projects / GitHub / mt8127 / android_kernel_alcatel_ttab.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
reiserfs: Remove unnecessary quota functions
[GitHub/mt8127/android_kernel_alcatel_ttab.git] / fs / ext4 / super.c
CommitLineData
ac27a0ec 1/*
617ba13b 2 * linux/fs/ext4/super.c
ac27a0ec
DK		 3 *
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
8 *
9 * from
10 *
11 * linux/fs/minix/inode.c
12 *
13 * Copyright (C) 1991, 1992 Linus Torvalds
14 *
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
17 */
18
19#include <linux/module.h>
20#include <linux/string.h>
21#include <linux/fs.h>
22#include <linux/time.h>
dab291af 23#include <linux/jbd2.h>
ac27a0ec
DK		 24#include <linux/slab.h>
25#include <linux/init.h>
26#include <linux/blkdev.h>
27#include <linux/parser.h>
28#include <linux/smp_lock.h>
29#include <linux/buffer_head.h>
a5694255 30#include <linux/exportfs.h>
ac27a0ec
DK		 31#include <linux/vfs.h>
32#include <linux/random.h>
33#include <linux/mount.h>
34#include <linux/namei.h>
35#include <linux/quotaops.h>
36#include <linux/seq_file.h>
9f6200bb 37#include <linux/proc_fs.h>
ede86cc4 38#include <linux/marker.h>
1330593e 39#include <linux/log2.h>
717d50e4 40#include <linux/crc16.h>
ac27a0ec
DK		 41#include <asm/uaccess.h>
42
3dcf5451
CH		 43#include "ext4.h"
44#include "ext4_jbd2.h"
ac27a0ec
DK		 45#include "xattr.h"
46#include "acl.h"
47#include "namei.h"
717d50e4 48#include "group.h"
ac27a0ec 49
9f6200bb
TT		 50struct proc_dir_entry *ext4_proc_root;
51
617ba13b 52static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
ac27a0ec 53 unsigned long journal_devnum);
c4be0c1d 54static int ext4_commit_super(struct super_block *sb,
2b2d6d01
TT		 55 struct ext4_super_block *es, int sync);
56static void ext4_mark_recovery_complete(struct super_block *sb,
57 struct ext4_super_block *es);
58static void ext4_clear_journal_err(struct super_block *sb,
59 struct ext4_super_block *es);
617ba13b 60static int ext4_sync_fs(struct super_block *sb, int wait);
2b2d6d01 61static const char *ext4_decode_error(struct super_block *sb, int errno,
ac27a0ec 62 char nbuf[16]);
2b2d6d01
TT		 63static int ext4_remount(struct super_block *sb, int *flags, char *data);
64static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
c4be0c1d 65static int ext4_unfreeze(struct super_block *sb);
2b2d6d01 66static void ext4_write_super(struct super_block *sb);
c4be0c1d 67static int ext4_freeze(struct super_block *sb);
ac27a0ec 68
bd81d8ee 69
8fadc143
AR		 70ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
71 struct ext4_group_desc *bg)
bd81d8ee 72{
3a14589c 73 return le32_to_cpu(bg->bg_block_bitmap_lo) |
8fadc143 74 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
3a14589c 75 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 76}
77
8fadc143
AR		 78ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
79 struct ext4_group_desc *bg)
bd81d8ee 80{
5272f837 81 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
8fadc143 82 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
5272f837 83 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 84}
85
8fadc143
AR		 86ext4_fsblk_t ext4_inode_table(struct super_block *sb,
87 struct ext4_group_desc *bg)
bd81d8ee 88{
5272f837 89 return le32_to_cpu(bg->bg_inode_table_lo) |
8fadc143 90 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
5272f837 91 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
bd81d8ee
LV		 92}
93
560671a0
AK		 94__u32 ext4_free_blks_count(struct super_block *sb,
95 struct ext4_group_desc *bg)
96{
97 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
98 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
99 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
100}
101
102__u32 ext4_free_inodes_count(struct super_block *sb,
103 struct ext4_group_desc *bg)
104{
105 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
106 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
107 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
108}
109
110__u32 ext4_used_dirs_count(struct super_block *sb,
111 struct ext4_group_desc *bg)
112{
113 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
114 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
115 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
116}
117
118__u32 ext4_itable_unused_count(struct super_block *sb,
119 struct ext4_group_desc *bg)
120{
121 return le16_to_cpu(bg->bg_itable_unused_lo) |
122 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
123 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
124}
125
8fadc143
AR		 126void ext4_block_bitmap_set(struct super_block *sb,
127 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 128{
3a14589c 129 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 130 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
131 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 132}
133
8fadc143
AR		 134void ext4_inode_bitmap_set(struct super_block *sb,
135 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 136{
5272f837 137 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 138 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
139 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 140}
141
8fadc143
AR		 142void ext4_inode_table_set(struct super_block *sb,
143 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 144{
5272f837 145 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 146 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
147 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 148}
149
560671a0
AK		 150void ext4_free_blks_set(struct super_block *sb,
151 struct ext4_group_desc *bg, __u32 count)
152{
153 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
154 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
155 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
156}
157
158void ext4_free_inodes_set(struct super_block *sb,
159 struct ext4_group_desc *bg, __u32 count)
160{
161 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
162 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
163 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
164}
165
166void ext4_used_dirs_set(struct super_block *sb,
167 struct ext4_group_desc *bg, __u32 count)
168{
169 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
170 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
171 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
172}
173
174void ext4_itable_unused_set(struct super_block *sb,
175 struct ext4_group_desc *bg, __u32 count)
176{
177 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
178 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
179 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
180}
181
ac27a0ec 182/*
dab291af 183 * Wrappers for jbd2_journal_start/end.
ac27a0ec
DK		 184 *
185 * The only special thing we need to do here is to make sure that all
186 * journal_end calls result in the superblock being marked dirty, so
187 * that sync() will call the filesystem's write_super callback if
188 * appropriate.
189 */
617ba13b 190handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
ac27a0ec
DK		 191{
192 journal_t *journal;
193
194 if (sb->s_flags & MS_RDONLY)
195 return ERR_PTR(-EROFS);
196
197 /* Special case here: if the journal has aborted behind our
198 * backs (eg. EIO in the commit thread), then we still need to
199 * take the FS itself readonly cleanly. */
617ba13b 200 journal = EXT4_SB(sb)->s_journal;
0390131b
FM		 201 if (journal) {
202 if (is_journal_aborted(journal)) {
203 ext4_abort(sb, __func__,
204 "Detected aborted journal");
205 return ERR_PTR(-EROFS);
206 }
207 return jbd2_journal_start(journal, nblocks);
ac27a0ec 208 }
0390131b
FM		 209 /*
210 * We're not journaling, return the appropriate indication.
211 */
212 current->journal_info = EXT4_NOJOURNAL_HANDLE;
213 return current->journal_info;
ac27a0ec
DK		 214}
215
216/*
217 * The only special thing we need to do here is to make sure that all
dab291af 218 * jbd2_journal_stop calls result in the superblock being marked dirty, so
ac27a0ec
DK		 219 * that sync() will call the filesystem's write_super callback if
220 * appropriate.
221 */
617ba13b 222int __ext4_journal_stop(const char *where, handle_t *handle)
ac27a0ec
DK		 223{
224 struct super_block *sb;
225 int err;
226 int rc;
227
0390131b
FM		 228 if (!ext4_handle_valid(handle)) {
229 /*
230 * Do this here since we don't call jbd2_journal_stop() in
231 * no-journal mode.
232 */
233 current->journal_info = NULL;
234 return 0;
235 }
ac27a0ec
DK		 236 sb = handle->h_transaction->t_journal->j_private;
237 err = handle->h_err;
dab291af 238 rc = jbd2_journal_stop(handle);
ac27a0ec
DK		 239
240 if (!err)
241 err = rc;
242 if (err)
617ba13b 243 __ext4_std_error(sb, where, err);
ac27a0ec
DK		 244 return err;
245}
246
617ba13b 247void ext4_journal_abort_handle(const char *caller, const char *err_fn,
ac27a0ec
DK		 248 struct buffer_head *bh, handle_t *handle, int err)
249{
250 char nbuf[16];
617ba13b 251 const char *errstr = ext4_decode_error(NULL, err, nbuf);
ac27a0ec 252
0390131b
FM		 253 BUG_ON(!ext4_handle_valid(handle));
254
ac27a0ec
DK		 255 if (bh)
256 BUFFER_TRACE(bh, "abort");
257
258 if (!handle->h_err)
259 handle->h_err = err;
260
261 if (is_handle_aborted(handle))
262 return;
263
264 printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
265 caller, errstr, err_fn);
266
dab291af 267 jbd2_journal_abort_handle(handle);
ac27a0ec
DK		 268}
269
270/* Deal with the reporting of failure conditions on a filesystem such as
271 * inconsistencies detected or read IO failures.
272 *
273 * On ext2, we can store the error state of the filesystem in the
617ba13b 274 * superblock. That is not possible on ext4, because we may have other
ac27a0ec
DK		 275 * write ordering constraints on the superblock which prevent us from
276 * writing it out straight away; and given that the journal is about to
277 * be aborted, we can't rely on the current, or future, transactions to
278 * write out the superblock safely.
279 *
dab291af 280 * We'll just use the jbd2_journal_abort() error code to record an error in
ac27a0ec
DK		 281 * the journal instead. On recovery, the journal will compain about
282 * that error until we've noted it down and cleared it.
283 */
284
617ba13b 285static void ext4_handle_error(struct super_block *sb)
ac27a0ec 286{
617ba13b 287 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 288
617ba13b
MC		 289 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
290 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
ac27a0ec
DK		 291
292 if (sb->s_flags & MS_RDONLY)
293 return;
294
2b2d6d01 295 if (!test_opt(sb, ERRORS_CONT)) {
617ba13b 296 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 297
617ba13b 298 EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
ac27a0ec 299 if (journal)
dab291af 300 jbd2_journal_abort(journal, -EIO);
ac27a0ec 301 }
2b2d6d01
TT		 302 if (test_opt(sb, ERRORS_RO)) {
303 printk(KERN_CRIT "Remounting filesystem read-only\n");
ac27a0ec
DK		 304 sb->s_flags |= MS_RDONLY;
305 }
617ba13b 306 ext4_commit_super(sb, es, 1);
ac27a0ec 307 if (test_opt(sb, ERRORS_PANIC))
617ba13b 308 panic("EXT4-fs (device %s): panic forced after error\n",
ac27a0ec
DK		 309 sb->s_id);
310}
311
2b2d6d01
TT		 312void ext4_error(struct super_block *sb, const char *function,
313 const char *fmt, ...)
ac27a0ec
DK		 314{
315 va_list args;
316
317 va_start(args, fmt);
2b2d6d01 318 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
ac27a0ec
DK		 319 vprintk(fmt, args);
320 printk("\n");
321 va_end(args);
322
617ba13b 323 ext4_handle_error(sb);
ac27a0ec
DK		 324}
325
2b2d6d01 326static const char *ext4_decode_error(struct super_block *sb, int errno,
ac27a0ec
DK		 327 char nbuf[16])
328{
329 char *errstr = NULL;
330
331 switch (errno) {
332 case -EIO:
333 errstr = "IO failure";
334 break;
335 case -ENOMEM:
336 errstr = "Out of memory";
337 break;
338 case -EROFS:
dab291af 339 if (!sb || EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT)
ac27a0ec
DK		 340 errstr = "Journal has aborted";
341 else
342 errstr = "Readonly filesystem";
343 break;
344 default:
345 /* If the caller passed in an extra buffer for unknown
346 * errors, textualise them now. Else we just return
347 * NULL. */
348 if (nbuf) {
349 /* Check for truncated error codes... */
350 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
351 errstr = nbuf;
352 }
353 break;
354 }
355
356 return errstr;
357}
358
617ba13b 359/* __ext4_std_error decodes expected errors from journaling functions
ac27a0ec
DK		 360 * automatically and invokes the appropriate error response. */
361
2b2d6d01 362void __ext4_std_error(struct super_block *sb, const char *function, int errno)
ac27a0ec
DK		 363{
364 char nbuf[16];
365 const char *errstr;
366
367 /* Special case: if the error is EROFS, and we're not already
368 * inside a transaction, then there's really no point in logging
369 * an error. */
370 if (errno == -EROFS && journal_current_handle() == NULL &&
371 (sb->s_flags & MS_RDONLY))
372 return;
373
617ba13b 374 errstr = ext4_decode_error(sb, errno, nbuf);
2b2d6d01
TT		 375 printk(KERN_CRIT "EXT4-fs error (device %s) in %s: %s\n",
376 sb->s_id, function, errstr);
ac27a0ec 377
617ba13b 378 ext4_handle_error(sb);
ac27a0ec
DK		 379}
380
381/*
617ba13b 382 * ext4_abort is a much stronger failure handler than ext4_error. The
ac27a0ec
DK		 383 * abort function may be used to deal with unrecoverable failures such
384 * as journal IO errors or ENOMEM at a critical moment in log management.
385 *
386 * We unconditionally force the filesystem into an ABORT|READONLY state,
387 * unless the error response on the fs has been set to panic in which
388 * case we take the easy way out and panic immediately.
389 */
390
2b2d6d01
TT		 391void ext4_abort(struct super_block *sb, const char *function,
392 const char *fmt, ...)
ac27a0ec
DK		 393{
394 va_list args;
395
2b2d6d01 396 printk(KERN_CRIT "ext4_abort called.\n");
ac27a0ec
DK		 397
398 va_start(args, fmt);
2b2d6d01 399 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
ac27a0ec
DK		 400 vprintk(fmt, args);
401 printk("\n");
402 va_end(args);
403
404 if (test_opt(sb, ERRORS_PANIC))
617ba13b 405 panic("EXT4-fs panic from previous error\n");
ac27a0ec
DK		 406
407 if (sb->s_flags & MS_RDONLY)
408 return;
409
410 printk(KERN_CRIT "Remounting filesystem read-only\n");
617ba13b 411 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
ac27a0ec 412 sb->s_flags |= MS_RDONLY;
617ba13b 413 EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
ef2cabf7
HK		 414 if (EXT4_SB(sb)->s_journal)
415 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
ac27a0ec
DK		 416}
417
2b2d6d01
TT		 418void ext4_warning(struct super_block *sb, const char *function,
419 const char *fmt, ...)
ac27a0ec
DK		 420{
421 va_list args;
422
423 va_start(args, fmt);
617ba13b 424 printk(KERN_WARNING "EXT4-fs warning (device %s): %s: ",
ac27a0ec
DK		 425 sb->s_id, function);
426 vprintk(fmt, args);
427 printk("\n");
428 va_end(args);
429}
430
5d1b1b3f
AK		 431void ext4_grp_locked_error(struct super_block *sb, ext4_group_t grp,
432 const char *function, const char *fmt, ...)
433__releases(bitlock)
434__acquires(bitlock)
435{
436 va_list args;
437 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
438
439 va_start(args, fmt);
440 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
441 vprintk(fmt, args);
442 printk("\n");
443 va_end(args);
444
445 if (test_opt(sb, ERRORS_CONT)) {
446 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
447 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
448 ext4_commit_super(sb, es, 0);
449 return;
450 }
451 ext4_unlock_group(sb, grp);
452 ext4_handle_error(sb);
453 /*
454 * We only get here in the ERRORS_RO case; relocking the group
455 * may be dangerous, but nothing bad will happen since the
456 * filesystem will have already been marked read/only and the
457 * journal has been aborted. We return 1 as a hint to callers
458 * who might what to use the return value from
459 * ext4_grp_locked_error() to distinguish beween the
460 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
461 * aggressively from the ext4 function in question, with a
462 * more appropriate error code.
463 */
464 ext4_lock_group(sb, grp);
465 return;
466}
467
468
617ba13b 469void ext4_update_dynamic_rev(struct super_block *sb)
ac27a0ec 470{
617ba13b 471 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 472
617ba13b 473 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
ac27a0ec
DK		 474 return;
475
46e665e9 476 ext4_warning(sb, __func__,
ac27a0ec
DK		 477 "updating to rev %d because of new feature flag, "
478 "running e2fsck is recommended",
617ba13b 479 EXT4_DYNAMIC_REV);
ac27a0ec 480
617ba13b
MC		 481 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
482 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
483 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
ac27a0ec
DK		 484 /* leave es->s_feature_*compat flags alone */
485 /* es->s_uuid will be set by e2fsck if empty */
486
487 /*
488 * The rest of the superblock fields should be zero, and if not it
489 * means they are likely already in use, so leave them alone. We
490 * can leave it up to e2fsck to clean up any inconsistencies there.
491 */
492}
493
494/*
495 * Open the external journal device
496 */
617ba13b 497static struct block_device *ext4_blkdev_get(dev_t dev)
ac27a0ec
DK		 498{
499 struct block_device *bdev;
500 char b[BDEVNAME_SIZE];
501
502 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
503 if (IS_ERR(bdev))
504 goto fail;
505 return bdev;
506
507fail:
abda1418 508 printk(KERN_ERR "EXT4-fs: failed to open journal device %s: %ld\n",
ac27a0ec
DK		 509 __bdevname(dev, b), PTR_ERR(bdev));
510 return NULL;
511}
512
513/*
514 * Release the journal device
515 */
617ba13b 516static int ext4_blkdev_put(struct block_device *bdev)
ac27a0ec
DK		 517{
518 bd_release(bdev);
9a1c3542 519 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
ac27a0ec
DK		 520}
521
617ba13b 522static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
ac27a0ec
DK		 523{
524 struct block_device *bdev;
525 int ret = -ENODEV;
526
527 bdev = sbi->journal_bdev;
528 if (bdev) {
617ba13b 529 ret = ext4_blkdev_put(bdev);
ac27a0ec
DK		 530 sbi->journal_bdev = NULL;
531 }
532 return ret;
533}
534
535static inline struct inode *orphan_list_entry(struct list_head *l)
536{
617ba13b 537 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
ac27a0ec
DK		 538}
539
617ba13b 540static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
ac27a0ec
DK		 541{
542 struct list_head *l;
543
544 printk(KERN_ERR "sb orphan head is %d\n",
545 le32_to_cpu(sbi->s_es->s_last_orphan));
546
547 printk(KERN_ERR "sb_info orphan list:\n");
548 list_for_each(l, &sbi->s_orphan) {
549 struct inode *inode = orphan_list_entry(l);
550 printk(KERN_ERR " "
551 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
552 inode->i_sb->s_id, inode->i_ino, inode,
553 inode->i_mode, inode->i_nlink,
554 NEXT_ORPHAN(inode));
555 }
556}
557
2b2d6d01 558static void ext4_put_super(struct super_block *sb)
ac27a0ec 559{
617ba13b
MC		 560 struct ext4_sb_info *sbi = EXT4_SB(sb);
561 struct ext4_super_block *es = sbi->s_es;
ef2cabf7 562 int i, err;
ac27a0ec 563
c9de560d 564 ext4_mb_release(sb);
a86c6181 565 ext4_ext_release(sb);
617ba13b 566 ext4_xattr_put_super(sb);
0390131b
FM		 567 if (sbi->s_journal) {
568 err = jbd2_journal_destroy(sbi->s_journal);
569 sbi->s_journal = NULL;
570 if (err < 0)
571 ext4_abort(sb, __func__,
572 "Couldn't clean up the journal");
573 }
ac27a0ec 574 if (!(sb->s_flags & MS_RDONLY)) {
617ba13b 575 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 576 es->s_state = cpu_to_le16(sbi->s_mount_state);
617ba13b 577 ext4_commit_super(sb, es, 1);
ac27a0ec 578 }
240799cd
TT		 579 if (sbi->s_proc) {
580 remove_proc_entry("inode_readahead_blks", sbi->s_proc);
9f6200bb 581 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 582 }
ac27a0ec
DK		 583
584 for (i = 0; i < sbi->s_gdb_count; i++)
585 brelse(sbi->s_group_desc[i]);
586 kfree(sbi->s_group_desc);
772cb7c8 587 kfree(sbi->s_flex_groups);
ac27a0ec
DK		 588 percpu_counter_destroy(&sbi->s_freeblocks_counter);
589 percpu_counter_destroy(&sbi->s_freeinodes_counter);
590 percpu_counter_destroy(&sbi->s_dirs_counter);
6bc6e63f 591 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
ac27a0ec
DK		 592 brelse(sbi->s_sbh);
593#ifdef CONFIG_QUOTA
594 for (i = 0; i < MAXQUOTAS; i++)
595 kfree(sbi->s_qf_names[i]);
596#endif
597
598 /* Debugging code just in case the in-memory inode orphan list
599 * isn't empty. The on-disk one can be non-empty if we've
600 * detected an error and taken the fs readonly, but the
601 * in-memory list had better be clean by this point. */
602 if (!list_empty(&sbi->s_orphan))
603 dump_orphan_list(sb, sbi);
604 J_ASSERT(list_empty(&sbi->s_orphan));
605
f98393a6 606 invalidate_bdev(sb->s_bdev);
ac27a0ec
DK		 607 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
608 /*
609 * Invalidate the journal device's buffers. We don't want them
610 * floating about in memory - the physical journal device may
611 * hotswapped, and it breaks the `ro-after' testing code.
612 */
613 sync_blockdev(sbi->journal_bdev);
f98393a6 614 invalidate_bdev(sbi->journal_bdev);
617ba13b 615 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 616 }
617 sb->s_fs_info = NULL;
618 kfree(sbi);
619 return;
620}
621
e18b890b 622static struct kmem_cache *ext4_inode_cachep;
ac27a0ec
DK		 623
624/*
625 * Called inside transaction, so use GFP_NOFS
626 */
617ba13b 627static struct inode *ext4_alloc_inode(struct super_block *sb)
ac27a0ec 628{
617ba13b 629 struct ext4_inode_info *ei;
ac27a0ec 630
e6b4f8da 631 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
ac27a0ec
DK		 632 if (!ei)
633 return NULL;
03010a33 634#ifdef CONFIG_EXT4_FS_POSIX_ACL
617ba13b
MC		 635 ei->i_acl = EXT4_ACL_NOT_CACHED;
636 ei->i_default_acl = EXT4_ACL_NOT_CACHED;
ac27a0ec 637#endif
ac27a0ec 638 ei->vfs_inode.i_version = 1;
91246c00 639 ei->vfs_inode.i_data.writeback_index = 0;
a86c6181 640 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
c9de560d
AT		 641 INIT_LIST_HEAD(&ei->i_prealloc_list);
642 spin_lock_init(&ei->i_prealloc_lock);
0390131b
FM		 643 /*
644 * Note: We can be called before EXT4_SB(sb)->s_journal is set,
645 * therefore it can be null here. Don't check it, just initialize
646 * jinode.
647 */
678aaf48 648 jbd2_journal_init_jbd_inode(&ei->jinode, &ei->vfs_inode);
d2a17637
MC		 649 ei->i_reserved_data_blocks = 0;
650 ei->i_reserved_meta_blocks = 0;
651 ei->i_allocated_meta_blocks = 0;
652 ei->i_delalloc_reserved_flag = 0;
653 spin_lock_init(&(ei->i_block_reservation_lock));
ac27a0ec
DK		 654 return &ei->vfs_inode;
655}
656
617ba13b 657static void ext4_destroy_inode(struct inode *inode)
ac27a0ec 658{
9f7dd93d
VA		 659 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
660 printk("EXT4 Inode %p: orphan list check failed!\n",
661 EXT4_I(inode));
662 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
663 EXT4_I(inode), sizeof(struct ext4_inode_info),
664 true);
665 dump_stack();
666 }
617ba13b 667 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
ac27a0ec
DK		 668}
669
51cc5068 670static void init_once(void *foo)
ac27a0ec 671{
617ba13b 672 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
ac27a0ec 673
a35afb83 674 INIT_LIST_HEAD(&ei->i_orphan);
03010a33 675#ifdef CONFIG_EXT4_FS_XATTR
a35afb83 676 init_rwsem(&ei->xattr_sem);
ac27a0ec 677#endif
0e855ac8 678 init_rwsem(&ei->i_data_sem);
a35afb83 679 inode_init_once(&ei->vfs_inode);
ac27a0ec
DK		 680}
681
682static int init_inodecache(void)
683{
617ba13b
MC		 684 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
685 sizeof(struct ext4_inode_info),
ac27a0ec
DK		 686 0, (SLAB_RECLAIM_ACCOUNT|
687 SLAB_MEM_SPREAD),
20c2df83 688 init_once);
617ba13b 689 if (ext4_inode_cachep == NULL)
ac27a0ec
DK		 690 return -ENOMEM;
691 return 0;
692}
693
694static void destroy_inodecache(void)
695{
617ba13b 696 kmem_cache_destroy(ext4_inode_cachep);
ac27a0ec
DK		 697}
698
617ba13b 699static void ext4_clear_inode(struct inode *inode)
ac27a0ec 700{
03010a33 701#ifdef CONFIG_EXT4_FS_POSIX_ACL
617ba13b
MC		 702 if (EXT4_I(inode)->i_acl &&
703 EXT4_I(inode)->i_acl != EXT4_ACL_NOT_CACHED) {
704 posix_acl_release(EXT4_I(inode)->i_acl);
705 EXT4_I(inode)->i_acl = EXT4_ACL_NOT_CACHED;
706 }
707 if (EXT4_I(inode)->i_default_acl &&
708 EXT4_I(inode)->i_default_acl != EXT4_ACL_NOT_CACHED) {
709 posix_acl_release(EXT4_I(inode)->i_default_acl);
710 EXT4_I(inode)->i_default_acl = EXT4_ACL_NOT_CACHED;
ac27a0ec
DK		 711 }
712#endif
c2ea3fde 713 ext4_discard_preallocations(inode);
0390131b
FM		 714 if (EXT4_JOURNAL(inode))
715 jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal,
678aaf48 716 &EXT4_I(inode)->jinode);
ac27a0ec
DK		 717}
718
2b2d6d01
TT		 719static inline void ext4_show_quota_options(struct seq_file *seq,
720 struct super_block *sb)
ac27a0ec
DK		 721{
722#if defined(CONFIG_QUOTA)
617ba13b 723 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 724
725 if (sbi->s_jquota_fmt)
726 seq_printf(seq, ",jqfmt=%s",
af5bc92d 727 (sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold" : "vfsv0");
ac27a0ec
DK		 728
729 if (sbi->s_qf_names[USRQUOTA])
730 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
731
732 if (sbi->s_qf_names[GRPQUOTA])
733 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
734
617ba13b 735 if (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA)
ac27a0ec
DK		 736 seq_puts(seq, ",usrquota");
737
617ba13b 738 if (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)
ac27a0ec
DK		 739 seq_puts(seq, ",grpquota");
740#endif
741}
742
d9c9bef1
MS		 743/*
744 * Show an option if
745 * - it's set to a non-default value OR
746 * - if the per-sb default is different from the global default
747 */
617ba13b 748static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
ac27a0ec 749{
aa22df2c
AK		 750 int def_errors;
751 unsigned long def_mount_opts;
ac27a0ec 752 struct super_block *sb = vfs->mnt_sb;
d9c9bef1
MS		 753 struct ext4_sb_info *sbi = EXT4_SB(sb);
754 struct ext4_super_block *es = sbi->s_es;
d9c9bef1
MS		 755
756 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
aa22df2c 757 def_errors = le16_to_cpu(es->s_errors);
d9c9bef1
MS		 758
759 if (sbi->s_sb_block != 1)
760 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
761 if (test_opt(sb, MINIX_DF))
762 seq_puts(seq, ",minixdf");
aa22df2c 763 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
d9c9bef1
MS		 764 seq_puts(seq, ",grpid");
765 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
766 seq_puts(seq, ",nogrpid");
767 if (sbi->s_resuid != EXT4_DEF_RESUID ||
768 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
769 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
770 }
771 if (sbi->s_resgid != EXT4_DEF_RESGID ||
772 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
773 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
774 }
bb4f397a 775 if (test_opt(sb, ERRORS_RO)) {
d9c9bef1 776 if (def_errors == EXT4_ERRORS_PANIC ||
bb4f397a
AK		 777 def_errors == EXT4_ERRORS_CONTINUE) {
778 seq_puts(seq, ",errors=remount-ro");
d9c9bef1
MS		 779 }
780 }
aa22df2c 781 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
bb4f397a 782 seq_puts(seq, ",errors=continue");
aa22df2c 783 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
d9c9bef1 784 seq_puts(seq, ",errors=panic");
aa22df2c 785 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
d9c9bef1 786 seq_puts(seq, ",nouid32");
aa22df2c 787 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
d9c9bef1
MS		 788 seq_puts(seq, ",debug");
789 if (test_opt(sb, OLDALLOC))
790 seq_puts(seq, ",oldalloc");
03010a33 791#ifdef CONFIG_EXT4_FS_XATTR
aa22df2c
AK		 792 if (test_opt(sb, XATTR_USER) &&
793 !(def_mount_opts & EXT4_DEFM_XATTR_USER))
d9c9bef1
MS		 794 seq_puts(seq, ",user_xattr");
795 if (!test_opt(sb, XATTR_USER) &&
796 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
797 seq_puts(seq, ",nouser_xattr");
798 }
799#endif
03010a33 800#ifdef CONFIG_EXT4_FS_POSIX_ACL
aa22df2c 801 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
d9c9bef1
MS		 802 seq_puts(seq, ",acl");
803 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
804 seq_puts(seq, ",noacl");
805#endif
806 if (!test_opt(sb, RESERVATION))
807 seq_puts(seq, ",noreservation");
30773840 808 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
d9c9bef1
MS		 809 seq_printf(seq, ",commit=%u",
810 (unsigned) (sbi->s_commit_interval / HZ));
811 }
30773840
TT		 812 if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
813 seq_printf(seq, ",min_batch_time=%u",
814 (unsigned) sbi->s_min_batch_time);
815 }
816 if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
817 seq_printf(seq, ",max_batch_time=%u",
818 (unsigned) sbi->s_min_batch_time);
819 }
820
571640ca
ES		 821 /*
822 * We're changing the default of barrier mount option, so
823 * let's always display its mount state so it's clear what its
824 * status is.
825 */
826 seq_puts(seq, ",barrier=");
827 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
cd0b6a39
TT		 828 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
829 seq_puts(seq, ",journal_async_commit");
d9c9bef1
MS		 830 if (test_opt(sb, NOBH))
831 seq_puts(seq, ",nobh");
25ec56b5
JNC		 832 if (test_opt(sb, I_VERSION))
833 seq_puts(seq, ",i_version");
dd919b98
AK		 834 if (!test_opt(sb, DELALLOC))
835 seq_puts(seq, ",nodelalloc");
836
ac27a0ec 837
cb45bbe4
MS		 838 if (sbi->s_stripe)
839 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
aa22df2c
AK		 840 /*
841 * journal mode get enabled in different ways
842 * So just print the value even if we didn't specify it
843 */
617ba13b 844 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
ac27a0ec 845 seq_puts(seq, ",data=journal");
617ba13b 846 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
ac27a0ec 847 seq_puts(seq, ",data=ordered");
617ba13b 848 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
ac27a0ec
DK		 849 seq_puts(seq, ",data=writeback");
850
240799cd
TT		 851 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
852 seq_printf(seq, ",inode_readahead_blks=%u",
853 sbi->s_inode_readahead_blks);
854
5bf5683a
HK		 855 if (test_opt(sb, DATA_ERR_ABORT))
856 seq_puts(seq, ",data_err=abort");
857
617ba13b 858 ext4_show_quota_options(seq, sb);
ac27a0ec
DK		 859 return 0;
860}
861
862
1b961ac0
CH		 863static struct inode *ext4_nfs_get_inode(struct super_block *sb,
864 u64 ino, u32 generation)
ac27a0ec 865{
ac27a0ec 866 struct inode *inode;
ac27a0ec 867
617ba13b 868 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
ac27a0ec 869 return ERR_PTR(-ESTALE);
617ba13b 870 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
ac27a0ec
DK		 871 return ERR_PTR(-ESTALE);
872
873 /* iget isn't really right if the inode is currently unallocated!!
874 *
617ba13b 875 * ext4_read_inode will return a bad_inode if the inode had been
ac27a0ec
DK		 876 * deleted, so we should be safe.
877 *
878 * Currently we don't know the generation for parent directory, so
879 * a generation of 0 means "accept any"
880 */
1d1fe1ee
DH		 881 inode = ext4_iget(sb, ino);
882 if (IS_ERR(inode))
883 return ERR_CAST(inode);
884 if (generation && inode->i_generation != generation) {
ac27a0ec
DK		 885 iput(inode);
886 return ERR_PTR(-ESTALE);
887 }
1b961ac0
CH		 888
889 return inode;
890}
891
892static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
893 int fh_len, int fh_type)
894{
895 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
896 ext4_nfs_get_inode);
897}
898
899static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
900 int fh_len, int fh_type)
901{
902 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
903 ext4_nfs_get_inode);
ac27a0ec
DK		 904}
905
c39a7f84
TO		 906/*
907 * Try to release metadata pages (indirect blocks, directories) which are
908 * mapped via the block device. Since these pages could have journal heads
909 * which would prevent try_to_free_buffers() from freeing them, we must use
910 * jbd2 layer's try_to_free_buffers() function to release them.
911 */
912static int bdev_try_to_free_page(struct super_block *sb, struct page *page, gfp_t wait)
913{
914 journal_t *journal = EXT4_SB(sb)->s_journal;
915
916 WARN_ON(PageChecked(page));
917 if (!page_has_buffers(page))
918 return 0;
919 if (journal)
920 return jbd2_journal_try_to_free_buffers(journal, page,
921 wait & ~__GFP_WAIT);
922 return try_to_free_buffers(page);
923}
924
ac27a0ec 925#ifdef CONFIG_QUOTA
af5bc92d 926#define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
2b2d6d01 927#define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
ac27a0ec 928
617ba13b
MC		 929static int ext4_write_dquot(struct dquot *dquot);
930static int ext4_acquire_dquot(struct dquot *dquot);
931static int ext4_release_dquot(struct dquot *dquot);
932static int ext4_mark_dquot_dirty(struct dquot *dquot);
933static int ext4_write_info(struct super_block *sb, int type);
6f28e087
JK		 934static int ext4_quota_on(struct super_block *sb, int type, int format_id,
935 char *path, int remount);
617ba13b
MC		 936static int ext4_quota_on_mount(struct super_block *sb, int type);
937static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 938 size_t len, loff_t off);
617ba13b 939static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 940 const char *data, size_t len, loff_t off);
941
617ba13b 942static struct dquot_operations ext4_quota_operations = {
edf72453
JK		 943 .initialize = dquot_initialize,
944 .drop = dquot_drop,
ac27a0ec
DK		 945 .alloc_space = dquot_alloc_space,
946 .alloc_inode = dquot_alloc_inode,
947 .free_space = dquot_free_space,
948 .free_inode = dquot_free_inode,
949 .transfer = dquot_transfer,
617ba13b
MC		 950 .write_dquot = ext4_write_dquot,
951 .acquire_dquot = ext4_acquire_dquot,
952 .release_dquot = ext4_release_dquot,
953 .mark_dirty = ext4_mark_dquot_dirty,
a5b5ee32
JK		 954 .write_info = ext4_write_info,
955 .alloc_dquot = dquot_alloc,
956 .destroy_dquot = dquot_destroy,
ac27a0ec
DK		 957};
958
617ba13b
MC		 959static struct quotactl_ops ext4_qctl_operations = {
960 .quota_on = ext4_quota_on,
ac27a0ec
DK		 961 .quota_off = vfs_quota_off,
962 .quota_sync = vfs_quota_sync,
963 .get_info = vfs_get_dqinfo,
964 .set_info = vfs_set_dqinfo,
965 .get_dqblk = vfs_get_dqblk,
966 .set_dqblk = vfs_set_dqblk
967};
968#endif
969
ee9b6d61 970static const struct super_operations ext4_sops = {
617ba13b
MC		 971 .alloc_inode = ext4_alloc_inode,
972 .destroy_inode = ext4_destroy_inode,
617ba13b
MC		 973 .write_inode = ext4_write_inode,
974 .dirty_inode = ext4_dirty_inode,
975 .delete_inode = ext4_delete_inode,
976 .put_super = ext4_put_super,
977 .write_super = ext4_write_super,
978 .sync_fs = ext4_sync_fs,
c4be0c1d
TS		 979 .freeze_fs = ext4_freeze,
980 .unfreeze_fs = ext4_unfreeze,
617ba13b
MC		 981 .statfs = ext4_statfs,
982 .remount_fs = ext4_remount,
983 .clear_inode = ext4_clear_inode,
984 .show_options = ext4_show_options,
ac27a0ec 985#ifdef CONFIG_QUOTA
617ba13b
MC		 986 .quota_read = ext4_quota_read,
987 .quota_write = ext4_quota_write,
ac27a0ec 988#endif
c39a7f84 989 .bdev_try_to_free_page = bdev_try_to_free_page,
ac27a0ec
DK		 990};
991
39655164 992static const struct export_operations ext4_export_ops = {
1b961ac0
CH		 993 .fh_to_dentry = ext4_fh_to_dentry,
994 .fh_to_parent = ext4_fh_to_parent,
617ba13b 995 .get_parent = ext4_get_parent,
ac27a0ec
DK		 996};
997
998enum {
999 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1000 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
01436ef2 1001 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
ac27a0ec
DK		 1002 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1003 Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh,
30773840 1004 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
c3191067 1005 Opt_journal_update, Opt_journal_dev,
818d276c 1006 Opt_journal_checksum, Opt_journal_async_commit,
ac27a0ec 1007 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
5bf5683a 1008 Opt_data_err_abort, Opt_data_err_ignore,
ac27a0ec
DK		 1009 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1010 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
1011 Opt_ignore, Opt_barrier, Opt_err, Opt_resize, Opt_usrquota,
83982b6f 1012 Opt_grpquota, Opt_i_version,
01436ef2 1013 Opt_stripe, Opt_delalloc, Opt_nodelalloc,
b3881f74 1014 Opt_inode_readahead_blks, Opt_journal_ioprio
ac27a0ec
DK		 1015};
1016
a447c093 1017static const match_table_t tokens = {
ac27a0ec
DK		 1018 {Opt_bsd_df, "bsddf"},
1019 {Opt_minix_df, "minixdf"},
1020 {Opt_grpid, "grpid"},
1021 {Opt_grpid, "bsdgroups"},
1022 {Opt_nogrpid, "nogrpid"},
1023 {Opt_nogrpid, "sysvgroups"},
1024 {Opt_resgid, "resgid=%u"},
1025 {Opt_resuid, "resuid=%u"},
1026 {Opt_sb, "sb=%u"},
1027 {Opt_err_cont, "errors=continue"},
1028 {Opt_err_panic, "errors=panic"},
1029 {Opt_err_ro, "errors=remount-ro"},
1030 {Opt_nouid32, "nouid32"},
ac27a0ec
DK		 1031 {Opt_debug, "debug"},
1032 {Opt_oldalloc, "oldalloc"},
1033 {Opt_orlov, "orlov"},
1034 {Opt_user_xattr, "user_xattr"},
1035 {Opt_nouser_xattr, "nouser_xattr"},
1036 {Opt_acl, "acl"},
1037 {Opt_noacl, "noacl"},
1038 {Opt_reservation, "reservation"},
1039 {Opt_noreservation, "noreservation"},
1040 {Opt_noload, "noload"},
1041 {Opt_nobh, "nobh"},
1042 {Opt_bh, "bh"},
1043 {Opt_commit, "commit=%u"},
30773840
TT		 1044 {Opt_min_batch_time, "min_batch_time=%u"},
1045 {Opt_max_batch_time, "max_batch_time=%u"},
ac27a0ec 1046 {Opt_journal_update, "journal=update"},
ac27a0ec 1047 {Opt_journal_dev, "journal_dev=%u"},
818d276c
GS		 1048 {Opt_journal_checksum, "journal_checksum"},
1049 {Opt_journal_async_commit, "journal_async_commit"},
ac27a0ec
DK		 1050 {Opt_abort, "abort"},
1051 {Opt_data_journal, "data=journal"},
1052 {Opt_data_ordered, "data=ordered"},
1053 {Opt_data_writeback, "data=writeback"},
5bf5683a
HK		 1054 {Opt_data_err_abort, "data_err=abort"},
1055 {Opt_data_err_ignore, "data_err=ignore"},
ac27a0ec
DK		 1056 {Opt_offusrjquota, "usrjquota="},
1057 {Opt_usrjquota, "usrjquota=%s"},
1058 {Opt_offgrpjquota, "grpjquota="},
1059 {Opt_grpjquota, "grpjquota=%s"},
1060 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1061 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1062 {Opt_grpquota, "grpquota"},
1063 {Opt_noquota, "noquota"},
1064 {Opt_quota, "quota"},
1065 {Opt_usrquota, "usrquota"},
1066 {Opt_barrier, "barrier=%u"},
25ec56b5 1067 {Opt_i_version, "i_version"},
c9de560d 1068 {Opt_stripe, "stripe=%u"},
ac27a0ec 1069 {Opt_resize, "resize"},
64769240 1070 {Opt_delalloc, "delalloc"},
dd919b98 1071 {Opt_nodelalloc, "nodelalloc"},
240799cd 1072 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
b3881f74 1073 {Opt_journal_ioprio, "journal_ioprio=%u"},
f3f12faa 1074 {Opt_err, NULL},
ac27a0ec
DK		 1075};
1076
617ba13b 1077static ext4_fsblk_t get_sb_block(void **data)
ac27a0ec 1078{
617ba13b 1079 ext4_fsblk_t sb_block;
ac27a0ec
DK		 1080 char *options = (char *) *data;
1081
1082 if (!options || strncmp(options, "sb=", 3) != 0)
1083 return 1; /* Default location */
1084 options += 3;
617ba13b 1085 /*todo: use simple_strtoll with >32bit ext4 */
ac27a0ec
DK		 1086 sb_block = simple_strtoul(options, &options, 0);
1087 if (*options && *options != ',') {
4776004f 1088 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
ac27a0ec
DK		 1089 (char *) *data);
1090 return 1;
1091 }
1092 if (*options == ',')
1093 options++;
1094 *data = (void *) options;
1095 return sb_block;
1096}
1097
b3881f74
TT		 1098#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1099
2b2d6d01 1100static int parse_options(char *options, struct super_block *sb,
c3191067 1101 unsigned long *journal_devnum,
b3881f74 1102 unsigned int *journal_ioprio,
2b2d6d01 1103 ext4_fsblk_t *n_blocks_count, int is_remount)
ac27a0ec 1104{
617ba13b 1105 struct ext4_sb_info *sbi = EXT4_SB(sb);
2b2d6d01 1106 char *p;
ac27a0ec
DK		 1107 substring_t args[MAX_OPT_ARGS];
1108 int data_opt = 0;
1109 int option;
1110#ifdef CONFIG_QUOTA
dfc5d03f 1111 int qtype, qfmt;
ac27a0ec
DK		 1112 char *qname;
1113#endif
1114
1115 if (!options)
1116 return 1;
1117
2b2d6d01 1118 while ((p = strsep(&options, ",")) != NULL) {
ac27a0ec
DK		 1119 int token;
1120 if (!*p)
1121 continue;
1122
1123 token = match_token(p, tokens, args);
1124 switch (token) {
1125 case Opt_bsd_df:
2b2d6d01 1126 clear_opt(sbi->s_mount_opt, MINIX_DF);
ac27a0ec
DK		 1127 break;
1128 case Opt_minix_df:
2b2d6d01 1129 set_opt(sbi->s_mount_opt, MINIX_DF);
ac27a0ec
DK		 1130 break;
1131 case Opt_grpid:
2b2d6d01 1132 set_opt(sbi->s_mount_opt, GRPID);
ac27a0ec
DK		 1133 break;
1134 case Opt_nogrpid:
2b2d6d01 1135 clear_opt(sbi->s_mount_opt, GRPID);
ac27a0ec
DK		 1136 break;
1137 case Opt_resuid:
1138 if (match_int(&args[0], &option))
1139 return 0;
1140 sbi->s_resuid = option;
1141 break;
1142 case Opt_resgid:
1143 if (match_int(&args[0], &option))
1144 return 0;
1145 sbi->s_resgid = option;
1146 break;
1147 case Opt_sb:
1148 /* handled by get_sb_block() instead of here */
1149 /* *sb_block = match_int(&args[0]); */
1150 break;
1151 case Opt_err_panic:
2b2d6d01
TT		 1152 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1153 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1154 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
ac27a0ec
DK		 1155 break;
1156 case Opt_err_ro:
2b2d6d01
TT		 1157 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1158 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1159 set_opt(sbi->s_mount_opt, ERRORS_RO);
ac27a0ec
DK		 1160 break;
1161 case Opt_err_cont:
2b2d6d01
TT		 1162 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1163 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1164 set_opt(sbi->s_mount_opt, ERRORS_CONT);
ac27a0ec
DK		 1165 break;
1166 case Opt_nouid32:
2b2d6d01 1167 set_opt(sbi->s_mount_opt, NO_UID32);
ac27a0ec 1168 break;
ac27a0ec 1169 case Opt_debug:
2b2d6d01 1170 set_opt(sbi->s_mount_opt, DEBUG);
ac27a0ec
DK		 1171 break;
1172 case Opt_oldalloc:
2b2d6d01 1173 set_opt(sbi->s_mount_opt, OLDALLOC);
ac27a0ec
DK		 1174 break;
1175 case Opt_orlov:
2b2d6d01 1176 clear_opt(sbi->s_mount_opt, OLDALLOC);
ac27a0ec 1177 break;
03010a33 1178#ifdef CONFIG_EXT4_FS_XATTR
ac27a0ec 1179 case Opt_user_xattr:
2b2d6d01 1180 set_opt(sbi->s_mount_opt, XATTR_USER);
ac27a0ec
DK		 1181 break;
1182 case Opt_nouser_xattr:
2b2d6d01 1183 clear_opt(sbi->s_mount_opt, XATTR_USER);
ac27a0ec
DK		 1184 break;
1185#else
1186 case Opt_user_xattr:
1187 case Opt_nouser_xattr:
4776004f
TT		 1188 printk(KERN_ERR "EXT4 (no)user_xattr options "
1189 "not supported\n");
ac27a0ec
DK		 1190 break;
1191#endif
03010a33 1192#ifdef CONFIG_EXT4_FS_POSIX_ACL
ac27a0ec
DK		 1193 case Opt_acl:
1194 set_opt(sbi->s_mount_opt, POSIX_ACL);
1195 break;
1196 case Opt_noacl:
1197 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1198 break;
1199#else
1200 case Opt_acl:
1201 case Opt_noacl:
4776004f
TT		 1202 printk(KERN_ERR "EXT4 (no)acl options "
1203 "not supported\n");
ac27a0ec
DK		 1204 break;
1205#endif
1206 case Opt_reservation:
1207 set_opt(sbi->s_mount_opt, RESERVATION);
1208 break;
1209 case Opt_noreservation:
1210 clear_opt(sbi->s_mount_opt, RESERVATION);
1211 break;
1212 case Opt_journal_update:
1213 /* @@@ FIXME */
1214 /* Eventually we will want to be able to create
1215 a journal file here. For now, only allow the
1216 user to specify an existing inode to be the
1217 journal file. */
1218 if (is_remount) {
617ba13b 1219 printk(KERN_ERR "EXT4-fs: cannot specify "
ac27a0ec
DK		 1220 "journal on remount\n");
1221 return 0;
1222 }
2b2d6d01 1223 set_opt(sbi->s_mount_opt, UPDATE_JOURNAL);
ac27a0ec 1224 break;
ac27a0ec
DK		 1225 case Opt_journal_dev:
1226 if (is_remount) {
617ba13b 1227 printk(KERN_ERR "EXT4-fs: cannot specify "
ac27a0ec
DK		 1228 "journal on remount\n");
1229 return 0;
1230 }
1231 if (match_int(&args[0], &option))
1232 return 0;
1233 *journal_devnum = option;
1234 break;
818d276c
GS		 1235 case Opt_journal_checksum:
1236 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1237 break;
1238 case Opt_journal_async_commit:
1239 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
1240 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1241 break;
ac27a0ec 1242 case Opt_noload:
2b2d6d01 1243 set_opt(sbi->s_mount_opt, NOLOAD);
ac27a0ec
DK		 1244 break;
1245 case Opt_commit:
1246 if (match_int(&args[0], &option))
1247 return 0;
1248 if (option < 0)
1249 return 0;
1250 if (option == 0)
cd02ff0b 1251 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
ac27a0ec
DK		 1252 sbi->s_commit_interval = HZ * option;
1253 break;
30773840
TT		 1254 case Opt_max_batch_time:
1255 if (match_int(&args[0], &option))
1256 return 0;
1257 if (option < 0)
1258 return 0;
1259 if (option == 0)
1260 option = EXT4_DEF_MAX_BATCH_TIME;
1261 sbi->s_max_batch_time = option;
1262 break;
1263 case Opt_min_batch_time:
1264 if (match_int(&args[0], &option))
1265 return 0;
1266 if (option < 0)
1267 return 0;
1268 sbi->s_min_batch_time = option;
1269 break;
ac27a0ec 1270 case Opt_data_journal:
617ba13b 1271 data_opt = EXT4_MOUNT_JOURNAL_DATA;
ac27a0ec
DK		 1272 goto datacheck;
1273 case Opt_data_ordered:
617ba13b 1274 data_opt = EXT4_MOUNT_ORDERED_DATA;
ac27a0ec
DK		 1275 goto datacheck;
1276 case Opt_data_writeback:
617ba13b 1277 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
ac27a0ec
DK		 1278 datacheck:
1279 if (is_remount) {
617ba13b 1280 if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
ac27a0ec
DK		 1281 != data_opt) {
1282 printk(KERN_ERR
617ba13b 1283 "EXT4-fs: cannot change data "
ac27a0ec
DK		 1284 "mode on remount\n");
1285 return 0;
1286 }
1287 } else {
617ba13b 1288 sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
ac27a0ec
DK		 1289 sbi->s_mount_opt |= data_opt;
1290 }
1291 break;
5bf5683a
HK		 1292 case Opt_data_err_abort:
1293 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1294 break;
1295 case Opt_data_err_ignore:
1296 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1297 break;
ac27a0ec
DK		 1298#ifdef CONFIG_QUOTA
1299 case Opt_usrjquota:
1300 qtype = USRQUOTA;
1301 goto set_qf_name;
1302 case Opt_grpjquota:
1303 qtype = GRPQUOTA;
1304set_qf_name:
17bd13b3 1305 if (sb_any_quota_loaded(sb) &&
dfc5d03f 1306 !sbi->s_qf_names[qtype]) {
ac27a0ec 1307 printk(KERN_ERR
4776004f
TT		 1308 "EXT4-fs: Cannot change journaled "
1309 "quota options when quota turned on.\n");
ac27a0ec
DK		 1310 return 0;
1311 }
1312 qname = match_strdup(&args[0]);
1313 if (!qname) {
1314 printk(KERN_ERR
617ba13b 1315 "EXT4-fs: not enough memory for "
ac27a0ec
DK		 1316 "storing quotafile name.\n");
1317 return 0;
1318 }
1319 if (sbi->s_qf_names[qtype] &&
1320 strcmp(sbi->s_qf_names[qtype], qname)) {
1321 printk(KERN_ERR
617ba13b 1322 "EXT4-fs: %s quota file already "
ac27a0ec
DK		 1323 "specified.\n", QTYPE2NAME(qtype));
1324 kfree(qname);
1325 return 0;
1326 }
1327 sbi->s_qf_names[qtype] = qname;
1328 if (strchr(sbi->s_qf_names[qtype], '/')) {
1329 printk(KERN_ERR
617ba13b 1330 "EXT4-fs: quotafile must be on "
ac27a0ec
DK		 1331 "filesystem root.\n");
1332 kfree(sbi->s_qf_names[qtype]);
1333 sbi->s_qf_names[qtype] = NULL;
1334 return 0;
1335 }
1336 set_opt(sbi->s_mount_opt, QUOTA);
1337 break;
1338 case Opt_offusrjquota:
1339 qtype = USRQUOTA;
1340 goto clear_qf_name;
1341 case Opt_offgrpjquota:
1342 qtype = GRPQUOTA;
1343clear_qf_name:
17bd13b3 1344 if (sb_any_quota_loaded(sb) &&
dfc5d03f 1345 sbi->s_qf_names[qtype]) {
617ba13b 1346 printk(KERN_ERR "EXT4-fs: Cannot change "
2c8be6b2 1347 "journaled quota options when "
ac27a0ec
DK		 1348 "quota turned on.\n");
1349 return 0;
1350 }
1351 /*
1352 * The space will be released later when all options
1353 * are confirmed to be correct
1354 */
1355 sbi->s_qf_names[qtype] = NULL;
1356 break;
1357 case Opt_jqfmt_vfsold:
dfc5d03f
JK		 1358 qfmt = QFMT_VFS_OLD;
1359 goto set_qf_format;
ac27a0ec 1360 case Opt_jqfmt_vfsv0:
dfc5d03f
JK		 1361 qfmt = QFMT_VFS_V0;
1362set_qf_format:
17bd13b3 1363 if (sb_any_quota_loaded(sb) &&
dfc5d03f
JK		 1364 sbi->s_jquota_fmt != qfmt) {
1365 printk(KERN_ERR "EXT4-fs: Cannot change "
1366 "journaled quota options when "
1367 "quota turned on.\n");
1368 return 0;
1369 }
1370 sbi->s_jquota_fmt = qfmt;
ac27a0ec
DK		 1371 break;
1372 case Opt_quota:
1373 case Opt_usrquota:
1374 set_opt(sbi->s_mount_opt, QUOTA);
1375 set_opt(sbi->s_mount_opt, USRQUOTA);
1376 break;
1377 case Opt_grpquota:
1378 set_opt(sbi->s_mount_opt, QUOTA);
1379 set_opt(sbi->s_mount_opt, GRPQUOTA);
1380 break;
1381 case Opt_noquota:
17bd13b3 1382 if (sb_any_quota_loaded(sb)) {
617ba13b 1383 printk(KERN_ERR "EXT4-fs: Cannot change quota "
ac27a0ec
DK		 1384 "options when quota turned on.\n");
1385 return 0;
1386 }
1387 clear_opt(sbi->s_mount_opt, QUOTA);
1388 clear_opt(sbi->s_mount_opt, USRQUOTA);
1389 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1390 break;
1391#else
1392 case Opt_quota:
1393 case Opt_usrquota:
1394 case Opt_grpquota:
cd59e7b9
JK		 1395 printk(KERN_ERR
1396 "EXT4-fs: quota options not supported.\n");
1397 break;
ac27a0ec
DK		 1398 case Opt_usrjquota:
1399 case Opt_grpjquota:
1400 case Opt_offusrjquota:
1401 case Opt_offgrpjquota:
1402 case Opt_jqfmt_vfsold:
1403 case Opt_jqfmt_vfsv0:
1404 printk(KERN_ERR
cd59e7b9 1405 "EXT4-fs: journaled quota options not "
ac27a0ec
DK		 1406 "supported.\n");
1407 break;
1408 case Opt_noquota:
1409 break;
1410#endif
1411 case Opt_abort:
1412 set_opt(sbi->s_mount_opt, ABORT);
1413 break;
1414 case Opt_barrier:
1415 if (match_int(&args[0], &option))
1416 return 0;
1417 if (option)
1418 set_opt(sbi->s_mount_opt, BARRIER);
1419 else
1420 clear_opt(sbi->s_mount_opt, BARRIER);
1421 break;
1422 case Opt_ignore:
1423 break;
1424 case Opt_resize:
1425 if (!is_remount) {
617ba13b 1426 printk("EXT4-fs: resize option only available "
ac27a0ec
DK		 1427 "for remount\n");
1428 return 0;
1429 }
1430 if (match_int(&args[0], &option) != 0)
1431 return 0;
1432 *n_blocks_count = option;
1433 break;
1434 case Opt_nobh:
1435 set_opt(sbi->s_mount_opt, NOBH);
1436 break;
1437 case Opt_bh:
1438 clear_opt(sbi->s_mount_opt, NOBH);
1439 break;
25ec56b5
JNC		 1440 case Opt_i_version:
1441 set_opt(sbi->s_mount_opt, I_VERSION);
1442 sb->s_flags |= MS_I_VERSION;
1443 break;
dd919b98
AK		 1444 case Opt_nodelalloc:
1445 clear_opt(sbi->s_mount_opt, DELALLOC);
1446 break;
c9de560d
AT		 1447 case Opt_stripe:
1448 if (match_int(&args[0], &option))
1449 return 0;
1450 if (option < 0)
1451 return 0;
1452 sbi->s_stripe = option;
1453 break;
64769240
AT		 1454 case Opt_delalloc:
1455 set_opt(sbi->s_mount_opt, DELALLOC);
1456 break;
240799cd
TT		 1457 case Opt_inode_readahead_blks:
1458 if (match_int(&args[0], &option))
1459 return 0;
1460 if (option < 0 || option > (1 << 30))
1461 return 0;
1462 sbi->s_inode_readahead_blks = option;
1463 break;
b3881f74
TT		 1464 case Opt_journal_ioprio:
1465 if (match_int(&args[0], &option))
1466 return 0;
1467 if (option < 0 || option > 7)
1468 break;
1469 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1470 option);
1471 break;
ac27a0ec 1472 default:
2b2d6d01
TT		 1473 printk(KERN_ERR
1474 "EXT4-fs: Unrecognized mount option \"%s\" "
1475 "or missing value\n", p);
ac27a0ec
DK		 1476 return 0;
1477 }
1478 }
1479#ifdef CONFIG_QUOTA
1480 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
617ba13b 1481 if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
ac27a0ec
DK		 1482 sbi->s_qf_names[USRQUOTA])
1483 clear_opt(sbi->s_mount_opt, USRQUOTA);
1484
617ba13b 1485 if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
ac27a0ec
DK		 1486 sbi->s_qf_names[GRPQUOTA])
1487 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1488
1489 if ((sbi->s_qf_names[USRQUOTA] &&
617ba13b 1490 (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
ac27a0ec 1491 (sbi->s_qf_names[GRPQUOTA] &&
617ba13b
MC		 1492 (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
1493 printk(KERN_ERR "EXT4-fs: old and new quota "
ac27a0ec
DK		 1494 "format mixing.\n");
1495 return 0;
1496 }
1497
1498 if (!sbi->s_jquota_fmt) {
2c8be6b2 1499 printk(KERN_ERR "EXT4-fs: journaled quota format "
ac27a0ec
DK		 1500 "not specified.\n");
1501 return 0;
1502 }
1503 } else {
1504 if (sbi->s_jquota_fmt) {
2c8be6b2
JK		 1505 printk(KERN_ERR "EXT4-fs: journaled quota format "
1506 "specified with no journaling "
ac27a0ec
DK		 1507 "enabled.\n");
1508 return 0;
1509 }
1510 }
1511#endif
1512 return 1;
1513}
1514
617ba13b 1515static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 1516 int read_only)
1517{
617ba13b 1518 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 1519 int res = 0;
1520
617ba13b 1521 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
2b2d6d01
TT		 1522 printk(KERN_ERR "EXT4-fs warning: revision level too high, "
1523 "forcing read-only mode\n");
ac27a0ec
DK		 1524 res = MS_RDONLY;
1525 }
1526 if (read_only)
1527 return res;
617ba13b 1528 if (!(sbi->s_mount_state & EXT4_VALID_FS))
2b2d6d01
TT		 1529 printk(KERN_WARNING "EXT4-fs warning: mounting unchecked fs, "
1530 "running e2fsck is recommended\n");
617ba13b 1531 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
2b2d6d01
TT		 1532 printk(KERN_WARNING
1533 "EXT4-fs warning: mounting fs with errors, "
1534 "running e2fsck is recommended\n");
ac27a0ec
DK		 1535 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1536 le16_to_cpu(es->s_mnt_count) >=
1537 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
2b2d6d01
TT		 1538 printk(KERN_WARNING
1539 "EXT4-fs warning: maximal mount count reached, "
1540 "running e2fsck is recommended\n");
ac27a0ec
DK		 1541 else if (le32_to_cpu(es->s_checkinterval) &&
1542 (le32_to_cpu(es->s_lastcheck) +
1543 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
2b2d6d01
TT		 1544 printk(KERN_WARNING
1545 "EXT4-fs warning: checktime reached, "
1546 "running e2fsck is recommended\n");
0390131b
FM		 1547 if (!sbi->s_journal)
1548 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec 1549 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 1550 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 1551 le16_add_cpu(&es->s_mnt_count, 1);
ac27a0ec 1552 es->s_mtime = cpu_to_le32(get_seconds());
617ba13b 1553 ext4_update_dynamic_rev(sb);
0390131b
FM		 1554 if (sbi->s_journal)
1555 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 1556
617ba13b 1557 ext4_commit_super(sb, es, 1);
ac27a0ec 1558 if (test_opt(sb, DEBUG))
a9df9a49 1559 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
ac27a0ec
DK		 1560 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
1561 sb->s_blocksize,
1562 sbi->s_groups_count,
617ba13b
MC		 1563 EXT4_BLOCKS_PER_GROUP(sb),
1564 EXT4_INODES_PER_GROUP(sb),
ac27a0ec
DK		 1565 sbi->s_mount_opt);
1566
0390131b
FM		 1567 if (EXT4_SB(sb)->s_journal) {
1568 printk(KERN_INFO "EXT4 FS on %s, %s journal on %s\n",
1569 sb->s_id, EXT4_SB(sb)->s_journal->j_inode ? "internal" :
1570 "external", EXT4_SB(sb)->s_journal->j_devname);
1571 } else {
1572 printk(KERN_INFO "EXT4 FS on %s, no journal\n", sb->s_id);
1573 }
ac27a0ec
DK		 1574 return res;
1575}
1576
772cb7c8
JS		 1577static int ext4_fill_flex_info(struct super_block *sb)
1578{
1579 struct ext4_sb_info *sbi = EXT4_SB(sb);
1580 struct ext4_group_desc *gdp = NULL;
1581 struct buffer_head *bh;
1582 ext4_group_t flex_group_count;
1583 ext4_group_t flex_group;
1584 int groups_per_flex = 0;
772cb7c8
JS		 1585 int i;
1586
1587 if (!sbi->s_es->s_log_groups_per_flex) {
1588 sbi->s_log_groups_per_flex = 0;
1589 return 1;
1590 }
1591
1592 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1593 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1594
c62a11fd
FB		 1595 /* We allocate both existing and potentially added groups */
1596 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
d94e99a6
AK		 1597 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1598 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
ec05e868 1599 sbi->s_flex_groups = kzalloc(flex_group_count *
772cb7c8
JS		 1600 sizeof(struct flex_groups), GFP_KERNEL);
1601 if (sbi->s_flex_groups == NULL) {
ec05e868 1602 printk(KERN_ERR "EXT4-fs: not enough memory for "
a9df9a49 1603 "%u flex groups\n", flex_group_count);
772cb7c8
JS		 1604 goto failed;
1605 }
772cb7c8 1606
772cb7c8
JS		 1607 for (i = 0; i < sbi->s_groups_count; i++) {
1608 gdp = ext4_get_group_desc(sb, i, &bh);
1609
1610 flex_group = ext4_flex_group(sbi, i);
1611 sbi->s_flex_groups[flex_group].free_inodes +=
560671a0 1612 ext4_free_inodes_count(sb, gdp);
772cb7c8 1613 sbi->s_flex_groups[flex_group].free_blocks +=
560671a0 1614 ext4_free_blks_count(sb, gdp);
772cb7c8
JS		 1615 }
1616
1617 return 1;
1618failed:
1619 return 0;
1620}
1621
717d50e4
AD		 1622__le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1623 struct ext4_group_desc *gdp)
1624{
1625 __u16 crc = 0;
1626
1627 if (sbi->s_es->s_feature_ro_compat &
1628 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1629 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1630 __le32 le_group = cpu_to_le32(block_group);
1631
1632 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1633 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1634 crc = crc16(crc, (__u8 *)gdp, offset);
1635 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1636 /* for checksum of struct ext4_group_desc do the rest...*/
1637 if ((sbi->s_es->s_feature_incompat &
1638 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1639 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1640 crc = crc16(crc, (__u8 *)gdp + offset,
1641 le16_to_cpu(sbi->s_es->s_desc_size) -
1642 offset);
1643 }
1644
1645 return cpu_to_le16(crc);
1646}
1647
1648int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1649 struct ext4_group_desc *gdp)
1650{
1651 if ((sbi->s_es->s_feature_ro_compat &
1652 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1653 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1654 return 0;
1655
1656 return 1;
1657}
1658
ac27a0ec 1659/* Called at mount-time, super-block is locked */
197cd65a 1660static int ext4_check_descriptors(struct super_block *sb)
ac27a0ec 1661{
617ba13b
MC		 1662 struct ext4_sb_info *sbi = EXT4_SB(sb);
1663 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1664 ext4_fsblk_t last_block;
bd81d8ee
LV		 1665 ext4_fsblk_t block_bitmap;
1666 ext4_fsblk_t inode_bitmap;
1667 ext4_fsblk_t inode_table;
ce421581 1668 int flexbg_flag = 0;
fd2d4291 1669 ext4_group_t i;
ac27a0ec 1670
ce421581
JS		 1671 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1672 flexbg_flag = 1;
1673
af5bc92d 1674 ext4_debug("Checking group descriptors");
ac27a0ec 1675
197cd65a
AM		 1676 for (i = 0; i < sbi->s_groups_count; i++) {
1677 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1678
ce421581 1679 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 1680 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 1681 else
1682 last_block = first_block +
617ba13b 1683 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 1684
8fadc143 1685 block_bitmap = ext4_block_bitmap(sb, gdp);
2b2d6d01 1686 if (block_bitmap < first_block || block_bitmap > last_block) {
c19204b0 1687 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
a9df9a49 1688 "Block bitmap for group %u not in group "
5128273a 1689 "(block %llu)!\n", i, block_bitmap);
ac27a0ec
DK		 1690 return 0;
1691 }
8fadc143 1692 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2b2d6d01 1693 if (inode_bitmap < first_block || inode_bitmap > last_block) {
c19204b0 1694 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
a9df9a49 1695 "Inode bitmap for group %u not in group "
5128273a 1696 "(block %llu)!\n", i, inode_bitmap);
ac27a0ec
DK		 1697 return 0;
1698 }
8fadc143 1699 inode_table = ext4_inode_table(sb, gdp);
bd81d8ee 1700 if (inode_table < first_block ||
2b2d6d01 1701 inode_table + sbi->s_itb_per_group - 1 > last_block) {
c19204b0 1702 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
a9df9a49 1703 "Inode table for group %u not in group "
5128273a 1704 "(block %llu)!\n", i, inode_table);
ac27a0ec
DK		 1705 return 0;
1706 }
b5f10eed 1707 spin_lock(sb_bgl_lock(sbi, i));
717d50e4 1708 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
c19204b0 1709 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
a9df9a49 1710 "Checksum for group %u failed (%u!=%u)\n",
c19204b0
JB		 1711 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1712 gdp)), le16_to_cpu(gdp->bg_checksum));
7ee1ec4c
LZ		 1713 if (!(sb->s_flags & MS_RDONLY)) {
1714 spin_unlock(sb_bgl_lock(sbi, i));
8a266467 1715 return 0;
7ee1ec4c 1716 }
717d50e4 1717 }
b5f10eed 1718 spin_unlock(sb_bgl_lock(sbi, i));
ce421581
JS		 1719 if (!flexbg_flag)
1720 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec
DK		 1721 }
1722
bd81d8ee 1723 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
2b2d6d01 1724 sbi->s_es->s_free_inodes_count = cpu_to_le32(ext4_count_free_inodes(sb));
ac27a0ec
DK		 1725 return 1;
1726}
1727
617ba13b 1728/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 1729 * the superblock) which were deleted from all directories, but held open by
1730 * a process at the time of a crash. We walk the list and try to delete these
1731 * inodes at recovery time (only with a read-write filesystem).
1732 *
1733 * In order to keep the orphan inode chain consistent during traversal (in
1734 * case of crash during recovery), we link each inode into the superblock
1735 * orphan list_head and handle it the same way as an inode deletion during
1736 * normal operation (which journals the operations for us).
1737 *
1738 * We only do an iget() and an iput() on each inode, which is very safe if we
1739 * accidentally point at an in-use or already deleted inode. The worst that
1740 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 1741 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 1742 * e2fsck was run on this filesystem, and it must have already done the orphan
1743 * inode cleanup for us, so we can safely abort without any further action.
1744 */
2b2d6d01
TT		 1745static void ext4_orphan_cleanup(struct super_block *sb,
1746 struct ext4_super_block *es)
ac27a0ec
DK		 1747{
1748 unsigned int s_flags = sb->s_flags;
1749 int nr_orphans = 0, nr_truncates = 0;
1750#ifdef CONFIG_QUOTA
1751 int i;
1752#endif
1753 if (!es->s_last_orphan) {
1754 jbd_debug(4, "no orphan inodes to clean up\n");
1755 return;
1756 }
1757
a8f48a95
ES		 1758 if (bdev_read_only(sb->s_bdev)) {
1759 printk(KERN_ERR "EXT4-fs: write access "
1760 "unavailable, skipping orphan cleanup.\n");
1761 return;
1762 }
1763
617ba13b 1764 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
ac27a0ec
DK		 1765 if (es->s_last_orphan)
1766 jbd_debug(1, "Errors on filesystem, "
1767 "clearing orphan list.\n");
1768 es->s_last_orphan = 0;
1769 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1770 return;
1771 }
1772
1773 if (s_flags & MS_RDONLY) {
617ba13b 1774 printk(KERN_INFO "EXT4-fs: %s: orphan cleanup on readonly fs\n",
ac27a0ec
DK		 1775 sb->s_id);
1776 sb->s_flags &= ~MS_RDONLY;
1777 }
1778#ifdef CONFIG_QUOTA
1779 /* Needed for iput() to work correctly and not trash data */
1780 sb->s_flags |= MS_ACTIVE;
1781 /* Turn on quotas so that they are updated correctly */
1782 for (i = 0; i < MAXQUOTAS; i++) {
617ba13b
MC		 1783 if (EXT4_SB(sb)->s_qf_names[i]) {
1784 int ret = ext4_quota_on_mount(sb, i);
ac27a0ec
DK		 1785 if (ret < 0)
1786 printk(KERN_ERR
2c8be6b2 1787 "EXT4-fs: Cannot turn on journaled "
ac27a0ec
DK		 1788 "quota: error %d\n", ret);
1789 }
1790 }
1791#endif
1792
1793 while (es->s_last_orphan) {
1794 struct inode *inode;
1795
97bd42b9
JB		 1796 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1797 if (IS_ERR(inode)) {
ac27a0ec
DK		 1798 es->s_last_orphan = 0;
1799 break;
1800 }
1801
617ba13b 1802 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
ac27a0ec
DK		 1803 DQUOT_INIT(inode);
1804 if (inode->i_nlink) {
1805 printk(KERN_DEBUG
e5f8eab8 1806 "%s: truncating inode %lu to %lld bytes\n",
46e665e9 1807 __func__, inode->i_ino, inode->i_size);
e5f8eab8 1808 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
ac27a0ec 1809 inode->i_ino, inode->i_size);
617ba13b 1810 ext4_truncate(inode);
ac27a0ec
DK		 1811 nr_truncates++;
1812 } else {
1813 printk(KERN_DEBUG
1814 "%s: deleting unreferenced inode %lu\n",
46e665e9 1815 __func__, inode->i_ino);
ac27a0ec
DK		 1816 jbd_debug(2, "deleting unreferenced inode %lu\n",
1817 inode->i_ino);
1818 nr_orphans++;
1819 }
1820 iput(inode); /* The delete magic happens here! */
1821 }
1822
2b2d6d01 1823#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
ac27a0ec
DK		 1824
1825 if (nr_orphans)
617ba13b 1826 printk(KERN_INFO "EXT4-fs: %s: %d orphan inode%s deleted\n",
ac27a0ec
DK		 1827 sb->s_id, PLURAL(nr_orphans));
1828 if (nr_truncates)
617ba13b 1829 printk(KERN_INFO "EXT4-fs: %s: %d truncate%s cleaned up\n",
ac27a0ec
DK		 1830 sb->s_id, PLURAL(nr_truncates));
1831#ifdef CONFIG_QUOTA
1832 /* Turn quotas off */
1833 for (i = 0; i < MAXQUOTAS; i++) {
1834 if (sb_dqopt(sb)->files[i])
6f28e087 1835 vfs_quota_off(sb, i, 0);
ac27a0ec
DK		 1836 }
1837#endif
1838 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1839}
cd2291a4
ES		 1840/*
1841 * Maximal extent format file size.
1842 * Resulting logical blkno at s_maxbytes must fit in our on-disk
1843 * extent format containers, within a sector_t, and within i_blocks
1844 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
1845 * so that won't be a limiting factor.
1846 *
1847 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1848 */
f287a1a5 1849static loff_t ext4_max_size(int blkbits, int has_huge_files)
cd2291a4
ES		 1850{
1851 loff_t res;
1852 loff_t upper_limit = MAX_LFS_FILESIZE;
1853
1854 /* small i_blocks in vfs inode? */
f287a1a5 1855 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
cd2291a4 1856 /*
b3a6ffe1 1857 * CONFIG_LBD is not enabled implies the inode
cd2291a4
ES		 1858 * i_block represent total blocks in 512 bytes
1859 * 32 == size of vfs inode i_blocks * 8
1860 */
1861 upper_limit = (1LL << 32) - 1;
1862
1863 /* total blocks in file system block size */
1864 upper_limit >>= (blkbits - 9);
1865 upper_limit <<= blkbits;
1866 }
1867
1868 /* 32-bit extent-start container, ee_block */
1869 res = 1LL << 32;
1870 res <<= blkbits;
1871 res -= 1;
1872
1873 /* Sanity check against vm- & vfs- imposed limits */
1874 if (res > upper_limit)
1875 res = upper_limit;
1876
1877 return res;
1878}
ac27a0ec 1879
ac27a0ec 1880/*
cd2291a4 1881 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 1882 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
1883 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 1884 */
f287a1a5 1885static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
ac27a0ec 1886{
617ba13b 1887 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 1888 int meta_blocks;
1889 loff_t upper_limit;
1890 /* This is calculated to be the largest file size for a
cd2291a4 1891 * dense, bitmapped file such that the total number of
ac27a0ec 1892 * sectors in the file, including data and all indirect blocks,
0fc1b451
AK		 1893 * does not exceed 2^48 -1
1894 * __u32 i_blocks_lo and _u16 i_blocks_high representing the
1895 * total number of 512 bytes blocks of the file
1896 */
1897
f287a1a5 1898 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
0fc1b451 1899 /*
b3a6ffe1 1900 * !has_huge_files or CONFIG_LBD is not enabled
f287a1a5
TT		 1901 * implies the inode i_block represent total blocks in
1902 * 512 bytes 32 == size of vfs inode i_blocks * 8
0fc1b451
AK		 1903 */
1904 upper_limit = (1LL << 32) - 1;
1905
1906 /* total blocks in file system block size */
1907 upper_limit >>= (bits - 9);
1908
1909 } else {
8180a562
AK		 1910 /*
1911 * We use 48 bit ext4_inode i_blocks
1912 * With EXT4_HUGE_FILE_FL set the i_blocks
1913 * represent total number of blocks in
1914 * file system block size
1915 */
0fc1b451
AK		 1916 upper_limit = (1LL << 48) - 1;
1917
0fc1b451
AK		 1918 }
1919
1920 /* indirect blocks */
1921 meta_blocks = 1;
1922 /* double indirect blocks */
1923 meta_blocks += 1 + (1LL << (bits-2));
1924 /* tripple indirect blocks */
1925 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
1926
1927 upper_limit -= meta_blocks;
1928 upper_limit <<= bits;
ac27a0ec
DK		 1929
1930 res += 1LL << (bits-2);
1931 res += 1LL << (2*(bits-2));
1932 res += 1LL << (3*(bits-2));
1933 res <<= bits;
1934 if (res > upper_limit)
1935 res = upper_limit;
0fc1b451
AK		 1936
1937 if (res > MAX_LFS_FILESIZE)
1938 res = MAX_LFS_FILESIZE;
1939
ac27a0ec
DK		 1940 return res;
1941}
1942
617ba13b 1943static ext4_fsblk_t descriptor_loc(struct super_block *sb,
70bbb3e0 1944 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 1945{
617ba13b 1946 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 1947 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 1948 int has_super = 0;
1949
1950 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1951
617ba13b 1952 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
ac27a0ec 1953 nr < first_meta_bg)
70bbb3e0 1954 return logical_sb_block + nr + 1;
ac27a0ec 1955 bg = sbi->s_desc_per_block * nr;
617ba13b 1956 if (ext4_bg_has_super(sb, bg))
ac27a0ec 1957 has_super = 1;
617ba13b 1958 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 1959}
1960
c9de560d
AT		 1961/**
1962 * ext4_get_stripe_size: Get the stripe size.
1963 * @sbi: In memory super block info
1964 *
1965 * If we have specified it via mount option, then
1966 * use the mount option value. If the value specified at mount time is
1967 * greater than the blocks per group use the super block value.
1968 * If the super block value is greater than blocks per group return 0.
1969 * Allocator needs it be less than blocks per group.
1970 *
1971 */
1972static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
1973{
1974 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
1975 unsigned long stripe_width =
1976 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
1977
1978 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
1979 return sbi->s_stripe;
1980
1981 if (stripe_width <= sbi->s_blocks_per_group)
1982 return stripe_width;
1983
1984 if (stride <= sbi->s_blocks_per_group)
1985 return stride;
1986
1987 return 0;
1988}
ac27a0ec 1989
2b2d6d01 1990static int ext4_fill_super(struct super_block *sb, void *data, int silent)
7477827f
AK		 1991 __releases(kernel_lock)
1992 __acquires(kernel_lock)
1d03ec98 1993
ac27a0ec 1994{
2b2d6d01 1995 struct buffer_head *bh;
617ba13b
MC		 1996 struct ext4_super_block *es = NULL;
1997 struct ext4_sb_info *sbi;
1998 ext4_fsblk_t block;
1999 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 2000 ext4_fsblk_t logical_sb_block;
ac27a0ec 2001 unsigned long offset = 0;
ac27a0ec
DK		 2002 unsigned long journal_devnum = 0;
2003 unsigned long def_mount_opts;
2004 struct inode *root;
9f6200bb 2005 char *cp;
0390131b 2006 const char *descr;
1d1fe1ee 2007 int ret = -EINVAL;
ac27a0ec 2008 int blocksize;
4ec11028
TT		 2009 unsigned int db_count;
2010 unsigned int i;
f287a1a5 2011 int needs_recovery, has_huge_files;
3a06d778 2012 int features;
bd81d8ee 2013 __u64 blocks_count;
833f4077 2014 int err;
b3881f74 2015 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
ac27a0ec
DK		 2016
2017 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
2018 if (!sbi)
2019 return -ENOMEM;
2020 sb->s_fs_info = sbi;
2021 sbi->s_mount_opt = 0;
617ba13b
MC		 2022 sbi->s_resuid = EXT4_DEF_RESUID;
2023 sbi->s_resgid = EXT4_DEF_RESGID;
240799cd 2024 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
d9c9bef1 2025 sbi->s_sb_block = sb_block;
ac27a0ec
DK		 2026
2027 unlock_kernel();
2028
9f6200bb
TT		 2029 /* Cleanup superblock name */
2030 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
2031 *cp = '!';
2032
617ba13b 2033 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 2034 if (!blocksize) {
617ba13b 2035 printk(KERN_ERR "EXT4-fs: unable to set blocksize\n");
ac27a0ec
DK		 2036 goto out_fail;
2037 }
2038
2039 /*
617ba13b 2040 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 2041 * block sizes. We need to calculate the offset from buffer start.
2042 */
617ba13b 2043 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 2044 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2045 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 2046 } else {
70bbb3e0 2047 logical_sb_block = sb_block;
ac27a0ec
DK		 2048 }
2049
70bbb3e0 2050 if (!(bh = sb_bread(sb, logical_sb_block))) {
2b2d6d01 2051 printk(KERN_ERR "EXT4-fs: unable to read superblock\n");
ac27a0ec
DK		 2052 goto out_fail;
2053 }
2054 /*
2055 * Note: s_es must be initialized as soon as possible because
617ba13b 2056 * some ext4 macro-instructions depend on its value
ac27a0ec 2057 */
617ba13b 2058 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
ac27a0ec
DK		 2059 sbi->s_es = es;
2060 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 2061 if (sb->s_magic != EXT4_SUPER_MAGIC)
2062 goto cantfind_ext4;
ac27a0ec
DK		 2063
2064 /* Set defaults before we parse the mount options */
2065 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
617ba13b 2066 if (def_mount_opts & EXT4_DEFM_DEBUG)
ac27a0ec 2067 set_opt(sbi->s_mount_opt, DEBUG);
617ba13b 2068 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
ac27a0ec 2069 set_opt(sbi->s_mount_opt, GRPID);
617ba13b 2070 if (def_mount_opts & EXT4_DEFM_UID16)
ac27a0ec 2071 set_opt(sbi->s_mount_opt, NO_UID32);
03010a33 2072#ifdef CONFIG_EXT4_FS_XATTR
617ba13b 2073 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
ac27a0ec 2074 set_opt(sbi->s_mount_opt, XATTR_USER);
2e7842b8 2075#endif
03010a33 2076#ifdef CONFIG_EXT4_FS_POSIX_ACL
617ba13b 2077 if (def_mount_opts & EXT4_DEFM_ACL)
ac27a0ec 2078 set_opt(sbi->s_mount_opt, POSIX_ACL);
2e7842b8 2079#endif
617ba13b
MC		 2080 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
2081 sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
2082 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
2083 sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
2084 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
2085 sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
2086
2087 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
ac27a0ec 2088 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
bb4f397a 2089 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
ceea16bf 2090 set_opt(sbi->s_mount_opt, ERRORS_CONT);
bb4f397a
AK		 2091 else
2092 set_opt(sbi->s_mount_opt, ERRORS_RO);
ac27a0ec
DK		 2093
2094 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
2095 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
30773840
TT		 2096 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
2097 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
2098 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
ac27a0ec
DK		 2099
2100 set_opt(sbi->s_mount_opt, RESERVATION);
571640ca 2101 set_opt(sbi->s_mount_opt, BARRIER);
ac27a0ec 2102
dd919b98
AK		 2103 /*
2104 * enable delayed allocation by default
2105 * Use -o nodelalloc to turn it off
2106 */
2107 set_opt(sbi->s_mount_opt, DELALLOC);
2108
2109
b3881f74
TT		 2110 if (!parse_options((char *) data, sb, &journal_devnum,
2111 &journal_ioprio, NULL, 0))
ac27a0ec
DK		 2112 goto failed_mount;
2113
2114 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
617ba13b 2115 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec 2116
617ba13b
MC		 2117 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
2118 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
2119 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
2120 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
ac27a0ec 2121 printk(KERN_WARNING
617ba13b 2122 "EXT4-fs warning: feature flags set on rev 0 fs, "
ac27a0ec 2123 "running e2fsck is recommended\n");
469108ff 2124
ac27a0ec
DK		 2125 /*
2126 * Check feature flags regardless of the revision level, since we
2127 * previously didn't change the revision level when setting the flags,
2128 * so there is a chance incompat flags are set on a rev 0 filesystem.
2129 */
617ba13b 2130 features = EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP);
ac27a0ec 2131 if (features) {
617ba13b 2132 printk(KERN_ERR "EXT4-fs: %s: couldn't mount because of "
3a06d778
AK		 2133 "unsupported optional features (%x).\n", sb->s_id,
2134 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2135 ~EXT4_FEATURE_INCOMPAT_SUPP));
ac27a0ec
DK		 2136 goto failed_mount;
2137 }
617ba13b 2138 features = EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP);
ac27a0ec 2139 if (!(sb->s_flags & MS_RDONLY) && features) {
617ba13b 2140 printk(KERN_ERR "EXT4-fs: %s: couldn't mount RDWR because of "
3a06d778
AK		 2141 "unsupported optional features (%x).\n", sb->s_id,
2142 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2143 ~EXT4_FEATURE_RO_COMPAT_SUPP));
ac27a0ec
DK		 2144 goto failed_mount;
2145 }
f287a1a5
TT		 2146 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2147 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
2148 if (has_huge_files) {
0fc1b451
AK		 2149 /*
2150 * Large file size enabled file system can only be
b3a6ffe1 2151 * mount if kernel is build with CONFIG_LBD
0fc1b451
AK		 2152 */
2153 if (sizeof(root->i_blocks) < sizeof(u64) &&
2154 !(sb->s_flags & MS_RDONLY)) {
2155 printk(KERN_ERR "EXT4-fs: %s: Filesystem with huge "
2156 "files cannot be mounted read-write "
b3a6ffe1 2157 "without CONFIG_LBD.\n", sb->s_id);
0fc1b451
AK		 2158 goto failed_mount;
2159 }
2160 }
ac27a0ec
DK		 2161 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
2162
617ba13b
MC		 2163 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
2164 blocksize > EXT4_MAX_BLOCK_SIZE) {
ac27a0ec 2165 printk(KERN_ERR
617ba13b 2166 "EXT4-fs: Unsupported filesystem blocksize %d on %s.\n",
ac27a0ec
DK		 2167 blocksize, sb->s_id);
2168 goto failed_mount;
2169 }
2170
ac27a0ec 2171 if (sb->s_blocksize != blocksize) {
ce40733c
AK		 2172
2173 /* Validate the filesystem blocksize */
2174 if (!sb_set_blocksize(sb, blocksize)) {
2175 printk(KERN_ERR "EXT4-fs: bad block size %d.\n",
2176 blocksize);
ac27a0ec
DK		 2177 goto failed_mount;
2178 }
2179
2b2d6d01 2180 brelse(bh);
70bbb3e0
AM		 2181 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2182 offset = do_div(logical_sb_block, blocksize);
2183 bh = sb_bread(sb, logical_sb_block);
ac27a0ec
DK		 2184 if (!bh) {
2185 printk(KERN_ERR
617ba13b 2186 "EXT4-fs: Can't read superblock on 2nd try.\n");
ac27a0ec
DK		 2187 goto failed_mount;
2188 }
617ba13b 2189 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
ac27a0ec 2190 sbi->s_es = es;
617ba13b 2191 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
2b2d6d01
TT		 2192 printk(KERN_ERR
2193 "EXT4-fs: Magic mismatch, very weird !\n");
ac27a0ec
DK		 2194 goto failed_mount;
2195 }
2196 }
2197
f287a1a5
TT		 2198 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
2199 has_huge_files);
2200 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
ac27a0ec 2201
617ba13b
MC		 2202 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2203 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2204 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
ac27a0ec
DK		 2205 } else {
2206 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2207 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
617ba13b 2208 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1330593e 2209 (!is_power_of_2(sbi->s_inode_size)) ||
ac27a0ec 2210 (sbi->s_inode_size > blocksize)) {
2b2d6d01
TT		 2211 printk(KERN_ERR
2212 "EXT4-fs: unsupported inode size: %d\n",
2213 sbi->s_inode_size);
ac27a0ec
DK		 2214 goto failed_mount;
2215 }
ef7f3835
KS		 2216 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2217 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
ac27a0ec 2218 }
0d1ee42f
AR		 2219 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2220 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
8fadc143 2221 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 2222 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 2223 !is_power_of_2(sbi->s_desc_size)) {
0d1ee42f 2224 printk(KERN_ERR
8fadc143 2225 "EXT4-fs: unsupported descriptor size %lu\n",
0d1ee42f
AR		 2226 sbi->s_desc_size);
2227 goto failed_mount;
2228 }
2229 } else
2230 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
ac27a0ec 2231 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 2232 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
b47b6f38 2233 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
617ba13b
MC		 2234 goto cantfind_ext4;
2235 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 2236 if (sbi->s_inodes_per_block == 0)
617ba13b 2237 goto cantfind_ext4;
ac27a0ec
DK		 2238 sbi->s_itb_per_group = sbi->s_inodes_per_group /
2239 sbi->s_inodes_per_block;
0d1ee42f 2240 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 2241 sbi->s_sbh = bh;
2242 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 2243 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2244 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
2b2d6d01 2245 for (i = 0; i < 4; i++)
ac27a0ec
DK		 2246 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2247 sbi->s_def_hash_version = es->s_def_hash_version;
f99b2589
TT		 2248 i = le32_to_cpu(es->s_flags);
2249 if (i & EXT2_FLAGS_UNSIGNED_HASH)
2250 sbi->s_hash_unsigned = 3;
2251 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
2252#ifdef __CHAR_UNSIGNED__
2253 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
2254 sbi->s_hash_unsigned = 3;
2255#else
2256 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
2257#endif
2258 sb->s_dirt = 1;
2259 }
ac27a0ec
DK		 2260
2261 if (sbi->s_blocks_per_group > blocksize * 8) {
2b2d6d01
TT		 2262 printk(KERN_ERR
2263 "EXT4-fs: #blocks per group too big: %lu\n",
2264 sbi->s_blocks_per_group);
ac27a0ec
DK		 2265 goto failed_mount;
2266 }
ac27a0ec 2267 if (sbi->s_inodes_per_group > blocksize * 8) {
2b2d6d01
TT		 2268 printk(KERN_ERR
2269 "EXT4-fs: #inodes per group too big: %lu\n",
2270 sbi->s_inodes_per_group);
ac27a0ec
DK		 2271 goto failed_mount;
2272 }
2273
bd81d8ee 2274 if (ext4_blocks_count(es) >
ac27a0ec 2275 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
617ba13b 2276 printk(KERN_ERR "EXT4-fs: filesystem on %s:"
ac27a0ec
DK		 2277 " too large to mount safely\n", sb->s_id);
2278 if (sizeof(sector_t) < 8)
617ba13b 2279 printk(KERN_WARNING "EXT4-fs: CONFIG_LBD not "
ac27a0ec
DK		 2280 "enabled\n");
2281 goto failed_mount;
2282 }
2283
617ba13b
MC		 2284 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2285 goto cantfind_ext4;
e7c95593 2286
4ec11028
TT		 2287 /*
2288 * It makes no sense for the first data block to be beyond the end
2289 * of the filesystem.
2290 */
2291 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
2292 printk(KERN_WARNING "EXT4-fs: bad geometry: first data"
2293 "block %u is beyond end of filesystem (%llu)\n",
2294 le32_to_cpu(es->s_first_data_block),
2295 ext4_blocks_count(es));
e7c95593
ES		 2296 goto failed_mount;
2297 }
bd81d8ee
LV		 2298 blocks_count = (ext4_blocks_count(es) -
2299 le32_to_cpu(es->s_first_data_block) +
2300 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2301 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4ec11028
TT		 2302 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
2303 printk(KERN_WARNING "EXT4-fs: groups count too large: %u "
2304 "(block count %llu, first data block %u, "
2305 "blocks per group %lu)\n", sbi->s_groups_count,
2306 ext4_blocks_count(es),
2307 le32_to_cpu(es->s_first_data_block),
2308 EXT4_BLOCKS_PER_GROUP(sb));
2309 goto failed_mount;
2310 }
bd81d8ee 2311 sbi->s_groups_count = blocks_count;
617ba13b
MC		 2312 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2313 EXT4_DESC_PER_BLOCK(sb);
2b2d6d01 2314 sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
ac27a0ec
DK		 2315 GFP_KERNEL);
2316 if (sbi->s_group_desc == NULL) {
2b2d6d01 2317 printk(KERN_ERR "EXT4-fs: not enough memory\n");
ac27a0ec
DK		 2318 goto failed_mount;
2319 }
2320
3244fcb1 2321#ifdef CONFIG_PROC_FS
9f6200bb
TT		 2322 if (ext4_proc_root)
2323 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
2324
240799cd
TT		 2325 if (sbi->s_proc)
2326 proc_create_data("inode_readahead_blks", 0644, sbi->s_proc,
2327 &ext4_ui_proc_fops,
2328 &sbi->s_inode_readahead_blks);
3244fcb1 2329#endif
240799cd 2330
ac27a0ec
DK		 2331 bgl_lock_init(&sbi->s_blockgroup_lock);
2332
2333 for (i = 0; i < db_count; i++) {
70bbb3e0 2334 block = descriptor_loc(sb, logical_sb_block, i);
ac27a0ec
DK		 2335 sbi->s_group_desc[i] = sb_bread(sb, block);
2336 if (!sbi->s_group_desc[i]) {
2b2d6d01
TT		 2337 printk(KERN_ERR "EXT4-fs: "
2338 "can't read group descriptor %d\n", i);
ac27a0ec
DK		 2339 db_count = i;
2340 goto failed_mount2;
2341 }
2342 }
2b2d6d01 2343 if (!ext4_check_descriptors(sb)) {
617ba13b 2344 printk(KERN_ERR "EXT4-fs: group descriptors corrupted!\n");
ac27a0ec
DK		 2345 goto failed_mount2;
2346 }
772cb7c8
JS		 2347 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2348 if (!ext4_fill_flex_info(sb)) {
2349 printk(KERN_ERR
2350 "EXT4-fs: unable to initialize "
2351 "flex_bg meta info!\n");
2352 goto failed_mount2;
2353 }
2354
ac27a0ec
DK		 2355 sbi->s_gdb_count = db_count;
2356 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2357 spin_lock_init(&sbi->s_next_gen_lock);
2358
833f4077
PZ		 2359 err = percpu_counter_init(&sbi->s_freeblocks_counter,
2360 ext4_count_free_blocks(sb));
2361 if (!err) {
2362 err = percpu_counter_init(&sbi->s_freeinodes_counter,
2363 ext4_count_free_inodes(sb));
2364 }
2365 if (!err) {
2366 err = percpu_counter_init(&sbi->s_dirs_counter,
2367 ext4_count_dirs(sb));
2368 }
6bc6e63f
AK		 2369 if (!err) {
2370 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
2371 }
833f4077
PZ		 2372 if (err) {
2373 printk(KERN_ERR "EXT4-fs: insufficient memory\n");
2374 goto failed_mount3;
2375 }
ac27a0ec 2376
c9de560d
AT		 2377 sbi->s_stripe = ext4_get_stripe_size(sbi);
2378
ac27a0ec
DK		 2379 /*
2380 * set up enough so that it can read an inode
2381 */
617ba13b
MC		 2382 sb->s_op = &ext4_sops;
2383 sb->s_export_op = &ext4_export_ops;
2384 sb->s_xattr = ext4_xattr_handlers;
ac27a0ec 2385#ifdef CONFIG_QUOTA
617ba13b
MC		 2386 sb->s_qcop = &ext4_qctl_operations;
2387 sb->dq_op = &ext4_quota_operations;
ac27a0ec
DK		 2388#endif
2389 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
2390
2391 sb->s_root = NULL;
2392
2393 needs_recovery = (es->s_last_orphan != 0 ||
617ba13b
MC		 2394 EXT4_HAS_INCOMPAT_FEATURE(sb,
2395 EXT4_FEATURE_INCOMPAT_RECOVER));
ac27a0ec
DK		 2396
2397 /*
2398 * The first inode we look at is the journal inode. Don't try
2399 * root first: it may be modified in the journal!
2400 */
2401 if (!test_opt(sb, NOLOAD) &&
617ba13b
MC		 2402 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2403 if (ext4_load_journal(sb, es, journal_devnum))
ac27a0ec 2404 goto failed_mount3;
624080ed
TT		 2405 if (!(sb->s_flags & MS_RDONLY) &&
2406 EXT4_SB(sb)->s_journal->j_failed_commit) {
2407 printk(KERN_CRIT "EXT4-fs error (device %s): "
2408 "ext4_fill_super: Journal transaction "
2b2d6d01 2409 "%u is corrupt\n", sb->s_id,
624080ed 2410 EXT4_SB(sb)->s_journal->j_failed_commit);
2b2d6d01
TT		 2411 if (test_opt(sb, ERRORS_RO)) {
2412 printk(KERN_CRIT
2413 "Mounting filesystem read-only\n");
624080ed
TT		 2414 sb->s_flags |= MS_RDONLY;
2415 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2416 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2417 }
2418 if (test_opt(sb, ERRORS_PANIC)) {
2419 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2420 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2421 ext4_commit_super(sb, es, 1);
624080ed
TT		 2422 goto failed_mount4;
2423 }
2424 }
0390131b
FM		 2425 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
2426 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
2427 printk(KERN_ERR "EXT4-fs: required journal recovery "
2428 "suppressed and not mounted read-only\n");
2429 goto failed_mount4;
ac27a0ec 2430 } else {
0390131b
FM		 2431 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
2432 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
2433 sbi->s_journal = NULL;
2434 needs_recovery = 0;
2435 goto no_journal;
ac27a0ec
DK		 2436 }
2437
eb40a09c
JS		 2438 if (ext4_blocks_count(es) > 0xffffffffULL &&
2439 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2440 JBD2_FEATURE_INCOMPAT_64BIT)) {
abda1418 2441 printk(KERN_ERR "EXT4-fs: Failed to set 64-bit journal feature\n");
eb40a09c
JS		 2442 goto failed_mount4;
2443 }
2444
818d276c
GS		 2445 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2446 jbd2_journal_set_features(sbi->s_journal,
2447 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2448 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2449 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2450 jbd2_journal_set_features(sbi->s_journal,
2451 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2452 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2453 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2454 } else {
2455 jbd2_journal_clear_features(sbi->s_journal,
2456 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2457 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2458 }
2459
ac27a0ec
DK		 2460 /* We have now updated the journal if required, so we can
2461 * validate the data journaling mode. */
2462 switch (test_opt(sb, DATA_FLAGS)) {
2463 case 0:
2464 /* No mode set, assume a default based on the journal
63f57933
AM		 2465 * capabilities: ORDERED_DATA if the journal can
2466 * cope, else JOURNAL_DATA
2467 */
dab291af
MC		 2468 if (jbd2_journal_check_available_features
2469 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
ac27a0ec
DK		 2470 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2471 else
2472 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2473 break;
2474
617ba13b
MC		 2475 case EXT4_MOUNT_ORDERED_DATA:
2476 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 2477 if (!jbd2_journal_check_available_features
2478 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
617ba13b 2479 printk(KERN_ERR "EXT4-fs: Journal does not support "
ac27a0ec
DK		 2480 "requested data journaling mode\n");
2481 goto failed_mount4;
2482 }
2483 default:
2484 break;
2485 }
b3881f74 2486 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
ac27a0ec 2487
0390131b 2488no_journal:
ac27a0ec
DK		 2489
2490 if (test_opt(sb, NOBH)) {
617ba13b
MC		 2491 if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
2492 printk(KERN_WARNING "EXT4-fs: Ignoring nobh option - "
ac27a0ec
DK		 2493 "its supported only with writeback mode\n");
2494 clear_opt(sbi->s_mount_opt, NOBH);
2495 }
2496 }
2497 /*
dab291af 2498 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 2499 * so we can safely mount the rest of the filesystem now.
2500 */
2501
1d1fe1ee
DH		 2502 root = ext4_iget(sb, EXT4_ROOT_INO);
2503 if (IS_ERR(root)) {
617ba13b 2504 printk(KERN_ERR "EXT4-fs: get root inode failed\n");
1d1fe1ee 2505 ret = PTR_ERR(root);
ac27a0ec
DK		 2506 goto failed_mount4;
2507 }
2508 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1d1fe1ee 2509 iput(root);
617ba13b 2510 printk(KERN_ERR "EXT4-fs: corrupt root inode, run e2fsck\n");
ac27a0ec
DK		 2511 goto failed_mount4;
2512 }
1d1fe1ee
DH		 2513 sb->s_root = d_alloc_root(root);
2514 if (!sb->s_root) {
2515 printk(KERN_ERR "EXT4-fs: get root dentry failed\n");
2516 iput(root);
2517 ret = -ENOMEM;
2518 goto failed_mount4;
2519 }
ac27a0ec 2520
2b2d6d01 2521 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
ef7f3835
KS		 2522
2523 /* determine the minimum size of new large inodes, if present */
2524 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2525 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2526 EXT4_GOOD_OLD_INODE_SIZE;
2527 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2528 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2529 if (sbi->s_want_extra_isize <
2530 le16_to_cpu(es->s_want_extra_isize))
2531 sbi->s_want_extra_isize =
2532 le16_to_cpu(es->s_want_extra_isize);
2533 if (sbi->s_want_extra_isize <
2534 le16_to_cpu(es->s_min_extra_isize))
2535 sbi->s_want_extra_isize =
2536 le16_to_cpu(es->s_min_extra_isize);
2537 }
2538 }
2539 /* Check if enough inode space is available */
2540 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2541 sbi->s_inode_size) {
2542 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2543 EXT4_GOOD_OLD_INODE_SIZE;
2544 printk(KERN_INFO "EXT4-fs: required extra inode space not"
2545 "available.\n");
2546 }
2547
c2774d84
AK		 2548 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
2549 printk(KERN_WARNING "EXT4-fs: Ignoring delalloc option - "
2550 "requested data journaling mode\n");
2551 clear_opt(sbi->s_mount_opt, DELALLOC);
2552 } else if (test_opt(sb, DELALLOC))
2553 printk(KERN_INFO "EXT4-fs: delayed allocation enabled\n");
2554
2555 ext4_ext_init(sb);
2556 err = ext4_mb_init(sb, needs_recovery);
2557 if (err) {
2558 printk(KERN_ERR "EXT4-fs: failed to initalize mballoc (%d)\n",
2559 err);
2560 goto failed_mount4;
2561 }
2562
ac27a0ec
DK		 2563 /*
2564 * akpm: core read_super() calls in here with the superblock locked.
2565 * That deadlocks, because orphan cleanup needs to lock the superblock
2566 * in numerous places. Here we just pop the lock - it's relatively
2567 * harmless, because we are now ready to accept write_super() requests,
2568 * and aviro says that's the only reason for hanging onto the
2569 * superblock lock.
2570 */
617ba13b
MC		 2571 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
2572 ext4_orphan_cleanup(sb, es);
2573 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
0390131b 2574 if (needs_recovery) {
2b2d6d01 2575 printk(KERN_INFO "EXT4-fs: recovery complete.\n");
0390131b
FM		 2576 ext4_mark_recovery_complete(sb, es);
2577 }
2578 if (EXT4_SB(sb)->s_journal) {
2579 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2580 descr = " journalled data mode";
2581 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2582 descr = " ordered data mode";
2583 else
2584 descr = " writeback data mode";
2585 } else
2586 descr = "out journal";
2587
2588 printk(KERN_INFO "EXT4-fs: mounted filesystem %s with%s\n",
2589 sb->s_id, descr);
ac27a0ec
DK		 2590
2591 lock_kernel();
2592 return 0;
2593
617ba13b 2594cantfind_ext4:
ac27a0ec 2595 if (!silent)
617ba13b 2596 printk(KERN_ERR "VFS: Can't find ext4 filesystem on dev %s.\n",
ac27a0ec
DK		 2597 sb->s_id);
2598 goto failed_mount;
2599
2600failed_mount4:
0390131b
FM		 2601 printk(KERN_ERR "EXT4-fs (device %s): mount failed\n", sb->s_id);
2602 if (sbi->s_journal) {
2603 jbd2_journal_destroy(sbi->s_journal);
2604 sbi->s_journal = NULL;
2605 }
ac27a0ec
DK		 2606failed_mount3:
2607 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2608 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2609 percpu_counter_destroy(&sbi->s_dirs_counter);
6bc6e63f 2610 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
ac27a0ec
DK		 2611failed_mount2:
2612 for (i = 0; i < db_count; i++)
2613 brelse(sbi->s_group_desc[i]);
2614 kfree(sbi->s_group_desc);
2615failed_mount:
240799cd
TT		 2616 if (sbi->s_proc) {
2617 remove_proc_entry("inode_readahead_blks", sbi->s_proc);
9f6200bb 2618 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 2619 }
ac27a0ec
DK		 2620#ifdef CONFIG_QUOTA
2621 for (i = 0; i < MAXQUOTAS; i++)
2622 kfree(sbi->s_qf_names[i]);
2623#endif
617ba13b 2624 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 2625 brelse(bh);
2626out_fail:
2627 sb->s_fs_info = NULL;
2628 kfree(sbi);
2629 lock_kernel();
1d1fe1ee 2630 return ret;
ac27a0ec
DK		 2631}
2632
2633/*
2634 * Setup any per-fs journal parameters now. We'll do this both on
2635 * initial mount, once the journal has been initialised but before we've
2636 * done any recovery; and again on any subsequent remount.
2637 */
617ba13b 2638static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 2639{
617ba13b 2640 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 2641
30773840
TT		 2642 journal->j_commit_interval = sbi->s_commit_interval;
2643 journal->j_min_batch_time = sbi->s_min_batch_time;
2644 journal->j_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 2645
2646 spin_lock(&journal->j_state_lock);
2647 if (test_opt(sb, BARRIER))
dab291af 2648 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 2649 else
dab291af 2650 journal->j_flags &= ~JBD2_BARRIER;
5bf5683a
HK		 2651 if (test_opt(sb, DATA_ERR_ABORT))
2652 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
2653 else
2654 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
ac27a0ec
DK		 2655 spin_unlock(&journal->j_state_lock);
2656}
2657
617ba13b 2658static journal_t *ext4_get_journal(struct super_block *sb,
ac27a0ec
DK		 2659 unsigned int journal_inum)
2660{
2661 struct inode *journal_inode;
2662 journal_t *journal;
2663
0390131b
FM		 2664 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
2665
ac27a0ec
DK		 2666 /* First, test for the existence of a valid inode on disk. Bad
2667 * things happen if we iget() an unused inode, as the subsequent
2668 * iput() will try to delete it. */
2669
1d1fe1ee
DH		 2670 journal_inode = ext4_iget(sb, journal_inum);
2671 if (IS_ERR(journal_inode)) {
617ba13b 2672 printk(KERN_ERR "EXT4-fs: no journal found.\n");
ac27a0ec
DK		 2673 return NULL;
2674 }
2675 if (!journal_inode->i_nlink) {
2676 make_bad_inode(journal_inode);
2677 iput(journal_inode);
617ba13b 2678 printk(KERN_ERR "EXT4-fs: journal inode is deleted.\n");
ac27a0ec
DK		 2679 return NULL;
2680 }
2681
e5f8eab8 2682 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
ac27a0ec 2683 journal_inode, journal_inode->i_size);
1d1fe1ee 2684 if (!S_ISREG(journal_inode->i_mode)) {
617ba13b 2685 printk(KERN_ERR "EXT4-fs: invalid journal inode.\n");
ac27a0ec
DK		 2686 iput(journal_inode);
2687 return NULL;
2688 }
2689
dab291af 2690 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 2691 if (!journal) {
617ba13b 2692 printk(KERN_ERR "EXT4-fs: Could not load journal inode\n");
ac27a0ec
DK		 2693 iput(journal_inode);
2694 return NULL;
2695 }
2696 journal->j_private = sb;
617ba13b 2697 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 2698 return journal;
2699}
2700
617ba13b 2701static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 2702 dev_t j_dev)
2703{
2b2d6d01 2704 struct buffer_head *bh;
ac27a0ec 2705 journal_t *journal;
617ba13b
MC		 2706 ext4_fsblk_t start;
2707 ext4_fsblk_t len;
ac27a0ec 2708 int hblock, blocksize;
617ba13b 2709 ext4_fsblk_t sb_block;
ac27a0ec 2710 unsigned long offset;
2b2d6d01 2711 struct ext4_super_block *es;
ac27a0ec
DK		 2712 struct block_device *bdev;
2713
0390131b
FM		 2714 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
2715
617ba13b 2716 bdev = ext4_blkdev_get(j_dev);
ac27a0ec
DK		 2717 if (bdev == NULL)
2718 return NULL;
2719
2720 if (bd_claim(bdev, sb)) {
2721 printk(KERN_ERR
abda1418 2722 "EXT4-fs: failed to claim external journal device.\n");
9a1c3542 2723 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
ac27a0ec
DK		 2724 return NULL;
2725 }
2726
2727 blocksize = sb->s_blocksize;
2728 hblock = bdev_hardsect_size(bdev);
2729 if (blocksize < hblock) {
2730 printk(KERN_ERR
617ba13b 2731 "EXT4-fs: blocksize too small for journal device.\n");
ac27a0ec
DK		 2732 goto out_bdev;
2733 }
2734
617ba13b
MC		 2735 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
2736 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 2737 set_blocksize(bdev, blocksize);
2738 if (!(bh = __bread(bdev, sb_block, blocksize))) {
617ba13b 2739 printk(KERN_ERR "EXT4-fs: couldn't read superblock of "
ac27a0ec
DK		 2740 "external journal\n");
2741 goto out_bdev;
2742 }
2743
617ba13b
MC		 2744 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2745 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 2746 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b
MC		 2747 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
2748 printk(KERN_ERR "EXT4-fs: external journal has "
ac27a0ec
DK		 2749 "bad superblock\n");
2750 brelse(bh);
2751 goto out_bdev;
2752 }
2753
617ba13b
MC		 2754 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
2755 printk(KERN_ERR "EXT4-fs: journal UUID does not match\n");
ac27a0ec
DK		 2756 brelse(bh);
2757 goto out_bdev;
2758 }
2759
bd81d8ee 2760 len = ext4_blocks_count(es);
ac27a0ec
DK		 2761 start = sb_block + 1;
2762 brelse(bh); /* we're done with the superblock */
2763
dab291af 2764 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 2765 start, len, blocksize);
2766 if (!journal) {
617ba13b 2767 printk(KERN_ERR "EXT4-fs: failed to create device journal\n");
ac27a0ec
DK		 2768 goto out_bdev;
2769 }
2770 journal->j_private = sb;
2771 ll_rw_block(READ, 1, &journal->j_sb_buffer);
2772 wait_on_buffer(journal->j_sb_buffer);
2773 if (!buffer_uptodate(journal->j_sb_buffer)) {
617ba13b 2774 printk(KERN_ERR "EXT4-fs: I/O error on journal device\n");
ac27a0ec
DK		 2775 goto out_journal;
2776 }
2777 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
617ba13b 2778 printk(KERN_ERR "EXT4-fs: External journal has more than one "
ac27a0ec
DK		 2779 "user (unsupported) - %d\n",
2780 be32_to_cpu(journal->j_superblock->s_nr_users));
2781 goto out_journal;
2782 }
617ba13b
MC		 2783 EXT4_SB(sb)->journal_bdev = bdev;
2784 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 2785 return journal;
2786out_journal:
dab291af 2787 jbd2_journal_destroy(journal);
ac27a0ec 2788out_bdev:
617ba13b 2789 ext4_blkdev_put(bdev);
ac27a0ec
DK		 2790 return NULL;
2791}
2792
617ba13b
MC		 2793static int ext4_load_journal(struct super_block *sb,
2794 struct ext4_super_block *es,
ac27a0ec
DK		 2795 unsigned long journal_devnum)
2796{
2797 journal_t *journal;
2798 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
2799 dev_t journal_dev;
2800 int err = 0;
2801 int really_read_only;
2802
0390131b
FM		 2803 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
2804
ac27a0ec
DK		 2805 if (journal_devnum &&
2806 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
617ba13b 2807 printk(KERN_INFO "EXT4-fs: external journal device major/minor "
ac27a0ec
DK		 2808 "numbers have changed\n");
2809 journal_dev = new_decode_dev(journal_devnum);
2810 } else
2811 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
2812
2813 really_read_only = bdev_read_only(sb->s_bdev);
2814
2815 /*
2816 * Are we loading a blank journal or performing recovery after a
2817 * crash? For recovery, we need to check in advance whether we
2818 * can get read-write access to the device.
2819 */
2820
617ba13b 2821 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
ac27a0ec 2822 if (sb->s_flags & MS_RDONLY) {
617ba13b 2823 printk(KERN_INFO "EXT4-fs: INFO: recovery "
ac27a0ec
DK		 2824 "required on readonly filesystem.\n");
2825 if (really_read_only) {
617ba13b 2826 printk(KERN_ERR "EXT4-fs: write access "
ac27a0ec
DK		 2827 "unavailable, cannot proceed.\n");
2828 return -EROFS;
2829 }
2b2d6d01
TT		 2830 printk(KERN_INFO "EXT4-fs: write access will "
2831 "be enabled during recovery.\n");
ac27a0ec
DK		 2832 }
2833 }
2834
2835 if (journal_inum && journal_dev) {
617ba13b 2836 printk(KERN_ERR "EXT4-fs: filesystem has both journal "
ac27a0ec
DK		 2837 "and inode journals!\n");
2838 return -EINVAL;
2839 }
2840
2841 if (journal_inum) {
617ba13b 2842 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 2843 return -EINVAL;
2844 } else {
617ba13b 2845 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 2846 return -EINVAL;
2847 }
2848
4776004f
TT		 2849 if (journal->j_flags & JBD2_BARRIER)
2850 printk(KERN_INFO "EXT4-fs: barriers enabled\n");
2851 else
2852 printk(KERN_INFO "EXT4-fs: barriers disabled\n");
2853
ac27a0ec 2854 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
dab291af 2855 err = jbd2_journal_update_format(journal);
ac27a0ec 2856 if (err) {
617ba13b 2857 printk(KERN_ERR "EXT4-fs: error updating journal.\n");
dab291af 2858 jbd2_journal_destroy(journal);
ac27a0ec
DK		 2859 return err;
2860 }
2861 }
2862
617ba13b 2863 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
dab291af 2864 err = jbd2_journal_wipe(journal, !really_read_only);
ac27a0ec 2865 if (!err)
dab291af 2866 err = jbd2_journal_load(journal);
ac27a0ec
DK		 2867
2868 if (err) {
617ba13b 2869 printk(KERN_ERR "EXT4-fs: error loading journal.\n");
dab291af 2870 jbd2_journal_destroy(journal);
ac27a0ec
DK		 2871 return err;
2872 }
2873
617ba13b
MC		 2874 EXT4_SB(sb)->s_journal = journal;
2875 ext4_clear_journal_err(sb, es);
ac27a0ec
DK		 2876
2877 if (journal_devnum &&
2878 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2879 es->s_journal_dev = cpu_to_le32(journal_devnum);
2880 sb->s_dirt = 1;
2881
2882 /* Make sure we flush the recovery flag to disk. */
617ba13b 2883 ext4_commit_super(sb, es, 1);
ac27a0ec
DK		 2884 }
2885
2886 return 0;
2887}
2888
c4be0c1d 2889static int ext4_commit_super(struct super_block *sb,
2b2d6d01 2890 struct ext4_super_block *es, int sync)
ac27a0ec 2891{
617ba13b 2892 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
c4be0c1d 2893 int error = 0;
ac27a0ec
DK		 2894
2895 if (!sbh)
c4be0c1d 2896 return error;
914258bf
TT		 2897 if (buffer_write_io_error(sbh)) {
2898 /*
2899 * Oh, dear. A previous attempt to write the
2900 * superblock failed. This could happen because the
2901 * USB device was yanked out. Or it could happen to
2902 * be a transient write error and maybe the block will
2903 * be remapped. Nothing we can do but to retry the
2904 * write and hope for the best.
2905 */
abda1418 2906 printk(KERN_ERR "EXT4-fs: previous I/O error to "
914258bf
TT		 2907 "superblock detected for %s.\n", sb->s_id);
2908 clear_buffer_write_io_error(sbh);
2909 set_buffer_uptodate(sbh);
2910 }
ac27a0ec 2911 es->s_wtime = cpu_to_le32(get_seconds());
5d1b1b3f
AK		 2912 ext4_free_blocks_count_set(es, percpu_counter_sum_positive(
2913 &EXT4_SB(sb)->s_freeblocks_counter));
2914 es->s_free_inodes_count = cpu_to_le32(percpu_counter_sum_positive(
2915 &EXT4_SB(sb)->s_freeinodes_counter));
2916
ac27a0ec
DK		 2917 BUFFER_TRACE(sbh, "marking dirty");
2918 mark_buffer_dirty(sbh);
914258bf 2919 if (sync) {
c4be0c1d
TS		 2920 error = sync_dirty_buffer(sbh);
2921 if (error)
2922 return error;
2923
2924 error = buffer_write_io_error(sbh);
2925 if (error) {
abda1418 2926 printk(KERN_ERR "EXT4-fs: I/O error while writing "
914258bf
TT		 2927 "superblock for %s.\n", sb->s_id);
2928 clear_buffer_write_io_error(sbh);
2929 set_buffer_uptodate(sbh);
2930 }
2931 }
c4be0c1d 2932 return error;
ac27a0ec
DK		 2933}
2934
2935
2936/*
2937 * Have we just finished recovery? If so, and if we are mounting (or
2938 * remounting) the filesystem readonly, then we will end up with a
2939 * consistent fs on disk. Record that fact.
2940 */
2b2d6d01
TT		 2941static void ext4_mark_recovery_complete(struct super_block *sb,
2942 struct ext4_super_block *es)
ac27a0ec 2943{
617ba13b 2944 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 2945
0390131b
FM		 2946 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2947 BUG_ON(journal != NULL);
2948 return;
2949 }
dab291af 2950 jbd2_journal_lock_updates(journal);
7ffe1ea8
HK		 2951 if (jbd2_journal_flush(journal) < 0)
2952 goto out;
2953
32c37730 2954 lock_super(sb);
617ba13b 2955 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
ac27a0ec 2956 sb->s_flags & MS_RDONLY) {
617ba13b 2957 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 2958 sb->s_dirt = 0;
617ba13b 2959 ext4_commit_super(sb, es, 1);
ac27a0ec 2960 }
32c37730 2961 unlock_super(sb);
7ffe1ea8
HK		 2962
2963out:
dab291af 2964 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 2965}
2966
2967/*
2968 * If we are mounting (or read-write remounting) a filesystem whose journal
2969 * has recorded an error from a previous lifetime, move that error to the
2970 * main filesystem now.
2971 */
2b2d6d01
TT		 2972static void ext4_clear_journal_err(struct super_block *sb,
2973 struct ext4_super_block *es)
ac27a0ec
DK		 2974{
2975 journal_t *journal;
2976 int j_errno;
2977 const char *errstr;
2978
0390131b
FM		 2979 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
2980
617ba13b 2981 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 2982
2983 /*
2984 * Now check for any error status which may have been recorded in the
617ba13b 2985 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 2986 */
2987
dab291af 2988 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 2989 if (j_errno) {
2990 char nbuf[16];
2991
617ba13b 2992 errstr = ext4_decode_error(sb, j_errno, nbuf);
46e665e9 2993 ext4_warning(sb, __func__, "Filesystem error recorded "
ac27a0ec 2994 "from previous mount: %s", errstr);
46e665e9 2995 ext4_warning(sb, __func__, "Marking fs in need of "
ac27a0ec
DK		 2996 "filesystem check.");
2997
617ba13b
MC		 2998 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2999 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2b2d6d01 3000 ext4_commit_super(sb, es, 1);
ac27a0ec 3001
dab291af 3002 jbd2_journal_clear_err(journal);
ac27a0ec
DK		 3003 }
3004}
3005
3006/*
3007 * Force the running and committing transactions to commit,
3008 * and wait on the commit.
3009 */
617ba13b 3010int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 3011{
3012 journal_t *journal;
0390131b 3013 int ret = 0;
ac27a0ec
DK		 3014
3015 if (sb->s_flags & MS_RDONLY)
3016 return 0;
3017
617ba13b 3018 journal = EXT4_SB(sb)->s_journal;
0390131b
FM		 3019 if (journal) {
3020 sb->s_dirt = 0;
3021 ret = ext4_journal_force_commit(journal);
3022 }
3023
ac27a0ec
DK		 3024 return ret;
3025}
3026
3027/*
617ba13b 3028 * Ext4 always journals updates to the superblock itself, so we don't
ac27a0ec 3029 * have to propagate any other updates to the superblock on disk at this
14ce0cb4
TT		 3030 * point. (We can probably nuke this function altogether, and remove
3031 * any mention to sb->s_dirt in all of fs/ext4; eventual cleanup...)
ac27a0ec 3032 */
2b2d6d01 3033static void ext4_write_super(struct super_block *sb)
ac27a0ec 3034{
0390131b
FM		 3035 if (EXT4_SB(sb)->s_journal) {
3036 if (mutex_trylock(&sb->s_lock) != 0)
3037 BUG();
3038 sb->s_dirt = 0;
3039 } else {
3040 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
3041 }
ac27a0ec
DK		 3042}
3043
617ba13b 3044static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec 3045{
14ce0cb4 3046 int ret = 0;
9eddacf9 3047 tid_t target;
ac27a0ec 3048
ede86cc4 3049 trace_mark(ext4_sync_fs, "dev %s wait %d", sb->s_id, wait);
ac27a0ec 3050 sb->s_dirt = 0;
0390131b 3051 if (EXT4_SB(sb)->s_journal) {
9eddacf9
JK		 3052 if (jbd2_journal_start_commit(EXT4_SB(sb)->s_journal,
3053 &target)) {
3054 if (wait)
3055 jbd2_log_wait_commit(EXT4_SB(sb)->s_journal,
3056 target);
3057 }
0390131b
FM		 3058 } else {
3059 ext4_commit_super(sb, EXT4_SB(sb)->s_es, wait);
3060 }
14ce0cb4 3061 return ret;
ac27a0ec
DK		 3062}
3063
3064/*
3065 * LVM calls this function before a (read-only) snapshot is created. This
3066 * gives us a chance to flush the journal completely and mark the fs clean.
3067 */
c4be0c1d 3068static int ext4_freeze(struct super_block *sb)
ac27a0ec 3069{
c4be0c1d
TS		 3070 int error = 0;
3071 journal_t *journal;
ac27a0ec
DK		 3072 sb->s_dirt = 0;
3073
3074 if (!(sb->s_flags & MS_RDONLY)) {
c4be0c1d 3075 journal = EXT4_SB(sb)->s_journal;
ac27a0ec 3076
0390131b
FM		 3077 if (journal) {
3078 /* Now we set up the journal barrier. */
3079 jbd2_journal_lock_updates(journal);
7ffe1ea8 3080
0390131b
FM		 3081 /*
3082 * We don't want to clear needs_recovery flag when we
3083 * failed to flush the journal.
3084 */
c4be0c1d
TS		 3085 error = jbd2_journal_flush(journal);
3086 if (error < 0)
3087 goto out;
0390131b 3088 }
ac27a0ec
DK		 3089
3090 /* Journal blocked and flushed, clear needs_recovery flag. */
617ba13b 3091 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
c4be0c1d
TS		 3092 error = ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
3093 if (error)
3094 goto out;
ac27a0ec 3095 }
c4be0c1d
TS		 3096 return 0;
3097out:
3098 jbd2_journal_unlock_updates(journal);
3099 return error;
ac27a0ec
DK		 3100}
3101
3102/*
3103 * Called by LVM after the snapshot is done. We need to reset the RECOVER
3104 * flag here, even though the filesystem is not technically dirty yet.
3105 */
c4be0c1d 3106static int ext4_unfreeze(struct super_block *sb)
ac27a0ec 3107{
0390131b 3108 if (EXT4_SB(sb)->s_journal && !(sb->s_flags & MS_RDONLY)) {
ac27a0ec
DK		 3109 lock_super(sb);
3110 /* Reser the needs_recovery flag before the fs is unlocked. */
617ba13b
MC		 3111 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3112 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
ac27a0ec 3113 unlock_super(sb);
dab291af 3114 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
ac27a0ec 3115 }
c4be0c1d 3116 return 0;
ac27a0ec
DK		 3117}
3118
2b2d6d01 3119static int ext4_remount(struct super_block *sb, int *flags, char *data)
ac27a0ec 3120{
2b2d6d01 3121 struct ext4_super_block *es;
617ba13b
MC		 3122 struct ext4_sb_info *sbi = EXT4_SB(sb);
3123 ext4_fsblk_t n_blocks_count = 0;
ac27a0ec 3124 unsigned long old_sb_flags;
617ba13b 3125 struct ext4_mount_options old_opts;
8a266467 3126 ext4_group_t g;
b3881f74 3127 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
ac27a0ec
DK		 3128 int err;
3129#ifdef CONFIG_QUOTA
3130 int i;
3131#endif
3132
3133 /* Store the original options */
3134 old_sb_flags = sb->s_flags;
3135 old_opts.s_mount_opt = sbi->s_mount_opt;
3136 old_opts.s_resuid = sbi->s_resuid;
3137 old_opts.s_resgid = sbi->s_resgid;
3138 old_opts.s_commit_interval = sbi->s_commit_interval;
30773840
TT		 3139 old_opts.s_min_batch_time = sbi->s_min_batch_time;
3140 old_opts.s_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 3141#ifdef CONFIG_QUOTA
3142 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
3143 for (i = 0; i < MAXQUOTAS; i++)
3144 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
3145#endif
b3881f74
TT		 3146 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
3147 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
ac27a0ec
DK		 3148
3149 /*
3150 * Allow the "check" option to be passed as a remount option.
3151 */
b3881f74
TT		 3152 if (!parse_options(data, sb, NULL, &journal_ioprio,
3153 &n_blocks_count, 1)) {
ac27a0ec
DK		 3154 err = -EINVAL;
3155 goto restore_opts;
3156 }
3157
617ba13b 3158 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT)
46e665e9 3159 ext4_abort(sb, __func__, "Abort forced by user");
ac27a0ec
DK		 3160
3161 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
617ba13b 3162 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec
DK		 3163
3164 es = sbi->s_es;
3165
b3881f74 3166 if (sbi->s_journal) {
0390131b 3167 ext4_init_journal_params(sb, sbi->s_journal);
b3881f74
TT		 3168 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3169 }
ac27a0ec
DK		 3170
3171 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
bd81d8ee 3172 n_blocks_count > ext4_blocks_count(es)) {
617ba13b 3173 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT) {
ac27a0ec
DK		 3174 err = -EROFS;
3175 goto restore_opts;
3176 }
3177
3178 if (*flags & MS_RDONLY) {
3179 /*
3180 * First of all, the unconditional stuff we have to do
3181 * to disable replay of the journal when we next remount
3182 */
3183 sb->s_flags |= MS_RDONLY;
3184
3185 /*
3186 * OK, test if we are remounting a valid rw partition
3187 * readonly, and if so set the rdonly flag and then
3188 * mark the partition as valid again.
3189 */
617ba13b
MC		 3190 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
3191 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 3192 es->s_state = cpu_to_le16(sbi->s_mount_state);
3193
32c37730
JK		 3194 /*
3195 * We have to unlock super so that we can wait for
3196 * transactions.
3197 */
0390131b
FM		 3198 if (sbi->s_journal) {
3199 unlock_super(sb);
3200 ext4_mark_recovery_complete(sb, es);
3201 lock_super(sb);
3202 }
ac27a0ec 3203 } else {
3a06d778 3204 int ret;
617ba13b
MC		 3205 if ((ret = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3206 ~EXT4_FEATURE_RO_COMPAT_SUPP))) {
3207 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
ac27a0ec 3208 "remount RDWR because of unsupported "
3a06d778
AK		 3209 "optional features (%x).\n", sb->s_id,
3210 (le32_to_cpu(sbi->s_es->s_feature_ro_compat) &
3211 ~EXT4_FEATURE_RO_COMPAT_SUPP));
ac27a0ec
DK		 3212 err = -EROFS;
3213 goto restore_opts;
3214 }
ead6596b 3215
8a266467
TT		 3216 /*
3217 * Make sure the group descriptor checksums
3218 * are sane. If they aren't, refuse to
3219 * remount r/w.
3220 */
3221 for (g = 0; g < sbi->s_groups_count; g++) {
3222 struct ext4_group_desc *gdp =
3223 ext4_get_group_desc(sb, g, NULL);
3224
3225 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
3226 printk(KERN_ERR
3227 "EXT4-fs: ext4_remount: "
a9df9a49 3228 "Checksum for group %u failed (%u!=%u)\n",
8a266467
TT		 3229 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
3230 le16_to_cpu(gdp->bg_checksum));
3231 err = -EINVAL;
3232 goto restore_opts;
3233 }
3234 }
3235
ead6596b
ES		 3236 /*
3237 * If we have an unprocessed orphan list hanging
3238 * around from a previously readonly bdev mount,
3239 * require a full umount/remount for now.
3240 */
3241 if (es->s_last_orphan) {
3242 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
3243 "remount RDWR because of unprocessed "
3244 "orphan inode list. Please "
3245 "umount/remount instead.\n",
3246 sb->s_id);
3247 err = -EINVAL;
3248 goto restore_opts;
3249 }
3250
ac27a0ec
DK		 3251 /*
3252 * Mounting a RDONLY partition read-write, so reread
3253 * and store the current valid flag. (It may have
3254 * been changed by e2fsck since we originally mounted
3255 * the partition.)
3256 */
0390131b
FM		 3257 if (sbi->s_journal)
3258 ext4_clear_journal_err(sb, es);
ac27a0ec 3259 sbi->s_mount_state = le16_to_cpu(es->s_state);
617ba13b 3260 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
ac27a0ec 3261 goto restore_opts;
2b2d6d01 3262 if (!ext4_setup_super(sb, es, 0))
ac27a0ec
DK		 3263 sb->s_flags &= ~MS_RDONLY;
3264 }
3265 }
0390131b
FM		 3266 if (sbi->s_journal == NULL)
3267 ext4_commit_super(sb, es, 1);
3268
ac27a0ec
DK		 3269#ifdef CONFIG_QUOTA
3270 /* Release old quota file names */
3271 for (i = 0; i < MAXQUOTAS; i++)
3272 if (old_opts.s_qf_names[i] &&
3273 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3274 kfree(old_opts.s_qf_names[i]);
3275#endif
3276 return 0;
3277restore_opts:
3278 sb->s_flags = old_sb_flags;
3279 sbi->s_mount_opt = old_opts.s_mount_opt;
3280 sbi->s_resuid = old_opts.s_resuid;
3281 sbi->s_resgid = old_opts.s_resgid;
3282 sbi->s_commit_interval = old_opts.s_commit_interval;
30773840
TT		 3283 sbi->s_min_batch_time = old_opts.s_min_batch_time;
3284 sbi->s_max_batch_time = old_opts.s_max_batch_time;
ac27a0ec
DK		 3285#ifdef CONFIG_QUOTA
3286 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
3287 for (i = 0; i < MAXQUOTAS; i++) {
3288 if (sbi->s_qf_names[i] &&
3289 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3290 kfree(sbi->s_qf_names[i]);
3291 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
3292 }
3293#endif
3294 return err;
3295}
3296
2b2d6d01 3297static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
ac27a0ec
DK		 3298{
3299 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 3300 struct ext4_sb_info *sbi = EXT4_SB(sb);
3301 struct ext4_super_block *es = sbi->s_es;
960cc398 3302 u64 fsid;
ac27a0ec 3303
5e70030d
BP		 3304 if (test_opt(sb, MINIX_DF)) {
3305 sbi->s_overhead_last = 0;
6bc9feff 3306 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
fd2d4291 3307 ext4_group_t ngroups = sbi->s_groups_count, i;
5e70030d 3308 ext4_fsblk_t overhead = 0;
ac27a0ec
DK		 3309 smp_rmb();
3310
3311 /*
5e70030d
BP		 3312 * Compute the overhead (FS structures). This is constant
3313 * for a given filesystem unless the number of block groups
3314 * changes so we cache the previous value until it does.
ac27a0ec
DK		 3315 */
3316
3317 /*
3318 * All of the blocks before first_data_block are
3319 * overhead
3320 */
3321 overhead = le32_to_cpu(es->s_first_data_block);
3322
3323 /*
3324 * Add the overhead attributed to the superblock and
3325 * block group descriptors. If the sparse superblocks
3326 * feature is turned on, then not all groups have this.
3327 */
3328 for (i = 0; i < ngroups; i++) {
617ba13b
MC		 3329 overhead += ext4_bg_has_super(sb, i) +
3330 ext4_bg_num_gdb(sb, i);
ac27a0ec
DK		 3331 cond_resched();
3332 }
3333
3334 /*
3335 * Every block group has an inode bitmap, a block
3336 * bitmap, and an inode table.
3337 */
5e70030d
BP		 3338 overhead += ngroups * (2 + sbi->s_itb_per_group);
3339 sbi->s_overhead_last = overhead;
3340 smp_wmb();
6bc9feff 3341 sbi->s_blocks_last = ext4_blocks_count(es);
ac27a0ec
DK		 3342 }
3343
617ba13b 3344 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 3345 buf->f_bsize = sb->s_blocksize;
5e70030d 3346 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
6bc6e63f
AK		 3347 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
3348 percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
308ba3ec 3349 ext4_free_blocks_count_set(es, buf->f_bfree);
bd81d8ee
LV		 3350 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3351 if (buf->f_bfree < ext4_r_blocks_count(es))
ac27a0ec
DK		 3352 buf->f_bavail = 0;
3353 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 3354 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
5e70030d 3355 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
617ba13b 3356 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 3357 fsid = le64_to_cpup((void *)es->s_uuid) ^
3358 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3359 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3360 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
ac27a0ec
DK		 3361 return 0;
3362}
3363
3364/* Helper function for writing quotas on sync - we need to start transaction before quota file
3365 * is locked for write. Otherwise the are possible deadlocks:
3366 * Process 1 Process 2
617ba13b 3367 * ext4_create() quota_sync()
dab291af 3368 * jbd2_journal_start() write_dquot()
ac27a0ec 3369 * DQUOT_INIT() down(dqio_mutex)
dab291af 3370 * down(dqio_mutex) jbd2_journal_start()
ac27a0ec
DK		 3371 *
3372 */
3373
3374#ifdef CONFIG_QUOTA
3375
3376static inline struct inode *dquot_to_inode(struct dquot *dquot)
3377{
3378 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3379}
3380
617ba13b 3381static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 3382{
3383 int ret, err;
3384 handle_t *handle;
3385 struct inode *inode;
3386
3387 inode = dquot_to_inode(dquot);
617ba13b
MC		 3388 handle = ext4_journal_start(inode,
3389 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 3390 if (IS_ERR(handle))
3391 return PTR_ERR(handle);
3392 ret = dquot_commit(dquot);
617ba13b 3393 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3394 if (!ret)
3395 ret = err;
3396 return ret;
3397}
3398
617ba13b 3399static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 3400{
3401 int ret, err;
3402 handle_t *handle;
3403
617ba13b
MC		 3404 handle = ext4_journal_start(dquot_to_inode(dquot),
3405 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 3406 if (IS_ERR(handle))
3407 return PTR_ERR(handle);
3408 ret = dquot_acquire(dquot);
617ba13b 3409 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3410 if (!ret)
3411 ret = err;
3412 return ret;
3413}
3414
617ba13b 3415static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 3416{
3417 int ret, err;
3418 handle_t *handle;
3419
617ba13b
MC		 3420 handle = ext4_journal_start(dquot_to_inode(dquot),
3421 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 3422 if (IS_ERR(handle)) {
3423 /* Release dquot anyway to avoid endless cycle in dqput() */
3424 dquot_release(dquot);
ac27a0ec 3425 return PTR_ERR(handle);
9c3013e9 3426 }
ac27a0ec 3427 ret = dquot_release(dquot);
617ba13b 3428 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3429 if (!ret)
3430 ret = err;
3431 return ret;
3432}
3433
617ba13b 3434static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec 3435{
2c8be6b2 3436 /* Are we journaling quotas? */
617ba13b
MC		 3437 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3438 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
ac27a0ec 3439 dquot_mark_dquot_dirty(dquot);
617ba13b 3440 return ext4_write_dquot(dquot);
ac27a0ec
DK		 3441 } else {
3442 return dquot_mark_dquot_dirty(dquot);
3443 }
3444}
3445
617ba13b 3446static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 3447{
3448 int ret, err;
3449 handle_t *handle;
3450
3451 /* Data block + inode block */
617ba13b 3452 handle = ext4_journal_start(sb->s_root->d_inode, 2);
ac27a0ec
DK		 3453 if (IS_ERR(handle))
3454 return PTR_ERR(handle);
3455 ret = dquot_commit_info(sb, type);
617ba13b 3456 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3457 if (!ret)
3458 ret = err;
3459 return ret;
3460}
3461
3462/*
3463 * Turn on quotas during mount time - we need to find
3464 * the quota file and such...
3465 */
617ba13b 3466static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 3467{
617ba13b
MC		 3468 return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
3469 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 3470}
3471
3472/*
3473 * Standard function to be called on quota_on
3474 */
617ba13b 3475static int ext4_quota_on(struct super_block *sb, int type, int format_id,
8264613d 3476 char *name, int remount)
ac27a0ec
DK		 3477{
3478 int err;
8264613d 3479 struct path path;
ac27a0ec
DK		 3480
3481 if (!test_opt(sb, QUOTA))
3482 return -EINVAL;
8264613d 3483 /* When remounting, no checks are needed and in fact, name is NULL */
0623543b 3484 if (remount)
8264613d 3485 return vfs_quota_on(sb, type, format_id, name, remount);
0623543b 3486
8264613d 3487 err = kern_path(name, LOOKUP_FOLLOW, &path);
ac27a0ec
DK		 3488 if (err)
3489 return err;
0623543b 3490
ac27a0ec 3491 /* Quotafile not on the same filesystem? */
8264613d
AV		 3492 if (path.mnt->mnt_sb != sb) {
3493 path_put(&path);
ac27a0ec
DK		 3494 return -EXDEV;
3495 }
0623543b
JK		 3496 /* Journaling quota? */
3497 if (EXT4_SB(sb)->s_qf_names[type]) {
2b2d6d01 3498 /* Quotafile not in fs root? */
8264613d 3499 if (path.dentry->d_parent != sb->s_root)
0623543b
JK		 3500 printk(KERN_WARNING
3501 "EXT4-fs: Quota file not on filesystem root. "
3502 "Journaled quota will not work.\n");
2b2d6d01 3503 }
0623543b
JK		 3504
3505 /*
3506 * When we journal data on quota file, we have to flush journal to see
3507 * all updates to the file when we bypass pagecache...
3508 */
0390131b
FM		 3509 if (EXT4_SB(sb)->s_journal &&
3510 ext4_should_journal_data(path.dentry->d_inode)) {
0623543b
JK		 3511 /*
3512 * We don't need to lock updates but journal_flush() could
3513 * otherwise be livelocked...
3514 */
3515 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 3516 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
0623543b 3517 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 3518 if (err) {
8264613d 3519 path_put(&path);
7ffe1ea8
HK		 3520 return err;
3521 }
0623543b
JK		 3522 }
3523
8264613d
AV		 3524 err = vfs_quota_on_path(sb, type, format_id, &path);
3525 path_put(&path);
77e69dac 3526 return err;
ac27a0ec
DK		 3527}
3528
3529/* Read data from quotafile - avoid pagecache and such because we cannot afford
3530 * acquiring the locks... As quota files are never truncated and quota code
3531 * itself serializes the operations (and noone else should touch the files)
3532 * we don't have to be afraid of races */
617ba13b 3533static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 3534 size_t len, loff_t off)
3535{
3536 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 3537 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 3538 int err = 0;
3539 int offset = off & (sb->s_blocksize - 1);
3540 int tocopy;
3541 size_t toread;
3542 struct buffer_head *bh;
3543 loff_t i_size = i_size_read(inode);
3544
3545 if (off > i_size)
3546 return 0;
3547 if (off+len > i_size)
3548 len = i_size-off;
3549 toread = len;
3550 while (toread > 0) {
3551 tocopy = sb->s_blocksize - offset < toread ?
3552 sb->s_blocksize - offset : toread;
617ba13b 3553 bh = ext4_bread(NULL, inode, blk, 0, &err);
ac27a0ec
DK		 3554 if (err)
3555 return err;
3556 if (!bh) /* A hole? */
3557 memset(data, 0, tocopy);
3558 else
3559 memcpy(data, bh->b_data+offset, tocopy);
3560 brelse(bh);
3561 offset = 0;
3562 toread -= tocopy;
3563 data += tocopy;
3564 blk++;
3565 }
3566 return len;
3567}
3568
3569/* Write to quotafile (we know the transaction is already started and has
3570 * enough credits) */
617ba13b 3571static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 3572 const char *data, size_t len, loff_t off)
3573{
3574 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 3575 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 3576 int err = 0;
3577 int offset = off & (sb->s_blocksize - 1);
3578 int tocopy;
617ba13b 3579 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
ac27a0ec
DK		 3580 size_t towrite = len;
3581 struct buffer_head *bh;
3582 handle_t *handle = journal_current_handle();
3583
0390131b 3584 if (EXT4_SB(sb)->s_journal && !handle) {
e5f8eab8 3585 printk(KERN_WARNING "EXT4-fs: Quota write (off=%llu, len=%llu)"
9c3013e9
JK		 3586 " cancelled because transaction is not started.\n",
3587 (unsigned long long)off, (unsigned long long)len);
3588 return -EIO;
3589 }
ac27a0ec
DK		 3590 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
3591 while (towrite > 0) {
3592 tocopy = sb->s_blocksize - offset < towrite ?
3593 sb->s_blocksize - offset : towrite;
617ba13b 3594 bh = ext4_bread(handle, inode, blk, 1, &err);
ac27a0ec
DK		 3595 if (!bh)
3596 goto out;
3597 if (journal_quota) {
617ba13b 3598 err = ext4_journal_get_write_access(handle, bh);
ac27a0ec
DK		 3599 if (err) {
3600 brelse(bh);
3601 goto out;
3602 }
3603 }
3604 lock_buffer(bh);
3605 memcpy(bh->b_data+offset, data, tocopy);
3606 flush_dcache_page(bh->b_page);
3607 unlock_buffer(bh);
3608 if (journal_quota)
0390131b 3609 err = ext4_handle_dirty_metadata(handle, NULL, bh);
ac27a0ec
DK		 3610 else {
3611 /* Always do at least ordered writes for quotas */
678aaf48 3612 err = ext4_jbd2_file_inode(handle, inode);
ac27a0ec
DK		 3613 mark_buffer_dirty(bh);
3614 }
3615 brelse(bh);
3616 if (err)
3617 goto out;
3618 offset = 0;
3619 towrite -= tocopy;
3620 data += tocopy;
3621 blk++;
3622 }
3623out:
4d04e4fb
JK		 3624 if (len == towrite) {
3625 mutex_unlock(&inode->i_mutex);
ac27a0ec 3626 return err;
4d04e4fb 3627 }
ac27a0ec
DK		 3628 if (inode->i_size < off+len-towrite) {
3629 i_size_write(inode, off+len-towrite);
617ba13b 3630 EXT4_I(inode)->i_disksize = inode->i_size;
ac27a0ec 3631 }
ac27a0ec 3632 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
617ba13b 3633 ext4_mark_inode_dirty(handle, inode);
ac27a0ec
DK		 3634 mutex_unlock(&inode->i_mutex);
3635 return len - towrite;
3636}
3637
3638#endif
3639
617ba13b 3640static int ext4_get_sb(struct file_system_type *fs_type,
ac27a0ec
DK		 3641 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3642{
617ba13b 3643 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
ac27a0ec
DK		 3644}
3645
5e8814f2
TT		 3646#ifdef CONFIG_PROC_FS
3647static int ext4_ui_proc_show(struct seq_file *m, void *v)
3648{
3649 unsigned int *p = m->private;
3650
3651 seq_printf(m, "%u\n", *p);
3652 return 0;
3653}
3654
3655static int ext4_ui_proc_open(struct inode *inode, struct file *file)
3656{
3657 return single_open(file, ext4_ui_proc_show, PDE(inode)->data);
3658}
3659
3660static ssize_t ext4_ui_proc_write(struct file *file, const char __user *buf,
3661 size_t cnt, loff_t *ppos)
3662{
23475e26 3663 unsigned long *p = PDE(file->f_path.dentry->d_inode)->data;
5e8814f2 3664 char str[32];
5e8814f2
TT		 3665
3666 if (cnt >= sizeof(str))
3667 return -EINVAL;
3668 if (copy_from_user(str, buf, cnt))
3669 return -EFAULT;
23475e26
RK		 3670
3671 *p = simple_strtoul(str, NULL, 0);
5e8814f2
TT		 3672 return cnt;
3673}
3674
3675const struct file_operations ext4_ui_proc_fops = {
3676 .owner = THIS_MODULE,
3677 .open = ext4_ui_proc_open,
3678 .read = seq_read,
3679 .llseek = seq_lseek,
3680 .release = single_release,
3681 .write = ext4_ui_proc_write,
3682};
3683#endif
3684
03010a33
TT		 3685static struct file_system_type ext4_fs_type = {
3686 .owner = THIS_MODULE,
3687 .name = "ext4",
3688 .get_sb = ext4_get_sb,
3689 .kill_sb = kill_block_super,
3690 .fs_flags = FS_REQUIRES_DEV,
3691};
3692
3693#ifdef CONFIG_EXT4DEV_COMPAT
3694static int ext4dev_get_sb(struct file_system_type *fs_type,
3695 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3696{
3697 printk(KERN_WARNING "EXT4-fs: Update your userspace programs "
3698 "to mount using ext4\n");
3699 printk(KERN_WARNING "EXT4-fs: ext4dev backwards compatibility "
3700 "will go away by 2.6.31\n");
3701 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
3702}
3703
617ba13b 3704static struct file_system_type ext4dev_fs_type = {
ac27a0ec 3705 .owner = THIS_MODULE,
617ba13b 3706 .name = "ext4dev",
03010a33 3707 .get_sb = ext4dev_get_sb,
ac27a0ec
DK		 3708 .kill_sb = kill_block_super,
3709 .fs_flags = FS_REQUIRES_DEV,
3710};
03010a33
TT		 3711MODULE_ALIAS("ext4dev");
3712#endif
ac27a0ec 3713
617ba13b 3714static int __init init_ext4_fs(void)
ac27a0ec 3715{
c9de560d
AT		 3716 int err;
3717
9f6200bb 3718 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
c9de560d 3719 err = init_ext4_mballoc();
ac27a0ec
DK		 3720 if (err)
3721 return err;
c9de560d
AT		 3722
3723 err = init_ext4_xattr();
3724 if (err)
3725 goto out2;
ac27a0ec
DK		 3726 err = init_inodecache();
3727 if (err)
3728 goto out1;
03010a33 3729 err = register_filesystem(&ext4_fs_type);
ac27a0ec
DK		 3730 if (err)
3731 goto out;
03010a33
TT		 3732#ifdef CONFIG_EXT4DEV_COMPAT
3733 err = register_filesystem(&ext4dev_fs_type);
3734 if (err) {
3735 unregister_filesystem(&ext4_fs_type);
3736 goto out;
3737 }
3738#endif
ac27a0ec
DK		 3739 return 0;
3740out:
3741 destroy_inodecache();
3742out1:
617ba13b 3743 exit_ext4_xattr();
c9de560d
AT		 3744out2:
3745 exit_ext4_mballoc();
ac27a0ec
DK		 3746 return err;
3747}
3748
617ba13b 3749static void __exit exit_ext4_fs(void)
ac27a0ec 3750{
03010a33
TT		 3751 unregister_filesystem(&ext4_fs_type);
3752#ifdef CONFIG_EXT4DEV_COMPAT
617ba13b 3753 unregister_filesystem(&ext4dev_fs_type);
03010a33 3754#endif
ac27a0ec 3755 destroy_inodecache();
617ba13b 3756 exit_ext4_xattr();
c9de560d 3757 exit_ext4_mballoc();
9f6200bb 3758 remove_proc_entry("fs/ext4", NULL);
ac27a0ec
DK		 3759}
3760
3761MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
83982b6f 3762MODULE_DESCRIPTION("Fourth Extended Filesystem");
ac27a0ec 3763MODULE_LICENSE("GPL");
617ba13b
MC		 3764module_init(init_ext4_fs)
3765module_exit(exit_ext4_fs)
kernel source for telekom puls (alcatel/mediatek)