]> git.proxmox.com Git - mirror_ubuntu-artful-kernel.git/blame - fs/ext4/super.c
projects / mirror_ubuntu-artful-kernel.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
ext4: display average commit time
[mirror_ubuntu-artful-kernel.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);
617ba13b 54static int ext4_create_journal(struct super_block *, struct ext4_super_block *,
ac27a0ec 55 unsigned int);
2b2d6d01
TT		 56static void ext4_commit_super(struct super_block *sb,
57 struct ext4_super_block *es, int sync);
58static void ext4_mark_recovery_complete(struct super_block *sb,
59 struct ext4_super_block *es);
60static void ext4_clear_journal_err(struct super_block *sb,
61 struct ext4_super_block *es);
617ba13b 62static int ext4_sync_fs(struct super_block *sb, int wait);
2b2d6d01 63static const char *ext4_decode_error(struct super_block *sb, int errno,
ac27a0ec 64 char nbuf[16]);
2b2d6d01
TT		 65static int ext4_remount(struct super_block *sb, int *flags, char *data);
66static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
617ba13b 67static void ext4_unlockfs(struct super_block *sb);
2b2d6d01 68static void ext4_write_super(struct super_block *sb);
617ba13b 69static void ext4_write_super_lockfs(struct super_block *sb);
ac27a0ec 70
bd81d8ee 71
8fadc143
AR		 72ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
73 struct ext4_group_desc *bg)
bd81d8ee 74{
3a14589c 75 return le32_to_cpu(bg->bg_block_bitmap_lo) |
8fadc143 76 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
3a14589c 77 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 78}
79
8fadc143
AR		 80ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
81 struct ext4_group_desc *bg)
bd81d8ee 82{
5272f837 83 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
8fadc143 84 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
5272f837 85 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 86}
87
8fadc143
AR		 88ext4_fsblk_t ext4_inode_table(struct super_block *sb,
89 struct ext4_group_desc *bg)
bd81d8ee 90{
5272f837 91 return le32_to_cpu(bg->bg_inode_table_lo) |
8fadc143 92 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
5272f837 93 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
bd81d8ee
LV		 94}
95
8fadc143
AR		 96void ext4_block_bitmap_set(struct super_block *sb,
97 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 98{
3a14589c 99 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 100 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
101 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 102}
103
8fadc143
AR		 104void ext4_inode_bitmap_set(struct super_block *sb,
105 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 106{
5272f837 107 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 108 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
109 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 110}
111
8fadc143
AR		 112void ext4_inode_table_set(struct super_block *sb,
113 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 114{
5272f837 115 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 116 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
117 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 118}
119
ac27a0ec 120/*
dab291af 121 * Wrappers for jbd2_journal_start/end.
ac27a0ec
DK		 122 *
123 * The only special thing we need to do here is to make sure that all
124 * journal_end calls result in the superblock being marked dirty, so
125 * that sync() will call the filesystem's write_super callback if
126 * appropriate.
127 */
617ba13b 128handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
ac27a0ec
DK		 129{
130 journal_t *journal;
131
132 if (sb->s_flags & MS_RDONLY)
133 return ERR_PTR(-EROFS);
134
135 /* Special case here: if the journal has aborted behind our
136 * backs (eg. EIO in the commit thread), then we still need to
137 * take the FS itself readonly cleanly. */
617ba13b 138 journal = EXT4_SB(sb)->s_journal;
0390131b
FM		 139 if (journal) {
140 if (is_journal_aborted(journal)) {
141 ext4_abort(sb, __func__,
142 "Detected aborted journal");
143 return ERR_PTR(-EROFS);
144 }
145 return jbd2_journal_start(journal, nblocks);
ac27a0ec 146 }
0390131b
FM		 147 /*
148 * We're not journaling, return the appropriate indication.
149 */
150 current->journal_info = EXT4_NOJOURNAL_HANDLE;
151 return current->journal_info;
ac27a0ec
DK		 152}
153
154/*
155 * The only special thing we need to do here is to make sure that all
dab291af 156 * jbd2_journal_stop calls result in the superblock being marked dirty, so
ac27a0ec
DK		 157 * that sync() will call the filesystem's write_super callback if
158 * appropriate.
159 */
617ba13b 160int __ext4_journal_stop(const char *where, handle_t *handle)
ac27a0ec
DK		 161{
162 struct super_block *sb;
163 int err;
164 int rc;
165
0390131b
FM		 166 if (!ext4_handle_valid(handle)) {
167 /*
168 * Do this here since we don't call jbd2_journal_stop() in
169 * no-journal mode.
170 */
171 current->journal_info = NULL;
172 return 0;
173 }
ac27a0ec
DK		 174 sb = handle->h_transaction->t_journal->j_private;
175 err = handle->h_err;
dab291af 176 rc = jbd2_journal_stop(handle);
ac27a0ec
DK		 177
178 if (!err)
179 err = rc;
180 if (err)
617ba13b 181 __ext4_std_error(sb, where, err);
ac27a0ec
DK		 182 return err;
183}
184
617ba13b 185void ext4_journal_abort_handle(const char *caller, const char *err_fn,
ac27a0ec
DK		 186 struct buffer_head *bh, handle_t *handle, int err)
187{
188 char nbuf[16];
617ba13b 189 const char *errstr = ext4_decode_error(NULL, err, nbuf);
ac27a0ec 190
0390131b
FM		 191 BUG_ON(!ext4_handle_valid(handle));
192
ac27a0ec
DK		 193 if (bh)
194 BUFFER_TRACE(bh, "abort");
195
196 if (!handle->h_err)
197 handle->h_err = err;
198
199 if (is_handle_aborted(handle))
200 return;
201
202 printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
203 caller, errstr, err_fn);
204
dab291af 205 jbd2_journal_abort_handle(handle);
ac27a0ec
DK		 206}
207
208/* Deal with the reporting of failure conditions on a filesystem such as
209 * inconsistencies detected or read IO failures.
210 *
211 * On ext2, we can store the error state of the filesystem in the
617ba13b 212 * superblock. That is not possible on ext4, because we may have other
ac27a0ec
DK		 213 * write ordering constraints on the superblock which prevent us from
214 * writing it out straight away; and given that the journal is about to
215 * be aborted, we can't rely on the current, or future, transactions to
216 * write out the superblock safely.
217 *
dab291af 218 * We'll just use the jbd2_journal_abort() error code to record an error in
ac27a0ec
DK		 219 * the journal instead. On recovery, the journal will compain about
220 * that error until we've noted it down and cleared it.
221 */
222
617ba13b 223static void ext4_handle_error(struct super_block *sb)
ac27a0ec 224{
617ba13b 225 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 226
617ba13b
MC		 227 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
228 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
ac27a0ec
DK		 229
230 if (sb->s_flags & MS_RDONLY)
231 return;
232
2b2d6d01 233 if (!test_opt(sb, ERRORS_CONT)) {
617ba13b 234 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 235
617ba13b 236 EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
ac27a0ec 237 if (journal)
dab291af 238 jbd2_journal_abort(journal, -EIO);
ac27a0ec 239 }
2b2d6d01
TT		 240 if (test_opt(sb, ERRORS_RO)) {
241 printk(KERN_CRIT "Remounting filesystem read-only\n");
ac27a0ec
DK		 242 sb->s_flags |= MS_RDONLY;
243 }
617ba13b 244 ext4_commit_super(sb, es, 1);
ac27a0ec 245 if (test_opt(sb, ERRORS_PANIC))
617ba13b 246 panic("EXT4-fs (device %s): panic forced after error\n",
ac27a0ec
DK		 247 sb->s_id);
248}
249
2b2d6d01
TT		 250void ext4_error(struct super_block *sb, const char *function,
251 const char *fmt, ...)
ac27a0ec
DK		 252{
253 va_list args;
254
255 va_start(args, fmt);
2b2d6d01 256 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
ac27a0ec
DK		 257 vprintk(fmt, args);
258 printk("\n");
259 va_end(args);
260
617ba13b 261 ext4_handle_error(sb);
ac27a0ec
DK		 262}
263
2b2d6d01 264static const char *ext4_decode_error(struct super_block *sb, int errno,
ac27a0ec
DK		 265 char nbuf[16])
266{
267 char *errstr = NULL;
268
269 switch (errno) {
270 case -EIO:
271 errstr = "IO failure";
272 break;
273 case -ENOMEM:
274 errstr = "Out of memory";
275 break;
276 case -EROFS:
dab291af 277 if (!sb || EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT)
ac27a0ec
DK		 278 errstr = "Journal has aborted";
279 else
280 errstr = "Readonly filesystem";
281 break;
282 default:
283 /* If the caller passed in an extra buffer for unknown
284 * errors, textualise them now. Else we just return
285 * NULL. */
286 if (nbuf) {
287 /* Check for truncated error codes... */
288 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
289 errstr = nbuf;
290 }
291 break;
292 }
293
294 return errstr;
295}
296
617ba13b 297/* __ext4_std_error decodes expected errors from journaling functions
ac27a0ec
DK		 298 * automatically and invokes the appropriate error response. */
299
2b2d6d01 300void __ext4_std_error(struct super_block *sb, const char *function, int errno)
ac27a0ec
DK		 301{
302 char nbuf[16];
303 const char *errstr;
304
305 /* Special case: if the error is EROFS, and we're not already
306 * inside a transaction, then there's really no point in logging
307 * an error. */
308 if (errno == -EROFS && journal_current_handle() == NULL &&
309 (sb->s_flags & MS_RDONLY))
310 return;
311
617ba13b 312 errstr = ext4_decode_error(sb, errno, nbuf);
2b2d6d01
TT		 313 printk(KERN_CRIT "EXT4-fs error (device %s) in %s: %s\n",
314 sb->s_id, function, errstr);
ac27a0ec 315
617ba13b 316 ext4_handle_error(sb);
ac27a0ec
DK		 317}
318
319/*
617ba13b 320 * ext4_abort is a much stronger failure handler than ext4_error. The
ac27a0ec
DK		 321 * abort function may be used to deal with unrecoverable failures such
322 * as journal IO errors or ENOMEM at a critical moment in log management.
323 *
324 * We unconditionally force the filesystem into an ABORT|READONLY state,
325 * unless the error response on the fs has been set to panic in which
326 * case we take the easy way out and panic immediately.
327 */
328
2b2d6d01
TT		 329void ext4_abort(struct super_block *sb, const char *function,
330 const char *fmt, ...)
ac27a0ec
DK		 331{
332 va_list args;
333
2b2d6d01 334 printk(KERN_CRIT "ext4_abort called.\n");
ac27a0ec
DK		 335
336 va_start(args, fmt);
2b2d6d01 337 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
ac27a0ec
DK		 338 vprintk(fmt, args);
339 printk("\n");
340 va_end(args);
341
342 if (test_opt(sb, ERRORS_PANIC))
617ba13b 343 panic("EXT4-fs panic from previous error\n");
ac27a0ec
DK		 344
345 if (sb->s_flags & MS_RDONLY)
346 return;
347
348 printk(KERN_CRIT "Remounting filesystem read-only\n");
617ba13b 349 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
ac27a0ec 350 sb->s_flags |= MS_RDONLY;
617ba13b 351 EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
ef2cabf7
HK		 352 if (EXT4_SB(sb)->s_journal)
353 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
ac27a0ec
DK		 354}
355
2b2d6d01
TT		 356void ext4_warning(struct super_block *sb, const char *function,
357 const char *fmt, ...)
ac27a0ec
DK		 358{
359 va_list args;
360
361 va_start(args, fmt);
617ba13b 362 printk(KERN_WARNING "EXT4-fs warning (device %s): %s: ",
ac27a0ec
DK		 363 sb->s_id, function);
364 vprintk(fmt, args);
365 printk("\n");
366 va_end(args);
367}
368
617ba13b 369void ext4_update_dynamic_rev(struct super_block *sb)
ac27a0ec 370{
617ba13b 371 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 372
617ba13b 373 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
ac27a0ec
DK		 374 return;
375
46e665e9 376 ext4_warning(sb, __func__,
ac27a0ec
DK		 377 "updating to rev %d because of new feature flag, "
378 "running e2fsck is recommended",
617ba13b 379 EXT4_DYNAMIC_REV);
ac27a0ec 380
617ba13b
MC		 381 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
382 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
383 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
ac27a0ec
DK		 384 /* leave es->s_feature_*compat flags alone */
385 /* es->s_uuid will be set by e2fsck if empty */
386
387 /*
388 * The rest of the superblock fields should be zero, and if not it
389 * means they are likely already in use, so leave them alone. We
390 * can leave it up to e2fsck to clean up any inconsistencies there.
391 */
392}
393
394/*
395 * Open the external journal device
396 */
617ba13b 397static struct block_device *ext4_blkdev_get(dev_t dev)
ac27a0ec
DK		 398{
399 struct block_device *bdev;
400 char b[BDEVNAME_SIZE];
401
402 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
403 if (IS_ERR(bdev))
404 goto fail;
405 return bdev;
406
407fail:
617ba13b 408 printk(KERN_ERR "EXT4: failed to open journal device %s: %ld\n",
ac27a0ec
DK		 409 __bdevname(dev, b), PTR_ERR(bdev));
410 return NULL;
411}
412
413/*
414 * Release the journal device
415 */
617ba13b 416static int ext4_blkdev_put(struct block_device *bdev)
ac27a0ec
DK		 417{
418 bd_release(bdev);
9a1c3542 419 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
ac27a0ec
DK		 420}
421
617ba13b 422static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
ac27a0ec
DK		 423{
424 struct block_device *bdev;
425 int ret = -ENODEV;
426
427 bdev = sbi->journal_bdev;
428 if (bdev) {
617ba13b 429 ret = ext4_blkdev_put(bdev);
ac27a0ec
DK		 430 sbi->journal_bdev = NULL;
431 }
432 return ret;
433}
434
435static inline struct inode *orphan_list_entry(struct list_head *l)
436{
617ba13b 437 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
ac27a0ec
DK		 438}
439
617ba13b 440static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
ac27a0ec
DK		 441{
442 struct list_head *l;
443
444 printk(KERN_ERR "sb orphan head is %d\n",
445 le32_to_cpu(sbi->s_es->s_last_orphan));
446
447 printk(KERN_ERR "sb_info orphan list:\n");
448 list_for_each(l, &sbi->s_orphan) {
449 struct inode *inode = orphan_list_entry(l);
450 printk(KERN_ERR " "
451 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
452 inode->i_sb->s_id, inode->i_ino, inode,
453 inode->i_mode, inode->i_nlink,
454 NEXT_ORPHAN(inode));
455 }
456}
457
2b2d6d01 458static void ext4_put_super(struct super_block *sb)
ac27a0ec 459{
617ba13b
MC		 460 struct ext4_sb_info *sbi = EXT4_SB(sb);
461 struct ext4_super_block *es = sbi->s_es;
ef2cabf7 462 int i, err;
ac27a0ec 463
c9de560d 464 ext4_mb_release(sb);
a86c6181 465 ext4_ext_release(sb);
617ba13b 466 ext4_xattr_put_super(sb);
0390131b
FM		 467 if (sbi->s_journal) {
468 err = jbd2_journal_destroy(sbi->s_journal);
469 sbi->s_journal = NULL;
470 if (err < 0)
471 ext4_abort(sb, __func__,
472 "Couldn't clean up the journal");
473 }
ac27a0ec 474 if (!(sb->s_flags & MS_RDONLY)) {
617ba13b 475 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 476 es->s_state = cpu_to_le16(sbi->s_mount_state);
617ba13b 477 ext4_commit_super(sb, es, 1);
ac27a0ec 478 }
240799cd
TT		 479 if (sbi->s_proc) {
480 remove_proc_entry("inode_readahead_blks", sbi->s_proc);
9f6200bb 481 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 482 }
ac27a0ec
DK		 483
484 for (i = 0; i < sbi->s_gdb_count; i++)
485 brelse(sbi->s_group_desc[i]);
486 kfree(sbi->s_group_desc);
772cb7c8 487 kfree(sbi->s_flex_groups);
ac27a0ec
DK		 488 percpu_counter_destroy(&sbi->s_freeblocks_counter);
489 percpu_counter_destroy(&sbi->s_freeinodes_counter);
490 percpu_counter_destroy(&sbi->s_dirs_counter);
6bc6e63f 491 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
ac27a0ec
DK		 492 brelse(sbi->s_sbh);
493#ifdef CONFIG_QUOTA
494 for (i = 0; i < MAXQUOTAS; i++)
495 kfree(sbi->s_qf_names[i]);
496#endif
497
498 /* Debugging code just in case the in-memory inode orphan list
499 * isn't empty. The on-disk one can be non-empty if we've
500 * detected an error and taken the fs readonly, but the
501 * in-memory list had better be clean by this point. */
502 if (!list_empty(&sbi->s_orphan))
503 dump_orphan_list(sb, sbi);
504 J_ASSERT(list_empty(&sbi->s_orphan));
505
f98393a6 506 invalidate_bdev(sb->s_bdev);
ac27a0ec
DK		 507 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
508 /*
509 * Invalidate the journal device's buffers. We don't want them
510 * floating about in memory - the physical journal device may
511 * hotswapped, and it breaks the `ro-after' testing code.
512 */
513 sync_blockdev(sbi->journal_bdev);
f98393a6 514 invalidate_bdev(sbi->journal_bdev);
617ba13b 515 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 516 }
517 sb->s_fs_info = NULL;
518 kfree(sbi);
519 return;
520}
521
e18b890b 522static struct kmem_cache *ext4_inode_cachep;
ac27a0ec
DK		 523
524/*
525 * Called inside transaction, so use GFP_NOFS
526 */
617ba13b 527static struct inode *ext4_alloc_inode(struct super_block *sb)
ac27a0ec 528{
617ba13b 529 struct ext4_inode_info *ei;
ac27a0ec 530
e6b4f8da 531 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
ac27a0ec
DK		 532 if (!ei)
533 return NULL;
03010a33 534#ifdef CONFIG_EXT4_FS_POSIX_ACL
617ba13b
MC		 535 ei->i_acl = EXT4_ACL_NOT_CACHED;
536 ei->i_default_acl = EXT4_ACL_NOT_CACHED;
ac27a0ec 537#endif
ac27a0ec 538 ei->vfs_inode.i_version = 1;
91246c00 539 ei->vfs_inode.i_data.writeback_index = 0;
a86c6181 540 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
c9de560d
AT		 541 INIT_LIST_HEAD(&ei->i_prealloc_list);
542 spin_lock_init(&ei->i_prealloc_lock);
0390131b
FM		 543 /*
544 * Note: We can be called before EXT4_SB(sb)->s_journal is set,
545 * therefore it can be null here. Don't check it, just initialize
546 * jinode.
547 */
678aaf48 548 jbd2_journal_init_jbd_inode(&ei->jinode, &ei->vfs_inode);
d2a17637
MC		 549 ei->i_reserved_data_blocks = 0;
550 ei->i_reserved_meta_blocks = 0;
551 ei->i_allocated_meta_blocks = 0;
552 ei->i_delalloc_reserved_flag = 0;
553 spin_lock_init(&(ei->i_block_reservation_lock));
ac27a0ec
DK		 554 return &ei->vfs_inode;
555}
556
617ba13b 557static void ext4_destroy_inode(struct inode *inode)
ac27a0ec 558{
9f7dd93d
VA		 559 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
560 printk("EXT4 Inode %p: orphan list check failed!\n",
561 EXT4_I(inode));
562 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
563 EXT4_I(inode), sizeof(struct ext4_inode_info),
564 true);
565 dump_stack();
566 }
617ba13b 567 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
ac27a0ec
DK		 568}
569
51cc5068 570static void init_once(void *foo)
ac27a0ec 571{
617ba13b 572 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
ac27a0ec 573
a35afb83 574 INIT_LIST_HEAD(&ei->i_orphan);
03010a33 575#ifdef CONFIG_EXT4_FS_XATTR
a35afb83 576 init_rwsem(&ei->xattr_sem);
ac27a0ec 577#endif
0e855ac8 578 init_rwsem(&ei->i_data_sem);
a35afb83 579 inode_init_once(&ei->vfs_inode);
ac27a0ec
DK		 580}
581
582static int init_inodecache(void)
583{
617ba13b
MC		 584 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
585 sizeof(struct ext4_inode_info),
ac27a0ec
DK		 586 0, (SLAB_RECLAIM_ACCOUNT|
587 SLAB_MEM_SPREAD),
20c2df83 588 init_once);
617ba13b 589 if (ext4_inode_cachep == NULL)
ac27a0ec
DK		 590 return -ENOMEM;
591 return 0;
592}
593
594static void destroy_inodecache(void)
595{
617ba13b 596 kmem_cache_destroy(ext4_inode_cachep);
ac27a0ec
DK		 597}
598
617ba13b 599static void ext4_clear_inode(struct inode *inode)
ac27a0ec 600{
03010a33 601#ifdef CONFIG_EXT4_FS_POSIX_ACL
617ba13b
MC		 602 if (EXT4_I(inode)->i_acl &&
603 EXT4_I(inode)->i_acl != EXT4_ACL_NOT_CACHED) {
604 posix_acl_release(EXT4_I(inode)->i_acl);
605 EXT4_I(inode)->i_acl = EXT4_ACL_NOT_CACHED;
606 }
607 if (EXT4_I(inode)->i_default_acl &&
608 EXT4_I(inode)->i_default_acl != EXT4_ACL_NOT_CACHED) {
609 posix_acl_release(EXT4_I(inode)->i_default_acl);
610 EXT4_I(inode)->i_default_acl = EXT4_ACL_NOT_CACHED;
ac27a0ec
DK		 611 }
612#endif
c2ea3fde 613 ext4_discard_preallocations(inode);
0390131b
FM		 614 if (EXT4_JOURNAL(inode))
615 jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal,
678aaf48 616 &EXT4_I(inode)->jinode);
ac27a0ec
DK		 617}
618
2b2d6d01
TT		 619static inline void ext4_show_quota_options(struct seq_file *seq,
620 struct super_block *sb)
ac27a0ec
DK		 621{
622#if defined(CONFIG_QUOTA)
617ba13b 623 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 624
625 if (sbi->s_jquota_fmt)
626 seq_printf(seq, ",jqfmt=%s",
af5bc92d 627 (sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold" : "vfsv0");
ac27a0ec
DK		 628
629 if (sbi->s_qf_names[USRQUOTA])
630 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
631
632 if (sbi->s_qf_names[GRPQUOTA])
633 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
634
617ba13b 635 if (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA)
ac27a0ec
DK		 636 seq_puts(seq, ",usrquota");
637
617ba13b 638 if (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)
ac27a0ec
DK		 639 seq_puts(seq, ",grpquota");
640#endif
641}
642
d9c9bef1
MS		 643/*
644 * Show an option if
645 * - it's set to a non-default value OR
646 * - if the per-sb default is different from the global default
647 */
617ba13b 648static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
ac27a0ec 649{
aa22df2c
AK		 650 int def_errors;
651 unsigned long def_mount_opts;
ac27a0ec 652 struct super_block *sb = vfs->mnt_sb;
d9c9bef1
MS		 653 struct ext4_sb_info *sbi = EXT4_SB(sb);
654 struct ext4_super_block *es = sbi->s_es;
d9c9bef1
MS		 655
656 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
aa22df2c 657 def_errors = le16_to_cpu(es->s_errors);
d9c9bef1
MS		 658
659 if (sbi->s_sb_block != 1)
660 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
661 if (test_opt(sb, MINIX_DF))
662 seq_puts(seq, ",minixdf");
aa22df2c 663 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
d9c9bef1
MS		 664 seq_puts(seq, ",grpid");
665 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
666 seq_puts(seq, ",nogrpid");
667 if (sbi->s_resuid != EXT4_DEF_RESUID ||
668 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
669 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
670 }
671 if (sbi->s_resgid != EXT4_DEF_RESGID ||
672 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
673 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
674 }
bb4f397a 675 if (test_opt(sb, ERRORS_RO)) {
d9c9bef1 676 if (def_errors == EXT4_ERRORS_PANIC ||
bb4f397a
AK		 677 def_errors == EXT4_ERRORS_CONTINUE) {
678 seq_puts(seq, ",errors=remount-ro");
d9c9bef1
MS		 679 }
680 }
aa22df2c 681 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
bb4f397a 682 seq_puts(seq, ",errors=continue");
aa22df2c 683 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
d9c9bef1 684 seq_puts(seq, ",errors=panic");
aa22df2c 685 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
d9c9bef1 686 seq_puts(seq, ",nouid32");
aa22df2c 687 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
d9c9bef1
MS		 688 seq_puts(seq, ",debug");
689 if (test_opt(sb, OLDALLOC))
690 seq_puts(seq, ",oldalloc");
03010a33 691#ifdef CONFIG_EXT4_FS_XATTR
aa22df2c
AK		 692 if (test_opt(sb, XATTR_USER) &&
693 !(def_mount_opts & EXT4_DEFM_XATTR_USER))
d9c9bef1
MS		 694 seq_puts(seq, ",user_xattr");
695 if (!test_opt(sb, XATTR_USER) &&
696 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
697 seq_puts(seq, ",nouser_xattr");
698 }
699#endif
03010a33 700#ifdef CONFIG_EXT4_FS_POSIX_ACL
aa22df2c 701 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
d9c9bef1
MS		 702 seq_puts(seq, ",acl");
703 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
704 seq_puts(seq, ",noacl");
705#endif
706 if (!test_opt(sb, RESERVATION))
707 seq_puts(seq, ",noreservation");
708 if (sbi->s_commit_interval) {
709 seq_printf(seq, ",commit=%u",
710 (unsigned) (sbi->s_commit_interval / HZ));
711 }
571640ca
ES		 712 /*
713 * We're changing the default of barrier mount option, so
714 * let's always display its mount state so it's clear what its
715 * status is.
716 */
717 seq_puts(seq, ",barrier=");
718 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
cd0b6a39
TT		 719 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
720 seq_puts(seq, ",journal_async_commit");
d9c9bef1
MS		 721 if (test_opt(sb, NOBH))
722 seq_puts(seq, ",nobh");
723 if (!test_opt(sb, EXTENTS))
724 seq_puts(seq, ",noextents");
25ec56b5
JNC		 725 if (test_opt(sb, I_VERSION))
726 seq_puts(seq, ",i_version");
dd919b98
AK		 727 if (!test_opt(sb, DELALLOC))
728 seq_puts(seq, ",nodelalloc");
729
ac27a0ec 730
cb45bbe4
MS		 731 if (sbi->s_stripe)
732 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
aa22df2c
AK		 733 /*
734 * journal mode get enabled in different ways
735 * So just print the value even if we didn't specify it
736 */
617ba13b 737 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
ac27a0ec 738 seq_puts(seq, ",data=journal");
617ba13b 739 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
ac27a0ec 740 seq_puts(seq, ",data=ordered");
617ba13b 741 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
ac27a0ec
DK		 742 seq_puts(seq, ",data=writeback");
743
240799cd
TT		 744 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
745 seq_printf(seq, ",inode_readahead_blks=%u",
746 sbi->s_inode_readahead_blks);
747
5bf5683a
HK		 748 if (test_opt(sb, DATA_ERR_ABORT))
749 seq_puts(seq, ",data_err=abort");
750
617ba13b 751 ext4_show_quota_options(seq, sb);
ac27a0ec
DK		 752 return 0;
753}
754
755
1b961ac0
CH		 756static struct inode *ext4_nfs_get_inode(struct super_block *sb,
757 u64 ino, u32 generation)
ac27a0ec 758{
ac27a0ec 759 struct inode *inode;
ac27a0ec 760
617ba13b 761 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
ac27a0ec 762 return ERR_PTR(-ESTALE);
617ba13b 763 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
ac27a0ec
DK		 764 return ERR_PTR(-ESTALE);
765
766 /* iget isn't really right if the inode is currently unallocated!!
767 *
617ba13b 768 * ext4_read_inode will return a bad_inode if the inode had been
ac27a0ec
DK		 769 * deleted, so we should be safe.
770 *
771 * Currently we don't know the generation for parent directory, so
772 * a generation of 0 means "accept any"
773 */
1d1fe1ee
DH		 774 inode = ext4_iget(sb, ino);
775 if (IS_ERR(inode))
776 return ERR_CAST(inode);
777 if (generation && inode->i_generation != generation) {
ac27a0ec
DK		 778 iput(inode);
779 return ERR_PTR(-ESTALE);
780 }
1b961ac0
CH		 781
782 return inode;
783}
784
785static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
786 int fh_len, int fh_type)
787{
788 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
789 ext4_nfs_get_inode);
790}
791
792static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
793 int fh_len, int fh_type)
794{
795 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
796 ext4_nfs_get_inode);
ac27a0ec
DK		 797}
798
799#ifdef CONFIG_QUOTA
af5bc92d 800#define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
2b2d6d01 801#define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
ac27a0ec 802
617ba13b
MC		 803static int ext4_dquot_initialize(struct inode *inode, int type);
804static int ext4_dquot_drop(struct inode *inode);
805static int ext4_write_dquot(struct dquot *dquot);
806static int ext4_acquire_dquot(struct dquot *dquot);
807static int ext4_release_dquot(struct dquot *dquot);
808static int ext4_mark_dquot_dirty(struct dquot *dquot);
809static int ext4_write_info(struct super_block *sb, int type);
6f28e087
JK		 810static int ext4_quota_on(struct super_block *sb, int type, int format_id,
811 char *path, int remount);
617ba13b
MC		 812static int ext4_quota_on_mount(struct super_block *sb, int type);
813static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 814 size_t len, loff_t off);
617ba13b 815static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 816 const char *data, size_t len, loff_t off);
817
617ba13b
MC		 818static struct dquot_operations ext4_quota_operations = {
819 .initialize = ext4_dquot_initialize,
820 .drop = ext4_dquot_drop,
ac27a0ec
DK		 821 .alloc_space = dquot_alloc_space,
822 .alloc_inode = dquot_alloc_inode,
823 .free_space = dquot_free_space,
824 .free_inode = dquot_free_inode,
825 .transfer = dquot_transfer,
617ba13b
MC		 826 .write_dquot = ext4_write_dquot,
827 .acquire_dquot = ext4_acquire_dquot,
828 .release_dquot = ext4_release_dquot,
829 .mark_dirty = ext4_mark_dquot_dirty,
830 .write_info = ext4_write_info
ac27a0ec
DK		 831};
832
617ba13b
MC		 833static struct quotactl_ops ext4_qctl_operations = {
834 .quota_on = ext4_quota_on,
ac27a0ec
DK		 835 .quota_off = vfs_quota_off,
836 .quota_sync = vfs_quota_sync,
837 .get_info = vfs_get_dqinfo,
838 .set_info = vfs_set_dqinfo,
839 .get_dqblk = vfs_get_dqblk,
840 .set_dqblk = vfs_set_dqblk
841};
842#endif
843
ee9b6d61 844static const struct super_operations ext4_sops = {
617ba13b
MC		 845 .alloc_inode = ext4_alloc_inode,
846 .destroy_inode = ext4_destroy_inode,
617ba13b
MC		 847 .write_inode = ext4_write_inode,
848 .dirty_inode = ext4_dirty_inode,
849 .delete_inode = ext4_delete_inode,
850 .put_super = ext4_put_super,
851 .write_super = ext4_write_super,
852 .sync_fs = ext4_sync_fs,
853 .write_super_lockfs = ext4_write_super_lockfs,
854 .unlockfs = ext4_unlockfs,
855 .statfs = ext4_statfs,
856 .remount_fs = ext4_remount,
857 .clear_inode = ext4_clear_inode,
858 .show_options = ext4_show_options,
ac27a0ec 859#ifdef CONFIG_QUOTA
617ba13b
MC		 860 .quota_read = ext4_quota_read,
861 .quota_write = ext4_quota_write,
ac27a0ec
DK		 862#endif
863};
864
39655164 865static const struct export_operations ext4_export_ops = {
1b961ac0
CH		 866 .fh_to_dentry = ext4_fh_to_dentry,
867 .fh_to_parent = ext4_fh_to_parent,
617ba13b 868 .get_parent = ext4_get_parent,
ac27a0ec
DK		 869};
870
871enum {
872 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
873 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
01436ef2 874 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
ac27a0ec
DK		 875 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
876 Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh,
877 Opt_commit, Opt_journal_update, Opt_journal_inum, Opt_journal_dev,
818d276c 878 Opt_journal_checksum, Opt_journal_async_commit,
ac27a0ec 879 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
5bf5683a 880 Opt_data_err_abort, Opt_data_err_ignore,
ac27a0ec
DK		 881 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
882 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
883 Opt_ignore, Opt_barrier, Opt_err, Opt_resize, Opt_usrquota,
25ec56b5 884 Opt_grpquota, Opt_extents, Opt_noextents, Opt_i_version,
01436ef2 885 Opt_stripe, Opt_delalloc, Opt_nodelalloc,
240799cd 886 Opt_inode_readahead_blks
ac27a0ec
DK		 887};
888
a447c093 889static const match_table_t tokens = {
ac27a0ec
DK		 890 {Opt_bsd_df, "bsddf"},
891 {Opt_minix_df, "minixdf"},
892 {Opt_grpid, "grpid"},
893 {Opt_grpid, "bsdgroups"},
894 {Opt_nogrpid, "nogrpid"},
895 {Opt_nogrpid, "sysvgroups"},
896 {Opt_resgid, "resgid=%u"},
897 {Opt_resuid, "resuid=%u"},
898 {Opt_sb, "sb=%u"},
899 {Opt_err_cont, "errors=continue"},
900 {Opt_err_panic, "errors=panic"},
901 {Opt_err_ro, "errors=remount-ro"},
902 {Opt_nouid32, "nouid32"},
ac27a0ec
DK		 903 {Opt_debug, "debug"},
904 {Opt_oldalloc, "oldalloc"},
905 {Opt_orlov, "orlov"},
906 {Opt_user_xattr, "user_xattr"},
907 {Opt_nouser_xattr, "nouser_xattr"},
908 {Opt_acl, "acl"},
909 {Opt_noacl, "noacl"},
910 {Opt_reservation, "reservation"},
911 {Opt_noreservation, "noreservation"},
912 {Opt_noload, "noload"},
913 {Opt_nobh, "nobh"},
914 {Opt_bh, "bh"},
915 {Opt_commit, "commit=%u"},
916 {Opt_journal_update, "journal=update"},
917 {Opt_journal_inum, "journal=%u"},
918 {Opt_journal_dev, "journal_dev=%u"},
818d276c
GS		 919 {Opt_journal_checksum, "journal_checksum"},
920 {Opt_journal_async_commit, "journal_async_commit"},
ac27a0ec
DK		 921 {Opt_abort, "abort"},
922 {Opt_data_journal, "data=journal"},
923 {Opt_data_ordered, "data=ordered"},
924 {Opt_data_writeback, "data=writeback"},
5bf5683a
HK		 925 {Opt_data_err_abort, "data_err=abort"},
926 {Opt_data_err_ignore, "data_err=ignore"},
ac27a0ec
DK		 927 {Opt_offusrjquota, "usrjquota="},
928 {Opt_usrjquota, "usrjquota=%s"},
929 {Opt_offgrpjquota, "grpjquota="},
930 {Opt_grpjquota, "grpjquota=%s"},
931 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
932 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
933 {Opt_grpquota, "grpquota"},
934 {Opt_noquota, "noquota"},
935 {Opt_quota, "quota"},
936 {Opt_usrquota, "usrquota"},
937 {Opt_barrier, "barrier=%u"},
a86c6181 938 {Opt_extents, "extents"},
1e2462f9 939 {Opt_noextents, "noextents"},
25ec56b5 940 {Opt_i_version, "i_version"},
c9de560d 941 {Opt_stripe, "stripe=%u"},
ac27a0ec 942 {Opt_resize, "resize"},
64769240 943 {Opt_delalloc, "delalloc"},
dd919b98 944 {Opt_nodelalloc, "nodelalloc"},
240799cd 945 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
f3f12faa 946 {Opt_err, NULL},
ac27a0ec
DK		 947};
948
617ba13b 949static ext4_fsblk_t get_sb_block(void **data)
ac27a0ec 950{
617ba13b 951 ext4_fsblk_t sb_block;
ac27a0ec
DK		 952 char *options = (char *) *data;
953
954 if (!options || strncmp(options, "sb=", 3) != 0)
955 return 1; /* Default location */
956 options += 3;
617ba13b 957 /*todo: use simple_strtoll with >32bit ext4 */
ac27a0ec
DK		 958 sb_block = simple_strtoul(options, &options, 0);
959 if (*options && *options != ',') {
4776004f 960 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
ac27a0ec
DK		 961 (char *) *data);
962 return 1;
963 }
964 if (*options == ',')
965 options++;
966 *data = (void *) options;
967 return sb_block;
968}
969
2b2d6d01
TT		 970static int parse_options(char *options, struct super_block *sb,
971 unsigned int *inum, unsigned long *journal_devnum,
972 ext4_fsblk_t *n_blocks_count, int is_remount)
ac27a0ec 973{
617ba13b 974 struct ext4_sb_info *sbi = EXT4_SB(sb);
2b2d6d01 975 char *p;
ac27a0ec
DK		 976 substring_t args[MAX_OPT_ARGS];
977 int data_opt = 0;
978 int option;
979#ifdef CONFIG_QUOTA
dfc5d03f 980 int qtype, qfmt;
ac27a0ec
DK		 981 char *qname;
982#endif
c07651b5 983 ext4_fsblk_t last_block;
ac27a0ec
DK		 984
985 if (!options)
986 return 1;
987
2b2d6d01 988 while ((p = strsep(&options, ",")) != NULL) {
ac27a0ec
DK		 989 int token;
990 if (!*p)
991 continue;
992
993 token = match_token(p, tokens, args);
994 switch (token) {
995 case Opt_bsd_df:
2b2d6d01 996 clear_opt(sbi->s_mount_opt, MINIX_DF);
ac27a0ec
DK		 997 break;
998 case Opt_minix_df:
2b2d6d01 999 set_opt(sbi->s_mount_opt, MINIX_DF);
ac27a0ec
DK		 1000 break;
1001 case Opt_grpid:
2b2d6d01 1002 set_opt(sbi->s_mount_opt, GRPID);
ac27a0ec
DK		 1003 break;
1004 case Opt_nogrpid:
2b2d6d01 1005 clear_opt(sbi->s_mount_opt, GRPID);
ac27a0ec
DK		 1006 break;
1007 case Opt_resuid:
1008 if (match_int(&args[0], &option))
1009 return 0;
1010 sbi->s_resuid = option;
1011 break;
1012 case Opt_resgid:
1013 if (match_int(&args[0], &option))
1014 return 0;
1015 sbi->s_resgid = option;
1016 break;
1017 case Opt_sb:
1018 /* handled by get_sb_block() instead of here */
1019 /* *sb_block = match_int(&args[0]); */
1020 break;
1021 case Opt_err_panic:
2b2d6d01
TT		 1022 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1023 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1024 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
ac27a0ec
DK		 1025 break;
1026 case Opt_err_ro:
2b2d6d01
TT		 1027 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1028 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1029 set_opt(sbi->s_mount_opt, ERRORS_RO);
ac27a0ec
DK		 1030 break;
1031 case Opt_err_cont:
2b2d6d01
TT		 1032 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1033 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1034 set_opt(sbi->s_mount_opt, ERRORS_CONT);
ac27a0ec
DK		 1035 break;
1036 case Opt_nouid32:
2b2d6d01 1037 set_opt(sbi->s_mount_opt, NO_UID32);
ac27a0ec 1038 break;
ac27a0ec 1039 case Opt_debug:
2b2d6d01 1040 set_opt(sbi->s_mount_opt, DEBUG);
ac27a0ec
DK		 1041 break;
1042 case Opt_oldalloc:
2b2d6d01 1043 set_opt(sbi->s_mount_opt, OLDALLOC);
ac27a0ec
DK		 1044 break;
1045 case Opt_orlov:
2b2d6d01 1046 clear_opt(sbi->s_mount_opt, OLDALLOC);
ac27a0ec 1047 break;
03010a33 1048#ifdef CONFIG_EXT4_FS_XATTR
ac27a0ec 1049 case Opt_user_xattr:
2b2d6d01 1050 set_opt(sbi->s_mount_opt, XATTR_USER);
ac27a0ec
DK		 1051 break;
1052 case Opt_nouser_xattr:
2b2d6d01 1053 clear_opt(sbi->s_mount_opt, XATTR_USER);
ac27a0ec
DK		 1054 break;
1055#else
1056 case Opt_user_xattr:
1057 case Opt_nouser_xattr:
4776004f
TT		 1058 printk(KERN_ERR "EXT4 (no)user_xattr options "
1059 "not supported\n");
ac27a0ec
DK		 1060 break;
1061#endif
03010a33 1062#ifdef CONFIG_EXT4_FS_POSIX_ACL
ac27a0ec
DK		 1063 case Opt_acl:
1064 set_opt(sbi->s_mount_opt, POSIX_ACL);
1065 break;
1066 case Opt_noacl:
1067 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1068 break;
1069#else
1070 case Opt_acl:
1071 case Opt_noacl:
4776004f
TT		 1072 printk(KERN_ERR "EXT4 (no)acl options "
1073 "not supported\n");
ac27a0ec
DK		 1074 break;
1075#endif
1076 case Opt_reservation:
1077 set_opt(sbi->s_mount_opt, RESERVATION);
1078 break;
1079 case Opt_noreservation:
1080 clear_opt(sbi->s_mount_opt, RESERVATION);
1081 break;
1082 case Opt_journal_update:
1083 /* @@@ FIXME */
1084 /* Eventually we will want to be able to create
1085 a journal file here. For now, only allow the
1086 user to specify an existing inode to be the
1087 journal file. */
1088 if (is_remount) {
617ba13b 1089 printk(KERN_ERR "EXT4-fs: cannot specify "
ac27a0ec
DK		 1090 "journal on remount\n");
1091 return 0;
1092 }
2b2d6d01 1093 set_opt(sbi->s_mount_opt, UPDATE_JOURNAL);
ac27a0ec
DK		 1094 break;
1095 case Opt_journal_inum:
1096 if (is_remount) {
617ba13b 1097 printk(KERN_ERR "EXT4-fs: cannot specify "
ac27a0ec
DK		 1098 "journal on remount\n");
1099 return 0;
1100 }
1101 if (match_int(&args[0], &option))
1102 return 0;
1103 *inum = option;
1104 break;
1105 case Opt_journal_dev:
1106 if (is_remount) {
617ba13b 1107 printk(KERN_ERR "EXT4-fs: cannot specify "
ac27a0ec
DK		 1108 "journal on remount\n");
1109 return 0;
1110 }
1111 if (match_int(&args[0], &option))
1112 return 0;
1113 *journal_devnum = option;
1114 break;
818d276c
GS		 1115 case Opt_journal_checksum:
1116 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1117 break;
1118 case Opt_journal_async_commit:
1119 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
1120 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1121 break;
ac27a0ec 1122 case Opt_noload:
2b2d6d01 1123 set_opt(sbi->s_mount_opt, NOLOAD);
ac27a0ec
DK		 1124 break;
1125 case Opt_commit:
1126 if (match_int(&args[0], &option))
1127 return 0;
1128 if (option < 0)
1129 return 0;
1130 if (option == 0)
cd02ff0b 1131 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
ac27a0ec
DK		 1132 sbi->s_commit_interval = HZ * option;
1133 break;
1134 case Opt_data_journal:
617ba13b 1135 data_opt = EXT4_MOUNT_JOURNAL_DATA;
ac27a0ec
DK		 1136 goto datacheck;
1137 case Opt_data_ordered:
617ba13b 1138 data_opt = EXT4_MOUNT_ORDERED_DATA;
ac27a0ec
DK		 1139 goto datacheck;
1140 case Opt_data_writeback:
617ba13b 1141 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
ac27a0ec
DK		 1142 datacheck:
1143 if (is_remount) {
617ba13b 1144 if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
ac27a0ec
DK		 1145 != data_opt) {
1146 printk(KERN_ERR
617ba13b 1147 "EXT4-fs: cannot change data "
ac27a0ec
DK		 1148 "mode on remount\n");
1149 return 0;
1150 }
1151 } else {
617ba13b 1152 sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
ac27a0ec
DK		 1153 sbi->s_mount_opt |= data_opt;
1154 }
1155 break;
5bf5683a
HK		 1156 case Opt_data_err_abort:
1157 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1158 break;
1159 case Opt_data_err_ignore:
1160 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1161 break;
ac27a0ec
DK		 1162#ifdef CONFIG_QUOTA
1163 case Opt_usrjquota:
1164 qtype = USRQUOTA;
1165 goto set_qf_name;
1166 case Opt_grpjquota:
1167 qtype = GRPQUOTA;
1168set_qf_name:
dfc5d03f
JK		 1169 if ((sb_any_quota_enabled(sb) ||
1170 sb_any_quota_suspended(sb)) &&
1171 !sbi->s_qf_names[qtype]) {
ac27a0ec 1172 printk(KERN_ERR
4776004f
TT		 1173 "EXT4-fs: Cannot change journaled "
1174 "quota options when quota turned on.\n");
ac27a0ec
DK		 1175 return 0;
1176 }
1177 qname = match_strdup(&args[0]);
1178 if (!qname) {
1179 printk(KERN_ERR
617ba13b 1180 "EXT4-fs: not enough memory for "
ac27a0ec
DK		 1181 "storing quotafile name.\n");
1182 return 0;
1183 }
1184 if (sbi->s_qf_names[qtype] &&
1185 strcmp(sbi->s_qf_names[qtype], qname)) {
1186 printk(KERN_ERR
617ba13b 1187 "EXT4-fs: %s quota file already "
ac27a0ec
DK		 1188 "specified.\n", QTYPE2NAME(qtype));
1189 kfree(qname);
1190 return 0;
1191 }
1192 sbi->s_qf_names[qtype] = qname;
1193 if (strchr(sbi->s_qf_names[qtype], '/')) {
1194 printk(KERN_ERR
617ba13b 1195 "EXT4-fs: quotafile must be on "
ac27a0ec
DK		 1196 "filesystem root.\n");
1197 kfree(sbi->s_qf_names[qtype]);
1198 sbi->s_qf_names[qtype] = NULL;
1199 return 0;
1200 }
1201 set_opt(sbi->s_mount_opt, QUOTA);
1202 break;
1203 case Opt_offusrjquota:
1204 qtype = USRQUOTA;
1205 goto clear_qf_name;
1206 case Opt_offgrpjquota:
1207 qtype = GRPQUOTA;
1208clear_qf_name:
dfc5d03f
JK		 1209 if ((sb_any_quota_enabled(sb) ||
1210 sb_any_quota_suspended(sb)) &&
1211 sbi->s_qf_names[qtype]) {
617ba13b 1212 printk(KERN_ERR "EXT4-fs: Cannot change "
2c8be6b2 1213 "journaled quota options when "
ac27a0ec
DK		 1214 "quota turned on.\n");
1215 return 0;
1216 }
1217 /*
1218 * The space will be released later when all options
1219 * are confirmed to be correct
1220 */
1221 sbi->s_qf_names[qtype] = NULL;
1222 break;
1223 case Opt_jqfmt_vfsold:
dfc5d03f
JK		 1224 qfmt = QFMT_VFS_OLD;
1225 goto set_qf_format;
ac27a0ec 1226 case Opt_jqfmt_vfsv0:
dfc5d03f
JK		 1227 qfmt = QFMT_VFS_V0;
1228set_qf_format:
1229 if ((sb_any_quota_enabled(sb) ||
1230 sb_any_quota_suspended(sb)) &&
1231 sbi->s_jquota_fmt != qfmt) {
1232 printk(KERN_ERR "EXT4-fs: Cannot change "
1233 "journaled quota options when "
1234 "quota turned on.\n");
1235 return 0;
1236 }
1237 sbi->s_jquota_fmt = qfmt;
ac27a0ec
DK		 1238 break;
1239 case Opt_quota:
1240 case Opt_usrquota:
1241 set_opt(sbi->s_mount_opt, QUOTA);
1242 set_opt(sbi->s_mount_opt, USRQUOTA);
1243 break;
1244 case Opt_grpquota:
1245 set_opt(sbi->s_mount_opt, QUOTA);
1246 set_opt(sbi->s_mount_opt, GRPQUOTA);
1247 break;
1248 case Opt_noquota:
1249 if (sb_any_quota_enabled(sb)) {
617ba13b 1250 printk(KERN_ERR "EXT4-fs: Cannot change quota "
ac27a0ec
DK		 1251 "options when quota turned on.\n");
1252 return 0;
1253 }
1254 clear_opt(sbi->s_mount_opt, QUOTA);
1255 clear_opt(sbi->s_mount_opt, USRQUOTA);
1256 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1257 break;
1258#else
1259 case Opt_quota:
1260 case Opt_usrquota:
1261 case Opt_grpquota:
cd59e7b9
JK		 1262 printk(KERN_ERR
1263 "EXT4-fs: quota options not supported.\n");
1264 break;
ac27a0ec
DK		 1265 case Opt_usrjquota:
1266 case Opt_grpjquota:
1267 case Opt_offusrjquota:
1268 case Opt_offgrpjquota:
1269 case Opt_jqfmt_vfsold:
1270 case Opt_jqfmt_vfsv0:
1271 printk(KERN_ERR
cd59e7b9 1272 "EXT4-fs: journaled quota options not "
ac27a0ec
DK		 1273 "supported.\n");
1274 break;
1275 case Opt_noquota:
1276 break;
1277#endif
1278 case Opt_abort:
1279 set_opt(sbi->s_mount_opt, ABORT);
1280 break;
1281 case Opt_barrier:
1282 if (match_int(&args[0], &option))
1283 return 0;
1284 if (option)
1285 set_opt(sbi->s_mount_opt, BARRIER);
1286 else
1287 clear_opt(sbi->s_mount_opt, BARRIER);
1288 break;
1289 case Opt_ignore:
1290 break;
1291 case Opt_resize:
1292 if (!is_remount) {
617ba13b 1293 printk("EXT4-fs: resize option only available "
ac27a0ec
DK		 1294 "for remount\n");
1295 return 0;
1296 }
1297 if (match_int(&args[0], &option) != 0)
1298 return 0;
1299 *n_blocks_count = option;
1300 break;
1301 case Opt_nobh:
1302 set_opt(sbi->s_mount_opt, NOBH);
1303 break;
1304 case Opt_bh:
1305 clear_opt(sbi->s_mount_opt, NOBH);
1306 break;
a86c6181 1307 case Opt_extents:
e4079a11
ES		 1308 if (!EXT4_HAS_INCOMPAT_FEATURE(sb,
1309 EXT4_FEATURE_INCOMPAT_EXTENTS)) {
1310 ext4_warning(sb, __func__,
1311 "extents feature not enabled "
fde4d95a 1312 "on this filesystem, use tune2fs");
e4079a11
ES		 1313 return 0;
1314 }
2b2d6d01 1315 set_opt(sbi->s_mount_opt, EXTENTS);
a86c6181 1316 break;
1e2462f9 1317 case Opt_noextents:
c07651b5
AK		 1318 /*
1319 * When e2fsprogs support resizing an already existing
1320 * ext3 file system to greater than 2**32 we need to
1321 * add support to block allocator to handle growing
1322 * already existing block mapped inode so that blocks
1323 * allocated for them fall within 2**32
1324 */
1325 last_block = ext4_blocks_count(sbi->s_es) - 1;
1326 if (last_block > 0xffffffffULL) {
1327 printk(KERN_ERR "EXT4-fs: Filesystem too "
1328 "large to mount with "
1329 "-o noextents options\n");
1330 return 0;
1331 }
2b2d6d01 1332 clear_opt(sbi->s_mount_opt, EXTENTS);
1e2462f9 1333 break;
25ec56b5
JNC		 1334 case Opt_i_version:
1335 set_opt(sbi->s_mount_opt, I_VERSION);
1336 sb->s_flags |= MS_I_VERSION;
1337 break;
dd919b98
AK		 1338 case Opt_nodelalloc:
1339 clear_opt(sbi->s_mount_opt, DELALLOC);
1340 break;
c9de560d
AT		 1341 case Opt_stripe:
1342 if (match_int(&args[0], &option))
1343 return 0;
1344 if (option < 0)
1345 return 0;
1346 sbi->s_stripe = option;
1347 break;
64769240
AT		 1348 case Opt_delalloc:
1349 set_opt(sbi->s_mount_opt, DELALLOC);
1350 break;
240799cd
TT		 1351 case Opt_inode_readahead_blks:
1352 if (match_int(&args[0], &option))
1353 return 0;
1354 if (option < 0 || option > (1 << 30))
1355 return 0;
1356 sbi->s_inode_readahead_blks = option;
1357 break;
ac27a0ec 1358 default:
2b2d6d01
TT		 1359 printk(KERN_ERR
1360 "EXT4-fs: Unrecognized mount option \"%s\" "
1361 "or missing value\n", p);
ac27a0ec
DK		 1362 return 0;
1363 }
1364 }
1365#ifdef CONFIG_QUOTA
1366 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
617ba13b 1367 if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
ac27a0ec
DK		 1368 sbi->s_qf_names[USRQUOTA])
1369 clear_opt(sbi->s_mount_opt, USRQUOTA);
1370
617ba13b 1371 if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
ac27a0ec
DK		 1372 sbi->s_qf_names[GRPQUOTA])
1373 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1374
1375 if ((sbi->s_qf_names[USRQUOTA] &&
617ba13b 1376 (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
ac27a0ec 1377 (sbi->s_qf_names[GRPQUOTA] &&
617ba13b
MC		 1378 (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
1379 printk(KERN_ERR "EXT4-fs: old and new quota "
ac27a0ec
DK		 1380 "format mixing.\n");
1381 return 0;
1382 }
1383
1384 if (!sbi->s_jquota_fmt) {
2c8be6b2 1385 printk(KERN_ERR "EXT4-fs: journaled quota format "
ac27a0ec
DK		 1386 "not specified.\n");
1387 return 0;
1388 }
1389 } else {
1390 if (sbi->s_jquota_fmt) {
2c8be6b2
JK		 1391 printk(KERN_ERR "EXT4-fs: journaled quota format "
1392 "specified with no journaling "
ac27a0ec
DK		 1393 "enabled.\n");
1394 return 0;
1395 }
1396 }
1397#endif
1398 return 1;
1399}
1400
617ba13b 1401static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 1402 int read_only)
1403{
617ba13b 1404 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 1405 int res = 0;
1406
617ba13b 1407 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
2b2d6d01
TT		 1408 printk(KERN_ERR "EXT4-fs warning: revision level too high, "
1409 "forcing read-only mode\n");
ac27a0ec
DK		 1410 res = MS_RDONLY;
1411 }
1412 if (read_only)
1413 return res;
617ba13b 1414 if (!(sbi->s_mount_state & EXT4_VALID_FS))
2b2d6d01
TT		 1415 printk(KERN_WARNING "EXT4-fs warning: mounting unchecked fs, "
1416 "running e2fsck is recommended\n");
617ba13b 1417 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
2b2d6d01
TT		 1418 printk(KERN_WARNING
1419 "EXT4-fs warning: mounting fs with errors, "
1420 "running e2fsck is recommended\n");
ac27a0ec
DK		 1421 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1422 le16_to_cpu(es->s_mnt_count) >=
1423 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
2b2d6d01
TT		 1424 printk(KERN_WARNING
1425 "EXT4-fs warning: maximal mount count reached, "
1426 "running e2fsck is recommended\n");
ac27a0ec
DK		 1427 else if (le32_to_cpu(es->s_checkinterval) &&
1428 (le32_to_cpu(es->s_lastcheck) +
1429 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
2b2d6d01
TT		 1430 printk(KERN_WARNING
1431 "EXT4-fs warning: checktime reached, "
1432 "running e2fsck is recommended\n");
0390131b
FM		 1433 if (!sbi->s_journal)
1434 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec 1435 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 1436 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 1437 le16_add_cpu(&es->s_mnt_count, 1);
ac27a0ec 1438 es->s_mtime = cpu_to_le32(get_seconds());
617ba13b 1439 ext4_update_dynamic_rev(sb);
0390131b
FM		 1440 if (sbi->s_journal)
1441 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 1442
617ba13b 1443 ext4_commit_super(sb, es, 1);
ac27a0ec 1444 if (test_opt(sb, DEBUG))
617ba13b 1445 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%lu, "
ac27a0ec
DK		 1446 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
1447 sb->s_blocksize,
1448 sbi->s_groups_count,
617ba13b
MC		 1449 EXT4_BLOCKS_PER_GROUP(sb),
1450 EXT4_INODES_PER_GROUP(sb),
ac27a0ec
DK		 1451 sbi->s_mount_opt);
1452
0390131b
FM		 1453 if (EXT4_SB(sb)->s_journal) {
1454 printk(KERN_INFO "EXT4 FS on %s, %s journal on %s\n",
1455 sb->s_id, EXT4_SB(sb)->s_journal->j_inode ? "internal" :
1456 "external", EXT4_SB(sb)->s_journal->j_devname);
1457 } else {
1458 printk(KERN_INFO "EXT4 FS on %s, no journal\n", sb->s_id);
1459 }
ac27a0ec
DK		 1460 return res;
1461}
1462
772cb7c8
JS		 1463static int ext4_fill_flex_info(struct super_block *sb)
1464{
1465 struct ext4_sb_info *sbi = EXT4_SB(sb);
1466 struct ext4_group_desc *gdp = NULL;
1467 struct buffer_head *bh;
1468 ext4_group_t flex_group_count;
1469 ext4_group_t flex_group;
1470 int groups_per_flex = 0;
772cb7c8
JS		 1471 int i;
1472
1473 if (!sbi->s_es->s_log_groups_per_flex) {
1474 sbi->s_log_groups_per_flex = 0;
1475 return 1;
1476 }
1477
1478 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1479 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1480
c62a11fd
FB		 1481 /* We allocate both existing and potentially added groups */
1482 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
d94e99a6
AK		 1483 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1484 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
ec05e868 1485 sbi->s_flex_groups = kzalloc(flex_group_count *
772cb7c8
JS		 1486 sizeof(struct flex_groups), GFP_KERNEL);
1487 if (sbi->s_flex_groups == NULL) {
ec05e868
LZ		 1488 printk(KERN_ERR "EXT4-fs: not enough memory for "
1489 "%lu flex groups\n", flex_group_count);
772cb7c8
JS		 1490 goto failed;
1491 }
772cb7c8 1492
772cb7c8
JS		 1493 for (i = 0; i < sbi->s_groups_count; i++) {
1494 gdp = ext4_get_group_desc(sb, i, &bh);
1495
1496 flex_group = ext4_flex_group(sbi, i);
1497 sbi->s_flex_groups[flex_group].free_inodes +=
1498 le16_to_cpu(gdp->bg_free_inodes_count);
1499 sbi->s_flex_groups[flex_group].free_blocks +=
1500 le16_to_cpu(gdp->bg_free_blocks_count);
1501 }
1502
1503 return 1;
1504failed:
1505 return 0;
1506}
1507
717d50e4
AD		 1508__le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1509 struct ext4_group_desc *gdp)
1510{
1511 __u16 crc = 0;
1512
1513 if (sbi->s_es->s_feature_ro_compat &
1514 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1515 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1516 __le32 le_group = cpu_to_le32(block_group);
1517
1518 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1519 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1520 crc = crc16(crc, (__u8 *)gdp, offset);
1521 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1522 /* for checksum of struct ext4_group_desc do the rest...*/
1523 if ((sbi->s_es->s_feature_incompat &
1524 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1525 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1526 crc = crc16(crc, (__u8 *)gdp + offset,
1527 le16_to_cpu(sbi->s_es->s_desc_size) -
1528 offset);
1529 }
1530
1531 return cpu_to_le16(crc);
1532}
1533
1534int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1535 struct ext4_group_desc *gdp)
1536{
1537 if ((sbi->s_es->s_feature_ro_compat &
1538 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1539 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1540 return 0;
1541
1542 return 1;
1543}
1544
ac27a0ec 1545/* Called at mount-time, super-block is locked */
197cd65a 1546static int ext4_check_descriptors(struct super_block *sb)
ac27a0ec 1547{
617ba13b
MC		 1548 struct ext4_sb_info *sbi = EXT4_SB(sb);
1549 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1550 ext4_fsblk_t last_block;
bd81d8ee
LV		 1551 ext4_fsblk_t block_bitmap;
1552 ext4_fsblk_t inode_bitmap;
1553 ext4_fsblk_t inode_table;
ce421581 1554 int flexbg_flag = 0;
fd2d4291 1555 ext4_group_t i;
ac27a0ec 1556
ce421581
JS		 1557 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1558 flexbg_flag = 1;
1559
af5bc92d 1560 ext4_debug("Checking group descriptors");
ac27a0ec 1561
197cd65a
AM		 1562 for (i = 0; i < sbi->s_groups_count; i++) {
1563 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1564
ce421581 1565 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 1566 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 1567 else
1568 last_block = first_block +
617ba13b 1569 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 1570
8fadc143 1571 block_bitmap = ext4_block_bitmap(sb, gdp);
2b2d6d01 1572 if (block_bitmap < first_block || block_bitmap > last_block) {
c19204b0
JB		 1573 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1574 "Block bitmap for group %lu not in group "
5128273a 1575 "(block %llu)!\n", i, block_bitmap);
ac27a0ec
DK		 1576 return 0;
1577 }
8fadc143 1578 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2b2d6d01 1579 if (inode_bitmap < first_block || inode_bitmap > last_block) {
c19204b0
JB		 1580 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1581 "Inode bitmap for group %lu not in group "
5128273a 1582 "(block %llu)!\n", i, inode_bitmap);
ac27a0ec
DK		 1583 return 0;
1584 }
8fadc143 1585 inode_table = ext4_inode_table(sb, gdp);
bd81d8ee 1586 if (inode_table < first_block ||
2b2d6d01 1587 inode_table + sbi->s_itb_per_group - 1 > last_block) {
c19204b0
JB		 1588 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1589 "Inode table for group %lu not in group "
5128273a 1590 "(block %llu)!\n", i, inode_table);
ac27a0ec
DK		 1591 return 0;
1592 }
b5f10eed 1593 spin_lock(sb_bgl_lock(sbi, i));
717d50e4 1594 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
c19204b0
JB		 1595 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1596 "Checksum for group %lu failed (%u!=%u)\n",
1597 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1598 gdp)), le16_to_cpu(gdp->bg_checksum));
7ee1ec4c
LZ		 1599 if (!(sb->s_flags & MS_RDONLY)) {
1600 spin_unlock(sb_bgl_lock(sbi, i));
8a266467 1601 return 0;
7ee1ec4c 1602 }
717d50e4 1603 }
b5f10eed 1604 spin_unlock(sb_bgl_lock(sbi, i));
ce421581
JS		 1605 if (!flexbg_flag)
1606 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec
DK		 1607 }
1608
bd81d8ee 1609 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
2b2d6d01 1610 sbi->s_es->s_free_inodes_count = cpu_to_le32(ext4_count_free_inodes(sb));
ac27a0ec
DK		 1611 return 1;
1612}
1613
617ba13b 1614/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 1615 * the superblock) which were deleted from all directories, but held open by
1616 * a process at the time of a crash. We walk the list and try to delete these
1617 * inodes at recovery time (only with a read-write filesystem).
1618 *
1619 * In order to keep the orphan inode chain consistent during traversal (in
1620 * case of crash during recovery), we link each inode into the superblock
1621 * orphan list_head and handle it the same way as an inode deletion during
1622 * normal operation (which journals the operations for us).
1623 *
1624 * We only do an iget() and an iput() on each inode, which is very safe if we
1625 * accidentally point at an in-use or already deleted inode. The worst that
1626 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 1627 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 1628 * e2fsck was run on this filesystem, and it must have already done the orphan
1629 * inode cleanup for us, so we can safely abort without any further action.
1630 */
2b2d6d01
TT		 1631static void ext4_orphan_cleanup(struct super_block *sb,
1632 struct ext4_super_block *es)
ac27a0ec
DK		 1633{
1634 unsigned int s_flags = sb->s_flags;
1635 int nr_orphans = 0, nr_truncates = 0;
1636#ifdef CONFIG_QUOTA
1637 int i;
1638#endif
1639 if (!es->s_last_orphan) {
1640 jbd_debug(4, "no orphan inodes to clean up\n");
1641 return;
1642 }
1643
a8f48a95
ES		 1644 if (bdev_read_only(sb->s_bdev)) {
1645 printk(KERN_ERR "EXT4-fs: write access "
1646 "unavailable, skipping orphan cleanup.\n");
1647 return;
1648 }
1649
617ba13b 1650 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
ac27a0ec
DK		 1651 if (es->s_last_orphan)
1652 jbd_debug(1, "Errors on filesystem, "
1653 "clearing orphan list.\n");
1654 es->s_last_orphan = 0;
1655 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1656 return;
1657 }
1658
1659 if (s_flags & MS_RDONLY) {
617ba13b 1660 printk(KERN_INFO "EXT4-fs: %s: orphan cleanup on readonly fs\n",
ac27a0ec
DK		 1661 sb->s_id);
1662 sb->s_flags &= ~MS_RDONLY;
1663 }
1664#ifdef CONFIG_QUOTA
1665 /* Needed for iput() to work correctly and not trash data */
1666 sb->s_flags |= MS_ACTIVE;
1667 /* Turn on quotas so that they are updated correctly */
1668 for (i = 0; i < MAXQUOTAS; i++) {
617ba13b
MC		 1669 if (EXT4_SB(sb)->s_qf_names[i]) {
1670 int ret = ext4_quota_on_mount(sb, i);
ac27a0ec
DK		 1671 if (ret < 0)
1672 printk(KERN_ERR
2c8be6b2 1673 "EXT4-fs: Cannot turn on journaled "
ac27a0ec
DK		 1674 "quota: error %d\n", ret);
1675 }
1676 }
1677#endif
1678
1679 while (es->s_last_orphan) {
1680 struct inode *inode;
1681
97bd42b9
JB		 1682 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1683 if (IS_ERR(inode)) {
ac27a0ec
DK		 1684 es->s_last_orphan = 0;
1685 break;
1686 }
1687
617ba13b 1688 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
ac27a0ec
DK		 1689 DQUOT_INIT(inode);
1690 if (inode->i_nlink) {
1691 printk(KERN_DEBUG
e5f8eab8 1692 "%s: truncating inode %lu to %lld bytes\n",
46e665e9 1693 __func__, inode->i_ino, inode->i_size);
e5f8eab8 1694 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
ac27a0ec 1695 inode->i_ino, inode->i_size);
617ba13b 1696 ext4_truncate(inode);
ac27a0ec
DK		 1697 nr_truncates++;
1698 } else {
1699 printk(KERN_DEBUG
1700 "%s: deleting unreferenced inode %lu\n",
46e665e9 1701 __func__, inode->i_ino);
ac27a0ec
DK		 1702 jbd_debug(2, "deleting unreferenced inode %lu\n",
1703 inode->i_ino);
1704 nr_orphans++;
1705 }
1706 iput(inode); /* The delete magic happens here! */
1707 }
1708
2b2d6d01 1709#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
ac27a0ec
DK		 1710
1711 if (nr_orphans)
617ba13b 1712 printk(KERN_INFO "EXT4-fs: %s: %d orphan inode%s deleted\n",
ac27a0ec
DK		 1713 sb->s_id, PLURAL(nr_orphans));
1714 if (nr_truncates)
617ba13b 1715 printk(KERN_INFO "EXT4-fs: %s: %d truncate%s cleaned up\n",
ac27a0ec
DK		 1716 sb->s_id, PLURAL(nr_truncates));
1717#ifdef CONFIG_QUOTA
1718 /* Turn quotas off */
1719 for (i = 0; i < MAXQUOTAS; i++) {
1720 if (sb_dqopt(sb)->files[i])
6f28e087 1721 vfs_quota_off(sb, i, 0);
ac27a0ec
DK		 1722 }
1723#endif
1724 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1725}
cd2291a4
ES		 1726/*
1727 * Maximal extent format file size.
1728 * Resulting logical blkno at s_maxbytes must fit in our on-disk
1729 * extent format containers, within a sector_t, and within i_blocks
1730 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
1731 * so that won't be a limiting factor.
1732 *
1733 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1734 */
f287a1a5 1735static loff_t ext4_max_size(int blkbits, int has_huge_files)
cd2291a4
ES		 1736{
1737 loff_t res;
1738 loff_t upper_limit = MAX_LFS_FILESIZE;
1739
1740 /* small i_blocks in vfs inode? */
f287a1a5 1741 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
cd2291a4 1742 /*
b3a6ffe1 1743 * CONFIG_LBD is not enabled implies the inode
cd2291a4
ES		 1744 * i_block represent total blocks in 512 bytes
1745 * 32 == size of vfs inode i_blocks * 8
1746 */
1747 upper_limit = (1LL << 32) - 1;
1748
1749 /* total blocks in file system block size */
1750 upper_limit >>= (blkbits - 9);
1751 upper_limit <<= blkbits;
1752 }
1753
1754 /* 32-bit extent-start container, ee_block */
1755 res = 1LL << 32;
1756 res <<= blkbits;
1757 res -= 1;
1758
1759 /* Sanity check against vm- & vfs- imposed limits */
1760 if (res > upper_limit)
1761 res = upper_limit;
1762
1763 return res;
1764}
ac27a0ec 1765
ac27a0ec 1766/*
cd2291a4 1767 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 1768 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
1769 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 1770 */
f287a1a5 1771static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
ac27a0ec 1772{
617ba13b 1773 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 1774 int meta_blocks;
1775 loff_t upper_limit;
1776 /* This is calculated to be the largest file size for a
cd2291a4 1777 * dense, bitmapped file such that the total number of
ac27a0ec 1778 * sectors in the file, including data and all indirect blocks,
0fc1b451
AK		 1779 * does not exceed 2^48 -1
1780 * __u32 i_blocks_lo and _u16 i_blocks_high representing the
1781 * total number of 512 bytes blocks of the file
1782 */
1783
f287a1a5 1784 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
0fc1b451 1785 /*
b3a6ffe1 1786 * !has_huge_files or CONFIG_LBD is not enabled
f287a1a5
TT		 1787 * implies the inode i_block represent total blocks in
1788 * 512 bytes 32 == size of vfs inode i_blocks * 8
0fc1b451
AK		 1789 */
1790 upper_limit = (1LL << 32) - 1;
1791
1792 /* total blocks in file system block size */
1793 upper_limit >>= (bits - 9);
1794
1795 } else {
8180a562
AK		 1796 /*
1797 * We use 48 bit ext4_inode i_blocks
1798 * With EXT4_HUGE_FILE_FL set the i_blocks
1799 * represent total number of blocks in
1800 * file system block size
1801 */
0fc1b451
AK		 1802 upper_limit = (1LL << 48) - 1;
1803
0fc1b451
AK		 1804 }
1805
1806 /* indirect blocks */
1807 meta_blocks = 1;
1808 /* double indirect blocks */
1809 meta_blocks += 1 + (1LL << (bits-2));
1810 /* tripple indirect blocks */
1811 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
1812
1813 upper_limit -= meta_blocks;
1814 upper_limit <<= bits;
ac27a0ec
DK		 1815
1816 res += 1LL << (bits-2);
1817 res += 1LL << (2*(bits-2));
1818 res += 1LL << (3*(bits-2));
1819 res <<= bits;
1820 if (res > upper_limit)
1821 res = upper_limit;
0fc1b451
AK		 1822
1823 if (res > MAX_LFS_FILESIZE)
1824 res = MAX_LFS_FILESIZE;
1825
ac27a0ec
DK		 1826 return res;
1827}
1828
617ba13b 1829static ext4_fsblk_t descriptor_loc(struct super_block *sb,
70bbb3e0 1830 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 1831{
617ba13b 1832 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 1833 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 1834 int has_super = 0;
1835
1836 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1837
617ba13b 1838 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
ac27a0ec 1839 nr < first_meta_bg)
70bbb3e0 1840 return logical_sb_block + nr + 1;
ac27a0ec 1841 bg = sbi->s_desc_per_block * nr;
617ba13b 1842 if (ext4_bg_has_super(sb, bg))
ac27a0ec 1843 has_super = 1;
617ba13b 1844 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 1845}
1846
c9de560d
AT		 1847/**
1848 * ext4_get_stripe_size: Get the stripe size.
1849 * @sbi: In memory super block info
1850 *
1851 * If we have specified it via mount option, then
1852 * use the mount option value. If the value specified at mount time is
1853 * greater than the blocks per group use the super block value.
1854 * If the super block value is greater than blocks per group return 0.
1855 * Allocator needs it be less than blocks per group.
1856 *
1857 */
1858static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
1859{
1860 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
1861 unsigned long stripe_width =
1862 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
1863
1864 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
1865 return sbi->s_stripe;
1866
1867 if (stripe_width <= sbi->s_blocks_per_group)
1868 return stripe_width;
1869
1870 if (stride <= sbi->s_blocks_per_group)
1871 return stride;
1872
1873 return 0;
1874}
ac27a0ec 1875
2b2d6d01 1876static int ext4_fill_super(struct super_block *sb, void *data, int silent)
7477827f
AK		 1877 __releases(kernel_lock)
1878 __acquires(kernel_lock)
1d03ec98 1879
ac27a0ec 1880{
2b2d6d01 1881 struct buffer_head *bh;
617ba13b
MC		 1882 struct ext4_super_block *es = NULL;
1883 struct ext4_sb_info *sbi;
1884 ext4_fsblk_t block;
1885 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 1886 ext4_fsblk_t logical_sb_block;
ac27a0ec
DK		 1887 unsigned long offset = 0;
1888 unsigned int journal_inum = 0;
1889 unsigned long journal_devnum = 0;
1890 unsigned long def_mount_opts;
1891 struct inode *root;
9f6200bb 1892 char *cp;
0390131b 1893 const char *descr;
1d1fe1ee 1894 int ret = -EINVAL;
ac27a0ec 1895 int blocksize;
ac27a0ec
DK		 1896 int db_count;
1897 int i;
f287a1a5 1898 int needs_recovery, has_huge_files;
ac27a0ec 1899 __le32 features;
bd81d8ee 1900 __u64 blocks_count;
833f4077 1901 int err;
ac27a0ec
DK		 1902
1903 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
1904 if (!sbi)
1905 return -ENOMEM;
1906 sb->s_fs_info = sbi;
1907 sbi->s_mount_opt = 0;
617ba13b
MC		 1908 sbi->s_resuid = EXT4_DEF_RESUID;
1909 sbi->s_resgid = EXT4_DEF_RESGID;
240799cd 1910 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
d9c9bef1 1911 sbi->s_sb_block = sb_block;
ac27a0ec
DK		 1912
1913 unlock_kernel();
1914
9f6200bb
TT		 1915 /* Cleanup superblock name */
1916 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
1917 *cp = '!';
1918
617ba13b 1919 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 1920 if (!blocksize) {
617ba13b 1921 printk(KERN_ERR "EXT4-fs: unable to set blocksize\n");
ac27a0ec
DK		 1922 goto out_fail;
1923 }
1924
1925 /*
617ba13b 1926 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 1927 * block sizes. We need to calculate the offset from buffer start.
1928 */
617ba13b 1929 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 1930 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
1931 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 1932 } else {
70bbb3e0 1933 logical_sb_block = sb_block;
ac27a0ec
DK		 1934 }
1935
70bbb3e0 1936 if (!(bh = sb_bread(sb, logical_sb_block))) {
2b2d6d01 1937 printk(KERN_ERR "EXT4-fs: unable to read superblock\n");
ac27a0ec
DK		 1938 goto out_fail;
1939 }
1940 /*
1941 * Note: s_es must be initialized as soon as possible because
617ba13b 1942 * some ext4 macro-instructions depend on its value
ac27a0ec 1943 */
617ba13b 1944 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
ac27a0ec
DK		 1945 sbi->s_es = es;
1946 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 1947 if (sb->s_magic != EXT4_SUPER_MAGIC)
1948 goto cantfind_ext4;
ac27a0ec
DK		 1949
1950 /* Set defaults before we parse the mount options */
1951 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
617ba13b 1952 if (def_mount_opts & EXT4_DEFM_DEBUG)
ac27a0ec 1953 set_opt(sbi->s_mount_opt, DEBUG);
617ba13b 1954 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
ac27a0ec 1955 set_opt(sbi->s_mount_opt, GRPID);
617ba13b 1956 if (def_mount_opts & EXT4_DEFM_UID16)
ac27a0ec 1957 set_opt(sbi->s_mount_opt, NO_UID32);
03010a33 1958#ifdef CONFIG_EXT4_FS_XATTR
617ba13b 1959 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
ac27a0ec 1960 set_opt(sbi->s_mount_opt, XATTR_USER);
2e7842b8 1961#endif
03010a33 1962#ifdef CONFIG_EXT4_FS_POSIX_ACL
617ba13b 1963 if (def_mount_opts & EXT4_DEFM_ACL)
ac27a0ec 1964 set_opt(sbi->s_mount_opt, POSIX_ACL);
2e7842b8 1965#endif
617ba13b
MC		 1966 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
1967 sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
1968 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
1969 sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
1970 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
1971 sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
1972
1973 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
ac27a0ec 1974 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
bb4f397a 1975 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
ceea16bf 1976 set_opt(sbi->s_mount_opt, ERRORS_CONT);
bb4f397a
AK		 1977 else
1978 set_opt(sbi->s_mount_opt, ERRORS_RO);
ac27a0ec
DK		 1979
1980 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
1981 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
1982
1983 set_opt(sbi->s_mount_opt, RESERVATION);
571640ca 1984 set_opt(sbi->s_mount_opt, BARRIER);
ac27a0ec 1985
1e2462f9
MC		 1986 /*
1987 * turn on extents feature by default in ext4 filesystem
e4079a11
ES		 1988 * only if feature flag already set by mkfs or tune2fs.
1989 * Use -o noextents to turn it off
1e2462f9 1990 */
e4079a11
ES		 1991 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS))
1992 set_opt(sbi->s_mount_opt, EXTENTS);
1993 else
1994 ext4_warning(sb, __func__,
1995 "extents feature not enabled on this filesystem, "
fde4d95a 1996 "use tune2fs.");
1e2462f9 1997
dd919b98
AK		 1998 /*
1999 * enable delayed allocation by default
2000 * Use -o nodelalloc to turn it off
2001 */
2002 set_opt(sbi->s_mount_opt, DELALLOC);
2003
2004
2b2d6d01
TT		 2005 if (!parse_options((char *) data, sb, &journal_inum, &journal_devnum,
2006 NULL, 0))
ac27a0ec
DK		 2007 goto failed_mount;
2008
2009 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
617ba13b 2010 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec 2011
617ba13b
MC		 2012 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
2013 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
2014 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
2015 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
ac27a0ec 2016 printk(KERN_WARNING
617ba13b 2017 "EXT4-fs warning: feature flags set on rev 0 fs, "
ac27a0ec 2018 "running e2fsck is recommended\n");
469108ff 2019
ac27a0ec
DK		 2020 /*
2021 * Check feature flags regardless of the revision level, since we
2022 * previously didn't change the revision level when setting the flags,
2023 * so there is a chance incompat flags are set on a rev 0 filesystem.
2024 */
617ba13b 2025 features = EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP);
ac27a0ec 2026 if (features) {
617ba13b 2027 printk(KERN_ERR "EXT4-fs: %s: couldn't mount because of "
ac27a0ec
DK		 2028 "unsupported optional features (%x).\n",
2029 sb->s_id, le32_to_cpu(features));
2030 goto failed_mount;
2031 }
617ba13b 2032 features = EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP);
ac27a0ec 2033 if (!(sb->s_flags & MS_RDONLY) && features) {
617ba13b 2034 printk(KERN_ERR "EXT4-fs: %s: couldn't mount RDWR because of "
ac27a0ec
DK		 2035 "unsupported optional features (%x).\n",
2036 sb->s_id, le32_to_cpu(features));
2037 goto failed_mount;
2038 }
f287a1a5
TT		 2039 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2040 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
2041 if (has_huge_files) {
0fc1b451
AK		 2042 /*
2043 * Large file size enabled file system can only be
b3a6ffe1 2044 * mount if kernel is build with CONFIG_LBD
0fc1b451
AK		 2045 */
2046 if (sizeof(root->i_blocks) < sizeof(u64) &&
2047 !(sb->s_flags & MS_RDONLY)) {
2048 printk(KERN_ERR "EXT4-fs: %s: Filesystem with huge "
2049 "files cannot be mounted read-write "
b3a6ffe1 2050 "without CONFIG_LBD.\n", sb->s_id);
0fc1b451
AK		 2051 goto failed_mount;
2052 }
2053 }
ac27a0ec
DK		 2054 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
2055
617ba13b
MC		 2056 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
2057 blocksize > EXT4_MAX_BLOCK_SIZE) {
ac27a0ec 2058 printk(KERN_ERR
617ba13b 2059 "EXT4-fs: Unsupported filesystem blocksize %d on %s.\n",
ac27a0ec
DK		 2060 blocksize, sb->s_id);
2061 goto failed_mount;
2062 }
2063
ac27a0ec 2064 if (sb->s_blocksize != blocksize) {
ce40733c
AK		 2065
2066 /* Validate the filesystem blocksize */
2067 if (!sb_set_blocksize(sb, blocksize)) {
2068 printk(KERN_ERR "EXT4-fs: bad block size %d.\n",
2069 blocksize);
ac27a0ec
DK		 2070 goto failed_mount;
2071 }
2072
2b2d6d01 2073 brelse(bh);
70bbb3e0
AM		 2074 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2075 offset = do_div(logical_sb_block, blocksize);
2076 bh = sb_bread(sb, logical_sb_block);
ac27a0ec
DK		 2077 if (!bh) {
2078 printk(KERN_ERR
617ba13b 2079 "EXT4-fs: Can't read superblock on 2nd try.\n");
ac27a0ec
DK		 2080 goto failed_mount;
2081 }
617ba13b 2082 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
ac27a0ec 2083 sbi->s_es = es;
617ba13b 2084 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
2b2d6d01
TT		 2085 printk(KERN_ERR
2086 "EXT4-fs: Magic mismatch, very weird !\n");
ac27a0ec
DK		 2087 goto failed_mount;
2088 }
2089 }
2090
f287a1a5
TT		 2091 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
2092 has_huge_files);
2093 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
ac27a0ec 2094
617ba13b
MC		 2095 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2096 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2097 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
ac27a0ec
DK		 2098 } else {
2099 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2100 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
617ba13b 2101 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1330593e 2102 (!is_power_of_2(sbi->s_inode_size)) ||
ac27a0ec 2103 (sbi->s_inode_size > blocksize)) {
2b2d6d01
TT		 2104 printk(KERN_ERR
2105 "EXT4-fs: unsupported inode size: %d\n",
2106 sbi->s_inode_size);
ac27a0ec
DK		 2107 goto failed_mount;
2108 }
ef7f3835
KS		 2109 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2110 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
ac27a0ec 2111 }
0d1ee42f
AR		 2112 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2113 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
8fadc143 2114 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 2115 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 2116 !is_power_of_2(sbi->s_desc_size)) {
0d1ee42f 2117 printk(KERN_ERR
8fadc143 2118 "EXT4-fs: unsupported descriptor size %lu\n",
0d1ee42f
AR		 2119 sbi->s_desc_size);
2120 goto failed_mount;
2121 }
2122 } else
2123 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
ac27a0ec 2124 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 2125 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
b47b6f38 2126 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
617ba13b
MC		 2127 goto cantfind_ext4;
2128 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 2129 if (sbi->s_inodes_per_block == 0)
617ba13b 2130 goto cantfind_ext4;
ac27a0ec
DK		 2131 sbi->s_itb_per_group = sbi->s_inodes_per_group /
2132 sbi->s_inodes_per_block;
0d1ee42f 2133 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 2134 sbi->s_sbh = bh;
2135 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 2136 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2137 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
2b2d6d01 2138 for (i = 0; i < 4; i++)
ac27a0ec
DK		 2139 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2140 sbi->s_def_hash_version = es->s_def_hash_version;
f99b2589
TT		 2141 i = le32_to_cpu(es->s_flags);
2142 if (i & EXT2_FLAGS_UNSIGNED_HASH)
2143 sbi->s_hash_unsigned = 3;
2144 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
2145#ifdef __CHAR_UNSIGNED__
2146 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
2147 sbi->s_hash_unsigned = 3;
2148#else
2149 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
2150#endif
2151 sb->s_dirt = 1;
2152 }
ac27a0ec
DK		 2153
2154 if (sbi->s_blocks_per_group > blocksize * 8) {
2b2d6d01
TT		 2155 printk(KERN_ERR
2156 "EXT4-fs: #blocks per group too big: %lu\n",
2157 sbi->s_blocks_per_group);
ac27a0ec
DK		 2158 goto failed_mount;
2159 }
ac27a0ec 2160 if (sbi->s_inodes_per_group > blocksize * 8) {
2b2d6d01
TT		 2161 printk(KERN_ERR
2162 "EXT4-fs: #inodes per group too big: %lu\n",
2163 sbi->s_inodes_per_group);
ac27a0ec
DK		 2164 goto failed_mount;
2165 }
2166
bd81d8ee 2167 if (ext4_blocks_count(es) >
ac27a0ec 2168 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
617ba13b 2169 printk(KERN_ERR "EXT4-fs: filesystem on %s:"
ac27a0ec
DK		 2170 " too large to mount safely\n", sb->s_id);
2171 if (sizeof(sector_t) < 8)
617ba13b 2172 printk(KERN_WARNING "EXT4-fs: CONFIG_LBD not "
ac27a0ec
DK		 2173 "enabled\n");
2174 goto failed_mount;
2175 }
2176
617ba13b
MC		 2177 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2178 goto cantfind_ext4;
e7c95593
ES		 2179
2180 /* ensure blocks_count calculation below doesn't sign-extend */
2181 if (ext4_blocks_count(es) + EXT4_BLOCKS_PER_GROUP(sb) <
2182 le32_to_cpu(es->s_first_data_block) + 1) {
2183 printk(KERN_WARNING "EXT4-fs: bad geometry: block count %llu, "
2184 "first data block %u, blocks per group %lu\n",
2185 ext4_blocks_count(es),
2186 le32_to_cpu(es->s_first_data_block),
2187 EXT4_BLOCKS_PER_GROUP(sb));
2188 goto failed_mount;
2189 }
bd81d8ee
LV		 2190 blocks_count = (ext4_blocks_count(es) -
2191 le32_to_cpu(es->s_first_data_block) +
2192 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2193 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
2194 sbi->s_groups_count = blocks_count;
617ba13b
MC		 2195 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2196 EXT4_DESC_PER_BLOCK(sb);
2b2d6d01 2197 sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
ac27a0ec
DK		 2198 GFP_KERNEL);
2199 if (sbi->s_group_desc == NULL) {
2b2d6d01 2200 printk(KERN_ERR "EXT4-fs: not enough memory\n");
ac27a0ec
DK		 2201 goto failed_mount;
2202 }
2203
3244fcb1 2204#ifdef CONFIG_PROC_FS
9f6200bb
TT		 2205 if (ext4_proc_root)
2206 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
2207
240799cd
TT		 2208 if (sbi->s_proc)
2209 proc_create_data("inode_readahead_blks", 0644, sbi->s_proc,
2210 &ext4_ui_proc_fops,
2211 &sbi->s_inode_readahead_blks);
3244fcb1 2212#endif
240799cd 2213
ac27a0ec
DK		 2214 bgl_lock_init(&sbi->s_blockgroup_lock);
2215
2216 for (i = 0; i < db_count; i++) {
70bbb3e0 2217 block = descriptor_loc(sb, logical_sb_block, i);
ac27a0ec
DK		 2218 sbi->s_group_desc[i] = sb_bread(sb, block);
2219 if (!sbi->s_group_desc[i]) {
2b2d6d01
TT		 2220 printk(KERN_ERR "EXT4-fs: "
2221 "can't read group descriptor %d\n", i);
ac27a0ec
DK		 2222 db_count = i;
2223 goto failed_mount2;
2224 }
2225 }
2b2d6d01 2226 if (!ext4_check_descriptors(sb)) {
617ba13b 2227 printk(KERN_ERR "EXT4-fs: group descriptors corrupted!\n");
ac27a0ec
DK		 2228 goto failed_mount2;
2229 }
772cb7c8
JS		 2230 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2231 if (!ext4_fill_flex_info(sb)) {
2232 printk(KERN_ERR
2233 "EXT4-fs: unable to initialize "
2234 "flex_bg meta info!\n");
2235 goto failed_mount2;
2236 }
2237
ac27a0ec
DK		 2238 sbi->s_gdb_count = db_count;
2239 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2240 spin_lock_init(&sbi->s_next_gen_lock);
2241
833f4077
PZ		 2242 err = percpu_counter_init(&sbi->s_freeblocks_counter,
2243 ext4_count_free_blocks(sb));
2244 if (!err) {
2245 err = percpu_counter_init(&sbi->s_freeinodes_counter,
2246 ext4_count_free_inodes(sb));
2247 }
2248 if (!err) {
2249 err = percpu_counter_init(&sbi->s_dirs_counter,
2250 ext4_count_dirs(sb));
2251 }
6bc6e63f
AK		 2252 if (!err) {
2253 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
2254 }
833f4077
PZ		 2255 if (err) {
2256 printk(KERN_ERR "EXT4-fs: insufficient memory\n");
2257 goto failed_mount3;
2258 }
ac27a0ec 2259
c9de560d
AT		 2260 sbi->s_stripe = ext4_get_stripe_size(sbi);
2261
ac27a0ec
DK		 2262 /*
2263 * set up enough so that it can read an inode
2264 */
617ba13b
MC		 2265 sb->s_op = &ext4_sops;
2266 sb->s_export_op = &ext4_export_ops;
2267 sb->s_xattr = ext4_xattr_handlers;
ac27a0ec 2268#ifdef CONFIG_QUOTA
617ba13b
MC		 2269 sb->s_qcop = &ext4_qctl_operations;
2270 sb->dq_op = &ext4_quota_operations;
ac27a0ec
DK		 2271#endif
2272 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
2273
2274 sb->s_root = NULL;
2275
2276 needs_recovery = (es->s_last_orphan != 0 ||
617ba13b
MC		 2277 EXT4_HAS_INCOMPAT_FEATURE(sb,
2278 EXT4_FEATURE_INCOMPAT_RECOVER));
ac27a0ec
DK		 2279
2280 /*
2281 * The first inode we look at is the journal inode. Don't try
2282 * root first: it may be modified in the journal!
2283 */
2284 if (!test_opt(sb, NOLOAD) &&
617ba13b
MC		 2285 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2286 if (ext4_load_journal(sb, es, journal_devnum))
ac27a0ec 2287 goto failed_mount3;
624080ed
TT		 2288 if (!(sb->s_flags & MS_RDONLY) &&
2289 EXT4_SB(sb)->s_journal->j_failed_commit) {
2290 printk(KERN_CRIT "EXT4-fs error (device %s): "
2291 "ext4_fill_super: Journal transaction "
2b2d6d01 2292 "%u is corrupt\n", sb->s_id,
624080ed 2293 EXT4_SB(sb)->s_journal->j_failed_commit);
2b2d6d01
TT		 2294 if (test_opt(sb, ERRORS_RO)) {
2295 printk(KERN_CRIT
2296 "Mounting filesystem read-only\n");
624080ed
TT		 2297 sb->s_flags |= MS_RDONLY;
2298 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2299 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2300 }
2301 if (test_opt(sb, ERRORS_PANIC)) {
2302 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2303 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2304 ext4_commit_super(sb, es, 1);
624080ed
TT		 2305 goto failed_mount4;
2306 }
2307 }
ac27a0ec 2308 } else if (journal_inum) {
617ba13b 2309 if (ext4_create_journal(sb, es, journal_inum))
ac27a0ec 2310 goto failed_mount3;
0390131b
FM		 2311 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
2312 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
2313 printk(KERN_ERR "EXT4-fs: required journal recovery "
2314 "suppressed and not mounted read-only\n");
2315 goto failed_mount4;
ac27a0ec 2316 } else {
0390131b
FM		 2317 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
2318 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
2319 sbi->s_journal = NULL;
2320 needs_recovery = 0;
2321 goto no_journal;
ac27a0ec
DK		 2322 }
2323
eb40a09c
JS		 2324 if (ext4_blocks_count(es) > 0xffffffffULL &&
2325 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2326 JBD2_FEATURE_INCOMPAT_64BIT)) {
2327 printk(KERN_ERR "ext4: Failed to set 64-bit journal feature\n");
2328 goto failed_mount4;
2329 }
2330
818d276c
GS		 2331 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2332 jbd2_journal_set_features(sbi->s_journal,
2333 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2334 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2335 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2336 jbd2_journal_set_features(sbi->s_journal,
2337 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2338 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2339 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2340 } else {
2341 jbd2_journal_clear_features(sbi->s_journal,
2342 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2343 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2344 }
2345
ac27a0ec
DK		 2346 /* We have now updated the journal if required, so we can
2347 * validate the data journaling mode. */
2348 switch (test_opt(sb, DATA_FLAGS)) {
2349 case 0:
2350 /* No mode set, assume a default based on the journal
63f57933
AM		 2351 * capabilities: ORDERED_DATA if the journal can
2352 * cope, else JOURNAL_DATA
2353 */
dab291af
MC		 2354 if (jbd2_journal_check_available_features
2355 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
ac27a0ec
DK		 2356 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2357 else
2358 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2359 break;
2360
617ba13b
MC		 2361 case EXT4_MOUNT_ORDERED_DATA:
2362 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 2363 if (!jbd2_journal_check_available_features
2364 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
617ba13b 2365 printk(KERN_ERR "EXT4-fs: Journal does not support "
ac27a0ec
DK		 2366 "requested data journaling mode\n");
2367 goto failed_mount4;
2368 }
2369 default:
2370 break;
2371 }
2372
0390131b
FM		 2373no_journal:
2374
ac27a0ec 2375 if (test_opt(sb, NOBH)) {
617ba13b
MC		 2376 if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
2377 printk(KERN_WARNING "EXT4-fs: Ignoring nobh option - "
ac27a0ec
DK		 2378 "its supported only with writeback mode\n");
2379 clear_opt(sbi->s_mount_opt, NOBH);
2380 }
2381 }
2382 /*
dab291af 2383 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 2384 * so we can safely mount the rest of the filesystem now.
2385 */
2386
1d1fe1ee
DH		 2387 root = ext4_iget(sb, EXT4_ROOT_INO);
2388 if (IS_ERR(root)) {
617ba13b 2389 printk(KERN_ERR "EXT4-fs: get root inode failed\n");
1d1fe1ee 2390 ret = PTR_ERR(root);
ac27a0ec
DK		 2391 goto failed_mount4;
2392 }
2393 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1d1fe1ee 2394 iput(root);
617ba13b 2395 printk(KERN_ERR "EXT4-fs: corrupt root inode, run e2fsck\n");
ac27a0ec
DK		 2396 goto failed_mount4;
2397 }
1d1fe1ee
DH		 2398 sb->s_root = d_alloc_root(root);
2399 if (!sb->s_root) {
2400 printk(KERN_ERR "EXT4-fs: get root dentry failed\n");
2401 iput(root);
2402 ret = -ENOMEM;
2403 goto failed_mount4;
2404 }
ac27a0ec 2405
2b2d6d01 2406 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
ef7f3835
KS		 2407
2408 /* determine the minimum size of new large inodes, if present */
2409 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2410 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2411 EXT4_GOOD_OLD_INODE_SIZE;
2412 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2413 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2414 if (sbi->s_want_extra_isize <
2415 le16_to_cpu(es->s_want_extra_isize))
2416 sbi->s_want_extra_isize =
2417 le16_to_cpu(es->s_want_extra_isize);
2418 if (sbi->s_want_extra_isize <
2419 le16_to_cpu(es->s_min_extra_isize))
2420 sbi->s_want_extra_isize =
2421 le16_to_cpu(es->s_min_extra_isize);
2422 }
2423 }
2424 /* Check if enough inode space is available */
2425 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2426 sbi->s_inode_size) {
2427 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2428 EXT4_GOOD_OLD_INODE_SIZE;
2429 printk(KERN_INFO "EXT4-fs: required extra inode space not"
2430 "available.\n");
2431 }
2432
c2774d84
AK		 2433 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
2434 printk(KERN_WARNING "EXT4-fs: Ignoring delalloc option - "
2435 "requested data journaling mode\n");
2436 clear_opt(sbi->s_mount_opt, DELALLOC);
2437 } else if (test_opt(sb, DELALLOC))
2438 printk(KERN_INFO "EXT4-fs: delayed allocation enabled\n");
2439
2440 ext4_ext_init(sb);
2441 err = ext4_mb_init(sb, needs_recovery);
2442 if (err) {
2443 printk(KERN_ERR "EXT4-fs: failed to initalize mballoc (%d)\n",
2444 err);
2445 goto failed_mount4;
2446 }
2447
ac27a0ec
DK		 2448 /*
2449 * akpm: core read_super() calls in here with the superblock locked.
2450 * That deadlocks, because orphan cleanup needs to lock the superblock
2451 * in numerous places. Here we just pop the lock - it's relatively
2452 * harmless, because we are now ready to accept write_super() requests,
2453 * and aviro says that's the only reason for hanging onto the
2454 * superblock lock.
2455 */
617ba13b
MC		 2456 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
2457 ext4_orphan_cleanup(sb, es);
2458 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
0390131b 2459 if (needs_recovery) {
2b2d6d01 2460 printk(KERN_INFO "EXT4-fs: recovery complete.\n");
0390131b
FM		 2461 ext4_mark_recovery_complete(sb, es);
2462 }
2463 if (EXT4_SB(sb)->s_journal) {
2464 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2465 descr = " journalled data mode";
2466 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2467 descr = " ordered data mode";
2468 else
2469 descr = " writeback data mode";
2470 } else
2471 descr = "out journal";
2472
2473 printk(KERN_INFO "EXT4-fs: mounted filesystem %s with%s\n",
2474 sb->s_id, descr);
ac27a0ec
DK		 2475
2476 lock_kernel();
2477 return 0;
2478
617ba13b 2479cantfind_ext4:
ac27a0ec 2480 if (!silent)
617ba13b 2481 printk(KERN_ERR "VFS: Can't find ext4 filesystem on dev %s.\n",
ac27a0ec
DK		 2482 sb->s_id);
2483 goto failed_mount;
2484
2485failed_mount4:
0390131b
FM		 2486 printk(KERN_ERR "EXT4-fs (device %s): mount failed\n", sb->s_id);
2487 if (sbi->s_journal) {
2488 jbd2_journal_destroy(sbi->s_journal);
2489 sbi->s_journal = NULL;
2490 }
ac27a0ec
DK		 2491failed_mount3:
2492 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2493 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2494 percpu_counter_destroy(&sbi->s_dirs_counter);
6bc6e63f 2495 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
ac27a0ec
DK		 2496failed_mount2:
2497 for (i = 0; i < db_count; i++)
2498 brelse(sbi->s_group_desc[i]);
2499 kfree(sbi->s_group_desc);
2500failed_mount:
240799cd
TT		 2501 if (sbi->s_proc) {
2502 remove_proc_entry("inode_readahead_blks", sbi->s_proc);
9f6200bb 2503 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 2504 }
ac27a0ec
DK		 2505#ifdef CONFIG_QUOTA
2506 for (i = 0; i < MAXQUOTAS; i++)
2507 kfree(sbi->s_qf_names[i]);
2508#endif
617ba13b 2509 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 2510 brelse(bh);
2511out_fail:
2512 sb->s_fs_info = NULL;
2513 kfree(sbi);
2514 lock_kernel();
1d1fe1ee 2515 return ret;
ac27a0ec
DK		 2516}
2517
2518/*
2519 * Setup any per-fs journal parameters now. We'll do this both on
2520 * initial mount, once the journal has been initialised but before we've
2521 * done any recovery; and again on any subsequent remount.
2522 */
617ba13b 2523static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 2524{
617ba13b 2525 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 2526
2527 if (sbi->s_commit_interval)
2528 journal->j_commit_interval = sbi->s_commit_interval;
617ba13b 2529 /* We could also set up an ext4-specific default for the commit
ac27a0ec
DK		 2530 * interval here, but for now we'll just fall back to the jbd
2531 * default. */
2532
2533 spin_lock(&journal->j_state_lock);
2534 if (test_opt(sb, BARRIER))
dab291af 2535 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 2536 else
dab291af 2537 journal->j_flags &= ~JBD2_BARRIER;
5bf5683a
HK		 2538 if (test_opt(sb, DATA_ERR_ABORT))
2539 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
2540 else
2541 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
ac27a0ec
DK		 2542 spin_unlock(&journal->j_state_lock);
2543}
2544
617ba13b 2545static journal_t *ext4_get_journal(struct super_block *sb,
ac27a0ec
DK		 2546 unsigned int journal_inum)
2547{
2548 struct inode *journal_inode;
2549 journal_t *journal;
2550
0390131b
FM		 2551 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
2552
ac27a0ec
DK		 2553 /* First, test for the existence of a valid inode on disk. Bad
2554 * things happen if we iget() an unused inode, as the subsequent
2555 * iput() will try to delete it. */
2556
1d1fe1ee
DH		 2557 journal_inode = ext4_iget(sb, journal_inum);
2558 if (IS_ERR(journal_inode)) {
617ba13b 2559 printk(KERN_ERR "EXT4-fs: no journal found.\n");
ac27a0ec
DK		 2560 return NULL;
2561 }
2562 if (!journal_inode->i_nlink) {
2563 make_bad_inode(journal_inode);
2564 iput(journal_inode);
617ba13b 2565 printk(KERN_ERR "EXT4-fs: journal inode is deleted.\n");
ac27a0ec
DK		 2566 return NULL;
2567 }
2568
e5f8eab8 2569 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
ac27a0ec 2570 journal_inode, journal_inode->i_size);
1d1fe1ee 2571 if (!S_ISREG(journal_inode->i_mode)) {
617ba13b 2572 printk(KERN_ERR "EXT4-fs: invalid journal inode.\n");
ac27a0ec
DK		 2573 iput(journal_inode);
2574 return NULL;
2575 }
2576
dab291af 2577 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 2578 if (!journal) {
617ba13b 2579 printk(KERN_ERR "EXT4-fs: Could not load journal inode\n");
ac27a0ec
DK		 2580 iput(journal_inode);
2581 return NULL;
2582 }
2583 journal->j_private = sb;
617ba13b 2584 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 2585 return journal;
2586}
2587
617ba13b 2588static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 2589 dev_t j_dev)
2590{
2b2d6d01 2591 struct buffer_head *bh;
ac27a0ec 2592 journal_t *journal;
617ba13b
MC		 2593 ext4_fsblk_t start;
2594 ext4_fsblk_t len;
ac27a0ec 2595 int hblock, blocksize;
617ba13b 2596 ext4_fsblk_t sb_block;
ac27a0ec 2597 unsigned long offset;
2b2d6d01 2598 struct ext4_super_block *es;
ac27a0ec
DK		 2599 struct block_device *bdev;
2600
0390131b
FM		 2601 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
2602
617ba13b 2603 bdev = ext4_blkdev_get(j_dev);
ac27a0ec
DK		 2604 if (bdev == NULL)
2605 return NULL;
2606
2607 if (bd_claim(bdev, sb)) {
2608 printk(KERN_ERR
8c55e204 2609 "EXT4: failed to claim external journal device.\n");
9a1c3542 2610 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
ac27a0ec
DK		 2611 return NULL;
2612 }
2613
2614 blocksize = sb->s_blocksize;
2615 hblock = bdev_hardsect_size(bdev);
2616 if (blocksize < hblock) {
2617 printk(KERN_ERR
617ba13b 2618 "EXT4-fs: blocksize too small for journal device.\n");
ac27a0ec
DK		 2619 goto out_bdev;
2620 }
2621
617ba13b
MC		 2622 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
2623 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 2624 set_blocksize(bdev, blocksize);
2625 if (!(bh = __bread(bdev, sb_block, blocksize))) {
617ba13b 2626 printk(KERN_ERR "EXT4-fs: couldn't read superblock of "
ac27a0ec
DK		 2627 "external journal\n");
2628 goto out_bdev;
2629 }
2630
617ba13b
MC		 2631 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2632 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 2633 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b
MC		 2634 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
2635 printk(KERN_ERR "EXT4-fs: external journal has "
ac27a0ec
DK		 2636 "bad superblock\n");
2637 brelse(bh);
2638 goto out_bdev;
2639 }
2640
617ba13b
MC		 2641 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
2642 printk(KERN_ERR "EXT4-fs: journal UUID does not match\n");
ac27a0ec
DK		 2643 brelse(bh);
2644 goto out_bdev;
2645 }
2646
bd81d8ee 2647 len = ext4_blocks_count(es);
ac27a0ec
DK		 2648 start = sb_block + 1;
2649 brelse(bh); /* we're done with the superblock */
2650
dab291af 2651 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 2652 start, len, blocksize);
2653 if (!journal) {
617ba13b 2654 printk(KERN_ERR "EXT4-fs: failed to create device journal\n");
ac27a0ec
DK		 2655 goto out_bdev;
2656 }
2657 journal->j_private = sb;
2658 ll_rw_block(READ, 1, &journal->j_sb_buffer);
2659 wait_on_buffer(journal->j_sb_buffer);
2660 if (!buffer_uptodate(journal->j_sb_buffer)) {
617ba13b 2661 printk(KERN_ERR "EXT4-fs: I/O error on journal device\n");
ac27a0ec
DK		 2662 goto out_journal;
2663 }
2664 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
617ba13b 2665 printk(KERN_ERR "EXT4-fs: External journal has more than one "
ac27a0ec
DK		 2666 "user (unsupported) - %d\n",
2667 be32_to_cpu(journal->j_superblock->s_nr_users));
2668 goto out_journal;
2669 }
617ba13b
MC		 2670 EXT4_SB(sb)->journal_bdev = bdev;
2671 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 2672 return journal;
2673out_journal:
dab291af 2674 jbd2_journal_destroy(journal);
ac27a0ec 2675out_bdev:
617ba13b 2676 ext4_blkdev_put(bdev);
ac27a0ec
DK		 2677 return NULL;
2678}
2679
617ba13b
MC		 2680static int ext4_load_journal(struct super_block *sb,
2681 struct ext4_super_block *es,
ac27a0ec
DK		 2682 unsigned long journal_devnum)
2683{
2684 journal_t *journal;
2685 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
2686 dev_t journal_dev;
2687 int err = 0;
2688 int really_read_only;
2689
0390131b
FM		 2690 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
2691
ac27a0ec
DK		 2692 if (journal_devnum &&
2693 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
617ba13b 2694 printk(KERN_INFO "EXT4-fs: external journal device major/minor "
ac27a0ec
DK		 2695 "numbers have changed\n");
2696 journal_dev = new_decode_dev(journal_devnum);
2697 } else
2698 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
2699
2700 really_read_only = bdev_read_only(sb->s_bdev);
2701
2702 /*
2703 * Are we loading a blank journal or performing recovery after a
2704 * crash? For recovery, we need to check in advance whether we
2705 * can get read-write access to the device.
2706 */
2707
617ba13b 2708 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
ac27a0ec 2709 if (sb->s_flags & MS_RDONLY) {
617ba13b 2710 printk(KERN_INFO "EXT4-fs: INFO: recovery "
ac27a0ec
DK		 2711 "required on readonly filesystem.\n");
2712 if (really_read_only) {
617ba13b 2713 printk(KERN_ERR "EXT4-fs: write access "
ac27a0ec
DK		 2714 "unavailable, cannot proceed.\n");
2715 return -EROFS;
2716 }
2b2d6d01
TT		 2717 printk(KERN_INFO "EXT4-fs: write access will "
2718 "be enabled during recovery.\n");
ac27a0ec
DK		 2719 }
2720 }
2721
2722 if (journal_inum && journal_dev) {
617ba13b 2723 printk(KERN_ERR "EXT4-fs: filesystem has both journal "
ac27a0ec
DK		 2724 "and inode journals!\n");
2725 return -EINVAL;
2726 }
2727
2728 if (journal_inum) {
617ba13b 2729 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 2730 return -EINVAL;
2731 } else {
617ba13b 2732 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 2733 return -EINVAL;
2734 }
2735
4776004f
TT		 2736 if (journal->j_flags & JBD2_BARRIER)
2737 printk(KERN_INFO "EXT4-fs: barriers enabled\n");
2738 else
2739 printk(KERN_INFO "EXT4-fs: barriers disabled\n");
2740
ac27a0ec 2741 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
dab291af 2742 err = jbd2_journal_update_format(journal);
ac27a0ec 2743 if (err) {
617ba13b 2744 printk(KERN_ERR "EXT4-fs: error updating journal.\n");
dab291af 2745 jbd2_journal_destroy(journal);
ac27a0ec
DK		 2746 return err;
2747 }
2748 }
2749
617ba13b 2750 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
dab291af 2751 err = jbd2_journal_wipe(journal, !really_read_only);
ac27a0ec 2752 if (!err)
dab291af 2753 err = jbd2_journal_load(journal);
ac27a0ec
DK		 2754
2755 if (err) {
617ba13b 2756 printk(KERN_ERR "EXT4-fs: error loading journal.\n");
dab291af 2757 jbd2_journal_destroy(journal);
ac27a0ec
DK		 2758 return err;
2759 }
2760
617ba13b
MC		 2761 EXT4_SB(sb)->s_journal = journal;
2762 ext4_clear_journal_err(sb, es);
ac27a0ec
DK		 2763
2764 if (journal_devnum &&
2765 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2766 es->s_journal_dev = cpu_to_le32(journal_devnum);
2767 sb->s_dirt = 1;
2768
2769 /* Make sure we flush the recovery flag to disk. */
617ba13b 2770 ext4_commit_super(sb, es, 1);
ac27a0ec
DK		 2771 }
2772
2773 return 0;
2774}
2775
2b2d6d01
TT		 2776static int ext4_create_journal(struct super_block *sb,
2777 struct ext4_super_block *es,
ac27a0ec
DK		 2778 unsigned int journal_inum)
2779{
2780 journal_t *journal;
6c675bd4 2781 int err;
ac27a0ec
DK		 2782
2783 if (sb->s_flags & MS_RDONLY) {
617ba13b 2784 printk(KERN_ERR "EXT4-fs: readonly filesystem when trying to "
ac27a0ec
DK		 2785 "create journal.\n");
2786 return -EROFS;
2787 }
2788
6c675bd4
BP		 2789 journal = ext4_get_journal(sb, journal_inum);
2790 if (!journal)
ac27a0ec
DK		 2791 return -EINVAL;
2792
617ba13b 2793 printk(KERN_INFO "EXT4-fs: creating new journal on inode %u\n",
ac27a0ec
DK		 2794 journal_inum);
2795
6c675bd4
BP		 2796 err = jbd2_journal_create(journal);
2797 if (err) {
617ba13b 2798 printk(KERN_ERR "EXT4-fs: error creating journal.\n");
dab291af 2799 jbd2_journal_destroy(journal);
ac27a0ec
DK		 2800 return -EIO;
2801 }
2802
617ba13b 2803 EXT4_SB(sb)->s_journal = journal;
ac27a0ec 2804
617ba13b
MC		 2805 ext4_update_dynamic_rev(sb);
2806 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2807 EXT4_SET_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL);
ac27a0ec
DK		 2808
2809 es->s_journal_inum = cpu_to_le32(journal_inum);
2810 sb->s_dirt = 1;
2811
2812 /* Make sure we flush the recovery flag to disk. */
617ba13b 2813 ext4_commit_super(sb, es, 1);
ac27a0ec
DK		 2814
2815 return 0;
2816}
2817
2b2d6d01
TT		 2818static void ext4_commit_super(struct super_block *sb,
2819 struct ext4_super_block *es, int sync)
ac27a0ec 2820{
617ba13b 2821 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
ac27a0ec
DK		 2822
2823 if (!sbh)
2824 return;
914258bf
TT		 2825 if (buffer_write_io_error(sbh)) {
2826 /*
2827 * Oh, dear. A previous attempt to write the
2828 * superblock failed. This could happen because the
2829 * USB device was yanked out. Or it could happen to
2830 * be a transient write error and maybe the block will
2831 * be remapped. Nothing we can do but to retry the
2832 * write and hope for the best.
2833 */
2834 printk(KERN_ERR "ext4: previous I/O error to "
2835 "superblock detected for %s.\n", sb->s_id);
2836 clear_buffer_write_io_error(sbh);
2837 set_buffer_uptodate(sbh);
2838 }
ac27a0ec 2839 es->s_wtime = cpu_to_le32(get_seconds());
bd81d8ee 2840 ext4_free_blocks_count_set(es, ext4_count_free_blocks(sb));
617ba13b 2841 es->s_free_inodes_count = cpu_to_le32(ext4_count_free_inodes(sb));
ac27a0ec
DK		 2842 BUFFER_TRACE(sbh, "marking dirty");
2843 mark_buffer_dirty(sbh);
914258bf 2844 if (sync) {
ac27a0ec 2845 sync_dirty_buffer(sbh);
914258bf
TT		 2846 if (buffer_write_io_error(sbh)) {
2847 printk(KERN_ERR "ext4: I/O error while writing "
2848 "superblock for %s.\n", sb->s_id);
2849 clear_buffer_write_io_error(sbh);
2850 set_buffer_uptodate(sbh);
2851 }
2852 }
ac27a0ec
DK		 2853}
2854
2855
2856/*
2857 * Have we just finished recovery? If so, and if we are mounting (or
2858 * remounting) the filesystem readonly, then we will end up with a
2859 * consistent fs on disk. Record that fact.
2860 */
2b2d6d01
TT		 2861static void ext4_mark_recovery_complete(struct super_block *sb,
2862 struct ext4_super_block *es)
ac27a0ec 2863{
617ba13b 2864 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 2865
0390131b
FM		 2866 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2867 BUG_ON(journal != NULL);
2868 return;
2869 }
dab291af 2870 jbd2_journal_lock_updates(journal);
7ffe1ea8
HK		 2871 if (jbd2_journal_flush(journal) < 0)
2872 goto out;
2873
32c37730 2874 lock_super(sb);
617ba13b 2875 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
ac27a0ec 2876 sb->s_flags & MS_RDONLY) {
617ba13b 2877 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 2878 sb->s_dirt = 0;
617ba13b 2879 ext4_commit_super(sb, es, 1);
ac27a0ec 2880 }
32c37730 2881 unlock_super(sb);
7ffe1ea8
HK		 2882
2883out:
dab291af 2884 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 2885}
2886
2887/*
2888 * If we are mounting (or read-write remounting) a filesystem whose journal
2889 * has recorded an error from a previous lifetime, move that error to the
2890 * main filesystem now.
2891 */
2b2d6d01
TT		 2892static void ext4_clear_journal_err(struct super_block *sb,
2893 struct ext4_super_block *es)
ac27a0ec
DK		 2894{
2895 journal_t *journal;
2896 int j_errno;
2897 const char *errstr;
2898
0390131b
FM		 2899 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
2900
617ba13b 2901 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 2902
2903 /*
2904 * Now check for any error status which may have been recorded in the
617ba13b 2905 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 2906 */
2907
dab291af 2908 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 2909 if (j_errno) {
2910 char nbuf[16];
2911
617ba13b 2912 errstr = ext4_decode_error(sb, j_errno, nbuf);
46e665e9 2913 ext4_warning(sb, __func__, "Filesystem error recorded "
ac27a0ec 2914 "from previous mount: %s", errstr);
46e665e9 2915 ext4_warning(sb, __func__, "Marking fs in need of "
ac27a0ec
DK		 2916 "filesystem check.");
2917
617ba13b
MC		 2918 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2919 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2b2d6d01 2920 ext4_commit_super(sb, es, 1);
ac27a0ec 2921
dab291af 2922 jbd2_journal_clear_err(journal);
ac27a0ec
DK		 2923 }
2924}
2925
2926/*
2927 * Force the running and committing transactions to commit,
2928 * and wait on the commit.
2929 */
617ba13b 2930int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 2931{
2932 journal_t *journal;
0390131b 2933 int ret = 0;
ac27a0ec
DK		 2934
2935 if (sb->s_flags & MS_RDONLY)
2936 return 0;
2937
617ba13b 2938 journal = EXT4_SB(sb)->s_journal;
0390131b
FM		 2939 if (journal) {
2940 sb->s_dirt = 0;
2941 ret = ext4_journal_force_commit(journal);
2942 }
2943
ac27a0ec
DK		 2944 return ret;
2945}
2946
2947/*
617ba13b 2948 * Ext4 always journals updates to the superblock itself, so we don't
ac27a0ec 2949 * have to propagate any other updates to the superblock on disk at this
14ce0cb4
TT		 2950 * point. (We can probably nuke this function altogether, and remove
2951 * any mention to sb->s_dirt in all of fs/ext4; eventual cleanup...)
ac27a0ec 2952 */
2b2d6d01 2953static void ext4_write_super(struct super_block *sb)
ac27a0ec 2954{
0390131b
FM		 2955 if (EXT4_SB(sb)->s_journal) {
2956 if (mutex_trylock(&sb->s_lock) != 0)
2957 BUG();
2958 sb->s_dirt = 0;
2959 } else {
2960 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
2961 }
ac27a0ec
DK		 2962}
2963
617ba13b 2964static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec 2965{
14ce0cb4 2966 int ret = 0;
ac27a0ec 2967
ede86cc4 2968 trace_mark(ext4_sync_fs, "dev %s wait %d", sb->s_id, wait);
ac27a0ec 2969 sb->s_dirt = 0;
0390131b
FM		 2970 if (EXT4_SB(sb)->s_journal) {
2971 if (wait)
2972 ret = ext4_force_commit(sb);
2973 else
2974 jbd2_journal_start_commit(EXT4_SB(sb)->s_journal, NULL);
2975 } else {
2976 ext4_commit_super(sb, EXT4_SB(sb)->s_es, wait);
2977 }
14ce0cb4 2978 return ret;
ac27a0ec
DK		 2979}
2980
2981/*
2982 * LVM calls this function before a (read-only) snapshot is created. This
2983 * gives us a chance to flush the journal completely and mark the fs clean.
2984 */
617ba13b 2985static void ext4_write_super_lockfs(struct super_block *sb)
ac27a0ec
DK		 2986{
2987 sb->s_dirt = 0;
2988
2989 if (!(sb->s_flags & MS_RDONLY)) {
617ba13b 2990 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 2991
0390131b
FM		 2992 if (journal) {
2993 /* Now we set up the journal barrier. */
2994 jbd2_journal_lock_updates(journal);
7ffe1ea8 2995
0390131b
FM		 2996 /*
2997 * We don't want to clear needs_recovery flag when we
2998 * failed to flush the journal.
2999 */
3000 if (jbd2_journal_flush(journal) < 0)
3001 return;
3002 }
ac27a0ec
DK		 3003
3004 /* Journal blocked and flushed, clear needs_recovery flag. */
617ba13b
MC		 3005 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3006 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
ac27a0ec
DK		 3007 }
3008}
3009
3010/*
3011 * Called by LVM after the snapshot is done. We need to reset the RECOVER
3012 * flag here, even though the filesystem is not technically dirty yet.
3013 */
617ba13b 3014static void ext4_unlockfs(struct super_block *sb)
ac27a0ec 3015{
0390131b 3016 if (EXT4_SB(sb)->s_journal && !(sb->s_flags & MS_RDONLY)) {
ac27a0ec
DK		 3017 lock_super(sb);
3018 /* Reser the needs_recovery flag before the fs is unlocked. */
617ba13b
MC		 3019 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3020 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
ac27a0ec 3021 unlock_super(sb);
dab291af 3022 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
ac27a0ec
DK		 3023 }
3024}
3025
2b2d6d01 3026static int ext4_remount(struct super_block *sb, int *flags, char *data)
ac27a0ec 3027{
2b2d6d01 3028 struct ext4_super_block *es;
617ba13b
MC		 3029 struct ext4_sb_info *sbi = EXT4_SB(sb);
3030 ext4_fsblk_t n_blocks_count = 0;
ac27a0ec 3031 unsigned long old_sb_flags;
617ba13b 3032 struct ext4_mount_options old_opts;
8a266467 3033 ext4_group_t g;
ac27a0ec
DK		 3034 int err;
3035#ifdef CONFIG_QUOTA
3036 int i;
3037#endif
3038
3039 /* Store the original options */
3040 old_sb_flags = sb->s_flags;
3041 old_opts.s_mount_opt = sbi->s_mount_opt;
3042 old_opts.s_resuid = sbi->s_resuid;
3043 old_opts.s_resgid = sbi->s_resgid;
3044 old_opts.s_commit_interval = sbi->s_commit_interval;
3045#ifdef CONFIG_QUOTA
3046 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
3047 for (i = 0; i < MAXQUOTAS; i++)
3048 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
3049#endif
3050
3051 /*
3052 * Allow the "check" option to be passed as a remount option.
3053 */
3054 if (!parse_options(data, sb, NULL, NULL, &n_blocks_count, 1)) {
3055 err = -EINVAL;
3056 goto restore_opts;
3057 }
3058
617ba13b 3059 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT)
46e665e9 3060 ext4_abort(sb, __func__, "Abort forced by user");
ac27a0ec
DK		 3061
3062 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
617ba13b 3063 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec
DK		 3064
3065 es = sbi->s_es;
3066
0390131b
FM		 3067 if (sbi->s_journal)
3068 ext4_init_journal_params(sb, sbi->s_journal);
ac27a0ec
DK		 3069
3070 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
bd81d8ee 3071 n_blocks_count > ext4_blocks_count(es)) {
617ba13b 3072 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT) {
ac27a0ec
DK		 3073 err = -EROFS;
3074 goto restore_opts;
3075 }
3076
3077 if (*flags & MS_RDONLY) {
3078 /*
3079 * First of all, the unconditional stuff we have to do
3080 * to disable replay of the journal when we next remount
3081 */
3082 sb->s_flags |= MS_RDONLY;
3083
3084 /*
3085 * OK, test if we are remounting a valid rw partition
3086 * readonly, and if so set the rdonly flag and then
3087 * mark the partition as valid again.
3088 */
617ba13b
MC		 3089 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
3090 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 3091 es->s_state = cpu_to_le16(sbi->s_mount_state);
3092
32c37730
JK		 3093 /*
3094 * We have to unlock super so that we can wait for
3095 * transactions.
3096 */
0390131b
FM		 3097 if (sbi->s_journal) {
3098 unlock_super(sb);
3099 ext4_mark_recovery_complete(sb, es);
3100 lock_super(sb);
3101 }
ac27a0ec
DK		 3102 } else {
3103 __le32 ret;
617ba13b
MC		 3104 if ((ret = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3105 ~EXT4_FEATURE_RO_COMPAT_SUPP))) {
3106 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
ac27a0ec
DK		 3107 "remount RDWR because of unsupported "
3108 "optional features (%x).\n",
3109 sb->s_id, le32_to_cpu(ret));
3110 err = -EROFS;
3111 goto restore_opts;
3112 }
ead6596b 3113
8a266467
TT		 3114 /*
3115 * Make sure the group descriptor checksums
3116 * are sane. If they aren't, refuse to
3117 * remount r/w.
3118 */
3119 for (g = 0; g < sbi->s_groups_count; g++) {
3120 struct ext4_group_desc *gdp =
3121 ext4_get_group_desc(sb, g, NULL);
3122
3123 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
3124 printk(KERN_ERR
3125 "EXT4-fs: ext4_remount: "
3126 "Checksum for group %lu failed (%u!=%u)\n",
3127 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
3128 le16_to_cpu(gdp->bg_checksum));
3129 err = -EINVAL;
3130 goto restore_opts;
3131 }
3132 }
3133
ead6596b
ES		 3134 /*
3135 * If we have an unprocessed orphan list hanging
3136 * around from a previously readonly bdev mount,
3137 * require a full umount/remount for now.
3138 */
3139 if (es->s_last_orphan) {
3140 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
3141 "remount RDWR because of unprocessed "
3142 "orphan inode list. Please "
3143 "umount/remount instead.\n",
3144 sb->s_id);
3145 err = -EINVAL;
3146 goto restore_opts;
3147 }
3148
ac27a0ec
DK		 3149 /*
3150 * Mounting a RDONLY partition read-write, so reread
3151 * and store the current valid flag. (It may have
3152 * been changed by e2fsck since we originally mounted
3153 * the partition.)
3154 */
0390131b
FM		 3155 if (sbi->s_journal)
3156 ext4_clear_journal_err(sb, es);
ac27a0ec 3157 sbi->s_mount_state = le16_to_cpu(es->s_state);
617ba13b 3158 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
ac27a0ec 3159 goto restore_opts;
2b2d6d01 3160 if (!ext4_setup_super(sb, es, 0))
ac27a0ec
DK		 3161 sb->s_flags &= ~MS_RDONLY;
3162 }
3163 }
0390131b
FM		 3164 if (sbi->s_journal == NULL)
3165 ext4_commit_super(sb, es, 1);
3166
ac27a0ec
DK		 3167#ifdef CONFIG_QUOTA
3168 /* Release old quota file names */
3169 for (i = 0; i < MAXQUOTAS; i++)
3170 if (old_opts.s_qf_names[i] &&
3171 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3172 kfree(old_opts.s_qf_names[i]);
3173#endif
3174 return 0;
3175restore_opts:
3176 sb->s_flags = old_sb_flags;
3177 sbi->s_mount_opt = old_opts.s_mount_opt;
3178 sbi->s_resuid = old_opts.s_resuid;
3179 sbi->s_resgid = old_opts.s_resgid;
3180 sbi->s_commit_interval = old_opts.s_commit_interval;
3181#ifdef CONFIG_QUOTA
3182 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
3183 for (i = 0; i < MAXQUOTAS; i++) {
3184 if (sbi->s_qf_names[i] &&
3185 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3186 kfree(sbi->s_qf_names[i]);
3187 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
3188 }
3189#endif
3190 return err;
3191}
3192
2b2d6d01 3193static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
ac27a0ec
DK		 3194{
3195 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 3196 struct ext4_sb_info *sbi = EXT4_SB(sb);
3197 struct ext4_super_block *es = sbi->s_es;
960cc398 3198 u64 fsid;
ac27a0ec 3199
5e70030d
BP		 3200 if (test_opt(sb, MINIX_DF)) {
3201 sbi->s_overhead_last = 0;
6bc9feff 3202 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
fd2d4291 3203 ext4_group_t ngroups = sbi->s_groups_count, i;
5e70030d 3204 ext4_fsblk_t overhead = 0;
ac27a0ec
DK		 3205 smp_rmb();
3206
3207 /*
5e70030d
BP		 3208 * Compute the overhead (FS structures). This is constant
3209 * for a given filesystem unless the number of block groups
3210 * changes so we cache the previous value until it does.
ac27a0ec
DK		 3211 */
3212
3213 /*
3214 * All of the blocks before first_data_block are
3215 * overhead
3216 */
3217 overhead = le32_to_cpu(es->s_first_data_block);
3218
3219 /*
3220 * Add the overhead attributed to the superblock and
3221 * block group descriptors. If the sparse superblocks
3222 * feature is turned on, then not all groups have this.
3223 */
3224 for (i = 0; i < ngroups; i++) {
617ba13b
MC		 3225 overhead += ext4_bg_has_super(sb, i) +
3226 ext4_bg_num_gdb(sb, i);
ac27a0ec
DK		 3227 cond_resched();
3228 }
3229
3230 /*
3231 * Every block group has an inode bitmap, a block
3232 * bitmap, and an inode table.
3233 */
5e70030d
BP		 3234 overhead += ngroups * (2 + sbi->s_itb_per_group);
3235 sbi->s_overhead_last = overhead;
3236 smp_wmb();
6bc9feff 3237 sbi->s_blocks_last = ext4_blocks_count(es);
ac27a0ec
DK		 3238 }
3239
617ba13b 3240 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 3241 buf->f_bsize = sb->s_blocksize;
5e70030d 3242 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
6bc6e63f
AK		 3243 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
3244 percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
308ba3ec 3245 ext4_free_blocks_count_set(es, buf->f_bfree);
bd81d8ee
LV		 3246 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3247 if (buf->f_bfree < ext4_r_blocks_count(es))
ac27a0ec
DK		 3248 buf->f_bavail = 0;
3249 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 3250 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
5e70030d 3251 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
617ba13b 3252 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 3253 fsid = le64_to_cpup((void *)es->s_uuid) ^
3254 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3255 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3256 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
ac27a0ec
DK		 3257 return 0;
3258}
3259
3260/* Helper function for writing quotas on sync - we need to start transaction before quota file
3261 * is locked for write. Otherwise the are possible deadlocks:
3262 * Process 1 Process 2
617ba13b 3263 * ext4_create() quota_sync()
dab291af 3264 * jbd2_journal_start() write_dquot()
ac27a0ec 3265 * DQUOT_INIT() down(dqio_mutex)
dab291af 3266 * down(dqio_mutex) jbd2_journal_start()
ac27a0ec
DK		 3267 *
3268 */
3269
3270#ifdef CONFIG_QUOTA
3271
3272static inline struct inode *dquot_to_inode(struct dquot *dquot)
3273{
3274 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3275}
3276
617ba13b 3277static int ext4_dquot_initialize(struct inode *inode, int type)
ac27a0ec
DK		 3278{
3279 handle_t *handle;
3280 int ret, err;
3281
3282 /* We may create quota structure so we need to reserve enough blocks */
617ba13b 3283 handle = ext4_journal_start(inode, 2*EXT4_QUOTA_INIT_BLOCKS(inode->i_sb));
ac27a0ec
DK		 3284 if (IS_ERR(handle))
3285 return PTR_ERR(handle);
3286 ret = dquot_initialize(inode, type);
617ba13b 3287 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3288 if (!ret)
3289 ret = err;
3290 return ret;
3291}
3292
617ba13b 3293static int ext4_dquot_drop(struct inode *inode)
ac27a0ec
DK		 3294{
3295 handle_t *handle;
3296 int ret, err;
3297
3298 /* We may delete quota structure so we need to reserve enough blocks */
617ba13b 3299 handle = ext4_journal_start(inode, 2*EXT4_QUOTA_DEL_BLOCKS(inode->i_sb));
2887df13
JK		 3300 if (IS_ERR(handle)) {
3301 /*
3302 * We call dquot_drop() anyway to at least release references
3303 * to quota structures so that umount does not hang.
3304 */
3305 dquot_drop(inode);
ac27a0ec 3306 return PTR_ERR(handle);
2887df13 3307 }
ac27a0ec 3308 ret = dquot_drop(inode);
617ba13b 3309 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3310 if (!ret)
3311 ret = err;
3312 return ret;
3313}
3314
617ba13b 3315static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 3316{
3317 int ret, err;
3318 handle_t *handle;
3319 struct inode *inode;
3320
3321 inode = dquot_to_inode(dquot);
617ba13b
MC		 3322 handle = ext4_journal_start(inode,
3323 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 3324 if (IS_ERR(handle))
3325 return PTR_ERR(handle);
3326 ret = dquot_commit(dquot);
617ba13b 3327 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3328 if (!ret)
3329 ret = err;
3330 return ret;
3331}
3332
617ba13b 3333static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 3334{
3335 int ret, err;
3336 handle_t *handle;
3337
617ba13b
MC		 3338 handle = ext4_journal_start(dquot_to_inode(dquot),
3339 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 3340 if (IS_ERR(handle))
3341 return PTR_ERR(handle);
3342 ret = dquot_acquire(dquot);
617ba13b 3343 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3344 if (!ret)
3345 ret = err;
3346 return ret;
3347}
3348
617ba13b 3349static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 3350{
3351 int ret, err;
3352 handle_t *handle;
3353
617ba13b
MC		 3354 handle = ext4_journal_start(dquot_to_inode(dquot),
3355 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 3356 if (IS_ERR(handle)) {
3357 /* Release dquot anyway to avoid endless cycle in dqput() */
3358 dquot_release(dquot);
ac27a0ec 3359 return PTR_ERR(handle);
9c3013e9 3360 }
ac27a0ec 3361 ret = dquot_release(dquot);
617ba13b 3362 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3363 if (!ret)
3364 ret = err;
3365 return ret;
3366}
3367
617ba13b 3368static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec 3369{
2c8be6b2 3370 /* Are we journaling quotas? */
617ba13b
MC		 3371 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3372 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
ac27a0ec 3373 dquot_mark_dquot_dirty(dquot);
617ba13b 3374 return ext4_write_dquot(dquot);
ac27a0ec
DK		 3375 } else {
3376 return dquot_mark_dquot_dirty(dquot);
3377 }
3378}
3379
617ba13b 3380static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 3381{
3382 int ret, err;
3383 handle_t *handle;
3384
3385 /* Data block + inode block */
617ba13b 3386 handle = ext4_journal_start(sb->s_root->d_inode, 2);
ac27a0ec
DK		 3387 if (IS_ERR(handle))
3388 return PTR_ERR(handle);
3389 ret = dquot_commit_info(sb, type);
617ba13b 3390 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3391 if (!ret)
3392 ret = err;
3393 return ret;
3394}
3395
3396/*
3397 * Turn on quotas during mount time - we need to find
3398 * the quota file and such...
3399 */
617ba13b 3400static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 3401{
617ba13b
MC		 3402 return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
3403 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 3404}
3405
3406/*
3407 * Standard function to be called on quota_on
3408 */
617ba13b 3409static int ext4_quota_on(struct super_block *sb, int type, int format_id,
8264613d 3410 char *name, int remount)
ac27a0ec
DK		 3411{
3412 int err;
8264613d 3413 struct path path;
ac27a0ec
DK		 3414
3415 if (!test_opt(sb, QUOTA))
3416 return -EINVAL;
8264613d 3417 /* When remounting, no checks are needed and in fact, name is NULL */
0623543b 3418 if (remount)
8264613d 3419 return vfs_quota_on(sb, type, format_id, name, remount);
0623543b 3420
8264613d 3421 err = kern_path(name, LOOKUP_FOLLOW, &path);
ac27a0ec
DK		 3422 if (err)
3423 return err;
0623543b 3424
ac27a0ec 3425 /* Quotafile not on the same filesystem? */
8264613d
AV		 3426 if (path.mnt->mnt_sb != sb) {
3427 path_put(&path);
ac27a0ec
DK		 3428 return -EXDEV;
3429 }
0623543b
JK		 3430 /* Journaling quota? */
3431 if (EXT4_SB(sb)->s_qf_names[type]) {
2b2d6d01 3432 /* Quotafile not in fs root? */
8264613d 3433 if (path.dentry->d_parent != sb->s_root)
0623543b
JK		 3434 printk(KERN_WARNING
3435 "EXT4-fs: Quota file not on filesystem root. "
3436 "Journaled quota will not work.\n");
2b2d6d01 3437 }
0623543b
JK		 3438
3439 /*
3440 * When we journal data on quota file, we have to flush journal to see
3441 * all updates to the file when we bypass pagecache...
3442 */
0390131b
FM		 3443 if (EXT4_SB(sb)->s_journal &&
3444 ext4_should_journal_data(path.dentry->d_inode)) {
0623543b
JK		 3445 /*
3446 * We don't need to lock updates but journal_flush() could
3447 * otherwise be livelocked...
3448 */
3449 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 3450 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
0623543b 3451 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 3452 if (err) {
8264613d 3453 path_put(&path);
7ffe1ea8
HK		 3454 return err;
3455 }
0623543b
JK		 3456 }
3457
8264613d
AV		 3458 err = vfs_quota_on_path(sb, type, format_id, &path);
3459 path_put(&path);
77e69dac 3460 return err;
ac27a0ec
DK		 3461}
3462
3463/* Read data from quotafile - avoid pagecache and such because we cannot afford
3464 * acquiring the locks... As quota files are never truncated and quota code
3465 * itself serializes the operations (and noone else should touch the files)
3466 * we don't have to be afraid of races */
617ba13b 3467static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 3468 size_t len, loff_t off)
3469{
3470 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 3471 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 3472 int err = 0;
3473 int offset = off & (sb->s_blocksize - 1);
3474 int tocopy;
3475 size_t toread;
3476 struct buffer_head *bh;
3477 loff_t i_size = i_size_read(inode);
3478
3479 if (off > i_size)
3480 return 0;
3481 if (off+len > i_size)
3482 len = i_size-off;
3483 toread = len;
3484 while (toread > 0) {
3485 tocopy = sb->s_blocksize - offset < toread ?
3486 sb->s_blocksize - offset : toread;
617ba13b 3487 bh = ext4_bread(NULL, inode, blk, 0, &err);
ac27a0ec
DK		 3488 if (err)
3489 return err;
3490 if (!bh) /* A hole? */
3491 memset(data, 0, tocopy);
3492 else
3493 memcpy(data, bh->b_data+offset, tocopy);
3494 brelse(bh);
3495 offset = 0;
3496 toread -= tocopy;
3497 data += tocopy;
3498 blk++;
3499 }
3500 return len;
3501}
3502
3503/* Write to quotafile (we know the transaction is already started and has
3504 * enough credits) */
617ba13b 3505static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 3506 const char *data, size_t len, loff_t off)
3507{
3508 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 3509 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 3510 int err = 0;
3511 int offset = off & (sb->s_blocksize - 1);
3512 int tocopy;
617ba13b 3513 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
ac27a0ec
DK		 3514 size_t towrite = len;
3515 struct buffer_head *bh;
3516 handle_t *handle = journal_current_handle();
3517
0390131b 3518 if (EXT4_SB(sb)->s_journal && !handle) {
e5f8eab8 3519 printk(KERN_WARNING "EXT4-fs: Quota write (off=%llu, len=%llu)"
9c3013e9
JK		 3520 " cancelled because transaction is not started.\n",
3521 (unsigned long long)off, (unsigned long long)len);
3522 return -EIO;
3523 }
ac27a0ec
DK		 3524 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
3525 while (towrite > 0) {
3526 tocopy = sb->s_blocksize - offset < towrite ?
3527 sb->s_blocksize - offset : towrite;
617ba13b 3528 bh = ext4_bread(handle, inode, blk, 1, &err);
ac27a0ec
DK		 3529 if (!bh)
3530 goto out;
3531 if (journal_quota) {
617ba13b 3532 err = ext4_journal_get_write_access(handle, bh);
ac27a0ec
DK		 3533 if (err) {
3534 brelse(bh);
3535 goto out;
3536 }
3537 }
3538 lock_buffer(bh);
3539 memcpy(bh->b_data+offset, data, tocopy);
3540 flush_dcache_page(bh->b_page);
3541 unlock_buffer(bh);
3542 if (journal_quota)
0390131b 3543 err = ext4_handle_dirty_metadata(handle, NULL, bh);
ac27a0ec
DK		 3544 else {
3545 /* Always do at least ordered writes for quotas */
678aaf48 3546 err = ext4_jbd2_file_inode(handle, inode);
ac27a0ec
DK		 3547 mark_buffer_dirty(bh);
3548 }
3549 brelse(bh);
3550 if (err)
3551 goto out;
3552 offset = 0;
3553 towrite -= tocopy;
3554 data += tocopy;
3555 blk++;
3556 }
3557out:
4d04e4fb
JK		 3558 if (len == towrite) {
3559 mutex_unlock(&inode->i_mutex);
ac27a0ec 3560 return err;
4d04e4fb 3561 }
ac27a0ec
DK		 3562 if (inode->i_size < off+len-towrite) {
3563 i_size_write(inode, off+len-towrite);
617ba13b 3564 EXT4_I(inode)->i_disksize = inode->i_size;
ac27a0ec 3565 }
ac27a0ec 3566 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
617ba13b 3567 ext4_mark_inode_dirty(handle, inode);
ac27a0ec
DK		 3568 mutex_unlock(&inode->i_mutex);
3569 return len - towrite;
3570}
3571
3572#endif
3573
617ba13b 3574static int ext4_get_sb(struct file_system_type *fs_type,
ac27a0ec
DK		 3575 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3576{
617ba13b 3577 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
ac27a0ec
DK		 3578}
3579
5e8814f2
TT		 3580#ifdef CONFIG_PROC_FS
3581static int ext4_ui_proc_show(struct seq_file *m, void *v)
3582{
3583 unsigned int *p = m->private;
3584
3585 seq_printf(m, "%u\n", *p);
3586 return 0;
3587}
3588
3589static int ext4_ui_proc_open(struct inode *inode, struct file *file)
3590{
3591 return single_open(file, ext4_ui_proc_show, PDE(inode)->data);
3592}
3593
3594static ssize_t ext4_ui_proc_write(struct file *file, const char __user *buf,
3595 size_t cnt, loff_t *ppos)
3596{
23475e26 3597 unsigned long *p = PDE(file->f_path.dentry->d_inode)->data;
5e8814f2 3598 char str[32];
5e8814f2
TT		 3599
3600 if (cnt >= sizeof(str))
3601 return -EINVAL;
3602 if (copy_from_user(str, buf, cnt))
3603 return -EFAULT;
23475e26
RK		 3604
3605 *p = simple_strtoul(str, NULL, 0);
5e8814f2
TT		 3606 return cnt;
3607}
3608
3609const struct file_operations ext4_ui_proc_fops = {
3610 .owner = THIS_MODULE,
3611 .open = ext4_ui_proc_open,
3612 .read = seq_read,
3613 .llseek = seq_lseek,
3614 .release = single_release,
3615 .write = ext4_ui_proc_write,
3616};
3617#endif
3618
03010a33
TT		 3619static struct file_system_type ext4_fs_type = {
3620 .owner = THIS_MODULE,
3621 .name = "ext4",
3622 .get_sb = ext4_get_sb,
3623 .kill_sb = kill_block_super,
3624 .fs_flags = FS_REQUIRES_DEV,
3625};
3626
3627#ifdef CONFIG_EXT4DEV_COMPAT
3628static int ext4dev_get_sb(struct file_system_type *fs_type,
3629 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3630{
3631 printk(KERN_WARNING "EXT4-fs: Update your userspace programs "
3632 "to mount using ext4\n");
3633 printk(KERN_WARNING "EXT4-fs: ext4dev backwards compatibility "
3634 "will go away by 2.6.31\n");
3635 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
3636}
3637
617ba13b 3638static struct file_system_type ext4dev_fs_type = {
ac27a0ec 3639 .owner = THIS_MODULE,
617ba13b 3640 .name = "ext4dev",
03010a33 3641 .get_sb = ext4dev_get_sb,
ac27a0ec
DK		 3642 .kill_sb = kill_block_super,
3643 .fs_flags = FS_REQUIRES_DEV,
3644};
03010a33
TT		 3645MODULE_ALIAS("ext4dev");
3646#endif
ac27a0ec 3647
617ba13b 3648static int __init init_ext4_fs(void)
ac27a0ec 3649{
c9de560d
AT		 3650 int err;
3651
9f6200bb 3652 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
c9de560d 3653 err = init_ext4_mballoc();
ac27a0ec
DK		 3654 if (err)
3655 return err;
c9de560d
AT		 3656
3657 err = init_ext4_xattr();
3658 if (err)
3659 goto out2;
ac27a0ec
DK		 3660 err = init_inodecache();
3661 if (err)
3662 goto out1;
03010a33 3663 err = register_filesystem(&ext4_fs_type);
ac27a0ec
DK		 3664 if (err)
3665 goto out;
03010a33
TT		 3666#ifdef CONFIG_EXT4DEV_COMPAT
3667 err = register_filesystem(&ext4dev_fs_type);
3668 if (err) {
3669 unregister_filesystem(&ext4_fs_type);
3670 goto out;
3671 }
3672#endif
ac27a0ec
DK		 3673 return 0;
3674out:
3675 destroy_inodecache();
3676out1:
617ba13b 3677 exit_ext4_xattr();
c9de560d
AT		 3678out2:
3679 exit_ext4_mballoc();
ac27a0ec
DK		 3680 return err;
3681}
3682
617ba13b 3683static void __exit exit_ext4_fs(void)
ac27a0ec 3684{
03010a33
TT		 3685 unregister_filesystem(&ext4_fs_type);
3686#ifdef CONFIG_EXT4DEV_COMPAT
617ba13b 3687 unregister_filesystem(&ext4dev_fs_type);
03010a33 3688#endif
ac27a0ec 3689 destroy_inodecache();
617ba13b 3690 exit_ext4_xattr();
c9de560d 3691 exit_ext4_mballoc();
9f6200bb 3692 remove_proc_entry("fs/ext4", NULL);
ac27a0ec
DK		 3693}
3694
3695MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
617ba13b 3696MODULE_DESCRIPTION("Fourth Extended Filesystem with extents");
ac27a0ec 3697MODULE_LICENSE("GPL");
617ba13b
MC		 3698module_init(init_ext4_fs)
3699module_exit(exit_ext4_fs)
Ubuntu Artful kernel mirror