]> git.ipfire.org Git - thirdparty/kernel/linux.git/blame - fs/ext4/super.c
projects / thirdparty / kernel / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
ext4: Fix typo in ext4/Kconfig
[thirdparty/kernel/linux.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>
29#include <linux/smp_lock.h>
30#include <linux/buffer_head.h>
a5694255 31#include <linux/exportfs.h>
ac27a0ec
DK		 32#include <linux/vfs.h>
33#include <linux/random.h>
34#include <linux/mount.h>
35#include <linux/namei.h>
36#include <linux/quotaops.h>
37#include <linux/seq_file.h>
9f6200bb 38#include <linux/proc_fs.h>
3197ebdb 39#include <linux/ctype.h>
1330593e 40#include <linux/log2.h>
717d50e4 41#include <linux/crc16.h>
ac27a0ec
DK		 42#include <asm/uaccess.h>
43
3dcf5451
CH		 44#include "ext4.h"
45#include "ext4_jbd2.h"
ac27a0ec
DK		 46#include "xattr.h"
47#include "acl.h"
ac27a0ec 48
9bffad1e
TT		 49#define CREATE_TRACE_POINTS
50#include <trace/events/ext4.h>
51
f4033903
CW		 52static int default_mb_history_length = 1000;
53
54module_param_named(default_mb_history_length, default_mb_history_length,
55 int, 0644);
56MODULE_PARM_DESC(default_mb_history_length,
57 "Default number of entries saved for mb_history");
58
9f6200bb 59struct proc_dir_entry *ext4_proc_root;
3197ebdb 60static struct kset *ext4_kset;
9f6200bb 61
617ba13b 62static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
ac27a0ec 63 unsigned long journal_devnum);
e2d67052 64static int ext4_commit_super(struct super_block *sb, int sync);
2b2d6d01
TT		 65static void ext4_mark_recovery_complete(struct super_block *sb,
66 struct ext4_super_block *es);
67static void ext4_clear_journal_err(struct super_block *sb,
68 struct ext4_super_block *es);
617ba13b 69static int ext4_sync_fs(struct super_block *sb, int wait);
2b2d6d01 70static const char *ext4_decode_error(struct super_block *sb, int errno,
ac27a0ec 71 char nbuf[16]);
2b2d6d01
TT		 72static int ext4_remount(struct super_block *sb, int *flags, char *data);
73static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
c4be0c1d 74static int ext4_unfreeze(struct super_block *sb);
2b2d6d01 75static void ext4_write_super(struct super_block *sb);
c4be0c1d 76static int ext4_freeze(struct super_block *sb);
ac27a0ec 77
bd81d8ee 78
8fadc143
AR		 79ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
80 struct ext4_group_desc *bg)
bd81d8ee 81{
3a14589c 82 return le32_to_cpu(bg->bg_block_bitmap_lo) |
8fadc143 83 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 84 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 85}
86
8fadc143
AR		 87ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
88 struct ext4_group_desc *bg)
bd81d8ee 89{
5272f837 90 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
8fadc143 91 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 92 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 93}
94
8fadc143
AR		 95ext4_fsblk_t ext4_inode_table(struct super_block *sb,
96 struct ext4_group_desc *bg)
bd81d8ee 97{
5272f837 98 return le32_to_cpu(bg->bg_inode_table_lo) |
8fadc143 99 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 100 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
bd81d8ee
LV		 101}
102
560671a0
AK		 103__u32 ext4_free_blks_count(struct super_block *sb,
104 struct ext4_group_desc *bg)
105{
106 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
107 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 108 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
560671a0
AK		 109}
110
111__u32 ext4_free_inodes_count(struct super_block *sb,
112 struct ext4_group_desc *bg)
113{
114 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
115 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 116 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
560671a0
AK		 117}
118
119__u32 ext4_used_dirs_count(struct super_block *sb,
120 struct ext4_group_desc *bg)
121{
122 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
123 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 124 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
560671a0
AK		 125}
126
127__u32 ext4_itable_unused_count(struct super_block *sb,
128 struct ext4_group_desc *bg)
129{
130 return le16_to_cpu(bg->bg_itable_unused_lo) |
131 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 132 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
560671a0
AK		 133}
134
8fadc143
AR		 135void ext4_block_bitmap_set(struct super_block *sb,
136 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 137{
3a14589c 138 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 139 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
140 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 141}
142
8fadc143
AR		 143void ext4_inode_bitmap_set(struct super_block *sb,
144 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 145{
5272f837 146 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 147 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
148 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 149}
150
8fadc143
AR		 151void ext4_inode_table_set(struct super_block *sb,
152 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 153{
5272f837 154 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 155 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
156 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 157}
158
560671a0
AK		 159void ext4_free_blks_set(struct super_block *sb,
160 struct ext4_group_desc *bg, __u32 count)
161{
162 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
163 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
164 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
165}
166
167void ext4_free_inodes_set(struct super_block *sb,
168 struct ext4_group_desc *bg, __u32 count)
169{
170 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
171 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
172 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
173}
174
175void ext4_used_dirs_set(struct super_block *sb,
176 struct ext4_group_desc *bg, __u32 count)
177{
178 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
179 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
180 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
181}
182
183void ext4_itable_unused_set(struct super_block *sb,
184 struct ext4_group_desc *bg, __u32 count)
185{
186 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
187 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
188 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
189}
190
ac27a0ec 191/*
dab291af 192 * Wrappers for jbd2_journal_start/end.
ac27a0ec
DK		 193 *
194 * The only special thing we need to do here is to make sure that all
195 * journal_end calls result in the superblock being marked dirty, so
196 * that sync() will call the filesystem's write_super callback if
197 * appropriate.
198 */
617ba13b 199handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
ac27a0ec
DK		 200{
201 journal_t *journal;
202
203 if (sb->s_flags & MS_RDONLY)
204 return ERR_PTR(-EROFS);
205
206 /* Special case here: if the journal has aborted behind our
207 * backs (eg. EIO in the commit thread), then we still need to
208 * take the FS itself readonly cleanly. */
617ba13b 209 journal = EXT4_SB(sb)->s_journal;
0390131b
FM		 210 if (journal) {
211 if (is_journal_aborted(journal)) {
0b8e58a1 212 ext4_abort(sb, __func__, "Detected aborted journal");
0390131b
FM		 213 return ERR_PTR(-EROFS);
214 }
215 return jbd2_journal_start(journal, nblocks);
ac27a0ec 216 }
0390131b
FM		 217 /*
218 * We're not journaling, return the appropriate indication.
219 */
220 current->journal_info = EXT4_NOJOURNAL_HANDLE;
221 return current->journal_info;
ac27a0ec
DK		 222}
223
224/*
225 * The only special thing we need to do here is to make sure that all
dab291af 226 * jbd2_journal_stop calls result in the superblock being marked dirty, so
ac27a0ec
DK		 227 * that sync() will call the filesystem's write_super callback if
228 * appropriate.
229 */
617ba13b 230int __ext4_journal_stop(const char *where, handle_t *handle)
ac27a0ec
DK		 231{
232 struct super_block *sb;
233 int err;
234 int rc;
235
0390131b
FM		 236 if (!ext4_handle_valid(handle)) {
237 /*
238 * Do this here since we don't call jbd2_journal_stop() in
239 * no-journal mode.
240 */
241 current->journal_info = NULL;
242 return 0;
243 }
ac27a0ec
DK		 244 sb = handle->h_transaction->t_journal->j_private;
245 err = handle->h_err;
dab291af 246 rc = jbd2_journal_stop(handle);
ac27a0ec
DK		 247
248 if (!err)
249 err = rc;
250 if (err)
617ba13b 251 __ext4_std_error(sb, where, err);
ac27a0ec
DK		 252 return err;
253}
254
617ba13b 255void ext4_journal_abort_handle(const char *caller, const char *err_fn,
ac27a0ec
DK		 256 struct buffer_head *bh, handle_t *handle, int err)
257{
258 char nbuf[16];
617ba13b 259 const char *errstr = ext4_decode_error(NULL, err, nbuf);
ac27a0ec 260
0390131b
FM		 261 BUG_ON(!ext4_handle_valid(handle));
262
ac27a0ec
DK		 263 if (bh)
264 BUFFER_TRACE(bh, "abort");
265
266 if (!handle->h_err)
267 handle->h_err = err;
268
269 if (is_handle_aborted(handle))
270 return;
271
272 printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
273 caller, errstr, err_fn);
274
dab291af 275 jbd2_journal_abort_handle(handle);
ac27a0ec
DK		 276}
277
278/* Deal with the reporting of failure conditions on a filesystem such as
279 * inconsistencies detected or read IO failures.
280 *
281 * On ext2, we can store the error state of the filesystem in the
617ba13b 282 * superblock. That is not possible on ext4, because we may have other
ac27a0ec
DK		 283 * write ordering constraints on the superblock which prevent us from
284 * writing it out straight away; and given that the journal is about to
285 * be aborted, we can't rely on the current, or future, transactions to
286 * write out the superblock safely.
287 *
dab291af 288 * We'll just use the jbd2_journal_abort() error code to record an error in
ac27a0ec
DK		 289 * the journal instead. On recovery, the journal will compain about
290 * that error until we've noted it down and cleared it.
291 */
292
617ba13b 293static void ext4_handle_error(struct super_block *sb)
ac27a0ec 294{
617ba13b 295 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 296
617ba13b
MC		 297 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
298 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
ac27a0ec
DK		 299
300 if (sb->s_flags & MS_RDONLY)
301 return;
302
2b2d6d01 303 if (!test_opt(sb, ERRORS_CONT)) {
617ba13b 304 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 305
4ab2f15b 306 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
ac27a0ec 307 if (journal)
dab291af 308 jbd2_journal_abort(journal, -EIO);
ac27a0ec 309 }
2b2d6d01 310 if (test_opt(sb, ERRORS_RO)) {
b31e1552 311 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
ac27a0ec
DK		 312 sb->s_flags |= MS_RDONLY;
313 }
e2d67052 314 ext4_commit_super(sb, 1);
ac27a0ec 315 if (test_opt(sb, ERRORS_PANIC))
617ba13b 316 panic("EXT4-fs (device %s): panic forced after error\n",
ac27a0ec
DK		 317 sb->s_id);
318}
319
2b2d6d01
TT		 320void ext4_error(struct super_block *sb, const char *function,
321 const char *fmt, ...)
ac27a0ec
DK		 322{
323 va_list args;
324
325 va_start(args, fmt);
2b2d6d01 326 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
ac27a0ec
DK		 327 vprintk(fmt, args);
328 printk("\n");
329 va_end(args);
330
617ba13b 331 ext4_handle_error(sb);
ac27a0ec
DK		 332}
333
2b2d6d01 334static const char *ext4_decode_error(struct super_block *sb, int errno,
ac27a0ec
DK		 335 char nbuf[16])
336{
337 char *errstr = NULL;
338
339 switch (errno) {
340 case -EIO:
341 errstr = "IO failure";
342 break;
343 case -ENOMEM:
344 errstr = "Out of memory";
345 break;
346 case -EROFS:
dab291af 347 if (!sb || EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT)
ac27a0ec
DK		 348 errstr = "Journal has aborted";
349 else
350 errstr = "Readonly filesystem";
351 break;
352 default:
353 /* If the caller passed in an extra buffer for unknown
354 * errors, textualise them now. Else we just return
355 * NULL. */
356 if (nbuf) {
357 /* Check for truncated error codes... */
358 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
359 errstr = nbuf;
360 }
361 break;
362 }
363
364 return errstr;
365}
366
617ba13b 367/* __ext4_std_error decodes expected errors from journaling functions
ac27a0ec
DK		 368 * automatically and invokes the appropriate error response. */
369
2b2d6d01 370void __ext4_std_error(struct super_block *sb, const char *function, int errno)
ac27a0ec
DK		 371{
372 char nbuf[16];
373 const char *errstr;
374
375 /* Special case: if the error is EROFS, and we're not already
376 * inside a transaction, then there's really no point in logging
377 * an error. */
378 if (errno == -EROFS && journal_current_handle() == NULL &&
379 (sb->s_flags & MS_RDONLY))
380 return;
381
617ba13b 382 errstr = ext4_decode_error(sb, errno, nbuf);
2b2d6d01
TT		 383 printk(KERN_CRIT "EXT4-fs error (device %s) in %s: %s\n",
384 sb->s_id, function, errstr);
ac27a0ec 385
617ba13b 386 ext4_handle_error(sb);
ac27a0ec
DK		 387}
388
389/*
617ba13b 390 * ext4_abort is a much stronger failure handler than ext4_error. The
ac27a0ec
DK		 391 * abort function may be used to deal with unrecoverable failures such
392 * as journal IO errors or ENOMEM at a critical moment in log management.
393 *
394 * We unconditionally force the filesystem into an ABORT|READONLY state,
395 * unless the error response on the fs has been set to panic in which
396 * case we take the easy way out and panic immediately.
397 */
398
2b2d6d01
TT		 399void ext4_abort(struct super_block *sb, const char *function,
400 const char *fmt, ...)
ac27a0ec
DK		 401{
402 va_list args;
403
ac27a0ec 404 va_start(args, fmt);
2b2d6d01 405 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
ac27a0ec
DK		 406 vprintk(fmt, args);
407 printk("\n");
408 va_end(args);
409
410 if (test_opt(sb, ERRORS_PANIC))
617ba13b 411 panic("EXT4-fs panic from previous error\n");
ac27a0ec
DK		 412
413 if (sb->s_flags & MS_RDONLY)
414 return;
415
b31e1552 416 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
617ba13b 417 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
ac27a0ec 418 sb->s_flags |= MS_RDONLY;
4ab2f15b 419 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
ef2cabf7
HK		 420 if (EXT4_SB(sb)->s_journal)
421 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
ac27a0ec
DK		 422}
423
b31e1552
ES		 424void ext4_msg (struct super_block * sb, const char *prefix,
425 const char *fmt, ...)
426{
427 va_list args;
428
429 va_start(args, fmt);
430 printk("%sEXT4-fs (%s): ", prefix, sb->s_id);
431 vprintk(fmt, args);
432 printk("\n");
433 va_end(args);
434}
435
2b2d6d01
TT		 436void ext4_warning(struct super_block *sb, const char *function,
437 const char *fmt, ...)
ac27a0ec
DK		 438{
439 va_list args;
440
441 va_start(args, fmt);
617ba13b 442 printk(KERN_WARNING "EXT4-fs warning (device %s): %s: ",
ac27a0ec
DK		 443 sb->s_id, function);
444 vprintk(fmt, args);
445 printk("\n");
446 va_end(args);
447}
448
5d1b1b3f 449void ext4_grp_locked_error(struct super_block *sb, ext4_group_t grp,
0b8e58a1 450 const char *function, const char *fmt, ...)
5d1b1b3f
AK		 451__releases(bitlock)
452__acquires(bitlock)
453{
454 va_list args;
455 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
456
457 va_start(args, fmt);
458 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
459 vprintk(fmt, args);
460 printk("\n");
461 va_end(args);
462
463 if (test_opt(sb, ERRORS_CONT)) {
464 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
465 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
e2d67052 466 ext4_commit_super(sb, 0);
5d1b1b3f
AK		 467 return;
468 }
469 ext4_unlock_group(sb, grp);
470 ext4_handle_error(sb);
471 /*
472 * We only get here in the ERRORS_RO case; relocking the group
473 * may be dangerous, but nothing bad will happen since the
474 * filesystem will have already been marked read/only and the
475 * journal has been aborted. We return 1 as a hint to callers
476 * who might what to use the return value from
477 * ext4_grp_locked_error() to distinguish beween the
478 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
479 * aggressively from the ext4 function in question, with a
480 * more appropriate error code.
481 */
482 ext4_lock_group(sb, grp);
483 return;
484}
485
617ba13b 486void ext4_update_dynamic_rev(struct super_block *sb)
ac27a0ec 487{
617ba13b 488 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 489
617ba13b 490 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
ac27a0ec
DK		 491 return;
492
46e665e9 493 ext4_warning(sb, __func__,
ac27a0ec
DK		 494 "updating to rev %d because of new feature flag, "
495 "running e2fsck is recommended",
617ba13b 496 EXT4_DYNAMIC_REV);
ac27a0ec 497
617ba13b
MC		 498 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
499 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
500 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
ac27a0ec
DK		 501 /* leave es->s_feature_*compat flags alone */
502 /* es->s_uuid will be set by e2fsck if empty */
503
504 /*
505 * The rest of the superblock fields should be zero, and if not it
506 * means they are likely already in use, so leave them alone. We
507 * can leave it up to e2fsck to clean up any inconsistencies there.
508 */
509}
510
511/*
512 * Open the external journal device
513 */
b31e1552 514static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
ac27a0ec
DK		 515{
516 struct block_device *bdev;
517 char b[BDEVNAME_SIZE];
518
519 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
520 if (IS_ERR(bdev))
521 goto fail;
522 return bdev;
523
524fail:
b31e1552 525 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
ac27a0ec
DK		 526 __bdevname(dev, b), PTR_ERR(bdev));
527 return NULL;
528}
529
530/*
531 * Release the journal device
532 */
617ba13b 533static int ext4_blkdev_put(struct block_device *bdev)
ac27a0ec
DK		 534{
535 bd_release(bdev);
9a1c3542 536 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
ac27a0ec
DK		 537}
538
617ba13b 539static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
ac27a0ec
DK		 540{
541 struct block_device *bdev;
542 int ret = -ENODEV;
543
544 bdev = sbi->journal_bdev;
545 if (bdev) {
617ba13b 546 ret = ext4_blkdev_put(bdev);
ac27a0ec
DK		 547 sbi->journal_bdev = NULL;
548 }
549 return ret;
550}
551
552static inline struct inode *orphan_list_entry(struct list_head *l)
553{
617ba13b 554 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
ac27a0ec
DK		 555}
556
617ba13b 557static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
ac27a0ec
DK		 558{
559 struct list_head *l;
560
b31e1552
ES		 561 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
562 le32_to_cpu(sbi->s_es->s_last_orphan));
ac27a0ec
DK		 563
564 printk(KERN_ERR "sb_info orphan list:\n");
565 list_for_each(l, &sbi->s_orphan) {
566 struct inode *inode = orphan_list_entry(l);
567 printk(KERN_ERR " "
568 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
569 inode->i_sb->s_id, inode->i_ino, inode,
570 inode->i_mode, inode->i_nlink,
571 NEXT_ORPHAN(inode));
572 }
573}
574
2b2d6d01 575static void ext4_put_super(struct super_block *sb)
ac27a0ec 576{
617ba13b
MC		 577 struct ext4_sb_info *sbi = EXT4_SB(sb);
578 struct ext4_super_block *es = sbi->s_es;
ef2cabf7 579 int i, err;
ac27a0ec 580
a9e220f8 581 lock_super(sb);
6cfd0148 582 lock_kernel();
8c85e125 583 if (sb->s_dirt)
ebc1ac16 584 ext4_commit_super(sb, 1);
8c85e125 585
6fd058f7 586 ext4_release_system_zone(sb);
c9de560d 587 ext4_mb_release(sb);
a86c6181 588 ext4_ext_release(sb);
617ba13b 589 ext4_xattr_put_super(sb);
0390131b
FM		 590 if (sbi->s_journal) {
591 err = jbd2_journal_destroy(sbi->s_journal);
592 sbi->s_journal = NULL;
593 if (err < 0)
594 ext4_abort(sb, __func__,
595 "Couldn't clean up the journal");
596 }
ac27a0ec 597 if (!(sb->s_flags & MS_RDONLY)) {
617ba13b 598 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 599 es->s_state = cpu_to_le16(sbi->s_mount_state);
e2d67052 600 ext4_commit_super(sb, 1);
ac27a0ec 601 }
240799cd 602 if (sbi->s_proc) {
9f6200bb 603 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 604 }
3197ebdb 605 kobject_del(&sbi->s_kobj);
ac27a0ec
DK		 606
607 for (i = 0; i < sbi->s_gdb_count; i++)
608 brelse(sbi->s_group_desc[i]);
609 kfree(sbi->s_group_desc);
c5ca7c76
TT		 610 if (is_vmalloc_addr(sbi->s_flex_groups))
611 vfree(sbi->s_flex_groups);
612 else
613 kfree(sbi->s_flex_groups);
ac27a0ec
DK		 614 percpu_counter_destroy(&sbi->s_freeblocks_counter);
615 percpu_counter_destroy(&sbi->s_freeinodes_counter);
616 percpu_counter_destroy(&sbi->s_dirs_counter);
6bc6e63f 617 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
ac27a0ec
DK		 618 brelse(sbi->s_sbh);
619#ifdef CONFIG_QUOTA
620 for (i = 0; i < MAXQUOTAS; i++)
621 kfree(sbi->s_qf_names[i]);
622#endif
623
624 /* Debugging code just in case the in-memory inode orphan list
625 * isn't empty. The on-disk one can be non-empty if we've
626 * detected an error and taken the fs readonly, but the
627 * in-memory list had better be clean by this point. */
628 if (!list_empty(&sbi->s_orphan))
629 dump_orphan_list(sb, sbi);
630 J_ASSERT(list_empty(&sbi->s_orphan));
631
f98393a6 632 invalidate_bdev(sb->s_bdev);
ac27a0ec
DK		 633 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
634 /*
635 * Invalidate the journal device's buffers. We don't want them
636 * floating about in memory - the physical journal device may
637 * hotswapped, and it breaks the `ro-after' testing code.
638 */
639 sync_blockdev(sbi->journal_bdev);
f98393a6 640 invalidate_bdev(sbi->journal_bdev);
617ba13b 641 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 642 }
643 sb->s_fs_info = NULL;
3197ebdb
TT		 644 /*
645 * Now that we are completely done shutting down the
646 * superblock, we need to actually destroy the kobject.
647 */
648 unlock_kernel();
649 unlock_super(sb);
650 kobject_put(&sbi->s_kobj);
651 wait_for_completion(&sbi->s_kobj_unregister);
705895b6 652 kfree(sbi->s_blockgroup_lock);
ac27a0ec 653 kfree(sbi);
ac27a0ec
DK		 654}
655
e18b890b 656static struct kmem_cache *ext4_inode_cachep;
ac27a0ec
DK		 657
658/*
659 * Called inside transaction, so use GFP_NOFS
660 */
617ba13b 661static struct inode *ext4_alloc_inode(struct super_block *sb)
ac27a0ec 662{
617ba13b 663 struct ext4_inode_info *ei;
ac27a0ec 664
e6b4f8da 665 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
ac27a0ec
DK		 666 if (!ei)
667 return NULL;
0b8e58a1 668
ac27a0ec 669 ei->vfs_inode.i_version = 1;
91246c00 670 ei->vfs_inode.i_data.writeback_index = 0;
a86c6181 671 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
c9de560d
AT		 672 INIT_LIST_HEAD(&ei->i_prealloc_list);
673 spin_lock_init(&ei->i_prealloc_lock);
0390131b
FM		 674 /*
675 * Note: We can be called before EXT4_SB(sb)->s_journal is set,
676 * therefore it can be null here. Don't check it, just initialize
677 * jinode.
678 */
678aaf48 679 jbd2_journal_init_jbd_inode(&ei->jinode, &ei->vfs_inode);
d2a17637
MC		 680 ei->i_reserved_data_blocks = 0;
681 ei->i_reserved_meta_blocks = 0;
682 ei->i_allocated_meta_blocks = 0;
683 ei->i_delalloc_reserved_flag = 0;
684 spin_lock_init(&(ei->i_block_reservation_lock));
0b8e58a1 685
ac27a0ec
DK		 686 return &ei->vfs_inode;
687}
688
617ba13b 689static void ext4_destroy_inode(struct inode *inode)
ac27a0ec 690{
9f7dd93d 691 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
b31e1552
ES		 692 ext4_msg(inode->i_sb, KERN_ERR,
693 "Inode %lu (%p): orphan list check failed!",
694 inode->i_ino, EXT4_I(inode));
9f7dd93d
VA		 695 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
696 EXT4_I(inode), sizeof(struct ext4_inode_info),
697 true);
698 dump_stack();
699 }
617ba13b 700 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
ac27a0ec
DK		 701}
702
51cc5068 703static void init_once(void *foo)
ac27a0ec 704{
617ba13b 705 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
ac27a0ec 706
a35afb83 707 INIT_LIST_HEAD(&ei->i_orphan);
03010a33 708#ifdef CONFIG_EXT4_FS_XATTR
a35afb83 709 init_rwsem(&ei->xattr_sem);
ac27a0ec 710#endif
0e855ac8 711 init_rwsem(&ei->i_data_sem);
a35afb83 712 inode_init_once(&ei->vfs_inode);
ac27a0ec
DK		 713}
714
715static int init_inodecache(void)
716{
617ba13b
MC		 717 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
718 sizeof(struct ext4_inode_info),
ac27a0ec
DK		 719 0, (SLAB_RECLAIM_ACCOUNT|
720 SLAB_MEM_SPREAD),
20c2df83 721 init_once);
617ba13b 722 if (ext4_inode_cachep == NULL)
ac27a0ec
DK		 723 return -ENOMEM;
724 return 0;
725}
726
727static void destroy_inodecache(void)
728{
617ba13b 729 kmem_cache_destroy(ext4_inode_cachep);
ac27a0ec
DK		 730}
731
617ba13b 732static void ext4_clear_inode(struct inode *inode)
ac27a0ec 733{
c2ea3fde 734 ext4_discard_preallocations(inode);
0390131b
FM		 735 if (EXT4_JOURNAL(inode))
736 jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal,
678aaf48 737 &EXT4_I(inode)->jinode);
ac27a0ec
DK		 738}
739
2b2d6d01
TT		 740static inline void ext4_show_quota_options(struct seq_file *seq,
741 struct super_block *sb)
ac27a0ec
DK		 742{
743#if defined(CONFIG_QUOTA)
617ba13b 744 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 745
746 if (sbi->s_jquota_fmt)
747 seq_printf(seq, ",jqfmt=%s",
af5bc92d 748 (sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold" : "vfsv0");
ac27a0ec
DK		 749
750 if (sbi->s_qf_names[USRQUOTA])
751 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
752
753 if (sbi->s_qf_names[GRPQUOTA])
754 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
755
617ba13b 756 if (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA)
ac27a0ec
DK		 757 seq_puts(seq, ",usrquota");
758
617ba13b 759 if (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)
ac27a0ec
DK		 760 seq_puts(seq, ",grpquota");
761#endif
762}
763
d9c9bef1
MS		 764/*
765 * Show an option if
766 * - it's set to a non-default value OR
767 * - if the per-sb default is different from the global default
768 */
617ba13b 769static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
ac27a0ec 770{
aa22df2c
AK		 771 int def_errors;
772 unsigned long def_mount_opts;
ac27a0ec 773 struct super_block *sb = vfs->mnt_sb;
d9c9bef1
MS		 774 struct ext4_sb_info *sbi = EXT4_SB(sb);
775 struct ext4_super_block *es = sbi->s_es;
d9c9bef1
MS		 776
777 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
aa22df2c 778 def_errors = le16_to_cpu(es->s_errors);
d9c9bef1
MS		 779
780 if (sbi->s_sb_block != 1)
781 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
782 if (test_opt(sb, MINIX_DF))
783 seq_puts(seq, ",minixdf");
aa22df2c 784 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
d9c9bef1
MS		 785 seq_puts(seq, ",grpid");
786 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
787 seq_puts(seq, ",nogrpid");
788 if (sbi->s_resuid != EXT4_DEF_RESUID ||
789 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
790 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
791 }
792 if (sbi->s_resgid != EXT4_DEF_RESGID ||
793 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
794 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
795 }
bb4f397a 796 if (test_opt(sb, ERRORS_RO)) {
d9c9bef1 797 if (def_errors == EXT4_ERRORS_PANIC ||
bb4f397a
AK		 798 def_errors == EXT4_ERRORS_CONTINUE) {
799 seq_puts(seq, ",errors=remount-ro");
d9c9bef1
MS		 800 }
801 }
aa22df2c 802 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
bb4f397a 803 seq_puts(seq, ",errors=continue");
aa22df2c 804 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
d9c9bef1 805 seq_puts(seq, ",errors=panic");
aa22df2c 806 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
d9c9bef1 807 seq_puts(seq, ",nouid32");
aa22df2c 808 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
d9c9bef1
MS		 809 seq_puts(seq, ",debug");
810 if (test_opt(sb, OLDALLOC))
811 seq_puts(seq, ",oldalloc");
03010a33 812#ifdef CONFIG_EXT4_FS_XATTR
aa22df2c
AK		 813 if (test_opt(sb, XATTR_USER) &&
814 !(def_mount_opts & EXT4_DEFM_XATTR_USER))
d9c9bef1
MS		 815 seq_puts(seq, ",user_xattr");
816 if (!test_opt(sb, XATTR_USER) &&
817 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
818 seq_puts(seq, ",nouser_xattr");
819 }
820#endif
03010a33 821#ifdef CONFIG_EXT4_FS_POSIX_ACL
aa22df2c 822 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
d9c9bef1
MS		 823 seq_puts(seq, ",acl");
824 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
825 seq_puts(seq, ",noacl");
826#endif
30773840 827 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
d9c9bef1
MS		 828 seq_printf(seq, ",commit=%u",
829 (unsigned) (sbi->s_commit_interval / HZ));
830 }
30773840
TT		 831 if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
832 seq_printf(seq, ",min_batch_time=%u",
833 (unsigned) sbi->s_min_batch_time);
834 }
835 if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
836 seq_printf(seq, ",max_batch_time=%u",
837 (unsigned) sbi->s_min_batch_time);
838 }
839
571640ca
ES		 840 /*
841 * We're changing the default of barrier mount option, so
842 * let's always display its mount state so it's clear what its
843 * status is.
844 */
845 seq_puts(seq, ",barrier=");
846 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
cd0b6a39
TT		 847 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
848 seq_puts(seq, ",journal_async_commit");
d9c9bef1
MS		 849 if (test_opt(sb, NOBH))
850 seq_puts(seq, ",nobh");
25ec56b5
JNC		 851 if (test_opt(sb, I_VERSION))
852 seq_puts(seq, ",i_version");
dd919b98
AK		 853 if (!test_opt(sb, DELALLOC))
854 seq_puts(seq, ",nodelalloc");
855
ac27a0ec 856
cb45bbe4
MS		 857 if (sbi->s_stripe)
858 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
aa22df2c
AK		 859 /*
860 * journal mode get enabled in different ways
861 * So just print the value even if we didn't specify it
862 */
617ba13b 863 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
ac27a0ec 864 seq_puts(seq, ",data=journal");
617ba13b 865 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
ac27a0ec 866 seq_puts(seq, ",data=ordered");
617ba13b 867 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
ac27a0ec
DK		 868 seq_puts(seq, ",data=writeback");
869
240799cd
TT		 870 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
871 seq_printf(seq, ",inode_readahead_blks=%u",
872 sbi->s_inode_readahead_blks);
873
5bf5683a
HK		 874 if (test_opt(sb, DATA_ERR_ABORT))
875 seq_puts(seq, ",data_err=abort");
876
afd4672d 877 if (test_opt(sb, NO_AUTO_DA_ALLOC))
06705bff 878 seq_puts(seq, ",noauto_da_alloc");
afd4672d 879
617ba13b 880 ext4_show_quota_options(seq, sb);
0b8e58a1 881
ac27a0ec
DK		 882 return 0;
883}
884
1b961ac0 885static struct inode *ext4_nfs_get_inode(struct super_block *sb,
0b8e58a1 886 u64 ino, u32 generation)
ac27a0ec 887{
ac27a0ec 888 struct inode *inode;
ac27a0ec 889
617ba13b 890 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
ac27a0ec 891 return ERR_PTR(-ESTALE);
617ba13b 892 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
ac27a0ec
DK		 893 return ERR_PTR(-ESTALE);
894
895 /* iget isn't really right if the inode is currently unallocated!!
896 *
617ba13b 897 * ext4_read_inode will return a bad_inode if the inode had been
ac27a0ec
DK		 898 * deleted, so we should be safe.
899 *
900 * Currently we don't know the generation for parent directory, so
901 * a generation of 0 means "accept any"
902 */
1d1fe1ee
DH		 903 inode = ext4_iget(sb, ino);
904 if (IS_ERR(inode))
905 return ERR_CAST(inode);
906 if (generation && inode->i_generation != generation) {
ac27a0ec
DK		 907 iput(inode);
908 return ERR_PTR(-ESTALE);
909 }
1b961ac0
CH		 910
911 return inode;
912}
913
914static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
0b8e58a1 915 int fh_len, int fh_type)
1b961ac0
CH		 916{
917 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
918 ext4_nfs_get_inode);
919}
920
921static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
0b8e58a1 922 int fh_len, int fh_type)
1b961ac0
CH		 923{
924 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
925 ext4_nfs_get_inode);
ac27a0ec
DK		 926}
927
c39a7f84
TO		 928/*
929 * Try to release metadata pages (indirect blocks, directories) which are
930 * mapped via the block device. Since these pages could have journal heads
931 * which would prevent try_to_free_buffers() from freeing them, we must use
932 * jbd2 layer's try_to_free_buffers() function to release them.
933 */
0b8e58a1
AD		 934static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
935 gfp_t wait)
c39a7f84
TO		 936{
937 journal_t *journal = EXT4_SB(sb)->s_journal;
938
939 WARN_ON(PageChecked(page));
940 if (!page_has_buffers(page))
941 return 0;
942 if (journal)
943 return jbd2_journal_try_to_free_buffers(journal, page,
944 wait & ~__GFP_WAIT);
945 return try_to_free_buffers(page);
946}
947
ac27a0ec 948#ifdef CONFIG_QUOTA
af5bc92d 949#define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
2b2d6d01 950#define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
ac27a0ec 951
617ba13b
MC		 952static int ext4_write_dquot(struct dquot *dquot);
953static int ext4_acquire_dquot(struct dquot *dquot);
954static int ext4_release_dquot(struct dquot *dquot);
955static int ext4_mark_dquot_dirty(struct dquot *dquot);
956static int ext4_write_info(struct super_block *sb, int type);
6f28e087
JK		 957static int ext4_quota_on(struct super_block *sb, int type, int format_id,
958 char *path, int remount);
617ba13b
MC		 959static int ext4_quota_on_mount(struct super_block *sb, int type);
960static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 961 size_t len, loff_t off);
617ba13b 962static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 963 const char *data, size_t len, loff_t off);
964
617ba13b 965static struct dquot_operations ext4_quota_operations = {
edf72453
JK		 966 .initialize = dquot_initialize,
967 .drop = dquot_drop,
ac27a0ec 968 .alloc_space = dquot_alloc_space,
60e58e0f
MC		 969 .reserve_space = dquot_reserve_space,
970 .claim_space = dquot_claim_space,
971 .release_rsv = dquot_release_reserved_space,
972 .get_reserved_space = ext4_get_reserved_space,
ac27a0ec
DK		 973 .alloc_inode = dquot_alloc_inode,
974 .free_space = dquot_free_space,
975 .free_inode = dquot_free_inode,
976 .transfer = dquot_transfer,
617ba13b
MC		 977 .write_dquot = ext4_write_dquot,
978 .acquire_dquot = ext4_acquire_dquot,
979 .release_dquot = ext4_release_dquot,
980 .mark_dirty = ext4_mark_dquot_dirty,
a5b5ee32
JK		 981 .write_info = ext4_write_info,
982 .alloc_dquot = dquot_alloc,
983 .destroy_dquot = dquot_destroy,
ac27a0ec
DK		 984};
985
617ba13b
MC		 986static struct quotactl_ops ext4_qctl_operations = {
987 .quota_on = ext4_quota_on,
ac27a0ec
DK		 988 .quota_off = vfs_quota_off,
989 .quota_sync = vfs_quota_sync,
990 .get_info = vfs_get_dqinfo,
991 .set_info = vfs_set_dqinfo,
992 .get_dqblk = vfs_get_dqblk,
993 .set_dqblk = vfs_set_dqblk
994};
995#endif
996
ee9b6d61 997static const struct super_operations ext4_sops = {
617ba13b
MC		 998 .alloc_inode = ext4_alloc_inode,
999 .destroy_inode = ext4_destroy_inode,
617ba13b
MC		 1000 .write_inode = ext4_write_inode,
1001 .dirty_inode = ext4_dirty_inode,
1002 .delete_inode = ext4_delete_inode,
1003 .put_super = ext4_put_super,
617ba13b 1004 .sync_fs = ext4_sync_fs,
c4be0c1d
TS		 1005 .freeze_fs = ext4_freeze,
1006 .unfreeze_fs = ext4_unfreeze,
617ba13b
MC		 1007 .statfs = ext4_statfs,
1008 .remount_fs = ext4_remount,
1009 .clear_inode = ext4_clear_inode,
1010 .show_options = ext4_show_options,
ac27a0ec 1011#ifdef CONFIG_QUOTA
617ba13b
MC		 1012 .quota_read = ext4_quota_read,
1013 .quota_write = ext4_quota_write,
ac27a0ec 1014#endif
c39a7f84 1015 .bdev_try_to_free_page = bdev_try_to_free_page,
ac27a0ec
DK		 1016};
1017
9ca92389
TT		 1018static const struct super_operations ext4_nojournal_sops = {
1019 .alloc_inode = ext4_alloc_inode,
1020 .destroy_inode = ext4_destroy_inode,
1021 .write_inode = ext4_write_inode,
1022 .dirty_inode = ext4_dirty_inode,
1023 .delete_inode = ext4_delete_inode,
1024 .write_super = ext4_write_super,
1025 .put_super = ext4_put_super,
1026 .statfs = ext4_statfs,
1027 .remount_fs = ext4_remount,
1028 .clear_inode = ext4_clear_inode,
1029 .show_options = ext4_show_options,
1030#ifdef CONFIG_QUOTA
1031 .quota_read = ext4_quota_read,
1032 .quota_write = ext4_quota_write,
1033#endif
1034 .bdev_try_to_free_page = bdev_try_to_free_page,
1035};
1036
39655164 1037static const struct export_operations ext4_export_ops = {
1b961ac0
CH		 1038 .fh_to_dentry = ext4_fh_to_dentry,
1039 .fh_to_parent = ext4_fh_to_parent,
617ba13b 1040 .get_parent = ext4_get_parent,
ac27a0ec
DK		 1041};
1042
1043enum {
1044 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1045 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
01436ef2 1046 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
ac27a0ec 1047 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
06705bff 1048 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh,
30773840 1049 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
c3191067 1050 Opt_journal_update, Opt_journal_dev,
818d276c 1051 Opt_journal_checksum, Opt_journal_async_commit,
ac27a0ec 1052 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
f4033903 1053 Opt_data_err_abort, Opt_data_err_ignore, Opt_mb_history_length,
ac27a0ec
DK		 1054 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1055 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
06705bff
TT		 1056 Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err, Opt_resize,
1057 Opt_usrquota, Opt_grpquota, Opt_i_version,
01436ef2 1058 Opt_stripe, Opt_delalloc, Opt_nodelalloc,
6fd058f7 1059 Opt_block_validity, Opt_noblock_validity,
b3881f74 1060 Opt_inode_readahead_blks, Opt_journal_ioprio
ac27a0ec
DK		 1061};
1062
a447c093 1063static const match_table_t tokens = {
ac27a0ec
DK		 1064 {Opt_bsd_df, "bsddf"},
1065 {Opt_minix_df, "minixdf"},
1066 {Opt_grpid, "grpid"},
1067 {Opt_grpid, "bsdgroups"},
1068 {Opt_nogrpid, "nogrpid"},
1069 {Opt_nogrpid, "sysvgroups"},
1070 {Opt_resgid, "resgid=%u"},
1071 {Opt_resuid, "resuid=%u"},
1072 {Opt_sb, "sb=%u"},
1073 {Opt_err_cont, "errors=continue"},
1074 {Opt_err_panic, "errors=panic"},
1075 {Opt_err_ro, "errors=remount-ro"},
1076 {Opt_nouid32, "nouid32"},
ac27a0ec
DK		 1077 {Opt_debug, "debug"},
1078 {Opt_oldalloc, "oldalloc"},
1079 {Opt_orlov, "orlov"},
1080 {Opt_user_xattr, "user_xattr"},
1081 {Opt_nouser_xattr, "nouser_xattr"},
1082 {Opt_acl, "acl"},
1083 {Opt_noacl, "noacl"},
ac27a0ec
DK		 1084 {Opt_noload, "noload"},
1085 {Opt_nobh, "nobh"},
1086 {Opt_bh, "bh"},
1087 {Opt_commit, "commit=%u"},
30773840
TT		 1088 {Opt_min_batch_time, "min_batch_time=%u"},
1089 {Opt_max_batch_time, "max_batch_time=%u"},
ac27a0ec 1090 {Opt_journal_update, "journal=update"},
ac27a0ec 1091 {Opt_journal_dev, "journal_dev=%u"},
818d276c
GS		 1092 {Opt_journal_checksum, "journal_checksum"},
1093 {Opt_journal_async_commit, "journal_async_commit"},
ac27a0ec
DK		 1094 {Opt_abort, "abort"},
1095 {Opt_data_journal, "data=journal"},
1096 {Opt_data_ordered, "data=ordered"},
1097 {Opt_data_writeback, "data=writeback"},
5bf5683a
HK		 1098 {Opt_data_err_abort, "data_err=abort"},
1099 {Opt_data_err_ignore, "data_err=ignore"},
f4033903 1100 {Opt_mb_history_length, "mb_history_length=%u"},
ac27a0ec
DK		 1101 {Opt_offusrjquota, "usrjquota="},
1102 {Opt_usrjquota, "usrjquota=%s"},
1103 {Opt_offgrpjquota, "grpjquota="},
1104 {Opt_grpjquota, "grpjquota=%s"},
1105 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1106 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1107 {Opt_grpquota, "grpquota"},
1108 {Opt_noquota, "noquota"},
1109 {Opt_quota, "quota"},
1110 {Opt_usrquota, "usrquota"},
1111 {Opt_barrier, "barrier=%u"},
06705bff
TT		 1112 {Opt_barrier, "barrier"},
1113 {Opt_nobarrier, "nobarrier"},
25ec56b5 1114 {Opt_i_version, "i_version"},
c9de560d 1115 {Opt_stripe, "stripe=%u"},
ac27a0ec 1116 {Opt_resize, "resize"},
64769240 1117 {Opt_delalloc, "delalloc"},
dd919b98 1118 {Opt_nodelalloc, "nodelalloc"},
6fd058f7
TT		 1119 {Opt_block_validity, "block_validity"},
1120 {Opt_noblock_validity, "noblock_validity"},
240799cd 1121 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
b3881f74 1122 {Opt_journal_ioprio, "journal_ioprio=%u"},
afd4672d 1123 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
06705bff
TT		 1124 {Opt_auto_da_alloc, "auto_da_alloc"},
1125 {Opt_noauto_da_alloc, "noauto_da_alloc"},
f3f12faa 1126 {Opt_err, NULL},
ac27a0ec
DK		 1127};
1128
617ba13b 1129static ext4_fsblk_t get_sb_block(void **data)
ac27a0ec 1130{
617ba13b 1131 ext4_fsblk_t sb_block;
ac27a0ec
DK		 1132 char *options = (char *) *data;
1133
1134 if (!options || strncmp(options, "sb=", 3) != 0)
1135 return 1; /* Default location */
0b8e58a1 1136
ac27a0ec 1137 options += 3;
0b8e58a1 1138 /* TODO: use simple_strtoll with >32bit ext4 */
ac27a0ec
DK		 1139 sb_block = simple_strtoul(options, &options, 0);
1140 if (*options && *options != ',') {
4776004f 1141 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
ac27a0ec
DK		 1142 (char *) *data);
1143 return 1;
1144 }
1145 if (*options == ',')
1146 options++;
1147 *data = (void *) options;
0b8e58a1 1148
ac27a0ec
DK		 1149 return sb_block;
1150}
1151
b3881f74
TT		 1152#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1153
2b2d6d01 1154static int parse_options(char *options, struct super_block *sb,
c3191067 1155 unsigned long *journal_devnum,
b3881f74 1156 unsigned int *journal_ioprio,
2b2d6d01 1157 ext4_fsblk_t *n_blocks_count, int is_remount)
ac27a0ec 1158{
617ba13b 1159 struct ext4_sb_info *sbi = EXT4_SB(sb);
2b2d6d01 1160 char *p;
ac27a0ec
DK		 1161 substring_t args[MAX_OPT_ARGS];
1162 int data_opt = 0;
1163 int option;
1164#ifdef CONFIG_QUOTA
dfc5d03f 1165 int qtype, qfmt;
ac27a0ec
DK		 1166 char *qname;
1167#endif
1168
1169 if (!options)
1170 return 1;
1171
2b2d6d01 1172 while ((p = strsep(&options, ",")) != NULL) {
ac27a0ec
DK		 1173 int token;
1174 if (!*p)
1175 continue;
1176
1177 token = match_token(p, tokens, args);
1178 switch (token) {
1179 case Opt_bsd_df:
2b2d6d01 1180 clear_opt(sbi->s_mount_opt, MINIX_DF);
ac27a0ec
DK		 1181 break;
1182 case Opt_minix_df:
2b2d6d01 1183 set_opt(sbi->s_mount_opt, MINIX_DF);
ac27a0ec
DK		 1184 break;
1185 case Opt_grpid:
2b2d6d01 1186 set_opt(sbi->s_mount_opt, GRPID);
ac27a0ec
DK		 1187 break;
1188 case Opt_nogrpid:
2b2d6d01 1189 clear_opt(sbi->s_mount_opt, GRPID);
ac27a0ec
DK		 1190 break;
1191 case Opt_resuid:
1192 if (match_int(&args[0], &option))
1193 return 0;
1194 sbi->s_resuid = option;
1195 break;
1196 case Opt_resgid:
1197 if (match_int(&args[0], &option))
1198 return 0;
1199 sbi->s_resgid = option;
1200 break;
1201 case Opt_sb:
1202 /* handled by get_sb_block() instead of here */
1203 /* *sb_block = match_int(&args[0]); */
1204 break;
1205 case Opt_err_panic:
2b2d6d01
TT		 1206 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1207 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1208 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
ac27a0ec
DK		 1209 break;
1210 case Opt_err_ro:
2b2d6d01
TT		 1211 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1212 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1213 set_opt(sbi->s_mount_opt, ERRORS_RO);
ac27a0ec
DK		 1214 break;
1215 case Opt_err_cont:
2b2d6d01
TT		 1216 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1217 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1218 set_opt(sbi->s_mount_opt, ERRORS_CONT);
ac27a0ec
DK		 1219 break;
1220 case Opt_nouid32:
2b2d6d01 1221 set_opt(sbi->s_mount_opt, NO_UID32);
ac27a0ec 1222 break;
ac27a0ec 1223 case Opt_debug:
2b2d6d01 1224 set_opt(sbi->s_mount_opt, DEBUG);
ac27a0ec
DK		 1225 break;
1226 case Opt_oldalloc:
2b2d6d01 1227 set_opt(sbi->s_mount_opt, OLDALLOC);
ac27a0ec
DK		 1228 break;
1229 case Opt_orlov:
2b2d6d01 1230 clear_opt(sbi->s_mount_opt, OLDALLOC);
ac27a0ec 1231 break;
03010a33 1232#ifdef CONFIG_EXT4_FS_XATTR
ac27a0ec 1233 case Opt_user_xattr:
2b2d6d01 1234 set_opt(sbi->s_mount_opt, XATTR_USER);
ac27a0ec
DK		 1235 break;
1236 case Opt_nouser_xattr:
2b2d6d01 1237 clear_opt(sbi->s_mount_opt, XATTR_USER);
ac27a0ec
DK		 1238 break;
1239#else
1240 case Opt_user_xattr:
1241 case Opt_nouser_xattr:
b31e1552 1242 ext4_msg(sb, KERN_ERR, "(no)user_xattr options not supported");
ac27a0ec
DK		 1243 break;
1244#endif
03010a33 1245#ifdef CONFIG_EXT4_FS_POSIX_ACL
ac27a0ec
DK		 1246 case Opt_acl:
1247 set_opt(sbi->s_mount_opt, POSIX_ACL);
1248 break;
1249 case Opt_noacl:
1250 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1251 break;
1252#else
1253 case Opt_acl:
1254 case Opt_noacl:
b31e1552 1255 ext4_msg(sb, KERN_ERR, "(no)acl options not supported");
ac27a0ec
DK		 1256 break;
1257#endif
ac27a0ec
DK		 1258 case Opt_journal_update:
1259 /* @@@ FIXME */
1260 /* Eventually we will want to be able to create
1261 a journal file here. For now, only allow the
1262 user to specify an existing inode to be the
1263 journal file. */
1264 if (is_remount) {
b31e1552
ES		 1265 ext4_msg(sb, KERN_ERR,
1266 "Cannot specify journal on remount");
ac27a0ec
DK		 1267 return 0;
1268 }
2b2d6d01 1269 set_opt(sbi->s_mount_opt, UPDATE_JOURNAL);
ac27a0ec 1270 break;
ac27a0ec
DK		 1271 case Opt_journal_dev:
1272 if (is_remount) {
b31e1552
ES		 1273 ext4_msg(sb, KERN_ERR,
1274 "Cannot specify journal on remount");
ac27a0ec
DK		 1275 return 0;
1276 }
1277 if (match_int(&args[0], &option))
1278 return 0;
1279 *journal_devnum = option;
1280 break;
818d276c
GS		 1281 case Opt_journal_checksum:
1282 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1283 break;
1284 case Opt_journal_async_commit:
1285 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
1286 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1287 break;
ac27a0ec 1288 case Opt_noload:
2b2d6d01 1289 set_opt(sbi->s_mount_opt, NOLOAD);
ac27a0ec
DK		 1290 break;
1291 case Opt_commit:
1292 if (match_int(&args[0], &option))
1293 return 0;
1294 if (option < 0)
1295 return 0;
1296 if (option == 0)
cd02ff0b 1297 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
ac27a0ec
DK		 1298 sbi->s_commit_interval = HZ * option;
1299 break;
30773840
TT		 1300 case Opt_max_batch_time:
1301 if (match_int(&args[0], &option))
1302 return 0;
1303 if (option < 0)
1304 return 0;
1305 if (option == 0)
1306 option = EXT4_DEF_MAX_BATCH_TIME;
1307 sbi->s_max_batch_time = option;
1308 break;
1309 case Opt_min_batch_time:
1310 if (match_int(&args[0], &option))
1311 return 0;
1312 if (option < 0)
1313 return 0;
1314 sbi->s_min_batch_time = option;
1315 break;
ac27a0ec 1316 case Opt_data_journal:
617ba13b 1317 data_opt = EXT4_MOUNT_JOURNAL_DATA;
ac27a0ec
DK		 1318 goto datacheck;
1319 case Opt_data_ordered:
617ba13b 1320 data_opt = EXT4_MOUNT_ORDERED_DATA;
ac27a0ec
DK		 1321 goto datacheck;
1322 case Opt_data_writeback:
617ba13b 1323 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
ac27a0ec
DK		 1324 datacheck:
1325 if (is_remount) {
617ba13b 1326 if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
ac27a0ec 1327 != data_opt) {
b31e1552
ES		 1328 ext4_msg(sb, KERN_ERR,
1329 "Cannot change data mode on remount");
ac27a0ec
DK		 1330 return 0;
1331 }
1332 } else {
617ba13b 1333 sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
ac27a0ec
DK		 1334 sbi->s_mount_opt |= data_opt;
1335 }
1336 break;
5bf5683a
HK		 1337 case Opt_data_err_abort:
1338 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1339 break;
1340 case Opt_data_err_ignore:
1341 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1342 break;
f4033903
CW		 1343 case Opt_mb_history_length:
1344 if (match_int(&args[0], &option))
1345 return 0;
1346 if (option < 0)
1347 return 0;
1348 sbi->s_mb_history_max = option;
1349 break;
ac27a0ec
DK		 1350#ifdef CONFIG_QUOTA
1351 case Opt_usrjquota:
1352 qtype = USRQUOTA;
1353 goto set_qf_name;
1354 case Opt_grpjquota:
1355 qtype = GRPQUOTA;
1356set_qf_name:
17bd13b3 1357 if (sb_any_quota_loaded(sb) &&
dfc5d03f 1358 !sbi->s_qf_names[qtype]) {
b31e1552
ES		 1359 ext4_msg(sb, KERN_ERR,
1360 "Cannot change journaled "
1361 "quota options when quota turned on");
ac27a0ec
DK		 1362 return 0;
1363 }
1364 qname = match_strdup(&args[0]);
1365 if (!qname) {
b31e1552
ES		 1366 ext4_msg(sb, KERN_ERR,
1367 "Not enough memory for "
1368 "storing quotafile name");
ac27a0ec
DK		 1369 return 0;
1370 }
1371 if (sbi->s_qf_names[qtype] &&
1372 strcmp(sbi->s_qf_names[qtype], qname)) {
b31e1552
ES		 1373 ext4_msg(sb, KERN_ERR,
1374 "%s quota file already "
1375 "specified", QTYPE2NAME(qtype));
ac27a0ec
DK		 1376 kfree(qname);
1377 return 0;
1378 }
1379 sbi->s_qf_names[qtype] = qname;
1380 if (strchr(sbi->s_qf_names[qtype], '/')) {
b31e1552
ES		 1381 ext4_msg(sb, KERN_ERR,
1382 "quotafile must be on "
1383 "filesystem root");
ac27a0ec
DK		 1384 kfree(sbi->s_qf_names[qtype]);
1385 sbi->s_qf_names[qtype] = NULL;
1386 return 0;
1387 }
1388 set_opt(sbi->s_mount_opt, QUOTA);
1389 break;
1390 case Opt_offusrjquota:
1391 qtype = USRQUOTA;
1392 goto clear_qf_name;
1393 case Opt_offgrpjquota:
1394 qtype = GRPQUOTA;
1395clear_qf_name:
17bd13b3 1396 if (sb_any_quota_loaded(sb) &&
dfc5d03f 1397 sbi->s_qf_names[qtype]) {
b31e1552 1398 ext4_msg(sb, KERN_ERR, "Cannot change "
2c8be6b2 1399 "journaled quota options when "
b31e1552 1400 "quota turned on");
ac27a0ec
DK		 1401 return 0;
1402 }
1403 /*
1404 * The space will be released later when all options
1405 * are confirmed to be correct
1406 */
1407 sbi->s_qf_names[qtype] = NULL;
1408 break;
1409 case Opt_jqfmt_vfsold:
dfc5d03f
JK		 1410 qfmt = QFMT_VFS_OLD;
1411 goto set_qf_format;
ac27a0ec 1412 case Opt_jqfmt_vfsv0:
dfc5d03f
JK		 1413 qfmt = QFMT_VFS_V0;
1414set_qf_format:
17bd13b3 1415 if (sb_any_quota_loaded(sb) &&
dfc5d03f 1416 sbi->s_jquota_fmt != qfmt) {
b31e1552 1417 ext4_msg(sb, KERN_ERR, "Cannot change "
dfc5d03f 1418 "journaled quota options when "
b31e1552 1419 "quota turned on");
dfc5d03f
JK		 1420 return 0;
1421 }
1422 sbi->s_jquota_fmt = qfmt;
ac27a0ec
DK		 1423 break;
1424 case Opt_quota:
1425 case Opt_usrquota:
1426 set_opt(sbi->s_mount_opt, QUOTA);
1427 set_opt(sbi->s_mount_opt, USRQUOTA);
1428 break;
1429 case Opt_grpquota:
1430 set_opt(sbi->s_mount_opt, QUOTA);
1431 set_opt(sbi->s_mount_opt, GRPQUOTA);
1432 break;
1433 case Opt_noquota:
17bd13b3 1434 if (sb_any_quota_loaded(sb)) {
b31e1552
ES		 1435 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1436 "options when quota turned on");
ac27a0ec
DK		 1437 return 0;
1438 }
1439 clear_opt(sbi->s_mount_opt, QUOTA);
1440 clear_opt(sbi->s_mount_opt, USRQUOTA);
1441 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1442 break;
1443#else
1444 case Opt_quota:
1445 case Opt_usrquota:
1446 case Opt_grpquota:
b31e1552
ES		 1447 ext4_msg(sb, KERN_ERR,
1448 "quota options not supported");
cd59e7b9 1449 break;
ac27a0ec
DK		 1450 case Opt_usrjquota:
1451 case Opt_grpjquota:
1452 case Opt_offusrjquota:
1453 case Opt_offgrpjquota:
1454 case Opt_jqfmt_vfsold:
1455 case Opt_jqfmt_vfsv0:
b31e1552
ES		 1456 ext4_msg(sb, KERN_ERR,
1457 "journaled quota options not supported");
ac27a0ec
DK		 1458 break;
1459 case Opt_noquota:
1460 break;
1461#endif
1462 case Opt_abort:
4ab2f15b 1463 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
ac27a0ec 1464 break;
06705bff
TT		 1465 case Opt_nobarrier:
1466 clear_opt(sbi->s_mount_opt, BARRIER);
1467 break;
ac27a0ec 1468 case Opt_barrier:
06705bff
TT		 1469 if (match_int(&args[0], &option)) {
1470 set_opt(sbi->s_mount_opt, BARRIER);
1471 break;
1472 }
ac27a0ec
DK		 1473 if (option)
1474 set_opt(sbi->s_mount_opt, BARRIER);
1475 else
1476 clear_opt(sbi->s_mount_opt, BARRIER);
1477 break;
1478 case Opt_ignore:
1479 break;
1480 case Opt_resize:
1481 if (!is_remount) {
b31e1552
ES		 1482 ext4_msg(sb, KERN_ERR,
1483 "resize option only available "
1484 "for remount");
ac27a0ec
DK		 1485 return 0;
1486 }
1487 if (match_int(&args[0], &option) != 0)
1488 return 0;
1489 *n_blocks_count = option;
1490 break;
1491 case Opt_nobh:
1492 set_opt(sbi->s_mount_opt, NOBH);
1493 break;
1494 case Opt_bh:
1495 clear_opt(sbi->s_mount_opt, NOBH);
1496 break;
25ec56b5
JNC		 1497 case Opt_i_version:
1498 set_opt(sbi->s_mount_opt, I_VERSION);
1499 sb->s_flags |= MS_I_VERSION;
1500 break;
dd919b98
AK		 1501 case Opt_nodelalloc:
1502 clear_opt(sbi->s_mount_opt, DELALLOC);
1503 break;
c9de560d
AT		 1504 case Opt_stripe:
1505 if (match_int(&args[0], &option))
1506 return 0;
1507 if (option < 0)
1508 return 0;
1509 sbi->s_stripe = option;
1510 break;
64769240
AT		 1511 case Opt_delalloc:
1512 set_opt(sbi->s_mount_opt, DELALLOC);
1513 break;
6fd058f7
TT		 1514 case Opt_block_validity:
1515 set_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1516 break;
1517 case Opt_noblock_validity:
1518 clear_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1519 break;
240799cd
TT		 1520 case Opt_inode_readahead_blks:
1521 if (match_int(&args[0], &option))
1522 return 0;
1523 if (option < 0 || option > (1 << 30))
1524 return 0;
f7c43950 1525 if (!is_power_of_2(option)) {
b31e1552
ES		 1526 ext4_msg(sb, KERN_ERR,
1527 "EXT4-fs: inode_readahead_blks"
1528 " must be a power of 2");
3197ebdb
TT		 1529 return 0;
1530 }
240799cd
TT		 1531 sbi->s_inode_readahead_blks = option;
1532 break;
b3881f74
TT		 1533 case Opt_journal_ioprio:
1534 if (match_int(&args[0], &option))
1535 return 0;
1536 if (option < 0 || option > 7)
1537 break;
1538 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1539 option);
1540 break;
06705bff
TT		 1541 case Opt_noauto_da_alloc:
1542 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1543 break;
afd4672d 1544 case Opt_auto_da_alloc:
06705bff
TT		 1545 if (match_int(&args[0], &option)) {
1546 clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1547 break;
1548 }
afd4672d
TT		 1549 if (option)
1550 clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1551 else
1552 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1553 break;
ac27a0ec 1554 default:
b31e1552
ES		 1555 ext4_msg(sb, KERN_ERR,
1556 "Unrecognized mount option \"%s\" "
1557 "or missing value", p);
ac27a0ec
DK		 1558 return 0;
1559 }
1560 }
1561#ifdef CONFIG_QUOTA
1562 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
617ba13b 1563 if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
ac27a0ec
DK		 1564 sbi->s_qf_names[USRQUOTA])
1565 clear_opt(sbi->s_mount_opt, USRQUOTA);
1566
617ba13b 1567 if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
ac27a0ec
DK		 1568 sbi->s_qf_names[GRPQUOTA])
1569 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1570
1571 if ((sbi->s_qf_names[USRQUOTA] &&
617ba13b 1572 (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
ac27a0ec 1573 (sbi->s_qf_names[GRPQUOTA] &&
617ba13b 1574 (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
b31e1552
ES		 1575 ext4_msg(sb, KERN_ERR, "old and new quota "
1576 "format mixing");
ac27a0ec
DK		 1577 return 0;
1578 }
1579
1580 if (!sbi->s_jquota_fmt) {
b31e1552
ES		 1581 ext4_msg(sb, KERN_ERR, "journaled quota format "
1582 "not specified");
ac27a0ec
DK		 1583 return 0;
1584 }
1585 } else {
1586 if (sbi->s_jquota_fmt) {
b31e1552 1587 ext4_msg(sb, KERN_ERR, "journaled quota format "
2c8be6b2 1588 "specified with no journaling "
b31e1552 1589 "enabled");
ac27a0ec
DK		 1590 return 0;
1591 }
1592 }
1593#endif
1594 return 1;
1595}
1596
617ba13b 1597static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 1598 int read_only)
1599{
617ba13b 1600 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 1601 int res = 0;
1602
617ba13b 1603 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
b31e1552
ES		 1604 ext4_msg(sb, KERN_ERR, "revision level too high, "
1605 "forcing read-only mode");
ac27a0ec
DK		 1606 res = MS_RDONLY;
1607 }
1608 if (read_only)
1609 return res;
617ba13b 1610 if (!(sbi->s_mount_state & EXT4_VALID_FS))
b31e1552
ES		 1611 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1612 "running e2fsck is recommended");
617ba13b 1613 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
b31e1552
ES		 1614 ext4_msg(sb, KERN_WARNING,
1615 "warning: mounting fs with errors, "
1616 "running e2fsck is recommended");
ac27a0ec
DK		 1617 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1618 le16_to_cpu(es->s_mnt_count) >=
1619 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
b31e1552
ES		 1620 ext4_msg(sb, KERN_WARNING,
1621 "warning: maximal mount count reached, "
1622 "running e2fsck is recommended");
ac27a0ec
DK		 1623 else if (le32_to_cpu(es->s_checkinterval) &&
1624 (le32_to_cpu(es->s_lastcheck) +
1625 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
b31e1552
ES		 1626 ext4_msg(sb, KERN_WARNING,
1627 "warning: checktime reached, "
1628 "running e2fsck is recommended");
0b8e58a1 1629 if (!sbi->s_journal)
0390131b 1630 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec 1631 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 1632 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 1633 le16_add_cpu(&es->s_mnt_count, 1);
ac27a0ec 1634 es->s_mtime = cpu_to_le32(get_seconds());
617ba13b 1635 ext4_update_dynamic_rev(sb);
0390131b
FM		 1636 if (sbi->s_journal)
1637 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 1638
e2d67052 1639 ext4_commit_super(sb, 1);
ac27a0ec 1640 if (test_opt(sb, DEBUG))
a9df9a49 1641 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
7f4520cc 1642 "bpg=%lu, ipg=%lu, mo=%04x]\n",
ac27a0ec
DK		 1643 sb->s_blocksize,
1644 sbi->s_groups_count,
617ba13b
MC		 1645 EXT4_BLOCKS_PER_GROUP(sb),
1646 EXT4_INODES_PER_GROUP(sb),
ac27a0ec
DK		 1647 sbi->s_mount_opt);
1648
0390131b 1649 if (EXT4_SB(sb)->s_journal) {
b31e1552
ES		 1650 ext4_msg(sb, KERN_INFO, "%s journal on %s",
1651 EXT4_SB(sb)->s_journal->j_inode ? "internal" :
0390131b
FM		 1652 "external", EXT4_SB(sb)->s_journal->j_devname);
1653 } else {
b31e1552 1654 ext4_msg(sb, KERN_INFO, "no journal");
0390131b 1655 }
ac27a0ec
DK		 1656 return res;
1657}
1658
772cb7c8
JS		 1659static int ext4_fill_flex_info(struct super_block *sb)
1660{
1661 struct ext4_sb_info *sbi = EXT4_SB(sb);
1662 struct ext4_group_desc *gdp = NULL;
772cb7c8
JS		 1663 ext4_group_t flex_group_count;
1664 ext4_group_t flex_group;
1665 int groups_per_flex = 0;
c5ca7c76 1666 size_t size;
772cb7c8
JS		 1667 int i;
1668
1669 if (!sbi->s_es->s_log_groups_per_flex) {
1670 sbi->s_log_groups_per_flex = 0;
1671 return 1;
1672 }
1673
1674 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1675 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1676
c62a11fd
FB		 1677 /* We allocate both existing and potentially added groups */
1678 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
d94e99a6
AK		 1679 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1680 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
c5ca7c76
TT		 1681 size = flex_group_count * sizeof(struct flex_groups);
1682 sbi->s_flex_groups = kzalloc(size, GFP_KERNEL);
1683 if (sbi->s_flex_groups == NULL) {
1684 sbi->s_flex_groups = vmalloc(size);
1685 if (sbi->s_flex_groups)
1686 memset(sbi->s_flex_groups, 0, size);
1687 }
772cb7c8 1688 if (sbi->s_flex_groups == NULL) {
b31e1552
ES		 1689 ext4_msg(sb, KERN_ERR, "not enough memory for "
1690 "%u flex groups", flex_group_count);
772cb7c8
JS		 1691 goto failed;
1692 }
772cb7c8 1693
772cb7c8 1694 for (i = 0; i < sbi->s_groups_count; i++) {
88b6edd1 1695 gdp = ext4_get_group_desc(sb, i, NULL);
772cb7c8
JS		 1696
1697 flex_group = ext4_flex_group(sbi, i);
9f24e420
TT		 1698 atomic_set(&sbi->s_flex_groups[flex_group].free_inodes,
1699 ext4_free_inodes_count(sb, gdp));
1700 atomic_set(&sbi->s_flex_groups[flex_group].free_blocks,
1701 ext4_free_blks_count(sb, gdp));
7d39db14
TT		 1702 atomic_set(&sbi->s_flex_groups[flex_group].used_dirs,
1703 ext4_used_dirs_count(sb, gdp));
772cb7c8
JS		 1704 }
1705
1706 return 1;
1707failed:
1708 return 0;
1709}
1710
717d50e4
AD		 1711__le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1712 struct ext4_group_desc *gdp)
1713{
1714 __u16 crc = 0;
1715
1716 if (sbi->s_es->s_feature_ro_compat &
1717 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1718 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1719 __le32 le_group = cpu_to_le32(block_group);
1720
1721 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1722 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1723 crc = crc16(crc, (__u8 *)gdp, offset);
1724 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1725 /* for checksum of struct ext4_group_desc do the rest...*/
1726 if ((sbi->s_es->s_feature_incompat &
1727 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1728 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1729 crc = crc16(crc, (__u8 *)gdp + offset,
1730 le16_to_cpu(sbi->s_es->s_desc_size) -
1731 offset);
1732 }
1733
1734 return cpu_to_le16(crc);
1735}
1736
1737int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1738 struct ext4_group_desc *gdp)
1739{
1740 if ((sbi->s_es->s_feature_ro_compat &
1741 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1742 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1743 return 0;
1744
1745 return 1;
1746}
1747
ac27a0ec 1748/* Called at mount-time, super-block is locked */
197cd65a 1749static int ext4_check_descriptors(struct super_block *sb)
ac27a0ec 1750{
617ba13b
MC		 1751 struct ext4_sb_info *sbi = EXT4_SB(sb);
1752 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1753 ext4_fsblk_t last_block;
bd81d8ee
LV		 1754 ext4_fsblk_t block_bitmap;
1755 ext4_fsblk_t inode_bitmap;
1756 ext4_fsblk_t inode_table;
ce421581 1757 int flexbg_flag = 0;
fd2d4291 1758 ext4_group_t i;
ac27a0ec 1759
ce421581
JS		 1760 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1761 flexbg_flag = 1;
1762
af5bc92d 1763 ext4_debug("Checking group descriptors");
ac27a0ec 1764
197cd65a
AM		 1765 for (i = 0; i < sbi->s_groups_count; i++) {
1766 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1767
ce421581 1768 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 1769 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 1770 else
1771 last_block = first_block +
617ba13b 1772 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 1773
8fadc143 1774 block_bitmap = ext4_block_bitmap(sb, gdp);
2b2d6d01 1775 if (block_bitmap < first_block || block_bitmap > last_block) {
b31e1552 1776 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 1777 "Block bitmap for group %u not in group "
b31e1552 1778 "(block %llu)!", i, block_bitmap);
ac27a0ec
DK		 1779 return 0;
1780 }
8fadc143 1781 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2b2d6d01 1782 if (inode_bitmap < first_block || inode_bitmap > last_block) {
b31e1552 1783 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 1784 "Inode bitmap for group %u not in group "
b31e1552 1785 "(block %llu)!", i, inode_bitmap);
ac27a0ec
DK		 1786 return 0;
1787 }
8fadc143 1788 inode_table = ext4_inode_table(sb, gdp);
bd81d8ee 1789 if (inode_table < first_block ||
2b2d6d01 1790 inode_table + sbi->s_itb_per_group - 1 > last_block) {
b31e1552 1791 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 1792 "Inode table for group %u not in group "
b31e1552 1793 "(block %llu)!", i, inode_table);
ac27a0ec
DK		 1794 return 0;
1795 }
955ce5f5 1796 ext4_lock_group(sb, i);
717d50e4 1797 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
b31e1552
ES		 1798 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1799 "Checksum for group %u failed (%u!=%u)",
1800 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1801 gdp)), le16_to_cpu(gdp->bg_checksum));
7ee1ec4c 1802 if (!(sb->s_flags & MS_RDONLY)) {
955ce5f5 1803 ext4_unlock_group(sb, i);
8a266467 1804 return 0;
7ee1ec4c 1805 }
717d50e4 1806 }
955ce5f5 1807 ext4_unlock_group(sb, i);
ce421581
JS		 1808 if (!flexbg_flag)
1809 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec
DK		 1810 }
1811
bd81d8ee 1812 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
0b8e58a1 1813 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
ac27a0ec
DK		 1814 return 1;
1815}
1816
617ba13b 1817/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 1818 * the superblock) which were deleted from all directories, but held open by
1819 * a process at the time of a crash. We walk the list and try to delete these
1820 * inodes at recovery time (only with a read-write filesystem).
1821 *
1822 * In order to keep the orphan inode chain consistent during traversal (in
1823 * case of crash during recovery), we link each inode into the superblock
1824 * orphan list_head and handle it the same way as an inode deletion during
1825 * normal operation (which journals the operations for us).
1826 *
1827 * We only do an iget() and an iput() on each inode, which is very safe if we
1828 * accidentally point at an in-use or already deleted inode. The worst that
1829 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 1830 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 1831 * e2fsck was run on this filesystem, and it must have already done the orphan
1832 * inode cleanup for us, so we can safely abort without any further action.
1833 */
2b2d6d01
TT		 1834static void ext4_orphan_cleanup(struct super_block *sb,
1835 struct ext4_super_block *es)
ac27a0ec
DK		 1836{
1837 unsigned int s_flags = sb->s_flags;
1838 int nr_orphans = 0, nr_truncates = 0;
1839#ifdef CONFIG_QUOTA
1840 int i;
1841#endif
1842 if (!es->s_last_orphan) {
1843 jbd_debug(4, "no orphan inodes to clean up\n");
1844 return;
1845 }
1846
a8f48a95 1847 if (bdev_read_only(sb->s_bdev)) {
b31e1552
ES		 1848 ext4_msg(sb, KERN_ERR, "write access "
1849 "unavailable, skipping orphan cleanup");
a8f48a95
ES		 1850 return;
1851 }
1852
617ba13b 1853 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
ac27a0ec
DK		 1854 if (es->s_last_orphan)
1855 jbd_debug(1, "Errors on filesystem, "
1856 "clearing orphan list.\n");
1857 es->s_last_orphan = 0;
1858 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1859 return;
1860 }
1861
1862 if (s_flags & MS_RDONLY) {
b31e1552 1863 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
ac27a0ec
DK		 1864 sb->s_flags &= ~MS_RDONLY;
1865 }
1866#ifdef CONFIG_QUOTA
1867 /* Needed for iput() to work correctly and not trash data */
1868 sb->s_flags |= MS_ACTIVE;
1869 /* Turn on quotas so that they are updated correctly */
1870 for (i = 0; i < MAXQUOTAS; i++) {
617ba13b
MC		 1871 if (EXT4_SB(sb)->s_qf_names[i]) {
1872 int ret = ext4_quota_on_mount(sb, i);
ac27a0ec 1873 if (ret < 0)
b31e1552
ES		 1874 ext4_msg(sb, KERN_ERR,
1875 "Cannot turn on journaled "
1876 "quota: error %d", ret);
ac27a0ec
DK		 1877 }
1878 }
1879#endif
1880
1881 while (es->s_last_orphan) {
1882 struct inode *inode;
1883
97bd42b9
JB		 1884 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1885 if (IS_ERR(inode)) {
ac27a0ec
DK		 1886 es->s_last_orphan = 0;
1887 break;
1888 }
1889
617ba13b 1890 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
a269eb18 1891 vfs_dq_init(inode);
ac27a0ec 1892 if (inode->i_nlink) {
b31e1552
ES		 1893 ext4_msg(sb, KERN_DEBUG,
1894 "%s: truncating inode %lu to %lld bytes",
46e665e9 1895 __func__, inode->i_ino, inode->i_size);
e5f8eab8 1896 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
ac27a0ec 1897 inode->i_ino, inode->i_size);
617ba13b 1898 ext4_truncate(inode);
ac27a0ec
DK		 1899 nr_truncates++;
1900 } else {
b31e1552
ES		 1901 ext4_msg(sb, KERN_DEBUG,
1902 "%s: deleting unreferenced inode %lu",
46e665e9 1903 __func__, inode->i_ino);
ac27a0ec
DK		 1904 jbd_debug(2, "deleting unreferenced inode %lu\n",
1905 inode->i_ino);
1906 nr_orphans++;
1907 }
1908 iput(inode); /* The delete magic happens here! */
1909 }
1910
2b2d6d01 1911#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
ac27a0ec
DK		 1912
1913 if (nr_orphans)
b31e1552
ES		 1914 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
1915 PLURAL(nr_orphans));
ac27a0ec 1916 if (nr_truncates)
b31e1552
ES		 1917 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
1918 PLURAL(nr_truncates));
ac27a0ec
DK		 1919#ifdef CONFIG_QUOTA
1920 /* Turn quotas off */
1921 for (i = 0; i < MAXQUOTAS; i++) {
1922 if (sb_dqopt(sb)->files[i])
6f28e087 1923 vfs_quota_off(sb, i, 0);
ac27a0ec
DK		 1924 }
1925#endif
1926 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1927}
0b8e58a1 1928
cd2291a4
ES		 1929/*
1930 * Maximal extent format file size.
1931 * Resulting logical blkno at s_maxbytes must fit in our on-disk
1932 * extent format containers, within a sector_t, and within i_blocks
1933 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
1934 * so that won't be a limiting factor.
1935 *
1936 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1937 */
f287a1a5 1938static loff_t ext4_max_size(int blkbits, int has_huge_files)
cd2291a4
ES		 1939{
1940 loff_t res;
1941 loff_t upper_limit = MAX_LFS_FILESIZE;
1942
1943 /* small i_blocks in vfs inode? */
f287a1a5 1944 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
cd2291a4 1945 /*
90c699a9 1946 * CONFIG_LBDAF is not enabled implies the inode
cd2291a4
ES		 1947 * i_block represent total blocks in 512 bytes
1948 * 32 == size of vfs inode i_blocks * 8
1949 */
1950 upper_limit = (1LL << 32) - 1;
1951
1952 /* total blocks in file system block size */
1953 upper_limit >>= (blkbits - 9);
1954 upper_limit <<= blkbits;
1955 }
1956
1957 /* 32-bit extent-start container, ee_block */
1958 res = 1LL << 32;
1959 res <<= blkbits;
1960 res -= 1;
1961
1962 /* Sanity check against vm- & vfs- imposed limits */
1963 if (res > upper_limit)
1964 res = upper_limit;
1965
1966 return res;
1967}
ac27a0ec 1968
ac27a0ec 1969/*
cd2291a4 1970 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 1971 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
1972 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 1973 */
f287a1a5 1974static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
ac27a0ec 1975{
617ba13b 1976 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 1977 int meta_blocks;
1978 loff_t upper_limit;
0b8e58a1
AD		 1979 /* This is calculated to be the largest file size for a dense, block
1980 * mapped file such that the file's total number of 512-byte sectors,
1981 * including data and all indirect blocks, does not exceed (2^48 - 1).
1982 *
1983 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
1984 * number of 512-byte sectors of the file.
0fc1b451
AK		 1985 */
1986
f287a1a5 1987 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
0fc1b451 1988 /*
90c699a9 1989 * !has_huge_files or CONFIG_LBDAF not enabled implies that
0b8e58a1
AD		 1990 * the inode i_block field represents total file blocks in
1991 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
0fc1b451
AK		 1992 */
1993 upper_limit = (1LL << 32) - 1;
1994
1995 /* total blocks in file system block size */
1996 upper_limit >>= (bits - 9);
1997
1998 } else {
8180a562
AK		 1999 /*
2000 * We use 48 bit ext4_inode i_blocks
2001 * With EXT4_HUGE_FILE_FL set the i_blocks
2002 * represent total number of blocks in
2003 * file system block size
2004 */
0fc1b451
AK		 2005 upper_limit = (1LL << 48) - 1;
2006
0fc1b451
AK		 2007 }
2008
2009 /* indirect blocks */
2010 meta_blocks = 1;
2011 /* double indirect blocks */
2012 meta_blocks += 1 + (1LL << (bits-2));
2013 /* tripple indirect blocks */
2014 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2015
2016 upper_limit -= meta_blocks;
2017 upper_limit <<= bits;
ac27a0ec
DK		 2018
2019 res += 1LL << (bits-2);
2020 res += 1LL << (2*(bits-2));
2021 res += 1LL << (3*(bits-2));
2022 res <<= bits;
2023 if (res > upper_limit)
2024 res = upper_limit;
0fc1b451
AK		 2025
2026 if (res > MAX_LFS_FILESIZE)
2027 res = MAX_LFS_FILESIZE;
2028
ac27a0ec
DK		 2029 return res;
2030}
2031
617ba13b 2032static ext4_fsblk_t descriptor_loc(struct super_block *sb,
0b8e58a1 2033 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 2034{
617ba13b 2035 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 2036 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 2037 int has_super = 0;
2038
2039 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2040
617ba13b 2041 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
ac27a0ec 2042 nr < first_meta_bg)
70bbb3e0 2043 return logical_sb_block + nr + 1;
ac27a0ec 2044 bg = sbi->s_desc_per_block * nr;
617ba13b 2045 if (ext4_bg_has_super(sb, bg))
ac27a0ec 2046 has_super = 1;
0b8e58a1 2047
617ba13b 2048 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 2049}
2050
c9de560d
AT		 2051/**
2052 * ext4_get_stripe_size: Get the stripe size.
2053 * @sbi: In memory super block info
2054 *
2055 * If we have specified it via mount option, then
2056 * use the mount option value. If the value specified at mount time is
2057 * greater than the blocks per group use the super block value.
2058 * If the super block value is greater than blocks per group return 0.
2059 * Allocator needs it be less than blocks per group.
2060 *
2061 */
2062static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2063{
2064 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2065 unsigned long stripe_width =
2066 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2067
2068 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2069 return sbi->s_stripe;
2070
2071 if (stripe_width <= sbi->s_blocks_per_group)
2072 return stripe_width;
2073
2074 if (stride <= sbi->s_blocks_per_group)
2075 return stride;
2076
2077 return 0;
2078}
ac27a0ec 2079
3197ebdb
TT		 2080/* sysfs supprt */
2081
2082struct ext4_attr {
2083 struct attribute attr;
2084 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2085 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2086 const char *, size_t);
2087 int offset;
2088};
2089
2090static int parse_strtoul(const char *buf,
2091 unsigned long max, unsigned long *value)
2092{
2093 char *endp;
2094
2095 while (*buf && isspace(*buf))
2096 buf++;
2097 *value = simple_strtoul(buf, &endp, 0);
2098 while (*endp && isspace(*endp))
2099 endp++;
2100 if (*endp || *value > max)
2101 return -EINVAL;
2102
2103 return 0;
2104}
2105
2106static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2107 struct ext4_sb_info *sbi,
2108 char *buf)
2109{
2110 return snprintf(buf, PAGE_SIZE, "%llu\n",
2111 (s64) percpu_counter_sum(&sbi->s_dirtyblocks_counter));
2112}
2113
2114static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2115 struct ext4_sb_info *sbi, char *buf)
2116{
2117 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2118
2119 return snprintf(buf, PAGE_SIZE, "%lu\n",
2120 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2121 sbi->s_sectors_written_start) >> 1);
2122}
2123
2124static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2125 struct ext4_sb_info *sbi, char *buf)
2126{
2127 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2128
2129 return snprintf(buf, PAGE_SIZE, "%llu\n",
2130 sbi->s_kbytes_written +
2131 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2132 EXT4_SB(sb)->s_sectors_written_start) >> 1));
2133}
2134
2135static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2136 struct ext4_sb_info *sbi,
2137 const char *buf, size_t count)
2138{
2139 unsigned long t;
2140
2141 if (parse_strtoul(buf, 0x40000000, &t))
2142 return -EINVAL;
2143
f7c43950 2144 if (!is_power_of_2(t))
3197ebdb
TT		 2145 return -EINVAL;
2146
2147 sbi->s_inode_readahead_blks = t;
2148 return count;
2149}
2150
2151static ssize_t sbi_ui_show(struct ext4_attr *a,
0b8e58a1 2152 struct ext4_sb_info *sbi, char *buf)
3197ebdb
TT		 2153{
2154 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2155
2156 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2157}
2158
2159static ssize_t sbi_ui_store(struct ext4_attr *a,
2160 struct ext4_sb_info *sbi,
2161 const char *buf, size_t count)
2162{
2163 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2164 unsigned long t;
2165
2166 if (parse_strtoul(buf, 0xffffffff, &t))
2167 return -EINVAL;
2168 *ui = t;
2169 return count;
2170}
2171
2172#define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2173static struct ext4_attr ext4_attr_##_name = { \
2174 .attr = {.name = __stringify(_name), .mode = _mode }, \
2175 .show = _show, \
2176 .store = _store, \
2177 .offset = offsetof(struct ext4_sb_info, _elname), \
2178}
2179#define EXT4_ATTR(name, mode, show, store) \
2180static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2181
2182#define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2183#define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2184#define EXT4_RW_ATTR_SBI_UI(name, elname) \
2185 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2186#define ATTR_LIST(name) &ext4_attr_##name.attr
2187
2188EXT4_RO_ATTR(delayed_allocation_blocks);
2189EXT4_RO_ATTR(session_write_kbytes);
2190EXT4_RO_ATTR(lifetime_write_kbytes);
2191EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2192 inode_readahead_blks_store, s_inode_readahead_blks);
11013911 2193EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
3197ebdb
TT		 2194EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2195EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2196EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2197EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2198EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2199EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2200
2201static struct attribute *ext4_attrs[] = {
2202 ATTR_LIST(delayed_allocation_blocks),
2203 ATTR_LIST(session_write_kbytes),
2204 ATTR_LIST(lifetime_write_kbytes),
2205 ATTR_LIST(inode_readahead_blks),
11013911 2206 ATTR_LIST(inode_goal),
3197ebdb
TT		 2207 ATTR_LIST(mb_stats),
2208 ATTR_LIST(mb_max_to_scan),
2209 ATTR_LIST(mb_min_to_scan),
2210 ATTR_LIST(mb_order2_req),
2211 ATTR_LIST(mb_stream_req),
2212 ATTR_LIST(mb_group_prealloc),
2213 NULL,
2214};
2215
2216static ssize_t ext4_attr_show(struct kobject *kobj,
2217 struct attribute *attr, char *buf)
2218{
2219 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2220 s_kobj);
2221 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2222
2223 return a->show ? a->show(a, sbi, buf) : 0;
2224}
2225
2226static ssize_t ext4_attr_store(struct kobject *kobj,
2227 struct attribute *attr,
2228 const char *buf, size_t len)
2229{
2230 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2231 s_kobj);
2232 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2233
2234 return a->store ? a->store(a, sbi, buf, len) : 0;
2235}
2236
2237static void ext4_sb_release(struct kobject *kobj)
2238{
2239 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2240 s_kobj);
2241 complete(&sbi->s_kobj_unregister);
2242}
2243
2244
2245static struct sysfs_ops ext4_attr_ops = {
2246 .show = ext4_attr_show,
2247 .store = ext4_attr_store,
2248};
2249
2250static struct kobj_type ext4_ktype = {
2251 .default_attrs = ext4_attrs,
2252 .sysfs_ops = &ext4_attr_ops,
2253 .release = ext4_sb_release,
2254};
2255
2b2d6d01 2256static int ext4_fill_super(struct super_block *sb, void *data, int silent)
7477827f
AK		 2257 __releases(kernel_lock)
2258 __acquires(kernel_lock)
ac27a0ec 2259{
2b2d6d01 2260 struct buffer_head *bh;
617ba13b
MC		 2261 struct ext4_super_block *es = NULL;
2262 struct ext4_sb_info *sbi;
2263 ext4_fsblk_t block;
2264 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 2265 ext4_fsblk_t logical_sb_block;
ac27a0ec 2266 unsigned long offset = 0;
ac27a0ec
DK		 2267 unsigned long journal_devnum = 0;
2268 unsigned long def_mount_opts;
2269 struct inode *root;
9f6200bb 2270 char *cp;
0390131b 2271 const char *descr;
1d1fe1ee 2272 int ret = -EINVAL;
ac27a0ec 2273 int blocksize;
4ec11028
TT		 2274 unsigned int db_count;
2275 unsigned int i;
f287a1a5 2276 int needs_recovery, has_huge_files;
3a06d778 2277 int features;
bd81d8ee 2278 __u64 blocks_count;
833f4077 2279 int err;
b3881f74 2280 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
ac27a0ec
DK		 2281
2282 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
2283 if (!sbi)
2284 return -ENOMEM;
705895b6
PE		 2285
2286 sbi->s_blockgroup_lock =
2287 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
2288 if (!sbi->s_blockgroup_lock) {
2289 kfree(sbi);
2290 return -ENOMEM;
2291 }
ac27a0ec
DK		 2292 sb->s_fs_info = sbi;
2293 sbi->s_mount_opt = 0;
617ba13b
MC		 2294 sbi->s_resuid = EXT4_DEF_RESUID;
2295 sbi->s_resgid = EXT4_DEF_RESGID;
240799cd 2296 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
d9c9bef1 2297 sbi->s_sb_block = sb_block;
afc32f7e
TT		 2298 sbi->s_sectors_written_start = part_stat_read(sb->s_bdev->bd_part,
2299 sectors[1]);
ac27a0ec
DK		 2300
2301 unlock_kernel();
2302
9f6200bb
TT		 2303 /* Cleanup superblock name */
2304 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
2305 *cp = '!';
2306
617ba13b 2307 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 2308 if (!blocksize) {
b31e1552 2309 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
ac27a0ec
DK		 2310 goto out_fail;
2311 }
2312
2313 /*
617ba13b 2314 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 2315 * block sizes. We need to calculate the offset from buffer start.
2316 */
617ba13b 2317 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 2318 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2319 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 2320 } else {
70bbb3e0 2321 logical_sb_block = sb_block;
ac27a0ec
DK		 2322 }
2323
70bbb3e0 2324 if (!(bh = sb_bread(sb, logical_sb_block))) {
b31e1552 2325 ext4_msg(sb, KERN_ERR, "unable to read superblock");
ac27a0ec
DK		 2326 goto out_fail;
2327 }
2328 /*
2329 * Note: s_es must be initialized as soon as possible because
617ba13b 2330 * some ext4 macro-instructions depend on its value
ac27a0ec 2331 */
617ba13b 2332 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
ac27a0ec
DK		 2333 sbi->s_es = es;
2334 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 2335 if (sb->s_magic != EXT4_SUPER_MAGIC)
2336 goto cantfind_ext4;
afc32f7e 2337 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
ac27a0ec
DK		 2338
2339 /* Set defaults before we parse the mount options */
2340 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
617ba13b 2341 if (def_mount_opts & EXT4_DEFM_DEBUG)
ac27a0ec 2342 set_opt(sbi->s_mount_opt, DEBUG);
617ba13b 2343 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
ac27a0ec 2344 set_opt(sbi->s_mount_opt, GRPID);
617ba13b 2345 if (def_mount_opts & EXT4_DEFM_UID16)
ac27a0ec 2346 set_opt(sbi->s_mount_opt, NO_UID32);
03010a33 2347#ifdef CONFIG_EXT4_FS_XATTR
617ba13b 2348 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
ac27a0ec 2349 set_opt(sbi->s_mount_opt, XATTR_USER);
2e7842b8 2350#endif
03010a33 2351#ifdef CONFIG_EXT4_FS_POSIX_ACL
617ba13b 2352 if (def_mount_opts & EXT4_DEFM_ACL)
ac27a0ec 2353 set_opt(sbi->s_mount_opt, POSIX_ACL);
2e7842b8 2354#endif
617ba13b
MC		 2355 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
2356 sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
2357 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
2358 sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
2359 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
2360 sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
2361
2362 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
ac27a0ec 2363 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
bb4f397a 2364 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
ceea16bf 2365 set_opt(sbi->s_mount_opt, ERRORS_CONT);
bb4f397a
AK		 2366 else
2367 set_opt(sbi->s_mount_opt, ERRORS_RO);
ac27a0ec
DK		 2368
2369 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
2370 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
30773840
TT		 2371 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
2372 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
2373 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
f4033903 2374 sbi->s_mb_history_max = default_mb_history_length;
ac27a0ec 2375
571640ca 2376 set_opt(sbi->s_mount_opt, BARRIER);
ac27a0ec 2377
dd919b98
AK		 2378 /*
2379 * enable delayed allocation by default
2380 * Use -o nodelalloc to turn it off
2381 */
2382 set_opt(sbi->s_mount_opt, DELALLOC);
2383
b3881f74
TT		 2384 if (!parse_options((char *) data, sb, &journal_devnum,
2385 &journal_ioprio, NULL, 0))
ac27a0ec
DK		 2386 goto failed_mount;
2387
2388 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
617ba13b 2389 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec 2390
617ba13b
MC		 2391 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
2392 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
2393 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
2394 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
b31e1552
ES		 2395 ext4_msg(sb, KERN_WARNING,
2396 "feature flags set on rev 0 fs, "
2397 "running e2fsck is recommended");
469108ff 2398
ac27a0ec
DK		 2399 /*
2400 * Check feature flags regardless of the revision level, since we
2401 * previously didn't change the revision level when setting the flags,
2402 * so there is a chance incompat flags are set on a rev 0 filesystem.
2403 */
617ba13b 2404 features = EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP);
ac27a0ec 2405 if (features) {
b31e1552
ES		 2406 ext4_msg(sb, KERN_ERR,
2407 "Couldn't mount because of "
2408 "unsupported optional features (%x)",
3a06d778
AK		 2409 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2410 ~EXT4_FEATURE_INCOMPAT_SUPP));
ac27a0ec
DK		 2411 goto failed_mount;
2412 }
617ba13b 2413 features = EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP);
ac27a0ec 2414 if (!(sb->s_flags & MS_RDONLY) && features) {
b31e1552
ES		 2415 ext4_msg(sb, KERN_ERR,
2416 "Couldn't mount RDWR because of "
2417 "unsupported optional features (%x)",
3a06d778
AK		 2418 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2419 ~EXT4_FEATURE_RO_COMPAT_SUPP));
ac27a0ec
DK		 2420 goto failed_mount;
2421 }
f287a1a5
TT		 2422 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2423 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
2424 if (has_huge_files) {
0fc1b451
AK		 2425 /*
2426 * Large file size enabled file system can only be
90c699a9 2427 * mount if kernel is build with CONFIG_LBDAF
0fc1b451
AK		 2428 */
2429 if (sizeof(root->i_blocks) < sizeof(u64) &&
2430 !(sb->s_flags & MS_RDONLY)) {
b31e1552 2431 ext4_msg(sb, KERN_ERR, "Filesystem with huge "
0fc1b451 2432 "files cannot be mounted read-write "
90c699a9 2433 "without CONFIG_LBDAF");
0fc1b451
AK		 2434 goto failed_mount;
2435 }
2436 }
ac27a0ec
DK		 2437 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
2438
617ba13b
MC		 2439 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
2440 blocksize > EXT4_MAX_BLOCK_SIZE) {
b31e1552
ES		 2441 ext4_msg(sb, KERN_ERR,
2442 "Unsupported filesystem blocksize %d", blocksize);
ac27a0ec
DK		 2443 goto failed_mount;
2444 }
2445
ac27a0ec 2446 if (sb->s_blocksize != blocksize) {
ce40733c
AK		 2447 /* Validate the filesystem blocksize */
2448 if (!sb_set_blocksize(sb, blocksize)) {
b31e1552 2449 ext4_msg(sb, KERN_ERR, "bad block size %d",
ce40733c 2450 blocksize);
ac27a0ec
DK		 2451 goto failed_mount;
2452 }
2453
2b2d6d01 2454 brelse(bh);
70bbb3e0
AM		 2455 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2456 offset = do_div(logical_sb_block, blocksize);
2457 bh = sb_bread(sb, logical_sb_block);
ac27a0ec 2458 if (!bh) {
b31e1552
ES		 2459 ext4_msg(sb, KERN_ERR,
2460 "Can't read superblock on 2nd try");
ac27a0ec
DK		 2461 goto failed_mount;
2462 }
617ba13b 2463 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
ac27a0ec 2464 sbi->s_es = es;
617ba13b 2465 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
b31e1552
ES		 2466 ext4_msg(sb, KERN_ERR,
2467 "Magic mismatch, very weird!");
ac27a0ec
DK		 2468 goto failed_mount;
2469 }
2470 }
2471
f287a1a5
TT		 2472 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
2473 has_huge_files);
2474 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
ac27a0ec 2475
617ba13b
MC		 2476 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2477 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2478 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
ac27a0ec
DK		 2479 } else {
2480 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2481 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
617ba13b 2482 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1330593e 2483 (!is_power_of_2(sbi->s_inode_size)) ||
ac27a0ec 2484 (sbi->s_inode_size > blocksize)) {
b31e1552
ES		 2485 ext4_msg(sb, KERN_ERR,
2486 "unsupported inode size: %d",
2b2d6d01 2487 sbi->s_inode_size);
ac27a0ec
DK		 2488 goto failed_mount;
2489 }
ef7f3835
KS		 2490 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2491 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
ac27a0ec 2492 }
0b8e58a1 2493
0d1ee42f
AR		 2494 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2495 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
8fadc143 2496 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 2497 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 2498 !is_power_of_2(sbi->s_desc_size)) {
b31e1552
ES		 2499 ext4_msg(sb, KERN_ERR,
2500 "unsupported descriptor size %lu",
0d1ee42f
AR		 2501 sbi->s_desc_size);
2502 goto failed_mount;
2503 }
2504 } else
2505 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
0b8e58a1 2506
ac27a0ec 2507 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 2508 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
b47b6f38 2509 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
617ba13b 2510 goto cantfind_ext4;
0b8e58a1 2511
617ba13b 2512 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 2513 if (sbi->s_inodes_per_block == 0)
617ba13b 2514 goto cantfind_ext4;
ac27a0ec
DK		 2515 sbi->s_itb_per_group = sbi->s_inodes_per_group /
2516 sbi->s_inodes_per_block;
0d1ee42f 2517 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 2518 sbi->s_sbh = bh;
2519 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 2520 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2521 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
0b8e58a1 2522
2b2d6d01 2523 for (i = 0; i < 4; i++)
ac27a0ec
DK		 2524 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2525 sbi->s_def_hash_version = es->s_def_hash_version;
f99b2589
TT		 2526 i = le32_to_cpu(es->s_flags);
2527 if (i & EXT2_FLAGS_UNSIGNED_HASH)
2528 sbi->s_hash_unsigned = 3;
2529 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
2530#ifdef __CHAR_UNSIGNED__
2531 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
2532 sbi->s_hash_unsigned = 3;
2533#else
2534 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
2535#endif
2536 sb->s_dirt = 1;
2537 }
ac27a0ec
DK		 2538
2539 if (sbi->s_blocks_per_group > blocksize * 8) {
b31e1552
ES		 2540 ext4_msg(sb, KERN_ERR,
2541 "#blocks per group too big: %lu",
2b2d6d01 2542 sbi->s_blocks_per_group);
ac27a0ec
DK		 2543 goto failed_mount;
2544 }
ac27a0ec 2545 if (sbi->s_inodes_per_group > blocksize * 8) {
b31e1552
ES		 2546 ext4_msg(sb, KERN_ERR,
2547 "#inodes per group too big: %lu",
2b2d6d01 2548 sbi->s_inodes_per_group);
ac27a0ec
DK		 2549 goto failed_mount;
2550 }
2551
bd81d8ee 2552 if (ext4_blocks_count(es) >
ac27a0ec 2553 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
b31e1552
ES		 2554 ext4_msg(sb, KERN_ERR, "filesystem"
2555 " too large to mount safely");
ac27a0ec 2556 if (sizeof(sector_t) < 8)
90c699a9 2557 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
ac27a0ec
DK		 2558 goto failed_mount;
2559 }
2560
617ba13b
MC		 2561 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2562 goto cantfind_ext4;
e7c95593 2563
0f2ddca6
FTN		 2564 /* check blocks count against device size */
2565 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
2566 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
b31e1552
ES		 2567 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
2568 "exceeds size of device (%llu blocks)",
0f2ddca6
FTN		 2569 ext4_blocks_count(es), blocks_count);
2570 goto failed_mount;
2571 }
2572
0b8e58a1
AD		 2573 /*
2574 * It makes no sense for the first data block to be beyond the end
2575 * of the filesystem.
2576 */
2577 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
b31e1552
ES		 2578 ext4_msg(sb, KERN_WARNING, "bad geometry: first data"
2579 "block %u is beyond end of filesystem (%llu)",
2580 le32_to_cpu(es->s_first_data_block),
2581 ext4_blocks_count(es));
e7c95593
ES		 2582 goto failed_mount;
2583 }
bd81d8ee
LV		 2584 blocks_count = (ext4_blocks_count(es) -
2585 le32_to_cpu(es->s_first_data_block) +
2586 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2587 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4ec11028 2588 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
b31e1552 2589 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
4ec11028 2590 "(block count %llu, first data block %u, "
b31e1552 2591 "blocks per group %lu)", sbi->s_groups_count,
4ec11028
TT		 2592 ext4_blocks_count(es),
2593 le32_to_cpu(es->s_first_data_block),
2594 EXT4_BLOCKS_PER_GROUP(sb));
2595 goto failed_mount;
2596 }
bd81d8ee 2597 sbi->s_groups_count = blocks_count;
617ba13b
MC		 2598 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2599 EXT4_DESC_PER_BLOCK(sb);
2b2d6d01 2600 sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
ac27a0ec
DK		 2601 GFP_KERNEL);
2602 if (sbi->s_group_desc == NULL) {
b31e1552 2603 ext4_msg(sb, KERN_ERR, "not enough memory");
ac27a0ec
DK		 2604 goto failed_mount;
2605 }
2606
3244fcb1 2607#ifdef CONFIG_PROC_FS
9f6200bb
TT		 2608 if (ext4_proc_root)
2609 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
3244fcb1 2610#endif
240799cd 2611
705895b6 2612 bgl_lock_init(sbi->s_blockgroup_lock);
ac27a0ec
DK		 2613
2614 for (i = 0; i < db_count; i++) {
70bbb3e0 2615 block = descriptor_loc(sb, logical_sb_block, i);
ac27a0ec
DK		 2616 sbi->s_group_desc[i] = sb_bread(sb, block);
2617 if (!sbi->s_group_desc[i]) {
b31e1552
ES		 2618 ext4_msg(sb, KERN_ERR,
2619 "can't read group descriptor %d", i);
ac27a0ec
DK		 2620 db_count = i;
2621 goto failed_mount2;
2622 }
2623 }
2b2d6d01 2624 if (!ext4_check_descriptors(sb)) {
b31e1552 2625 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
ac27a0ec
DK		 2626 goto failed_mount2;
2627 }
772cb7c8
JS		 2628 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2629 if (!ext4_fill_flex_info(sb)) {
b31e1552
ES		 2630 ext4_msg(sb, KERN_ERR,
2631 "unable to initialize "
2632 "flex_bg meta info!");
772cb7c8
JS		 2633 goto failed_mount2;
2634 }
2635
ac27a0ec
DK		 2636 sbi->s_gdb_count = db_count;
2637 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2638 spin_lock_init(&sbi->s_next_gen_lock);
2639
833f4077
PZ		 2640 err = percpu_counter_init(&sbi->s_freeblocks_counter,
2641 ext4_count_free_blocks(sb));
2642 if (!err) {
2643 err = percpu_counter_init(&sbi->s_freeinodes_counter,
2644 ext4_count_free_inodes(sb));
2645 }
2646 if (!err) {
2647 err = percpu_counter_init(&sbi->s_dirs_counter,
2648 ext4_count_dirs(sb));
2649 }
6bc6e63f
AK		 2650 if (!err) {
2651 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
2652 }
833f4077 2653 if (err) {
b31e1552 2654 ext4_msg(sb, KERN_ERR, "insufficient memory");
833f4077
PZ		 2655 goto failed_mount3;
2656 }
ac27a0ec 2657
c9de560d
AT		 2658 sbi->s_stripe = ext4_get_stripe_size(sbi);
2659
ac27a0ec
DK		 2660 /*
2661 * set up enough so that it can read an inode
2662 */
9ca92389
TT		 2663 if (!test_opt(sb, NOLOAD) &&
2664 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
2665 sb->s_op = &ext4_sops;
2666 else
2667 sb->s_op = &ext4_nojournal_sops;
617ba13b
MC		 2668 sb->s_export_op = &ext4_export_ops;
2669 sb->s_xattr = ext4_xattr_handlers;
ac27a0ec 2670#ifdef CONFIG_QUOTA
617ba13b
MC		 2671 sb->s_qcop = &ext4_qctl_operations;
2672 sb->dq_op = &ext4_quota_operations;
ac27a0ec
DK		 2673#endif
2674 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3b9d4ed2 2675 mutex_init(&sbi->s_orphan_lock);
32ed5058 2676 mutex_init(&sbi->s_resize_lock);
ac27a0ec
DK		 2677
2678 sb->s_root = NULL;
2679
2680 needs_recovery = (es->s_last_orphan != 0 ||
617ba13b
MC		 2681 EXT4_HAS_INCOMPAT_FEATURE(sb,
2682 EXT4_FEATURE_INCOMPAT_RECOVER));
ac27a0ec
DK		 2683
2684 /*
2685 * The first inode we look at is the journal inode. Don't try
2686 * root first: it may be modified in the journal!
2687 */
2688 if (!test_opt(sb, NOLOAD) &&
617ba13b
MC		 2689 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2690 if (ext4_load_journal(sb, es, journal_devnum))
ac27a0ec 2691 goto failed_mount3;
624080ed
TT		 2692 if (!(sb->s_flags & MS_RDONLY) &&
2693 EXT4_SB(sb)->s_journal->j_failed_commit) {
b31e1552 2694 ext4_msg(sb, KERN_CRIT, "error: "
624080ed 2695 "ext4_fill_super: Journal transaction "
b31e1552 2696 "%u is corrupt",
624080ed 2697 EXT4_SB(sb)->s_journal->j_failed_commit);
2b2d6d01 2698 if (test_opt(sb, ERRORS_RO)) {
b31e1552
ES		 2699 ext4_msg(sb, KERN_CRIT,
2700 "Mounting filesystem read-only");
624080ed
TT		 2701 sb->s_flags |= MS_RDONLY;
2702 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2703 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2704 }
2705 if (test_opt(sb, ERRORS_PANIC)) {
2706 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2707 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
e2d67052 2708 ext4_commit_super(sb, 1);
624080ed
TT		 2709 goto failed_mount4;
2710 }
2711 }
0390131b
FM		 2712 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
2713 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
b31e1552
ES		 2714 ext4_msg(sb, KERN_ERR, "required journal recovery "
2715 "suppressed and not mounted read-only");
0390131b 2716 goto failed_mount4;
ac27a0ec 2717 } else {
0390131b
FM		 2718 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
2719 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
2720 sbi->s_journal = NULL;
2721 needs_recovery = 0;
2722 goto no_journal;
ac27a0ec
DK		 2723 }
2724
eb40a09c
JS		 2725 if (ext4_blocks_count(es) > 0xffffffffULL &&
2726 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2727 JBD2_FEATURE_INCOMPAT_64BIT)) {
b31e1552 2728 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
eb40a09c
JS		 2729 goto failed_mount4;
2730 }
2731
818d276c
GS		 2732 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2733 jbd2_journal_set_features(sbi->s_journal,
2734 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2735 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2736 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2737 jbd2_journal_set_features(sbi->s_journal,
2738 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2739 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2740 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2741 } else {
2742 jbd2_journal_clear_features(sbi->s_journal,
2743 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2744 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2745 }
2746
ac27a0ec
DK		 2747 /* We have now updated the journal if required, so we can
2748 * validate the data journaling mode. */
2749 switch (test_opt(sb, DATA_FLAGS)) {
2750 case 0:
2751 /* No mode set, assume a default based on the journal
63f57933
AM		 2752 * capabilities: ORDERED_DATA if the journal can
2753 * cope, else JOURNAL_DATA
2754 */
dab291af
MC		 2755 if (jbd2_journal_check_available_features
2756 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
ac27a0ec
DK		 2757 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2758 else
2759 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2760 break;
2761
617ba13b
MC		 2762 case EXT4_MOUNT_ORDERED_DATA:
2763 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 2764 if (!jbd2_journal_check_available_features
2765 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
b31e1552
ES		 2766 ext4_msg(sb, KERN_ERR, "Journal does not support "
2767 "requested data journaling mode");
ac27a0ec
DK		 2768 goto failed_mount4;
2769 }
2770 default:
2771 break;
2772 }
b3881f74 2773 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
ac27a0ec 2774
0390131b 2775no_journal:
ac27a0ec
DK		 2776
2777 if (test_opt(sb, NOBH)) {
617ba13b 2778 if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
b31e1552
ES		 2779 ext4_msg(sb, KERN_WARNING, "Ignoring nobh option - "
2780 "its supported only with writeback mode");
ac27a0ec
DK		 2781 clear_opt(sbi->s_mount_opt, NOBH);
2782 }
2783 }
2784 /*
dab291af 2785 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 2786 * so we can safely mount the rest of the filesystem now.
2787 */
2788
1d1fe1ee
DH		 2789 root = ext4_iget(sb, EXT4_ROOT_INO);
2790 if (IS_ERR(root)) {
b31e1552 2791 ext4_msg(sb, KERN_ERR, "get root inode failed");
1d1fe1ee 2792 ret = PTR_ERR(root);
ac27a0ec
DK		 2793 goto failed_mount4;
2794 }
2795 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1d1fe1ee 2796 iput(root);
b31e1552 2797 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
ac27a0ec
DK		 2798 goto failed_mount4;
2799 }
1d1fe1ee
DH		 2800 sb->s_root = d_alloc_root(root);
2801 if (!sb->s_root) {
b31e1552 2802 ext4_msg(sb, KERN_ERR, "get root dentry failed");
1d1fe1ee
DH		 2803 iput(root);
2804 ret = -ENOMEM;
2805 goto failed_mount4;
2806 }
ac27a0ec 2807
2b2d6d01 2808 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
ef7f3835
KS		 2809
2810 /* determine the minimum size of new large inodes, if present */
2811 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2812 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2813 EXT4_GOOD_OLD_INODE_SIZE;
2814 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2815 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2816 if (sbi->s_want_extra_isize <
2817 le16_to_cpu(es->s_want_extra_isize))
2818 sbi->s_want_extra_isize =
2819 le16_to_cpu(es->s_want_extra_isize);
2820 if (sbi->s_want_extra_isize <
2821 le16_to_cpu(es->s_min_extra_isize))
2822 sbi->s_want_extra_isize =
2823 le16_to_cpu(es->s_min_extra_isize);
2824 }
2825 }
2826 /* Check if enough inode space is available */
2827 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2828 sbi->s_inode_size) {
2829 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2830 EXT4_GOOD_OLD_INODE_SIZE;
b31e1552
ES		 2831 ext4_msg(sb, KERN_INFO, "required extra inode space not"
2832 "available");
ef7f3835
KS		 2833 }
2834
c2774d84 2835 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
b31e1552
ES		 2836 ext4_msg(sb, KERN_WARNING, "Ignoring delalloc option - "
2837 "requested data journaling mode");
c2774d84
AK		 2838 clear_opt(sbi->s_mount_opt, DELALLOC);
2839 } else if (test_opt(sb, DELALLOC))
b31e1552 2840 ext4_msg(sb, KERN_INFO, "delayed allocation enabled");
c2774d84 2841
6fd058f7
TT		 2842 err = ext4_setup_system_zone(sb);
2843 if (err) {
b31e1552
ES		 2844 ext4_msg(sb, KERN_ERR, "failed to initialize system "
2845 "zone (%d)\n", err);
6fd058f7
TT		 2846 goto failed_mount4;
2847 }
2848
c2774d84
AK		 2849 ext4_ext_init(sb);
2850 err = ext4_mb_init(sb, needs_recovery);
2851 if (err) {
b31e1552
ES		 2852 ext4_msg(sb, KERN_ERR, "failed to initalize mballoc (%d)",
2853 err);
c2774d84
AK		 2854 goto failed_mount4;
2855 }
2856
3197ebdb
TT		 2857 sbi->s_kobj.kset = ext4_kset;
2858 init_completion(&sbi->s_kobj_unregister);
2859 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
2860 "%s", sb->s_id);
2861 if (err) {
2862 ext4_mb_release(sb);
2863 ext4_ext_release(sb);
2864 goto failed_mount4;
2865 };
2866
617ba13b
MC		 2867 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
2868 ext4_orphan_cleanup(sb, es);
2869 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
0390131b 2870 if (needs_recovery) {
b31e1552 2871 ext4_msg(sb, KERN_INFO, "recovery complete");
0390131b
FM		 2872 ext4_mark_recovery_complete(sb, es);
2873 }
2874 if (EXT4_SB(sb)->s_journal) {
2875 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2876 descr = " journalled data mode";
2877 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2878 descr = " ordered data mode";
2879 else
2880 descr = " writeback data mode";
2881 } else
2882 descr = "out journal";
2883
b31e1552 2884 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s", descr);
ac27a0ec
DK		 2885
2886 lock_kernel();
2887 return 0;
2888
617ba13b 2889cantfind_ext4:
ac27a0ec 2890 if (!silent)
b31e1552 2891 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
ac27a0ec
DK		 2892 goto failed_mount;
2893
2894failed_mount4:
b31e1552 2895 ext4_msg(sb, KERN_ERR, "mount failed");
6fd058f7 2896 ext4_release_system_zone(sb);
0390131b
FM		 2897 if (sbi->s_journal) {
2898 jbd2_journal_destroy(sbi->s_journal);
2899 sbi->s_journal = NULL;
2900 }
ac27a0ec 2901failed_mount3:
c5ca7c76
TT		 2902 if (sbi->s_flex_groups) {
2903 if (is_vmalloc_addr(sbi->s_flex_groups))
2904 vfree(sbi->s_flex_groups);
2905 else
2906 kfree(sbi->s_flex_groups);
2907 }
ac27a0ec
DK		 2908 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2909 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2910 percpu_counter_destroy(&sbi->s_dirs_counter);
6bc6e63f 2911 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
ac27a0ec
DK		 2912failed_mount2:
2913 for (i = 0; i < db_count; i++)
2914 brelse(sbi->s_group_desc[i]);
2915 kfree(sbi->s_group_desc);
2916failed_mount:
240799cd 2917 if (sbi->s_proc) {
9f6200bb 2918 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 2919 }
ac27a0ec
DK		 2920#ifdef CONFIG_QUOTA
2921 for (i = 0; i < MAXQUOTAS; i++)
2922 kfree(sbi->s_qf_names[i]);
2923#endif
617ba13b 2924 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 2925 brelse(bh);
2926out_fail:
2927 sb->s_fs_info = NULL;
f6830165 2928 kfree(sbi->s_blockgroup_lock);
ac27a0ec
DK		 2929 kfree(sbi);
2930 lock_kernel();
1d1fe1ee 2931 return ret;
ac27a0ec
DK		 2932}
2933
2934/*
2935 * Setup any per-fs journal parameters now. We'll do this both on
2936 * initial mount, once the journal has been initialised but before we've
2937 * done any recovery; and again on any subsequent remount.
2938 */
617ba13b 2939static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 2940{
617ba13b 2941 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 2942
30773840
TT		 2943 journal->j_commit_interval = sbi->s_commit_interval;
2944 journal->j_min_batch_time = sbi->s_min_batch_time;
2945 journal->j_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 2946
2947 spin_lock(&journal->j_state_lock);
2948 if (test_opt(sb, BARRIER))
dab291af 2949 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 2950 else
dab291af 2951 journal->j_flags &= ~JBD2_BARRIER;
5bf5683a
HK		 2952 if (test_opt(sb, DATA_ERR_ABORT))
2953 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
2954 else
2955 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
ac27a0ec
DK		 2956 spin_unlock(&journal->j_state_lock);
2957}
2958
617ba13b 2959static journal_t *ext4_get_journal(struct super_block *sb,
ac27a0ec
DK		 2960 unsigned int journal_inum)
2961{
2962 struct inode *journal_inode;
2963 journal_t *journal;
2964
0390131b
FM		 2965 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
2966
ac27a0ec
DK		 2967 /* First, test for the existence of a valid inode on disk. Bad
2968 * things happen if we iget() an unused inode, as the subsequent
2969 * iput() will try to delete it. */
2970
1d1fe1ee
DH		 2971 journal_inode = ext4_iget(sb, journal_inum);
2972 if (IS_ERR(journal_inode)) {
b31e1552 2973 ext4_msg(sb, KERN_ERR, "no journal found");
ac27a0ec
DK		 2974 return NULL;
2975 }
2976 if (!journal_inode->i_nlink) {
2977 make_bad_inode(journal_inode);
2978 iput(journal_inode);
b31e1552 2979 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
ac27a0ec
DK		 2980 return NULL;
2981 }
2982
e5f8eab8 2983 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
ac27a0ec 2984 journal_inode, journal_inode->i_size);
1d1fe1ee 2985 if (!S_ISREG(journal_inode->i_mode)) {
b31e1552 2986 ext4_msg(sb, KERN_ERR, "invalid journal inode");
ac27a0ec
DK		 2987 iput(journal_inode);
2988 return NULL;
2989 }
2990
dab291af 2991 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 2992 if (!journal) {
b31e1552 2993 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
ac27a0ec
DK		 2994 iput(journal_inode);
2995 return NULL;
2996 }
2997 journal->j_private = sb;
617ba13b 2998 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 2999 return journal;
3000}
3001
617ba13b 3002static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 3003 dev_t j_dev)
3004{
2b2d6d01 3005 struct buffer_head *bh;
ac27a0ec 3006 journal_t *journal;
617ba13b
MC		 3007 ext4_fsblk_t start;
3008 ext4_fsblk_t len;
ac27a0ec 3009 int hblock, blocksize;
617ba13b 3010 ext4_fsblk_t sb_block;
ac27a0ec 3011 unsigned long offset;
2b2d6d01 3012 struct ext4_super_block *es;
ac27a0ec
DK		 3013 struct block_device *bdev;
3014
0390131b
FM		 3015 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3016
b31e1552 3017 bdev = ext4_blkdev_get(j_dev, sb);
ac27a0ec
DK		 3018 if (bdev == NULL)
3019 return NULL;
3020
3021 if (bd_claim(bdev, sb)) {
b31e1552
ES		 3022 ext4_msg(sb, KERN_ERR,
3023 "failed to claim external journal device");
9a1c3542 3024 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
ac27a0ec
DK		 3025 return NULL;
3026 }
3027
3028 blocksize = sb->s_blocksize;
e1defc4f 3029 hblock = bdev_logical_block_size(bdev);
ac27a0ec 3030 if (blocksize < hblock) {
b31e1552
ES		 3031 ext4_msg(sb, KERN_ERR,
3032 "blocksize too small for journal device");
ac27a0ec
DK		 3033 goto out_bdev;
3034 }
3035
617ba13b
MC		 3036 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3037 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 3038 set_blocksize(bdev, blocksize);
3039 if (!(bh = __bread(bdev, sb_block, blocksize))) {
b31e1552
ES		 3040 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
3041 "external journal");
ac27a0ec
DK		 3042 goto out_bdev;
3043 }
3044
617ba13b
MC		 3045 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3046 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 3047 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b 3048 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
b31e1552
ES		 3049 ext4_msg(sb, KERN_ERR, "external journal has "
3050 "bad superblock");
ac27a0ec
DK		 3051 brelse(bh);
3052 goto out_bdev;
3053 }
3054
617ba13b 3055 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
b31e1552 3056 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
ac27a0ec
DK		 3057 brelse(bh);
3058 goto out_bdev;
3059 }
3060
bd81d8ee 3061 len = ext4_blocks_count(es);
ac27a0ec
DK		 3062 start = sb_block + 1;
3063 brelse(bh); /* we're done with the superblock */
3064
dab291af 3065 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 3066 start, len, blocksize);
3067 if (!journal) {
b31e1552 3068 ext4_msg(sb, KERN_ERR, "failed to create device journal");
ac27a0ec
DK		 3069 goto out_bdev;
3070 }
3071 journal->j_private = sb;
3072 ll_rw_block(READ, 1, &journal->j_sb_buffer);
3073 wait_on_buffer(journal->j_sb_buffer);
3074 if (!buffer_uptodate(journal->j_sb_buffer)) {
b31e1552 3075 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
ac27a0ec
DK		 3076 goto out_journal;
3077 }
3078 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
b31e1552
ES		 3079 ext4_msg(sb, KERN_ERR, "External journal has more than one "
3080 "user (unsupported) - %d",
ac27a0ec
DK		 3081 be32_to_cpu(journal->j_superblock->s_nr_users));
3082 goto out_journal;
3083 }
617ba13b
MC		 3084 EXT4_SB(sb)->journal_bdev = bdev;
3085 ext4_init_journal_params(sb, journal);
ac27a0ec 3086 return journal;
0b8e58a1 3087
ac27a0ec 3088out_journal:
dab291af 3089 jbd2_journal_destroy(journal);
ac27a0ec 3090out_bdev:
617ba13b 3091 ext4_blkdev_put(bdev);
ac27a0ec
DK		 3092 return NULL;
3093}
3094
617ba13b
MC		 3095static int ext4_load_journal(struct super_block *sb,
3096 struct ext4_super_block *es,
ac27a0ec
DK		 3097 unsigned long journal_devnum)
3098{
3099 journal_t *journal;
3100 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
3101 dev_t journal_dev;
3102 int err = 0;
3103 int really_read_only;
3104
0390131b
FM		 3105 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3106
ac27a0ec
DK		 3107 if (journal_devnum &&
3108 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
b31e1552
ES		 3109 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
3110 "numbers have changed");
ac27a0ec
DK		 3111 journal_dev = new_decode_dev(journal_devnum);
3112 } else
3113 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
3114
3115 really_read_only = bdev_read_only(sb->s_bdev);
3116
3117 /*
3118 * Are we loading a blank journal or performing recovery after a
3119 * crash? For recovery, we need to check in advance whether we
3120 * can get read-write access to the device.
3121 */
617ba13b 3122 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
ac27a0ec 3123 if (sb->s_flags & MS_RDONLY) {
b31e1552
ES		 3124 ext4_msg(sb, KERN_INFO, "INFO: recovery "
3125 "required on readonly filesystem");
ac27a0ec 3126 if (really_read_only) {
b31e1552
ES		 3127 ext4_msg(sb, KERN_ERR, "write access "
3128 "unavailable, cannot proceed");
ac27a0ec
DK		 3129 return -EROFS;
3130 }
b31e1552
ES		 3131 ext4_msg(sb, KERN_INFO, "write access will "
3132 "be enabled during recovery");
ac27a0ec
DK		 3133 }
3134 }
3135
3136 if (journal_inum && journal_dev) {
b31e1552
ES		 3137 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
3138 "and inode journals!");
ac27a0ec
DK		 3139 return -EINVAL;
3140 }
3141
3142 if (journal_inum) {
617ba13b 3143 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 3144 return -EINVAL;
3145 } else {
617ba13b 3146 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 3147 return -EINVAL;
3148 }
3149
4776004f 3150 if (journal->j_flags & JBD2_BARRIER)
b31e1552 3151 ext4_msg(sb, KERN_INFO, "barriers enabled");
4776004f 3152 else
b31e1552 3153 ext4_msg(sb, KERN_INFO, "barriers disabled");
4776004f 3154
ac27a0ec 3155 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
dab291af 3156 err = jbd2_journal_update_format(journal);
ac27a0ec 3157 if (err) {
b31e1552 3158 ext4_msg(sb, KERN_ERR, "error updating journal");
dab291af 3159 jbd2_journal_destroy(journal);
ac27a0ec
DK		 3160 return err;
3161 }
3162 }
3163
617ba13b 3164 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
dab291af 3165 err = jbd2_journal_wipe(journal, !really_read_only);
ac27a0ec 3166 if (!err)
dab291af 3167 err = jbd2_journal_load(journal);
ac27a0ec
DK		 3168
3169 if (err) {
b31e1552 3170 ext4_msg(sb, KERN_ERR, "error loading journal");
dab291af 3171 jbd2_journal_destroy(journal);
ac27a0ec
DK		 3172 return err;
3173 }
3174
617ba13b
MC		 3175 EXT4_SB(sb)->s_journal = journal;
3176 ext4_clear_journal_err(sb, es);
ac27a0ec
DK		 3177
3178 if (journal_devnum &&
3179 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3180 es->s_journal_dev = cpu_to_le32(journal_devnum);
ac27a0ec
DK		 3181
3182 /* Make sure we flush the recovery flag to disk. */
e2d67052 3183 ext4_commit_super(sb, 1);
ac27a0ec
DK		 3184 }
3185
3186 return 0;
3187}
3188
e2d67052 3189static int ext4_commit_super(struct super_block *sb, int sync)
ac27a0ec 3190{
e2d67052 3191 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
617ba13b 3192 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
c4be0c1d 3193 int error = 0;
ac27a0ec
DK		 3194
3195 if (!sbh)
c4be0c1d 3196 return error;
914258bf
TT		 3197 if (buffer_write_io_error(sbh)) {
3198 /*
3199 * Oh, dear. A previous attempt to write the
3200 * superblock failed. This could happen because the
3201 * USB device was yanked out. Or it could happen to
3202 * be a transient write error and maybe the block will
3203 * be remapped. Nothing we can do but to retry the
3204 * write and hope for the best.
3205 */
b31e1552
ES		 3206 ext4_msg(sb, KERN_ERR, "previous I/O error to "
3207 "superblock detected");
914258bf
TT		 3208 clear_buffer_write_io_error(sbh);
3209 set_buffer_uptodate(sbh);
3210 }
ac27a0ec 3211 es->s_wtime = cpu_to_le32(get_seconds());
afc32f7e
TT		 3212 es->s_kbytes_written =
3213 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
3214 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
3215 EXT4_SB(sb)->s_sectors_written_start) >> 1));
5d1b1b3f
AK		 3216 ext4_free_blocks_count_set(es, percpu_counter_sum_positive(
3217 &EXT4_SB(sb)->s_freeblocks_counter));
3218 es->s_free_inodes_count = cpu_to_le32(percpu_counter_sum_positive(
3219 &EXT4_SB(sb)->s_freeinodes_counter));
7234ab2a 3220 sb->s_dirt = 0;
ac27a0ec
DK		 3221 BUFFER_TRACE(sbh, "marking dirty");
3222 mark_buffer_dirty(sbh);
914258bf 3223 if (sync) {
c4be0c1d
TS		 3224 error = sync_dirty_buffer(sbh);
3225 if (error)
3226 return error;
3227
3228 error = buffer_write_io_error(sbh);
3229 if (error) {
b31e1552
ES		 3230 ext4_msg(sb, KERN_ERR, "I/O error while writing "
3231 "superblock");
914258bf
TT		 3232 clear_buffer_write_io_error(sbh);
3233 set_buffer_uptodate(sbh);
3234 }
3235 }
c4be0c1d 3236 return error;
ac27a0ec
DK		 3237}
3238
ac27a0ec
DK		 3239/*
3240 * Have we just finished recovery? If so, and if we are mounting (or
3241 * remounting) the filesystem readonly, then we will end up with a
3242 * consistent fs on disk. Record that fact.
3243 */
2b2d6d01
TT		 3244static void ext4_mark_recovery_complete(struct super_block *sb,
3245 struct ext4_super_block *es)
ac27a0ec 3246{
617ba13b 3247 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 3248
0390131b
FM		 3249 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3250 BUG_ON(journal != NULL);
3251 return;
3252 }
dab291af 3253 jbd2_journal_lock_updates(journal);
7ffe1ea8
HK		 3254 if (jbd2_journal_flush(journal) < 0)
3255 goto out;
3256
617ba13b 3257 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
ac27a0ec 3258 sb->s_flags & MS_RDONLY) {
617ba13b 3259 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
e2d67052 3260 ext4_commit_super(sb, 1);
ac27a0ec 3261 }
7ffe1ea8
HK		 3262
3263out:
dab291af 3264 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 3265}
3266
3267/*
3268 * If we are mounting (or read-write remounting) a filesystem whose journal
3269 * has recorded an error from a previous lifetime, move that error to the
3270 * main filesystem now.
3271 */
2b2d6d01
TT		 3272static void ext4_clear_journal_err(struct super_block *sb,
3273 struct ext4_super_block *es)
ac27a0ec
DK		 3274{
3275 journal_t *journal;
3276 int j_errno;
3277 const char *errstr;
3278
0390131b
FM		 3279 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3280
617ba13b 3281 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 3282
3283 /*
3284 * Now check for any error status which may have been recorded in the
617ba13b 3285 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 3286 */
3287
dab291af 3288 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 3289 if (j_errno) {
3290 char nbuf[16];
3291
617ba13b 3292 errstr = ext4_decode_error(sb, j_errno, nbuf);
46e665e9 3293 ext4_warning(sb, __func__, "Filesystem error recorded "
ac27a0ec 3294 "from previous mount: %s", errstr);
46e665e9 3295 ext4_warning(sb, __func__, "Marking fs in need of "
ac27a0ec
DK		 3296 "filesystem check.");
3297
617ba13b
MC		 3298 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
3299 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
e2d67052 3300 ext4_commit_super(sb, 1);
ac27a0ec 3301
dab291af 3302 jbd2_journal_clear_err(journal);
ac27a0ec
DK		 3303 }
3304}
3305
3306/*
3307 * Force the running and committing transactions to commit,
3308 * and wait on the commit.
3309 */
617ba13b 3310int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 3311{
3312 journal_t *journal;
0390131b 3313 int ret = 0;
ac27a0ec
DK		 3314
3315 if (sb->s_flags & MS_RDONLY)
3316 return 0;
3317
617ba13b 3318 journal = EXT4_SB(sb)->s_journal;
7234ab2a 3319 if (journal)
0390131b 3320 ret = ext4_journal_force_commit(journal);
0390131b 3321
ac27a0ec
DK		 3322 return ret;
3323}
3324
2b2d6d01 3325static void ext4_write_super(struct super_block *sb)
ac27a0ec 3326{
ebc1ac16 3327 lock_super(sb);
9ca92389 3328 ext4_commit_super(sb, 1);
ebc1ac16 3329 unlock_super(sb);
ac27a0ec
DK		 3330}
3331
617ba13b 3332static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec 3333{
14ce0cb4 3334 int ret = 0;
9eddacf9 3335 tid_t target;
ac27a0ec 3336
9bffad1e 3337 trace_ext4_sync_fs(sb, wait);
9ca92389
TT		 3338 if (jbd2_journal_start_commit(EXT4_SB(sb)->s_journal, &target)) {
3339 if (wait)
3340 jbd2_log_wait_commit(EXT4_SB(sb)->s_journal, target);
0390131b 3341 }
14ce0cb4 3342 return ret;
ac27a0ec
DK		 3343}
3344
3345/*
3346 * LVM calls this function before a (read-only) snapshot is created. This
3347 * gives us a chance to flush the journal completely and mark the fs clean.
3348 */
c4be0c1d 3349static int ext4_freeze(struct super_block *sb)
ac27a0ec 3350{
c4be0c1d
TS		 3351 int error = 0;
3352 journal_t *journal;
ac27a0ec 3353
9ca92389
TT		 3354 if (sb->s_flags & MS_RDONLY)
3355 return 0;
ac27a0ec 3356
9ca92389 3357 journal = EXT4_SB(sb)->s_journal;
7ffe1ea8 3358
9ca92389
TT		 3359 /* Now we set up the journal barrier. */
3360 jbd2_journal_lock_updates(journal);
ac27a0ec 3361
9ca92389
TT		 3362 /*
3363 * Don't clear the needs_recovery flag if we failed to flush
3364 * the journal.
3365 */
3366 error = jbd2_journal_flush(journal);
3367 if (error < 0) {
3368 out:
3369 jbd2_journal_unlock_updates(journal);
3370 return error;
ac27a0ec 3371 }
9ca92389
TT		 3372
3373 /* Journal blocked and flushed, clear needs_recovery flag. */
3374 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3375 error = ext4_commit_super(sb, 1);
3376 if (error)
3377 goto out;
c4be0c1d 3378 return 0;
ac27a0ec
DK		 3379}
3380
3381/*
3382 * Called by LVM after the snapshot is done. We need to reset the RECOVER
3383 * flag here, even though the filesystem is not technically dirty yet.
3384 */
c4be0c1d 3385static int ext4_unfreeze(struct super_block *sb)
ac27a0ec 3386{
9ca92389
TT		 3387 if (sb->s_flags & MS_RDONLY)
3388 return 0;
3389
3390 lock_super(sb);
3391 /* Reset the needs_recovery flag before the fs is unlocked. */
3392 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3393 ext4_commit_super(sb, 1);
3394 unlock_super(sb);
3395 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
c4be0c1d 3396 return 0;
ac27a0ec
DK		 3397}
3398
2b2d6d01 3399static int ext4_remount(struct super_block *sb, int *flags, char *data)
ac27a0ec 3400{
2b2d6d01 3401 struct ext4_super_block *es;
617ba13b
MC		 3402 struct ext4_sb_info *sbi = EXT4_SB(sb);
3403 ext4_fsblk_t n_blocks_count = 0;
ac27a0ec 3404 unsigned long old_sb_flags;
617ba13b 3405 struct ext4_mount_options old_opts;
8a266467 3406 ext4_group_t g;
b3881f74 3407 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
ac27a0ec
DK		 3408 int err;
3409#ifdef CONFIG_QUOTA
3410 int i;
3411#endif
3412
337eb00a
AIB		 3413 lock_kernel();
3414
ac27a0ec 3415 /* Store the original options */
bbd6851a 3416 lock_super(sb);
ac27a0ec
DK		 3417 old_sb_flags = sb->s_flags;
3418 old_opts.s_mount_opt = sbi->s_mount_opt;
3419 old_opts.s_resuid = sbi->s_resuid;
3420 old_opts.s_resgid = sbi->s_resgid;
3421 old_opts.s_commit_interval = sbi->s_commit_interval;
30773840
TT		 3422 old_opts.s_min_batch_time = sbi->s_min_batch_time;
3423 old_opts.s_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 3424#ifdef CONFIG_QUOTA
3425 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
3426 for (i = 0; i < MAXQUOTAS; i++)
3427 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
3428#endif
b3881f74
TT		 3429 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
3430 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
ac27a0ec
DK		 3431
3432 /*
3433 * Allow the "check" option to be passed as a remount option.
3434 */
b3881f74
TT		 3435 if (!parse_options(data, sb, NULL, &journal_ioprio,
3436 &n_blocks_count, 1)) {
ac27a0ec
DK		 3437 err = -EINVAL;
3438 goto restore_opts;
3439 }
3440
4ab2f15b 3441 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
46e665e9 3442 ext4_abort(sb, __func__, "Abort forced by user");
ac27a0ec
DK		 3443
3444 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
617ba13b 3445 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec
DK		 3446
3447 es = sbi->s_es;
3448
b3881f74 3449 if (sbi->s_journal) {
0390131b 3450 ext4_init_journal_params(sb, sbi->s_journal);
b3881f74
TT		 3451 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3452 }
ac27a0ec
DK		 3453
3454 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
bd81d8ee 3455 n_blocks_count > ext4_blocks_count(es)) {
4ab2f15b 3456 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
ac27a0ec
DK		 3457 err = -EROFS;
3458 goto restore_opts;
3459 }
3460
3461 if (*flags & MS_RDONLY) {
3462 /*
3463 * First of all, the unconditional stuff we have to do
3464 * to disable replay of the journal when we next remount
3465 */
3466 sb->s_flags |= MS_RDONLY;
3467
3468 /*
3469 * OK, test if we are remounting a valid rw partition
3470 * readonly, and if so set the rdonly flag and then
3471 * mark the partition as valid again.
3472 */
617ba13b
MC		 3473 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
3474 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 3475 es->s_state = cpu_to_le16(sbi->s_mount_state);
3476
a63c9eb2 3477 if (sbi->s_journal)
0390131b 3478 ext4_mark_recovery_complete(sb, es);
ac27a0ec 3479 } else {
3a06d778 3480 int ret;
617ba13b
MC		 3481 if ((ret = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3482 ~EXT4_FEATURE_RO_COMPAT_SUPP))) {
b31e1552 3483 ext4_msg(sb, KERN_WARNING, "couldn't "
ac27a0ec 3484 "remount RDWR because of unsupported "
b31e1552 3485 "optional features (%x)",
3a06d778
AK		 3486 (le32_to_cpu(sbi->s_es->s_feature_ro_compat) &
3487 ~EXT4_FEATURE_RO_COMPAT_SUPP));
ac27a0ec
DK		 3488 err = -EROFS;
3489 goto restore_opts;
3490 }
ead6596b 3491
8a266467
TT		 3492 /*
3493 * Make sure the group descriptor checksums
0b8e58a1 3494 * are sane. If they aren't, refuse to remount r/w.
8a266467
TT		 3495 */
3496 for (g = 0; g < sbi->s_groups_count; g++) {
3497 struct ext4_group_desc *gdp =
3498 ext4_get_group_desc(sb, g, NULL);
3499
3500 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
b31e1552
ES		 3501 ext4_msg(sb, KERN_ERR,
3502 "ext4_remount: Checksum for group %u failed (%u!=%u)",
8a266467
TT		 3503 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
3504 le16_to_cpu(gdp->bg_checksum));
3505 err = -EINVAL;
3506 goto restore_opts;
3507 }
3508 }
3509
ead6596b
ES		 3510 /*
3511 * If we have an unprocessed orphan list hanging
3512 * around from a previously readonly bdev mount,
3513 * require a full umount/remount for now.
3514 */
3515 if (es->s_last_orphan) {
b31e1552 3516 ext4_msg(sb, KERN_WARNING, "Couldn't "
ead6596b
ES		 3517 "remount RDWR because of unprocessed "
3518 "orphan inode list. Please "
b31e1552 3519 "umount/remount instead");
ead6596b
ES		 3520 err = -EINVAL;
3521 goto restore_opts;
3522 }
3523
ac27a0ec
DK		 3524 /*
3525 * Mounting a RDONLY partition read-write, so reread
3526 * and store the current valid flag. (It may have
3527 * been changed by e2fsck since we originally mounted
3528 * the partition.)
3529 */
0390131b
FM		 3530 if (sbi->s_journal)
3531 ext4_clear_journal_err(sb, es);
ac27a0ec 3532 sbi->s_mount_state = le16_to_cpu(es->s_state);
617ba13b 3533 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
ac27a0ec 3534 goto restore_opts;
2b2d6d01 3535 if (!ext4_setup_super(sb, es, 0))
ac27a0ec
DK		 3536 sb->s_flags &= ~MS_RDONLY;
3537 }
3538 }
6fd058f7 3539 ext4_setup_system_zone(sb);
0390131b 3540 if (sbi->s_journal == NULL)
e2d67052 3541 ext4_commit_super(sb, 1);
0390131b 3542
ac27a0ec
DK		 3543#ifdef CONFIG_QUOTA
3544 /* Release old quota file names */
3545 for (i = 0; i < MAXQUOTAS; i++)
3546 if (old_opts.s_qf_names[i] &&
3547 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3548 kfree(old_opts.s_qf_names[i]);
3549#endif
bbd6851a 3550 unlock_super(sb);
337eb00a 3551 unlock_kernel();
ac27a0ec 3552 return 0;
0b8e58a1 3553
ac27a0ec
DK		 3554restore_opts:
3555 sb->s_flags = old_sb_flags;
3556 sbi->s_mount_opt = old_opts.s_mount_opt;
3557 sbi->s_resuid = old_opts.s_resuid;
3558 sbi->s_resgid = old_opts.s_resgid;
3559 sbi->s_commit_interval = old_opts.s_commit_interval;
30773840
TT		 3560 sbi->s_min_batch_time = old_opts.s_min_batch_time;
3561 sbi->s_max_batch_time = old_opts.s_max_batch_time;
ac27a0ec
DK		 3562#ifdef CONFIG_QUOTA
3563 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
3564 for (i = 0; i < MAXQUOTAS; i++) {
3565 if (sbi->s_qf_names[i] &&
3566 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3567 kfree(sbi->s_qf_names[i]);
3568 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
3569 }
3570#endif
bbd6851a 3571 unlock_super(sb);
337eb00a 3572 unlock_kernel();
ac27a0ec
DK		 3573 return err;
3574}
3575
2b2d6d01 3576static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
ac27a0ec
DK		 3577{
3578 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 3579 struct ext4_sb_info *sbi = EXT4_SB(sb);
3580 struct ext4_super_block *es = sbi->s_es;
960cc398 3581 u64 fsid;
ac27a0ec 3582
5e70030d
BP		 3583 if (test_opt(sb, MINIX_DF)) {
3584 sbi->s_overhead_last = 0;
6bc9feff 3585 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
8df9675f 3586 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
5e70030d 3587 ext4_fsblk_t overhead = 0;
ac27a0ec
DK		 3588
3589 /*
5e70030d
BP		 3590 * Compute the overhead (FS structures). This is constant
3591 * for a given filesystem unless the number of block groups
3592 * changes so we cache the previous value until it does.
ac27a0ec
DK		 3593 */
3594
3595 /*
3596 * All of the blocks before first_data_block are
3597 * overhead
3598 */
3599 overhead = le32_to_cpu(es->s_first_data_block);
3600
3601 /*
3602 * Add the overhead attributed to the superblock and
3603 * block group descriptors. If the sparse superblocks
3604 * feature is turned on, then not all groups have this.
3605 */
3606 for (i = 0; i < ngroups; i++) {
617ba13b
MC		 3607 overhead += ext4_bg_has_super(sb, i) +
3608 ext4_bg_num_gdb(sb, i);
ac27a0ec
DK		 3609 cond_resched();
3610 }
3611
3612 /*
3613 * Every block group has an inode bitmap, a block
3614 * bitmap, and an inode table.
3615 */
5e70030d
BP		 3616 overhead += ngroups * (2 + sbi->s_itb_per_group);
3617 sbi->s_overhead_last = overhead;
3618 smp_wmb();
6bc9feff 3619 sbi->s_blocks_last = ext4_blocks_count(es);
ac27a0ec
DK		 3620 }
3621
617ba13b 3622 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 3623 buf->f_bsize = sb->s_blocksize;
5e70030d 3624 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
6bc6e63f
AK		 3625 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
3626 percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
308ba3ec 3627 ext4_free_blocks_count_set(es, buf->f_bfree);
bd81d8ee
LV		 3628 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3629 if (buf->f_bfree < ext4_r_blocks_count(es))
ac27a0ec
DK		 3630 buf->f_bavail = 0;
3631 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 3632 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
5e70030d 3633 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
617ba13b 3634 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 3635 fsid = le64_to_cpup((void *)es->s_uuid) ^
3636 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3637 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3638 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
0b8e58a1 3639
ac27a0ec
DK		 3640 return 0;
3641}
3642
0b8e58a1
AD		 3643/* Helper function for writing quotas on sync - we need to start transaction
3644 * before quota file is locked for write. Otherwise the are possible deadlocks:
ac27a0ec 3645 * Process 1 Process 2
617ba13b 3646 * ext4_create() quota_sync()
a269eb18
JK		 3647 * jbd2_journal_start() write_dquot()
3648 * vfs_dq_init() down(dqio_mutex)
dab291af 3649 * down(dqio_mutex) jbd2_journal_start()
ac27a0ec
DK		 3650 *
3651 */
3652
3653#ifdef CONFIG_QUOTA
3654
3655static inline struct inode *dquot_to_inode(struct dquot *dquot)
3656{
3657 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3658}
3659
617ba13b 3660static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 3661{
3662 int ret, err;
3663 handle_t *handle;
3664 struct inode *inode;
3665
3666 inode = dquot_to_inode(dquot);
617ba13b 3667 handle = ext4_journal_start(inode,
0b8e58a1 3668 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 3669 if (IS_ERR(handle))
3670 return PTR_ERR(handle);
3671 ret = dquot_commit(dquot);
617ba13b 3672 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3673 if (!ret)
3674 ret = err;
3675 return ret;
3676}
3677
617ba13b 3678static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 3679{
3680 int ret, err;
3681 handle_t *handle;
3682
617ba13b 3683 handle = ext4_journal_start(dquot_to_inode(dquot),
0b8e58a1 3684 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 3685 if (IS_ERR(handle))
3686 return PTR_ERR(handle);
3687 ret = dquot_acquire(dquot);
617ba13b 3688 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3689 if (!ret)
3690 ret = err;
3691 return ret;
3692}
3693
617ba13b 3694static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 3695{
3696 int ret, err;
3697 handle_t *handle;
3698
617ba13b 3699 handle = ext4_journal_start(dquot_to_inode(dquot),
0b8e58a1 3700 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 3701 if (IS_ERR(handle)) {
3702 /* Release dquot anyway to avoid endless cycle in dqput() */
3703 dquot_release(dquot);
ac27a0ec 3704 return PTR_ERR(handle);
9c3013e9 3705 }
ac27a0ec 3706 ret = dquot_release(dquot);
617ba13b 3707 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3708 if (!ret)
3709 ret = err;
3710 return ret;
3711}
3712
617ba13b 3713static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec 3714{
2c8be6b2 3715 /* Are we journaling quotas? */
617ba13b
MC		 3716 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3717 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
ac27a0ec 3718 dquot_mark_dquot_dirty(dquot);
617ba13b 3719 return ext4_write_dquot(dquot);
ac27a0ec
DK		 3720 } else {
3721 return dquot_mark_dquot_dirty(dquot);
3722 }
3723}
3724
617ba13b 3725static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 3726{
3727 int ret, err;
3728 handle_t *handle;
3729
3730 /* Data block + inode block */
617ba13b 3731 handle = ext4_journal_start(sb->s_root->d_inode, 2);
ac27a0ec
DK		 3732 if (IS_ERR(handle))
3733 return PTR_ERR(handle);
3734 ret = dquot_commit_info(sb, type);
617ba13b 3735 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3736 if (!ret)
3737 ret = err;
3738 return ret;
3739}
3740
3741/*
3742 * Turn on quotas during mount time - we need to find
3743 * the quota file and such...
3744 */
617ba13b 3745static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 3746{
617ba13b 3747 return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
0b8e58a1 3748 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 3749}
3750
3751/*
3752 * Standard function to be called on quota_on
3753 */
617ba13b 3754static int ext4_quota_on(struct super_block *sb, int type, int format_id,
8264613d 3755 char *name, int remount)
ac27a0ec
DK		 3756{
3757 int err;
8264613d 3758 struct path path;
ac27a0ec
DK		 3759
3760 if (!test_opt(sb, QUOTA))
3761 return -EINVAL;
8264613d 3762 /* When remounting, no checks are needed and in fact, name is NULL */
0623543b 3763 if (remount)
8264613d 3764 return vfs_quota_on(sb, type, format_id, name, remount);
0623543b 3765
8264613d 3766 err = kern_path(name, LOOKUP_FOLLOW, &path);
ac27a0ec
DK		 3767 if (err)
3768 return err;
0623543b 3769
ac27a0ec 3770 /* Quotafile not on the same filesystem? */
8264613d
AV		 3771 if (path.mnt->mnt_sb != sb) {
3772 path_put(&path);
ac27a0ec
DK		 3773 return -EXDEV;
3774 }
0623543b
JK		 3775 /* Journaling quota? */
3776 if (EXT4_SB(sb)->s_qf_names[type]) {
2b2d6d01 3777 /* Quotafile not in fs root? */
8264613d 3778 if (path.dentry->d_parent != sb->s_root)
b31e1552
ES		 3779 ext4_msg(sb, KERN_WARNING,
3780 "Quota file not on filesystem root. "
3781 "Journaled quota will not work");
2b2d6d01 3782 }
0623543b
JK		 3783
3784 /*
3785 * When we journal data on quota file, we have to flush journal to see
3786 * all updates to the file when we bypass pagecache...
3787 */
0390131b
FM		 3788 if (EXT4_SB(sb)->s_journal &&
3789 ext4_should_journal_data(path.dentry->d_inode)) {
0623543b
JK		 3790 /*
3791 * We don't need to lock updates but journal_flush() could
3792 * otherwise be livelocked...
3793 */
3794 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 3795 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
0623543b 3796 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 3797 if (err) {
8264613d 3798 path_put(&path);
7ffe1ea8
HK		 3799 return err;
3800 }
0623543b
JK		 3801 }
3802
8264613d
AV		 3803 err = vfs_quota_on_path(sb, type, format_id, &path);
3804 path_put(&path);
77e69dac 3805 return err;
ac27a0ec
DK		 3806}
3807
3808/* Read data from quotafile - avoid pagecache and such because we cannot afford
3809 * acquiring the locks... As quota files are never truncated and quota code
3810 * itself serializes the operations (and noone else should touch the files)
3811 * we don't have to be afraid of races */
617ba13b 3812static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 3813 size_t len, loff_t off)
3814{
3815 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 3816 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 3817 int err = 0;
3818 int offset = off & (sb->s_blocksize - 1);
3819 int tocopy;
3820 size_t toread;
3821 struct buffer_head *bh;
3822 loff_t i_size = i_size_read(inode);
3823
3824 if (off > i_size)
3825 return 0;
3826 if (off+len > i_size)
3827 len = i_size-off;
3828 toread = len;
3829 while (toread > 0) {
3830 tocopy = sb->s_blocksize - offset < toread ?
3831 sb->s_blocksize - offset : toread;
617ba13b 3832 bh = ext4_bread(NULL, inode, blk, 0, &err);
ac27a0ec
DK		 3833 if (err)
3834 return err;
3835 if (!bh) /* A hole? */
3836 memset(data, 0, tocopy);
3837 else
3838 memcpy(data, bh->b_data+offset, tocopy);
3839 brelse(bh);
3840 offset = 0;
3841 toread -= tocopy;
3842 data += tocopy;
3843 blk++;
3844 }
3845 return len;
3846}
3847
3848/* Write to quotafile (we know the transaction is already started and has
3849 * enough credits) */
617ba13b 3850static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 3851 const char *data, size_t len, loff_t off)
3852{
3853 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 3854 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 3855 int err = 0;
3856 int offset = off & (sb->s_blocksize - 1);
3857 int tocopy;
617ba13b 3858 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
ac27a0ec
DK		 3859 size_t towrite = len;
3860 struct buffer_head *bh;
3861 handle_t *handle = journal_current_handle();
3862
0390131b 3863 if (EXT4_SB(sb)->s_journal && !handle) {
b31e1552
ES		 3864 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
3865 " cancelled because transaction is not started",
9c3013e9
JK		 3866 (unsigned long long)off, (unsigned long long)len);
3867 return -EIO;
3868 }
ac27a0ec
DK		 3869 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
3870 while (towrite > 0) {
3871 tocopy = sb->s_blocksize - offset < towrite ?
3872 sb->s_blocksize - offset : towrite;
617ba13b 3873 bh = ext4_bread(handle, inode, blk, 1, &err);
ac27a0ec
DK		 3874 if (!bh)
3875 goto out;
3876 if (journal_quota) {
617ba13b 3877 err = ext4_journal_get_write_access(handle, bh);
ac27a0ec
DK		 3878 if (err) {
3879 brelse(bh);
3880 goto out;
3881 }
3882 }
3883 lock_buffer(bh);
3884 memcpy(bh->b_data+offset, data, tocopy);
3885 flush_dcache_page(bh->b_page);
3886 unlock_buffer(bh);
3887 if (journal_quota)
0390131b 3888 err = ext4_handle_dirty_metadata(handle, NULL, bh);
ac27a0ec
DK		 3889 else {
3890 /* Always do at least ordered writes for quotas */
678aaf48 3891 err = ext4_jbd2_file_inode(handle, inode);
ac27a0ec
DK		 3892 mark_buffer_dirty(bh);
3893 }
3894 brelse(bh);
3895 if (err)
3896 goto out;
3897 offset = 0;
3898 towrite -= tocopy;
3899 data += tocopy;
3900 blk++;
3901 }
3902out:
4d04e4fb
JK		 3903 if (len == towrite) {
3904 mutex_unlock(&inode->i_mutex);
ac27a0ec 3905 return err;
4d04e4fb 3906 }
ac27a0ec
DK		 3907 if (inode->i_size < off+len-towrite) {
3908 i_size_write(inode, off+len-towrite);
617ba13b 3909 EXT4_I(inode)->i_disksize = inode->i_size;
ac27a0ec 3910 }
ac27a0ec 3911 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
617ba13b 3912 ext4_mark_inode_dirty(handle, inode);
ac27a0ec
DK		 3913 mutex_unlock(&inode->i_mutex);
3914 return len - towrite;
3915}
3916
3917#endif
3918
0b8e58a1
AD		 3919static int ext4_get_sb(struct file_system_type *fs_type, int flags,
3920 const char *dev_name, void *data, struct vfsmount *mnt)
ac27a0ec 3921{
0b8e58a1 3922 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super,mnt);
ac27a0ec
DK		 3923}
3924
03010a33
TT		 3925static struct file_system_type ext4_fs_type = {
3926 .owner = THIS_MODULE,
3927 .name = "ext4",
3928 .get_sb = ext4_get_sb,
3929 .kill_sb = kill_block_super,
3930 .fs_flags = FS_REQUIRES_DEV,
3931};
3932
3933#ifdef CONFIG_EXT4DEV_COMPAT
0b8e58a1
AD		 3934static int ext4dev_get_sb(struct file_system_type *fs_type, int flags,
3935 const char *dev_name, void *data,struct vfsmount *mnt)
03010a33 3936{
b31e1552
ES		 3937 printk(KERN_WARNING "EXT4-fs (%s): Update your userspace programs "
3938 "to mount using ext4\n", dev_name);
3939 printk(KERN_WARNING "EXT4-fs (%s): ext4dev backwards compatibility "
3940 "will go away by 2.6.31\n", dev_name);
0b8e58a1 3941 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super,mnt);
03010a33
TT		 3942}
3943
617ba13b 3944static struct file_system_type ext4dev_fs_type = {
ac27a0ec 3945 .owner = THIS_MODULE,
617ba13b 3946 .name = "ext4dev",
03010a33 3947 .get_sb = ext4dev_get_sb,
ac27a0ec
DK		 3948 .kill_sb = kill_block_super,
3949 .fs_flags = FS_REQUIRES_DEV,
3950};
03010a33
TT		 3951MODULE_ALIAS("ext4dev");
3952#endif
ac27a0ec 3953
617ba13b 3954static int __init init_ext4_fs(void)
ac27a0ec 3955{
c9de560d
AT		 3956 int err;
3957
6fd058f7
TT		 3958 err = init_ext4_system_zone();
3959 if (err)
3960 return err;
3197ebdb
TT		 3961 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
3962 if (!ext4_kset)
6fd058f7 3963 goto out4;
9f6200bb 3964 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
c9de560d 3965 err = init_ext4_mballoc();
ac27a0ec 3966 if (err)
6fd058f7 3967 goto out3;
c9de560d
AT		 3968
3969 err = init_ext4_xattr();
3970 if (err)
3971 goto out2;
ac27a0ec
DK		 3972 err = init_inodecache();
3973 if (err)
3974 goto out1;
03010a33 3975 err = register_filesystem(&ext4_fs_type);
ac27a0ec
DK		 3976 if (err)
3977 goto out;
03010a33
TT		 3978#ifdef CONFIG_EXT4DEV_COMPAT
3979 err = register_filesystem(&ext4dev_fs_type);
3980 if (err) {
3981 unregister_filesystem(&ext4_fs_type);
3982 goto out;
3983 }
3984#endif
ac27a0ec
DK		 3985 return 0;
3986out:
3987 destroy_inodecache();
3988out1:
617ba13b 3989 exit_ext4_xattr();
c9de560d
AT		 3990out2:
3991 exit_ext4_mballoc();
6fd058f7
TT		 3992out3:
3993 remove_proc_entry("fs/ext4", NULL);
3994 kset_unregister(ext4_kset);
3995out4:
3996 exit_ext4_system_zone();
ac27a0ec
DK		 3997 return err;
3998}
3999
617ba13b 4000static void __exit exit_ext4_fs(void)
ac27a0ec 4001{
03010a33
TT		 4002 unregister_filesystem(&ext4_fs_type);
4003#ifdef CONFIG_EXT4DEV_COMPAT
617ba13b 4004 unregister_filesystem(&ext4dev_fs_type);
03010a33 4005#endif
ac27a0ec 4006 destroy_inodecache();
617ba13b 4007 exit_ext4_xattr();
c9de560d 4008 exit_ext4_mballoc();
9f6200bb 4009 remove_proc_entry("fs/ext4", NULL);
3197ebdb 4010 kset_unregister(ext4_kset);
6fd058f7 4011 exit_ext4_system_zone();
ac27a0ec
DK		 4012}
4013
4014MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
83982b6f 4015MODULE_DESCRIPTION("Fourth Extended Filesystem");
ac27a0ec 4016MODULE_LICENSE("GPL");
617ba13b
MC		 4017module_init(init_ext4_fs)
4018module_exit(exit_ext4_fs)
A mirror of Linus' kernel repository