projects / GitHub / mt8127 / android_kernel_alcatel_ttab.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
ext4: fdatasync should skip metadata writeout when overwriting
[GitHub/mt8127/android_kernel_alcatel_ttab.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>
617ba13b 24#include <linux/ext4_fs.h>
dab291af 25#include <linux/ext4_jbd2.h>
ac27a0ec
DK		 26#include <linux/slab.h>
27#include <linux/init.h>
28#include <linux/blkdev.h>
29#include <linux/parser.h>
30#include <linux/smp_lock.h>
31#include <linux/buffer_head.h>
a5694255 32#include <linux/exportfs.h>
ac27a0ec
DK		 33#include <linux/vfs.h>
34#include <linux/random.h>
35#include <linux/mount.h>
36#include <linux/namei.h>
37#include <linux/quotaops.h>
38#include <linux/seq_file.h>
1330593e 39#include <linux/log2.h>
717d50e4 40#include <linux/crc16.h>
ac27a0ec
DK		 41
42#include <asm/uaccess.h>
43
44#include "xattr.h"
45#include "acl.h"
46#include "namei.h"
717d50e4 47#include "group.h"
ac27a0ec 48
617ba13b 49static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
ac27a0ec 50 unsigned long journal_devnum);
617ba13b 51static int ext4_create_journal(struct super_block *, struct ext4_super_block *,
ac27a0ec 52 unsigned int);
617ba13b
MC		 53static void ext4_commit_super (struct super_block * sb,
54 struct ext4_super_block * es,
ac27a0ec 55 int sync);
617ba13b
MC		 56static void ext4_mark_recovery_complete(struct super_block * sb,
57 struct ext4_super_block * es);
58static void ext4_clear_journal_err(struct super_block * sb,
59 struct ext4_super_block * es);
60static int ext4_sync_fs(struct super_block *sb, int wait);
61static const char *ext4_decode_error(struct super_block * sb, int errno,
ac27a0ec 62 char nbuf[16]);
617ba13b
MC		 63static int ext4_remount (struct super_block * sb, int * flags, char * data);
64static int ext4_statfs (struct dentry * dentry, struct kstatfs * buf);
65static void ext4_unlockfs(struct super_block *sb);
66static void ext4_write_super (struct super_block * sb);
67static void ext4_write_super_lockfs(struct super_block *sb);
ac27a0ec 68
bd81d8ee 69
8fadc143
AR		 70ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
71 struct ext4_group_desc *bg)
bd81d8ee 72{
3a14589c 73 return le32_to_cpu(bg->bg_block_bitmap_lo) |
8fadc143 74 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
3a14589c 75 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 76}
77
8fadc143
AR		 78ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
79 struct ext4_group_desc *bg)
bd81d8ee 80{
5272f837 81 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
8fadc143 82 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
5272f837 83 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 84}
85
8fadc143
AR		 86ext4_fsblk_t ext4_inode_table(struct super_block *sb,
87 struct ext4_group_desc *bg)
bd81d8ee 88{
5272f837 89 return le32_to_cpu(bg->bg_inode_table_lo) |
8fadc143 90 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
5272f837 91 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
bd81d8ee
LV		 92}
93
8fadc143
AR		 94void ext4_block_bitmap_set(struct super_block *sb,
95 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 96{
3a14589c 97 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 98 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
99 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 100}
101
8fadc143
AR		 102void ext4_inode_bitmap_set(struct super_block *sb,
103 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 104{
5272f837 105 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 106 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
107 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 108}
109
8fadc143
AR		 110void ext4_inode_table_set(struct super_block *sb,
111 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 112{
5272f837 113 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 114 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
115 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 116}
117
ac27a0ec 118/*
dab291af 119 * Wrappers for jbd2_journal_start/end.
ac27a0ec
DK		 120 *
121 * The only special thing we need to do here is to make sure that all
122 * journal_end calls result in the superblock being marked dirty, so
123 * that sync() will call the filesystem's write_super callback if
124 * appropriate.
125 */
617ba13b 126handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
ac27a0ec
DK		 127{
128 journal_t *journal;
129
130 if (sb->s_flags & MS_RDONLY)
131 return ERR_PTR(-EROFS);
132
133 /* Special case here: if the journal has aborted behind our
134 * backs (eg. EIO in the commit thread), then we still need to
135 * take the FS itself readonly cleanly. */
617ba13b 136 journal = EXT4_SB(sb)->s_journal;
ac27a0ec 137 if (is_journal_aborted(journal)) {
617ba13b 138 ext4_abort(sb, __FUNCTION__,
ac27a0ec
DK		 139 "Detected aborted journal");
140 return ERR_PTR(-EROFS);
141 }
142
dab291af 143 return jbd2_journal_start(journal, nblocks);
ac27a0ec
DK		 144}
145
146/*
147 * The only special thing we need to do here is to make sure that all
dab291af 148 * jbd2_journal_stop calls result in the superblock being marked dirty, so
ac27a0ec
DK		 149 * that sync() will call the filesystem's write_super callback if
150 * appropriate.
151 */
617ba13b 152int __ext4_journal_stop(const char *where, handle_t *handle)
ac27a0ec
DK		 153{
154 struct super_block *sb;
155 int err;
156 int rc;
157
158 sb = handle->h_transaction->t_journal->j_private;
159 err = handle->h_err;
dab291af 160 rc = jbd2_journal_stop(handle);
ac27a0ec
DK		 161
162 if (!err)
163 err = rc;
164 if (err)
617ba13b 165 __ext4_std_error(sb, where, err);
ac27a0ec
DK		 166 return err;
167}
168
617ba13b 169void ext4_journal_abort_handle(const char *caller, const char *err_fn,
ac27a0ec
DK		 170 struct buffer_head *bh, handle_t *handle, int err)
171{
172 char nbuf[16];
617ba13b 173 const char *errstr = ext4_decode_error(NULL, err, nbuf);
ac27a0ec
DK		 174
175 if (bh)
176 BUFFER_TRACE(bh, "abort");
177
178 if (!handle->h_err)
179 handle->h_err = err;
180
181 if (is_handle_aborted(handle))
182 return;
183
184 printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
185 caller, errstr, err_fn);
186
dab291af 187 jbd2_journal_abort_handle(handle);
ac27a0ec
DK		 188}
189
190/* Deal with the reporting of failure conditions on a filesystem such as
191 * inconsistencies detected or read IO failures.
192 *
193 * On ext2, we can store the error state of the filesystem in the
617ba13b 194 * superblock. That is not possible on ext4, because we may have other
ac27a0ec
DK		 195 * write ordering constraints on the superblock which prevent us from
196 * writing it out straight away; and given that the journal is about to
197 * be aborted, we can't rely on the current, or future, transactions to
198 * write out the superblock safely.
199 *
dab291af 200 * We'll just use the jbd2_journal_abort() error code to record an error in
ac27a0ec
DK		 201 * the journal instead. On recovery, the journal will compain about
202 * that error until we've noted it down and cleared it.
203 */
204
617ba13b 205static void ext4_handle_error(struct super_block *sb)
ac27a0ec 206{
617ba13b 207 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 208
617ba13b
MC		 209 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
210 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
ac27a0ec
DK		 211
212 if (sb->s_flags & MS_RDONLY)
213 return;
214
215 if (!test_opt (sb, ERRORS_CONT)) {
617ba13b 216 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 217
617ba13b 218 EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
ac27a0ec 219 if (journal)
dab291af 220 jbd2_journal_abort(journal, -EIO);
ac27a0ec
DK		 221 }
222 if (test_opt (sb, ERRORS_RO)) {
223 printk (KERN_CRIT "Remounting filesystem read-only\n");
224 sb->s_flags |= MS_RDONLY;
225 }
617ba13b 226 ext4_commit_super(sb, es, 1);
ac27a0ec 227 if (test_opt(sb, ERRORS_PANIC))
617ba13b 228 panic("EXT4-fs (device %s): panic forced after error\n",
ac27a0ec
DK		 229 sb->s_id);
230}
231
617ba13b 232void ext4_error (struct super_block * sb, const char * function,
ac27a0ec
DK		 233 const char * fmt, ...)
234{
235 va_list args;
236
237 va_start(args, fmt);
617ba13b 238 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ",sb->s_id, function);
ac27a0ec
DK		 239 vprintk(fmt, args);
240 printk("\n");
241 va_end(args);
242
617ba13b 243 ext4_handle_error(sb);
ac27a0ec
DK		 244}
245
617ba13b 246static const char *ext4_decode_error(struct super_block * sb, int errno,
ac27a0ec
DK		 247 char nbuf[16])
248{
249 char *errstr = NULL;
250
251 switch (errno) {
252 case -EIO:
253 errstr = "IO failure";
254 break;
255 case -ENOMEM:
256 errstr = "Out of memory";
257 break;
258 case -EROFS:
dab291af 259 if (!sb || EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT)
ac27a0ec
DK		 260 errstr = "Journal has aborted";
261 else
262 errstr = "Readonly filesystem";
263 break;
264 default:
265 /* If the caller passed in an extra buffer for unknown
266 * errors, textualise them now. Else we just return
267 * NULL. */
268 if (nbuf) {
269 /* Check for truncated error codes... */
270 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
271 errstr = nbuf;
272 }
273 break;
274 }
275
276 return errstr;
277}
278
617ba13b 279/* __ext4_std_error decodes expected errors from journaling functions
ac27a0ec
DK		 280 * automatically and invokes the appropriate error response. */
281
617ba13b 282void __ext4_std_error (struct super_block * sb, const char * function,
ac27a0ec
DK		 283 int errno)
284{
285 char nbuf[16];
286 const char *errstr;
287
288 /* Special case: if the error is EROFS, and we're not already
289 * inside a transaction, then there's really no point in logging
290 * an error. */
291 if (errno == -EROFS && journal_current_handle() == NULL &&
292 (sb->s_flags & MS_RDONLY))
293 return;
294
617ba13b
MC		 295 errstr = ext4_decode_error(sb, errno, nbuf);
296 printk (KERN_CRIT "EXT4-fs error (device %s) in %s: %s\n",
ac27a0ec
DK		 297 sb->s_id, function, errstr);
298
617ba13b 299 ext4_handle_error(sb);
ac27a0ec
DK		 300}
301
302/*
617ba13b 303 * ext4_abort is a much stronger failure handler than ext4_error. The
ac27a0ec
DK		 304 * abort function may be used to deal with unrecoverable failures such
305 * as journal IO errors or ENOMEM at a critical moment in log management.
306 *
307 * We unconditionally force the filesystem into an ABORT|READONLY state,
308 * unless the error response on the fs has been set to panic in which
309 * case we take the easy way out and panic immediately.
310 */
311
617ba13b 312void ext4_abort (struct super_block * sb, const char * function,
ac27a0ec
DK		 313 const char * fmt, ...)
314{
315 va_list args;
316
617ba13b 317 printk (KERN_CRIT "ext4_abort called.\n");
ac27a0ec
DK		 318
319 va_start(args, fmt);
617ba13b 320 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ",sb->s_id, function);
ac27a0ec
DK		 321 vprintk(fmt, args);
322 printk("\n");
323 va_end(args);
324
325 if (test_opt(sb, ERRORS_PANIC))
617ba13b 326 panic("EXT4-fs panic from previous error\n");
ac27a0ec
DK		 327
328 if (sb->s_flags & MS_RDONLY)
329 return;
330
331 printk(KERN_CRIT "Remounting filesystem read-only\n");
617ba13b 332 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
ac27a0ec 333 sb->s_flags |= MS_RDONLY;
617ba13b 334 EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
dab291af 335 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
ac27a0ec
DK		 336}
337
617ba13b 338void ext4_warning (struct super_block * sb, const char * function,
ac27a0ec
DK		 339 const char * fmt, ...)
340{
341 va_list args;
342
343 va_start(args, fmt);
617ba13b 344 printk(KERN_WARNING "EXT4-fs warning (device %s): %s: ",
ac27a0ec
DK		 345 sb->s_id, function);
346 vprintk(fmt, args);
347 printk("\n");
348 va_end(args);
349}
350
617ba13b 351void ext4_update_dynamic_rev(struct super_block *sb)
ac27a0ec 352{
617ba13b 353 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 354
617ba13b 355 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
ac27a0ec
DK		 356 return;
357
617ba13b 358 ext4_warning(sb, __FUNCTION__,
ac27a0ec
DK		 359 "updating to rev %d because of new feature flag, "
360 "running e2fsck is recommended",
617ba13b 361 EXT4_DYNAMIC_REV);
ac27a0ec 362
617ba13b
MC		 363 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
364 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
365 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
ac27a0ec
DK		 366 /* leave es->s_feature_*compat flags alone */
367 /* es->s_uuid will be set by e2fsck if empty */
368
369 /*
370 * The rest of the superblock fields should be zero, and if not it
371 * means they are likely already in use, so leave them alone. We
372 * can leave it up to e2fsck to clean up any inconsistencies there.
373 */
374}
375
99e6f829
AK		 376int ext4_update_compat_feature(handle_t *handle,
377 struct super_block *sb, __u32 compat)
378{
379 int err = 0;
380 if (!EXT4_HAS_COMPAT_FEATURE(sb, compat)) {
381 err = ext4_journal_get_write_access(handle,
382 EXT4_SB(sb)->s_sbh);
383 if (err)
384 return err;
385 EXT4_SET_COMPAT_FEATURE(sb, compat);
386 sb->s_dirt = 1;
387 handle->h_sync = 1;
388 BUFFER_TRACE(EXT4_SB(sb)->s_sbh,
389 "call ext4_journal_dirty_met adata");
390 err = ext4_journal_dirty_metadata(handle,
391 EXT4_SB(sb)->s_sbh);
392 }
393 return err;
394}
395
396int ext4_update_rocompat_feature(handle_t *handle,
397 struct super_block *sb, __u32 rocompat)
398{
399 int err = 0;
400 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, rocompat)) {
401 err = ext4_journal_get_write_access(handle,
402 EXT4_SB(sb)->s_sbh);
403 if (err)
404 return err;
405 EXT4_SET_RO_COMPAT_FEATURE(sb, rocompat);
406 sb->s_dirt = 1;
407 handle->h_sync = 1;
408 BUFFER_TRACE(EXT4_SB(sb)->s_sbh,
409 "call ext4_journal_dirty_met adata");
410 err = ext4_journal_dirty_metadata(handle,
411 EXT4_SB(sb)->s_sbh);
412 }
413 return err;
414}
415
416int ext4_update_incompat_feature(handle_t *handle,
417 struct super_block *sb, __u32 incompat)
418{
419 int err = 0;
420 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, incompat)) {
421 err = ext4_journal_get_write_access(handle,
422 EXT4_SB(sb)->s_sbh);
423 if (err)
424 return err;
425 EXT4_SET_INCOMPAT_FEATURE(sb, incompat);
426 sb->s_dirt = 1;
427 handle->h_sync = 1;
428 BUFFER_TRACE(EXT4_SB(sb)->s_sbh,
429 "call ext4_journal_dirty_met adata");
430 err = ext4_journal_dirty_metadata(handle,
431 EXT4_SB(sb)->s_sbh);
432 }
433 return err;
434}
435
ac27a0ec
DK		 436/*
437 * Open the external journal device
438 */
617ba13b 439static struct block_device *ext4_blkdev_get(dev_t dev)
ac27a0ec
DK		 440{
441 struct block_device *bdev;
442 char b[BDEVNAME_SIZE];
443
444 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
445 if (IS_ERR(bdev))
446 goto fail;
447 return bdev;
448
449fail:
617ba13b 450 printk(KERN_ERR "EXT4: failed to open journal device %s: %ld\n",
ac27a0ec
DK		 451 __bdevname(dev, b), PTR_ERR(bdev));
452 return NULL;
453}
454
455/*
456 * Release the journal device
457 */
617ba13b 458static int ext4_blkdev_put(struct block_device *bdev)
ac27a0ec
DK		 459{
460 bd_release(bdev);
461 return blkdev_put(bdev);
462}
463
617ba13b 464static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
ac27a0ec
DK		 465{
466 struct block_device *bdev;
467 int ret = -ENODEV;
468
469 bdev = sbi->journal_bdev;
470 if (bdev) {
617ba13b 471 ret = ext4_blkdev_put(bdev);
ac27a0ec
DK		 472 sbi->journal_bdev = NULL;
473 }
474 return ret;
475}
476
477static inline struct inode *orphan_list_entry(struct list_head *l)
478{
617ba13b 479 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
ac27a0ec
DK		 480}
481
617ba13b 482static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
ac27a0ec
DK		 483{
484 struct list_head *l;
485
486 printk(KERN_ERR "sb orphan head is %d\n",
487 le32_to_cpu(sbi->s_es->s_last_orphan));
488
489 printk(KERN_ERR "sb_info orphan list:\n");
490 list_for_each(l, &sbi->s_orphan) {
491 struct inode *inode = orphan_list_entry(l);
492 printk(KERN_ERR " "
493 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
494 inode->i_sb->s_id, inode->i_ino, inode,
495 inode->i_mode, inode->i_nlink,
496 NEXT_ORPHAN(inode));
497 }
498}
499
617ba13b 500static void ext4_put_super (struct super_block * sb)
ac27a0ec 501{
617ba13b
MC		 502 struct ext4_sb_info *sbi = EXT4_SB(sb);
503 struct ext4_super_block *es = sbi->s_es;
ac27a0ec
DK		 504 int i;
505
c9de560d 506 ext4_mb_release(sb);
a86c6181 507 ext4_ext_release(sb);
617ba13b 508 ext4_xattr_put_super(sb);
dab291af 509 jbd2_journal_destroy(sbi->s_journal);
ac27a0ec 510 if (!(sb->s_flags & MS_RDONLY)) {
617ba13b 511 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec
DK		 512 es->s_state = cpu_to_le16(sbi->s_mount_state);
513 BUFFER_TRACE(sbi->s_sbh, "marking dirty");
514 mark_buffer_dirty(sbi->s_sbh);
617ba13b 515 ext4_commit_super(sb, es, 1);
ac27a0ec
DK		 516 }
517
518 for (i = 0; i < sbi->s_gdb_count; i++)
519 brelse(sbi->s_group_desc[i]);
520 kfree(sbi->s_group_desc);
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 }
733 if (test_opt(sb, BARRIER))
734 seq_puts(seq, ",barrier=1");
735 if (test_opt(sb, NOBH))
736 seq_puts(seq, ",nobh");
737 if (!test_opt(sb, EXTENTS))
738 seq_puts(seq, ",noextents");
3dbd0ede
AK		 739 if (!test_opt(sb, MBALLOC))
740 seq_puts(seq, ",nomballoc");
25ec56b5
JNC		 741 if (test_opt(sb, I_VERSION))
742 seq_puts(seq, ",i_version");
ac27a0ec 743
cb45bbe4
MS		 744 if (sbi->s_stripe)
745 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
aa22df2c
AK		 746 /*
747 * journal mode get enabled in different ways
748 * So just print the value even if we didn't specify it
749 */
617ba13b 750 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
ac27a0ec 751 seq_puts(seq, ",data=journal");
617ba13b 752 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
ac27a0ec 753 seq_puts(seq, ",data=ordered");
617ba13b 754 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
ac27a0ec
DK		 755 seq_puts(seq, ",data=writeback");
756
617ba13b 757 ext4_show_quota_options(seq, sb);
ac27a0ec
DK		 758 return 0;
759}
760
761
1b961ac0
CH		 762static struct inode *ext4_nfs_get_inode(struct super_block *sb,
763 u64 ino, u32 generation)
ac27a0ec 764{
ac27a0ec 765 struct inode *inode;
ac27a0ec 766
617ba13b 767 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
ac27a0ec 768 return ERR_PTR(-ESTALE);
617ba13b 769 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
ac27a0ec
DK		 770 return ERR_PTR(-ESTALE);
771
772 /* iget isn't really right if the inode is currently unallocated!!
773 *
617ba13b 774 * ext4_read_inode will return a bad_inode if the inode had been
ac27a0ec
DK		 775 * deleted, so we should be safe.
776 *
777 * Currently we don't know the generation for parent directory, so
778 * a generation of 0 means "accept any"
779 */
1d1fe1ee
DH		 780 inode = ext4_iget(sb, ino);
781 if (IS_ERR(inode))
782 return ERR_CAST(inode);
783 if (generation && inode->i_generation != generation) {
ac27a0ec
DK		 784 iput(inode);
785 return ERR_PTR(-ESTALE);
786 }
1b961ac0
CH		 787
788 return inode;
789}
790
791static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
792 int fh_len, int fh_type)
793{
794 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
795 ext4_nfs_get_inode);
796}
797
798static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
799 int fh_len, int fh_type)
800{
801 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
802 ext4_nfs_get_inode);
ac27a0ec
DK		 803}
804
805#ifdef CONFIG_QUOTA
806#define QTYPE2NAME(t) ((t)==USRQUOTA?"user":"group")
807#define QTYPE2MOPT(on, t) ((t)==USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
808
617ba13b
MC		 809static int ext4_dquot_initialize(struct inode *inode, int type);
810static int ext4_dquot_drop(struct inode *inode);
811static int ext4_write_dquot(struct dquot *dquot);
812static int ext4_acquire_dquot(struct dquot *dquot);
813static int ext4_release_dquot(struct dquot *dquot);
814static int ext4_mark_dquot_dirty(struct dquot *dquot);
815static int ext4_write_info(struct super_block *sb, int type);
6f28e087
JK		 816static int ext4_quota_on(struct super_block *sb, int type, int format_id,
817 char *path, int remount);
617ba13b
MC		 818static int ext4_quota_on_mount(struct super_block *sb, int type);
819static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 820 size_t len, loff_t off);
617ba13b 821static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 822 const char *data, size_t len, loff_t off);
823
617ba13b
MC		 824static struct dquot_operations ext4_quota_operations = {
825 .initialize = ext4_dquot_initialize,
826 .drop = ext4_dquot_drop,
ac27a0ec
DK		 827 .alloc_space = dquot_alloc_space,
828 .alloc_inode = dquot_alloc_inode,
829 .free_space = dquot_free_space,
830 .free_inode = dquot_free_inode,
831 .transfer = dquot_transfer,
617ba13b
MC		 832 .write_dquot = ext4_write_dquot,
833 .acquire_dquot = ext4_acquire_dquot,
834 .release_dquot = ext4_release_dquot,
835 .mark_dirty = ext4_mark_dquot_dirty,
836 .write_info = ext4_write_info
ac27a0ec
DK		 837};
838
617ba13b
MC		 839static struct quotactl_ops ext4_qctl_operations = {
840 .quota_on = ext4_quota_on,
ac27a0ec
DK		 841 .quota_off = vfs_quota_off,
842 .quota_sync = vfs_quota_sync,
843 .get_info = vfs_get_dqinfo,
844 .set_info = vfs_set_dqinfo,
845 .get_dqblk = vfs_get_dqblk,
846 .set_dqblk = vfs_set_dqblk
847};
848#endif
849
ee9b6d61 850static const struct super_operations ext4_sops = {
617ba13b
MC		 851 .alloc_inode = ext4_alloc_inode,
852 .destroy_inode = ext4_destroy_inode,
617ba13b
MC		 853 .write_inode = ext4_write_inode,
854 .dirty_inode = ext4_dirty_inode,
855 .delete_inode = ext4_delete_inode,
856 .put_super = ext4_put_super,
857 .write_super = ext4_write_super,
858 .sync_fs = ext4_sync_fs,
859 .write_super_lockfs = ext4_write_super_lockfs,
860 .unlockfs = ext4_unlockfs,
861 .statfs = ext4_statfs,
862 .remount_fs = ext4_remount,
863 .clear_inode = ext4_clear_inode,
864 .show_options = ext4_show_options,
ac27a0ec 865#ifdef CONFIG_QUOTA
617ba13b
MC		 866 .quota_read = ext4_quota_read,
867 .quota_write = ext4_quota_write,
ac27a0ec
DK		 868#endif
869};
870
39655164 871static const struct export_operations ext4_export_ops = {
1b961ac0
CH		 872 .fh_to_dentry = ext4_fh_to_dentry,
873 .fh_to_parent = ext4_fh_to_parent,
617ba13b 874 .get_parent = ext4_get_parent,
ac27a0ec
DK		 875};
876
877enum {
878 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
879 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
880 Opt_nouid32, Opt_nocheck, Opt_debug, Opt_oldalloc, Opt_orlov,
881 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
882 Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh,
883 Opt_commit, Opt_journal_update, Opt_journal_inum, Opt_journal_dev,
818d276c 884 Opt_journal_checksum, Opt_journal_async_commit,
ac27a0ec
DK		 885 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
886 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
887 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
888 Opt_ignore, Opt_barrier, Opt_err, Opt_resize, Opt_usrquota,
25ec56b5 889 Opt_grpquota, Opt_extents, Opt_noextents, Opt_i_version,
c9de560d 890 Opt_mballoc, Opt_nomballoc, Opt_stripe,
ac27a0ec
DK		 891};
892
893static match_table_t tokens = {
894 {Opt_bsd_df, "bsddf"},
895 {Opt_minix_df, "minixdf"},
896 {Opt_grpid, "grpid"},
897 {Opt_grpid, "bsdgroups"},
898 {Opt_nogrpid, "nogrpid"},
899 {Opt_nogrpid, "sysvgroups"},
900 {Opt_resgid, "resgid=%u"},
901 {Opt_resuid, "resuid=%u"},
902 {Opt_sb, "sb=%u"},
903 {Opt_err_cont, "errors=continue"},
904 {Opt_err_panic, "errors=panic"},
905 {Opt_err_ro, "errors=remount-ro"},
906 {Opt_nouid32, "nouid32"},
907 {Opt_nocheck, "nocheck"},
908 {Opt_nocheck, "check=none"},
909 {Opt_debug, "debug"},
910 {Opt_oldalloc, "oldalloc"},
911 {Opt_orlov, "orlov"},
912 {Opt_user_xattr, "user_xattr"},
913 {Opt_nouser_xattr, "nouser_xattr"},
914 {Opt_acl, "acl"},
915 {Opt_noacl, "noacl"},
916 {Opt_reservation, "reservation"},
917 {Opt_noreservation, "noreservation"},
918 {Opt_noload, "noload"},
919 {Opt_nobh, "nobh"},
920 {Opt_bh, "bh"},
921 {Opt_commit, "commit=%u"},
922 {Opt_journal_update, "journal=update"},
923 {Opt_journal_inum, "journal=%u"},
924 {Opt_journal_dev, "journal_dev=%u"},
818d276c
GS		 925 {Opt_journal_checksum, "journal_checksum"},
926 {Opt_journal_async_commit, "journal_async_commit"},
ac27a0ec
DK		 927 {Opt_abort, "abort"},
928 {Opt_data_journal, "data=journal"},
929 {Opt_data_ordered, "data=ordered"},
930 {Opt_data_writeback, "data=writeback"},
931 {Opt_offusrjquota, "usrjquota="},
932 {Opt_usrjquota, "usrjquota=%s"},
933 {Opt_offgrpjquota, "grpjquota="},
934 {Opt_grpjquota, "grpjquota=%s"},
935 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
936 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
937 {Opt_grpquota, "grpquota"},
938 {Opt_noquota, "noquota"},
939 {Opt_quota, "quota"},
940 {Opt_usrquota, "usrquota"},
941 {Opt_barrier, "barrier=%u"},
a86c6181 942 {Opt_extents, "extents"},
1e2462f9 943 {Opt_noextents, "noextents"},
25ec56b5 944 {Opt_i_version, "i_version"},
c9de560d
AT		 945 {Opt_mballoc, "mballoc"},
946 {Opt_nomballoc, "nomballoc"},
947 {Opt_stripe, "stripe=%u"},
ac27a0ec
DK		 948 {Opt_err, NULL},
949 {Opt_resize, "resize"},
950};
951
617ba13b 952static ext4_fsblk_t get_sb_block(void **data)
ac27a0ec 953{
617ba13b 954 ext4_fsblk_t sb_block;
ac27a0ec
DK		 955 char *options = (char *) *data;
956
957 if (!options || strncmp(options, "sb=", 3) != 0)
958 return 1; /* Default location */
959 options += 3;
617ba13b 960 /*todo: use simple_strtoll with >32bit ext4 */
ac27a0ec
DK		 961 sb_block = simple_strtoul(options, &options, 0);
962 if (*options && *options != ',') {
617ba13b 963 printk("EXT4-fs: Invalid sb specification: %s\n",
ac27a0ec
DK		 964 (char *) *data);
965 return 1;
966 }
967 if (*options == ',')
968 options++;
969 *data = (void *) options;
970 return sb_block;
971}
972
973static int parse_options (char *options, struct super_block *sb,
974 unsigned int *inum, unsigned long *journal_devnum,
617ba13b 975 ext4_fsblk_t *n_blocks_count, int is_remount)
ac27a0ec 976{
617ba13b 977 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 978 char * p;
979 substring_t args[MAX_OPT_ARGS];
980 int data_opt = 0;
981 int option;
982#ifdef CONFIG_QUOTA
983 int qtype;
984 char *qname;
985#endif
986
987 if (!options)
988 return 1;
989
990 while ((p = strsep (&options, ",")) != NULL) {
991 int token;
992 if (!*p)
993 continue;
994
995 token = match_token(p, tokens, args);
996 switch (token) {
997 case Opt_bsd_df:
998 clear_opt (sbi->s_mount_opt, MINIX_DF);
999 break;
1000 case Opt_minix_df:
1001 set_opt (sbi->s_mount_opt, MINIX_DF);
1002 break;
1003 case Opt_grpid:
1004 set_opt (sbi->s_mount_opt, GRPID);
1005 break;
1006 case Opt_nogrpid:
1007 clear_opt (sbi->s_mount_opt, GRPID);
1008 break;
1009 case Opt_resuid:
1010 if (match_int(&args[0], &option))
1011 return 0;
1012 sbi->s_resuid = option;
1013 break;
1014 case Opt_resgid:
1015 if (match_int(&args[0], &option))
1016 return 0;
1017 sbi->s_resgid = option;
1018 break;
1019 case Opt_sb:
1020 /* handled by get_sb_block() instead of here */
1021 /* *sb_block = match_int(&args[0]); */
1022 break;
1023 case Opt_err_panic:
1024 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
1025 clear_opt (sbi->s_mount_opt, ERRORS_RO);
1026 set_opt (sbi->s_mount_opt, ERRORS_PANIC);
1027 break;
1028 case Opt_err_ro:
1029 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
1030 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
1031 set_opt (sbi->s_mount_opt, ERRORS_RO);
1032 break;
1033 case Opt_err_cont:
1034 clear_opt (sbi->s_mount_opt, ERRORS_RO);
1035 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
1036 set_opt (sbi->s_mount_opt, ERRORS_CONT);
1037 break;
1038 case Opt_nouid32:
1039 set_opt (sbi->s_mount_opt, NO_UID32);
1040 break;
1041 case Opt_nocheck:
1042 clear_opt (sbi->s_mount_opt, CHECK);
1043 break;
1044 case Opt_debug:
1045 set_opt (sbi->s_mount_opt, DEBUG);
1046 break;
1047 case Opt_oldalloc:
1048 set_opt (sbi->s_mount_opt, OLDALLOC);
1049 break;
1050 case Opt_orlov:
1051 clear_opt (sbi->s_mount_opt, OLDALLOC);
1052 break;
617ba13b 1053#ifdef CONFIG_EXT4DEV_FS_XATTR
ac27a0ec
DK		 1054 case Opt_user_xattr:
1055 set_opt (sbi->s_mount_opt, XATTR_USER);
1056 break;
1057 case Opt_nouser_xattr:
1058 clear_opt (sbi->s_mount_opt, XATTR_USER);
1059 break;
1060#else
1061 case Opt_user_xattr:
1062 case Opt_nouser_xattr:
617ba13b 1063 printk("EXT4 (no)user_xattr options not supported\n");
ac27a0ec
DK		 1064 break;
1065#endif
617ba13b 1066#ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
ac27a0ec
DK		 1067 case Opt_acl:
1068 set_opt(sbi->s_mount_opt, POSIX_ACL);
1069 break;
1070 case Opt_noacl:
1071 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1072 break;
1073#else
1074 case Opt_acl:
1075 case Opt_noacl:
617ba13b 1076 printk("EXT4 (no)acl options not supported\n");
ac27a0ec
DK		 1077 break;
1078#endif
1079 case Opt_reservation:
1080 set_opt(sbi->s_mount_opt, RESERVATION);
1081 break;
1082 case Opt_noreservation:
1083 clear_opt(sbi->s_mount_opt, RESERVATION);
1084 break;
1085 case Opt_journal_update:
1086 /* @@@ FIXME */
1087 /* Eventually we will want to be able to create
1088 a journal file here. For now, only allow the
1089 user to specify an existing inode to be the
1090 journal file. */
1091 if (is_remount) {
617ba13b 1092 printk(KERN_ERR "EXT4-fs: cannot specify "
ac27a0ec
DK		 1093 "journal on remount\n");
1094 return 0;
1095 }
1096 set_opt (sbi->s_mount_opt, UPDATE_JOURNAL);
1097 break;
1098 case Opt_journal_inum:
1099 if (is_remount) {
617ba13b 1100 printk(KERN_ERR "EXT4-fs: cannot specify "
ac27a0ec
DK		 1101 "journal on remount\n");
1102 return 0;
1103 }
1104 if (match_int(&args[0], &option))
1105 return 0;
1106 *inum = option;
1107 break;
1108 case Opt_journal_dev:
1109 if (is_remount) {
617ba13b 1110 printk(KERN_ERR "EXT4-fs: cannot specify "
ac27a0ec
DK		 1111 "journal on remount\n");
1112 return 0;
1113 }
1114 if (match_int(&args[0], &option))
1115 return 0;
1116 *journal_devnum = option;
1117 break;
818d276c
GS		 1118 case Opt_journal_checksum:
1119 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1120 break;
1121 case Opt_journal_async_commit:
1122 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
1123 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1124 break;
ac27a0ec
DK		 1125 case Opt_noload:
1126 set_opt (sbi->s_mount_opt, NOLOAD);
1127 break;
1128 case Opt_commit:
1129 if (match_int(&args[0], &option))
1130 return 0;
1131 if (option < 0)
1132 return 0;
1133 if (option == 0)
cd02ff0b 1134 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
ac27a0ec
DK		 1135 sbi->s_commit_interval = HZ * option;
1136 break;
1137 case Opt_data_journal:
617ba13b 1138 data_opt = EXT4_MOUNT_JOURNAL_DATA;
ac27a0ec
DK		 1139 goto datacheck;
1140 case Opt_data_ordered:
617ba13b 1141 data_opt = EXT4_MOUNT_ORDERED_DATA;
ac27a0ec
DK		 1142 goto datacheck;
1143 case Opt_data_writeback:
617ba13b 1144 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
ac27a0ec
DK		 1145 datacheck:
1146 if (is_remount) {
617ba13b 1147 if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
ac27a0ec
DK		 1148 != data_opt) {
1149 printk(KERN_ERR
617ba13b 1150 "EXT4-fs: cannot change data "
ac27a0ec
DK		 1151 "mode on remount\n");
1152 return 0;
1153 }
1154 } else {
617ba13b 1155 sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
ac27a0ec
DK		 1156 sbi->s_mount_opt |= data_opt;
1157 }
1158 break;
1159#ifdef CONFIG_QUOTA
1160 case Opt_usrjquota:
1161 qtype = USRQUOTA;
1162 goto set_qf_name;
1163 case Opt_grpjquota:
1164 qtype = GRPQUOTA;
1165set_qf_name:
1166 if (sb_any_quota_enabled(sb)) {
1167 printk(KERN_ERR
617ba13b 1168 "EXT4-fs: Cannot change journalled "
ac27a0ec
DK		 1169 "quota options when quota turned on.\n");
1170 return 0;
1171 }
1172 qname = match_strdup(&args[0]);
1173 if (!qname) {
1174 printk(KERN_ERR
617ba13b 1175 "EXT4-fs: not enough memory for "
ac27a0ec
DK		 1176 "storing quotafile name.\n");
1177 return 0;
1178 }
1179 if (sbi->s_qf_names[qtype] &&
1180 strcmp(sbi->s_qf_names[qtype], qname)) {
1181 printk(KERN_ERR
617ba13b 1182 "EXT4-fs: %s quota file already "
ac27a0ec
DK		 1183 "specified.\n", QTYPE2NAME(qtype));
1184 kfree(qname);
1185 return 0;
1186 }
1187 sbi->s_qf_names[qtype] = qname;
1188 if (strchr(sbi->s_qf_names[qtype], '/')) {
1189 printk(KERN_ERR
617ba13b 1190 "EXT4-fs: quotafile must be on "
ac27a0ec
DK		 1191 "filesystem root.\n");
1192 kfree(sbi->s_qf_names[qtype]);
1193 sbi->s_qf_names[qtype] = NULL;
1194 return 0;
1195 }
1196 set_opt(sbi->s_mount_opt, QUOTA);
1197 break;
1198 case Opt_offusrjquota:
1199 qtype = USRQUOTA;
1200 goto clear_qf_name;
1201 case Opt_offgrpjquota:
1202 qtype = GRPQUOTA;
1203clear_qf_name:
1204 if (sb_any_quota_enabled(sb)) {
617ba13b 1205 printk(KERN_ERR "EXT4-fs: Cannot change "
ac27a0ec
DK		 1206 "journalled quota options when "
1207 "quota turned on.\n");
1208 return 0;
1209 }
1210 /*
1211 * The space will be released later when all options
1212 * are confirmed to be correct
1213 */
1214 sbi->s_qf_names[qtype] = NULL;
1215 break;
1216 case Opt_jqfmt_vfsold:
1217 sbi->s_jquota_fmt = QFMT_VFS_OLD;
1218 break;
1219 case Opt_jqfmt_vfsv0:
1220 sbi->s_jquota_fmt = QFMT_VFS_V0;
1221 break;
1222 case Opt_quota:
1223 case Opt_usrquota:
1224 set_opt(sbi->s_mount_opt, QUOTA);
1225 set_opt(sbi->s_mount_opt, USRQUOTA);
1226 break;
1227 case Opt_grpquota:
1228 set_opt(sbi->s_mount_opt, QUOTA);
1229 set_opt(sbi->s_mount_opt, GRPQUOTA);
1230 break;
1231 case Opt_noquota:
1232 if (sb_any_quota_enabled(sb)) {
617ba13b 1233 printk(KERN_ERR "EXT4-fs: Cannot change quota "
ac27a0ec
DK		 1234 "options when quota turned on.\n");
1235 return 0;
1236 }
1237 clear_opt(sbi->s_mount_opt, QUOTA);
1238 clear_opt(sbi->s_mount_opt, USRQUOTA);
1239 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1240 break;
1241#else
1242 case Opt_quota:
1243 case Opt_usrquota:
1244 case Opt_grpquota:
1245 case Opt_usrjquota:
1246 case Opt_grpjquota:
1247 case Opt_offusrjquota:
1248 case Opt_offgrpjquota:
1249 case Opt_jqfmt_vfsold:
1250 case Opt_jqfmt_vfsv0:
1251 printk(KERN_ERR
617ba13b 1252 "EXT4-fs: journalled quota options not "
ac27a0ec
DK		 1253 "supported.\n");
1254 break;
1255 case Opt_noquota:
1256 break;
1257#endif
1258 case Opt_abort:
1259 set_opt(sbi->s_mount_opt, ABORT);
1260 break;
1261 case Opt_barrier:
1262 if (match_int(&args[0], &option))
1263 return 0;
1264 if (option)
1265 set_opt(sbi->s_mount_opt, BARRIER);
1266 else
1267 clear_opt(sbi->s_mount_opt, BARRIER);
1268 break;
1269 case Opt_ignore:
1270 break;
1271 case Opt_resize:
1272 if (!is_remount) {
617ba13b 1273 printk("EXT4-fs: resize option only available "
ac27a0ec
DK		 1274 "for remount\n");
1275 return 0;
1276 }
1277 if (match_int(&args[0], &option) != 0)
1278 return 0;
1279 *n_blocks_count = option;
1280 break;
1281 case Opt_nobh:
1282 set_opt(sbi->s_mount_opt, NOBH);
1283 break;
1284 case Opt_bh:
1285 clear_opt(sbi->s_mount_opt, NOBH);
1286 break;
a86c6181
AT		 1287 case Opt_extents:
1288 set_opt (sbi->s_mount_opt, EXTENTS);
1289 break;
1e2462f9
MC		 1290 case Opt_noextents:
1291 clear_opt (sbi->s_mount_opt, EXTENTS);
1292 break;
25ec56b5
JNC		 1293 case Opt_i_version:
1294 set_opt(sbi->s_mount_opt, I_VERSION);
1295 sb->s_flags |= MS_I_VERSION;
1296 break;
c9de560d
AT		 1297 case Opt_mballoc:
1298 set_opt(sbi->s_mount_opt, MBALLOC);
1299 break;
1300 case Opt_nomballoc:
1301 clear_opt(sbi->s_mount_opt, MBALLOC);
1302 break;
1303 case Opt_stripe:
1304 if (match_int(&args[0], &option))
1305 return 0;
1306 if (option < 0)
1307 return 0;
1308 sbi->s_stripe = option;
1309 break;
ac27a0ec
DK		 1310 default:
1311 printk (KERN_ERR
617ba13b 1312 "EXT4-fs: Unrecognized mount option \"%s\" "
ac27a0ec
DK		 1313 "or missing value\n", p);
1314 return 0;
1315 }
1316 }
1317#ifdef CONFIG_QUOTA
1318 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
617ba13b 1319 if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
ac27a0ec
DK		 1320 sbi->s_qf_names[USRQUOTA])
1321 clear_opt(sbi->s_mount_opt, USRQUOTA);
1322
617ba13b 1323 if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
ac27a0ec
DK		 1324 sbi->s_qf_names[GRPQUOTA])
1325 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1326
1327 if ((sbi->s_qf_names[USRQUOTA] &&
617ba13b 1328 (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
ac27a0ec 1329 (sbi->s_qf_names[GRPQUOTA] &&
617ba13b
MC		 1330 (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
1331 printk(KERN_ERR "EXT4-fs: old and new quota "
ac27a0ec
DK		 1332 "format mixing.\n");
1333 return 0;
1334 }
1335
1336 if (!sbi->s_jquota_fmt) {
617ba13b 1337 printk(KERN_ERR "EXT4-fs: journalled quota format "
ac27a0ec
DK		 1338 "not specified.\n");
1339 return 0;
1340 }
1341 } else {
1342 if (sbi->s_jquota_fmt) {
617ba13b 1343 printk(KERN_ERR "EXT4-fs: journalled quota format "
ac27a0ec
DK		 1344 "specified with no journalling "
1345 "enabled.\n");
1346 return 0;
1347 }
1348 }
1349#endif
1350 return 1;
1351}
1352
617ba13b 1353static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 1354 int read_only)
1355{
617ba13b 1356 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 1357 int res = 0;
1358
617ba13b
MC		 1359 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1360 printk (KERN_ERR "EXT4-fs warning: revision level too high, "
ac27a0ec
DK		 1361 "forcing read-only mode\n");
1362 res = MS_RDONLY;
1363 }
1364 if (read_only)
1365 return res;
617ba13b
MC		 1366 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1367 printk (KERN_WARNING "EXT4-fs warning: mounting unchecked fs, "
ac27a0ec 1368 "running e2fsck is recommended\n");
617ba13b 1369 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
ac27a0ec 1370 printk (KERN_WARNING
617ba13b 1371 "EXT4-fs warning: mounting fs with errors, "
ac27a0ec
DK		 1372 "running e2fsck is recommended\n");
1373 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1374 le16_to_cpu(es->s_mnt_count) >=
1375 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1376 printk (KERN_WARNING
617ba13b 1377 "EXT4-fs warning: maximal mount count reached, "
ac27a0ec
DK		 1378 "running e2fsck is recommended\n");
1379 else if (le32_to_cpu(es->s_checkinterval) &&
1380 (le32_to_cpu(es->s_lastcheck) +
1381 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1382 printk (KERN_WARNING
617ba13b 1383 "EXT4-fs warning: checktime reached, "
ac27a0ec
DK		 1384 "running e2fsck is recommended\n");
1385#if 0
1386 /* @@@ We _will_ want to clear the valid bit if we find
63f57933
AM		 1387 * inconsistencies, to force a fsck at reboot. But for
1388 * a plain journaled filesystem we can keep it set as
1389 * valid forever! :)
1390 */
617ba13b 1391 es->s_state = cpu_to_le16(le16_to_cpu(es->s_state) & ~EXT4_VALID_FS);
ac27a0ec
DK		 1392#endif
1393 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 1394 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
ac27a0ec
DK		 1395 es->s_mnt_count=cpu_to_le16(le16_to_cpu(es->s_mnt_count) + 1);
1396 es->s_mtime = cpu_to_le32(get_seconds());
617ba13b
MC		 1397 ext4_update_dynamic_rev(sb);
1398 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 1399
617ba13b 1400 ext4_commit_super(sb, es, 1);
ac27a0ec 1401 if (test_opt(sb, DEBUG))
617ba13b 1402 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%lu, "
ac27a0ec
DK		 1403 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
1404 sb->s_blocksize,
1405 sbi->s_groups_count,
617ba13b
MC		 1406 EXT4_BLOCKS_PER_GROUP(sb),
1407 EXT4_INODES_PER_GROUP(sb),
ac27a0ec
DK		 1408 sbi->s_mount_opt);
1409
617ba13b
MC		 1410 printk(KERN_INFO "EXT4 FS on %s, ", sb->s_id);
1411 if (EXT4_SB(sb)->s_journal->j_inode == NULL) {
ac27a0ec
DK		 1412 char b[BDEVNAME_SIZE];
1413
1414 printk("external journal on %s\n",
617ba13b 1415 bdevname(EXT4_SB(sb)->s_journal->j_dev, b));
ac27a0ec
DK		 1416 } else {
1417 printk("internal journal\n");
1418 }
1419 return res;
1420}
1421
717d50e4
AD		 1422__le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1423 struct ext4_group_desc *gdp)
1424{
1425 __u16 crc = 0;
1426
1427 if (sbi->s_es->s_feature_ro_compat &
1428 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1429 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1430 __le32 le_group = cpu_to_le32(block_group);
1431
1432 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1433 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1434 crc = crc16(crc, (__u8 *)gdp, offset);
1435 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1436 /* for checksum of struct ext4_group_desc do the rest...*/
1437 if ((sbi->s_es->s_feature_incompat &
1438 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1439 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1440 crc = crc16(crc, (__u8 *)gdp + offset,
1441 le16_to_cpu(sbi->s_es->s_desc_size) -
1442 offset);
1443 }
1444
1445 return cpu_to_le16(crc);
1446}
1447
1448int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1449 struct ext4_group_desc *gdp)
1450{
1451 if ((sbi->s_es->s_feature_ro_compat &
1452 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1453 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1454 return 0;
1455
1456 return 1;
1457}
1458
ac27a0ec 1459/* Called at mount-time, super-block is locked */
197cd65a 1460static int ext4_check_descriptors(struct super_block *sb)
ac27a0ec 1461{
617ba13b
MC		 1462 struct ext4_sb_info *sbi = EXT4_SB(sb);
1463 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1464 ext4_fsblk_t last_block;
bd81d8ee
LV		 1465 ext4_fsblk_t block_bitmap;
1466 ext4_fsblk_t inode_bitmap;
1467 ext4_fsblk_t inode_table;
ce421581 1468 int flexbg_flag = 0;
fd2d4291 1469 ext4_group_t i;
ac27a0ec 1470
ce421581
JS		 1471 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1472 flexbg_flag = 1;
1473
617ba13b 1474 ext4_debug ("Checking group descriptors");
ac27a0ec 1475
197cd65a
AM		 1476 for (i = 0; i < sbi->s_groups_count; i++) {
1477 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1478
ce421581 1479 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 1480 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 1481 else
1482 last_block = first_block +
617ba13b 1483 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 1484
8fadc143 1485 block_bitmap = ext4_block_bitmap(sb, gdp);
bd81d8ee 1486 if (block_bitmap < first_block || block_bitmap > last_block)
ac27a0ec 1487 {
617ba13b 1488 ext4_error (sb, "ext4_check_descriptors",
fd2d4291 1489 "Block bitmap for group %lu"
2ae02107 1490 " not in group (block %llu)!",
bd81d8ee 1491 i, block_bitmap);
ac27a0ec
DK		 1492 return 0;
1493 }
8fadc143 1494 inode_bitmap = ext4_inode_bitmap(sb, gdp);
bd81d8ee 1495 if (inode_bitmap < first_block || inode_bitmap > last_block)
ac27a0ec 1496 {
617ba13b 1497 ext4_error (sb, "ext4_check_descriptors",
fd2d4291 1498 "Inode bitmap for group %lu"
2ae02107 1499 " not in group (block %llu)!",
bd81d8ee 1500 i, inode_bitmap);
ac27a0ec
DK		 1501 return 0;
1502 }
8fadc143 1503 inode_table = ext4_inode_table(sb, gdp);
bd81d8ee 1504 if (inode_table < first_block ||
780dcdb2 1505 inode_table + sbi->s_itb_per_group - 1 > last_block)
ac27a0ec 1506 {
617ba13b 1507 ext4_error (sb, "ext4_check_descriptors",
fd2d4291 1508 "Inode table for group %lu"
2ae02107 1509 " not in group (block %llu)!",
bd81d8ee 1510 i, inode_table);
ac27a0ec
DK		 1511 return 0;
1512 }
717d50e4
AD		 1513 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
1514 ext4_error(sb, __FUNCTION__,
fd2d4291
AM		 1515 "Checksum for group %lu failed (%u!=%u)\n",
1516 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1517 gdp)), le16_to_cpu(gdp->bg_checksum));
717d50e4
AD		 1518 return 0;
1519 }
ce421581
JS		 1520 if (!flexbg_flag)
1521 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec
DK		 1522 }
1523
bd81d8ee 1524 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
617ba13b 1525 sbi->s_es->s_free_inodes_count=cpu_to_le32(ext4_count_free_inodes(sb));
ac27a0ec
DK		 1526 return 1;
1527}
1528
617ba13b 1529/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 1530 * the superblock) which were deleted from all directories, but held open by
1531 * a process at the time of a crash. We walk the list and try to delete these
1532 * inodes at recovery time (only with a read-write filesystem).
1533 *
1534 * In order to keep the orphan inode chain consistent during traversal (in
1535 * case of crash during recovery), we link each inode into the superblock
1536 * orphan list_head and handle it the same way as an inode deletion during
1537 * normal operation (which journals the operations for us).
1538 *
1539 * We only do an iget() and an iput() on each inode, which is very safe if we
1540 * accidentally point at an in-use or already deleted inode. The worst that
1541 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 1542 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 1543 * e2fsck was run on this filesystem, and it must have already done the orphan
1544 * inode cleanup for us, so we can safely abort without any further action.
1545 */
617ba13b
MC		 1546static void ext4_orphan_cleanup (struct super_block * sb,
1547 struct ext4_super_block * es)
ac27a0ec
DK		 1548{
1549 unsigned int s_flags = sb->s_flags;
1550 int nr_orphans = 0, nr_truncates = 0;
1551#ifdef CONFIG_QUOTA
1552 int i;
1553#endif
1554 if (!es->s_last_orphan) {
1555 jbd_debug(4, "no orphan inodes to clean up\n");
1556 return;
1557 }
1558
a8f48a95
ES		 1559 if (bdev_read_only(sb->s_bdev)) {
1560 printk(KERN_ERR "EXT4-fs: write access "
1561 "unavailable, skipping orphan cleanup.\n");
1562 return;
1563 }
1564
617ba13b 1565 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
ac27a0ec
DK		 1566 if (es->s_last_orphan)
1567 jbd_debug(1, "Errors on filesystem, "
1568 "clearing orphan list.\n");
1569 es->s_last_orphan = 0;
1570 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1571 return;
1572 }
1573
1574 if (s_flags & MS_RDONLY) {
617ba13b 1575 printk(KERN_INFO "EXT4-fs: %s: orphan cleanup on readonly fs\n",
ac27a0ec
DK		 1576 sb->s_id);
1577 sb->s_flags &= ~MS_RDONLY;
1578 }
1579#ifdef CONFIG_QUOTA
1580 /* Needed for iput() to work correctly and not trash data */
1581 sb->s_flags |= MS_ACTIVE;
1582 /* Turn on quotas so that they are updated correctly */
1583 for (i = 0; i < MAXQUOTAS; i++) {
617ba13b
MC		 1584 if (EXT4_SB(sb)->s_qf_names[i]) {
1585 int ret = ext4_quota_on_mount(sb, i);
ac27a0ec
DK		 1586 if (ret < 0)
1587 printk(KERN_ERR
617ba13b 1588 "EXT4-fs: Cannot turn on journalled "
ac27a0ec
DK		 1589 "quota: error %d\n", ret);
1590 }
1591 }
1592#endif
1593
1594 while (es->s_last_orphan) {
1595 struct inode *inode;
1596
97bd42b9
JB		 1597 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1598 if (IS_ERR(inode)) {
ac27a0ec
DK		 1599 es->s_last_orphan = 0;
1600 break;
1601 }
1602
617ba13b 1603 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
ac27a0ec
DK		 1604 DQUOT_INIT(inode);
1605 if (inode->i_nlink) {
1606 printk(KERN_DEBUG
1607 "%s: truncating inode %lu to %Ld bytes\n",
1608 __FUNCTION__, inode->i_ino, inode->i_size);
1609 jbd_debug(2, "truncating inode %lu to %Ld bytes\n",
1610 inode->i_ino, inode->i_size);
617ba13b 1611 ext4_truncate(inode);
ac27a0ec
DK		 1612 nr_truncates++;
1613 } else {
1614 printk(KERN_DEBUG
1615 "%s: deleting unreferenced inode %lu\n",
1616 __FUNCTION__, inode->i_ino);
1617 jbd_debug(2, "deleting unreferenced inode %lu\n",
1618 inode->i_ino);
1619 nr_orphans++;
1620 }
1621 iput(inode); /* The delete magic happens here! */
1622 }
1623
1624#define PLURAL(x) (x), ((x)==1) ? "" : "s"
1625
1626 if (nr_orphans)
617ba13b 1627 printk(KERN_INFO "EXT4-fs: %s: %d orphan inode%s deleted\n",
ac27a0ec
DK		 1628 sb->s_id, PLURAL(nr_orphans));
1629 if (nr_truncates)
617ba13b 1630 printk(KERN_INFO "EXT4-fs: %s: %d truncate%s cleaned up\n",
ac27a0ec
DK		 1631 sb->s_id, PLURAL(nr_truncates));
1632#ifdef CONFIG_QUOTA
1633 /* Turn quotas off */
1634 for (i = 0; i < MAXQUOTAS; i++) {
1635 if (sb_dqopt(sb)->files[i])
6f28e087 1636 vfs_quota_off(sb, i, 0);
ac27a0ec
DK		 1637 }
1638#endif
1639 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1640}
cd2291a4
ES		 1641/*
1642 * Maximal extent format file size.
1643 * Resulting logical blkno at s_maxbytes must fit in our on-disk
1644 * extent format containers, within a sector_t, and within i_blocks
1645 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
1646 * so that won't be a limiting factor.
1647 *
1648 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1649 */
1650static loff_t ext4_max_size(int blkbits)
1651{
1652 loff_t res;
1653 loff_t upper_limit = MAX_LFS_FILESIZE;
1654
1655 /* small i_blocks in vfs inode? */
1656 if (sizeof(blkcnt_t) < sizeof(u64)) {
1657 /*
1658 * CONFIG_LSF is not enabled implies the inode
1659 * i_block represent total blocks in 512 bytes
1660 * 32 == size of vfs inode i_blocks * 8
1661 */
1662 upper_limit = (1LL << 32) - 1;
1663
1664 /* total blocks in file system block size */
1665 upper_limit >>= (blkbits - 9);
1666 upper_limit <<= blkbits;
1667 }
1668
1669 /* 32-bit extent-start container, ee_block */
1670 res = 1LL << 32;
1671 res <<= blkbits;
1672 res -= 1;
1673
1674 /* Sanity check against vm- & vfs- imposed limits */
1675 if (res > upper_limit)
1676 res = upper_limit;
1677
1678 return res;
1679}
ac27a0ec 1680
ac27a0ec 1681/*
cd2291a4 1682 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 1683 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
1684 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 1685 */
cd2291a4 1686static loff_t ext4_max_bitmap_size(int bits)
ac27a0ec 1687{
617ba13b 1688 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 1689 int meta_blocks;
1690 loff_t upper_limit;
1691 /* This is calculated to be the largest file size for a
cd2291a4 1692 * dense, bitmapped file such that the total number of
ac27a0ec 1693 * sectors in the file, including data and all indirect blocks,
0fc1b451
AK		 1694 * does not exceed 2^48 -1
1695 * __u32 i_blocks_lo and _u16 i_blocks_high representing the
1696 * total number of 512 bytes blocks of the file
1697 */
1698
1699 if (sizeof(blkcnt_t) < sizeof(u64)) {
1700 /*
1701 * CONFIG_LSF is not enabled implies the inode
1702 * i_block represent total blocks in 512 bytes
1703 * 32 == size of vfs inode i_blocks * 8
1704 */
1705 upper_limit = (1LL << 32) - 1;
1706
1707 /* total blocks in file system block size */
1708 upper_limit >>= (bits - 9);
1709
1710 } else {
8180a562
AK		 1711 /*
1712 * We use 48 bit ext4_inode i_blocks
1713 * With EXT4_HUGE_FILE_FL set the i_blocks
1714 * represent total number of blocks in
1715 * file system block size
1716 */
0fc1b451
AK		 1717 upper_limit = (1LL << 48) - 1;
1718
0fc1b451
AK		 1719 }
1720
1721 /* indirect blocks */
1722 meta_blocks = 1;
1723 /* double indirect blocks */
1724 meta_blocks += 1 + (1LL << (bits-2));
1725 /* tripple indirect blocks */
1726 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
1727
1728 upper_limit -= meta_blocks;
1729 upper_limit <<= bits;
ac27a0ec
DK		 1730
1731 res += 1LL << (bits-2);
1732 res += 1LL << (2*(bits-2));
1733 res += 1LL << (3*(bits-2));
1734 res <<= bits;
1735 if (res > upper_limit)
1736 res = upper_limit;
0fc1b451
AK		 1737
1738 if (res > MAX_LFS_FILESIZE)
1739 res = MAX_LFS_FILESIZE;
1740
ac27a0ec
DK		 1741 return res;
1742}
1743
617ba13b 1744static ext4_fsblk_t descriptor_loc(struct super_block *sb,
70bbb3e0 1745 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 1746{
617ba13b 1747 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 1748 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 1749 int has_super = 0;
1750
1751 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1752
617ba13b 1753 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
ac27a0ec 1754 nr < first_meta_bg)
70bbb3e0 1755 return logical_sb_block + nr + 1;
ac27a0ec 1756 bg = sbi->s_desc_per_block * nr;
617ba13b 1757 if (ext4_bg_has_super(sb, bg))
ac27a0ec 1758 has_super = 1;
617ba13b 1759 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 1760}
1761
c9de560d
AT		 1762/**
1763 * ext4_get_stripe_size: Get the stripe size.
1764 * @sbi: In memory super block info
1765 *
1766 * If we have specified it via mount option, then
1767 * use the mount option value. If the value specified at mount time is
1768 * greater than the blocks per group use the super block value.
1769 * If the super block value is greater than blocks per group return 0.
1770 * Allocator needs it be less than blocks per group.
1771 *
1772 */
1773static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
1774{
1775 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
1776 unsigned long stripe_width =
1777 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
1778
1779 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
1780 return sbi->s_stripe;
1781
1782 if (stripe_width <= sbi->s_blocks_per_group)
1783 return stripe_width;
1784
1785 if (stride <= sbi->s_blocks_per_group)
1786 return stride;
1787
1788 return 0;
1789}
ac27a0ec 1790
617ba13b 1791static int ext4_fill_super (struct super_block *sb, void *data, int silent)
1d03ec98
AK		 1792 __releases(kernel_sem)
1793 __acquires(kernel_sem)
1794
ac27a0ec
DK		 1795{
1796 struct buffer_head * bh;
617ba13b
MC		 1797 struct ext4_super_block *es = NULL;
1798 struct ext4_sb_info *sbi;
1799 ext4_fsblk_t block;
1800 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 1801 ext4_fsblk_t logical_sb_block;
ac27a0ec
DK		 1802 unsigned long offset = 0;
1803 unsigned int journal_inum = 0;
1804 unsigned long journal_devnum = 0;
1805 unsigned long def_mount_opts;
1806 struct inode *root;
1d1fe1ee 1807 int ret = -EINVAL;
ac27a0ec 1808 int blocksize;
ac27a0ec
DK		 1809 int db_count;
1810 int i;
1811 int needs_recovery;
1812 __le32 features;
bd81d8ee 1813 __u64 blocks_count;
833f4077 1814 int err;
ac27a0ec
DK		 1815
1816 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
1817 if (!sbi)
1818 return -ENOMEM;
1819 sb->s_fs_info = sbi;
1820 sbi->s_mount_opt = 0;
617ba13b
MC		 1821 sbi->s_resuid = EXT4_DEF_RESUID;
1822 sbi->s_resgid = EXT4_DEF_RESGID;
d9c9bef1 1823 sbi->s_sb_block = sb_block;
ac27a0ec
DK		 1824
1825 unlock_kernel();
1826
617ba13b 1827 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 1828 if (!blocksize) {
617ba13b 1829 printk(KERN_ERR "EXT4-fs: unable to set blocksize\n");
ac27a0ec
DK		 1830 goto out_fail;
1831 }
1832
afc7cbca
TS		 1833 if (!sb_set_blocksize(sb, blocksize)) {
1834 printk(KERN_ERR "EXT4-fs: bad blocksize %d.\n", blocksize);
1835 goto out_fail;
1836 }
1837
ac27a0ec 1838 /*
617ba13b 1839 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 1840 * block sizes. We need to calculate the offset from buffer start.
1841 */
617ba13b 1842 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 1843 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
1844 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 1845 } else {
70bbb3e0 1846 logical_sb_block = sb_block;
ac27a0ec
DK		 1847 }
1848
70bbb3e0 1849 if (!(bh = sb_bread(sb, logical_sb_block))) {
617ba13b 1850 printk (KERN_ERR "EXT4-fs: unable to read superblock\n");
ac27a0ec
DK		 1851 goto out_fail;
1852 }
1853 /*
1854 * Note: s_es must be initialized as soon as possible because
617ba13b 1855 * some ext4 macro-instructions depend on its value
ac27a0ec 1856 */
617ba13b 1857 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
ac27a0ec
DK		 1858 sbi->s_es = es;
1859 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 1860 if (sb->s_magic != EXT4_SUPER_MAGIC)
1861 goto cantfind_ext4;
ac27a0ec
DK		 1862
1863 /* Set defaults before we parse the mount options */
1864 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
617ba13b 1865 if (def_mount_opts & EXT4_DEFM_DEBUG)
ac27a0ec 1866 set_opt(sbi->s_mount_opt, DEBUG);
617ba13b 1867 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
ac27a0ec 1868 set_opt(sbi->s_mount_opt, GRPID);
617ba13b 1869 if (def_mount_opts & EXT4_DEFM_UID16)
ac27a0ec 1870 set_opt(sbi->s_mount_opt, NO_UID32);
2e7842b8 1871#ifdef CONFIG_EXT4DEV_FS_XATTR
617ba13b 1872 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
ac27a0ec 1873 set_opt(sbi->s_mount_opt, XATTR_USER);
2e7842b8
HD		 1874#endif
1875#ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
617ba13b 1876 if (def_mount_opts & EXT4_DEFM_ACL)
ac27a0ec 1877 set_opt(sbi->s_mount_opt, POSIX_ACL);
2e7842b8 1878#endif
617ba13b
MC		 1879 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
1880 sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
1881 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
1882 sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
1883 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
1884 sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
1885
1886 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
ac27a0ec 1887 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
bb4f397a 1888 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
ceea16bf 1889 set_opt(sbi->s_mount_opt, ERRORS_CONT);
bb4f397a
AK		 1890 else
1891 set_opt(sbi->s_mount_opt, ERRORS_RO);
ac27a0ec
DK		 1892
1893 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
1894 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
1895
1896 set_opt(sbi->s_mount_opt, RESERVATION);
1897
1e2462f9
MC		 1898 /*
1899 * turn on extents feature by default in ext4 filesystem
1900 * User -o noextents to turn it off
1901 */
1902 set_opt(sbi->s_mount_opt, EXTENTS);
3dbd0ede
AK		 1903 /*
1904 * turn on mballoc feature by default in ext4 filesystem
1905 * User -o nomballoc to turn it off
1906 */
1907 set_opt(sbi->s_mount_opt, MBALLOC);
1e2462f9 1908
ac27a0ec
DK		 1909 if (!parse_options ((char *) data, sb, &journal_inum, &journal_devnum,
1910 NULL, 0))
1911 goto failed_mount;
1912
1913 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
617ba13b 1914 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec 1915
617ba13b
MC		 1916 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
1917 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
1918 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
1919 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
ac27a0ec 1920 printk(KERN_WARNING
617ba13b 1921 "EXT4-fs warning: feature flags set on rev 0 fs, "
ac27a0ec 1922 "running e2fsck is recommended\n");
469108ff
TT		 1923
1924 /*
1925 * Since ext4 is still considered development code, we require
1926 * that the TEST_FILESYS flag in s->flags be set.
1927 */
1928 if (!(le32_to_cpu(es->s_flags) & EXT2_FLAGS_TEST_FILESYS)) {
1929 printk(KERN_WARNING "EXT4-fs: %s: not marked "
1930 "OK to use with test code.\n", sb->s_id);
1931 goto failed_mount;
1932 }
1933
ac27a0ec
DK		 1934 /*
1935 * Check feature flags regardless of the revision level, since we
1936 * previously didn't change the revision level when setting the flags,
1937 * so there is a chance incompat flags are set on a rev 0 filesystem.
1938 */
617ba13b 1939 features = EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP);
ac27a0ec 1940 if (features) {
617ba13b 1941 printk(KERN_ERR "EXT4-fs: %s: couldn't mount because of "
ac27a0ec
DK		 1942 "unsupported optional features (%x).\n",
1943 sb->s_id, le32_to_cpu(features));
1944 goto failed_mount;
1945 }
617ba13b 1946 features = EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP);
ac27a0ec 1947 if (!(sb->s_flags & MS_RDONLY) && features) {
617ba13b 1948 printk(KERN_ERR "EXT4-fs: %s: couldn't mount RDWR because of "
ac27a0ec
DK		 1949 "unsupported optional features (%x).\n",
1950 sb->s_id, le32_to_cpu(features));
1951 goto failed_mount;
1952 }
0fc1b451
AK		 1953 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
1954 /*
1955 * Large file size enabled file system can only be
1956 * mount if kernel is build with CONFIG_LSF
1957 */
1958 if (sizeof(root->i_blocks) < sizeof(u64) &&
1959 !(sb->s_flags & MS_RDONLY)) {
1960 printk(KERN_ERR "EXT4-fs: %s: Filesystem with huge "
1961 "files cannot be mounted read-write "
1962 "without CONFIG_LSF.\n", sb->s_id);
1963 goto failed_mount;
1964 }
1965 }
ac27a0ec
DK		 1966 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
1967
617ba13b
MC		 1968 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
1969 blocksize > EXT4_MAX_BLOCK_SIZE) {
ac27a0ec 1970 printk(KERN_ERR
617ba13b 1971 "EXT4-fs: Unsupported filesystem blocksize %d on %s.\n",
ac27a0ec
DK		 1972 blocksize, sb->s_id);
1973 goto failed_mount;
1974 }
1975
ac27a0ec 1976 if (sb->s_blocksize != blocksize) {
ce40733c
AK		 1977
1978 /* Validate the filesystem blocksize */
1979 if (!sb_set_blocksize(sb, blocksize)) {
1980 printk(KERN_ERR "EXT4-fs: bad block size %d.\n",
1981 blocksize);
ac27a0ec
DK		 1982 goto failed_mount;
1983 }
1984
1985 brelse (bh);
70bbb3e0
AM		 1986 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
1987 offset = do_div(logical_sb_block, blocksize);
1988 bh = sb_bread(sb, logical_sb_block);
ac27a0ec
DK		 1989 if (!bh) {
1990 printk(KERN_ERR
617ba13b 1991 "EXT4-fs: Can't read superblock on 2nd try.\n");
ac27a0ec
DK		 1992 goto failed_mount;
1993 }
617ba13b 1994 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
ac27a0ec 1995 sbi->s_es = es;
617ba13b 1996 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
ac27a0ec 1997 printk (KERN_ERR
617ba13b 1998 "EXT4-fs: Magic mismatch, very weird !\n");
ac27a0ec
DK		 1999 goto failed_mount;
2000 }
2001 }
2002
e2b46574 2003 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits);
617ba13b 2004 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits);
ac27a0ec 2005
617ba13b
MC		 2006 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2007 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2008 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
ac27a0ec
DK		 2009 } else {
2010 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2011 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
617ba13b 2012 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1330593e 2013 (!is_power_of_2(sbi->s_inode_size)) ||
ac27a0ec
DK		 2014 (sbi->s_inode_size > blocksize)) {
2015 printk (KERN_ERR
617ba13b 2016 "EXT4-fs: unsupported inode size: %d\n",
ac27a0ec
DK		 2017 sbi->s_inode_size);
2018 goto failed_mount;
2019 }
ef7f3835
KS		 2020 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2021 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
ac27a0ec 2022 }
0d1ee42f
AR		 2023 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2024 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
8fadc143 2025 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 2026 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 2027 !is_power_of_2(sbi->s_desc_size)) {
0d1ee42f 2028 printk(KERN_ERR
8fadc143 2029 "EXT4-fs: unsupported descriptor size %lu\n",
0d1ee42f
AR		 2030 sbi->s_desc_size);
2031 goto failed_mount;
2032 }
2033 } else
2034 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
ac27a0ec 2035 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 2036 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
b47b6f38 2037 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
617ba13b
MC		 2038 goto cantfind_ext4;
2039 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 2040 if (sbi->s_inodes_per_block == 0)
617ba13b 2041 goto cantfind_ext4;
ac27a0ec
DK		 2042 sbi->s_itb_per_group = sbi->s_inodes_per_group /
2043 sbi->s_inodes_per_block;
0d1ee42f 2044 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 2045 sbi->s_sbh = bh;
2046 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 2047 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2048 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
ac27a0ec
DK		 2049 for (i=0; i < 4; i++)
2050 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2051 sbi->s_def_hash_version = es->s_def_hash_version;
2052
2053 if (sbi->s_blocks_per_group > blocksize * 8) {
2054 printk (KERN_ERR
617ba13b 2055 "EXT4-fs: #blocks per group too big: %lu\n",
ac27a0ec
DK		 2056 sbi->s_blocks_per_group);
2057 goto failed_mount;
2058 }
ac27a0ec
DK		 2059 if (sbi->s_inodes_per_group > blocksize * 8) {
2060 printk (KERN_ERR
617ba13b 2061 "EXT4-fs: #inodes per group too big: %lu\n",
ac27a0ec
DK		 2062 sbi->s_inodes_per_group);
2063 goto failed_mount;
2064 }
2065
bd81d8ee 2066 if (ext4_blocks_count(es) >
ac27a0ec 2067 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
617ba13b 2068 printk(KERN_ERR "EXT4-fs: filesystem on %s:"
ac27a0ec
DK		 2069 " too large to mount safely\n", sb->s_id);
2070 if (sizeof(sector_t) < 8)
617ba13b 2071 printk(KERN_WARNING "EXT4-fs: CONFIG_LBD not "
ac27a0ec
DK		 2072 "enabled\n");
2073 goto failed_mount;
2074 }
2075
617ba13b
MC		 2076 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2077 goto cantfind_ext4;
e7c95593
ES		 2078
2079 /* ensure blocks_count calculation below doesn't sign-extend */
2080 if (ext4_blocks_count(es) + EXT4_BLOCKS_PER_GROUP(sb) <
2081 le32_to_cpu(es->s_first_data_block) + 1) {
2082 printk(KERN_WARNING "EXT4-fs: bad geometry: block count %llu, "
2083 "first data block %u, blocks per group %lu\n",
2084 ext4_blocks_count(es),
2085 le32_to_cpu(es->s_first_data_block),
2086 EXT4_BLOCKS_PER_GROUP(sb));
2087 goto failed_mount;
2088 }
bd81d8ee
LV		 2089 blocks_count = (ext4_blocks_count(es) -
2090 le32_to_cpu(es->s_first_data_block) +
2091 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2092 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
2093 sbi->s_groups_count = blocks_count;
617ba13b
MC		 2094 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2095 EXT4_DESC_PER_BLOCK(sb);
ac27a0ec
DK		 2096 sbi->s_group_desc = kmalloc(db_count * sizeof (struct buffer_head *),
2097 GFP_KERNEL);
2098 if (sbi->s_group_desc == NULL) {
617ba13b 2099 printk (KERN_ERR "EXT4-fs: not enough memory\n");
ac27a0ec
DK		 2100 goto failed_mount;
2101 }
2102
2103 bgl_lock_init(&sbi->s_blockgroup_lock);
2104
2105 for (i = 0; i < db_count; i++) {
70bbb3e0 2106 block = descriptor_loc(sb, logical_sb_block, i);
ac27a0ec
DK		 2107 sbi->s_group_desc[i] = sb_bread(sb, block);
2108 if (!sbi->s_group_desc[i]) {
617ba13b 2109 printk (KERN_ERR "EXT4-fs: "
ac27a0ec
DK		 2110 "can't read group descriptor %d\n", i);
2111 db_count = i;
2112 goto failed_mount2;
2113 }
2114 }
617ba13b
MC		 2115 if (!ext4_check_descriptors (sb)) {
2116 printk(KERN_ERR "EXT4-fs: group descriptors corrupted!\n");
ac27a0ec
DK		 2117 goto failed_mount2;
2118 }
2119 sbi->s_gdb_count = db_count;
2120 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2121 spin_lock_init(&sbi->s_next_gen_lock);
2122
833f4077
PZ		 2123 err = percpu_counter_init(&sbi->s_freeblocks_counter,
2124 ext4_count_free_blocks(sb));
2125 if (!err) {
2126 err = percpu_counter_init(&sbi->s_freeinodes_counter,
2127 ext4_count_free_inodes(sb));
2128 }
2129 if (!err) {
2130 err = percpu_counter_init(&sbi->s_dirs_counter,
2131 ext4_count_dirs(sb));
2132 }
2133 if (err) {
2134 printk(KERN_ERR "EXT4-fs: insufficient memory\n");
2135 goto failed_mount3;
2136 }
ac27a0ec
DK		 2137
2138 /* per fileystem reservation list head & lock */
2139 spin_lock_init(&sbi->s_rsv_window_lock);
2140 sbi->s_rsv_window_root = RB_ROOT;
2141 /* Add a single, static dummy reservation to the start of the
2142 * reservation window list --- it gives us a placeholder for
2143 * append-at-start-of-list which makes the allocation logic
2144 * _much_ simpler. */
617ba13b
MC		 2145 sbi->s_rsv_window_head.rsv_start = EXT4_RESERVE_WINDOW_NOT_ALLOCATED;
2146 sbi->s_rsv_window_head.rsv_end = EXT4_RESERVE_WINDOW_NOT_ALLOCATED;
ac27a0ec
DK		 2147 sbi->s_rsv_window_head.rsv_alloc_hit = 0;
2148 sbi->s_rsv_window_head.rsv_goal_size = 0;
617ba13b 2149 ext4_rsv_window_add(sb, &sbi->s_rsv_window_head);
ac27a0ec 2150
c9de560d
AT		 2151 sbi->s_stripe = ext4_get_stripe_size(sbi);
2152
ac27a0ec
DK		 2153 /*
2154 * set up enough so that it can read an inode
2155 */
617ba13b
MC		 2156 sb->s_op = &ext4_sops;
2157 sb->s_export_op = &ext4_export_ops;
2158 sb->s_xattr = ext4_xattr_handlers;
ac27a0ec 2159#ifdef CONFIG_QUOTA
617ba13b
MC		 2160 sb->s_qcop = &ext4_qctl_operations;
2161 sb->dq_op = &ext4_quota_operations;
ac27a0ec
DK		 2162#endif
2163 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
2164
2165 sb->s_root = NULL;
2166
2167 needs_recovery = (es->s_last_orphan != 0 ||
617ba13b
MC		 2168 EXT4_HAS_INCOMPAT_FEATURE(sb,
2169 EXT4_FEATURE_INCOMPAT_RECOVER));
ac27a0ec
DK		 2170
2171 /*
2172 * The first inode we look at is the journal inode. Don't try
2173 * root first: it may be modified in the journal!
2174 */
2175 if (!test_opt(sb, NOLOAD) &&
617ba13b
MC		 2176 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2177 if (ext4_load_journal(sb, es, journal_devnum))
ac27a0ec
DK		 2178 goto failed_mount3;
2179 } else if (journal_inum) {
617ba13b 2180 if (ext4_create_journal(sb, es, journal_inum))
ac27a0ec
DK		 2181 goto failed_mount3;
2182 } else {
2183 if (!silent)
2184 printk (KERN_ERR
617ba13b 2185 "ext4: No journal on filesystem on %s\n",
ac27a0ec
DK		 2186 sb->s_id);
2187 goto failed_mount3;
2188 }
2189
eb40a09c
JS		 2190 if (ext4_blocks_count(es) > 0xffffffffULL &&
2191 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2192 JBD2_FEATURE_INCOMPAT_64BIT)) {
2193 printk(KERN_ERR "ext4: Failed to set 64-bit journal feature\n");
2194 goto failed_mount4;
2195 }
2196
818d276c
GS		 2197 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2198 jbd2_journal_set_features(sbi->s_journal,
2199 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2200 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2201 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2202 jbd2_journal_set_features(sbi->s_journal,
2203 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2204 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2205 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2206 } else {
2207 jbd2_journal_clear_features(sbi->s_journal,
2208 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2209 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2210 }
2211
ac27a0ec
DK		 2212 /* We have now updated the journal if required, so we can
2213 * validate the data journaling mode. */
2214 switch (test_opt(sb, DATA_FLAGS)) {
2215 case 0:
2216 /* No mode set, assume a default based on the journal
63f57933
AM		 2217 * capabilities: ORDERED_DATA if the journal can
2218 * cope, else JOURNAL_DATA
2219 */
dab291af
MC		 2220 if (jbd2_journal_check_available_features
2221 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
ac27a0ec
DK		 2222 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2223 else
2224 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2225 break;
2226
617ba13b
MC		 2227 case EXT4_MOUNT_ORDERED_DATA:
2228 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 2229 if (!jbd2_journal_check_available_features
2230 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
617ba13b 2231 printk(KERN_ERR "EXT4-fs: Journal does not support "
ac27a0ec
DK		 2232 "requested data journaling mode\n");
2233 goto failed_mount4;
2234 }
2235 default:
2236 break;
2237 }
2238
2239 if (test_opt(sb, NOBH)) {
617ba13b
MC		 2240 if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
2241 printk(KERN_WARNING "EXT4-fs: Ignoring nobh option - "
ac27a0ec
DK		 2242 "its supported only with writeback mode\n");
2243 clear_opt(sbi->s_mount_opt, NOBH);
2244 }
2245 }
2246 /*
dab291af 2247 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 2248 * so we can safely mount the rest of the filesystem now.
2249 */
2250
1d1fe1ee
DH		 2251 root = ext4_iget(sb, EXT4_ROOT_INO);
2252 if (IS_ERR(root)) {
617ba13b 2253 printk(KERN_ERR "EXT4-fs: get root inode failed\n");
1d1fe1ee 2254 ret = PTR_ERR(root);
ac27a0ec
DK		 2255 goto failed_mount4;
2256 }
2257 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1d1fe1ee 2258 iput(root);
617ba13b 2259 printk(KERN_ERR "EXT4-fs: corrupt root inode, run e2fsck\n");
ac27a0ec
DK		 2260 goto failed_mount4;
2261 }
1d1fe1ee
DH		 2262 sb->s_root = d_alloc_root(root);
2263 if (!sb->s_root) {
2264 printk(KERN_ERR "EXT4-fs: get root dentry failed\n");
2265 iput(root);
2266 ret = -ENOMEM;
2267 goto failed_mount4;
2268 }
ac27a0ec 2269
617ba13b 2270 ext4_setup_super (sb, es, sb->s_flags & MS_RDONLY);
ef7f3835
KS		 2271
2272 /* determine the minimum size of new large inodes, if present */
2273 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2274 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2275 EXT4_GOOD_OLD_INODE_SIZE;
2276 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2277 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2278 if (sbi->s_want_extra_isize <
2279 le16_to_cpu(es->s_want_extra_isize))
2280 sbi->s_want_extra_isize =
2281 le16_to_cpu(es->s_want_extra_isize);
2282 if (sbi->s_want_extra_isize <
2283 le16_to_cpu(es->s_min_extra_isize))
2284 sbi->s_want_extra_isize =
2285 le16_to_cpu(es->s_min_extra_isize);
2286 }
2287 }
2288 /* Check if enough inode space is available */
2289 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2290 sbi->s_inode_size) {
2291 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2292 EXT4_GOOD_OLD_INODE_SIZE;
2293 printk(KERN_INFO "EXT4-fs: required extra inode space not"
2294 "available.\n");
2295 }
2296
ac27a0ec
DK		 2297 /*
2298 * akpm: core read_super() calls in here with the superblock locked.
2299 * That deadlocks, because orphan cleanup needs to lock the superblock
2300 * in numerous places. Here we just pop the lock - it's relatively
2301 * harmless, because we are now ready to accept write_super() requests,
2302 * and aviro says that's the only reason for hanging onto the
2303 * superblock lock.
2304 */
617ba13b
MC		 2305 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
2306 ext4_orphan_cleanup(sb, es);
2307 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
ac27a0ec 2308 if (needs_recovery)
617ba13b
MC		 2309 printk (KERN_INFO "EXT4-fs: recovery complete.\n");
2310 ext4_mark_recovery_complete(sb, es);
2311 printk (KERN_INFO "EXT4-fs: mounted filesystem with %s data mode.\n",
2312 test_opt(sb,DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA ? "journal":
2313 test_opt(sb,DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA ? "ordered":
ac27a0ec
DK		 2314 "writeback");
2315
a86c6181 2316 ext4_ext_init(sb);
c9de560d 2317 ext4_mb_init(sb, needs_recovery);
a86c6181 2318
ac27a0ec
DK		 2319 lock_kernel();
2320 return 0;
2321
617ba13b 2322cantfind_ext4:
ac27a0ec 2323 if (!silent)
617ba13b 2324 printk(KERN_ERR "VFS: Can't find ext4 filesystem on dev %s.\n",
ac27a0ec
DK		 2325 sb->s_id);
2326 goto failed_mount;
2327
2328failed_mount4:
dab291af 2329 jbd2_journal_destroy(sbi->s_journal);
ac27a0ec
DK		 2330failed_mount3:
2331 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2332 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2333 percpu_counter_destroy(&sbi->s_dirs_counter);
2334failed_mount2:
2335 for (i = 0; i < db_count; i++)
2336 brelse(sbi->s_group_desc[i]);
2337 kfree(sbi->s_group_desc);
2338failed_mount:
2339#ifdef CONFIG_QUOTA
2340 for (i = 0; i < MAXQUOTAS; i++)
2341 kfree(sbi->s_qf_names[i]);
2342#endif
617ba13b 2343 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 2344 brelse(bh);
2345out_fail:
2346 sb->s_fs_info = NULL;
2347 kfree(sbi);
2348 lock_kernel();
1d1fe1ee 2349 return ret;
ac27a0ec
DK		 2350}
2351
2352/*
2353 * Setup any per-fs journal parameters now. We'll do this both on
2354 * initial mount, once the journal has been initialised but before we've
2355 * done any recovery; and again on any subsequent remount.
2356 */
617ba13b 2357static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 2358{
617ba13b 2359 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 2360
2361 if (sbi->s_commit_interval)
2362 journal->j_commit_interval = sbi->s_commit_interval;
617ba13b 2363 /* We could also set up an ext4-specific default for the commit
ac27a0ec
DK		 2364 * interval here, but for now we'll just fall back to the jbd
2365 * default. */
2366
2367 spin_lock(&journal->j_state_lock);
2368 if (test_opt(sb, BARRIER))
dab291af 2369 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 2370 else
dab291af 2371 journal->j_flags &= ~JBD2_BARRIER;
ac27a0ec
DK		 2372 spin_unlock(&journal->j_state_lock);
2373}
2374
617ba13b 2375static journal_t *ext4_get_journal(struct super_block *sb,
ac27a0ec
DK		 2376 unsigned int journal_inum)
2377{
2378 struct inode *journal_inode;
2379 journal_t *journal;
2380
2381 /* First, test for the existence of a valid inode on disk. Bad
2382 * things happen if we iget() an unused inode, as the subsequent
2383 * iput() will try to delete it. */
2384
1d1fe1ee
DH		 2385 journal_inode = ext4_iget(sb, journal_inum);
2386 if (IS_ERR(journal_inode)) {
617ba13b 2387 printk(KERN_ERR "EXT4-fs: no journal found.\n");
ac27a0ec
DK		 2388 return NULL;
2389 }
2390 if (!journal_inode->i_nlink) {
2391 make_bad_inode(journal_inode);
2392 iput(journal_inode);
617ba13b 2393 printk(KERN_ERR "EXT4-fs: journal inode is deleted.\n");
ac27a0ec
DK		 2394 return NULL;
2395 }
2396
2397 jbd_debug(2, "Journal inode found at %p: %Ld bytes\n",
2398 journal_inode, journal_inode->i_size);
1d1fe1ee 2399 if (!S_ISREG(journal_inode->i_mode)) {
617ba13b 2400 printk(KERN_ERR "EXT4-fs: invalid journal inode.\n");
ac27a0ec
DK		 2401 iput(journal_inode);
2402 return NULL;
2403 }
2404
dab291af 2405 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 2406 if (!journal) {
617ba13b 2407 printk(KERN_ERR "EXT4-fs: Could not load journal inode\n");
ac27a0ec
DK		 2408 iput(journal_inode);
2409 return NULL;
2410 }
2411 journal->j_private = sb;
617ba13b 2412 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 2413 return journal;
2414}
2415
617ba13b 2416static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 2417 dev_t j_dev)
2418{
2419 struct buffer_head * bh;
2420 journal_t *journal;
617ba13b
MC		 2421 ext4_fsblk_t start;
2422 ext4_fsblk_t len;
ac27a0ec 2423 int hblock, blocksize;
617ba13b 2424 ext4_fsblk_t sb_block;
ac27a0ec 2425 unsigned long offset;
617ba13b 2426 struct ext4_super_block * es;
ac27a0ec
DK		 2427 struct block_device *bdev;
2428
617ba13b 2429 bdev = ext4_blkdev_get(j_dev);
ac27a0ec
DK		 2430 if (bdev == NULL)
2431 return NULL;
2432
2433 if (bd_claim(bdev, sb)) {
2434 printk(KERN_ERR
8c55e204 2435 "EXT4: failed to claim external journal device.\n");
ac27a0ec
DK		 2436 blkdev_put(bdev);
2437 return NULL;
2438 }
2439
2440 blocksize = sb->s_blocksize;
2441 hblock = bdev_hardsect_size(bdev);
2442 if (blocksize < hblock) {
2443 printk(KERN_ERR
617ba13b 2444 "EXT4-fs: blocksize too small for journal device.\n");
ac27a0ec
DK		 2445 goto out_bdev;
2446 }
2447
617ba13b
MC		 2448 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
2449 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 2450 set_blocksize(bdev, blocksize);
2451 if (!(bh = __bread(bdev, sb_block, blocksize))) {
617ba13b 2452 printk(KERN_ERR "EXT4-fs: couldn't read superblock of "
ac27a0ec
DK		 2453 "external journal\n");
2454 goto out_bdev;
2455 }
2456
617ba13b
MC		 2457 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2458 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 2459 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b
MC		 2460 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
2461 printk(KERN_ERR "EXT4-fs: external journal has "
ac27a0ec
DK		 2462 "bad superblock\n");
2463 brelse(bh);
2464 goto out_bdev;
2465 }
2466
617ba13b
MC		 2467 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
2468 printk(KERN_ERR "EXT4-fs: journal UUID does not match\n");
ac27a0ec
DK		 2469 brelse(bh);
2470 goto out_bdev;
2471 }
2472
bd81d8ee 2473 len = ext4_blocks_count(es);
ac27a0ec
DK		 2474 start = sb_block + 1;
2475 brelse(bh); /* we're done with the superblock */
2476
dab291af 2477 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 2478 start, len, blocksize);
2479 if (!journal) {
617ba13b 2480 printk(KERN_ERR "EXT4-fs: failed to create device journal\n");
ac27a0ec
DK		 2481 goto out_bdev;
2482 }
2483 journal->j_private = sb;
2484 ll_rw_block(READ, 1, &journal->j_sb_buffer);
2485 wait_on_buffer(journal->j_sb_buffer);
2486 if (!buffer_uptodate(journal->j_sb_buffer)) {
617ba13b 2487 printk(KERN_ERR "EXT4-fs: I/O error on journal device\n");
ac27a0ec
DK		 2488 goto out_journal;
2489 }
2490 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
617ba13b 2491 printk(KERN_ERR "EXT4-fs: External journal has more than one "
ac27a0ec
DK		 2492 "user (unsupported) - %d\n",
2493 be32_to_cpu(journal->j_superblock->s_nr_users));
2494 goto out_journal;
2495 }
617ba13b
MC		 2496 EXT4_SB(sb)->journal_bdev = bdev;
2497 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 2498 return journal;
2499out_journal:
dab291af 2500 jbd2_journal_destroy(journal);
ac27a0ec 2501out_bdev:
617ba13b 2502 ext4_blkdev_put(bdev);
ac27a0ec
DK		 2503 return NULL;
2504}
2505
617ba13b
MC		 2506static int ext4_load_journal(struct super_block *sb,
2507 struct ext4_super_block *es,
ac27a0ec
DK		 2508 unsigned long journal_devnum)
2509{
2510 journal_t *journal;
2511 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
2512 dev_t journal_dev;
2513 int err = 0;
2514 int really_read_only;
2515
2516 if (journal_devnum &&
2517 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
617ba13b 2518 printk(KERN_INFO "EXT4-fs: external journal device major/minor "
ac27a0ec
DK		 2519 "numbers have changed\n");
2520 journal_dev = new_decode_dev(journal_devnum);
2521 } else
2522 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
2523
2524 really_read_only = bdev_read_only(sb->s_bdev);
2525
2526 /*
2527 * Are we loading a blank journal or performing recovery after a
2528 * crash? For recovery, we need to check in advance whether we
2529 * can get read-write access to the device.
2530 */
2531
617ba13b 2532 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
ac27a0ec 2533 if (sb->s_flags & MS_RDONLY) {
617ba13b 2534 printk(KERN_INFO "EXT4-fs: INFO: recovery "
ac27a0ec
DK		 2535 "required on readonly filesystem.\n");
2536 if (really_read_only) {
617ba13b 2537 printk(KERN_ERR "EXT4-fs: write access "
ac27a0ec
DK		 2538 "unavailable, cannot proceed.\n");
2539 return -EROFS;
2540 }
617ba13b 2541 printk (KERN_INFO "EXT4-fs: write access will "
ac27a0ec
DK		 2542 "be enabled during recovery.\n");
2543 }
2544 }
2545
2546 if (journal_inum && journal_dev) {
617ba13b 2547 printk(KERN_ERR "EXT4-fs: filesystem has both journal "
ac27a0ec
DK		 2548 "and inode journals!\n");
2549 return -EINVAL;
2550 }
2551
2552 if (journal_inum) {
617ba13b 2553 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 2554 return -EINVAL;
2555 } else {
617ba13b 2556 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 2557 return -EINVAL;
2558 }
2559
2560 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
dab291af 2561 err = jbd2_journal_update_format(journal);
ac27a0ec 2562 if (err) {
617ba13b 2563 printk(KERN_ERR "EXT4-fs: error updating journal.\n");
dab291af 2564 jbd2_journal_destroy(journal);
ac27a0ec
DK		 2565 return err;
2566 }
2567 }
2568
617ba13b 2569 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
dab291af 2570 err = jbd2_journal_wipe(journal, !really_read_only);
ac27a0ec 2571 if (!err)
dab291af 2572 err = jbd2_journal_load(journal);
ac27a0ec
DK		 2573
2574 if (err) {
617ba13b 2575 printk(KERN_ERR "EXT4-fs: error loading journal.\n");
dab291af 2576 jbd2_journal_destroy(journal);
ac27a0ec
DK		 2577 return err;
2578 }
2579
617ba13b
MC		 2580 EXT4_SB(sb)->s_journal = journal;
2581 ext4_clear_journal_err(sb, es);
ac27a0ec
DK		 2582
2583 if (journal_devnum &&
2584 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2585 es->s_journal_dev = cpu_to_le32(journal_devnum);
2586 sb->s_dirt = 1;
2587
2588 /* Make sure we flush the recovery flag to disk. */
617ba13b 2589 ext4_commit_super(sb, es, 1);
ac27a0ec
DK		 2590 }
2591
2592 return 0;
2593}
2594
617ba13b
MC		 2595static int ext4_create_journal(struct super_block * sb,
2596 struct ext4_super_block * es,
ac27a0ec
DK		 2597 unsigned int journal_inum)
2598{
2599 journal_t *journal;
6c675bd4 2600 int err;
ac27a0ec
DK		 2601
2602 if (sb->s_flags & MS_RDONLY) {
617ba13b 2603 printk(KERN_ERR "EXT4-fs: readonly filesystem when trying to "
ac27a0ec
DK		 2604 "create journal.\n");
2605 return -EROFS;
2606 }
2607
6c675bd4
BP		 2608 journal = ext4_get_journal(sb, journal_inum);
2609 if (!journal)
ac27a0ec
DK		 2610 return -EINVAL;
2611
617ba13b 2612 printk(KERN_INFO "EXT4-fs: creating new journal on inode %u\n",
ac27a0ec
DK		 2613 journal_inum);
2614
6c675bd4
BP		 2615 err = jbd2_journal_create(journal);
2616 if (err) {
617ba13b 2617 printk(KERN_ERR "EXT4-fs: error creating journal.\n");
dab291af 2618 jbd2_journal_destroy(journal);
ac27a0ec
DK		 2619 return -EIO;
2620 }
2621
617ba13b 2622 EXT4_SB(sb)->s_journal = journal;
ac27a0ec 2623
617ba13b
MC		 2624 ext4_update_dynamic_rev(sb);
2625 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2626 EXT4_SET_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL);
ac27a0ec
DK		 2627
2628 es->s_journal_inum = cpu_to_le32(journal_inum);
2629 sb->s_dirt = 1;
2630
2631 /* Make sure we flush the recovery flag to disk. */
617ba13b 2632 ext4_commit_super(sb, es, 1);
ac27a0ec
DK		 2633
2634 return 0;
2635}
2636
617ba13b
MC		 2637static void ext4_commit_super (struct super_block * sb,
2638 struct ext4_super_block * es,
ac27a0ec
DK		 2639 int sync)
2640{
617ba13b 2641 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
ac27a0ec
DK		 2642
2643 if (!sbh)
2644 return;
2645 es->s_wtime = cpu_to_le32(get_seconds());
bd81d8ee 2646 ext4_free_blocks_count_set(es, ext4_count_free_blocks(sb));
617ba13b 2647 es->s_free_inodes_count = cpu_to_le32(ext4_count_free_inodes(sb));
ac27a0ec
DK		 2648 BUFFER_TRACE(sbh, "marking dirty");
2649 mark_buffer_dirty(sbh);
2650 if (sync)
2651 sync_dirty_buffer(sbh);
2652}
2653
2654
2655/*
2656 * Have we just finished recovery? If so, and if we are mounting (or
2657 * remounting) the filesystem readonly, then we will end up with a
2658 * consistent fs on disk. Record that fact.
2659 */
617ba13b
MC		 2660static void ext4_mark_recovery_complete(struct super_block * sb,
2661 struct ext4_super_block * es)
ac27a0ec 2662{
617ba13b 2663 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 2664
dab291af
MC		 2665 jbd2_journal_lock_updates(journal);
2666 jbd2_journal_flush(journal);
32c37730 2667 lock_super(sb);
617ba13b 2668 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
ac27a0ec 2669 sb->s_flags & MS_RDONLY) {
617ba13b 2670 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 2671 sb->s_dirt = 0;
617ba13b 2672 ext4_commit_super(sb, es, 1);
ac27a0ec 2673 }
32c37730 2674 unlock_super(sb);
dab291af 2675 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 2676}
2677
2678/*
2679 * If we are mounting (or read-write remounting) a filesystem whose journal
2680 * has recorded an error from a previous lifetime, move that error to the
2681 * main filesystem now.
2682 */
617ba13b
MC		 2683static void ext4_clear_journal_err(struct super_block * sb,
2684 struct ext4_super_block * es)
ac27a0ec
DK		 2685{
2686 journal_t *journal;
2687 int j_errno;
2688 const char *errstr;
2689
617ba13b 2690 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 2691
2692 /*
2693 * Now check for any error status which may have been recorded in the
617ba13b 2694 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 2695 */
2696
dab291af 2697 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 2698 if (j_errno) {
2699 char nbuf[16];
2700
617ba13b
MC		 2701 errstr = ext4_decode_error(sb, j_errno, nbuf);
2702 ext4_warning(sb, __FUNCTION__, "Filesystem error recorded "
ac27a0ec 2703 "from previous mount: %s", errstr);
617ba13b 2704 ext4_warning(sb, __FUNCTION__, "Marking fs in need of "
ac27a0ec
DK		 2705 "filesystem check.");
2706
617ba13b
MC		 2707 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2708 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2709 ext4_commit_super (sb, es, 1);
ac27a0ec 2710
dab291af 2711 jbd2_journal_clear_err(journal);
ac27a0ec
DK		 2712 }
2713}
2714
2715/*
2716 * Force the running and committing transactions to commit,
2717 * and wait on the commit.
2718 */
617ba13b 2719int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 2720{
2721 journal_t *journal;
2722 int ret;
2723
2724 if (sb->s_flags & MS_RDONLY)
2725 return 0;
2726
617ba13b 2727 journal = EXT4_SB(sb)->s_journal;
ac27a0ec 2728 sb->s_dirt = 0;
617ba13b 2729 ret = ext4_journal_force_commit(journal);
ac27a0ec
DK		 2730 return ret;
2731}
2732
2733/*
617ba13b 2734 * Ext4 always journals updates to the superblock itself, so we don't
ac27a0ec
DK		 2735 * have to propagate any other updates to the superblock on disk at this
2736 * point. Just start an async writeback to get the buffers on their way
2737 * to the disk.
2738 *
2739 * This implicitly triggers the writebehind on sync().
2740 */
2741
617ba13b 2742static void ext4_write_super (struct super_block * sb)
ac27a0ec
DK		 2743{
2744 if (mutex_trylock(&sb->s_lock) != 0)
2745 BUG();
2746 sb->s_dirt = 0;
2747}
2748
617ba13b 2749static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec
DK		 2750{
2751 tid_t target;
2752
2753 sb->s_dirt = 0;
dab291af 2754 if (jbd2_journal_start_commit(EXT4_SB(sb)->s_journal, &target)) {
ac27a0ec 2755 if (wait)
dab291af 2756 jbd2_log_wait_commit(EXT4_SB(sb)->s_journal, target);
ac27a0ec
DK		 2757 }
2758 return 0;
2759}
2760
2761/*
2762 * LVM calls this function before a (read-only) snapshot is created. This
2763 * gives us a chance to flush the journal completely and mark the fs clean.
2764 */
617ba13b 2765static void ext4_write_super_lockfs(struct super_block *sb)
ac27a0ec
DK		 2766{
2767 sb->s_dirt = 0;
2768
2769 if (!(sb->s_flags & MS_RDONLY)) {
617ba13b 2770 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 2771
2772 /* Now we set up the journal barrier. */
dab291af
MC		 2773 jbd2_journal_lock_updates(journal);
2774 jbd2_journal_flush(journal);
ac27a0ec
DK		 2775
2776 /* Journal blocked and flushed, clear needs_recovery flag. */
617ba13b
MC		 2777 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2778 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
ac27a0ec
DK		 2779 }
2780}
2781
2782/*
2783 * Called by LVM after the snapshot is done. We need to reset the RECOVER
2784 * flag here, even though the filesystem is not technically dirty yet.
2785 */
617ba13b 2786static void ext4_unlockfs(struct super_block *sb)
ac27a0ec
DK		 2787{
2788 if (!(sb->s_flags & MS_RDONLY)) {
2789 lock_super(sb);
2790 /* Reser the needs_recovery flag before the fs is unlocked. */
617ba13b
MC		 2791 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2792 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
ac27a0ec 2793 unlock_super(sb);
dab291af 2794 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
ac27a0ec
DK		 2795 }
2796}
2797
617ba13b 2798static int ext4_remount (struct super_block * sb, int * flags, char * data)
ac27a0ec 2799{
617ba13b
MC		 2800 struct ext4_super_block * es;
2801 struct ext4_sb_info *sbi = EXT4_SB(sb);
2802 ext4_fsblk_t n_blocks_count = 0;
ac27a0ec 2803 unsigned long old_sb_flags;
617ba13b 2804 struct ext4_mount_options old_opts;
ac27a0ec
DK		 2805 int err;
2806#ifdef CONFIG_QUOTA
2807 int i;
2808#endif
2809
2810 /* Store the original options */
2811 old_sb_flags = sb->s_flags;
2812 old_opts.s_mount_opt = sbi->s_mount_opt;
2813 old_opts.s_resuid = sbi->s_resuid;
2814 old_opts.s_resgid = sbi->s_resgid;
2815 old_opts.s_commit_interval = sbi->s_commit_interval;
2816#ifdef CONFIG_QUOTA
2817 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
2818 for (i = 0; i < MAXQUOTAS; i++)
2819 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
2820#endif
2821
2822 /*
2823 * Allow the "check" option to be passed as a remount option.
2824 */
2825 if (!parse_options(data, sb, NULL, NULL, &n_blocks_count, 1)) {
2826 err = -EINVAL;
2827 goto restore_opts;
2828 }
2829
617ba13b
MC		 2830 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT)
2831 ext4_abort(sb, __FUNCTION__, "Abort forced by user");
ac27a0ec
DK		 2832
2833 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
617ba13b 2834 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec
DK		 2835
2836 es = sbi->s_es;
2837
617ba13b 2838 ext4_init_journal_params(sb, sbi->s_journal);
ac27a0ec
DK		 2839
2840 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
bd81d8ee 2841 n_blocks_count > ext4_blocks_count(es)) {
617ba13b 2842 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT) {
ac27a0ec
DK		 2843 err = -EROFS;
2844 goto restore_opts;
2845 }
2846
2847 if (*flags & MS_RDONLY) {
2848 /*
2849 * First of all, the unconditional stuff we have to do
2850 * to disable replay of the journal when we next remount
2851 */
2852 sb->s_flags |= MS_RDONLY;
2853
2854 /*
2855 * OK, test if we are remounting a valid rw partition
2856 * readonly, and if so set the rdonly flag and then
2857 * mark the partition as valid again.
2858 */
617ba13b
MC		 2859 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
2860 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 2861 es->s_state = cpu_to_le16(sbi->s_mount_state);
2862
32c37730
JK		 2863 /*
2864 * We have to unlock super so that we can wait for
2865 * transactions.
2866 */
2867 unlock_super(sb);
617ba13b 2868 ext4_mark_recovery_complete(sb, es);
32c37730 2869 lock_super(sb);
ac27a0ec
DK		 2870 } else {
2871 __le32 ret;
617ba13b
MC		 2872 if ((ret = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2873 ~EXT4_FEATURE_RO_COMPAT_SUPP))) {
2874 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
ac27a0ec
DK		 2875 "remount RDWR because of unsupported "
2876 "optional features (%x).\n",
2877 sb->s_id, le32_to_cpu(ret));
2878 err = -EROFS;
2879 goto restore_opts;
2880 }
ead6596b
ES		 2881
2882 /*
2883 * If we have an unprocessed orphan list hanging
2884 * around from a previously readonly bdev mount,
2885 * require a full umount/remount for now.
2886 */
2887 if (es->s_last_orphan) {
2888 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
2889 "remount RDWR because of unprocessed "
2890 "orphan inode list. Please "
2891 "umount/remount instead.\n",
2892 sb->s_id);
2893 err = -EINVAL;
2894 goto restore_opts;
2895 }
2896
ac27a0ec
DK		 2897 /*
2898 * Mounting a RDONLY partition read-write, so reread
2899 * and store the current valid flag. (It may have
2900 * been changed by e2fsck since we originally mounted
2901 * the partition.)
2902 */
617ba13b 2903 ext4_clear_journal_err(sb, es);
ac27a0ec 2904 sbi->s_mount_state = le16_to_cpu(es->s_state);
617ba13b 2905 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
ac27a0ec 2906 goto restore_opts;
617ba13b 2907 if (!ext4_setup_super (sb, es, 0))
ac27a0ec
DK		 2908 sb->s_flags &= ~MS_RDONLY;
2909 }
2910 }
2911#ifdef CONFIG_QUOTA
2912 /* Release old quota file names */
2913 for (i = 0; i < MAXQUOTAS; i++)
2914 if (old_opts.s_qf_names[i] &&
2915 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2916 kfree(old_opts.s_qf_names[i]);
2917#endif
2918 return 0;
2919restore_opts:
2920 sb->s_flags = old_sb_flags;
2921 sbi->s_mount_opt = old_opts.s_mount_opt;
2922 sbi->s_resuid = old_opts.s_resuid;
2923 sbi->s_resgid = old_opts.s_resgid;
2924 sbi->s_commit_interval = old_opts.s_commit_interval;
2925#ifdef CONFIG_QUOTA
2926 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
2927 for (i = 0; i < MAXQUOTAS; i++) {
2928 if (sbi->s_qf_names[i] &&
2929 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2930 kfree(sbi->s_qf_names[i]);
2931 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
2932 }
2933#endif
2934 return err;
2935}
2936
617ba13b 2937static int ext4_statfs (struct dentry * dentry, struct kstatfs * buf)
ac27a0ec
DK		 2938{
2939 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 2940 struct ext4_sb_info *sbi = EXT4_SB(sb);
2941 struct ext4_super_block *es = sbi->s_es;
960cc398 2942 u64 fsid;
ac27a0ec 2943
5e70030d
BP		 2944 if (test_opt(sb, MINIX_DF)) {
2945 sbi->s_overhead_last = 0;
6bc9feff 2946 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
fd2d4291 2947 ext4_group_t ngroups = sbi->s_groups_count, i;
5e70030d 2948 ext4_fsblk_t overhead = 0;
ac27a0ec
DK		 2949 smp_rmb();
2950
2951 /*
5e70030d
BP		 2952 * Compute the overhead (FS structures). This is constant
2953 * for a given filesystem unless the number of block groups
2954 * changes so we cache the previous value until it does.
ac27a0ec
DK		 2955 */
2956
2957 /*
2958 * All of the blocks before first_data_block are
2959 * overhead
2960 */
2961 overhead = le32_to_cpu(es->s_first_data_block);
2962
2963 /*
2964 * Add the overhead attributed to the superblock and
2965 * block group descriptors. If the sparse superblocks
2966 * feature is turned on, then not all groups have this.
2967 */
2968 for (i = 0; i < ngroups; i++) {
617ba13b
MC		 2969 overhead += ext4_bg_has_super(sb, i) +
2970 ext4_bg_num_gdb(sb, i);
ac27a0ec
DK		 2971 cond_resched();
2972 }
2973
2974 /*
2975 * Every block group has an inode bitmap, a block
2976 * bitmap, and an inode table.
2977 */
5e70030d
BP		 2978 overhead += ngroups * (2 + sbi->s_itb_per_group);
2979 sbi->s_overhead_last = overhead;
2980 smp_wmb();
6bc9feff 2981 sbi->s_blocks_last = ext4_blocks_count(es);
ac27a0ec
DK		 2982 }
2983
617ba13b 2984 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 2985 buf->f_bsize = sb->s_blocksize;
5e70030d 2986 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
52d9f3b4 2987 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter);
308ba3ec 2988 ext4_free_blocks_count_set(es, buf->f_bfree);
bd81d8ee
LV		 2989 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
2990 if (buf->f_bfree < ext4_r_blocks_count(es))
ac27a0ec
DK		 2991 buf->f_bavail = 0;
2992 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 2993 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
5e70030d 2994 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
617ba13b 2995 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 2996 fsid = le64_to_cpup((void *)es->s_uuid) ^
2997 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
2998 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
2999 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
ac27a0ec
DK		 3000 return 0;
3001}
3002
3003/* Helper function for writing quotas on sync - we need to start transaction before quota file
3004 * is locked for write. Otherwise the are possible deadlocks:
3005 * Process 1 Process 2
617ba13b 3006 * ext4_create() quota_sync()
dab291af 3007 * jbd2_journal_start() write_dquot()
ac27a0ec 3008 * DQUOT_INIT() down(dqio_mutex)
dab291af 3009 * down(dqio_mutex) jbd2_journal_start()
ac27a0ec
DK		 3010 *
3011 */
3012
3013#ifdef CONFIG_QUOTA
3014
3015static inline struct inode *dquot_to_inode(struct dquot *dquot)
3016{
3017 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3018}
3019
617ba13b 3020static int ext4_dquot_initialize(struct inode *inode, int type)
ac27a0ec
DK		 3021{
3022 handle_t *handle;
3023 int ret, err;
3024
3025 /* We may create quota structure so we need to reserve enough blocks */
617ba13b 3026 handle = ext4_journal_start(inode, 2*EXT4_QUOTA_INIT_BLOCKS(inode->i_sb));
ac27a0ec
DK		 3027 if (IS_ERR(handle))
3028 return PTR_ERR(handle);
3029 ret = dquot_initialize(inode, type);
617ba13b 3030 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3031 if (!ret)
3032 ret = err;
3033 return ret;
3034}
3035
617ba13b 3036static int ext4_dquot_drop(struct inode *inode)
ac27a0ec
DK		 3037{
3038 handle_t *handle;
3039 int ret, err;
3040
3041 /* We may delete quota structure so we need to reserve enough blocks */
617ba13b 3042 handle = ext4_journal_start(inode, 2*EXT4_QUOTA_DEL_BLOCKS(inode->i_sb));
ac27a0ec
DK		 3043 if (IS_ERR(handle))
3044 return PTR_ERR(handle);
3045 ret = dquot_drop(inode);
617ba13b 3046 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3047 if (!ret)
3048 ret = err;
3049 return ret;
3050}
3051
617ba13b 3052static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 3053{
3054 int ret, err;
3055 handle_t *handle;
3056 struct inode *inode;
3057
3058 inode = dquot_to_inode(dquot);
617ba13b
MC		 3059 handle = ext4_journal_start(inode,
3060 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 3061 if (IS_ERR(handle))
3062 return PTR_ERR(handle);
3063 ret = dquot_commit(dquot);
617ba13b 3064 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3065 if (!ret)
3066 ret = err;
3067 return ret;
3068}
3069
617ba13b 3070static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 3071{
3072 int ret, err;
3073 handle_t *handle;
3074
617ba13b
MC		 3075 handle = ext4_journal_start(dquot_to_inode(dquot),
3076 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 3077 if (IS_ERR(handle))
3078 return PTR_ERR(handle);
3079 ret = dquot_acquire(dquot);
617ba13b 3080 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3081 if (!ret)
3082 ret = err;
3083 return ret;
3084}
3085
617ba13b 3086static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 3087{
3088 int ret, err;
3089 handle_t *handle;
3090
617ba13b
MC		 3091 handle = ext4_journal_start(dquot_to_inode(dquot),
3092 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 3093 if (IS_ERR(handle)) {
3094 /* Release dquot anyway to avoid endless cycle in dqput() */
3095 dquot_release(dquot);
ac27a0ec 3096 return PTR_ERR(handle);
9c3013e9 3097 }
ac27a0ec 3098 ret = dquot_release(dquot);
617ba13b 3099 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3100 if (!ret)
3101 ret = err;
3102 return ret;
3103}
3104
617ba13b 3105static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec
DK		 3106{
3107 /* Are we journalling quotas? */
617ba13b
MC		 3108 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3109 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
ac27a0ec 3110 dquot_mark_dquot_dirty(dquot);
617ba13b 3111 return ext4_write_dquot(dquot);
ac27a0ec
DK		 3112 } else {
3113 return dquot_mark_dquot_dirty(dquot);
3114 }
3115}
3116
617ba13b 3117static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 3118{
3119 int ret, err;
3120 handle_t *handle;
3121
3122 /* Data block + inode block */
617ba13b 3123 handle = ext4_journal_start(sb->s_root->d_inode, 2);
ac27a0ec
DK		 3124 if (IS_ERR(handle))
3125 return PTR_ERR(handle);
3126 ret = dquot_commit_info(sb, type);
617ba13b 3127 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3128 if (!ret)
3129 ret = err;
3130 return ret;
3131}
3132
3133/*
3134 * Turn on quotas during mount time - we need to find
3135 * the quota file and such...
3136 */
617ba13b 3137static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 3138{
617ba13b
MC		 3139 return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
3140 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 3141}
3142
3143/*
3144 * Standard function to be called on quota_on
3145 */
617ba13b 3146static int ext4_quota_on(struct super_block *sb, int type, int format_id,
6f28e087 3147 char *path, int remount)
ac27a0ec
DK		 3148{
3149 int err;
3150 struct nameidata nd;
3151
3152 if (!test_opt(sb, QUOTA))
3153 return -EINVAL;
3154 /* Not journalling quota? */
6f28e087
JK		 3155 if ((!EXT4_SB(sb)->s_qf_names[USRQUOTA] &&
3156 !EXT4_SB(sb)->s_qf_names[GRPQUOTA]) || remount)
3157 return vfs_quota_on(sb, type, format_id, path, remount);
ac27a0ec
DK		 3158 err = path_lookup(path, LOOKUP_FOLLOW, &nd);
3159 if (err)
3160 return err;
3161 /* Quotafile not on the same filesystem? */
4ac91378 3162 if (nd.path.mnt->mnt_sb != sb) {
1d957f9b 3163 path_put(&nd.path);
ac27a0ec
DK		 3164 return -EXDEV;
3165 }
3166 /* Quotafile not of fs root? */
4ac91378 3167 if (nd.path.dentry->d_parent->d_inode != sb->s_root->d_inode)
ac27a0ec 3168 printk(KERN_WARNING
617ba13b 3169 "EXT4-fs: Quota file not on filesystem root. "
ac27a0ec 3170 "Journalled quota will not work.\n");
1d957f9b 3171 path_put(&nd.path);
6f28e087 3172 return vfs_quota_on(sb, type, format_id, path, remount);
ac27a0ec
DK		 3173}
3174
3175/* Read data from quotafile - avoid pagecache and such because we cannot afford
3176 * acquiring the locks... As quota files are never truncated and quota code
3177 * itself serializes the operations (and noone else should touch the files)
3178 * we don't have to be afraid of races */
617ba13b 3179static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 3180 size_t len, loff_t off)
3181{
3182 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 3183 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 3184 int err = 0;
3185 int offset = off & (sb->s_blocksize - 1);
3186 int tocopy;
3187 size_t toread;
3188 struct buffer_head *bh;
3189 loff_t i_size = i_size_read(inode);
3190
3191 if (off > i_size)
3192 return 0;
3193 if (off+len > i_size)
3194 len = i_size-off;
3195 toread = len;
3196 while (toread > 0) {
3197 tocopy = sb->s_blocksize - offset < toread ?
3198 sb->s_blocksize - offset : toread;
617ba13b 3199 bh = ext4_bread(NULL, inode, blk, 0, &err);
ac27a0ec
DK		 3200 if (err)
3201 return err;
3202 if (!bh) /* A hole? */
3203 memset(data, 0, tocopy);
3204 else
3205 memcpy(data, bh->b_data+offset, tocopy);
3206 brelse(bh);
3207 offset = 0;
3208 toread -= tocopy;
3209 data += tocopy;
3210 blk++;
3211 }
3212 return len;
3213}
3214
3215/* Write to quotafile (we know the transaction is already started and has
3216 * enough credits) */
617ba13b 3217static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 3218 const char *data, size_t len, loff_t off)
3219{
3220 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 3221 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 3222 int err = 0;
3223 int offset = off & (sb->s_blocksize - 1);
3224 int tocopy;
617ba13b 3225 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
ac27a0ec
DK		 3226 size_t towrite = len;
3227 struct buffer_head *bh;
3228 handle_t *handle = journal_current_handle();
3229
9c3013e9
JK		 3230 if (!handle) {
3231 printk(KERN_WARNING "EXT4-fs: Quota write (off=%Lu, len=%Lu)"
3232 " cancelled because transaction is not started.\n",
3233 (unsigned long long)off, (unsigned long long)len);
3234 return -EIO;
3235 }
ac27a0ec
DK		 3236 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
3237 while (towrite > 0) {
3238 tocopy = sb->s_blocksize - offset < towrite ?
3239 sb->s_blocksize - offset : towrite;
617ba13b 3240 bh = ext4_bread(handle, inode, blk, 1, &err);
ac27a0ec
DK		 3241 if (!bh)
3242 goto out;
3243 if (journal_quota) {
617ba13b 3244 err = ext4_journal_get_write_access(handle, bh);
ac27a0ec
DK		 3245 if (err) {
3246 brelse(bh);
3247 goto out;
3248 }
3249 }
3250 lock_buffer(bh);
3251 memcpy(bh->b_data+offset, data, tocopy);
3252 flush_dcache_page(bh->b_page);
3253 unlock_buffer(bh);
3254 if (journal_quota)
617ba13b 3255 err = ext4_journal_dirty_metadata(handle, bh);
ac27a0ec
DK		 3256 else {
3257 /* Always do at least ordered writes for quotas */
617ba13b 3258 err = ext4_journal_dirty_data(handle, bh);
ac27a0ec
DK		 3259 mark_buffer_dirty(bh);
3260 }
3261 brelse(bh);
3262 if (err)
3263 goto out;
3264 offset = 0;
3265 towrite -= tocopy;
3266 data += tocopy;
3267 blk++;
3268 }
3269out:
3270 if (len == towrite)
3271 return err;
3272 if (inode->i_size < off+len-towrite) {
3273 i_size_write(inode, off+len-towrite);
617ba13b 3274 EXT4_I(inode)->i_disksize = inode->i_size;
ac27a0ec 3275 }
ac27a0ec 3276 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
617ba13b 3277 ext4_mark_inode_dirty(handle, inode);
ac27a0ec
DK		 3278 mutex_unlock(&inode->i_mutex);
3279 return len - towrite;
3280}
3281
3282#endif
3283
617ba13b 3284static int ext4_get_sb(struct file_system_type *fs_type,
ac27a0ec
DK		 3285 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3286{
617ba13b 3287 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
ac27a0ec
DK		 3288}
3289
617ba13b 3290static struct file_system_type ext4dev_fs_type = {
ac27a0ec 3291 .owner = THIS_MODULE,
617ba13b
MC		 3292 .name = "ext4dev",
3293 .get_sb = ext4_get_sb,
ac27a0ec
DK		 3294 .kill_sb = kill_block_super,
3295 .fs_flags = FS_REQUIRES_DEV,
3296};
3297
617ba13b 3298static int __init init_ext4_fs(void)
ac27a0ec 3299{
c9de560d
AT		 3300 int err;
3301
3302 err = init_ext4_mballoc();
ac27a0ec
DK		 3303 if (err)
3304 return err;
c9de560d
AT		 3305
3306 err = init_ext4_xattr();
3307 if (err)
3308 goto out2;
ac27a0ec
DK		 3309 err = init_inodecache();
3310 if (err)
3311 goto out1;
63f57933 3312 err = register_filesystem(&ext4dev_fs_type);
ac27a0ec
DK		 3313 if (err)
3314 goto out;
3315 return 0;
3316out:
3317 destroy_inodecache();
3318out1:
617ba13b 3319 exit_ext4_xattr();
c9de560d
AT		 3320out2:
3321 exit_ext4_mballoc();
ac27a0ec
DK		 3322 return err;
3323}
3324
617ba13b 3325static void __exit exit_ext4_fs(void)
ac27a0ec 3326{
617ba13b 3327 unregister_filesystem(&ext4dev_fs_type);
ac27a0ec 3328 destroy_inodecache();
617ba13b 3329 exit_ext4_xattr();
c9de560d 3330 exit_ext4_mballoc();
ac27a0ec
DK		 3331}
3332
3333MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
617ba13b 3334MODULE_DESCRIPTION("Fourth Extended Filesystem with extents");
ac27a0ec 3335MODULE_LICENSE("GPL");
617ba13b
MC		 3336module_init(init_ext4_fs)
3337module_exit(exit_ext4_fs)
kernel source for telekom puls (alcatel/mediatek)