]> bbs.cooldavid.org Git - net-next-2.6.git/blame - fs/ext4/super.c
git://bbs.cooldavid.org / net-next-2.6.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
jbd2: Fix return value of jbd2_journal_start_commit()
[net-next-2.6.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_dquot_initialize(struct inode *inode, int type);
930static int ext4_dquot_drop(struct inode *inode);
931static int ext4_write_dquot(struct dquot *dquot);
932static int ext4_acquire_dquot(struct dquot *dquot);
933static int ext4_release_dquot(struct dquot *dquot);
934static int ext4_mark_dquot_dirty(struct dquot *dquot);
935static int ext4_write_info(struct super_block *sb, int type);
6f28e087
JK		 936static int ext4_quota_on(struct super_block *sb, int type, int format_id,
937 char *path, int remount);
617ba13b
MC		 938static int ext4_quota_on_mount(struct super_block *sb, int type);
939static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 940 size_t len, loff_t off);
617ba13b 941static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 942 const char *data, size_t len, loff_t off);
943
617ba13b
MC		 944static struct dquot_operations ext4_quota_operations = {
945 .initialize = ext4_dquot_initialize,
946 .drop = ext4_dquot_drop,
ac27a0ec
DK		 947 .alloc_space = dquot_alloc_space,
948 .alloc_inode = dquot_alloc_inode,
949 .free_space = dquot_free_space,
950 .free_inode = dquot_free_inode,
951 .transfer = dquot_transfer,
617ba13b
MC		 952 .write_dquot = ext4_write_dquot,
953 .acquire_dquot = ext4_acquire_dquot,
954 .release_dquot = ext4_release_dquot,
955 .mark_dirty = ext4_mark_dquot_dirty,
a5b5ee32
JK		 956 .write_info = ext4_write_info,
957 .alloc_dquot = dquot_alloc,
958 .destroy_dquot = dquot_destroy,
ac27a0ec
DK		 959};
960
617ba13b
MC		 961static struct quotactl_ops ext4_qctl_operations = {
962 .quota_on = ext4_quota_on,
ac27a0ec
DK		 963 .quota_off = vfs_quota_off,
964 .quota_sync = vfs_quota_sync,
965 .get_info = vfs_get_dqinfo,
966 .set_info = vfs_set_dqinfo,
967 .get_dqblk = vfs_get_dqblk,
968 .set_dqblk = vfs_set_dqblk
969};
970#endif
971
ee9b6d61 972static const struct super_operations ext4_sops = {
617ba13b
MC		 973 .alloc_inode = ext4_alloc_inode,
974 .destroy_inode = ext4_destroy_inode,
617ba13b
MC		 975 .write_inode = ext4_write_inode,
976 .dirty_inode = ext4_dirty_inode,
977 .delete_inode = ext4_delete_inode,
978 .put_super = ext4_put_super,
979 .write_super = ext4_write_super,
980 .sync_fs = ext4_sync_fs,
c4be0c1d
TS		 981 .freeze_fs = ext4_freeze,
982 .unfreeze_fs = ext4_unfreeze,
617ba13b
MC		 983 .statfs = ext4_statfs,
984 .remount_fs = ext4_remount,
985 .clear_inode = ext4_clear_inode,
986 .show_options = ext4_show_options,
ac27a0ec 987#ifdef CONFIG_QUOTA
617ba13b
MC		 988 .quota_read = ext4_quota_read,
989 .quota_write = ext4_quota_write,
ac27a0ec 990#endif
c39a7f84 991 .bdev_try_to_free_page = bdev_try_to_free_page,
ac27a0ec
DK		 992};
993
39655164 994static const struct export_operations ext4_export_ops = {
1b961ac0
CH		 995 .fh_to_dentry = ext4_fh_to_dentry,
996 .fh_to_parent = ext4_fh_to_parent,
617ba13b 997 .get_parent = ext4_get_parent,
ac27a0ec
DK		 998};
999
1000enum {
1001 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1002 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
01436ef2 1003 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
ac27a0ec
DK		 1004 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1005 Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh,
30773840 1006 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
c3191067 1007 Opt_journal_update, Opt_journal_dev,
818d276c 1008 Opt_journal_checksum, Opt_journal_async_commit,
ac27a0ec 1009 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
5bf5683a 1010 Opt_data_err_abort, Opt_data_err_ignore,
ac27a0ec
DK		 1011 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1012 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
1013 Opt_ignore, Opt_barrier, Opt_err, Opt_resize, Opt_usrquota,
83982b6f 1014 Opt_grpquota, Opt_i_version,
01436ef2 1015 Opt_stripe, Opt_delalloc, Opt_nodelalloc,
b3881f74 1016 Opt_inode_readahead_blks, Opt_journal_ioprio
ac27a0ec
DK		 1017};
1018
a447c093 1019static const match_table_t tokens = {
ac27a0ec
DK		 1020 {Opt_bsd_df, "bsddf"},
1021 {Opt_minix_df, "minixdf"},
1022 {Opt_grpid, "grpid"},
1023 {Opt_grpid, "bsdgroups"},
1024 {Opt_nogrpid, "nogrpid"},
1025 {Opt_nogrpid, "sysvgroups"},
1026 {Opt_resgid, "resgid=%u"},
1027 {Opt_resuid, "resuid=%u"},
1028 {Opt_sb, "sb=%u"},
1029 {Opt_err_cont, "errors=continue"},
1030 {Opt_err_panic, "errors=panic"},
1031 {Opt_err_ro, "errors=remount-ro"},
1032 {Opt_nouid32, "nouid32"},
ac27a0ec
DK		 1033 {Opt_debug, "debug"},
1034 {Opt_oldalloc, "oldalloc"},
1035 {Opt_orlov, "orlov"},
1036 {Opt_user_xattr, "user_xattr"},
1037 {Opt_nouser_xattr, "nouser_xattr"},
1038 {Opt_acl, "acl"},
1039 {Opt_noacl, "noacl"},
1040 {Opt_reservation, "reservation"},
1041 {Opt_noreservation, "noreservation"},
1042 {Opt_noload, "noload"},
1043 {Opt_nobh, "nobh"},
1044 {Opt_bh, "bh"},
1045 {Opt_commit, "commit=%u"},
30773840
TT		 1046 {Opt_min_batch_time, "min_batch_time=%u"},
1047 {Opt_max_batch_time, "max_batch_time=%u"},
ac27a0ec 1048 {Opt_journal_update, "journal=update"},
ac27a0ec 1049 {Opt_journal_dev, "journal_dev=%u"},
818d276c
GS		 1050 {Opt_journal_checksum, "journal_checksum"},
1051 {Opt_journal_async_commit, "journal_async_commit"},
ac27a0ec
DK		 1052 {Opt_abort, "abort"},
1053 {Opt_data_journal, "data=journal"},
1054 {Opt_data_ordered, "data=ordered"},
1055 {Opt_data_writeback, "data=writeback"},
5bf5683a
HK		 1056 {Opt_data_err_abort, "data_err=abort"},
1057 {Opt_data_err_ignore, "data_err=ignore"},
ac27a0ec
DK		 1058 {Opt_offusrjquota, "usrjquota="},
1059 {Opt_usrjquota, "usrjquota=%s"},
1060 {Opt_offgrpjquota, "grpjquota="},
1061 {Opt_grpjquota, "grpjquota=%s"},
1062 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1063 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1064 {Opt_grpquota, "grpquota"},
1065 {Opt_noquota, "noquota"},
1066 {Opt_quota, "quota"},
1067 {Opt_usrquota, "usrquota"},
1068 {Opt_barrier, "barrier=%u"},
25ec56b5 1069 {Opt_i_version, "i_version"},
c9de560d 1070 {Opt_stripe, "stripe=%u"},
ac27a0ec 1071 {Opt_resize, "resize"},
64769240 1072 {Opt_delalloc, "delalloc"},
dd919b98 1073 {Opt_nodelalloc, "nodelalloc"},
240799cd 1074 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
b3881f74 1075 {Opt_journal_ioprio, "journal_ioprio=%u"},
f3f12faa 1076 {Opt_err, NULL},
ac27a0ec
DK		 1077};
1078
617ba13b 1079static ext4_fsblk_t get_sb_block(void **data)
ac27a0ec 1080{
617ba13b 1081 ext4_fsblk_t sb_block;
ac27a0ec
DK		 1082 char *options = (char *) *data;
1083
1084 if (!options || strncmp(options, "sb=", 3) != 0)
1085 return 1; /* Default location */
1086 options += 3;
617ba13b 1087 /*todo: use simple_strtoll with >32bit ext4 */
ac27a0ec
DK		 1088 sb_block = simple_strtoul(options, &options, 0);
1089 if (*options && *options != ',') {
4776004f 1090 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
ac27a0ec
DK		 1091 (char *) *data);
1092 return 1;
1093 }
1094 if (*options == ',')
1095 options++;
1096 *data = (void *) options;
1097 return sb_block;
1098}
1099
b3881f74
TT		 1100#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1101
2b2d6d01 1102static int parse_options(char *options, struct super_block *sb,
c3191067 1103 unsigned long *journal_devnum,
b3881f74 1104 unsigned int *journal_ioprio,
2b2d6d01 1105 ext4_fsblk_t *n_blocks_count, int is_remount)
ac27a0ec 1106{
617ba13b 1107 struct ext4_sb_info *sbi = EXT4_SB(sb);
2b2d6d01 1108 char *p;
ac27a0ec
DK		 1109 substring_t args[MAX_OPT_ARGS];
1110 int data_opt = 0;
1111 int option;
1112#ifdef CONFIG_QUOTA
dfc5d03f 1113 int qtype, qfmt;
ac27a0ec
DK		 1114 char *qname;
1115#endif
1116
1117 if (!options)
1118 return 1;
1119
2b2d6d01 1120 while ((p = strsep(&options, ",")) != NULL) {
ac27a0ec
DK		 1121 int token;
1122 if (!*p)
1123 continue;
1124
1125 token = match_token(p, tokens, args);
1126 switch (token) {
1127 case Opt_bsd_df:
2b2d6d01 1128 clear_opt(sbi->s_mount_opt, MINIX_DF);
ac27a0ec
DK		 1129 break;
1130 case Opt_minix_df:
2b2d6d01 1131 set_opt(sbi->s_mount_opt, MINIX_DF);
ac27a0ec
DK		 1132 break;
1133 case Opt_grpid:
2b2d6d01 1134 set_opt(sbi->s_mount_opt, GRPID);
ac27a0ec
DK		 1135 break;
1136 case Opt_nogrpid:
2b2d6d01 1137 clear_opt(sbi->s_mount_opt, GRPID);
ac27a0ec
DK		 1138 break;
1139 case Opt_resuid:
1140 if (match_int(&args[0], &option))
1141 return 0;
1142 sbi->s_resuid = option;
1143 break;
1144 case Opt_resgid:
1145 if (match_int(&args[0], &option))
1146 return 0;
1147 sbi->s_resgid = option;
1148 break;
1149 case Opt_sb:
1150 /* handled by get_sb_block() instead of here */
1151 /* *sb_block = match_int(&args[0]); */
1152 break;
1153 case Opt_err_panic:
2b2d6d01
TT		 1154 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1155 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1156 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
ac27a0ec
DK		 1157 break;
1158 case Opt_err_ro:
2b2d6d01
TT		 1159 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1160 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1161 set_opt(sbi->s_mount_opt, ERRORS_RO);
ac27a0ec
DK		 1162 break;
1163 case Opt_err_cont:
2b2d6d01
TT		 1164 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1165 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1166 set_opt(sbi->s_mount_opt, ERRORS_CONT);
ac27a0ec
DK		 1167 break;
1168 case Opt_nouid32:
2b2d6d01 1169 set_opt(sbi->s_mount_opt, NO_UID32);
ac27a0ec 1170 break;
ac27a0ec 1171 case Opt_debug:
2b2d6d01 1172 set_opt(sbi->s_mount_opt, DEBUG);
ac27a0ec
DK		 1173 break;
1174 case Opt_oldalloc:
2b2d6d01 1175 set_opt(sbi->s_mount_opt, OLDALLOC);
ac27a0ec
DK		 1176 break;
1177 case Opt_orlov:
2b2d6d01 1178 clear_opt(sbi->s_mount_opt, OLDALLOC);
ac27a0ec 1179 break;
03010a33 1180#ifdef CONFIG_EXT4_FS_XATTR
ac27a0ec 1181 case Opt_user_xattr:
2b2d6d01 1182 set_opt(sbi->s_mount_opt, XATTR_USER);
ac27a0ec
DK		 1183 break;
1184 case Opt_nouser_xattr:
2b2d6d01 1185 clear_opt(sbi->s_mount_opt, XATTR_USER);
ac27a0ec
DK		 1186 break;
1187#else
1188 case Opt_user_xattr:
1189 case Opt_nouser_xattr:
4776004f
TT		 1190 printk(KERN_ERR "EXT4 (no)user_xattr options "
1191 "not supported\n");
ac27a0ec
DK		 1192 break;
1193#endif
03010a33 1194#ifdef CONFIG_EXT4_FS_POSIX_ACL
ac27a0ec
DK		 1195 case Opt_acl:
1196 set_opt(sbi->s_mount_opt, POSIX_ACL);
1197 break;
1198 case Opt_noacl:
1199 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1200 break;
1201#else
1202 case Opt_acl:
1203 case Opt_noacl:
4776004f
TT		 1204 printk(KERN_ERR "EXT4 (no)acl options "
1205 "not supported\n");
ac27a0ec
DK		 1206 break;
1207#endif
1208 case Opt_reservation:
1209 set_opt(sbi->s_mount_opt, RESERVATION);
1210 break;
1211 case Opt_noreservation:
1212 clear_opt(sbi->s_mount_opt, RESERVATION);
1213 break;
1214 case Opt_journal_update:
1215 /* @@@ FIXME */
1216 /* Eventually we will want to be able to create
1217 a journal file here. For now, only allow the
1218 user to specify an existing inode to be the
1219 journal file. */
1220 if (is_remount) {
617ba13b 1221 printk(KERN_ERR "EXT4-fs: cannot specify "
ac27a0ec
DK		 1222 "journal on remount\n");
1223 return 0;
1224 }
2b2d6d01 1225 set_opt(sbi->s_mount_opt, UPDATE_JOURNAL);
ac27a0ec 1226 break;
ac27a0ec
DK		 1227 case Opt_journal_dev:
1228 if (is_remount) {
617ba13b 1229 printk(KERN_ERR "EXT4-fs: cannot specify "
ac27a0ec
DK		 1230 "journal on remount\n");
1231 return 0;
1232 }
1233 if (match_int(&args[0], &option))
1234 return 0;
1235 *journal_devnum = option;
1236 break;
818d276c
GS		 1237 case Opt_journal_checksum:
1238 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1239 break;
1240 case Opt_journal_async_commit:
1241 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
1242 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1243 break;
ac27a0ec 1244 case Opt_noload:
2b2d6d01 1245 set_opt(sbi->s_mount_opt, NOLOAD);
ac27a0ec
DK		 1246 break;
1247 case Opt_commit:
1248 if (match_int(&args[0], &option))
1249 return 0;
1250 if (option < 0)
1251 return 0;
1252 if (option == 0)
cd02ff0b 1253 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
ac27a0ec
DK		 1254 sbi->s_commit_interval = HZ * option;
1255 break;
30773840
TT		 1256 case Opt_max_batch_time:
1257 if (match_int(&args[0], &option))
1258 return 0;
1259 if (option < 0)
1260 return 0;
1261 if (option == 0)
1262 option = EXT4_DEF_MAX_BATCH_TIME;
1263 sbi->s_max_batch_time = option;
1264 break;
1265 case Opt_min_batch_time:
1266 if (match_int(&args[0], &option))
1267 return 0;
1268 if (option < 0)
1269 return 0;
1270 sbi->s_min_batch_time = option;
1271 break;
ac27a0ec 1272 case Opt_data_journal:
617ba13b 1273 data_opt = EXT4_MOUNT_JOURNAL_DATA;
ac27a0ec
DK		 1274 goto datacheck;
1275 case Opt_data_ordered:
617ba13b 1276 data_opt = EXT4_MOUNT_ORDERED_DATA;
ac27a0ec
DK		 1277 goto datacheck;
1278 case Opt_data_writeback:
617ba13b 1279 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
ac27a0ec
DK		 1280 datacheck:
1281 if (is_remount) {
617ba13b 1282 if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
ac27a0ec
DK		 1283 != data_opt) {
1284 printk(KERN_ERR
617ba13b 1285 "EXT4-fs: cannot change data "
ac27a0ec
DK		 1286 "mode on remount\n");
1287 return 0;
1288 }
1289 } else {
617ba13b 1290 sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
ac27a0ec
DK		 1291 sbi->s_mount_opt |= data_opt;
1292 }
1293 break;
5bf5683a
HK		 1294 case Opt_data_err_abort:
1295 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1296 break;
1297 case Opt_data_err_ignore:
1298 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1299 break;
ac27a0ec
DK		 1300#ifdef CONFIG_QUOTA
1301 case Opt_usrjquota:
1302 qtype = USRQUOTA;
1303 goto set_qf_name;
1304 case Opt_grpjquota:
1305 qtype = GRPQUOTA;
1306set_qf_name:
17bd13b3 1307 if (sb_any_quota_loaded(sb) &&
dfc5d03f 1308 !sbi->s_qf_names[qtype]) {
ac27a0ec 1309 printk(KERN_ERR
4776004f
TT		 1310 "EXT4-fs: Cannot change journaled "
1311 "quota options when quota turned on.\n");
ac27a0ec
DK		 1312 return 0;
1313 }
1314 qname = match_strdup(&args[0]);
1315 if (!qname) {
1316 printk(KERN_ERR
617ba13b 1317 "EXT4-fs: not enough memory for "
ac27a0ec
DK		 1318 "storing quotafile name.\n");
1319 return 0;
1320 }
1321 if (sbi->s_qf_names[qtype] &&
1322 strcmp(sbi->s_qf_names[qtype], qname)) {
1323 printk(KERN_ERR
617ba13b 1324 "EXT4-fs: %s quota file already "
ac27a0ec
DK		 1325 "specified.\n", QTYPE2NAME(qtype));
1326 kfree(qname);
1327 return 0;
1328 }
1329 sbi->s_qf_names[qtype] = qname;
1330 if (strchr(sbi->s_qf_names[qtype], '/')) {
1331 printk(KERN_ERR
617ba13b 1332 "EXT4-fs: quotafile must be on "
ac27a0ec
DK		 1333 "filesystem root.\n");
1334 kfree(sbi->s_qf_names[qtype]);
1335 sbi->s_qf_names[qtype] = NULL;
1336 return 0;
1337 }
1338 set_opt(sbi->s_mount_opt, QUOTA);
1339 break;
1340 case Opt_offusrjquota:
1341 qtype = USRQUOTA;
1342 goto clear_qf_name;
1343 case Opt_offgrpjquota:
1344 qtype = GRPQUOTA;
1345clear_qf_name:
17bd13b3 1346 if (sb_any_quota_loaded(sb) &&
dfc5d03f 1347 sbi->s_qf_names[qtype]) {
617ba13b 1348 printk(KERN_ERR "EXT4-fs: Cannot change "
2c8be6b2 1349 "journaled quota options when "
ac27a0ec
DK		 1350 "quota turned on.\n");
1351 return 0;
1352 }
1353 /*
1354 * The space will be released later when all options
1355 * are confirmed to be correct
1356 */
1357 sbi->s_qf_names[qtype] = NULL;
1358 break;
1359 case Opt_jqfmt_vfsold:
dfc5d03f
JK		 1360 qfmt = QFMT_VFS_OLD;
1361 goto set_qf_format;
ac27a0ec 1362 case Opt_jqfmt_vfsv0:
dfc5d03f
JK		 1363 qfmt = QFMT_VFS_V0;
1364set_qf_format:
17bd13b3 1365 if (sb_any_quota_loaded(sb) &&
dfc5d03f
JK		 1366 sbi->s_jquota_fmt != qfmt) {
1367 printk(KERN_ERR "EXT4-fs: Cannot change "
1368 "journaled quota options when "
1369 "quota turned on.\n");
1370 return 0;
1371 }
1372 sbi->s_jquota_fmt = qfmt;
ac27a0ec
DK		 1373 break;
1374 case Opt_quota:
1375 case Opt_usrquota:
1376 set_opt(sbi->s_mount_opt, QUOTA);
1377 set_opt(sbi->s_mount_opt, USRQUOTA);
1378 break;
1379 case Opt_grpquota:
1380 set_opt(sbi->s_mount_opt, QUOTA);
1381 set_opt(sbi->s_mount_opt, GRPQUOTA);
1382 break;
1383 case Opt_noquota:
17bd13b3 1384 if (sb_any_quota_loaded(sb)) {
617ba13b 1385 printk(KERN_ERR "EXT4-fs: Cannot change quota "
ac27a0ec
DK		 1386 "options when quota turned on.\n");
1387 return 0;
1388 }
1389 clear_opt(sbi->s_mount_opt, QUOTA);
1390 clear_opt(sbi->s_mount_opt, USRQUOTA);
1391 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1392 break;
1393#else
1394 case Opt_quota:
1395 case Opt_usrquota:
1396 case Opt_grpquota:
cd59e7b9
JK		 1397 printk(KERN_ERR
1398 "EXT4-fs: quota options not supported.\n");
1399 break;
ac27a0ec
DK		 1400 case Opt_usrjquota:
1401 case Opt_grpjquota:
1402 case Opt_offusrjquota:
1403 case Opt_offgrpjquota:
1404 case Opt_jqfmt_vfsold:
1405 case Opt_jqfmt_vfsv0:
1406 printk(KERN_ERR
cd59e7b9 1407 "EXT4-fs: journaled quota options not "
ac27a0ec
DK		 1408 "supported.\n");
1409 break;
1410 case Opt_noquota:
1411 break;
1412#endif
1413 case Opt_abort:
1414 set_opt(sbi->s_mount_opt, ABORT);
1415 break;
1416 case Opt_barrier:
1417 if (match_int(&args[0], &option))
1418 return 0;
1419 if (option)
1420 set_opt(sbi->s_mount_opt, BARRIER);
1421 else
1422 clear_opt(sbi->s_mount_opt, BARRIER);
1423 break;
1424 case Opt_ignore:
1425 break;
1426 case Opt_resize:
1427 if (!is_remount) {
617ba13b 1428 printk("EXT4-fs: resize option only available "
ac27a0ec
DK		 1429 "for remount\n");
1430 return 0;
1431 }
1432 if (match_int(&args[0], &option) != 0)
1433 return 0;
1434 *n_blocks_count = option;
1435 break;
1436 case Opt_nobh:
1437 set_opt(sbi->s_mount_opt, NOBH);
1438 break;
1439 case Opt_bh:
1440 clear_opt(sbi->s_mount_opt, NOBH);
1441 break;
25ec56b5
JNC		 1442 case Opt_i_version:
1443 set_opt(sbi->s_mount_opt, I_VERSION);
1444 sb->s_flags |= MS_I_VERSION;
1445 break;
dd919b98
AK		 1446 case Opt_nodelalloc:
1447 clear_opt(sbi->s_mount_opt, DELALLOC);
1448 break;
c9de560d
AT		 1449 case Opt_stripe:
1450 if (match_int(&args[0], &option))
1451 return 0;
1452 if (option < 0)
1453 return 0;
1454 sbi->s_stripe = option;
1455 break;
64769240
AT		 1456 case Opt_delalloc:
1457 set_opt(sbi->s_mount_opt, DELALLOC);
1458 break;
240799cd
TT		 1459 case Opt_inode_readahead_blks:
1460 if (match_int(&args[0], &option))
1461 return 0;
1462 if (option < 0 || option > (1 << 30))
1463 return 0;
1464 sbi->s_inode_readahead_blks = option;
1465 break;
b3881f74
TT		 1466 case Opt_journal_ioprio:
1467 if (match_int(&args[0], &option))
1468 return 0;
1469 if (option < 0 || option > 7)
1470 break;
1471 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1472 option);
1473 break;
ac27a0ec 1474 default:
2b2d6d01
TT		 1475 printk(KERN_ERR
1476 "EXT4-fs: Unrecognized mount option \"%s\" "
1477 "or missing value\n", p);
ac27a0ec
DK		 1478 return 0;
1479 }
1480 }
1481#ifdef CONFIG_QUOTA
1482 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
617ba13b 1483 if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
ac27a0ec
DK		 1484 sbi->s_qf_names[USRQUOTA])
1485 clear_opt(sbi->s_mount_opt, USRQUOTA);
1486
617ba13b 1487 if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
ac27a0ec
DK		 1488 sbi->s_qf_names[GRPQUOTA])
1489 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1490
1491 if ((sbi->s_qf_names[USRQUOTA] &&
617ba13b 1492 (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
ac27a0ec 1493 (sbi->s_qf_names[GRPQUOTA] &&
617ba13b
MC		 1494 (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
1495 printk(KERN_ERR "EXT4-fs: old and new quota "
ac27a0ec
DK		 1496 "format mixing.\n");
1497 return 0;
1498 }
1499
1500 if (!sbi->s_jquota_fmt) {
2c8be6b2 1501 printk(KERN_ERR "EXT4-fs: journaled quota format "
ac27a0ec
DK		 1502 "not specified.\n");
1503 return 0;
1504 }
1505 } else {
1506 if (sbi->s_jquota_fmt) {
2c8be6b2
JK		 1507 printk(KERN_ERR "EXT4-fs: journaled quota format "
1508 "specified with no journaling "
ac27a0ec
DK		 1509 "enabled.\n");
1510 return 0;
1511 }
1512 }
1513#endif
1514 return 1;
1515}
1516
617ba13b 1517static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 1518 int read_only)
1519{
617ba13b 1520 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 1521 int res = 0;
1522
617ba13b 1523 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
2b2d6d01
TT		 1524 printk(KERN_ERR "EXT4-fs warning: revision level too high, "
1525 "forcing read-only mode\n");
ac27a0ec
DK		 1526 res = MS_RDONLY;
1527 }
1528 if (read_only)
1529 return res;
617ba13b 1530 if (!(sbi->s_mount_state & EXT4_VALID_FS))
2b2d6d01
TT		 1531 printk(KERN_WARNING "EXT4-fs warning: mounting unchecked fs, "
1532 "running e2fsck is recommended\n");
617ba13b 1533 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
2b2d6d01
TT		 1534 printk(KERN_WARNING
1535 "EXT4-fs warning: mounting fs with errors, "
1536 "running e2fsck is recommended\n");
ac27a0ec
DK		 1537 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1538 le16_to_cpu(es->s_mnt_count) >=
1539 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
2b2d6d01
TT		 1540 printk(KERN_WARNING
1541 "EXT4-fs warning: maximal mount count reached, "
1542 "running e2fsck is recommended\n");
ac27a0ec
DK		 1543 else if (le32_to_cpu(es->s_checkinterval) &&
1544 (le32_to_cpu(es->s_lastcheck) +
1545 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
2b2d6d01
TT		 1546 printk(KERN_WARNING
1547 "EXT4-fs warning: checktime reached, "
1548 "running e2fsck is recommended\n");
0390131b
FM		 1549 if (!sbi->s_journal)
1550 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec 1551 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 1552 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 1553 le16_add_cpu(&es->s_mnt_count, 1);
ac27a0ec 1554 es->s_mtime = cpu_to_le32(get_seconds());
617ba13b 1555 ext4_update_dynamic_rev(sb);
0390131b
FM		 1556 if (sbi->s_journal)
1557 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 1558
617ba13b 1559 ext4_commit_super(sb, es, 1);
ac27a0ec 1560 if (test_opt(sb, DEBUG))
a9df9a49 1561 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
ac27a0ec
DK		 1562 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
1563 sb->s_blocksize,
1564 sbi->s_groups_count,
617ba13b
MC		 1565 EXT4_BLOCKS_PER_GROUP(sb),
1566 EXT4_INODES_PER_GROUP(sb),
ac27a0ec
DK		 1567 sbi->s_mount_opt);
1568
0390131b
FM		 1569 if (EXT4_SB(sb)->s_journal) {
1570 printk(KERN_INFO "EXT4 FS on %s, %s journal on %s\n",
1571 sb->s_id, EXT4_SB(sb)->s_journal->j_inode ? "internal" :
1572 "external", EXT4_SB(sb)->s_journal->j_devname);
1573 } else {
1574 printk(KERN_INFO "EXT4 FS on %s, no journal\n", sb->s_id);
1575 }
ac27a0ec
DK		 1576 return res;
1577}
1578
772cb7c8
JS		 1579static int ext4_fill_flex_info(struct super_block *sb)
1580{
1581 struct ext4_sb_info *sbi = EXT4_SB(sb);
1582 struct ext4_group_desc *gdp = NULL;
1583 struct buffer_head *bh;
1584 ext4_group_t flex_group_count;
1585 ext4_group_t flex_group;
1586 int groups_per_flex = 0;
772cb7c8
JS		 1587 int i;
1588
1589 if (!sbi->s_es->s_log_groups_per_flex) {
1590 sbi->s_log_groups_per_flex = 0;
1591 return 1;
1592 }
1593
1594 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1595 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1596
c62a11fd
FB		 1597 /* We allocate both existing and potentially added groups */
1598 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
d94e99a6
AK		 1599 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1600 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
ec05e868 1601 sbi->s_flex_groups = kzalloc(flex_group_count *
772cb7c8
JS		 1602 sizeof(struct flex_groups), GFP_KERNEL);
1603 if (sbi->s_flex_groups == NULL) {
ec05e868 1604 printk(KERN_ERR "EXT4-fs: not enough memory for "
a9df9a49 1605 "%u flex groups\n", flex_group_count);
772cb7c8
JS		 1606 goto failed;
1607 }
772cb7c8 1608
772cb7c8
JS		 1609 for (i = 0; i < sbi->s_groups_count; i++) {
1610 gdp = ext4_get_group_desc(sb, i, &bh);
1611
1612 flex_group = ext4_flex_group(sbi, i);
1613 sbi->s_flex_groups[flex_group].free_inodes +=
560671a0 1614 ext4_free_inodes_count(sb, gdp);
772cb7c8 1615 sbi->s_flex_groups[flex_group].free_blocks +=
560671a0 1616 ext4_free_blks_count(sb, gdp);
772cb7c8
JS		 1617 }
1618
1619 return 1;
1620failed:
1621 return 0;
1622}
1623
717d50e4
AD		 1624__le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1625 struct ext4_group_desc *gdp)
1626{
1627 __u16 crc = 0;
1628
1629 if (sbi->s_es->s_feature_ro_compat &
1630 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1631 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1632 __le32 le_group = cpu_to_le32(block_group);
1633
1634 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1635 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1636 crc = crc16(crc, (__u8 *)gdp, offset);
1637 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1638 /* for checksum of struct ext4_group_desc do the rest...*/
1639 if ((sbi->s_es->s_feature_incompat &
1640 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1641 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1642 crc = crc16(crc, (__u8 *)gdp + offset,
1643 le16_to_cpu(sbi->s_es->s_desc_size) -
1644 offset);
1645 }
1646
1647 return cpu_to_le16(crc);
1648}
1649
1650int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1651 struct ext4_group_desc *gdp)
1652{
1653 if ((sbi->s_es->s_feature_ro_compat &
1654 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1655 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1656 return 0;
1657
1658 return 1;
1659}
1660
ac27a0ec 1661/* Called at mount-time, super-block is locked */
197cd65a 1662static int ext4_check_descriptors(struct super_block *sb)
ac27a0ec 1663{
617ba13b
MC		 1664 struct ext4_sb_info *sbi = EXT4_SB(sb);
1665 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1666 ext4_fsblk_t last_block;
bd81d8ee
LV		 1667 ext4_fsblk_t block_bitmap;
1668 ext4_fsblk_t inode_bitmap;
1669 ext4_fsblk_t inode_table;
ce421581 1670 int flexbg_flag = 0;
fd2d4291 1671 ext4_group_t i;
ac27a0ec 1672
ce421581
JS		 1673 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1674 flexbg_flag = 1;
1675
af5bc92d 1676 ext4_debug("Checking group descriptors");
ac27a0ec 1677
197cd65a
AM		 1678 for (i = 0; i < sbi->s_groups_count; i++) {
1679 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1680
ce421581 1681 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 1682 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 1683 else
1684 last_block = first_block +
617ba13b 1685 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 1686
8fadc143 1687 block_bitmap = ext4_block_bitmap(sb, gdp);
2b2d6d01 1688 if (block_bitmap < first_block || block_bitmap > last_block) {
c19204b0 1689 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
a9df9a49 1690 "Block bitmap for group %u not in group "
5128273a 1691 "(block %llu)!\n", i, block_bitmap);
ac27a0ec
DK		 1692 return 0;
1693 }
8fadc143 1694 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2b2d6d01 1695 if (inode_bitmap < first_block || inode_bitmap > last_block) {
c19204b0 1696 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
a9df9a49 1697 "Inode bitmap for group %u not in group "
5128273a 1698 "(block %llu)!\n", i, inode_bitmap);
ac27a0ec
DK		 1699 return 0;
1700 }
8fadc143 1701 inode_table = ext4_inode_table(sb, gdp);
bd81d8ee 1702 if (inode_table < first_block ||
2b2d6d01 1703 inode_table + sbi->s_itb_per_group - 1 > last_block) {
c19204b0 1704 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
a9df9a49 1705 "Inode table for group %u not in group "
5128273a 1706 "(block %llu)!\n", i, inode_table);
ac27a0ec
DK		 1707 return 0;
1708 }
b5f10eed 1709 spin_lock(sb_bgl_lock(sbi, i));
717d50e4 1710 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
c19204b0 1711 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
a9df9a49 1712 "Checksum for group %u failed (%u!=%u)\n",
c19204b0
JB		 1713 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1714 gdp)), le16_to_cpu(gdp->bg_checksum));
7ee1ec4c
LZ		 1715 if (!(sb->s_flags & MS_RDONLY)) {
1716 spin_unlock(sb_bgl_lock(sbi, i));
8a266467 1717 return 0;
7ee1ec4c 1718 }
717d50e4 1719 }
b5f10eed 1720 spin_unlock(sb_bgl_lock(sbi, i));
ce421581
JS		 1721 if (!flexbg_flag)
1722 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec
DK		 1723 }
1724
bd81d8ee 1725 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
2b2d6d01 1726 sbi->s_es->s_free_inodes_count = cpu_to_le32(ext4_count_free_inodes(sb));
ac27a0ec
DK		 1727 return 1;
1728}
1729
617ba13b 1730/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 1731 * the superblock) which were deleted from all directories, but held open by
1732 * a process at the time of a crash. We walk the list and try to delete these
1733 * inodes at recovery time (only with a read-write filesystem).
1734 *
1735 * In order to keep the orphan inode chain consistent during traversal (in
1736 * case of crash during recovery), we link each inode into the superblock
1737 * orphan list_head and handle it the same way as an inode deletion during
1738 * normal operation (which journals the operations for us).
1739 *
1740 * We only do an iget() and an iput() on each inode, which is very safe if we
1741 * accidentally point at an in-use or already deleted inode. The worst that
1742 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 1743 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 1744 * e2fsck was run on this filesystem, and it must have already done the orphan
1745 * inode cleanup for us, so we can safely abort without any further action.
1746 */
2b2d6d01
TT		 1747static void ext4_orphan_cleanup(struct super_block *sb,
1748 struct ext4_super_block *es)
ac27a0ec
DK		 1749{
1750 unsigned int s_flags = sb->s_flags;
1751 int nr_orphans = 0, nr_truncates = 0;
1752#ifdef CONFIG_QUOTA
1753 int i;
1754#endif
1755 if (!es->s_last_orphan) {
1756 jbd_debug(4, "no orphan inodes to clean up\n");
1757 return;
1758 }
1759
a8f48a95
ES		 1760 if (bdev_read_only(sb->s_bdev)) {
1761 printk(KERN_ERR "EXT4-fs: write access "
1762 "unavailable, skipping orphan cleanup.\n");
1763 return;
1764 }
1765
617ba13b 1766 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
ac27a0ec
DK		 1767 if (es->s_last_orphan)
1768 jbd_debug(1, "Errors on filesystem, "
1769 "clearing orphan list.\n");
1770 es->s_last_orphan = 0;
1771 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1772 return;
1773 }
1774
1775 if (s_flags & MS_RDONLY) {
617ba13b 1776 printk(KERN_INFO "EXT4-fs: %s: orphan cleanup on readonly fs\n",
ac27a0ec
DK		 1777 sb->s_id);
1778 sb->s_flags &= ~MS_RDONLY;
1779 }
1780#ifdef CONFIG_QUOTA
1781 /* Needed for iput() to work correctly and not trash data */
1782 sb->s_flags |= MS_ACTIVE;
1783 /* Turn on quotas so that they are updated correctly */
1784 for (i = 0; i < MAXQUOTAS; i++) {
617ba13b
MC		 1785 if (EXT4_SB(sb)->s_qf_names[i]) {
1786 int ret = ext4_quota_on_mount(sb, i);
ac27a0ec
DK		 1787 if (ret < 0)
1788 printk(KERN_ERR
2c8be6b2 1789 "EXT4-fs: Cannot turn on journaled "
ac27a0ec
DK		 1790 "quota: error %d\n", ret);
1791 }
1792 }
1793#endif
1794
1795 while (es->s_last_orphan) {
1796 struct inode *inode;
1797
97bd42b9
JB		 1798 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1799 if (IS_ERR(inode)) {
ac27a0ec
DK		 1800 es->s_last_orphan = 0;
1801 break;
1802 }
1803
617ba13b 1804 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
ac27a0ec
DK		 1805 DQUOT_INIT(inode);
1806 if (inode->i_nlink) {
1807 printk(KERN_DEBUG
e5f8eab8 1808 "%s: truncating inode %lu to %lld bytes\n",
46e665e9 1809 __func__, inode->i_ino, inode->i_size);
e5f8eab8 1810 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
ac27a0ec 1811 inode->i_ino, inode->i_size);
617ba13b 1812 ext4_truncate(inode);
ac27a0ec
DK		 1813 nr_truncates++;
1814 } else {
1815 printk(KERN_DEBUG
1816 "%s: deleting unreferenced inode %lu\n",
46e665e9 1817 __func__, inode->i_ino);
ac27a0ec
DK		 1818 jbd_debug(2, "deleting unreferenced inode %lu\n",
1819 inode->i_ino);
1820 nr_orphans++;
1821 }
1822 iput(inode); /* The delete magic happens here! */
1823 }
1824
2b2d6d01 1825#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
ac27a0ec
DK		 1826
1827 if (nr_orphans)
617ba13b 1828 printk(KERN_INFO "EXT4-fs: %s: %d orphan inode%s deleted\n",
ac27a0ec
DK		 1829 sb->s_id, PLURAL(nr_orphans));
1830 if (nr_truncates)
617ba13b 1831 printk(KERN_INFO "EXT4-fs: %s: %d truncate%s cleaned up\n",
ac27a0ec
DK		 1832 sb->s_id, PLURAL(nr_truncates));
1833#ifdef CONFIG_QUOTA
1834 /* Turn quotas off */
1835 for (i = 0; i < MAXQUOTAS; i++) {
1836 if (sb_dqopt(sb)->files[i])
6f28e087 1837 vfs_quota_off(sb, i, 0);
ac27a0ec
DK		 1838 }
1839#endif
1840 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1841}
cd2291a4
ES		 1842/*
1843 * Maximal extent format file size.
1844 * Resulting logical blkno at s_maxbytes must fit in our on-disk
1845 * extent format containers, within a sector_t, and within i_blocks
1846 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
1847 * so that won't be a limiting factor.
1848 *
1849 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1850 */
f287a1a5 1851static loff_t ext4_max_size(int blkbits, int has_huge_files)
cd2291a4
ES		 1852{
1853 loff_t res;
1854 loff_t upper_limit = MAX_LFS_FILESIZE;
1855
1856 /* small i_blocks in vfs inode? */
f287a1a5 1857 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
cd2291a4 1858 /*
b3a6ffe1 1859 * CONFIG_LBD is not enabled implies the inode
cd2291a4
ES		 1860 * i_block represent total blocks in 512 bytes
1861 * 32 == size of vfs inode i_blocks * 8
1862 */
1863 upper_limit = (1LL << 32) - 1;
1864
1865 /* total blocks in file system block size */
1866 upper_limit >>= (blkbits - 9);
1867 upper_limit <<= blkbits;
1868 }
1869
1870 /* 32-bit extent-start container, ee_block */
1871 res = 1LL << 32;
1872 res <<= blkbits;
1873 res -= 1;
1874
1875 /* Sanity check against vm- & vfs- imposed limits */
1876 if (res > upper_limit)
1877 res = upper_limit;
1878
1879 return res;
1880}
ac27a0ec 1881
ac27a0ec 1882/*
cd2291a4 1883 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 1884 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
1885 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 1886 */
f287a1a5 1887static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
ac27a0ec 1888{
617ba13b 1889 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 1890 int meta_blocks;
1891 loff_t upper_limit;
1892 /* This is calculated to be the largest file size for a
cd2291a4 1893 * dense, bitmapped file such that the total number of
ac27a0ec 1894 * sectors in the file, including data and all indirect blocks,
0fc1b451
AK		 1895 * does not exceed 2^48 -1
1896 * __u32 i_blocks_lo and _u16 i_blocks_high representing the
1897 * total number of 512 bytes blocks of the file
1898 */
1899
f287a1a5 1900 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
0fc1b451 1901 /*
b3a6ffe1 1902 * !has_huge_files or CONFIG_LBD is not enabled
f287a1a5
TT		 1903 * implies the inode i_block represent total blocks in
1904 * 512 bytes 32 == size of vfs inode i_blocks * 8
0fc1b451
AK		 1905 */
1906 upper_limit = (1LL << 32) - 1;
1907
1908 /* total blocks in file system block size */
1909 upper_limit >>= (bits - 9);
1910
1911 } else {
8180a562
AK		 1912 /*
1913 * We use 48 bit ext4_inode i_blocks
1914 * With EXT4_HUGE_FILE_FL set the i_blocks
1915 * represent total number of blocks in
1916 * file system block size
1917 */
0fc1b451
AK		 1918 upper_limit = (1LL << 48) - 1;
1919
0fc1b451
AK		 1920 }
1921
1922 /* indirect blocks */
1923 meta_blocks = 1;
1924 /* double indirect blocks */
1925 meta_blocks += 1 + (1LL << (bits-2));
1926 /* tripple indirect blocks */
1927 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
1928
1929 upper_limit -= meta_blocks;
1930 upper_limit <<= bits;
ac27a0ec
DK		 1931
1932 res += 1LL << (bits-2);
1933 res += 1LL << (2*(bits-2));
1934 res += 1LL << (3*(bits-2));
1935 res <<= bits;
1936 if (res > upper_limit)
1937 res = upper_limit;
0fc1b451
AK		 1938
1939 if (res > MAX_LFS_FILESIZE)
1940 res = MAX_LFS_FILESIZE;
1941
ac27a0ec
DK		 1942 return res;
1943}
1944
617ba13b 1945static ext4_fsblk_t descriptor_loc(struct super_block *sb,
70bbb3e0 1946 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 1947{
617ba13b 1948 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 1949 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 1950 int has_super = 0;
1951
1952 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1953
617ba13b 1954 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
ac27a0ec 1955 nr < first_meta_bg)
70bbb3e0 1956 return logical_sb_block + nr + 1;
ac27a0ec 1957 bg = sbi->s_desc_per_block * nr;
617ba13b 1958 if (ext4_bg_has_super(sb, bg))
ac27a0ec 1959 has_super = 1;
617ba13b 1960 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 1961}
1962
c9de560d
AT		 1963/**
1964 * ext4_get_stripe_size: Get the stripe size.
1965 * @sbi: In memory super block info
1966 *
1967 * If we have specified it via mount option, then
1968 * use the mount option value. If the value specified at mount time is
1969 * greater than the blocks per group use the super block value.
1970 * If the super block value is greater than blocks per group return 0.
1971 * Allocator needs it be less than blocks per group.
1972 *
1973 */
1974static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
1975{
1976 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
1977 unsigned long stripe_width =
1978 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
1979
1980 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
1981 return sbi->s_stripe;
1982
1983 if (stripe_width <= sbi->s_blocks_per_group)
1984 return stripe_width;
1985
1986 if (stride <= sbi->s_blocks_per_group)
1987 return stride;
1988
1989 return 0;
1990}
ac27a0ec 1991
2b2d6d01 1992static int ext4_fill_super(struct super_block *sb, void *data, int silent)
7477827f
AK		 1993 __releases(kernel_lock)
1994 __acquires(kernel_lock)
1d03ec98 1995
ac27a0ec 1996{
2b2d6d01 1997 struct buffer_head *bh;
617ba13b
MC		 1998 struct ext4_super_block *es = NULL;
1999 struct ext4_sb_info *sbi;
2000 ext4_fsblk_t block;
2001 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 2002 ext4_fsblk_t logical_sb_block;
ac27a0ec 2003 unsigned long offset = 0;
ac27a0ec
DK		 2004 unsigned long journal_devnum = 0;
2005 unsigned long def_mount_opts;
2006 struct inode *root;
9f6200bb 2007 char *cp;
0390131b 2008 const char *descr;
1d1fe1ee 2009 int ret = -EINVAL;
ac27a0ec 2010 int blocksize;
4ec11028
TT		 2011 unsigned int db_count;
2012 unsigned int i;
f287a1a5 2013 int needs_recovery, has_huge_files;
3a06d778 2014 int features;
bd81d8ee 2015 __u64 blocks_count;
833f4077 2016 int err;
b3881f74 2017 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
ac27a0ec
DK		 2018
2019 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
2020 if (!sbi)
2021 return -ENOMEM;
2022 sb->s_fs_info = sbi;
2023 sbi->s_mount_opt = 0;
617ba13b
MC		 2024 sbi->s_resuid = EXT4_DEF_RESUID;
2025 sbi->s_resgid = EXT4_DEF_RESGID;
240799cd 2026 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
d9c9bef1 2027 sbi->s_sb_block = sb_block;
ac27a0ec
DK		 2028
2029 unlock_kernel();
2030
9f6200bb
TT		 2031 /* Cleanup superblock name */
2032 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
2033 *cp = '!';
2034
617ba13b 2035 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 2036 if (!blocksize) {
617ba13b 2037 printk(KERN_ERR "EXT4-fs: unable to set blocksize\n");
ac27a0ec
DK		 2038 goto out_fail;
2039 }
2040
2041 /*
617ba13b 2042 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 2043 * block sizes. We need to calculate the offset from buffer start.
2044 */
617ba13b 2045 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 2046 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2047 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 2048 } else {
70bbb3e0 2049 logical_sb_block = sb_block;
ac27a0ec
DK		 2050 }
2051
70bbb3e0 2052 if (!(bh = sb_bread(sb, logical_sb_block))) {
2b2d6d01 2053 printk(KERN_ERR "EXT4-fs: unable to read superblock\n");
ac27a0ec
DK		 2054 goto out_fail;
2055 }
2056 /*
2057 * Note: s_es must be initialized as soon as possible because
617ba13b 2058 * some ext4 macro-instructions depend on its value
ac27a0ec 2059 */
617ba13b 2060 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
ac27a0ec
DK		 2061 sbi->s_es = es;
2062 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 2063 if (sb->s_magic != EXT4_SUPER_MAGIC)
2064 goto cantfind_ext4;
ac27a0ec
DK		 2065
2066 /* Set defaults before we parse the mount options */
2067 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
617ba13b 2068 if (def_mount_opts & EXT4_DEFM_DEBUG)
ac27a0ec 2069 set_opt(sbi->s_mount_opt, DEBUG);
617ba13b 2070 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
ac27a0ec 2071 set_opt(sbi->s_mount_opt, GRPID);
617ba13b 2072 if (def_mount_opts & EXT4_DEFM_UID16)
ac27a0ec 2073 set_opt(sbi->s_mount_opt, NO_UID32);
03010a33 2074#ifdef CONFIG_EXT4_FS_XATTR
617ba13b 2075 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
ac27a0ec 2076 set_opt(sbi->s_mount_opt, XATTR_USER);
2e7842b8 2077#endif
03010a33 2078#ifdef CONFIG_EXT4_FS_POSIX_ACL
617ba13b 2079 if (def_mount_opts & EXT4_DEFM_ACL)
ac27a0ec 2080 set_opt(sbi->s_mount_opt, POSIX_ACL);
2e7842b8 2081#endif
617ba13b
MC		 2082 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
2083 sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
2084 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
2085 sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
2086 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
2087 sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
2088
2089 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
ac27a0ec 2090 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
bb4f397a 2091 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
ceea16bf 2092 set_opt(sbi->s_mount_opt, ERRORS_CONT);
bb4f397a
AK		 2093 else
2094 set_opt(sbi->s_mount_opt, ERRORS_RO);
ac27a0ec
DK		 2095
2096 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
2097 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
30773840
TT		 2098 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
2099 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
2100 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
ac27a0ec
DK		 2101
2102 set_opt(sbi->s_mount_opt, RESERVATION);
571640ca 2103 set_opt(sbi->s_mount_opt, BARRIER);
ac27a0ec 2104
dd919b98
AK		 2105 /*
2106 * enable delayed allocation by default
2107 * Use -o nodelalloc to turn it off
2108 */
2109 set_opt(sbi->s_mount_opt, DELALLOC);
2110
2111
b3881f74
TT		 2112 if (!parse_options((char *) data, sb, &journal_devnum,
2113 &journal_ioprio, NULL, 0))
ac27a0ec
DK		 2114 goto failed_mount;
2115
2116 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
617ba13b 2117 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec 2118
617ba13b
MC		 2119 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
2120 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
2121 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
2122 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
ac27a0ec 2123 printk(KERN_WARNING
617ba13b 2124 "EXT4-fs warning: feature flags set on rev 0 fs, "
ac27a0ec 2125 "running e2fsck is recommended\n");
469108ff 2126
ac27a0ec
DK		 2127 /*
2128 * Check feature flags regardless of the revision level, since we
2129 * previously didn't change the revision level when setting the flags,
2130 * so there is a chance incompat flags are set on a rev 0 filesystem.
2131 */
617ba13b 2132 features = EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP);
ac27a0ec 2133 if (features) {
617ba13b 2134 printk(KERN_ERR "EXT4-fs: %s: couldn't mount because of "
3a06d778
AK		 2135 "unsupported optional features (%x).\n", sb->s_id,
2136 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2137 ~EXT4_FEATURE_INCOMPAT_SUPP));
ac27a0ec
DK		 2138 goto failed_mount;
2139 }
617ba13b 2140 features = EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP);
ac27a0ec 2141 if (!(sb->s_flags & MS_RDONLY) && features) {
617ba13b 2142 printk(KERN_ERR "EXT4-fs: %s: couldn't mount RDWR because of "
3a06d778
AK		 2143 "unsupported optional features (%x).\n", sb->s_id,
2144 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2145 ~EXT4_FEATURE_RO_COMPAT_SUPP));
ac27a0ec
DK		 2146 goto failed_mount;
2147 }
f287a1a5
TT		 2148 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2149 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
2150 if (has_huge_files) {
0fc1b451
AK		 2151 /*
2152 * Large file size enabled file system can only be
b3a6ffe1 2153 * mount if kernel is build with CONFIG_LBD
0fc1b451
AK		 2154 */
2155 if (sizeof(root->i_blocks) < sizeof(u64) &&
2156 !(sb->s_flags & MS_RDONLY)) {
2157 printk(KERN_ERR "EXT4-fs: %s: Filesystem with huge "
2158 "files cannot be mounted read-write "
b3a6ffe1 2159 "without CONFIG_LBD.\n", sb->s_id);
0fc1b451
AK		 2160 goto failed_mount;
2161 }
2162 }
ac27a0ec
DK		 2163 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
2164
617ba13b
MC		 2165 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
2166 blocksize > EXT4_MAX_BLOCK_SIZE) {
ac27a0ec 2167 printk(KERN_ERR
617ba13b 2168 "EXT4-fs: Unsupported filesystem blocksize %d on %s.\n",
ac27a0ec
DK		 2169 blocksize, sb->s_id);
2170 goto failed_mount;
2171 }
2172
ac27a0ec 2173 if (sb->s_blocksize != blocksize) {
ce40733c
AK		 2174
2175 /* Validate the filesystem blocksize */
2176 if (!sb_set_blocksize(sb, blocksize)) {
2177 printk(KERN_ERR "EXT4-fs: bad block size %d.\n",
2178 blocksize);
ac27a0ec
DK		 2179 goto failed_mount;
2180 }
2181
2b2d6d01 2182 brelse(bh);
70bbb3e0
AM		 2183 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2184 offset = do_div(logical_sb_block, blocksize);
2185 bh = sb_bread(sb, logical_sb_block);
ac27a0ec
DK		 2186 if (!bh) {
2187 printk(KERN_ERR
617ba13b 2188 "EXT4-fs: Can't read superblock on 2nd try.\n");
ac27a0ec
DK		 2189 goto failed_mount;
2190 }
617ba13b 2191 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
ac27a0ec 2192 sbi->s_es = es;
617ba13b 2193 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
2b2d6d01
TT		 2194 printk(KERN_ERR
2195 "EXT4-fs: Magic mismatch, very weird !\n");
ac27a0ec
DK		 2196 goto failed_mount;
2197 }
2198 }
2199
f287a1a5
TT		 2200 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
2201 has_huge_files);
2202 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
ac27a0ec 2203
617ba13b
MC		 2204 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2205 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2206 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
ac27a0ec
DK		 2207 } else {
2208 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2209 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
617ba13b 2210 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1330593e 2211 (!is_power_of_2(sbi->s_inode_size)) ||
ac27a0ec 2212 (sbi->s_inode_size > blocksize)) {
2b2d6d01
TT		 2213 printk(KERN_ERR
2214 "EXT4-fs: unsupported inode size: %d\n",
2215 sbi->s_inode_size);
ac27a0ec
DK		 2216 goto failed_mount;
2217 }
ef7f3835
KS		 2218 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2219 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
ac27a0ec 2220 }
0d1ee42f
AR		 2221 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2222 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
8fadc143 2223 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 2224 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 2225 !is_power_of_2(sbi->s_desc_size)) {
0d1ee42f 2226 printk(KERN_ERR
8fadc143 2227 "EXT4-fs: unsupported descriptor size %lu\n",
0d1ee42f
AR		 2228 sbi->s_desc_size);
2229 goto failed_mount;
2230 }
2231 } else
2232 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
ac27a0ec 2233 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 2234 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
b47b6f38 2235 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
617ba13b
MC		 2236 goto cantfind_ext4;
2237 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 2238 if (sbi->s_inodes_per_block == 0)
617ba13b 2239 goto cantfind_ext4;
ac27a0ec
DK		 2240 sbi->s_itb_per_group = sbi->s_inodes_per_group /
2241 sbi->s_inodes_per_block;
0d1ee42f 2242 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 2243 sbi->s_sbh = bh;
2244 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 2245 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2246 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
2b2d6d01 2247 for (i = 0; i < 4; i++)
ac27a0ec
DK		 2248 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2249 sbi->s_def_hash_version = es->s_def_hash_version;
f99b2589
TT		 2250 i = le32_to_cpu(es->s_flags);
2251 if (i & EXT2_FLAGS_UNSIGNED_HASH)
2252 sbi->s_hash_unsigned = 3;
2253 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
2254#ifdef __CHAR_UNSIGNED__
2255 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
2256 sbi->s_hash_unsigned = 3;
2257#else
2258 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
2259#endif
2260 sb->s_dirt = 1;
2261 }
ac27a0ec
DK		 2262
2263 if (sbi->s_blocks_per_group > blocksize * 8) {
2b2d6d01
TT		 2264 printk(KERN_ERR
2265 "EXT4-fs: #blocks per group too big: %lu\n",
2266 sbi->s_blocks_per_group);
ac27a0ec
DK		 2267 goto failed_mount;
2268 }
ac27a0ec 2269 if (sbi->s_inodes_per_group > blocksize * 8) {
2b2d6d01
TT		 2270 printk(KERN_ERR
2271 "EXT4-fs: #inodes per group too big: %lu\n",
2272 sbi->s_inodes_per_group);
ac27a0ec
DK		 2273 goto failed_mount;
2274 }
2275
bd81d8ee 2276 if (ext4_blocks_count(es) >
ac27a0ec 2277 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
617ba13b 2278 printk(KERN_ERR "EXT4-fs: filesystem on %s:"
ac27a0ec
DK		 2279 " too large to mount safely\n", sb->s_id);
2280 if (sizeof(sector_t) < 8)
617ba13b 2281 printk(KERN_WARNING "EXT4-fs: CONFIG_LBD not "
ac27a0ec
DK		 2282 "enabled\n");
2283 goto failed_mount;
2284 }
2285
617ba13b
MC		 2286 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2287 goto cantfind_ext4;
e7c95593 2288
4ec11028
TT		 2289 /*
2290 * It makes no sense for the first data block to be beyond the end
2291 * of the filesystem.
2292 */
2293 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
2294 printk(KERN_WARNING "EXT4-fs: bad geometry: first data"
2295 "block %u is beyond end of filesystem (%llu)\n",
2296 le32_to_cpu(es->s_first_data_block),
2297 ext4_blocks_count(es));
e7c95593
ES		 2298 goto failed_mount;
2299 }
bd81d8ee
LV		 2300 blocks_count = (ext4_blocks_count(es) -
2301 le32_to_cpu(es->s_first_data_block) +
2302 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2303 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4ec11028
TT		 2304 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
2305 printk(KERN_WARNING "EXT4-fs: groups count too large: %u "
2306 "(block count %llu, first data block %u, "
2307 "blocks per group %lu)\n", sbi->s_groups_count,
2308 ext4_blocks_count(es),
2309 le32_to_cpu(es->s_first_data_block),
2310 EXT4_BLOCKS_PER_GROUP(sb));
2311 goto failed_mount;
2312 }
bd81d8ee 2313 sbi->s_groups_count = blocks_count;
617ba13b
MC		 2314 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2315 EXT4_DESC_PER_BLOCK(sb);
2b2d6d01 2316 sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
ac27a0ec
DK		 2317 GFP_KERNEL);
2318 if (sbi->s_group_desc == NULL) {
2b2d6d01 2319 printk(KERN_ERR "EXT4-fs: not enough memory\n");
ac27a0ec
DK		 2320 goto failed_mount;
2321 }
2322
3244fcb1 2323#ifdef CONFIG_PROC_FS
9f6200bb
TT		 2324 if (ext4_proc_root)
2325 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
2326
240799cd
TT		 2327 if (sbi->s_proc)
2328 proc_create_data("inode_readahead_blks", 0644, sbi->s_proc,
2329 &ext4_ui_proc_fops,
2330 &sbi->s_inode_readahead_blks);
3244fcb1 2331#endif
240799cd 2332
ac27a0ec
DK		 2333 bgl_lock_init(&sbi->s_blockgroup_lock);
2334
2335 for (i = 0; i < db_count; i++) {
70bbb3e0 2336 block = descriptor_loc(sb, logical_sb_block, i);
ac27a0ec
DK		 2337 sbi->s_group_desc[i] = sb_bread(sb, block);
2338 if (!sbi->s_group_desc[i]) {
2b2d6d01
TT		 2339 printk(KERN_ERR "EXT4-fs: "
2340 "can't read group descriptor %d\n", i);
ac27a0ec
DK		 2341 db_count = i;
2342 goto failed_mount2;
2343 }
2344 }
2b2d6d01 2345 if (!ext4_check_descriptors(sb)) {
617ba13b 2346 printk(KERN_ERR "EXT4-fs: group descriptors corrupted!\n");
ac27a0ec
DK		 2347 goto failed_mount2;
2348 }
772cb7c8
JS		 2349 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2350 if (!ext4_fill_flex_info(sb)) {
2351 printk(KERN_ERR
2352 "EXT4-fs: unable to initialize "
2353 "flex_bg meta info!\n");
2354 goto failed_mount2;
2355 }
2356
ac27a0ec
DK		 2357 sbi->s_gdb_count = db_count;
2358 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2359 spin_lock_init(&sbi->s_next_gen_lock);
2360
833f4077
PZ		 2361 err = percpu_counter_init(&sbi->s_freeblocks_counter,
2362 ext4_count_free_blocks(sb));
2363 if (!err) {
2364 err = percpu_counter_init(&sbi->s_freeinodes_counter,
2365 ext4_count_free_inodes(sb));
2366 }
2367 if (!err) {
2368 err = percpu_counter_init(&sbi->s_dirs_counter,
2369 ext4_count_dirs(sb));
2370 }
6bc6e63f
AK		 2371 if (!err) {
2372 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
2373 }
833f4077
PZ		 2374 if (err) {
2375 printk(KERN_ERR "EXT4-fs: insufficient memory\n");
2376 goto failed_mount3;
2377 }
ac27a0ec 2378
c9de560d
AT		 2379 sbi->s_stripe = ext4_get_stripe_size(sbi);
2380
ac27a0ec
DK		 2381 /*
2382 * set up enough so that it can read an inode
2383 */
617ba13b
MC		 2384 sb->s_op = &ext4_sops;
2385 sb->s_export_op = &ext4_export_ops;
2386 sb->s_xattr = ext4_xattr_handlers;
ac27a0ec 2387#ifdef CONFIG_QUOTA
617ba13b
MC		 2388 sb->s_qcop = &ext4_qctl_operations;
2389 sb->dq_op = &ext4_quota_operations;
ac27a0ec
DK		 2390#endif
2391 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
2392
2393 sb->s_root = NULL;
2394
2395 needs_recovery = (es->s_last_orphan != 0 ||
617ba13b
MC		 2396 EXT4_HAS_INCOMPAT_FEATURE(sb,
2397 EXT4_FEATURE_INCOMPAT_RECOVER));
ac27a0ec
DK		 2398
2399 /*
2400 * The first inode we look at is the journal inode. Don't try
2401 * root first: it may be modified in the journal!
2402 */
2403 if (!test_opt(sb, NOLOAD) &&
617ba13b
MC		 2404 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2405 if (ext4_load_journal(sb, es, journal_devnum))
ac27a0ec 2406 goto failed_mount3;
624080ed
TT		 2407 if (!(sb->s_flags & MS_RDONLY) &&
2408 EXT4_SB(sb)->s_journal->j_failed_commit) {
2409 printk(KERN_CRIT "EXT4-fs error (device %s): "
2410 "ext4_fill_super: Journal transaction "
2b2d6d01 2411 "%u is corrupt\n", sb->s_id,
624080ed 2412 EXT4_SB(sb)->s_journal->j_failed_commit);
2b2d6d01
TT		 2413 if (test_opt(sb, ERRORS_RO)) {
2414 printk(KERN_CRIT
2415 "Mounting filesystem read-only\n");
624080ed
TT		 2416 sb->s_flags |= MS_RDONLY;
2417 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2418 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2419 }
2420 if (test_opt(sb, ERRORS_PANIC)) {
2421 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2422 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2423 ext4_commit_super(sb, es, 1);
624080ed
TT		 2424 goto failed_mount4;
2425 }
2426 }
0390131b
FM		 2427 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
2428 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
2429 printk(KERN_ERR "EXT4-fs: required journal recovery "
2430 "suppressed and not mounted read-only\n");
2431 goto failed_mount4;
ac27a0ec 2432 } else {
0390131b
FM		 2433 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
2434 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
2435 sbi->s_journal = NULL;
2436 needs_recovery = 0;
2437 goto no_journal;
ac27a0ec
DK		 2438 }
2439
eb40a09c
JS		 2440 if (ext4_blocks_count(es) > 0xffffffffULL &&
2441 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2442 JBD2_FEATURE_INCOMPAT_64BIT)) {
abda1418 2443 printk(KERN_ERR "EXT4-fs: Failed to set 64-bit journal feature\n");
eb40a09c
JS		 2444 goto failed_mount4;
2445 }
2446
818d276c
GS		 2447 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2448 jbd2_journal_set_features(sbi->s_journal,
2449 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2450 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2451 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2452 jbd2_journal_set_features(sbi->s_journal,
2453 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2454 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2455 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2456 } else {
2457 jbd2_journal_clear_features(sbi->s_journal,
2458 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2459 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2460 }
2461
ac27a0ec
DK		 2462 /* We have now updated the journal if required, so we can
2463 * validate the data journaling mode. */
2464 switch (test_opt(sb, DATA_FLAGS)) {
2465 case 0:
2466 /* No mode set, assume a default based on the journal
63f57933
AM		 2467 * capabilities: ORDERED_DATA if the journal can
2468 * cope, else JOURNAL_DATA
2469 */
dab291af
MC		 2470 if (jbd2_journal_check_available_features
2471 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
ac27a0ec
DK		 2472 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2473 else
2474 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2475 break;
2476
617ba13b
MC		 2477 case EXT4_MOUNT_ORDERED_DATA:
2478 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 2479 if (!jbd2_journal_check_available_features
2480 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
617ba13b 2481 printk(KERN_ERR "EXT4-fs: Journal does not support "
ac27a0ec
DK		 2482 "requested data journaling mode\n");
2483 goto failed_mount4;
2484 }
2485 default:
2486 break;
2487 }
b3881f74 2488 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
ac27a0ec 2489
0390131b 2490no_journal:
ac27a0ec
DK		 2491
2492 if (test_opt(sb, NOBH)) {
617ba13b
MC		 2493 if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
2494 printk(KERN_WARNING "EXT4-fs: Ignoring nobh option - "
ac27a0ec
DK		 2495 "its supported only with writeback mode\n");
2496 clear_opt(sbi->s_mount_opt, NOBH);
2497 }
2498 }
2499 /*
dab291af 2500 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 2501 * so we can safely mount the rest of the filesystem now.
2502 */
2503
1d1fe1ee
DH		 2504 root = ext4_iget(sb, EXT4_ROOT_INO);
2505 if (IS_ERR(root)) {
617ba13b 2506 printk(KERN_ERR "EXT4-fs: get root inode failed\n");
1d1fe1ee 2507 ret = PTR_ERR(root);
ac27a0ec
DK		 2508 goto failed_mount4;
2509 }
2510 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1d1fe1ee 2511 iput(root);
617ba13b 2512 printk(KERN_ERR "EXT4-fs: corrupt root inode, run e2fsck\n");
ac27a0ec
DK		 2513 goto failed_mount4;
2514 }
1d1fe1ee
DH		 2515 sb->s_root = d_alloc_root(root);
2516 if (!sb->s_root) {
2517 printk(KERN_ERR "EXT4-fs: get root dentry failed\n");
2518 iput(root);
2519 ret = -ENOMEM;
2520 goto failed_mount4;
2521 }
ac27a0ec 2522
2b2d6d01 2523 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
ef7f3835
KS		 2524
2525 /* determine the minimum size of new large inodes, if present */
2526 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2527 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2528 EXT4_GOOD_OLD_INODE_SIZE;
2529 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2530 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2531 if (sbi->s_want_extra_isize <
2532 le16_to_cpu(es->s_want_extra_isize))
2533 sbi->s_want_extra_isize =
2534 le16_to_cpu(es->s_want_extra_isize);
2535 if (sbi->s_want_extra_isize <
2536 le16_to_cpu(es->s_min_extra_isize))
2537 sbi->s_want_extra_isize =
2538 le16_to_cpu(es->s_min_extra_isize);
2539 }
2540 }
2541 /* Check if enough inode space is available */
2542 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2543 sbi->s_inode_size) {
2544 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2545 EXT4_GOOD_OLD_INODE_SIZE;
2546 printk(KERN_INFO "EXT4-fs: required extra inode space not"
2547 "available.\n");
2548 }
2549
c2774d84
AK		 2550 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
2551 printk(KERN_WARNING "EXT4-fs: Ignoring delalloc option - "
2552 "requested data journaling mode\n");
2553 clear_opt(sbi->s_mount_opt, DELALLOC);
2554 } else if (test_opt(sb, DELALLOC))
2555 printk(KERN_INFO "EXT4-fs: delayed allocation enabled\n");
2556
2557 ext4_ext_init(sb);
2558 err = ext4_mb_init(sb, needs_recovery);
2559 if (err) {
2560 printk(KERN_ERR "EXT4-fs: failed to initalize mballoc (%d)\n",
2561 err);
2562 goto failed_mount4;
2563 }
2564
ac27a0ec
DK		 2565 /*
2566 * akpm: core read_super() calls in here with the superblock locked.
2567 * That deadlocks, because orphan cleanup needs to lock the superblock
2568 * in numerous places. Here we just pop the lock - it's relatively
2569 * harmless, because we are now ready to accept write_super() requests,
2570 * and aviro says that's the only reason for hanging onto the
2571 * superblock lock.
2572 */
617ba13b
MC		 2573 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
2574 ext4_orphan_cleanup(sb, es);
2575 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
0390131b 2576 if (needs_recovery) {
2b2d6d01 2577 printk(KERN_INFO "EXT4-fs: recovery complete.\n");
0390131b
FM		 2578 ext4_mark_recovery_complete(sb, es);
2579 }
2580 if (EXT4_SB(sb)->s_journal) {
2581 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2582 descr = " journalled data mode";
2583 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2584 descr = " ordered data mode";
2585 else
2586 descr = " writeback data mode";
2587 } else
2588 descr = "out journal";
2589
2590 printk(KERN_INFO "EXT4-fs: mounted filesystem %s with%s\n",
2591 sb->s_id, descr);
ac27a0ec
DK		 2592
2593 lock_kernel();
2594 return 0;
2595
617ba13b 2596cantfind_ext4:
ac27a0ec 2597 if (!silent)
617ba13b 2598 printk(KERN_ERR "VFS: Can't find ext4 filesystem on dev %s.\n",
ac27a0ec
DK		 2599 sb->s_id);
2600 goto failed_mount;
2601
2602failed_mount4:
0390131b
FM		 2603 printk(KERN_ERR "EXT4-fs (device %s): mount failed\n", sb->s_id);
2604 if (sbi->s_journal) {
2605 jbd2_journal_destroy(sbi->s_journal);
2606 sbi->s_journal = NULL;
2607 }
ac27a0ec
DK		 2608failed_mount3:
2609 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2610 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2611 percpu_counter_destroy(&sbi->s_dirs_counter);
6bc6e63f 2612 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
ac27a0ec
DK		 2613failed_mount2:
2614 for (i = 0; i < db_count; i++)
2615 brelse(sbi->s_group_desc[i]);
2616 kfree(sbi->s_group_desc);
2617failed_mount:
240799cd
TT		 2618 if (sbi->s_proc) {
2619 remove_proc_entry("inode_readahead_blks", sbi->s_proc);
9f6200bb 2620 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 2621 }
ac27a0ec
DK		 2622#ifdef CONFIG_QUOTA
2623 for (i = 0; i < MAXQUOTAS; i++)
2624 kfree(sbi->s_qf_names[i]);
2625#endif
617ba13b 2626 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 2627 brelse(bh);
2628out_fail:
2629 sb->s_fs_info = NULL;
2630 kfree(sbi);
2631 lock_kernel();
1d1fe1ee 2632 return ret;
ac27a0ec
DK		 2633}
2634
2635/*
2636 * Setup any per-fs journal parameters now. We'll do this both on
2637 * initial mount, once the journal has been initialised but before we've
2638 * done any recovery; and again on any subsequent remount.
2639 */
617ba13b 2640static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 2641{
617ba13b 2642 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 2643
30773840
TT		 2644 journal->j_commit_interval = sbi->s_commit_interval;
2645 journal->j_min_batch_time = sbi->s_min_batch_time;
2646 journal->j_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 2647
2648 spin_lock(&journal->j_state_lock);
2649 if (test_opt(sb, BARRIER))
dab291af 2650 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 2651 else
dab291af 2652 journal->j_flags &= ~JBD2_BARRIER;
5bf5683a
HK		 2653 if (test_opt(sb, DATA_ERR_ABORT))
2654 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
2655 else
2656 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
ac27a0ec
DK		 2657 spin_unlock(&journal->j_state_lock);
2658}
2659
617ba13b 2660static journal_t *ext4_get_journal(struct super_block *sb,
ac27a0ec
DK		 2661 unsigned int journal_inum)
2662{
2663 struct inode *journal_inode;
2664 journal_t *journal;
2665
0390131b
FM		 2666 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
2667
ac27a0ec
DK		 2668 /* First, test for the existence of a valid inode on disk. Bad
2669 * things happen if we iget() an unused inode, as the subsequent
2670 * iput() will try to delete it. */
2671
1d1fe1ee
DH		 2672 journal_inode = ext4_iget(sb, journal_inum);
2673 if (IS_ERR(journal_inode)) {
617ba13b 2674 printk(KERN_ERR "EXT4-fs: no journal found.\n");
ac27a0ec
DK		 2675 return NULL;
2676 }
2677 if (!journal_inode->i_nlink) {
2678 make_bad_inode(journal_inode);
2679 iput(journal_inode);
617ba13b 2680 printk(KERN_ERR "EXT4-fs: journal inode is deleted.\n");
ac27a0ec
DK		 2681 return NULL;
2682 }
2683
e5f8eab8 2684 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
ac27a0ec 2685 journal_inode, journal_inode->i_size);
1d1fe1ee 2686 if (!S_ISREG(journal_inode->i_mode)) {
617ba13b 2687 printk(KERN_ERR "EXT4-fs: invalid journal inode.\n");
ac27a0ec
DK		 2688 iput(journal_inode);
2689 return NULL;
2690 }
2691
dab291af 2692 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 2693 if (!journal) {
617ba13b 2694 printk(KERN_ERR "EXT4-fs: Could not load journal inode\n");
ac27a0ec
DK		 2695 iput(journal_inode);
2696 return NULL;
2697 }
2698 journal->j_private = sb;
617ba13b 2699 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 2700 return journal;
2701}
2702
617ba13b 2703static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 2704 dev_t j_dev)
2705{
2b2d6d01 2706 struct buffer_head *bh;
ac27a0ec 2707 journal_t *journal;
617ba13b
MC		 2708 ext4_fsblk_t start;
2709 ext4_fsblk_t len;
ac27a0ec 2710 int hblock, blocksize;
617ba13b 2711 ext4_fsblk_t sb_block;
ac27a0ec 2712 unsigned long offset;
2b2d6d01 2713 struct ext4_super_block *es;
ac27a0ec
DK		 2714 struct block_device *bdev;
2715
0390131b
FM		 2716 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
2717
617ba13b 2718 bdev = ext4_blkdev_get(j_dev);
ac27a0ec
DK		 2719 if (bdev == NULL)
2720 return NULL;
2721
2722 if (bd_claim(bdev, sb)) {
2723 printk(KERN_ERR
abda1418 2724 "EXT4-fs: failed to claim external journal device.\n");
9a1c3542 2725 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
ac27a0ec
DK		 2726 return NULL;
2727 }
2728
2729 blocksize = sb->s_blocksize;
2730 hblock = bdev_hardsect_size(bdev);
2731 if (blocksize < hblock) {
2732 printk(KERN_ERR
617ba13b 2733 "EXT4-fs: blocksize too small for journal device.\n");
ac27a0ec
DK		 2734 goto out_bdev;
2735 }
2736
617ba13b
MC		 2737 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
2738 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 2739 set_blocksize(bdev, blocksize);
2740 if (!(bh = __bread(bdev, sb_block, blocksize))) {
617ba13b 2741 printk(KERN_ERR "EXT4-fs: couldn't read superblock of "
ac27a0ec
DK		 2742 "external journal\n");
2743 goto out_bdev;
2744 }
2745
617ba13b
MC		 2746 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2747 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 2748 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b
MC		 2749 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
2750 printk(KERN_ERR "EXT4-fs: external journal has "
ac27a0ec
DK		 2751 "bad superblock\n");
2752 brelse(bh);
2753 goto out_bdev;
2754 }
2755
617ba13b
MC		 2756 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
2757 printk(KERN_ERR "EXT4-fs: journal UUID does not match\n");
ac27a0ec
DK		 2758 brelse(bh);
2759 goto out_bdev;
2760 }
2761
bd81d8ee 2762 len = ext4_blocks_count(es);
ac27a0ec
DK		 2763 start = sb_block + 1;
2764 brelse(bh); /* we're done with the superblock */
2765
dab291af 2766 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 2767 start, len, blocksize);
2768 if (!journal) {
617ba13b 2769 printk(KERN_ERR "EXT4-fs: failed to create device journal\n");
ac27a0ec
DK		 2770 goto out_bdev;
2771 }
2772 journal->j_private = sb;
2773 ll_rw_block(READ, 1, &journal->j_sb_buffer);
2774 wait_on_buffer(journal->j_sb_buffer);
2775 if (!buffer_uptodate(journal->j_sb_buffer)) {
617ba13b 2776 printk(KERN_ERR "EXT4-fs: I/O error on journal device\n");
ac27a0ec
DK		 2777 goto out_journal;
2778 }
2779 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
617ba13b 2780 printk(KERN_ERR "EXT4-fs: External journal has more than one "
ac27a0ec
DK		 2781 "user (unsupported) - %d\n",
2782 be32_to_cpu(journal->j_superblock->s_nr_users));
2783 goto out_journal;
2784 }
617ba13b
MC		 2785 EXT4_SB(sb)->journal_bdev = bdev;
2786 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 2787 return journal;
2788out_journal:
dab291af 2789 jbd2_journal_destroy(journal);
ac27a0ec 2790out_bdev:
617ba13b 2791 ext4_blkdev_put(bdev);
ac27a0ec
DK		 2792 return NULL;
2793}
2794
617ba13b
MC		 2795static int ext4_load_journal(struct super_block *sb,
2796 struct ext4_super_block *es,
ac27a0ec
DK		 2797 unsigned long journal_devnum)
2798{
2799 journal_t *journal;
2800 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
2801 dev_t journal_dev;
2802 int err = 0;
2803 int really_read_only;
2804
0390131b
FM		 2805 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
2806
ac27a0ec
DK		 2807 if (journal_devnum &&
2808 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
617ba13b 2809 printk(KERN_INFO "EXT4-fs: external journal device major/minor "
ac27a0ec
DK		 2810 "numbers have changed\n");
2811 journal_dev = new_decode_dev(journal_devnum);
2812 } else
2813 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
2814
2815 really_read_only = bdev_read_only(sb->s_bdev);
2816
2817 /*
2818 * Are we loading a blank journal or performing recovery after a
2819 * crash? For recovery, we need to check in advance whether we
2820 * can get read-write access to the device.
2821 */
2822
617ba13b 2823 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
ac27a0ec 2824 if (sb->s_flags & MS_RDONLY) {
617ba13b 2825 printk(KERN_INFO "EXT4-fs: INFO: recovery "
ac27a0ec
DK		 2826 "required on readonly filesystem.\n");
2827 if (really_read_only) {
617ba13b 2828 printk(KERN_ERR "EXT4-fs: write access "
ac27a0ec
DK		 2829 "unavailable, cannot proceed.\n");
2830 return -EROFS;
2831 }
2b2d6d01
TT		 2832 printk(KERN_INFO "EXT4-fs: write access will "
2833 "be enabled during recovery.\n");
ac27a0ec
DK		 2834 }
2835 }
2836
2837 if (journal_inum && journal_dev) {
617ba13b 2838 printk(KERN_ERR "EXT4-fs: filesystem has both journal "
ac27a0ec
DK		 2839 "and inode journals!\n");
2840 return -EINVAL;
2841 }
2842
2843 if (journal_inum) {
617ba13b 2844 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 2845 return -EINVAL;
2846 } else {
617ba13b 2847 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 2848 return -EINVAL;
2849 }
2850
4776004f
TT		 2851 if (journal->j_flags & JBD2_BARRIER)
2852 printk(KERN_INFO "EXT4-fs: barriers enabled\n");
2853 else
2854 printk(KERN_INFO "EXT4-fs: barriers disabled\n");
2855
ac27a0ec 2856 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
dab291af 2857 err = jbd2_journal_update_format(journal);
ac27a0ec 2858 if (err) {
617ba13b 2859 printk(KERN_ERR "EXT4-fs: error updating journal.\n");
dab291af 2860 jbd2_journal_destroy(journal);
ac27a0ec
DK		 2861 return err;
2862 }
2863 }
2864
617ba13b 2865 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
dab291af 2866 err = jbd2_journal_wipe(journal, !really_read_only);
ac27a0ec 2867 if (!err)
dab291af 2868 err = jbd2_journal_load(journal);
ac27a0ec
DK		 2869
2870 if (err) {
617ba13b 2871 printk(KERN_ERR "EXT4-fs: error loading journal.\n");
dab291af 2872 jbd2_journal_destroy(journal);
ac27a0ec
DK		 2873 return err;
2874 }
2875
617ba13b
MC		 2876 EXT4_SB(sb)->s_journal = journal;
2877 ext4_clear_journal_err(sb, es);
ac27a0ec
DK		 2878
2879 if (journal_devnum &&
2880 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2881 es->s_journal_dev = cpu_to_le32(journal_devnum);
2882 sb->s_dirt = 1;
2883
2884 /* Make sure we flush the recovery flag to disk. */
617ba13b 2885 ext4_commit_super(sb, es, 1);
ac27a0ec
DK		 2886 }
2887
2888 return 0;
2889}
2890
c4be0c1d 2891static int ext4_commit_super(struct super_block *sb,
2b2d6d01 2892 struct ext4_super_block *es, int sync)
ac27a0ec 2893{
617ba13b 2894 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
c4be0c1d 2895 int error = 0;
ac27a0ec
DK		 2896
2897 if (!sbh)
c4be0c1d 2898 return error;
914258bf
TT		 2899 if (buffer_write_io_error(sbh)) {
2900 /*
2901 * Oh, dear. A previous attempt to write the
2902 * superblock failed. This could happen because the
2903 * USB device was yanked out. Or it could happen to
2904 * be a transient write error and maybe the block will
2905 * be remapped. Nothing we can do but to retry the
2906 * write and hope for the best.
2907 */
abda1418 2908 printk(KERN_ERR "EXT4-fs: previous I/O error to "
914258bf
TT		 2909 "superblock detected for %s.\n", sb->s_id);
2910 clear_buffer_write_io_error(sbh);
2911 set_buffer_uptodate(sbh);
2912 }
ac27a0ec 2913 es->s_wtime = cpu_to_le32(get_seconds());
5d1b1b3f
AK		 2914 ext4_free_blocks_count_set(es, percpu_counter_sum_positive(
2915 &EXT4_SB(sb)->s_freeblocks_counter));
2916 es->s_free_inodes_count = cpu_to_le32(percpu_counter_sum_positive(
2917 &EXT4_SB(sb)->s_freeinodes_counter));
2918
ac27a0ec
DK		 2919 BUFFER_TRACE(sbh, "marking dirty");
2920 mark_buffer_dirty(sbh);
914258bf 2921 if (sync) {
c4be0c1d
TS		 2922 error = sync_dirty_buffer(sbh);
2923 if (error)
2924 return error;
2925
2926 error = buffer_write_io_error(sbh);
2927 if (error) {
abda1418 2928 printk(KERN_ERR "EXT4-fs: I/O error while writing "
914258bf
TT		 2929 "superblock for %s.\n", sb->s_id);
2930 clear_buffer_write_io_error(sbh);
2931 set_buffer_uptodate(sbh);
2932 }
2933 }
c4be0c1d 2934 return error;
ac27a0ec
DK		 2935}
2936
2937
2938/*
2939 * Have we just finished recovery? If so, and if we are mounting (or
2940 * remounting) the filesystem readonly, then we will end up with a
2941 * consistent fs on disk. Record that fact.
2942 */
2b2d6d01
TT		 2943static void ext4_mark_recovery_complete(struct super_block *sb,
2944 struct ext4_super_block *es)
ac27a0ec 2945{
617ba13b 2946 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 2947
0390131b
FM		 2948 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2949 BUG_ON(journal != NULL);
2950 return;
2951 }
dab291af 2952 jbd2_journal_lock_updates(journal);
7ffe1ea8
HK		 2953 if (jbd2_journal_flush(journal) < 0)
2954 goto out;
2955
32c37730 2956 lock_super(sb);
617ba13b 2957 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
ac27a0ec 2958 sb->s_flags & MS_RDONLY) {
617ba13b 2959 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 2960 sb->s_dirt = 0;
617ba13b 2961 ext4_commit_super(sb, es, 1);
ac27a0ec 2962 }
32c37730 2963 unlock_super(sb);
7ffe1ea8
HK		 2964
2965out:
dab291af 2966 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 2967}
2968
2969/*
2970 * If we are mounting (or read-write remounting) a filesystem whose journal
2971 * has recorded an error from a previous lifetime, move that error to the
2972 * main filesystem now.
2973 */
2b2d6d01
TT		 2974static void ext4_clear_journal_err(struct super_block *sb,
2975 struct ext4_super_block *es)
ac27a0ec
DK		 2976{
2977 journal_t *journal;
2978 int j_errno;
2979 const char *errstr;
2980
0390131b
FM		 2981 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
2982
617ba13b 2983 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 2984
2985 /*
2986 * Now check for any error status which may have been recorded in the
617ba13b 2987 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 2988 */
2989
dab291af 2990 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 2991 if (j_errno) {
2992 char nbuf[16];
2993
617ba13b 2994 errstr = ext4_decode_error(sb, j_errno, nbuf);
46e665e9 2995 ext4_warning(sb, __func__, "Filesystem error recorded "
ac27a0ec 2996 "from previous mount: %s", errstr);
46e665e9 2997 ext4_warning(sb, __func__, "Marking fs in need of "
ac27a0ec
DK		 2998 "filesystem check.");
2999
617ba13b
MC		 3000 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
3001 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2b2d6d01 3002 ext4_commit_super(sb, es, 1);
ac27a0ec 3003
dab291af 3004 jbd2_journal_clear_err(journal);
ac27a0ec
DK		 3005 }
3006}
3007
3008/*
3009 * Force the running and committing transactions to commit,
3010 * and wait on the commit.
3011 */
617ba13b 3012int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 3013{
3014 journal_t *journal;
0390131b 3015 int ret = 0;
ac27a0ec
DK		 3016
3017 if (sb->s_flags & MS_RDONLY)
3018 return 0;
3019
617ba13b 3020 journal = EXT4_SB(sb)->s_journal;
0390131b
FM		 3021 if (journal) {
3022 sb->s_dirt = 0;
3023 ret = ext4_journal_force_commit(journal);
3024 }
3025
ac27a0ec
DK		 3026 return ret;
3027}
3028
3029/*
617ba13b 3030 * Ext4 always journals updates to the superblock itself, so we don't
ac27a0ec 3031 * have to propagate any other updates to the superblock on disk at this
14ce0cb4
TT		 3032 * point. (We can probably nuke this function altogether, and remove
3033 * any mention to sb->s_dirt in all of fs/ext4; eventual cleanup...)
ac27a0ec 3034 */
2b2d6d01 3035static void ext4_write_super(struct super_block *sb)
ac27a0ec 3036{
0390131b
FM		 3037 if (EXT4_SB(sb)->s_journal) {
3038 if (mutex_trylock(&sb->s_lock) != 0)
3039 BUG();
3040 sb->s_dirt = 0;
3041 } else {
3042 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
3043 }
ac27a0ec
DK		 3044}
3045
617ba13b 3046static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec 3047{
14ce0cb4 3048 int ret = 0;
ac27a0ec 3049
ede86cc4 3050 trace_mark(ext4_sync_fs, "dev %s wait %d", sb->s_id, wait);
ac27a0ec 3051 sb->s_dirt = 0;
0390131b
FM		 3052 if (EXT4_SB(sb)->s_journal) {
3053 if (wait)
3054 ret = ext4_force_commit(sb);
3055 else
3056 jbd2_journal_start_commit(EXT4_SB(sb)->s_journal, NULL);
3057 } else {
3058 ext4_commit_super(sb, EXT4_SB(sb)->s_es, wait);
3059 }
14ce0cb4 3060 return ret;
ac27a0ec
DK		 3061}
3062
3063/*
3064 * LVM calls this function before a (read-only) snapshot is created. This
3065 * gives us a chance to flush the journal completely and mark the fs clean.
3066 */
c4be0c1d 3067static int ext4_freeze(struct super_block *sb)
ac27a0ec 3068{
c4be0c1d
TS		 3069 int error = 0;
3070 journal_t *journal;
ac27a0ec
DK		 3071 sb->s_dirt = 0;
3072
3073 if (!(sb->s_flags & MS_RDONLY)) {
c4be0c1d 3074 journal = EXT4_SB(sb)->s_journal;
ac27a0ec 3075
0390131b
FM		 3076 if (journal) {
3077 /* Now we set up the journal barrier. */
3078 jbd2_journal_lock_updates(journal);
7ffe1ea8 3079
0390131b
FM		 3080 /*
3081 * We don't want to clear needs_recovery flag when we
3082 * failed to flush the journal.
3083 */
c4be0c1d
TS		 3084 error = jbd2_journal_flush(journal);
3085 if (error < 0)
3086 goto out;
0390131b 3087 }
ac27a0ec
DK		 3088
3089 /* Journal blocked and flushed, clear needs_recovery flag. */
617ba13b
MC		 3090 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3091 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
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_dquot_initialize(struct inode *inode, int type)
ac27a0ec
DK		 3382{
3383 handle_t *handle;
3384 int ret, err;
3385
3386 /* We may create quota structure so we need to reserve enough blocks */
617ba13b 3387 handle = ext4_journal_start(inode, 2*EXT4_QUOTA_INIT_BLOCKS(inode->i_sb));
ac27a0ec
DK		 3388 if (IS_ERR(handle))
3389 return PTR_ERR(handle);
3390 ret = dquot_initialize(inode, type);
617ba13b 3391 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3392 if (!ret)
3393 ret = err;
3394 return ret;
3395}
3396
617ba13b 3397static int ext4_dquot_drop(struct inode *inode)
ac27a0ec
DK		 3398{
3399 handle_t *handle;
3400 int ret, err;
3401
3402 /* We may delete quota structure so we need to reserve enough blocks */
617ba13b 3403 handle = ext4_journal_start(inode, 2*EXT4_QUOTA_DEL_BLOCKS(inode->i_sb));
2887df13
JK		 3404 if (IS_ERR(handle)) {
3405 /*
3406 * We call dquot_drop() anyway to at least release references
3407 * to quota structures so that umount does not hang.
3408 */
3409 dquot_drop(inode);
ac27a0ec 3410 return PTR_ERR(handle);
2887df13 3411 }
ac27a0ec 3412 ret = dquot_drop(inode);
617ba13b 3413 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3414 if (!ret)
3415 ret = err;
3416 return ret;
3417}
3418
617ba13b 3419static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 3420{
3421 int ret, err;
3422 handle_t *handle;
3423 struct inode *inode;
3424
3425 inode = dquot_to_inode(dquot);
617ba13b
MC		 3426 handle = ext4_journal_start(inode,
3427 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 3428 if (IS_ERR(handle))
3429 return PTR_ERR(handle);
3430 ret = dquot_commit(dquot);
617ba13b 3431 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3432 if (!ret)
3433 ret = err;
3434 return ret;
3435}
3436
617ba13b 3437static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 3438{
3439 int ret, err;
3440 handle_t *handle;
3441
617ba13b
MC		 3442 handle = ext4_journal_start(dquot_to_inode(dquot),
3443 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 3444 if (IS_ERR(handle))
3445 return PTR_ERR(handle);
3446 ret = dquot_acquire(dquot);
617ba13b 3447 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3448 if (!ret)
3449 ret = err;
3450 return ret;
3451}
3452
617ba13b 3453static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 3454{
3455 int ret, err;
3456 handle_t *handle;
3457
617ba13b
MC		 3458 handle = ext4_journal_start(dquot_to_inode(dquot),
3459 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 3460 if (IS_ERR(handle)) {
3461 /* Release dquot anyway to avoid endless cycle in dqput() */
3462 dquot_release(dquot);
ac27a0ec 3463 return PTR_ERR(handle);
9c3013e9 3464 }
ac27a0ec 3465 ret = dquot_release(dquot);
617ba13b 3466 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3467 if (!ret)
3468 ret = err;
3469 return ret;
3470}
3471
617ba13b 3472static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec 3473{
2c8be6b2 3474 /* Are we journaling quotas? */
617ba13b
MC		 3475 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3476 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
ac27a0ec 3477 dquot_mark_dquot_dirty(dquot);
617ba13b 3478 return ext4_write_dquot(dquot);
ac27a0ec
DK		 3479 } else {
3480 return dquot_mark_dquot_dirty(dquot);
3481 }
3482}
3483
617ba13b 3484static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 3485{
3486 int ret, err;
3487 handle_t *handle;
3488
3489 /* Data block + inode block */
617ba13b 3490 handle = ext4_journal_start(sb->s_root->d_inode, 2);
ac27a0ec
DK		 3491 if (IS_ERR(handle))
3492 return PTR_ERR(handle);
3493 ret = dquot_commit_info(sb, type);
617ba13b 3494 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3495 if (!ret)
3496 ret = err;
3497 return ret;
3498}
3499
3500/*
3501 * Turn on quotas during mount time - we need to find
3502 * the quota file and such...
3503 */
617ba13b 3504static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 3505{
617ba13b
MC		 3506 return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
3507 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 3508}
3509
3510/*
3511 * Standard function to be called on quota_on
3512 */
617ba13b 3513static int ext4_quota_on(struct super_block *sb, int type, int format_id,
8264613d 3514 char *name, int remount)
ac27a0ec
DK		 3515{
3516 int err;
8264613d 3517 struct path path;
ac27a0ec
DK		 3518
3519 if (!test_opt(sb, QUOTA))
3520 return -EINVAL;
8264613d 3521 /* When remounting, no checks are needed and in fact, name is NULL */
0623543b 3522 if (remount)
8264613d 3523 return vfs_quota_on(sb, type, format_id, name, remount);
0623543b 3524
8264613d 3525 err = kern_path(name, LOOKUP_FOLLOW, &path);
ac27a0ec
DK		 3526 if (err)
3527 return err;
0623543b 3528
ac27a0ec 3529 /* Quotafile not on the same filesystem? */
8264613d
AV		 3530 if (path.mnt->mnt_sb != sb) {
3531 path_put(&path);
ac27a0ec
DK		 3532 return -EXDEV;
3533 }
0623543b
JK		 3534 /* Journaling quota? */
3535 if (EXT4_SB(sb)->s_qf_names[type]) {
2b2d6d01 3536 /* Quotafile not in fs root? */
8264613d 3537 if (path.dentry->d_parent != sb->s_root)
0623543b
JK		 3538 printk(KERN_WARNING
3539 "EXT4-fs: Quota file not on filesystem root. "
3540 "Journaled quota will not work.\n");
2b2d6d01 3541 }
0623543b
JK		 3542
3543 /*
3544 * When we journal data on quota file, we have to flush journal to see
3545 * all updates to the file when we bypass pagecache...
3546 */
0390131b
FM		 3547 if (EXT4_SB(sb)->s_journal &&
3548 ext4_should_journal_data(path.dentry->d_inode)) {
0623543b
JK		 3549 /*
3550 * We don't need to lock updates but journal_flush() could
3551 * otherwise be livelocked...
3552 */
3553 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 3554 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
0623543b 3555 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 3556 if (err) {
8264613d 3557 path_put(&path);
7ffe1ea8
HK		 3558 return err;
3559 }
0623543b
JK		 3560 }
3561
8264613d
AV		 3562 err = vfs_quota_on_path(sb, type, format_id, &path);
3563 path_put(&path);
77e69dac 3564 return err;
ac27a0ec
DK		 3565}
3566
3567/* Read data from quotafile - avoid pagecache and such because we cannot afford
3568 * acquiring the locks... As quota files are never truncated and quota code
3569 * itself serializes the operations (and noone else should touch the files)
3570 * we don't have to be afraid of races */
617ba13b 3571static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 3572 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;
3579 size_t toread;
3580 struct buffer_head *bh;
3581 loff_t i_size = i_size_read(inode);
3582
3583 if (off > i_size)
3584 return 0;
3585 if (off+len > i_size)
3586 len = i_size-off;
3587 toread = len;
3588 while (toread > 0) {
3589 tocopy = sb->s_blocksize - offset < toread ?
3590 sb->s_blocksize - offset : toread;
617ba13b 3591 bh = ext4_bread(NULL, inode, blk, 0, &err);
ac27a0ec
DK		 3592 if (err)
3593 return err;
3594 if (!bh) /* A hole? */
3595 memset(data, 0, tocopy);
3596 else
3597 memcpy(data, bh->b_data+offset, tocopy);
3598 brelse(bh);
3599 offset = 0;
3600 toread -= tocopy;
3601 data += tocopy;
3602 blk++;
3603 }
3604 return len;
3605}
3606
3607/* Write to quotafile (we know the transaction is already started and has
3608 * enough credits) */
617ba13b 3609static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 3610 const char *data, size_t len, loff_t off)
3611{
3612 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 3613 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 3614 int err = 0;
3615 int offset = off & (sb->s_blocksize - 1);
3616 int tocopy;
617ba13b 3617 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
ac27a0ec
DK		 3618 size_t towrite = len;
3619 struct buffer_head *bh;
3620 handle_t *handle = journal_current_handle();
3621
0390131b 3622 if (EXT4_SB(sb)->s_journal && !handle) {
e5f8eab8 3623 printk(KERN_WARNING "EXT4-fs: Quota write (off=%llu, len=%llu)"
9c3013e9
JK		 3624 " cancelled because transaction is not started.\n",
3625 (unsigned long long)off, (unsigned long long)len);
3626 return -EIO;
3627 }
ac27a0ec
DK		 3628 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
3629 while (towrite > 0) {
3630 tocopy = sb->s_blocksize - offset < towrite ?
3631 sb->s_blocksize - offset : towrite;
617ba13b 3632 bh = ext4_bread(handle, inode, blk, 1, &err);
ac27a0ec
DK		 3633 if (!bh)
3634 goto out;
3635 if (journal_quota) {
617ba13b 3636 err = ext4_journal_get_write_access(handle, bh);
ac27a0ec
DK		 3637 if (err) {
3638 brelse(bh);
3639 goto out;
3640 }
3641 }
3642 lock_buffer(bh);
3643 memcpy(bh->b_data+offset, data, tocopy);
3644 flush_dcache_page(bh->b_page);
3645 unlock_buffer(bh);
3646 if (journal_quota)
0390131b 3647 err = ext4_handle_dirty_metadata(handle, NULL, bh);
ac27a0ec
DK		 3648 else {
3649 /* Always do at least ordered writes for quotas */
678aaf48 3650 err = ext4_jbd2_file_inode(handle, inode);
ac27a0ec
DK		 3651 mark_buffer_dirty(bh);
3652 }
3653 brelse(bh);
3654 if (err)
3655 goto out;
3656 offset = 0;
3657 towrite -= tocopy;
3658 data += tocopy;
3659 blk++;
3660 }
3661out:
4d04e4fb
JK		 3662 if (len == towrite) {
3663 mutex_unlock(&inode->i_mutex);
ac27a0ec 3664 return err;
4d04e4fb 3665 }
ac27a0ec
DK		 3666 if (inode->i_size < off+len-towrite) {
3667 i_size_write(inode, off+len-towrite);
617ba13b 3668 EXT4_I(inode)->i_disksize = inode->i_size;
ac27a0ec 3669 }
ac27a0ec 3670 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
617ba13b 3671 ext4_mark_inode_dirty(handle, inode);
ac27a0ec
DK		 3672 mutex_unlock(&inode->i_mutex);
3673 return len - towrite;
3674}
3675
3676#endif
3677
617ba13b 3678static int ext4_get_sb(struct file_system_type *fs_type,
ac27a0ec
DK		 3679 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3680{
617ba13b 3681 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
ac27a0ec
DK		 3682}
3683
5e8814f2
TT		 3684#ifdef CONFIG_PROC_FS
3685static int ext4_ui_proc_show(struct seq_file *m, void *v)
3686{
3687 unsigned int *p = m->private;
3688
3689 seq_printf(m, "%u\n", *p);
3690 return 0;
3691}
3692
3693static int ext4_ui_proc_open(struct inode *inode, struct file *file)
3694{
3695 return single_open(file, ext4_ui_proc_show, PDE(inode)->data);
3696}
3697
3698static ssize_t ext4_ui_proc_write(struct file *file, const char __user *buf,
3699 size_t cnt, loff_t *ppos)
3700{
23475e26 3701 unsigned long *p = PDE(file->f_path.dentry->d_inode)->data;
5e8814f2 3702 char str[32];
5e8814f2
TT		 3703
3704 if (cnt >= sizeof(str))
3705 return -EINVAL;
3706 if (copy_from_user(str, buf, cnt))
3707 return -EFAULT;
23475e26
RK		 3708
3709 *p = simple_strtoul(str, NULL, 0);
5e8814f2
TT		 3710 return cnt;
3711}
3712
3713const struct file_operations ext4_ui_proc_fops = {
3714 .owner = THIS_MODULE,
3715 .open = ext4_ui_proc_open,
3716 .read = seq_read,
3717 .llseek = seq_lseek,
3718 .release = single_release,
3719 .write = ext4_ui_proc_write,
3720};
3721#endif
3722
03010a33
TT		 3723static struct file_system_type ext4_fs_type = {
3724 .owner = THIS_MODULE,
3725 .name = "ext4",
3726 .get_sb = ext4_get_sb,
3727 .kill_sb = kill_block_super,
3728 .fs_flags = FS_REQUIRES_DEV,
3729};
3730
3731#ifdef CONFIG_EXT4DEV_COMPAT
3732static int ext4dev_get_sb(struct file_system_type *fs_type,
3733 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3734{
3735 printk(KERN_WARNING "EXT4-fs: Update your userspace programs "
3736 "to mount using ext4\n");
3737 printk(KERN_WARNING "EXT4-fs: ext4dev backwards compatibility "
3738 "will go away by 2.6.31\n");
3739 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
3740}
3741
617ba13b 3742static struct file_system_type ext4dev_fs_type = {
ac27a0ec 3743 .owner = THIS_MODULE,
617ba13b 3744 .name = "ext4dev",
03010a33 3745 .get_sb = ext4dev_get_sb,
ac27a0ec
DK		 3746 .kill_sb = kill_block_super,
3747 .fs_flags = FS_REQUIRES_DEV,
3748};
03010a33
TT		 3749MODULE_ALIAS("ext4dev");
3750#endif
ac27a0ec 3751
617ba13b 3752static int __init init_ext4_fs(void)
ac27a0ec 3753{
c9de560d
AT		 3754 int err;
3755
9f6200bb 3756 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
c9de560d 3757 err = init_ext4_mballoc();
ac27a0ec
DK		 3758 if (err)
3759 return err;
c9de560d
AT		 3760
3761 err = init_ext4_xattr();
3762 if (err)
3763 goto out2;
ac27a0ec
DK		 3764 err = init_inodecache();
3765 if (err)
3766 goto out1;
03010a33 3767 err = register_filesystem(&ext4_fs_type);
ac27a0ec
DK		 3768 if (err)
3769 goto out;
03010a33
TT		 3770#ifdef CONFIG_EXT4DEV_COMPAT
3771 err = register_filesystem(&ext4dev_fs_type);
3772 if (err) {
3773 unregister_filesystem(&ext4_fs_type);
3774 goto out;
3775 }
3776#endif
ac27a0ec
DK		 3777 return 0;
3778out:
3779 destroy_inodecache();
3780out1:
617ba13b 3781 exit_ext4_xattr();
c9de560d
AT		 3782out2:
3783 exit_ext4_mballoc();
ac27a0ec
DK		 3784 return err;
3785}
3786
617ba13b 3787static void __exit exit_ext4_fs(void)
ac27a0ec 3788{
03010a33
TT		 3789 unregister_filesystem(&ext4_fs_type);
3790#ifdef CONFIG_EXT4DEV_COMPAT
617ba13b 3791 unregister_filesystem(&ext4dev_fs_type);
03010a33 3792#endif
ac27a0ec 3793 destroy_inodecache();
617ba13b 3794 exit_ext4_xattr();
c9de560d 3795 exit_ext4_mballoc();
9f6200bb 3796 remove_proc_entry("fs/ext4", NULL);
ac27a0ec
DK		 3797}
3798
3799MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
83982b6f 3800MODULE_DESCRIPTION("Fourth Extended Filesystem");
ac27a0ec 3801MODULE_LICENSE("GPL");
617ba13b
MC		 3802module_init(init_ext4_fs)
3803module_exit(exit_ext4_fs)
Clone of davem's net-next-2.6 tree