]> 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
ext4: Add mount option to set kjournald's I/O priority
[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);
2b2d6d01
TT		 54static void ext4_commit_super(struct super_block *sb,
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);
617ba13b 65static void ext4_unlockfs(struct super_block *sb);
2b2d6d01 66static void ext4_write_super(struct super_block *sb);
617ba13b 67static void ext4_write_super_lockfs(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:
617ba13b 508 printk(KERN_ERR "EXT4: 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");
832 if (!test_opt(sb, EXTENTS))
833 seq_puts(seq, ",noextents");
25ec56b5
JNC		 834 if (test_opt(sb, I_VERSION))
835 seq_puts(seq, ",i_version");
dd919b98
AK		 836 if (!test_opt(sb, DELALLOC))
837 seq_puts(seq, ",nodelalloc");
838
ac27a0ec 839
cb45bbe4
MS		 840 if (sbi->s_stripe)
841 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
aa22df2c
AK		 842 /*
843 * journal mode get enabled in different ways
844 * So just print the value even if we didn't specify it
845 */
617ba13b 846 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
ac27a0ec 847 seq_puts(seq, ",data=journal");
617ba13b 848 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
ac27a0ec 849 seq_puts(seq, ",data=ordered");
617ba13b 850 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
ac27a0ec
DK		 851 seq_puts(seq, ",data=writeback");
852
240799cd
TT		 853 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
854 seq_printf(seq, ",inode_readahead_blks=%u",
855 sbi->s_inode_readahead_blks);
856
5bf5683a
HK		 857 if (test_opt(sb, DATA_ERR_ABORT))
858 seq_puts(seq, ",data_err=abort");
859
617ba13b 860 ext4_show_quota_options(seq, sb);
ac27a0ec
DK		 861 return 0;
862}
863
864
1b961ac0
CH		 865static struct inode *ext4_nfs_get_inode(struct super_block *sb,
866 u64 ino, u32 generation)
ac27a0ec 867{
ac27a0ec 868 struct inode *inode;
ac27a0ec 869
617ba13b 870 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
ac27a0ec 871 return ERR_PTR(-ESTALE);
617ba13b 872 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
ac27a0ec
DK		 873 return ERR_PTR(-ESTALE);
874
875 /* iget isn't really right if the inode is currently unallocated!!
876 *
617ba13b 877 * ext4_read_inode will return a bad_inode if the inode had been
ac27a0ec
DK		 878 * deleted, so we should be safe.
879 *
880 * Currently we don't know the generation for parent directory, so
881 * a generation of 0 means "accept any"
882 */
1d1fe1ee
DH		 883 inode = ext4_iget(sb, ino);
884 if (IS_ERR(inode))
885 return ERR_CAST(inode);
886 if (generation && inode->i_generation != generation) {
ac27a0ec
DK		 887 iput(inode);
888 return ERR_PTR(-ESTALE);
889 }
1b961ac0
CH		 890
891 return inode;
892}
893
894static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
895 int fh_len, int fh_type)
896{
897 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
898 ext4_nfs_get_inode);
899}
900
901static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
902 int fh_len, int fh_type)
903{
904 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
905 ext4_nfs_get_inode);
ac27a0ec
DK		 906}
907
c39a7f84
TO		 908/*
909 * Try to release metadata pages (indirect blocks, directories) which are
910 * mapped via the block device. Since these pages could have journal heads
911 * which would prevent try_to_free_buffers() from freeing them, we must use
912 * jbd2 layer's try_to_free_buffers() function to release them.
913 */
914static int bdev_try_to_free_page(struct super_block *sb, struct page *page, gfp_t wait)
915{
916 journal_t *journal = EXT4_SB(sb)->s_journal;
917
918 WARN_ON(PageChecked(page));
919 if (!page_has_buffers(page))
920 return 0;
921 if (journal)
922 return jbd2_journal_try_to_free_buffers(journal, page,
923 wait & ~__GFP_WAIT);
924 return try_to_free_buffers(page);
925}
926
ac27a0ec 927#ifdef CONFIG_QUOTA
af5bc92d 928#define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
2b2d6d01 929#define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
ac27a0ec 930
617ba13b
MC		 931static int ext4_dquot_initialize(struct inode *inode, int type);
932static int ext4_dquot_drop(struct inode *inode);
933static int ext4_write_dquot(struct dquot *dquot);
934static int ext4_acquire_dquot(struct dquot *dquot);
935static int ext4_release_dquot(struct dquot *dquot);
936static int ext4_mark_dquot_dirty(struct dquot *dquot);
937static int ext4_write_info(struct super_block *sb, int type);
6f28e087
JK		 938static int ext4_quota_on(struct super_block *sb, int type, int format_id,
939 char *path, int remount);
617ba13b
MC		 940static int ext4_quota_on_mount(struct super_block *sb, int type);
941static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 942 size_t len, loff_t off);
617ba13b 943static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 944 const char *data, size_t len, loff_t off);
945
617ba13b
MC		 946static struct dquot_operations ext4_quota_operations = {
947 .initialize = ext4_dquot_initialize,
948 .drop = ext4_dquot_drop,
ac27a0ec
DK		 949 .alloc_space = dquot_alloc_space,
950 .alloc_inode = dquot_alloc_inode,
951 .free_space = dquot_free_space,
952 .free_inode = dquot_free_inode,
953 .transfer = dquot_transfer,
617ba13b
MC		 954 .write_dquot = ext4_write_dquot,
955 .acquire_dquot = ext4_acquire_dquot,
956 .release_dquot = ext4_release_dquot,
957 .mark_dirty = ext4_mark_dquot_dirty,
958 .write_info = ext4_write_info
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,
981 .write_super_lockfs = ext4_write_super_lockfs,
982 .unlockfs = ext4_unlockfs,
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,
25ec56b5 1014 Opt_grpquota, Opt_extents, Opt_noextents, 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"},
a86c6181 1069 {Opt_extents, "extents"},
1e2462f9 1070 {Opt_noextents, "noextents"},
25ec56b5 1071 {Opt_i_version, "i_version"},
c9de560d 1072 {Opt_stripe, "stripe=%u"},
ac27a0ec 1073 {Opt_resize, "resize"},
64769240 1074 {Opt_delalloc, "delalloc"},
dd919b98 1075 {Opt_nodelalloc, "nodelalloc"},
240799cd 1076 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
b3881f74 1077 {Opt_journal_ioprio, "journal_ioprio=%u"},
f3f12faa 1078 {Opt_err, NULL},
ac27a0ec
DK		 1079};
1080
617ba13b 1081static ext4_fsblk_t get_sb_block(void **data)
ac27a0ec 1082{
617ba13b 1083 ext4_fsblk_t sb_block;
ac27a0ec
DK		 1084 char *options = (char *) *data;
1085
1086 if (!options || strncmp(options, "sb=", 3) != 0)
1087 return 1; /* Default location */
1088 options += 3;
617ba13b 1089 /*todo: use simple_strtoll with >32bit ext4 */
ac27a0ec
DK		 1090 sb_block = simple_strtoul(options, &options, 0);
1091 if (*options && *options != ',') {
4776004f 1092 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
ac27a0ec
DK		 1093 (char *) *data);
1094 return 1;
1095 }
1096 if (*options == ',')
1097 options++;
1098 *data = (void *) options;
1099 return sb_block;
1100}
1101
b3881f74
TT		 1102#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1103
2b2d6d01 1104static int parse_options(char *options, struct super_block *sb,
c3191067 1105 unsigned long *journal_devnum,
b3881f74 1106 unsigned int *journal_ioprio,
2b2d6d01 1107 ext4_fsblk_t *n_blocks_count, int is_remount)
ac27a0ec 1108{
617ba13b 1109 struct ext4_sb_info *sbi = EXT4_SB(sb);
2b2d6d01 1110 char *p;
ac27a0ec
DK		 1111 substring_t args[MAX_OPT_ARGS];
1112 int data_opt = 0;
1113 int option;
1114#ifdef CONFIG_QUOTA
dfc5d03f 1115 int qtype, qfmt;
ac27a0ec
DK		 1116 char *qname;
1117#endif
c07651b5 1118 ext4_fsblk_t last_block;
ac27a0ec
DK		 1119
1120 if (!options)
1121 return 1;
1122
2b2d6d01 1123 while ((p = strsep(&options, ",")) != NULL) {
ac27a0ec
DK		 1124 int token;
1125 if (!*p)
1126 continue;
1127
1128 token = match_token(p, tokens, args);
1129 switch (token) {
1130 case Opt_bsd_df:
2b2d6d01 1131 clear_opt(sbi->s_mount_opt, MINIX_DF);
ac27a0ec
DK		 1132 break;
1133 case Opt_minix_df:
2b2d6d01 1134 set_opt(sbi->s_mount_opt, MINIX_DF);
ac27a0ec
DK		 1135 break;
1136 case Opt_grpid:
2b2d6d01 1137 set_opt(sbi->s_mount_opt, GRPID);
ac27a0ec
DK		 1138 break;
1139 case Opt_nogrpid:
2b2d6d01 1140 clear_opt(sbi->s_mount_opt, GRPID);
ac27a0ec
DK		 1141 break;
1142 case Opt_resuid:
1143 if (match_int(&args[0], &option))
1144 return 0;
1145 sbi->s_resuid = option;
1146 break;
1147 case Opt_resgid:
1148 if (match_int(&args[0], &option))
1149 return 0;
1150 sbi->s_resgid = option;
1151 break;
1152 case Opt_sb:
1153 /* handled by get_sb_block() instead of here */
1154 /* *sb_block = match_int(&args[0]); */
1155 break;
1156 case Opt_err_panic:
2b2d6d01
TT		 1157 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1158 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1159 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
ac27a0ec
DK		 1160 break;
1161 case Opt_err_ro:
2b2d6d01
TT		 1162 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1163 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1164 set_opt(sbi->s_mount_opt, ERRORS_RO);
ac27a0ec
DK		 1165 break;
1166 case Opt_err_cont:
2b2d6d01
TT		 1167 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1168 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1169 set_opt(sbi->s_mount_opt, ERRORS_CONT);
ac27a0ec
DK		 1170 break;
1171 case Opt_nouid32:
2b2d6d01 1172 set_opt(sbi->s_mount_opt, NO_UID32);
ac27a0ec 1173 break;
ac27a0ec 1174 case Opt_debug:
2b2d6d01 1175 set_opt(sbi->s_mount_opt, DEBUG);
ac27a0ec
DK		 1176 break;
1177 case Opt_oldalloc:
2b2d6d01 1178 set_opt(sbi->s_mount_opt, OLDALLOC);
ac27a0ec
DK		 1179 break;
1180 case Opt_orlov:
2b2d6d01 1181 clear_opt(sbi->s_mount_opt, OLDALLOC);
ac27a0ec 1182 break;
03010a33 1183#ifdef CONFIG_EXT4_FS_XATTR
ac27a0ec 1184 case Opt_user_xattr:
2b2d6d01 1185 set_opt(sbi->s_mount_opt, XATTR_USER);
ac27a0ec
DK		 1186 break;
1187 case Opt_nouser_xattr:
2b2d6d01 1188 clear_opt(sbi->s_mount_opt, XATTR_USER);
ac27a0ec
DK		 1189 break;
1190#else
1191 case Opt_user_xattr:
1192 case Opt_nouser_xattr:
4776004f
TT		 1193 printk(KERN_ERR "EXT4 (no)user_xattr options "
1194 "not supported\n");
ac27a0ec
DK		 1195 break;
1196#endif
03010a33 1197#ifdef CONFIG_EXT4_FS_POSIX_ACL
ac27a0ec
DK		 1198 case Opt_acl:
1199 set_opt(sbi->s_mount_opt, POSIX_ACL);
1200 break;
1201 case Opt_noacl:
1202 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1203 break;
1204#else
1205 case Opt_acl:
1206 case Opt_noacl:
4776004f
TT		 1207 printk(KERN_ERR "EXT4 (no)acl options "
1208 "not supported\n");
ac27a0ec
DK		 1209 break;
1210#endif
1211 case Opt_reservation:
1212 set_opt(sbi->s_mount_opt, RESERVATION);
1213 break;
1214 case Opt_noreservation:
1215 clear_opt(sbi->s_mount_opt, RESERVATION);
1216 break;
1217 case Opt_journal_update:
1218 /* @@@ FIXME */
1219 /* Eventually we will want to be able to create
1220 a journal file here. For now, only allow the
1221 user to specify an existing inode to be the
1222 journal file. */
1223 if (is_remount) {
617ba13b 1224 printk(KERN_ERR "EXT4-fs: cannot specify "
ac27a0ec
DK		 1225 "journal on remount\n");
1226 return 0;
1227 }
2b2d6d01 1228 set_opt(sbi->s_mount_opt, UPDATE_JOURNAL);
ac27a0ec 1229 break;
ac27a0ec
DK		 1230 case Opt_journal_dev:
1231 if (is_remount) {
617ba13b 1232 printk(KERN_ERR "EXT4-fs: cannot specify "
ac27a0ec
DK		 1233 "journal on remount\n");
1234 return 0;
1235 }
1236 if (match_int(&args[0], &option))
1237 return 0;
1238 *journal_devnum = option;
1239 break;
818d276c
GS		 1240 case Opt_journal_checksum:
1241 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1242 break;
1243 case Opt_journal_async_commit:
1244 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
1245 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1246 break;
ac27a0ec 1247 case Opt_noload:
2b2d6d01 1248 set_opt(sbi->s_mount_opt, NOLOAD);
ac27a0ec
DK		 1249 break;
1250 case Opt_commit:
1251 if (match_int(&args[0], &option))
1252 return 0;
1253 if (option < 0)
1254 return 0;
1255 if (option == 0)
cd02ff0b 1256 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
ac27a0ec
DK		 1257 sbi->s_commit_interval = HZ * option;
1258 break;
30773840
TT		 1259 case Opt_max_batch_time:
1260 if (match_int(&args[0], &option))
1261 return 0;
1262 if (option < 0)
1263 return 0;
1264 if (option == 0)
1265 option = EXT4_DEF_MAX_BATCH_TIME;
1266 sbi->s_max_batch_time = option;
1267 break;
1268 case Opt_min_batch_time:
1269 if (match_int(&args[0], &option))
1270 return 0;
1271 if (option < 0)
1272 return 0;
1273 sbi->s_min_batch_time = option;
1274 break;
ac27a0ec 1275 case Opt_data_journal:
617ba13b 1276 data_opt = EXT4_MOUNT_JOURNAL_DATA;
ac27a0ec
DK		 1277 goto datacheck;
1278 case Opt_data_ordered:
617ba13b 1279 data_opt = EXT4_MOUNT_ORDERED_DATA;
ac27a0ec
DK		 1280 goto datacheck;
1281 case Opt_data_writeback:
617ba13b 1282 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
ac27a0ec
DK		 1283 datacheck:
1284 if (is_remount) {
617ba13b 1285 if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
ac27a0ec
DK		 1286 != data_opt) {
1287 printk(KERN_ERR
617ba13b 1288 "EXT4-fs: cannot change data "
ac27a0ec
DK		 1289 "mode on remount\n");
1290 return 0;
1291 }
1292 } else {
617ba13b 1293 sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
ac27a0ec
DK		 1294 sbi->s_mount_opt |= data_opt;
1295 }
1296 break;
5bf5683a
HK		 1297 case Opt_data_err_abort:
1298 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1299 break;
1300 case Opt_data_err_ignore:
1301 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1302 break;
ac27a0ec
DK		 1303#ifdef CONFIG_QUOTA
1304 case Opt_usrjquota:
1305 qtype = USRQUOTA;
1306 goto set_qf_name;
1307 case Opt_grpjquota:
1308 qtype = GRPQUOTA;
1309set_qf_name:
dfc5d03f
JK		 1310 if ((sb_any_quota_enabled(sb) ||
1311 sb_any_quota_suspended(sb)) &&
1312 !sbi->s_qf_names[qtype]) {
ac27a0ec 1313 printk(KERN_ERR
4776004f
TT		 1314 "EXT4-fs: Cannot change journaled "
1315 "quota options when quota turned on.\n");
ac27a0ec
DK		 1316 return 0;
1317 }
1318 qname = match_strdup(&args[0]);
1319 if (!qname) {
1320 printk(KERN_ERR
617ba13b 1321 "EXT4-fs: not enough memory for "
ac27a0ec
DK		 1322 "storing quotafile name.\n");
1323 return 0;
1324 }
1325 if (sbi->s_qf_names[qtype] &&
1326 strcmp(sbi->s_qf_names[qtype], qname)) {
1327 printk(KERN_ERR
617ba13b 1328 "EXT4-fs: %s quota file already "
ac27a0ec
DK		 1329 "specified.\n", QTYPE2NAME(qtype));
1330 kfree(qname);
1331 return 0;
1332 }
1333 sbi->s_qf_names[qtype] = qname;
1334 if (strchr(sbi->s_qf_names[qtype], '/')) {
1335 printk(KERN_ERR
617ba13b 1336 "EXT4-fs: quotafile must be on "
ac27a0ec
DK		 1337 "filesystem root.\n");
1338 kfree(sbi->s_qf_names[qtype]);
1339 sbi->s_qf_names[qtype] = NULL;
1340 return 0;
1341 }
1342 set_opt(sbi->s_mount_opt, QUOTA);
1343 break;
1344 case Opt_offusrjquota:
1345 qtype = USRQUOTA;
1346 goto clear_qf_name;
1347 case Opt_offgrpjquota:
1348 qtype = GRPQUOTA;
1349clear_qf_name:
dfc5d03f
JK		 1350 if ((sb_any_quota_enabled(sb) ||
1351 sb_any_quota_suspended(sb)) &&
1352 sbi->s_qf_names[qtype]) {
617ba13b 1353 printk(KERN_ERR "EXT4-fs: Cannot change "
2c8be6b2 1354 "journaled quota options when "
ac27a0ec
DK		 1355 "quota turned on.\n");
1356 return 0;
1357 }
1358 /*
1359 * The space will be released later when all options
1360 * are confirmed to be correct
1361 */
1362 sbi->s_qf_names[qtype] = NULL;
1363 break;
1364 case Opt_jqfmt_vfsold:
dfc5d03f
JK		 1365 qfmt = QFMT_VFS_OLD;
1366 goto set_qf_format;
ac27a0ec 1367 case Opt_jqfmt_vfsv0:
dfc5d03f
JK		 1368 qfmt = QFMT_VFS_V0;
1369set_qf_format:
1370 if ((sb_any_quota_enabled(sb) ||
1371 sb_any_quota_suspended(sb)) &&
1372 sbi->s_jquota_fmt != qfmt) {
1373 printk(KERN_ERR "EXT4-fs: Cannot change "
1374 "journaled quota options when "
1375 "quota turned on.\n");
1376 return 0;
1377 }
1378 sbi->s_jquota_fmt = qfmt;
ac27a0ec
DK		 1379 break;
1380 case Opt_quota:
1381 case Opt_usrquota:
1382 set_opt(sbi->s_mount_opt, QUOTA);
1383 set_opt(sbi->s_mount_opt, USRQUOTA);
1384 break;
1385 case Opt_grpquota:
1386 set_opt(sbi->s_mount_opt, QUOTA);
1387 set_opt(sbi->s_mount_opt, GRPQUOTA);
1388 break;
1389 case Opt_noquota:
1390 if (sb_any_quota_enabled(sb)) {
617ba13b 1391 printk(KERN_ERR "EXT4-fs: Cannot change quota "
ac27a0ec
DK		 1392 "options when quota turned on.\n");
1393 return 0;
1394 }
1395 clear_opt(sbi->s_mount_opt, QUOTA);
1396 clear_opt(sbi->s_mount_opt, USRQUOTA);
1397 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1398 break;
1399#else
1400 case Opt_quota:
1401 case Opt_usrquota:
1402 case Opt_grpquota:
cd59e7b9
JK		 1403 printk(KERN_ERR
1404 "EXT4-fs: quota options not supported.\n");
1405 break;
ac27a0ec
DK		 1406 case Opt_usrjquota:
1407 case Opt_grpjquota:
1408 case Opt_offusrjquota:
1409 case Opt_offgrpjquota:
1410 case Opt_jqfmt_vfsold:
1411 case Opt_jqfmt_vfsv0:
1412 printk(KERN_ERR
cd59e7b9 1413 "EXT4-fs: journaled quota options not "
ac27a0ec
DK		 1414 "supported.\n");
1415 break;
1416 case Opt_noquota:
1417 break;
1418#endif
1419 case Opt_abort:
1420 set_opt(sbi->s_mount_opt, ABORT);
1421 break;
1422 case Opt_barrier:
1423 if (match_int(&args[0], &option))
1424 return 0;
1425 if (option)
1426 set_opt(sbi->s_mount_opt, BARRIER);
1427 else
1428 clear_opt(sbi->s_mount_opt, BARRIER);
1429 break;
1430 case Opt_ignore:
1431 break;
1432 case Opt_resize:
1433 if (!is_remount) {
617ba13b 1434 printk("EXT4-fs: resize option only available "
ac27a0ec
DK		 1435 "for remount\n");
1436 return 0;
1437 }
1438 if (match_int(&args[0], &option) != 0)
1439 return 0;
1440 *n_blocks_count = option;
1441 break;
1442 case Opt_nobh:
1443 set_opt(sbi->s_mount_opt, NOBH);
1444 break;
1445 case Opt_bh:
1446 clear_opt(sbi->s_mount_opt, NOBH);
1447 break;
a86c6181 1448 case Opt_extents:
e4079a11
ES		 1449 if (!EXT4_HAS_INCOMPAT_FEATURE(sb,
1450 EXT4_FEATURE_INCOMPAT_EXTENTS)) {
1451 ext4_warning(sb, __func__,
1452 "extents feature not enabled "
fde4d95a 1453 "on this filesystem, use tune2fs");
e4079a11
ES		 1454 return 0;
1455 }
2b2d6d01 1456 set_opt(sbi->s_mount_opt, EXTENTS);
a86c6181 1457 break;
1e2462f9 1458 case Opt_noextents:
c07651b5
AK		 1459 /*
1460 * When e2fsprogs support resizing an already existing
1461 * ext3 file system to greater than 2**32 we need to
1462 * add support to block allocator to handle growing
1463 * already existing block mapped inode so that blocks
1464 * allocated for them fall within 2**32
1465 */
1466 last_block = ext4_blocks_count(sbi->s_es) - 1;
1467 if (last_block > 0xffffffffULL) {
1468 printk(KERN_ERR "EXT4-fs: Filesystem too "
1469 "large to mount with "
1470 "-o noextents options\n");
1471 return 0;
1472 }
2b2d6d01 1473 clear_opt(sbi->s_mount_opt, EXTENTS);
1e2462f9 1474 break;
25ec56b5
JNC		 1475 case Opt_i_version:
1476 set_opt(sbi->s_mount_opt, I_VERSION);
1477 sb->s_flags |= MS_I_VERSION;
1478 break;
dd919b98
AK		 1479 case Opt_nodelalloc:
1480 clear_opt(sbi->s_mount_opt, DELALLOC);
1481 break;
c9de560d
AT		 1482 case Opt_stripe:
1483 if (match_int(&args[0], &option))
1484 return 0;
1485 if (option < 0)
1486 return 0;
1487 sbi->s_stripe = option;
1488 break;
64769240
AT		 1489 case Opt_delalloc:
1490 set_opt(sbi->s_mount_opt, DELALLOC);
1491 break;
240799cd
TT		 1492 case Opt_inode_readahead_blks:
1493 if (match_int(&args[0], &option))
1494 return 0;
1495 if (option < 0 || option > (1 << 30))
1496 return 0;
1497 sbi->s_inode_readahead_blks = option;
1498 break;
b3881f74
TT		 1499 case Opt_journal_ioprio:
1500 if (match_int(&args[0], &option))
1501 return 0;
1502 if (option < 0 || option > 7)
1503 break;
1504 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1505 option);
1506 break;
ac27a0ec 1507 default:
2b2d6d01
TT		 1508 printk(KERN_ERR
1509 "EXT4-fs: Unrecognized mount option \"%s\" "
1510 "or missing value\n", p);
ac27a0ec
DK		 1511 return 0;
1512 }
1513 }
1514#ifdef CONFIG_QUOTA
1515 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
617ba13b 1516 if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
ac27a0ec
DK		 1517 sbi->s_qf_names[USRQUOTA])
1518 clear_opt(sbi->s_mount_opt, USRQUOTA);
1519
617ba13b 1520 if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
ac27a0ec
DK		 1521 sbi->s_qf_names[GRPQUOTA])
1522 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1523
1524 if ((sbi->s_qf_names[USRQUOTA] &&
617ba13b 1525 (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
ac27a0ec 1526 (sbi->s_qf_names[GRPQUOTA] &&
617ba13b
MC		 1527 (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
1528 printk(KERN_ERR "EXT4-fs: old and new quota "
ac27a0ec
DK		 1529 "format mixing.\n");
1530 return 0;
1531 }
1532
1533 if (!sbi->s_jquota_fmt) {
2c8be6b2 1534 printk(KERN_ERR "EXT4-fs: journaled quota format "
ac27a0ec
DK		 1535 "not specified.\n");
1536 return 0;
1537 }
1538 } else {
1539 if (sbi->s_jquota_fmt) {
2c8be6b2
JK		 1540 printk(KERN_ERR "EXT4-fs: journaled quota format "
1541 "specified with no journaling "
ac27a0ec
DK		 1542 "enabled.\n");
1543 return 0;
1544 }
1545 }
1546#endif
1547 return 1;
1548}
1549
617ba13b 1550static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 1551 int read_only)
1552{
617ba13b 1553 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 1554 int res = 0;
1555
617ba13b 1556 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
2b2d6d01
TT		 1557 printk(KERN_ERR "EXT4-fs warning: revision level too high, "
1558 "forcing read-only mode\n");
ac27a0ec
DK		 1559 res = MS_RDONLY;
1560 }
1561 if (read_only)
1562 return res;
617ba13b 1563 if (!(sbi->s_mount_state & EXT4_VALID_FS))
2b2d6d01
TT		 1564 printk(KERN_WARNING "EXT4-fs warning: mounting unchecked fs, "
1565 "running e2fsck is recommended\n");
617ba13b 1566 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
2b2d6d01
TT		 1567 printk(KERN_WARNING
1568 "EXT4-fs warning: mounting fs with errors, "
1569 "running e2fsck is recommended\n");
ac27a0ec
DK		 1570 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1571 le16_to_cpu(es->s_mnt_count) >=
1572 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
2b2d6d01
TT		 1573 printk(KERN_WARNING
1574 "EXT4-fs warning: maximal mount count reached, "
1575 "running e2fsck is recommended\n");
ac27a0ec
DK		 1576 else if (le32_to_cpu(es->s_checkinterval) &&
1577 (le32_to_cpu(es->s_lastcheck) +
1578 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
2b2d6d01
TT		 1579 printk(KERN_WARNING
1580 "EXT4-fs warning: checktime reached, "
1581 "running e2fsck is recommended\n");
0390131b
FM		 1582 if (!sbi->s_journal)
1583 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec 1584 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 1585 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 1586 le16_add_cpu(&es->s_mnt_count, 1);
ac27a0ec 1587 es->s_mtime = cpu_to_le32(get_seconds());
617ba13b 1588 ext4_update_dynamic_rev(sb);
0390131b
FM		 1589 if (sbi->s_journal)
1590 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 1591
617ba13b 1592 ext4_commit_super(sb, es, 1);
ac27a0ec 1593 if (test_opt(sb, DEBUG))
a9df9a49 1594 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
ac27a0ec
DK		 1595 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
1596 sb->s_blocksize,
1597 sbi->s_groups_count,
617ba13b
MC		 1598 EXT4_BLOCKS_PER_GROUP(sb),
1599 EXT4_INODES_PER_GROUP(sb),
ac27a0ec
DK		 1600 sbi->s_mount_opt);
1601
0390131b
FM		 1602 if (EXT4_SB(sb)->s_journal) {
1603 printk(KERN_INFO "EXT4 FS on %s, %s journal on %s\n",
1604 sb->s_id, EXT4_SB(sb)->s_journal->j_inode ? "internal" :
1605 "external", EXT4_SB(sb)->s_journal->j_devname);
1606 } else {
1607 printk(KERN_INFO "EXT4 FS on %s, no journal\n", sb->s_id);
1608 }
ac27a0ec
DK		 1609 return res;
1610}
1611
772cb7c8
JS		 1612static int ext4_fill_flex_info(struct super_block *sb)
1613{
1614 struct ext4_sb_info *sbi = EXT4_SB(sb);
1615 struct ext4_group_desc *gdp = NULL;
1616 struct buffer_head *bh;
1617 ext4_group_t flex_group_count;
1618 ext4_group_t flex_group;
1619 int groups_per_flex = 0;
772cb7c8
JS		 1620 int i;
1621
1622 if (!sbi->s_es->s_log_groups_per_flex) {
1623 sbi->s_log_groups_per_flex = 0;
1624 return 1;
1625 }
1626
1627 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1628 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1629
c62a11fd
FB		 1630 /* We allocate both existing and potentially added groups */
1631 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
d94e99a6
AK		 1632 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1633 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
ec05e868 1634 sbi->s_flex_groups = kzalloc(flex_group_count *
772cb7c8
JS		 1635 sizeof(struct flex_groups), GFP_KERNEL);
1636 if (sbi->s_flex_groups == NULL) {
ec05e868 1637 printk(KERN_ERR "EXT4-fs: not enough memory for "
a9df9a49 1638 "%u flex groups\n", flex_group_count);
772cb7c8
JS		 1639 goto failed;
1640 }
772cb7c8 1641
772cb7c8
JS		 1642 for (i = 0; i < sbi->s_groups_count; i++) {
1643 gdp = ext4_get_group_desc(sb, i, &bh);
1644
1645 flex_group = ext4_flex_group(sbi, i);
1646 sbi->s_flex_groups[flex_group].free_inodes +=
560671a0 1647 ext4_free_inodes_count(sb, gdp);
772cb7c8 1648 sbi->s_flex_groups[flex_group].free_blocks +=
560671a0 1649 ext4_free_blks_count(sb, gdp);
772cb7c8
JS		 1650 }
1651
1652 return 1;
1653failed:
1654 return 0;
1655}
1656
717d50e4
AD		 1657__le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1658 struct ext4_group_desc *gdp)
1659{
1660 __u16 crc = 0;
1661
1662 if (sbi->s_es->s_feature_ro_compat &
1663 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1664 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1665 __le32 le_group = cpu_to_le32(block_group);
1666
1667 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1668 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1669 crc = crc16(crc, (__u8 *)gdp, offset);
1670 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1671 /* for checksum of struct ext4_group_desc do the rest...*/
1672 if ((sbi->s_es->s_feature_incompat &
1673 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1674 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1675 crc = crc16(crc, (__u8 *)gdp + offset,
1676 le16_to_cpu(sbi->s_es->s_desc_size) -
1677 offset);
1678 }
1679
1680 return cpu_to_le16(crc);
1681}
1682
1683int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1684 struct ext4_group_desc *gdp)
1685{
1686 if ((sbi->s_es->s_feature_ro_compat &
1687 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1688 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1689 return 0;
1690
1691 return 1;
1692}
1693
ac27a0ec 1694/* Called at mount-time, super-block is locked */
197cd65a 1695static int ext4_check_descriptors(struct super_block *sb)
ac27a0ec 1696{
617ba13b
MC		 1697 struct ext4_sb_info *sbi = EXT4_SB(sb);
1698 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1699 ext4_fsblk_t last_block;
bd81d8ee
LV		 1700 ext4_fsblk_t block_bitmap;
1701 ext4_fsblk_t inode_bitmap;
1702 ext4_fsblk_t inode_table;
ce421581 1703 int flexbg_flag = 0;
fd2d4291 1704 ext4_group_t i;
ac27a0ec 1705
ce421581
JS		 1706 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1707 flexbg_flag = 1;
1708
af5bc92d 1709 ext4_debug("Checking group descriptors");
ac27a0ec 1710
197cd65a
AM		 1711 for (i = 0; i < sbi->s_groups_count; i++) {
1712 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1713
ce421581 1714 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 1715 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 1716 else
1717 last_block = first_block +
617ba13b 1718 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 1719
8fadc143 1720 block_bitmap = ext4_block_bitmap(sb, gdp);
2b2d6d01 1721 if (block_bitmap < first_block || block_bitmap > last_block) {
c19204b0 1722 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
a9df9a49 1723 "Block bitmap for group %u not in group "
5128273a 1724 "(block %llu)!\n", i, block_bitmap);
ac27a0ec
DK		 1725 return 0;
1726 }
8fadc143 1727 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2b2d6d01 1728 if (inode_bitmap < first_block || inode_bitmap > last_block) {
c19204b0 1729 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
a9df9a49 1730 "Inode bitmap for group %u not in group "
5128273a 1731 "(block %llu)!\n", i, inode_bitmap);
ac27a0ec
DK		 1732 return 0;
1733 }
8fadc143 1734 inode_table = ext4_inode_table(sb, gdp);
bd81d8ee 1735 if (inode_table < first_block ||
2b2d6d01 1736 inode_table + sbi->s_itb_per_group - 1 > last_block) {
c19204b0 1737 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
a9df9a49 1738 "Inode table for group %u not in group "
5128273a 1739 "(block %llu)!\n", i, inode_table);
ac27a0ec
DK		 1740 return 0;
1741 }
b5f10eed 1742 spin_lock(sb_bgl_lock(sbi, i));
717d50e4 1743 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
c19204b0 1744 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
a9df9a49 1745 "Checksum for group %u failed (%u!=%u)\n",
c19204b0
JB		 1746 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1747 gdp)), le16_to_cpu(gdp->bg_checksum));
7ee1ec4c
LZ		 1748 if (!(sb->s_flags & MS_RDONLY)) {
1749 spin_unlock(sb_bgl_lock(sbi, i));
8a266467 1750 return 0;
7ee1ec4c 1751 }
717d50e4 1752 }
b5f10eed 1753 spin_unlock(sb_bgl_lock(sbi, i));
ce421581
JS		 1754 if (!flexbg_flag)
1755 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec
DK		 1756 }
1757
bd81d8ee 1758 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
2b2d6d01 1759 sbi->s_es->s_free_inodes_count = cpu_to_le32(ext4_count_free_inodes(sb));
ac27a0ec
DK		 1760 return 1;
1761}
1762
617ba13b 1763/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 1764 * the superblock) which were deleted from all directories, but held open by
1765 * a process at the time of a crash. We walk the list and try to delete these
1766 * inodes at recovery time (only with a read-write filesystem).
1767 *
1768 * In order to keep the orphan inode chain consistent during traversal (in
1769 * case of crash during recovery), we link each inode into the superblock
1770 * orphan list_head and handle it the same way as an inode deletion during
1771 * normal operation (which journals the operations for us).
1772 *
1773 * We only do an iget() and an iput() on each inode, which is very safe if we
1774 * accidentally point at an in-use or already deleted inode. The worst that
1775 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 1776 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 1777 * e2fsck was run on this filesystem, and it must have already done the orphan
1778 * inode cleanup for us, so we can safely abort without any further action.
1779 */
2b2d6d01
TT		 1780static void ext4_orphan_cleanup(struct super_block *sb,
1781 struct ext4_super_block *es)
ac27a0ec
DK		 1782{
1783 unsigned int s_flags = sb->s_flags;
1784 int nr_orphans = 0, nr_truncates = 0;
1785#ifdef CONFIG_QUOTA
1786 int i;
1787#endif
1788 if (!es->s_last_orphan) {
1789 jbd_debug(4, "no orphan inodes to clean up\n");
1790 return;
1791 }
1792
a8f48a95
ES		 1793 if (bdev_read_only(sb->s_bdev)) {
1794 printk(KERN_ERR "EXT4-fs: write access "
1795 "unavailable, skipping orphan cleanup.\n");
1796 return;
1797 }
1798
617ba13b 1799 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
ac27a0ec
DK		 1800 if (es->s_last_orphan)
1801 jbd_debug(1, "Errors on filesystem, "
1802 "clearing orphan list.\n");
1803 es->s_last_orphan = 0;
1804 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1805 return;
1806 }
1807
1808 if (s_flags & MS_RDONLY) {
617ba13b 1809 printk(KERN_INFO "EXT4-fs: %s: orphan cleanup on readonly fs\n",
ac27a0ec
DK		 1810 sb->s_id);
1811 sb->s_flags &= ~MS_RDONLY;
1812 }
1813#ifdef CONFIG_QUOTA
1814 /* Needed for iput() to work correctly and not trash data */
1815 sb->s_flags |= MS_ACTIVE;
1816 /* Turn on quotas so that they are updated correctly */
1817 for (i = 0; i < MAXQUOTAS; i++) {
617ba13b
MC		 1818 if (EXT4_SB(sb)->s_qf_names[i]) {
1819 int ret = ext4_quota_on_mount(sb, i);
ac27a0ec
DK		 1820 if (ret < 0)
1821 printk(KERN_ERR
2c8be6b2 1822 "EXT4-fs: Cannot turn on journaled "
ac27a0ec
DK		 1823 "quota: error %d\n", ret);
1824 }
1825 }
1826#endif
1827
1828 while (es->s_last_orphan) {
1829 struct inode *inode;
1830
97bd42b9
JB		 1831 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1832 if (IS_ERR(inode)) {
ac27a0ec
DK		 1833 es->s_last_orphan = 0;
1834 break;
1835 }
1836
617ba13b 1837 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
ac27a0ec
DK		 1838 DQUOT_INIT(inode);
1839 if (inode->i_nlink) {
1840 printk(KERN_DEBUG
e5f8eab8 1841 "%s: truncating inode %lu to %lld bytes\n",
46e665e9 1842 __func__, inode->i_ino, inode->i_size);
e5f8eab8 1843 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
ac27a0ec 1844 inode->i_ino, inode->i_size);
617ba13b 1845 ext4_truncate(inode);
ac27a0ec
DK		 1846 nr_truncates++;
1847 } else {
1848 printk(KERN_DEBUG
1849 "%s: deleting unreferenced inode %lu\n",
46e665e9 1850 __func__, inode->i_ino);
ac27a0ec
DK		 1851 jbd_debug(2, "deleting unreferenced inode %lu\n",
1852 inode->i_ino);
1853 nr_orphans++;
1854 }
1855 iput(inode); /* The delete magic happens here! */
1856 }
1857
2b2d6d01 1858#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
ac27a0ec
DK		 1859
1860 if (nr_orphans)
617ba13b 1861 printk(KERN_INFO "EXT4-fs: %s: %d orphan inode%s deleted\n",
ac27a0ec
DK		 1862 sb->s_id, PLURAL(nr_orphans));
1863 if (nr_truncates)
617ba13b 1864 printk(KERN_INFO "EXT4-fs: %s: %d truncate%s cleaned up\n",
ac27a0ec
DK		 1865 sb->s_id, PLURAL(nr_truncates));
1866#ifdef CONFIG_QUOTA
1867 /* Turn quotas off */
1868 for (i = 0; i < MAXQUOTAS; i++) {
1869 if (sb_dqopt(sb)->files[i])
6f28e087 1870 vfs_quota_off(sb, i, 0);
ac27a0ec
DK		 1871 }
1872#endif
1873 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1874}
cd2291a4
ES		 1875/*
1876 * Maximal extent format file size.
1877 * Resulting logical blkno at s_maxbytes must fit in our on-disk
1878 * extent format containers, within a sector_t, and within i_blocks
1879 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
1880 * so that won't be a limiting factor.
1881 *
1882 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1883 */
f287a1a5 1884static loff_t ext4_max_size(int blkbits, int has_huge_files)
cd2291a4
ES		 1885{
1886 loff_t res;
1887 loff_t upper_limit = MAX_LFS_FILESIZE;
1888
1889 /* small i_blocks in vfs inode? */
f287a1a5 1890 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
cd2291a4 1891 /*
b3a6ffe1 1892 * CONFIG_LBD is not enabled implies the inode
cd2291a4
ES		 1893 * i_block represent total blocks in 512 bytes
1894 * 32 == size of vfs inode i_blocks * 8
1895 */
1896 upper_limit = (1LL << 32) - 1;
1897
1898 /* total blocks in file system block size */
1899 upper_limit >>= (blkbits - 9);
1900 upper_limit <<= blkbits;
1901 }
1902
1903 /* 32-bit extent-start container, ee_block */
1904 res = 1LL << 32;
1905 res <<= blkbits;
1906 res -= 1;
1907
1908 /* Sanity check against vm- & vfs- imposed limits */
1909 if (res > upper_limit)
1910 res = upper_limit;
1911
1912 return res;
1913}
ac27a0ec 1914
ac27a0ec 1915/*
cd2291a4 1916 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 1917 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
1918 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 1919 */
f287a1a5 1920static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
ac27a0ec 1921{
617ba13b 1922 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 1923 int meta_blocks;
1924 loff_t upper_limit;
1925 /* This is calculated to be the largest file size for a
cd2291a4 1926 * dense, bitmapped file such that the total number of
ac27a0ec 1927 * sectors in the file, including data and all indirect blocks,
0fc1b451
AK		 1928 * does not exceed 2^48 -1
1929 * __u32 i_blocks_lo and _u16 i_blocks_high representing the
1930 * total number of 512 bytes blocks of the file
1931 */
1932
f287a1a5 1933 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
0fc1b451 1934 /*
b3a6ffe1 1935 * !has_huge_files or CONFIG_LBD is not enabled
f287a1a5
TT		 1936 * implies the inode i_block represent total blocks in
1937 * 512 bytes 32 == size of vfs inode i_blocks * 8
0fc1b451
AK		 1938 */
1939 upper_limit = (1LL << 32) - 1;
1940
1941 /* total blocks in file system block size */
1942 upper_limit >>= (bits - 9);
1943
1944 } else {
8180a562
AK		 1945 /*
1946 * We use 48 bit ext4_inode i_blocks
1947 * With EXT4_HUGE_FILE_FL set the i_blocks
1948 * represent total number of blocks in
1949 * file system block size
1950 */
0fc1b451
AK		 1951 upper_limit = (1LL << 48) - 1;
1952
0fc1b451
AK		 1953 }
1954
1955 /* indirect blocks */
1956 meta_blocks = 1;
1957 /* double indirect blocks */
1958 meta_blocks += 1 + (1LL << (bits-2));
1959 /* tripple indirect blocks */
1960 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
1961
1962 upper_limit -= meta_blocks;
1963 upper_limit <<= bits;
ac27a0ec
DK		 1964
1965 res += 1LL << (bits-2);
1966 res += 1LL << (2*(bits-2));
1967 res += 1LL << (3*(bits-2));
1968 res <<= bits;
1969 if (res > upper_limit)
1970 res = upper_limit;
0fc1b451
AK		 1971
1972 if (res > MAX_LFS_FILESIZE)
1973 res = MAX_LFS_FILESIZE;
1974
ac27a0ec
DK		 1975 return res;
1976}
1977
617ba13b 1978static ext4_fsblk_t descriptor_loc(struct super_block *sb,
70bbb3e0 1979 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 1980{
617ba13b 1981 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 1982 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 1983 int has_super = 0;
1984
1985 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1986
617ba13b 1987 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
ac27a0ec 1988 nr < first_meta_bg)
70bbb3e0 1989 return logical_sb_block + nr + 1;
ac27a0ec 1990 bg = sbi->s_desc_per_block * nr;
617ba13b 1991 if (ext4_bg_has_super(sb, bg))
ac27a0ec 1992 has_super = 1;
617ba13b 1993 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 1994}
1995
c9de560d
AT		 1996/**
1997 * ext4_get_stripe_size: Get the stripe size.
1998 * @sbi: In memory super block info
1999 *
2000 * If we have specified it via mount option, then
2001 * use the mount option value. If the value specified at mount time is
2002 * greater than the blocks per group use the super block value.
2003 * If the super block value is greater than blocks per group return 0.
2004 * Allocator needs it be less than blocks per group.
2005 *
2006 */
2007static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2008{
2009 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2010 unsigned long stripe_width =
2011 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2012
2013 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2014 return sbi->s_stripe;
2015
2016 if (stripe_width <= sbi->s_blocks_per_group)
2017 return stripe_width;
2018
2019 if (stride <= sbi->s_blocks_per_group)
2020 return stride;
2021
2022 return 0;
2023}
ac27a0ec 2024
2b2d6d01 2025static int ext4_fill_super(struct super_block *sb, void *data, int silent)
7477827f
AK		 2026 __releases(kernel_lock)
2027 __acquires(kernel_lock)
1d03ec98 2028
ac27a0ec 2029{
2b2d6d01 2030 struct buffer_head *bh;
617ba13b
MC		 2031 struct ext4_super_block *es = NULL;
2032 struct ext4_sb_info *sbi;
2033 ext4_fsblk_t block;
2034 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 2035 ext4_fsblk_t logical_sb_block;
ac27a0ec 2036 unsigned long offset = 0;
ac27a0ec
DK		 2037 unsigned long journal_devnum = 0;
2038 unsigned long def_mount_opts;
2039 struct inode *root;
9f6200bb 2040 char *cp;
0390131b 2041 const char *descr;
1d1fe1ee 2042 int ret = -EINVAL;
ac27a0ec 2043 int blocksize;
ac27a0ec
DK		 2044 int db_count;
2045 int i;
f287a1a5 2046 int needs_recovery, has_huge_files;
3a06d778 2047 int features;
bd81d8ee 2048 __u64 blocks_count;
833f4077 2049 int err;
b3881f74 2050 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
ac27a0ec
DK		 2051
2052 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
2053 if (!sbi)
2054 return -ENOMEM;
2055 sb->s_fs_info = sbi;
2056 sbi->s_mount_opt = 0;
617ba13b
MC		 2057 sbi->s_resuid = EXT4_DEF_RESUID;
2058 sbi->s_resgid = EXT4_DEF_RESGID;
240799cd 2059 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
d9c9bef1 2060 sbi->s_sb_block = sb_block;
ac27a0ec
DK		 2061
2062 unlock_kernel();
2063
9f6200bb
TT		 2064 /* Cleanup superblock name */
2065 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
2066 *cp = '!';
2067
617ba13b 2068 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 2069 if (!blocksize) {
617ba13b 2070 printk(KERN_ERR "EXT4-fs: unable to set blocksize\n");
ac27a0ec
DK		 2071 goto out_fail;
2072 }
2073
2074 /*
617ba13b 2075 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 2076 * block sizes. We need to calculate the offset from buffer start.
2077 */
617ba13b 2078 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 2079 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2080 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 2081 } else {
70bbb3e0 2082 logical_sb_block = sb_block;
ac27a0ec
DK		 2083 }
2084
70bbb3e0 2085 if (!(bh = sb_bread(sb, logical_sb_block))) {
2b2d6d01 2086 printk(KERN_ERR "EXT4-fs: unable to read superblock\n");
ac27a0ec
DK		 2087 goto out_fail;
2088 }
2089 /*
2090 * Note: s_es must be initialized as soon as possible because
617ba13b 2091 * some ext4 macro-instructions depend on its value
ac27a0ec 2092 */
617ba13b 2093 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
ac27a0ec
DK		 2094 sbi->s_es = es;
2095 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 2096 if (sb->s_magic != EXT4_SUPER_MAGIC)
2097 goto cantfind_ext4;
ac27a0ec
DK		 2098
2099 /* Set defaults before we parse the mount options */
2100 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
617ba13b 2101 if (def_mount_opts & EXT4_DEFM_DEBUG)
ac27a0ec 2102 set_opt(sbi->s_mount_opt, DEBUG);
617ba13b 2103 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
ac27a0ec 2104 set_opt(sbi->s_mount_opt, GRPID);
617ba13b 2105 if (def_mount_opts & EXT4_DEFM_UID16)
ac27a0ec 2106 set_opt(sbi->s_mount_opt, NO_UID32);
03010a33 2107#ifdef CONFIG_EXT4_FS_XATTR
617ba13b 2108 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
ac27a0ec 2109 set_opt(sbi->s_mount_opt, XATTR_USER);
2e7842b8 2110#endif
03010a33 2111#ifdef CONFIG_EXT4_FS_POSIX_ACL
617ba13b 2112 if (def_mount_opts & EXT4_DEFM_ACL)
ac27a0ec 2113 set_opt(sbi->s_mount_opt, POSIX_ACL);
2e7842b8 2114#endif
617ba13b
MC		 2115 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
2116 sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
2117 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
2118 sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
2119 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
2120 sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
2121
2122 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
ac27a0ec 2123 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
bb4f397a 2124 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
ceea16bf 2125 set_opt(sbi->s_mount_opt, ERRORS_CONT);
bb4f397a
AK		 2126 else
2127 set_opt(sbi->s_mount_opt, ERRORS_RO);
ac27a0ec
DK		 2128
2129 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
2130 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
30773840
TT		 2131 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
2132 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
2133 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
ac27a0ec
DK		 2134
2135 set_opt(sbi->s_mount_opt, RESERVATION);
571640ca 2136 set_opt(sbi->s_mount_opt, BARRIER);
ac27a0ec 2137
1e2462f9
MC		 2138 /*
2139 * turn on extents feature by default in ext4 filesystem
e4079a11
ES		 2140 * only if feature flag already set by mkfs or tune2fs.
2141 * Use -o noextents to turn it off
1e2462f9 2142 */
e4079a11
ES		 2143 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS))
2144 set_opt(sbi->s_mount_opt, EXTENTS);
2145 else
2146 ext4_warning(sb, __func__,
2147 "extents feature not enabled on this filesystem, "
fde4d95a 2148 "use tune2fs.");
1e2462f9 2149
dd919b98
AK		 2150 /*
2151 * enable delayed allocation by default
2152 * Use -o nodelalloc to turn it off
2153 */
2154 set_opt(sbi->s_mount_opt, DELALLOC);
2155
2156
b3881f74
TT		 2157 if (!parse_options((char *) data, sb, &journal_devnum,
2158 &journal_ioprio, NULL, 0))
ac27a0ec
DK		 2159 goto failed_mount;
2160
2161 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
617ba13b 2162 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec 2163
617ba13b
MC		 2164 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
2165 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
2166 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
2167 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
ac27a0ec 2168 printk(KERN_WARNING
617ba13b 2169 "EXT4-fs warning: feature flags set on rev 0 fs, "
ac27a0ec 2170 "running e2fsck is recommended\n");
469108ff 2171
ac27a0ec
DK		 2172 /*
2173 * Check feature flags regardless of the revision level, since we
2174 * previously didn't change the revision level when setting the flags,
2175 * so there is a chance incompat flags are set on a rev 0 filesystem.
2176 */
617ba13b 2177 features = EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP);
ac27a0ec 2178 if (features) {
617ba13b 2179 printk(KERN_ERR "EXT4-fs: %s: couldn't mount because of "
3a06d778
AK		 2180 "unsupported optional features (%x).\n", sb->s_id,
2181 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2182 ~EXT4_FEATURE_INCOMPAT_SUPP));
ac27a0ec
DK		 2183 goto failed_mount;
2184 }
617ba13b 2185 features = EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP);
ac27a0ec 2186 if (!(sb->s_flags & MS_RDONLY) && features) {
617ba13b 2187 printk(KERN_ERR "EXT4-fs: %s: couldn't mount RDWR because of "
3a06d778
AK		 2188 "unsupported optional features (%x).\n", sb->s_id,
2189 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2190 ~EXT4_FEATURE_RO_COMPAT_SUPP));
ac27a0ec
DK		 2191 goto failed_mount;
2192 }
f287a1a5
TT		 2193 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2194 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
2195 if (has_huge_files) {
0fc1b451
AK		 2196 /*
2197 * Large file size enabled file system can only be
b3a6ffe1 2198 * mount if kernel is build with CONFIG_LBD
0fc1b451
AK		 2199 */
2200 if (sizeof(root->i_blocks) < sizeof(u64) &&
2201 !(sb->s_flags & MS_RDONLY)) {
2202 printk(KERN_ERR "EXT4-fs: %s: Filesystem with huge "
2203 "files cannot be mounted read-write "
b3a6ffe1 2204 "without CONFIG_LBD.\n", sb->s_id);
0fc1b451
AK		 2205 goto failed_mount;
2206 }
2207 }
ac27a0ec
DK		 2208 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
2209
617ba13b
MC		 2210 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
2211 blocksize > EXT4_MAX_BLOCK_SIZE) {
ac27a0ec 2212 printk(KERN_ERR
617ba13b 2213 "EXT4-fs: Unsupported filesystem blocksize %d on %s.\n",
ac27a0ec
DK		 2214 blocksize, sb->s_id);
2215 goto failed_mount;
2216 }
2217
ac27a0ec 2218 if (sb->s_blocksize != blocksize) {
ce40733c
AK		 2219
2220 /* Validate the filesystem blocksize */
2221 if (!sb_set_blocksize(sb, blocksize)) {
2222 printk(KERN_ERR "EXT4-fs: bad block size %d.\n",
2223 blocksize);
ac27a0ec
DK		 2224 goto failed_mount;
2225 }
2226
2b2d6d01 2227 brelse(bh);
70bbb3e0
AM		 2228 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2229 offset = do_div(logical_sb_block, blocksize);
2230 bh = sb_bread(sb, logical_sb_block);
ac27a0ec
DK		 2231 if (!bh) {
2232 printk(KERN_ERR
617ba13b 2233 "EXT4-fs: Can't read superblock on 2nd try.\n");
ac27a0ec
DK		 2234 goto failed_mount;
2235 }
617ba13b 2236 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
ac27a0ec 2237 sbi->s_es = es;
617ba13b 2238 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
2b2d6d01
TT		 2239 printk(KERN_ERR
2240 "EXT4-fs: Magic mismatch, very weird !\n");
ac27a0ec
DK		 2241 goto failed_mount;
2242 }
2243 }
2244
f287a1a5
TT		 2245 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
2246 has_huge_files);
2247 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
ac27a0ec 2248
617ba13b
MC		 2249 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2250 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2251 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
ac27a0ec
DK		 2252 } else {
2253 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2254 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
617ba13b 2255 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1330593e 2256 (!is_power_of_2(sbi->s_inode_size)) ||
ac27a0ec 2257 (sbi->s_inode_size > blocksize)) {
2b2d6d01
TT		 2258 printk(KERN_ERR
2259 "EXT4-fs: unsupported inode size: %d\n",
2260 sbi->s_inode_size);
ac27a0ec
DK		 2261 goto failed_mount;
2262 }
ef7f3835
KS		 2263 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2264 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
ac27a0ec 2265 }
0d1ee42f
AR		 2266 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2267 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
8fadc143 2268 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 2269 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 2270 !is_power_of_2(sbi->s_desc_size)) {
0d1ee42f 2271 printk(KERN_ERR
8fadc143 2272 "EXT4-fs: unsupported descriptor size %lu\n",
0d1ee42f
AR		 2273 sbi->s_desc_size);
2274 goto failed_mount;
2275 }
2276 } else
2277 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
ac27a0ec 2278 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 2279 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
b47b6f38 2280 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
617ba13b
MC		 2281 goto cantfind_ext4;
2282 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 2283 if (sbi->s_inodes_per_block == 0)
617ba13b 2284 goto cantfind_ext4;
ac27a0ec
DK		 2285 sbi->s_itb_per_group = sbi->s_inodes_per_group /
2286 sbi->s_inodes_per_block;
0d1ee42f 2287 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 2288 sbi->s_sbh = bh;
2289 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 2290 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2291 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
2b2d6d01 2292 for (i = 0; i < 4; i++)
ac27a0ec
DK		 2293 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2294 sbi->s_def_hash_version = es->s_def_hash_version;
f99b2589
TT		 2295 i = le32_to_cpu(es->s_flags);
2296 if (i & EXT2_FLAGS_UNSIGNED_HASH)
2297 sbi->s_hash_unsigned = 3;
2298 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
2299#ifdef __CHAR_UNSIGNED__
2300 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
2301 sbi->s_hash_unsigned = 3;
2302#else
2303 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
2304#endif
2305 sb->s_dirt = 1;
2306 }
ac27a0ec
DK		 2307
2308 if (sbi->s_blocks_per_group > blocksize * 8) {
2b2d6d01
TT		 2309 printk(KERN_ERR
2310 "EXT4-fs: #blocks per group too big: %lu\n",
2311 sbi->s_blocks_per_group);
ac27a0ec
DK		 2312 goto failed_mount;
2313 }
ac27a0ec 2314 if (sbi->s_inodes_per_group > blocksize * 8) {
2b2d6d01
TT		 2315 printk(KERN_ERR
2316 "EXT4-fs: #inodes per group too big: %lu\n",
2317 sbi->s_inodes_per_group);
ac27a0ec
DK		 2318 goto failed_mount;
2319 }
2320
bd81d8ee 2321 if (ext4_blocks_count(es) >
ac27a0ec 2322 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
617ba13b 2323 printk(KERN_ERR "EXT4-fs: filesystem on %s:"
ac27a0ec
DK		 2324 " too large to mount safely\n", sb->s_id);
2325 if (sizeof(sector_t) < 8)
617ba13b 2326 printk(KERN_WARNING "EXT4-fs: CONFIG_LBD not "
ac27a0ec
DK		 2327 "enabled\n");
2328 goto failed_mount;
2329 }
2330
617ba13b
MC		 2331 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2332 goto cantfind_ext4;
e7c95593
ES		 2333
2334 /* ensure blocks_count calculation below doesn't sign-extend */
2335 if (ext4_blocks_count(es) + EXT4_BLOCKS_PER_GROUP(sb) <
2336 le32_to_cpu(es->s_first_data_block) + 1) {
2337 printk(KERN_WARNING "EXT4-fs: bad geometry: block count %llu, "
2338 "first data block %u, blocks per group %lu\n",
2339 ext4_blocks_count(es),
2340 le32_to_cpu(es->s_first_data_block),
2341 EXT4_BLOCKS_PER_GROUP(sb));
2342 goto failed_mount;
2343 }
bd81d8ee
LV		 2344 blocks_count = (ext4_blocks_count(es) -
2345 le32_to_cpu(es->s_first_data_block) +
2346 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2347 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
2348 sbi->s_groups_count = blocks_count;
617ba13b
MC		 2349 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2350 EXT4_DESC_PER_BLOCK(sb);
2b2d6d01 2351 sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
ac27a0ec
DK		 2352 GFP_KERNEL);
2353 if (sbi->s_group_desc == NULL) {
2b2d6d01 2354 printk(KERN_ERR "EXT4-fs: not enough memory\n");
ac27a0ec
DK		 2355 goto failed_mount;
2356 }
2357
3244fcb1 2358#ifdef CONFIG_PROC_FS
9f6200bb
TT		 2359 if (ext4_proc_root)
2360 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
2361
240799cd
TT		 2362 if (sbi->s_proc)
2363 proc_create_data("inode_readahead_blks", 0644, sbi->s_proc,
2364 &ext4_ui_proc_fops,
2365 &sbi->s_inode_readahead_blks);
3244fcb1 2366#endif
240799cd 2367
ac27a0ec
DK		 2368 bgl_lock_init(&sbi->s_blockgroup_lock);
2369
2370 for (i = 0; i < db_count; i++) {
70bbb3e0 2371 block = descriptor_loc(sb, logical_sb_block, i);
ac27a0ec
DK		 2372 sbi->s_group_desc[i] = sb_bread(sb, block);
2373 if (!sbi->s_group_desc[i]) {
2b2d6d01
TT		 2374 printk(KERN_ERR "EXT4-fs: "
2375 "can't read group descriptor %d\n", i);
ac27a0ec
DK		 2376 db_count = i;
2377 goto failed_mount2;
2378 }
2379 }
2b2d6d01 2380 if (!ext4_check_descriptors(sb)) {
617ba13b 2381 printk(KERN_ERR "EXT4-fs: group descriptors corrupted!\n");
ac27a0ec
DK		 2382 goto failed_mount2;
2383 }
772cb7c8
JS		 2384 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2385 if (!ext4_fill_flex_info(sb)) {
2386 printk(KERN_ERR
2387 "EXT4-fs: unable to initialize "
2388 "flex_bg meta info!\n");
2389 goto failed_mount2;
2390 }
2391
ac27a0ec
DK		 2392 sbi->s_gdb_count = db_count;
2393 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2394 spin_lock_init(&sbi->s_next_gen_lock);
2395
833f4077
PZ		 2396 err = percpu_counter_init(&sbi->s_freeblocks_counter,
2397 ext4_count_free_blocks(sb));
2398 if (!err) {
2399 err = percpu_counter_init(&sbi->s_freeinodes_counter,
2400 ext4_count_free_inodes(sb));
2401 }
2402 if (!err) {
2403 err = percpu_counter_init(&sbi->s_dirs_counter,
2404 ext4_count_dirs(sb));
2405 }
6bc6e63f
AK		 2406 if (!err) {
2407 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
2408 }
833f4077
PZ		 2409 if (err) {
2410 printk(KERN_ERR "EXT4-fs: insufficient memory\n");
2411 goto failed_mount3;
2412 }
ac27a0ec 2413
c9de560d
AT		 2414 sbi->s_stripe = ext4_get_stripe_size(sbi);
2415
ac27a0ec
DK		 2416 /*
2417 * set up enough so that it can read an inode
2418 */
617ba13b
MC		 2419 sb->s_op = &ext4_sops;
2420 sb->s_export_op = &ext4_export_ops;
2421 sb->s_xattr = ext4_xattr_handlers;
ac27a0ec 2422#ifdef CONFIG_QUOTA
617ba13b
MC		 2423 sb->s_qcop = &ext4_qctl_operations;
2424 sb->dq_op = &ext4_quota_operations;
ac27a0ec
DK		 2425#endif
2426 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
2427
2428 sb->s_root = NULL;
2429
2430 needs_recovery = (es->s_last_orphan != 0 ||
617ba13b
MC		 2431 EXT4_HAS_INCOMPAT_FEATURE(sb,
2432 EXT4_FEATURE_INCOMPAT_RECOVER));
ac27a0ec
DK		 2433
2434 /*
2435 * The first inode we look at is the journal inode. Don't try
2436 * root first: it may be modified in the journal!
2437 */
2438 if (!test_opt(sb, NOLOAD) &&
617ba13b
MC		 2439 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2440 if (ext4_load_journal(sb, es, journal_devnum))
ac27a0ec 2441 goto failed_mount3;
624080ed
TT		 2442 if (!(sb->s_flags & MS_RDONLY) &&
2443 EXT4_SB(sb)->s_journal->j_failed_commit) {
2444 printk(KERN_CRIT "EXT4-fs error (device %s): "
2445 "ext4_fill_super: Journal transaction "
2b2d6d01 2446 "%u is corrupt\n", sb->s_id,
624080ed 2447 EXT4_SB(sb)->s_journal->j_failed_commit);
2b2d6d01
TT		 2448 if (test_opt(sb, ERRORS_RO)) {
2449 printk(KERN_CRIT
2450 "Mounting filesystem read-only\n");
624080ed
TT		 2451 sb->s_flags |= MS_RDONLY;
2452 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2453 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2454 }
2455 if (test_opt(sb, ERRORS_PANIC)) {
2456 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2457 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2458 ext4_commit_super(sb, es, 1);
624080ed
TT		 2459 goto failed_mount4;
2460 }
2461 }
0390131b
FM		 2462 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
2463 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
2464 printk(KERN_ERR "EXT4-fs: required journal recovery "
2465 "suppressed and not mounted read-only\n");
2466 goto failed_mount4;
ac27a0ec 2467 } else {
0390131b
FM		 2468 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
2469 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
2470 sbi->s_journal = NULL;
2471 needs_recovery = 0;
2472 goto no_journal;
ac27a0ec
DK		 2473 }
2474
eb40a09c
JS		 2475 if (ext4_blocks_count(es) > 0xffffffffULL &&
2476 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2477 JBD2_FEATURE_INCOMPAT_64BIT)) {
2478 printk(KERN_ERR "ext4: Failed to set 64-bit journal feature\n");
2479 goto failed_mount4;
2480 }
2481
818d276c
GS		 2482 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2483 jbd2_journal_set_features(sbi->s_journal,
2484 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2485 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2486 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2487 jbd2_journal_set_features(sbi->s_journal,
2488 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2489 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2490 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2491 } else {
2492 jbd2_journal_clear_features(sbi->s_journal,
2493 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2494 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2495 }
2496
ac27a0ec
DK		 2497 /* We have now updated the journal if required, so we can
2498 * validate the data journaling mode. */
2499 switch (test_opt(sb, DATA_FLAGS)) {
2500 case 0:
2501 /* No mode set, assume a default based on the journal
63f57933
AM		 2502 * capabilities: ORDERED_DATA if the journal can
2503 * cope, else JOURNAL_DATA
2504 */
dab291af
MC		 2505 if (jbd2_journal_check_available_features
2506 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
ac27a0ec
DK		 2507 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2508 else
2509 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2510 break;
2511
617ba13b
MC		 2512 case EXT4_MOUNT_ORDERED_DATA:
2513 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 2514 if (!jbd2_journal_check_available_features
2515 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
617ba13b 2516 printk(KERN_ERR "EXT4-fs: Journal does not support "
ac27a0ec
DK		 2517 "requested data journaling mode\n");
2518 goto failed_mount4;
2519 }
2520 default:
2521 break;
2522 }
b3881f74 2523 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
ac27a0ec 2524
0390131b
FM		 2525no_journal:
2526
ac27a0ec 2527 if (test_opt(sb, NOBH)) {
617ba13b
MC		 2528 if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
2529 printk(KERN_WARNING "EXT4-fs: Ignoring nobh option - "
ac27a0ec
DK		 2530 "its supported only with writeback mode\n");
2531 clear_opt(sbi->s_mount_opt, NOBH);
2532 }
2533 }
2534 /*
dab291af 2535 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 2536 * so we can safely mount the rest of the filesystem now.
2537 */
2538
1d1fe1ee
DH		 2539 root = ext4_iget(sb, EXT4_ROOT_INO);
2540 if (IS_ERR(root)) {
617ba13b 2541 printk(KERN_ERR "EXT4-fs: get root inode failed\n");
1d1fe1ee 2542 ret = PTR_ERR(root);
ac27a0ec
DK		 2543 goto failed_mount4;
2544 }
2545 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1d1fe1ee 2546 iput(root);
617ba13b 2547 printk(KERN_ERR "EXT4-fs: corrupt root inode, run e2fsck\n");
ac27a0ec
DK		 2548 goto failed_mount4;
2549 }
1d1fe1ee
DH		 2550 sb->s_root = d_alloc_root(root);
2551 if (!sb->s_root) {
2552 printk(KERN_ERR "EXT4-fs: get root dentry failed\n");
2553 iput(root);
2554 ret = -ENOMEM;
2555 goto failed_mount4;
2556 }
ac27a0ec 2557
2b2d6d01 2558 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
ef7f3835
KS		 2559
2560 /* determine the minimum size of new large inodes, if present */
2561 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2562 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2563 EXT4_GOOD_OLD_INODE_SIZE;
2564 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2565 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2566 if (sbi->s_want_extra_isize <
2567 le16_to_cpu(es->s_want_extra_isize))
2568 sbi->s_want_extra_isize =
2569 le16_to_cpu(es->s_want_extra_isize);
2570 if (sbi->s_want_extra_isize <
2571 le16_to_cpu(es->s_min_extra_isize))
2572 sbi->s_want_extra_isize =
2573 le16_to_cpu(es->s_min_extra_isize);
2574 }
2575 }
2576 /* Check if enough inode space is available */
2577 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2578 sbi->s_inode_size) {
2579 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2580 EXT4_GOOD_OLD_INODE_SIZE;
2581 printk(KERN_INFO "EXT4-fs: required extra inode space not"
2582 "available.\n");
2583 }
2584
c2774d84
AK		 2585 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
2586 printk(KERN_WARNING "EXT4-fs: Ignoring delalloc option - "
2587 "requested data journaling mode\n");
2588 clear_opt(sbi->s_mount_opt, DELALLOC);
2589 } else if (test_opt(sb, DELALLOC))
2590 printk(KERN_INFO "EXT4-fs: delayed allocation enabled\n");
2591
2592 ext4_ext_init(sb);
2593 err = ext4_mb_init(sb, needs_recovery);
2594 if (err) {
2595 printk(KERN_ERR "EXT4-fs: failed to initalize mballoc (%d)\n",
2596 err);
2597 goto failed_mount4;
2598 }
2599
ac27a0ec
DK		 2600 /*
2601 * akpm: core read_super() calls in here with the superblock locked.
2602 * That deadlocks, because orphan cleanup needs to lock the superblock
2603 * in numerous places. Here we just pop the lock - it's relatively
2604 * harmless, because we are now ready to accept write_super() requests,
2605 * and aviro says that's the only reason for hanging onto the
2606 * superblock lock.
2607 */
617ba13b
MC		 2608 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
2609 ext4_orphan_cleanup(sb, es);
2610 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
0390131b 2611 if (needs_recovery) {
2b2d6d01 2612 printk(KERN_INFO "EXT4-fs: recovery complete.\n");
0390131b
FM		 2613 ext4_mark_recovery_complete(sb, es);
2614 }
2615 if (EXT4_SB(sb)->s_journal) {
2616 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2617 descr = " journalled data mode";
2618 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2619 descr = " ordered data mode";
2620 else
2621 descr = " writeback data mode";
2622 } else
2623 descr = "out journal";
2624
2625 printk(KERN_INFO "EXT4-fs: mounted filesystem %s with%s\n",
2626 sb->s_id, descr);
ac27a0ec
DK		 2627
2628 lock_kernel();
2629 return 0;
2630
617ba13b 2631cantfind_ext4:
ac27a0ec 2632 if (!silent)
617ba13b 2633 printk(KERN_ERR "VFS: Can't find ext4 filesystem on dev %s.\n",
ac27a0ec
DK		 2634 sb->s_id);
2635 goto failed_mount;
2636
2637failed_mount4:
0390131b
FM		 2638 printk(KERN_ERR "EXT4-fs (device %s): mount failed\n", sb->s_id);
2639 if (sbi->s_journal) {
2640 jbd2_journal_destroy(sbi->s_journal);
2641 sbi->s_journal = NULL;
2642 }
ac27a0ec
DK		 2643failed_mount3:
2644 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2645 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2646 percpu_counter_destroy(&sbi->s_dirs_counter);
6bc6e63f 2647 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
ac27a0ec
DK		 2648failed_mount2:
2649 for (i = 0; i < db_count; i++)
2650 brelse(sbi->s_group_desc[i]);
2651 kfree(sbi->s_group_desc);
2652failed_mount:
240799cd
TT		 2653 if (sbi->s_proc) {
2654 remove_proc_entry("inode_readahead_blks", sbi->s_proc);
9f6200bb 2655 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 2656 }
ac27a0ec
DK		 2657#ifdef CONFIG_QUOTA
2658 for (i = 0; i < MAXQUOTAS; i++)
2659 kfree(sbi->s_qf_names[i]);
2660#endif
617ba13b 2661 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 2662 brelse(bh);
2663out_fail:
2664 sb->s_fs_info = NULL;
2665 kfree(sbi);
2666 lock_kernel();
1d1fe1ee 2667 return ret;
ac27a0ec
DK		 2668}
2669
2670/*
2671 * Setup any per-fs journal parameters now. We'll do this both on
2672 * initial mount, once the journal has been initialised but before we've
2673 * done any recovery; and again on any subsequent remount.
2674 */
617ba13b 2675static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 2676{
617ba13b 2677 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 2678
30773840
TT		 2679 journal->j_commit_interval = sbi->s_commit_interval;
2680 journal->j_min_batch_time = sbi->s_min_batch_time;
2681 journal->j_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 2682
2683 spin_lock(&journal->j_state_lock);
2684 if (test_opt(sb, BARRIER))
dab291af 2685 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 2686 else
dab291af 2687 journal->j_flags &= ~JBD2_BARRIER;
5bf5683a
HK		 2688 if (test_opt(sb, DATA_ERR_ABORT))
2689 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
2690 else
2691 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
ac27a0ec
DK		 2692 spin_unlock(&journal->j_state_lock);
2693}
2694
617ba13b 2695static journal_t *ext4_get_journal(struct super_block *sb,
ac27a0ec
DK		 2696 unsigned int journal_inum)
2697{
2698 struct inode *journal_inode;
2699 journal_t *journal;
2700
0390131b
FM		 2701 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
2702
ac27a0ec
DK		 2703 /* First, test for the existence of a valid inode on disk. Bad
2704 * things happen if we iget() an unused inode, as the subsequent
2705 * iput() will try to delete it. */
2706
1d1fe1ee
DH		 2707 journal_inode = ext4_iget(sb, journal_inum);
2708 if (IS_ERR(journal_inode)) {
617ba13b 2709 printk(KERN_ERR "EXT4-fs: no journal found.\n");
ac27a0ec
DK		 2710 return NULL;
2711 }
2712 if (!journal_inode->i_nlink) {
2713 make_bad_inode(journal_inode);
2714 iput(journal_inode);
617ba13b 2715 printk(KERN_ERR "EXT4-fs: journal inode is deleted.\n");
ac27a0ec
DK		 2716 return NULL;
2717 }
2718
e5f8eab8 2719 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
ac27a0ec 2720 journal_inode, journal_inode->i_size);
1d1fe1ee 2721 if (!S_ISREG(journal_inode->i_mode)) {
617ba13b 2722 printk(KERN_ERR "EXT4-fs: invalid journal inode.\n");
ac27a0ec
DK		 2723 iput(journal_inode);
2724 return NULL;
2725 }
2726
dab291af 2727 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 2728 if (!journal) {
617ba13b 2729 printk(KERN_ERR "EXT4-fs: Could not load journal inode\n");
ac27a0ec
DK		 2730 iput(journal_inode);
2731 return NULL;
2732 }
2733 journal->j_private = sb;
617ba13b 2734 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 2735 return journal;
2736}
2737
617ba13b 2738static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 2739 dev_t j_dev)
2740{
2b2d6d01 2741 struct buffer_head *bh;
ac27a0ec 2742 journal_t *journal;
617ba13b
MC		 2743 ext4_fsblk_t start;
2744 ext4_fsblk_t len;
ac27a0ec 2745 int hblock, blocksize;
617ba13b 2746 ext4_fsblk_t sb_block;
ac27a0ec 2747 unsigned long offset;
2b2d6d01 2748 struct ext4_super_block *es;
ac27a0ec
DK		 2749 struct block_device *bdev;
2750
0390131b
FM		 2751 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
2752
617ba13b 2753 bdev = ext4_blkdev_get(j_dev);
ac27a0ec
DK		 2754 if (bdev == NULL)
2755 return NULL;
2756
2757 if (bd_claim(bdev, sb)) {
2758 printk(KERN_ERR
8c55e204 2759 "EXT4: failed to claim external journal device.\n");
9a1c3542 2760 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
ac27a0ec
DK		 2761 return NULL;
2762 }
2763
2764 blocksize = sb->s_blocksize;
2765 hblock = bdev_hardsect_size(bdev);
2766 if (blocksize < hblock) {
2767 printk(KERN_ERR
617ba13b 2768 "EXT4-fs: blocksize too small for journal device.\n");
ac27a0ec
DK		 2769 goto out_bdev;
2770 }
2771
617ba13b
MC		 2772 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
2773 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 2774 set_blocksize(bdev, blocksize);
2775 if (!(bh = __bread(bdev, sb_block, blocksize))) {
617ba13b 2776 printk(KERN_ERR "EXT4-fs: couldn't read superblock of "
ac27a0ec
DK		 2777 "external journal\n");
2778 goto out_bdev;
2779 }
2780
617ba13b
MC		 2781 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2782 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 2783 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b
MC		 2784 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
2785 printk(KERN_ERR "EXT4-fs: external journal has "
ac27a0ec
DK		 2786 "bad superblock\n");
2787 brelse(bh);
2788 goto out_bdev;
2789 }
2790
617ba13b
MC		 2791 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
2792 printk(KERN_ERR "EXT4-fs: journal UUID does not match\n");
ac27a0ec
DK		 2793 brelse(bh);
2794 goto out_bdev;
2795 }
2796
bd81d8ee 2797 len = ext4_blocks_count(es);
ac27a0ec
DK		 2798 start = sb_block + 1;
2799 brelse(bh); /* we're done with the superblock */
2800
dab291af 2801 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 2802 start, len, blocksize);
2803 if (!journal) {
617ba13b 2804 printk(KERN_ERR "EXT4-fs: failed to create device journal\n");
ac27a0ec
DK		 2805 goto out_bdev;
2806 }
2807 journal->j_private = sb;
2808 ll_rw_block(READ, 1, &journal->j_sb_buffer);
2809 wait_on_buffer(journal->j_sb_buffer);
2810 if (!buffer_uptodate(journal->j_sb_buffer)) {
617ba13b 2811 printk(KERN_ERR "EXT4-fs: I/O error on journal device\n");
ac27a0ec
DK		 2812 goto out_journal;
2813 }
2814 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
617ba13b 2815 printk(KERN_ERR "EXT4-fs: External journal has more than one "
ac27a0ec
DK		 2816 "user (unsupported) - %d\n",
2817 be32_to_cpu(journal->j_superblock->s_nr_users));
2818 goto out_journal;
2819 }
617ba13b
MC		 2820 EXT4_SB(sb)->journal_bdev = bdev;
2821 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 2822 return journal;
2823out_journal:
dab291af 2824 jbd2_journal_destroy(journal);
ac27a0ec 2825out_bdev:
617ba13b 2826 ext4_blkdev_put(bdev);
ac27a0ec
DK		 2827 return NULL;
2828}
2829
617ba13b
MC		 2830static int ext4_load_journal(struct super_block *sb,
2831 struct ext4_super_block *es,
ac27a0ec
DK		 2832 unsigned long journal_devnum)
2833{
2834 journal_t *journal;
2835 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
2836 dev_t journal_dev;
2837 int err = 0;
2838 int really_read_only;
2839
0390131b
FM		 2840 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
2841
ac27a0ec
DK		 2842 if (journal_devnum &&
2843 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
617ba13b 2844 printk(KERN_INFO "EXT4-fs: external journal device major/minor "
ac27a0ec
DK		 2845 "numbers have changed\n");
2846 journal_dev = new_decode_dev(journal_devnum);
2847 } else
2848 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
2849
2850 really_read_only = bdev_read_only(sb->s_bdev);
2851
2852 /*
2853 * Are we loading a blank journal or performing recovery after a
2854 * crash? For recovery, we need to check in advance whether we
2855 * can get read-write access to the device.
2856 */
2857
617ba13b 2858 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
ac27a0ec 2859 if (sb->s_flags & MS_RDONLY) {
617ba13b 2860 printk(KERN_INFO "EXT4-fs: INFO: recovery "
ac27a0ec
DK		 2861 "required on readonly filesystem.\n");
2862 if (really_read_only) {
617ba13b 2863 printk(KERN_ERR "EXT4-fs: write access "
ac27a0ec
DK		 2864 "unavailable, cannot proceed.\n");
2865 return -EROFS;
2866 }
2b2d6d01
TT		 2867 printk(KERN_INFO "EXT4-fs: write access will "
2868 "be enabled during recovery.\n");
ac27a0ec
DK		 2869 }
2870 }
2871
2872 if (journal_inum && journal_dev) {
617ba13b 2873 printk(KERN_ERR "EXT4-fs: filesystem has both journal "
ac27a0ec
DK		 2874 "and inode journals!\n");
2875 return -EINVAL;
2876 }
2877
2878 if (journal_inum) {
617ba13b 2879 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 2880 return -EINVAL;
2881 } else {
617ba13b 2882 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 2883 return -EINVAL;
2884 }
2885
4776004f
TT		 2886 if (journal->j_flags & JBD2_BARRIER)
2887 printk(KERN_INFO "EXT4-fs: barriers enabled\n");
2888 else
2889 printk(KERN_INFO "EXT4-fs: barriers disabled\n");
2890
ac27a0ec 2891 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
dab291af 2892 err = jbd2_journal_update_format(journal);
ac27a0ec 2893 if (err) {
617ba13b 2894 printk(KERN_ERR "EXT4-fs: error updating journal.\n");
dab291af 2895 jbd2_journal_destroy(journal);
ac27a0ec
DK		 2896 return err;
2897 }
2898 }
2899
617ba13b 2900 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
dab291af 2901 err = jbd2_journal_wipe(journal, !really_read_only);
ac27a0ec 2902 if (!err)
dab291af 2903 err = jbd2_journal_load(journal);
ac27a0ec
DK		 2904
2905 if (err) {
617ba13b 2906 printk(KERN_ERR "EXT4-fs: error loading journal.\n");
dab291af 2907 jbd2_journal_destroy(journal);
ac27a0ec
DK		 2908 return err;
2909 }
2910
617ba13b
MC		 2911 EXT4_SB(sb)->s_journal = journal;
2912 ext4_clear_journal_err(sb, es);
ac27a0ec
DK		 2913
2914 if (journal_devnum &&
2915 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2916 es->s_journal_dev = cpu_to_le32(journal_devnum);
2917 sb->s_dirt = 1;
2918
2919 /* Make sure we flush the recovery flag to disk. */
617ba13b 2920 ext4_commit_super(sb, es, 1);
ac27a0ec
DK		 2921 }
2922
2923 return 0;
2924}
2925
2b2d6d01
TT		 2926static void ext4_commit_super(struct super_block *sb,
2927 struct ext4_super_block *es, int sync)
ac27a0ec 2928{
617ba13b 2929 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
ac27a0ec
DK		 2930
2931 if (!sbh)
2932 return;
914258bf
TT		 2933 if (buffer_write_io_error(sbh)) {
2934 /*
2935 * Oh, dear. A previous attempt to write the
2936 * superblock failed. This could happen because the
2937 * USB device was yanked out. Or it could happen to
2938 * be a transient write error and maybe the block will
2939 * be remapped. Nothing we can do but to retry the
2940 * write and hope for the best.
2941 */
2942 printk(KERN_ERR "ext4: previous I/O error to "
2943 "superblock detected for %s.\n", sb->s_id);
2944 clear_buffer_write_io_error(sbh);
2945 set_buffer_uptodate(sbh);
2946 }
ac27a0ec 2947 es->s_wtime = cpu_to_le32(get_seconds());
5d1b1b3f
AK		 2948 ext4_free_blocks_count_set(es, percpu_counter_sum_positive(
2949 &EXT4_SB(sb)->s_freeblocks_counter));
2950 es->s_free_inodes_count = cpu_to_le32(percpu_counter_sum_positive(
2951 &EXT4_SB(sb)->s_freeinodes_counter));
2952
ac27a0ec
DK		 2953 BUFFER_TRACE(sbh, "marking dirty");
2954 mark_buffer_dirty(sbh);
914258bf 2955 if (sync) {
ac27a0ec 2956 sync_dirty_buffer(sbh);
914258bf
TT		 2957 if (buffer_write_io_error(sbh)) {
2958 printk(KERN_ERR "ext4: I/O error while writing "
2959 "superblock for %s.\n", sb->s_id);
2960 clear_buffer_write_io_error(sbh);
2961 set_buffer_uptodate(sbh);
2962 }
2963 }
ac27a0ec
DK		 2964}
2965
2966
2967/*
2968 * Have we just finished recovery? If so, and if we are mounting (or
2969 * remounting) the filesystem readonly, then we will end up with a
2970 * consistent fs on disk. Record that fact.
2971 */
2b2d6d01
TT		 2972static void ext4_mark_recovery_complete(struct super_block *sb,
2973 struct ext4_super_block *es)
ac27a0ec 2974{
617ba13b 2975 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 2976
0390131b
FM		 2977 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2978 BUG_ON(journal != NULL);
2979 return;
2980 }
dab291af 2981 jbd2_journal_lock_updates(journal);
7ffe1ea8
HK		 2982 if (jbd2_journal_flush(journal) < 0)
2983 goto out;
2984
32c37730 2985 lock_super(sb);
617ba13b 2986 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
ac27a0ec 2987 sb->s_flags & MS_RDONLY) {
617ba13b 2988 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 2989 sb->s_dirt = 0;
617ba13b 2990 ext4_commit_super(sb, es, 1);
ac27a0ec 2991 }
32c37730 2992 unlock_super(sb);
7ffe1ea8
HK		 2993
2994out:
dab291af 2995 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 2996}
2997
2998/*
2999 * If we are mounting (or read-write remounting) a filesystem whose journal
3000 * has recorded an error from a previous lifetime, move that error to the
3001 * main filesystem now.
3002 */
2b2d6d01
TT		 3003static void ext4_clear_journal_err(struct super_block *sb,
3004 struct ext4_super_block *es)
ac27a0ec
DK		 3005{
3006 journal_t *journal;
3007 int j_errno;
3008 const char *errstr;
3009
0390131b
FM		 3010 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3011
617ba13b 3012 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 3013
3014 /*
3015 * Now check for any error status which may have been recorded in the
617ba13b 3016 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 3017 */
3018
dab291af 3019 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 3020 if (j_errno) {
3021 char nbuf[16];
3022
617ba13b 3023 errstr = ext4_decode_error(sb, j_errno, nbuf);
46e665e9 3024 ext4_warning(sb, __func__, "Filesystem error recorded "
ac27a0ec 3025 "from previous mount: %s", errstr);
46e665e9 3026 ext4_warning(sb, __func__, "Marking fs in need of "
ac27a0ec
DK		 3027 "filesystem check.");
3028
617ba13b
MC		 3029 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
3030 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2b2d6d01 3031 ext4_commit_super(sb, es, 1);
ac27a0ec 3032
dab291af 3033 jbd2_journal_clear_err(journal);
ac27a0ec
DK		 3034 }
3035}
3036
3037/*
3038 * Force the running and committing transactions to commit,
3039 * and wait on the commit.
3040 */
617ba13b 3041int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 3042{
3043 journal_t *journal;
0390131b 3044 int ret = 0;
ac27a0ec
DK		 3045
3046 if (sb->s_flags & MS_RDONLY)
3047 return 0;
3048
617ba13b 3049 journal = EXT4_SB(sb)->s_journal;
0390131b
FM		 3050 if (journal) {
3051 sb->s_dirt = 0;
3052 ret = ext4_journal_force_commit(journal);
3053 }
3054
ac27a0ec
DK		 3055 return ret;
3056}
3057
3058/*
617ba13b 3059 * Ext4 always journals updates to the superblock itself, so we don't
ac27a0ec 3060 * have to propagate any other updates to the superblock on disk at this
14ce0cb4
TT		 3061 * point. (We can probably nuke this function altogether, and remove
3062 * any mention to sb->s_dirt in all of fs/ext4; eventual cleanup...)
ac27a0ec 3063 */
2b2d6d01 3064static void ext4_write_super(struct super_block *sb)
ac27a0ec 3065{
0390131b
FM		 3066 if (EXT4_SB(sb)->s_journal) {
3067 if (mutex_trylock(&sb->s_lock) != 0)
3068 BUG();
3069 sb->s_dirt = 0;
3070 } else {
3071 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
3072 }
ac27a0ec
DK		 3073}
3074
617ba13b 3075static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec 3076{
14ce0cb4 3077 int ret = 0;
ac27a0ec 3078
ede86cc4 3079 trace_mark(ext4_sync_fs, "dev %s wait %d", sb->s_id, wait);
ac27a0ec 3080 sb->s_dirt = 0;
0390131b
FM		 3081 if (EXT4_SB(sb)->s_journal) {
3082 if (wait)
3083 ret = ext4_force_commit(sb);
3084 else
3085 jbd2_journal_start_commit(EXT4_SB(sb)->s_journal, NULL);
3086 } else {
3087 ext4_commit_super(sb, EXT4_SB(sb)->s_es, wait);
3088 }
14ce0cb4 3089 return ret;
ac27a0ec
DK		 3090}
3091
3092/*
3093 * LVM calls this function before a (read-only) snapshot is created. This
3094 * gives us a chance to flush the journal completely and mark the fs clean.
3095 */
617ba13b 3096static void ext4_write_super_lockfs(struct super_block *sb)
ac27a0ec
DK		 3097{
3098 sb->s_dirt = 0;
3099
3100 if (!(sb->s_flags & MS_RDONLY)) {
617ba13b 3101 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 3102
0390131b
FM		 3103 if (journal) {
3104 /* Now we set up the journal barrier. */
3105 jbd2_journal_lock_updates(journal);
7ffe1ea8 3106
0390131b
FM		 3107 /*
3108 * We don't want to clear needs_recovery flag when we
3109 * failed to flush the journal.
3110 */
3111 if (jbd2_journal_flush(journal) < 0)
3112 return;
3113 }
ac27a0ec
DK		 3114
3115 /* Journal blocked and flushed, clear needs_recovery flag. */
617ba13b
MC		 3116 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3117 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
ac27a0ec
DK		 3118 }
3119}
3120
3121/*
3122 * Called by LVM after the snapshot is done. We need to reset the RECOVER
3123 * flag here, even though the filesystem is not technically dirty yet.
3124 */
617ba13b 3125static void ext4_unlockfs(struct super_block *sb)
ac27a0ec 3126{
0390131b 3127 if (EXT4_SB(sb)->s_journal && !(sb->s_flags & MS_RDONLY)) {
ac27a0ec
DK		 3128 lock_super(sb);
3129 /* Reser the needs_recovery flag before the fs is unlocked. */
617ba13b
MC		 3130 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3131 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
ac27a0ec 3132 unlock_super(sb);
dab291af 3133 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
ac27a0ec
DK		 3134 }
3135}
3136
2b2d6d01 3137static int ext4_remount(struct super_block *sb, int *flags, char *data)
ac27a0ec 3138{
2b2d6d01 3139 struct ext4_super_block *es;
617ba13b
MC		 3140 struct ext4_sb_info *sbi = EXT4_SB(sb);
3141 ext4_fsblk_t n_blocks_count = 0;
ac27a0ec 3142 unsigned long old_sb_flags;
617ba13b 3143 struct ext4_mount_options old_opts;
8a266467 3144 ext4_group_t g;
b3881f74 3145 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
ac27a0ec
DK		 3146 int err;
3147#ifdef CONFIG_QUOTA
3148 int i;
3149#endif
3150
3151 /* Store the original options */
3152 old_sb_flags = sb->s_flags;
3153 old_opts.s_mount_opt = sbi->s_mount_opt;
3154 old_opts.s_resuid = sbi->s_resuid;
3155 old_opts.s_resgid = sbi->s_resgid;
3156 old_opts.s_commit_interval = sbi->s_commit_interval;
30773840
TT		 3157 old_opts.s_min_batch_time = sbi->s_min_batch_time;
3158 old_opts.s_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 3159#ifdef CONFIG_QUOTA
3160 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
3161 for (i = 0; i < MAXQUOTAS; i++)
3162 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
3163#endif
b3881f74
TT		 3164 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
3165 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
ac27a0ec
DK		 3166
3167 /*
3168 * Allow the "check" option to be passed as a remount option.
3169 */
b3881f74
TT		 3170 if (!parse_options(data, sb, NULL, &journal_ioprio,
3171 &n_blocks_count, 1)) {
ac27a0ec
DK		 3172 err = -EINVAL;
3173 goto restore_opts;
3174 }
3175
617ba13b 3176 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT)
46e665e9 3177 ext4_abort(sb, __func__, "Abort forced by user");
ac27a0ec
DK		 3178
3179 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
617ba13b 3180 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec
DK		 3181
3182 es = sbi->s_es;
3183
b3881f74 3184 if (sbi->s_journal) {
0390131b 3185 ext4_init_journal_params(sb, sbi->s_journal);
b3881f74
TT		 3186 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3187 }
ac27a0ec
DK		 3188
3189 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
bd81d8ee 3190 n_blocks_count > ext4_blocks_count(es)) {
617ba13b 3191 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT) {
ac27a0ec
DK		 3192 err = -EROFS;
3193 goto restore_opts;
3194 }
3195
3196 if (*flags & MS_RDONLY) {
3197 /*
3198 * First of all, the unconditional stuff we have to do
3199 * to disable replay of the journal when we next remount
3200 */
3201 sb->s_flags |= MS_RDONLY;
3202
3203 /*
3204 * OK, test if we are remounting a valid rw partition
3205 * readonly, and if so set the rdonly flag and then
3206 * mark the partition as valid again.
3207 */
617ba13b
MC		 3208 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
3209 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 3210 es->s_state = cpu_to_le16(sbi->s_mount_state);
3211
32c37730
JK		 3212 /*
3213 * We have to unlock super so that we can wait for
3214 * transactions.
3215 */
0390131b
FM		 3216 if (sbi->s_journal) {
3217 unlock_super(sb);
3218 ext4_mark_recovery_complete(sb, es);
3219 lock_super(sb);
3220 }
ac27a0ec 3221 } else {
3a06d778 3222 int ret;
617ba13b
MC		 3223 if ((ret = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3224 ~EXT4_FEATURE_RO_COMPAT_SUPP))) {
3225 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
ac27a0ec 3226 "remount RDWR because of unsupported "
3a06d778
AK		 3227 "optional features (%x).\n", sb->s_id,
3228 (le32_to_cpu(sbi->s_es->s_feature_ro_compat) &
3229 ~EXT4_FEATURE_RO_COMPAT_SUPP));
ac27a0ec
DK		 3230 err = -EROFS;
3231 goto restore_opts;
3232 }
ead6596b 3233
8a266467
TT		 3234 /*
3235 * Make sure the group descriptor checksums
3236 * are sane. If they aren't, refuse to
3237 * remount r/w.
3238 */
3239 for (g = 0; g < sbi->s_groups_count; g++) {
3240 struct ext4_group_desc *gdp =
3241 ext4_get_group_desc(sb, g, NULL);
3242
3243 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
3244 printk(KERN_ERR
3245 "EXT4-fs: ext4_remount: "
a9df9a49 3246 "Checksum for group %u failed (%u!=%u)\n",
8a266467
TT		 3247 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
3248 le16_to_cpu(gdp->bg_checksum));
3249 err = -EINVAL;
3250 goto restore_opts;
3251 }
3252 }
3253
ead6596b
ES		 3254 /*
3255 * If we have an unprocessed orphan list hanging
3256 * around from a previously readonly bdev mount,
3257 * require a full umount/remount for now.
3258 */
3259 if (es->s_last_orphan) {
3260 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
3261 "remount RDWR because of unprocessed "
3262 "orphan inode list. Please "
3263 "umount/remount instead.\n",
3264 sb->s_id);
3265 err = -EINVAL;
3266 goto restore_opts;
3267 }
3268
ac27a0ec
DK		 3269 /*
3270 * Mounting a RDONLY partition read-write, so reread
3271 * and store the current valid flag. (It may have
3272 * been changed by e2fsck since we originally mounted
3273 * the partition.)
3274 */
0390131b
FM		 3275 if (sbi->s_journal)
3276 ext4_clear_journal_err(sb, es);
ac27a0ec 3277 sbi->s_mount_state = le16_to_cpu(es->s_state);
617ba13b 3278 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
ac27a0ec 3279 goto restore_opts;
2b2d6d01 3280 if (!ext4_setup_super(sb, es, 0))
ac27a0ec
DK		 3281 sb->s_flags &= ~MS_RDONLY;
3282 }
3283 }
0390131b
FM		 3284 if (sbi->s_journal == NULL)
3285 ext4_commit_super(sb, es, 1);
3286
ac27a0ec
DK		 3287#ifdef CONFIG_QUOTA
3288 /* Release old quota file names */
3289 for (i = 0; i < MAXQUOTAS; i++)
3290 if (old_opts.s_qf_names[i] &&
3291 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3292 kfree(old_opts.s_qf_names[i]);
3293#endif
3294 return 0;
3295restore_opts:
3296 sb->s_flags = old_sb_flags;
3297 sbi->s_mount_opt = old_opts.s_mount_opt;
3298 sbi->s_resuid = old_opts.s_resuid;
3299 sbi->s_resgid = old_opts.s_resgid;
3300 sbi->s_commit_interval = old_opts.s_commit_interval;
30773840
TT		 3301 sbi->s_min_batch_time = old_opts.s_min_batch_time;
3302 sbi->s_max_batch_time = old_opts.s_max_batch_time;
ac27a0ec
DK		 3303#ifdef CONFIG_QUOTA
3304 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
3305 for (i = 0; i < MAXQUOTAS; i++) {
3306 if (sbi->s_qf_names[i] &&
3307 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3308 kfree(sbi->s_qf_names[i]);
3309 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
3310 }
3311#endif
3312 return err;
3313}
3314
2b2d6d01 3315static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
ac27a0ec
DK		 3316{
3317 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 3318 struct ext4_sb_info *sbi = EXT4_SB(sb);
3319 struct ext4_super_block *es = sbi->s_es;
960cc398 3320 u64 fsid;
ac27a0ec 3321
5e70030d
BP		 3322 if (test_opt(sb, MINIX_DF)) {
3323 sbi->s_overhead_last = 0;
6bc9feff 3324 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
fd2d4291 3325 ext4_group_t ngroups = sbi->s_groups_count, i;
5e70030d 3326 ext4_fsblk_t overhead = 0;
ac27a0ec
DK		 3327 smp_rmb();
3328
3329 /*
5e70030d
BP		 3330 * Compute the overhead (FS structures). This is constant
3331 * for a given filesystem unless the number of block groups
3332 * changes so we cache the previous value until it does.
ac27a0ec
DK		 3333 */
3334
3335 /*
3336 * All of the blocks before first_data_block are
3337 * overhead
3338 */
3339 overhead = le32_to_cpu(es->s_first_data_block);
3340
3341 /*
3342 * Add the overhead attributed to the superblock and
3343 * block group descriptors. If the sparse superblocks
3344 * feature is turned on, then not all groups have this.
3345 */
3346 for (i = 0; i < ngroups; i++) {
617ba13b
MC		 3347 overhead += ext4_bg_has_super(sb, i) +
3348 ext4_bg_num_gdb(sb, i);
ac27a0ec
DK		 3349 cond_resched();
3350 }
3351
3352 /*
3353 * Every block group has an inode bitmap, a block
3354 * bitmap, and an inode table.
3355 */
5e70030d
BP		 3356 overhead += ngroups * (2 + sbi->s_itb_per_group);
3357 sbi->s_overhead_last = overhead;
3358 smp_wmb();
6bc9feff 3359 sbi->s_blocks_last = ext4_blocks_count(es);
ac27a0ec
DK		 3360 }
3361
617ba13b 3362 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 3363 buf->f_bsize = sb->s_blocksize;
5e70030d 3364 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
6bc6e63f
AK		 3365 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
3366 percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
308ba3ec 3367 ext4_free_blocks_count_set(es, buf->f_bfree);
bd81d8ee
LV		 3368 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3369 if (buf->f_bfree < ext4_r_blocks_count(es))
ac27a0ec
DK		 3370 buf->f_bavail = 0;
3371 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 3372 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
5e70030d 3373 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
617ba13b 3374 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 3375 fsid = le64_to_cpup((void *)es->s_uuid) ^
3376 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3377 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3378 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
ac27a0ec
DK		 3379 return 0;
3380}
3381
3382/* Helper function for writing quotas on sync - we need to start transaction before quota file
3383 * is locked for write. Otherwise the are possible deadlocks:
3384 * Process 1 Process 2
617ba13b 3385 * ext4_create() quota_sync()
dab291af 3386 * jbd2_journal_start() write_dquot()
ac27a0ec 3387 * DQUOT_INIT() down(dqio_mutex)
dab291af 3388 * down(dqio_mutex) jbd2_journal_start()
ac27a0ec
DK		 3389 *
3390 */
3391
3392#ifdef CONFIG_QUOTA
3393
3394static inline struct inode *dquot_to_inode(struct dquot *dquot)
3395{
3396 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3397}
3398
617ba13b 3399static int ext4_dquot_initialize(struct inode *inode, int type)
ac27a0ec
DK		 3400{
3401 handle_t *handle;
3402 int ret, err;
3403
3404 /* We may create quota structure so we need to reserve enough blocks */
617ba13b 3405 handle = ext4_journal_start(inode, 2*EXT4_QUOTA_INIT_BLOCKS(inode->i_sb));
ac27a0ec
DK		 3406 if (IS_ERR(handle))
3407 return PTR_ERR(handle);
3408 ret = dquot_initialize(inode, type);
617ba13b 3409 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3410 if (!ret)
3411 ret = err;
3412 return ret;
3413}
3414
617ba13b 3415static int ext4_dquot_drop(struct inode *inode)
ac27a0ec
DK		 3416{
3417 handle_t *handle;
3418 int ret, err;
3419
3420 /* We may delete quota structure so we need to reserve enough blocks */
617ba13b 3421 handle = ext4_journal_start(inode, 2*EXT4_QUOTA_DEL_BLOCKS(inode->i_sb));
2887df13
JK		 3422 if (IS_ERR(handle)) {
3423 /*
3424 * We call dquot_drop() anyway to at least release references
3425 * to quota structures so that umount does not hang.
3426 */
3427 dquot_drop(inode);
ac27a0ec 3428 return PTR_ERR(handle);
2887df13 3429 }
ac27a0ec 3430 ret = dquot_drop(inode);
617ba13b 3431 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3432 if (!ret)
3433 ret = err;
3434 return ret;
3435}
3436
617ba13b 3437static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 3438{
3439 int ret, err;
3440 handle_t *handle;
3441 struct inode *inode;
3442
3443 inode = dquot_to_inode(dquot);
617ba13b
MC		 3444 handle = ext4_journal_start(inode,
3445 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 3446 if (IS_ERR(handle))
3447 return PTR_ERR(handle);
3448 ret = dquot_commit(dquot);
617ba13b 3449 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3450 if (!ret)
3451 ret = err;
3452 return ret;
3453}
3454
617ba13b 3455static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 3456{
3457 int ret, err;
3458 handle_t *handle;
3459
617ba13b
MC		 3460 handle = ext4_journal_start(dquot_to_inode(dquot),
3461 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 3462 if (IS_ERR(handle))
3463 return PTR_ERR(handle);
3464 ret = dquot_acquire(dquot);
617ba13b 3465 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3466 if (!ret)
3467 ret = err;
3468 return ret;
3469}
3470
617ba13b 3471static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 3472{
3473 int ret, err;
3474 handle_t *handle;
3475
617ba13b
MC		 3476 handle = ext4_journal_start(dquot_to_inode(dquot),
3477 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 3478 if (IS_ERR(handle)) {
3479 /* Release dquot anyway to avoid endless cycle in dqput() */
3480 dquot_release(dquot);
ac27a0ec 3481 return PTR_ERR(handle);
9c3013e9 3482 }
ac27a0ec 3483 ret = dquot_release(dquot);
617ba13b 3484 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3485 if (!ret)
3486 ret = err;
3487 return ret;
3488}
3489
617ba13b 3490static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec 3491{
2c8be6b2 3492 /* Are we journaling quotas? */
617ba13b
MC		 3493 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3494 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
ac27a0ec 3495 dquot_mark_dquot_dirty(dquot);
617ba13b 3496 return ext4_write_dquot(dquot);
ac27a0ec
DK		 3497 } else {
3498 return dquot_mark_dquot_dirty(dquot);
3499 }
3500}
3501
617ba13b 3502static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 3503{
3504 int ret, err;
3505 handle_t *handle;
3506
3507 /* Data block + inode block */
617ba13b 3508 handle = ext4_journal_start(sb->s_root->d_inode, 2);
ac27a0ec
DK		 3509 if (IS_ERR(handle))
3510 return PTR_ERR(handle);
3511 ret = dquot_commit_info(sb, type);
617ba13b 3512 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3513 if (!ret)
3514 ret = err;
3515 return ret;
3516}
3517
3518/*
3519 * Turn on quotas during mount time - we need to find
3520 * the quota file and such...
3521 */
617ba13b 3522static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 3523{
617ba13b
MC		 3524 return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
3525 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 3526}
3527
3528/*
3529 * Standard function to be called on quota_on
3530 */
617ba13b 3531static int ext4_quota_on(struct super_block *sb, int type, int format_id,
8264613d 3532 char *name, int remount)
ac27a0ec
DK		 3533{
3534 int err;
8264613d 3535 struct path path;
ac27a0ec
DK		 3536
3537 if (!test_opt(sb, QUOTA))
3538 return -EINVAL;
8264613d 3539 /* When remounting, no checks are needed and in fact, name is NULL */
0623543b 3540 if (remount)
8264613d 3541 return vfs_quota_on(sb, type, format_id, name, remount);
0623543b 3542
8264613d 3543 err = kern_path(name, LOOKUP_FOLLOW, &path);
ac27a0ec
DK		 3544 if (err)
3545 return err;
0623543b 3546
ac27a0ec 3547 /* Quotafile not on the same filesystem? */
8264613d
AV		 3548 if (path.mnt->mnt_sb != sb) {
3549 path_put(&path);
ac27a0ec
DK		 3550 return -EXDEV;
3551 }
0623543b
JK		 3552 /* Journaling quota? */
3553 if (EXT4_SB(sb)->s_qf_names[type]) {
2b2d6d01 3554 /* Quotafile not in fs root? */
8264613d 3555 if (path.dentry->d_parent != sb->s_root)
0623543b
JK		 3556 printk(KERN_WARNING
3557 "EXT4-fs: Quota file not on filesystem root. "
3558 "Journaled quota will not work.\n");
2b2d6d01 3559 }
0623543b
JK		 3560
3561 /*
3562 * When we journal data on quota file, we have to flush journal to see
3563 * all updates to the file when we bypass pagecache...
3564 */
0390131b
FM		 3565 if (EXT4_SB(sb)->s_journal &&
3566 ext4_should_journal_data(path.dentry->d_inode)) {
0623543b
JK		 3567 /*
3568 * We don't need to lock updates but journal_flush() could
3569 * otherwise be livelocked...
3570 */
3571 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 3572 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
0623543b 3573 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 3574 if (err) {
8264613d 3575 path_put(&path);
7ffe1ea8
HK		 3576 return err;
3577 }
0623543b
JK		 3578 }
3579
8264613d
AV		 3580 err = vfs_quota_on_path(sb, type, format_id, &path);
3581 path_put(&path);
77e69dac 3582 return err;
ac27a0ec
DK		 3583}
3584
3585/* Read data from quotafile - avoid pagecache and such because we cannot afford
3586 * acquiring the locks... As quota files are never truncated and quota code
3587 * itself serializes the operations (and noone else should touch the files)
3588 * we don't have to be afraid of races */
617ba13b 3589static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 3590 size_t len, loff_t off)
3591{
3592 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 3593 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 3594 int err = 0;
3595 int offset = off & (sb->s_blocksize - 1);
3596 int tocopy;
3597 size_t toread;
3598 struct buffer_head *bh;
3599 loff_t i_size = i_size_read(inode);
3600
3601 if (off > i_size)
3602 return 0;
3603 if (off+len > i_size)
3604 len = i_size-off;
3605 toread = len;
3606 while (toread > 0) {
3607 tocopy = sb->s_blocksize - offset < toread ?
3608 sb->s_blocksize - offset : toread;
617ba13b 3609 bh = ext4_bread(NULL, inode, blk, 0, &err);
ac27a0ec
DK		 3610 if (err)
3611 return err;
3612 if (!bh) /* A hole? */
3613 memset(data, 0, tocopy);
3614 else
3615 memcpy(data, bh->b_data+offset, tocopy);
3616 brelse(bh);
3617 offset = 0;
3618 toread -= tocopy;
3619 data += tocopy;
3620 blk++;
3621 }
3622 return len;
3623}
3624
3625/* Write to quotafile (we know the transaction is already started and has
3626 * enough credits) */
617ba13b 3627static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 3628 const char *data, size_t len, loff_t off)
3629{
3630 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 3631 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 3632 int err = 0;
3633 int offset = off & (sb->s_blocksize - 1);
3634 int tocopy;
617ba13b 3635 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
ac27a0ec
DK		 3636 size_t towrite = len;
3637 struct buffer_head *bh;
3638 handle_t *handle = journal_current_handle();
3639
0390131b 3640 if (EXT4_SB(sb)->s_journal && !handle) {
e5f8eab8 3641 printk(KERN_WARNING "EXT4-fs: Quota write (off=%llu, len=%llu)"
9c3013e9
JK		 3642 " cancelled because transaction is not started.\n",
3643 (unsigned long long)off, (unsigned long long)len);
3644 return -EIO;
3645 }
ac27a0ec
DK		 3646 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
3647 while (towrite > 0) {
3648 tocopy = sb->s_blocksize - offset < towrite ?
3649 sb->s_blocksize - offset : towrite;
617ba13b 3650 bh = ext4_bread(handle, inode, blk, 1, &err);
ac27a0ec
DK		 3651 if (!bh)
3652 goto out;
3653 if (journal_quota) {
617ba13b 3654 err = ext4_journal_get_write_access(handle, bh);
ac27a0ec
DK		 3655 if (err) {
3656 brelse(bh);
3657 goto out;
3658 }
3659 }
3660 lock_buffer(bh);
3661 memcpy(bh->b_data+offset, data, tocopy);
3662 flush_dcache_page(bh->b_page);
3663 unlock_buffer(bh);
3664 if (journal_quota)
0390131b 3665 err = ext4_handle_dirty_metadata(handle, NULL, bh);
ac27a0ec
DK		 3666 else {
3667 /* Always do at least ordered writes for quotas */
678aaf48 3668 err = ext4_jbd2_file_inode(handle, inode);
ac27a0ec
DK		 3669 mark_buffer_dirty(bh);
3670 }
3671 brelse(bh);
3672 if (err)
3673 goto out;
3674 offset = 0;
3675 towrite -= tocopy;
3676 data += tocopy;
3677 blk++;
3678 }
3679out:
4d04e4fb
JK		 3680 if (len == towrite) {
3681 mutex_unlock(&inode->i_mutex);
ac27a0ec 3682 return err;
4d04e4fb 3683 }
ac27a0ec
DK		 3684 if (inode->i_size < off+len-towrite) {
3685 i_size_write(inode, off+len-towrite);
617ba13b 3686 EXT4_I(inode)->i_disksize = inode->i_size;
ac27a0ec 3687 }
ac27a0ec 3688 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
617ba13b 3689 ext4_mark_inode_dirty(handle, inode);
ac27a0ec
DK		 3690 mutex_unlock(&inode->i_mutex);
3691 return len - towrite;
3692}
3693
3694#endif
3695
617ba13b 3696static int ext4_get_sb(struct file_system_type *fs_type,
ac27a0ec
DK		 3697 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3698{
617ba13b 3699 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
ac27a0ec
DK		 3700}
3701
5e8814f2
TT		 3702#ifdef CONFIG_PROC_FS
3703static int ext4_ui_proc_show(struct seq_file *m, void *v)
3704{
3705 unsigned int *p = m->private;
3706
3707 seq_printf(m, "%u\n", *p);
3708 return 0;
3709}
3710
3711static int ext4_ui_proc_open(struct inode *inode, struct file *file)
3712{
3713 return single_open(file, ext4_ui_proc_show, PDE(inode)->data);
3714}
3715
3716static ssize_t ext4_ui_proc_write(struct file *file, const char __user *buf,
3717 size_t cnt, loff_t *ppos)
3718{
23475e26 3719 unsigned long *p = PDE(file->f_path.dentry->d_inode)->data;
5e8814f2 3720 char str[32];
5e8814f2
TT		 3721
3722 if (cnt >= sizeof(str))
3723 return -EINVAL;
3724 if (copy_from_user(str, buf, cnt))
3725 return -EFAULT;
23475e26
RK		 3726
3727 *p = simple_strtoul(str, NULL, 0);
5e8814f2
TT		 3728 return cnt;
3729}
3730
3731const struct file_operations ext4_ui_proc_fops = {
3732 .owner = THIS_MODULE,
3733 .open = ext4_ui_proc_open,
3734 .read = seq_read,
3735 .llseek = seq_lseek,
3736 .release = single_release,
3737 .write = ext4_ui_proc_write,
3738};
3739#endif
3740
03010a33
TT		 3741static struct file_system_type ext4_fs_type = {
3742 .owner = THIS_MODULE,
3743 .name = "ext4",
3744 .get_sb = ext4_get_sb,
3745 .kill_sb = kill_block_super,
3746 .fs_flags = FS_REQUIRES_DEV,
3747};
3748
3749#ifdef CONFIG_EXT4DEV_COMPAT
3750static int ext4dev_get_sb(struct file_system_type *fs_type,
3751 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3752{
3753 printk(KERN_WARNING "EXT4-fs: Update your userspace programs "
3754 "to mount using ext4\n");
3755 printk(KERN_WARNING "EXT4-fs: ext4dev backwards compatibility "
3756 "will go away by 2.6.31\n");
3757 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
3758}
3759
617ba13b 3760static struct file_system_type ext4dev_fs_type = {
ac27a0ec 3761 .owner = THIS_MODULE,
617ba13b 3762 .name = "ext4dev",
03010a33 3763 .get_sb = ext4dev_get_sb,
ac27a0ec
DK		 3764 .kill_sb = kill_block_super,
3765 .fs_flags = FS_REQUIRES_DEV,
3766};
03010a33
TT		 3767MODULE_ALIAS("ext4dev");
3768#endif
ac27a0ec 3769
617ba13b 3770static int __init init_ext4_fs(void)
ac27a0ec 3771{
c9de560d
AT		 3772 int err;
3773
9f6200bb 3774 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
c9de560d 3775 err = init_ext4_mballoc();
ac27a0ec
DK		 3776 if (err)
3777 return err;
c9de560d
AT		 3778
3779 err = init_ext4_xattr();
3780 if (err)
3781 goto out2;
ac27a0ec
DK		 3782 err = init_inodecache();
3783 if (err)
3784 goto out1;
03010a33 3785 err = register_filesystem(&ext4_fs_type);
ac27a0ec
DK		 3786 if (err)
3787 goto out;
03010a33
TT		 3788#ifdef CONFIG_EXT4DEV_COMPAT
3789 err = register_filesystem(&ext4dev_fs_type);
3790 if (err) {
3791 unregister_filesystem(&ext4_fs_type);
3792 goto out;
3793 }
3794#endif
ac27a0ec
DK		 3795 return 0;
3796out:
3797 destroy_inodecache();
3798out1:
617ba13b 3799 exit_ext4_xattr();
c9de560d
AT		 3800out2:
3801 exit_ext4_mballoc();
ac27a0ec
DK		 3802 return err;
3803}
3804
617ba13b 3805static void __exit exit_ext4_fs(void)
ac27a0ec 3806{
03010a33
TT		 3807 unregister_filesystem(&ext4_fs_type);
3808#ifdef CONFIG_EXT4DEV_COMPAT
617ba13b 3809 unregister_filesystem(&ext4dev_fs_type);
03010a33 3810#endif
ac27a0ec 3811 destroy_inodecache();
617ba13b 3812 exit_ext4_xattr();
c9de560d 3813 exit_ext4_mballoc();
9f6200bb 3814 remove_proc_entry("fs/ext4", NULL);
ac27a0ec
DK		 3815}
3816
3817MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
617ba13b 3818MODULE_DESCRIPTION("Fourth Extended Filesystem with extents");
ac27a0ec 3819MODULE_LICENSE("GPL");
617ba13b
MC		 3820module_init(init_ext4_fs)
3821module_exit(exit_ext4_fs)
Clone of davem's net-next-2.6 tree