + offset);
ac27a0ec
DK		 3277 sbi->s_es = es;
3278 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 3279 if (sb->s_magic != EXT4_SUPER_MAGIC)
3280 goto cantfind_ext4;
afc32f7e 3281 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
ac27a0ec 3282
feb0ab32
DW		 3283 /* Warn if metadata_csum and gdt_csum are both set. */
3284 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3285 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
3286 EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_GDT_CSUM))
3287 ext4_warning(sb, KERN_INFO "metadata_csum and uninit_bg are "
3288 "redundant flags; please run fsck.");
3289
d25425f8
DW		 3290 /* Check for a known checksum algorithm */
3291 if (!ext4_verify_csum_type(sb, es)) {
3292 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3293 "unknown checksum algorithm.");
3294 silent = 1;
3295 goto cantfind_ext4;
3296 }
3297
0441984a
DW		 3298 /* Load the checksum driver */
3299 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3300 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3301 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3302 if (IS_ERR(sbi->s_chksum_driver)) {
3303 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3304 ret = PTR_ERR(sbi->s_chksum_driver);
3305 sbi->s_chksum_driver = NULL;
3306 goto failed_mount;
3307 }
3308 }
3309
a9c47317
DW		 3310 /* Check superblock checksum */
3311 if (!ext4_superblock_csum_verify(sb, es)) {
3312 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3313 "invalid superblock checksum. Run e2fsck?");
3314 silent = 1;
3315 goto cantfind_ext4;
3316 }
3317
3318 /* Precompute checksum seed for all metadata */
3319 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3320 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
3321 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3322 sizeof(es->s_uuid));
3323
ac27a0ec
DK		 3324 /* Set defaults before we parse the mount options */
3325 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
fd8c37ec 3326 set_opt(sb, INIT_INODE_TABLE);
617ba13b 3327 if (def_mount_opts & EXT4_DEFM_DEBUG)
fd8c37ec 3328 set_opt(sb, DEBUG);
87f26807 3329 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
fd8c37ec 3330 set_opt(sb, GRPID);
617ba13b 3331 if (def_mount_opts & EXT4_DEFM_UID16)
fd8c37ec 3332 set_opt(sb, NO_UID32);
ea663336 3333 /* xattr user namespace & acls are now defaulted on */
03010a33 3334#ifdef CONFIG_EXT4_FS_XATTR
ea663336 3335 set_opt(sb, XATTR_USER);
2e7842b8 3336#endif
03010a33 3337#ifdef CONFIG_EXT4_FS_POSIX_ACL
ea663336 3338 set_opt(sb, POSIX_ACL);
2e7842b8 3339#endif
6fd7a467 3340 set_opt(sb, MBLK_IO_SUBMIT);
617ba13b 3341 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
fd8c37ec 3342 set_opt(sb, JOURNAL_DATA);
617ba13b 3343 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
fd8c37ec 3344 set_opt(sb, ORDERED_DATA);
617ba13b 3345 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
fd8c37ec 3346 set_opt(sb, WRITEBACK_DATA);
617ba13b
MC		 3347
3348 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
fd8c37ec 3349 set_opt(sb, ERRORS_PANIC);
bb4f397a 3350 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
fd8c37ec 3351 set_opt(sb, ERRORS_CONT);
bb4f397a 3352 else
fd8c37ec 3353 set_opt(sb, ERRORS_RO);
8b67f04a 3354 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
fd8c37ec 3355 set_opt(sb, BLOCK_VALIDITY);
8b67f04a 3356 if (def_mount_opts & EXT4_DEFM_DISCARD)
fd8c37ec 3357 set_opt(sb, DISCARD);
ac27a0ec 3358
08cefc7a
EB		 3359 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3360 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
30773840
TT		 3361 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3362 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3363 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
ac27a0ec 3364
8b67f04a 3365 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
fd8c37ec 3366 set_opt(sb, BARRIER);
ac27a0ec 3367
dd919b98
AK		 3368 /*
3369 * enable delayed allocation by default
3370 * Use -o nodelalloc to turn it off
3371 */
8b67f04a
TT		 3372 if (!IS_EXT3_SB(sb) &&
3373 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
fd8c37ec 3374 set_opt(sb, DELALLOC);
dd919b98 3375
51ce6511
LC		 3376 /*
3377 * set default s_li_wait_mult for lazyinit, for the case there is
3378 * no mount option specified.
3379 */
3380 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3381
8b67f04a 3382 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
661aa520 3383 &journal_devnum, &journal_ioprio, 0)) {
8b67f04a
TT		 3384 ext4_msg(sb, KERN_WARNING,
3385 "failed to parse options in superblock: %s",
3386 sbi->s_es->s_mount_opts);
3387 }
5a916be1 3388 sbi->s_def_mount_opt = sbi->s_mount_opt;
b3881f74 3389 if (!parse_options((char *) data, sb, &journal_devnum,
661aa520 3390 &journal_ioprio, 0))
ac27a0ec
DK		 3391 goto failed_mount;
3392
56889787
TT		 3393 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3394 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3395 "with data=journal disables delayed "
3396 "allocation and O_DIRECT support!\n");
3397 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3398 ext4_msg(sb, KERN_ERR, "can't mount with "
3399 "both data=journal and delalloc");
3400 goto failed_mount;
3401 }
3402 if (test_opt(sb, DIOREAD_NOLOCK)) {
3403 ext4_msg(sb, KERN_ERR, "can't mount with "
3404 "both data=journal and delalloc");
3405 goto failed_mount;
3406 }
3407 if (test_opt(sb, DELALLOC))
3408 clear_opt(sb, DELALLOC);
3409 }
3410
3411 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3412 if (test_opt(sb, DIOREAD_NOLOCK)) {
3413 if (blocksize < PAGE_SIZE) {
3414 ext4_msg(sb, KERN_ERR, "can't mount with "
3415 "dioread_nolock if block size != PAGE_SIZE");
3416 goto failed_mount;
3417 }
3418 }
3419
ac27a0ec 3420 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 3421 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec 3422
617ba13b
MC		 3423 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3424 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3425 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3426 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
b31e1552
ES		 3427 ext4_msg(sb, KERN_WARNING,
3428 "feature flags set on rev 0 fs, "
3429 "running e2fsck is recommended");
469108ff 3430
2035e776
TT		 3431 if (IS_EXT2_SB(sb)) {
3432 if (ext2_feature_set_ok(sb))
3433 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3434 "using the ext4 subsystem");
3435 else {
3436 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3437 "to feature incompatibilities");
3438 goto failed_mount;
3439 }
3440 }
3441
3442 if (IS_EXT3_SB(sb)) {
3443 if (ext3_feature_set_ok(sb))
3444 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3445 "using the ext4 subsystem");
3446 else {
3447 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3448 "to feature incompatibilities");
3449 goto failed_mount;
3450 }
3451 }
3452
ac27a0ec
DK		 3453 /*
3454 * Check feature flags regardless of the revision level, since we
3455 * previously didn't change the revision level when setting the flags,
3456 * so there is a chance incompat flags are set on a rev 0 filesystem.
3457 */
a13fb1a4 3458 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
ac27a0ec 3459 goto failed_mount;
a13fb1a4 3460
617ba13b
MC		 3461 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3462 blocksize > EXT4_MAX_BLOCK_SIZE) {
b31e1552
ES		 3463 ext4_msg(sb, KERN_ERR,
3464 "Unsupported filesystem blocksize %d", blocksize);
ac27a0ec
DK		 3465 goto failed_mount;
3466 }
3467
ac27a0ec 3468 if (sb->s_blocksize != blocksize) {
ce40733c
AK		 3469 /* Validate the filesystem blocksize */
3470 if (!sb_set_blocksize(sb, blocksize)) {
b31e1552 3471 ext4_msg(sb, KERN_ERR, "bad block size %d",
ce40733c 3472 blocksize);
ac27a0ec
DK		 3473 goto failed_mount;
3474 }
3475
2b2d6d01 3476 brelse(bh);
70bbb3e0
AM		 3477 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3478 offset = do_div(logical_sb_block, blocksize);
3479 bh = sb_bread(sb, logical_sb_block);
ac27a0ec 3480 if (!bh) {
b31e1552
ES		 3481 ext4_msg(sb, KERN_ERR,
3482 "Can't read superblock on 2nd try");
ac27a0ec
DK		 3483 goto failed_mount;
3484 }
2716b802 3485 es = (struct ext4_super_block *)(bh->b_data + offset);
ac27a0ec 3486 sbi->s_es = es;
617ba13b 3487 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
b31e1552
ES		 3488 ext4_msg(sb, KERN_ERR,
3489 "Magic mismatch, very weird!");
ac27a0ec
DK		 3490 goto failed_mount;
3491 }
3492 }
3493
a13fb1a4
ES		 3494 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3495 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
f287a1a5
TT		 3496 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3497 has_huge_files);
3498 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
ac27a0ec 3499
617ba13b
MC		 3500 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3501 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3502 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
ac27a0ec
DK		 3503 } else {
3504 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3505 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
617ba13b 3506 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1330593e 3507 (!is_power_of_2(sbi->s_inode_size)) ||
ac27a0ec 3508 (sbi->s_inode_size > blocksize)) {
b31e1552
ES		 3509 ext4_msg(sb, KERN_ERR,
3510 "unsupported inode size: %d",
2b2d6d01 3511 sbi->s_inode_size);
ac27a0ec
DK		 3512 goto failed_mount;
3513 }
ef7f3835
KS		 3514 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3515 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
ac27a0ec 3516 }
0b8e58a1 3517
0d1ee42f
AR		 3518 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3519 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
8fadc143 3520 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 3521 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 3522 !is_power_of_2(sbi->s_desc_size)) {
b31e1552
ES		 3523 ext4_msg(sb, KERN_ERR,
3524 "unsupported descriptor size %lu",
0d1ee42f
AR		 3525 sbi->s_desc_size);
3526 goto failed_mount;
3527 }
3528 } else
3529 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
0b8e58a1 3530
ac27a0ec 3531 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 3532 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
b47b6f38 3533 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
617ba13b 3534 goto cantfind_ext4;
0b8e58a1 3535
617ba13b 3536 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 3537 if (sbi->s_inodes_per_block == 0)
617ba13b 3538 goto cantfind_ext4;
ac27a0ec
DK		 3539 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3540 sbi->s_inodes_per_block;
0d1ee42f 3541 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 3542 sbi->s_sbh = bh;
3543 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 3544 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3545 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
0b8e58a1 3546
2b2d6d01 3547 for (i = 0; i < 4; i++)
ac27a0ec
DK		 3548 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3549 sbi->s_def_hash_version = es->s_def_hash_version;
f99b2589
TT		 3550 i = le32_to_cpu(es->s_flags);
3551 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3552 sbi->s_hash_unsigned = 3;
3553 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3554#ifdef __CHAR_UNSIGNED__
3555 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3556 sbi->s_hash_unsigned = 3;
3557#else
3558 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3559#endif
f99b2589 3560 }
ac27a0ec 3561
281b5995
TT		 3562 /* Handle clustersize */
3563 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3564 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3565 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3566 if (has_bigalloc) {
3567 if (clustersize < blocksize) {
3568 ext4_msg(sb, KERN_ERR,
3569 "cluster size (%d) smaller than "
3570 "block size (%d)", clustersize, blocksize);
3571 goto failed_mount;
3572 }
3573 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3574 le32_to_cpu(es->s_log_block_size);
3575 sbi->s_clusters_per_group =
3576 le32_to_cpu(es->s_clusters_per_group);
3577 if (sbi->s_clusters_per_group > blocksize * 8) {
3578 ext4_msg(sb, KERN_ERR,
3579 "#clusters per group too big: %lu",
3580 sbi->s_clusters_per_group);
3581 goto failed_mount;
3582 }
3583 if (sbi->s_blocks_per_group !=
3584 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3585 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3586 "clusters per group (%lu) inconsistent",
3587 sbi->s_blocks_per_group,
3588 sbi->s_clusters_per_group);
3589 goto failed_mount;
3590 }
3591 } else {
3592 if (clustersize != blocksize) {
3593 ext4_warning(sb, "fragment/cluster size (%d) != "
3594 "block size (%d)", clustersize,
3595 blocksize);
3596 clustersize = blocksize;
3597 }
3598 if (sbi->s_blocks_per_group > blocksize * 8) {
3599 ext4_msg(sb, KERN_ERR,
3600 "#blocks per group too big: %lu",
3601 sbi->s_blocks_per_group);
3602 goto failed_mount;
3603 }
3604 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3605 sbi->s_cluster_bits = 0;
ac27a0ec 3606 }
281b5995
TT		 3607 sbi->s_cluster_ratio = clustersize / blocksize;
3608
ac27a0ec 3609 if (sbi->s_inodes_per_group > blocksize * 8) {
b31e1552
ES		 3610 ext4_msg(sb, KERN_ERR,
3611 "#inodes per group too big: %lu",
2b2d6d01 3612 sbi->s_inodes_per_group);
ac27a0ec
DK		 3613 goto failed_mount;
3614 }
3615
bf43d84b
ES		 3616 /*
3617 * Test whether we have more sectors than will fit in sector_t,
3618 * and whether the max offset is addressable by the page cache.
3619 */
5a9ae68a 3620 err = generic_check_addressable(sb->s_blocksize_bits,
30ca22c7 3621 ext4_blocks_count(es));
5a9ae68a 3622 if (err) {
b31e1552 3623 ext4_msg(sb, KERN_ERR, "filesystem"
bf43d84b 3624 " too large to mount safely on this system");
ac27a0ec 3625 if (sizeof(sector_t) < 8)
90c699a9 3626 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
5a9ae68a 3627 ret = err;
ac27a0ec
DK		 3628 goto failed_mount;
3629 }
3630
617ba13b
MC		 3631 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3632 goto cantfind_ext4;
e7c95593 3633
0f2ddca6
FTN		 3634 /* check blocks count against device size */
3635 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3636 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
b31e1552
ES		 3637 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3638 "exceeds size of device (%llu blocks)",
0f2ddca6
FTN		 3639 ext4_blocks_count(es), blocks_count);
3640 goto failed_mount;
3641 }
3642
0b8e58a1
AD		 3643 /*
3644 * It makes no sense for the first data block to be beyond the end
3645 * of the filesystem.
3646 */
3647 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
5635a62b 3648 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
b31e1552
ES		 3649 "block %u is beyond end of filesystem (%llu)",
3650 le32_to_cpu(es->s_first_data_block),
3651 ext4_blocks_count(es));
e7c95593
ES		 3652 goto failed_mount;
3653 }
bd81d8ee
LV		 3654 blocks_count = (ext4_blocks_count(es) -
3655 le32_to_cpu(es->s_first_data_block) +
3656 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3657 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4ec11028 3658 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
b31e1552 3659 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
4ec11028 3660 "(block count %llu, first data block %u, "
b31e1552 3661 "blocks per group %lu)", sbi->s_groups_count,
4ec11028
TT		 3662 ext4_blocks_count(es),
3663 le32_to_cpu(es->s_first_data_block),
3664 EXT4_BLOCKS_PER_GROUP(sb));
3665 goto failed_mount;
3666 }
bd81d8ee 3667 sbi->s_groups_count = blocks_count;
fb0a387d
ES		 3668 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3669 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
617ba13b
MC		 3670 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3671 EXT4_DESC_PER_BLOCK(sb);
f18a5f21
TT		 3672 sbi->s_group_desc = ext4_kvmalloc(db_count *
3673 sizeof(struct buffer_head *),
3674 GFP_KERNEL);
ac27a0ec 3675 if (sbi->s_group_desc == NULL) {
b31e1552 3676 ext4_msg(sb, KERN_ERR, "not enough memory");
2cde417d 3677 ret = -ENOMEM;
ac27a0ec
DK		 3678 goto failed_mount;
3679 }
3680
9f6200bb
TT		 3681 if (ext4_proc_root)
3682 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
240799cd 3683
66acdcf4
TT		 3684 if (sbi->s_proc)
3685 proc_create_data("options", S_IRUGO, sbi->s_proc,
3686 &ext4_seq_options_fops, sb);
3687
705895b6 3688 bgl_lock_init(sbi->s_blockgroup_lock);
ac27a0ec
DK		 3689
3690 for (i = 0; i < db_count; i++) {
70bbb3e0 3691 block = descriptor_loc(sb, logical_sb_block, i);
ac27a0ec
DK		 3692 sbi->s_group_desc[i] = sb_bread(sb, block);
3693 if (!sbi->s_group_desc[i]) {
b31e1552
ES		 3694 ext4_msg(sb, KERN_ERR,
3695 "can't read group descriptor %d", i);
ac27a0ec
DK		 3696 db_count = i;
3697 goto failed_mount2;
3698 }
3699 }
bfff6873 3700 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
b31e1552 3701 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
ac27a0ec
DK		 3702 goto failed_mount2;
3703 }
772cb7c8
JS		 3704 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3705 if (!ext4_fill_flex_info(sb)) {
b31e1552
ES		 3706 ext4_msg(sb, KERN_ERR,
3707 "unable to initialize "
3708 "flex_bg meta info!");
772cb7c8
JS		 3709 goto failed_mount2;
3710 }
3711
ac27a0ec
DK		 3712 sbi->s_gdb_count = db_count;
3713 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3714 spin_lock_init(&sbi->s_next_gen_lock);
3715
04496411
TM		 3716 init_timer(&sbi->s_err_report);
3717 sbi->s_err_report.function = print_daily_error_info;
3718 sbi->s_err_report.data = (unsigned long) sb;
3719
57042651 3720 err = percpu_counter_init(&sbi->s_freeclusters_counter,
5dee5437 3721 ext4_count_free_clusters(sb));
ce7e010a
TT		 3722 if (!err) {
3723 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3724 ext4_count_free_inodes(sb));
3725 }
3726 if (!err) {
3727 err = percpu_counter_init(&sbi->s_dirs_counter,
3728 ext4_count_dirs(sb));
3729 }
3730 if (!err) {
57042651 3731 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0);
ce7e010a
TT		 3732 }
3733 if (err) {
3734 ext4_msg(sb, KERN_ERR, "insufficient memory");
2cde417d 3735 ret = err;
ce7e010a
TT		 3736 goto failed_mount3;
3737 }
3738
c9de560d 3739 sbi->s_stripe = ext4_get_stripe_size(sbi);
55138e0b 3740 sbi->s_max_writeback_mb_bump = 128;
c9de560d 3741
ac27a0ec
DK		 3742 /*
3743 * set up enough so that it can read an inode
3744 */
9ca92389
TT		 3745 if (!test_opt(sb, NOLOAD) &&
3746 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3747 sb->s_op = &ext4_sops;
3748 else
3749 sb->s_op = &ext4_nojournal_sops;
617ba13b
MC		 3750 sb->s_export_op = &ext4_export_ops;
3751 sb->s_xattr = ext4_xattr_handlers;
ac27a0ec 3752#ifdef CONFIG_QUOTA
617ba13b
MC		 3753 sb->s_qcop = &ext4_qctl_operations;
3754 sb->dq_op = &ext4_quota_operations;
7c319d32
AK		 3755
3756 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) {
3757 /* Use qctl operations for hidden quota files. */
3758 sb->s_qcop = &ext4_qctl_sysfile_operations;
3759 }
ac27a0ec 3760#endif
f2fa2ffc
AK		 3761 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3762
ac27a0ec 3763 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3b9d4ed2 3764 mutex_init(&sbi->s_orphan_lock);
8f82f840 3765 sbi->s_resize_flags = 0;
ac27a0ec
DK		 3766
3767 sb->s_root = NULL;
3768
3769 needs_recovery = (es->s_last_orphan != 0 ||
617ba13b
MC		 3770 EXT4_HAS_INCOMPAT_FEATURE(sb,
3771 EXT4_FEATURE_INCOMPAT_RECOVER));
ac27a0ec 3772
c5e06d10
JL		 3773 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3774 !(sb->s_flags & MS_RDONLY))
3775 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3776 goto failed_mount3;
3777
ac27a0ec
DK		 3778 /*
3779 * The first inode we look at is the journal inode. Don't try
3780 * root first: it may be modified in the journal!
3781 */
3782 if (!test_opt(sb, NOLOAD) &&
617ba13b
MC		 3783 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3784 if (ext4_load_journal(sb, es, journal_devnum))
ac27a0ec 3785 goto failed_mount3;
0390131b
FM		 3786 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3787 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
b31e1552
ES		 3788 ext4_msg(sb, KERN_ERR, "required journal recovery "
3789 "suppressed and not mounted read-only");
744692dc 3790 goto failed_mount_wq;
ac27a0ec 3791 } else {
fd8c37ec 3792 clear_opt(sb, DATA_FLAGS);
0390131b
FM		 3793 sbi->s_journal = NULL;
3794 needs_recovery = 0;
3795 goto no_journal;
ac27a0ec
DK		 3796 }
3797
f32aaf2d 3798 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT) &&
eb40a09c
JS		 3799 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3800 JBD2_FEATURE_INCOMPAT_64BIT)) {
b31e1552 3801 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
744692dc 3802 goto failed_mount_wq;
eb40a09c
JS		 3803 }
3804
25ed6e8a
DW		 3805 if (!set_journal_csum_feature_set(sb)) {
3806 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
3807 "feature set");
3808 goto failed_mount_wq;
d4da6c9c 3809 }
818d276c 3810
ac27a0ec
DK		 3811 /* We have now updated the journal if required, so we can
3812 * validate the data journaling mode. */
3813 switch (test_opt(sb, DATA_FLAGS)) {
3814 case 0:
3815 /* No mode set, assume a default based on the journal
63f57933
AM		 3816 * capabilities: ORDERED_DATA if the journal can
3817 * cope, else JOURNAL_DATA
3818 */
dab291af
MC		 3819 if (jbd2_journal_check_available_features
3820 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
fd8c37ec 3821 set_opt(sb, ORDERED_DATA);
ac27a0ec 3822 else
fd8c37ec 3823 set_opt(sb, JOURNAL_DATA);
ac27a0ec
DK		 3824 break;
3825
617ba13b
MC		 3826 case EXT4_MOUNT_ORDERED_DATA:
3827 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 3828 if (!jbd2_journal_check_available_features
3829 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
b31e1552
ES		 3830 ext4_msg(sb, KERN_ERR, "Journal does not support "
3831 "requested data journaling mode");
744692dc 3832 goto failed_mount_wq;
ac27a0ec
DK		 3833 }
3834 default:
3835 break;
3836 }
b3881f74 3837 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
ac27a0ec 3838
18aadd47
BJ		 3839 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
3840
ce7e010a
TT		 3841 /*
3842 * The journal may have updated the bg summary counts, so we
3843 * need to update the global counters.
3844 */
57042651 3845 percpu_counter_set(&sbi->s_freeclusters_counter,
5dee5437 3846 ext4_count_free_clusters(sb));
ce7e010a
TT		 3847 percpu_counter_set(&sbi->s_freeinodes_counter,
3848 ext4_count_free_inodes(sb));
3849 percpu_counter_set(&sbi->s_dirs_counter,
3850 ext4_count_dirs(sb));
57042651 3851 percpu_counter_set(&sbi->s_dirtyclusters_counter, 0);
206f7ab4 3852
ce7e010a 3853no_journal:
952fc18e
TT		 3854 /*
3855 * Get the # of file system overhead blocks from the
3856 * superblock if present.
3857 */
3858 if (es->s_overhead_clusters)
3859 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
3860 else {
3861 ret = ext4_calculate_overhead(sb);
3862 if (ret)
3863 goto failed_mount_wq;
3864 }
3865
fd89d5f2
TH		 3866 /*
3867 * The maximum number of concurrent works can be high and
3868 * concurrency isn't really necessary. Limit it to 1.
3869 */
3870 EXT4_SB(sb)->dio_unwritten_wq =
ae005cbe 3871 alloc_workqueue("ext4-dio-unwritten", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4c0425ff
MC		 3872 if (!EXT4_SB(sb)->dio_unwritten_wq) {
3873 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
3874 goto failed_mount_wq;
3875 }
3876
ac27a0ec 3877 /*
dab291af 3878 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 3879 * so we can safely mount the rest of the filesystem now.
3880 */
3881
1d1fe1ee
DH		 3882 root = ext4_iget(sb, EXT4_ROOT_INO);
3883 if (IS_ERR(root)) {
b31e1552 3884 ext4_msg(sb, KERN_ERR, "get root inode failed");
1d1fe1ee 3885 ret = PTR_ERR(root);
32a9bb57 3886 root = NULL;
ac27a0ec
DK		 3887 goto failed_mount4;
3888 }
3889 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
b31e1552 3890 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
94bf608a 3891 iput(root);
ac27a0ec
DK		 3892 goto failed_mount4;
3893 }
48fde701 3894 sb->s_root = d_make_root(root);
1d1fe1ee 3895 if (!sb->s_root) {
b31e1552 3896 ext4_msg(sb, KERN_ERR, "get root dentry failed");
1d1fe1ee
DH		 3897 ret = -ENOMEM;
3898 goto failed_mount4;
3899 }
ac27a0ec 3900
7e84b621
ES		 3901 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
3902 sb->s_flags |= MS_RDONLY;
ef7f3835
KS		 3903
3904 /* determine the minimum size of new large inodes, if present */
3905 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3906 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3907 EXT4_GOOD_OLD_INODE_SIZE;
3908 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3909 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
3910 if (sbi->s_want_extra_isize <
3911 le16_to_cpu(es->s_want_extra_isize))
3912 sbi->s_want_extra_isize =
3913 le16_to_cpu(es->s_want_extra_isize);
3914 if (sbi->s_want_extra_isize <
3915 le16_to_cpu(es->s_min_extra_isize))
3916 sbi->s_want_extra_isize =
3917 le16_to_cpu(es->s_min_extra_isize);
3918 }
3919 }
3920 /* Check if enough inode space is available */
3921 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3922 sbi->s_inode_size) {
3923 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3924 EXT4_GOOD_OLD_INODE_SIZE;
b31e1552
ES		 3925 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3926 "available");
ef7f3835
KS		 3927 }
3928
6fd058f7
TT		 3929 err = ext4_setup_system_zone(sb);
3930 if (err) {
b31e1552 3931 ext4_msg(sb, KERN_ERR, "failed to initialize system "
fbe845dd 3932 "zone (%d)", err);
94bf608a 3933 goto failed_mount4a;
6fd058f7
TT		 3934 }
3935
c2774d84 3936 ext4_ext_init(sb);
9d99012f 3937 err = ext4_mb_init(sb);
c2774d84 3938 if (err) {
421f91d2 3939 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
b31e1552 3940 err);
dcf2d804 3941 goto failed_mount5;
c2774d84
AK		 3942 }
3943
bfff6873
LC		 3944 err = ext4_register_li_request(sb, first_not_zeroed);
3945 if (err)
dcf2d804 3946 goto failed_mount6;
bfff6873 3947
3197ebdb
TT		 3948 sbi->s_kobj.kset = ext4_kset;
3949 init_completion(&sbi->s_kobj_unregister);
3950 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
3951 "%s", sb->s_id);
dcf2d804
TM		 3952 if (err)
3953 goto failed_mount7;
3197ebdb 3954
617ba13b
MC		 3955 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3956 ext4_orphan_cleanup(sb, es);
3957 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
0390131b 3958 if (needs_recovery) {
b31e1552 3959 ext4_msg(sb, KERN_INFO, "recovery complete");
0390131b
FM		 3960 ext4_mark_recovery_complete(sb, es);
3961 }
3962 if (EXT4_SB(sb)->s_journal) {
3963 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3964 descr = " journalled data mode";
3965 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3966 descr = " ordered data mode";
3967 else
3968 descr = " writeback data mode";
3969 } else
3970 descr = "out journal";
3971
7c319d32
AK		 3972#ifdef CONFIG_QUOTA
3973 /* Enable quota usage during mount. */
3974 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
3975 !(sb->s_flags & MS_RDONLY)) {
3976 ret = ext4_enable_quotas(sb);
3977 if (ret)
3978 goto failed_mount7;
3979 }
3980#endif /* CONFIG_QUOTA */
3981
d4c402d9 3982 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
8b67f04a
TT		 3983 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
3984 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
ac27a0ec 3985
66e61a9e
TT		 3986 if (es->s_error_count)
3987 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
ac27a0ec 3988
d4c402d9 3989 kfree(orig_data);
ac27a0ec
DK		 3990 return 0;
3991
617ba13b 3992cantfind_ext4:
ac27a0ec 3993 if (!silent)
b31e1552 3994 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
ac27a0ec
DK		 3995 goto failed_mount;
3996
dcf2d804
TM		 3997failed_mount7:
3998 ext4_unregister_li_request(sb);
3999failed_mount6:
dcf2d804 4000 ext4_mb_release(sb);
94bf608a
AV		 4001failed_mount5:
4002 ext4_ext_release(sb);
dcf2d804 4003 ext4_release_system_zone(sb);
94bf608a
AV		 4004failed_mount4a:
4005 dput(sb->s_root);
32a9bb57 4006 sb->s_root = NULL;
94bf608a 4007failed_mount4:
b31e1552 4008 ext4_msg(sb, KERN_ERR, "mount failed");
4c0425ff
MC		 4009 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
4010failed_mount_wq:
0390131b
FM		 4011 if (sbi->s_journal) {
4012 jbd2_journal_destroy(sbi->s_journal);
4013 sbi->s_journal = NULL;
4014 }
ac27a0ec 4015failed_mount3:
04496411 4016 del_timer(&sbi->s_err_report);
9933fc0a
TT		 4017 if (sbi->s_flex_groups)
4018 ext4_kvfree(sbi->s_flex_groups);
57042651 4019 percpu_counter_destroy(&sbi->s_freeclusters_counter);
ce7e010a
TT		 4020 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4021 percpu_counter_destroy(&sbi->s_dirs_counter);
57042651 4022 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
c5e06d10
JL		 4023 if (sbi->s_mmp_tsk)
4024 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec
DK		 4025failed_mount2:
4026 for (i = 0; i < db_count; i++)
4027 brelse(sbi->s_group_desc[i]);
f18a5f21 4028 ext4_kvfree(sbi->s_group_desc);
ac27a0ec 4029failed_mount:
0441984a
DW		 4030 if (sbi->s_chksum_driver)
4031 crypto_free_shash(sbi->s_chksum_driver);
240799cd 4032 if (sbi->s_proc) {
66acdcf4 4033 remove_proc_entry("options", sbi->s_proc);
9f6200bb 4034 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 4035 }
ac27a0ec
DK		 4036#ifdef CONFIG_QUOTA
4037 for (i = 0; i < MAXQUOTAS; i++)
4038 kfree(sbi->s_qf_names[i]);
4039#endif
617ba13b 4040 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 4041 brelse(bh);
4042out_fail:
4043 sb->s_fs_info = NULL;
f6830165 4044 kfree(sbi->s_blockgroup_lock);
ac27a0ec 4045 kfree(sbi);
dcc7dae3 4046out_free_orig:
d4c402d9 4047 kfree(orig_data);
1d1fe1ee 4048 return ret;
ac27a0ec
DK		 4049}
4050
4051/*
4052 * Setup any per-fs journal parameters now. We'll do this both on
4053 * initial mount, once the journal has been initialised but before we've
4054 * done any recovery; and again on any subsequent remount.
4055 */
617ba13b 4056static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 4057{
617ba13b 4058 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4059
30773840
TT		 4060 journal->j_commit_interval = sbi->s_commit_interval;
4061 journal->j_min_batch_time = sbi->s_min_batch_time;
4062 journal->j_max_batch_time = sbi->s_max_batch_time;
ac27a0ec 4063
a931da6a 4064 write_lock(&journal->j_state_lock);
ac27a0ec 4065 if (test_opt(sb, BARRIER))
dab291af 4066 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 4067 else
dab291af 4068 journal->j_flags &= ~JBD2_BARRIER;
5bf5683a
HK		 4069 if (test_opt(sb, DATA_ERR_ABORT))
4070 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4071 else
4072 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
a931da6a 4073 write_unlock(&journal->j_state_lock);
ac27a0ec
DK		 4074}
4075
617ba13b 4076static journal_t *ext4_get_journal(struct super_block *sb,
ac27a0ec
DK		 4077 unsigned int journal_inum)
4078{
4079 struct inode *journal_inode;
4080 journal_t *journal;
4081
0390131b
FM		 4082 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4083
ac27a0ec
DK		 4084 /* First, test for the existence of a valid inode on disk. Bad
4085 * things happen if we iget() an unused inode, as the subsequent
4086 * iput() will try to delete it. */
4087
1d1fe1ee
DH		 4088 journal_inode = ext4_iget(sb, journal_inum);
4089 if (IS_ERR(journal_inode)) {
b31e1552 4090 ext4_msg(sb, KERN_ERR, "no journal found");
ac27a0ec
DK		 4091 return NULL;
4092 }
4093 if (!journal_inode->i_nlink) {
4094 make_bad_inode(journal_inode);
4095 iput(journal_inode);
b31e1552 4096 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
ac27a0ec
DK		 4097 return NULL;
4098 }
4099
e5f8eab8 4100 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
ac27a0ec 4101 journal_inode, journal_inode->i_size);
1d1fe1ee 4102 if (!S_ISREG(journal_inode->i_mode)) {
b31e1552 4103 ext4_msg(sb, KERN_ERR, "invalid journal inode");
ac27a0ec
DK		 4104 iput(journal_inode);
4105 return NULL;
4106 }
4107
dab291af 4108 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 4109 if (!journal) {
b31e1552 4110 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
ac27a0ec
DK		 4111 iput(journal_inode);
4112 return NULL;
4113 }
4114 journal->j_private = sb;
617ba13b 4115 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 4116 return journal;
4117}
4118
617ba13b 4119static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 4120 dev_t j_dev)
4121{
2b2d6d01 4122 struct buffer_head *bh;
ac27a0ec 4123 journal_t *journal;
617ba13b
MC		 4124 ext4_fsblk_t start;
4125 ext4_fsblk_t len;
ac27a0ec 4126 int hblock, blocksize;
617ba13b 4127 ext4_fsblk_t sb_block;
ac27a0ec 4128 unsigned long offset;
2b2d6d01 4129 struct ext4_super_block *es;
ac27a0ec
DK		 4130 struct block_device *bdev;
4131
0390131b
FM		 4132 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4133
b31e1552 4134 bdev = ext4_blkdev_get(j_dev, sb);
ac27a0ec
DK		 4135 if (bdev == NULL)
4136 return NULL;
4137
ac27a0ec 4138 blocksize = sb->s_blocksize;
e1defc4f 4139 hblock = bdev_logical_block_size(bdev);
ac27a0ec 4140 if (blocksize < hblock) {
b31e1552
ES		 4141 ext4_msg(sb, KERN_ERR,
4142 "blocksize too small for journal device");
ac27a0ec
DK		 4143 goto out_bdev;
4144 }
4145
617ba13b
MC		 4146 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4147 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 4148 set_blocksize(bdev, blocksize);
4149 if (!(bh = __bread(bdev, sb_block, blocksize))) {
b31e1552
ES		 4150 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4151 "external journal");
ac27a0ec
DK		 4152 goto out_bdev;
4153 }
4154
2716b802 4155 es = (struct ext4_super_block *) (bh->b_data + offset);
617ba13b 4156 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 4157 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b 4158 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
b31e1552
ES		 4159 ext4_msg(sb, KERN_ERR, "external journal has "
4160 "bad superblock");
ac27a0ec
DK		 4161 brelse(bh);
4162 goto out_bdev;
4163 }
4164
617ba13b 4165 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
b31e1552 4166 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
ac27a0ec
DK		 4167 brelse(bh);
4168 goto out_bdev;
4169 }
4170
bd81d8ee 4171 len = ext4_blocks_count(es);
ac27a0ec
DK		 4172 start = sb_block + 1;
4173 brelse(bh); /* we're done with the superblock */
4174
dab291af 4175 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 4176 start, len, blocksize);
4177 if (!journal) {
b31e1552 4178 ext4_msg(sb, KERN_ERR, "failed to create device journal");
ac27a0ec
DK		 4179 goto out_bdev;
4180 }
4181 journal->j_private = sb;
4182 ll_rw_block(READ, 1, &journal->j_sb_buffer);
4183 wait_on_buffer(journal->j_sb_buffer);
4184 if (!buffer_uptodate(journal->j_sb_buffer)) {
b31e1552 4185 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
ac27a0ec
DK		 4186 goto out_journal;
4187 }
4188 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
b31e1552
ES		 4189 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4190 "user (unsupported) - %d",
ac27a0ec
DK		 4191 be32_to_cpu(journal->j_superblock->s_nr_users));
4192 goto out_journal;
4193 }
617ba13b
MC		 4194 EXT4_SB(sb)->journal_bdev = bdev;
4195 ext4_init_journal_params(sb, journal);
ac27a0ec 4196 return journal;
0b8e58a1 4197
ac27a0ec 4198out_journal:
dab291af 4199 jbd2_journal_destroy(journal);
ac27a0ec 4200out_bdev:
617ba13b 4201 ext4_blkdev_put(bdev);
ac27a0ec
DK		 4202 return NULL;
4203}
4204
617ba13b
MC		 4205static int ext4_load_journal(struct super_block *sb,
4206 struct ext4_super_block *es,
ac27a0ec
DK		 4207 unsigned long journal_devnum)
4208{
4209 journal_t *journal;
4210 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4211 dev_t journal_dev;
4212 int err = 0;
4213 int really_read_only;
4214
0390131b
FM		 4215 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4216
ac27a0ec
DK		 4217 if (journal_devnum &&
4218 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
b31e1552
ES		 4219 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4220 "numbers have changed");
ac27a0ec
DK		 4221 journal_dev = new_decode_dev(journal_devnum);
4222 } else
4223 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4224
4225 really_read_only = bdev_read_only(sb->s_bdev);
4226
4227 /*
4228 * Are we loading a blank journal or performing recovery after a
4229 * crash? For recovery, we need to check in advance whether we
4230 * can get read-write access to the device.
4231 */
617ba13b 4232 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
ac27a0ec 4233 if (sb->s_flags & MS_RDONLY) {
b31e1552
ES		 4234 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4235 "required on readonly filesystem");
ac27a0ec 4236 if (really_read_only) {
b31e1552
ES		 4237 ext4_msg(sb, KERN_ERR, "write access "
4238 "unavailable, cannot proceed");
ac27a0ec
DK		 4239 return -EROFS;
4240 }
b31e1552
ES		 4241 ext4_msg(sb, KERN_INFO, "write access will "
4242 "be enabled during recovery");
ac27a0ec
DK		 4243 }
4244 }
4245
4246 if (journal_inum && journal_dev) {
b31e1552
ES		 4247 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4248 "and inode journals!");
ac27a0ec
DK		 4249 return -EINVAL;
4250 }
4251
4252 if (journal_inum) {
617ba13b 4253 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 4254 return -EINVAL;
4255 } else {
617ba13b 4256 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 4257 return -EINVAL;
4258 }
4259
90576c0b 4260 if (!(journal->j_flags & JBD2_BARRIER))
b31e1552 4261 ext4_msg(sb, KERN_INFO, "barriers disabled");
4776004f 4262
617ba13b 4263 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
dab291af 4264 err = jbd2_journal_wipe(journal, !really_read_only);
1c13d5c0
TT		 4265 if (!err) {
4266 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4267 if (save)
4268 memcpy(save, ((char *) es) +
4269 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
dab291af 4270 err = jbd2_journal_load(journal);
1c13d5c0
TT		 4271 if (save)
4272 memcpy(((char *) es) + EXT4_S_ERR_START,
4273 save, EXT4_S_ERR_LEN);
4274 kfree(save);
4275 }
ac27a0ec
DK		 4276
4277 if (err) {
b31e1552 4278 ext4_msg(sb, KERN_ERR, "error loading journal");
dab291af 4279 jbd2_journal_destroy(journal);
ac27a0ec
DK		 4280 return err;
4281 }
4282
617ba13b
MC		 4283 EXT4_SB(sb)->s_journal = journal;
4284 ext4_clear_journal_err(sb, es);
ac27a0ec 4285
c41303ce 4286 if (!really_read_only && journal_devnum &&
ac27a0ec
DK		 4287 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4288 es->s_journal_dev = cpu_to_le32(journal_devnum);
ac27a0ec
DK		 4289
4290 /* Make sure we flush the recovery flag to disk. */
e2d67052 4291 ext4_commit_super(sb, 1);
ac27a0ec
DK		 4292 }
4293
4294 return 0;
4295}
4296
e2d67052 4297static int ext4_commit_super(struct super_block *sb, int sync)
ac27a0ec 4298{
e2d67052 4299 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
617ba13b 4300 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
c4be0c1d 4301 int error = 0;
ac27a0ec 4302
7c2e7087 4303 if (!sbh || block_device_ejected(sb))
c4be0c1d 4304 return error;
914258bf
TT		 4305 if (buffer_write_io_error(sbh)) {
4306 /*
4307 * Oh, dear. A previous attempt to write the
4308 * superblock failed. This could happen because the
4309 * USB device was yanked out. Or it could happen to
4310 * be a transient write error and maybe the block will
4311 * be remapped. Nothing we can do but to retry the
4312 * write and hope for the best.
4313 */
b31e1552
ES		 4314 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4315 "superblock detected");
914258bf
TT		 4316 clear_buffer_write_io_error(sbh);
4317 set_buffer_uptodate(sbh);
4318 }
71290b36
TT		 4319 /*
4320 * If the file system is mounted read-only, don't update the
4321 * superblock write time. This avoids updating the superblock
4322 * write time when we are mounting the root file system
4323 * read/only but we need to replay the journal; at that point,
4324 * for people who are east of GMT and who make their clock
4325 * tick in localtime for Windows bug-for-bug compatibility,
4326 * the clock is set in the future, and this will cause e2fsck
4327 * to complain and force a full file system check.
4328 */
4329 if (!(sb->s_flags & MS_RDONLY))
4330 es->s_wtime = cpu_to_le32(get_seconds());
f613dfcb
TT		 4331 if (sb->s_bdev->bd_part)
4332 es->s_kbytes_written =
4333 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
afc32f7e
TT		 4334 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4335 EXT4_SB(sb)->s_sectors_written_start) >> 1));
f613dfcb
TT		 4336 else
4337 es->s_kbytes_written =
4338 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
57042651
TT		 4339 ext4_free_blocks_count_set(es,
4340 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4341 &EXT4_SB(sb)->s_freeclusters_counter)));
ce7e010a
TT		 4342 es->s_free_inodes_count =
4343 cpu_to_le32(percpu_counter_sum_positive(
4344 &EXT4_SB(sb)->s_freeinodes_counter));
ac27a0ec 4345 BUFFER_TRACE(sbh, "marking dirty");
a9c47317 4346 ext4_superblock_csum_set(sb, es);
ac27a0ec 4347 mark_buffer_dirty(sbh);
914258bf 4348 if (sync) {
c4be0c1d
TS		 4349 error = sync_dirty_buffer(sbh);
4350 if (error)
4351 return error;
4352
4353 error = buffer_write_io_error(sbh);
4354 if (error) {
b31e1552
ES		 4355 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4356 "superblock");
914258bf
TT		 4357 clear_buffer_write_io_error(sbh);
4358 set_buffer_uptodate(sbh);
4359 }
4360 }
c4be0c1d 4361 return error;
ac27a0ec
DK		 4362}
4363
ac27a0ec
DK		 4364/*
4365 * Have we just finished recovery? If so, and if we are mounting (or
4366 * remounting) the filesystem readonly, then we will end up with a
4367 * consistent fs on disk. Record that fact.
4368 */
2b2d6d01
TT		 4369static void ext4_mark_recovery_complete(struct super_block *sb,
4370 struct ext4_super_block *es)
ac27a0ec 4371{
617ba13b 4372 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 4373
0390131b
FM		 4374 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4375 BUG_ON(journal != NULL);
4376 return;
4377 }
dab291af 4378 jbd2_journal_lock_updates(journal);
7ffe1ea8
HK		 4379 if (jbd2_journal_flush(journal) < 0)
4380 goto out;
4381
617ba13b 4382 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
ac27a0ec 4383 sb->s_flags & MS_RDONLY) {
617ba13b 4384 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
e2d67052 4385 ext4_commit_super(sb, 1);
ac27a0ec 4386 }
7ffe1ea8
HK		 4387
4388out:
dab291af 4389 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 4390}
4391
4392/*
4393 * If we are mounting (or read-write remounting) a filesystem whose journal
4394 * has recorded an error from a previous lifetime, move that error to the
4395 * main filesystem now.
4396 */
2b2d6d01
TT		 4397static void ext4_clear_journal_err(struct super_block *sb,
4398 struct ext4_super_block *es)
ac27a0ec
DK		 4399{
4400 journal_t *journal;
4401 int j_errno;
4402 const char *errstr;
4403
0390131b
FM		 4404 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4405
617ba13b 4406 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 4407
4408 /*
4409 * Now check for any error status which may have been recorded in the
617ba13b 4410 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 4411 */
4412
dab291af 4413 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 4414 if (j_errno) {
4415 char nbuf[16];
4416
617ba13b 4417 errstr = ext4_decode_error(sb, j_errno, nbuf);
12062ddd 4418 ext4_warning(sb, "Filesystem error recorded "
ac27a0ec 4419 "from previous mount: %s", errstr);
12062ddd 4420 ext4_warning(sb, "Marking fs in need of filesystem check.");
ac27a0ec 4421
617ba13b
MC		 4422 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4423 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
e2d67052 4424 ext4_commit_super(sb, 1);
ac27a0ec 4425
dab291af 4426 jbd2_journal_clear_err(journal);
d796c52e 4427 jbd2_journal_update_sb_errno(journal);
ac27a0ec
DK		 4428 }
4429}
4430
4431/*
4432 * Force the running and committing transactions to commit,
4433 * and wait on the commit.
4434 */
617ba13b 4435int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 4436{
4437 journal_t *journal;
0390131b 4438 int ret = 0;
ac27a0ec
DK		 4439
4440 if (sb->s_flags & MS_RDONLY)
4441 return 0;
4442
617ba13b 4443 journal = EXT4_SB(sb)->s_journal;
8e8ad8a5 4444 if (journal)
0390131b 4445 ret = ext4_journal_force_commit(journal);
0390131b 4446
ac27a0ec
DK		 4447 return ret;
4448}
4449
617ba13b 4450static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec 4451{
14ce0cb4 4452 int ret = 0;
9eddacf9 4453 tid_t target;
8d5d02e6 4454 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4455
9bffad1e 4456 trace_ext4_sync_fs(sb, wait);
8d5d02e6 4457 flush_workqueue(sbi->dio_unwritten_wq);
a1177825
JK		 4458 /*
4459 * Writeback quota in non-journalled quota case - journalled quota has
4460 * no dirty dquots
4461 */
4462 dquot_writeback_dquots(sb, -1);
8d5d02e6 4463 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
9ca92389 4464 if (wait)
8d5d02e6 4465 jbd2_log_wait_commit(sbi->s_journal, target);
0390131b 4466 }
14ce0cb4 4467 return ret;
ac27a0ec
DK		 4468}
4469
4470/*
4471 * LVM calls this function before a (read-only) snapshot is created. This
4472 * gives us a chance to flush the journal completely and mark the fs clean.
be4f27d3
YY		 4473 *
4474 * Note that only this function cannot bring a filesystem to be in a clean
8e8ad8a5
JK		 4475 * state independently. It relies on upper layer to stop all data & metadata
4476 * modifications.
ac27a0ec 4477 */
c4be0c1d 4478static int ext4_freeze(struct super_block *sb)
ac27a0ec 4479{
c4be0c1d
TS		 4480 int error = 0;
4481 journal_t *journal;
ac27a0ec 4482
9ca92389
TT		 4483 if (sb->s_flags & MS_RDONLY)
4484 return 0;
ac27a0ec 4485
9ca92389 4486 journal = EXT4_SB(sb)->s_journal;
7ffe1ea8 4487
9ca92389
TT		 4488 /* Now we set up the journal barrier. */
4489 jbd2_journal_lock_updates(journal);
ac27a0ec 4490
9ca92389
TT		 4491 /*
4492 * Don't clear the needs_recovery flag if we failed to flush
4493 * the journal.
4494 */
4495 error = jbd2_journal_flush(journal);
6b0310fb
ES		 4496 if (error < 0)
4497 goto out;
9ca92389
TT		 4498
4499 /* Journal blocked and flushed, clear needs_recovery flag. */
4500 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4501 error = ext4_commit_super(sb, 1);
6b0310fb 4502out:
8e8ad8a5 4503 /* we rely on upper layer to stop further updates */
6b0310fb
ES		 4504 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4505 return error;
ac27a0ec
DK		 4506}
4507
4508/*
4509 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4510 * flag here, even though the filesystem is not technically dirty yet.
4511 */
c4be0c1d 4512static int ext4_unfreeze(struct super_block *sb)
ac27a0ec 4513{
9ca92389
TT		 4514 if (sb->s_flags & MS_RDONLY)
4515 return 0;
4516
4517 lock_super(sb);
4518 /* Reset the needs_recovery flag before the fs is unlocked. */
4519 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4520 ext4_commit_super(sb, 1);
4521 unlock_super(sb);
c4be0c1d 4522 return 0;
ac27a0ec
DK		 4523}
4524
673c6100
TT		 4525/*
4526 * Structure to save mount options for ext4_remount's benefit
4527 */
4528struct ext4_mount_options {
4529 unsigned long s_mount_opt;
a2595b8a 4530 unsigned long s_mount_opt2;
08cefc7a
EB		 4531 kuid_t s_resuid;
4532 kgid_t s_resgid;
673c6100
TT		 4533 unsigned long s_commit_interval;
4534 u32 s_min_batch_time, s_max_batch_time;
4535#ifdef CONFIG_QUOTA
4536 int s_jquota_fmt;
4537 char *s_qf_names[MAXQUOTAS];
4538#endif
4539};
4540
2b2d6d01 4541static int ext4_remount(struct super_block *sb, int *flags, char *data)
ac27a0ec 4542{
2b2d6d01 4543 struct ext4_super_block *es;
617ba13b 4544 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4545 unsigned long old_sb_flags;
617ba13b 4546 struct ext4_mount_options old_opts;
c79d967d 4547 int enable_quota = 0;
8a266467 4548 ext4_group_t g;
b3881f74 4549 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
c5e06d10 4550 int err = 0;
ac27a0ec
DK		 4551#ifdef CONFIG_QUOTA
4552 int i;
4553#endif
d4c402d9 4554 char *orig_data = kstrdup(data, GFP_KERNEL);
ac27a0ec
DK		 4555
4556 /* Store the original options */
bbd6851a 4557 lock_super(sb);
ac27a0ec
DK		 4558 old_sb_flags = sb->s_flags;
4559 old_opts.s_mount_opt = sbi->s_mount_opt;
a2595b8a 4560 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
ac27a0ec
DK		 4561 old_opts.s_resuid = sbi->s_resuid;
4562 old_opts.s_resgid = sbi->s_resgid;
4563 old_opts.s_commit_interval = sbi->s_commit_interval;
30773840
TT		 4564 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4565 old_opts.s_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 4566#ifdef CONFIG_QUOTA
4567 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4568 for (i = 0; i < MAXQUOTAS; i++)
4569 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
4570#endif
b3881f74
TT		 4571 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4572 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
ac27a0ec
DK		 4573
4574 /*
4575 * Allow the "check" option to be passed as a remount option.
4576 */
661aa520 4577 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
ac27a0ec
DK		 4578 err = -EINVAL;
4579 goto restore_opts;
4580 }
4581
4ab2f15b 4582 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
c67d859e 4583 ext4_abort(sb, "Abort forced by user");
ac27a0ec
DK		 4584
4585 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 4586 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec
DK		 4587
4588 es = sbi->s_es;
4589
b3881f74 4590 if (sbi->s_journal) {
0390131b 4591 ext4_init_journal_params(sb, sbi->s_journal);
b3881f74
TT		 4592 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4593 }
ac27a0ec 4594
661aa520 4595 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4ab2f15b 4596 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
ac27a0ec
DK		 4597 err = -EROFS;
4598 goto restore_opts;
4599 }
4600
4601 if (*flags & MS_RDONLY) {
0f0dd62f
CH		 4602 err = dquot_suspend(sb, -1);
4603 if (err < 0)
c79d967d 4604 goto restore_opts;
c79d967d 4605
ac27a0ec
DK		 4606 /*
4607 * First of all, the unconditional stuff we have to do
4608 * to disable replay of the journal when we next remount
4609 */
4610 sb->s_flags |= MS_RDONLY;
4611
4612 /*
4613 * OK, test if we are remounting a valid rw partition
4614 * readonly, and if so set the rdonly flag and then
4615 * mark the partition as valid again.
4616 */
617ba13b
MC		 4617 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4618 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 4619 es->s_state = cpu_to_le16(sbi->s_mount_state);
4620
a63c9eb2 4621 if (sbi->s_journal)
0390131b 4622 ext4_mark_recovery_complete(sb, es);
ac27a0ec 4623 } else {
a13fb1a4
ES		 4624 /* Make sure we can mount this feature set readwrite */
4625 if (!ext4_feature_set_ok(sb, 0)) {
ac27a0ec
DK		 4626 err = -EROFS;
4627 goto restore_opts;
4628 }
8a266467
TT		 4629 /*
4630 * Make sure the group descriptor checksums
0b8e58a1 4631 * are sane. If they aren't, refuse to remount r/w.
8a266467
TT		 4632 */
4633 for (g = 0; g < sbi->s_groups_count; g++) {
4634 struct ext4_group_desc *gdp =
4635 ext4_get_group_desc(sb, g, NULL);
4636
feb0ab32 4637 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
b31e1552
ES		 4638 ext4_msg(sb, KERN_ERR,
4639 "ext4_remount: Checksum for group %u failed (%u!=%u)",
8a266467
TT		 4640 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4641 le16_to_cpu(gdp->bg_checksum));
4642 err = -EINVAL;
4643 goto restore_opts;
4644 }
4645 }
4646
ead6596b
ES		 4647 /*
4648 * If we have an unprocessed orphan list hanging
4649 * around from a previously readonly bdev mount,
4650 * require a full umount/remount for now.
4651 */
4652 if (es->s_last_orphan) {
b31e1552 4653 ext4_msg(sb, KERN_WARNING, "Couldn't "
ead6596b
ES		 4654 "remount RDWR because of unprocessed "
4655 "orphan inode list. Please "
b31e1552 4656 "umount/remount instead");
ead6596b
ES		 4657 err = -EINVAL;
4658 goto restore_opts;
4659 }
4660
ac27a0ec
DK		 4661 /*
4662 * Mounting a RDONLY partition read-write, so reread
4663 * and store the current valid flag. (It may have
4664 * been changed by e2fsck since we originally mounted
4665 * the partition.)
4666 */
0390131b
FM		 4667 if (sbi->s_journal)
4668 ext4_clear_journal_err(sb, es);
ac27a0ec 4669 sbi->s_mount_state = le16_to_cpu(es->s_state);
2b2d6d01 4670 if (!ext4_setup_super(sb, es, 0))
ac27a0ec 4671 sb->s_flags &= ~MS_RDONLY;
c5e06d10
JL		 4672 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
4673 EXT4_FEATURE_INCOMPAT_MMP))
4674 if (ext4_multi_mount_protect(sb,
4675 le64_to_cpu(es->s_mmp_block))) {
4676 err = -EROFS;
4677 goto restore_opts;
4678 }
c79d967d 4679 enable_quota = 1;
ac27a0ec
DK		 4680 }
4681 }
bfff6873
LC		 4682
4683 /*
4684 * Reinitialize lazy itable initialization thread based on
4685 * current settings
4686 */
4687 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4688 ext4_unregister_li_request(sb);
4689 else {
4690 ext4_group_t first_not_zeroed;
4691 first_not_zeroed = ext4_has_uninit_itable(sb);
4692 ext4_register_li_request(sb, first_not_zeroed);
4693 }
4694
6fd058f7 4695 ext4_setup_system_zone(sb);
0390131b 4696 if (sbi->s_journal == NULL)
e2d67052 4697 ext4_commit_super(sb, 1);
0390131b 4698
7c319d32 4699 unlock_super(sb);
ac27a0ec
DK		 4700#ifdef CONFIG_QUOTA
4701 /* Release old quota file names */
4702 for (i = 0; i < MAXQUOTAS; i++)
4703 if (old_opts.s_qf_names[i] &&
4704 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4705 kfree(old_opts.s_qf_names[i]);
7c319d32
AK		 4706 if (enable_quota) {
4707 if (sb_any_quota_suspended(sb))
4708 dquot_resume(sb, -1);
4709 else if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4710 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
4711 err = ext4_enable_quotas(sb);
4712 if (err) {
4713 lock_super(sb);
4714 goto restore_opts;
4715 }
4716 }
4717 }
ac27a0ec 4718#endif
d4c402d9
CW		 4719
4720 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4721 kfree(orig_data);
ac27a0ec 4722 return 0;
0b8e58a1 4723
ac27a0ec
DK		 4724restore_opts:
4725 sb->s_flags = old_sb_flags;
4726 sbi->s_mount_opt = old_opts.s_mount_opt;
a2595b8a 4727 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
ac27a0ec
DK		 4728 sbi->s_resuid = old_opts.s_resuid;
4729 sbi->s_resgid = old_opts.s_resgid;
4730 sbi->s_commit_interval = old_opts.s_commit_interval;
30773840
TT		 4731 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4732 sbi->s_max_batch_time = old_opts.s_max_batch_time;
ac27a0ec
DK		 4733#ifdef CONFIG_QUOTA
4734 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4735 for (i = 0; i < MAXQUOTAS; i++) {
4736 if (sbi->s_qf_names[i] &&
4737 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4738 kfree(sbi->s_qf_names[i]);
4739 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4740 }
4741#endif
bbd6851a 4742 unlock_super(sb);
d4c402d9 4743 kfree(orig_data);
ac27a0ec
DK		 4744 return err;
4745}
4746
2b2d6d01 4747static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
ac27a0ec
DK		 4748{
4749 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 4750 struct ext4_sb_info *sbi = EXT4_SB(sb);
4751 struct ext4_super_block *es = sbi->s_es;
952fc18e 4752 ext4_fsblk_t overhead = 0;
960cc398 4753 u64 fsid;
d02a9391 4754 s64 bfree;
ac27a0ec 4755
952fc18e
TT		 4756 if (!test_opt(sb, MINIX_DF))
4757 overhead = sbi->s_overhead;
ac27a0ec 4758
617ba13b 4759 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 4760 buf->f_bsize = sb->s_blocksize;
952fc18e 4761 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, sbi->s_overhead);
57042651
TT		 4762 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
4763 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
d02a9391 4764 /* prevent underflow in case that few free space is available */
57042651 4765 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
bd81d8ee
LV		 4766 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
4767 if (buf->f_bfree < ext4_r_blocks_count(es))
ac27a0ec
DK		 4768 buf->f_bavail = 0;
4769 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 4770 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
617ba13b 4771 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 4772 fsid = le64_to_cpup((void *)es->s_uuid) ^
4773 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4774 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4775 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
0b8e58a1 4776
ac27a0ec
DK		 4777 return 0;
4778}
4779
0b8e58a1
AD		 4780/* Helper function for writing quotas on sync - we need to start transaction
4781 * before quota file is locked for write. Otherwise the are possible deadlocks:
ac27a0ec 4782 * Process 1 Process 2
617ba13b 4783 * ext4_create() quota_sync()
a269eb18 4784 * jbd2_journal_start() write_dquot()
871a2931 4785 * dquot_initialize() down(dqio_mutex)
dab291af 4786 * down(dqio_mutex) jbd2_journal_start()
ac27a0ec
DK		 4787 *
4788 */
4789
4790#ifdef CONFIG_QUOTA
4791
4792static inline struct inode *dquot_to_inode(struct dquot *dquot)
4793{
4794 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
4795}
4796
617ba13b 4797static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 4798{
4799 int ret, err;
4800 handle_t *handle;
4801 struct inode *inode;
4802
4803 inode = dquot_to_inode(dquot);
617ba13b 4804 handle = ext4_journal_start(inode,
0b8e58a1 4805 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 4806 if (IS_ERR(handle))
4807 return PTR_ERR(handle);
4808 ret = dquot_commit(dquot);
617ba13b 4809 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4810 if (!ret)
4811 ret = err;
4812 return ret;
4813}
4814
617ba13b 4815static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 4816{
4817 int ret, err;
4818 handle_t *handle;
4819
617ba13b 4820 handle = ext4_journal_start(dquot_to_inode(dquot),
0b8e58a1 4821 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 4822 if (IS_ERR(handle))
4823 return PTR_ERR(handle);
4824 ret = dquot_acquire(dquot);
617ba13b 4825 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4826 if (!ret)
4827 ret = err;
4828 return ret;
4829}
4830
617ba13b 4831static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 4832{
4833 int ret, err;
4834 handle_t *handle;
4835
617ba13b 4836 handle = ext4_journal_start(dquot_to_inode(dquot),
0b8e58a1 4837 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 4838 if (IS_ERR(handle)) {
4839 /* Release dquot anyway to avoid endless cycle in dqput() */
4840 dquot_release(dquot);
ac27a0ec 4841 return PTR_ERR(handle);
9c3013e9 4842 }
ac27a0ec 4843 ret = dquot_release(dquot);
617ba13b 4844 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4845 if (!ret)
4846 ret = err;
4847 return ret;
4848}
4849
617ba13b 4850static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec 4851{
2c8be6b2 4852 /* Are we journaling quotas? */
617ba13b
MC		 4853 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
4854 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
ac27a0ec 4855 dquot_mark_dquot_dirty(dquot);
617ba13b 4856 return ext4_write_dquot(dquot);
ac27a0ec
DK		 4857 } else {
4858 return dquot_mark_dquot_dirty(dquot);
4859 }
4860}
4861
617ba13b 4862static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 4863{
4864 int ret, err;
4865 handle_t *handle;
4866
4867 /* Data block + inode block */
617ba13b 4868 handle = ext4_journal_start(sb->s_root->d_inode, 2);
ac27a0ec
DK		 4869 if (IS_ERR(handle))
4870 return PTR_ERR(handle);
4871 ret = dquot_commit_info(sb, type);
617ba13b 4872 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4873 if (!ret)
4874 ret = err;
4875 return ret;
4876}
4877
4878/*
4879 * Turn on quotas during mount time - we need to find
4880 * the quota file and such...
4881 */
617ba13b 4882static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 4883{
287a8095
CH		 4884 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4885 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 4886}
4887
4888/*
4889 * Standard function to be called on quota_on
4890 */
617ba13b 4891static int ext4_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 4892 struct path *path)
ac27a0ec
DK		 4893{
4894 int err;
ac27a0ec
DK		 4895
4896 if (!test_opt(sb, QUOTA))
4897 return -EINVAL;
0623543b 4898
ac27a0ec 4899 /* Quotafile not on the same filesystem? */
d8c9584e 4900 if (path->dentry->d_sb != sb)
ac27a0ec 4901 return -EXDEV;
0623543b
JK		 4902 /* Journaling quota? */
4903 if (EXT4_SB(sb)->s_qf_names[type]) {
2b2d6d01 4904 /* Quotafile not in fs root? */
f00c9e44 4905 if (path->dentry->d_parent != sb->s_root)
b31e1552
ES		 4906 ext4_msg(sb, KERN_WARNING,
4907 "Quota file not on filesystem root. "
4908 "Journaled quota will not work");
2b2d6d01 4909 }
0623543b
JK		 4910
4911 /*
4912 * When we journal data on quota file, we have to flush journal to see
4913 * all updates to the file when we bypass pagecache...
4914 */
0390131b 4915 if (EXT4_SB(sb)->s_journal &&
f00c9e44 4916 ext4_should_journal_data(path->dentry->d_inode)) {
0623543b
JK		 4917 /*
4918 * We don't need to lock updates but journal_flush() could
4919 * otherwise be livelocked...
4920 */
4921 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 4922 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
0623543b 4923 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
f00c9e44 4924 if (err)
7ffe1ea8 4925 return err;
0623543b
JK		 4926 }
4927
f00c9e44 4928 return dquot_quota_on(sb, type, format_id, path);
ac27a0ec
DK		 4929}
4930
7c319d32
AK		 4931static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
4932 unsigned int flags)
4933{
4934 int err;
4935 struct inode *qf_inode;
4936 unsigned long qf_inums[MAXQUOTAS] = {
4937 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
4938 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
4939 };
4940
4941 BUG_ON(!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA));
4942
4943 if (!qf_inums[type])
4944 return -EPERM;
4945
4946 qf_inode = ext4_iget(sb, qf_inums[type]);
4947 if (IS_ERR(qf_inode)) {
4948 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
4949 return PTR_ERR(qf_inode);
4950 }
4951
4952 err = dquot_enable(qf_inode, type, format_id, flags);
4953 iput(qf_inode);
4954
4955 return err;
4956}
4957
4958/* Enable usage tracking for all quota types. */
4959static int ext4_enable_quotas(struct super_block *sb)
4960{
4961 int type, err = 0;
4962 unsigned long qf_inums[MAXQUOTAS] = {
4963 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
4964 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
4965 };
4966
4967 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
4968 for (type = 0; type < MAXQUOTAS; type++) {
4969 if (qf_inums[type]) {
4970 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
4971 DQUOT_USAGE_ENABLED);
4972 if (err) {
4973 ext4_warning(sb,
4974 "Failed to enable quota (type=%d) "
4975 "tracking. Please run e2fsck to fix.",
4976 type);
4977 return err;
4978 }
4979 }
4980 }
4981 return 0;
4982}
4983
4984/*
4985 * quota_on function that is used when QUOTA feature is set.
4986 */
4987static int ext4_quota_on_sysfile(struct super_block *sb, int type,
4988 int format_id)
4989{
4990 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
4991 return -EINVAL;
4992
4993 /*
4994 * USAGE was enabled at mount time. Only need to enable LIMITS now.
4995 */
4996 return ext4_quota_enable(sb, type, format_id, DQUOT_LIMITS_ENABLED);
4997}
4998
ca0e05e4
DM		 4999static int ext4_quota_off(struct super_block *sb, int type)
5000{
21f97697
JK		 5001 struct inode *inode = sb_dqopt(sb)->files[type];
5002 handle_t *handle;
5003
87009d86
DM		 5004 /* Force all delayed allocation blocks to be allocated.
5005 * Caller already holds s_umount sem */
5006 if (test_opt(sb, DELALLOC))
ca0e05e4 5007 sync_filesystem(sb);
ca0e05e4 5008
0b268590
AG		 5009 if (!inode)
5010 goto out;
5011
21f97697
JK		 5012 /* Update modification times of quota files when userspace can
5013 * start looking at them */
5014 handle = ext4_journal_start(inode, 1);
5015 if (IS_ERR(handle))
5016 goto out;
5017 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
5018 ext4_mark_inode_dirty(handle, inode);
5019 ext4_journal_stop(handle);
5020
5021out:
ca0e05e4
DM		 5022 return dquot_quota_off(sb, type);
5023}
5024
7c319d32
AK		 5025/*
5026 * quota_off function that is used when QUOTA feature is set.
5027 */
5028static int ext4_quota_off_sysfile(struct super_block *sb, int type)
5029{
5030 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5031 return -EINVAL;
5032
5033 /* Disable only the limits. */
5034 return dquot_disable(sb, type, DQUOT_LIMITS_ENABLED);
5035}
5036
ac27a0ec
DK		 5037/* Read data from quotafile - avoid pagecache and such because we cannot afford
5038 * acquiring the locks... As quota files are never truncated and quota code
25985edc 5039 * itself serializes the operations (and no one else should touch the files)
ac27a0ec 5040 * we don't have to be afraid of races */
617ba13b 5041static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 5042 size_t len, loff_t off)
5043{
5044 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 5045 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 5046 int err = 0;
5047 int offset = off & (sb->s_blocksize - 1);
5048 int tocopy;
5049 size_t toread;
5050 struct buffer_head *bh;
5051 loff_t i_size = i_size_read(inode);
5052
5053 if (off > i_size)
5054 return 0;
5055 if (off+len > i_size)
5056 len = i_size-off;
5057 toread = len;
5058 while (toread > 0) {
5059 tocopy = sb->s_blocksize - offset < toread ?
5060 sb->s_blocksize - offset : toread;
617ba13b 5061 bh = ext4_bread(NULL, inode, blk, 0, &err);
ac27a0ec
DK		 5062 if (err)
5063 return err;
5064 if (!bh) /* A hole? */
5065 memset(data, 0, tocopy);
5066 else
5067 memcpy(data, bh->b_data+offset, tocopy);
5068 brelse(bh);
5069 offset = 0;
5070 toread -= tocopy;
5071 data += tocopy;
5072 blk++;
5073 }
5074 return len;
5075}
5076
5077/* Write to quotafile (we know the transaction is already started and has
5078 * enough credits) */
617ba13b 5079static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 5080 const char *data, size_t len, loff_t off)
5081{
5082 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 5083 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 5084 int err = 0;
5085 int offset = off & (sb->s_blocksize - 1);
ac27a0ec
DK		 5086 struct buffer_head *bh;
5087 handle_t *handle = journal_current_handle();
5088
0390131b 5089 if (EXT4_SB(sb)->s_journal && !handle) {
b31e1552
ES		 5090 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5091 " cancelled because transaction is not started",
9c3013e9
JK		 5092 (unsigned long long)off, (unsigned long long)len);
5093 return -EIO;
5094 }
67eeb568
DM		 5095 /*
5096 * Since we account only one data block in transaction credits,
5097 * then it is impossible to cross a block boundary.
5098 */
5099 if (sb->s_blocksize - offset < len) {
5100 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5101 " cancelled because not block aligned",
5102 (unsigned long long)off, (unsigned long long)len);
5103 return -EIO;
5104 }
5105
67eeb568
DM		 5106 bh = ext4_bread(handle, inode, blk, 1, &err);
5107 if (!bh)
5108 goto out;
62d2b5f2
JK		 5109 err = ext4_journal_get_write_access(handle, bh);
5110 if (err) {
5111 brelse(bh);
5112 goto out;
ac27a0ec 5113 }
67eeb568
DM		 5114 lock_buffer(bh);
5115 memcpy(bh->b_data+offset, data, len);
5116 flush_dcache_page(bh->b_page);
5117 unlock_buffer(bh);
62d2b5f2 5118 err = ext4_handle_dirty_metadata(handle, NULL, bh);
67eeb568 5119 brelse(bh);
ac27a0ec 5120out:
0b7f7cef 5121 if (err)
ac27a0ec 5122 return err;
67eeb568
DM		 5123 if (inode->i_size < off + len) {
5124 i_size_write(inode, off + len);
617ba13b 5125 EXT4_I(inode)->i_disksize = inode->i_size;
21f97697 5126 ext4_mark_inode_dirty(handle, inode);
ac27a0ec 5127 }
67eeb568 5128 return len;
ac27a0ec
DK		 5129}
5130
5131#endif
5132
152a0836
AV		 5133static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5134 const char *dev_name, void *data)
ac27a0ec 5135{
152a0836 5136 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
ac27a0ec
DK		 5137}
5138
37f328eb 5139#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
24b58424
TT		 5140static inline void register_as_ext2(void)
5141{
5142 int err = register_filesystem(&ext2_fs_type);
5143 if (err)
5144 printk(KERN_WARNING
5145 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5146}
5147
5148static inline void unregister_as_ext2(void)
5149{
5150 unregister_filesystem(&ext2_fs_type);
5151}
2035e776
TT		 5152
5153static inline int ext2_feature_set_ok(struct super_block *sb)
5154{
5155 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
5156 return 0;
5157 if (sb->s_flags & MS_RDONLY)
5158 return 1;
5159 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
5160 return 0;
5161 return 1;
5162}
51b7e3c9 5163MODULE_ALIAS("ext2");
24b58424
TT		 5164#else
5165static inline void register_as_ext2(void) { }
5166static inline void unregister_as_ext2(void) { }
2035e776 5167static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 5168#endif
5169
37f328eb 5170#if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
24b58424
TT		 5171static inline void register_as_ext3(void)
5172{
5173 int err = register_filesystem(&ext3_fs_type);
5174 if (err)
5175 printk(KERN_WARNING
5176 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5177}
5178
5179static inline void unregister_as_ext3(void)
5180{
5181 unregister_filesystem(&ext3_fs_type);
5182}
2035e776
TT		 5183
5184static inline int ext3_feature_set_ok(struct super_block *sb)
5185{
5186 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
5187 return 0;
5188 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
5189 return 0;
5190 if (sb->s_flags & MS_RDONLY)
5191 return 1;
5192 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
5193 return 0;
5194 return 1;
5195}
51b7e3c9 5196MODULE_ALIAS("ext3");
24b58424
TT		 5197#else
5198static inline void register_as_ext3(void) { }
5199static inline void unregister_as_ext3(void) { }
2035e776 5200static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 5201#endif
5202
03010a33
TT		 5203static struct file_system_type ext4_fs_type = {
5204 .owner = THIS_MODULE,
5205 .name = "ext4",
152a0836 5206 .mount = ext4_mount,
03010a33
TT		 5207 .kill_sb = kill_block_super,
5208 .fs_flags = FS_REQUIRES_DEV,
5209};
5210
8f021222 5211static int __init ext4_init_feat_adverts(void)
857ac889
LC		 5212{
5213 struct ext4_features *ef;
5214 int ret = -ENOMEM;
5215
5216 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
5217 if (!ef)
5218 goto out;
5219
5220 ef->f_kobj.kset = ext4_kset;
5221 init_completion(&ef->f_kobj_unregister);
5222 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
5223 "features");
5224 if (ret) {
5225 kfree(ef);
5226 goto out;
5227 }
5228
5229 ext4_feat = ef;
5230 ret = 0;
5231out:
5232 return ret;
5233}
5234
8f021222
LC		 5235static void ext4_exit_feat_adverts(void)
5236{
5237 kobject_put(&ext4_feat->f_kobj);
5238 wait_for_completion(&ext4_feat->f_kobj_unregister);
5239 kfree(ext4_feat);
5240}
5241
e9e3bcec
ES		 5242/* Shared across all ext4 file systems */
5243wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5244struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5245
5dabfc78 5246static int __init ext4_init_fs(void)
ac27a0ec 5247{
e9e3bcec 5248 int i, err;
c9de560d 5249
07c0c5d8
AV		 5250 ext4_li_info = NULL;
5251 mutex_init(&ext4_li_mtx);
5252
12e9b892 5253 ext4_check_flag_values();
e9e3bcec
ES		 5254
5255 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5256 mutex_init(&ext4__aio_mutex[i]);
5257 init_waitqueue_head(&ext4__ioend_wq[i]);
5258 }
5259
5dabfc78 5260 err = ext4_init_pageio();
6fd058f7
TT		 5261 if (err)
5262 return err;
5dabfc78 5263 err = ext4_init_system_zone();
bd2d0210 5264 if (err)
d44651d0 5265 goto out6;
3197ebdb
TT		 5266 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
5267 if (!ext4_kset)
dd68314c 5268 goto out5;
d44651d0 5269 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
857ac889
LC		 5270
5271 err = ext4_init_feat_adverts();
dd68314c
TT		 5272 if (err)
5273 goto out4;
857ac889 5274
5dabfc78 5275 err = ext4_init_mballoc();
ac27a0ec 5276 if (err)
6fd058f7 5277 goto out3;
c9de560d 5278
5dabfc78 5279 err = ext4_init_xattr();
c9de560d
AT		 5280 if (err)
5281 goto out2;
ac27a0ec
DK		 5282 err = init_inodecache();
5283 if (err)
5284 goto out1;
24b58424 5285 register_as_ext3();
2035e776 5286 register_as_ext2();
03010a33 5287 err = register_filesystem(&ext4_fs_type);
ac27a0ec
DK		 5288 if (err)
5289 goto out;
bfff6873 5290
ac27a0ec
DK		 5291 return 0;
5292out:
24b58424
TT		 5293 unregister_as_ext2();
5294 unregister_as_ext3();
ac27a0ec
DK		 5295 destroy_inodecache();
5296out1:
5dabfc78 5297 ext4_exit_xattr();
c9de560d 5298out2:
5dabfc78 5299 ext4_exit_mballoc();
6fd058f7 5300out3:
8f021222 5301 ext4_exit_feat_adverts();
dd68314c 5302out4:
d44651d0
FJ		 5303 if (ext4_proc_root)
5304 remove_proc_entry("fs/ext4", NULL);
6fd058f7 5305 kset_unregister(ext4_kset);
d44651d0 5306out5:
5dabfc78 5307 ext4_exit_system_zone();
d44651d0 5308out6:
5dabfc78 5309 ext4_exit_pageio();
ac27a0ec
DK		 5310 return err;
5311}
5312
5dabfc78 5313static void __exit ext4_exit_fs(void)
ac27a0ec 5314{
bfff6873 5315 ext4_destroy_lazyinit_thread();
24b58424
TT		 5316 unregister_as_ext2();
5317 unregister_as_ext3();
03010a33 5318 unregister_filesystem(&ext4_fs_type);
ac27a0ec 5319 destroy_inodecache();
5dabfc78
TT		 5320 ext4_exit_xattr();
5321 ext4_exit_mballoc();
8f021222 5322 ext4_exit_feat_adverts();
9f6200bb 5323 remove_proc_entry("fs/ext4", NULL);
3197ebdb 5324 kset_unregister(ext4_kset);
5dabfc78
TT		 5325 ext4_exit_system_zone();
5326 ext4_exit_pageio();
ac27a0ec
DK		 5327}
5328
5329MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
83982b6f 5330MODULE_DESCRIPTION("Fourth Extended Filesystem");
ac27a0ec 5331MODULE_LICENSE("GPL");
5dabfc78
TT		 5332module_init(ext4_init_fs)
5333module_exit(ext4_exit_fs)
Linux 6.x block layer and io_uring tree(s)