projects / deliverable / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
ext[234]: fix comment for nonexistent variable
[deliverable/linux.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 */
780 inode = iget(sb, ino);
781 if (inode == NULL)
782 return ERR_PTR(-ENOMEM);
783 if (is_bad_inode(inode) ||
784 (generation && inode->i_generation != generation)) {
785 iput(inode);
786 return ERR_PTR(-ESTALE);
787 }
1b961ac0
CH		 788
789 return inode;
790}
791
792static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
793 int fh_len, int fh_type)
794{
795 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
796 ext4_nfs_get_inode);
797}
798
799static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
800 int fh_len, int fh_type)
801{
802 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
803 ext4_nfs_get_inode);
ac27a0ec
DK		 804}
805
806#ifdef CONFIG_QUOTA
807#define QTYPE2NAME(t) ((t)==USRQUOTA?"user":"group")
808#define QTYPE2MOPT(on, t) ((t)==USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
809
617ba13b
MC		 810static int ext4_dquot_initialize(struct inode *inode, int type);
811static int ext4_dquot_drop(struct inode *inode);
812static int ext4_write_dquot(struct dquot *dquot);
813static int ext4_acquire_dquot(struct dquot *dquot);
814static int ext4_release_dquot(struct dquot *dquot);
815static int ext4_mark_dquot_dirty(struct dquot *dquot);
816static int ext4_write_info(struct super_block *sb, int type);
817static int ext4_quota_on(struct super_block *sb, int type, int format_id, char *path);
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,
853 .read_inode = ext4_read_inode,
854 .write_inode = ext4_write_inode,
855 .dirty_inode = ext4_dirty_inode,
856 .delete_inode = ext4_delete_inode,
857 .put_super = ext4_put_super,
858 .write_super = ext4_write_super,
859 .sync_fs = ext4_sync_fs,
860 .write_super_lockfs = ext4_write_super_lockfs,
861 .unlockfs = ext4_unlockfs,
862 .statfs = ext4_statfs,
863 .remount_fs = ext4_remount,
864 .clear_inode = ext4_clear_inode,
865 .show_options = ext4_show_options,
ac27a0ec 866#ifdef CONFIG_QUOTA
617ba13b
MC		 867 .quota_read = ext4_quota_read,
868 .quota_write = ext4_quota_write,
ac27a0ec
DK		 869#endif
870};
871
39655164 872static const struct export_operations ext4_export_ops = {
1b961ac0
CH		 873 .fh_to_dentry = ext4_fh_to_dentry,
874 .fh_to_parent = ext4_fh_to_parent,
617ba13b 875 .get_parent = ext4_get_parent,
ac27a0ec
DK		 876};
877
878enum {
879 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
880 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
881 Opt_nouid32, Opt_nocheck, Opt_debug, Opt_oldalloc, Opt_orlov,
882 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
883 Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh,
884 Opt_commit, Opt_journal_update, Opt_journal_inum, Opt_journal_dev,
818d276c 885 Opt_journal_checksum, Opt_journal_async_commit,
ac27a0ec
DK		 886 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
887 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
888 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
889 Opt_ignore, Opt_barrier, Opt_err, Opt_resize, Opt_usrquota,
25ec56b5 890 Opt_grpquota, Opt_extents, Opt_noextents, Opt_i_version,
c9de560d 891 Opt_mballoc, Opt_nomballoc, Opt_stripe,
ac27a0ec
DK		 892};
893
894static match_table_t tokens = {
895 {Opt_bsd_df, "bsddf"},
896 {Opt_minix_df, "minixdf"},
897 {Opt_grpid, "grpid"},
898 {Opt_grpid, "bsdgroups"},
899 {Opt_nogrpid, "nogrpid"},
900 {Opt_nogrpid, "sysvgroups"},
901 {Opt_resgid, "resgid=%u"},
902 {Opt_resuid, "resuid=%u"},
903 {Opt_sb, "sb=%u"},
904 {Opt_err_cont, "errors=continue"},
905 {Opt_err_panic, "errors=panic"},
906 {Opt_err_ro, "errors=remount-ro"},
907 {Opt_nouid32, "nouid32"},
908 {Opt_nocheck, "nocheck"},
909 {Opt_nocheck, "check=none"},
910 {Opt_debug, "debug"},
911 {Opt_oldalloc, "oldalloc"},
912 {Opt_orlov, "orlov"},
913 {Opt_user_xattr, "user_xattr"},
914 {Opt_nouser_xattr, "nouser_xattr"},
915 {Opt_acl, "acl"},
916 {Opt_noacl, "noacl"},
917 {Opt_reservation, "reservation"},
918 {Opt_noreservation, "noreservation"},
919 {Opt_noload, "noload"},
920 {Opt_nobh, "nobh"},
921 {Opt_bh, "bh"},
922 {Opt_commit, "commit=%u"},
923 {Opt_journal_update, "journal=update"},
924 {Opt_journal_inum, "journal=%u"},
925 {Opt_journal_dev, "journal_dev=%u"},
818d276c
GS		 926 {Opt_journal_checksum, "journal_checksum"},
927 {Opt_journal_async_commit, "journal_async_commit"},
ac27a0ec
DK		 928 {Opt_abort, "abort"},
929 {Opt_data_journal, "data=journal"},
930 {Opt_data_ordered, "data=ordered"},
931 {Opt_data_writeback, "data=writeback"},
932 {Opt_offusrjquota, "usrjquota="},
933 {Opt_usrjquota, "usrjquota=%s"},
934 {Opt_offgrpjquota, "grpjquota="},
935 {Opt_grpjquota, "grpjquota=%s"},
936 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
937 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
938 {Opt_grpquota, "grpquota"},
939 {Opt_noquota, "noquota"},
940 {Opt_quota, "quota"},
941 {Opt_usrquota, "usrquota"},
942 {Opt_barrier, "barrier=%u"},
a86c6181 943 {Opt_extents, "extents"},
1e2462f9 944 {Opt_noextents, "noextents"},
25ec56b5 945 {Opt_i_version, "i_version"},
c9de560d
AT		 946 {Opt_mballoc, "mballoc"},
947 {Opt_nomballoc, "nomballoc"},
948 {Opt_stripe, "stripe=%u"},
ac27a0ec
DK		 949 {Opt_err, NULL},
950 {Opt_resize, "resize"},
951};
952
617ba13b 953static ext4_fsblk_t get_sb_block(void **data)
ac27a0ec 954{
617ba13b 955 ext4_fsblk_t sb_block;
ac27a0ec
DK		 956 char *options = (char *) *data;
957
958 if (!options || strncmp(options, "sb=", 3) != 0)
959 return 1; /* Default location */
960 options += 3;
617ba13b 961 /*todo: use simple_strtoll with >32bit ext4 */
ac27a0ec
DK		 962 sb_block = simple_strtoul(options, &options, 0);
963 if (*options && *options != ',') {
617ba13b 964 printk("EXT4-fs: Invalid sb specification: %s\n",
ac27a0ec
DK		 965 (char *) *data);
966 return 1;
967 }
968 if (*options == ',')
969 options++;
970 *data = (void *) options;
971 return sb_block;
972}
973
974static int parse_options (char *options, struct super_block *sb,
975 unsigned int *inum, unsigned long *journal_devnum,
617ba13b 976 ext4_fsblk_t *n_blocks_count, int is_remount)
ac27a0ec 977{
617ba13b 978 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 979 char * p;
980 substring_t args[MAX_OPT_ARGS];
981 int data_opt = 0;
982 int option;
983#ifdef CONFIG_QUOTA
984 int qtype;
985 char *qname;
986#endif
987
988 if (!options)
989 return 1;
990
991 while ((p = strsep (&options, ",")) != NULL) {
992 int token;
993 if (!*p)
994 continue;
995
996 token = match_token(p, tokens, args);
997 switch (token) {
998 case Opt_bsd_df:
999 clear_opt (sbi->s_mount_opt, MINIX_DF);
1000 break;
1001 case Opt_minix_df:
1002 set_opt (sbi->s_mount_opt, MINIX_DF);
1003 break;
1004 case Opt_grpid:
1005 set_opt (sbi->s_mount_opt, GRPID);
1006 break;
1007 case Opt_nogrpid:
1008 clear_opt (sbi->s_mount_opt, GRPID);
1009 break;
1010 case Opt_resuid:
1011 if (match_int(&args[0], &option))
1012 return 0;
1013 sbi->s_resuid = option;
1014 break;
1015 case Opt_resgid:
1016 if (match_int(&args[0], &option))
1017 return 0;
1018 sbi->s_resgid = option;
1019 break;
1020 case Opt_sb:
1021 /* handled by get_sb_block() instead of here */
1022 /* *sb_block = match_int(&args[0]); */
1023 break;
1024 case Opt_err_panic:
1025 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
1026 clear_opt (sbi->s_mount_opt, ERRORS_RO);
1027 set_opt (sbi->s_mount_opt, ERRORS_PANIC);
1028 break;
1029 case Opt_err_ro:
1030 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
1031 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
1032 set_opt (sbi->s_mount_opt, ERRORS_RO);
1033 break;
1034 case Opt_err_cont:
1035 clear_opt (sbi->s_mount_opt, ERRORS_RO);
1036 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
1037 set_opt (sbi->s_mount_opt, ERRORS_CONT);
1038 break;
1039 case Opt_nouid32:
1040 set_opt (sbi->s_mount_opt, NO_UID32);
1041 break;
1042 case Opt_nocheck:
1043 clear_opt (sbi->s_mount_opt, CHECK);
1044 break;
1045 case Opt_debug:
1046 set_opt (sbi->s_mount_opt, DEBUG);
1047 break;
1048 case Opt_oldalloc:
1049 set_opt (sbi->s_mount_opt, OLDALLOC);
1050 break;
1051 case Opt_orlov:
1052 clear_opt (sbi->s_mount_opt, OLDALLOC);
1053 break;
617ba13b 1054#ifdef CONFIG_EXT4DEV_FS_XATTR
ac27a0ec
DK		 1055 case Opt_user_xattr:
1056 set_opt (sbi->s_mount_opt, XATTR_USER);
1057 break;
1058 case Opt_nouser_xattr:
1059 clear_opt (sbi->s_mount_opt, XATTR_USER);
1060 break;
1061#else
1062 case Opt_user_xattr:
1063 case Opt_nouser_xattr:
617ba13b 1064 printk("EXT4 (no)user_xattr options not supported\n");
ac27a0ec
DK		 1065 break;
1066#endif
617ba13b 1067#ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
ac27a0ec
DK		 1068 case Opt_acl:
1069 set_opt(sbi->s_mount_opt, POSIX_ACL);
1070 break;
1071 case Opt_noacl:
1072 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1073 break;
1074#else
1075 case Opt_acl:
1076 case Opt_noacl:
617ba13b 1077 printk("EXT4 (no)acl options not supported\n");
ac27a0ec
DK		 1078 break;
1079#endif
1080 case Opt_reservation:
1081 set_opt(sbi->s_mount_opt, RESERVATION);
1082 break;
1083 case Opt_noreservation:
1084 clear_opt(sbi->s_mount_opt, RESERVATION);
1085 break;
1086 case Opt_journal_update:
1087 /* @@@ FIXME */
1088 /* Eventually we will want to be able to create
1089 a journal file here. For now, only allow the
1090 user to specify an existing inode to be the
1091 journal file. */
1092 if (is_remount) {
617ba13b 1093 printk(KERN_ERR "EXT4-fs: cannot specify "
ac27a0ec
DK		 1094 "journal on remount\n");
1095 return 0;
1096 }
1097 set_opt (sbi->s_mount_opt, UPDATE_JOURNAL);
1098 break;
1099 case Opt_journal_inum:
1100 if (is_remount) {
617ba13b 1101 printk(KERN_ERR "EXT4-fs: cannot specify "
ac27a0ec
DK		 1102 "journal on remount\n");
1103 return 0;
1104 }
1105 if (match_int(&args[0], &option))
1106 return 0;
1107 *inum = option;
1108 break;
1109 case Opt_journal_dev:
1110 if (is_remount) {
617ba13b 1111 printk(KERN_ERR "EXT4-fs: cannot specify "
ac27a0ec
DK		 1112 "journal on remount\n");
1113 return 0;
1114 }
1115 if (match_int(&args[0], &option))
1116 return 0;
1117 *journal_devnum = option;
1118 break;
818d276c
GS		 1119 case Opt_journal_checksum:
1120 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1121 break;
1122 case Opt_journal_async_commit:
1123 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
1124 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1125 break;
ac27a0ec
DK		 1126 case Opt_noload:
1127 set_opt (sbi->s_mount_opt, NOLOAD);
1128 break;
1129 case Opt_commit:
1130 if (match_int(&args[0], &option))
1131 return 0;
1132 if (option < 0)
1133 return 0;
1134 if (option == 0)
cd02ff0b 1135 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
ac27a0ec
DK		 1136 sbi->s_commit_interval = HZ * option;
1137 break;
1138 case Opt_data_journal:
617ba13b 1139 data_opt = EXT4_MOUNT_JOURNAL_DATA;
ac27a0ec
DK		 1140 goto datacheck;
1141 case Opt_data_ordered:
617ba13b 1142 data_opt = EXT4_MOUNT_ORDERED_DATA;
ac27a0ec
DK		 1143 goto datacheck;
1144 case Opt_data_writeback:
617ba13b 1145 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
ac27a0ec
DK		 1146 datacheck:
1147 if (is_remount) {
617ba13b 1148 if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
ac27a0ec
DK		 1149 != data_opt) {
1150 printk(KERN_ERR
617ba13b 1151 "EXT4-fs: cannot change data "
ac27a0ec
DK		 1152 "mode on remount\n");
1153 return 0;
1154 }
1155 } else {
617ba13b 1156 sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
ac27a0ec
DK		 1157 sbi->s_mount_opt |= data_opt;
1158 }
1159 break;
1160#ifdef CONFIG_QUOTA
1161 case Opt_usrjquota:
1162 qtype = USRQUOTA;
1163 goto set_qf_name;
1164 case Opt_grpjquota:
1165 qtype = GRPQUOTA;
1166set_qf_name:
1167 if (sb_any_quota_enabled(sb)) {
1168 printk(KERN_ERR
617ba13b 1169 "EXT4-fs: Cannot change journalled "
ac27a0ec
DK		 1170 "quota options when quota turned on.\n");
1171 return 0;
1172 }
1173 qname = match_strdup(&args[0]);
1174 if (!qname) {
1175 printk(KERN_ERR
617ba13b 1176 "EXT4-fs: not enough memory for "
ac27a0ec
DK		 1177 "storing quotafile name.\n");
1178 return 0;
1179 }
1180 if (sbi->s_qf_names[qtype] &&
1181 strcmp(sbi->s_qf_names[qtype], qname)) {
1182 printk(KERN_ERR
617ba13b 1183 "EXT4-fs: %s quota file already "
ac27a0ec
DK		 1184 "specified.\n", QTYPE2NAME(qtype));
1185 kfree(qname);
1186 return 0;
1187 }
1188 sbi->s_qf_names[qtype] = qname;
1189 if (strchr(sbi->s_qf_names[qtype], '/')) {
1190 printk(KERN_ERR
617ba13b 1191 "EXT4-fs: quotafile must be on "
ac27a0ec
DK		 1192 "filesystem root.\n");
1193 kfree(sbi->s_qf_names[qtype]);
1194 sbi->s_qf_names[qtype] = NULL;
1195 return 0;
1196 }
1197 set_opt(sbi->s_mount_opt, QUOTA);
1198 break;
1199 case Opt_offusrjquota:
1200 qtype = USRQUOTA;
1201 goto clear_qf_name;
1202 case Opt_offgrpjquota:
1203 qtype = GRPQUOTA;
1204clear_qf_name:
1205 if (sb_any_quota_enabled(sb)) {
617ba13b 1206 printk(KERN_ERR "EXT4-fs: Cannot change "
ac27a0ec
DK		 1207 "journalled quota options when "
1208 "quota turned on.\n");
1209 return 0;
1210 }
1211 /*
1212 * The space will be released later when all options
1213 * are confirmed to be correct
1214 */
1215 sbi->s_qf_names[qtype] = NULL;
1216 break;
1217 case Opt_jqfmt_vfsold:
1218 sbi->s_jquota_fmt = QFMT_VFS_OLD;
1219 break;
1220 case Opt_jqfmt_vfsv0:
1221 sbi->s_jquota_fmt = QFMT_VFS_V0;
1222 break;
1223 case Opt_quota:
1224 case Opt_usrquota:
1225 set_opt(sbi->s_mount_opt, QUOTA);
1226 set_opt(sbi->s_mount_opt, USRQUOTA);
1227 break;
1228 case Opt_grpquota:
1229 set_opt(sbi->s_mount_opt, QUOTA);
1230 set_opt(sbi->s_mount_opt, GRPQUOTA);
1231 break;
1232 case Opt_noquota:
1233 if (sb_any_quota_enabled(sb)) {
617ba13b 1234 printk(KERN_ERR "EXT4-fs: Cannot change quota "
ac27a0ec
DK		 1235 "options when quota turned on.\n");
1236 return 0;
1237 }
1238 clear_opt(sbi->s_mount_opt, QUOTA);
1239 clear_opt(sbi->s_mount_opt, USRQUOTA);
1240 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1241 break;
1242#else
1243 case Opt_quota:
1244 case Opt_usrquota:
1245 case Opt_grpquota:
1246 case Opt_usrjquota:
1247 case Opt_grpjquota:
1248 case Opt_offusrjquota:
1249 case Opt_offgrpjquota:
1250 case Opt_jqfmt_vfsold:
1251 case Opt_jqfmt_vfsv0:
1252 printk(KERN_ERR
617ba13b 1253 "EXT4-fs: journalled quota options not "
ac27a0ec
DK		 1254 "supported.\n");
1255 break;
1256 case Opt_noquota:
1257 break;
1258#endif
1259 case Opt_abort:
1260 set_opt(sbi->s_mount_opt, ABORT);
1261 break;
1262 case Opt_barrier:
1263 if (match_int(&args[0], &option))
1264 return 0;
1265 if (option)
1266 set_opt(sbi->s_mount_opt, BARRIER);
1267 else
1268 clear_opt(sbi->s_mount_opt, BARRIER);
1269 break;
1270 case Opt_ignore:
1271 break;
1272 case Opt_resize:
1273 if (!is_remount) {
617ba13b 1274 printk("EXT4-fs: resize option only available "
ac27a0ec
DK		 1275 "for remount\n");
1276 return 0;
1277 }
1278 if (match_int(&args[0], &option) != 0)
1279 return 0;
1280 *n_blocks_count = option;
1281 break;
1282 case Opt_nobh:
1283 set_opt(sbi->s_mount_opt, NOBH);
1284 break;
1285 case Opt_bh:
1286 clear_opt(sbi->s_mount_opt, NOBH);
1287 break;
a86c6181
AT		 1288 case Opt_extents:
1289 set_opt (sbi->s_mount_opt, EXTENTS);
1290 break;
1e2462f9
MC		 1291 case Opt_noextents:
1292 clear_opt (sbi->s_mount_opt, EXTENTS);
1293 break;
25ec56b5
JNC		 1294 case Opt_i_version:
1295 set_opt(sbi->s_mount_opt, I_VERSION);
1296 sb->s_flags |= MS_I_VERSION;
1297 break;
c9de560d
AT		 1298 case Opt_mballoc:
1299 set_opt(sbi->s_mount_opt, MBALLOC);
1300 break;
1301 case Opt_nomballoc:
1302 clear_opt(sbi->s_mount_opt, MBALLOC);
1303 break;
1304 case Opt_stripe:
1305 if (match_int(&args[0], &option))
1306 return 0;
1307 if (option < 0)
1308 return 0;
1309 sbi->s_stripe = option;
1310 break;
ac27a0ec
DK		 1311 default:
1312 printk (KERN_ERR
617ba13b 1313 "EXT4-fs: Unrecognized mount option \"%s\" "
ac27a0ec
DK		 1314 "or missing value\n", p);
1315 return 0;
1316 }
1317 }
1318#ifdef CONFIG_QUOTA
1319 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
617ba13b 1320 if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
ac27a0ec
DK		 1321 sbi->s_qf_names[USRQUOTA])
1322 clear_opt(sbi->s_mount_opt, USRQUOTA);
1323
617ba13b 1324 if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
ac27a0ec
DK		 1325 sbi->s_qf_names[GRPQUOTA])
1326 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1327
1328 if ((sbi->s_qf_names[USRQUOTA] &&
617ba13b 1329 (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
ac27a0ec 1330 (sbi->s_qf_names[GRPQUOTA] &&
617ba13b
MC		 1331 (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
1332 printk(KERN_ERR "EXT4-fs: old and new quota "
ac27a0ec
DK		 1333 "format mixing.\n");
1334 return 0;
1335 }
1336
1337 if (!sbi->s_jquota_fmt) {
617ba13b 1338 printk(KERN_ERR "EXT4-fs: journalled quota format "
ac27a0ec
DK		 1339 "not specified.\n");
1340 return 0;
1341 }
1342 } else {
1343 if (sbi->s_jquota_fmt) {
617ba13b 1344 printk(KERN_ERR "EXT4-fs: journalled quota format "
ac27a0ec
DK		 1345 "specified with no journalling "
1346 "enabled.\n");
1347 return 0;
1348 }
1349 }
1350#endif
1351 return 1;
1352}
1353
617ba13b 1354static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 1355 int read_only)
1356{
617ba13b 1357 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 1358 int res = 0;
1359
617ba13b
MC		 1360 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1361 printk (KERN_ERR "EXT4-fs warning: revision level too high, "
ac27a0ec
DK		 1362 "forcing read-only mode\n");
1363 res = MS_RDONLY;
1364 }
1365 if (read_only)
1366 return res;
617ba13b
MC		 1367 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1368 printk (KERN_WARNING "EXT4-fs warning: mounting unchecked fs, "
ac27a0ec 1369 "running e2fsck is recommended\n");
617ba13b 1370 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
ac27a0ec 1371 printk (KERN_WARNING
617ba13b 1372 "EXT4-fs warning: mounting fs with errors, "
ac27a0ec
DK		 1373 "running e2fsck is recommended\n");
1374 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1375 le16_to_cpu(es->s_mnt_count) >=
1376 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1377 printk (KERN_WARNING
617ba13b 1378 "EXT4-fs warning: maximal mount count reached, "
ac27a0ec
DK		 1379 "running e2fsck is recommended\n");
1380 else if (le32_to_cpu(es->s_checkinterval) &&
1381 (le32_to_cpu(es->s_lastcheck) +
1382 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1383 printk (KERN_WARNING
617ba13b 1384 "EXT4-fs warning: checktime reached, "
ac27a0ec
DK		 1385 "running e2fsck is recommended\n");
1386#if 0
1387 /* @@@ We _will_ want to clear the valid bit if we find
63f57933
AM		 1388 * inconsistencies, to force a fsck at reboot. But for
1389 * a plain journaled filesystem we can keep it set as
1390 * valid forever! :)
1391 */
617ba13b 1392 es->s_state = cpu_to_le16(le16_to_cpu(es->s_state) & ~EXT4_VALID_FS);
ac27a0ec
DK		 1393#endif
1394 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 1395 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
ac27a0ec
DK		 1396 es->s_mnt_count=cpu_to_le16(le16_to_cpu(es->s_mnt_count) + 1);
1397 es->s_mtime = cpu_to_le32(get_seconds());
617ba13b
MC		 1398 ext4_update_dynamic_rev(sb);
1399 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 1400
617ba13b 1401 ext4_commit_super(sb, es, 1);
ac27a0ec 1402 if (test_opt(sb, DEBUG))
617ba13b 1403 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%lu, "
ac27a0ec
DK		 1404 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
1405 sb->s_blocksize,
1406 sbi->s_groups_count,
617ba13b
MC		 1407 EXT4_BLOCKS_PER_GROUP(sb),
1408 EXT4_INODES_PER_GROUP(sb),
ac27a0ec
DK		 1409 sbi->s_mount_opt);
1410
617ba13b
MC		 1411 printk(KERN_INFO "EXT4 FS on %s, ", sb->s_id);
1412 if (EXT4_SB(sb)->s_journal->j_inode == NULL) {
ac27a0ec
DK		 1413 char b[BDEVNAME_SIZE];
1414
1415 printk("external journal on %s\n",
617ba13b 1416 bdevname(EXT4_SB(sb)->s_journal->j_dev, b));
ac27a0ec
DK		 1417 } else {
1418 printk("internal journal\n");
1419 }
1420 return res;
1421}
1422
717d50e4
AD		 1423__le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1424 struct ext4_group_desc *gdp)
1425{
1426 __u16 crc = 0;
1427
1428 if (sbi->s_es->s_feature_ro_compat &
1429 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1430 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1431 __le32 le_group = cpu_to_le32(block_group);
1432
1433 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1434 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1435 crc = crc16(crc, (__u8 *)gdp, offset);
1436 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1437 /* for checksum of struct ext4_group_desc do the rest...*/
1438 if ((sbi->s_es->s_feature_incompat &
1439 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1440 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1441 crc = crc16(crc, (__u8 *)gdp + offset,
1442 le16_to_cpu(sbi->s_es->s_desc_size) -
1443 offset);
1444 }
1445
1446 return cpu_to_le16(crc);
1447}
1448
1449int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1450 struct ext4_group_desc *gdp)
1451{
1452 if ((sbi->s_es->s_feature_ro_compat &
1453 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1454 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1455 return 0;
1456
1457 return 1;
1458}
1459
ac27a0ec 1460/* Called at mount-time, super-block is locked */
617ba13b 1461static int ext4_check_descriptors (struct super_block * sb)
ac27a0ec 1462{
617ba13b
MC		 1463 struct ext4_sb_info *sbi = EXT4_SB(sb);
1464 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1465 ext4_fsblk_t last_block;
bd81d8ee
LV		 1466 ext4_fsblk_t block_bitmap;
1467 ext4_fsblk_t inode_bitmap;
1468 ext4_fsblk_t inode_table;
617ba13b 1469 struct ext4_group_desc * gdp = NULL;
ac27a0ec 1470 int desc_block = 0;
ce421581 1471 int flexbg_flag = 0;
fd2d4291 1472 ext4_group_t i;
ac27a0ec 1473
ce421581
JS		 1474 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1475 flexbg_flag = 1;
1476
617ba13b 1477 ext4_debug ("Checking group descriptors");
ac27a0ec
DK		 1478
1479 for (i = 0; i < sbi->s_groups_count; i++)
1480 {
ce421581 1481 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 1482 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 1483 else
1484 last_block = first_block +
617ba13b 1485 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 1486
617ba13b
MC		 1487 if ((i % EXT4_DESC_PER_BLOCK(sb)) == 0)
1488 gdp = (struct ext4_group_desc *)
ac27a0ec 1489 sbi->s_group_desc[desc_block++]->b_data;
8fadc143 1490 block_bitmap = ext4_block_bitmap(sb, gdp);
bd81d8ee 1491 if (block_bitmap < first_block || block_bitmap > last_block)
ac27a0ec 1492 {
617ba13b 1493 ext4_error (sb, "ext4_check_descriptors",
fd2d4291 1494 "Block bitmap for group %lu"
2ae02107 1495 " not in group (block %llu)!",
bd81d8ee 1496 i, block_bitmap);
ac27a0ec
DK		 1497 return 0;
1498 }
8fadc143 1499 inode_bitmap = ext4_inode_bitmap(sb, gdp);
bd81d8ee 1500 if (inode_bitmap < first_block || inode_bitmap > last_block)
ac27a0ec 1501 {
617ba13b 1502 ext4_error (sb, "ext4_check_descriptors",
fd2d4291 1503 "Inode bitmap for group %lu"
2ae02107 1504 " not in group (block %llu)!",
bd81d8ee 1505 i, inode_bitmap);
ac27a0ec
DK		 1506 return 0;
1507 }
8fadc143 1508 inode_table = ext4_inode_table(sb, gdp);
bd81d8ee 1509 if (inode_table < first_block ||
780dcdb2 1510 inode_table + sbi->s_itb_per_group - 1 > last_block)
ac27a0ec 1511 {
617ba13b 1512 ext4_error (sb, "ext4_check_descriptors",
fd2d4291 1513 "Inode table for group %lu"
2ae02107 1514 " not in group (block %llu)!",
bd81d8ee 1515 i, inode_table);
ac27a0ec
DK		 1516 return 0;
1517 }
717d50e4
AD		 1518 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
1519 ext4_error(sb, __FUNCTION__,
fd2d4291
AM		 1520 "Checksum for group %lu failed (%u!=%u)\n",
1521 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1522 gdp)), le16_to_cpu(gdp->bg_checksum));
717d50e4
AD		 1523 return 0;
1524 }
ce421581
JS		 1525 if (!flexbg_flag)
1526 first_block += EXT4_BLOCKS_PER_GROUP(sb);
0d1ee42f
AR		 1527 gdp = (struct ext4_group_desc *)
1528 ((__u8 *)gdp + EXT4_DESC_SIZE(sb));
ac27a0ec
DK		 1529 }
1530
bd81d8ee 1531 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
617ba13b 1532 sbi->s_es->s_free_inodes_count=cpu_to_le32(ext4_count_free_inodes(sb));
ac27a0ec
DK		 1533 return 1;
1534}
1535
617ba13b 1536/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 1537 * the superblock) which were deleted from all directories, but held open by
1538 * a process at the time of a crash. We walk the list and try to delete these
1539 * inodes at recovery time (only with a read-write filesystem).
1540 *
1541 * In order to keep the orphan inode chain consistent during traversal (in
1542 * case of crash during recovery), we link each inode into the superblock
1543 * orphan list_head and handle it the same way as an inode deletion during
1544 * normal operation (which journals the operations for us).
1545 *
1546 * We only do an iget() and an iput() on each inode, which is very safe if we
1547 * accidentally point at an in-use or already deleted inode. The worst that
1548 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 1549 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 1550 * e2fsck was run on this filesystem, and it must have already done the orphan
1551 * inode cleanup for us, so we can safely abort without any further action.
1552 */
617ba13b
MC		 1553static void ext4_orphan_cleanup (struct super_block * sb,
1554 struct ext4_super_block * es)
ac27a0ec
DK		 1555{
1556 unsigned int s_flags = sb->s_flags;
1557 int nr_orphans = 0, nr_truncates = 0;
1558#ifdef CONFIG_QUOTA
1559 int i;
1560#endif
1561 if (!es->s_last_orphan) {
1562 jbd_debug(4, "no orphan inodes to clean up\n");
1563 return;
1564 }
1565
a8f48a95
ES		 1566 if (bdev_read_only(sb->s_bdev)) {
1567 printk(KERN_ERR "EXT4-fs: write access "
1568 "unavailable, skipping orphan cleanup.\n");
1569 return;
1570 }
1571
617ba13b 1572 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
ac27a0ec
DK		 1573 if (es->s_last_orphan)
1574 jbd_debug(1, "Errors on filesystem, "
1575 "clearing orphan list.\n");
1576 es->s_last_orphan = 0;
1577 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1578 return;
1579 }
1580
1581 if (s_flags & MS_RDONLY) {
617ba13b 1582 printk(KERN_INFO "EXT4-fs: %s: orphan cleanup on readonly fs\n",
ac27a0ec
DK		 1583 sb->s_id);
1584 sb->s_flags &= ~MS_RDONLY;
1585 }
1586#ifdef CONFIG_QUOTA
1587 /* Needed for iput() to work correctly and not trash data */
1588 sb->s_flags |= MS_ACTIVE;
1589 /* Turn on quotas so that they are updated correctly */
1590 for (i = 0; i < MAXQUOTAS; i++) {
617ba13b
MC		 1591 if (EXT4_SB(sb)->s_qf_names[i]) {
1592 int ret = ext4_quota_on_mount(sb, i);
ac27a0ec
DK		 1593 if (ret < 0)
1594 printk(KERN_ERR
617ba13b 1595 "EXT4-fs: Cannot turn on journalled "
ac27a0ec
DK		 1596 "quota: error %d\n", ret);
1597 }
1598 }
1599#endif
1600
1601 while (es->s_last_orphan) {
1602 struct inode *inode;
1603
1604 if (!(inode =
617ba13b 1605 ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan)))) {
ac27a0ec
DK		 1606 es->s_last_orphan = 0;
1607 break;
1608 }
1609
617ba13b 1610 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
ac27a0ec
DK		 1611 DQUOT_INIT(inode);
1612 if (inode->i_nlink) {
1613 printk(KERN_DEBUG
1614 "%s: truncating inode %lu to %Ld bytes\n",
1615 __FUNCTION__, inode->i_ino, inode->i_size);
1616 jbd_debug(2, "truncating inode %lu to %Ld bytes\n",
1617 inode->i_ino, inode->i_size);
617ba13b 1618 ext4_truncate(inode);
ac27a0ec
DK		 1619 nr_truncates++;
1620 } else {
1621 printk(KERN_DEBUG
1622 "%s: deleting unreferenced inode %lu\n",
1623 __FUNCTION__, inode->i_ino);
1624 jbd_debug(2, "deleting unreferenced inode %lu\n",
1625 inode->i_ino);
1626 nr_orphans++;
1627 }
1628 iput(inode); /* The delete magic happens here! */
1629 }
1630
1631#define PLURAL(x) (x), ((x)==1) ? "" : "s"
1632
1633 if (nr_orphans)
617ba13b 1634 printk(KERN_INFO "EXT4-fs: %s: %d orphan inode%s deleted\n",
ac27a0ec
DK		 1635 sb->s_id, PLURAL(nr_orphans));
1636 if (nr_truncates)
617ba13b 1637 printk(KERN_INFO "EXT4-fs: %s: %d truncate%s cleaned up\n",
ac27a0ec
DK		 1638 sb->s_id, PLURAL(nr_truncates));
1639#ifdef CONFIG_QUOTA
1640 /* Turn quotas off */
1641 for (i = 0; i < MAXQUOTAS; i++) {
1642 if (sb_dqopt(sb)->files[i])
1643 vfs_quota_off(sb, i);
1644 }
1645#endif
1646 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1647}
cd2291a4
ES		 1648/*
1649 * Maximal extent format file size.
1650 * Resulting logical blkno at s_maxbytes must fit in our on-disk
1651 * extent format containers, within a sector_t, and within i_blocks
1652 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
1653 * so that won't be a limiting factor.
1654 *
1655 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1656 */
1657static loff_t ext4_max_size(int blkbits)
1658{
1659 loff_t res;
1660 loff_t upper_limit = MAX_LFS_FILESIZE;
1661
1662 /* small i_blocks in vfs inode? */
1663 if (sizeof(blkcnt_t) < sizeof(u64)) {
1664 /*
1665 * CONFIG_LSF is not enabled implies the inode
1666 * i_block represent total blocks in 512 bytes
1667 * 32 == size of vfs inode i_blocks * 8
1668 */
1669 upper_limit = (1LL << 32) - 1;
1670
1671 /* total blocks in file system block size */
1672 upper_limit >>= (blkbits - 9);
1673 upper_limit <<= blkbits;
1674 }
1675
1676 /* 32-bit extent-start container, ee_block */
1677 res = 1LL << 32;
1678 res <<= blkbits;
1679 res -= 1;
1680
1681 /* Sanity check against vm- & vfs- imposed limits */
1682 if (res > upper_limit)
1683 res = upper_limit;
1684
1685 return res;
1686}
ac27a0ec 1687
ac27a0ec 1688/*
cd2291a4 1689 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 1690 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
1691 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 1692 */
cd2291a4 1693static loff_t ext4_max_bitmap_size(int bits)
ac27a0ec 1694{
617ba13b 1695 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 1696 int meta_blocks;
1697 loff_t upper_limit;
1698 /* This is calculated to be the largest file size for a
cd2291a4 1699 * dense, bitmapped file such that the total number of
ac27a0ec 1700 * sectors in the file, including data and all indirect blocks,
0fc1b451
AK		 1701 * does not exceed 2^48 -1
1702 * __u32 i_blocks_lo and _u16 i_blocks_high representing the
1703 * total number of 512 bytes blocks of the file
1704 */
1705
1706 if (sizeof(blkcnt_t) < sizeof(u64)) {
1707 /*
1708 * CONFIG_LSF is not enabled implies the inode
1709 * i_block represent total blocks in 512 bytes
1710 * 32 == size of vfs inode i_blocks * 8
1711 */
1712 upper_limit = (1LL << 32) - 1;
1713
1714 /* total blocks in file system block size */
1715 upper_limit >>= (bits - 9);
1716
1717 } else {
8180a562
AK		 1718 /*
1719 * We use 48 bit ext4_inode i_blocks
1720 * With EXT4_HUGE_FILE_FL set the i_blocks
1721 * represent total number of blocks in
1722 * file system block size
1723 */
0fc1b451
AK		 1724 upper_limit = (1LL << 48) - 1;
1725
0fc1b451
AK		 1726 }
1727
1728 /* indirect blocks */
1729 meta_blocks = 1;
1730 /* double indirect blocks */
1731 meta_blocks += 1 + (1LL << (bits-2));
1732 /* tripple indirect blocks */
1733 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
1734
1735 upper_limit -= meta_blocks;
1736 upper_limit <<= bits;
ac27a0ec
DK		 1737
1738 res += 1LL << (bits-2);
1739 res += 1LL << (2*(bits-2));
1740 res += 1LL << (3*(bits-2));
1741 res <<= bits;
1742 if (res > upper_limit)
1743 res = upper_limit;
0fc1b451
AK		 1744
1745 if (res > MAX_LFS_FILESIZE)
1746 res = MAX_LFS_FILESIZE;
1747
ac27a0ec
DK		 1748 return res;
1749}
1750
617ba13b 1751static ext4_fsblk_t descriptor_loc(struct super_block *sb,
70bbb3e0 1752 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 1753{
617ba13b 1754 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 1755 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 1756 int has_super = 0;
1757
1758 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1759
617ba13b 1760 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
ac27a0ec 1761 nr < first_meta_bg)
70bbb3e0 1762 return logical_sb_block + nr + 1;
ac27a0ec 1763 bg = sbi->s_desc_per_block * nr;
617ba13b 1764 if (ext4_bg_has_super(sb, bg))
ac27a0ec 1765 has_super = 1;
617ba13b 1766 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 1767}
1768
c9de560d
AT		 1769/**
1770 * ext4_get_stripe_size: Get the stripe size.
1771 * @sbi: In memory super block info
1772 *
1773 * If we have specified it via mount option, then
1774 * use the mount option value. If the value specified at mount time is
1775 * greater than the blocks per group use the super block value.
1776 * If the super block value is greater than blocks per group return 0.
1777 * Allocator needs it be less than blocks per group.
1778 *
1779 */
1780static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
1781{
1782 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
1783 unsigned long stripe_width =
1784 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
1785
1786 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
1787 return sbi->s_stripe;
1788
1789 if (stripe_width <= sbi->s_blocks_per_group)
1790 return stripe_width;
1791
1792 if (stride <= sbi->s_blocks_per_group)
1793 return stride;
1794
1795 return 0;
1796}
ac27a0ec 1797
617ba13b 1798static int ext4_fill_super (struct super_block *sb, void *data, int silent)
1d03ec98
AK		 1799 __releases(kernel_sem)
1800 __acquires(kernel_sem)
1801
ac27a0ec
DK		 1802{
1803 struct buffer_head * bh;
617ba13b
MC		 1804 struct ext4_super_block *es = NULL;
1805 struct ext4_sb_info *sbi;
1806 ext4_fsblk_t block;
1807 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 1808 ext4_fsblk_t logical_sb_block;
ac27a0ec
DK		 1809 unsigned long offset = 0;
1810 unsigned int journal_inum = 0;
1811 unsigned long journal_devnum = 0;
1812 unsigned long def_mount_opts;
1813 struct inode *root;
1814 int blocksize;
ac27a0ec
DK		 1815 int db_count;
1816 int i;
1817 int needs_recovery;
1818 __le32 features;
bd81d8ee 1819 __u64 blocks_count;
833f4077 1820 int err;
ac27a0ec
DK		 1821
1822 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
1823 if (!sbi)
1824 return -ENOMEM;
1825 sb->s_fs_info = sbi;
1826 sbi->s_mount_opt = 0;
617ba13b
MC		 1827 sbi->s_resuid = EXT4_DEF_RESUID;
1828 sbi->s_resgid = EXT4_DEF_RESGID;
d9c9bef1 1829 sbi->s_sb_block = sb_block;
ac27a0ec
DK		 1830
1831 unlock_kernel();
1832
617ba13b 1833 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 1834 if (!blocksize) {
617ba13b 1835 printk(KERN_ERR "EXT4-fs: unable to set blocksize\n");
ac27a0ec
DK		 1836 goto out_fail;
1837 }
1838
afc7cbca
TS		 1839 if (!sb_set_blocksize(sb, blocksize)) {
1840 printk(KERN_ERR "EXT4-fs: bad blocksize %d.\n", blocksize);
1841 goto out_fail;
1842 }
1843
ac27a0ec 1844 /*
617ba13b 1845 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 1846 * block sizes. We need to calculate the offset from buffer start.
1847 */
617ba13b 1848 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 1849 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
1850 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 1851 } else {
70bbb3e0 1852 logical_sb_block = sb_block;
ac27a0ec
DK		 1853 }
1854
70bbb3e0 1855 if (!(bh = sb_bread(sb, logical_sb_block))) {
617ba13b 1856 printk (KERN_ERR "EXT4-fs: unable to read superblock\n");
ac27a0ec
DK		 1857 goto out_fail;
1858 }
1859 /*
1860 * Note: s_es must be initialized as soon as possible because
617ba13b 1861 * some ext4 macro-instructions depend on its value
ac27a0ec 1862 */
617ba13b 1863 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
ac27a0ec
DK		 1864 sbi->s_es = es;
1865 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 1866 if (sb->s_magic != EXT4_SUPER_MAGIC)
1867 goto cantfind_ext4;
ac27a0ec
DK		 1868
1869 /* Set defaults before we parse the mount options */
1870 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
617ba13b 1871 if (def_mount_opts & EXT4_DEFM_DEBUG)
ac27a0ec 1872 set_opt(sbi->s_mount_opt, DEBUG);
617ba13b 1873 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
ac27a0ec 1874 set_opt(sbi->s_mount_opt, GRPID);
617ba13b 1875 if (def_mount_opts & EXT4_DEFM_UID16)
ac27a0ec 1876 set_opt(sbi->s_mount_opt, NO_UID32);
2e7842b8 1877#ifdef CONFIG_EXT4DEV_FS_XATTR
617ba13b 1878 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
ac27a0ec 1879 set_opt(sbi->s_mount_opt, XATTR_USER);
2e7842b8
HD		 1880#endif
1881#ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
617ba13b 1882 if (def_mount_opts & EXT4_DEFM_ACL)
ac27a0ec 1883 set_opt(sbi->s_mount_opt, POSIX_ACL);
2e7842b8 1884#endif
617ba13b
MC		 1885 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
1886 sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
1887 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
1888 sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
1889 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
1890 sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
1891
1892 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
ac27a0ec 1893 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
bb4f397a 1894 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
ceea16bf 1895 set_opt(sbi->s_mount_opt, ERRORS_CONT);
bb4f397a
AK		 1896 else
1897 set_opt(sbi->s_mount_opt, ERRORS_RO);
ac27a0ec
DK		 1898
1899 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
1900 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
1901
1902 set_opt(sbi->s_mount_opt, RESERVATION);
1903
1e2462f9
MC		 1904 /*
1905 * turn on extents feature by default in ext4 filesystem
1906 * User -o noextents to turn it off
1907 */
1908 set_opt(sbi->s_mount_opt, EXTENTS);
3dbd0ede
AK		 1909 /*
1910 * turn on mballoc feature by default in ext4 filesystem
1911 * User -o nomballoc to turn it off
1912 */
1913 set_opt(sbi->s_mount_opt, MBALLOC);
1e2462f9 1914
ac27a0ec
DK		 1915 if (!parse_options ((char *) data, sb, &journal_inum, &journal_devnum,
1916 NULL, 0))
1917 goto failed_mount;
1918
1919 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
617ba13b 1920 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec 1921
617ba13b
MC		 1922 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
1923 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
1924 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
1925 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
ac27a0ec 1926 printk(KERN_WARNING
617ba13b 1927 "EXT4-fs warning: feature flags set on rev 0 fs, "
ac27a0ec
DK		 1928 "running e2fsck is recommended\n");
1929 /*
1930 * Check feature flags regardless of the revision level, since we
1931 * previously didn't change the revision level when setting the flags,
1932 * so there is a chance incompat flags are set on a rev 0 filesystem.
1933 */
617ba13b 1934 features = EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP);
ac27a0ec 1935 if (features) {
617ba13b 1936 printk(KERN_ERR "EXT4-fs: %s: couldn't mount because of "
ac27a0ec
DK		 1937 "unsupported optional features (%x).\n",
1938 sb->s_id, le32_to_cpu(features));
1939 goto failed_mount;
1940 }
617ba13b 1941 features = EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP);
ac27a0ec 1942 if (!(sb->s_flags & MS_RDONLY) && features) {
617ba13b 1943 printk(KERN_ERR "EXT4-fs: %s: couldn't mount RDWR because of "
ac27a0ec
DK		 1944 "unsupported optional features (%x).\n",
1945 sb->s_id, le32_to_cpu(features));
1946 goto failed_mount;
1947 }
0fc1b451
AK		 1948 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
1949 /*
1950 * Large file size enabled file system can only be
1951 * mount if kernel is build with CONFIG_LSF
1952 */
1953 if (sizeof(root->i_blocks) < sizeof(u64) &&
1954 !(sb->s_flags & MS_RDONLY)) {
1955 printk(KERN_ERR "EXT4-fs: %s: Filesystem with huge "
1956 "files cannot be mounted read-write "
1957 "without CONFIG_LSF.\n", sb->s_id);
1958 goto failed_mount;
1959 }
1960 }
ac27a0ec
DK		 1961 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
1962
617ba13b
MC		 1963 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
1964 blocksize > EXT4_MAX_BLOCK_SIZE) {
ac27a0ec 1965 printk(KERN_ERR
617ba13b 1966 "EXT4-fs: Unsupported filesystem blocksize %d on %s.\n",
ac27a0ec
DK		 1967 blocksize, sb->s_id);
1968 goto failed_mount;
1969 }
1970
ac27a0ec 1971 if (sb->s_blocksize != blocksize) {
ce40733c
AK		 1972
1973 /* Validate the filesystem blocksize */
1974 if (!sb_set_blocksize(sb, blocksize)) {
1975 printk(KERN_ERR "EXT4-fs: bad block size %d.\n",
1976 blocksize);
ac27a0ec
DK		 1977 goto failed_mount;
1978 }
1979
1980 brelse (bh);
70bbb3e0
AM		 1981 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
1982 offset = do_div(logical_sb_block, blocksize);
1983 bh = sb_bread(sb, logical_sb_block);
ac27a0ec
DK		 1984 if (!bh) {
1985 printk(KERN_ERR
617ba13b 1986 "EXT4-fs: Can't read superblock on 2nd try.\n");
ac27a0ec
DK		 1987 goto failed_mount;
1988 }
617ba13b 1989 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
ac27a0ec 1990 sbi->s_es = es;
617ba13b 1991 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
ac27a0ec 1992 printk (KERN_ERR
617ba13b 1993 "EXT4-fs: Magic mismatch, very weird !\n");
ac27a0ec
DK		 1994 goto failed_mount;
1995 }
1996 }
1997
e2b46574 1998 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits);
617ba13b 1999 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits);
ac27a0ec 2000
617ba13b
MC		 2001 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2002 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2003 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
ac27a0ec
DK		 2004 } else {
2005 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2006 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
617ba13b 2007 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1330593e 2008 (!is_power_of_2(sbi->s_inode_size)) ||
ac27a0ec
DK		 2009 (sbi->s_inode_size > blocksize)) {
2010 printk (KERN_ERR
617ba13b 2011 "EXT4-fs: unsupported inode size: %d\n",
ac27a0ec
DK		 2012 sbi->s_inode_size);
2013 goto failed_mount;
2014 }
ef7f3835
KS		 2015 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2016 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
ac27a0ec 2017 }
0d1ee42f
AR		 2018 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2019 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
8fadc143 2020 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 2021 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 2022 !is_power_of_2(sbi->s_desc_size)) {
0d1ee42f 2023 printk(KERN_ERR
8fadc143 2024 "EXT4-fs: unsupported descriptor size %lu\n",
0d1ee42f
AR		 2025 sbi->s_desc_size);
2026 goto failed_mount;
2027 }
2028 } else
2029 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
ac27a0ec 2030 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 2031 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
b47b6f38 2032 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
617ba13b
MC		 2033 goto cantfind_ext4;
2034 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 2035 if (sbi->s_inodes_per_block == 0)
617ba13b 2036 goto cantfind_ext4;
ac27a0ec
DK		 2037 sbi->s_itb_per_group = sbi->s_inodes_per_group /
2038 sbi->s_inodes_per_block;
0d1ee42f 2039 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 2040 sbi->s_sbh = bh;
2041 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 2042 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2043 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
ac27a0ec
DK		 2044 for (i=0; i < 4; i++)
2045 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2046 sbi->s_def_hash_version = es->s_def_hash_version;
2047
2048 if (sbi->s_blocks_per_group > blocksize * 8) {
2049 printk (KERN_ERR
617ba13b 2050 "EXT4-fs: #blocks per group too big: %lu\n",
ac27a0ec
DK		 2051 sbi->s_blocks_per_group);
2052 goto failed_mount;
2053 }
ac27a0ec
DK		 2054 if (sbi->s_inodes_per_group > blocksize * 8) {
2055 printk (KERN_ERR
617ba13b 2056 "EXT4-fs: #inodes per group too big: %lu\n",
ac27a0ec
DK		 2057 sbi->s_inodes_per_group);
2058 goto failed_mount;
2059 }
2060
bd81d8ee 2061 if (ext4_blocks_count(es) >
ac27a0ec 2062 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
617ba13b 2063 printk(KERN_ERR "EXT4-fs: filesystem on %s:"
ac27a0ec
DK		 2064 " too large to mount safely\n", sb->s_id);
2065 if (sizeof(sector_t) < 8)
617ba13b 2066 printk(KERN_WARNING "EXT4-fs: CONFIG_LBD not "
ac27a0ec
DK		 2067 "enabled\n");
2068 goto failed_mount;
2069 }
2070
617ba13b
MC		 2071 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2072 goto cantfind_ext4;
e7c95593
ES		 2073
2074 /* ensure blocks_count calculation below doesn't sign-extend */
2075 if (ext4_blocks_count(es) + EXT4_BLOCKS_PER_GROUP(sb) <
2076 le32_to_cpu(es->s_first_data_block) + 1) {
2077 printk(KERN_WARNING "EXT4-fs: bad geometry: block count %llu, "
2078 "first data block %u, blocks per group %lu\n",
2079 ext4_blocks_count(es),
2080 le32_to_cpu(es->s_first_data_block),
2081 EXT4_BLOCKS_PER_GROUP(sb));
2082 goto failed_mount;
2083 }
bd81d8ee
LV		 2084 blocks_count = (ext4_blocks_count(es) -
2085 le32_to_cpu(es->s_first_data_block) +
2086 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2087 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
2088 sbi->s_groups_count = blocks_count;
617ba13b
MC		 2089 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2090 EXT4_DESC_PER_BLOCK(sb);
ac27a0ec
DK		 2091 sbi->s_group_desc = kmalloc(db_count * sizeof (struct buffer_head *),
2092 GFP_KERNEL);
2093 if (sbi->s_group_desc == NULL) {
617ba13b 2094 printk (KERN_ERR "EXT4-fs: not enough memory\n");
ac27a0ec
DK		 2095 goto failed_mount;
2096 }
2097
2098 bgl_lock_init(&sbi->s_blockgroup_lock);
2099
2100 for (i = 0; i < db_count; i++) {
70bbb3e0 2101 block = descriptor_loc(sb, logical_sb_block, i);
ac27a0ec
DK		 2102 sbi->s_group_desc[i] = sb_bread(sb, block);
2103 if (!sbi->s_group_desc[i]) {
617ba13b 2104 printk (KERN_ERR "EXT4-fs: "
ac27a0ec
DK		 2105 "can't read group descriptor %d\n", i);
2106 db_count = i;
2107 goto failed_mount2;
2108 }
2109 }
617ba13b
MC		 2110 if (!ext4_check_descriptors (sb)) {
2111 printk(KERN_ERR "EXT4-fs: group descriptors corrupted!\n");
ac27a0ec
DK		 2112 goto failed_mount2;
2113 }
2114 sbi->s_gdb_count = db_count;
2115 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2116 spin_lock_init(&sbi->s_next_gen_lock);
2117
833f4077
PZ		 2118 err = percpu_counter_init(&sbi->s_freeblocks_counter,
2119 ext4_count_free_blocks(sb));
2120 if (!err) {
2121 err = percpu_counter_init(&sbi->s_freeinodes_counter,
2122 ext4_count_free_inodes(sb));
2123 }
2124 if (!err) {
2125 err = percpu_counter_init(&sbi->s_dirs_counter,
2126 ext4_count_dirs(sb));
2127 }
2128 if (err) {
2129 printk(KERN_ERR "EXT4-fs: insufficient memory\n");
2130 goto failed_mount3;
2131 }
ac27a0ec
DK		 2132
2133 /* per fileystem reservation list head & lock */
2134 spin_lock_init(&sbi->s_rsv_window_lock);
2135 sbi->s_rsv_window_root = RB_ROOT;
2136 /* Add a single, static dummy reservation to the start of the
2137 * reservation window list --- it gives us a placeholder for
2138 * append-at-start-of-list which makes the allocation logic
2139 * _much_ simpler. */
617ba13b
MC		 2140 sbi->s_rsv_window_head.rsv_start = EXT4_RESERVE_WINDOW_NOT_ALLOCATED;
2141 sbi->s_rsv_window_head.rsv_end = EXT4_RESERVE_WINDOW_NOT_ALLOCATED;
ac27a0ec
DK		 2142 sbi->s_rsv_window_head.rsv_alloc_hit = 0;
2143 sbi->s_rsv_window_head.rsv_goal_size = 0;
617ba13b 2144 ext4_rsv_window_add(sb, &sbi->s_rsv_window_head);
ac27a0ec 2145
c9de560d
AT		 2146 sbi->s_stripe = ext4_get_stripe_size(sbi);
2147
ac27a0ec
DK		 2148 /*
2149 * set up enough so that it can read an inode
2150 */
617ba13b
MC		 2151 sb->s_op = &ext4_sops;
2152 sb->s_export_op = &ext4_export_ops;
2153 sb->s_xattr = ext4_xattr_handlers;
ac27a0ec 2154#ifdef CONFIG_QUOTA
617ba13b
MC		 2155 sb->s_qcop = &ext4_qctl_operations;
2156 sb->dq_op = &ext4_quota_operations;
ac27a0ec
DK		 2157#endif
2158 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
2159
2160 sb->s_root = NULL;
2161
2162 needs_recovery = (es->s_last_orphan != 0 ||
617ba13b
MC		 2163 EXT4_HAS_INCOMPAT_FEATURE(sb,
2164 EXT4_FEATURE_INCOMPAT_RECOVER));
ac27a0ec
DK		 2165
2166 /*
2167 * The first inode we look at is the journal inode. Don't try
2168 * root first: it may be modified in the journal!
2169 */
2170 if (!test_opt(sb, NOLOAD) &&
617ba13b
MC		 2171 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2172 if (ext4_load_journal(sb, es, journal_devnum))
ac27a0ec
DK		 2173 goto failed_mount3;
2174 } else if (journal_inum) {
617ba13b 2175 if (ext4_create_journal(sb, es, journal_inum))
ac27a0ec
DK		 2176 goto failed_mount3;
2177 } else {
2178 if (!silent)
2179 printk (KERN_ERR
617ba13b 2180 "ext4: No journal on filesystem on %s\n",
ac27a0ec
DK		 2181 sb->s_id);
2182 goto failed_mount3;
2183 }
2184
eb40a09c
JS		 2185 if (ext4_blocks_count(es) > 0xffffffffULL &&
2186 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2187 JBD2_FEATURE_INCOMPAT_64BIT)) {
2188 printk(KERN_ERR "ext4: Failed to set 64-bit journal feature\n");
2189 goto failed_mount4;
2190 }
2191
818d276c
GS		 2192 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2193 jbd2_journal_set_features(sbi->s_journal,
2194 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2195 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2196 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2197 jbd2_journal_set_features(sbi->s_journal,
2198 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2199 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2200 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2201 } else {
2202 jbd2_journal_clear_features(sbi->s_journal,
2203 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2204 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2205 }
2206
ac27a0ec
DK		 2207 /* We have now updated the journal if required, so we can
2208 * validate the data journaling mode. */
2209 switch (test_opt(sb, DATA_FLAGS)) {
2210 case 0:
2211 /* No mode set, assume a default based on the journal
63f57933
AM		 2212 * capabilities: ORDERED_DATA if the journal can
2213 * cope, else JOURNAL_DATA
2214 */
dab291af
MC		 2215 if (jbd2_journal_check_available_features
2216 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
ac27a0ec
DK		 2217 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2218 else
2219 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2220 break;
2221
617ba13b
MC		 2222 case EXT4_MOUNT_ORDERED_DATA:
2223 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 2224 if (!jbd2_journal_check_available_features
2225 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
617ba13b 2226 printk(KERN_ERR "EXT4-fs: Journal does not support "
ac27a0ec
DK		 2227 "requested data journaling mode\n");
2228 goto failed_mount4;
2229 }
2230 default:
2231 break;
2232 }
2233
2234 if (test_opt(sb, NOBH)) {
617ba13b
MC		 2235 if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
2236 printk(KERN_WARNING "EXT4-fs: Ignoring nobh option - "
ac27a0ec
DK		 2237 "its supported only with writeback mode\n");
2238 clear_opt(sbi->s_mount_opt, NOBH);
2239 }
2240 }
2241 /*
dab291af 2242 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 2243 * so we can safely mount the rest of the filesystem now.
2244 */
2245
617ba13b 2246 root = iget(sb, EXT4_ROOT_INO);
ac27a0ec
DK		 2247 sb->s_root = d_alloc_root(root);
2248 if (!sb->s_root) {
617ba13b 2249 printk(KERN_ERR "EXT4-fs: get root inode failed\n");
ac27a0ec
DK		 2250 iput(root);
2251 goto failed_mount4;
2252 }
2253 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
2254 dput(sb->s_root);
2255 sb->s_root = NULL;
617ba13b 2256 printk(KERN_ERR "EXT4-fs: corrupt root inode, run e2fsck\n");
ac27a0ec
DK		 2257 goto failed_mount4;
2258 }
2259
617ba13b 2260 ext4_setup_super (sb, es, sb->s_flags & MS_RDONLY);
ef7f3835
KS		 2261
2262 /* determine the minimum size of new large inodes, if present */
2263 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2264 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2265 EXT4_GOOD_OLD_INODE_SIZE;
2266 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2267 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2268 if (sbi->s_want_extra_isize <
2269 le16_to_cpu(es->s_want_extra_isize))
2270 sbi->s_want_extra_isize =
2271 le16_to_cpu(es->s_want_extra_isize);
2272 if (sbi->s_want_extra_isize <
2273 le16_to_cpu(es->s_min_extra_isize))
2274 sbi->s_want_extra_isize =
2275 le16_to_cpu(es->s_min_extra_isize);
2276 }
2277 }
2278 /* Check if enough inode space is available */
2279 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2280 sbi->s_inode_size) {
2281 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2282 EXT4_GOOD_OLD_INODE_SIZE;
2283 printk(KERN_INFO "EXT4-fs: required extra inode space not"
2284 "available.\n");
2285 }
2286
ac27a0ec
DK		 2287 /*
2288 * akpm: core read_super() calls in here with the superblock locked.
2289 * That deadlocks, because orphan cleanup needs to lock the superblock
2290 * in numerous places. Here we just pop the lock - it's relatively
2291 * harmless, because we are now ready to accept write_super() requests,
2292 * and aviro says that's the only reason for hanging onto the
2293 * superblock lock.
2294 */
617ba13b
MC		 2295 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
2296 ext4_orphan_cleanup(sb, es);
2297 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
ac27a0ec 2298 if (needs_recovery)
617ba13b
MC		 2299 printk (KERN_INFO "EXT4-fs: recovery complete.\n");
2300 ext4_mark_recovery_complete(sb, es);
2301 printk (KERN_INFO "EXT4-fs: mounted filesystem with %s data mode.\n",
2302 test_opt(sb,DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA ? "journal":
2303 test_opt(sb,DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA ? "ordered":
ac27a0ec
DK		 2304 "writeback");
2305
a86c6181 2306 ext4_ext_init(sb);
c9de560d 2307 ext4_mb_init(sb, needs_recovery);
a86c6181 2308
ac27a0ec
DK		 2309 lock_kernel();
2310 return 0;
2311
617ba13b 2312cantfind_ext4:
ac27a0ec 2313 if (!silent)
617ba13b 2314 printk(KERN_ERR "VFS: Can't find ext4 filesystem on dev %s.\n",
ac27a0ec
DK		 2315 sb->s_id);
2316 goto failed_mount;
2317
2318failed_mount4:
dab291af 2319 jbd2_journal_destroy(sbi->s_journal);
ac27a0ec
DK		 2320failed_mount3:
2321 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2322 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2323 percpu_counter_destroy(&sbi->s_dirs_counter);
2324failed_mount2:
2325 for (i = 0; i < db_count; i++)
2326 brelse(sbi->s_group_desc[i]);
2327 kfree(sbi->s_group_desc);
2328failed_mount:
2329#ifdef CONFIG_QUOTA
2330 for (i = 0; i < MAXQUOTAS; i++)
2331 kfree(sbi->s_qf_names[i]);
2332#endif
617ba13b 2333 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 2334 brelse(bh);
2335out_fail:
2336 sb->s_fs_info = NULL;
2337 kfree(sbi);
2338 lock_kernel();
2339 return -EINVAL;
2340}
2341
2342/*
2343 * Setup any per-fs journal parameters now. We'll do this both on
2344 * initial mount, once the journal has been initialised but before we've
2345 * done any recovery; and again on any subsequent remount.
2346 */
617ba13b 2347static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 2348{
617ba13b 2349 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 2350
2351 if (sbi->s_commit_interval)
2352 journal->j_commit_interval = sbi->s_commit_interval;
617ba13b 2353 /* We could also set up an ext4-specific default for the commit
ac27a0ec
DK		 2354 * interval here, but for now we'll just fall back to the jbd
2355 * default. */
2356
2357 spin_lock(&journal->j_state_lock);
2358 if (test_opt(sb, BARRIER))
dab291af 2359 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 2360 else
dab291af 2361 journal->j_flags &= ~JBD2_BARRIER;
ac27a0ec
DK		 2362 spin_unlock(&journal->j_state_lock);
2363}
2364
617ba13b 2365static journal_t *ext4_get_journal(struct super_block *sb,
ac27a0ec
DK		 2366 unsigned int journal_inum)
2367{
2368 struct inode *journal_inode;
2369 journal_t *journal;
2370
2371 /* First, test for the existence of a valid inode on disk. Bad
2372 * things happen if we iget() an unused inode, as the subsequent
2373 * iput() will try to delete it. */
2374
2375 journal_inode = iget(sb, journal_inum);
2376 if (!journal_inode) {
617ba13b 2377 printk(KERN_ERR "EXT4-fs: no journal found.\n");
ac27a0ec
DK		 2378 return NULL;
2379 }
2380 if (!journal_inode->i_nlink) {
2381 make_bad_inode(journal_inode);
2382 iput(journal_inode);
617ba13b 2383 printk(KERN_ERR "EXT4-fs: journal inode is deleted.\n");
ac27a0ec
DK		 2384 return NULL;
2385 }
2386
2387 jbd_debug(2, "Journal inode found at %p: %Ld bytes\n",
2388 journal_inode, journal_inode->i_size);
2389 if (is_bad_inode(journal_inode) || !S_ISREG(journal_inode->i_mode)) {
617ba13b 2390 printk(KERN_ERR "EXT4-fs: invalid journal inode.\n");
ac27a0ec
DK		 2391 iput(journal_inode);
2392 return NULL;
2393 }
2394
dab291af 2395 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 2396 if (!journal) {
617ba13b 2397 printk(KERN_ERR "EXT4-fs: Could not load journal inode\n");
ac27a0ec
DK		 2398 iput(journal_inode);
2399 return NULL;
2400 }
2401 journal->j_private = sb;
617ba13b 2402 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 2403 return journal;
2404}
2405
617ba13b 2406static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 2407 dev_t j_dev)
2408{
2409 struct buffer_head * bh;
2410 journal_t *journal;
617ba13b
MC		 2411 ext4_fsblk_t start;
2412 ext4_fsblk_t len;
ac27a0ec 2413 int hblock, blocksize;
617ba13b 2414 ext4_fsblk_t sb_block;
ac27a0ec 2415 unsigned long offset;
617ba13b 2416 struct ext4_super_block * es;
ac27a0ec
DK		 2417 struct block_device *bdev;
2418
617ba13b 2419 bdev = ext4_blkdev_get(j_dev);
ac27a0ec
DK		 2420 if (bdev == NULL)
2421 return NULL;
2422
2423 if (bd_claim(bdev, sb)) {
2424 printk(KERN_ERR
8c55e204 2425 "EXT4: failed to claim external journal device.\n");
ac27a0ec
DK		 2426 blkdev_put(bdev);
2427 return NULL;
2428 }
2429
2430 blocksize = sb->s_blocksize;
2431 hblock = bdev_hardsect_size(bdev);
2432 if (blocksize < hblock) {
2433 printk(KERN_ERR
617ba13b 2434 "EXT4-fs: blocksize too small for journal device.\n");
ac27a0ec
DK		 2435 goto out_bdev;
2436 }
2437
617ba13b
MC		 2438 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
2439 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 2440 set_blocksize(bdev, blocksize);
2441 if (!(bh = __bread(bdev, sb_block, blocksize))) {
617ba13b 2442 printk(KERN_ERR "EXT4-fs: couldn't read superblock of "
ac27a0ec
DK		 2443 "external journal\n");
2444 goto out_bdev;
2445 }
2446
617ba13b
MC		 2447 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2448 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 2449 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b
MC		 2450 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
2451 printk(KERN_ERR "EXT4-fs: external journal has "
ac27a0ec
DK		 2452 "bad superblock\n");
2453 brelse(bh);
2454 goto out_bdev;
2455 }
2456
617ba13b
MC		 2457 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
2458 printk(KERN_ERR "EXT4-fs: journal UUID does not match\n");
ac27a0ec
DK		 2459 brelse(bh);
2460 goto out_bdev;
2461 }
2462
bd81d8ee 2463 len = ext4_blocks_count(es);
ac27a0ec
DK		 2464 start = sb_block + 1;
2465 brelse(bh); /* we're done with the superblock */
2466
dab291af 2467 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 2468 start, len, blocksize);
2469 if (!journal) {
617ba13b 2470 printk(KERN_ERR "EXT4-fs: failed to create device journal\n");
ac27a0ec
DK		 2471 goto out_bdev;
2472 }
2473 journal->j_private = sb;
2474 ll_rw_block(READ, 1, &journal->j_sb_buffer);
2475 wait_on_buffer(journal->j_sb_buffer);
2476 if (!buffer_uptodate(journal->j_sb_buffer)) {
617ba13b 2477 printk(KERN_ERR "EXT4-fs: I/O error on journal device\n");
ac27a0ec
DK		 2478 goto out_journal;
2479 }
2480 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
617ba13b 2481 printk(KERN_ERR "EXT4-fs: External journal has more than one "
ac27a0ec
DK		 2482 "user (unsupported) - %d\n",
2483 be32_to_cpu(journal->j_superblock->s_nr_users));
2484 goto out_journal;
2485 }
617ba13b
MC		 2486 EXT4_SB(sb)->journal_bdev = bdev;
2487 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 2488 return journal;
2489out_journal:
dab291af 2490 jbd2_journal_destroy(journal);
ac27a0ec 2491out_bdev:
617ba13b 2492 ext4_blkdev_put(bdev);
ac27a0ec
DK		 2493 return NULL;
2494}
2495
617ba13b
MC		 2496static int ext4_load_journal(struct super_block *sb,
2497 struct ext4_super_block *es,
ac27a0ec
DK		 2498 unsigned long journal_devnum)
2499{
2500 journal_t *journal;
2501 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
2502 dev_t journal_dev;
2503 int err = 0;
2504 int really_read_only;
2505
2506 if (journal_devnum &&
2507 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
617ba13b 2508 printk(KERN_INFO "EXT4-fs: external journal device major/minor "
ac27a0ec
DK		 2509 "numbers have changed\n");
2510 journal_dev = new_decode_dev(journal_devnum);
2511 } else
2512 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
2513
2514 really_read_only = bdev_read_only(sb->s_bdev);
2515
2516 /*
2517 * Are we loading a blank journal or performing recovery after a
2518 * crash? For recovery, we need to check in advance whether we
2519 * can get read-write access to the device.
2520 */
2521
617ba13b 2522 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
ac27a0ec 2523 if (sb->s_flags & MS_RDONLY) {
617ba13b 2524 printk(KERN_INFO "EXT4-fs: INFO: recovery "
ac27a0ec
DK		 2525 "required on readonly filesystem.\n");
2526 if (really_read_only) {
617ba13b 2527 printk(KERN_ERR "EXT4-fs: write access "
ac27a0ec
DK		 2528 "unavailable, cannot proceed.\n");
2529 return -EROFS;
2530 }
617ba13b 2531 printk (KERN_INFO "EXT4-fs: write access will "
ac27a0ec
DK		 2532 "be enabled during recovery.\n");
2533 }
2534 }
2535
2536 if (journal_inum && journal_dev) {
617ba13b 2537 printk(KERN_ERR "EXT4-fs: filesystem has both journal "
ac27a0ec
DK		 2538 "and inode journals!\n");
2539 return -EINVAL;
2540 }
2541
2542 if (journal_inum) {
617ba13b 2543 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 2544 return -EINVAL;
2545 } else {
617ba13b 2546 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 2547 return -EINVAL;
2548 }
2549
2550 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
dab291af 2551 err = jbd2_journal_update_format(journal);
ac27a0ec 2552 if (err) {
617ba13b 2553 printk(KERN_ERR "EXT4-fs: error updating journal.\n");
dab291af 2554 jbd2_journal_destroy(journal);
ac27a0ec
DK		 2555 return err;
2556 }
2557 }
2558
617ba13b 2559 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
dab291af 2560 err = jbd2_journal_wipe(journal, !really_read_only);
ac27a0ec 2561 if (!err)
dab291af 2562 err = jbd2_journal_load(journal);
ac27a0ec
DK		 2563
2564 if (err) {
617ba13b 2565 printk(KERN_ERR "EXT4-fs: error loading journal.\n");
dab291af 2566 jbd2_journal_destroy(journal);
ac27a0ec
DK		 2567 return err;
2568 }
2569
617ba13b
MC		 2570 EXT4_SB(sb)->s_journal = journal;
2571 ext4_clear_journal_err(sb, es);
ac27a0ec
DK		 2572
2573 if (journal_devnum &&
2574 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2575 es->s_journal_dev = cpu_to_le32(journal_devnum);
2576 sb->s_dirt = 1;
2577
2578 /* Make sure we flush the recovery flag to disk. */
617ba13b 2579 ext4_commit_super(sb, es, 1);
ac27a0ec
DK		 2580 }
2581
2582 return 0;
2583}
2584
617ba13b
MC		 2585static int ext4_create_journal(struct super_block * sb,
2586 struct ext4_super_block * es,
ac27a0ec
DK		 2587 unsigned int journal_inum)
2588{
2589 journal_t *journal;
6c675bd4 2590 int err;
ac27a0ec
DK		 2591
2592 if (sb->s_flags & MS_RDONLY) {
617ba13b 2593 printk(KERN_ERR "EXT4-fs: readonly filesystem when trying to "
ac27a0ec
DK		 2594 "create journal.\n");
2595 return -EROFS;
2596 }
2597
6c675bd4
BP		 2598 journal = ext4_get_journal(sb, journal_inum);
2599 if (!journal)
ac27a0ec
DK		 2600 return -EINVAL;
2601
617ba13b 2602 printk(KERN_INFO "EXT4-fs: creating new journal on inode %u\n",
ac27a0ec
DK		 2603 journal_inum);
2604
6c675bd4
BP		 2605 err = jbd2_journal_create(journal);
2606 if (err) {
617ba13b 2607 printk(KERN_ERR "EXT4-fs: error creating journal.\n");
dab291af 2608 jbd2_journal_destroy(journal);
ac27a0ec
DK		 2609 return -EIO;
2610 }
2611
617ba13b 2612 EXT4_SB(sb)->s_journal = journal;
ac27a0ec 2613
617ba13b
MC		 2614 ext4_update_dynamic_rev(sb);
2615 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2616 EXT4_SET_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL);
ac27a0ec
DK		 2617
2618 es->s_journal_inum = cpu_to_le32(journal_inum);
2619 sb->s_dirt = 1;
2620
2621 /* Make sure we flush the recovery flag to disk. */
617ba13b 2622 ext4_commit_super(sb, es, 1);
ac27a0ec
DK		 2623
2624 return 0;
2625}
2626
617ba13b
MC		 2627static void ext4_commit_super (struct super_block * sb,
2628 struct ext4_super_block * es,
ac27a0ec
DK		 2629 int sync)
2630{
617ba13b 2631 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
ac27a0ec
DK		 2632
2633 if (!sbh)
2634 return;
2635 es->s_wtime = cpu_to_le32(get_seconds());
bd81d8ee 2636 ext4_free_blocks_count_set(es, ext4_count_free_blocks(sb));
617ba13b 2637 es->s_free_inodes_count = cpu_to_le32(ext4_count_free_inodes(sb));
ac27a0ec
DK		 2638 BUFFER_TRACE(sbh, "marking dirty");
2639 mark_buffer_dirty(sbh);
2640 if (sync)
2641 sync_dirty_buffer(sbh);
2642}
2643
2644
2645/*
2646 * Have we just finished recovery? If so, and if we are mounting (or
2647 * remounting) the filesystem readonly, then we will end up with a
2648 * consistent fs on disk. Record that fact.
2649 */
617ba13b
MC		 2650static void ext4_mark_recovery_complete(struct super_block * sb,
2651 struct ext4_super_block * es)
ac27a0ec 2652{
617ba13b 2653 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 2654
dab291af
MC		 2655 jbd2_journal_lock_updates(journal);
2656 jbd2_journal_flush(journal);
32c37730 2657 lock_super(sb);
617ba13b 2658 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
ac27a0ec 2659 sb->s_flags & MS_RDONLY) {
617ba13b 2660 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 2661 sb->s_dirt = 0;
617ba13b 2662 ext4_commit_super(sb, es, 1);
ac27a0ec 2663 }
32c37730 2664 unlock_super(sb);
dab291af 2665 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 2666}
2667
2668/*
2669 * If we are mounting (or read-write remounting) a filesystem whose journal
2670 * has recorded an error from a previous lifetime, move that error to the
2671 * main filesystem now.
2672 */
617ba13b
MC		 2673static void ext4_clear_journal_err(struct super_block * sb,
2674 struct ext4_super_block * es)
ac27a0ec
DK		 2675{
2676 journal_t *journal;
2677 int j_errno;
2678 const char *errstr;
2679
617ba13b 2680 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 2681
2682 /*
2683 * Now check for any error status which may have been recorded in the
617ba13b 2684 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 2685 */
2686
dab291af 2687 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 2688 if (j_errno) {
2689 char nbuf[16];
2690
617ba13b
MC		 2691 errstr = ext4_decode_error(sb, j_errno, nbuf);
2692 ext4_warning(sb, __FUNCTION__, "Filesystem error recorded "
ac27a0ec 2693 "from previous mount: %s", errstr);
617ba13b 2694 ext4_warning(sb, __FUNCTION__, "Marking fs in need of "
ac27a0ec
DK		 2695 "filesystem check.");
2696
617ba13b
MC		 2697 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2698 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2699 ext4_commit_super (sb, es, 1);
ac27a0ec 2700
dab291af 2701 jbd2_journal_clear_err(journal);
ac27a0ec
DK		 2702 }
2703}
2704
2705/*
2706 * Force the running and committing transactions to commit,
2707 * and wait on the commit.
2708 */
617ba13b 2709int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 2710{
2711 journal_t *journal;
2712 int ret;
2713
2714 if (sb->s_flags & MS_RDONLY)
2715 return 0;
2716
617ba13b 2717 journal = EXT4_SB(sb)->s_journal;
ac27a0ec 2718 sb->s_dirt = 0;
617ba13b 2719 ret = ext4_journal_force_commit(journal);
ac27a0ec
DK		 2720 return ret;
2721}
2722
2723/*
617ba13b 2724 * Ext4 always journals updates to the superblock itself, so we don't
ac27a0ec
DK		 2725 * have to propagate any other updates to the superblock on disk at this
2726 * point. Just start an async writeback to get the buffers on their way
2727 * to the disk.
2728 *
2729 * This implicitly triggers the writebehind on sync().
2730 */
2731
617ba13b 2732static void ext4_write_super (struct super_block * sb)
ac27a0ec
DK		 2733{
2734 if (mutex_trylock(&sb->s_lock) != 0)
2735 BUG();
2736 sb->s_dirt = 0;
2737}
2738
617ba13b 2739static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec
DK		 2740{
2741 tid_t target;
2742
2743 sb->s_dirt = 0;
dab291af 2744 if (jbd2_journal_start_commit(EXT4_SB(sb)->s_journal, &target)) {
ac27a0ec 2745 if (wait)
dab291af 2746 jbd2_log_wait_commit(EXT4_SB(sb)->s_journal, target);
ac27a0ec
DK		 2747 }
2748 return 0;
2749}
2750
2751/*
2752 * LVM calls this function before a (read-only) snapshot is created. This
2753 * gives us a chance to flush the journal completely and mark the fs clean.
2754 */
617ba13b 2755static void ext4_write_super_lockfs(struct super_block *sb)
ac27a0ec
DK		 2756{
2757 sb->s_dirt = 0;
2758
2759 if (!(sb->s_flags & MS_RDONLY)) {
617ba13b 2760 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 2761
2762 /* Now we set up the journal barrier. */
dab291af
MC		 2763 jbd2_journal_lock_updates(journal);
2764 jbd2_journal_flush(journal);
ac27a0ec
DK		 2765
2766 /* Journal blocked and flushed, clear needs_recovery flag. */
617ba13b
MC		 2767 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2768 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
ac27a0ec
DK		 2769 }
2770}
2771
2772/*
2773 * Called by LVM after the snapshot is done. We need to reset the RECOVER
2774 * flag here, even though the filesystem is not technically dirty yet.
2775 */
617ba13b 2776static void ext4_unlockfs(struct super_block *sb)
ac27a0ec
DK		 2777{
2778 if (!(sb->s_flags & MS_RDONLY)) {
2779 lock_super(sb);
2780 /* Reser the needs_recovery flag before the fs is unlocked. */
617ba13b
MC		 2781 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2782 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
ac27a0ec 2783 unlock_super(sb);
dab291af 2784 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
ac27a0ec
DK		 2785 }
2786}
2787
617ba13b 2788static int ext4_remount (struct super_block * sb, int * flags, char * data)
ac27a0ec 2789{
617ba13b
MC		 2790 struct ext4_super_block * es;
2791 struct ext4_sb_info *sbi = EXT4_SB(sb);
2792 ext4_fsblk_t n_blocks_count = 0;
ac27a0ec 2793 unsigned long old_sb_flags;
617ba13b 2794 struct ext4_mount_options old_opts;
ac27a0ec
DK		 2795 int err;
2796#ifdef CONFIG_QUOTA
2797 int i;
2798#endif
2799
2800 /* Store the original options */
2801 old_sb_flags = sb->s_flags;
2802 old_opts.s_mount_opt = sbi->s_mount_opt;
2803 old_opts.s_resuid = sbi->s_resuid;
2804 old_opts.s_resgid = sbi->s_resgid;
2805 old_opts.s_commit_interval = sbi->s_commit_interval;
2806#ifdef CONFIG_QUOTA
2807 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
2808 for (i = 0; i < MAXQUOTAS; i++)
2809 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
2810#endif
2811
2812 /*
2813 * Allow the "check" option to be passed as a remount option.
2814 */
2815 if (!parse_options(data, sb, NULL, NULL, &n_blocks_count, 1)) {
2816 err = -EINVAL;
2817 goto restore_opts;
2818 }
2819
617ba13b
MC		 2820 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT)
2821 ext4_abort(sb, __FUNCTION__, "Abort forced by user");
ac27a0ec
DK		 2822
2823 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
617ba13b 2824 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec
DK		 2825
2826 es = sbi->s_es;
2827
617ba13b 2828 ext4_init_journal_params(sb, sbi->s_journal);
ac27a0ec
DK		 2829
2830 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
bd81d8ee 2831 n_blocks_count > ext4_blocks_count(es)) {
617ba13b 2832 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT) {
ac27a0ec
DK		 2833 err = -EROFS;
2834 goto restore_opts;
2835 }
2836
2837 if (*flags & MS_RDONLY) {
2838 /*
2839 * First of all, the unconditional stuff we have to do
2840 * to disable replay of the journal when we next remount
2841 */
2842 sb->s_flags |= MS_RDONLY;
2843
2844 /*
2845 * OK, test if we are remounting a valid rw partition
2846 * readonly, and if so set the rdonly flag and then
2847 * mark the partition as valid again.
2848 */
617ba13b
MC		 2849 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
2850 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 2851 es->s_state = cpu_to_le16(sbi->s_mount_state);
2852
32c37730
JK		 2853 /*
2854 * We have to unlock super so that we can wait for
2855 * transactions.
2856 */
2857 unlock_super(sb);
617ba13b 2858 ext4_mark_recovery_complete(sb, es);
32c37730 2859 lock_super(sb);
ac27a0ec
DK		 2860 } else {
2861 __le32 ret;
617ba13b
MC		 2862 if ((ret = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2863 ~EXT4_FEATURE_RO_COMPAT_SUPP))) {
2864 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
ac27a0ec
DK		 2865 "remount RDWR because of unsupported "
2866 "optional features (%x).\n",
2867 sb->s_id, le32_to_cpu(ret));
2868 err = -EROFS;
2869 goto restore_opts;
2870 }
ead6596b
ES		 2871
2872 /*
2873 * If we have an unprocessed orphan list hanging
2874 * around from a previously readonly bdev mount,
2875 * require a full umount/remount for now.
2876 */
2877 if (es->s_last_orphan) {
2878 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
2879 "remount RDWR because of unprocessed "
2880 "orphan inode list. Please "
2881 "umount/remount instead.\n",
2882 sb->s_id);
2883 err = -EINVAL;
2884 goto restore_opts;
2885 }
2886
ac27a0ec
DK		 2887 /*
2888 * Mounting a RDONLY partition read-write, so reread
2889 * and store the current valid flag. (It may have
2890 * been changed by e2fsck since we originally mounted
2891 * the partition.)
2892 */
617ba13b 2893 ext4_clear_journal_err(sb, es);
ac27a0ec 2894 sbi->s_mount_state = le16_to_cpu(es->s_state);
617ba13b 2895 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
ac27a0ec 2896 goto restore_opts;
617ba13b 2897 if (!ext4_setup_super (sb, es, 0))
ac27a0ec
DK		 2898 sb->s_flags &= ~MS_RDONLY;
2899 }
2900 }
2901#ifdef CONFIG_QUOTA
2902 /* Release old quota file names */
2903 for (i = 0; i < MAXQUOTAS; i++)
2904 if (old_opts.s_qf_names[i] &&
2905 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2906 kfree(old_opts.s_qf_names[i]);
2907#endif
2908 return 0;
2909restore_opts:
2910 sb->s_flags = old_sb_flags;
2911 sbi->s_mount_opt = old_opts.s_mount_opt;
2912 sbi->s_resuid = old_opts.s_resuid;
2913 sbi->s_resgid = old_opts.s_resgid;
2914 sbi->s_commit_interval = old_opts.s_commit_interval;
2915#ifdef CONFIG_QUOTA
2916 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
2917 for (i = 0; i < MAXQUOTAS; i++) {
2918 if (sbi->s_qf_names[i] &&
2919 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2920 kfree(sbi->s_qf_names[i]);
2921 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
2922 }
2923#endif
2924 return err;
2925}
2926
617ba13b 2927static int ext4_statfs (struct dentry * dentry, struct kstatfs * buf)
ac27a0ec
DK		 2928{
2929 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 2930 struct ext4_sb_info *sbi = EXT4_SB(sb);
2931 struct ext4_super_block *es = sbi->s_es;
960cc398 2932 u64 fsid;
ac27a0ec 2933
5e70030d
BP		 2934 if (test_opt(sb, MINIX_DF)) {
2935 sbi->s_overhead_last = 0;
6bc9feff 2936 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
fd2d4291 2937 ext4_group_t ngroups = sbi->s_groups_count, i;
5e70030d 2938 ext4_fsblk_t overhead = 0;
ac27a0ec
DK		 2939 smp_rmb();
2940
2941 /*
5e70030d
BP		 2942 * Compute the overhead (FS structures). This is constant
2943 * for a given filesystem unless the number of block groups
2944 * changes so we cache the previous value until it does.
ac27a0ec
DK		 2945 */
2946
2947 /*
2948 * All of the blocks before first_data_block are
2949 * overhead
2950 */
2951 overhead = le32_to_cpu(es->s_first_data_block);
2952
2953 /*
2954 * Add the overhead attributed to the superblock and
2955 * block group descriptors. If the sparse superblocks
2956 * feature is turned on, then not all groups have this.
2957 */
2958 for (i = 0; i < ngroups; i++) {
617ba13b
MC		 2959 overhead += ext4_bg_has_super(sb, i) +
2960 ext4_bg_num_gdb(sb, i);
ac27a0ec
DK		 2961 cond_resched();
2962 }
2963
2964 /*
2965 * Every block group has an inode bitmap, a block
2966 * bitmap, and an inode table.
2967 */
5e70030d
BP		 2968 overhead += ngroups * (2 + sbi->s_itb_per_group);
2969 sbi->s_overhead_last = overhead;
2970 smp_wmb();
6bc9feff 2971 sbi->s_blocks_last = ext4_blocks_count(es);
ac27a0ec
DK		 2972 }
2973
617ba13b 2974 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 2975 buf->f_bsize = sb->s_blocksize;
5e70030d 2976 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
52d9f3b4 2977 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter);
308ba3ec 2978 ext4_free_blocks_count_set(es, buf->f_bfree);
bd81d8ee
LV		 2979 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
2980 if (buf->f_bfree < ext4_r_blocks_count(es))
ac27a0ec
DK		 2981 buf->f_bavail = 0;
2982 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 2983 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
5e70030d 2984 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
617ba13b 2985 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 2986 fsid = le64_to_cpup((void *)es->s_uuid) ^
2987 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
2988 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
2989 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
ac27a0ec
DK		 2990 return 0;
2991}
2992
2993/* Helper function for writing quotas on sync - we need to start transaction before quota file
2994 * is locked for write. Otherwise the are possible deadlocks:
2995 * Process 1 Process 2
617ba13b 2996 * ext4_create() quota_sync()
dab291af 2997 * jbd2_journal_start() write_dquot()
ac27a0ec 2998 * DQUOT_INIT() down(dqio_mutex)
dab291af 2999 * down(dqio_mutex) jbd2_journal_start()
ac27a0ec
DK		 3000 *
3001 */
3002
3003#ifdef CONFIG_QUOTA
3004
3005static inline struct inode *dquot_to_inode(struct dquot *dquot)
3006{
3007 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3008}
3009
617ba13b 3010static int ext4_dquot_initialize(struct inode *inode, int type)
ac27a0ec
DK		 3011{
3012 handle_t *handle;
3013 int ret, err;
3014
3015 /* We may create quota structure so we need to reserve enough blocks */
617ba13b 3016 handle = ext4_journal_start(inode, 2*EXT4_QUOTA_INIT_BLOCKS(inode->i_sb));
ac27a0ec
DK		 3017 if (IS_ERR(handle))
3018 return PTR_ERR(handle);
3019 ret = dquot_initialize(inode, type);
617ba13b 3020 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3021 if (!ret)
3022 ret = err;
3023 return ret;
3024}
3025
617ba13b 3026static int ext4_dquot_drop(struct inode *inode)
ac27a0ec
DK		 3027{
3028 handle_t *handle;
3029 int ret, err;
3030
3031 /* We may delete quota structure so we need to reserve enough blocks */
617ba13b 3032 handle = ext4_journal_start(inode, 2*EXT4_QUOTA_DEL_BLOCKS(inode->i_sb));
ac27a0ec
DK		 3033 if (IS_ERR(handle))
3034 return PTR_ERR(handle);
3035 ret = dquot_drop(inode);
617ba13b 3036 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3037 if (!ret)
3038 ret = err;
3039 return ret;
3040}
3041
617ba13b 3042static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 3043{
3044 int ret, err;
3045 handle_t *handle;
3046 struct inode *inode;
3047
3048 inode = dquot_to_inode(dquot);
617ba13b
MC		 3049 handle = ext4_journal_start(inode,
3050 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 3051 if (IS_ERR(handle))
3052 return PTR_ERR(handle);
3053 ret = dquot_commit(dquot);
617ba13b 3054 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3055 if (!ret)
3056 ret = err;
3057 return ret;
3058}
3059
617ba13b 3060static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 3061{
3062 int ret, err;
3063 handle_t *handle;
3064
617ba13b
MC		 3065 handle = ext4_journal_start(dquot_to_inode(dquot),
3066 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 3067 if (IS_ERR(handle))
3068 return PTR_ERR(handle);
3069 ret = dquot_acquire(dquot);
617ba13b 3070 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3071 if (!ret)
3072 ret = err;
3073 return ret;
3074}
3075
617ba13b 3076static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 3077{
3078 int ret, err;
3079 handle_t *handle;
3080
617ba13b
MC		 3081 handle = ext4_journal_start(dquot_to_inode(dquot),
3082 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 3083 if (IS_ERR(handle)) {
3084 /* Release dquot anyway to avoid endless cycle in dqput() */
3085 dquot_release(dquot);
ac27a0ec 3086 return PTR_ERR(handle);
9c3013e9 3087 }
ac27a0ec 3088 ret = dquot_release(dquot);
617ba13b 3089 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3090 if (!ret)
3091 ret = err;
3092 return ret;
3093}
3094
617ba13b 3095static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec
DK		 3096{
3097 /* Are we journalling quotas? */
617ba13b
MC		 3098 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3099 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
ac27a0ec 3100 dquot_mark_dquot_dirty(dquot);
617ba13b 3101 return ext4_write_dquot(dquot);
ac27a0ec
DK		 3102 } else {
3103 return dquot_mark_dquot_dirty(dquot);
3104 }
3105}
3106
617ba13b 3107static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 3108{
3109 int ret, err;
3110 handle_t *handle;
3111
3112 /* Data block + inode block */
617ba13b 3113 handle = ext4_journal_start(sb->s_root->d_inode, 2);
ac27a0ec
DK		 3114 if (IS_ERR(handle))
3115 return PTR_ERR(handle);
3116 ret = dquot_commit_info(sb, type);
617ba13b 3117 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3118 if (!ret)
3119 ret = err;
3120 return ret;
3121}
3122
3123/*
3124 * Turn on quotas during mount time - we need to find
3125 * the quota file and such...
3126 */
617ba13b 3127static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 3128{
617ba13b
MC		 3129 return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
3130 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 3131}
3132
3133/*
3134 * Standard function to be called on quota_on
3135 */
617ba13b 3136static int ext4_quota_on(struct super_block *sb, int type, int format_id,
ac27a0ec
DK		 3137 char *path)
3138{
3139 int err;
3140 struct nameidata nd;
3141
3142 if (!test_opt(sb, QUOTA))
3143 return -EINVAL;
3144 /* Not journalling quota? */
617ba13b
MC		 3145 if (!EXT4_SB(sb)->s_qf_names[USRQUOTA] &&
3146 !EXT4_SB(sb)->s_qf_names[GRPQUOTA])
ac27a0ec
DK		 3147 return vfs_quota_on(sb, type, format_id, path);
3148 err = path_lookup(path, LOOKUP_FOLLOW, &nd);
3149 if (err)
3150 return err;
3151 /* Quotafile not on the same filesystem? */
3152 if (nd.mnt->mnt_sb != sb) {
3153 path_release(&nd);
3154 return -EXDEV;
3155 }
3156 /* Quotafile not of fs root? */
3157 if (nd.dentry->d_parent->d_inode != sb->s_root->d_inode)
3158 printk(KERN_WARNING
617ba13b 3159 "EXT4-fs: Quota file not on filesystem root. "
ac27a0ec
DK		 3160 "Journalled quota will not work.\n");
3161 path_release(&nd);
3162 return vfs_quota_on(sb, type, format_id, path);
3163}
3164
3165/* Read data from quotafile - avoid pagecache and such because we cannot afford
3166 * acquiring the locks... As quota files are never truncated and quota code
3167 * itself serializes the operations (and noone else should touch the files)
3168 * we don't have to be afraid of races */
617ba13b 3169static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 3170 size_t len, loff_t off)
3171{
3172 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 3173 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 3174 int err = 0;
3175 int offset = off & (sb->s_blocksize - 1);
3176 int tocopy;
3177 size_t toread;
3178 struct buffer_head *bh;
3179 loff_t i_size = i_size_read(inode);
3180
3181 if (off > i_size)
3182 return 0;
3183 if (off+len > i_size)
3184 len = i_size-off;
3185 toread = len;
3186 while (toread > 0) {
3187 tocopy = sb->s_blocksize - offset < toread ?
3188 sb->s_blocksize - offset : toread;
617ba13b 3189 bh = ext4_bread(NULL, inode, blk, 0, &err);
ac27a0ec
DK		 3190 if (err)
3191 return err;
3192 if (!bh) /* A hole? */
3193 memset(data, 0, tocopy);
3194 else
3195 memcpy(data, bh->b_data+offset, tocopy);
3196 brelse(bh);
3197 offset = 0;
3198 toread -= tocopy;
3199 data += tocopy;
3200 blk++;
3201 }
3202 return len;
3203}
3204
3205/* Write to quotafile (we know the transaction is already started and has
3206 * enough credits) */
617ba13b 3207static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 3208 const char *data, size_t len, loff_t off)
3209{
3210 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 3211 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 3212 int err = 0;
3213 int offset = off & (sb->s_blocksize - 1);
3214 int tocopy;
617ba13b 3215 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
ac27a0ec
DK		 3216 size_t towrite = len;
3217 struct buffer_head *bh;
3218 handle_t *handle = journal_current_handle();
3219
9c3013e9
JK		 3220 if (!handle) {
3221 printk(KERN_WARNING "EXT4-fs: Quota write (off=%Lu, len=%Lu)"
3222 " cancelled because transaction is not started.\n",
3223 (unsigned long long)off, (unsigned long long)len);
3224 return -EIO;
3225 }
ac27a0ec
DK		 3226 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
3227 while (towrite > 0) {
3228 tocopy = sb->s_blocksize - offset < towrite ?
3229 sb->s_blocksize - offset : towrite;
617ba13b 3230 bh = ext4_bread(handle, inode, blk, 1, &err);
ac27a0ec
DK		 3231 if (!bh)
3232 goto out;
3233 if (journal_quota) {
617ba13b 3234 err = ext4_journal_get_write_access(handle, bh);
ac27a0ec
DK		 3235 if (err) {
3236 brelse(bh);
3237 goto out;
3238 }
3239 }
3240 lock_buffer(bh);
3241 memcpy(bh->b_data+offset, data, tocopy);
3242 flush_dcache_page(bh->b_page);
3243 unlock_buffer(bh);
3244 if (journal_quota)
617ba13b 3245 err = ext4_journal_dirty_metadata(handle, bh);
ac27a0ec
DK		 3246 else {
3247 /* Always do at least ordered writes for quotas */
617ba13b 3248 err = ext4_journal_dirty_data(handle, bh);
ac27a0ec
DK		 3249 mark_buffer_dirty(bh);
3250 }
3251 brelse(bh);
3252 if (err)
3253 goto out;
3254 offset = 0;
3255 towrite -= tocopy;
3256 data += tocopy;
3257 blk++;
3258 }
3259out:
3260 if (len == towrite)
3261 return err;
3262 if (inode->i_size < off+len-towrite) {
3263 i_size_write(inode, off+len-towrite);
617ba13b 3264 EXT4_I(inode)->i_disksize = inode->i_size;
ac27a0ec 3265 }
ac27a0ec 3266 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
617ba13b 3267 ext4_mark_inode_dirty(handle, inode);
ac27a0ec
DK		 3268 mutex_unlock(&inode->i_mutex);
3269 return len - towrite;
3270}
3271
3272#endif
3273
617ba13b 3274static int ext4_get_sb(struct file_system_type *fs_type,
ac27a0ec
DK		 3275 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3276{
617ba13b 3277 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
ac27a0ec
DK		 3278}
3279
617ba13b 3280static struct file_system_type ext4dev_fs_type = {
ac27a0ec 3281 .owner = THIS_MODULE,
617ba13b
MC		 3282 .name = "ext4dev",
3283 .get_sb = ext4_get_sb,
ac27a0ec
DK		 3284 .kill_sb = kill_block_super,
3285 .fs_flags = FS_REQUIRES_DEV,
3286};
3287
617ba13b 3288static int __init init_ext4_fs(void)
ac27a0ec 3289{
c9de560d
AT		 3290 int err;
3291
3292 err = init_ext4_mballoc();
ac27a0ec
DK		 3293 if (err)
3294 return err;
c9de560d
AT		 3295
3296 err = init_ext4_xattr();
3297 if (err)
3298 goto out2;
ac27a0ec
DK		 3299 err = init_inodecache();
3300 if (err)
3301 goto out1;
63f57933 3302 err = register_filesystem(&ext4dev_fs_type);
ac27a0ec
DK		 3303 if (err)
3304 goto out;
3305 return 0;
3306out:
3307 destroy_inodecache();
3308out1:
617ba13b 3309 exit_ext4_xattr();
c9de560d
AT		 3310out2:
3311 exit_ext4_mballoc();
ac27a0ec
DK		 3312 return err;
3313}
3314
617ba13b 3315static void __exit exit_ext4_fs(void)
ac27a0ec 3316{
617ba13b 3317 unregister_filesystem(&ext4dev_fs_type);
ac27a0ec 3318 destroy_inodecache();
617ba13b 3319 exit_ext4_xattr();
c9de560d 3320 exit_ext4_mballoc();
ac27a0ec
DK		 3321}
3322
3323MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
617ba13b 3324MODULE_DESCRIPTION("Fourth Extended Filesystem with extents");
ac27a0ec 3325MODULE_LICENSE("GPL");
617ba13b
MC		 3326module_init(init_ext4_fs)
3327module_exit(exit_ext4_fs)
Linux kernel - Deliverables
This page took 0.490375 seconds and 5 git commands to generate.