]> 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
ext4: mballoc avoid use root reserved blocks for non root allocation
[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>
1330593e 37#include <linux/log2.h>
717d50e4 38#include <linux/crc16.h>
ac27a0ec
DK		 39#include <asm/uaccess.h>
40
3dcf5451
CH		 41#include "ext4.h"
42#include "ext4_jbd2.h"
ac27a0ec
DK		 43#include "xattr.h"
44#include "acl.h"
45#include "namei.h"
717d50e4 46#include "group.h"
ac27a0ec 47
617ba13b 48static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
ac27a0ec 49 unsigned long journal_devnum);
617ba13b 50static int ext4_create_journal(struct super_block *, struct ext4_super_block *,
ac27a0ec 51 unsigned int);
617ba13b
MC		 52static void ext4_commit_super (struct super_block * sb,
53 struct ext4_super_block * es,
ac27a0ec 54 int sync);
617ba13b
MC		 55static void ext4_mark_recovery_complete(struct super_block * sb,
56 struct ext4_super_block * es);
57static void ext4_clear_journal_err(struct super_block * sb,
58 struct ext4_super_block * es);
59static int ext4_sync_fs(struct super_block *sb, int wait);
60static const char *ext4_decode_error(struct super_block * sb, int errno,
ac27a0ec 61 char nbuf[16]);
617ba13b
MC		 62static int ext4_remount (struct super_block * sb, int * flags, char * data);
63static int ext4_statfs (struct dentry * dentry, struct kstatfs * buf);
64static void ext4_unlockfs(struct super_block *sb);
65static void ext4_write_super (struct super_block * sb);
66static void ext4_write_super_lockfs(struct super_block *sb);
ac27a0ec 67
bd81d8ee 68
8fadc143
AR		 69ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
70 struct ext4_group_desc *bg)
bd81d8ee 71{
3a14589c 72 return le32_to_cpu(bg->bg_block_bitmap_lo) |
8fadc143 73 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
3a14589c 74 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 75}
76
8fadc143
AR		 77ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
78 struct ext4_group_desc *bg)
bd81d8ee 79{
5272f837 80 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
8fadc143 81 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
5272f837 82 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 83}
84
8fadc143
AR		 85ext4_fsblk_t ext4_inode_table(struct super_block *sb,
86 struct ext4_group_desc *bg)
bd81d8ee 87{
5272f837 88 return le32_to_cpu(bg->bg_inode_table_lo) |
8fadc143 89 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
5272f837 90 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
bd81d8ee
LV		 91}
92
8fadc143
AR		 93void ext4_block_bitmap_set(struct super_block *sb,
94 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 95{
3a14589c 96 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 97 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
98 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 99}
100
8fadc143
AR		 101void ext4_inode_bitmap_set(struct super_block *sb,
102 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 103{
5272f837 104 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 105 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
106 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 107}
108
8fadc143
AR		 109void ext4_inode_table_set(struct super_block *sb,
110 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 111{
5272f837 112 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 113 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
114 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 115}
116
ac27a0ec 117/*
dab291af 118 * Wrappers for jbd2_journal_start/end.
ac27a0ec
DK		 119 *
120 * The only special thing we need to do here is to make sure that all
121 * journal_end calls result in the superblock being marked dirty, so
122 * that sync() will call the filesystem's write_super callback if
123 * appropriate.
124 */
617ba13b 125handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
ac27a0ec
DK		 126{
127 journal_t *journal;
128
129 if (sb->s_flags & MS_RDONLY)
130 return ERR_PTR(-EROFS);
131
132 /* Special case here: if the journal has aborted behind our
133 * backs (eg. EIO in the commit thread), then we still need to
134 * take the FS itself readonly cleanly. */
617ba13b 135 journal = EXT4_SB(sb)->s_journal;
ac27a0ec 136 if (is_journal_aborted(journal)) {
46e665e9 137 ext4_abort(sb, __func__,
ac27a0ec
DK		 138 "Detected aborted journal");
139 return ERR_PTR(-EROFS);
140 }
141
dab291af 142 return jbd2_journal_start(journal, nblocks);
ac27a0ec
DK		 143}
144
145/*
146 * The only special thing we need to do here is to make sure that all
dab291af 147 * jbd2_journal_stop calls result in the superblock being marked dirty, so
ac27a0ec
DK		 148 * that sync() will call the filesystem's write_super callback if
149 * appropriate.
150 */
617ba13b 151int __ext4_journal_stop(const char *where, handle_t *handle)
ac27a0ec
DK		 152{
153 struct super_block *sb;
154 int err;
155 int rc;
156
157 sb = handle->h_transaction->t_journal->j_private;
158 err = handle->h_err;
dab291af 159 rc = jbd2_journal_stop(handle);
ac27a0ec
DK		 160
161 if (!err)
162 err = rc;
163 if (err)
617ba13b 164 __ext4_std_error(sb, where, err);
ac27a0ec
DK		 165 return err;
166}
167
617ba13b 168void ext4_journal_abort_handle(const char *caller, const char *err_fn,
ac27a0ec
DK		 169 struct buffer_head *bh, handle_t *handle, int err)
170{
171 char nbuf[16];
617ba13b 172 const char *errstr = ext4_decode_error(NULL, err, nbuf);
ac27a0ec
DK		 173
174 if (bh)
175 BUFFER_TRACE(bh, "abort");
176
177 if (!handle->h_err)
178 handle->h_err = err;
179
180 if (is_handle_aborted(handle))
181 return;
182
183 printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
184 caller, errstr, err_fn);
185
dab291af 186 jbd2_journal_abort_handle(handle);
ac27a0ec
DK		 187}
188
189/* Deal with the reporting of failure conditions on a filesystem such as
190 * inconsistencies detected or read IO failures.
191 *
192 * On ext2, we can store the error state of the filesystem in the
617ba13b 193 * superblock. That is not possible on ext4, because we may have other
ac27a0ec
DK		 194 * write ordering constraints on the superblock which prevent us from
195 * writing it out straight away; and given that the journal is about to
196 * be aborted, we can't rely on the current, or future, transactions to
197 * write out the superblock safely.
198 *
dab291af 199 * We'll just use the jbd2_journal_abort() error code to record an error in
ac27a0ec
DK		 200 * the journal instead. On recovery, the journal will compain about
201 * that error until we've noted it down and cleared it.
202 */
203
617ba13b 204static void ext4_handle_error(struct super_block *sb)
ac27a0ec 205{
617ba13b 206 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 207
617ba13b
MC		 208 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
209 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
ac27a0ec
DK		 210
211 if (sb->s_flags & MS_RDONLY)
212 return;
213
214 if (!test_opt (sb, ERRORS_CONT)) {
617ba13b 215 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 216
617ba13b 217 EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
ac27a0ec 218 if (journal)
dab291af 219 jbd2_journal_abort(journal, -EIO);
ac27a0ec
DK		 220 }
221 if (test_opt (sb, ERRORS_RO)) {
222 printk (KERN_CRIT "Remounting filesystem read-only\n");
223 sb->s_flags |= MS_RDONLY;
224 }
617ba13b 225 ext4_commit_super(sb, es, 1);
ac27a0ec 226 if (test_opt(sb, ERRORS_PANIC))
617ba13b 227 panic("EXT4-fs (device %s): panic forced after error\n",
ac27a0ec
DK		 228 sb->s_id);
229}
230
617ba13b 231void ext4_error (struct super_block * sb, const char * function,
ac27a0ec
DK		 232 const char * fmt, ...)
233{
234 va_list args;
235
236 va_start(args, fmt);
617ba13b 237 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ",sb->s_id, function);
ac27a0ec
DK		 238 vprintk(fmt, args);
239 printk("\n");
240 va_end(args);
241
617ba13b 242 ext4_handle_error(sb);
ac27a0ec
DK		 243}
244
617ba13b 245static const char *ext4_decode_error(struct super_block * sb, int errno,
ac27a0ec
DK		 246 char nbuf[16])
247{
248 char *errstr = NULL;
249
250 switch (errno) {
251 case -EIO:
252 errstr = "IO failure";
253 break;
254 case -ENOMEM:
255 errstr = "Out of memory";
256 break;
257 case -EROFS:
dab291af 258 if (!sb || EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT)
ac27a0ec
DK		 259 errstr = "Journal has aborted";
260 else
261 errstr = "Readonly filesystem";
262 break;
263 default:
264 /* If the caller passed in an extra buffer for unknown
265 * errors, textualise them now. Else we just return
266 * NULL. */
267 if (nbuf) {
268 /* Check for truncated error codes... */
269 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
270 errstr = nbuf;
271 }
272 break;
273 }
274
275 return errstr;
276}
277
617ba13b 278/* __ext4_std_error decodes expected errors from journaling functions
ac27a0ec
DK		 279 * automatically and invokes the appropriate error response. */
280
617ba13b 281void __ext4_std_error (struct super_block * sb, const char * function,
ac27a0ec
DK		 282 int errno)
283{
284 char nbuf[16];
285 const char *errstr;
286
287 /* Special case: if the error is EROFS, and we're not already
288 * inside a transaction, then there's really no point in logging
289 * an error. */
290 if (errno == -EROFS && journal_current_handle() == NULL &&
291 (sb->s_flags & MS_RDONLY))
292 return;
293
617ba13b
MC		 294 errstr = ext4_decode_error(sb, errno, nbuf);
295 printk (KERN_CRIT "EXT4-fs error (device %s) in %s: %s\n",
ac27a0ec
DK		 296 sb->s_id, function, errstr);
297
617ba13b 298 ext4_handle_error(sb);
ac27a0ec
DK		 299}
300
301/*
617ba13b 302 * ext4_abort is a much stronger failure handler than ext4_error. The
ac27a0ec
DK		 303 * abort function may be used to deal with unrecoverable failures such
304 * as journal IO errors or ENOMEM at a critical moment in log management.
305 *
306 * We unconditionally force the filesystem into an ABORT|READONLY state,
307 * unless the error response on the fs has been set to panic in which
308 * case we take the easy way out and panic immediately.
309 */
310
617ba13b 311void ext4_abort (struct super_block * sb, const char * function,
ac27a0ec
DK		 312 const char * fmt, ...)
313{
314 va_list args;
315
617ba13b 316 printk (KERN_CRIT "ext4_abort called.\n");
ac27a0ec
DK		 317
318 va_start(args, fmt);
617ba13b 319 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ",sb->s_id, function);
ac27a0ec
DK		 320 vprintk(fmt, args);
321 printk("\n");
322 va_end(args);
323
324 if (test_opt(sb, ERRORS_PANIC))
617ba13b 325 panic("EXT4-fs panic from previous error\n");
ac27a0ec
DK		 326
327 if (sb->s_flags & MS_RDONLY)
328 return;
329
330 printk(KERN_CRIT "Remounting filesystem read-only\n");
617ba13b 331 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
ac27a0ec 332 sb->s_flags |= MS_RDONLY;
617ba13b 333 EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
dab291af 334 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
ac27a0ec
DK		 335}
336
617ba13b 337void ext4_warning (struct super_block * sb, const char * function,
ac27a0ec
DK		 338 const char * fmt, ...)
339{
340 va_list args;
341
342 va_start(args, fmt);
617ba13b 343 printk(KERN_WARNING "EXT4-fs warning (device %s): %s: ",
ac27a0ec
DK		 344 sb->s_id, function);
345 vprintk(fmt, args);
346 printk("\n");
347 va_end(args);
348}
349
617ba13b 350void ext4_update_dynamic_rev(struct super_block *sb)
ac27a0ec 351{
617ba13b 352 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 353
617ba13b 354 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
ac27a0ec
DK		 355 return;
356
46e665e9 357 ext4_warning(sb, __func__,
ac27a0ec
DK		 358 "updating to rev %d because of new feature flag, "
359 "running e2fsck is recommended",
617ba13b 360 EXT4_DYNAMIC_REV);
ac27a0ec 361
617ba13b
MC		 362 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
363 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
364 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
ac27a0ec
DK		 365 /* leave es->s_feature_*compat flags alone */
366 /* es->s_uuid will be set by e2fsck if empty */
367
368 /*
369 * The rest of the superblock fields should be zero, and if not it
370 * means they are likely already in use, so leave them alone. We
371 * can leave it up to e2fsck to clean up any inconsistencies there.
372 */
373}
374
99e6f829
AK		 375int ext4_update_compat_feature(handle_t *handle,
376 struct super_block *sb, __u32 compat)
377{
378 int err = 0;
379 if (!EXT4_HAS_COMPAT_FEATURE(sb, compat)) {
380 err = ext4_journal_get_write_access(handle,
381 EXT4_SB(sb)->s_sbh);
382 if (err)
383 return err;
384 EXT4_SET_COMPAT_FEATURE(sb, compat);
385 sb->s_dirt = 1;
386 handle->h_sync = 1;
387 BUFFER_TRACE(EXT4_SB(sb)->s_sbh,
388 "call ext4_journal_dirty_met adata");
389 err = ext4_journal_dirty_metadata(handle,
390 EXT4_SB(sb)->s_sbh);
391 }
392 return err;
393}
394
395int ext4_update_rocompat_feature(handle_t *handle,
396 struct super_block *sb, __u32 rocompat)
397{
398 int err = 0;
399 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, rocompat)) {
400 err = ext4_journal_get_write_access(handle,
401 EXT4_SB(sb)->s_sbh);
402 if (err)
403 return err;
404 EXT4_SET_RO_COMPAT_FEATURE(sb, rocompat);
405 sb->s_dirt = 1;
406 handle->h_sync = 1;
407 BUFFER_TRACE(EXT4_SB(sb)->s_sbh,
408 "call ext4_journal_dirty_met adata");
409 err = ext4_journal_dirty_metadata(handle,
410 EXT4_SB(sb)->s_sbh);
411 }
412 return err;
413}
414
415int ext4_update_incompat_feature(handle_t *handle,
416 struct super_block *sb, __u32 incompat)
417{
418 int err = 0;
419 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, incompat)) {
420 err = ext4_journal_get_write_access(handle,
421 EXT4_SB(sb)->s_sbh);
422 if (err)
423 return err;
424 EXT4_SET_INCOMPAT_FEATURE(sb, incompat);
425 sb->s_dirt = 1;
426 handle->h_sync = 1;
427 BUFFER_TRACE(EXT4_SB(sb)->s_sbh,
428 "call ext4_journal_dirty_met adata");
429 err = ext4_journal_dirty_metadata(handle,
430 EXT4_SB(sb)->s_sbh);
431 }
432 return err;
433}
434
ac27a0ec
DK		 435/*
436 * Open the external journal device
437 */
617ba13b 438static struct block_device *ext4_blkdev_get(dev_t dev)
ac27a0ec
DK		 439{
440 struct block_device *bdev;
441 char b[BDEVNAME_SIZE];
442
443 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
444 if (IS_ERR(bdev))
445 goto fail;
446 return bdev;
447
448fail:
617ba13b 449 printk(KERN_ERR "EXT4: failed to open journal device %s: %ld\n",
ac27a0ec
DK		 450 __bdevname(dev, b), PTR_ERR(bdev));
451 return NULL;
452}
453
454/*
455 * Release the journal device
456 */
617ba13b 457static int ext4_blkdev_put(struct block_device *bdev)
ac27a0ec
DK		 458{
459 bd_release(bdev);
460 return blkdev_put(bdev);
461}
462
617ba13b 463static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
ac27a0ec
DK		 464{
465 struct block_device *bdev;
466 int ret = -ENODEV;
467
468 bdev = sbi->journal_bdev;
469 if (bdev) {
617ba13b 470 ret = ext4_blkdev_put(bdev);
ac27a0ec
DK		 471 sbi->journal_bdev = NULL;
472 }
473 return ret;
474}
475
476static inline struct inode *orphan_list_entry(struct list_head *l)
477{
617ba13b 478 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
ac27a0ec
DK		 479}
480
617ba13b 481static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
ac27a0ec
DK		 482{
483 struct list_head *l;
484
485 printk(KERN_ERR "sb orphan head is %d\n",
486 le32_to_cpu(sbi->s_es->s_last_orphan));
487
488 printk(KERN_ERR "sb_info orphan list:\n");
489 list_for_each(l, &sbi->s_orphan) {
490 struct inode *inode = orphan_list_entry(l);
491 printk(KERN_ERR " "
492 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
493 inode->i_sb->s_id, inode->i_ino, inode,
494 inode->i_mode, inode->i_nlink,
495 NEXT_ORPHAN(inode));
496 }
497}
498
617ba13b 499static void ext4_put_super (struct super_block * sb)
ac27a0ec 500{
617ba13b
MC		 501 struct ext4_sb_info *sbi = EXT4_SB(sb);
502 struct ext4_super_block *es = sbi->s_es;
ac27a0ec
DK		 503 int i;
504
c9de560d 505 ext4_mb_release(sb);
a86c6181 506 ext4_ext_release(sb);
617ba13b 507 ext4_xattr_put_super(sb);
dab291af 508 jbd2_journal_destroy(sbi->s_journal);
ac27a0ec 509 if (!(sb->s_flags & MS_RDONLY)) {
617ba13b 510 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec
DK		 511 es->s_state = cpu_to_le16(sbi->s_mount_state);
512 BUFFER_TRACE(sbi->s_sbh, "marking dirty");
513 mark_buffer_dirty(sbi->s_sbh);
617ba13b 514 ext4_commit_super(sb, es, 1);
ac27a0ec
DK		 515 }
516
517 for (i = 0; i < sbi->s_gdb_count; i++)
518 brelse(sbi->s_group_desc[i]);
519 kfree(sbi->s_group_desc);
772cb7c8 520 kfree(sbi->s_flex_groups);
ac27a0ec
DK		 521 percpu_counter_destroy(&sbi->s_freeblocks_counter);
522 percpu_counter_destroy(&sbi->s_freeinodes_counter);
523 percpu_counter_destroy(&sbi->s_dirs_counter);
524 brelse(sbi->s_sbh);
525#ifdef CONFIG_QUOTA
526 for (i = 0; i < MAXQUOTAS; i++)
527 kfree(sbi->s_qf_names[i]);
528#endif
529
530 /* Debugging code just in case the in-memory inode orphan list
531 * isn't empty. The on-disk one can be non-empty if we've
532 * detected an error and taken the fs readonly, but the
533 * in-memory list had better be clean by this point. */
534 if (!list_empty(&sbi->s_orphan))
535 dump_orphan_list(sb, sbi);
536 J_ASSERT(list_empty(&sbi->s_orphan));
537
f98393a6 538 invalidate_bdev(sb->s_bdev);
ac27a0ec
DK		 539 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
540 /*
541 * Invalidate the journal device's buffers. We don't want them
542 * floating about in memory - the physical journal device may
543 * hotswapped, and it breaks the `ro-after' testing code.
544 */
545 sync_blockdev(sbi->journal_bdev);
f98393a6 546 invalidate_bdev(sbi->journal_bdev);
617ba13b 547 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 548 }
549 sb->s_fs_info = NULL;
550 kfree(sbi);
551 return;
552}
553
e18b890b 554static struct kmem_cache *ext4_inode_cachep;
ac27a0ec
DK		 555
556/*
557 * Called inside transaction, so use GFP_NOFS
558 */
617ba13b 559static struct inode *ext4_alloc_inode(struct super_block *sb)
ac27a0ec 560{
617ba13b 561 struct ext4_inode_info *ei;
ac27a0ec 562
e6b4f8da 563 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
ac27a0ec
DK		 564 if (!ei)
565 return NULL;
617ba13b
MC		 566#ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
567 ei->i_acl = EXT4_ACL_NOT_CACHED;
568 ei->i_default_acl = EXT4_ACL_NOT_CACHED;
ac27a0ec
DK		 569#endif
570 ei->i_block_alloc_info = NULL;
571 ei->vfs_inode.i_version = 1;
a86c6181 572 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
c9de560d
AT		 573 INIT_LIST_HEAD(&ei->i_prealloc_list);
574 spin_lock_init(&ei->i_prealloc_lock);
ac27a0ec
DK		 575 return &ei->vfs_inode;
576}
577
617ba13b 578static void ext4_destroy_inode(struct inode *inode)
ac27a0ec 579{
9f7dd93d
VA		 580 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
581 printk("EXT4 Inode %p: orphan list check failed!\n",
582 EXT4_I(inode));
583 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
584 EXT4_I(inode), sizeof(struct ext4_inode_info),
585 true);
586 dump_stack();
587 }
617ba13b 588 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
ac27a0ec
DK		 589}
590
4ba9b9d0 591static void init_once(struct kmem_cache *cachep, void *foo)
ac27a0ec 592{
617ba13b 593 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
ac27a0ec 594
a35afb83 595 INIT_LIST_HEAD(&ei->i_orphan);
617ba13b 596#ifdef CONFIG_EXT4DEV_FS_XATTR
a35afb83 597 init_rwsem(&ei->xattr_sem);
ac27a0ec 598#endif
0e855ac8 599 init_rwsem(&ei->i_data_sem);
a35afb83 600 inode_init_once(&ei->vfs_inode);
ac27a0ec
DK		 601}
602
603static int init_inodecache(void)
604{
617ba13b
MC		 605 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
606 sizeof(struct ext4_inode_info),
ac27a0ec
DK		 607 0, (SLAB_RECLAIM_ACCOUNT|
608 SLAB_MEM_SPREAD),
20c2df83 609 init_once);
617ba13b 610 if (ext4_inode_cachep == NULL)
ac27a0ec
DK		 611 return -ENOMEM;
612 return 0;
613}
614
615static void destroy_inodecache(void)
616{
617ba13b 617 kmem_cache_destroy(ext4_inode_cachep);
ac27a0ec
DK		 618}
619
617ba13b 620static void ext4_clear_inode(struct inode *inode)
ac27a0ec 621{
617ba13b
MC		 622 struct ext4_block_alloc_info *rsv = EXT4_I(inode)->i_block_alloc_info;
623#ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
624 if (EXT4_I(inode)->i_acl &&
625 EXT4_I(inode)->i_acl != EXT4_ACL_NOT_CACHED) {
626 posix_acl_release(EXT4_I(inode)->i_acl);
627 EXT4_I(inode)->i_acl = EXT4_ACL_NOT_CACHED;
628 }
629 if (EXT4_I(inode)->i_default_acl &&
630 EXT4_I(inode)->i_default_acl != EXT4_ACL_NOT_CACHED) {
631 posix_acl_release(EXT4_I(inode)->i_default_acl);
632 EXT4_I(inode)->i_default_acl = EXT4_ACL_NOT_CACHED;
ac27a0ec
DK		 633 }
634#endif
617ba13b
MC		 635 ext4_discard_reservation(inode);
636 EXT4_I(inode)->i_block_alloc_info = NULL;
ac27a0ec
DK		 637 if (unlikely(rsv))
638 kfree(rsv);
639}
640
617ba13b 641static inline void ext4_show_quota_options(struct seq_file *seq, struct super_block *sb)
ac27a0ec
DK		 642{
643#if defined(CONFIG_QUOTA)
617ba13b 644 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 645
646 if (sbi->s_jquota_fmt)
647 seq_printf(seq, ",jqfmt=%s",
648 (sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold": "vfsv0");
649
650 if (sbi->s_qf_names[USRQUOTA])
651 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
652
653 if (sbi->s_qf_names[GRPQUOTA])
654 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
655
617ba13b 656 if (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA)
ac27a0ec
DK		 657 seq_puts(seq, ",usrquota");
658
617ba13b 659 if (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)
ac27a0ec
DK		 660 seq_puts(seq, ",grpquota");
661#endif
662}
663
d9c9bef1
MS		 664/*
665 * Show an option if
666 * - it's set to a non-default value OR
667 * - if the per-sb default is different from the global default
668 */
617ba13b 669static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
ac27a0ec 670{
aa22df2c
AK		 671 int def_errors;
672 unsigned long def_mount_opts;
ac27a0ec 673 struct super_block *sb = vfs->mnt_sb;
d9c9bef1
MS		 674 struct ext4_sb_info *sbi = EXT4_SB(sb);
675 struct ext4_super_block *es = sbi->s_es;
d9c9bef1
MS		 676
677 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
aa22df2c 678 def_errors = le16_to_cpu(es->s_errors);
d9c9bef1
MS		 679
680 if (sbi->s_sb_block != 1)
681 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
682 if (test_opt(sb, MINIX_DF))
683 seq_puts(seq, ",minixdf");
aa22df2c 684 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
d9c9bef1
MS		 685 seq_puts(seq, ",grpid");
686 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
687 seq_puts(seq, ",nogrpid");
688 if (sbi->s_resuid != EXT4_DEF_RESUID ||
689 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
690 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
691 }
692 if (sbi->s_resgid != EXT4_DEF_RESGID ||
693 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
694 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
695 }
bb4f397a 696 if (test_opt(sb, ERRORS_RO)) {
d9c9bef1 697 if (def_errors == EXT4_ERRORS_PANIC ||
bb4f397a
AK		 698 def_errors == EXT4_ERRORS_CONTINUE) {
699 seq_puts(seq, ",errors=remount-ro");
d9c9bef1
MS		 700 }
701 }
aa22df2c 702 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
bb4f397a 703 seq_puts(seq, ",errors=continue");
aa22df2c 704 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
d9c9bef1 705 seq_puts(seq, ",errors=panic");
aa22df2c 706 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
d9c9bef1 707 seq_puts(seq, ",nouid32");
aa22df2c 708 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
d9c9bef1
MS		 709 seq_puts(seq, ",debug");
710 if (test_opt(sb, OLDALLOC))
711 seq_puts(seq, ",oldalloc");
07620f69 712#ifdef CONFIG_EXT4DEV_FS_XATTR
aa22df2c
AK		 713 if (test_opt(sb, XATTR_USER) &&
714 !(def_mount_opts & EXT4_DEFM_XATTR_USER))
d9c9bef1
MS		 715 seq_puts(seq, ",user_xattr");
716 if (!test_opt(sb, XATTR_USER) &&
717 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
718 seq_puts(seq, ",nouser_xattr");
719 }
720#endif
07620f69 721#ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
aa22df2c 722 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
d9c9bef1
MS		 723 seq_puts(seq, ",acl");
724 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
725 seq_puts(seq, ",noacl");
726#endif
727 if (!test_opt(sb, RESERVATION))
728 seq_puts(seq, ",noreservation");
729 if (sbi->s_commit_interval) {
730 seq_printf(seq, ",commit=%u",
731 (unsigned) (sbi->s_commit_interval / HZ));
732 }
571640ca
ES		 733 /*
734 * We're changing the default of barrier mount option, so
735 * let's always display its mount state so it's clear what its
736 * status is.
737 */
738 seq_puts(seq, ",barrier=");
739 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
cd0b6a39
TT		 740 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
741 seq_puts(seq, ",journal_async_commit");
d9c9bef1
MS		 742 if (test_opt(sb, NOBH))
743 seq_puts(seq, ",nobh");
744 if (!test_opt(sb, EXTENTS))
745 seq_puts(seq, ",noextents");
3dbd0ede
AK		 746 if (!test_opt(sb, MBALLOC))
747 seq_puts(seq, ",nomballoc");
25ec56b5
JNC		 748 if (test_opt(sb, I_VERSION))
749 seq_puts(seq, ",i_version");
ac27a0ec 750
cb45bbe4
MS		 751 if (sbi->s_stripe)
752 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
aa22df2c
AK		 753 /*
754 * journal mode get enabled in different ways
755 * So just print the value even if we didn't specify it
756 */
617ba13b 757 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
ac27a0ec 758 seq_puts(seq, ",data=journal");
617ba13b 759 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
ac27a0ec 760 seq_puts(seq, ",data=ordered");
617ba13b 761 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
ac27a0ec
DK		 762 seq_puts(seq, ",data=writeback");
763
617ba13b 764 ext4_show_quota_options(seq, sb);
ac27a0ec
DK		 765 return 0;
766}
767
768
1b961ac0
CH		 769static struct inode *ext4_nfs_get_inode(struct super_block *sb,
770 u64 ino, u32 generation)
ac27a0ec 771{
ac27a0ec 772 struct inode *inode;
ac27a0ec 773
617ba13b 774 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
ac27a0ec 775 return ERR_PTR(-ESTALE);
617ba13b 776 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
ac27a0ec
DK		 777 return ERR_PTR(-ESTALE);
778
779 /* iget isn't really right if the inode is currently unallocated!!
780 *
617ba13b 781 * ext4_read_inode will return a bad_inode if the inode had been
ac27a0ec
DK		 782 * deleted, so we should be safe.
783 *
784 * Currently we don't know the generation for parent directory, so
785 * a generation of 0 means "accept any"
786 */
1d1fe1ee
DH		 787 inode = ext4_iget(sb, ino);
788 if (IS_ERR(inode))
789 return ERR_CAST(inode);
790 if (generation && inode->i_generation != generation) {
ac27a0ec
DK		 791 iput(inode);
792 return ERR_PTR(-ESTALE);
793 }
1b961ac0
CH		 794
795 return inode;
796}
797
798static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
799 int fh_len, int fh_type)
800{
801 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
802 ext4_nfs_get_inode);
803}
804
805static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
806 int fh_len, int fh_type)
807{
808 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
809 ext4_nfs_get_inode);
ac27a0ec
DK		 810}
811
812#ifdef CONFIG_QUOTA
813#define QTYPE2NAME(t) ((t)==USRQUOTA?"user":"group")
814#define QTYPE2MOPT(on, t) ((t)==USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
815
617ba13b
MC		 816static int ext4_dquot_initialize(struct inode *inode, int type);
817static int ext4_dquot_drop(struct inode *inode);
818static int ext4_write_dquot(struct dquot *dquot);
819static int ext4_acquire_dquot(struct dquot *dquot);
820static int ext4_release_dquot(struct dquot *dquot);
821static int ext4_mark_dquot_dirty(struct dquot *dquot);
822static int ext4_write_info(struct super_block *sb, int type);
6f28e087
JK		 823static int ext4_quota_on(struct super_block *sb, int type, int format_id,
824 char *path, int remount);
617ba13b
MC		 825static int ext4_quota_on_mount(struct super_block *sb, int type);
826static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 827 size_t len, loff_t off);
617ba13b 828static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 829 const char *data, size_t len, loff_t off);
830
617ba13b
MC		 831static struct dquot_operations ext4_quota_operations = {
832 .initialize = ext4_dquot_initialize,
833 .drop = ext4_dquot_drop,
ac27a0ec
DK		 834 .alloc_space = dquot_alloc_space,
835 .alloc_inode = dquot_alloc_inode,
836 .free_space = dquot_free_space,
837 .free_inode = dquot_free_inode,
838 .transfer = dquot_transfer,
617ba13b
MC		 839 .write_dquot = ext4_write_dquot,
840 .acquire_dquot = ext4_acquire_dquot,
841 .release_dquot = ext4_release_dquot,
842 .mark_dirty = ext4_mark_dquot_dirty,
843 .write_info = ext4_write_info
ac27a0ec
DK		 844};
845
617ba13b
MC		 846static struct quotactl_ops ext4_qctl_operations = {
847 .quota_on = ext4_quota_on,
ac27a0ec
DK		 848 .quota_off = vfs_quota_off,
849 .quota_sync = vfs_quota_sync,
850 .get_info = vfs_get_dqinfo,
851 .set_info = vfs_set_dqinfo,
852 .get_dqblk = vfs_get_dqblk,
853 .set_dqblk = vfs_set_dqblk
854};
855#endif
856
ee9b6d61 857static const struct super_operations ext4_sops = {
617ba13b
MC		 858 .alloc_inode = ext4_alloc_inode,
859 .destroy_inode = ext4_destroy_inode,
617ba13b
MC		 860 .write_inode = ext4_write_inode,
861 .dirty_inode = ext4_dirty_inode,
862 .delete_inode = ext4_delete_inode,
863 .put_super = ext4_put_super,
864 .write_super = ext4_write_super,
865 .sync_fs = ext4_sync_fs,
866 .write_super_lockfs = ext4_write_super_lockfs,
867 .unlockfs = ext4_unlockfs,
868 .statfs = ext4_statfs,
869 .remount_fs = ext4_remount,
870 .clear_inode = ext4_clear_inode,
871 .show_options = ext4_show_options,
ac27a0ec 872#ifdef CONFIG_QUOTA
617ba13b
MC		 873 .quota_read = ext4_quota_read,
874 .quota_write = ext4_quota_write,
ac27a0ec
DK		 875#endif
876};
877
39655164 878static const struct export_operations ext4_export_ops = {
1b961ac0
CH		 879 .fh_to_dentry = ext4_fh_to_dentry,
880 .fh_to_parent = ext4_fh_to_parent,
617ba13b 881 .get_parent = ext4_get_parent,
ac27a0ec
DK		 882};
883
884enum {
885 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
886 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
887 Opt_nouid32, Opt_nocheck, Opt_debug, Opt_oldalloc, Opt_orlov,
888 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
889 Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh,
890 Opt_commit, Opt_journal_update, Opt_journal_inum, Opt_journal_dev,
818d276c 891 Opt_journal_checksum, Opt_journal_async_commit,
ac27a0ec
DK		 892 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
893 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
894 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
895 Opt_ignore, Opt_barrier, Opt_err, Opt_resize, Opt_usrquota,
25ec56b5 896 Opt_grpquota, Opt_extents, Opt_noextents, Opt_i_version,
c9de560d 897 Opt_mballoc, Opt_nomballoc, Opt_stripe,
ac27a0ec
DK		 898};
899
900static match_table_t tokens = {
901 {Opt_bsd_df, "bsddf"},
902 {Opt_minix_df, "minixdf"},
903 {Opt_grpid, "grpid"},
904 {Opt_grpid, "bsdgroups"},
905 {Opt_nogrpid, "nogrpid"},
906 {Opt_nogrpid, "sysvgroups"},
907 {Opt_resgid, "resgid=%u"},
908 {Opt_resuid, "resuid=%u"},
909 {Opt_sb, "sb=%u"},
910 {Opt_err_cont, "errors=continue"},
911 {Opt_err_panic, "errors=panic"},
912 {Opt_err_ro, "errors=remount-ro"},
913 {Opt_nouid32, "nouid32"},
914 {Opt_nocheck, "nocheck"},
915 {Opt_nocheck, "check=none"},
916 {Opt_debug, "debug"},
917 {Opt_oldalloc, "oldalloc"},
918 {Opt_orlov, "orlov"},
919 {Opt_user_xattr, "user_xattr"},
920 {Opt_nouser_xattr, "nouser_xattr"},
921 {Opt_acl, "acl"},
922 {Opt_noacl, "noacl"},
923 {Opt_reservation, "reservation"},
924 {Opt_noreservation, "noreservation"},
925 {Opt_noload, "noload"},
926 {Opt_nobh, "nobh"},
927 {Opt_bh, "bh"},
928 {Opt_commit, "commit=%u"},
929 {Opt_journal_update, "journal=update"},
930 {Opt_journal_inum, "journal=%u"},
931 {Opt_journal_dev, "journal_dev=%u"},
818d276c
GS		 932 {Opt_journal_checksum, "journal_checksum"},
933 {Opt_journal_async_commit, "journal_async_commit"},
ac27a0ec
DK		 934 {Opt_abort, "abort"},
935 {Opt_data_journal, "data=journal"},
936 {Opt_data_ordered, "data=ordered"},
937 {Opt_data_writeback, "data=writeback"},
938 {Opt_offusrjquota, "usrjquota="},
939 {Opt_usrjquota, "usrjquota=%s"},
940 {Opt_offgrpjquota, "grpjquota="},
941 {Opt_grpjquota, "grpjquota=%s"},
942 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
943 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
944 {Opt_grpquota, "grpquota"},
945 {Opt_noquota, "noquota"},
946 {Opt_quota, "quota"},
947 {Opt_usrquota, "usrquota"},
948 {Opt_barrier, "barrier=%u"},
a86c6181 949 {Opt_extents, "extents"},
1e2462f9 950 {Opt_noextents, "noextents"},
25ec56b5 951 {Opt_i_version, "i_version"},
c9de560d
AT		 952 {Opt_mballoc, "mballoc"},
953 {Opt_nomballoc, "nomballoc"},
954 {Opt_stripe, "stripe=%u"},
ac27a0ec 955 {Opt_resize, "resize"},
f3f12faa 956 {Opt_err, NULL},
ac27a0ec
DK		 957};
958
617ba13b 959static ext4_fsblk_t get_sb_block(void **data)
ac27a0ec 960{
617ba13b 961 ext4_fsblk_t sb_block;
ac27a0ec
DK		 962 char *options = (char *) *data;
963
964 if (!options || strncmp(options, "sb=", 3) != 0)
965 return 1; /* Default location */
966 options += 3;
617ba13b 967 /*todo: use simple_strtoll with >32bit ext4 */
ac27a0ec
DK		 968 sb_block = simple_strtoul(options, &options, 0);
969 if (*options && *options != ',') {
617ba13b 970 printk("EXT4-fs: Invalid sb specification: %s\n",
ac27a0ec
DK		 971 (char *) *data);
972 return 1;
973 }
974 if (*options == ',')
975 options++;
976 *data = (void *) options;
977 return sb_block;
978}
979
980static int parse_options (char *options, struct super_block *sb,
981 unsigned int *inum, unsigned long *journal_devnum,
617ba13b 982 ext4_fsblk_t *n_blocks_count, int is_remount)
ac27a0ec 983{
617ba13b 984 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 985 char * p;
986 substring_t args[MAX_OPT_ARGS];
987 int data_opt = 0;
988 int option;
989#ifdef CONFIG_QUOTA
dfc5d03f 990 int qtype, qfmt;
ac27a0ec
DK		 991 char *qname;
992#endif
993
994 if (!options)
995 return 1;
996
997 while ((p = strsep (&options, ",")) != NULL) {
998 int token;
999 if (!*p)
1000 continue;
1001
1002 token = match_token(p, tokens, args);
1003 switch (token) {
1004 case Opt_bsd_df:
1005 clear_opt (sbi->s_mount_opt, MINIX_DF);
1006 break;
1007 case Opt_minix_df:
1008 set_opt (sbi->s_mount_opt, MINIX_DF);
1009 break;
1010 case Opt_grpid:
1011 set_opt (sbi->s_mount_opt, GRPID);
1012 break;
1013 case Opt_nogrpid:
1014 clear_opt (sbi->s_mount_opt, GRPID);
1015 break;
1016 case Opt_resuid:
1017 if (match_int(&args[0], &option))
1018 return 0;
1019 sbi->s_resuid = option;
1020 break;
1021 case Opt_resgid:
1022 if (match_int(&args[0], &option))
1023 return 0;
1024 sbi->s_resgid = option;
1025 break;
1026 case Opt_sb:
1027 /* handled by get_sb_block() instead of here */
1028 /* *sb_block = match_int(&args[0]); */
1029 break;
1030 case Opt_err_panic:
1031 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
1032 clear_opt (sbi->s_mount_opt, ERRORS_RO);
1033 set_opt (sbi->s_mount_opt, ERRORS_PANIC);
1034 break;
1035 case Opt_err_ro:
1036 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
1037 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
1038 set_opt (sbi->s_mount_opt, ERRORS_RO);
1039 break;
1040 case Opt_err_cont:
1041 clear_opt (sbi->s_mount_opt, ERRORS_RO);
1042 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
1043 set_opt (sbi->s_mount_opt, ERRORS_CONT);
1044 break;
1045 case Opt_nouid32:
1046 set_opt (sbi->s_mount_opt, NO_UID32);
1047 break;
1048 case Opt_nocheck:
1049 clear_opt (sbi->s_mount_opt, CHECK);
1050 break;
1051 case Opt_debug:
1052 set_opt (sbi->s_mount_opt, DEBUG);
1053 break;
1054 case Opt_oldalloc:
1055 set_opt (sbi->s_mount_opt, OLDALLOC);
1056 break;
1057 case Opt_orlov:
1058 clear_opt (sbi->s_mount_opt, OLDALLOC);
1059 break;
617ba13b 1060#ifdef CONFIG_EXT4DEV_FS_XATTR
ac27a0ec
DK		 1061 case Opt_user_xattr:
1062 set_opt (sbi->s_mount_opt, XATTR_USER);
1063 break;
1064 case Opt_nouser_xattr:
1065 clear_opt (sbi->s_mount_opt, XATTR_USER);
1066 break;
1067#else
1068 case Opt_user_xattr:
1069 case Opt_nouser_xattr:
617ba13b 1070 printk("EXT4 (no)user_xattr options not supported\n");
ac27a0ec
DK		 1071 break;
1072#endif
617ba13b 1073#ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
ac27a0ec
DK		 1074 case Opt_acl:
1075 set_opt(sbi->s_mount_opt, POSIX_ACL);
1076 break;
1077 case Opt_noacl:
1078 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1079 break;
1080#else
1081 case Opt_acl:
1082 case Opt_noacl:
617ba13b 1083 printk("EXT4 (no)acl options not supported\n");
ac27a0ec
DK		 1084 break;
1085#endif
1086 case Opt_reservation:
1087 set_opt(sbi->s_mount_opt, RESERVATION);
1088 break;
1089 case Opt_noreservation:
1090 clear_opt(sbi->s_mount_opt, RESERVATION);
1091 break;
1092 case Opt_journal_update:
1093 /* @@@ FIXME */
1094 /* Eventually we will want to be able to create
1095 a journal file here. For now, only allow the
1096 user to specify an existing inode to be the
1097 journal file. */
1098 if (is_remount) {
617ba13b 1099 printk(KERN_ERR "EXT4-fs: cannot specify "
ac27a0ec
DK		 1100 "journal on remount\n");
1101 return 0;
1102 }
1103 set_opt (sbi->s_mount_opt, UPDATE_JOURNAL);
1104 break;
1105 case Opt_journal_inum:
1106 if (is_remount) {
617ba13b 1107 printk(KERN_ERR "EXT4-fs: cannot specify "
ac27a0ec
DK		 1108 "journal on remount\n");
1109 return 0;
1110 }
1111 if (match_int(&args[0], &option))
1112 return 0;
1113 *inum = option;
1114 break;
1115 case Opt_journal_dev:
1116 if (is_remount) {
617ba13b 1117 printk(KERN_ERR "EXT4-fs: cannot specify "
ac27a0ec
DK		 1118 "journal on remount\n");
1119 return 0;
1120 }
1121 if (match_int(&args[0], &option))
1122 return 0;
1123 *journal_devnum = option;
1124 break;
818d276c
GS		 1125 case Opt_journal_checksum:
1126 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1127 break;
1128 case Opt_journal_async_commit:
1129 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
1130 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1131 break;
ac27a0ec
DK		 1132 case Opt_noload:
1133 set_opt (sbi->s_mount_opt, NOLOAD);
1134 break;
1135 case Opt_commit:
1136 if (match_int(&args[0], &option))
1137 return 0;
1138 if (option < 0)
1139 return 0;
1140 if (option == 0)
cd02ff0b 1141 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
ac27a0ec
DK		 1142 sbi->s_commit_interval = HZ * option;
1143 break;
1144 case Opt_data_journal:
617ba13b 1145 data_opt = EXT4_MOUNT_JOURNAL_DATA;
ac27a0ec
DK		 1146 goto datacheck;
1147 case Opt_data_ordered:
617ba13b 1148 data_opt = EXT4_MOUNT_ORDERED_DATA;
ac27a0ec
DK		 1149 goto datacheck;
1150 case Opt_data_writeback:
617ba13b 1151 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
ac27a0ec
DK		 1152 datacheck:
1153 if (is_remount) {
617ba13b 1154 if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
ac27a0ec
DK		 1155 != data_opt) {
1156 printk(KERN_ERR
617ba13b 1157 "EXT4-fs: cannot change data "
ac27a0ec
DK		 1158 "mode on remount\n");
1159 return 0;
1160 }
1161 } else {
617ba13b 1162 sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
ac27a0ec
DK		 1163 sbi->s_mount_opt |= data_opt;
1164 }
1165 break;
1166#ifdef CONFIG_QUOTA
1167 case Opt_usrjquota:
1168 qtype = USRQUOTA;
1169 goto set_qf_name;
1170 case Opt_grpjquota:
1171 qtype = GRPQUOTA;
1172set_qf_name:
dfc5d03f
JK		 1173 if ((sb_any_quota_enabled(sb) ||
1174 sb_any_quota_suspended(sb)) &&
1175 !sbi->s_qf_names[qtype]) {
ac27a0ec 1176 printk(KERN_ERR
2c8be6b2 1177 "EXT4-fs: Cannot change journaled "
ac27a0ec
DK		 1178 "quota options when quota turned on.\n");
1179 return 0;
1180 }
1181 qname = match_strdup(&args[0]);
1182 if (!qname) {
1183 printk(KERN_ERR
617ba13b 1184 "EXT4-fs: not enough memory for "
ac27a0ec
DK		 1185 "storing quotafile name.\n");
1186 return 0;
1187 }
1188 if (sbi->s_qf_names[qtype] &&
1189 strcmp(sbi->s_qf_names[qtype], qname)) {
1190 printk(KERN_ERR
617ba13b 1191 "EXT4-fs: %s quota file already "
ac27a0ec
DK		 1192 "specified.\n", QTYPE2NAME(qtype));
1193 kfree(qname);
1194 return 0;
1195 }
1196 sbi->s_qf_names[qtype] = qname;
1197 if (strchr(sbi->s_qf_names[qtype], '/')) {
1198 printk(KERN_ERR
617ba13b 1199 "EXT4-fs: quotafile must be on "
ac27a0ec
DK		 1200 "filesystem root.\n");
1201 kfree(sbi->s_qf_names[qtype]);
1202 sbi->s_qf_names[qtype] = NULL;
1203 return 0;
1204 }
1205 set_opt(sbi->s_mount_opt, QUOTA);
1206 break;
1207 case Opt_offusrjquota:
1208 qtype = USRQUOTA;
1209 goto clear_qf_name;
1210 case Opt_offgrpjquota:
1211 qtype = GRPQUOTA;
1212clear_qf_name:
dfc5d03f
JK		 1213 if ((sb_any_quota_enabled(sb) ||
1214 sb_any_quota_suspended(sb)) &&
1215 sbi->s_qf_names[qtype]) {
617ba13b 1216 printk(KERN_ERR "EXT4-fs: Cannot change "
2c8be6b2 1217 "journaled quota options when "
ac27a0ec
DK		 1218 "quota turned on.\n");
1219 return 0;
1220 }
1221 /*
1222 * The space will be released later when all options
1223 * are confirmed to be correct
1224 */
1225 sbi->s_qf_names[qtype] = NULL;
1226 break;
1227 case Opt_jqfmt_vfsold:
dfc5d03f
JK		 1228 qfmt = QFMT_VFS_OLD;
1229 goto set_qf_format;
ac27a0ec 1230 case Opt_jqfmt_vfsv0:
dfc5d03f
JK		 1231 qfmt = QFMT_VFS_V0;
1232set_qf_format:
1233 if ((sb_any_quota_enabled(sb) ||
1234 sb_any_quota_suspended(sb)) &&
1235 sbi->s_jquota_fmt != qfmt) {
1236 printk(KERN_ERR "EXT4-fs: Cannot change "
1237 "journaled quota options when "
1238 "quota turned on.\n");
1239 return 0;
1240 }
1241 sbi->s_jquota_fmt = qfmt;
ac27a0ec
DK		 1242 break;
1243 case Opt_quota:
1244 case Opt_usrquota:
1245 set_opt(sbi->s_mount_opt, QUOTA);
1246 set_opt(sbi->s_mount_opt, USRQUOTA);
1247 break;
1248 case Opt_grpquota:
1249 set_opt(sbi->s_mount_opt, QUOTA);
1250 set_opt(sbi->s_mount_opt, GRPQUOTA);
1251 break;
1252 case Opt_noquota:
1253 if (sb_any_quota_enabled(sb)) {
617ba13b 1254 printk(KERN_ERR "EXT4-fs: Cannot change quota "
ac27a0ec
DK		 1255 "options when quota turned on.\n");
1256 return 0;
1257 }
1258 clear_opt(sbi->s_mount_opt, QUOTA);
1259 clear_opt(sbi->s_mount_opt, USRQUOTA);
1260 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1261 break;
1262#else
1263 case Opt_quota:
1264 case Opt_usrquota:
1265 case Opt_grpquota:
cd59e7b9
JK		 1266 printk(KERN_ERR
1267 "EXT4-fs: quota options not supported.\n");
1268 break;
ac27a0ec
DK		 1269 case Opt_usrjquota:
1270 case Opt_grpjquota:
1271 case Opt_offusrjquota:
1272 case Opt_offgrpjquota:
1273 case Opt_jqfmt_vfsold:
1274 case Opt_jqfmt_vfsv0:
1275 printk(KERN_ERR
cd59e7b9 1276 "EXT4-fs: journaled quota options not "
ac27a0ec
DK		 1277 "supported.\n");
1278 break;
1279 case Opt_noquota:
1280 break;
1281#endif
1282 case Opt_abort:
1283 set_opt(sbi->s_mount_opt, ABORT);
1284 break;
1285 case Opt_barrier:
1286 if (match_int(&args[0], &option))
1287 return 0;
1288 if (option)
1289 set_opt(sbi->s_mount_opt, BARRIER);
1290 else
1291 clear_opt(sbi->s_mount_opt, BARRIER);
1292 break;
1293 case Opt_ignore:
1294 break;
1295 case Opt_resize:
1296 if (!is_remount) {
617ba13b 1297 printk("EXT4-fs: resize option only available "
ac27a0ec
DK		 1298 "for remount\n");
1299 return 0;
1300 }
1301 if (match_int(&args[0], &option) != 0)
1302 return 0;
1303 *n_blocks_count = option;
1304 break;
1305 case Opt_nobh:
1306 set_opt(sbi->s_mount_opt, NOBH);
1307 break;
1308 case Opt_bh:
1309 clear_opt(sbi->s_mount_opt, NOBH);
1310 break;
a86c6181
AT		 1311 case Opt_extents:
1312 set_opt (sbi->s_mount_opt, EXTENTS);
1313 break;
1e2462f9
MC		 1314 case Opt_noextents:
1315 clear_opt (sbi->s_mount_opt, EXTENTS);
1316 break;
25ec56b5
JNC		 1317 case Opt_i_version:
1318 set_opt(sbi->s_mount_opt, I_VERSION);
1319 sb->s_flags |= MS_I_VERSION;
1320 break;
c9de560d
AT		 1321 case Opt_mballoc:
1322 set_opt(sbi->s_mount_opt, MBALLOC);
1323 break;
1324 case Opt_nomballoc:
1325 clear_opt(sbi->s_mount_opt, MBALLOC);
1326 break;
1327 case Opt_stripe:
1328 if (match_int(&args[0], &option))
1329 return 0;
1330 if (option < 0)
1331 return 0;
1332 sbi->s_stripe = option;
1333 break;
ac27a0ec
DK		 1334 default:
1335 printk (KERN_ERR
617ba13b 1336 "EXT4-fs: Unrecognized mount option \"%s\" "
ac27a0ec
DK		 1337 "or missing value\n", p);
1338 return 0;
1339 }
1340 }
1341#ifdef CONFIG_QUOTA
1342 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
617ba13b 1343 if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
ac27a0ec
DK		 1344 sbi->s_qf_names[USRQUOTA])
1345 clear_opt(sbi->s_mount_opt, USRQUOTA);
1346
617ba13b 1347 if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
ac27a0ec
DK		 1348 sbi->s_qf_names[GRPQUOTA])
1349 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1350
1351 if ((sbi->s_qf_names[USRQUOTA] &&
617ba13b 1352 (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
ac27a0ec 1353 (sbi->s_qf_names[GRPQUOTA] &&
617ba13b
MC		 1354 (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
1355 printk(KERN_ERR "EXT4-fs: old and new quota "
ac27a0ec
DK		 1356 "format mixing.\n");
1357 return 0;
1358 }
1359
1360 if (!sbi->s_jquota_fmt) {
2c8be6b2 1361 printk(KERN_ERR "EXT4-fs: journaled quota format "
ac27a0ec
DK		 1362 "not specified.\n");
1363 return 0;
1364 }
1365 } else {
1366 if (sbi->s_jquota_fmt) {
2c8be6b2
JK		 1367 printk(KERN_ERR "EXT4-fs: journaled quota format "
1368 "specified with no journaling "
ac27a0ec
DK		 1369 "enabled.\n");
1370 return 0;
1371 }
1372 }
1373#endif
1374 return 1;
1375}
1376
617ba13b 1377static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 1378 int read_only)
1379{
617ba13b 1380 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 1381 int res = 0;
1382
617ba13b
MC		 1383 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1384 printk (KERN_ERR "EXT4-fs warning: revision level too high, "
ac27a0ec
DK		 1385 "forcing read-only mode\n");
1386 res = MS_RDONLY;
1387 }
1388 if (read_only)
1389 return res;
617ba13b
MC		 1390 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1391 printk (KERN_WARNING "EXT4-fs warning: mounting unchecked fs, "
ac27a0ec 1392 "running e2fsck is recommended\n");
617ba13b 1393 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
ac27a0ec 1394 printk (KERN_WARNING
617ba13b 1395 "EXT4-fs warning: mounting fs with errors, "
ac27a0ec
DK		 1396 "running e2fsck is recommended\n");
1397 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1398 le16_to_cpu(es->s_mnt_count) >=
1399 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1400 printk (KERN_WARNING
617ba13b 1401 "EXT4-fs warning: maximal mount count reached, "
ac27a0ec
DK		 1402 "running e2fsck is recommended\n");
1403 else if (le32_to_cpu(es->s_checkinterval) &&
1404 (le32_to_cpu(es->s_lastcheck) +
1405 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1406 printk (KERN_WARNING
617ba13b 1407 "EXT4-fs warning: checktime reached, "
ac27a0ec
DK		 1408 "running e2fsck is recommended\n");
1409#if 0
1410 /* @@@ We _will_ want to clear the valid bit if we find
63f57933
AM		 1411 * inconsistencies, to force a fsck at reboot. But for
1412 * a plain journaled filesystem we can keep it set as
1413 * valid forever! :)
1414 */
216c34b2 1415 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec
DK		 1416#endif
1417 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 1418 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 1419 le16_add_cpu(&es->s_mnt_count, 1);
ac27a0ec 1420 es->s_mtime = cpu_to_le32(get_seconds());
617ba13b
MC		 1421 ext4_update_dynamic_rev(sb);
1422 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 1423
617ba13b 1424 ext4_commit_super(sb, es, 1);
ac27a0ec 1425 if (test_opt(sb, DEBUG))
617ba13b 1426 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%lu, "
ac27a0ec
DK		 1427 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
1428 sb->s_blocksize,
1429 sbi->s_groups_count,
617ba13b
MC		 1430 EXT4_BLOCKS_PER_GROUP(sb),
1431 EXT4_INODES_PER_GROUP(sb),
ac27a0ec
DK		 1432 sbi->s_mount_opt);
1433
617ba13b
MC		 1434 printk(KERN_INFO "EXT4 FS on %s, ", sb->s_id);
1435 if (EXT4_SB(sb)->s_journal->j_inode == NULL) {
ac27a0ec
DK		 1436 char b[BDEVNAME_SIZE];
1437
1438 printk("external journal on %s\n",
617ba13b 1439 bdevname(EXT4_SB(sb)->s_journal->j_dev, b));
ac27a0ec
DK		 1440 } else {
1441 printk("internal journal\n");
1442 }
1443 return res;
1444}
1445
772cb7c8
JS		 1446static int ext4_fill_flex_info(struct super_block *sb)
1447{
1448 struct ext4_sb_info *sbi = EXT4_SB(sb);
1449 struct ext4_group_desc *gdp = NULL;
1450 struct buffer_head *bh;
1451 ext4_group_t flex_group_count;
1452 ext4_group_t flex_group;
1453 int groups_per_flex = 0;
1454 __u64 block_bitmap = 0;
1455 int i;
1456
1457 if (!sbi->s_es->s_log_groups_per_flex) {
1458 sbi->s_log_groups_per_flex = 0;
1459 return 1;
1460 }
1461
1462 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1463 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1464
1465 flex_group_count = (sbi->s_groups_count + groups_per_flex - 1) /
1466 groups_per_flex;
1467 sbi->s_flex_groups = kmalloc(flex_group_count *
1468 sizeof(struct flex_groups), GFP_KERNEL);
1469 if (sbi->s_flex_groups == NULL) {
1470 printk(KERN_ERR "EXT4-fs: not enough memory\n");
1471 goto failed;
1472 }
1473 memset(sbi->s_flex_groups, 0, flex_group_count *
1474 sizeof(struct flex_groups));
1475
1476 gdp = ext4_get_group_desc(sb, 1, &bh);
1477 block_bitmap = ext4_block_bitmap(sb, gdp) - 1;
1478
1479 for (i = 0; i < sbi->s_groups_count; i++) {
1480 gdp = ext4_get_group_desc(sb, i, &bh);
1481
1482 flex_group = ext4_flex_group(sbi, i);
1483 sbi->s_flex_groups[flex_group].free_inodes +=
1484 le16_to_cpu(gdp->bg_free_inodes_count);
1485 sbi->s_flex_groups[flex_group].free_blocks +=
1486 le16_to_cpu(gdp->bg_free_blocks_count);
1487 }
1488
1489 return 1;
1490failed:
1491 return 0;
1492}
1493
717d50e4
AD		 1494__le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1495 struct ext4_group_desc *gdp)
1496{
1497 __u16 crc = 0;
1498
1499 if (sbi->s_es->s_feature_ro_compat &
1500 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1501 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1502 __le32 le_group = cpu_to_le32(block_group);
1503
1504 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1505 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1506 crc = crc16(crc, (__u8 *)gdp, offset);
1507 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1508 /* for checksum of struct ext4_group_desc do the rest...*/
1509 if ((sbi->s_es->s_feature_incompat &
1510 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1511 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1512 crc = crc16(crc, (__u8 *)gdp + offset,
1513 le16_to_cpu(sbi->s_es->s_desc_size) -
1514 offset);
1515 }
1516
1517 return cpu_to_le16(crc);
1518}
1519
1520int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1521 struct ext4_group_desc *gdp)
1522{
1523 if ((sbi->s_es->s_feature_ro_compat &
1524 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1525 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1526 return 0;
1527
1528 return 1;
1529}
1530
ac27a0ec 1531/* Called at mount-time, super-block is locked */
197cd65a 1532static int ext4_check_descriptors(struct super_block *sb)
ac27a0ec 1533{
617ba13b
MC		 1534 struct ext4_sb_info *sbi = EXT4_SB(sb);
1535 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1536 ext4_fsblk_t last_block;
bd81d8ee
LV		 1537 ext4_fsblk_t block_bitmap;
1538 ext4_fsblk_t inode_bitmap;
1539 ext4_fsblk_t inode_table;
ce421581 1540 int flexbg_flag = 0;
fd2d4291 1541 ext4_group_t i;
ac27a0ec 1542
ce421581
JS		 1543 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1544 flexbg_flag = 1;
1545
617ba13b 1546 ext4_debug ("Checking group descriptors");
ac27a0ec 1547
197cd65a
AM		 1548 for (i = 0; i < sbi->s_groups_count; i++) {
1549 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1550
ce421581 1551 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 1552 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 1553 else
1554 last_block = first_block +
617ba13b 1555 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 1556
8fadc143 1557 block_bitmap = ext4_block_bitmap(sb, gdp);
bd81d8ee 1558 if (block_bitmap < first_block || block_bitmap > last_block)
ac27a0ec 1559 {
c19204b0
JB		 1560 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1561 "Block bitmap for group %lu not in group "
1562 "(block %llu)!", i, block_bitmap);
ac27a0ec
DK		 1563 return 0;
1564 }
8fadc143 1565 inode_bitmap = ext4_inode_bitmap(sb, gdp);
bd81d8ee 1566 if (inode_bitmap < first_block || inode_bitmap > last_block)
ac27a0ec 1567 {
c19204b0
JB		 1568 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1569 "Inode bitmap for group %lu not in group "
1570 "(block %llu)!", i, inode_bitmap);
ac27a0ec
DK		 1571 return 0;
1572 }
8fadc143 1573 inode_table = ext4_inode_table(sb, gdp);
bd81d8ee 1574 if (inode_table < first_block ||
780dcdb2 1575 inode_table + sbi->s_itb_per_group - 1 > last_block)
ac27a0ec 1576 {
c19204b0
JB		 1577 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1578 "Inode table for group %lu not in group "
1579 "(block %llu)!", i, inode_table);
ac27a0ec
DK		 1580 return 0;
1581 }
717d50e4 1582 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
c19204b0
JB		 1583 printk(KERN_ERR "EXT4-fs: ext4_check_descriptors: "
1584 "Checksum for group %lu failed (%u!=%u)\n",
1585 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1586 gdp)), le16_to_cpu(gdp->bg_checksum));
717d50e4
AD		 1587 return 0;
1588 }
ce421581
JS		 1589 if (!flexbg_flag)
1590 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec
DK		 1591 }
1592
bd81d8ee 1593 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
617ba13b 1594 sbi->s_es->s_free_inodes_count=cpu_to_le32(ext4_count_free_inodes(sb));
ac27a0ec
DK		 1595 return 1;
1596}
1597
617ba13b 1598/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 1599 * the superblock) which were deleted from all directories, but held open by
1600 * a process at the time of a crash. We walk the list and try to delete these
1601 * inodes at recovery time (only with a read-write filesystem).
1602 *
1603 * In order to keep the orphan inode chain consistent during traversal (in
1604 * case of crash during recovery), we link each inode into the superblock
1605 * orphan list_head and handle it the same way as an inode deletion during
1606 * normal operation (which journals the operations for us).
1607 *
1608 * We only do an iget() and an iput() on each inode, which is very safe if we
1609 * accidentally point at an in-use or already deleted inode. The worst that
1610 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 1611 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 1612 * e2fsck was run on this filesystem, and it must have already done the orphan
1613 * inode cleanup for us, so we can safely abort without any further action.
1614 */
617ba13b
MC		 1615static void ext4_orphan_cleanup (struct super_block * sb,
1616 struct ext4_super_block * es)
ac27a0ec
DK		 1617{
1618 unsigned int s_flags = sb->s_flags;
1619 int nr_orphans = 0, nr_truncates = 0;
1620#ifdef CONFIG_QUOTA
1621 int i;
1622#endif
1623 if (!es->s_last_orphan) {
1624 jbd_debug(4, "no orphan inodes to clean up\n");
1625 return;
1626 }
1627
a8f48a95
ES		 1628 if (bdev_read_only(sb->s_bdev)) {
1629 printk(KERN_ERR "EXT4-fs: write access "
1630 "unavailable, skipping orphan cleanup.\n");
1631 return;
1632 }
1633
617ba13b 1634 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
ac27a0ec
DK		 1635 if (es->s_last_orphan)
1636 jbd_debug(1, "Errors on filesystem, "
1637 "clearing orphan list.\n");
1638 es->s_last_orphan = 0;
1639 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1640 return;
1641 }
1642
1643 if (s_flags & MS_RDONLY) {
617ba13b 1644 printk(KERN_INFO "EXT4-fs: %s: orphan cleanup on readonly fs\n",
ac27a0ec
DK		 1645 sb->s_id);
1646 sb->s_flags &= ~MS_RDONLY;
1647 }
1648#ifdef CONFIG_QUOTA
1649 /* Needed for iput() to work correctly and not trash data */
1650 sb->s_flags |= MS_ACTIVE;
1651 /* Turn on quotas so that they are updated correctly */
1652 for (i = 0; i < MAXQUOTAS; i++) {
617ba13b
MC		 1653 if (EXT4_SB(sb)->s_qf_names[i]) {
1654 int ret = ext4_quota_on_mount(sb, i);
ac27a0ec
DK		 1655 if (ret < 0)
1656 printk(KERN_ERR
2c8be6b2 1657 "EXT4-fs: Cannot turn on journaled "
ac27a0ec
DK		 1658 "quota: error %d\n", ret);
1659 }
1660 }
1661#endif
1662
1663 while (es->s_last_orphan) {
1664 struct inode *inode;
1665
97bd42b9
JB		 1666 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1667 if (IS_ERR(inode)) {
ac27a0ec
DK		 1668 es->s_last_orphan = 0;
1669 break;
1670 }
1671
617ba13b 1672 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
ac27a0ec
DK		 1673 DQUOT_INIT(inode);
1674 if (inode->i_nlink) {
1675 printk(KERN_DEBUG
1676 "%s: truncating inode %lu to %Ld bytes\n",
46e665e9 1677 __func__, inode->i_ino, inode->i_size);
ac27a0ec
DK		 1678 jbd_debug(2, "truncating inode %lu to %Ld bytes\n",
1679 inode->i_ino, inode->i_size);
617ba13b 1680 ext4_truncate(inode);
ac27a0ec
DK		 1681 nr_truncates++;
1682 } else {
1683 printk(KERN_DEBUG
1684 "%s: deleting unreferenced inode %lu\n",
46e665e9 1685 __func__, inode->i_ino);
ac27a0ec
DK		 1686 jbd_debug(2, "deleting unreferenced inode %lu\n",
1687 inode->i_ino);
1688 nr_orphans++;
1689 }
1690 iput(inode); /* The delete magic happens here! */
1691 }
1692
1693#define PLURAL(x) (x), ((x)==1) ? "" : "s"
1694
1695 if (nr_orphans)
617ba13b 1696 printk(KERN_INFO "EXT4-fs: %s: %d orphan inode%s deleted\n",
ac27a0ec
DK		 1697 sb->s_id, PLURAL(nr_orphans));
1698 if (nr_truncates)
617ba13b 1699 printk(KERN_INFO "EXT4-fs: %s: %d truncate%s cleaned up\n",
ac27a0ec
DK		 1700 sb->s_id, PLURAL(nr_truncates));
1701#ifdef CONFIG_QUOTA
1702 /* Turn quotas off */
1703 for (i = 0; i < MAXQUOTAS; i++) {
1704 if (sb_dqopt(sb)->files[i])
6f28e087 1705 vfs_quota_off(sb, i, 0);
ac27a0ec
DK		 1706 }
1707#endif
1708 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1709}
cd2291a4
ES		 1710/*
1711 * Maximal extent format file size.
1712 * Resulting logical blkno at s_maxbytes must fit in our on-disk
1713 * extent format containers, within a sector_t, and within i_blocks
1714 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
1715 * so that won't be a limiting factor.
1716 *
1717 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1718 */
1719static loff_t ext4_max_size(int blkbits)
1720{
1721 loff_t res;
1722 loff_t upper_limit = MAX_LFS_FILESIZE;
1723
1724 /* small i_blocks in vfs inode? */
1725 if (sizeof(blkcnt_t) < sizeof(u64)) {
1726 /*
1727 * CONFIG_LSF is not enabled implies the inode
1728 * i_block represent total blocks in 512 bytes
1729 * 32 == size of vfs inode i_blocks * 8
1730 */
1731 upper_limit = (1LL << 32) - 1;
1732
1733 /* total blocks in file system block size */
1734 upper_limit >>= (blkbits - 9);
1735 upper_limit <<= blkbits;
1736 }
1737
1738 /* 32-bit extent-start container, ee_block */
1739 res = 1LL << 32;
1740 res <<= blkbits;
1741 res -= 1;
1742
1743 /* Sanity check against vm- & vfs- imposed limits */
1744 if (res > upper_limit)
1745 res = upper_limit;
1746
1747 return res;
1748}
ac27a0ec 1749
ac27a0ec 1750/*
cd2291a4 1751 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 1752 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
1753 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 1754 */
cd2291a4 1755static loff_t ext4_max_bitmap_size(int bits)
ac27a0ec 1756{
617ba13b 1757 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 1758 int meta_blocks;
1759 loff_t upper_limit;
1760 /* This is calculated to be the largest file size for a
cd2291a4 1761 * dense, bitmapped file such that the total number of
ac27a0ec 1762 * sectors in the file, including data and all indirect blocks,
0fc1b451
AK		 1763 * does not exceed 2^48 -1
1764 * __u32 i_blocks_lo and _u16 i_blocks_high representing the
1765 * total number of 512 bytes blocks of the file
1766 */
1767
1768 if (sizeof(blkcnt_t) < sizeof(u64)) {
1769 /*
1770 * CONFIG_LSF is not enabled implies the inode
1771 * i_block represent total blocks in 512 bytes
1772 * 32 == size of vfs inode i_blocks * 8
1773 */
1774 upper_limit = (1LL << 32) - 1;
1775
1776 /* total blocks in file system block size */
1777 upper_limit >>= (bits - 9);
1778
1779 } else {
8180a562
AK		 1780 /*
1781 * We use 48 bit ext4_inode i_blocks
1782 * With EXT4_HUGE_FILE_FL set the i_blocks
1783 * represent total number of blocks in
1784 * file system block size
1785 */
0fc1b451
AK		 1786 upper_limit = (1LL << 48) - 1;
1787
0fc1b451
AK		 1788 }
1789
1790 /* indirect blocks */
1791 meta_blocks = 1;
1792 /* double indirect blocks */
1793 meta_blocks += 1 + (1LL << (bits-2));
1794 /* tripple indirect blocks */
1795 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
1796
1797 upper_limit -= meta_blocks;
1798 upper_limit <<= bits;
ac27a0ec
DK		 1799
1800 res += 1LL << (bits-2);
1801 res += 1LL << (2*(bits-2));
1802 res += 1LL << (3*(bits-2));
1803 res <<= bits;
1804 if (res > upper_limit)
1805 res = upper_limit;
0fc1b451
AK		 1806
1807 if (res > MAX_LFS_FILESIZE)
1808 res = MAX_LFS_FILESIZE;
1809
ac27a0ec
DK		 1810 return res;
1811}
1812
617ba13b 1813static ext4_fsblk_t descriptor_loc(struct super_block *sb,
70bbb3e0 1814 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 1815{
617ba13b 1816 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 1817 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 1818 int has_super = 0;
1819
1820 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1821
617ba13b 1822 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
ac27a0ec 1823 nr < first_meta_bg)
70bbb3e0 1824 return logical_sb_block + nr + 1;
ac27a0ec 1825 bg = sbi->s_desc_per_block * nr;
617ba13b 1826 if (ext4_bg_has_super(sb, bg))
ac27a0ec 1827 has_super = 1;
617ba13b 1828 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 1829}
1830
c9de560d
AT		 1831/**
1832 * ext4_get_stripe_size: Get the stripe size.
1833 * @sbi: In memory super block info
1834 *
1835 * If we have specified it via mount option, then
1836 * use the mount option value. If the value specified at mount time is
1837 * greater than the blocks per group use the super block value.
1838 * If the super block value is greater than blocks per group return 0.
1839 * Allocator needs it be less than blocks per group.
1840 *
1841 */
1842static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
1843{
1844 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
1845 unsigned long stripe_width =
1846 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
1847
1848 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
1849 return sbi->s_stripe;
1850
1851 if (stripe_width <= sbi->s_blocks_per_group)
1852 return stripe_width;
1853
1854 if (stride <= sbi->s_blocks_per_group)
1855 return stride;
1856
1857 return 0;
1858}
ac27a0ec 1859
617ba13b 1860static int ext4_fill_super (struct super_block *sb, void *data, int silent)
7477827f
AK		 1861 __releases(kernel_lock)
1862 __acquires(kernel_lock)
1d03ec98 1863
ac27a0ec
DK		 1864{
1865 struct buffer_head * bh;
617ba13b
MC		 1866 struct ext4_super_block *es = NULL;
1867 struct ext4_sb_info *sbi;
1868 ext4_fsblk_t block;
1869 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 1870 ext4_fsblk_t logical_sb_block;
ac27a0ec
DK		 1871 unsigned long offset = 0;
1872 unsigned int journal_inum = 0;
1873 unsigned long journal_devnum = 0;
1874 unsigned long def_mount_opts;
1875 struct inode *root;
1d1fe1ee 1876 int ret = -EINVAL;
ac27a0ec 1877 int blocksize;
ac27a0ec
DK		 1878 int db_count;
1879 int i;
1880 int needs_recovery;
1881 __le32 features;
bd81d8ee 1882 __u64 blocks_count;
833f4077 1883 int err;
ac27a0ec
DK		 1884
1885 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
1886 if (!sbi)
1887 return -ENOMEM;
1888 sb->s_fs_info = sbi;
1889 sbi->s_mount_opt = 0;
617ba13b
MC		 1890 sbi->s_resuid = EXT4_DEF_RESUID;
1891 sbi->s_resgid = EXT4_DEF_RESGID;
d9c9bef1 1892 sbi->s_sb_block = sb_block;
ac27a0ec
DK		 1893
1894 unlock_kernel();
1895
617ba13b 1896 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 1897 if (!blocksize) {
617ba13b 1898 printk(KERN_ERR "EXT4-fs: unable to set blocksize\n");
ac27a0ec
DK		 1899 goto out_fail;
1900 }
1901
1902 /*
617ba13b 1903 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 1904 * block sizes. We need to calculate the offset from buffer start.
1905 */
617ba13b 1906 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 1907 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
1908 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 1909 } else {
70bbb3e0 1910 logical_sb_block = sb_block;
ac27a0ec
DK		 1911 }
1912
70bbb3e0 1913 if (!(bh = sb_bread(sb, logical_sb_block))) {
617ba13b 1914 printk (KERN_ERR "EXT4-fs: unable to read superblock\n");
ac27a0ec
DK		 1915 goto out_fail;
1916 }
1917 /*
1918 * Note: s_es must be initialized as soon as possible because
617ba13b 1919 * some ext4 macro-instructions depend on its value
ac27a0ec 1920 */
617ba13b 1921 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
ac27a0ec
DK		 1922 sbi->s_es = es;
1923 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 1924 if (sb->s_magic != EXT4_SUPER_MAGIC)
1925 goto cantfind_ext4;
ac27a0ec
DK		 1926
1927 /* Set defaults before we parse the mount options */
1928 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
617ba13b 1929 if (def_mount_opts & EXT4_DEFM_DEBUG)
ac27a0ec 1930 set_opt(sbi->s_mount_opt, DEBUG);
617ba13b 1931 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
ac27a0ec 1932 set_opt(sbi->s_mount_opt, GRPID);
617ba13b 1933 if (def_mount_opts & EXT4_DEFM_UID16)
ac27a0ec 1934 set_opt(sbi->s_mount_opt, NO_UID32);
2e7842b8 1935#ifdef CONFIG_EXT4DEV_FS_XATTR
617ba13b 1936 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
ac27a0ec 1937 set_opt(sbi->s_mount_opt, XATTR_USER);
2e7842b8
HD		 1938#endif
1939#ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
617ba13b 1940 if (def_mount_opts & EXT4_DEFM_ACL)
ac27a0ec 1941 set_opt(sbi->s_mount_opt, POSIX_ACL);
2e7842b8 1942#endif
617ba13b
MC		 1943 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
1944 sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
1945 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
1946 sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
1947 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
1948 sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
1949
1950 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
ac27a0ec 1951 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
bb4f397a 1952 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
ceea16bf 1953 set_opt(sbi->s_mount_opt, ERRORS_CONT);
bb4f397a
AK		 1954 else
1955 set_opt(sbi->s_mount_opt, ERRORS_RO);
ac27a0ec
DK		 1956
1957 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
1958 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
1959
1960 set_opt(sbi->s_mount_opt, RESERVATION);
571640ca 1961 set_opt(sbi->s_mount_opt, BARRIER);
ac27a0ec 1962
1e2462f9
MC		 1963 /*
1964 * turn on extents feature by default in ext4 filesystem
1965 * User -o noextents to turn it off
1966 */
1967 set_opt(sbi->s_mount_opt, EXTENTS);
3dbd0ede
AK		 1968 /*
1969 * turn on mballoc feature by default in ext4 filesystem
1970 * User -o nomballoc to turn it off
1971 */
1972 set_opt(sbi->s_mount_opt, MBALLOC);
1e2462f9 1973
ac27a0ec
DK		 1974 if (!parse_options ((char *) data, sb, &journal_inum, &journal_devnum,
1975 NULL, 0))
1976 goto failed_mount;
1977
1978 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
617ba13b 1979 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec 1980
617ba13b
MC		 1981 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
1982 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
1983 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
1984 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
ac27a0ec 1985 printk(KERN_WARNING
617ba13b 1986 "EXT4-fs warning: feature flags set on rev 0 fs, "
ac27a0ec 1987 "running e2fsck is recommended\n");
469108ff
TT		 1988
1989 /*
1990 * Since ext4 is still considered development code, we require
1991 * that the TEST_FILESYS flag in s->flags be set.
1992 */
1993 if (!(le32_to_cpu(es->s_flags) & EXT2_FLAGS_TEST_FILESYS)) {
1994 printk(KERN_WARNING "EXT4-fs: %s: not marked "
1995 "OK to use with test code.\n", sb->s_id);
1996 goto failed_mount;
1997 }
1998
ac27a0ec
DK		 1999 /*
2000 * Check feature flags regardless of the revision level, since we
2001 * previously didn't change the revision level when setting the flags,
2002 * so there is a chance incompat flags are set on a rev 0 filesystem.
2003 */
617ba13b 2004 features = EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP);
ac27a0ec 2005 if (features) {
617ba13b 2006 printk(KERN_ERR "EXT4-fs: %s: couldn't mount because of "
ac27a0ec
DK		 2007 "unsupported optional features (%x).\n",
2008 sb->s_id, le32_to_cpu(features));
2009 goto failed_mount;
2010 }
617ba13b 2011 features = EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP);
ac27a0ec 2012 if (!(sb->s_flags & MS_RDONLY) && features) {
617ba13b 2013 printk(KERN_ERR "EXT4-fs: %s: couldn't mount RDWR because of "
ac27a0ec
DK		 2014 "unsupported optional features (%x).\n",
2015 sb->s_id, le32_to_cpu(features));
2016 goto failed_mount;
2017 }
0fc1b451
AK		 2018 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2019 /*
2020 * Large file size enabled file system can only be
2021 * mount if kernel is build with CONFIG_LSF
2022 */
2023 if (sizeof(root->i_blocks) < sizeof(u64) &&
2024 !(sb->s_flags & MS_RDONLY)) {
2025 printk(KERN_ERR "EXT4-fs: %s: Filesystem with huge "
2026 "files cannot be mounted read-write "
2027 "without CONFIG_LSF.\n", sb->s_id);
2028 goto failed_mount;
2029 }
2030 }
ac27a0ec
DK		 2031 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
2032
617ba13b
MC		 2033 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
2034 blocksize > EXT4_MAX_BLOCK_SIZE) {
ac27a0ec 2035 printk(KERN_ERR
617ba13b 2036 "EXT4-fs: Unsupported filesystem blocksize %d on %s.\n",
ac27a0ec
DK		 2037 blocksize, sb->s_id);
2038 goto failed_mount;
2039 }
2040
ac27a0ec 2041 if (sb->s_blocksize != blocksize) {
ce40733c
AK		 2042
2043 /* Validate the filesystem blocksize */
2044 if (!sb_set_blocksize(sb, blocksize)) {
2045 printk(KERN_ERR "EXT4-fs: bad block size %d.\n",
2046 blocksize);
ac27a0ec
DK		 2047 goto failed_mount;
2048 }
2049
2050 brelse (bh);
70bbb3e0
AM		 2051 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2052 offset = do_div(logical_sb_block, blocksize);
2053 bh = sb_bread(sb, logical_sb_block);
ac27a0ec
DK		 2054 if (!bh) {
2055 printk(KERN_ERR
617ba13b 2056 "EXT4-fs: Can't read superblock on 2nd try.\n");
ac27a0ec
DK		 2057 goto failed_mount;
2058 }
617ba13b 2059 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
ac27a0ec 2060 sbi->s_es = es;
617ba13b 2061 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
ac27a0ec 2062 printk (KERN_ERR
617ba13b 2063 "EXT4-fs: Magic mismatch, very weird !\n");
ac27a0ec
DK		 2064 goto failed_mount;
2065 }
2066 }
2067
e2b46574 2068 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits);
617ba13b 2069 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits);
ac27a0ec 2070
617ba13b
MC		 2071 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2072 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2073 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
ac27a0ec
DK		 2074 } else {
2075 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2076 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
617ba13b 2077 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1330593e 2078 (!is_power_of_2(sbi->s_inode_size)) ||
ac27a0ec
DK		 2079 (sbi->s_inode_size > blocksize)) {
2080 printk (KERN_ERR
617ba13b 2081 "EXT4-fs: unsupported inode size: %d\n",
ac27a0ec
DK		 2082 sbi->s_inode_size);
2083 goto failed_mount;
2084 }
ef7f3835
KS		 2085 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2086 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
ac27a0ec 2087 }
0d1ee42f
AR		 2088 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2089 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
8fadc143 2090 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 2091 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 2092 !is_power_of_2(sbi->s_desc_size)) {
0d1ee42f 2093 printk(KERN_ERR
8fadc143 2094 "EXT4-fs: unsupported descriptor size %lu\n",
0d1ee42f
AR		 2095 sbi->s_desc_size);
2096 goto failed_mount;
2097 }
2098 } else
2099 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
ac27a0ec 2100 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 2101 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
b47b6f38 2102 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
617ba13b
MC		 2103 goto cantfind_ext4;
2104 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 2105 if (sbi->s_inodes_per_block == 0)
617ba13b 2106 goto cantfind_ext4;
ac27a0ec
DK		 2107 sbi->s_itb_per_group = sbi->s_inodes_per_group /
2108 sbi->s_inodes_per_block;
0d1ee42f 2109 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 2110 sbi->s_sbh = bh;
2111 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 2112 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2113 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
ac27a0ec
DK		 2114 for (i=0; i < 4; i++)
2115 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2116 sbi->s_def_hash_version = es->s_def_hash_version;
2117
2118 if (sbi->s_blocks_per_group > blocksize * 8) {
2119 printk (KERN_ERR
617ba13b 2120 "EXT4-fs: #blocks per group too big: %lu\n",
ac27a0ec
DK		 2121 sbi->s_blocks_per_group);
2122 goto failed_mount;
2123 }
ac27a0ec
DK		 2124 if (sbi->s_inodes_per_group > blocksize * 8) {
2125 printk (KERN_ERR
617ba13b 2126 "EXT4-fs: #inodes per group too big: %lu\n",
ac27a0ec
DK		 2127 sbi->s_inodes_per_group);
2128 goto failed_mount;
2129 }
2130
bd81d8ee 2131 if (ext4_blocks_count(es) >
ac27a0ec 2132 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
617ba13b 2133 printk(KERN_ERR "EXT4-fs: filesystem on %s:"
ac27a0ec
DK		 2134 " too large to mount safely\n", sb->s_id);
2135 if (sizeof(sector_t) < 8)
617ba13b 2136 printk(KERN_WARNING "EXT4-fs: CONFIG_LBD not "
ac27a0ec
DK		 2137 "enabled\n");
2138 goto failed_mount;
2139 }
2140
617ba13b
MC		 2141 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2142 goto cantfind_ext4;
e7c95593
ES		 2143
2144 /* ensure blocks_count calculation below doesn't sign-extend */
2145 if (ext4_blocks_count(es) + EXT4_BLOCKS_PER_GROUP(sb) <
2146 le32_to_cpu(es->s_first_data_block) + 1) {
2147 printk(KERN_WARNING "EXT4-fs: bad geometry: block count %llu, "
2148 "first data block %u, blocks per group %lu\n",
2149 ext4_blocks_count(es),
2150 le32_to_cpu(es->s_first_data_block),
2151 EXT4_BLOCKS_PER_GROUP(sb));
2152 goto failed_mount;
2153 }
bd81d8ee
LV		 2154 blocks_count = (ext4_blocks_count(es) -
2155 le32_to_cpu(es->s_first_data_block) +
2156 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2157 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
2158 sbi->s_groups_count = blocks_count;
617ba13b
MC		 2159 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2160 EXT4_DESC_PER_BLOCK(sb);
ac27a0ec
DK		 2161 sbi->s_group_desc = kmalloc(db_count * sizeof (struct buffer_head *),
2162 GFP_KERNEL);
2163 if (sbi->s_group_desc == NULL) {
617ba13b 2164 printk (KERN_ERR "EXT4-fs: not enough memory\n");
ac27a0ec
DK		 2165 goto failed_mount;
2166 }
2167
2168 bgl_lock_init(&sbi->s_blockgroup_lock);
2169
2170 for (i = 0; i < db_count; i++) {
70bbb3e0 2171 block = descriptor_loc(sb, logical_sb_block, i);
ac27a0ec
DK		 2172 sbi->s_group_desc[i] = sb_bread(sb, block);
2173 if (!sbi->s_group_desc[i]) {
617ba13b 2174 printk (KERN_ERR "EXT4-fs: "
ac27a0ec
DK		 2175 "can't read group descriptor %d\n", i);
2176 db_count = i;
2177 goto failed_mount2;
2178 }
2179 }
617ba13b
MC		 2180 if (!ext4_check_descriptors (sb)) {
2181 printk(KERN_ERR "EXT4-fs: group descriptors corrupted!\n");
ac27a0ec
DK		 2182 goto failed_mount2;
2183 }
772cb7c8
JS		 2184 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2185 if (!ext4_fill_flex_info(sb)) {
2186 printk(KERN_ERR
2187 "EXT4-fs: unable to initialize "
2188 "flex_bg meta info!\n");
2189 goto failed_mount2;
2190 }
2191
ac27a0ec
DK		 2192 sbi->s_gdb_count = db_count;
2193 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2194 spin_lock_init(&sbi->s_next_gen_lock);
2195
833f4077
PZ		 2196 err = percpu_counter_init(&sbi->s_freeblocks_counter,
2197 ext4_count_free_blocks(sb));
2198 if (!err) {
2199 err = percpu_counter_init(&sbi->s_freeinodes_counter,
2200 ext4_count_free_inodes(sb));
2201 }
2202 if (!err) {
2203 err = percpu_counter_init(&sbi->s_dirs_counter,
2204 ext4_count_dirs(sb));
2205 }
2206 if (err) {
2207 printk(KERN_ERR "EXT4-fs: insufficient memory\n");
2208 goto failed_mount3;
2209 }
ac27a0ec
DK		 2210
2211 /* per fileystem reservation list head & lock */
2212 spin_lock_init(&sbi->s_rsv_window_lock);
2213 sbi->s_rsv_window_root = RB_ROOT;
2214 /* Add a single, static dummy reservation to the start of the
2215 * reservation window list --- it gives us a placeholder for
2216 * append-at-start-of-list which makes the allocation logic
2217 * _much_ simpler. */
617ba13b
MC		 2218 sbi->s_rsv_window_head.rsv_start = EXT4_RESERVE_WINDOW_NOT_ALLOCATED;
2219 sbi->s_rsv_window_head.rsv_end = EXT4_RESERVE_WINDOW_NOT_ALLOCATED;
ac27a0ec
DK		 2220 sbi->s_rsv_window_head.rsv_alloc_hit = 0;
2221 sbi->s_rsv_window_head.rsv_goal_size = 0;
617ba13b 2222 ext4_rsv_window_add(sb, &sbi->s_rsv_window_head);
ac27a0ec 2223
c9de560d
AT		 2224 sbi->s_stripe = ext4_get_stripe_size(sbi);
2225
ac27a0ec
DK		 2226 /*
2227 * set up enough so that it can read an inode
2228 */
617ba13b
MC		 2229 sb->s_op = &ext4_sops;
2230 sb->s_export_op = &ext4_export_ops;
2231 sb->s_xattr = ext4_xattr_handlers;
ac27a0ec 2232#ifdef CONFIG_QUOTA
617ba13b
MC		 2233 sb->s_qcop = &ext4_qctl_operations;
2234 sb->dq_op = &ext4_quota_operations;
ac27a0ec
DK		 2235#endif
2236 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
2237
2238 sb->s_root = NULL;
2239
2240 needs_recovery = (es->s_last_orphan != 0 ||
617ba13b
MC		 2241 EXT4_HAS_INCOMPAT_FEATURE(sb,
2242 EXT4_FEATURE_INCOMPAT_RECOVER));
ac27a0ec
DK		 2243
2244 /*
2245 * The first inode we look at is the journal inode. Don't try
2246 * root first: it may be modified in the journal!
2247 */
2248 if (!test_opt(sb, NOLOAD) &&
617ba13b
MC		 2249 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2250 if (ext4_load_journal(sb, es, journal_devnum))
ac27a0ec 2251 goto failed_mount3;
624080ed
TT		 2252 if (!(sb->s_flags & MS_RDONLY) &&
2253 EXT4_SB(sb)->s_journal->j_failed_commit) {
2254 printk(KERN_CRIT "EXT4-fs error (device %s): "
2255 "ext4_fill_super: Journal transaction "
2256 "%u is corrupt\n", sb->s_id,
2257 EXT4_SB(sb)->s_journal->j_failed_commit);
2258 if (test_opt (sb, ERRORS_RO)) {
2259 printk (KERN_CRIT
2260 "Mounting filesystem read-only\n");
2261 sb->s_flags |= MS_RDONLY;
2262 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2263 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2264 }
2265 if (test_opt(sb, ERRORS_PANIC)) {
2266 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2267 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2268 ext4_commit_super(sb, es, 1);
2269 printk(KERN_CRIT
2270 "EXT4-fs (device %s): mount failed\n",
2271 sb->s_id);
2272 goto failed_mount4;
2273 }
2274 }
ac27a0ec 2275 } else if (journal_inum) {
617ba13b 2276 if (ext4_create_journal(sb, es, journal_inum))
ac27a0ec
DK		 2277 goto failed_mount3;
2278 } else {
2279 if (!silent)
2280 printk (KERN_ERR
617ba13b 2281 "ext4: No journal on filesystem on %s\n",
ac27a0ec
DK		 2282 sb->s_id);
2283 goto failed_mount3;
2284 }
2285
eb40a09c
JS		 2286 if (ext4_blocks_count(es) > 0xffffffffULL &&
2287 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2288 JBD2_FEATURE_INCOMPAT_64BIT)) {
2289 printk(KERN_ERR "ext4: Failed to set 64-bit journal feature\n");
2290 goto failed_mount4;
2291 }
2292
818d276c
GS		 2293 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2294 jbd2_journal_set_features(sbi->s_journal,
2295 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2296 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2297 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2298 jbd2_journal_set_features(sbi->s_journal,
2299 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2300 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2301 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2302 } else {
2303 jbd2_journal_clear_features(sbi->s_journal,
2304 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2305 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2306 }
2307
ac27a0ec
DK		 2308 /* We have now updated the journal if required, so we can
2309 * validate the data journaling mode. */
2310 switch (test_opt(sb, DATA_FLAGS)) {
2311 case 0:
2312 /* No mode set, assume a default based on the journal
63f57933
AM		 2313 * capabilities: ORDERED_DATA if the journal can
2314 * cope, else JOURNAL_DATA
2315 */
dab291af
MC		 2316 if (jbd2_journal_check_available_features
2317 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
ac27a0ec
DK		 2318 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2319 else
2320 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2321 break;
2322
617ba13b
MC		 2323 case EXT4_MOUNT_ORDERED_DATA:
2324 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 2325 if (!jbd2_journal_check_available_features
2326 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
617ba13b 2327 printk(KERN_ERR "EXT4-fs: Journal does not support "
ac27a0ec
DK		 2328 "requested data journaling mode\n");
2329 goto failed_mount4;
2330 }
2331 default:
2332 break;
2333 }
2334
2335 if (test_opt(sb, NOBH)) {
617ba13b
MC		 2336 if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
2337 printk(KERN_WARNING "EXT4-fs: Ignoring nobh option - "
ac27a0ec
DK		 2338 "its supported only with writeback mode\n");
2339 clear_opt(sbi->s_mount_opt, NOBH);
2340 }
2341 }
2342 /*
dab291af 2343 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 2344 * so we can safely mount the rest of the filesystem now.
2345 */
2346
1d1fe1ee
DH		 2347 root = ext4_iget(sb, EXT4_ROOT_INO);
2348 if (IS_ERR(root)) {
617ba13b 2349 printk(KERN_ERR "EXT4-fs: get root inode failed\n");
1d1fe1ee 2350 ret = PTR_ERR(root);
ac27a0ec
DK		 2351 goto failed_mount4;
2352 }
2353 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1d1fe1ee 2354 iput(root);
617ba13b 2355 printk(KERN_ERR "EXT4-fs: corrupt root inode, run e2fsck\n");
ac27a0ec
DK		 2356 goto failed_mount4;
2357 }
1d1fe1ee
DH		 2358 sb->s_root = d_alloc_root(root);
2359 if (!sb->s_root) {
2360 printk(KERN_ERR "EXT4-fs: get root dentry failed\n");
2361 iput(root);
2362 ret = -ENOMEM;
2363 goto failed_mount4;
2364 }
ac27a0ec 2365
617ba13b 2366 ext4_setup_super (sb, es, sb->s_flags & MS_RDONLY);
ef7f3835
KS		 2367
2368 /* determine the minimum size of new large inodes, if present */
2369 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2370 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2371 EXT4_GOOD_OLD_INODE_SIZE;
2372 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2373 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2374 if (sbi->s_want_extra_isize <
2375 le16_to_cpu(es->s_want_extra_isize))
2376 sbi->s_want_extra_isize =
2377 le16_to_cpu(es->s_want_extra_isize);
2378 if (sbi->s_want_extra_isize <
2379 le16_to_cpu(es->s_min_extra_isize))
2380 sbi->s_want_extra_isize =
2381 le16_to_cpu(es->s_min_extra_isize);
2382 }
2383 }
2384 /* Check if enough inode space is available */
2385 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2386 sbi->s_inode_size) {
2387 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2388 EXT4_GOOD_OLD_INODE_SIZE;
2389 printk(KERN_INFO "EXT4-fs: required extra inode space not"
2390 "available.\n");
2391 }
2392
ac27a0ec
DK		 2393 /*
2394 * akpm: core read_super() calls in here with the superblock locked.
2395 * That deadlocks, because orphan cleanup needs to lock the superblock
2396 * in numerous places. Here we just pop the lock - it's relatively
2397 * harmless, because we are now ready to accept write_super() requests,
2398 * and aviro says that's the only reason for hanging onto the
2399 * superblock lock.
2400 */
617ba13b
MC		 2401 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
2402 ext4_orphan_cleanup(sb, es);
2403 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
ac27a0ec 2404 if (needs_recovery)
617ba13b
MC		 2405 printk (KERN_INFO "EXT4-fs: recovery complete.\n");
2406 ext4_mark_recovery_complete(sb, es);
2407 printk (KERN_INFO "EXT4-fs: mounted filesystem with %s data mode.\n",
2408 test_opt(sb,DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA ? "journal":
2409 test_opt(sb,DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA ? "ordered":
ac27a0ec
DK		 2410 "writeback");
2411
a86c6181 2412 ext4_ext_init(sb);
c9de560d 2413 ext4_mb_init(sb, needs_recovery);
a86c6181 2414
ac27a0ec
DK		 2415 lock_kernel();
2416 return 0;
2417
617ba13b 2418cantfind_ext4:
ac27a0ec 2419 if (!silent)
617ba13b 2420 printk(KERN_ERR "VFS: Can't find ext4 filesystem on dev %s.\n",
ac27a0ec
DK		 2421 sb->s_id);
2422 goto failed_mount;
2423
2424failed_mount4:
dab291af 2425 jbd2_journal_destroy(sbi->s_journal);
ac27a0ec
DK		 2426failed_mount3:
2427 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2428 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2429 percpu_counter_destroy(&sbi->s_dirs_counter);
2430failed_mount2:
2431 for (i = 0; i < db_count; i++)
2432 brelse(sbi->s_group_desc[i]);
2433 kfree(sbi->s_group_desc);
2434failed_mount:
2435#ifdef CONFIG_QUOTA
2436 for (i = 0; i < MAXQUOTAS; i++)
2437 kfree(sbi->s_qf_names[i]);
2438#endif
617ba13b 2439 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 2440 brelse(bh);
2441out_fail:
2442 sb->s_fs_info = NULL;
2443 kfree(sbi);
2444 lock_kernel();
1d1fe1ee 2445 return ret;
ac27a0ec
DK		 2446}
2447
2448/*
2449 * Setup any per-fs journal parameters now. We'll do this both on
2450 * initial mount, once the journal has been initialised but before we've
2451 * done any recovery; and again on any subsequent remount.
2452 */
617ba13b 2453static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 2454{
617ba13b 2455 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 2456
2457 if (sbi->s_commit_interval)
2458 journal->j_commit_interval = sbi->s_commit_interval;
617ba13b 2459 /* We could also set up an ext4-specific default for the commit
ac27a0ec
DK		 2460 * interval here, but for now we'll just fall back to the jbd
2461 * default. */
2462
2463 spin_lock(&journal->j_state_lock);
2464 if (test_opt(sb, BARRIER))
dab291af 2465 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 2466 else
dab291af 2467 journal->j_flags &= ~JBD2_BARRIER;
ac27a0ec
DK		 2468 spin_unlock(&journal->j_state_lock);
2469}
2470
617ba13b 2471static journal_t *ext4_get_journal(struct super_block *sb,
ac27a0ec
DK		 2472 unsigned int journal_inum)
2473{
2474 struct inode *journal_inode;
2475 journal_t *journal;
2476
2477 /* First, test for the existence of a valid inode on disk. Bad
2478 * things happen if we iget() an unused inode, as the subsequent
2479 * iput() will try to delete it. */
2480
1d1fe1ee
DH		 2481 journal_inode = ext4_iget(sb, journal_inum);
2482 if (IS_ERR(journal_inode)) {
617ba13b 2483 printk(KERN_ERR "EXT4-fs: no journal found.\n");
ac27a0ec
DK		 2484 return NULL;
2485 }
2486 if (!journal_inode->i_nlink) {
2487 make_bad_inode(journal_inode);
2488 iput(journal_inode);
617ba13b 2489 printk(KERN_ERR "EXT4-fs: journal inode is deleted.\n");
ac27a0ec
DK		 2490 return NULL;
2491 }
2492
2493 jbd_debug(2, "Journal inode found at %p: %Ld bytes\n",
2494 journal_inode, journal_inode->i_size);
1d1fe1ee 2495 if (!S_ISREG(journal_inode->i_mode)) {
617ba13b 2496 printk(KERN_ERR "EXT4-fs: invalid journal inode.\n");
ac27a0ec
DK		 2497 iput(journal_inode);
2498 return NULL;
2499 }
2500
dab291af 2501 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 2502 if (!journal) {
617ba13b 2503 printk(KERN_ERR "EXT4-fs: Could not load journal inode\n");
ac27a0ec
DK		 2504 iput(journal_inode);
2505 return NULL;
2506 }
2507 journal->j_private = sb;
617ba13b 2508 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 2509 return journal;
2510}
2511
617ba13b 2512static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 2513 dev_t j_dev)
2514{
2515 struct buffer_head * bh;
2516 journal_t *journal;
617ba13b
MC		 2517 ext4_fsblk_t start;
2518 ext4_fsblk_t len;
ac27a0ec 2519 int hblock, blocksize;
617ba13b 2520 ext4_fsblk_t sb_block;
ac27a0ec 2521 unsigned long offset;
617ba13b 2522 struct ext4_super_block * es;
ac27a0ec
DK		 2523 struct block_device *bdev;
2524
617ba13b 2525 bdev = ext4_blkdev_get(j_dev);
ac27a0ec
DK		 2526 if (bdev == NULL)
2527 return NULL;
2528
2529 if (bd_claim(bdev, sb)) {
2530 printk(KERN_ERR
8c55e204 2531 "EXT4: failed to claim external journal device.\n");
ac27a0ec
DK		 2532 blkdev_put(bdev);
2533 return NULL;
2534 }
2535
2536 blocksize = sb->s_blocksize;
2537 hblock = bdev_hardsect_size(bdev);
2538 if (blocksize < hblock) {
2539 printk(KERN_ERR
617ba13b 2540 "EXT4-fs: blocksize too small for journal device.\n");
ac27a0ec
DK		 2541 goto out_bdev;
2542 }
2543
617ba13b
MC		 2544 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
2545 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 2546 set_blocksize(bdev, blocksize);
2547 if (!(bh = __bread(bdev, sb_block, blocksize))) {
617ba13b 2548 printk(KERN_ERR "EXT4-fs: couldn't read superblock of "
ac27a0ec
DK		 2549 "external journal\n");
2550 goto out_bdev;
2551 }
2552
617ba13b
MC		 2553 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2554 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 2555 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b
MC		 2556 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
2557 printk(KERN_ERR "EXT4-fs: external journal has "
ac27a0ec
DK		 2558 "bad superblock\n");
2559 brelse(bh);
2560 goto out_bdev;
2561 }
2562
617ba13b
MC		 2563 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
2564 printk(KERN_ERR "EXT4-fs: journal UUID does not match\n");
ac27a0ec
DK		 2565 brelse(bh);
2566 goto out_bdev;
2567 }
2568
bd81d8ee 2569 len = ext4_blocks_count(es);
ac27a0ec
DK		 2570 start = sb_block + 1;
2571 brelse(bh); /* we're done with the superblock */
2572
dab291af 2573 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 2574 start, len, blocksize);
2575 if (!journal) {
617ba13b 2576 printk(KERN_ERR "EXT4-fs: failed to create device journal\n");
ac27a0ec
DK		 2577 goto out_bdev;
2578 }
2579 journal->j_private = sb;
2580 ll_rw_block(READ, 1, &journal->j_sb_buffer);
2581 wait_on_buffer(journal->j_sb_buffer);
2582 if (!buffer_uptodate(journal->j_sb_buffer)) {
617ba13b 2583 printk(KERN_ERR "EXT4-fs: I/O error on journal device\n");
ac27a0ec
DK		 2584 goto out_journal;
2585 }
2586 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
617ba13b 2587 printk(KERN_ERR "EXT4-fs: External journal has more than one "
ac27a0ec
DK		 2588 "user (unsupported) - %d\n",
2589 be32_to_cpu(journal->j_superblock->s_nr_users));
2590 goto out_journal;
2591 }
617ba13b
MC		 2592 EXT4_SB(sb)->journal_bdev = bdev;
2593 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 2594 return journal;
2595out_journal:
dab291af 2596 jbd2_journal_destroy(journal);
ac27a0ec 2597out_bdev:
617ba13b 2598 ext4_blkdev_put(bdev);
ac27a0ec
DK		 2599 return NULL;
2600}
2601
617ba13b
MC		 2602static int ext4_load_journal(struct super_block *sb,
2603 struct ext4_super_block *es,
ac27a0ec
DK		 2604 unsigned long journal_devnum)
2605{
2606 journal_t *journal;
2607 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
2608 dev_t journal_dev;
2609 int err = 0;
2610 int really_read_only;
2611
2612 if (journal_devnum &&
2613 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
617ba13b 2614 printk(KERN_INFO "EXT4-fs: external journal device major/minor "
ac27a0ec
DK		 2615 "numbers have changed\n");
2616 journal_dev = new_decode_dev(journal_devnum);
2617 } else
2618 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
2619
2620 really_read_only = bdev_read_only(sb->s_bdev);
2621
2622 /*
2623 * Are we loading a blank journal or performing recovery after a
2624 * crash? For recovery, we need to check in advance whether we
2625 * can get read-write access to the device.
2626 */
2627
617ba13b 2628 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
ac27a0ec 2629 if (sb->s_flags & MS_RDONLY) {
617ba13b 2630 printk(KERN_INFO "EXT4-fs: INFO: recovery "
ac27a0ec
DK		 2631 "required on readonly filesystem.\n");
2632 if (really_read_only) {
617ba13b 2633 printk(KERN_ERR "EXT4-fs: write access "
ac27a0ec
DK		 2634 "unavailable, cannot proceed.\n");
2635 return -EROFS;
2636 }
617ba13b 2637 printk (KERN_INFO "EXT4-fs: write access will "
ac27a0ec
DK		 2638 "be enabled during recovery.\n");
2639 }
2640 }
2641
2642 if (journal_inum && journal_dev) {
617ba13b 2643 printk(KERN_ERR "EXT4-fs: filesystem has both journal "
ac27a0ec
DK		 2644 "and inode journals!\n");
2645 return -EINVAL;
2646 }
2647
2648 if (journal_inum) {
617ba13b 2649 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 2650 return -EINVAL;
2651 } else {
617ba13b 2652 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 2653 return -EINVAL;
2654 }
2655
2656 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
dab291af 2657 err = jbd2_journal_update_format(journal);
ac27a0ec 2658 if (err) {
617ba13b 2659 printk(KERN_ERR "EXT4-fs: error updating journal.\n");
dab291af 2660 jbd2_journal_destroy(journal);
ac27a0ec
DK		 2661 return err;
2662 }
2663 }
2664
617ba13b 2665 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
dab291af 2666 err = jbd2_journal_wipe(journal, !really_read_only);
ac27a0ec 2667 if (!err)
dab291af 2668 err = jbd2_journal_load(journal);
ac27a0ec
DK		 2669
2670 if (err) {
617ba13b 2671 printk(KERN_ERR "EXT4-fs: error loading journal.\n");
dab291af 2672 jbd2_journal_destroy(journal);
ac27a0ec
DK		 2673 return err;
2674 }
2675
617ba13b
MC		 2676 EXT4_SB(sb)->s_journal = journal;
2677 ext4_clear_journal_err(sb, es);
ac27a0ec
DK		 2678
2679 if (journal_devnum &&
2680 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2681 es->s_journal_dev = cpu_to_le32(journal_devnum);
2682 sb->s_dirt = 1;
2683
2684 /* Make sure we flush the recovery flag to disk. */
617ba13b 2685 ext4_commit_super(sb, es, 1);
ac27a0ec
DK		 2686 }
2687
2688 return 0;
2689}
2690
617ba13b
MC		 2691static int ext4_create_journal(struct super_block * sb,
2692 struct ext4_super_block * es,
ac27a0ec
DK		 2693 unsigned int journal_inum)
2694{
2695 journal_t *journal;
6c675bd4 2696 int err;
ac27a0ec
DK		 2697
2698 if (sb->s_flags & MS_RDONLY) {
617ba13b 2699 printk(KERN_ERR "EXT4-fs: readonly filesystem when trying to "
ac27a0ec
DK		 2700 "create journal.\n");
2701 return -EROFS;
2702 }
2703
6c675bd4
BP		 2704 journal = ext4_get_journal(sb, journal_inum);
2705 if (!journal)
ac27a0ec
DK		 2706 return -EINVAL;
2707
617ba13b 2708 printk(KERN_INFO "EXT4-fs: creating new journal on inode %u\n",
ac27a0ec
DK		 2709 journal_inum);
2710
6c675bd4
BP		 2711 err = jbd2_journal_create(journal);
2712 if (err) {
617ba13b 2713 printk(KERN_ERR "EXT4-fs: error creating journal.\n");
dab291af 2714 jbd2_journal_destroy(journal);
ac27a0ec
DK		 2715 return -EIO;
2716 }
2717
617ba13b 2718 EXT4_SB(sb)->s_journal = journal;
ac27a0ec 2719
617ba13b
MC		 2720 ext4_update_dynamic_rev(sb);
2721 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2722 EXT4_SET_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL);
ac27a0ec
DK		 2723
2724 es->s_journal_inum = cpu_to_le32(journal_inum);
2725 sb->s_dirt = 1;
2726
2727 /* Make sure we flush the recovery flag to disk. */
617ba13b 2728 ext4_commit_super(sb, es, 1);
ac27a0ec
DK		 2729
2730 return 0;
2731}
2732
617ba13b
MC		 2733static void ext4_commit_super (struct super_block * sb,
2734 struct ext4_super_block * es,
ac27a0ec
DK		 2735 int sync)
2736{
617ba13b 2737 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
ac27a0ec
DK		 2738
2739 if (!sbh)
2740 return;
2741 es->s_wtime = cpu_to_le32(get_seconds());
bd81d8ee 2742 ext4_free_blocks_count_set(es, ext4_count_free_blocks(sb));
617ba13b 2743 es->s_free_inodes_count = cpu_to_le32(ext4_count_free_inodes(sb));
ac27a0ec
DK		 2744 BUFFER_TRACE(sbh, "marking dirty");
2745 mark_buffer_dirty(sbh);
2746 if (sync)
2747 sync_dirty_buffer(sbh);
2748}
2749
2750
2751/*
2752 * Have we just finished recovery? If so, and if we are mounting (or
2753 * remounting) the filesystem readonly, then we will end up with a
2754 * consistent fs on disk. Record that fact.
2755 */
617ba13b
MC		 2756static void ext4_mark_recovery_complete(struct super_block * sb,
2757 struct ext4_super_block * es)
ac27a0ec 2758{
617ba13b 2759 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 2760
dab291af
MC		 2761 jbd2_journal_lock_updates(journal);
2762 jbd2_journal_flush(journal);
32c37730 2763 lock_super(sb);
617ba13b 2764 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
ac27a0ec 2765 sb->s_flags & MS_RDONLY) {
617ba13b 2766 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 2767 sb->s_dirt = 0;
617ba13b 2768 ext4_commit_super(sb, es, 1);
ac27a0ec 2769 }
32c37730 2770 unlock_super(sb);
dab291af 2771 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 2772}
2773
2774/*
2775 * If we are mounting (or read-write remounting) a filesystem whose journal
2776 * has recorded an error from a previous lifetime, move that error to the
2777 * main filesystem now.
2778 */
617ba13b
MC		 2779static void ext4_clear_journal_err(struct super_block * sb,
2780 struct ext4_super_block * es)
ac27a0ec
DK		 2781{
2782 journal_t *journal;
2783 int j_errno;
2784 const char *errstr;
2785
617ba13b 2786 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 2787
2788 /*
2789 * Now check for any error status which may have been recorded in the
617ba13b 2790 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 2791 */
2792
dab291af 2793 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 2794 if (j_errno) {
2795 char nbuf[16];
2796
617ba13b 2797 errstr = ext4_decode_error(sb, j_errno, nbuf);
46e665e9 2798 ext4_warning(sb, __func__, "Filesystem error recorded "
ac27a0ec 2799 "from previous mount: %s", errstr);
46e665e9 2800 ext4_warning(sb, __func__, "Marking fs in need of "
ac27a0ec
DK		 2801 "filesystem check.");
2802
617ba13b
MC		 2803 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2804 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2805 ext4_commit_super (sb, es, 1);
ac27a0ec 2806
dab291af 2807 jbd2_journal_clear_err(journal);
ac27a0ec
DK		 2808 }
2809}
2810
2811/*
2812 * Force the running and committing transactions to commit,
2813 * and wait on the commit.
2814 */
617ba13b 2815int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 2816{
2817 journal_t *journal;
2818 int ret;
2819
2820 if (sb->s_flags & MS_RDONLY)
2821 return 0;
2822
617ba13b 2823 journal = EXT4_SB(sb)->s_journal;
ac27a0ec 2824 sb->s_dirt = 0;
617ba13b 2825 ret = ext4_journal_force_commit(journal);
ac27a0ec
DK		 2826 return ret;
2827}
2828
2829/*
617ba13b 2830 * Ext4 always journals updates to the superblock itself, so we don't
ac27a0ec
DK		 2831 * have to propagate any other updates to the superblock on disk at this
2832 * point. Just start an async writeback to get the buffers on their way
2833 * to the disk.
2834 *
2835 * This implicitly triggers the writebehind on sync().
2836 */
2837
617ba13b 2838static void ext4_write_super (struct super_block * sb)
ac27a0ec
DK		 2839{
2840 if (mutex_trylock(&sb->s_lock) != 0)
2841 BUG();
2842 sb->s_dirt = 0;
2843}
2844
617ba13b 2845static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec
DK		 2846{
2847 tid_t target;
2848
2849 sb->s_dirt = 0;
dab291af 2850 if (jbd2_journal_start_commit(EXT4_SB(sb)->s_journal, &target)) {
ac27a0ec 2851 if (wait)
dab291af 2852 jbd2_log_wait_commit(EXT4_SB(sb)->s_journal, target);
ac27a0ec
DK		 2853 }
2854 return 0;
2855}
2856
2857/*
2858 * LVM calls this function before a (read-only) snapshot is created. This
2859 * gives us a chance to flush the journal completely and mark the fs clean.
2860 */
617ba13b 2861static void ext4_write_super_lockfs(struct super_block *sb)
ac27a0ec
DK		 2862{
2863 sb->s_dirt = 0;
2864
2865 if (!(sb->s_flags & MS_RDONLY)) {
617ba13b 2866 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 2867
2868 /* Now we set up the journal barrier. */
dab291af
MC		 2869 jbd2_journal_lock_updates(journal);
2870 jbd2_journal_flush(journal);
ac27a0ec
DK		 2871
2872 /* Journal blocked and flushed, clear needs_recovery flag. */
617ba13b
MC		 2873 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2874 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
ac27a0ec
DK		 2875 }
2876}
2877
2878/*
2879 * Called by LVM after the snapshot is done. We need to reset the RECOVER
2880 * flag here, even though the filesystem is not technically dirty yet.
2881 */
617ba13b 2882static void ext4_unlockfs(struct super_block *sb)
ac27a0ec
DK		 2883{
2884 if (!(sb->s_flags & MS_RDONLY)) {
2885 lock_super(sb);
2886 /* Reser the needs_recovery flag before the fs is unlocked. */
617ba13b
MC		 2887 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2888 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
ac27a0ec 2889 unlock_super(sb);
dab291af 2890 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
ac27a0ec
DK		 2891 }
2892}
2893
617ba13b 2894static int ext4_remount (struct super_block * sb, int * flags, char * data)
ac27a0ec 2895{
617ba13b
MC		 2896 struct ext4_super_block * es;
2897 struct ext4_sb_info *sbi = EXT4_SB(sb);
2898 ext4_fsblk_t n_blocks_count = 0;
ac27a0ec 2899 unsigned long old_sb_flags;
617ba13b 2900 struct ext4_mount_options old_opts;
ac27a0ec
DK		 2901 int err;
2902#ifdef CONFIG_QUOTA
2903 int i;
2904#endif
2905
2906 /* Store the original options */
2907 old_sb_flags = sb->s_flags;
2908 old_opts.s_mount_opt = sbi->s_mount_opt;
2909 old_opts.s_resuid = sbi->s_resuid;
2910 old_opts.s_resgid = sbi->s_resgid;
2911 old_opts.s_commit_interval = sbi->s_commit_interval;
2912#ifdef CONFIG_QUOTA
2913 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
2914 for (i = 0; i < MAXQUOTAS; i++)
2915 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
2916#endif
2917
2918 /*
2919 * Allow the "check" option to be passed as a remount option.
2920 */
2921 if (!parse_options(data, sb, NULL, NULL, &n_blocks_count, 1)) {
2922 err = -EINVAL;
2923 goto restore_opts;
2924 }
2925
617ba13b 2926 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT)
46e665e9 2927 ext4_abort(sb, __func__, "Abort forced by user");
ac27a0ec
DK		 2928
2929 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
617ba13b 2930 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec
DK		 2931
2932 es = sbi->s_es;
2933
617ba13b 2934 ext4_init_journal_params(sb, sbi->s_journal);
ac27a0ec
DK		 2935
2936 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
bd81d8ee 2937 n_blocks_count > ext4_blocks_count(es)) {
617ba13b 2938 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT) {
ac27a0ec
DK		 2939 err = -EROFS;
2940 goto restore_opts;
2941 }
2942
2943 if (*flags & MS_RDONLY) {
2944 /*
2945 * First of all, the unconditional stuff we have to do
2946 * to disable replay of the journal when we next remount
2947 */
2948 sb->s_flags |= MS_RDONLY;
2949
2950 /*
2951 * OK, test if we are remounting a valid rw partition
2952 * readonly, and if so set the rdonly flag and then
2953 * mark the partition as valid again.
2954 */
617ba13b
MC		 2955 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
2956 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 2957 es->s_state = cpu_to_le16(sbi->s_mount_state);
2958
32c37730
JK		 2959 /*
2960 * We have to unlock super so that we can wait for
2961 * transactions.
2962 */
2963 unlock_super(sb);
617ba13b 2964 ext4_mark_recovery_complete(sb, es);
32c37730 2965 lock_super(sb);
ac27a0ec
DK		 2966 } else {
2967 __le32 ret;
617ba13b
MC		 2968 if ((ret = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2969 ~EXT4_FEATURE_RO_COMPAT_SUPP))) {
2970 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
ac27a0ec
DK		 2971 "remount RDWR because of unsupported "
2972 "optional features (%x).\n",
2973 sb->s_id, le32_to_cpu(ret));
2974 err = -EROFS;
2975 goto restore_opts;
2976 }
ead6596b
ES		 2977
2978 /*
2979 * If we have an unprocessed orphan list hanging
2980 * around from a previously readonly bdev mount,
2981 * require a full umount/remount for now.
2982 */
2983 if (es->s_last_orphan) {
2984 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
2985 "remount RDWR because of unprocessed "
2986 "orphan inode list. Please "
2987 "umount/remount instead.\n",
2988 sb->s_id);
2989 err = -EINVAL;
2990 goto restore_opts;
2991 }
2992
ac27a0ec
DK		 2993 /*
2994 * Mounting a RDONLY partition read-write, so reread
2995 * and store the current valid flag. (It may have
2996 * been changed by e2fsck since we originally mounted
2997 * the partition.)
2998 */
617ba13b 2999 ext4_clear_journal_err(sb, es);
ac27a0ec 3000 sbi->s_mount_state = le16_to_cpu(es->s_state);
617ba13b 3001 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
ac27a0ec 3002 goto restore_opts;
617ba13b 3003 if (!ext4_setup_super (sb, es, 0))
ac27a0ec
DK		 3004 sb->s_flags &= ~MS_RDONLY;
3005 }
3006 }
3007#ifdef CONFIG_QUOTA
3008 /* Release old quota file names */
3009 for (i = 0; i < MAXQUOTAS; i++)
3010 if (old_opts.s_qf_names[i] &&
3011 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3012 kfree(old_opts.s_qf_names[i]);
3013#endif
3014 return 0;
3015restore_opts:
3016 sb->s_flags = old_sb_flags;
3017 sbi->s_mount_opt = old_opts.s_mount_opt;
3018 sbi->s_resuid = old_opts.s_resuid;
3019 sbi->s_resgid = old_opts.s_resgid;
3020 sbi->s_commit_interval = old_opts.s_commit_interval;
3021#ifdef CONFIG_QUOTA
3022 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
3023 for (i = 0; i < MAXQUOTAS; i++) {
3024 if (sbi->s_qf_names[i] &&
3025 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3026 kfree(sbi->s_qf_names[i]);
3027 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
3028 }
3029#endif
3030 return err;
3031}
3032
617ba13b 3033static int ext4_statfs (struct dentry * dentry, struct kstatfs * buf)
ac27a0ec
DK		 3034{
3035 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 3036 struct ext4_sb_info *sbi = EXT4_SB(sb);
3037 struct ext4_super_block *es = sbi->s_es;
960cc398 3038 u64 fsid;
ac27a0ec 3039
5e70030d
BP		 3040 if (test_opt(sb, MINIX_DF)) {
3041 sbi->s_overhead_last = 0;
6bc9feff 3042 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
fd2d4291 3043 ext4_group_t ngroups = sbi->s_groups_count, i;
5e70030d 3044 ext4_fsblk_t overhead = 0;
ac27a0ec
DK		 3045 smp_rmb();
3046
3047 /*
5e70030d
BP		 3048 * Compute the overhead (FS structures). This is constant
3049 * for a given filesystem unless the number of block groups
3050 * changes so we cache the previous value until it does.
ac27a0ec
DK		 3051 */
3052
3053 /*
3054 * All of the blocks before first_data_block are
3055 * overhead
3056 */
3057 overhead = le32_to_cpu(es->s_first_data_block);
3058
3059 /*
3060 * Add the overhead attributed to the superblock and
3061 * block group descriptors. If the sparse superblocks
3062 * feature is turned on, then not all groups have this.
3063 */
3064 for (i = 0; i < ngroups; i++) {
617ba13b
MC		 3065 overhead += ext4_bg_has_super(sb, i) +
3066 ext4_bg_num_gdb(sb, i);
ac27a0ec
DK		 3067 cond_resched();
3068 }
3069
3070 /*
3071 * Every block group has an inode bitmap, a block
3072 * bitmap, and an inode table.
3073 */
5e70030d
BP		 3074 overhead += ngroups * (2 + sbi->s_itb_per_group);
3075 sbi->s_overhead_last = overhead;
3076 smp_wmb();
6bc9feff 3077 sbi->s_blocks_last = ext4_blocks_count(es);
ac27a0ec
DK		 3078 }
3079
617ba13b 3080 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 3081 buf->f_bsize = sb->s_blocksize;
5e70030d 3082 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
52d9f3b4 3083 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter);
308ba3ec 3084 ext4_free_blocks_count_set(es, buf->f_bfree);
bd81d8ee
LV		 3085 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3086 if (buf->f_bfree < ext4_r_blocks_count(es))
ac27a0ec
DK		 3087 buf->f_bavail = 0;
3088 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 3089 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
5e70030d 3090 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
617ba13b 3091 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 3092 fsid = le64_to_cpup((void *)es->s_uuid) ^
3093 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3094 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3095 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
ac27a0ec
DK		 3096 return 0;
3097}
3098
3099/* Helper function for writing quotas on sync - we need to start transaction before quota file
3100 * is locked for write. Otherwise the are possible deadlocks:
3101 * Process 1 Process 2
617ba13b 3102 * ext4_create() quota_sync()
dab291af 3103 * jbd2_journal_start() write_dquot()
ac27a0ec 3104 * DQUOT_INIT() down(dqio_mutex)
dab291af 3105 * down(dqio_mutex) jbd2_journal_start()
ac27a0ec
DK		 3106 *
3107 */
3108
3109#ifdef CONFIG_QUOTA
3110
3111static inline struct inode *dquot_to_inode(struct dquot *dquot)
3112{
3113 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3114}
3115
617ba13b 3116static int ext4_dquot_initialize(struct inode *inode, int type)
ac27a0ec
DK		 3117{
3118 handle_t *handle;
3119 int ret, err;
3120
3121 /* We may create quota structure so we need to reserve enough blocks */
617ba13b 3122 handle = ext4_journal_start(inode, 2*EXT4_QUOTA_INIT_BLOCKS(inode->i_sb));
ac27a0ec
DK		 3123 if (IS_ERR(handle))
3124 return PTR_ERR(handle);
3125 ret = dquot_initialize(inode, type);
617ba13b 3126 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3127 if (!ret)
3128 ret = err;
3129 return ret;
3130}
3131
617ba13b 3132static int ext4_dquot_drop(struct inode *inode)
ac27a0ec
DK		 3133{
3134 handle_t *handle;
3135 int ret, err;
3136
3137 /* We may delete quota structure so we need to reserve enough blocks */
617ba13b 3138 handle = ext4_journal_start(inode, 2*EXT4_QUOTA_DEL_BLOCKS(inode->i_sb));
2887df13
JK		 3139 if (IS_ERR(handle)) {
3140 /*
3141 * We call dquot_drop() anyway to at least release references
3142 * to quota structures so that umount does not hang.
3143 */
3144 dquot_drop(inode);
ac27a0ec 3145 return PTR_ERR(handle);
2887df13 3146 }
ac27a0ec 3147 ret = dquot_drop(inode);
617ba13b 3148 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3149 if (!ret)
3150 ret = err;
3151 return ret;
3152}
3153
617ba13b 3154static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 3155{
3156 int ret, err;
3157 handle_t *handle;
3158 struct inode *inode;
3159
3160 inode = dquot_to_inode(dquot);
617ba13b
MC		 3161 handle = ext4_journal_start(inode,
3162 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 3163 if (IS_ERR(handle))
3164 return PTR_ERR(handle);
3165 ret = dquot_commit(dquot);
617ba13b 3166 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3167 if (!ret)
3168 ret = err;
3169 return ret;
3170}
3171
617ba13b 3172static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 3173{
3174 int ret, err;
3175 handle_t *handle;
3176
617ba13b
MC		 3177 handle = ext4_journal_start(dquot_to_inode(dquot),
3178 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 3179 if (IS_ERR(handle))
3180 return PTR_ERR(handle);
3181 ret = dquot_acquire(dquot);
617ba13b 3182 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3183 if (!ret)
3184 ret = err;
3185 return ret;
3186}
3187
617ba13b 3188static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 3189{
3190 int ret, err;
3191 handle_t *handle;
3192
617ba13b
MC		 3193 handle = ext4_journal_start(dquot_to_inode(dquot),
3194 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 3195 if (IS_ERR(handle)) {
3196 /* Release dquot anyway to avoid endless cycle in dqput() */
3197 dquot_release(dquot);
ac27a0ec 3198 return PTR_ERR(handle);
9c3013e9 3199 }
ac27a0ec 3200 ret = dquot_release(dquot);
617ba13b 3201 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3202 if (!ret)
3203 ret = err;
3204 return ret;
3205}
3206
617ba13b 3207static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec 3208{
2c8be6b2 3209 /* Are we journaling quotas? */
617ba13b
MC		 3210 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3211 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
ac27a0ec 3212 dquot_mark_dquot_dirty(dquot);
617ba13b 3213 return ext4_write_dquot(dquot);
ac27a0ec
DK		 3214 } else {
3215 return dquot_mark_dquot_dirty(dquot);
3216 }
3217}
3218
617ba13b 3219static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 3220{
3221 int ret, err;
3222 handle_t *handle;
3223
3224 /* Data block + inode block */
617ba13b 3225 handle = ext4_journal_start(sb->s_root->d_inode, 2);
ac27a0ec
DK		 3226 if (IS_ERR(handle))
3227 return PTR_ERR(handle);
3228 ret = dquot_commit_info(sb, type);
617ba13b 3229 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3230 if (!ret)
3231 ret = err;
3232 return ret;
3233}
3234
3235/*
3236 * Turn on quotas during mount time - we need to find
3237 * the quota file and such...
3238 */
617ba13b 3239static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 3240{
617ba13b
MC		 3241 return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
3242 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 3243}
3244
3245/*
3246 * Standard function to be called on quota_on
3247 */
617ba13b 3248static int ext4_quota_on(struct super_block *sb, int type, int format_id,
6f28e087 3249 char *path, int remount)
ac27a0ec
DK		 3250{
3251 int err;
3252 struct nameidata nd;
3253
3254 if (!test_opt(sb, QUOTA))
3255 return -EINVAL;
0623543b
JK		 3256 /* When remounting, no checks are needed and in fact, path is NULL */
3257 if (remount)
6f28e087 3258 return vfs_quota_on(sb, type, format_id, path, remount);
0623543b 3259
ac27a0ec
DK		 3260 err = path_lookup(path, LOOKUP_FOLLOW, &nd);
3261 if (err)
3262 return err;
0623543b 3263
ac27a0ec 3264 /* Quotafile not on the same filesystem? */
4ac91378 3265 if (nd.path.mnt->mnt_sb != sb) {
1d957f9b 3266 path_put(&nd.path);
ac27a0ec
DK		 3267 return -EXDEV;
3268 }
0623543b
JK		 3269 /* Journaling quota? */
3270 if (EXT4_SB(sb)->s_qf_names[type]) {
3271 /* Quotafile not of fs root? */
3272 if (nd.path.dentry->d_parent->d_inode != sb->s_root->d_inode)
3273 printk(KERN_WARNING
3274 "EXT4-fs: Quota file not on filesystem root. "
3275 "Journaled quota will not work.\n");
3276 }
3277
3278 /*
3279 * When we journal data on quota file, we have to flush journal to see
3280 * all updates to the file when we bypass pagecache...
3281 */
3282 if (ext4_should_journal_data(nd.path.dentry->d_inode)) {
3283 /*
3284 * We don't need to lock updates but journal_flush() could
3285 * otherwise be livelocked...
3286 */
3287 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
3288 jbd2_journal_flush(EXT4_SB(sb)->s_journal);
3289 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3290 }
3291
1d957f9b 3292 path_put(&nd.path);
6f28e087 3293 return vfs_quota_on(sb, type, format_id, path, remount);
ac27a0ec
DK		 3294}
3295
3296/* Read data from quotafile - avoid pagecache and such because we cannot afford
3297 * acquiring the locks... As quota files are never truncated and quota code
3298 * itself serializes the operations (and noone else should touch the files)
3299 * we don't have to be afraid of races */
617ba13b 3300static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 3301 size_t len, loff_t off)
3302{
3303 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 3304 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 3305 int err = 0;
3306 int offset = off & (sb->s_blocksize - 1);
3307 int tocopy;
3308 size_t toread;
3309 struct buffer_head *bh;
3310 loff_t i_size = i_size_read(inode);
3311
3312 if (off > i_size)
3313 return 0;
3314 if (off+len > i_size)
3315 len = i_size-off;
3316 toread = len;
3317 while (toread > 0) {
3318 tocopy = sb->s_blocksize - offset < toread ?
3319 sb->s_blocksize - offset : toread;
617ba13b 3320 bh = ext4_bread(NULL, inode, blk, 0, &err);
ac27a0ec
DK		 3321 if (err)
3322 return err;
3323 if (!bh) /* A hole? */
3324 memset(data, 0, tocopy);
3325 else
3326 memcpy(data, bh->b_data+offset, tocopy);
3327 brelse(bh);
3328 offset = 0;
3329 toread -= tocopy;
3330 data += tocopy;
3331 blk++;
3332 }
3333 return len;
3334}
3335
3336/* Write to quotafile (we know the transaction is already started and has
3337 * enough credits) */
617ba13b 3338static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 3339 const char *data, size_t len, loff_t off)
3340{
3341 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 3342 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 3343 int err = 0;
3344 int offset = off & (sb->s_blocksize - 1);
3345 int tocopy;
617ba13b 3346 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
ac27a0ec
DK		 3347 size_t towrite = len;
3348 struct buffer_head *bh;
3349 handle_t *handle = journal_current_handle();
3350
9c3013e9
JK		 3351 if (!handle) {
3352 printk(KERN_WARNING "EXT4-fs: Quota write (off=%Lu, len=%Lu)"
3353 " cancelled because transaction is not started.\n",
3354 (unsigned long long)off, (unsigned long long)len);
3355 return -EIO;
3356 }
ac27a0ec
DK		 3357 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
3358 while (towrite > 0) {
3359 tocopy = sb->s_blocksize - offset < towrite ?
3360 sb->s_blocksize - offset : towrite;
617ba13b 3361 bh = ext4_bread(handle, inode, blk, 1, &err);
ac27a0ec
DK		 3362 if (!bh)
3363 goto out;
3364 if (journal_quota) {
617ba13b 3365 err = ext4_journal_get_write_access(handle, bh);
ac27a0ec
DK		 3366 if (err) {
3367 brelse(bh);
3368 goto out;
3369 }
3370 }
3371 lock_buffer(bh);
3372 memcpy(bh->b_data+offset, data, tocopy);
3373 flush_dcache_page(bh->b_page);
3374 unlock_buffer(bh);
3375 if (journal_quota)
617ba13b 3376 err = ext4_journal_dirty_metadata(handle, bh);
ac27a0ec
DK		 3377 else {
3378 /* Always do at least ordered writes for quotas */
617ba13b 3379 err = ext4_journal_dirty_data(handle, bh);
ac27a0ec
DK		 3380 mark_buffer_dirty(bh);
3381 }
3382 brelse(bh);
3383 if (err)
3384 goto out;
3385 offset = 0;
3386 towrite -= tocopy;
3387 data += tocopy;
3388 blk++;
3389 }
3390out:
4d04e4fb
JK		 3391 if (len == towrite) {
3392 mutex_unlock(&inode->i_mutex);
ac27a0ec 3393 return err;
4d04e4fb 3394 }
ac27a0ec
DK		 3395 if (inode->i_size < off+len-towrite) {
3396 i_size_write(inode, off+len-towrite);
617ba13b 3397 EXT4_I(inode)->i_disksize = inode->i_size;
ac27a0ec 3398 }
ac27a0ec 3399 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
617ba13b 3400 ext4_mark_inode_dirty(handle, inode);
ac27a0ec
DK		 3401 mutex_unlock(&inode->i_mutex);
3402 return len - towrite;
3403}
3404
3405#endif
3406
617ba13b 3407static int ext4_get_sb(struct file_system_type *fs_type,
ac27a0ec
DK		 3408 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3409{
617ba13b 3410 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
ac27a0ec
DK		 3411}
3412
617ba13b 3413static struct file_system_type ext4dev_fs_type = {
ac27a0ec 3414 .owner = THIS_MODULE,
617ba13b
MC		 3415 .name = "ext4dev",
3416 .get_sb = ext4_get_sb,
ac27a0ec
DK		 3417 .kill_sb = kill_block_super,
3418 .fs_flags = FS_REQUIRES_DEV,
3419};
3420
617ba13b 3421static int __init init_ext4_fs(void)
ac27a0ec 3422{
c9de560d
AT		 3423 int err;
3424
3425 err = init_ext4_mballoc();
ac27a0ec
DK		 3426 if (err)
3427 return err;
c9de560d
AT		 3428
3429 err = init_ext4_xattr();
3430 if (err)
3431 goto out2;
ac27a0ec
DK		 3432 err = init_inodecache();
3433 if (err)
3434 goto out1;
63f57933 3435 err = register_filesystem(&ext4dev_fs_type);
ac27a0ec
DK		 3436 if (err)
3437 goto out;
3438 return 0;
3439out:
3440 destroy_inodecache();
3441out1:
617ba13b 3442 exit_ext4_xattr();
c9de560d
AT		 3443out2:
3444 exit_ext4_mballoc();
ac27a0ec
DK		 3445 return err;
3446}
3447
617ba13b 3448static void __exit exit_ext4_fs(void)
ac27a0ec 3449{
617ba13b 3450 unregister_filesystem(&ext4dev_fs_type);
ac27a0ec 3451 destroy_inodecache();
617ba13b 3452 exit_ext4_xattr();
c9de560d 3453 exit_ext4_mballoc();
ac27a0ec
DK		 3454}
3455
3456MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
617ba13b 3457MODULE_DESCRIPTION("Fourth Extended Filesystem with extents");
ac27a0ec 3458MODULE_LICENSE("GPL");
617ba13b
MC		 3459module_init(init_ext4_fs)
3460module_exit(exit_ext4_fs)
ubuntu-bionic-kernel mirror