]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blame - fs/ext4/super.c
projects / mirror_ubuntu-bionic-kernel.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
jbd2: don't dirty original metadata buffer on abort
[mirror_ubuntu-bionic-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;
ac27a0ec 139 if (is_journal_aborted(journal)) {
46e665e9 140 ext4_abort(sb, __func__,
ac27a0ec
DK		 141 "Detected aborted journal");
142 return ERR_PTR(-EROFS);
143 }
144
dab291af 145 return jbd2_journal_start(journal, nblocks);
ac27a0ec
DK		 146}
147
148/*
149 * The only special thing we need to do here is to make sure that all
dab291af 150 * jbd2_journal_stop calls result in the superblock being marked dirty, so
ac27a0ec
DK		 151 * that sync() will call the filesystem's write_super callback if
152 * appropriate.
153 */
617ba13b 154int __ext4_journal_stop(const char *where, handle_t *handle)
ac27a0ec
DK		 155{
156 struct super_block *sb;
157 int err;
158 int rc;
159
160 sb = handle->h_transaction->t_journal->j_private;
161 err = handle->h_err;
dab291af 162 rc = jbd2_journal_stop(handle);
ac27a0ec
DK		 163
164 if (!err)
165 err = rc;
166 if (err)
617ba13b 167 __ext4_std_error(sb, where, err);
ac27a0ec
DK		 168 return err;
169}
170
617ba13b 171void ext4_journal_abort_handle(const char *caller, const char *err_fn,
ac27a0ec
DK		 172 struct buffer_head *bh, handle_t *handle, int err)
173{
174 char nbuf[16];
617ba13b 175 const char *errstr = ext4_decode_error(NULL, err, nbuf);
ac27a0ec
DK		 176
177 if (bh)
178 BUFFER_TRACE(bh, "abort");
179
180 if (!handle->h_err)
181 handle->h_err = err;
182
183 if (is_handle_aborted(handle))
184 return;
185
186 printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
187 caller, errstr, err_fn);
188
dab291af 189 jbd2_journal_abort_handle(handle);
ac27a0ec
DK		 190}
191
192/* Deal with the reporting of failure conditions on a filesystem such as
193 * inconsistencies detected or read IO failures.
194 *
195 * On ext2, we can store the error state of the filesystem in the
617ba13b 196 * superblock. That is not possible on ext4, because we may have other
ac27a0ec
DK		 197 * write ordering constraints on the superblock which prevent us from
198 * writing it out straight away; and given that the journal is about to
199 * be aborted, we can't rely on the current, or future, transactions to
200 * write out the superblock safely.
201 *
dab291af 202 * We'll just use the jbd2_journal_abort() error code to record an error in
ac27a0ec
DK		 203 * the journal instead. On recovery, the journal will compain about
204 * that error until we've noted it down and cleared it.
205 */
206
617ba13b 207static void ext4_handle_error(struct super_block *sb)
ac27a0ec 208{
617ba13b 209 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 210
617ba13b
MC		 211 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
212 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
ac27a0ec
DK		 213
214 if (sb->s_flags & MS_RDONLY)
215 return;
216
2b2d6d01 217 if (!test_opt(sb, ERRORS_CONT)) {
617ba13b 218 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 219
617ba13b 220 EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
ac27a0ec 221 if (journal)
dab291af 222 jbd2_journal_abort(journal, -EIO);
ac27a0ec 223 }
2b2d6d01
TT		 224 if (test_opt(sb, ERRORS_RO)) {
225 printk(KERN_CRIT "Remounting filesystem read-only\n");
ac27a0ec
DK		 226 sb->s_flags |= MS_RDONLY;
227 }
617ba13b 228 ext4_commit_super(sb, es, 1);
ac27a0ec 229 if (test_opt(sb, ERRORS_PANIC))
617ba13b 230 panic("EXT4-fs (device %s): panic forced after error\n",
ac27a0ec
DK		 231 sb->s_id);
232}
233
2b2d6d01
TT		 234void ext4_error(struct super_block *sb, const char *function,
235 const char *fmt, ...)
ac27a0ec
DK		 236{
237 va_list args;
238
239 va_start(args, fmt);
2b2d6d01 240 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
ac27a0ec
DK		 241 vprintk(fmt, args);
242 printk("\n");
243 va_end(args);
244
617ba13b 245 ext4_handle_error(sb);
ac27a0ec
DK		 246}
247
2b2d6d01 248static const char *ext4_decode_error(struct super_block *sb, int errno,
ac27a0ec
DK		 249 char nbuf[16])
250{
251 char *errstr = NULL;
252
253 switch (errno) {
254 case -EIO:
255 errstr = "IO failure";
256 break;
257 case -ENOMEM:
258 errstr = "Out of memory";
259 break;
260 case -EROFS:
dab291af 261 if (!sb || EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT)
ac27a0ec
DK		 262 errstr = "Journal has aborted";
263 else
264 errstr = "Readonly filesystem";
265 break;
266 default:
267 /* If the caller passed in an extra buffer for unknown
268 * errors, textualise them now. Else we just return
269 * NULL. */
270 if (nbuf) {
271 /* Check for truncated error codes... */
272 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
273 errstr = nbuf;
274 }
275 break;
276 }
277
278 return errstr;
279}
280
617ba13b 281/* __ext4_std_error decodes expected errors from journaling functions
ac27a0ec
DK		 282 * automatically and invokes the appropriate error response. */
283
2b2d6d01 284void __ext4_std_error(struct super_block *sb, const char *function, int errno)
ac27a0ec
DK		 285{
286 char nbuf[16];
287 const char *errstr;
288
289 /* Special case: if the error is EROFS, and we're not already
290 * inside a transaction, then there's really no point in logging
291 * an error. */
292 if (errno == -EROFS && journal_current_handle() == NULL &&
293 (sb->s_flags & MS_RDONLY))
294 return;
295
617ba13b 296 errstr = ext4_decode_error(sb, errno, nbuf);
2b2d6d01
TT		 297 printk(KERN_CRIT "EXT4-fs error (device %s) in %s: %s\n",
298 sb->s_id, function, errstr);
ac27a0ec 299
617ba13b 300 ext4_handle_error(sb);
ac27a0ec
DK		 301}
302
303/*
617ba13b 304 * ext4_abort is a much stronger failure handler than ext4_error. The
ac27a0ec
DK		 305 * abort function may be used to deal with unrecoverable failures such
306 * as journal IO errors or ENOMEM at a critical moment in log management.
307 *
308 * We unconditionally force the filesystem into an ABORT|READONLY state,
309 * unless the error response on the fs has been set to panic in which
310 * case we take the easy way out and panic immediately.
311 */
312
2b2d6d01
TT		 313void ext4_abort(struct super_block *sb, const char *function,
314 const char *fmt, ...)
ac27a0ec
DK		 315{
316 va_list args;
317
2b2d6d01 318 printk(KERN_CRIT "ext4_abort called.\n");
ac27a0ec
DK		 319
320 va_start(args, fmt);
2b2d6d01 321 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
ac27a0ec
DK		 322 vprintk(fmt, args);
323 printk("\n");
324 va_end(args);
325
326 if (test_opt(sb, ERRORS_PANIC))
617ba13b 327 panic("EXT4-fs panic from previous error\n");
ac27a0ec
DK		 328
329 if (sb->s_flags & MS_RDONLY)
330 return;
331
332 printk(KERN_CRIT "Remounting filesystem read-only\n");
617ba13b 333 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
ac27a0ec 334 sb->s_flags |= MS_RDONLY;
617ba13b 335 EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
dab291af 336 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
ac27a0ec
DK		 337}
338
2b2d6d01
TT		 339void ext4_warning(struct super_block *sb, const char *function,
340 const char *fmt, ...)
ac27a0ec
DK		 341{
342 va_list args;
343
344 va_start(args, fmt);
617ba13b 345 printk(KERN_WARNING "EXT4-fs warning (device %s): %s: ",
ac27a0ec
DK		 346 sb->s_id, function);
347 vprintk(fmt, args);
348 printk("\n");
349 va_end(args);
350}
351
617ba13b 352void ext4_update_dynamic_rev(struct super_block *sb)
ac27a0ec 353{
617ba13b 354 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 355
617ba13b 356 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
ac27a0ec
DK		 357 return;
358
46e665e9 359 ext4_warning(sb, __func__,
ac27a0ec
DK		 360 "updating to rev %d because of new feature flag, "
361 "running e2fsck is recommended",
617ba13b 362 EXT4_DYNAMIC_REV);
ac27a0ec 363
617ba13b
MC		 364 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
365 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
366 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
ac27a0ec
DK		 367 /* leave es->s_feature_*compat flags alone */
368 /* es->s_uuid will be set by e2fsck if empty */
369
370 /*
371 * The rest of the superblock fields should be zero, and if not it
372 * means they are likely already in use, so leave them alone. We
373 * can leave it up to e2fsck to clean up any inconsistencies there.
374 */
375}
376
99e6f829
AK		 377int ext4_update_compat_feature(handle_t *handle,
378 struct super_block *sb, __u32 compat)
379{
380 int err = 0;
381 if (!EXT4_HAS_COMPAT_FEATURE(sb, compat)) {
382 err = ext4_journal_get_write_access(handle,
383 EXT4_SB(sb)->s_sbh);
384 if (err)
385 return err;
386 EXT4_SET_COMPAT_FEATURE(sb, compat);
387 sb->s_dirt = 1;
388 handle->h_sync = 1;
389 BUFFER_TRACE(EXT4_SB(sb)->s_sbh,
390 "call ext4_journal_dirty_met adata");
391 err = ext4_journal_dirty_metadata(handle,
392 EXT4_SB(sb)->s_sbh);
393 }
394 return err;
395}
396
397int ext4_update_rocompat_feature(handle_t *handle,
398 struct super_block *sb, __u32 rocompat)
399{
400 int err = 0;
401 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, rocompat)) {
402 err = ext4_journal_get_write_access(handle,
403 EXT4_SB(sb)->s_sbh);
404 if (err)
405 return err;
406 EXT4_SET_RO_COMPAT_FEATURE(sb, rocompat);
407 sb->s_dirt = 1;
408 handle->h_sync = 1;
409 BUFFER_TRACE(EXT4_SB(sb)->s_sbh,
410 "call ext4_journal_dirty_met adata");
411 err = ext4_journal_dirty_metadata(handle,
412 EXT4_SB(sb)->s_sbh);
413 }
414 return err;
415}
416
417int ext4_update_incompat_feature(handle_t *handle,
418 struct super_block *sb, __u32 incompat)
419{
420 int err = 0;
421 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, incompat)) {
422 err = ext4_journal_get_write_access(handle,
423 EXT4_SB(sb)->s_sbh);
424 if (err)
425 return err;
426 EXT4_SET_INCOMPAT_FEATURE(sb, incompat);
427 sb->s_dirt = 1;
428 handle->h_sync = 1;
429 BUFFER_TRACE(EXT4_SB(sb)->s_sbh,
430 "call ext4_journal_dirty_met adata");
431 err = ext4_journal_dirty_metadata(handle,
432 EXT4_SB(sb)->s_sbh);
433 }
434 return err;
435}
436
ac27a0ec
DK		 437/*
438 * Open the external journal device
439 */
617ba13b 440static struct block_device *ext4_blkdev_get(dev_t dev)
ac27a0ec
DK		 441{
442 struct block_device *bdev;
443 char b[BDEVNAME_SIZE];
444
445 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
446 if (IS_ERR(bdev))
447 goto fail;
448 return bdev;
449
450fail:
617ba13b 451 printk(KERN_ERR "EXT4: failed to open journal device %s: %ld\n",
ac27a0ec
DK		 452 __bdevname(dev, b), PTR_ERR(bdev));
453 return NULL;
454}
455
456/*
457 * Release the journal device
458 */
617ba13b 459static int ext4_blkdev_put(struct block_device *bdev)
ac27a0ec
DK		 460{
461 bd_release(bdev);
462 return blkdev_put(bdev);
463}
464
617ba13b 465static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
ac27a0ec
DK		 466{
467 struct block_device *bdev;
468 int ret = -ENODEV;
469
470 bdev = sbi->journal_bdev;
471 if (bdev) {
617ba13b 472 ret = ext4_blkdev_put(bdev);
ac27a0ec
DK		 473 sbi->journal_bdev = NULL;
474 }
475 return ret;
476}
477
478static inline struct inode *orphan_list_entry(struct list_head *l)
479{
617ba13b 480 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
ac27a0ec
DK		 481}
482
617ba13b 483static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
ac27a0ec
DK		 484{
485 struct list_head *l;
486
487 printk(KERN_ERR "sb orphan head is %d\n",
488 le32_to_cpu(sbi->s_es->s_last_orphan));
489
490 printk(KERN_ERR "sb_info orphan list:\n");
491 list_for_each(l, &sbi->s_orphan) {
492 struct inode *inode = orphan_list_entry(l);
493 printk(KERN_ERR " "
494 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
495 inode->i_sb->s_id, inode->i_ino, inode,
496 inode->i_mode, inode->i_nlink,
497 NEXT_ORPHAN(inode));
498 }
499}
500
2b2d6d01 501static void ext4_put_super(struct super_block *sb)
ac27a0ec 502{
617ba13b
MC		 503 struct ext4_sb_info *sbi = EXT4_SB(sb);
504 struct ext4_super_block *es = sbi->s_es;
ac27a0ec
DK		 505 int i;
506
c9de560d 507 ext4_mb_release(sb);
a86c6181 508 ext4_ext_release(sb);
617ba13b 509 ext4_xattr_put_super(sb);
7ffe1ea8
HK		 510 if (jbd2_journal_destroy(sbi->s_journal) < 0)
511 ext4_abort(sb, __func__, "Couldn't clean up the journal");
47b4a50b 512 sbi->s_journal = NULL;
ac27a0ec 513 if (!(sb->s_flags & MS_RDONLY)) {
617ba13b 514 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 515 es->s_state = cpu_to_le16(sbi->s_mount_state);
617ba13b 516 ext4_commit_super(sb, es, 1);
ac27a0ec 517 }
240799cd
TT		 518 if (sbi->s_proc) {
519 remove_proc_entry("inode_readahead_blks", sbi->s_proc);
9f6200bb 520 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 521 }
ac27a0ec
DK		 522
523 for (i = 0; i < sbi->s_gdb_count; i++)
524 brelse(sbi->s_group_desc[i]);
525 kfree(sbi->s_group_desc);
772cb7c8 526 kfree(sbi->s_flex_groups);
ac27a0ec
DK		 527 percpu_counter_destroy(&sbi->s_freeblocks_counter);
528 percpu_counter_destroy(&sbi->s_freeinodes_counter);
529 percpu_counter_destroy(&sbi->s_dirs_counter);
6bc6e63f 530 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
ac27a0ec
DK		 531 brelse(sbi->s_sbh);
532#ifdef CONFIG_QUOTA
533 for (i = 0; i < MAXQUOTAS; i++)
534 kfree(sbi->s_qf_names[i]);
535#endif
536
537 /* Debugging code just in case the in-memory inode orphan list
538 * isn't empty. The on-disk one can be non-empty if we've
539 * detected an error and taken the fs readonly, but the
540 * in-memory list had better be clean by this point. */
541 if (!list_empty(&sbi->s_orphan))
542 dump_orphan_list(sb, sbi);
543 J_ASSERT(list_empty(&sbi->s_orphan));
544
f98393a6 545 invalidate_bdev(sb->s_bdev);
ac27a0ec
DK		 546 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
547 /*
548 * Invalidate the journal device's buffers. We don't want them
549 * floating about in memory - the physical journal device may
550 * hotswapped, and it breaks the `ro-after' testing code.
551 */
552 sync_blockdev(sbi->journal_bdev);
f98393a6 553 invalidate_bdev(sbi->journal_bdev);
617ba13b 554 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 555 }
556 sb->s_fs_info = NULL;
557 kfree(sbi);
558 return;
559}
560
e18b890b 561static struct kmem_cache *ext4_inode_cachep;
ac27a0ec
DK		 562
563/*
564 * Called inside transaction, so use GFP_NOFS
565 */
617ba13b 566static struct inode *ext4_alloc_inode(struct super_block *sb)
ac27a0ec 567{
617ba13b 568 struct ext4_inode_info *ei;
ac27a0ec 569
e6b4f8da 570 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
ac27a0ec
DK		 571 if (!ei)
572 return NULL;
03010a33 573#ifdef CONFIG_EXT4_FS_POSIX_ACL
617ba13b
MC		 574 ei->i_acl = EXT4_ACL_NOT_CACHED;
575 ei->i_default_acl = EXT4_ACL_NOT_CACHED;
ac27a0ec 576#endif
ac27a0ec 577 ei->vfs_inode.i_version = 1;
91246c00 578 ei->vfs_inode.i_data.writeback_index = 0;
a86c6181 579 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
c9de560d
AT		 580 INIT_LIST_HEAD(&ei->i_prealloc_list);
581 spin_lock_init(&ei->i_prealloc_lock);
678aaf48 582 jbd2_journal_init_jbd_inode(&ei->jinode, &ei->vfs_inode);
d2a17637
MC		 583 ei->i_reserved_data_blocks = 0;
584 ei->i_reserved_meta_blocks = 0;
585 ei->i_allocated_meta_blocks = 0;
586 ei->i_delalloc_reserved_flag = 0;
587 spin_lock_init(&(ei->i_block_reservation_lock));
ac27a0ec
DK		 588 return &ei->vfs_inode;
589}
590
617ba13b 591static void ext4_destroy_inode(struct inode *inode)
ac27a0ec 592{
9f7dd93d
VA		 593 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
594 printk("EXT4 Inode %p: orphan list check failed!\n",
595 EXT4_I(inode));
596 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
597 EXT4_I(inode), sizeof(struct ext4_inode_info),
598 true);
599 dump_stack();
600 }
617ba13b 601 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
ac27a0ec
DK		 602}
603
51cc5068 604static void init_once(void *foo)
ac27a0ec 605{
617ba13b 606 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
ac27a0ec 607
a35afb83 608 INIT_LIST_HEAD(&ei->i_orphan);
03010a33 609#ifdef CONFIG_EXT4_FS_XATTR
a35afb83 610 init_rwsem(&ei->xattr_sem);
ac27a0ec 611#endif
0e855ac8 612 init_rwsem(&ei->i_data_sem);
a35afb83 613 inode_init_once(&ei->vfs_inode);
ac27a0ec
DK		 614}
615
616static int init_inodecache(void)
617{
617ba13b
MC		 618 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
619 sizeof(struct ext4_inode_info),
ac27a0ec
DK		 620 0, (SLAB_RECLAIM_ACCOUNT|
621 SLAB_MEM_SPREAD),
20c2df83 622 init_once);
617ba13b 623 if (ext4_inode_cachep == NULL)
ac27a0ec
DK		 624 return -ENOMEM;
625 return 0;
626}
627
628static void destroy_inodecache(void)
629{
617ba13b 630 kmem_cache_destroy(ext4_inode_cachep);
ac27a0ec
DK		 631}
632
617ba13b 633static void ext4_clear_inode(struct inode *inode)
ac27a0ec 634{
03010a33 635#ifdef CONFIG_EXT4_FS_POSIX_ACL
617ba13b
MC		 636 if (EXT4_I(inode)->i_acl &&
637 EXT4_I(inode)->i_acl != EXT4_ACL_NOT_CACHED) {
638 posix_acl_release(EXT4_I(inode)->i_acl);
639 EXT4_I(inode)->i_acl = EXT4_ACL_NOT_CACHED;
640 }
641 if (EXT4_I(inode)->i_default_acl &&
642 EXT4_I(inode)->i_default_acl != EXT4_ACL_NOT_CACHED) {
643 posix_acl_release(EXT4_I(inode)->i_default_acl);
644 EXT4_I(inode)->i_default_acl = EXT4_ACL_NOT_CACHED;
ac27a0ec
DK		 645 }
646#endif
c2ea3fde 647 ext4_discard_preallocations(inode);
678aaf48
JK		 648 jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal,
649 &EXT4_I(inode)->jinode);
ac27a0ec
DK		 650}
651
2b2d6d01
TT		 652static inline void ext4_show_quota_options(struct seq_file *seq,
653 struct super_block *sb)
ac27a0ec
DK		 654{
655#if defined(CONFIG_QUOTA)
617ba13b 656 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 657
658 if (sbi->s_jquota_fmt)
659 seq_printf(seq, ",jqfmt=%s",
af5bc92d 660 (sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold" : "vfsv0");
ac27a0ec
DK		 661
662 if (sbi->s_qf_names[USRQUOTA])
663 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
664
665 if (sbi->s_qf_names[GRPQUOTA])
666 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
667
617ba13b 668 if (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA)
ac27a0ec
DK		 669 seq_puts(seq, ",usrquota");
670
617ba13b 671 if (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)
ac27a0ec
DK		 672 seq_puts(seq, ",grpquota");
673#endif
674}
675
d9c9bef1
MS		 676/*
677 * Show an option if
678 * - it's set to a non-default value OR
679 * - if the per-sb default is different from the global default
680 */
617ba13b 681static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
ac27a0ec 682{
aa22df2c
AK		 683 int def_errors;
684 unsigned long def_mount_opts;
ac27a0ec 685 struct super_block *sb = vfs->mnt_sb;
d9c9bef1
MS		 686 struct ext4_sb_info *sbi = EXT4_SB(sb);
687 struct ext4_super_block *es = sbi->s_es;
d9c9bef1
MS		 688
689 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
aa22df2c 690 def_errors = le16_to_cpu(es->s_errors);
d9c9bef1
MS		 691
692 if (sbi->s_sb_block != 1)
693 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
694 if (test_opt(sb, MINIX_DF))
695 seq_puts(seq, ",minixdf");
aa22df2c 696 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
d9c9bef1
MS		 697 seq_puts(seq, ",grpid");
698 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
699 seq_puts(seq, ",nogrpid");
700 if (sbi->s_resuid != EXT4_DEF_RESUID ||
701 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
702 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
703 }
704 if (sbi->s_resgid != EXT4_DEF_RESGID ||
705 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
706 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
707 }
bb4f397a 708 if (test_opt(sb, ERRORS_RO)) {
d9c9bef1 709 if (def_errors == EXT4_ERRORS_PANIC ||
bb4f397a
AK		 710 def_errors == EXT4_ERRORS_CONTINUE) {
711 seq_puts(seq, ",errors=remount-ro");
d9c9bef1
MS		 712 }
713 }
aa22df2c 714 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
bb4f397a 715 seq_puts(seq, ",errors=continue");
aa22df2c 716 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
d9c9bef1 717 seq_puts(seq, ",errors=panic");
aa22df2c 718 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
d9c9bef1 719 seq_puts(seq, ",nouid32");
aa22df2c 720 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
d9c9bef1
MS		 721 seq_puts(seq, ",debug");
722 if (test_opt(sb, OLDALLOC))
723 seq_puts(seq, ",oldalloc");
03010a33 724#ifdef CONFIG_EXT4_FS_XATTR
aa22df2c
AK		 725 if (test_opt(sb, XATTR_USER) &&
726 !(def_mount_opts & EXT4_DEFM_XATTR_USER))
d9c9bef1
MS		 727 seq_puts(seq, ",user_xattr");
728 if (!test_opt(sb, XATTR_USER) &&
729 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
730 seq_puts(seq, ",nouser_xattr");
731 }
732#endif
03010a33 733#ifdef CONFIG_EXT4_FS_POSIX_ACL
aa22df2c 734 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
d9c9bef1
MS		 735 seq_puts(seq, ",acl");
736 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
737 seq_puts(seq, ",noacl");
738#endif
739 if (!test_opt(sb, RESERVATION))
740 seq_puts(seq, ",noreservation");
741 if (sbi->s_commit_interval) {
742 seq_printf(seq, ",commit=%u",
743 (unsigned) (sbi->s_commit_interval / HZ));
744 }
571640ca
ES		 745 /*
746 * We're changing the default of barrier mount option, so
747 * let's always display its mount state so it's clear what its
748 * status is.
749 */
750 seq_puts(seq, ",barrier=");
751 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
cd0b6a39
TT		 752 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
753 seq_puts(seq, ",journal_async_commit");
d9c9bef1
MS		 754 if (test_opt(sb, NOBH))
755 seq_puts(seq, ",nobh");
756 if (!test_opt(sb, EXTENTS))
757 seq_puts(seq, ",noextents");
25ec56b5
JNC		 758 if (test_opt(sb, I_VERSION))
759 seq_puts(seq, ",i_version");
dd919b98
AK		 760 if (!test_opt(sb, DELALLOC))
761 seq_puts(seq, ",nodelalloc");
762
ac27a0ec 763
cb45bbe4
MS		 764 if (sbi->s_stripe)
765 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
aa22df2c
AK		 766 /*
767 * journal mode get enabled in different ways
768 * So just print the value even if we didn't specify it
769 */
617ba13b 770 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
ac27a0ec 771 seq_puts(seq, ",data=journal");
617ba13b 772 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
ac27a0ec 773 seq_puts(seq, ",data=ordered");
617ba13b 774 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
ac27a0ec
DK		 775 seq_puts(seq, ",data=writeback");
776
240799cd
TT		 777 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
778 seq_printf(seq, ",inode_readahead_blks=%u",
779 sbi->s_inode_readahead_blks);
780
617ba13b 781 ext4_show_quota_options(seq, sb);
ac27a0ec
DK		 782 return 0;
783}
784
785
1b961ac0
CH		 786static struct inode *ext4_nfs_get_inode(struct super_block *sb,
787 u64 ino, u32 generation)
ac27a0ec 788{
ac27a0ec 789 struct inode *inode;
ac27a0ec 790
617ba13b 791 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
ac27a0ec 792 return ERR_PTR(-ESTALE);
617ba13b 793 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
ac27a0ec
DK		 794 return ERR_PTR(-ESTALE);
795
796 /* iget isn't really right if the inode is currently unallocated!!
797 *
617ba13b 798 * ext4_read_inode will return a bad_inode if the inode had been
ac27a0ec
DK		 799 * deleted, so we should be safe.
800 *
801 * Currently we don't know the generation for parent directory, so
802 * a generation of 0 means "accept any"
803 */
1d1fe1ee
DH		 804 inode = ext4_iget(sb, ino);
805 if (IS_ERR(inode))
806 return ERR_CAST(inode);
807 if (generation && inode->i_generation != generation) {
ac27a0ec
DK		 808 iput(inode);
809 return ERR_PTR(-ESTALE);
810 }
1b961ac0
CH		 811
812 return inode;
813}
814
815static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
816 int fh_len, int fh_type)
817{
818 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
819 ext4_nfs_get_inode);
820}
821
822static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
823 int fh_len, int fh_type)
824{
825 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
826 ext4_nfs_get_inode);
ac27a0ec
DK		 827}
828
829#ifdef CONFIG_QUOTA
af5bc92d 830#define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
2b2d6d01 831#define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
ac27a0ec 832
617ba13b
MC		 833static int ext4_dquot_initialize(struct inode *inode, int type);
834static int ext4_dquot_drop(struct inode *inode);
835static int ext4_write_dquot(struct dquot *dquot);
836static int ext4_acquire_dquot(struct dquot *dquot);
837static int ext4_release_dquot(struct dquot *dquot);
838static int ext4_mark_dquot_dirty(struct dquot *dquot);
839static int ext4_write_info(struct super_block *sb, int type);
6f28e087
JK		 840static int ext4_quota_on(struct super_block *sb, int type, int format_id,
841 char *path, int remount);
617ba13b
MC		 842static int ext4_quota_on_mount(struct super_block *sb, int type);
843static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 844 size_t len, loff_t off);
617ba13b 845static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 846 const char *data, size_t len, loff_t off);
847
617ba13b
MC		 848static struct dquot_operations ext4_quota_operations = {
849 .initialize = ext4_dquot_initialize,
850 .drop = ext4_dquot_drop,
ac27a0ec
DK		 851 .alloc_space = dquot_alloc_space,
852 .alloc_inode = dquot_alloc_inode,
853 .free_space = dquot_free_space,
854 .free_inode = dquot_free_inode,
855 .transfer = dquot_transfer,
617ba13b
MC		 856 .write_dquot = ext4_write_dquot,
857 .acquire_dquot = ext4_acquire_dquot,
858 .release_dquot = ext4_release_dquot,
859 .mark_dirty = ext4_mark_dquot_dirty,
860 .write_info = ext4_write_info
ac27a0ec
DK		 861};
862
617ba13b
MC		 863static struct quotactl_ops ext4_qctl_operations = {
864 .quota_on = ext4_quota_on,
ac27a0ec
DK		 865 .quota_off = vfs_quota_off,
866 .quota_sync = vfs_quota_sync,
867 .get_info = vfs_get_dqinfo,
868 .set_info = vfs_set_dqinfo,
869 .get_dqblk = vfs_get_dqblk,
870 .set_dqblk = vfs_set_dqblk
871};
872#endif
873
ee9b6d61 874static const struct super_operations ext4_sops = {
617ba13b
MC		 875 .alloc_inode = ext4_alloc_inode,
876 .destroy_inode = ext4_destroy_inode,
617ba13b
MC		 877 .write_inode = ext4_write_inode,
878 .dirty_inode = ext4_dirty_inode,
879 .delete_inode = ext4_delete_inode,
880 .put_super = ext4_put_super,
881 .write_super = ext4_write_super,
882 .sync_fs = ext4_sync_fs,
883 .write_super_lockfs = ext4_write_super_lockfs,
884 .unlockfs = ext4_unlockfs,
885 .statfs = ext4_statfs,
886 .remount_fs = ext4_remount,
887 .clear_inode = ext4_clear_inode,
888 .show_options = ext4_show_options,
ac27a0ec 889#ifdef CONFIG_QUOTA
617ba13b
MC		 890 .quota_read = ext4_quota_read,
891 .quota_write = ext4_quota_write,
ac27a0ec
DK		 892#endif
893};
894
39655164 895static const struct export_operations ext4_export_ops = {
1b961ac0
CH		 896 .fh_to_dentry = ext4_fh_to_dentry,
897 .fh_to_parent = ext4_fh_to_parent,
617ba13b 898 .get_parent = ext4_get_parent,
ac27a0ec
DK		 899};
900
901enum {
902 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
903 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
904 Opt_nouid32, Opt_nocheck, Opt_debug, Opt_oldalloc, Opt_orlov,
905 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
906 Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh,
907 Opt_commit, Opt_journal_update, Opt_journal_inum, Opt_journal_dev,
818d276c 908 Opt_journal_checksum, Opt_journal_async_commit,
ac27a0ec
DK		 909 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
910 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
911 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
912 Opt_ignore, Opt_barrier, Opt_err, Opt_resize, Opt_usrquota,
25ec56b5 913 Opt_grpquota, Opt_extents, Opt_noextents, Opt_i_version,
dd919b98 914 Opt_mballoc, Opt_nomballoc, Opt_stripe, Opt_delalloc, Opt_nodelalloc,
240799cd 915 Opt_inode_readahead_blks
ac27a0ec
DK		 916};
917
918static match_table_t tokens = {
919 {Opt_bsd_df, "bsddf"},
920 {Opt_minix_df, "minixdf"},
921 {Opt_grpid, "grpid"},
922 {Opt_grpid, "bsdgroups"},
923 {Opt_nogrpid, "nogrpid"},
924 {Opt_nogrpid, "sysvgroups"},
925 {Opt_resgid, "resgid=%u"},
926 {Opt_resuid, "resuid=%u"},
927 {Opt_sb, "sb=%u"},
928 {Opt_err_cont, "errors=continue"},
929 {Opt_err_panic, "errors=panic"},
930 {Opt_err_ro, "errors=remount-ro"},
931 {Opt_nouid32, "nouid32"},
932 {Opt_nocheck, "nocheck"},
933 {Opt_nocheck, "check=none"},
934 {Opt_debug, "debug"},
935 {Opt_oldalloc, "oldalloc"},
936 {Opt_orlov, "orlov"},
937 {Opt_user_xattr, "user_xattr"},
938 {Opt_nouser_xattr, "nouser_xattr"},
939 {Opt_acl, "acl"},
940 {Opt_noacl, "noacl"},
941 {Opt_reservation, "reservation"},
942 {Opt_noreservation, "noreservation"},
943 {Opt_noload, "noload"},
944 {Opt_nobh, "nobh"},
945 {Opt_bh, "bh"},
946 {Opt_commit, "commit=%u"},
947 {Opt_journal_update, "journal=update"},
948 {Opt_journal_inum, "journal=%u"},
949 {Opt_journal_dev, "journal_dev=%u"},
818d276c
GS		 950 {Opt_journal_checksum, "journal_checksum"},
951 {Opt_journal_async_commit, "journal_async_commit"},
ac27a0ec
DK		 952 {Opt_abort, "abort"},
953 {Opt_data_journal, "data=journal"},
954 {Opt_data_ordered, "data=ordered"},
955 {Opt_data_writeback, "data=writeback"},
956 {Opt_offusrjquota, "usrjquota="},
957 {Opt_usrjquota, "usrjquota=%s"},
958 {Opt_offgrpjquota, "grpjquota="},
959 {Opt_grpjquota, "grpjquota=%s"},
960 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
961 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
962 {Opt_grpquota, "grpquota"},
963 {Opt_noquota, "noquota"},
964 {Opt_quota, "quota"},
965 {Opt_usrquota, "usrquota"},
966 {Opt_barrier, "barrier=%u"},
a86c6181 967 {Opt_extents, "extents"},
1e2462f9 968 {Opt_noextents, "noextents"},
25ec56b5 969 {Opt_i_version, "i_version"},
c9de560d
AT		 970 {Opt_mballoc, "mballoc"},
971 {Opt_nomballoc, "nomballoc"},
972 {Opt_stripe, "stripe=%u"},
ac27a0ec 973 {Opt_resize, "resize"},
64769240 974 {Opt_delalloc, "delalloc"},
dd919b98 975 {Opt_nodelalloc, "nodelalloc"},
240799cd 976 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
f3f12faa 977 {Opt_err, NULL},
ac27a0ec
DK		 978};
979
617ba13b 980static ext4_fsblk_t get_sb_block(void **data)
ac27a0ec 981{
617ba13b 982 ext4_fsblk_t sb_block;
ac27a0ec
DK		 983 char *options = (char *) *data;
984
985 if (!options || strncmp(options, "sb=", 3) != 0)
986 return 1; /* Default location */
987 options += 3;
617ba13b 988 /*todo: use simple_strtoll with >32bit ext4 */
ac27a0ec
DK		 989 sb_block = simple_strtoul(options, &options, 0);
990 if (*options && *options != ',') {
4776004f 991 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
ac27a0ec
DK		 992 (char *) *data);
993 return 1;
994 }
995 if (*options == ',')
996 options++;
997 *data = (void *) options;
998 return sb_block;
999}
1000
2b2d6d01
TT		 1001static int parse_options(char *options, struct super_block *sb,
1002 unsigned int *inum, unsigned long *journal_devnum,
1003 ext4_fsblk_t *n_blocks_count, int is_remount)
ac27a0ec 1004{
617ba13b 1005 struct ext4_sb_info *sbi = EXT4_SB(sb);
2b2d6d01 1006 char *p;
ac27a0ec
DK		 1007 substring_t args[MAX_OPT_ARGS];
1008 int data_opt = 0;
1009 int option;
1010#ifdef CONFIG_QUOTA
dfc5d03f 1011 int qtype, qfmt;
ac27a0ec
DK		 1012 char *qname;
1013#endif
c07651b5 1014 ext4_fsblk_t last_block;
ac27a0ec
DK		 1015
1016 if (!options)
1017 return 1;
1018
2b2d6d01 1019 while ((p = strsep(&options, ",")) != NULL) {
ac27a0ec
DK		 1020 int token;
1021 if (!*p)
1022 continue;
1023
1024 token = match_token(p, tokens, args);
1025 switch (token) {
1026 case Opt_bsd_df:
2b2d6d01 1027 clear_opt(sbi->s_mount_opt, MINIX_DF);
ac27a0ec
DK		 1028 break;
1029 case Opt_minix_df:
2b2d6d01 1030 set_opt(sbi->s_mount_opt, MINIX_DF);
ac27a0ec
DK		 1031 break;
1032 case Opt_grpid:
2b2d6d01 1033 set_opt(sbi->s_mount_opt, GRPID);
ac27a0ec
DK		 1034 break;
1035 case Opt_nogrpid:
2b2d6d01 1036 clear_opt(sbi->s_mount_opt, GRPID);
ac27a0ec
DK		 1037 break;
1038 case Opt_resuid:
1039 if (match_int(&args[0], &option))
1040 return 0;
1041 sbi->s_resuid = option;
1042 break;
1043 case Opt_resgid:
1044 if (match_int(&args[0], &option))
1045 return 0;
1046 sbi->s_resgid = option;
1047 break;
1048 case Opt_sb:
1049 /* handled by get_sb_block() instead of here */
1050 /* *sb_block = match_int(&args[0]); */
1051 break;
1052 case Opt_err_panic:
2b2d6d01
TT		 1053 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1054 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1055 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
ac27a0ec
DK		 1056 break;
1057 case Opt_err_ro:
2b2d6d01
TT		 1058 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1059 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1060 set_opt(sbi->s_mount_opt, ERRORS_RO);
ac27a0ec
DK		 1061 break;
1062 case Opt_err_cont:
2b2d6d01
TT		 1063 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1064 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1065 set_opt(sbi->s_mount_opt, ERRORS_CONT);
ac27a0ec
DK		 1066 break;
1067 case Opt_nouid32:
2b2d6d01 1068 set_opt(sbi->s_mount_opt, NO_UID32);
ac27a0ec
DK		 1069 break;
1070 case Opt_nocheck:
2b2d6d01 1071 clear_opt(sbi->s_mount_opt, CHECK);
ac27a0ec
DK		 1072 break;
1073 case Opt_debug:
2b2d6d01 1074 set_opt(sbi->s_mount_opt, DEBUG);
ac27a0ec
DK		 1075 break;
1076 case Opt_oldalloc:
2b2d6d01 1077 set_opt(sbi->s_mount_opt, OLDALLOC);
ac27a0ec
DK		 1078 break;
1079 case Opt_orlov:
2b2d6d01 1080 clear_opt(sbi->s_mount_opt, OLDALLOC);
ac27a0ec 1081 break;
03010a33 1082#ifdef CONFIG_EXT4_FS_XATTR
ac27a0ec 1083 case Opt_user_xattr:
2b2d6d01 1084 set_opt(sbi->s_mount_opt, XATTR_USER);
ac27a0ec
DK		 1085 break;
1086 case Opt_nouser_xattr:
2b2d6d01 1087 clear_opt(sbi->s_mount_opt, XATTR_USER);
ac27a0ec
DK		 1088 break;
1089#else
1090 case Opt_user_xattr:
1091 case Opt_nouser_xattr:
4776004f
TT		 1092 printk(KERN_ERR "EXT4 (no)user_xattr options "
1093 "not supported\n");
ac27a0ec
DK		 1094 break;
1095#endif
03010a33 1096#ifdef CONFIG_EXT4_FS_POSIX_ACL
ac27a0ec
DK		 1097 case Opt_acl:
1098 set_opt(sbi->s_mount_opt, POSIX_ACL);
1099 break;
1100 case Opt_noacl:
1101 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1102 break;
1103#else
1104 case Opt_acl:
1105 case Opt_noacl:
4776004f
TT		 1106 printk(KERN_ERR "EXT4 (no)acl options "
1107 "not supported\n");
ac27a0ec
DK		 1108 break;
1109#endif
1110 case Opt_reservation:
1111 set_opt(sbi->s_mount_opt, RESERVATION);
1112 break;
1113 case Opt_noreservation:
1114 clear_opt(sbi->s_mount_opt, RESERVATION);
1115 break;
1116 case Opt_journal_update:
1117 /* @@@ FIXME */
1118 /* Eventually we will want to be able to create
1119 a journal file here. For now, only allow the
1120 user to specify an existing inode to be the
1121 journal file. */
1122 if (is_remount) {
617ba13b 1123 printk(KERN_ERR "EXT4-fs: cannot specify "
ac27a0ec
DK		 1124 "journal on remount\n");
1125 return 0;
1126 }
2b2d6d01 1127 set_opt(sbi->s_mount_opt, UPDATE_JOURNAL);
ac27a0ec
DK		 1128 break;
1129 case Opt_journal_inum:
1130 if (is_remount) {
617ba13b 1131 printk(KERN_ERR "EXT4-fs: cannot specify "
ac27a0ec
DK		 1132 "journal on remount\n");
1133 return 0;
1134 }
1135 if (match_int(&args[0], &option))
1136 return 0;
1137 *inum = option;
1138 break;
1139 case Opt_journal_dev:
1140 if (is_remount) {
617ba13b 1141 printk(KERN_ERR "EXT4-fs: cannot specify "
ac27a0ec
DK		 1142 "journal on remount\n");
1143 return 0;
1144 }
1145 if (match_int(&args[0], &option))
1146 return 0;
1147 *journal_devnum = option;
1148 break;
818d276c
GS		 1149 case Opt_journal_checksum:
1150 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1151 break;
1152 case Opt_journal_async_commit:
1153 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
1154 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1155 break;
ac27a0ec 1156 case Opt_noload:
2b2d6d01 1157 set_opt(sbi->s_mount_opt, NOLOAD);
ac27a0ec
DK		 1158 break;
1159 case Opt_commit:
1160 if (match_int(&args[0], &option))
1161 return 0;
1162 if (option < 0)
1163 return 0;
1164 if (option == 0)
cd02ff0b 1165 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
ac27a0ec
DK		 1166 sbi->s_commit_interval = HZ * option;
1167 break;
1168 case Opt_data_journal:
617ba13b 1169 data_opt = EXT4_MOUNT_JOURNAL_DATA;
ac27a0ec
DK		 1170 goto datacheck;
1171 case Opt_data_ordered:
617ba13b 1172 data_opt = EXT4_MOUNT_ORDERED_DATA;
ac27a0ec
DK		 1173 goto datacheck;
1174 case Opt_data_writeback:
617ba13b 1175 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
ac27a0ec
DK		 1176 datacheck:
1177 if (is_remount) {
617ba13b 1178 if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
ac27a0ec
DK		 1179 != data_opt) {
1180 printk(KERN_ERR
617ba13b 1181 "EXT4-fs: cannot change data "
ac27a0ec
DK		 1182 "mode on remount\n");
1183 return 0;
1184 }
1185 } else {
617ba13b 1186 sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
ac27a0ec
DK		 1187 sbi->s_mount_opt |= data_opt;
1188 }
1189 break;
1190#ifdef CONFIG_QUOTA
1191 case Opt_usrjquota:
1192 qtype = USRQUOTA;
1193 goto set_qf_name;
1194 case Opt_grpjquota:
1195 qtype = GRPQUOTA;
1196set_qf_name:
dfc5d03f
JK		 1197 if ((sb_any_quota_enabled(sb) ||
1198 sb_any_quota_suspended(sb)) &&
1199 !sbi->s_qf_names[qtype]) {
ac27a0ec 1200 printk(KERN_ERR
4776004f
TT		 1201 "EXT4-fs: Cannot change journaled "
1202 "quota options when quota turned on.\n");
ac27a0ec
DK		 1203 return 0;
1204 }
1205 qname = match_strdup(&args[0]);
1206 if (!qname) {
1207 printk(KERN_ERR
617ba13b 1208 "EXT4-fs: not enough memory for "
ac27a0ec
DK		 1209 "storing quotafile name.\n");
1210 return 0;
1211 }
1212 if (sbi->s_qf_names[qtype] &&
1213 strcmp(sbi->s_qf_names[qtype], qname)) {
1214 printk(KERN_ERR
617ba13b 1215 "EXT4-fs: %s quota file already "
ac27a0ec
DK		 1216 "specified.\n", QTYPE2NAME(qtype));
1217 kfree(qname);
1218 return 0;
1219 }
1220 sbi->s_qf_names[qtype] = qname;
1221 if (strchr(sbi->s_qf_names[qtype], '/')) {
1222 printk(KERN_ERR
617ba13b 1223 "EXT4-fs: quotafile must be on "
ac27a0ec
DK		 1224 "filesystem root.\n");
1225 kfree(sbi->s_qf_names[qtype]);
1226 sbi->s_qf_names[qtype] = NULL;
1227 return 0;
1228 }
1229 set_opt(sbi->s_mount_opt, QUOTA);
1230 break;
1231 case Opt_offusrjquota:
1232 qtype = USRQUOTA;
1233 goto clear_qf_name;
1234 case Opt_offgrpjquota:
1235 qtype = GRPQUOTA;
1236clear_qf_name:
dfc5d03f
JK		 1237 if ((sb_any_quota_enabled(sb) ||
1238 sb_any_quota_suspended(sb)) &&
1239 sbi->s_qf_names[qtype]) {
617ba13b 1240 printk(KERN_ERR "EXT4-fs: Cannot change "
2c8be6b2 1241 "journaled quota options when "
ac27a0ec
DK		 1242 "quota turned on.\n");
1243 return 0;
1244 }
1245 /*
1246 * The space will be released later when all options
1247 * are confirmed to be correct
1248 */
1249 sbi->s_qf_names[qtype] = NULL;
1250 break;
1251 case Opt_jqfmt_vfsold:
dfc5d03f
JK		 1252 qfmt = QFMT_VFS_OLD;
1253 goto set_qf_format;
ac27a0ec 1254 case Opt_jqfmt_vfsv0:
dfc5d03f
JK		 1255 qfmt = QFMT_VFS_V0;
1256set_qf_format:
1257 if ((sb_any_quota_enabled(sb) ||
1258 sb_any_quota_suspended(sb)) &&
1259 sbi->s_jquota_fmt != qfmt) {
1260 printk(KERN_ERR "EXT4-fs: Cannot change "
1261 "journaled quota options when "
1262 "quota turned on.\n");
1263 return 0;
1264 }
1265 sbi->s_jquota_fmt = qfmt;
ac27a0ec
DK		 1266 break;
1267 case Opt_quota:
1268 case Opt_usrquota:
1269 set_opt(sbi->s_mount_opt, QUOTA);
1270 set_opt(sbi->s_mount_opt, USRQUOTA);
1271 break;
1272 case Opt_grpquota:
1273 set_opt(sbi->s_mount_opt, QUOTA);
1274 set_opt(sbi->s_mount_opt, GRPQUOTA);
1275 break;
1276 case Opt_noquota:
1277 if (sb_any_quota_enabled(sb)) {
617ba13b 1278 printk(KERN_ERR "EXT4-fs: Cannot change quota "
ac27a0ec
DK		 1279 "options when quota turned on.\n");
1280 return 0;
1281 }
1282 clear_opt(sbi->s_mount_opt, QUOTA);
1283 clear_opt(sbi->s_mount_opt, USRQUOTA);
1284 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1285 break;
1286#else
1287 case Opt_quota:
1288 case Opt_usrquota:
1289 case Opt_grpquota:
cd59e7b9
JK		 1290 printk(KERN_ERR
1291 "EXT4-fs: quota options not supported.\n");
1292 break;
ac27a0ec
DK		 1293 case Opt_usrjquota:
1294 case Opt_grpjquota:
1295 case Opt_offusrjquota:
1296 case Opt_offgrpjquota:
1297 case Opt_jqfmt_vfsold:
1298 case Opt_jqfmt_vfsv0:
1299 printk(KERN_ERR
cd59e7b9 1300 "EXT4-fs: journaled quota options not "
ac27a0ec
DK		 1301 "supported.\n");
1302 break;
1303 case Opt_noquota:
1304 break;
1305#endif
1306 case Opt_abort:
1307 set_opt(sbi->s_mount_opt, ABORT);
1308 break;
1309 case Opt_barrier:
1310 if (match_int(&args[0], &option))
1311 return 0;
1312 if (option)
1313 set_opt(sbi->s_mount_opt, BARRIER);
1314 else
1315 clear_opt(sbi->s_mount_opt, BARRIER);
1316 break;
1317 case Opt_ignore:
1318 break;
1319 case Opt_resize:
1320 if (!is_remount) {
617ba13b 1321 printk("EXT4-fs: resize option only available "
ac27a0ec
DK		 1322 "for remount\n");
1323 return 0;
1324 }
1325 if (match_int(&args[0], &option) != 0)
1326 return 0;
1327 *n_blocks_count = option;
1328 break;
1329 case Opt_nobh:
1330 set_opt(sbi->s_mount_opt, NOBH);
1331 break;
1332 case Opt_bh:
1333 clear_opt(sbi->s_mount_opt, NOBH);
1334 break;
a86c6181 1335 case Opt_extents:
e4079a11
ES		 1336 if (!EXT4_HAS_INCOMPAT_FEATURE(sb,
1337 EXT4_FEATURE_INCOMPAT_EXTENTS)) {
1338 ext4_warning(sb, __func__,
1339 "extents feature not enabled "
1340 "on this filesystem, use tune2fs\n");
1341 return 0;
1342 }
2b2d6d01 1343 set_opt(sbi->s_mount_opt, EXTENTS);
a86c6181 1344 break;
1e2462f9 1345 case Opt_noextents:
c07651b5
AK		 1346 /*
1347 * When e2fsprogs support resizing an already existing
1348 * ext3 file system to greater than 2**32 we need to
1349 * add support to block allocator to handle growing
1350 * already existing block mapped inode so that blocks
1351 * allocated for them fall within 2**32
1352 */
1353 last_block = ext4_blocks_count(sbi->s_es) - 1;
1354 if (last_block > 0xffffffffULL) {
1355 printk(KERN_ERR "EXT4-fs: Filesystem too "
1356 "large to mount with "
1357 "-o noextents options\n");
1358 return 0;
1359 }
2b2d6d01 1360 clear_opt(sbi->s_mount_opt, EXTENTS);
1e2462f9 1361 break;
25ec56b5
JNC		 1362 case Opt_i_version:
1363 set_opt(sbi->s_mount_opt, I_VERSION);
1364 sb->s_flags |= MS_I_VERSION;
1365 break;
dd919b98
AK		 1366 case Opt_nodelalloc:
1367 clear_opt(sbi->s_mount_opt, DELALLOC);
1368 break;
c9de560d
AT		 1369 case Opt_stripe:
1370 if (match_int(&args[0], &option))
1371 return 0;
1372 if (option < 0)
1373 return 0;
1374 sbi->s_stripe = option;
1375 break;
64769240
AT		 1376 case Opt_delalloc:
1377 set_opt(sbi->s_mount_opt, DELALLOC);
1378 break;
240799cd
TT		 1379 case Opt_inode_readahead_blks:
1380 if (match_int(&args[0], &option))
1381 return 0;
1382 if (option < 0 || option > (1 << 30))
1383 return 0;
1384 sbi->s_inode_readahead_blks = option;
1385 break;
ac27a0ec 1386 default:
2b2d6d01
TT		 1387 printk(KERN_ERR
1388 "EXT4-fs: Unrecognized mount option \"%s\" "
1389 "or missing value\n", p);
ac27a0ec
DK		 1390 return 0;
1391 }
1392 }
1393#ifdef CONFIG_QUOTA
1394 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
617ba13b 1395 if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
ac27a0ec
DK		 1396 sbi->s_qf_names[USRQUOTA])
1397 clear_opt(sbi->s_mount_opt, USRQUOTA);
1398
617ba13b 1399 if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
ac27a0ec
DK		 1400 sbi->s_qf_names[GRPQUOTA])
1401 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1402
1403 if ((sbi->s_qf_names[USRQUOTA] &&
617ba13b 1404 (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
ac27a0ec 1405 (sbi->s_qf_names[GRPQUOTA] &&
617ba13b
MC		 1406 (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
1407 printk(KERN_ERR "EXT4-fs: old and new quota "
ac27a0ec
DK		 1408 "format mixing.\n");
1409 return 0;
1410 }
1411
1412 if (!sbi->s_jquota_fmt) {
2c8be6b2 1413 printk(KERN_ERR "EXT4-fs: journaled quota format "
ac27a0ec
DK		 1414 "not specified.\n");
1415 return 0;
1416 }
1417 } else {
1418 if (sbi->s_jquota_fmt) {
2c8be6b2
JK		 1419 printk(KERN_ERR "EXT4-fs: journaled quota format "
1420 "specified with no journaling "
ac27a0ec
DK		 1421 "enabled.\n");
1422 return 0;
1423 }
1424 }
1425#endif
1426 return 1;
1427}
1428
617ba13b 1429static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 1430 int read_only)
1431{
617ba13b 1432 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 1433 int res = 0;
1434
617ba13b 1435 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
2b2d6d01
TT		 1436 printk(KERN_ERR "EXT4-fs warning: revision level too high, "
1437 "forcing read-only mode\n");
ac27a0ec
DK		 1438 res = MS_RDONLY;
1439 }
1440 if (read_only)
1441 return res;
617ba13b 1442 if (!(sbi->s_mount_state & EXT4_VALID_FS))
2b2d6d01
TT		 1443 printk(KERN_WARNING "EXT4-fs warning: mounting unchecked fs, "
1444 "running e2fsck is recommended\n");
617ba13b 1445 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
2b2d6d01
TT		 1446 printk(KERN_WARNING
1447 "EXT4-fs warning: mounting fs with errors, "
1448 "running e2fsck is recommended\n");
ac27a0ec
DK		 1449 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1450 le16_to_cpu(es->s_mnt_count) >=
1451 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
2b2d6d01
TT		 1452 printk(KERN_WARNING
1453 "EXT4-fs warning: maximal mount count reached, "
1454 "running e2fsck is recommended\n");
ac27a0ec
DK		 1455 else if (le32_to_cpu(es->s_checkinterval) &&
1456 (le32_to_cpu(es->s_lastcheck) +
1457 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
2b2d6d01
TT		 1458 printk(KERN_WARNING
1459 "EXT4-fs warning: checktime reached, "
1460 "running e2fsck is recommended\n");
ac27a0ec
DK		 1461#if 0
1462 /* @@@ We _will_ want to clear the valid bit if we find
63f57933
AM		 1463 * inconsistencies, to force a fsck at reboot. But for
1464 * a plain journaled filesystem we can keep it set as
1465 * valid forever! :)
1466 */
216c34b2 1467 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec
DK		 1468#endif
1469 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 1470 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 1471 le16_add_cpu(&es->s_mnt_count, 1);
ac27a0ec 1472 es->s_mtime = cpu_to_le32(get_seconds());
617ba13b
MC		 1473 ext4_update_dynamic_rev(sb);
1474 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 1475
617ba13b 1476 ext4_commit_super(sb, es, 1);
ac27a0ec 1477 if (test_opt(sb, DEBUG))
617ba13b 1478 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%lu, "
ac27a0ec
DK		 1479 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
1480 sb->s_blocksize,
1481 sbi->s_groups_count,
617ba13b
MC		 1482 EXT4_BLOCKS_PER_GROUP(sb),
1483 EXT4_INODES_PER_GROUP(sb),
ac27a0ec
DK		 1484 sbi->s_mount_opt);
1485
05496769
TT		 1486 printk(KERN_INFO "EXT4 FS on %s, %s journal on %s\n",
1487 sb->s_id, EXT4_SB(sb)->s_journal->j_inode ? "internal" :
1488 "external", EXT4_SB(sb)->s_journal->j_devname);
ac27a0ec
DK		 1489 return res;
1490}
1491
772cb7c8
JS		 1492static int ext4_fill_flex_info(struct super_block *sb)
1493{
1494 struct ext4_sb_info *sbi = EXT4_SB(sb);
1495 struct ext4_group_desc *gdp = NULL;
1496 struct buffer_head *bh;
1497 ext4_group_t flex_group_count;
1498 ext4_group_t flex_group;
1499 int groups_per_flex = 0;
1500 __u64 block_bitmap = 0;
1501 int i;
1502
1503 if (!sbi->s_es->s_log_groups_per_flex) {
1504 sbi->s_log_groups_per_flex = 0;
1505 return 1;
1506 }
1507
1508 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1509 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1510
c62a11fd
FB		 1511 /* We allocate both existing and potentially added groups */
1512 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
1513 ((sbi->s_es->s_reserved_gdt_blocks +1 ) <<
1514 EXT4_DESC_PER_BLOCK_BITS(sb))) /
1515 groups_per_flex;
ec05e868 1516 sbi->s_flex_groups = kzalloc(flex_group_count *
772cb7c8
JS		 1517 sizeof(struct flex_groups), GFP_KERNEL);
1518 if (sbi->s_flex_groups == NULL) {
ec05e868
LZ		 1519 printk(KERN_ERR "EXT4-fs: not enough memory for "
1520 "%lu flex groups\n", flex_group_count);
772cb7c8
JS		 1521 goto failed;
1522 }
772cb7c8
JS		 1523
1524 gdp = ext4_get_group_desc(sb, 1, &bh);
1525 block_bitmap = ext4_block_bitmap(sb, gdp) - 1;
1526
1527 for (i = 0; i < sbi->s_groups_count; i++) {
1528 gdp = ext4_get_group_desc(sb, i, &bh);
1529
1530 flex_group = ext4_flex_group(sbi, i);
1531 sbi->s_flex_groups[flex_group].free_inodes +=
1532 le16_to_cpu(gdp->bg_free_inodes_count);
1533 sbi->s_flex_groups[flex_group].free_blocks +=
1534 le16_to_cpu(gdp->bg_free_blocks_count);
1535 }
1536
1537 return 1;
1538failed:
1539 return 0;
1540}
1541
717d50e4
AD		 1542__le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1543 struct ext4_group_desc *gdp)
1544{
1545 __u16 crc = 0;
1546
1547 if (sbi->s_es->s_feature_ro_compat &
1548 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1549 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1550 __le32 le_group = cpu_to_le32(block_group);
1551
1552 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1553 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1554 crc = crc16(crc, (__u8 *)gdp, offset);
1555 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1556 /* for checksum of struct ext4_group_desc do the rest...*/
1557 if ((sbi->s_es->s_feature_incompat &
1558 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1559 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1560 crc = crc16(crc, (__u8 *)gdp + offset,
1561 le16_to_cpu(sbi->s_es->s_desc_size) -
1562 offset);
1563 }
1564
1565 return cpu_to_le16(crc);
1566}
1567
1568int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1569 struct ext4_group_desc *gdp)
1570{
1571 if ((sbi->s_es->s_feature_ro_compat &
1572 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1573 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1574 return 0;
1575
1576 return 1;
1577}
1578
ac27a0ec 1579/* Called at mount-time, super-block is locked */
197cd65a 1580static int ext4_check_descriptors(struct super_block *sb)
ac27a0ec 1581{
617ba13b
MC		 1582 struct ext4_sb_info *sbi = EXT4_SB(sb);
1583 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1584 ext4_fsblk_t last_block;
bd81d8ee
LV		 1585 ext4_fsblk_t block_bitmap;
1586 ext4_fsblk_t inode_bitmap;
1587 ext4_fsblk_t inode_table;
ce421581 1588 int flexbg_flag = 0;
fd2d4291 1589 ext4_group_t i;
ac27a0ec 1590
ce421581
JS		 1591 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1592 flexbg_flag = 1;
1593
af5bc92d 1594 ext4_debug("Checking group descriptors");
ac27a0ec 1595
197cd65a
AM		 1596 for (i = 0; i < sbi->s_groups_count; i++) {
1597 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1598
ce421581 1599 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 1600 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 1601 else
1602 last_block = first_block +
617ba13b 1603 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 1604
8fadc143 1605 block_bitmap = ext4_block_bitmap(sb, gdp);
2b2d6d01 1606 if (block_bitmap < first_block || block_bitmap > last_block) {
c19204b0
JB		 1607 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1608 "Block bitmap for group %lu not in group "
1609 "(block %llu)!", i, block_bitmap);
ac27a0ec
DK		 1610 return 0;
1611 }
8fadc143 1612 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2b2d6d01 1613 if (inode_bitmap < first_block || inode_bitmap > last_block) {
c19204b0
JB		 1614 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1615 "Inode bitmap for group %lu not in group "
1616 "(block %llu)!", i, inode_bitmap);
ac27a0ec
DK		 1617 return 0;
1618 }
8fadc143 1619 inode_table = ext4_inode_table(sb, gdp);
bd81d8ee 1620 if (inode_table < first_block ||
2b2d6d01 1621 inode_table + sbi->s_itb_per_group - 1 > last_block) {
c19204b0
JB		 1622 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1623 "Inode table for group %lu not in group "
1624 "(block %llu)!", i, inode_table);
ac27a0ec
DK		 1625 return 0;
1626 }
b5f10eed 1627 spin_lock(sb_bgl_lock(sbi, i));
717d50e4 1628 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
c19204b0
JB		 1629 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1630 "Checksum for group %lu failed (%u!=%u)\n",
1631 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1632 gdp)), le16_to_cpu(gdp->bg_checksum));
7ee1ec4c
LZ		 1633 if (!(sb->s_flags & MS_RDONLY)) {
1634 spin_unlock(sb_bgl_lock(sbi, i));
8a266467 1635 return 0;
7ee1ec4c 1636 }
717d50e4 1637 }
b5f10eed 1638 spin_unlock(sb_bgl_lock(sbi, i));
ce421581
JS		 1639 if (!flexbg_flag)
1640 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec
DK		 1641 }
1642
bd81d8ee 1643 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
2b2d6d01 1644 sbi->s_es->s_free_inodes_count = cpu_to_le32(ext4_count_free_inodes(sb));
ac27a0ec
DK		 1645 return 1;
1646}
1647
617ba13b 1648/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 1649 * the superblock) which were deleted from all directories, but held open by
1650 * a process at the time of a crash. We walk the list and try to delete these
1651 * inodes at recovery time (only with a read-write filesystem).
1652 *
1653 * In order to keep the orphan inode chain consistent during traversal (in
1654 * case of crash during recovery), we link each inode into the superblock
1655 * orphan list_head and handle it the same way as an inode deletion during
1656 * normal operation (which journals the operations for us).
1657 *
1658 * We only do an iget() and an iput() on each inode, which is very safe if we
1659 * accidentally point at an in-use or already deleted inode. The worst that
1660 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 1661 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 1662 * e2fsck was run on this filesystem, and it must have already done the orphan
1663 * inode cleanup for us, so we can safely abort without any further action.
1664 */
2b2d6d01
TT		 1665static void ext4_orphan_cleanup(struct super_block *sb,
1666 struct ext4_super_block *es)
ac27a0ec
DK		 1667{
1668 unsigned int s_flags = sb->s_flags;
1669 int nr_orphans = 0, nr_truncates = 0;
1670#ifdef CONFIG_QUOTA
1671 int i;
1672#endif
1673 if (!es->s_last_orphan) {
1674 jbd_debug(4, "no orphan inodes to clean up\n");
1675 return;
1676 }
1677
a8f48a95
ES		 1678 if (bdev_read_only(sb->s_bdev)) {
1679 printk(KERN_ERR "EXT4-fs: write access "
1680 "unavailable, skipping orphan cleanup.\n");
1681 return;
1682 }
1683
617ba13b 1684 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
ac27a0ec
DK		 1685 if (es->s_last_orphan)
1686 jbd_debug(1, "Errors on filesystem, "
1687 "clearing orphan list.\n");
1688 es->s_last_orphan = 0;
1689 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1690 return;
1691 }
1692
1693 if (s_flags & MS_RDONLY) {
617ba13b 1694 printk(KERN_INFO "EXT4-fs: %s: orphan cleanup on readonly fs\n",
ac27a0ec
DK		 1695 sb->s_id);
1696 sb->s_flags &= ~MS_RDONLY;
1697 }
1698#ifdef CONFIG_QUOTA
1699 /* Needed for iput() to work correctly and not trash data */
1700 sb->s_flags |= MS_ACTIVE;
1701 /* Turn on quotas so that they are updated correctly */
1702 for (i = 0; i < MAXQUOTAS; i++) {
617ba13b
MC		 1703 if (EXT4_SB(sb)->s_qf_names[i]) {
1704 int ret = ext4_quota_on_mount(sb, i);
ac27a0ec
DK		 1705 if (ret < 0)
1706 printk(KERN_ERR
2c8be6b2 1707 "EXT4-fs: Cannot turn on journaled "
ac27a0ec
DK		 1708 "quota: error %d\n", ret);
1709 }
1710 }
1711#endif
1712
1713 while (es->s_last_orphan) {
1714 struct inode *inode;
1715
97bd42b9
JB		 1716 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1717 if (IS_ERR(inode)) {
ac27a0ec
DK		 1718 es->s_last_orphan = 0;
1719 break;
1720 }
1721
617ba13b 1722 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
ac27a0ec
DK		 1723 DQUOT_INIT(inode);
1724 if (inode->i_nlink) {
1725 printk(KERN_DEBUG
e5f8eab8 1726 "%s: truncating inode %lu to %lld bytes\n",
46e665e9 1727 __func__, inode->i_ino, inode->i_size);
e5f8eab8 1728 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
ac27a0ec 1729 inode->i_ino, inode->i_size);
617ba13b 1730 ext4_truncate(inode);
ac27a0ec
DK		 1731 nr_truncates++;
1732 } else {
1733 printk(KERN_DEBUG
1734 "%s: deleting unreferenced inode %lu\n",
46e665e9 1735 __func__, inode->i_ino);
ac27a0ec
DK		 1736 jbd_debug(2, "deleting unreferenced inode %lu\n",
1737 inode->i_ino);
1738 nr_orphans++;
1739 }
1740 iput(inode); /* The delete magic happens here! */
1741 }
1742
2b2d6d01 1743#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
ac27a0ec
DK		 1744
1745 if (nr_orphans)
617ba13b 1746 printk(KERN_INFO "EXT4-fs: %s: %d orphan inode%s deleted\n",
ac27a0ec
DK		 1747 sb->s_id, PLURAL(nr_orphans));
1748 if (nr_truncates)
617ba13b 1749 printk(KERN_INFO "EXT4-fs: %s: %d truncate%s cleaned up\n",
ac27a0ec
DK		 1750 sb->s_id, PLURAL(nr_truncates));
1751#ifdef CONFIG_QUOTA
1752 /* Turn quotas off */
1753 for (i = 0; i < MAXQUOTAS; i++) {
1754 if (sb_dqopt(sb)->files[i])
6f28e087 1755 vfs_quota_off(sb, i, 0);
ac27a0ec
DK		 1756 }
1757#endif
1758 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1759}
cd2291a4
ES		 1760/*
1761 * Maximal extent format file size.
1762 * Resulting logical blkno at s_maxbytes must fit in our on-disk
1763 * extent format containers, within a sector_t, and within i_blocks
1764 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
1765 * so that won't be a limiting factor.
1766 *
1767 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1768 */
1769static loff_t ext4_max_size(int blkbits)
1770{
1771 loff_t res;
1772 loff_t upper_limit = MAX_LFS_FILESIZE;
1773
1774 /* small i_blocks in vfs inode? */
1775 if (sizeof(blkcnt_t) < sizeof(u64)) {
1776 /*
1777 * CONFIG_LSF is not enabled implies the inode
1778 * i_block represent total blocks in 512 bytes
1779 * 32 == size of vfs inode i_blocks * 8
1780 */
1781 upper_limit = (1LL << 32) - 1;
1782
1783 /* total blocks in file system block size */
1784 upper_limit >>= (blkbits - 9);
1785 upper_limit <<= blkbits;
1786 }
1787
1788 /* 32-bit extent-start container, ee_block */
1789 res = 1LL << 32;
1790 res <<= blkbits;
1791 res -= 1;
1792
1793 /* Sanity check against vm- & vfs- imposed limits */
1794 if (res > upper_limit)
1795 res = upper_limit;
1796
1797 return res;
1798}
ac27a0ec 1799
ac27a0ec 1800/*
cd2291a4 1801 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 1802 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
1803 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 1804 */
cd2291a4 1805static loff_t ext4_max_bitmap_size(int bits)
ac27a0ec 1806{
617ba13b 1807 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 1808 int meta_blocks;
1809 loff_t upper_limit;
1810 /* This is calculated to be the largest file size for a
cd2291a4 1811 * dense, bitmapped file such that the total number of
ac27a0ec 1812 * sectors in the file, including data and all indirect blocks,
0fc1b451
AK		 1813 * does not exceed 2^48 -1
1814 * __u32 i_blocks_lo and _u16 i_blocks_high representing the
1815 * total number of 512 bytes blocks of the file
1816 */
1817
1818 if (sizeof(blkcnt_t) < sizeof(u64)) {
1819 /*
1820 * CONFIG_LSF is not enabled implies the inode
1821 * i_block represent total blocks in 512 bytes
1822 * 32 == size of vfs inode i_blocks * 8
1823 */
1824 upper_limit = (1LL << 32) - 1;
1825
1826 /* total blocks in file system block size */
1827 upper_limit >>= (bits - 9);
1828
1829 } else {
8180a562
AK		 1830 /*
1831 * We use 48 bit ext4_inode i_blocks
1832 * With EXT4_HUGE_FILE_FL set the i_blocks
1833 * represent total number of blocks in
1834 * file system block size
1835 */
0fc1b451
AK		 1836 upper_limit = (1LL << 48) - 1;
1837
0fc1b451
AK		 1838 }
1839
1840 /* indirect blocks */
1841 meta_blocks = 1;
1842 /* double indirect blocks */
1843 meta_blocks += 1 + (1LL << (bits-2));
1844 /* tripple indirect blocks */
1845 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
1846
1847 upper_limit -= meta_blocks;
1848 upper_limit <<= bits;
ac27a0ec
DK		 1849
1850 res += 1LL << (bits-2);
1851 res += 1LL << (2*(bits-2));
1852 res += 1LL << (3*(bits-2));
1853 res <<= bits;
1854 if (res > upper_limit)
1855 res = upper_limit;
0fc1b451
AK		 1856
1857 if (res > MAX_LFS_FILESIZE)
1858 res = MAX_LFS_FILESIZE;
1859
ac27a0ec
DK		 1860 return res;
1861}
1862
617ba13b 1863static ext4_fsblk_t descriptor_loc(struct super_block *sb,
70bbb3e0 1864 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 1865{
617ba13b 1866 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 1867 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 1868 int has_super = 0;
1869
1870 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1871
617ba13b 1872 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
ac27a0ec 1873 nr < first_meta_bg)
70bbb3e0 1874 return logical_sb_block + nr + 1;
ac27a0ec 1875 bg = sbi->s_desc_per_block * nr;
617ba13b 1876 if (ext4_bg_has_super(sb, bg))
ac27a0ec 1877 has_super = 1;
617ba13b 1878 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 1879}
1880
c9de560d
AT		 1881/**
1882 * ext4_get_stripe_size: Get the stripe size.
1883 * @sbi: In memory super block info
1884 *
1885 * If we have specified it via mount option, then
1886 * use the mount option value. If the value specified at mount time is
1887 * greater than the blocks per group use the super block value.
1888 * If the super block value is greater than blocks per group return 0.
1889 * Allocator needs it be less than blocks per group.
1890 *
1891 */
1892static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
1893{
1894 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
1895 unsigned long stripe_width =
1896 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
1897
1898 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
1899 return sbi->s_stripe;
1900
1901 if (stripe_width <= sbi->s_blocks_per_group)
1902 return stripe_width;
1903
1904 if (stride <= sbi->s_blocks_per_group)
1905 return stride;
1906
1907 return 0;
1908}
ac27a0ec 1909
2b2d6d01 1910static int ext4_fill_super(struct super_block *sb, void *data, int silent)
7477827f
AK		 1911 __releases(kernel_lock)
1912 __acquires(kernel_lock)
1d03ec98 1913
ac27a0ec 1914{
2b2d6d01 1915 struct buffer_head *bh;
617ba13b
MC		 1916 struct ext4_super_block *es = NULL;
1917 struct ext4_sb_info *sbi;
1918 ext4_fsblk_t block;
1919 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 1920 ext4_fsblk_t logical_sb_block;
ac27a0ec
DK		 1921 unsigned long offset = 0;
1922 unsigned int journal_inum = 0;
1923 unsigned long journal_devnum = 0;
1924 unsigned long def_mount_opts;
1925 struct inode *root;
9f6200bb 1926 char *cp;
1d1fe1ee 1927 int ret = -EINVAL;
ac27a0ec 1928 int blocksize;
ac27a0ec
DK		 1929 int db_count;
1930 int i;
1931 int needs_recovery;
1932 __le32 features;
bd81d8ee 1933 __u64 blocks_count;
833f4077 1934 int err;
ac27a0ec
DK		 1935
1936 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
1937 if (!sbi)
1938 return -ENOMEM;
1939 sb->s_fs_info = sbi;
1940 sbi->s_mount_opt = 0;
617ba13b
MC		 1941 sbi->s_resuid = EXT4_DEF_RESUID;
1942 sbi->s_resgid = EXT4_DEF_RESGID;
240799cd 1943 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
d9c9bef1 1944 sbi->s_sb_block = sb_block;
ac27a0ec
DK		 1945
1946 unlock_kernel();
1947
9f6200bb
TT		 1948 /* Cleanup superblock name */
1949 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
1950 *cp = '!';
1951
617ba13b 1952 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 1953 if (!blocksize) {
617ba13b 1954 printk(KERN_ERR "EXT4-fs: unable to set blocksize\n");
ac27a0ec
DK		 1955 goto out_fail;
1956 }
1957
1958 /*
617ba13b 1959 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 1960 * block sizes. We need to calculate the offset from buffer start.
1961 */
617ba13b 1962 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 1963 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
1964 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 1965 } else {
70bbb3e0 1966 logical_sb_block = sb_block;
ac27a0ec
DK		 1967 }
1968
70bbb3e0 1969 if (!(bh = sb_bread(sb, logical_sb_block))) {
2b2d6d01 1970 printk(KERN_ERR "EXT4-fs: unable to read superblock\n");
ac27a0ec
DK		 1971 goto out_fail;
1972 }
1973 /*
1974 * Note: s_es must be initialized as soon as possible because
617ba13b 1975 * some ext4 macro-instructions depend on its value
ac27a0ec 1976 */
617ba13b 1977 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
ac27a0ec
DK		 1978 sbi->s_es = es;
1979 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 1980 if (sb->s_magic != EXT4_SUPER_MAGIC)
1981 goto cantfind_ext4;
ac27a0ec
DK		 1982
1983 /* Set defaults before we parse the mount options */
1984 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
617ba13b 1985 if (def_mount_opts & EXT4_DEFM_DEBUG)
ac27a0ec 1986 set_opt(sbi->s_mount_opt, DEBUG);
617ba13b 1987 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
ac27a0ec 1988 set_opt(sbi->s_mount_opt, GRPID);
617ba13b 1989 if (def_mount_opts & EXT4_DEFM_UID16)
ac27a0ec 1990 set_opt(sbi->s_mount_opt, NO_UID32);
03010a33 1991#ifdef CONFIG_EXT4_FS_XATTR
617ba13b 1992 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
ac27a0ec 1993 set_opt(sbi->s_mount_opt, XATTR_USER);
2e7842b8 1994#endif
03010a33 1995#ifdef CONFIG_EXT4_FS_POSIX_ACL
617ba13b 1996 if (def_mount_opts & EXT4_DEFM_ACL)
ac27a0ec 1997 set_opt(sbi->s_mount_opt, POSIX_ACL);
2e7842b8 1998#endif
617ba13b
MC		 1999 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
2000 sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
2001 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
2002 sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
2003 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
2004 sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
2005
2006 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
ac27a0ec 2007 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
bb4f397a 2008 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
ceea16bf 2009 set_opt(sbi->s_mount_opt, ERRORS_CONT);
bb4f397a
AK		 2010 else
2011 set_opt(sbi->s_mount_opt, ERRORS_RO);
ac27a0ec
DK		 2012
2013 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
2014 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
2015
2016 set_opt(sbi->s_mount_opt, RESERVATION);
571640ca 2017 set_opt(sbi->s_mount_opt, BARRIER);
ac27a0ec 2018
1e2462f9
MC		 2019 /*
2020 * turn on extents feature by default in ext4 filesystem
e4079a11
ES		 2021 * only if feature flag already set by mkfs or tune2fs.
2022 * Use -o noextents to turn it off
1e2462f9 2023 */
e4079a11
ES		 2024 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS))
2025 set_opt(sbi->s_mount_opt, EXTENTS);
2026 else
2027 ext4_warning(sb, __func__,
2028 "extents feature not enabled on this filesystem, "
2029 "use tune2fs.\n");
1e2462f9 2030
dd919b98
AK		 2031 /*
2032 * enable delayed allocation by default
2033 * Use -o nodelalloc to turn it off
2034 */
2035 set_opt(sbi->s_mount_opt, DELALLOC);
2036
2037
2b2d6d01
TT		 2038 if (!parse_options((char *) data, sb, &journal_inum, &journal_devnum,
2039 NULL, 0))
ac27a0ec
DK		 2040 goto failed_mount;
2041
2042 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
617ba13b 2043 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec 2044
617ba13b
MC		 2045 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
2046 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
2047 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
2048 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
ac27a0ec 2049 printk(KERN_WARNING
617ba13b 2050 "EXT4-fs warning: feature flags set on rev 0 fs, "
ac27a0ec 2051 "running e2fsck is recommended\n");
469108ff 2052
ac27a0ec
DK		 2053 /*
2054 * Check feature flags regardless of the revision level, since we
2055 * previously didn't change the revision level when setting the flags,
2056 * so there is a chance incompat flags are set on a rev 0 filesystem.
2057 */
617ba13b 2058 features = EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP);
ac27a0ec 2059 if (features) {
617ba13b 2060 printk(KERN_ERR "EXT4-fs: %s: couldn't mount because of "
ac27a0ec
DK		 2061 "unsupported optional features (%x).\n",
2062 sb->s_id, le32_to_cpu(features));
2063 goto failed_mount;
2064 }
617ba13b 2065 features = EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP);
ac27a0ec 2066 if (!(sb->s_flags & MS_RDONLY) && features) {
617ba13b 2067 printk(KERN_ERR "EXT4-fs: %s: couldn't mount RDWR because of "
ac27a0ec
DK		 2068 "unsupported optional features (%x).\n",
2069 sb->s_id, le32_to_cpu(features));
2070 goto failed_mount;
2071 }
0fc1b451
AK		 2072 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2073 /*
2074 * Large file size enabled file system can only be
2075 * mount if kernel is build with CONFIG_LSF
2076 */
2077 if (sizeof(root->i_blocks) < sizeof(u64) &&
2078 !(sb->s_flags & MS_RDONLY)) {
2079 printk(KERN_ERR "EXT4-fs: %s: Filesystem with huge "
2080 "files cannot be mounted read-write "
2081 "without CONFIG_LSF.\n", sb->s_id);
2082 goto failed_mount;
2083 }
2084 }
ac27a0ec
DK		 2085 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
2086
617ba13b
MC		 2087 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
2088 blocksize > EXT4_MAX_BLOCK_SIZE) {
ac27a0ec 2089 printk(KERN_ERR
617ba13b 2090 "EXT4-fs: Unsupported filesystem blocksize %d on %s.\n",
ac27a0ec
DK		 2091 blocksize, sb->s_id);
2092 goto failed_mount;
2093 }
2094
ac27a0ec 2095 if (sb->s_blocksize != blocksize) {
ce40733c
AK		 2096
2097 /* Validate the filesystem blocksize */
2098 if (!sb_set_blocksize(sb, blocksize)) {
2099 printk(KERN_ERR "EXT4-fs: bad block size %d.\n",
2100 blocksize);
ac27a0ec
DK		 2101 goto failed_mount;
2102 }
2103
2b2d6d01 2104 brelse(bh);
70bbb3e0
AM		 2105 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2106 offset = do_div(logical_sb_block, blocksize);
2107 bh = sb_bread(sb, logical_sb_block);
ac27a0ec
DK		 2108 if (!bh) {
2109 printk(KERN_ERR
617ba13b 2110 "EXT4-fs: Can't read superblock on 2nd try.\n");
ac27a0ec
DK		 2111 goto failed_mount;
2112 }
617ba13b 2113 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
ac27a0ec 2114 sbi->s_es = es;
617ba13b 2115 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
2b2d6d01
TT		 2116 printk(KERN_ERR
2117 "EXT4-fs: Magic mismatch, very weird !\n");
ac27a0ec
DK		 2118 goto failed_mount;
2119 }
2120 }
2121
e2b46574 2122 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits);
617ba13b 2123 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits);
ac27a0ec 2124
617ba13b
MC		 2125 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2126 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2127 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
ac27a0ec
DK		 2128 } else {
2129 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2130 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
617ba13b 2131 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1330593e 2132 (!is_power_of_2(sbi->s_inode_size)) ||
ac27a0ec 2133 (sbi->s_inode_size > blocksize)) {
2b2d6d01
TT		 2134 printk(KERN_ERR
2135 "EXT4-fs: unsupported inode size: %d\n",
2136 sbi->s_inode_size);
ac27a0ec
DK		 2137 goto failed_mount;
2138 }
ef7f3835
KS		 2139 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2140 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
ac27a0ec 2141 }
0d1ee42f
AR		 2142 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2143 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
8fadc143 2144 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 2145 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 2146 !is_power_of_2(sbi->s_desc_size)) {
0d1ee42f 2147 printk(KERN_ERR
8fadc143 2148 "EXT4-fs: unsupported descriptor size %lu\n",
0d1ee42f
AR		 2149 sbi->s_desc_size);
2150 goto failed_mount;
2151 }
2152 } else
2153 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
ac27a0ec 2154 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 2155 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
b47b6f38 2156 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
617ba13b
MC		 2157 goto cantfind_ext4;
2158 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 2159 if (sbi->s_inodes_per_block == 0)
617ba13b 2160 goto cantfind_ext4;
ac27a0ec
DK		 2161 sbi->s_itb_per_group = sbi->s_inodes_per_group /
2162 sbi->s_inodes_per_block;
0d1ee42f 2163 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 2164 sbi->s_sbh = bh;
2165 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 2166 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2167 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
2b2d6d01 2168 for (i = 0; i < 4; i++)
ac27a0ec
DK		 2169 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2170 sbi->s_def_hash_version = es->s_def_hash_version;
2171
2172 if (sbi->s_blocks_per_group > blocksize * 8) {
2b2d6d01
TT		 2173 printk(KERN_ERR
2174 "EXT4-fs: #blocks per group too big: %lu\n",
2175 sbi->s_blocks_per_group);
ac27a0ec
DK		 2176 goto failed_mount;
2177 }
ac27a0ec 2178 if (sbi->s_inodes_per_group > blocksize * 8) {
2b2d6d01
TT		 2179 printk(KERN_ERR
2180 "EXT4-fs: #inodes per group too big: %lu\n",
2181 sbi->s_inodes_per_group);
ac27a0ec
DK		 2182 goto failed_mount;
2183 }
2184
bd81d8ee 2185 if (ext4_blocks_count(es) >
ac27a0ec 2186 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
617ba13b 2187 printk(KERN_ERR "EXT4-fs: filesystem on %s:"
ac27a0ec
DK		 2188 " too large to mount safely\n", sb->s_id);
2189 if (sizeof(sector_t) < 8)
617ba13b 2190 printk(KERN_WARNING "EXT4-fs: CONFIG_LBD not "
ac27a0ec
DK		 2191 "enabled\n");
2192 goto failed_mount;
2193 }
2194
617ba13b
MC		 2195 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2196 goto cantfind_ext4;
e7c95593
ES		 2197
2198 /* ensure blocks_count calculation below doesn't sign-extend */
2199 if (ext4_blocks_count(es) + EXT4_BLOCKS_PER_GROUP(sb) <
2200 le32_to_cpu(es->s_first_data_block) + 1) {
2201 printk(KERN_WARNING "EXT4-fs: bad geometry: block count %llu, "
2202 "first data block %u, blocks per group %lu\n",
2203 ext4_blocks_count(es),
2204 le32_to_cpu(es->s_first_data_block),
2205 EXT4_BLOCKS_PER_GROUP(sb));
2206 goto failed_mount;
2207 }
bd81d8ee
LV		 2208 blocks_count = (ext4_blocks_count(es) -
2209 le32_to_cpu(es->s_first_data_block) +
2210 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2211 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
2212 sbi->s_groups_count = blocks_count;
617ba13b
MC		 2213 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2214 EXT4_DESC_PER_BLOCK(sb);
2b2d6d01 2215 sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
ac27a0ec
DK		 2216 GFP_KERNEL);
2217 if (sbi->s_group_desc == NULL) {
2b2d6d01 2218 printk(KERN_ERR "EXT4-fs: not enough memory\n");
ac27a0ec
DK		 2219 goto failed_mount;
2220 }
2221
9f6200bb
TT		 2222 if (ext4_proc_root)
2223 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
2224
240799cd
TT		 2225 if (sbi->s_proc)
2226 proc_create_data("inode_readahead_blks", 0644, sbi->s_proc,
2227 &ext4_ui_proc_fops,
2228 &sbi->s_inode_readahead_blks);
2229
ac27a0ec
DK		 2230 bgl_lock_init(&sbi->s_blockgroup_lock);
2231
2232 for (i = 0; i < db_count; i++) {
70bbb3e0 2233 block = descriptor_loc(sb, logical_sb_block, i);
ac27a0ec
DK		 2234 sbi->s_group_desc[i] = sb_bread(sb, block);
2235 if (!sbi->s_group_desc[i]) {
2b2d6d01
TT		 2236 printk(KERN_ERR "EXT4-fs: "
2237 "can't read group descriptor %d\n", i);
ac27a0ec
DK		 2238 db_count = i;
2239 goto failed_mount2;
2240 }
2241 }
2b2d6d01 2242 if (!ext4_check_descriptors(sb)) {
617ba13b 2243 printk(KERN_ERR "EXT4-fs: group descriptors corrupted!\n");
ac27a0ec
DK		 2244 goto failed_mount2;
2245 }
772cb7c8
JS		 2246 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2247 if (!ext4_fill_flex_info(sb)) {
2248 printk(KERN_ERR
2249 "EXT4-fs: unable to initialize "
2250 "flex_bg meta info!\n");
2251 goto failed_mount2;
2252 }
2253
ac27a0ec
DK		 2254 sbi->s_gdb_count = db_count;
2255 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2256 spin_lock_init(&sbi->s_next_gen_lock);
2257
833f4077
PZ		 2258 err = percpu_counter_init(&sbi->s_freeblocks_counter,
2259 ext4_count_free_blocks(sb));
2260 if (!err) {
2261 err = percpu_counter_init(&sbi->s_freeinodes_counter,
2262 ext4_count_free_inodes(sb));
2263 }
2264 if (!err) {
2265 err = percpu_counter_init(&sbi->s_dirs_counter,
2266 ext4_count_dirs(sb));
2267 }
6bc6e63f
AK		 2268 if (!err) {
2269 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
2270 }
833f4077
PZ		 2271 if (err) {
2272 printk(KERN_ERR "EXT4-fs: insufficient memory\n");
2273 goto failed_mount3;
2274 }
ac27a0ec 2275
c9de560d
AT		 2276 sbi->s_stripe = ext4_get_stripe_size(sbi);
2277
ac27a0ec
DK		 2278 /*
2279 * set up enough so that it can read an inode
2280 */
617ba13b
MC		 2281 sb->s_op = &ext4_sops;
2282 sb->s_export_op = &ext4_export_ops;
2283 sb->s_xattr = ext4_xattr_handlers;
ac27a0ec 2284#ifdef CONFIG_QUOTA
617ba13b
MC		 2285 sb->s_qcop = &ext4_qctl_operations;
2286 sb->dq_op = &ext4_quota_operations;
ac27a0ec
DK		 2287#endif
2288 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
2289
2290 sb->s_root = NULL;
2291
2292 needs_recovery = (es->s_last_orphan != 0 ||
617ba13b
MC		 2293 EXT4_HAS_INCOMPAT_FEATURE(sb,
2294 EXT4_FEATURE_INCOMPAT_RECOVER));
ac27a0ec
DK		 2295
2296 /*
2297 * The first inode we look at is the journal inode. Don't try
2298 * root first: it may be modified in the journal!
2299 */
2300 if (!test_opt(sb, NOLOAD) &&
617ba13b
MC		 2301 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2302 if (ext4_load_journal(sb, es, journal_devnum))
ac27a0ec 2303 goto failed_mount3;
624080ed
TT		 2304 if (!(sb->s_flags & MS_RDONLY) &&
2305 EXT4_SB(sb)->s_journal->j_failed_commit) {
2306 printk(KERN_CRIT "EXT4-fs error (device %s): "
2307 "ext4_fill_super: Journal transaction "
2b2d6d01 2308 "%u is corrupt\n", sb->s_id,
624080ed 2309 EXT4_SB(sb)->s_journal->j_failed_commit);
2b2d6d01
TT		 2310 if (test_opt(sb, ERRORS_RO)) {
2311 printk(KERN_CRIT
2312 "Mounting filesystem read-only\n");
624080ed
TT		 2313 sb->s_flags |= MS_RDONLY;
2314 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2315 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2316 }
2317 if (test_opt(sb, ERRORS_PANIC)) {
2318 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2319 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2320 ext4_commit_super(sb, es, 1);
2321 printk(KERN_CRIT
2322 "EXT4-fs (device %s): mount failed\n",
2323 sb->s_id);
2324 goto failed_mount4;
2325 }
2326 }
ac27a0ec 2327 } else if (journal_inum) {
617ba13b 2328 if (ext4_create_journal(sb, es, journal_inum))
ac27a0ec
DK		 2329 goto failed_mount3;
2330 } else {
2331 if (!silent)
2b2d6d01
TT		 2332 printk(KERN_ERR
2333 "ext4: No journal on filesystem on %s\n",
2334 sb->s_id);
ac27a0ec
DK		 2335 goto failed_mount3;
2336 }
2337
eb40a09c
JS		 2338 if (ext4_blocks_count(es) > 0xffffffffULL &&
2339 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2340 JBD2_FEATURE_INCOMPAT_64BIT)) {
2341 printk(KERN_ERR "ext4: Failed to set 64-bit journal feature\n");
2342 goto failed_mount4;
2343 }
2344
818d276c
GS		 2345 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2346 jbd2_journal_set_features(sbi->s_journal,
2347 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2348 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2349 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2350 jbd2_journal_set_features(sbi->s_journal,
2351 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2352 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2353 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2354 } else {
2355 jbd2_journal_clear_features(sbi->s_journal,
2356 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2357 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2358 }
2359
ac27a0ec
DK		 2360 /* We have now updated the journal if required, so we can
2361 * validate the data journaling mode. */
2362 switch (test_opt(sb, DATA_FLAGS)) {
2363 case 0:
2364 /* No mode set, assume a default based on the journal
63f57933
AM		 2365 * capabilities: ORDERED_DATA if the journal can
2366 * cope, else JOURNAL_DATA
2367 */
dab291af
MC		 2368 if (jbd2_journal_check_available_features
2369 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
ac27a0ec
DK		 2370 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2371 else
2372 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2373 break;
2374
617ba13b
MC		 2375 case EXT4_MOUNT_ORDERED_DATA:
2376 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 2377 if (!jbd2_journal_check_available_features
2378 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
617ba13b 2379 printk(KERN_ERR "EXT4-fs: Journal does not support "
ac27a0ec
DK		 2380 "requested data journaling mode\n");
2381 goto failed_mount4;
2382 }
2383 default:
2384 break;
2385 }
2386
2387 if (test_opt(sb, NOBH)) {
617ba13b
MC		 2388 if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
2389 printk(KERN_WARNING "EXT4-fs: Ignoring nobh option - "
ac27a0ec
DK		 2390 "its supported only with writeback mode\n");
2391 clear_opt(sbi->s_mount_opt, NOBH);
2392 }
2393 }
2394 /*
dab291af 2395 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 2396 * so we can safely mount the rest of the filesystem now.
2397 */
2398
1d1fe1ee
DH		 2399 root = ext4_iget(sb, EXT4_ROOT_INO);
2400 if (IS_ERR(root)) {
617ba13b 2401 printk(KERN_ERR "EXT4-fs: get root inode failed\n");
1d1fe1ee 2402 ret = PTR_ERR(root);
ac27a0ec
DK		 2403 goto failed_mount4;
2404 }
2405 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1d1fe1ee 2406 iput(root);
617ba13b 2407 printk(KERN_ERR "EXT4-fs: corrupt root inode, run e2fsck\n");
ac27a0ec
DK		 2408 goto failed_mount4;
2409 }
1d1fe1ee
DH		 2410 sb->s_root = d_alloc_root(root);
2411 if (!sb->s_root) {
2412 printk(KERN_ERR "EXT4-fs: get root dentry failed\n");
2413 iput(root);
2414 ret = -ENOMEM;
2415 goto failed_mount4;
2416 }
ac27a0ec 2417
2b2d6d01 2418 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
ef7f3835
KS		 2419
2420 /* determine the minimum size of new large inodes, if present */
2421 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2422 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2423 EXT4_GOOD_OLD_INODE_SIZE;
2424 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2425 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2426 if (sbi->s_want_extra_isize <
2427 le16_to_cpu(es->s_want_extra_isize))
2428 sbi->s_want_extra_isize =
2429 le16_to_cpu(es->s_want_extra_isize);
2430 if (sbi->s_want_extra_isize <
2431 le16_to_cpu(es->s_min_extra_isize))
2432 sbi->s_want_extra_isize =
2433 le16_to_cpu(es->s_min_extra_isize);
2434 }
2435 }
2436 /* Check if enough inode space is available */
2437 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2438 sbi->s_inode_size) {
2439 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2440 EXT4_GOOD_OLD_INODE_SIZE;
2441 printk(KERN_INFO "EXT4-fs: required extra inode space not"
2442 "available.\n");
2443 }
2444
ac27a0ec
DK		 2445 /*
2446 * akpm: core read_super() calls in here with the superblock locked.
2447 * That deadlocks, because orphan cleanup needs to lock the superblock
2448 * in numerous places. Here we just pop the lock - it's relatively
2449 * harmless, because we are now ready to accept write_super() requests,
2450 * and aviro says that's the only reason for hanging onto the
2451 * superblock lock.
2452 */
617ba13b
MC		 2453 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
2454 ext4_orphan_cleanup(sb, es);
2455 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
ac27a0ec 2456 if (needs_recovery)
2b2d6d01 2457 printk(KERN_INFO "EXT4-fs: recovery complete.\n");
617ba13b 2458 ext4_mark_recovery_complete(sb, es);
2b2d6d01
TT		 2459 printk(KERN_INFO "EXT4-fs: mounted filesystem with %s data mode.\n",
2460 test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA ? "journal":
2461 test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA ? "ordered":
2462 "writeback");
ac27a0ec 2463
dd919b98
AK		 2464 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
2465 printk(KERN_WARNING "EXT4-fs: Ignoring delalloc option - "
2466 "requested data journaling mode\n");
2467 clear_opt(sbi->s_mount_opt, DELALLOC);
2468 } else if (test_opt(sb, DELALLOC))
2469 printk(KERN_INFO "EXT4-fs: delayed allocation enabled\n");
2470
a86c6181 2471 ext4_ext_init(sb);
c2ea3fde
TT		 2472 err = ext4_mb_init(sb, needs_recovery);
2473 if (err) {
2474 printk(KERN_ERR "EXT4-fs: failed to initalize mballoc (%d)\n",
2475 err);
2476 goto failed_mount4;
2477 }
a86c6181 2478
ac27a0ec
DK		 2479 lock_kernel();
2480 return 0;
2481
617ba13b 2482cantfind_ext4:
ac27a0ec 2483 if (!silent)
617ba13b 2484 printk(KERN_ERR "VFS: Can't find ext4 filesystem on dev %s.\n",
ac27a0ec
DK		 2485 sb->s_id);
2486 goto failed_mount;
2487
2488failed_mount4:
dab291af 2489 jbd2_journal_destroy(sbi->s_journal);
47b4a50b 2490 sbi->s_journal = NULL;
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;
ac27a0ec
DK		 2538 spin_unlock(&journal->j_state_lock);
2539}
2540
617ba13b 2541static journal_t *ext4_get_journal(struct super_block *sb,
ac27a0ec
DK		 2542 unsigned int journal_inum)
2543{
2544 struct inode *journal_inode;
2545 journal_t *journal;
2546
2547 /* First, test for the existence of a valid inode on disk. Bad
2548 * things happen if we iget() an unused inode, as the subsequent
2549 * iput() will try to delete it. */
2550
1d1fe1ee
DH		 2551 journal_inode = ext4_iget(sb, journal_inum);
2552 if (IS_ERR(journal_inode)) {
617ba13b 2553 printk(KERN_ERR "EXT4-fs: no journal found.\n");
ac27a0ec
DK		 2554 return NULL;
2555 }
2556 if (!journal_inode->i_nlink) {
2557 make_bad_inode(journal_inode);
2558 iput(journal_inode);
617ba13b 2559 printk(KERN_ERR "EXT4-fs: journal inode is deleted.\n");
ac27a0ec
DK		 2560 return NULL;
2561 }
2562
e5f8eab8 2563 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
ac27a0ec 2564 journal_inode, journal_inode->i_size);
1d1fe1ee 2565 if (!S_ISREG(journal_inode->i_mode)) {
617ba13b 2566 printk(KERN_ERR "EXT4-fs: invalid journal inode.\n");
ac27a0ec
DK		 2567 iput(journal_inode);
2568 return NULL;
2569 }
2570
dab291af 2571 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 2572 if (!journal) {
617ba13b 2573 printk(KERN_ERR "EXT4-fs: Could not load journal inode\n");
ac27a0ec
DK		 2574 iput(journal_inode);
2575 return NULL;
2576 }
2577 journal->j_private = sb;
617ba13b 2578 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 2579 return journal;
2580}
2581
617ba13b 2582static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 2583 dev_t j_dev)
2584{
2b2d6d01 2585 struct buffer_head *bh;
ac27a0ec 2586 journal_t *journal;
617ba13b
MC		 2587 ext4_fsblk_t start;
2588 ext4_fsblk_t len;
ac27a0ec 2589 int hblock, blocksize;
617ba13b 2590 ext4_fsblk_t sb_block;
ac27a0ec 2591 unsigned long offset;
2b2d6d01 2592 struct ext4_super_block *es;
ac27a0ec
DK		 2593 struct block_device *bdev;
2594
617ba13b 2595 bdev = ext4_blkdev_get(j_dev);
ac27a0ec
DK		 2596 if (bdev == NULL)
2597 return NULL;
2598
2599 if (bd_claim(bdev, sb)) {
2600 printk(KERN_ERR
8c55e204 2601 "EXT4: failed to claim external journal device.\n");
ac27a0ec
DK		 2602 blkdev_put(bdev);
2603 return NULL;
2604 }
2605
2606 blocksize = sb->s_blocksize;
2607 hblock = bdev_hardsect_size(bdev);
2608 if (blocksize < hblock) {
2609 printk(KERN_ERR
617ba13b 2610 "EXT4-fs: blocksize too small for journal device.\n");
ac27a0ec
DK		 2611 goto out_bdev;
2612 }
2613
617ba13b
MC		 2614 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
2615 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 2616 set_blocksize(bdev, blocksize);
2617 if (!(bh = __bread(bdev, sb_block, blocksize))) {
617ba13b 2618 printk(KERN_ERR "EXT4-fs: couldn't read superblock of "
ac27a0ec
DK		 2619 "external journal\n");
2620 goto out_bdev;
2621 }
2622
617ba13b
MC		 2623 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2624 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 2625 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b
MC		 2626 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
2627 printk(KERN_ERR "EXT4-fs: external journal has "
ac27a0ec
DK		 2628 "bad superblock\n");
2629 brelse(bh);
2630 goto out_bdev;
2631 }
2632
617ba13b
MC		 2633 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
2634 printk(KERN_ERR "EXT4-fs: journal UUID does not match\n");
ac27a0ec
DK		 2635 brelse(bh);
2636 goto out_bdev;
2637 }
2638
bd81d8ee 2639 len = ext4_blocks_count(es);
ac27a0ec
DK		 2640 start = sb_block + 1;
2641 brelse(bh); /* we're done with the superblock */
2642
dab291af 2643 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 2644 start, len, blocksize);
2645 if (!journal) {
617ba13b 2646 printk(KERN_ERR "EXT4-fs: failed to create device journal\n");
ac27a0ec
DK		 2647 goto out_bdev;
2648 }
2649 journal->j_private = sb;
2650 ll_rw_block(READ, 1, &journal->j_sb_buffer);
2651 wait_on_buffer(journal->j_sb_buffer);
2652 if (!buffer_uptodate(journal->j_sb_buffer)) {
617ba13b 2653 printk(KERN_ERR "EXT4-fs: I/O error on journal device\n");
ac27a0ec
DK		 2654 goto out_journal;
2655 }
2656 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
617ba13b 2657 printk(KERN_ERR "EXT4-fs: External journal has more than one "
ac27a0ec
DK		 2658 "user (unsupported) - %d\n",
2659 be32_to_cpu(journal->j_superblock->s_nr_users));
2660 goto out_journal;
2661 }
617ba13b
MC		 2662 EXT4_SB(sb)->journal_bdev = bdev;
2663 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 2664 return journal;
2665out_journal:
dab291af 2666 jbd2_journal_destroy(journal);
ac27a0ec 2667out_bdev:
617ba13b 2668 ext4_blkdev_put(bdev);
ac27a0ec
DK		 2669 return NULL;
2670}
2671
617ba13b
MC		 2672static int ext4_load_journal(struct super_block *sb,
2673 struct ext4_super_block *es,
ac27a0ec
DK		 2674 unsigned long journal_devnum)
2675{
2676 journal_t *journal;
2677 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
2678 dev_t journal_dev;
2679 int err = 0;
2680 int really_read_only;
2681
2682 if (journal_devnum &&
2683 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
617ba13b 2684 printk(KERN_INFO "EXT4-fs: external journal device major/minor "
ac27a0ec
DK		 2685 "numbers have changed\n");
2686 journal_dev = new_decode_dev(journal_devnum);
2687 } else
2688 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
2689
2690 really_read_only = bdev_read_only(sb->s_bdev);
2691
2692 /*
2693 * Are we loading a blank journal or performing recovery after a
2694 * crash? For recovery, we need to check in advance whether we
2695 * can get read-write access to the device.
2696 */
2697
617ba13b 2698 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
ac27a0ec 2699 if (sb->s_flags & MS_RDONLY) {
617ba13b 2700 printk(KERN_INFO "EXT4-fs: INFO: recovery "
ac27a0ec
DK		 2701 "required on readonly filesystem.\n");
2702 if (really_read_only) {
617ba13b 2703 printk(KERN_ERR "EXT4-fs: write access "
ac27a0ec
DK		 2704 "unavailable, cannot proceed.\n");
2705 return -EROFS;
2706 }
2b2d6d01
TT		 2707 printk(KERN_INFO "EXT4-fs: write access will "
2708 "be enabled during recovery.\n");
ac27a0ec
DK		 2709 }
2710 }
2711
2712 if (journal_inum && journal_dev) {
617ba13b 2713 printk(KERN_ERR "EXT4-fs: filesystem has both journal "
ac27a0ec
DK		 2714 "and inode journals!\n");
2715 return -EINVAL;
2716 }
2717
2718 if (journal_inum) {
617ba13b 2719 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 2720 return -EINVAL;
2721 } else {
617ba13b 2722 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 2723 return -EINVAL;
2724 }
2725
4776004f
TT		 2726 if (journal->j_flags & JBD2_BARRIER)
2727 printk(KERN_INFO "EXT4-fs: barriers enabled\n");
2728 else
2729 printk(KERN_INFO "EXT4-fs: barriers disabled\n");
2730
ac27a0ec 2731 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
dab291af 2732 err = jbd2_journal_update_format(journal);
ac27a0ec 2733 if (err) {
617ba13b 2734 printk(KERN_ERR "EXT4-fs: error updating journal.\n");
dab291af 2735 jbd2_journal_destroy(journal);
ac27a0ec
DK		 2736 return err;
2737 }
2738 }
2739
617ba13b 2740 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
dab291af 2741 err = jbd2_journal_wipe(journal, !really_read_only);
ac27a0ec 2742 if (!err)
dab291af 2743 err = jbd2_journal_load(journal);
ac27a0ec
DK		 2744
2745 if (err) {
617ba13b 2746 printk(KERN_ERR "EXT4-fs: error loading journal.\n");
dab291af 2747 jbd2_journal_destroy(journal);
ac27a0ec
DK		 2748 return err;
2749 }
2750
617ba13b
MC		 2751 EXT4_SB(sb)->s_journal = journal;
2752 ext4_clear_journal_err(sb, es);
ac27a0ec
DK		 2753
2754 if (journal_devnum &&
2755 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2756 es->s_journal_dev = cpu_to_le32(journal_devnum);
2757 sb->s_dirt = 1;
2758
2759 /* Make sure we flush the recovery flag to disk. */
617ba13b 2760 ext4_commit_super(sb, es, 1);
ac27a0ec
DK		 2761 }
2762
2763 return 0;
2764}
2765
2b2d6d01
TT		 2766static int ext4_create_journal(struct super_block *sb,
2767 struct ext4_super_block *es,
ac27a0ec
DK		 2768 unsigned int journal_inum)
2769{
2770 journal_t *journal;
6c675bd4 2771 int err;
ac27a0ec
DK		 2772
2773 if (sb->s_flags & MS_RDONLY) {
617ba13b 2774 printk(KERN_ERR "EXT4-fs: readonly filesystem when trying to "
ac27a0ec
DK		 2775 "create journal.\n");
2776 return -EROFS;
2777 }
2778
6c675bd4
BP		 2779 journal = ext4_get_journal(sb, journal_inum);
2780 if (!journal)
ac27a0ec
DK		 2781 return -EINVAL;
2782
617ba13b 2783 printk(KERN_INFO "EXT4-fs: creating new journal on inode %u\n",
ac27a0ec
DK		 2784 journal_inum);
2785
6c675bd4
BP		 2786 err = jbd2_journal_create(journal);
2787 if (err) {
617ba13b 2788 printk(KERN_ERR "EXT4-fs: error creating journal.\n");
dab291af 2789 jbd2_journal_destroy(journal);
ac27a0ec
DK		 2790 return -EIO;
2791 }
2792
617ba13b 2793 EXT4_SB(sb)->s_journal = journal;
ac27a0ec 2794
617ba13b
MC		 2795 ext4_update_dynamic_rev(sb);
2796 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2797 EXT4_SET_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL);
ac27a0ec
DK		 2798
2799 es->s_journal_inum = cpu_to_le32(journal_inum);
2800 sb->s_dirt = 1;
2801
2802 /* Make sure we flush the recovery flag to disk. */
617ba13b 2803 ext4_commit_super(sb, es, 1);
ac27a0ec
DK		 2804
2805 return 0;
2806}
2807
2b2d6d01
TT		 2808static void ext4_commit_super(struct super_block *sb,
2809 struct ext4_super_block *es, int sync)
ac27a0ec 2810{
617ba13b 2811 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
ac27a0ec
DK		 2812
2813 if (!sbh)
2814 return;
914258bf
TT		 2815 if (buffer_write_io_error(sbh)) {
2816 /*
2817 * Oh, dear. A previous attempt to write the
2818 * superblock failed. This could happen because the
2819 * USB device was yanked out. Or it could happen to
2820 * be a transient write error and maybe the block will
2821 * be remapped. Nothing we can do but to retry the
2822 * write and hope for the best.
2823 */
2824 printk(KERN_ERR "ext4: previous I/O error to "
2825 "superblock detected for %s.\n", sb->s_id);
2826 clear_buffer_write_io_error(sbh);
2827 set_buffer_uptodate(sbh);
2828 }
ac27a0ec 2829 es->s_wtime = cpu_to_le32(get_seconds());
bd81d8ee 2830 ext4_free_blocks_count_set(es, ext4_count_free_blocks(sb));
617ba13b 2831 es->s_free_inodes_count = cpu_to_le32(ext4_count_free_inodes(sb));
ac27a0ec
DK		 2832 BUFFER_TRACE(sbh, "marking dirty");
2833 mark_buffer_dirty(sbh);
914258bf 2834 if (sync) {
ac27a0ec 2835 sync_dirty_buffer(sbh);
914258bf
TT		 2836 if (buffer_write_io_error(sbh)) {
2837 printk(KERN_ERR "ext4: I/O error while writing "
2838 "superblock for %s.\n", sb->s_id);
2839 clear_buffer_write_io_error(sbh);
2840 set_buffer_uptodate(sbh);
2841 }
2842 }
ac27a0ec
DK		 2843}
2844
2845
2846/*
2847 * Have we just finished recovery? If so, and if we are mounting (or
2848 * remounting) the filesystem readonly, then we will end up with a
2849 * consistent fs on disk. Record that fact.
2850 */
2b2d6d01
TT		 2851static void ext4_mark_recovery_complete(struct super_block *sb,
2852 struct ext4_super_block *es)
ac27a0ec 2853{
617ba13b 2854 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 2855
dab291af 2856 jbd2_journal_lock_updates(journal);
7ffe1ea8
HK		 2857 if (jbd2_journal_flush(journal) < 0)
2858 goto out;
2859
32c37730 2860 lock_super(sb);
617ba13b 2861 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
ac27a0ec 2862 sb->s_flags & MS_RDONLY) {
617ba13b 2863 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 2864 sb->s_dirt = 0;
617ba13b 2865 ext4_commit_super(sb, es, 1);
ac27a0ec 2866 }
32c37730 2867 unlock_super(sb);
7ffe1ea8
HK		 2868
2869out:
dab291af 2870 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 2871}
2872
2873/*
2874 * If we are mounting (or read-write remounting) a filesystem whose journal
2875 * has recorded an error from a previous lifetime, move that error to the
2876 * main filesystem now.
2877 */
2b2d6d01
TT		 2878static void ext4_clear_journal_err(struct super_block *sb,
2879 struct ext4_super_block *es)
ac27a0ec
DK		 2880{
2881 journal_t *journal;
2882 int j_errno;
2883 const char *errstr;
2884
617ba13b 2885 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 2886
2887 /*
2888 * Now check for any error status which may have been recorded in the
617ba13b 2889 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 2890 */
2891
dab291af 2892 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 2893 if (j_errno) {
2894 char nbuf[16];
2895
617ba13b 2896 errstr = ext4_decode_error(sb, j_errno, nbuf);
46e665e9 2897 ext4_warning(sb, __func__, "Filesystem error recorded "
ac27a0ec 2898 "from previous mount: %s", errstr);
46e665e9 2899 ext4_warning(sb, __func__, "Marking fs in need of "
ac27a0ec
DK		 2900 "filesystem check.");
2901
617ba13b
MC		 2902 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2903 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2b2d6d01 2904 ext4_commit_super(sb, es, 1);
ac27a0ec 2905
dab291af 2906 jbd2_journal_clear_err(journal);
ac27a0ec
DK		 2907 }
2908}
2909
2910/*
2911 * Force the running and committing transactions to commit,
2912 * and wait on the commit.
2913 */
617ba13b 2914int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 2915{
2916 journal_t *journal;
2917 int ret;
2918
2919 if (sb->s_flags & MS_RDONLY)
2920 return 0;
2921
617ba13b 2922 journal = EXT4_SB(sb)->s_journal;
ac27a0ec 2923 sb->s_dirt = 0;
617ba13b 2924 ret = ext4_journal_force_commit(journal);
ac27a0ec
DK		 2925 return ret;
2926}
2927
2928/*
617ba13b 2929 * Ext4 always journals updates to the superblock itself, so we don't
ac27a0ec
DK		 2930 * have to propagate any other updates to the superblock on disk at this
2931 * point. Just start an async writeback to get the buffers on their way
2932 * to the disk.
2933 *
2934 * This implicitly triggers the writebehind on sync().
2935 */
2936
2b2d6d01 2937static void ext4_write_super(struct super_block *sb)
ac27a0ec
DK		 2938{
2939 if (mutex_trylock(&sb->s_lock) != 0)
2940 BUG();
2941 sb->s_dirt = 0;
2942}
2943
617ba13b 2944static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec
DK		 2945{
2946 tid_t target;
2947
ede86cc4 2948 trace_mark(ext4_sync_fs, "dev %s wait %d", sb->s_id, wait);
ac27a0ec 2949 sb->s_dirt = 0;
dab291af 2950 if (jbd2_journal_start_commit(EXT4_SB(sb)->s_journal, &target)) {
ac27a0ec 2951 if (wait)
dab291af 2952 jbd2_log_wait_commit(EXT4_SB(sb)->s_journal, target);
ac27a0ec
DK		 2953 }
2954 return 0;
2955}
2956
2957/*
2958 * LVM calls this function before a (read-only) snapshot is created. This
2959 * gives us a chance to flush the journal completely and mark the fs clean.
2960 */
617ba13b 2961static void ext4_write_super_lockfs(struct super_block *sb)
ac27a0ec
DK		 2962{
2963 sb->s_dirt = 0;
2964
2965 if (!(sb->s_flags & MS_RDONLY)) {
617ba13b 2966 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 2967
2968 /* Now we set up the journal barrier. */
dab291af 2969 jbd2_journal_lock_updates(journal);
7ffe1ea8
HK		 2970
2971 /*
2972 * We don't want to clear needs_recovery flag when we failed
2973 * to flush the journal.
2974 */
2975 if (jbd2_journal_flush(journal) < 0)
2976 return;
ac27a0ec
DK		 2977
2978 /* Journal blocked and flushed, clear needs_recovery flag. */
617ba13b
MC		 2979 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2980 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
ac27a0ec
DK		 2981 }
2982}
2983
2984/*
2985 * Called by LVM after the snapshot is done. We need to reset the RECOVER
2986 * flag here, even though the filesystem is not technically dirty yet.
2987 */
617ba13b 2988static void ext4_unlockfs(struct super_block *sb)
ac27a0ec
DK		 2989{
2990 if (!(sb->s_flags & MS_RDONLY)) {
2991 lock_super(sb);
2992 /* Reser the needs_recovery flag before the fs is unlocked. */
617ba13b
MC		 2993 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2994 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
ac27a0ec 2995 unlock_super(sb);
dab291af 2996 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
ac27a0ec
DK		 2997 }
2998}
2999
2b2d6d01 3000static int ext4_remount(struct super_block *sb, int *flags, char *data)
ac27a0ec 3001{
2b2d6d01 3002 struct ext4_super_block *es;
617ba13b
MC		 3003 struct ext4_sb_info *sbi = EXT4_SB(sb);
3004 ext4_fsblk_t n_blocks_count = 0;
ac27a0ec 3005 unsigned long old_sb_flags;
617ba13b 3006 struct ext4_mount_options old_opts;
8a266467 3007 ext4_group_t g;
ac27a0ec
DK		 3008 int err;
3009#ifdef CONFIG_QUOTA
3010 int i;
3011#endif
3012
3013 /* Store the original options */
3014 old_sb_flags = sb->s_flags;
3015 old_opts.s_mount_opt = sbi->s_mount_opt;
3016 old_opts.s_resuid = sbi->s_resuid;
3017 old_opts.s_resgid = sbi->s_resgid;
3018 old_opts.s_commit_interval = sbi->s_commit_interval;
3019#ifdef CONFIG_QUOTA
3020 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
3021 for (i = 0; i < MAXQUOTAS; i++)
3022 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
3023#endif
3024
3025 /*
3026 * Allow the "check" option to be passed as a remount option.
3027 */
3028 if (!parse_options(data, sb, NULL, NULL, &n_blocks_count, 1)) {
3029 err = -EINVAL;
3030 goto restore_opts;
3031 }
3032
617ba13b 3033 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT)
46e665e9 3034 ext4_abort(sb, __func__, "Abort forced by user");
ac27a0ec
DK		 3035
3036 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
617ba13b 3037 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec
DK		 3038
3039 es = sbi->s_es;
3040
617ba13b 3041 ext4_init_journal_params(sb, sbi->s_journal);
ac27a0ec
DK		 3042
3043 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
bd81d8ee 3044 n_blocks_count > ext4_blocks_count(es)) {
617ba13b 3045 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT) {
ac27a0ec
DK		 3046 err = -EROFS;
3047 goto restore_opts;
3048 }
3049
3050 if (*flags & MS_RDONLY) {
3051 /*
3052 * First of all, the unconditional stuff we have to do
3053 * to disable replay of the journal when we next remount
3054 */
3055 sb->s_flags |= MS_RDONLY;
3056
3057 /*
3058 * OK, test if we are remounting a valid rw partition
3059 * readonly, and if so set the rdonly flag and then
3060 * mark the partition as valid again.
3061 */
617ba13b
MC		 3062 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
3063 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 3064 es->s_state = cpu_to_le16(sbi->s_mount_state);
3065
32c37730
JK		 3066 /*
3067 * We have to unlock super so that we can wait for
3068 * transactions.
3069 */
3070 unlock_super(sb);
617ba13b 3071 ext4_mark_recovery_complete(sb, es);
32c37730 3072 lock_super(sb);
ac27a0ec
DK		 3073 } else {
3074 __le32 ret;
617ba13b
MC		 3075 if ((ret = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3076 ~EXT4_FEATURE_RO_COMPAT_SUPP))) {
3077 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
ac27a0ec
DK		 3078 "remount RDWR because of unsupported "
3079 "optional features (%x).\n",
3080 sb->s_id, le32_to_cpu(ret));
3081 err = -EROFS;
3082 goto restore_opts;
3083 }
ead6596b 3084
8a266467
TT		 3085 /*
3086 * Make sure the group descriptor checksums
3087 * are sane. If they aren't, refuse to
3088 * remount r/w.
3089 */
3090 for (g = 0; g < sbi->s_groups_count; g++) {
3091 struct ext4_group_desc *gdp =
3092 ext4_get_group_desc(sb, g, NULL);
3093
3094 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
3095 printk(KERN_ERR
3096 "EXT4-fs: ext4_remount: "
3097 "Checksum for group %lu failed (%u!=%u)\n",
3098 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
3099 le16_to_cpu(gdp->bg_checksum));
3100 err = -EINVAL;
3101 goto restore_opts;
3102 }
3103 }
3104
ead6596b
ES		 3105 /*
3106 * If we have an unprocessed orphan list hanging
3107 * around from a previously readonly bdev mount,
3108 * require a full umount/remount for now.
3109 */
3110 if (es->s_last_orphan) {
3111 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
3112 "remount RDWR because of unprocessed "
3113 "orphan inode list. Please "
3114 "umount/remount instead.\n",
3115 sb->s_id);
3116 err = -EINVAL;
3117 goto restore_opts;
3118 }
3119
ac27a0ec
DK		 3120 /*
3121 * Mounting a RDONLY partition read-write, so reread
3122 * and store the current valid flag. (It may have
3123 * been changed by e2fsck since we originally mounted
3124 * the partition.)
3125 */
617ba13b 3126 ext4_clear_journal_err(sb, es);
ac27a0ec 3127 sbi->s_mount_state = le16_to_cpu(es->s_state);
617ba13b 3128 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
ac27a0ec 3129 goto restore_opts;
2b2d6d01 3130 if (!ext4_setup_super(sb, es, 0))
ac27a0ec
DK		 3131 sb->s_flags &= ~MS_RDONLY;
3132 }
3133 }
3134#ifdef CONFIG_QUOTA
3135 /* Release old quota file names */
3136 for (i = 0; i < MAXQUOTAS; i++)
3137 if (old_opts.s_qf_names[i] &&
3138 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3139 kfree(old_opts.s_qf_names[i]);
3140#endif
3141 return 0;
3142restore_opts:
3143 sb->s_flags = old_sb_flags;
3144 sbi->s_mount_opt = old_opts.s_mount_opt;
3145 sbi->s_resuid = old_opts.s_resuid;
3146 sbi->s_resgid = old_opts.s_resgid;
3147 sbi->s_commit_interval = old_opts.s_commit_interval;
3148#ifdef CONFIG_QUOTA
3149 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
3150 for (i = 0; i < MAXQUOTAS; i++) {
3151 if (sbi->s_qf_names[i] &&
3152 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3153 kfree(sbi->s_qf_names[i]);
3154 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
3155 }
3156#endif
3157 return err;
3158}
3159
2b2d6d01 3160static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
ac27a0ec
DK		 3161{
3162 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 3163 struct ext4_sb_info *sbi = EXT4_SB(sb);
3164 struct ext4_super_block *es = sbi->s_es;
960cc398 3165 u64 fsid;
ac27a0ec 3166
5e70030d
BP		 3167 if (test_opt(sb, MINIX_DF)) {
3168 sbi->s_overhead_last = 0;
6bc9feff 3169 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
fd2d4291 3170 ext4_group_t ngroups = sbi->s_groups_count, i;
5e70030d 3171 ext4_fsblk_t overhead = 0;
ac27a0ec
DK		 3172 smp_rmb();
3173
3174 /*
5e70030d
BP		 3175 * Compute the overhead (FS structures). This is constant
3176 * for a given filesystem unless the number of block groups
3177 * changes so we cache the previous value until it does.
ac27a0ec
DK		 3178 */
3179
3180 /*
3181 * All of the blocks before first_data_block are
3182 * overhead
3183 */
3184 overhead = le32_to_cpu(es->s_first_data_block);
3185
3186 /*
3187 * Add the overhead attributed to the superblock and
3188 * block group descriptors. If the sparse superblocks
3189 * feature is turned on, then not all groups have this.
3190 */
3191 for (i = 0; i < ngroups; i++) {
617ba13b
MC		 3192 overhead += ext4_bg_has_super(sb, i) +
3193 ext4_bg_num_gdb(sb, i);
ac27a0ec
DK		 3194 cond_resched();
3195 }
3196
3197 /*
3198 * Every block group has an inode bitmap, a block
3199 * bitmap, and an inode table.
3200 */
5e70030d
BP		 3201 overhead += ngroups * (2 + sbi->s_itb_per_group);
3202 sbi->s_overhead_last = overhead;
3203 smp_wmb();
6bc9feff 3204 sbi->s_blocks_last = ext4_blocks_count(es);
ac27a0ec
DK		 3205 }
3206
617ba13b 3207 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 3208 buf->f_bsize = sb->s_blocksize;
5e70030d 3209 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
6bc6e63f
AK		 3210 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
3211 percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
308ba3ec 3212 ext4_free_blocks_count_set(es, buf->f_bfree);
bd81d8ee
LV		 3213 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3214 if (buf->f_bfree < ext4_r_blocks_count(es))
ac27a0ec
DK		 3215 buf->f_bavail = 0;
3216 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 3217 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
5e70030d 3218 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
617ba13b 3219 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 3220 fsid = le64_to_cpup((void *)es->s_uuid) ^
3221 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3222 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3223 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
ac27a0ec
DK		 3224 return 0;
3225}
3226
3227/* Helper function for writing quotas on sync - we need to start transaction before quota file
3228 * is locked for write. Otherwise the are possible deadlocks:
3229 * Process 1 Process 2
617ba13b 3230 * ext4_create() quota_sync()
dab291af 3231 * jbd2_journal_start() write_dquot()
ac27a0ec 3232 * DQUOT_INIT() down(dqio_mutex)
dab291af 3233 * down(dqio_mutex) jbd2_journal_start()
ac27a0ec
DK		 3234 *
3235 */
3236
3237#ifdef CONFIG_QUOTA
3238
3239static inline struct inode *dquot_to_inode(struct dquot *dquot)
3240{
3241 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3242}
3243
617ba13b 3244static int ext4_dquot_initialize(struct inode *inode, int type)
ac27a0ec
DK		 3245{
3246 handle_t *handle;
3247 int ret, err;
3248
3249 /* We may create quota structure so we need to reserve enough blocks */
617ba13b 3250 handle = ext4_journal_start(inode, 2*EXT4_QUOTA_INIT_BLOCKS(inode->i_sb));
ac27a0ec
DK		 3251 if (IS_ERR(handle))
3252 return PTR_ERR(handle);
3253 ret = dquot_initialize(inode, type);
617ba13b 3254 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3255 if (!ret)
3256 ret = err;
3257 return ret;
3258}
3259
617ba13b 3260static int ext4_dquot_drop(struct inode *inode)
ac27a0ec
DK		 3261{
3262 handle_t *handle;
3263 int ret, err;
3264
3265 /* We may delete quota structure so we need to reserve enough blocks */
617ba13b 3266 handle = ext4_journal_start(inode, 2*EXT4_QUOTA_DEL_BLOCKS(inode->i_sb));
2887df13
JK		 3267 if (IS_ERR(handle)) {
3268 /*
3269 * We call dquot_drop() anyway to at least release references
3270 * to quota structures so that umount does not hang.
3271 */
3272 dquot_drop(inode);
ac27a0ec 3273 return PTR_ERR(handle);
2887df13 3274 }
ac27a0ec 3275 ret = dquot_drop(inode);
617ba13b 3276 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3277 if (!ret)
3278 ret = err;
3279 return ret;
3280}
3281
617ba13b 3282static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 3283{
3284 int ret, err;
3285 handle_t *handle;
3286 struct inode *inode;
3287
3288 inode = dquot_to_inode(dquot);
617ba13b
MC		 3289 handle = ext4_journal_start(inode,
3290 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 3291 if (IS_ERR(handle))
3292 return PTR_ERR(handle);
3293 ret = dquot_commit(dquot);
617ba13b 3294 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3295 if (!ret)
3296 ret = err;
3297 return ret;
3298}
3299
617ba13b 3300static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 3301{
3302 int ret, err;
3303 handle_t *handle;
3304
617ba13b
MC		 3305 handle = ext4_journal_start(dquot_to_inode(dquot),
3306 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 3307 if (IS_ERR(handle))
3308 return PTR_ERR(handle);
3309 ret = dquot_acquire(dquot);
617ba13b 3310 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3311 if (!ret)
3312 ret = err;
3313 return ret;
3314}
3315
617ba13b 3316static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 3317{
3318 int ret, err;
3319 handle_t *handle;
3320
617ba13b
MC		 3321 handle = ext4_journal_start(dquot_to_inode(dquot),
3322 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 3323 if (IS_ERR(handle)) {
3324 /* Release dquot anyway to avoid endless cycle in dqput() */
3325 dquot_release(dquot);
ac27a0ec 3326 return PTR_ERR(handle);
9c3013e9 3327 }
ac27a0ec 3328 ret = dquot_release(dquot);
617ba13b 3329 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3330 if (!ret)
3331 ret = err;
3332 return ret;
3333}
3334
617ba13b 3335static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec 3336{
2c8be6b2 3337 /* Are we journaling quotas? */
617ba13b
MC		 3338 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3339 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
ac27a0ec 3340 dquot_mark_dquot_dirty(dquot);
617ba13b 3341 return ext4_write_dquot(dquot);
ac27a0ec
DK		 3342 } else {
3343 return dquot_mark_dquot_dirty(dquot);
3344 }
3345}
3346
617ba13b 3347static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 3348{
3349 int ret, err;
3350 handle_t *handle;
3351
3352 /* Data block + inode block */
617ba13b 3353 handle = ext4_journal_start(sb->s_root->d_inode, 2);
ac27a0ec
DK		 3354 if (IS_ERR(handle))
3355 return PTR_ERR(handle);
3356 ret = dquot_commit_info(sb, type);
617ba13b 3357 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3358 if (!ret)
3359 ret = err;
3360 return ret;
3361}
3362
3363/*
3364 * Turn on quotas during mount time - we need to find
3365 * the quota file and such...
3366 */
617ba13b 3367static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 3368{
617ba13b
MC		 3369 return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
3370 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 3371}
3372
3373/*
3374 * Standard function to be called on quota_on
3375 */
617ba13b 3376static int ext4_quota_on(struct super_block *sb, int type, int format_id,
6f28e087 3377 char *path, int remount)
ac27a0ec
DK		 3378{
3379 int err;
3380 struct nameidata nd;
3381
3382 if (!test_opt(sb, QUOTA))
3383 return -EINVAL;
0623543b
JK		 3384 /* When remounting, no checks are needed and in fact, path is NULL */
3385 if (remount)
6f28e087 3386 return vfs_quota_on(sb, type, format_id, path, remount);
0623543b 3387
ac27a0ec
DK		 3388 err = path_lookup(path, LOOKUP_FOLLOW, &nd);
3389 if (err)
3390 return err;
0623543b 3391
ac27a0ec 3392 /* Quotafile not on the same filesystem? */
4ac91378 3393 if (nd.path.mnt->mnt_sb != sb) {
1d957f9b 3394 path_put(&nd.path);
ac27a0ec
DK		 3395 return -EXDEV;
3396 }
0623543b
JK		 3397 /* Journaling quota? */
3398 if (EXT4_SB(sb)->s_qf_names[type]) {
2b2d6d01 3399 /* Quotafile not in fs root? */
0623543b
JK		 3400 if (nd.path.dentry->d_parent->d_inode != sb->s_root->d_inode)
3401 printk(KERN_WARNING
3402 "EXT4-fs: Quota file not on filesystem root. "
3403 "Journaled quota will not work.\n");
2b2d6d01 3404 }
0623543b
JK		 3405
3406 /*
3407 * When we journal data on quota file, we have to flush journal to see
3408 * all updates to the file when we bypass pagecache...
3409 */
3410 if (ext4_should_journal_data(nd.path.dentry->d_inode)) {
3411 /*
3412 * We don't need to lock updates but journal_flush() could
3413 * otherwise be livelocked...
3414 */
3415 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 3416 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
0623543b 3417 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8
HK		 3418 if (err) {
3419 path_put(&nd.path);
3420 return err;
3421 }
0623543b
JK		 3422 }
3423
77e69dac 3424 err = vfs_quota_on_path(sb, type, format_id, &nd.path);
1d957f9b 3425 path_put(&nd.path);
77e69dac 3426 return err;
ac27a0ec
DK		 3427}
3428
3429/* Read data from quotafile - avoid pagecache and such because we cannot afford
3430 * acquiring the locks... As quota files are never truncated and quota code
3431 * itself serializes the operations (and noone else should touch the files)
3432 * we don't have to be afraid of races */
617ba13b 3433static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 3434 size_t len, loff_t off)
3435{
3436 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 3437 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 3438 int err = 0;
3439 int offset = off & (sb->s_blocksize - 1);
3440 int tocopy;
3441 size_t toread;
3442 struct buffer_head *bh;
3443 loff_t i_size = i_size_read(inode);
3444
3445 if (off > i_size)
3446 return 0;
3447 if (off+len > i_size)
3448 len = i_size-off;
3449 toread = len;
3450 while (toread > 0) {
3451 tocopy = sb->s_blocksize - offset < toread ?
3452 sb->s_blocksize - offset : toread;
617ba13b 3453 bh = ext4_bread(NULL, inode, blk, 0, &err);
ac27a0ec
DK		 3454 if (err)
3455 return err;
3456 if (!bh) /* A hole? */
3457 memset(data, 0, tocopy);
3458 else
3459 memcpy(data, bh->b_data+offset, tocopy);
3460 brelse(bh);
3461 offset = 0;
3462 toread -= tocopy;
3463 data += tocopy;
3464 blk++;
3465 }
3466 return len;
3467}
3468
3469/* Write to quotafile (we know the transaction is already started and has
3470 * enough credits) */
617ba13b 3471static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 3472 const char *data, size_t len, loff_t off)
3473{
3474 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 3475 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 3476 int err = 0;
3477 int offset = off & (sb->s_blocksize - 1);
3478 int tocopy;
617ba13b 3479 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
ac27a0ec
DK		 3480 size_t towrite = len;
3481 struct buffer_head *bh;
3482 handle_t *handle = journal_current_handle();
3483
9c3013e9 3484 if (!handle) {
e5f8eab8 3485 printk(KERN_WARNING "EXT4-fs: Quota write (off=%llu, len=%llu)"
9c3013e9
JK		 3486 " cancelled because transaction is not started.\n",
3487 (unsigned long long)off, (unsigned long long)len);
3488 return -EIO;
3489 }
ac27a0ec
DK		 3490 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
3491 while (towrite > 0) {
3492 tocopy = sb->s_blocksize - offset < towrite ?
3493 sb->s_blocksize - offset : towrite;
617ba13b 3494 bh = ext4_bread(handle, inode, blk, 1, &err);
ac27a0ec
DK		 3495 if (!bh)
3496 goto out;
3497 if (journal_quota) {
617ba13b 3498 err = ext4_journal_get_write_access(handle, bh);
ac27a0ec
DK		 3499 if (err) {
3500 brelse(bh);
3501 goto out;
3502 }
3503 }
3504 lock_buffer(bh);
3505 memcpy(bh->b_data+offset, data, tocopy);
3506 flush_dcache_page(bh->b_page);
3507 unlock_buffer(bh);
3508 if (journal_quota)
617ba13b 3509 err = ext4_journal_dirty_metadata(handle, bh);
ac27a0ec
DK		 3510 else {
3511 /* Always do at least ordered writes for quotas */
678aaf48 3512 err = ext4_jbd2_file_inode(handle, inode);
ac27a0ec
DK		 3513 mark_buffer_dirty(bh);
3514 }
3515 brelse(bh);
3516 if (err)
3517 goto out;
3518 offset = 0;
3519 towrite -= tocopy;
3520 data += tocopy;
3521 blk++;
3522 }
3523out:
4d04e4fb
JK		 3524 if (len == towrite) {
3525 mutex_unlock(&inode->i_mutex);
ac27a0ec 3526 return err;
4d04e4fb 3527 }
ac27a0ec
DK		 3528 if (inode->i_size < off+len-towrite) {
3529 i_size_write(inode, off+len-towrite);
617ba13b 3530 EXT4_I(inode)->i_disksize = inode->i_size;
ac27a0ec 3531 }
ac27a0ec 3532 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
617ba13b 3533 ext4_mark_inode_dirty(handle, inode);
ac27a0ec
DK		 3534 mutex_unlock(&inode->i_mutex);
3535 return len - towrite;
3536}
3537
3538#endif
3539
617ba13b 3540static int ext4_get_sb(struct file_system_type *fs_type,
ac27a0ec
DK		 3541 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3542{
617ba13b 3543 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
ac27a0ec
DK		 3544}
3545
5e8814f2
TT		 3546#ifdef CONFIG_PROC_FS
3547static int ext4_ui_proc_show(struct seq_file *m, void *v)
3548{
3549 unsigned int *p = m->private;
3550
3551 seq_printf(m, "%u\n", *p);
3552 return 0;
3553}
3554
3555static int ext4_ui_proc_open(struct inode *inode, struct file *file)
3556{
3557 return single_open(file, ext4_ui_proc_show, PDE(inode)->data);
3558}
3559
3560static ssize_t ext4_ui_proc_write(struct file *file, const char __user *buf,
3561 size_t cnt, loff_t *ppos)
3562{
3563 unsigned int *p = PDE(file->f_path.dentry->d_inode)->data;
3564 char str[32];
3565 unsigned long value;
3566
3567 if (cnt >= sizeof(str))
3568 return -EINVAL;
3569 if (copy_from_user(str, buf, cnt))
3570 return -EFAULT;
3571 value = simple_strtol(str, NULL, 0);
3572 if (value < 0)
3573 return -ERANGE;
3574 *p = value;
3575 return cnt;
3576}
3577
3578const struct file_operations ext4_ui_proc_fops = {
3579 .owner = THIS_MODULE,
3580 .open = ext4_ui_proc_open,
3581 .read = seq_read,
3582 .llseek = seq_lseek,
3583 .release = single_release,
3584 .write = ext4_ui_proc_write,
3585};
3586#endif
3587
03010a33
TT		 3588static struct file_system_type ext4_fs_type = {
3589 .owner = THIS_MODULE,
3590 .name = "ext4",
3591 .get_sb = ext4_get_sb,
3592 .kill_sb = kill_block_super,
3593 .fs_flags = FS_REQUIRES_DEV,
3594};
3595
3596#ifdef CONFIG_EXT4DEV_COMPAT
3597static int ext4dev_get_sb(struct file_system_type *fs_type,
3598 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3599{
3600 printk(KERN_WARNING "EXT4-fs: Update your userspace programs "
3601 "to mount using ext4\n");
3602 printk(KERN_WARNING "EXT4-fs: ext4dev backwards compatibility "
3603 "will go away by 2.6.31\n");
3604 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
3605}
3606
617ba13b 3607static struct file_system_type ext4dev_fs_type = {
ac27a0ec 3608 .owner = THIS_MODULE,
617ba13b 3609 .name = "ext4dev",
03010a33 3610 .get_sb = ext4dev_get_sb,
ac27a0ec
DK		 3611 .kill_sb = kill_block_super,
3612 .fs_flags = FS_REQUIRES_DEV,
3613};
03010a33
TT		 3614MODULE_ALIAS("ext4dev");
3615#endif
ac27a0ec 3616
617ba13b 3617static int __init init_ext4_fs(void)
ac27a0ec 3618{
c9de560d
AT		 3619 int err;
3620
9f6200bb 3621 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
c9de560d 3622 err = init_ext4_mballoc();
ac27a0ec
DK		 3623 if (err)
3624 return err;
c9de560d
AT		 3625
3626 err = init_ext4_xattr();
3627 if (err)
3628 goto out2;
ac27a0ec
DK		 3629 err = init_inodecache();
3630 if (err)
3631 goto out1;
03010a33 3632 err = register_filesystem(&ext4_fs_type);
ac27a0ec
DK		 3633 if (err)
3634 goto out;
03010a33
TT		 3635#ifdef CONFIG_EXT4DEV_COMPAT
3636 err = register_filesystem(&ext4dev_fs_type);
3637 if (err) {
3638 unregister_filesystem(&ext4_fs_type);
3639 goto out;
3640 }
3641#endif
ac27a0ec
DK		 3642 return 0;
3643out:
3644 destroy_inodecache();
3645out1:
617ba13b 3646 exit_ext4_xattr();
c9de560d
AT		 3647out2:
3648 exit_ext4_mballoc();
ac27a0ec
DK		 3649 return err;
3650}
3651
617ba13b 3652static void __exit exit_ext4_fs(void)
ac27a0ec 3653{
03010a33
TT		 3654 unregister_filesystem(&ext4_fs_type);
3655#ifdef CONFIG_EXT4DEV_COMPAT
617ba13b 3656 unregister_filesystem(&ext4dev_fs_type);
03010a33 3657#endif
ac27a0ec 3658 destroy_inodecache();
617ba13b 3659 exit_ext4_xattr();
c9de560d 3660 exit_ext4_mballoc();
9f6200bb 3661 remove_proc_entry("fs/ext4", NULL);
ac27a0ec
DK		 3662}
3663
3664MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
617ba13b 3665MODULE_DESCRIPTION("Fourth Extended Filesystem with extents");
ac27a0ec 3666MODULE_LICENSE("GPL");
617ba13b
MC		 3667module_init(init_ext4_fs)
3668module_exit(exit_ext4_fs)
ubuntu-bionic-kernel mirror