projects / GitHub / LineageOS / android_kernel_motorola_exynos9610.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
ext4: Fix different block exchange issue in EXT4_IOC_MOVE_EXT
[GitHub/LineageOS/android_kernel_motorola_exynos9610.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"
3661d286 48#include "mballoc.h"
ac27a0ec 49
9bffad1e
TT		 50#define CREATE_TRACE_POINTS
51#include <trace/events/ext4.h>
52
f4033903
CW		 53static int default_mb_history_length = 1000;
54
55module_param_named(default_mb_history_length, default_mb_history_length,
56 int, 0644);
57MODULE_PARM_DESC(default_mb_history_length,
58 "Default number of entries saved for mb_history");
59
9f6200bb 60struct proc_dir_entry *ext4_proc_root;
3197ebdb 61static struct kset *ext4_kset;
9f6200bb 62
617ba13b 63static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
ac27a0ec 64 unsigned long journal_devnum);
e2d67052 65static int ext4_commit_super(struct super_block *sb, int sync);
2b2d6d01
TT		 66static void ext4_mark_recovery_complete(struct super_block *sb,
67 struct ext4_super_block *es);
68static void ext4_clear_journal_err(struct super_block *sb,
69 struct ext4_super_block *es);
617ba13b 70static int ext4_sync_fs(struct super_block *sb, int wait);
2b2d6d01 71static const char *ext4_decode_error(struct super_block *sb, int errno,
ac27a0ec 72 char nbuf[16]);
2b2d6d01
TT		 73static int ext4_remount(struct super_block *sb, int *flags, char *data);
74static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
c4be0c1d 75static int ext4_unfreeze(struct super_block *sb);
2b2d6d01 76static void ext4_write_super(struct super_block *sb);
c4be0c1d 77static int ext4_freeze(struct super_block *sb);
ac27a0ec 78
bd81d8ee 79
8fadc143
AR		 80ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
81 struct ext4_group_desc *bg)
bd81d8ee 82{
3a14589c 83 return le32_to_cpu(bg->bg_block_bitmap_lo) |
8fadc143 84 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 85 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 86}
87
8fadc143
AR		 88ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
89 struct ext4_group_desc *bg)
bd81d8ee 90{
5272f837 91 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
8fadc143 92 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 93 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 94}
95
8fadc143
AR		 96ext4_fsblk_t ext4_inode_table(struct super_block *sb,
97 struct ext4_group_desc *bg)
bd81d8ee 98{
5272f837 99 return le32_to_cpu(bg->bg_inode_table_lo) |
8fadc143 100 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 101 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
bd81d8ee
LV		 102}
103
560671a0
AK		 104__u32 ext4_free_blks_count(struct super_block *sb,
105 struct ext4_group_desc *bg)
106{
107 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
108 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 109 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
560671a0
AK		 110}
111
112__u32 ext4_free_inodes_count(struct super_block *sb,
113 struct ext4_group_desc *bg)
114{
115 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
116 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 117 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
560671a0
AK		 118}
119
120__u32 ext4_used_dirs_count(struct super_block *sb,
121 struct ext4_group_desc *bg)
122{
123 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
124 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 125 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
560671a0
AK		 126}
127
128__u32 ext4_itable_unused_count(struct super_block *sb,
129 struct ext4_group_desc *bg)
130{
131 return le16_to_cpu(bg->bg_itable_unused_lo) |
132 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 133 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
560671a0
AK		 134}
135
8fadc143
AR		 136void ext4_block_bitmap_set(struct super_block *sb,
137 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 138{
3a14589c 139 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 140 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
141 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 142}
143
8fadc143
AR		 144void ext4_inode_bitmap_set(struct super_block *sb,
145 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 146{
5272f837 147 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 148 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
149 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 150}
151
8fadc143
AR		 152void ext4_inode_table_set(struct super_block *sb,
153 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 154{
5272f837 155 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 156 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
157 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 158}
159
560671a0
AK		 160void ext4_free_blks_set(struct super_block *sb,
161 struct ext4_group_desc *bg, __u32 count)
162{
163 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
164 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
165 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
166}
167
168void ext4_free_inodes_set(struct super_block *sb,
169 struct ext4_group_desc *bg, __u32 count)
170{
171 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
172 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
173 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
174}
175
176void ext4_used_dirs_set(struct super_block *sb,
177 struct ext4_group_desc *bg, __u32 count)
178{
179 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
180 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
181 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
182}
183
184void ext4_itable_unused_set(struct super_block *sb,
185 struct ext4_group_desc *bg, __u32 count)
186{
187 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
188 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
189 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
190}
191
ac27a0ec 192/*
dab291af 193 * Wrappers for jbd2_journal_start/end.
ac27a0ec
DK		 194 *
195 * The only special thing we need to do here is to make sure that all
196 * journal_end calls result in the superblock being marked dirty, so
197 * that sync() will call the filesystem's write_super callback if
198 * appropriate.
199 */
617ba13b 200handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
ac27a0ec
DK		 201{
202 journal_t *journal;
203
204 if (sb->s_flags & MS_RDONLY)
205 return ERR_PTR(-EROFS);
206
207 /* Special case here: if the journal has aborted behind our
208 * backs (eg. EIO in the commit thread), then we still need to
209 * take the FS itself readonly cleanly. */
617ba13b 210 journal = EXT4_SB(sb)->s_journal;
0390131b
FM		 211 if (journal) {
212 if (is_journal_aborted(journal)) {
0b8e58a1 213 ext4_abort(sb, __func__, "Detected aborted journal");
0390131b
FM		 214 return ERR_PTR(-EROFS);
215 }
216 return jbd2_journal_start(journal, nblocks);
ac27a0ec 217 }
0390131b
FM		 218 /*
219 * We're not journaling, return the appropriate indication.
220 */
221 current->journal_info = EXT4_NOJOURNAL_HANDLE;
222 return current->journal_info;
ac27a0ec
DK		 223}
224
225/*
226 * The only special thing we need to do here is to make sure that all
dab291af 227 * jbd2_journal_stop calls result in the superblock being marked dirty, so
ac27a0ec
DK		 228 * that sync() will call the filesystem's write_super callback if
229 * appropriate.
230 */
617ba13b 231int __ext4_journal_stop(const char *where, handle_t *handle)
ac27a0ec
DK		 232{
233 struct super_block *sb;
234 int err;
235 int rc;
236
0390131b
FM		 237 if (!ext4_handle_valid(handle)) {
238 /*
239 * Do this here since we don't call jbd2_journal_stop() in
240 * no-journal mode.
241 */
242 current->journal_info = NULL;
243 return 0;
244 }
ac27a0ec
DK		 245 sb = handle->h_transaction->t_journal->j_private;
246 err = handle->h_err;
dab291af 247 rc = jbd2_journal_stop(handle);
ac27a0ec
DK		 248
249 if (!err)
250 err = rc;
251 if (err)
617ba13b 252 __ext4_std_error(sb, where, err);
ac27a0ec
DK		 253 return err;
254}
255
617ba13b 256void ext4_journal_abort_handle(const char *caller, const char *err_fn,
ac27a0ec
DK		 257 struct buffer_head *bh, handle_t *handle, int err)
258{
259 char nbuf[16];
617ba13b 260 const char *errstr = ext4_decode_error(NULL, err, nbuf);
ac27a0ec 261
0390131b
FM		 262 BUG_ON(!ext4_handle_valid(handle));
263
ac27a0ec
DK		 264 if (bh)
265 BUFFER_TRACE(bh, "abort");
266
267 if (!handle->h_err)
268 handle->h_err = err;
269
270 if (is_handle_aborted(handle))
271 return;
272
273 printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
274 caller, errstr, err_fn);
275
dab291af 276 jbd2_journal_abort_handle(handle);
ac27a0ec
DK		 277}
278
279/* Deal with the reporting of failure conditions on a filesystem such as
280 * inconsistencies detected or read IO failures.
281 *
282 * On ext2, we can store the error state of the filesystem in the
617ba13b 283 * superblock. That is not possible on ext4, because we may have other
ac27a0ec
DK		 284 * write ordering constraints on the superblock which prevent us from
285 * writing it out straight away; and given that the journal is about to
286 * be aborted, we can't rely on the current, or future, transactions to
287 * write out the superblock safely.
288 *
dab291af 289 * We'll just use the jbd2_journal_abort() error code to record an error in
ac27a0ec
DK		 290 * the journal instead. On recovery, the journal will compain about
291 * that error until we've noted it down and cleared it.
292 */
293
617ba13b 294static void ext4_handle_error(struct super_block *sb)
ac27a0ec 295{
617ba13b 296 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 297
617ba13b
MC		 298 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
299 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
ac27a0ec
DK		 300
301 if (sb->s_flags & MS_RDONLY)
302 return;
303
2b2d6d01 304 if (!test_opt(sb, ERRORS_CONT)) {
617ba13b 305 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 306
4ab2f15b 307 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
ac27a0ec 308 if (journal)
dab291af 309 jbd2_journal_abort(journal, -EIO);
ac27a0ec 310 }
2b2d6d01 311 if (test_opt(sb, ERRORS_RO)) {
b31e1552 312 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
ac27a0ec
DK		 313 sb->s_flags |= MS_RDONLY;
314 }
e2d67052 315 ext4_commit_super(sb, 1);
ac27a0ec 316 if (test_opt(sb, ERRORS_PANIC))
617ba13b 317 panic("EXT4-fs (device %s): panic forced after error\n",
ac27a0ec
DK		 318 sb->s_id);
319}
320
2b2d6d01
TT		 321void ext4_error(struct super_block *sb, const char *function,
322 const char *fmt, ...)
ac27a0ec
DK		 323{
324 va_list args;
325
326 va_start(args, fmt);
2b2d6d01 327 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
ac27a0ec
DK		 328 vprintk(fmt, args);
329 printk("\n");
330 va_end(args);
331
617ba13b 332 ext4_handle_error(sb);
ac27a0ec
DK		 333}
334
2b2d6d01 335static const char *ext4_decode_error(struct super_block *sb, int errno,
ac27a0ec
DK		 336 char nbuf[16])
337{
338 char *errstr = NULL;
339
340 switch (errno) {
341 case -EIO:
342 errstr = "IO failure";
343 break;
344 case -ENOMEM:
345 errstr = "Out of memory";
346 break;
347 case -EROFS:
78f1ddbb
TT		 348 if (!sb || (EXT4_SB(sb)->s_journal &&
349 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
ac27a0ec
DK		 350 errstr = "Journal has aborted";
351 else
352 errstr = "Readonly filesystem";
353 break;
354 default:
355 /* If the caller passed in an extra buffer for unknown
356 * errors, textualise them now. Else we just return
357 * NULL. */
358 if (nbuf) {
359 /* Check for truncated error codes... */
360 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
361 errstr = nbuf;
362 }
363 break;
364 }
365
366 return errstr;
367}
368
617ba13b 369/* __ext4_std_error decodes expected errors from journaling functions
ac27a0ec
DK		 370 * automatically and invokes the appropriate error response. */
371
2b2d6d01 372void __ext4_std_error(struct super_block *sb, const char *function, int errno)
ac27a0ec
DK		 373{
374 char nbuf[16];
375 const char *errstr;
376
377 /* Special case: if the error is EROFS, and we're not already
378 * inside a transaction, then there's really no point in logging
379 * an error. */
380 if (errno == -EROFS && journal_current_handle() == NULL &&
381 (sb->s_flags & MS_RDONLY))
382 return;
383
617ba13b 384 errstr = ext4_decode_error(sb, errno, nbuf);
2b2d6d01
TT		 385 printk(KERN_CRIT "EXT4-fs error (device %s) in %s: %s\n",
386 sb->s_id, function, errstr);
ac27a0ec 387
617ba13b 388 ext4_handle_error(sb);
ac27a0ec
DK		 389}
390
391/*
617ba13b 392 * ext4_abort is a much stronger failure handler than ext4_error. The
ac27a0ec
DK		 393 * abort function may be used to deal with unrecoverable failures such
394 * as journal IO errors or ENOMEM at a critical moment in log management.
395 *
396 * We unconditionally force the filesystem into an ABORT|READONLY state,
397 * unless the error response on the fs has been set to panic in which
398 * case we take the easy way out and panic immediately.
399 */
400
2b2d6d01
TT		 401void ext4_abort(struct super_block *sb, const char *function,
402 const char *fmt, ...)
ac27a0ec
DK		 403{
404 va_list args;
405
ac27a0ec 406 va_start(args, fmt);
2b2d6d01 407 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
ac27a0ec
DK		 408 vprintk(fmt, args);
409 printk("\n");
410 va_end(args);
411
412 if (test_opt(sb, ERRORS_PANIC))
617ba13b 413 panic("EXT4-fs panic from previous error\n");
ac27a0ec
DK		 414
415 if (sb->s_flags & MS_RDONLY)
416 return;
417
b31e1552 418 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
617ba13b 419 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
ac27a0ec 420 sb->s_flags |= MS_RDONLY;
4ab2f15b 421 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
ef2cabf7
HK		 422 if (EXT4_SB(sb)->s_journal)
423 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
ac27a0ec
DK		 424}
425
b31e1552
ES		 426void ext4_msg (struct super_block * sb, const char *prefix,
427 const char *fmt, ...)
428{
429 va_list args;
430
431 va_start(args, fmt);
432 printk("%sEXT4-fs (%s): ", prefix, sb->s_id);
433 vprintk(fmt, args);
434 printk("\n");
435 va_end(args);
436}
437
2b2d6d01
TT		 438void ext4_warning(struct super_block *sb, const char *function,
439 const char *fmt, ...)
ac27a0ec
DK		 440{
441 va_list args;
442
443 va_start(args, fmt);
617ba13b 444 printk(KERN_WARNING "EXT4-fs warning (device %s): %s: ",
ac27a0ec
DK		 445 sb->s_id, function);
446 vprintk(fmt, args);
447 printk("\n");
448 va_end(args);
449}
450
5d1b1b3f 451void ext4_grp_locked_error(struct super_block *sb, ext4_group_t grp,
0b8e58a1 452 const char *function, const char *fmt, ...)
5d1b1b3f
AK		 453__releases(bitlock)
454__acquires(bitlock)
455{
456 va_list args;
457 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
458
459 va_start(args, fmt);
460 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
461 vprintk(fmt, args);
462 printk("\n");
463 va_end(args);
464
465 if (test_opt(sb, ERRORS_CONT)) {
466 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
467 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
e2d67052 468 ext4_commit_super(sb, 0);
5d1b1b3f
AK		 469 return;
470 }
471 ext4_unlock_group(sb, grp);
472 ext4_handle_error(sb);
473 /*
474 * We only get here in the ERRORS_RO case; relocking the group
475 * may be dangerous, but nothing bad will happen since the
476 * filesystem will have already been marked read/only and the
477 * journal has been aborted. We return 1 as a hint to callers
478 * who might what to use the return value from
479 * ext4_grp_locked_error() to distinguish beween the
480 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
481 * aggressively from the ext4 function in question, with a
482 * more appropriate error code.
483 */
484 ext4_lock_group(sb, grp);
485 return;
486}
487
617ba13b 488void ext4_update_dynamic_rev(struct super_block *sb)
ac27a0ec 489{
617ba13b 490 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 491
617ba13b 492 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
ac27a0ec
DK		 493 return;
494
46e665e9 495 ext4_warning(sb, __func__,
ac27a0ec
DK		 496 "updating to rev %d because of new feature flag, "
497 "running e2fsck is recommended",
617ba13b 498 EXT4_DYNAMIC_REV);
ac27a0ec 499
617ba13b
MC		 500 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
501 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
502 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
ac27a0ec
DK		 503 /* leave es->s_feature_*compat flags alone */
504 /* es->s_uuid will be set by e2fsck if empty */
505
506 /*
507 * The rest of the superblock fields should be zero, and if not it
508 * means they are likely already in use, so leave them alone. We
509 * can leave it up to e2fsck to clean up any inconsistencies there.
510 */
511}
512
513/*
514 * Open the external journal device
515 */
b31e1552 516static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
ac27a0ec
DK		 517{
518 struct block_device *bdev;
519 char b[BDEVNAME_SIZE];
520
521 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
522 if (IS_ERR(bdev))
523 goto fail;
524 return bdev;
525
526fail:
b31e1552 527 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
ac27a0ec
DK		 528 __bdevname(dev, b), PTR_ERR(bdev));
529 return NULL;
530}
531
532/*
533 * Release the journal device
534 */
617ba13b 535static int ext4_blkdev_put(struct block_device *bdev)
ac27a0ec
DK		 536{
537 bd_release(bdev);
9a1c3542 538 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
ac27a0ec
DK		 539}
540
617ba13b 541static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
ac27a0ec
DK		 542{
543 struct block_device *bdev;
544 int ret = -ENODEV;
545
546 bdev = sbi->journal_bdev;
547 if (bdev) {
617ba13b 548 ret = ext4_blkdev_put(bdev);
ac27a0ec
DK		 549 sbi->journal_bdev = NULL;
550 }
551 return ret;
552}
553
554static inline struct inode *orphan_list_entry(struct list_head *l)
555{
617ba13b 556 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
ac27a0ec
DK		 557}
558
617ba13b 559static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
ac27a0ec
DK		 560{
561 struct list_head *l;
562
b31e1552
ES		 563 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
564 le32_to_cpu(sbi->s_es->s_last_orphan));
ac27a0ec
DK		 565
566 printk(KERN_ERR "sb_info orphan list:\n");
567 list_for_each(l, &sbi->s_orphan) {
568 struct inode *inode = orphan_list_entry(l);
569 printk(KERN_ERR " "
570 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
571 inode->i_sb->s_id, inode->i_ino, inode,
572 inode->i_mode, inode->i_nlink,
573 NEXT_ORPHAN(inode));
574 }
575}
576
2b2d6d01 577static void ext4_put_super(struct super_block *sb)
ac27a0ec 578{
617ba13b
MC		 579 struct ext4_sb_info *sbi = EXT4_SB(sb);
580 struct ext4_super_block *es = sbi->s_es;
ef2cabf7 581 int i, err;
ac27a0ec 582
a9e220f8 583 lock_super(sb);
6cfd0148 584 lock_kernel();
8c85e125 585 if (sb->s_dirt)
ebc1ac16 586 ext4_commit_super(sb, 1);
8c85e125 587
6fd058f7 588 ext4_release_system_zone(sb);
c9de560d 589 ext4_mb_release(sb);
a86c6181 590 ext4_ext_release(sb);
617ba13b 591 ext4_xattr_put_super(sb);
0390131b
FM		 592 if (sbi->s_journal) {
593 err = jbd2_journal_destroy(sbi->s_journal);
594 sbi->s_journal = NULL;
595 if (err < 0)
596 ext4_abort(sb, __func__,
597 "Couldn't clean up the journal");
598 }
ac27a0ec 599 if (!(sb->s_flags & MS_RDONLY)) {
617ba13b 600 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 601 es->s_state = cpu_to_le16(sbi->s_mount_state);
e2d67052 602 ext4_commit_super(sb, 1);
ac27a0ec 603 }
240799cd 604 if (sbi->s_proc) {
9f6200bb 605 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 606 }
3197ebdb 607 kobject_del(&sbi->s_kobj);
ac27a0ec
DK		 608
609 for (i = 0; i < sbi->s_gdb_count; i++)
610 brelse(sbi->s_group_desc[i]);
611 kfree(sbi->s_group_desc);
c5ca7c76
TT		 612 if (is_vmalloc_addr(sbi->s_flex_groups))
613 vfree(sbi->s_flex_groups);
614 else
615 kfree(sbi->s_flex_groups);
ac27a0ec
DK		 616 percpu_counter_destroy(&sbi->s_freeblocks_counter);
617 percpu_counter_destroy(&sbi->s_freeinodes_counter);
618 percpu_counter_destroy(&sbi->s_dirs_counter);
6bc6e63f 619 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
ac27a0ec
DK		 620 brelse(sbi->s_sbh);
621#ifdef CONFIG_QUOTA
622 for (i = 0; i < MAXQUOTAS; i++)
623 kfree(sbi->s_qf_names[i]);
624#endif
625
626 /* Debugging code just in case the in-memory inode orphan list
627 * isn't empty. The on-disk one can be non-empty if we've
628 * detected an error and taken the fs readonly, but the
629 * in-memory list had better be clean by this point. */
630 if (!list_empty(&sbi->s_orphan))
631 dump_orphan_list(sb, sbi);
632 J_ASSERT(list_empty(&sbi->s_orphan));
633
f98393a6 634 invalidate_bdev(sb->s_bdev);
ac27a0ec
DK		 635 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
636 /*
637 * Invalidate the journal device's buffers. We don't want them
638 * floating about in memory - the physical journal device may
639 * hotswapped, and it breaks the `ro-after' testing code.
640 */
641 sync_blockdev(sbi->journal_bdev);
f98393a6 642 invalidate_bdev(sbi->journal_bdev);
617ba13b 643 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 644 }
645 sb->s_fs_info = NULL;
3197ebdb
TT		 646 /*
647 * Now that we are completely done shutting down the
648 * superblock, we need to actually destroy the kobject.
649 */
650 unlock_kernel();
651 unlock_super(sb);
652 kobject_put(&sbi->s_kobj);
653 wait_for_completion(&sbi->s_kobj_unregister);
705895b6 654 kfree(sbi->s_blockgroup_lock);
ac27a0ec 655 kfree(sbi);
ac27a0ec
DK		 656}
657
e18b890b 658static struct kmem_cache *ext4_inode_cachep;
ac27a0ec
DK		 659
660/*
661 * Called inside transaction, so use GFP_NOFS
662 */
617ba13b 663static struct inode *ext4_alloc_inode(struct super_block *sb)
ac27a0ec 664{
617ba13b 665 struct ext4_inode_info *ei;
ac27a0ec 666
e6b4f8da 667 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
ac27a0ec
DK		 668 if (!ei)
669 return NULL;
0b8e58a1 670
ac27a0ec 671 ei->vfs_inode.i_version = 1;
91246c00 672 ei->vfs_inode.i_data.writeback_index = 0;
a86c6181 673 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
c9de560d
AT		 674 INIT_LIST_HEAD(&ei->i_prealloc_list);
675 spin_lock_init(&ei->i_prealloc_lock);
0390131b
FM		 676 /*
677 * Note: We can be called before EXT4_SB(sb)->s_journal is set,
678 * therefore it can be null here. Don't check it, just initialize
679 * jinode.
680 */
678aaf48 681 jbd2_journal_init_jbd_inode(&ei->jinode, &ei->vfs_inode);
d2a17637
MC		 682 ei->i_reserved_data_blocks = 0;
683 ei->i_reserved_meta_blocks = 0;
684 ei->i_allocated_meta_blocks = 0;
685 ei->i_delalloc_reserved_flag = 0;
686 spin_lock_init(&(ei->i_block_reservation_lock));
0b8e58a1 687
ac27a0ec
DK		 688 return &ei->vfs_inode;
689}
690
617ba13b 691static void ext4_destroy_inode(struct inode *inode)
ac27a0ec 692{
9f7dd93d 693 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
b31e1552
ES		 694 ext4_msg(inode->i_sb, KERN_ERR,
695 "Inode %lu (%p): orphan list check failed!",
696 inode->i_ino, EXT4_I(inode));
9f7dd93d
VA		 697 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
698 EXT4_I(inode), sizeof(struct ext4_inode_info),
699 true);
700 dump_stack();
701 }
617ba13b 702 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
ac27a0ec
DK		 703}
704
51cc5068 705static void init_once(void *foo)
ac27a0ec 706{
617ba13b 707 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
ac27a0ec 708
a35afb83 709 INIT_LIST_HEAD(&ei->i_orphan);
03010a33 710#ifdef CONFIG_EXT4_FS_XATTR
a35afb83 711 init_rwsem(&ei->xattr_sem);
ac27a0ec 712#endif
0e855ac8 713 init_rwsem(&ei->i_data_sem);
a35afb83 714 inode_init_once(&ei->vfs_inode);
ac27a0ec
DK		 715}
716
717static int init_inodecache(void)
718{
617ba13b
MC		 719 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
720 sizeof(struct ext4_inode_info),
ac27a0ec
DK		 721 0, (SLAB_RECLAIM_ACCOUNT|
722 SLAB_MEM_SPREAD),
20c2df83 723 init_once);
617ba13b 724 if (ext4_inode_cachep == NULL)
ac27a0ec
DK		 725 return -ENOMEM;
726 return 0;
727}
728
729static void destroy_inodecache(void)
730{
617ba13b 731 kmem_cache_destroy(ext4_inode_cachep);
ac27a0ec
DK		 732}
733
617ba13b 734static void ext4_clear_inode(struct inode *inode)
ac27a0ec 735{
c2ea3fde 736 ext4_discard_preallocations(inode);
0390131b
FM		 737 if (EXT4_JOURNAL(inode))
738 jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal,
678aaf48 739 &EXT4_I(inode)->jinode);
ac27a0ec
DK		 740}
741
2b2d6d01
TT		 742static inline void ext4_show_quota_options(struct seq_file *seq,
743 struct super_block *sb)
ac27a0ec
DK		 744{
745#if defined(CONFIG_QUOTA)
617ba13b 746 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 747
748 if (sbi->s_jquota_fmt)
749 seq_printf(seq, ",jqfmt=%s",
af5bc92d 750 (sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold" : "vfsv0");
ac27a0ec
DK		 751
752 if (sbi->s_qf_names[USRQUOTA])
753 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
754
755 if (sbi->s_qf_names[GRPQUOTA])
756 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
757
617ba13b 758 if (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA)
ac27a0ec
DK		 759 seq_puts(seq, ",usrquota");
760
617ba13b 761 if (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)
ac27a0ec
DK		 762 seq_puts(seq, ",grpquota");
763#endif
764}
765
d9c9bef1
MS		 766/*
767 * Show an option if
768 * - it's set to a non-default value OR
769 * - if the per-sb default is different from the global default
770 */
617ba13b 771static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
ac27a0ec 772{
aa22df2c
AK		 773 int def_errors;
774 unsigned long def_mount_opts;
ac27a0ec 775 struct super_block *sb = vfs->mnt_sb;
d9c9bef1
MS		 776 struct ext4_sb_info *sbi = EXT4_SB(sb);
777 struct ext4_super_block *es = sbi->s_es;
d9c9bef1
MS		 778
779 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
aa22df2c 780 def_errors = le16_to_cpu(es->s_errors);
d9c9bef1
MS		 781
782 if (sbi->s_sb_block != 1)
783 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
784 if (test_opt(sb, MINIX_DF))
785 seq_puts(seq, ",minixdf");
aa22df2c 786 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
d9c9bef1
MS		 787 seq_puts(seq, ",grpid");
788 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
789 seq_puts(seq, ",nogrpid");
790 if (sbi->s_resuid != EXT4_DEF_RESUID ||
791 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
792 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
793 }
794 if (sbi->s_resgid != EXT4_DEF_RESGID ||
795 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
796 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
797 }
bb4f397a 798 if (test_opt(sb, ERRORS_RO)) {
d9c9bef1 799 if (def_errors == EXT4_ERRORS_PANIC ||
bb4f397a
AK		 800 def_errors == EXT4_ERRORS_CONTINUE) {
801 seq_puts(seq, ",errors=remount-ro");
d9c9bef1
MS		 802 }
803 }
aa22df2c 804 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
bb4f397a 805 seq_puts(seq, ",errors=continue");
aa22df2c 806 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
d9c9bef1 807 seq_puts(seq, ",errors=panic");
aa22df2c 808 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
d9c9bef1 809 seq_puts(seq, ",nouid32");
aa22df2c 810 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
d9c9bef1
MS		 811 seq_puts(seq, ",debug");
812 if (test_opt(sb, OLDALLOC))
813 seq_puts(seq, ",oldalloc");
03010a33 814#ifdef CONFIG_EXT4_FS_XATTR
aa22df2c
AK		 815 if (test_opt(sb, XATTR_USER) &&
816 !(def_mount_opts & EXT4_DEFM_XATTR_USER))
d9c9bef1
MS		 817 seq_puts(seq, ",user_xattr");
818 if (!test_opt(sb, XATTR_USER) &&
819 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
820 seq_puts(seq, ",nouser_xattr");
821 }
822#endif
03010a33 823#ifdef CONFIG_EXT4_FS_POSIX_ACL
aa22df2c 824 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
d9c9bef1
MS		 825 seq_puts(seq, ",acl");
826 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
827 seq_puts(seq, ",noacl");
828#endif
30773840 829 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
d9c9bef1
MS		 830 seq_printf(seq, ",commit=%u",
831 (unsigned) (sbi->s_commit_interval / HZ));
832 }
30773840
TT		 833 if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
834 seq_printf(seq, ",min_batch_time=%u",
835 (unsigned) sbi->s_min_batch_time);
836 }
837 if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
838 seq_printf(seq, ",max_batch_time=%u",
839 (unsigned) sbi->s_min_batch_time);
840 }
841
571640ca
ES		 842 /*
843 * We're changing the default of barrier mount option, so
844 * let's always display its mount state so it's clear what its
845 * status is.
846 */
847 seq_puts(seq, ",barrier=");
848 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
cd0b6a39
TT		 849 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
850 seq_puts(seq, ",journal_async_commit");
d9c9bef1
MS		 851 if (test_opt(sb, NOBH))
852 seq_puts(seq, ",nobh");
25ec56b5
JNC		 853 if (test_opt(sb, I_VERSION))
854 seq_puts(seq, ",i_version");
dd919b98
AK		 855 if (!test_opt(sb, DELALLOC))
856 seq_puts(seq, ",nodelalloc");
857
ac27a0ec 858
cb45bbe4
MS		 859 if (sbi->s_stripe)
860 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
aa22df2c
AK		 861 /*
862 * journal mode get enabled in different ways
863 * So just print the value even if we didn't specify it
864 */
617ba13b 865 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
ac27a0ec 866 seq_puts(seq, ",data=journal");
617ba13b 867 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
ac27a0ec 868 seq_puts(seq, ",data=ordered");
617ba13b 869 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
ac27a0ec
DK		 870 seq_puts(seq, ",data=writeback");
871
240799cd
TT		 872 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
873 seq_printf(seq, ",inode_readahead_blks=%u",
874 sbi->s_inode_readahead_blks);
875
5bf5683a
HK		 876 if (test_opt(sb, DATA_ERR_ABORT))
877 seq_puts(seq, ",data_err=abort");
878
afd4672d 879 if (test_opt(sb, NO_AUTO_DA_ALLOC))
06705bff 880 seq_puts(seq, ",noauto_da_alloc");
afd4672d 881
617ba13b 882 ext4_show_quota_options(seq, sb);
0b8e58a1 883
ac27a0ec
DK		 884 return 0;
885}
886
1b961ac0 887static struct inode *ext4_nfs_get_inode(struct super_block *sb,
0b8e58a1 888 u64 ino, u32 generation)
ac27a0ec 889{
ac27a0ec 890 struct inode *inode;
ac27a0ec 891
617ba13b 892 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
ac27a0ec 893 return ERR_PTR(-ESTALE);
617ba13b 894 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
ac27a0ec
DK		 895 return ERR_PTR(-ESTALE);
896
897 /* iget isn't really right if the inode is currently unallocated!!
898 *
617ba13b 899 * ext4_read_inode will return a bad_inode if the inode had been
ac27a0ec
DK		 900 * deleted, so we should be safe.
901 *
902 * Currently we don't know the generation for parent directory, so
903 * a generation of 0 means "accept any"
904 */
1d1fe1ee
DH		 905 inode = ext4_iget(sb, ino);
906 if (IS_ERR(inode))
907 return ERR_CAST(inode);
908 if (generation && inode->i_generation != generation) {
ac27a0ec
DK		 909 iput(inode);
910 return ERR_PTR(-ESTALE);
911 }
1b961ac0
CH		 912
913 return inode;
914}
915
916static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
0b8e58a1 917 int fh_len, int fh_type)
1b961ac0
CH		 918{
919 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
920 ext4_nfs_get_inode);
921}
922
923static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
0b8e58a1 924 int fh_len, int fh_type)
1b961ac0
CH		 925{
926 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
927 ext4_nfs_get_inode);
ac27a0ec
DK		 928}
929
c39a7f84
TO		 930/*
931 * Try to release metadata pages (indirect blocks, directories) which are
932 * mapped via the block device. Since these pages could have journal heads
933 * which would prevent try_to_free_buffers() from freeing them, we must use
934 * jbd2 layer's try_to_free_buffers() function to release them.
935 */
0b8e58a1
AD		 936static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
937 gfp_t wait)
c39a7f84
TO		 938{
939 journal_t *journal = EXT4_SB(sb)->s_journal;
940
941 WARN_ON(PageChecked(page));
942 if (!page_has_buffers(page))
943 return 0;
944 if (journal)
945 return jbd2_journal_try_to_free_buffers(journal, page,
946 wait & ~__GFP_WAIT);
947 return try_to_free_buffers(page);
948}
949
ac27a0ec 950#ifdef CONFIG_QUOTA
af5bc92d 951#define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
2b2d6d01 952#define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
ac27a0ec 953
617ba13b
MC		 954static int ext4_write_dquot(struct dquot *dquot);
955static int ext4_acquire_dquot(struct dquot *dquot);
956static int ext4_release_dquot(struct dquot *dquot);
957static int ext4_mark_dquot_dirty(struct dquot *dquot);
958static int ext4_write_info(struct super_block *sb, int type);
6f28e087
JK		 959static int ext4_quota_on(struct super_block *sb, int type, int format_id,
960 char *path, int remount);
617ba13b
MC		 961static int ext4_quota_on_mount(struct super_block *sb, int type);
962static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 963 size_t len, loff_t off);
617ba13b 964static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 965 const char *data, size_t len, loff_t off);
966
617ba13b 967static struct dquot_operations ext4_quota_operations = {
edf72453
JK		 968 .initialize = dquot_initialize,
969 .drop = dquot_drop,
ac27a0ec 970 .alloc_space = dquot_alloc_space,
60e58e0f
MC		 971 .reserve_space = dquot_reserve_space,
972 .claim_space = dquot_claim_space,
973 .release_rsv = dquot_release_reserved_space,
974 .get_reserved_space = ext4_get_reserved_space,
ac27a0ec
DK		 975 .alloc_inode = dquot_alloc_inode,
976 .free_space = dquot_free_space,
977 .free_inode = dquot_free_inode,
978 .transfer = dquot_transfer,
617ba13b
MC		 979 .write_dquot = ext4_write_dquot,
980 .acquire_dquot = ext4_acquire_dquot,
981 .release_dquot = ext4_release_dquot,
982 .mark_dirty = ext4_mark_dquot_dirty,
a5b5ee32
JK		 983 .write_info = ext4_write_info,
984 .alloc_dquot = dquot_alloc,
985 .destroy_dquot = dquot_destroy,
ac27a0ec
DK		 986};
987
617ba13b
MC		 988static struct quotactl_ops ext4_qctl_operations = {
989 .quota_on = ext4_quota_on,
ac27a0ec
DK		 990 .quota_off = vfs_quota_off,
991 .quota_sync = vfs_quota_sync,
992 .get_info = vfs_get_dqinfo,
993 .set_info = vfs_set_dqinfo,
994 .get_dqblk = vfs_get_dqblk,
995 .set_dqblk = vfs_set_dqblk
996};
997#endif
998
ee9b6d61 999static const struct super_operations ext4_sops = {
617ba13b
MC		 1000 .alloc_inode = ext4_alloc_inode,
1001 .destroy_inode = ext4_destroy_inode,
617ba13b
MC		 1002 .write_inode = ext4_write_inode,
1003 .dirty_inode = ext4_dirty_inode,
1004 .delete_inode = ext4_delete_inode,
1005 .put_super = ext4_put_super,
617ba13b 1006 .sync_fs = ext4_sync_fs,
c4be0c1d
TS		 1007 .freeze_fs = ext4_freeze,
1008 .unfreeze_fs = ext4_unfreeze,
617ba13b
MC		 1009 .statfs = ext4_statfs,
1010 .remount_fs = ext4_remount,
1011 .clear_inode = ext4_clear_inode,
1012 .show_options = ext4_show_options,
ac27a0ec 1013#ifdef CONFIG_QUOTA
617ba13b
MC		 1014 .quota_read = ext4_quota_read,
1015 .quota_write = ext4_quota_write,
ac27a0ec 1016#endif
c39a7f84 1017 .bdev_try_to_free_page = bdev_try_to_free_page,
ac27a0ec
DK		 1018};
1019
9ca92389
TT		 1020static const struct super_operations ext4_nojournal_sops = {
1021 .alloc_inode = ext4_alloc_inode,
1022 .destroy_inode = ext4_destroy_inode,
1023 .write_inode = ext4_write_inode,
1024 .dirty_inode = ext4_dirty_inode,
1025 .delete_inode = ext4_delete_inode,
1026 .write_super = ext4_write_super,
1027 .put_super = ext4_put_super,
1028 .statfs = ext4_statfs,
1029 .remount_fs = ext4_remount,
1030 .clear_inode = ext4_clear_inode,
1031 .show_options = ext4_show_options,
1032#ifdef CONFIG_QUOTA
1033 .quota_read = ext4_quota_read,
1034 .quota_write = ext4_quota_write,
1035#endif
1036 .bdev_try_to_free_page = bdev_try_to_free_page,
1037};
1038
39655164 1039static const struct export_operations ext4_export_ops = {
1b961ac0
CH		 1040 .fh_to_dentry = ext4_fh_to_dentry,
1041 .fh_to_parent = ext4_fh_to_parent,
617ba13b 1042 .get_parent = ext4_get_parent,
ac27a0ec
DK		 1043};
1044
1045enum {
1046 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1047 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
01436ef2 1048 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
ac27a0ec 1049 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
06705bff 1050 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh,
30773840 1051 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
c3191067 1052 Opt_journal_update, Opt_journal_dev,
818d276c 1053 Opt_journal_checksum, Opt_journal_async_commit,
ac27a0ec 1054 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
f4033903 1055 Opt_data_err_abort, Opt_data_err_ignore, Opt_mb_history_length,
ac27a0ec
DK		 1056 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1057 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
06705bff
TT		 1058 Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err, Opt_resize,
1059 Opt_usrquota, Opt_grpquota, Opt_i_version,
01436ef2 1060 Opt_stripe, Opt_delalloc, Opt_nodelalloc,
6fd058f7 1061 Opt_block_validity, Opt_noblock_validity,
b3881f74 1062 Opt_inode_readahead_blks, Opt_journal_ioprio
ac27a0ec
DK		 1063};
1064
a447c093 1065static const match_table_t tokens = {
ac27a0ec
DK		 1066 {Opt_bsd_df, "bsddf"},
1067 {Opt_minix_df, "minixdf"},
1068 {Opt_grpid, "grpid"},
1069 {Opt_grpid, "bsdgroups"},
1070 {Opt_nogrpid, "nogrpid"},
1071 {Opt_nogrpid, "sysvgroups"},
1072 {Opt_resgid, "resgid=%u"},
1073 {Opt_resuid, "resuid=%u"},
1074 {Opt_sb, "sb=%u"},
1075 {Opt_err_cont, "errors=continue"},
1076 {Opt_err_panic, "errors=panic"},
1077 {Opt_err_ro, "errors=remount-ro"},
1078 {Opt_nouid32, "nouid32"},
ac27a0ec
DK		 1079 {Opt_debug, "debug"},
1080 {Opt_oldalloc, "oldalloc"},
1081 {Opt_orlov, "orlov"},
1082 {Opt_user_xattr, "user_xattr"},
1083 {Opt_nouser_xattr, "nouser_xattr"},
1084 {Opt_acl, "acl"},
1085 {Opt_noacl, "noacl"},
ac27a0ec
DK		 1086 {Opt_noload, "noload"},
1087 {Opt_nobh, "nobh"},
1088 {Opt_bh, "bh"},
1089 {Opt_commit, "commit=%u"},
30773840
TT		 1090 {Opt_min_batch_time, "min_batch_time=%u"},
1091 {Opt_max_batch_time, "max_batch_time=%u"},
ac27a0ec 1092 {Opt_journal_update, "journal=update"},
ac27a0ec 1093 {Opt_journal_dev, "journal_dev=%u"},
818d276c
GS		 1094 {Opt_journal_checksum, "journal_checksum"},
1095 {Opt_journal_async_commit, "journal_async_commit"},
ac27a0ec
DK		 1096 {Opt_abort, "abort"},
1097 {Opt_data_journal, "data=journal"},
1098 {Opt_data_ordered, "data=ordered"},
1099 {Opt_data_writeback, "data=writeback"},
5bf5683a
HK		 1100 {Opt_data_err_abort, "data_err=abort"},
1101 {Opt_data_err_ignore, "data_err=ignore"},
f4033903 1102 {Opt_mb_history_length, "mb_history_length=%u"},
ac27a0ec
DK		 1103 {Opt_offusrjquota, "usrjquota="},
1104 {Opt_usrjquota, "usrjquota=%s"},
1105 {Opt_offgrpjquota, "grpjquota="},
1106 {Opt_grpjquota, "grpjquota=%s"},
1107 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1108 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1109 {Opt_grpquota, "grpquota"},
1110 {Opt_noquota, "noquota"},
1111 {Opt_quota, "quota"},
1112 {Opt_usrquota, "usrquota"},
1113 {Opt_barrier, "barrier=%u"},
06705bff
TT		 1114 {Opt_barrier, "barrier"},
1115 {Opt_nobarrier, "nobarrier"},
25ec56b5 1116 {Opt_i_version, "i_version"},
c9de560d 1117 {Opt_stripe, "stripe=%u"},
ac27a0ec 1118 {Opt_resize, "resize"},
64769240 1119 {Opt_delalloc, "delalloc"},
dd919b98 1120 {Opt_nodelalloc, "nodelalloc"},
6fd058f7
TT		 1121 {Opt_block_validity, "block_validity"},
1122 {Opt_noblock_validity, "noblock_validity"},
240799cd 1123 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
b3881f74 1124 {Opt_journal_ioprio, "journal_ioprio=%u"},
afd4672d 1125 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
06705bff
TT		 1126 {Opt_auto_da_alloc, "auto_da_alloc"},
1127 {Opt_noauto_da_alloc, "noauto_da_alloc"},
f3f12faa 1128 {Opt_err, NULL},
ac27a0ec
DK		 1129};
1130
617ba13b 1131static ext4_fsblk_t get_sb_block(void **data)
ac27a0ec 1132{
617ba13b 1133 ext4_fsblk_t sb_block;
ac27a0ec
DK		 1134 char *options = (char *) *data;
1135
1136 if (!options || strncmp(options, "sb=", 3) != 0)
1137 return 1; /* Default location */
0b8e58a1 1138
ac27a0ec 1139 options += 3;
0b8e58a1 1140 /* TODO: use simple_strtoll with >32bit ext4 */
ac27a0ec
DK		 1141 sb_block = simple_strtoul(options, &options, 0);
1142 if (*options && *options != ',') {
4776004f 1143 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
ac27a0ec
DK		 1144 (char *) *data);
1145 return 1;
1146 }
1147 if (*options == ',')
1148 options++;
1149 *data = (void *) options;
0b8e58a1 1150
ac27a0ec
DK		 1151 return sb_block;
1152}
1153
b3881f74
TT		 1154#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1155
2b2d6d01 1156static int parse_options(char *options, struct super_block *sb,
c3191067 1157 unsigned long *journal_devnum,
b3881f74 1158 unsigned int *journal_ioprio,
2b2d6d01 1159 ext4_fsblk_t *n_blocks_count, int is_remount)
ac27a0ec 1160{
617ba13b 1161 struct ext4_sb_info *sbi = EXT4_SB(sb);
2b2d6d01 1162 char *p;
ac27a0ec
DK		 1163 substring_t args[MAX_OPT_ARGS];
1164 int data_opt = 0;
1165 int option;
1166#ifdef CONFIG_QUOTA
dfc5d03f 1167 int qtype, qfmt;
ac27a0ec
DK		 1168 char *qname;
1169#endif
1170
1171 if (!options)
1172 return 1;
1173
2b2d6d01 1174 while ((p = strsep(&options, ",")) != NULL) {
ac27a0ec
DK		 1175 int token;
1176 if (!*p)
1177 continue;
1178
1179 token = match_token(p, tokens, args);
1180 switch (token) {
1181 case Opt_bsd_df:
2b2d6d01 1182 clear_opt(sbi->s_mount_opt, MINIX_DF);
ac27a0ec
DK		 1183 break;
1184 case Opt_minix_df:
2b2d6d01 1185 set_opt(sbi->s_mount_opt, MINIX_DF);
ac27a0ec
DK		 1186 break;
1187 case Opt_grpid:
2b2d6d01 1188 set_opt(sbi->s_mount_opt, GRPID);
ac27a0ec
DK		 1189 break;
1190 case Opt_nogrpid:
2b2d6d01 1191 clear_opt(sbi->s_mount_opt, GRPID);
ac27a0ec
DK		 1192 break;
1193 case Opt_resuid:
1194 if (match_int(&args[0], &option))
1195 return 0;
1196 sbi->s_resuid = option;
1197 break;
1198 case Opt_resgid:
1199 if (match_int(&args[0], &option))
1200 return 0;
1201 sbi->s_resgid = option;
1202 break;
1203 case Opt_sb:
1204 /* handled by get_sb_block() instead of here */
1205 /* *sb_block = match_int(&args[0]); */
1206 break;
1207 case Opt_err_panic:
2b2d6d01
TT		 1208 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1209 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1210 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
ac27a0ec
DK		 1211 break;
1212 case Opt_err_ro:
2b2d6d01
TT		 1213 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1214 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1215 set_opt(sbi->s_mount_opt, ERRORS_RO);
ac27a0ec
DK		 1216 break;
1217 case Opt_err_cont:
2b2d6d01
TT		 1218 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1219 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1220 set_opt(sbi->s_mount_opt, ERRORS_CONT);
ac27a0ec
DK		 1221 break;
1222 case Opt_nouid32:
2b2d6d01 1223 set_opt(sbi->s_mount_opt, NO_UID32);
ac27a0ec 1224 break;
ac27a0ec 1225 case Opt_debug:
2b2d6d01 1226 set_opt(sbi->s_mount_opt, DEBUG);
ac27a0ec
DK		 1227 break;
1228 case Opt_oldalloc:
2b2d6d01 1229 set_opt(sbi->s_mount_opt, OLDALLOC);
ac27a0ec
DK		 1230 break;
1231 case Opt_orlov:
2b2d6d01 1232 clear_opt(sbi->s_mount_opt, OLDALLOC);
ac27a0ec 1233 break;
03010a33 1234#ifdef CONFIG_EXT4_FS_XATTR
ac27a0ec 1235 case Opt_user_xattr:
2b2d6d01 1236 set_opt(sbi->s_mount_opt, XATTR_USER);
ac27a0ec
DK		 1237 break;
1238 case Opt_nouser_xattr:
2b2d6d01 1239 clear_opt(sbi->s_mount_opt, XATTR_USER);
ac27a0ec
DK		 1240 break;
1241#else
1242 case Opt_user_xattr:
1243 case Opt_nouser_xattr:
b31e1552 1244 ext4_msg(sb, KERN_ERR, "(no)user_xattr options not supported");
ac27a0ec
DK		 1245 break;
1246#endif
03010a33 1247#ifdef CONFIG_EXT4_FS_POSIX_ACL
ac27a0ec
DK		 1248 case Opt_acl:
1249 set_opt(sbi->s_mount_opt, POSIX_ACL);
1250 break;
1251 case Opt_noacl:
1252 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1253 break;
1254#else
1255 case Opt_acl:
1256 case Opt_noacl:
b31e1552 1257 ext4_msg(sb, KERN_ERR, "(no)acl options not supported");
ac27a0ec
DK		 1258 break;
1259#endif
ac27a0ec
DK		 1260 case Opt_journal_update:
1261 /* @@@ FIXME */
1262 /* Eventually we will want to be able to create
1263 a journal file here. For now, only allow the
1264 user to specify an existing inode to be the
1265 journal file. */
1266 if (is_remount) {
b31e1552
ES		 1267 ext4_msg(sb, KERN_ERR,
1268 "Cannot specify journal on remount");
ac27a0ec
DK		 1269 return 0;
1270 }
2b2d6d01 1271 set_opt(sbi->s_mount_opt, UPDATE_JOURNAL);
ac27a0ec 1272 break;
ac27a0ec
DK		 1273 case Opt_journal_dev:
1274 if (is_remount) {
b31e1552
ES		 1275 ext4_msg(sb, KERN_ERR,
1276 "Cannot specify journal on remount");
ac27a0ec
DK		 1277 return 0;
1278 }
1279 if (match_int(&args[0], &option))
1280 return 0;
1281 *journal_devnum = option;
1282 break;
818d276c 1283 case Opt_journal_checksum:
d0646f7b 1284 break; /* Kept for backwards compatibility */
818d276c
GS		 1285 case Opt_journal_async_commit:
1286 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
818d276c 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);
7ad9bb65
TT		 1698 atomic_add(ext4_free_inodes_count(sb, gdp),
1699 &sbi->s_flex_groups[flex_group].free_inodes);
1700 atomic_add(ext4_free_blks_count(sb, gdp),
1701 &sbi->s_flex_groups[flex_group].free_blocks);
1702 atomic_add(ext4_used_dirs_count(sb, gdp),
1703 &sbi->s_flex_groups[flex_group].used_dirs);
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
a13fb1a4
ES		 2256/*
2257 * Check whether this filesystem can be mounted based on
2258 * the features present and the RDONLY/RDWR mount requested.
2259 * Returns 1 if this filesystem can be mounted as requested,
2260 * 0 if it cannot be.
2261 */
2262static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2263{
2264 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2265 ext4_msg(sb, KERN_ERR,
2266 "Couldn't mount because of "
2267 "unsupported optional features (%x)",
2268 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2269 ~EXT4_FEATURE_INCOMPAT_SUPP));
2270 return 0;
2271 }
2272
2273 if (readonly)
2274 return 1;
2275
2276 /* Check that feature set is OK for a read-write mount */
2277 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2278 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2279 "unsupported optional features (%x)",
2280 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2281 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2282 return 0;
2283 }
2284 /*
2285 * Large file size enabled file system can only be mounted
2286 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2287 */
2288 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2289 if (sizeof(blkcnt_t) < sizeof(u64)) {
2290 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2291 "cannot be mounted RDWR without "
2292 "CONFIG_LBDAF");
2293 return 0;
2294 }
2295 }
2296 return 1;
2297}
2298
2b2d6d01 2299static int ext4_fill_super(struct super_block *sb, void *data, int silent)
7477827f
AK		 2300 __releases(kernel_lock)
2301 __acquires(kernel_lock)
ac27a0ec 2302{
2b2d6d01 2303 struct buffer_head *bh;
617ba13b
MC		 2304 struct ext4_super_block *es = NULL;
2305 struct ext4_sb_info *sbi;
2306 ext4_fsblk_t block;
2307 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 2308 ext4_fsblk_t logical_sb_block;
ac27a0ec 2309 unsigned long offset = 0;
ac27a0ec
DK		 2310 unsigned long journal_devnum = 0;
2311 unsigned long def_mount_opts;
2312 struct inode *root;
9f6200bb 2313 char *cp;
0390131b 2314 const char *descr;
1d1fe1ee 2315 int ret = -EINVAL;
ac27a0ec 2316 int blocksize;
4ec11028
TT		 2317 unsigned int db_count;
2318 unsigned int i;
f287a1a5 2319 int needs_recovery, has_huge_files;
bd81d8ee 2320 __u64 blocks_count;
833f4077 2321 int err;
b3881f74 2322 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
ac27a0ec
DK		 2323
2324 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
2325 if (!sbi)
2326 return -ENOMEM;
705895b6
PE		 2327
2328 sbi->s_blockgroup_lock =
2329 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
2330 if (!sbi->s_blockgroup_lock) {
2331 kfree(sbi);
2332 return -ENOMEM;
2333 }
ac27a0ec
DK		 2334 sb->s_fs_info = sbi;
2335 sbi->s_mount_opt = 0;
617ba13b
MC		 2336 sbi->s_resuid = EXT4_DEF_RESUID;
2337 sbi->s_resgid = EXT4_DEF_RESGID;
240799cd 2338 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
d9c9bef1 2339 sbi->s_sb_block = sb_block;
afc32f7e
TT		 2340 sbi->s_sectors_written_start = part_stat_read(sb->s_bdev->bd_part,
2341 sectors[1]);
ac27a0ec
DK		 2342
2343 unlock_kernel();
2344
9f6200bb
TT		 2345 /* Cleanup superblock name */
2346 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
2347 *cp = '!';
2348
617ba13b 2349 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 2350 if (!blocksize) {
b31e1552 2351 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
ac27a0ec
DK		 2352 goto out_fail;
2353 }
2354
2355 /*
617ba13b 2356 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 2357 * block sizes. We need to calculate the offset from buffer start.
2358 */
617ba13b 2359 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 2360 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2361 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 2362 } else {
70bbb3e0 2363 logical_sb_block = sb_block;
ac27a0ec
DK		 2364 }
2365
70bbb3e0 2366 if (!(bh = sb_bread(sb, logical_sb_block))) {
b31e1552 2367 ext4_msg(sb, KERN_ERR, "unable to read superblock");
ac27a0ec
DK		 2368 goto out_fail;
2369 }
2370 /*
2371 * Note: s_es must be initialized as soon as possible because
617ba13b 2372 * some ext4 macro-instructions depend on its value
ac27a0ec 2373 */
617ba13b 2374 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
ac27a0ec
DK		 2375 sbi->s_es = es;
2376 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 2377 if (sb->s_magic != EXT4_SUPER_MAGIC)
2378 goto cantfind_ext4;
afc32f7e 2379 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
ac27a0ec
DK		 2380
2381 /* Set defaults before we parse the mount options */
2382 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
617ba13b 2383 if (def_mount_opts & EXT4_DEFM_DEBUG)
ac27a0ec 2384 set_opt(sbi->s_mount_opt, DEBUG);
617ba13b 2385 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
ac27a0ec 2386 set_opt(sbi->s_mount_opt, GRPID);
617ba13b 2387 if (def_mount_opts & EXT4_DEFM_UID16)
ac27a0ec 2388 set_opt(sbi->s_mount_opt, NO_UID32);
03010a33 2389#ifdef CONFIG_EXT4_FS_XATTR
617ba13b 2390 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
ac27a0ec 2391 set_opt(sbi->s_mount_opt, XATTR_USER);
2e7842b8 2392#endif
03010a33 2393#ifdef CONFIG_EXT4_FS_POSIX_ACL
617ba13b 2394 if (def_mount_opts & EXT4_DEFM_ACL)
ac27a0ec 2395 set_opt(sbi->s_mount_opt, POSIX_ACL);
2e7842b8 2396#endif
617ba13b
MC		 2397 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
2398 sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
2399 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
2400 sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
2401 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
2402 sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
2403
2404 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
ac27a0ec 2405 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
bb4f397a 2406 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
ceea16bf 2407 set_opt(sbi->s_mount_opt, ERRORS_CONT);
bb4f397a
AK		 2408 else
2409 set_opt(sbi->s_mount_opt, ERRORS_RO);
ac27a0ec
DK		 2410
2411 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
2412 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
30773840
TT		 2413 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
2414 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
2415 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
f4033903 2416 sbi->s_mb_history_max = default_mb_history_length;
ac27a0ec 2417
571640ca 2418 set_opt(sbi->s_mount_opt, BARRIER);
ac27a0ec 2419
dd919b98
AK		 2420 /*
2421 * enable delayed allocation by default
2422 * Use -o nodelalloc to turn it off
2423 */
2424 set_opt(sbi->s_mount_opt, DELALLOC);
2425
b3881f74
TT		 2426 if (!parse_options((char *) data, sb, &journal_devnum,
2427 &journal_ioprio, NULL, 0))
ac27a0ec
DK		 2428 goto failed_mount;
2429
2430 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
617ba13b 2431 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec 2432
617ba13b
MC		 2433 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
2434 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
2435 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
2436 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
b31e1552
ES		 2437 ext4_msg(sb, KERN_WARNING,
2438 "feature flags set on rev 0 fs, "
2439 "running e2fsck is recommended");
469108ff 2440
ac27a0ec
DK		 2441 /*
2442 * Check feature flags regardless of the revision level, since we
2443 * previously didn't change the revision level when setting the flags,
2444 * so there is a chance incompat flags are set on a rev 0 filesystem.
2445 */
a13fb1a4 2446 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
ac27a0ec 2447 goto failed_mount;
a13fb1a4 2448
ac27a0ec
DK		 2449 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
2450
617ba13b
MC		 2451 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
2452 blocksize > EXT4_MAX_BLOCK_SIZE) {
b31e1552
ES		 2453 ext4_msg(sb, KERN_ERR,
2454 "Unsupported filesystem blocksize %d", blocksize);
ac27a0ec
DK		 2455 goto failed_mount;
2456 }
2457
ac27a0ec 2458 if (sb->s_blocksize != blocksize) {
ce40733c
AK		 2459 /* Validate the filesystem blocksize */
2460 if (!sb_set_blocksize(sb, blocksize)) {
b31e1552 2461 ext4_msg(sb, KERN_ERR, "bad block size %d",
ce40733c 2462 blocksize);
ac27a0ec
DK		 2463 goto failed_mount;
2464 }
2465
2b2d6d01 2466 brelse(bh);
70bbb3e0
AM		 2467 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2468 offset = do_div(logical_sb_block, blocksize);
2469 bh = sb_bread(sb, logical_sb_block);
ac27a0ec 2470 if (!bh) {
b31e1552
ES		 2471 ext4_msg(sb, KERN_ERR,
2472 "Can't read superblock on 2nd try");
ac27a0ec
DK		 2473 goto failed_mount;
2474 }
617ba13b 2475 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
ac27a0ec 2476 sbi->s_es = es;
617ba13b 2477 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
b31e1552
ES		 2478 ext4_msg(sb, KERN_ERR,
2479 "Magic mismatch, very weird!");
ac27a0ec
DK		 2480 goto failed_mount;
2481 }
2482 }
2483
a13fb1a4
ES		 2484 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2485 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
f287a1a5
TT		 2486 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
2487 has_huge_files);
2488 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
ac27a0ec 2489
617ba13b
MC		 2490 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2491 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2492 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
ac27a0ec
DK		 2493 } else {
2494 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2495 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
617ba13b 2496 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1330593e 2497 (!is_power_of_2(sbi->s_inode_size)) ||
ac27a0ec 2498 (sbi->s_inode_size > blocksize)) {
b31e1552
ES		 2499 ext4_msg(sb, KERN_ERR,
2500 "unsupported inode size: %d",
2b2d6d01 2501 sbi->s_inode_size);
ac27a0ec
DK		 2502 goto failed_mount;
2503 }
ef7f3835
KS		 2504 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2505 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
ac27a0ec 2506 }
0b8e58a1 2507
0d1ee42f
AR		 2508 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2509 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
8fadc143 2510 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 2511 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 2512 !is_power_of_2(sbi->s_desc_size)) {
b31e1552
ES		 2513 ext4_msg(sb, KERN_ERR,
2514 "unsupported descriptor size %lu",
0d1ee42f
AR		 2515 sbi->s_desc_size);
2516 goto failed_mount;
2517 }
2518 } else
2519 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
0b8e58a1 2520
ac27a0ec 2521 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 2522 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
b47b6f38 2523 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
617ba13b 2524 goto cantfind_ext4;
0b8e58a1 2525
617ba13b 2526 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 2527 if (sbi->s_inodes_per_block == 0)
617ba13b 2528 goto cantfind_ext4;
ac27a0ec
DK		 2529 sbi->s_itb_per_group = sbi->s_inodes_per_group /
2530 sbi->s_inodes_per_block;
0d1ee42f 2531 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 2532 sbi->s_sbh = bh;
2533 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 2534 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2535 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
0b8e58a1 2536
2b2d6d01 2537 for (i = 0; i < 4; i++)
ac27a0ec
DK		 2538 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2539 sbi->s_def_hash_version = es->s_def_hash_version;
f99b2589
TT		 2540 i = le32_to_cpu(es->s_flags);
2541 if (i & EXT2_FLAGS_UNSIGNED_HASH)
2542 sbi->s_hash_unsigned = 3;
2543 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
2544#ifdef __CHAR_UNSIGNED__
2545 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
2546 sbi->s_hash_unsigned = 3;
2547#else
2548 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
2549#endif
2550 sb->s_dirt = 1;
2551 }
ac27a0ec
DK		 2552
2553 if (sbi->s_blocks_per_group > blocksize * 8) {
b31e1552
ES		 2554 ext4_msg(sb, KERN_ERR,
2555 "#blocks per group too big: %lu",
2b2d6d01 2556 sbi->s_blocks_per_group);
ac27a0ec
DK		 2557 goto failed_mount;
2558 }
ac27a0ec 2559 if (sbi->s_inodes_per_group > blocksize * 8) {
b31e1552
ES		 2560 ext4_msg(sb, KERN_ERR,
2561 "#inodes per group too big: %lu",
2b2d6d01 2562 sbi->s_inodes_per_group);
ac27a0ec
DK		 2563 goto failed_mount;
2564 }
2565
bf43d84b
ES		 2566 /*
2567 * Test whether we have more sectors than will fit in sector_t,
2568 * and whether the max offset is addressable by the page cache.
2569 */
2570 if ((ext4_blocks_count(es) >
2571 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) ||
2572 (ext4_blocks_count(es) >
2573 (pgoff_t)(~0ULL) >> (PAGE_CACHE_SHIFT - sb->s_blocksize_bits))) {
b31e1552 2574 ext4_msg(sb, KERN_ERR, "filesystem"
bf43d84b 2575 " too large to mount safely on this system");
ac27a0ec 2576 if (sizeof(sector_t) < 8)
90c699a9 2577 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
bf43d84b 2578 ret = -EFBIG;
ac27a0ec
DK		 2579 goto failed_mount;
2580 }
2581
617ba13b
MC		 2582 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2583 goto cantfind_ext4;
e7c95593 2584
0f2ddca6
FTN		 2585 /* check blocks count against device size */
2586 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
2587 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
b31e1552
ES		 2588 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
2589 "exceeds size of device (%llu blocks)",
0f2ddca6
FTN		 2590 ext4_blocks_count(es), blocks_count);
2591 goto failed_mount;
2592 }
2593
0b8e58a1
AD		 2594 /*
2595 * It makes no sense for the first data block to be beyond the end
2596 * of the filesystem.
2597 */
2598 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
b31e1552
ES		 2599 ext4_msg(sb, KERN_WARNING, "bad geometry: first data"
2600 "block %u is beyond end of filesystem (%llu)",
2601 le32_to_cpu(es->s_first_data_block),
2602 ext4_blocks_count(es));
e7c95593
ES		 2603 goto failed_mount;
2604 }
bd81d8ee
LV		 2605 blocks_count = (ext4_blocks_count(es) -
2606 le32_to_cpu(es->s_first_data_block) +
2607 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2608 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4ec11028 2609 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
b31e1552 2610 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
4ec11028 2611 "(block count %llu, first data block %u, "
b31e1552 2612 "blocks per group %lu)", sbi->s_groups_count,
4ec11028
TT		 2613 ext4_blocks_count(es),
2614 le32_to_cpu(es->s_first_data_block),
2615 EXT4_BLOCKS_PER_GROUP(sb));
2616 goto failed_mount;
2617 }
bd81d8ee 2618 sbi->s_groups_count = blocks_count;
617ba13b
MC		 2619 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2620 EXT4_DESC_PER_BLOCK(sb);
2b2d6d01 2621 sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
ac27a0ec
DK		 2622 GFP_KERNEL);
2623 if (sbi->s_group_desc == NULL) {
b31e1552 2624 ext4_msg(sb, KERN_ERR, "not enough memory");
ac27a0ec
DK		 2625 goto failed_mount;
2626 }
2627
3244fcb1 2628#ifdef CONFIG_PROC_FS
9f6200bb
TT		 2629 if (ext4_proc_root)
2630 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
3244fcb1 2631#endif
240799cd 2632
705895b6 2633 bgl_lock_init(sbi->s_blockgroup_lock);
ac27a0ec
DK		 2634
2635 for (i = 0; i < db_count; i++) {
70bbb3e0 2636 block = descriptor_loc(sb, logical_sb_block, i);
ac27a0ec
DK		 2637 sbi->s_group_desc[i] = sb_bread(sb, block);
2638 if (!sbi->s_group_desc[i]) {
b31e1552
ES		 2639 ext4_msg(sb, KERN_ERR,
2640 "can't read group descriptor %d", i);
ac27a0ec
DK		 2641 db_count = i;
2642 goto failed_mount2;
2643 }
2644 }
2b2d6d01 2645 if (!ext4_check_descriptors(sb)) {
b31e1552 2646 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
ac27a0ec
DK		 2647 goto failed_mount2;
2648 }
772cb7c8
JS		 2649 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2650 if (!ext4_fill_flex_info(sb)) {
b31e1552
ES		 2651 ext4_msg(sb, KERN_ERR,
2652 "unable to initialize "
2653 "flex_bg meta info!");
772cb7c8
JS		 2654 goto failed_mount2;
2655 }
2656
ac27a0ec
DK		 2657 sbi->s_gdb_count = db_count;
2658 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2659 spin_lock_init(&sbi->s_next_gen_lock);
2660
833f4077
PZ		 2661 err = percpu_counter_init(&sbi->s_freeblocks_counter,
2662 ext4_count_free_blocks(sb));
2663 if (!err) {
2664 err = percpu_counter_init(&sbi->s_freeinodes_counter,
2665 ext4_count_free_inodes(sb));
2666 }
2667 if (!err) {
2668 err = percpu_counter_init(&sbi->s_dirs_counter,
2669 ext4_count_dirs(sb));
2670 }
6bc6e63f
AK		 2671 if (!err) {
2672 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
2673 }
833f4077 2674 if (err) {
b31e1552 2675 ext4_msg(sb, KERN_ERR, "insufficient memory");
833f4077
PZ		 2676 goto failed_mount3;
2677 }
ac27a0ec 2678
c9de560d
AT		 2679 sbi->s_stripe = ext4_get_stripe_size(sbi);
2680
ac27a0ec
DK		 2681 /*
2682 * set up enough so that it can read an inode
2683 */
9ca92389
TT		 2684 if (!test_opt(sb, NOLOAD) &&
2685 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
2686 sb->s_op = &ext4_sops;
2687 else
2688 sb->s_op = &ext4_nojournal_sops;
617ba13b
MC		 2689 sb->s_export_op = &ext4_export_ops;
2690 sb->s_xattr = ext4_xattr_handlers;
ac27a0ec 2691#ifdef CONFIG_QUOTA
617ba13b
MC		 2692 sb->s_qcop = &ext4_qctl_operations;
2693 sb->dq_op = &ext4_quota_operations;
ac27a0ec
DK		 2694#endif
2695 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3b9d4ed2 2696 mutex_init(&sbi->s_orphan_lock);
32ed5058 2697 mutex_init(&sbi->s_resize_lock);
ac27a0ec
DK		 2698
2699 sb->s_root = NULL;
2700
2701 needs_recovery = (es->s_last_orphan != 0 ||
617ba13b
MC		 2702 EXT4_HAS_INCOMPAT_FEATURE(sb,
2703 EXT4_FEATURE_INCOMPAT_RECOVER));
ac27a0ec
DK		 2704
2705 /*
2706 * The first inode we look at is the journal inode. Don't try
2707 * root first: it may be modified in the journal!
2708 */
2709 if (!test_opt(sb, NOLOAD) &&
617ba13b
MC		 2710 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2711 if (ext4_load_journal(sb, es, journal_devnum))
ac27a0ec 2712 goto failed_mount3;
624080ed
TT		 2713 if (!(sb->s_flags & MS_RDONLY) &&
2714 EXT4_SB(sb)->s_journal->j_failed_commit) {
b31e1552 2715 ext4_msg(sb, KERN_CRIT, "error: "
624080ed 2716 "ext4_fill_super: Journal transaction "
b31e1552 2717 "%u is corrupt",
624080ed 2718 EXT4_SB(sb)->s_journal->j_failed_commit);
2b2d6d01 2719 if (test_opt(sb, ERRORS_RO)) {
b31e1552
ES		 2720 ext4_msg(sb, KERN_CRIT,
2721 "Mounting filesystem read-only");
624080ed
TT		 2722 sb->s_flags |= MS_RDONLY;
2723 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2724 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2725 }
2726 if (test_opt(sb, ERRORS_PANIC)) {
2727 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2728 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
e2d67052 2729 ext4_commit_super(sb, 1);
624080ed
TT		 2730 goto failed_mount4;
2731 }
2732 }
0390131b
FM		 2733 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
2734 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
b31e1552
ES		 2735 ext4_msg(sb, KERN_ERR, "required journal recovery "
2736 "suppressed and not mounted read-only");
0390131b 2737 goto failed_mount4;
ac27a0ec 2738 } else {
0390131b
FM		 2739 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
2740 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
2741 sbi->s_journal = NULL;
2742 needs_recovery = 0;
2743 goto no_journal;
ac27a0ec
DK		 2744 }
2745
eb40a09c
JS		 2746 if (ext4_blocks_count(es) > 0xffffffffULL &&
2747 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2748 JBD2_FEATURE_INCOMPAT_64BIT)) {
b31e1552 2749 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
eb40a09c
JS		 2750 goto failed_mount4;
2751 }
2752
d0646f7b
TT		 2753 jbd2_journal_set_features(sbi->s_journal,
2754 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2755 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
2756 jbd2_journal_set_features(sbi->s_journal, 0, 0,
818d276c 2757 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
d0646f7b 2758 else
818d276c
GS		 2759 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2760 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
818d276c 2761
ac27a0ec
DK		 2762 /* We have now updated the journal if required, so we can
2763 * validate the data journaling mode. */
2764 switch (test_opt(sb, DATA_FLAGS)) {
2765 case 0:
2766 /* No mode set, assume a default based on the journal
63f57933
AM		 2767 * capabilities: ORDERED_DATA if the journal can
2768 * cope, else JOURNAL_DATA
2769 */
dab291af
MC		 2770 if (jbd2_journal_check_available_features
2771 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
ac27a0ec
DK		 2772 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2773 else
2774 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2775 break;
2776
617ba13b
MC		 2777 case EXT4_MOUNT_ORDERED_DATA:
2778 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 2779 if (!jbd2_journal_check_available_features
2780 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
b31e1552
ES		 2781 ext4_msg(sb, KERN_ERR, "Journal does not support "
2782 "requested data journaling mode");
ac27a0ec
DK		 2783 goto failed_mount4;
2784 }
2785 default:
2786 break;
2787 }
b3881f74 2788 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
ac27a0ec 2789
0390131b 2790no_journal:
ac27a0ec
DK		 2791
2792 if (test_opt(sb, NOBH)) {
617ba13b 2793 if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
b31e1552
ES		 2794 ext4_msg(sb, KERN_WARNING, "Ignoring nobh option - "
2795 "its supported only with writeback mode");
ac27a0ec
DK		 2796 clear_opt(sbi->s_mount_opt, NOBH);
2797 }
2798 }
2799 /*
dab291af 2800 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 2801 * so we can safely mount the rest of the filesystem now.
2802 */
2803
1d1fe1ee
DH		 2804 root = ext4_iget(sb, EXT4_ROOT_INO);
2805 if (IS_ERR(root)) {
b31e1552 2806 ext4_msg(sb, KERN_ERR, "get root inode failed");
1d1fe1ee 2807 ret = PTR_ERR(root);
ac27a0ec
DK		 2808 goto failed_mount4;
2809 }
2810 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1d1fe1ee 2811 iput(root);
b31e1552 2812 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
ac27a0ec
DK		 2813 goto failed_mount4;
2814 }
1d1fe1ee
DH		 2815 sb->s_root = d_alloc_root(root);
2816 if (!sb->s_root) {
b31e1552 2817 ext4_msg(sb, KERN_ERR, "get root dentry failed");
1d1fe1ee
DH		 2818 iput(root);
2819 ret = -ENOMEM;
2820 goto failed_mount4;
2821 }
ac27a0ec 2822
2b2d6d01 2823 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
ef7f3835
KS		 2824
2825 /* determine the minimum size of new large inodes, if present */
2826 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2827 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2828 EXT4_GOOD_OLD_INODE_SIZE;
2829 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2830 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2831 if (sbi->s_want_extra_isize <
2832 le16_to_cpu(es->s_want_extra_isize))
2833 sbi->s_want_extra_isize =
2834 le16_to_cpu(es->s_want_extra_isize);
2835 if (sbi->s_want_extra_isize <
2836 le16_to_cpu(es->s_min_extra_isize))
2837 sbi->s_want_extra_isize =
2838 le16_to_cpu(es->s_min_extra_isize);
2839 }
2840 }
2841 /* Check if enough inode space is available */
2842 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2843 sbi->s_inode_size) {
2844 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2845 EXT4_GOOD_OLD_INODE_SIZE;
b31e1552
ES		 2846 ext4_msg(sb, KERN_INFO, "required extra inode space not"
2847 "available");
ef7f3835
KS		 2848 }
2849
c2774d84 2850 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
b31e1552
ES		 2851 ext4_msg(sb, KERN_WARNING, "Ignoring delalloc option - "
2852 "requested data journaling mode");
c2774d84
AK		 2853 clear_opt(sbi->s_mount_opt, DELALLOC);
2854 } else if (test_opt(sb, DELALLOC))
b31e1552 2855 ext4_msg(sb, KERN_INFO, "delayed allocation enabled");
c2774d84 2856
6fd058f7
TT		 2857 err = ext4_setup_system_zone(sb);
2858 if (err) {
b31e1552
ES		 2859 ext4_msg(sb, KERN_ERR, "failed to initialize system "
2860 "zone (%d)\n", err);
6fd058f7
TT		 2861 goto failed_mount4;
2862 }
2863
c2774d84
AK		 2864 ext4_ext_init(sb);
2865 err = ext4_mb_init(sb, needs_recovery);
2866 if (err) {
b31e1552
ES		 2867 ext4_msg(sb, KERN_ERR, "failed to initalize mballoc (%d)",
2868 err);
c2774d84
AK		 2869 goto failed_mount4;
2870 }
2871
3197ebdb
TT		 2872 sbi->s_kobj.kset = ext4_kset;
2873 init_completion(&sbi->s_kobj_unregister);
2874 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
2875 "%s", sb->s_id);
2876 if (err) {
2877 ext4_mb_release(sb);
2878 ext4_ext_release(sb);
2879 goto failed_mount4;
2880 };
2881
617ba13b
MC		 2882 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
2883 ext4_orphan_cleanup(sb, es);
2884 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
0390131b 2885 if (needs_recovery) {
b31e1552 2886 ext4_msg(sb, KERN_INFO, "recovery complete");
0390131b
FM		 2887 ext4_mark_recovery_complete(sb, es);
2888 }
2889 if (EXT4_SB(sb)->s_journal) {
2890 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2891 descr = " journalled data mode";
2892 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2893 descr = " ordered data mode";
2894 else
2895 descr = " writeback data mode";
2896 } else
2897 descr = "out journal";
2898
b31e1552 2899 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s", descr);
ac27a0ec
DK		 2900
2901 lock_kernel();
2902 return 0;
2903
617ba13b 2904cantfind_ext4:
ac27a0ec 2905 if (!silent)
b31e1552 2906 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
ac27a0ec
DK		 2907 goto failed_mount;
2908
2909failed_mount4:
b31e1552 2910 ext4_msg(sb, KERN_ERR, "mount failed");
6fd058f7 2911 ext4_release_system_zone(sb);
0390131b
FM		 2912 if (sbi->s_journal) {
2913 jbd2_journal_destroy(sbi->s_journal);
2914 sbi->s_journal = NULL;
2915 }
ac27a0ec 2916failed_mount3:
c5ca7c76
TT		 2917 if (sbi->s_flex_groups) {
2918 if (is_vmalloc_addr(sbi->s_flex_groups))
2919 vfree(sbi->s_flex_groups);
2920 else
2921 kfree(sbi->s_flex_groups);
2922 }
ac27a0ec
DK		 2923 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2924 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2925 percpu_counter_destroy(&sbi->s_dirs_counter);
6bc6e63f 2926 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
ac27a0ec
DK		 2927failed_mount2:
2928 for (i = 0; i < db_count; i++)
2929 brelse(sbi->s_group_desc[i]);
2930 kfree(sbi->s_group_desc);
2931failed_mount:
240799cd 2932 if (sbi->s_proc) {
9f6200bb 2933 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 2934 }
ac27a0ec
DK		 2935#ifdef CONFIG_QUOTA
2936 for (i = 0; i < MAXQUOTAS; i++)
2937 kfree(sbi->s_qf_names[i]);
2938#endif
617ba13b 2939 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 2940 brelse(bh);
2941out_fail:
2942 sb->s_fs_info = NULL;
f6830165 2943 kfree(sbi->s_blockgroup_lock);
ac27a0ec
DK		 2944 kfree(sbi);
2945 lock_kernel();
1d1fe1ee 2946 return ret;
ac27a0ec
DK		 2947}
2948
2949/*
2950 * Setup any per-fs journal parameters now. We'll do this both on
2951 * initial mount, once the journal has been initialised but before we've
2952 * done any recovery; and again on any subsequent remount.
2953 */
617ba13b 2954static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 2955{
617ba13b 2956 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 2957
30773840
TT		 2958 journal->j_commit_interval = sbi->s_commit_interval;
2959 journal->j_min_batch_time = sbi->s_min_batch_time;
2960 journal->j_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 2961
2962 spin_lock(&journal->j_state_lock);
2963 if (test_opt(sb, BARRIER))
dab291af 2964 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 2965 else
dab291af 2966 journal->j_flags &= ~JBD2_BARRIER;
5bf5683a
HK		 2967 if (test_opt(sb, DATA_ERR_ABORT))
2968 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
2969 else
2970 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
ac27a0ec
DK		 2971 spin_unlock(&journal->j_state_lock);
2972}
2973
617ba13b 2974static journal_t *ext4_get_journal(struct super_block *sb,
ac27a0ec
DK		 2975 unsigned int journal_inum)
2976{
2977 struct inode *journal_inode;
2978 journal_t *journal;
2979
0390131b
FM		 2980 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
2981
ac27a0ec
DK		 2982 /* First, test for the existence of a valid inode on disk. Bad
2983 * things happen if we iget() an unused inode, as the subsequent
2984 * iput() will try to delete it. */
2985
1d1fe1ee
DH		 2986 journal_inode = ext4_iget(sb, journal_inum);
2987 if (IS_ERR(journal_inode)) {
b31e1552 2988 ext4_msg(sb, KERN_ERR, "no journal found");
ac27a0ec
DK		 2989 return NULL;
2990 }
2991 if (!journal_inode->i_nlink) {
2992 make_bad_inode(journal_inode);
2993 iput(journal_inode);
b31e1552 2994 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
ac27a0ec
DK		 2995 return NULL;
2996 }
2997
e5f8eab8 2998 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
ac27a0ec 2999 journal_inode, journal_inode->i_size);
1d1fe1ee 3000 if (!S_ISREG(journal_inode->i_mode)) {
b31e1552 3001 ext4_msg(sb, KERN_ERR, "invalid journal inode");
ac27a0ec
DK		 3002 iput(journal_inode);
3003 return NULL;
3004 }
3005
dab291af 3006 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 3007 if (!journal) {
b31e1552 3008 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
ac27a0ec
DK		 3009 iput(journal_inode);
3010 return NULL;
3011 }
3012 journal->j_private = sb;
617ba13b 3013 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 3014 return journal;
3015}
3016
617ba13b 3017static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 3018 dev_t j_dev)
3019{
2b2d6d01 3020 struct buffer_head *bh;
ac27a0ec 3021 journal_t *journal;
617ba13b
MC		 3022 ext4_fsblk_t start;
3023 ext4_fsblk_t len;
ac27a0ec 3024 int hblock, blocksize;
617ba13b 3025 ext4_fsblk_t sb_block;
ac27a0ec 3026 unsigned long offset;
2b2d6d01 3027 struct ext4_super_block *es;
ac27a0ec
DK		 3028 struct block_device *bdev;
3029
0390131b
FM		 3030 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3031
b31e1552 3032 bdev = ext4_blkdev_get(j_dev, sb);
ac27a0ec
DK		 3033 if (bdev == NULL)
3034 return NULL;
3035
3036 if (bd_claim(bdev, sb)) {
b31e1552
ES		 3037 ext4_msg(sb, KERN_ERR,
3038 "failed to claim external journal device");
9a1c3542 3039 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
ac27a0ec
DK		 3040 return NULL;
3041 }
3042
3043 blocksize = sb->s_blocksize;
e1defc4f 3044 hblock = bdev_logical_block_size(bdev);
ac27a0ec 3045 if (blocksize < hblock) {
b31e1552
ES		 3046 ext4_msg(sb, KERN_ERR,
3047 "blocksize too small for journal device");
ac27a0ec
DK		 3048 goto out_bdev;
3049 }
3050
617ba13b
MC		 3051 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3052 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 3053 set_blocksize(bdev, blocksize);
3054 if (!(bh = __bread(bdev, sb_block, blocksize))) {
b31e1552
ES		 3055 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
3056 "external journal");
ac27a0ec
DK		 3057 goto out_bdev;
3058 }
3059
617ba13b
MC		 3060 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3061 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 3062 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b 3063 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
b31e1552
ES		 3064 ext4_msg(sb, KERN_ERR, "external journal has "
3065 "bad superblock");
ac27a0ec
DK		 3066 brelse(bh);
3067 goto out_bdev;
3068 }
3069
617ba13b 3070 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
b31e1552 3071 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
ac27a0ec
DK		 3072 brelse(bh);
3073 goto out_bdev;
3074 }
3075
bd81d8ee 3076 len = ext4_blocks_count(es);
ac27a0ec
DK		 3077 start = sb_block + 1;
3078 brelse(bh); /* we're done with the superblock */
3079
dab291af 3080 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 3081 start, len, blocksize);
3082 if (!journal) {
b31e1552 3083 ext4_msg(sb, KERN_ERR, "failed to create device journal");
ac27a0ec
DK		 3084 goto out_bdev;
3085 }
3086 journal->j_private = sb;
3087 ll_rw_block(READ, 1, &journal->j_sb_buffer);
3088 wait_on_buffer(journal->j_sb_buffer);
3089 if (!buffer_uptodate(journal->j_sb_buffer)) {
b31e1552 3090 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
ac27a0ec
DK		 3091 goto out_journal;
3092 }
3093 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
b31e1552
ES		 3094 ext4_msg(sb, KERN_ERR, "External journal has more than one "
3095 "user (unsupported) - %d",
ac27a0ec
DK		 3096 be32_to_cpu(journal->j_superblock->s_nr_users));
3097 goto out_journal;
3098 }
617ba13b
MC		 3099 EXT4_SB(sb)->journal_bdev = bdev;
3100 ext4_init_journal_params(sb, journal);
ac27a0ec 3101 return journal;
0b8e58a1 3102
ac27a0ec 3103out_journal:
dab291af 3104 jbd2_journal_destroy(journal);
ac27a0ec 3105out_bdev:
617ba13b 3106 ext4_blkdev_put(bdev);
ac27a0ec
DK		 3107 return NULL;
3108}
3109
617ba13b
MC		 3110static int ext4_load_journal(struct super_block *sb,
3111 struct ext4_super_block *es,
ac27a0ec
DK		 3112 unsigned long journal_devnum)
3113{
3114 journal_t *journal;
3115 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
3116 dev_t journal_dev;
3117 int err = 0;
3118 int really_read_only;
3119
0390131b
FM		 3120 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3121
ac27a0ec
DK		 3122 if (journal_devnum &&
3123 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
b31e1552
ES		 3124 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
3125 "numbers have changed");
ac27a0ec
DK		 3126 journal_dev = new_decode_dev(journal_devnum);
3127 } else
3128 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
3129
3130 really_read_only = bdev_read_only(sb->s_bdev);
3131
3132 /*
3133 * Are we loading a blank journal or performing recovery after a
3134 * crash? For recovery, we need to check in advance whether we
3135 * can get read-write access to the device.
3136 */
617ba13b 3137 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
ac27a0ec 3138 if (sb->s_flags & MS_RDONLY) {
b31e1552
ES		 3139 ext4_msg(sb, KERN_INFO, "INFO: recovery "
3140 "required on readonly filesystem");
ac27a0ec 3141 if (really_read_only) {
b31e1552
ES		 3142 ext4_msg(sb, KERN_ERR, "write access "
3143 "unavailable, cannot proceed");
ac27a0ec
DK		 3144 return -EROFS;
3145 }
b31e1552
ES		 3146 ext4_msg(sb, KERN_INFO, "write access will "
3147 "be enabled during recovery");
ac27a0ec
DK		 3148 }
3149 }
3150
3151 if (journal_inum && journal_dev) {
b31e1552
ES		 3152 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
3153 "and inode journals!");
ac27a0ec
DK		 3154 return -EINVAL;
3155 }
3156
3157 if (journal_inum) {
617ba13b 3158 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 3159 return -EINVAL;
3160 } else {
617ba13b 3161 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 3162 return -EINVAL;
3163 }
3164
4776004f 3165 if (journal->j_flags & JBD2_BARRIER)
b31e1552 3166 ext4_msg(sb, KERN_INFO, "barriers enabled");
4776004f 3167 else
b31e1552 3168 ext4_msg(sb, KERN_INFO, "barriers disabled");
4776004f 3169
ac27a0ec 3170 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
dab291af 3171 err = jbd2_journal_update_format(journal);
ac27a0ec 3172 if (err) {
b31e1552 3173 ext4_msg(sb, KERN_ERR, "error updating journal");
dab291af 3174 jbd2_journal_destroy(journal);
ac27a0ec
DK		 3175 return err;
3176 }
3177 }
3178
617ba13b 3179 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
dab291af 3180 err = jbd2_journal_wipe(journal, !really_read_only);
ac27a0ec 3181 if (!err)
dab291af 3182 err = jbd2_journal_load(journal);
ac27a0ec
DK		 3183
3184 if (err) {
b31e1552 3185 ext4_msg(sb, KERN_ERR, "error loading journal");
dab291af 3186 jbd2_journal_destroy(journal);
ac27a0ec
DK		 3187 return err;
3188 }
3189
617ba13b
MC		 3190 EXT4_SB(sb)->s_journal = journal;
3191 ext4_clear_journal_err(sb, es);
ac27a0ec
DK		 3192
3193 if (journal_devnum &&
3194 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3195 es->s_journal_dev = cpu_to_le32(journal_devnum);
ac27a0ec
DK		 3196
3197 /* Make sure we flush the recovery flag to disk. */
e2d67052 3198 ext4_commit_super(sb, 1);
ac27a0ec
DK		 3199 }
3200
3201 return 0;
3202}
3203
e2d67052 3204static int ext4_commit_super(struct super_block *sb, int sync)
ac27a0ec 3205{
e2d67052 3206 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
617ba13b 3207 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
c4be0c1d 3208 int error = 0;
ac27a0ec
DK		 3209
3210 if (!sbh)
c4be0c1d 3211 return error;
914258bf
TT		 3212 if (buffer_write_io_error(sbh)) {
3213 /*
3214 * Oh, dear. A previous attempt to write the
3215 * superblock failed. This could happen because the
3216 * USB device was yanked out. Or it could happen to
3217 * be a transient write error and maybe the block will
3218 * be remapped. Nothing we can do but to retry the
3219 * write and hope for the best.
3220 */
b31e1552
ES		 3221 ext4_msg(sb, KERN_ERR, "previous I/O error to "
3222 "superblock detected");
914258bf
TT		 3223 clear_buffer_write_io_error(sbh);
3224 set_buffer_uptodate(sbh);
3225 }
71290b36
TT		 3226 /*
3227 * If the file system is mounted read-only, don't update the
3228 * superblock write time. This avoids updating the superblock
3229 * write time when we are mounting the root file system
3230 * read/only but we need to replay the journal; at that point,
3231 * for people who are east of GMT and who make their clock
3232 * tick in localtime for Windows bug-for-bug compatibility,
3233 * the clock is set in the future, and this will cause e2fsck
3234 * to complain and force a full file system check.
3235 */
3236 if (!(sb->s_flags & MS_RDONLY))
3237 es->s_wtime = cpu_to_le32(get_seconds());
afc32f7e
TT		 3238 es->s_kbytes_written =
3239 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
3240 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
3241 EXT4_SB(sb)->s_sectors_written_start) >> 1));
5d1b1b3f
AK		 3242 ext4_free_blocks_count_set(es, percpu_counter_sum_positive(
3243 &EXT4_SB(sb)->s_freeblocks_counter));
3244 es->s_free_inodes_count = cpu_to_le32(percpu_counter_sum_positive(
3245 &EXT4_SB(sb)->s_freeinodes_counter));
7234ab2a 3246 sb->s_dirt = 0;
ac27a0ec
DK		 3247 BUFFER_TRACE(sbh, "marking dirty");
3248 mark_buffer_dirty(sbh);
914258bf 3249 if (sync) {
c4be0c1d
TS		 3250 error = sync_dirty_buffer(sbh);
3251 if (error)
3252 return error;
3253
3254 error = buffer_write_io_error(sbh);
3255 if (error) {
b31e1552
ES		 3256 ext4_msg(sb, KERN_ERR, "I/O error while writing "
3257 "superblock");
914258bf
TT		 3258 clear_buffer_write_io_error(sbh);
3259 set_buffer_uptodate(sbh);
3260 }
3261 }
c4be0c1d 3262 return error;
ac27a0ec
DK		 3263}
3264
ac27a0ec
DK		 3265/*
3266 * Have we just finished recovery? If so, and if we are mounting (or
3267 * remounting) the filesystem readonly, then we will end up with a
3268 * consistent fs on disk. Record that fact.
3269 */
2b2d6d01
TT		 3270static void ext4_mark_recovery_complete(struct super_block *sb,
3271 struct ext4_super_block *es)
ac27a0ec 3272{
617ba13b 3273 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 3274
0390131b
FM		 3275 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3276 BUG_ON(journal != NULL);
3277 return;
3278 }
dab291af 3279 jbd2_journal_lock_updates(journal);
7ffe1ea8
HK		 3280 if (jbd2_journal_flush(journal) < 0)
3281 goto out;
3282
617ba13b 3283 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
ac27a0ec 3284 sb->s_flags & MS_RDONLY) {
617ba13b 3285 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
e2d67052 3286 ext4_commit_super(sb, 1);
ac27a0ec 3287 }
7ffe1ea8
HK		 3288
3289out:
dab291af 3290 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 3291}
3292
3293/*
3294 * If we are mounting (or read-write remounting) a filesystem whose journal
3295 * has recorded an error from a previous lifetime, move that error to the
3296 * main filesystem now.
3297 */
2b2d6d01
TT		 3298static void ext4_clear_journal_err(struct super_block *sb,
3299 struct ext4_super_block *es)
ac27a0ec
DK		 3300{
3301 journal_t *journal;
3302 int j_errno;
3303 const char *errstr;
3304
0390131b
FM		 3305 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3306
617ba13b 3307 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 3308
3309 /*
3310 * Now check for any error status which may have been recorded in the
617ba13b 3311 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 3312 */
3313
dab291af 3314 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 3315 if (j_errno) {
3316 char nbuf[16];
3317
617ba13b 3318 errstr = ext4_decode_error(sb, j_errno, nbuf);
46e665e9 3319 ext4_warning(sb, __func__, "Filesystem error recorded "
ac27a0ec 3320 "from previous mount: %s", errstr);
46e665e9 3321 ext4_warning(sb, __func__, "Marking fs in need of "
ac27a0ec
DK		 3322 "filesystem check.");
3323
617ba13b
MC		 3324 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
3325 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
e2d67052 3326 ext4_commit_super(sb, 1);
ac27a0ec 3327
dab291af 3328 jbd2_journal_clear_err(journal);
ac27a0ec
DK		 3329 }
3330}
3331
3332/*
3333 * Force the running and committing transactions to commit,
3334 * and wait on the commit.
3335 */
617ba13b 3336int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 3337{
3338 journal_t *journal;
0390131b 3339 int ret = 0;
ac27a0ec
DK		 3340
3341 if (sb->s_flags & MS_RDONLY)
3342 return 0;
3343
617ba13b 3344 journal = EXT4_SB(sb)->s_journal;
7234ab2a 3345 if (journal)
0390131b 3346 ret = ext4_journal_force_commit(journal);
0390131b 3347
ac27a0ec
DK		 3348 return ret;
3349}
3350
2b2d6d01 3351static void ext4_write_super(struct super_block *sb)
ac27a0ec 3352{
ebc1ac16 3353 lock_super(sb);
9ca92389 3354 ext4_commit_super(sb, 1);
ebc1ac16 3355 unlock_super(sb);
ac27a0ec
DK		 3356}
3357
617ba13b 3358static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec 3359{
14ce0cb4 3360 int ret = 0;
9eddacf9 3361 tid_t target;
ac27a0ec 3362
9bffad1e 3363 trace_ext4_sync_fs(sb, wait);
9ca92389
TT		 3364 if (jbd2_journal_start_commit(EXT4_SB(sb)->s_journal, &target)) {
3365 if (wait)
3366 jbd2_log_wait_commit(EXT4_SB(sb)->s_journal, target);
0390131b 3367 }
14ce0cb4 3368 return ret;
ac27a0ec
DK		 3369}
3370
3371/*
3372 * LVM calls this function before a (read-only) snapshot is created. This
3373 * gives us a chance to flush the journal completely and mark the fs clean.
3374 */
c4be0c1d 3375static int ext4_freeze(struct super_block *sb)
ac27a0ec 3376{
c4be0c1d
TS		 3377 int error = 0;
3378 journal_t *journal;
ac27a0ec 3379
9ca92389
TT		 3380 if (sb->s_flags & MS_RDONLY)
3381 return 0;
ac27a0ec 3382
9ca92389 3383 journal = EXT4_SB(sb)->s_journal;
7ffe1ea8 3384
9ca92389
TT		 3385 /* Now we set up the journal barrier. */
3386 jbd2_journal_lock_updates(journal);
ac27a0ec 3387
9ca92389
TT		 3388 /*
3389 * Don't clear the needs_recovery flag if we failed to flush
3390 * the journal.
3391 */
3392 error = jbd2_journal_flush(journal);
3393 if (error < 0) {
3394 out:
3395 jbd2_journal_unlock_updates(journal);
3396 return error;
ac27a0ec 3397 }
9ca92389
TT		 3398
3399 /* Journal blocked and flushed, clear needs_recovery flag. */
3400 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3401 error = ext4_commit_super(sb, 1);
3402 if (error)
3403 goto out;
c4be0c1d 3404 return 0;
ac27a0ec
DK		 3405}
3406
3407/*
3408 * Called by LVM after the snapshot is done. We need to reset the RECOVER
3409 * flag here, even though the filesystem is not technically dirty yet.
3410 */
c4be0c1d 3411static int ext4_unfreeze(struct super_block *sb)
ac27a0ec 3412{
9ca92389
TT		 3413 if (sb->s_flags & MS_RDONLY)
3414 return 0;
3415
3416 lock_super(sb);
3417 /* Reset the needs_recovery flag before the fs is unlocked. */
3418 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3419 ext4_commit_super(sb, 1);
3420 unlock_super(sb);
3421 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
c4be0c1d 3422 return 0;
ac27a0ec
DK		 3423}
3424
2b2d6d01 3425static int ext4_remount(struct super_block *sb, int *flags, char *data)
ac27a0ec 3426{
2b2d6d01 3427 struct ext4_super_block *es;
617ba13b
MC		 3428 struct ext4_sb_info *sbi = EXT4_SB(sb);
3429 ext4_fsblk_t n_blocks_count = 0;
ac27a0ec 3430 unsigned long old_sb_flags;
617ba13b 3431 struct ext4_mount_options old_opts;
8a266467 3432 ext4_group_t g;
b3881f74 3433 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
ac27a0ec
DK		 3434 int err;
3435#ifdef CONFIG_QUOTA
3436 int i;
3437#endif
3438
337eb00a
AIB		 3439 lock_kernel();
3440
ac27a0ec 3441 /* Store the original options */
bbd6851a 3442 lock_super(sb);
ac27a0ec
DK		 3443 old_sb_flags = sb->s_flags;
3444 old_opts.s_mount_opt = sbi->s_mount_opt;
3445 old_opts.s_resuid = sbi->s_resuid;
3446 old_opts.s_resgid = sbi->s_resgid;
3447 old_opts.s_commit_interval = sbi->s_commit_interval;
30773840
TT		 3448 old_opts.s_min_batch_time = sbi->s_min_batch_time;
3449 old_opts.s_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 3450#ifdef CONFIG_QUOTA
3451 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
3452 for (i = 0; i < MAXQUOTAS; i++)
3453 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
3454#endif
b3881f74
TT		 3455 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
3456 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
ac27a0ec
DK		 3457
3458 /*
3459 * Allow the "check" option to be passed as a remount option.
3460 */
b3881f74
TT		 3461 if (!parse_options(data, sb, NULL, &journal_ioprio,
3462 &n_blocks_count, 1)) {
ac27a0ec
DK		 3463 err = -EINVAL;
3464 goto restore_opts;
3465 }
3466
4ab2f15b 3467 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
46e665e9 3468 ext4_abort(sb, __func__, "Abort forced by user");
ac27a0ec
DK		 3469
3470 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
617ba13b 3471 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec
DK		 3472
3473 es = sbi->s_es;
3474
b3881f74 3475 if (sbi->s_journal) {
0390131b 3476 ext4_init_journal_params(sb, sbi->s_journal);
b3881f74
TT		 3477 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3478 }
ac27a0ec
DK		 3479
3480 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
bd81d8ee 3481 n_blocks_count > ext4_blocks_count(es)) {
4ab2f15b 3482 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
ac27a0ec
DK		 3483 err = -EROFS;
3484 goto restore_opts;
3485 }
3486
3487 if (*flags & MS_RDONLY) {
3488 /*
3489 * First of all, the unconditional stuff we have to do
3490 * to disable replay of the journal when we next remount
3491 */
3492 sb->s_flags |= MS_RDONLY;
3493
3494 /*
3495 * OK, test if we are remounting a valid rw partition
3496 * readonly, and if so set the rdonly flag and then
3497 * mark the partition as valid again.
3498 */
617ba13b
MC		 3499 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
3500 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 3501 es->s_state = cpu_to_le16(sbi->s_mount_state);
3502
a63c9eb2 3503 if (sbi->s_journal)
0390131b 3504 ext4_mark_recovery_complete(sb, es);
ac27a0ec 3505 } else {
a13fb1a4
ES		 3506 /* Make sure we can mount this feature set readwrite */
3507 if (!ext4_feature_set_ok(sb, 0)) {
ac27a0ec
DK		 3508 err = -EROFS;
3509 goto restore_opts;
3510 }
8a266467
TT		 3511 /*
3512 * Make sure the group descriptor checksums
0b8e58a1 3513 * are sane. If they aren't, refuse to remount r/w.
8a266467
TT		 3514 */
3515 for (g = 0; g < sbi->s_groups_count; g++) {
3516 struct ext4_group_desc *gdp =
3517 ext4_get_group_desc(sb, g, NULL);
3518
3519 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
b31e1552
ES		 3520 ext4_msg(sb, KERN_ERR,
3521 "ext4_remount: Checksum for group %u failed (%u!=%u)",
8a266467
TT		 3522 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
3523 le16_to_cpu(gdp->bg_checksum));
3524 err = -EINVAL;
3525 goto restore_opts;
3526 }
3527 }
3528
ead6596b
ES		 3529 /*
3530 * If we have an unprocessed orphan list hanging
3531 * around from a previously readonly bdev mount,
3532 * require a full umount/remount for now.
3533 */
3534 if (es->s_last_orphan) {
b31e1552 3535 ext4_msg(sb, KERN_WARNING, "Couldn't "
ead6596b
ES		 3536 "remount RDWR because of unprocessed "
3537 "orphan inode list. Please "
b31e1552 3538 "umount/remount instead");
ead6596b
ES		 3539 err = -EINVAL;
3540 goto restore_opts;
3541 }
3542
ac27a0ec
DK		 3543 /*
3544 * Mounting a RDONLY partition read-write, so reread
3545 * and store the current valid flag. (It may have
3546 * been changed by e2fsck since we originally mounted
3547 * the partition.)
3548 */
0390131b
FM		 3549 if (sbi->s_journal)
3550 ext4_clear_journal_err(sb, es);
ac27a0ec 3551 sbi->s_mount_state = le16_to_cpu(es->s_state);
617ba13b 3552 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
ac27a0ec 3553 goto restore_opts;
2b2d6d01 3554 if (!ext4_setup_super(sb, es, 0))
ac27a0ec
DK		 3555 sb->s_flags &= ~MS_RDONLY;
3556 }
3557 }
6fd058f7 3558 ext4_setup_system_zone(sb);
0390131b 3559 if (sbi->s_journal == NULL)
e2d67052 3560 ext4_commit_super(sb, 1);
0390131b 3561
ac27a0ec
DK		 3562#ifdef CONFIG_QUOTA
3563 /* Release old quota file names */
3564 for (i = 0; i < MAXQUOTAS; i++)
3565 if (old_opts.s_qf_names[i] &&
3566 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3567 kfree(old_opts.s_qf_names[i]);
3568#endif
bbd6851a 3569 unlock_super(sb);
337eb00a 3570 unlock_kernel();
ac27a0ec 3571 return 0;
0b8e58a1 3572
ac27a0ec
DK		 3573restore_opts:
3574 sb->s_flags = old_sb_flags;
3575 sbi->s_mount_opt = old_opts.s_mount_opt;
3576 sbi->s_resuid = old_opts.s_resuid;
3577 sbi->s_resgid = old_opts.s_resgid;
3578 sbi->s_commit_interval = old_opts.s_commit_interval;
30773840
TT		 3579 sbi->s_min_batch_time = old_opts.s_min_batch_time;
3580 sbi->s_max_batch_time = old_opts.s_max_batch_time;
ac27a0ec
DK		 3581#ifdef CONFIG_QUOTA
3582 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
3583 for (i = 0; i < MAXQUOTAS; i++) {
3584 if (sbi->s_qf_names[i] &&
3585 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3586 kfree(sbi->s_qf_names[i]);
3587 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
3588 }
3589#endif
bbd6851a 3590 unlock_super(sb);
337eb00a 3591 unlock_kernel();
ac27a0ec
DK		 3592 return err;
3593}
3594
2b2d6d01 3595static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
ac27a0ec
DK		 3596{
3597 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 3598 struct ext4_sb_info *sbi = EXT4_SB(sb);
3599 struct ext4_super_block *es = sbi->s_es;
960cc398 3600 u64 fsid;
ac27a0ec 3601
5e70030d
BP		 3602 if (test_opt(sb, MINIX_DF)) {
3603 sbi->s_overhead_last = 0;
6bc9feff 3604 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
8df9675f 3605 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
5e70030d 3606 ext4_fsblk_t overhead = 0;
ac27a0ec
DK		 3607
3608 /*
5e70030d
BP		 3609 * Compute the overhead (FS structures). This is constant
3610 * for a given filesystem unless the number of block groups
3611 * changes so we cache the previous value until it does.
ac27a0ec
DK		 3612 */
3613
3614 /*
3615 * All of the blocks before first_data_block are
3616 * overhead
3617 */
3618 overhead = le32_to_cpu(es->s_first_data_block);
3619
3620 /*
3621 * Add the overhead attributed to the superblock and
3622 * block group descriptors. If the sparse superblocks
3623 * feature is turned on, then not all groups have this.
3624 */
3625 for (i = 0; i < ngroups; i++) {
617ba13b
MC		 3626 overhead += ext4_bg_has_super(sb, i) +
3627 ext4_bg_num_gdb(sb, i);
ac27a0ec
DK		 3628 cond_resched();
3629 }
3630
3631 /*
3632 * Every block group has an inode bitmap, a block
3633 * bitmap, and an inode table.
3634 */
5e70030d
BP		 3635 overhead += ngroups * (2 + sbi->s_itb_per_group);
3636 sbi->s_overhead_last = overhead;
3637 smp_wmb();
6bc9feff 3638 sbi->s_blocks_last = ext4_blocks_count(es);
ac27a0ec
DK		 3639 }
3640
617ba13b 3641 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 3642 buf->f_bsize = sb->s_blocksize;
5e70030d 3643 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
6bc6e63f
AK		 3644 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
3645 percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
308ba3ec 3646 ext4_free_blocks_count_set(es, buf->f_bfree);
bd81d8ee
LV		 3647 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3648 if (buf->f_bfree < ext4_r_blocks_count(es))
ac27a0ec
DK		 3649 buf->f_bavail = 0;
3650 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 3651 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
5e70030d 3652 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
617ba13b 3653 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 3654 fsid = le64_to_cpup((void *)es->s_uuid) ^
3655 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3656 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3657 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
0b8e58a1 3658
ac27a0ec
DK		 3659 return 0;
3660}
3661
0b8e58a1
AD		 3662/* Helper function for writing quotas on sync - we need to start transaction
3663 * before quota file is locked for write. Otherwise the are possible deadlocks:
ac27a0ec 3664 * Process 1 Process 2
617ba13b 3665 * ext4_create() quota_sync()
a269eb18
JK		 3666 * jbd2_journal_start() write_dquot()
3667 * vfs_dq_init() down(dqio_mutex)
dab291af 3668 * down(dqio_mutex) jbd2_journal_start()
ac27a0ec
DK		 3669 *
3670 */
3671
3672#ifdef CONFIG_QUOTA
3673
3674static inline struct inode *dquot_to_inode(struct dquot *dquot)
3675{
3676 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3677}
3678
617ba13b 3679static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 3680{
3681 int ret, err;
3682 handle_t *handle;
3683 struct inode *inode;
3684
3685 inode = dquot_to_inode(dquot);
617ba13b 3686 handle = ext4_journal_start(inode,
0b8e58a1 3687 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 3688 if (IS_ERR(handle))
3689 return PTR_ERR(handle);
3690 ret = dquot_commit(dquot);
617ba13b 3691 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3692 if (!ret)
3693 ret = err;
3694 return ret;
3695}
3696
617ba13b 3697static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 3698{
3699 int ret, err;
3700 handle_t *handle;
3701
617ba13b 3702 handle = ext4_journal_start(dquot_to_inode(dquot),
0b8e58a1 3703 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 3704 if (IS_ERR(handle))
3705 return PTR_ERR(handle);
3706 ret = dquot_acquire(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_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 3714{
3715 int ret, err;
3716 handle_t *handle;
3717
617ba13b 3718 handle = ext4_journal_start(dquot_to_inode(dquot),
0b8e58a1 3719 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 3720 if (IS_ERR(handle)) {
3721 /* Release dquot anyway to avoid endless cycle in dqput() */
3722 dquot_release(dquot);
ac27a0ec 3723 return PTR_ERR(handle);
9c3013e9 3724 }
ac27a0ec 3725 ret = dquot_release(dquot);
617ba13b 3726 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3727 if (!ret)
3728 ret = err;
3729 return ret;
3730}
3731
617ba13b 3732static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec 3733{
2c8be6b2 3734 /* Are we journaling quotas? */
617ba13b
MC		 3735 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3736 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
ac27a0ec 3737 dquot_mark_dquot_dirty(dquot);
617ba13b 3738 return ext4_write_dquot(dquot);
ac27a0ec
DK		 3739 } else {
3740 return dquot_mark_dquot_dirty(dquot);
3741 }
3742}
3743
617ba13b 3744static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 3745{
3746 int ret, err;
3747 handle_t *handle;
3748
3749 /* Data block + inode block */
617ba13b 3750 handle = ext4_journal_start(sb->s_root->d_inode, 2);
ac27a0ec
DK		 3751 if (IS_ERR(handle))
3752 return PTR_ERR(handle);
3753 ret = dquot_commit_info(sb, type);
617ba13b 3754 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3755 if (!ret)
3756 ret = err;
3757 return ret;
3758}
3759
3760/*
3761 * Turn on quotas during mount time - we need to find
3762 * the quota file and such...
3763 */
617ba13b 3764static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 3765{
617ba13b 3766 return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
0b8e58a1 3767 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 3768}
3769
3770/*
3771 * Standard function to be called on quota_on
3772 */
617ba13b 3773static int ext4_quota_on(struct super_block *sb, int type, int format_id,
8264613d 3774 char *name, int remount)
ac27a0ec
DK		 3775{
3776 int err;
8264613d 3777 struct path path;
ac27a0ec
DK		 3778
3779 if (!test_opt(sb, QUOTA))
3780 return -EINVAL;
8264613d 3781 /* When remounting, no checks are needed and in fact, name is NULL */
0623543b 3782 if (remount)
8264613d 3783 return vfs_quota_on(sb, type, format_id, name, remount);
0623543b 3784
8264613d 3785 err = kern_path(name, LOOKUP_FOLLOW, &path);
ac27a0ec
DK		 3786 if (err)
3787 return err;
0623543b 3788
ac27a0ec 3789 /* Quotafile not on the same filesystem? */
8264613d
AV		 3790 if (path.mnt->mnt_sb != sb) {
3791 path_put(&path);
ac27a0ec
DK		 3792 return -EXDEV;
3793 }
0623543b
JK		 3794 /* Journaling quota? */
3795 if (EXT4_SB(sb)->s_qf_names[type]) {
2b2d6d01 3796 /* Quotafile not in fs root? */
8264613d 3797 if (path.dentry->d_parent != sb->s_root)
b31e1552
ES		 3798 ext4_msg(sb, KERN_WARNING,
3799 "Quota file not on filesystem root. "
3800 "Journaled quota will not work");
2b2d6d01 3801 }
0623543b
JK		 3802
3803 /*
3804 * When we journal data on quota file, we have to flush journal to see
3805 * all updates to the file when we bypass pagecache...
3806 */
0390131b
FM		 3807 if (EXT4_SB(sb)->s_journal &&
3808 ext4_should_journal_data(path.dentry->d_inode)) {
0623543b
JK		 3809 /*
3810 * We don't need to lock updates but journal_flush() could
3811 * otherwise be livelocked...
3812 */
3813 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 3814 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
0623543b 3815 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 3816 if (err) {
8264613d 3817 path_put(&path);
7ffe1ea8
HK		 3818 return err;
3819 }
0623543b
JK		 3820 }
3821
8264613d
AV		 3822 err = vfs_quota_on_path(sb, type, format_id, &path);
3823 path_put(&path);
77e69dac 3824 return err;
ac27a0ec
DK		 3825}
3826
3827/* Read data from quotafile - avoid pagecache and such because we cannot afford
3828 * acquiring the locks... As quota files are never truncated and quota code
3829 * itself serializes the operations (and noone else should touch the files)
3830 * we don't have to be afraid of races */
617ba13b 3831static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 3832 size_t len, loff_t off)
3833{
3834 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 3835 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 3836 int err = 0;
3837 int offset = off & (sb->s_blocksize - 1);
3838 int tocopy;
3839 size_t toread;
3840 struct buffer_head *bh;
3841 loff_t i_size = i_size_read(inode);
3842
3843 if (off > i_size)
3844 return 0;
3845 if (off+len > i_size)
3846 len = i_size-off;
3847 toread = len;
3848 while (toread > 0) {
3849 tocopy = sb->s_blocksize - offset < toread ?
3850 sb->s_blocksize - offset : toread;
617ba13b 3851 bh = ext4_bread(NULL, inode, blk, 0, &err);
ac27a0ec
DK		 3852 if (err)
3853 return err;
3854 if (!bh) /* A hole? */
3855 memset(data, 0, tocopy);
3856 else
3857 memcpy(data, bh->b_data+offset, tocopy);
3858 brelse(bh);
3859 offset = 0;
3860 toread -= tocopy;
3861 data += tocopy;
3862 blk++;
3863 }
3864 return len;
3865}
3866
3867/* Write to quotafile (we know the transaction is already started and has
3868 * enough credits) */
617ba13b 3869static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 3870 const char *data, size_t len, loff_t off)
3871{
3872 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 3873 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 3874 int err = 0;
3875 int offset = off & (sb->s_blocksize - 1);
3876 int tocopy;
617ba13b 3877 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
ac27a0ec
DK		 3878 size_t towrite = len;
3879 struct buffer_head *bh;
3880 handle_t *handle = journal_current_handle();
3881
0390131b 3882 if (EXT4_SB(sb)->s_journal && !handle) {
b31e1552
ES		 3883 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
3884 " cancelled because transaction is not started",
9c3013e9
JK		 3885 (unsigned long long)off, (unsigned long long)len);
3886 return -EIO;
3887 }
ac27a0ec
DK		 3888 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
3889 while (towrite > 0) {
3890 tocopy = sb->s_blocksize - offset < towrite ?
3891 sb->s_blocksize - offset : towrite;
617ba13b 3892 bh = ext4_bread(handle, inode, blk, 1, &err);
ac27a0ec
DK		 3893 if (!bh)
3894 goto out;
3895 if (journal_quota) {
617ba13b 3896 err = ext4_journal_get_write_access(handle, bh);
ac27a0ec
DK		 3897 if (err) {
3898 brelse(bh);
3899 goto out;
3900 }
3901 }
3902 lock_buffer(bh);
3903 memcpy(bh->b_data+offset, data, tocopy);
3904 flush_dcache_page(bh->b_page);
3905 unlock_buffer(bh);
3906 if (journal_quota)
0390131b 3907 err = ext4_handle_dirty_metadata(handle, NULL, bh);
ac27a0ec
DK		 3908 else {
3909 /* Always do at least ordered writes for quotas */
678aaf48 3910 err = ext4_jbd2_file_inode(handle, inode);
ac27a0ec
DK		 3911 mark_buffer_dirty(bh);
3912 }
3913 brelse(bh);
3914 if (err)
3915 goto out;
3916 offset = 0;
3917 towrite -= tocopy;
3918 data += tocopy;
3919 blk++;
3920 }
3921out:
4d04e4fb
JK		 3922 if (len == towrite) {
3923 mutex_unlock(&inode->i_mutex);
ac27a0ec 3924 return err;
4d04e4fb 3925 }
ac27a0ec
DK		 3926 if (inode->i_size < off+len-towrite) {
3927 i_size_write(inode, off+len-towrite);
617ba13b 3928 EXT4_I(inode)->i_disksize = inode->i_size;
ac27a0ec 3929 }
ac27a0ec 3930 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
617ba13b 3931 ext4_mark_inode_dirty(handle, inode);
ac27a0ec
DK		 3932 mutex_unlock(&inode->i_mutex);
3933 return len - towrite;
3934}
3935
3936#endif
3937
0b8e58a1
AD		 3938static int ext4_get_sb(struct file_system_type *fs_type, int flags,
3939 const char *dev_name, void *data, struct vfsmount *mnt)
ac27a0ec 3940{
0b8e58a1 3941 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super,mnt);
ac27a0ec
DK		 3942}
3943
03010a33
TT		 3944static struct file_system_type ext4_fs_type = {
3945 .owner = THIS_MODULE,
3946 .name = "ext4",
3947 .get_sb = ext4_get_sb,
3948 .kill_sb = kill_block_super,
3949 .fs_flags = FS_REQUIRES_DEV,
3950};
3951
3952#ifdef CONFIG_EXT4DEV_COMPAT
0b8e58a1
AD		 3953static int ext4dev_get_sb(struct file_system_type *fs_type, int flags,
3954 const char *dev_name, void *data,struct vfsmount *mnt)
03010a33 3955{
b31e1552
ES		 3956 printk(KERN_WARNING "EXT4-fs (%s): Update your userspace programs "
3957 "to mount using ext4\n", dev_name);
3958 printk(KERN_WARNING "EXT4-fs (%s): ext4dev backwards compatibility "
3959 "will go away by 2.6.31\n", dev_name);
0b8e58a1 3960 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super,mnt);
03010a33
TT		 3961}
3962
617ba13b 3963static struct file_system_type ext4dev_fs_type = {
ac27a0ec 3964 .owner = THIS_MODULE,
617ba13b 3965 .name = "ext4dev",
03010a33 3966 .get_sb = ext4dev_get_sb,
ac27a0ec
DK		 3967 .kill_sb = kill_block_super,
3968 .fs_flags = FS_REQUIRES_DEV,
3969};
03010a33
TT		 3970MODULE_ALIAS("ext4dev");
3971#endif
ac27a0ec 3972
617ba13b 3973static int __init init_ext4_fs(void)
ac27a0ec 3974{
c9de560d
AT		 3975 int err;
3976
6fd058f7
TT		 3977 err = init_ext4_system_zone();
3978 if (err)
3979 return err;
3197ebdb
TT		 3980 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
3981 if (!ext4_kset)
6fd058f7 3982 goto out4;
9f6200bb 3983 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
c9de560d 3984 err = init_ext4_mballoc();
ac27a0ec 3985 if (err)
6fd058f7 3986 goto out3;
c9de560d
AT		 3987
3988 err = init_ext4_xattr();
3989 if (err)
3990 goto out2;
ac27a0ec
DK		 3991 err = init_inodecache();
3992 if (err)
3993 goto out1;
03010a33 3994 err = register_filesystem(&ext4_fs_type);
ac27a0ec
DK		 3995 if (err)
3996 goto out;
03010a33
TT		 3997#ifdef CONFIG_EXT4DEV_COMPAT
3998 err = register_filesystem(&ext4dev_fs_type);
3999 if (err) {
4000 unregister_filesystem(&ext4_fs_type);
4001 goto out;
4002 }
4003#endif
ac27a0ec
DK		 4004 return 0;
4005out:
4006 destroy_inodecache();
4007out1:
617ba13b 4008 exit_ext4_xattr();
c9de560d
AT		 4009out2:
4010 exit_ext4_mballoc();
6fd058f7
TT		 4011out3:
4012 remove_proc_entry("fs/ext4", NULL);
4013 kset_unregister(ext4_kset);
4014out4:
4015 exit_ext4_system_zone();
ac27a0ec
DK		 4016 return err;
4017}
4018
617ba13b 4019static void __exit exit_ext4_fs(void)
ac27a0ec 4020{
03010a33
TT		 4021 unregister_filesystem(&ext4_fs_type);
4022#ifdef CONFIG_EXT4DEV_COMPAT
617ba13b 4023 unregister_filesystem(&ext4dev_fs_type);
03010a33 4024#endif
ac27a0ec 4025 destroy_inodecache();
617ba13b 4026 exit_ext4_xattr();
c9de560d 4027 exit_ext4_mballoc();
9f6200bb 4028 remove_proc_entry("fs/ext4", NULL);
3197ebdb 4029 kset_unregister(ext4_kset);
6fd058f7 4030 exit_ext4_system_zone();
ac27a0ec
DK		 4031}
4032
4033MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
83982b6f 4034MODULE_DESCRIPTION("Fourth Extended Filesystem");
ac27a0ec 4035MODULE_LICENSE("GPL");
617ba13b
MC		 4036module_init(init_ext4_fs)
4037module_exit(exit_ext4_fs)