]> git.proxmox.com Git - mirror_ubuntu-artful-kernel.git/blame - fs/ext4/super.c
projects / mirror_ubuntu-artful-kernel.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
ext4: Add new functions for searching extent tree
[mirror_ubuntu-artful-kernel.git] / fs / ext4 / super.c
CommitLineData
ac27a0ec 1/*
617ba13b 2 * linux/fs/ext4/super.c
ac27a0ec
DK		 3 *
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
8 *
9 * from
10 *
11 * linux/fs/minix/inode.c
12 *
13 * Copyright (C) 1991, 1992 Linus Torvalds
14 *
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
17 */
18
19#include <linux/module.h>
20#include <linux/string.h>
21#include <linux/fs.h>
22#include <linux/time.h>
dab291af 23#include <linux/jbd2.h>
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
a86c6181 506 ext4_ext_release(sb);
617ba13b 507 ext4_xattr_put_super(sb);
dab291af 508 jbd2_journal_destroy(sbi->s_journal);
ac27a0ec 509 if (!(sb->s_flags & MS_RDONLY)) {
617ba13b 510 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec
DK		 511 es->s_state = cpu_to_le16(sbi->s_mount_state);
512 BUFFER_TRACE(sbi->s_sbh, "marking dirty");
513 mark_buffer_dirty(sbi->s_sbh);
617ba13b 514 ext4_commit_super(sb, es, 1);
ac27a0ec
DK		 515 }
516
517 for (i = 0; i < sbi->s_gdb_count; i++)
518 brelse(sbi->s_group_desc[i]);
519 kfree(sbi->s_group_desc);
520 percpu_counter_destroy(&sbi->s_freeblocks_counter);
521 percpu_counter_destroy(&sbi->s_freeinodes_counter);
522 percpu_counter_destroy(&sbi->s_dirs_counter);
523 brelse(sbi->s_sbh);
524#ifdef CONFIG_QUOTA
525 for (i = 0; i < MAXQUOTAS; i++)
526 kfree(sbi->s_qf_names[i]);
527#endif
528
529 /* Debugging code just in case the in-memory inode orphan list
530 * isn't empty. The on-disk one can be non-empty if we've
531 * detected an error and taken the fs readonly, but the
532 * in-memory list had better be clean by this point. */
533 if (!list_empty(&sbi->s_orphan))
534 dump_orphan_list(sb, sbi);
535 J_ASSERT(list_empty(&sbi->s_orphan));
536
f98393a6 537 invalidate_bdev(sb->s_bdev);
ac27a0ec
DK		 538 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
539 /*
540 * Invalidate the journal device's buffers. We don't want them
541 * floating about in memory - the physical journal device may
542 * hotswapped, and it breaks the `ro-after' testing code.
543 */
544 sync_blockdev(sbi->journal_bdev);
f98393a6 545 invalidate_bdev(sbi->journal_bdev);
617ba13b 546 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 547 }
548 sb->s_fs_info = NULL;
549 kfree(sbi);
550 return;
551}
552
e18b890b 553static struct kmem_cache *ext4_inode_cachep;
ac27a0ec
DK		 554
555/*
556 * Called inside transaction, so use GFP_NOFS
557 */
617ba13b 558static struct inode *ext4_alloc_inode(struct super_block *sb)
ac27a0ec 559{
617ba13b 560 struct ext4_inode_info *ei;
ac27a0ec 561
e6b4f8da 562 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
ac27a0ec
DK		 563 if (!ei)
564 return NULL;
617ba13b
MC		 565#ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
566 ei->i_acl = EXT4_ACL_NOT_CACHED;
567 ei->i_default_acl = EXT4_ACL_NOT_CACHED;
ac27a0ec
DK		 568#endif
569 ei->i_block_alloc_info = NULL;
570 ei->vfs_inode.i_version = 1;
a86c6181 571 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
ac27a0ec
DK		 572 return &ei->vfs_inode;
573}
574
617ba13b 575static void ext4_destroy_inode(struct inode *inode)
ac27a0ec 576{
9f7dd93d
VA		 577 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
578 printk("EXT4 Inode %p: orphan list check failed!\n",
579 EXT4_I(inode));
580 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
581 EXT4_I(inode), sizeof(struct ext4_inode_info),
582 true);
583 dump_stack();
584 }
617ba13b 585 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
ac27a0ec
DK		 586}
587
4ba9b9d0 588static void init_once(struct kmem_cache *cachep, void *foo)
ac27a0ec 589{
617ba13b 590 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
ac27a0ec 591
a35afb83 592 INIT_LIST_HEAD(&ei->i_orphan);
617ba13b 593#ifdef CONFIG_EXT4DEV_FS_XATTR
a35afb83 594 init_rwsem(&ei->xattr_sem);
ac27a0ec 595#endif
0e855ac8 596 init_rwsem(&ei->i_data_sem);
a35afb83 597 inode_init_once(&ei->vfs_inode);
ac27a0ec
DK		 598}
599
600static int init_inodecache(void)
601{
617ba13b
MC		 602 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
603 sizeof(struct ext4_inode_info),
ac27a0ec
DK		 604 0, (SLAB_RECLAIM_ACCOUNT|
605 SLAB_MEM_SPREAD),
20c2df83 606 init_once);
617ba13b 607 if (ext4_inode_cachep == NULL)
ac27a0ec
DK		 608 return -ENOMEM;
609 return 0;
610}
611
612static void destroy_inodecache(void)
613{
617ba13b 614 kmem_cache_destroy(ext4_inode_cachep);
ac27a0ec
DK		 615}
616
617ba13b 617static void ext4_clear_inode(struct inode *inode)
ac27a0ec 618{
617ba13b
MC		 619 struct ext4_block_alloc_info *rsv = EXT4_I(inode)->i_block_alloc_info;
620#ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
621 if (EXT4_I(inode)->i_acl &&
622 EXT4_I(inode)->i_acl != EXT4_ACL_NOT_CACHED) {
623 posix_acl_release(EXT4_I(inode)->i_acl);
624 EXT4_I(inode)->i_acl = EXT4_ACL_NOT_CACHED;
625 }
626 if (EXT4_I(inode)->i_default_acl &&
627 EXT4_I(inode)->i_default_acl != EXT4_ACL_NOT_CACHED) {
628 posix_acl_release(EXT4_I(inode)->i_default_acl);
629 EXT4_I(inode)->i_default_acl = EXT4_ACL_NOT_CACHED;
ac27a0ec
DK		 630 }
631#endif
617ba13b
MC		 632 ext4_discard_reservation(inode);
633 EXT4_I(inode)->i_block_alloc_info = NULL;
ac27a0ec
DK		 634 if (unlikely(rsv))
635 kfree(rsv);
636}
637
617ba13b 638static inline void ext4_show_quota_options(struct seq_file *seq, struct super_block *sb)
ac27a0ec
DK		 639{
640#if defined(CONFIG_QUOTA)
617ba13b 641 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 642
643 if (sbi->s_jquota_fmt)
644 seq_printf(seq, ",jqfmt=%s",
645 (sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold": "vfsv0");
646
647 if (sbi->s_qf_names[USRQUOTA])
648 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
649
650 if (sbi->s_qf_names[GRPQUOTA])
651 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
652
617ba13b 653 if (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA)
ac27a0ec
DK		 654 seq_puts(seq, ",usrquota");
655
617ba13b 656 if (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)
ac27a0ec
DK		 657 seq_puts(seq, ",grpquota");
658#endif
659}
660
d9c9bef1
MS		 661/*
662 * Show an option if
663 * - it's set to a non-default value OR
664 * - if the per-sb default is different from the global default
665 */
617ba13b 666static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
ac27a0ec 667{
aa22df2c
AK		 668 int def_errors;
669 unsigned long def_mount_opts;
ac27a0ec 670 struct super_block *sb = vfs->mnt_sb;
d9c9bef1
MS		 671 struct ext4_sb_info *sbi = EXT4_SB(sb);
672 struct ext4_super_block *es = sbi->s_es;
d9c9bef1
MS		 673
674 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
aa22df2c 675 def_errors = le16_to_cpu(es->s_errors);
d9c9bef1
MS		 676
677 if (sbi->s_sb_block != 1)
678 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
679 if (test_opt(sb, MINIX_DF))
680 seq_puts(seq, ",minixdf");
aa22df2c 681 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
d9c9bef1
MS		 682 seq_puts(seq, ",grpid");
683 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
684 seq_puts(seq, ",nogrpid");
685 if (sbi->s_resuid != EXT4_DEF_RESUID ||
686 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
687 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
688 }
689 if (sbi->s_resgid != EXT4_DEF_RESGID ||
690 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
691 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
692 }
bb4f397a 693 if (test_opt(sb, ERRORS_RO)) {
d9c9bef1 694 if (def_errors == EXT4_ERRORS_PANIC ||
bb4f397a
AK		 695 def_errors == EXT4_ERRORS_CONTINUE) {
696 seq_puts(seq, ",errors=remount-ro");
d9c9bef1
MS		 697 }
698 }
aa22df2c 699 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
bb4f397a 700 seq_puts(seq, ",errors=continue");
aa22df2c 701 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
d9c9bef1 702 seq_puts(seq, ",errors=panic");
aa22df2c 703 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
d9c9bef1 704 seq_puts(seq, ",nouid32");
aa22df2c 705 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
d9c9bef1
MS		 706 seq_puts(seq, ",debug");
707 if (test_opt(sb, OLDALLOC))
708 seq_puts(seq, ",oldalloc");
07620f69 709#ifdef CONFIG_EXT4DEV_FS_XATTR
aa22df2c
AK		 710 if (test_opt(sb, XATTR_USER) &&
711 !(def_mount_opts & EXT4_DEFM_XATTR_USER))
d9c9bef1
MS		 712 seq_puts(seq, ",user_xattr");
713 if (!test_opt(sb, XATTR_USER) &&
714 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
715 seq_puts(seq, ",nouser_xattr");
716 }
717#endif
07620f69 718#ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
aa22df2c 719 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
d9c9bef1
MS		 720 seq_puts(seq, ",acl");
721 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
722 seq_puts(seq, ",noacl");
723#endif
724 if (!test_opt(sb, RESERVATION))
725 seq_puts(seq, ",noreservation");
726 if (sbi->s_commit_interval) {
727 seq_printf(seq, ",commit=%u",
728 (unsigned) (sbi->s_commit_interval / HZ));
729 }
730 if (test_opt(sb, BARRIER))
731 seq_puts(seq, ",barrier=1");
732 if (test_opt(sb, NOBH))
733 seq_puts(seq, ",nobh");
734 if (!test_opt(sb, EXTENTS))
735 seq_puts(seq, ",noextents");
25ec56b5
JNC		 736 if (test_opt(sb, I_VERSION))
737 seq_puts(seq, ",i_version");
ac27a0ec 738
aa22df2c
AK		 739 /*
740 * journal mode get enabled in different ways
741 * So just print the value even if we didn't specify it
742 */
617ba13b 743 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
ac27a0ec 744 seq_puts(seq, ",data=journal");
617ba13b 745 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
ac27a0ec 746 seq_puts(seq, ",data=ordered");
617ba13b 747 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
ac27a0ec
DK		 748 seq_puts(seq, ",data=writeback");
749
617ba13b 750 ext4_show_quota_options(seq, sb);
ac27a0ec
DK		 751 return 0;
752}
753
754
1b961ac0
CH		 755static struct inode *ext4_nfs_get_inode(struct super_block *sb,
756 u64 ino, u32 generation)
ac27a0ec 757{
ac27a0ec 758 struct inode *inode;
ac27a0ec 759
617ba13b 760 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
ac27a0ec 761 return ERR_PTR(-ESTALE);
617ba13b 762 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
ac27a0ec
DK		 763 return ERR_PTR(-ESTALE);
764
765 /* iget isn't really right if the inode is currently unallocated!!
766 *
617ba13b 767 * ext4_read_inode will return a bad_inode if the inode had been
ac27a0ec
DK		 768 * deleted, so we should be safe.
769 *
770 * Currently we don't know the generation for parent directory, so
771 * a generation of 0 means "accept any"
772 */
773 inode = iget(sb, ino);
774 if (inode == NULL)
775 return ERR_PTR(-ENOMEM);
776 if (is_bad_inode(inode) ||
777 (generation && inode->i_generation != generation)) {
778 iput(inode);
779 return ERR_PTR(-ESTALE);
780 }
1b961ac0
CH		 781
782 return inode;
783}
784
785static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
786 int fh_len, int fh_type)
787{
788 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
789 ext4_nfs_get_inode);
790}
791
792static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
793 int fh_len, int fh_type)
794{
795 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
796 ext4_nfs_get_inode);
ac27a0ec
DK		 797}
798
799#ifdef CONFIG_QUOTA
800#define QTYPE2NAME(t) ((t)==USRQUOTA?"user":"group")
801#define QTYPE2MOPT(on, t) ((t)==USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
802
617ba13b
MC		 803static int ext4_dquot_initialize(struct inode *inode, int type);
804static int ext4_dquot_drop(struct inode *inode);
805static int ext4_write_dquot(struct dquot *dquot);
806static int ext4_acquire_dquot(struct dquot *dquot);
807static int ext4_release_dquot(struct dquot *dquot);
808static int ext4_mark_dquot_dirty(struct dquot *dquot);
809static int ext4_write_info(struct super_block *sb, int type);
810static int ext4_quota_on(struct super_block *sb, int type, int format_id, char *path);
811static int ext4_quota_on_mount(struct super_block *sb, int type);
812static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 813 size_t len, loff_t off);
617ba13b 814static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 815 const char *data, size_t len, loff_t off);
816
617ba13b
MC		 817static struct dquot_operations ext4_quota_operations = {
818 .initialize = ext4_dquot_initialize,
819 .drop = ext4_dquot_drop,
ac27a0ec
DK		 820 .alloc_space = dquot_alloc_space,
821 .alloc_inode = dquot_alloc_inode,
822 .free_space = dquot_free_space,
823 .free_inode = dquot_free_inode,
824 .transfer = dquot_transfer,
617ba13b
MC		 825 .write_dquot = ext4_write_dquot,
826 .acquire_dquot = ext4_acquire_dquot,
827 .release_dquot = ext4_release_dquot,
828 .mark_dirty = ext4_mark_dquot_dirty,
829 .write_info = ext4_write_info
ac27a0ec
DK		 830};
831
617ba13b
MC		 832static struct quotactl_ops ext4_qctl_operations = {
833 .quota_on = ext4_quota_on,
ac27a0ec
DK		 834 .quota_off = vfs_quota_off,
835 .quota_sync = vfs_quota_sync,
836 .get_info = vfs_get_dqinfo,
837 .set_info = vfs_set_dqinfo,
838 .get_dqblk = vfs_get_dqblk,
839 .set_dqblk = vfs_set_dqblk
840};
841#endif
842
ee9b6d61 843static const struct super_operations ext4_sops = {
617ba13b
MC		 844 .alloc_inode = ext4_alloc_inode,
845 .destroy_inode = ext4_destroy_inode,
846 .read_inode = ext4_read_inode,
847 .write_inode = ext4_write_inode,
848 .dirty_inode = ext4_dirty_inode,
849 .delete_inode = ext4_delete_inode,
850 .put_super = ext4_put_super,
851 .write_super = ext4_write_super,
852 .sync_fs = ext4_sync_fs,
853 .write_super_lockfs = ext4_write_super_lockfs,
854 .unlockfs = ext4_unlockfs,
855 .statfs = ext4_statfs,
856 .remount_fs = ext4_remount,
857 .clear_inode = ext4_clear_inode,
858 .show_options = ext4_show_options,
ac27a0ec 859#ifdef CONFIG_QUOTA
617ba13b
MC		 860 .quota_read = ext4_quota_read,
861 .quota_write = ext4_quota_write,
ac27a0ec
DK		 862#endif
863};
864
39655164 865static const struct export_operations ext4_export_ops = {
1b961ac0
CH		 866 .fh_to_dentry = ext4_fh_to_dentry,
867 .fh_to_parent = ext4_fh_to_parent,
617ba13b 868 .get_parent = ext4_get_parent,
ac27a0ec
DK		 869};
870
871enum {
872 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
873 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
874 Opt_nouid32, Opt_nocheck, Opt_debug, Opt_oldalloc, Opt_orlov,
875 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
876 Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh,
877 Opt_commit, Opt_journal_update, Opt_journal_inum, Opt_journal_dev,
818d276c 878 Opt_journal_checksum, Opt_journal_async_commit,
ac27a0ec
DK		 879 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
880 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
881 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
882 Opt_ignore, Opt_barrier, Opt_err, Opt_resize, Opt_usrquota,
25ec56b5 883 Opt_grpquota, Opt_extents, Opt_noextents, Opt_i_version,
ac27a0ec
DK		 884};
885
886static match_table_t tokens = {
887 {Opt_bsd_df, "bsddf"},
888 {Opt_minix_df, "minixdf"},
889 {Opt_grpid, "grpid"},
890 {Opt_grpid, "bsdgroups"},
891 {Opt_nogrpid, "nogrpid"},
892 {Opt_nogrpid, "sysvgroups"},
893 {Opt_resgid, "resgid=%u"},
894 {Opt_resuid, "resuid=%u"},
895 {Opt_sb, "sb=%u"},
896 {Opt_err_cont, "errors=continue"},
897 {Opt_err_panic, "errors=panic"},
898 {Opt_err_ro, "errors=remount-ro"},
899 {Opt_nouid32, "nouid32"},
900 {Opt_nocheck, "nocheck"},
901 {Opt_nocheck, "check=none"},
902 {Opt_debug, "debug"},
903 {Opt_oldalloc, "oldalloc"},
904 {Opt_orlov, "orlov"},
905 {Opt_user_xattr, "user_xattr"},
906 {Opt_nouser_xattr, "nouser_xattr"},
907 {Opt_acl, "acl"},
908 {Opt_noacl, "noacl"},
909 {Opt_reservation, "reservation"},
910 {Opt_noreservation, "noreservation"},
911 {Opt_noload, "noload"},
912 {Opt_nobh, "nobh"},
913 {Opt_bh, "bh"},
914 {Opt_commit, "commit=%u"},
915 {Opt_journal_update, "journal=update"},
916 {Opt_journal_inum, "journal=%u"},
917 {Opt_journal_dev, "journal_dev=%u"},
818d276c
GS		 918 {Opt_journal_checksum, "journal_checksum"},
919 {Opt_journal_async_commit, "journal_async_commit"},
ac27a0ec
DK		 920 {Opt_abort, "abort"},
921 {Opt_data_journal, "data=journal"},
922 {Opt_data_ordered, "data=ordered"},
923 {Opt_data_writeback, "data=writeback"},
924 {Opt_offusrjquota, "usrjquota="},
925 {Opt_usrjquota, "usrjquota=%s"},
926 {Opt_offgrpjquota, "grpjquota="},
927 {Opt_grpjquota, "grpjquota=%s"},
928 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
929 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
930 {Opt_grpquota, "grpquota"},
931 {Opt_noquota, "noquota"},
932 {Opt_quota, "quota"},
933 {Opt_usrquota, "usrquota"},
934 {Opt_barrier, "barrier=%u"},
a86c6181 935 {Opt_extents, "extents"},
1e2462f9 936 {Opt_noextents, "noextents"},
25ec56b5 937 {Opt_i_version, "i_version"},
ac27a0ec
DK		 938 {Opt_err, NULL},
939 {Opt_resize, "resize"},
940};
941
617ba13b 942static ext4_fsblk_t get_sb_block(void **data)
ac27a0ec 943{
617ba13b 944 ext4_fsblk_t sb_block;
ac27a0ec
DK		 945 char *options = (char *) *data;
946
947 if (!options || strncmp(options, "sb=", 3) != 0)
948 return 1; /* Default location */
949 options += 3;
617ba13b 950 /*todo: use simple_strtoll with >32bit ext4 */
ac27a0ec
DK		 951 sb_block = simple_strtoul(options, &options, 0);
952 if (*options && *options != ',') {
617ba13b 953 printk("EXT4-fs: Invalid sb specification: %s\n",
ac27a0ec
DK		 954 (char *) *data);
955 return 1;
956 }
957 if (*options == ',')
958 options++;
959 *data = (void *) options;
960 return sb_block;
961}
962
963static int parse_options (char *options, struct super_block *sb,
964 unsigned int *inum, unsigned long *journal_devnum,
617ba13b 965 ext4_fsblk_t *n_blocks_count, int is_remount)
ac27a0ec 966{
617ba13b 967 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 968 char * p;
969 substring_t args[MAX_OPT_ARGS];
970 int data_opt = 0;
971 int option;
972#ifdef CONFIG_QUOTA
973 int qtype;
974 char *qname;
975#endif
976
977 if (!options)
978 return 1;
979
980 while ((p = strsep (&options, ",")) != NULL) {
981 int token;
982 if (!*p)
983 continue;
984
985 token = match_token(p, tokens, args);
986 switch (token) {
987 case Opt_bsd_df:
988 clear_opt (sbi->s_mount_opt, MINIX_DF);
989 break;
990 case Opt_minix_df:
991 set_opt (sbi->s_mount_opt, MINIX_DF);
992 break;
993 case Opt_grpid:
994 set_opt (sbi->s_mount_opt, GRPID);
995 break;
996 case Opt_nogrpid:
997 clear_opt (sbi->s_mount_opt, GRPID);
998 break;
999 case Opt_resuid:
1000 if (match_int(&args[0], &option))
1001 return 0;
1002 sbi->s_resuid = option;
1003 break;
1004 case Opt_resgid:
1005 if (match_int(&args[0], &option))
1006 return 0;
1007 sbi->s_resgid = option;
1008 break;
1009 case Opt_sb:
1010 /* handled by get_sb_block() instead of here */
1011 /* *sb_block = match_int(&args[0]); */
1012 break;
1013 case Opt_err_panic:
1014 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
1015 clear_opt (sbi->s_mount_opt, ERRORS_RO);
1016 set_opt (sbi->s_mount_opt, ERRORS_PANIC);
1017 break;
1018 case Opt_err_ro:
1019 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
1020 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
1021 set_opt (sbi->s_mount_opt, ERRORS_RO);
1022 break;
1023 case Opt_err_cont:
1024 clear_opt (sbi->s_mount_opt, ERRORS_RO);
1025 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
1026 set_opt (sbi->s_mount_opt, ERRORS_CONT);
1027 break;
1028 case Opt_nouid32:
1029 set_opt (sbi->s_mount_opt, NO_UID32);
1030 break;
1031 case Opt_nocheck:
1032 clear_opt (sbi->s_mount_opt, CHECK);
1033 break;
1034 case Opt_debug:
1035 set_opt (sbi->s_mount_opt, DEBUG);
1036 break;
1037 case Opt_oldalloc:
1038 set_opt (sbi->s_mount_opt, OLDALLOC);
1039 break;
1040 case Opt_orlov:
1041 clear_opt (sbi->s_mount_opt, OLDALLOC);
1042 break;
617ba13b 1043#ifdef CONFIG_EXT4DEV_FS_XATTR
ac27a0ec
DK		 1044 case Opt_user_xattr:
1045 set_opt (sbi->s_mount_opt, XATTR_USER);
1046 break;
1047 case Opt_nouser_xattr:
1048 clear_opt (sbi->s_mount_opt, XATTR_USER);
1049 break;
1050#else
1051 case Opt_user_xattr:
1052 case Opt_nouser_xattr:
617ba13b 1053 printk("EXT4 (no)user_xattr options not supported\n");
ac27a0ec
DK		 1054 break;
1055#endif
617ba13b 1056#ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
ac27a0ec
DK		 1057 case Opt_acl:
1058 set_opt(sbi->s_mount_opt, POSIX_ACL);
1059 break;
1060 case Opt_noacl:
1061 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1062 break;
1063#else
1064 case Opt_acl:
1065 case Opt_noacl:
617ba13b 1066 printk("EXT4 (no)acl options not supported\n");
ac27a0ec
DK		 1067 break;
1068#endif
1069 case Opt_reservation:
1070 set_opt(sbi->s_mount_opt, RESERVATION);
1071 break;
1072 case Opt_noreservation:
1073 clear_opt(sbi->s_mount_opt, RESERVATION);
1074 break;
1075 case Opt_journal_update:
1076 /* @@@ FIXME */
1077 /* Eventually we will want to be able to create
1078 a journal file here. For now, only allow the
1079 user to specify an existing inode to be the
1080 journal file. */
1081 if (is_remount) {
617ba13b 1082 printk(KERN_ERR "EXT4-fs: cannot specify "
ac27a0ec
DK		 1083 "journal on remount\n");
1084 return 0;
1085 }
1086 set_opt (sbi->s_mount_opt, UPDATE_JOURNAL);
1087 break;
1088 case Opt_journal_inum:
1089 if (is_remount) {
617ba13b 1090 printk(KERN_ERR "EXT4-fs: cannot specify "
ac27a0ec
DK		 1091 "journal on remount\n");
1092 return 0;
1093 }
1094 if (match_int(&args[0], &option))
1095 return 0;
1096 *inum = option;
1097 break;
1098 case Opt_journal_dev:
1099 if (is_remount) {
617ba13b 1100 printk(KERN_ERR "EXT4-fs: cannot specify "
ac27a0ec
DK		 1101 "journal on remount\n");
1102 return 0;
1103 }
1104 if (match_int(&args[0], &option))
1105 return 0;
1106 *journal_devnum = option;
1107 break;
818d276c
GS		 1108 case Opt_journal_checksum:
1109 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1110 break;
1111 case Opt_journal_async_commit:
1112 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
1113 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1114 break;
ac27a0ec
DK		 1115 case Opt_noload:
1116 set_opt (sbi->s_mount_opt, NOLOAD);
1117 break;
1118 case Opt_commit:
1119 if (match_int(&args[0], &option))
1120 return 0;
1121 if (option < 0)
1122 return 0;
1123 if (option == 0)
cd02ff0b 1124 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
ac27a0ec
DK		 1125 sbi->s_commit_interval = HZ * option;
1126 break;
1127 case Opt_data_journal:
617ba13b 1128 data_opt = EXT4_MOUNT_JOURNAL_DATA;
ac27a0ec
DK		 1129 goto datacheck;
1130 case Opt_data_ordered:
617ba13b 1131 data_opt = EXT4_MOUNT_ORDERED_DATA;
ac27a0ec
DK		 1132 goto datacheck;
1133 case Opt_data_writeback:
617ba13b 1134 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
ac27a0ec
DK		 1135 datacheck:
1136 if (is_remount) {
617ba13b 1137 if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
ac27a0ec
DK		 1138 != data_opt) {
1139 printk(KERN_ERR
617ba13b 1140 "EXT4-fs: cannot change data "
ac27a0ec
DK		 1141 "mode on remount\n");
1142 return 0;
1143 }
1144 } else {
617ba13b 1145 sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
ac27a0ec
DK		 1146 sbi->s_mount_opt |= data_opt;
1147 }
1148 break;
1149#ifdef CONFIG_QUOTA
1150 case Opt_usrjquota:
1151 qtype = USRQUOTA;
1152 goto set_qf_name;
1153 case Opt_grpjquota:
1154 qtype = GRPQUOTA;
1155set_qf_name:
1156 if (sb_any_quota_enabled(sb)) {
1157 printk(KERN_ERR
617ba13b 1158 "EXT4-fs: Cannot change journalled "
ac27a0ec
DK		 1159 "quota options when quota turned on.\n");
1160 return 0;
1161 }
1162 qname = match_strdup(&args[0]);
1163 if (!qname) {
1164 printk(KERN_ERR
617ba13b 1165 "EXT4-fs: not enough memory for "
ac27a0ec
DK		 1166 "storing quotafile name.\n");
1167 return 0;
1168 }
1169 if (sbi->s_qf_names[qtype] &&
1170 strcmp(sbi->s_qf_names[qtype], qname)) {
1171 printk(KERN_ERR
617ba13b 1172 "EXT4-fs: %s quota file already "
ac27a0ec
DK		 1173 "specified.\n", QTYPE2NAME(qtype));
1174 kfree(qname);
1175 return 0;
1176 }
1177 sbi->s_qf_names[qtype] = qname;
1178 if (strchr(sbi->s_qf_names[qtype], '/')) {
1179 printk(KERN_ERR
617ba13b 1180 "EXT4-fs: quotafile must be on "
ac27a0ec
DK		 1181 "filesystem root.\n");
1182 kfree(sbi->s_qf_names[qtype]);
1183 sbi->s_qf_names[qtype] = NULL;
1184 return 0;
1185 }
1186 set_opt(sbi->s_mount_opt, QUOTA);
1187 break;
1188 case Opt_offusrjquota:
1189 qtype = USRQUOTA;
1190 goto clear_qf_name;
1191 case Opt_offgrpjquota:
1192 qtype = GRPQUOTA;
1193clear_qf_name:
1194 if (sb_any_quota_enabled(sb)) {
617ba13b 1195 printk(KERN_ERR "EXT4-fs: Cannot change "
ac27a0ec
DK		 1196 "journalled quota options when "
1197 "quota turned on.\n");
1198 return 0;
1199 }
1200 /*
1201 * The space will be released later when all options
1202 * are confirmed to be correct
1203 */
1204 sbi->s_qf_names[qtype] = NULL;
1205 break;
1206 case Opt_jqfmt_vfsold:
1207 sbi->s_jquota_fmt = QFMT_VFS_OLD;
1208 break;
1209 case Opt_jqfmt_vfsv0:
1210 sbi->s_jquota_fmt = QFMT_VFS_V0;
1211 break;
1212 case Opt_quota:
1213 case Opt_usrquota:
1214 set_opt(sbi->s_mount_opt, QUOTA);
1215 set_opt(sbi->s_mount_opt, USRQUOTA);
1216 break;
1217 case Opt_grpquota:
1218 set_opt(sbi->s_mount_opt, QUOTA);
1219 set_opt(sbi->s_mount_opt, GRPQUOTA);
1220 break;
1221 case Opt_noquota:
1222 if (sb_any_quota_enabled(sb)) {
617ba13b 1223 printk(KERN_ERR "EXT4-fs: Cannot change quota "
ac27a0ec
DK		 1224 "options when quota turned on.\n");
1225 return 0;
1226 }
1227 clear_opt(sbi->s_mount_opt, QUOTA);
1228 clear_opt(sbi->s_mount_opt, USRQUOTA);
1229 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1230 break;
1231#else
1232 case Opt_quota:
1233 case Opt_usrquota:
1234 case Opt_grpquota:
1235 case Opt_usrjquota:
1236 case Opt_grpjquota:
1237 case Opt_offusrjquota:
1238 case Opt_offgrpjquota:
1239 case Opt_jqfmt_vfsold:
1240 case Opt_jqfmt_vfsv0:
1241 printk(KERN_ERR
617ba13b 1242 "EXT4-fs: journalled quota options not "
ac27a0ec
DK		 1243 "supported.\n");
1244 break;
1245 case Opt_noquota:
1246 break;
1247#endif
1248 case Opt_abort:
1249 set_opt(sbi->s_mount_opt, ABORT);
1250 break;
1251 case Opt_barrier:
1252 if (match_int(&args[0], &option))
1253 return 0;
1254 if (option)
1255 set_opt(sbi->s_mount_opt, BARRIER);
1256 else
1257 clear_opt(sbi->s_mount_opt, BARRIER);
1258 break;
1259 case Opt_ignore:
1260 break;
1261 case Opt_resize:
1262 if (!is_remount) {
617ba13b 1263 printk("EXT4-fs: resize option only available "
ac27a0ec
DK		 1264 "for remount\n");
1265 return 0;
1266 }
1267 if (match_int(&args[0], &option) != 0)
1268 return 0;
1269 *n_blocks_count = option;
1270 break;
1271 case Opt_nobh:
1272 set_opt(sbi->s_mount_opt, NOBH);
1273 break;
1274 case Opt_bh:
1275 clear_opt(sbi->s_mount_opt, NOBH);
1276 break;
a86c6181
AT		 1277 case Opt_extents:
1278 set_opt (sbi->s_mount_opt, EXTENTS);
1279 break;
1e2462f9
MC		 1280 case Opt_noextents:
1281 clear_opt (sbi->s_mount_opt, EXTENTS);
1282 break;
25ec56b5
JNC		 1283 case Opt_i_version:
1284 set_opt(sbi->s_mount_opt, I_VERSION);
1285 sb->s_flags |= MS_I_VERSION;
1286 break;
ac27a0ec
DK		 1287 default:
1288 printk (KERN_ERR
617ba13b 1289 "EXT4-fs: Unrecognized mount option \"%s\" "
ac27a0ec
DK		 1290 "or missing value\n", p);
1291 return 0;
1292 }
1293 }
1294#ifdef CONFIG_QUOTA
1295 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
617ba13b 1296 if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
ac27a0ec
DK		 1297 sbi->s_qf_names[USRQUOTA])
1298 clear_opt(sbi->s_mount_opt, USRQUOTA);
1299
617ba13b 1300 if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
ac27a0ec
DK		 1301 sbi->s_qf_names[GRPQUOTA])
1302 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1303
1304 if ((sbi->s_qf_names[USRQUOTA] &&
617ba13b 1305 (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
ac27a0ec 1306 (sbi->s_qf_names[GRPQUOTA] &&
617ba13b
MC		 1307 (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
1308 printk(KERN_ERR "EXT4-fs: old and new quota "
ac27a0ec
DK		 1309 "format mixing.\n");
1310 return 0;
1311 }
1312
1313 if (!sbi->s_jquota_fmt) {
617ba13b 1314 printk(KERN_ERR "EXT4-fs: journalled quota format "
ac27a0ec
DK		 1315 "not specified.\n");
1316 return 0;
1317 }
1318 } else {
1319 if (sbi->s_jquota_fmt) {
617ba13b 1320 printk(KERN_ERR "EXT4-fs: journalled quota format "
ac27a0ec
DK		 1321 "specified with no journalling "
1322 "enabled.\n");
1323 return 0;
1324 }
1325 }
1326#endif
1327 return 1;
1328}
1329
617ba13b 1330static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 1331 int read_only)
1332{
617ba13b 1333 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 1334 int res = 0;
1335
617ba13b
MC		 1336 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1337 printk (KERN_ERR "EXT4-fs warning: revision level too high, "
ac27a0ec
DK		 1338 "forcing read-only mode\n");
1339 res = MS_RDONLY;
1340 }
1341 if (read_only)
1342 return res;
617ba13b
MC		 1343 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1344 printk (KERN_WARNING "EXT4-fs warning: mounting unchecked fs, "
ac27a0ec 1345 "running e2fsck is recommended\n");
617ba13b 1346 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
ac27a0ec 1347 printk (KERN_WARNING
617ba13b 1348 "EXT4-fs warning: mounting fs with errors, "
ac27a0ec
DK		 1349 "running e2fsck is recommended\n");
1350 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1351 le16_to_cpu(es->s_mnt_count) >=
1352 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1353 printk (KERN_WARNING
617ba13b 1354 "EXT4-fs warning: maximal mount count reached, "
ac27a0ec
DK		 1355 "running e2fsck is recommended\n");
1356 else if (le32_to_cpu(es->s_checkinterval) &&
1357 (le32_to_cpu(es->s_lastcheck) +
1358 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1359 printk (KERN_WARNING
617ba13b 1360 "EXT4-fs warning: checktime reached, "
ac27a0ec
DK		 1361 "running e2fsck is recommended\n");
1362#if 0
1363 /* @@@ We _will_ want to clear the valid bit if we find
63f57933
AM		 1364 * inconsistencies, to force a fsck at reboot. But for
1365 * a plain journaled filesystem we can keep it set as
1366 * valid forever! :)
1367 */
617ba13b 1368 es->s_state = cpu_to_le16(le16_to_cpu(es->s_state) & ~EXT4_VALID_FS);
ac27a0ec
DK		 1369#endif
1370 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 1371 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
ac27a0ec
DK		 1372 es->s_mnt_count=cpu_to_le16(le16_to_cpu(es->s_mnt_count) + 1);
1373 es->s_mtime = cpu_to_le32(get_seconds());
617ba13b
MC		 1374 ext4_update_dynamic_rev(sb);
1375 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 1376
617ba13b 1377 ext4_commit_super(sb, es, 1);
ac27a0ec 1378 if (test_opt(sb, DEBUG))
617ba13b 1379 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%lu, "
ac27a0ec
DK		 1380 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
1381 sb->s_blocksize,
1382 sbi->s_groups_count,
617ba13b
MC		 1383 EXT4_BLOCKS_PER_GROUP(sb),
1384 EXT4_INODES_PER_GROUP(sb),
ac27a0ec
DK		 1385 sbi->s_mount_opt);
1386
617ba13b
MC		 1387 printk(KERN_INFO "EXT4 FS on %s, ", sb->s_id);
1388 if (EXT4_SB(sb)->s_journal->j_inode == NULL) {
ac27a0ec
DK		 1389 char b[BDEVNAME_SIZE];
1390
1391 printk("external journal on %s\n",
617ba13b 1392 bdevname(EXT4_SB(sb)->s_journal->j_dev, b));
ac27a0ec
DK		 1393 } else {
1394 printk("internal journal\n");
1395 }
1396 return res;
1397}
1398
717d50e4
AD		 1399__le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1400 struct ext4_group_desc *gdp)
1401{
1402 __u16 crc = 0;
1403
1404 if (sbi->s_es->s_feature_ro_compat &
1405 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1406 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1407 __le32 le_group = cpu_to_le32(block_group);
1408
1409 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1410 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1411 crc = crc16(crc, (__u8 *)gdp, offset);
1412 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1413 /* for checksum of struct ext4_group_desc do the rest...*/
1414 if ((sbi->s_es->s_feature_incompat &
1415 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1416 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1417 crc = crc16(crc, (__u8 *)gdp + offset,
1418 le16_to_cpu(sbi->s_es->s_desc_size) -
1419 offset);
1420 }
1421
1422 return cpu_to_le16(crc);
1423}
1424
1425int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1426 struct ext4_group_desc *gdp)
1427{
1428 if ((sbi->s_es->s_feature_ro_compat &
1429 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1430 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1431 return 0;
1432
1433 return 1;
1434}
1435
ac27a0ec 1436/* Called at mount-time, super-block is locked */
617ba13b 1437static int ext4_check_descriptors (struct super_block * sb)
ac27a0ec 1438{
617ba13b
MC		 1439 struct ext4_sb_info *sbi = EXT4_SB(sb);
1440 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1441 ext4_fsblk_t last_block;
bd81d8ee
LV		 1442 ext4_fsblk_t block_bitmap;
1443 ext4_fsblk_t inode_bitmap;
1444 ext4_fsblk_t inode_table;
617ba13b 1445 struct ext4_group_desc * gdp = NULL;
ac27a0ec 1446 int desc_block = 0;
ce421581 1447 int flexbg_flag = 0;
fd2d4291 1448 ext4_group_t i;
ac27a0ec 1449
ce421581
JS		 1450 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1451 flexbg_flag = 1;
1452
617ba13b 1453 ext4_debug ("Checking group descriptors");
ac27a0ec
DK		 1454
1455 for (i = 0; i < sbi->s_groups_count; i++)
1456 {
ce421581 1457 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 1458 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 1459 else
1460 last_block = first_block +
617ba13b 1461 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 1462
617ba13b
MC		 1463 if ((i % EXT4_DESC_PER_BLOCK(sb)) == 0)
1464 gdp = (struct ext4_group_desc *)
ac27a0ec 1465 sbi->s_group_desc[desc_block++]->b_data;
8fadc143 1466 block_bitmap = ext4_block_bitmap(sb, gdp);
bd81d8ee 1467 if (block_bitmap < first_block || block_bitmap > last_block)
ac27a0ec 1468 {
617ba13b 1469 ext4_error (sb, "ext4_check_descriptors",
fd2d4291 1470 "Block bitmap for group %lu"
2ae02107 1471 " not in group (block %llu)!",
bd81d8ee 1472 i, block_bitmap);
ac27a0ec
DK		 1473 return 0;
1474 }
8fadc143 1475 inode_bitmap = ext4_inode_bitmap(sb, gdp);
bd81d8ee 1476 if (inode_bitmap < first_block || inode_bitmap > last_block)
ac27a0ec 1477 {
617ba13b 1478 ext4_error (sb, "ext4_check_descriptors",
fd2d4291 1479 "Inode bitmap for group %lu"
2ae02107 1480 " not in group (block %llu)!",
bd81d8ee 1481 i, inode_bitmap);
ac27a0ec
DK		 1482 return 0;
1483 }
8fadc143 1484 inode_table = ext4_inode_table(sb, gdp);
bd81d8ee 1485 if (inode_table < first_block ||
780dcdb2 1486 inode_table + sbi->s_itb_per_group - 1 > last_block)
ac27a0ec 1487 {
617ba13b 1488 ext4_error (sb, "ext4_check_descriptors",
fd2d4291 1489 "Inode table for group %lu"
2ae02107 1490 " not in group (block %llu)!",
bd81d8ee 1491 i, inode_table);
ac27a0ec
DK		 1492 return 0;
1493 }
717d50e4
AD		 1494 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
1495 ext4_error(sb, __FUNCTION__,
fd2d4291
AM		 1496 "Checksum for group %lu failed (%u!=%u)\n",
1497 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1498 gdp)), le16_to_cpu(gdp->bg_checksum));
717d50e4
AD		 1499 return 0;
1500 }
ce421581
JS		 1501 if (!flexbg_flag)
1502 first_block += EXT4_BLOCKS_PER_GROUP(sb);
0d1ee42f
AR		 1503 gdp = (struct ext4_group_desc *)
1504 ((__u8 *)gdp + EXT4_DESC_SIZE(sb));
ac27a0ec
DK		 1505 }
1506
bd81d8ee 1507 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
617ba13b 1508 sbi->s_es->s_free_inodes_count=cpu_to_le32(ext4_count_free_inodes(sb));
ac27a0ec
DK		 1509 return 1;
1510}
1511
617ba13b 1512/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 1513 * the superblock) which were deleted from all directories, but held open by
1514 * a process at the time of a crash. We walk the list and try to delete these
1515 * inodes at recovery time (only with a read-write filesystem).
1516 *
1517 * In order to keep the orphan inode chain consistent during traversal (in
1518 * case of crash during recovery), we link each inode into the superblock
1519 * orphan list_head and handle it the same way as an inode deletion during
1520 * normal operation (which journals the operations for us).
1521 *
1522 * We only do an iget() and an iput() on each inode, which is very safe if we
1523 * accidentally point at an in-use or already deleted inode. The worst that
1524 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 1525 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 1526 * e2fsck was run on this filesystem, and it must have already done the orphan
1527 * inode cleanup for us, so we can safely abort without any further action.
1528 */
617ba13b
MC		 1529static void ext4_orphan_cleanup (struct super_block * sb,
1530 struct ext4_super_block * es)
ac27a0ec
DK		 1531{
1532 unsigned int s_flags = sb->s_flags;
1533 int nr_orphans = 0, nr_truncates = 0;
1534#ifdef CONFIG_QUOTA
1535 int i;
1536#endif
1537 if (!es->s_last_orphan) {
1538 jbd_debug(4, "no orphan inodes to clean up\n");
1539 return;
1540 }
1541
a8f48a95
ES		 1542 if (bdev_read_only(sb->s_bdev)) {
1543 printk(KERN_ERR "EXT4-fs: write access "
1544 "unavailable, skipping orphan cleanup.\n");
1545 return;
1546 }
1547
617ba13b 1548 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
ac27a0ec
DK		 1549 if (es->s_last_orphan)
1550 jbd_debug(1, "Errors on filesystem, "
1551 "clearing orphan list.\n");
1552 es->s_last_orphan = 0;
1553 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1554 return;
1555 }
1556
1557 if (s_flags & MS_RDONLY) {
617ba13b 1558 printk(KERN_INFO "EXT4-fs: %s: orphan cleanup on readonly fs\n",
ac27a0ec
DK		 1559 sb->s_id);
1560 sb->s_flags &= ~MS_RDONLY;
1561 }
1562#ifdef CONFIG_QUOTA
1563 /* Needed for iput() to work correctly and not trash data */
1564 sb->s_flags |= MS_ACTIVE;
1565 /* Turn on quotas so that they are updated correctly */
1566 for (i = 0; i < MAXQUOTAS; i++) {
617ba13b
MC		 1567 if (EXT4_SB(sb)->s_qf_names[i]) {
1568 int ret = ext4_quota_on_mount(sb, i);
ac27a0ec
DK		 1569 if (ret < 0)
1570 printk(KERN_ERR
617ba13b 1571 "EXT4-fs: Cannot turn on journalled "
ac27a0ec
DK		 1572 "quota: error %d\n", ret);
1573 }
1574 }
1575#endif
1576
1577 while (es->s_last_orphan) {
1578 struct inode *inode;
1579
1580 if (!(inode =
617ba13b 1581 ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan)))) {
ac27a0ec
DK		 1582 es->s_last_orphan = 0;
1583 break;
1584 }
1585
617ba13b 1586 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
ac27a0ec
DK		 1587 DQUOT_INIT(inode);
1588 if (inode->i_nlink) {
1589 printk(KERN_DEBUG
1590 "%s: truncating inode %lu to %Ld bytes\n",
1591 __FUNCTION__, inode->i_ino, inode->i_size);
1592 jbd_debug(2, "truncating inode %lu to %Ld bytes\n",
1593 inode->i_ino, inode->i_size);
617ba13b 1594 ext4_truncate(inode);
ac27a0ec
DK		 1595 nr_truncates++;
1596 } else {
1597 printk(KERN_DEBUG
1598 "%s: deleting unreferenced inode %lu\n",
1599 __FUNCTION__, inode->i_ino);
1600 jbd_debug(2, "deleting unreferenced inode %lu\n",
1601 inode->i_ino);
1602 nr_orphans++;
1603 }
1604 iput(inode); /* The delete magic happens here! */
1605 }
1606
1607#define PLURAL(x) (x), ((x)==1) ? "" : "s"
1608
1609 if (nr_orphans)
617ba13b 1610 printk(KERN_INFO "EXT4-fs: %s: %d orphan inode%s deleted\n",
ac27a0ec
DK		 1611 sb->s_id, PLURAL(nr_orphans));
1612 if (nr_truncates)
617ba13b 1613 printk(KERN_INFO "EXT4-fs: %s: %d truncate%s cleaned up\n",
ac27a0ec
DK		 1614 sb->s_id, PLURAL(nr_truncates));
1615#ifdef CONFIG_QUOTA
1616 /* Turn quotas off */
1617 for (i = 0; i < MAXQUOTAS; i++) {
1618 if (sb_dqopt(sb)->files[i])
1619 vfs_quota_off(sb, i);
1620 }
1621#endif
1622 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1623}
cd2291a4
ES		 1624/*
1625 * Maximal extent format file size.
1626 * Resulting logical blkno at s_maxbytes must fit in our on-disk
1627 * extent format containers, within a sector_t, and within i_blocks
1628 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
1629 * so that won't be a limiting factor.
1630 *
1631 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1632 */
1633static loff_t ext4_max_size(int blkbits)
1634{
1635 loff_t res;
1636 loff_t upper_limit = MAX_LFS_FILESIZE;
1637
1638 /* small i_blocks in vfs inode? */
1639 if (sizeof(blkcnt_t) < sizeof(u64)) {
1640 /*
1641 * CONFIG_LSF is not enabled implies the inode
1642 * i_block represent total blocks in 512 bytes
1643 * 32 == size of vfs inode i_blocks * 8
1644 */
1645 upper_limit = (1LL << 32) - 1;
1646
1647 /* total blocks in file system block size */
1648 upper_limit >>= (blkbits - 9);
1649 upper_limit <<= blkbits;
1650 }
1651
1652 /* 32-bit extent-start container, ee_block */
1653 res = 1LL << 32;
1654 res <<= blkbits;
1655 res -= 1;
1656
1657 /* Sanity check against vm- & vfs- imposed limits */
1658 if (res > upper_limit)
1659 res = upper_limit;
1660
1661 return res;
1662}
ac27a0ec 1663
ac27a0ec 1664/*
cd2291a4 1665 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 1666 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
1667 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 1668 */
cd2291a4 1669static loff_t ext4_max_bitmap_size(int bits)
ac27a0ec 1670{
617ba13b 1671 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 1672 int meta_blocks;
1673 loff_t upper_limit;
1674 /* This is calculated to be the largest file size for a
cd2291a4 1675 * dense, bitmapped file such that the total number of
ac27a0ec 1676 * sectors in the file, including data and all indirect blocks,
0fc1b451
AK		 1677 * does not exceed 2^48 -1
1678 * __u32 i_blocks_lo and _u16 i_blocks_high representing the
1679 * total number of 512 bytes blocks of the file
1680 */
1681
1682 if (sizeof(blkcnt_t) < sizeof(u64)) {
1683 /*
1684 * CONFIG_LSF is not enabled implies the inode
1685 * i_block represent total blocks in 512 bytes
1686 * 32 == size of vfs inode i_blocks * 8
1687 */
1688 upper_limit = (1LL << 32) - 1;
1689
1690 /* total blocks in file system block size */
1691 upper_limit >>= (bits - 9);
1692
1693 } else {
8180a562
AK		 1694 /*
1695 * We use 48 bit ext4_inode i_blocks
1696 * With EXT4_HUGE_FILE_FL set the i_blocks
1697 * represent total number of blocks in
1698 * file system block size
1699 */
0fc1b451
AK		 1700 upper_limit = (1LL << 48) - 1;
1701
0fc1b451
AK		 1702 }
1703
1704 /* indirect blocks */
1705 meta_blocks = 1;
1706 /* double indirect blocks */
1707 meta_blocks += 1 + (1LL << (bits-2));
1708 /* tripple indirect blocks */
1709 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
1710
1711 upper_limit -= meta_blocks;
1712 upper_limit <<= bits;
ac27a0ec
DK		 1713
1714 res += 1LL << (bits-2);
1715 res += 1LL << (2*(bits-2));
1716 res += 1LL << (3*(bits-2));
1717 res <<= bits;
1718 if (res > upper_limit)
1719 res = upper_limit;
0fc1b451
AK		 1720
1721 if (res > MAX_LFS_FILESIZE)
1722 res = MAX_LFS_FILESIZE;
1723
ac27a0ec
DK		 1724 return res;
1725}
1726
617ba13b 1727static ext4_fsblk_t descriptor_loc(struct super_block *sb,
70bbb3e0 1728 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 1729{
617ba13b 1730 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 1731 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 1732 int has_super = 0;
1733
1734 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1735
617ba13b 1736 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
ac27a0ec 1737 nr < first_meta_bg)
70bbb3e0 1738 return logical_sb_block + nr + 1;
ac27a0ec 1739 bg = sbi->s_desc_per_block * nr;
617ba13b 1740 if (ext4_bg_has_super(sb, bg))
ac27a0ec 1741 has_super = 1;
617ba13b 1742 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 1743}
1744
1745
617ba13b 1746static int ext4_fill_super (struct super_block *sb, void *data, int silent)
1d03ec98
AK		 1747 __releases(kernel_sem)
1748 __acquires(kernel_sem)
1749
ac27a0ec
DK		 1750{
1751 struct buffer_head * bh;
617ba13b
MC		 1752 struct ext4_super_block *es = NULL;
1753 struct ext4_sb_info *sbi;
1754 ext4_fsblk_t block;
1755 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 1756 ext4_fsblk_t logical_sb_block;
ac27a0ec
DK		 1757 unsigned long offset = 0;
1758 unsigned int journal_inum = 0;
1759 unsigned long journal_devnum = 0;
1760 unsigned long def_mount_opts;
1761 struct inode *root;
1762 int blocksize;
1763 int hblock;
1764 int db_count;
1765 int i;
1766 int needs_recovery;
1767 __le32 features;
bd81d8ee 1768 __u64 blocks_count;
833f4077 1769 int err;
ac27a0ec
DK		 1770
1771 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
1772 if (!sbi)
1773 return -ENOMEM;
1774 sb->s_fs_info = sbi;
1775 sbi->s_mount_opt = 0;
617ba13b
MC		 1776 sbi->s_resuid = EXT4_DEF_RESUID;
1777 sbi->s_resgid = EXT4_DEF_RESGID;
d9c9bef1 1778 sbi->s_sb_block = sb_block;
ac27a0ec
DK		 1779
1780 unlock_kernel();
1781
617ba13b 1782 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 1783 if (!blocksize) {
617ba13b 1784 printk(KERN_ERR "EXT4-fs: unable to set blocksize\n");
ac27a0ec
DK		 1785 goto out_fail;
1786 }
1787
afc7cbca
TS		 1788 if (!sb_set_blocksize(sb, blocksize)) {
1789 printk(KERN_ERR "EXT4-fs: bad blocksize %d.\n", blocksize);
1790 goto out_fail;
1791 }
1792
ac27a0ec 1793 /*
617ba13b 1794 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 1795 * block sizes. We need to calculate the offset from buffer start.
1796 */
617ba13b 1797 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 1798 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
1799 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 1800 } else {
70bbb3e0 1801 logical_sb_block = sb_block;
ac27a0ec
DK		 1802 }
1803
70bbb3e0 1804 if (!(bh = sb_bread(sb, logical_sb_block))) {
617ba13b 1805 printk (KERN_ERR "EXT4-fs: unable to read superblock\n");
ac27a0ec
DK		 1806 goto out_fail;
1807 }
1808 /*
1809 * Note: s_es must be initialized as soon as possible because
617ba13b 1810 * some ext4 macro-instructions depend on its value
ac27a0ec 1811 */
617ba13b 1812 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
ac27a0ec
DK		 1813 sbi->s_es = es;
1814 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 1815 if (sb->s_magic != EXT4_SUPER_MAGIC)
1816 goto cantfind_ext4;
ac27a0ec
DK		 1817
1818 /* Set defaults before we parse the mount options */
1819 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
617ba13b 1820 if (def_mount_opts & EXT4_DEFM_DEBUG)
ac27a0ec 1821 set_opt(sbi->s_mount_opt, DEBUG);
617ba13b 1822 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
ac27a0ec 1823 set_opt(sbi->s_mount_opt, GRPID);
617ba13b 1824 if (def_mount_opts & EXT4_DEFM_UID16)
ac27a0ec 1825 set_opt(sbi->s_mount_opt, NO_UID32);
2e7842b8 1826#ifdef CONFIG_EXT4DEV_FS_XATTR
617ba13b 1827 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
ac27a0ec 1828 set_opt(sbi->s_mount_opt, XATTR_USER);
2e7842b8
HD		 1829#endif
1830#ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
617ba13b 1831 if (def_mount_opts & EXT4_DEFM_ACL)
ac27a0ec 1832 set_opt(sbi->s_mount_opt, POSIX_ACL);
2e7842b8 1833#endif
617ba13b
MC		 1834 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
1835 sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
1836 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
1837 sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
1838 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
1839 sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
1840
1841 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
ac27a0ec 1842 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
bb4f397a 1843 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
ceea16bf 1844 set_opt(sbi->s_mount_opt, ERRORS_CONT);
bb4f397a
AK		 1845 else
1846 set_opt(sbi->s_mount_opt, ERRORS_RO);
ac27a0ec
DK		 1847
1848 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
1849 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
1850
1851 set_opt(sbi->s_mount_opt, RESERVATION);
1852
1e2462f9
MC		 1853 /*
1854 * turn on extents feature by default in ext4 filesystem
1855 * User -o noextents to turn it off
1856 */
1857 set_opt(sbi->s_mount_opt, EXTENTS);
1858
ac27a0ec
DK		 1859 if (!parse_options ((char *) data, sb, &journal_inum, &journal_devnum,
1860 NULL, 0))
1861 goto failed_mount;
1862
1863 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
617ba13b 1864 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec 1865
617ba13b
MC		 1866 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
1867 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
1868 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
1869 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
ac27a0ec 1870 printk(KERN_WARNING
617ba13b 1871 "EXT4-fs warning: feature flags set on rev 0 fs, "
ac27a0ec
DK		 1872 "running e2fsck is recommended\n");
1873 /*
1874 * Check feature flags regardless of the revision level, since we
1875 * previously didn't change the revision level when setting the flags,
1876 * so there is a chance incompat flags are set on a rev 0 filesystem.
1877 */
617ba13b 1878 features = EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP);
ac27a0ec 1879 if (features) {
617ba13b 1880 printk(KERN_ERR "EXT4-fs: %s: couldn't mount because of "
ac27a0ec
DK		 1881 "unsupported optional features (%x).\n",
1882 sb->s_id, le32_to_cpu(features));
1883 goto failed_mount;
1884 }
617ba13b 1885 features = EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP);
ac27a0ec 1886 if (!(sb->s_flags & MS_RDONLY) && features) {
617ba13b 1887 printk(KERN_ERR "EXT4-fs: %s: couldn't mount RDWR because of "
ac27a0ec
DK		 1888 "unsupported optional features (%x).\n",
1889 sb->s_id, le32_to_cpu(features));
1890 goto failed_mount;
1891 }
0fc1b451
AK		 1892 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
1893 /*
1894 * Large file size enabled file system can only be
1895 * mount if kernel is build with CONFIG_LSF
1896 */
1897 if (sizeof(root->i_blocks) < sizeof(u64) &&
1898 !(sb->s_flags & MS_RDONLY)) {
1899 printk(KERN_ERR "EXT4-fs: %s: Filesystem with huge "
1900 "files cannot be mounted read-write "
1901 "without CONFIG_LSF.\n", sb->s_id);
1902 goto failed_mount;
1903 }
1904 }
ac27a0ec
DK		 1905 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
1906
617ba13b
MC		 1907 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
1908 blocksize > EXT4_MAX_BLOCK_SIZE) {
ac27a0ec 1909 printk(KERN_ERR
617ba13b 1910 "EXT4-fs: Unsupported filesystem blocksize %d on %s.\n",
ac27a0ec
DK		 1911 blocksize, sb->s_id);
1912 goto failed_mount;
1913 }
1914
1915 hblock = bdev_hardsect_size(sb->s_bdev);
1916 if (sb->s_blocksize != blocksize) {
1917 /*
1918 * Make sure the blocksize for the filesystem is larger
1919 * than the hardware sectorsize for the machine.
1920 */
1921 if (blocksize < hblock) {
617ba13b 1922 printk(KERN_ERR "EXT4-fs: blocksize %d too small for "
ac27a0ec
DK		 1923 "device blocksize %d.\n", blocksize, hblock);
1924 goto failed_mount;
1925 }
1926
1927 brelse (bh);
1928 sb_set_blocksize(sb, blocksize);
70bbb3e0
AM		 1929 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
1930 offset = do_div(logical_sb_block, blocksize);
1931 bh = sb_bread(sb, logical_sb_block);
ac27a0ec
DK		 1932 if (!bh) {
1933 printk(KERN_ERR
617ba13b 1934 "EXT4-fs: Can't read superblock on 2nd try.\n");
ac27a0ec
DK		 1935 goto failed_mount;
1936 }
617ba13b 1937 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
ac27a0ec 1938 sbi->s_es = es;
617ba13b 1939 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
ac27a0ec 1940 printk (KERN_ERR
617ba13b 1941 "EXT4-fs: Magic mismatch, very weird !\n");
ac27a0ec
DK		 1942 goto failed_mount;
1943 }
1944 }
1945
e2b46574 1946 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits);
617ba13b 1947 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits);
ac27a0ec 1948
617ba13b
MC		 1949 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
1950 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
1951 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
ac27a0ec
DK		 1952 } else {
1953 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
1954 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
617ba13b 1955 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1330593e 1956 (!is_power_of_2(sbi->s_inode_size)) ||
ac27a0ec
DK		 1957 (sbi->s_inode_size > blocksize)) {
1958 printk (KERN_ERR
617ba13b 1959 "EXT4-fs: unsupported inode size: %d\n",
ac27a0ec
DK		 1960 sbi->s_inode_size);
1961 goto failed_mount;
1962 }
ef7f3835
KS		 1963 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
1964 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
ac27a0ec 1965 }
0d1ee42f
AR		 1966 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
1967 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
8fadc143 1968 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 1969 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 1970 !is_power_of_2(sbi->s_desc_size)) {
0d1ee42f 1971 printk(KERN_ERR
8fadc143 1972 "EXT4-fs: unsupported descriptor size %lu\n",
0d1ee42f
AR		 1973 sbi->s_desc_size);
1974 goto failed_mount;
1975 }
1976 } else
1977 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
ac27a0ec 1978 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 1979 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
b47b6f38 1980 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
617ba13b
MC		 1981 goto cantfind_ext4;
1982 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 1983 if (sbi->s_inodes_per_block == 0)
617ba13b 1984 goto cantfind_ext4;
ac27a0ec
DK		 1985 sbi->s_itb_per_group = sbi->s_inodes_per_group /
1986 sbi->s_inodes_per_block;
0d1ee42f 1987 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 1988 sbi->s_sbh = bh;
1989 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 1990 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
1991 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
ac27a0ec
DK		 1992 for (i=0; i < 4; i++)
1993 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
1994 sbi->s_def_hash_version = es->s_def_hash_version;
1995
1996 if (sbi->s_blocks_per_group > blocksize * 8) {
1997 printk (KERN_ERR
617ba13b 1998 "EXT4-fs: #blocks per group too big: %lu\n",
ac27a0ec
DK		 1999 sbi->s_blocks_per_group);
2000 goto failed_mount;
2001 }
ac27a0ec
DK		 2002 if (sbi->s_inodes_per_group > blocksize * 8) {
2003 printk (KERN_ERR
617ba13b 2004 "EXT4-fs: #inodes per group too big: %lu\n",
ac27a0ec
DK		 2005 sbi->s_inodes_per_group);
2006 goto failed_mount;
2007 }
2008
bd81d8ee 2009 if (ext4_blocks_count(es) >
ac27a0ec 2010 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
617ba13b 2011 printk(KERN_ERR "EXT4-fs: filesystem on %s:"
ac27a0ec
DK		 2012 " too large to mount safely\n", sb->s_id);
2013 if (sizeof(sector_t) < 8)
617ba13b 2014 printk(KERN_WARNING "EXT4-fs: CONFIG_LBD not "
ac27a0ec
DK		 2015 "enabled\n");
2016 goto failed_mount;
2017 }
2018
617ba13b
MC		 2019 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2020 goto cantfind_ext4;
e7c95593
ES		 2021
2022 /* ensure blocks_count calculation below doesn't sign-extend */
2023 if (ext4_blocks_count(es) + EXT4_BLOCKS_PER_GROUP(sb) <
2024 le32_to_cpu(es->s_first_data_block) + 1) {
2025 printk(KERN_WARNING "EXT4-fs: bad geometry: block count %llu, "
2026 "first data block %u, blocks per group %lu\n",
2027 ext4_blocks_count(es),
2028 le32_to_cpu(es->s_first_data_block),
2029 EXT4_BLOCKS_PER_GROUP(sb));
2030 goto failed_mount;
2031 }
bd81d8ee
LV		 2032 blocks_count = (ext4_blocks_count(es) -
2033 le32_to_cpu(es->s_first_data_block) +
2034 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2035 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
2036 sbi->s_groups_count = blocks_count;
617ba13b
MC		 2037 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2038 EXT4_DESC_PER_BLOCK(sb);
ac27a0ec
DK		 2039 sbi->s_group_desc = kmalloc(db_count * sizeof (struct buffer_head *),
2040 GFP_KERNEL);
2041 if (sbi->s_group_desc == NULL) {
617ba13b 2042 printk (KERN_ERR "EXT4-fs: not enough memory\n");
ac27a0ec
DK		 2043 goto failed_mount;
2044 }
2045
2046 bgl_lock_init(&sbi->s_blockgroup_lock);
2047
2048 for (i = 0; i < db_count; i++) {
70bbb3e0 2049 block = descriptor_loc(sb, logical_sb_block, i);
ac27a0ec
DK		 2050 sbi->s_group_desc[i] = sb_bread(sb, block);
2051 if (!sbi->s_group_desc[i]) {
617ba13b 2052 printk (KERN_ERR "EXT4-fs: "
ac27a0ec
DK		 2053 "can't read group descriptor %d\n", i);
2054 db_count = i;
2055 goto failed_mount2;
2056 }
2057 }
617ba13b
MC		 2058 if (!ext4_check_descriptors (sb)) {
2059 printk(KERN_ERR "EXT4-fs: group descriptors corrupted!\n");
ac27a0ec
DK		 2060 goto failed_mount2;
2061 }
2062 sbi->s_gdb_count = db_count;
2063 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2064 spin_lock_init(&sbi->s_next_gen_lock);
2065
833f4077
PZ		 2066 err = percpu_counter_init(&sbi->s_freeblocks_counter,
2067 ext4_count_free_blocks(sb));
2068 if (!err) {
2069 err = percpu_counter_init(&sbi->s_freeinodes_counter,
2070 ext4_count_free_inodes(sb));
2071 }
2072 if (!err) {
2073 err = percpu_counter_init(&sbi->s_dirs_counter,
2074 ext4_count_dirs(sb));
2075 }
2076 if (err) {
2077 printk(KERN_ERR "EXT4-fs: insufficient memory\n");
2078 goto failed_mount3;
2079 }
ac27a0ec
DK		 2080
2081 /* per fileystem reservation list head & lock */
2082 spin_lock_init(&sbi->s_rsv_window_lock);
2083 sbi->s_rsv_window_root = RB_ROOT;
2084 /* Add a single, static dummy reservation to the start of the
2085 * reservation window list --- it gives us a placeholder for
2086 * append-at-start-of-list which makes the allocation logic
2087 * _much_ simpler. */
617ba13b
MC		 2088 sbi->s_rsv_window_head.rsv_start = EXT4_RESERVE_WINDOW_NOT_ALLOCATED;
2089 sbi->s_rsv_window_head.rsv_end = EXT4_RESERVE_WINDOW_NOT_ALLOCATED;
ac27a0ec
DK		 2090 sbi->s_rsv_window_head.rsv_alloc_hit = 0;
2091 sbi->s_rsv_window_head.rsv_goal_size = 0;
617ba13b 2092 ext4_rsv_window_add(sb, &sbi->s_rsv_window_head);
ac27a0ec
DK		 2093
2094 /*
2095 * set up enough so that it can read an inode
2096 */
617ba13b
MC		 2097 sb->s_op = &ext4_sops;
2098 sb->s_export_op = &ext4_export_ops;
2099 sb->s_xattr = ext4_xattr_handlers;
ac27a0ec 2100#ifdef CONFIG_QUOTA
617ba13b
MC		 2101 sb->s_qcop = &ext4_qctl_operations;
2102 sb->dq_op = &ext4_quota_operations;
ac27a0ec
DK		 2103#endif
2104 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
2105
2106 sb->s_root = NULL;
2107
2108 needs_recovery = (es->s_last_orphan != 0 ||
617ba13b
MC		 2109 EXT4_HAS_INCOMPAT_FEATURE(sb,
2110 EXT4_FEATURE_INCOMPAT_RECOVER));
ac27a0ec
DK		 2111
2112 /*
2113 * The first inode we look at is the journal inode. Don't try
2114 * root first: it may be modified in the journal!
2115 */
2116 if (!test_opt(sb, NOLOAD) &&
617ba13b
MC		 2117 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2118 if (ext4_load_journal(sb, es, journal_devnum))
ac27a0ec
DK		 2119 goto failed_mount3;
2120 } else if (journal_inum) {
617ba13b 2121 if (ext4_create_journal(sb, es, journal_inum))
ac27a0ec
DK		 2122 goto failed_mount3;
2123 } else {
2124 if (!silent)
2125 printk (KERN_ERR
617ba13b 2126 "ext4: No journal on filesystem on %s\n",
ac27a0ec
DK		 2127 sb->s_id);
2128 goto failed_mount3;
2129 }
2130
eb40a09c
JS		 2131 if (ext4_blocks_count(es) > 0xffffffffULL &&
2132 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2133 JBD2_FEATURE_INCOMPAT_64BIT)) {
2134 printk(KERN_ERR "ext4: Failed to set 64-bit journal feature\n");
2135 goto failed_mount4;
2136 }
2137
818d276c
GS		 2138 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2139 jbd2_journal_set_features(sbi->s_journal,
2140 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2141 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2142 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2143 jbd2_journal_set_features(sbi->s_journal,
2144 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2145 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2146 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2147 } else {
2148 jbd2_journal_clear_features(sbi->s_journal,
2149 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2150 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2151 }
2152
ac27a0ec
DK		 2153 /* We have now updated the journal if required, so we can
2154 * validate the data journaling mode. */
2155 switch (test_opt(sb, DATA_FLAGS)) {
2156 case 0:
2157 /* No mode set, assume a default based on the journal
63f57933
AM		 2158 * capabilities: ORDERED_DATA if the journal can
2159 * cope, else JOURNAL_DATA
2160 */
dab291af
MC		 2161 if (jbd2_journal_check_available_features
2162 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
ac27a0ec
DK		 2163 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2164 else
2165 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2166 break;
2167
617ba13b
MC		 2168 case EXT4_MOUNT_ORDERED_DATA:
2169 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 2170 if (!jbd2_journal_check_available_features
2171 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
617ba13b 2172 printk(KERN_ERR "EXT4-fs: Journal does not support "
ac27a0ec
DK		 2173 "requested data journaling mode\n");
2174 goto failed_mount4;
2175 }
2176 default:
2177 break;
2178 }
2179
2180 if (test_opt(sb, NOBH)) {
617ba13b
MC		 2181 if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
2182 printk(KERN_WARNING "EXT4-fs: Ignoring nobh option - "
ac27a0ec
DK		 2183 "its supported only with writeback mode\n");
2184 clear_opt(sbi->s_mount_opt, NOBH);
2185 }
2186 }
2187 /*
dab291af 2188 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 2189 * so we can safely mount the rest of the filesystem now.
2190 */
2191
617ba13b 2192 root = iget(sb, EXT4_ROOT_INO);
ac27a0ec
DK		 2193 sb->s_root = d_alloc_root(root);
2194 if (!sb->s_root) {
617ba13b 2195 printk(KERN_ERR "EXT4-fs: get root inode failed\n");
ac27a0ec
DK		 2196 iput(root);
2197 goto failed_mount4;
2198 }
2199 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
2200 dput(sb->s_root);
2201 sb->s_root = NULL;
617ba13b 2202 printk(KERN_ERR "EXT4-fs: corrupt root inode, run e2fsck\n");
ac27a0ec
DK		 2203 goto failed_mount4;
2204 }
2205
617ba13b 2206 ext4_setup_super (sb, es, sb->s_flags & MS_RDONLY);
ef7f3835
KS		 2207
2208 /* determine the minimum size of new large inodes, if present */
2209 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2210 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2211 EXT4_GOOD_OLD_INODE_SIZE;
2212 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2213 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2214 if (sbi->s_want_extra_isize <
2215 le16_to_cpu(es->s_want_extra_isize))
2216 sbi->s_want_extra_isize =
2217 le16_to_cpu(es->s_want_extra_isize);
2218 if (sbi->s_want_extra_isize <
2219 le16_to_cpu(es->s_min_extra_isize))
2220 sbi->s_want_extra_isize =
2221 le16_to_cpu(es->s_min_extra_isize);
2222 }
2223 }
2224 /* Check if enough inode space is available */
2225 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2226 sbi->s_inode_size) {
2227 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2228 EXT4_GOOD_OLD_INODE_SIZE;
2229 printk(KERN_INFO "EXT4-fs: required extra inode space not"
2230 "available.\n");
2231 }
2232
ac27a0ec
DK		 2233 /*
2234 * akpm: core read_super() calls in here with the superblock locked.
2235 * That deadlocks, because orphan cleanup needs to lock the superblock
2236 * in numerous places. Here we just pop the lock - it's relatively
2237 * harmless, because we are now ready to accept write_super() requests,
2238 * and aviro says that's the only reason for hanging onto the
2239 * superblock lock.
2240 */
617ba13b
MC		 2241 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
2242 ext4_orphan_cleanup(sb, es);
2243 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
ac27a0ec 2244 if (needs_recovery)
617ba13b
MC		 2245 printk (KERN_INFO "EXT4-fs: recovery complete.\n");
2246 ext4_mark_recovery_complete(sb, es);
2247 printk (KERN_INFO "EXT4-fs: mounted filesystem with %s data mode.\n",
2248 test_opt(sb,DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA ? "journal":
2249 test_opt(sb,DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA ? "ordered":
ac27a0ec
DK		 2250 "writeback");
2251
a86c6181
AT		 2252 ext4_ext_init(sb);
2253
ac27a0ec
DK		 2254 lock_kernel();
2255 return 0;
2256
617ba13b 2257cantfind_ext4:
ac27a0ec 2258 if (!silent)
617ba13b 2259 printk(KERN_ERR "VFS: Can't find ext4 filesystem on dev %s.\n",
ac27a0ec
DK		 2260 sb->s_id);
2261 goto failed_mount;
2262
2263failed_mount4:
dab291af 2264 jbd2_journal_destroy(sbi->s_journal);
ac27a0ec
DK		 2265failed_mount3:
2266 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2267 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2268 percpu_counter_destroy(&sbi->s_dirs_counter);
2269failed_mount2:
2270 for (i = 0; i < db_count; i++)
2271 brelse(sbi->s_group_desc[i]);
2272 kfree(sbi->s_group_desc);
2273failed_mount:
2274#ifdef CONFIG_QUOTA
2275 for (i = 0; i < MAXQUOTAS; i++)
2276 kfree(sbi->s_qf_names[i]);
2277#endif
617ba13b 2278 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 2279 brelse(bh);
2280out_fail:
2281 sb->s_fs_info = NULL;
2282 kfree(sbi);
2283 lock_kernel();
2284 return -EINVAL;
2285}
2286
2287/*
2288 * Setup any per-fs journal parameters now. We'll do this both on
2289 * initial mount, once the journal has been initialised but before we've
2290 * done any recovery; and again on any subsequent remount.
2291 */
617ba13b 2292static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 2293{
617ba13b 2294 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 2295
2296 if (sbi->s_commit_interval)
2297 journal->j_commit_interval = sbi->s_commit_interval;
617ba13b 2298 /* We could also set up an ext4-specific default for the commit
ac27a0ec
DK		 2299 * interval here, but for now we'll just fall back to the jbd
2300 * default. */
2301
2302 spin_lock(&journal->j_state_lock);
2303 if (test_opt(sb, BARRIER))
dab291af 2304 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 2305 else
dab291af 2306 journal->j_flags &= ~JBD2_BARRIER;
ac27a0ec
DK		 2307 spin_unlock(&journal->j_state_lock);
2308}
2309
617ba13b 2310static journal_t *ext4_get_journal(struct super_block *sb,
ac27a0ec
DK		 2311 unsigned int journal_inum)
2312{
2313 struct inode *journal_inode;
2314 journal_t *journal;
2315
2316 /* First, test for the existence of a valid inode on disk. Bad
2317 * things happen if we iget() an unused inode, as the subsequent
2318 * iput() will try to delete it. */
2319
2320 journal_inode = iget(sb, journal_inum);
2321 if (!journal_inode) {
617ba13b 2322 printk(KERN_ERR "EXT4-fs: no journal found.\n");
ac27a0ec
DK		 2323 return NULL;
2324 }
2325 if (!journal_inode->i_nlink) {
2326 make_bad_inode(journal_inode);
2327 iput(journal_inode);
617ba13b 2328 printk(KERN_ERR "EXT4-fs: journal inode is deleted.\n");
ac27a0ec
DK		 2329 return NULL;
2330 }
2331
2332 jbd_debug(2, "Journal inode found at %p: %Ld bytes\n",
2333 journal_inode, journal_inode->i_size);
2334 if (is_bad_inode(journal_inode) || !S_ISREG(journal_inode->i_mode)) {
617ba13b 2335 printk(KERN_ERR "EXT4-fs: invalid journal inode.\n");
ac27a0ec
DK		 2336 iput(journal_inode);
2337 return NULL;
2338 }
2339
dab291af 2340 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 2341 if (!journal) {
617ba13b 2342 printk(KERN_ERR "EXT4-fs: Could not load journal inode\n");
ac27a0ec
DK		 2343 iput(journal_inode);
2344 return NULL;
2345 }
2346 journal->j_private = sb;
617ba13b 2347 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 2348 return journal;
2349}
2350
617ba13b 2351static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 2352 dev_t j_dev)
2353{
2354 struct buffer_head * bh;
2355 journal_t *journal;
617ba13b
MC		 2356 ext4_fsblk_t start;
2357 ext4_fsblk_t len;
ac27a0ec 2358 int hblock, blocksize;
617ba13b 2359 ext4_fsblk_t sb_block;
ac27a0ec 2360 unsigned long offset;
617ba13b 2361 struct ext4_super_block * es;
ac27a0ec
DK		 2362 struct block_device *bdev;
2363
617ba13b 2364 bdev = ext4_blkdev_get(j_dev);
ac27a0ec
DK		 2365 if (bdev == NULL)
2366 return NULL;
2367
2368 if (bd_claim(bdev, sb)) {
2369 printk(KERN_ERR
8c55e204 2370 "EXT4: failed to claim external journal device.\n");
ac27a0ec
DK		 2371 blkdev_put(bdev);
2372 return NULL;
2373 }
2374
2375 blocksize = sb->s_blocksize;
2376 hblock = bdev_hardsect_size(bdev);
2377 if (blocksize < hblock) {
2378 printk(KERN_ERR
617ba13b 2379 "EXT4-fs: blocksize too small for journal device.\n");
ac27a0ec
DK		 2380 goto out_bdev;
2381 }
2382
617ba13b
MC		 2383 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
2384 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 2385 set_blocksize(bdev, blocksize);
2386 if (!(bh = __bread(bdev, sb_block, blocksize))) {
617ba13b 2387 printk(KERN_ERR "EXT4-fs: couldn't read superblock of "
ac27a0ec
DK		 2388 "external journal\n");
2389 goto out_bdev;
2390 }
2391
617ba13b
MC		 2392 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2393 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 2394 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b
MC		 2395 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
2396 printk(KERN_ERR "EXT4-fs: external journal has "
ac27a0ec
DK		 2397 "bad superblock\n");
2398 brelse(bh);
2399 goto out_bdev;
2400 }
2401
617ba13b
MC		 2402 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
2403 printk(KERN_ERR "EXT4-fs: journal UUID does not match\n");
ac27a0ec
DK		 2404 brelse(bh);
2405 goto out_bdev;
2406 }
2407
bd81d8ee 2408 len = ext4_blocks_count(es);
ac27a0ec
DK		 2409 start = sb_block + 1;
2410 brelse(bh); /* we're done with the superblock */
2411
dab291af 2412 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 2413 start, len, blocksize);
2414 if (!journal) {
617ba13b 2415 printk(KERN_ERR "EXT4-fs: failed to create device journal\n");
ac27a0ec
DK		 2416 goto out_bdev;
2417 }
2418 journal->j_private = sb;
2419 ll_rw_block(READ, 1, &journal->j_sb_buffer);
2420 wait_on_buffer(journal->j_sb_buffer);
2421 if (!buffer_uptodate(journal->j_sb_buffer)) {
617ba13b 2422 printk(KERN_ERR "EXT4-fs: I/O error on journal device\n");
ac27a0ec
DK		 2423 goto out_journal;
2424 }
2425 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
617ba13b 2426 printk(KERN_ERR "EXT4-fs: External journal has more than one "
ac27a0ec
DK		 2427 "user (unsupported) - %d\n",
2428 be32_to_cpu(journal->j_superblock->s_nr_users));
2429 goto out_journal;
2430 }
617ba13b
MC		 2431 EXT4_SB(sb)->journal_bdev = bdev;
2432 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 2433 return journal;
2434out_journal:
dab291af 2435 jbd2_journal_destroy(journal);
ac27a0ec 2436out_bdev:
617ba13b 2437 ext4_blkdev_put(bdev);
ac27a0ec
DK		 2438 return NULL;
2439}
2440
617ba13b
MC		 2441static int ext4_load_journal(struct super_block *sb,
2442 struct ext4_super_block *es,
ac27a0ec
DK		 2443 unsigned long journal_devnum)
2444{
2445 journal_t *journal;
2446 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
2447 dev_t journal_dev;
2448 int err = 0;
2449 int really_read_only;
2450
2451 if (journal_devnum &&
2452 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
617ba13b 2453 printk(KERN_INFO "EXT4-fs: external journal device major/minor "
ac27a0ec
DK		 2454 "numbers have changed\n");
2455 journal_dev = new_decode_dev(journal_devnum);
2456 } else
2457 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
2458
2459 really_read_only = bdev_read_only(sb->s_bdev);
2460
2461 /*
2462 * Are we loading a blank journal or performing recovery after a
2463 * crash? For recovery, we need to check in advance whether we
2464 * can get read-write access to the device.
2465 */
2466
617ba13b 2467 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
ac27a0ec 2468 if (sb->s_flags & MS_RDONLY) {
617ba13b 2469 printk(KERN_INFO "EXT4-fs: INFO: recovery "
ac27a0ec
DK		 2470 "required on readonly filesystem.\n");
2471 if (really_read_only) {
617ba13b 2472 printk(KERN_ERR "EXT4-fs: write access "
ac27a0ec
DK		 2473 "unavailable, cannot proceed.\n");
2474 return -EROFS;
2475 }
617ba13b 2476 printk (KERN_INFO "EXT4-fs: write access will "
ac27a0ec
DK		 2477 "be enabled during recovery.\n");
2478 }
2479 }
2480
2481 if (journal_inum && journal_dev) {
617ba13b 2482 printk(KERN_ERR "EXT4-fs: filesystem has both journal "
ac27a0ec
DK		 2483 "and inode journals!\n");
2484 return -EINVAL;
2485 }
2486
2487 if (journal_inum) {
617ba13b 2488 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 2489 return -EINVAL;
2490 } else {
617ba13b 2491 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 2492 return -EINVAL;
2493 }
2494
2495 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
dab291af 2496 err = jbd2_journal_update_format(journal);
ac27a0ec 2497 if (err) {
617ba13b 2498 printk(KERN_ERR "EXT4-fs: error updating journal.\n");
dab291af 2499 jbd2_journal_destroy(journal);
ac27a0ec
DK		 2500 return err;
2501 }
2502 }
2503
617ba13b 2504 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
dab291af 2505 err = jbd2_journal_wipe(journal, !really_read_only);
ac27a0ec 2506 if (!err)
dab291af 2507 err = jbd2_journal_load(journal);
ac27a0ec
DK		 2508
2509 if (err) {
617ba13b 2510 printk(KERN_ERR "EXT4-fs: error loading journal.\n");
dab291af 2511 jbd2_journal_destroy(journal);
ac27a0ec
DK		 2512 return err;
2513 }
2514
617ba13b
MC		 2515 EXT4_SB(sb)->s_journal = journal;
2516 ext4_clear_journal_err(sb, es);
ac27a0ec
DK		 2517
2518 if (journal_devnum &&
2519 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2520 es->s_journal_dev = cpu_to_le32(journal_devnum);
2521 sb->s_dirt = 1;
2522
2523 /* Make sure we flush the recovery flag to disk. */
617ba13b 2524 ext4_commit_super(sb, es, 1);
ac27a0ec
DK		 2525 }
2526
2527 return 0;
2528}
2529
617ba13b
MC		 2530static int ext4_create_journal(struct super_block * sb,
2531 struct ext4_super_block * es,
ac27a0ec
DK		 2532 unsigned int journal_inum)
2533{
2534 journal_t *journal;
6c675bd4 2535 int err;
ac27a0ec
DK		 2536
2537 if (sb->s_flags & MS_RDONLY) {
617ba13b 2538 printk(KERN_ERR "EXT4-fs: readonly filesystem when trying to "
ac27a0ec
DK		 2539 "create journal.\n");
2540 return -EROFS;
2541 }
2542
6c675bd4
BP		 2543 journal = ext4_get_journal(sb, journal_inum);
2544 if (!journal)
ac27a0ec
DK		 2545 return -EINVAL;
2546
617ba13b 2547 printk(KERN_INFO "EXT4-fs: creating new journal on inode %u\n",
ac27a0ec
DK		 2548 journal_inum);
2549
6c675bd4
BP		 2550 err = jbd2_journal_create(journal);
2551 if (err) {
617ba13b 2552 printk(KERN_ERR "EXT4-fs: error creating journal.\n");
dab291af 2553 jbd2_journal_destroy(journal);
ac27a0ec
DK		 2554 return -EIO;
2555 }
2556
617ba13b 2557 EXT4_SB(sb)->s_journal = journal;
ac27a0ec 2558
617ba13b
MC		 2559 ext4_update_dynamic_rev(sb);
2560 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2561 EXT4_SET_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL);
ac27a0ec
DK		 2562
2563 es->s_journal_inum = cpu_to_le32(journal_inum);
2564 sb->s_dirt = 1;
2565
2566 /* Make sure we flush the recovery flag to disk. */
617ba13b 2567 ext4_commit_super(sb, es, 1);
ac27a0ec
DK		 2568
2569 return 0;
2570}
2571
617ba13b
MC		 2572static void ext4_commit_super (struct super_block * sb,
2573 struct ext4_super_block * es,
ac27a0ec
DK		 2574 int sync)
2575{
617ba13b 2576 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
ac27a0ec
DK		 2577
2578 if (!sbh)
2579 return;
2580 es->s_wtime = cpu_to_le32(get_seconds());
bd81d8ee 2581 ext4_free_blocks_count_set(es, ext4_count_free_blocks(sb));
617ba13b 2582 es->s_free_inodes_count = cpu_to_le32(ext4_count_free_inodes(sb));
ac27a0ec
DK		 2583 BUFFER_TRACE(sbh, "marking dirty");
2584 mark_buffer_dirty(sbh);
2585 if (sync)
2586 sync_dirty_buffer(sbh);
2587}
2588
2589
2590/*
2591 * Have we just finished recovery? If so, and if we are mounting (or
2592 * remounting) the filesystem readonly, then we will end up with a
2593 * consistent fs on disk. Record that fact.
2594 */
617ba13b
MC		 2595static void ext4_mark_recovery_complete(struct super_block * sb,
2596 struct ext4_super_block * es)
ac27a0ec 2597{
617ba13b 2598 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 2599
dab291af
MC		 2600 jbd2_journal_lock_updates(journal);
2601 jbd2_journal_flush(journal);
32c37730 2602 lock_super(sb);
617ba13b 2603 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
ac27a0ec 2604 sb->s_flags & MS_RDONLY) {
617ba13b 2605 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 2606 sb->s_dirt = 0;
617ba13b 2607 ext4_commit_super(sb, es, 1);
ac27a0ec 2608 }
32c37730 2609 unlock_super(sb);
dab291af 2610 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 2611}
2612
2613/*
2614 * If we are mounting (or read-write remounting) a filesystem whose journal
2615 * has recorded an error from a previous lifetime, move that error to the
2616 * main filesystem now.
2617 */
617ba13b
MC		 2618static void ext4_clear_journal_err(struct super_block * sb,
2619 struct ext4_super_block * es)
ac27a0ec
DK		 2620{
2621 journal_t *journal;
2622 int j_errno;
2623 const char *errstr;
2624
617ba13b 2625 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 2626
2627 /*
2628 * Now check for any error status which may have been recorded in the
617ba13b 2629 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 2630 */
2631
dab291af 2632 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 2633 if (j_errno) {
2634 char nbuf[16];
2635
617ba13b
MC		 2636 errstr = ext4_decode_error(sb, j_errno, nbuf);
2637 ext4_warning(sb, __FUNCTION__, "Filesystem error recorded "
ac27a0ec 2638 "from previous mount: %s", errstr);
617ba13b 2639 ext4_warning(sb, __FUNCTION__, "Marking fs in need of "
ac27a0ec
DK		 2640 "filesystem check.");
2641
617ba13b
MC		 2642 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2643 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2644 ext4_commit_super (sb, es, 1);
ac27a0ec 2645
dab291af 2646 jbd2_journal_clear_err(journal);
ac27a0ec
DK		 2647 }
2648}
2649
2650/*
2651 * Force the running and committing transactions to commit,
2652 * and wait on the commit.
2653 */
617ba13b 2654int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 2655{
2656 journal_t *journal;
2657 int ret;
2658
2659 if (sb->s_flags & MS_RDONLY)
2660 return 0;
2661
617ba13b 2662 journal = EXT4_SB(sb)->s_journal;
ac27a0ec 2663 sb->s_dirt = 0;
617ba13b 2664 ret = ext4_journal_force_commit(journal);
ac27a0ec
DK		 2665 return ret;
2666}
2667
2668/*
617ba13b 2669 * Ext4 always journals updates to the superblock itself, so we don't
ac27a0ec
DK		 2670 * have to propagate any other updates to the superblock on disk at this
2671 * point. Just start an async writeback to get the buffers on their way
2672 * to the disk.
2673 *
2674 * This implicitly triggers the writebehind on sync().
2675 */
2676
617ba13b 2677static void ext4_write_super (struct super_block * sb)
ac27a0ec
DK		 2678{
2679 if (mutex_trylock(&sb->s_lock) != 0)
2680 BUG();
2681 sb->s_dirt = 0;
2682}
2683
617ba13b 2684static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec
DK		 2685{
2686 tid_t target;
2687
2688 sb->s_dirt = 0;
dab291af 2689 if (jbd2_journal_start_commit(EXT4_SB(sb)->s_journal, &target)) {
ac27a0ec 2690 if (wait)
dab291af 2691 jbd2_log_wait_commit(EXT4_SB(sb)->s_journal, target);
ac27a0ec
DK		 2692 }
2693 return 0;
2694}
2695
2696/*
2697 * LVM calls this function before a (read-only) snapshot is created. This
2698 * gives us a chance to flush the journal completely and mark the fs clean.
2699 */
617ba13b 2700static void ext4_write_super_lockfs(struct super_block *sb)
ac27a0ec
DK		 2701{
2702 sb->s_dirt = 0;
2703
2704 if (!(sb->s_flags & MS_RDONLY)) {
617ba13b 2705 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 2706
2707 /* Now we set up the journal barrier. */
dab291af
MC		 2708 jbd2_journal_lock_updates(journal);
2709 jbd2_journal_flush(journal);
ac27a0ec
DK		 2710
2711 /* Journal blocked and flushed, clear needs_recovery flag. */
617ba13b
MC		 2712 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2713 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
ac27a0ec
DK		 2714 }
2715}
2716
2717/*
2718 * Called by LVM after the snapshot is done. We need to reset the RECOVER
2719 * flag here, even though the filesystem is not technically dirty yet.
2720 */
617ba13b 2721static void ext4_unlockfs(struct super_block *sb)
ac27a0ec
DK		 2722{
2723 if (!(sb->s_flags & MS_RDONLY)) {
2724 lock_super(sb);
2725 /* Reser the needs_recovery flag before the fs is unlocked. */
617ba13b
MC		 2726 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2727 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
ac27a0ec 2728 unlock_super(sb);
dab291af 2729 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
ac27a0ec
DK		 2730 }
2731}
2732
617ba13b 2733static int ext4_remount (struct super_block * sb, int * flags, char * data)
ac27a0ec 2734{
617ba13b
MC		 2735 struct ext4_super_block * es;
2736 struct ext4_sb_info *sbi = EXT4_SB(sb);
2737 ext4_fsblk_t n_blocks_count = 0;
ac27a0ec 2738 unsigned long old_sb_flags;
617ba13b 2739 struct ext4_mount_options old_opts;
ac27a0ec
DK		 2740 int err;
2741#ifdef CONFIG_QUOTA
2742 int i;
2743#endif
2744
2745 /* Store the original options */
2746 old_sb_flags = sb->s_flags;
2747 old_opts.s_mount_opt = sbi->s_mount_opt;
2748 old_opts.s_resuid = sbi->s_resuid;
2749 old_opts.s_resgid = sbi->s_resgid;
2750 old_opts.s_commit_interval = sbi->s_commit_interval;
2751#ifdef CONFIG_QUOTA
2752 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
2753 for (i = 0; i < MAXQUOTAS; i++)
2754 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
2755#endif
2756
2757 /*
2758 * Allow the "check" option to be passed as a remount option.
2759 */
2760 if (!parse_options(data, sb, NULL, NULL, &n_blocks_count, 1)) {
2761 err = -EINVAL;
2762 goto restore_opts;
2763 }
2764
617ba13b
MC		 2765 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT)
2766 ext4_abort(sb, __FUNCTION__, "Abort forced by user");
ac27a0ec
DK		 2767
2768 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
617ba13b 2769 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec
DK		 2770
2771 es = sbi->s_es;
2772
617ba13b 2773 ext4_init_journal_params(sb, sbi->s_journal);
ac27a0ec
DK		 2774
2775 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
bd81d8ee 2776 n_blocks_count > ext4_blocks_count(es)) {
617ba13b 2777 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT) {
ac27a0ec
DK		 2778 err = -EROFS;
2779 goto restore_opts;
2780 }
2781
2782 if (*flags & MS_RDONLY) {
2783 /*
2784 * First of all, the unconditional stuff we have to do
2785 * to disable replay of the journal when we next remount
2786 */
2787 sb->s_flags |= MS_RDONLY;
2788
2789 /*
2790 * OK, test if we are remounting a valid rw partition
2791 * readonly, and if so set the rdonly flag and then
2792 * mark the partition as valid again.
2793 */
617ba13b
MC		 2794 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
2795 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 2796 es->s_state = cpu_to_le16(sbi->s_mount_state);
2797
32c37730
JK		 2798 /*
2799 * We have to unlock super so that we can wait for
2800 * transactions.
2801 */
2802 unlock_super(sb);
617ba13b 2803 ext4_mark_recovery_complete(sb, es);
32c37730 2804 lock_super(sb);
ac27a0ec
DK		 2805 } else {
2806 __le32 ret;
617ba13b
MC		 2807 if ((ret = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2808 ~EXT4_FEATURE_RO_COMPAT_SUPP))) {
2809 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
ac27a0ec
DK		 2810 "remount RDWR because of unsupported "
2811 "optional features (%x).\n",
2812 sb->s_id, le32_to_cpu(ret));
2813 err = -EROFS;
2814 goto restore_opts;
2815 }
ead6596b
ES		 2816
2817 /*
2818 * If we have an unprocessed orphan list hanging
2819 * around from a previously readonly bdev mount,
2820 * require a full umount/remount for now.
2821 */
2822 if (es->s_last_orphan) {
2823 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
2824 "remount RDWR because of unprocessed "
2825 "orphan inode list. Please "
2826 "umount/remount instead.\n",
2827 sb->s_id);
2828 err = -EINVAL;
2829 goto restore_opts;
2830 }
2831
ac27a0ec
DK		 2832 /*
2833 * Mounting a RDONLY partition read-write, so reread
2834 * and store the current valid flag. (It may have
2835 * been changed by e2fsck since we originally mounted
2836 * the partition.)
2837 */
617ba13b 2838 ext4_clear_journal_err(sb, es);
ac27a0ec 2839 sbi->s_mount_state = le16_to_cpu(es->s_state);
617ba13b 2840 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
ac27a0ec 2841 goto restore_opts;
617ba13b 2842 if (!ext4_setup_super (sb, es, 0))
ac27a0ec
DK		 2843 sb->s_flags &= ~MS_RDONLY;
2844 }
2845 }
2846#ifdef CONFIG_QUOTA
2847 /* Release old quota file names */
2848 for (i = 0; i < MAXQUOTAS; i++)
2849 if (old_opts.s_qf_names[i] &&
2850 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2851 kfree(old_opts.s_qf_names[i]);
2852#endif
2853 return 0;
2854restore_opts:
2855 sb->s_flags = old_sb_flags;
2856 sbi->s_mount_opt = old_opts.s_mount_opt;
2857 sbi->s_resuid = old_opts.s_resuid;
2858 sbi->s_resgid = old_opts.s_resgid;
2859 sbi->s_commit_interval = old_opts.s_commit_interval;
2860#ifdef CONFIG_QUOTA
2861 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
2862 for (i = 0; i < MAXQUOTAS; i++) {
2863 if (sbi->s_qf_names[i] &&
2864 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2865 kfree(sbi->s_qf_names[i]);
2866 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
2867 }
2868#endif
2869 return err;
2870}
2871
617ba13b 2872static int ext4_statfs (struct dentry * dentry, struct kstatfs * buf)
ac27a0ec
DK		 2873{
2874 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 2875 struct ext4_sb_info *sbi = EXT4_SB(sb);
2876 struct ext4_super_block *es = sbi->s_es;
960cc398 2877 u64 fsid;
ac27a0ec 2878
5e70030d
BP		 2879 if (test_opt(sb, MINIX_DF)) {
2880 sbi->s_overhead_last = 0;
6bc9feff 2881 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
fd2d4291 2882 ext4_group_t ngroups = sbi->s_groups_count, i;
5e70030d 2883 ext4_fsblk_t overhead = 0;
ac27a0ec
DK		 2884 smp_rmb();
2885
2886 /*
5e70030d
BP		 2887 * Compute the overhead (FS structures). This is constant
2888 * for a given filesystem unless the number of block groups
2889 * changes so we cache the previous value until it does.
ac27a0ec
DK		 2890 */
2891
2892 /*
2893 * All of the blocks before first_data_block are
2894 * overhead
2895 */
2896 overhead = le32_to_cpu(es->s_first_data_block);
2897
2898 /*
2899 * Add the overhead attributed to the superblock and
2900 * block group descriptors. If the sparse superblocks
2901 * feature is turned on, then not all groups have this.
2902 */
2903 for (i = 0; i < ngroups; i++) {
617ba13b
MC		 2904 overhead += ext4_bg_has_super(sb, i) +
2905 ext4_bg_num_gdb(sb, i);
ac27a0ec
DK		 2906 cond_resched();
2907 }
2908
2909 /*
2910 * Every block group has an inode bitmap, a block
2911 * bitmap, and an inode table.
2912 */
5e70030d
BP		 2913 overhead += ngroups * (2 + sbi->s_itb_per_group);
2914 sbi->s_overhead_last = overhead;
2915 smp_wmb();
6bc9feff 2916 sbi->s_blocks_last = ext4_blocks_count(es);
ac27a0ec
DK		 2917 }
2918
617ba13b 2919 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 2920 buf->f_bsize = sb->s_blocksize;
5e70030d 2921 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
52d9f3b4 2922 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter);
308ba3ec 2923 ext4_free_blocks_count_set(es, buf->f_bfree);
bd81d8ee
LV		 2924 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
2925 if (buf->f_bfree < ext4_r_blocks_count(es))
ac27a0ec
DK		 2926 buf->f_bavail = 0;
2927 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 2928 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
5e70030d 2929 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
617ba13b 2930 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 2931 fsid = le64_to_cpup((void *)es->s_uuid) ^
2932 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
2933 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
2934 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
ac27a0ec
DK		 2935 return 0;
2936}
2937
2938/* Helper function for writing quotas on sync - we need to start transaction before quota file
2939 * is locked for write. Otherwise the are possible deadlocks:
2940 * Process 1 Process 2
617ba13b 2941 * ext4_create() quota_sync()
dab291af 2942 * jbd2_journal_start() write_dquot()
ac27a0ec 2943 * DQUOT_INIT() down(dqio_mutex)
dab291af 2944 * down(dqio_mutex) jbd2_journal_start()
ac27a0ec
DK		 2945 *
2946 */
2947
2948#ifdef CONFIG_QUOTA
2949
2950static inline struct inode *dquot_to_inode(struct dquot *dquot)
2951{
2952 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
2953}
2954
617ba13b 2955static int ext4_dquot_initialize(struct inode *inode, int type)
ac27a0ec
DK		 2956{
2957 handle_t *handle;
2958 int ret, err;
2959
2960 /* We may create quota structure so we need to reserve enough blocks */
617ba13b 2961 handle = ext4_journal_start(inode, 2*EXT4_QUOTA_INIT_BLOCKS(inode->i_sb));
ac27a0ec
DK		 2962 if (IS_ERR(handle))
2963 return PTR_ERR(handle);
2964 ret = dquot_initialize(inode, type);
617ba13b 2965 err = ext4_journal_stop(handle);
ac27a0ec
DK		 2966 if (!ret)
2967 ret = err;
2968 return ret;
2969}
2970
617ba13b 2971static int ext4_dquot_drop(struct inode *inode)
ac27a0ec
DK		 2972{
2973 handle_t *handle;
2974 int ret, err;
2975
2976 /* We may delete quota structure so we need to reserve enough blocks */
617ba13b 2977 handle = ext4_journal_start(inode, 2*EXT4_QUOTA_DEL_BLOCKS(inode->i_sb));
ac27a0ec
DK		 2978 if (IS_ERR(handle))
2979 return PTR_ERR(handle);
2980 ret = dquot_drop(inode);
617ba13b 2981 err = ext4_journal_stop(handle);
ac27a0ec
DK		 2982 if (!ret)
2983 ret = err;
2984 return ret;
2985}
2986
617ba13b 2987static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 2988{
2989 int ret, err;
2990 handle_t *handle;
2991 struct inode *inode;
2992
2993 inode = dquot_to_inode(dquot);
617ba13b
MC		 2994 handle = ext4_journal_start(inode,
2995 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 2996 if (IS_ERR(handle))
2997 return PTR_ERR(handle);
2998 ret = dquot_commit(dquot);
617ba13b 2999 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3000 if (!ret)
3001 ret = err;
3002 return ret;
3003}
3004
617ba13b 3005static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 3006{
3007 int ret, err;
3008 handle_t *handle;
3009
617ba13b
MC		 3010 handle = ext4_journal_start(dquot_to_inode(dquot),
3011 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 3012 if (IS_ERR(handle))
3013 return PTR_ERR(handle);
3014 ret = dquot_acquire(dquot);
617ba13b 3015 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3016 if (!ret)
3017 ret = err;
3018 return ret;
3019}
3020
617ba13b 3021static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 3022{
3023 int ret, err;
3024 handle_t *handle;
3025
617ba13b
MC		 3026 handle = ext4_journal_start(dquot_to_inode(dquot),
3027 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 3028 if (IS_ERR(handle)) {
3029 /* Release dquot anyway to avoid endless cycle in dqput() */
3030 dquot_release(dquot);
ac27a0ec 3031 return PTR_ERR(handle);
9c3013e9 3032 }
ac27a0ec 3033 ret = dquot_release(dquot);
617ba13b 3034 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3035 if (!ret)
3036 ret = err;
3037 return ret;
3038}
3039
617ba13b 3040static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec
DK		 3041{
3042 /* Are we journalling quotas? */
617ba13b
MC		 3043 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3044 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
ac27a0ec 3045 dquot_mark_dquot_dirty(dquot);
617ba13b 3046 return ext4_write_dquot(dquot);
ac27a0ec
DK		 3047 } else {
3048 return dquot_mark_dquot_dirty(dquot);
3049 }
3050}
3051
617ba13b 3052static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 3053{
3054 int ret, err;
3055 handle_t *handle;
3056
3057 /* Data block + inode block */
617ba13b 3058 handle = ext4_journal_start(sb->s_root->d_inode, 2);
ac27a0ec
DK		 3059 if (IS_ERR(handle))
3060 return PTR_ERR(handle);
3061 ret = dquot_commit_info(sb, type);
617ba13b 3062 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3063 if (!ret)
3064 ret = err;
3065 return ret;
3066}
3067
3068/*
3069 * Turn on quotas during mount time - we need to find
3070 * the quota file and such...
3071 */
617ba13b 3072static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 3073{
617ba13b
MC		 3074 return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
3075 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 3076}
3077
3078/*
3079 * Standard function to be called on quota_on
3080 */
617ba13b 3081static int ext4_quota_on(struct super_block *sb, int type, int format_id,
ac27a0ec
DK		 3082 char *path)
3083{
3084 int err;
3085 struct nameidata nd;
3086
3087 if (!test_opt(sb, QUOTA))
3088 return -EINVAL;
3089 /* Not journalling quota? */
617ba13b
MC		 3090 if (!EXT4_SB(sb)->s_qf_names[USRQUOTA] &&
3091 !EXT4_SB(sb)->s_qf_names[GRPQUOTA])
ac27a0ec
DK		 3092 return vfs_quota_on(sb, type, format_id, path);
3093 err = path_lookup(path, LOOKUP_FOLLOW, &nd);
3094 if (err)
3095 return err;
3096 /* Quotafile not on the same filesystem? */
3097 if (nd.mnt->mnt_sb != sb) {
3098 path_release(&nd);
3099 return -EXDEV;
3100 }
3101 /* Quotafile not of fs root? */
3102 if (nd.dentry->d_parent->d_inode != sb->s_root->d_inode)
3103 printk(KERN_WARNING
617ba13b 3104 "EXT4-fs: Quota file not on filesystem root. "
ac27a0ec
DK		 3105 "Journalled quota will not work.\n");
3106 path_release(&nd);
3107 return vfs_quota_on(sb, type, format_id, path);
3108}
3109
3110/* Read data from quotafile - avoid pagecache and such because we cannot afford
3111 * acquiring the locks... As quota files are never truncated and quota code
3112 * itself serializes the operations (and noone else should touch the files)
3113 * we don't have to be afraid of races */
617ba13b 3114static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 3115 size_t len, loff_t off)
3116{
3117 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 3118 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 3119 int err = 0;
3120 int offset = off & (sb->s_blocksize - 1);
3121 int tocopy;
3122 size_t toread;
3123 struct buffer_head *bh;
3124 loff_t i_size = i_size_read(inode);
3125
3126 if (off > i_size)
3127 return 0;
3128 if (off+len > i_size)
3129 len = i_size-off;
3130 toread = len;
3131 while (toread > 0) {
3132 tocopy = sb->s_blocksize - offset < toread ?
3133 sb->s_blocksize - offset : toread;
617ba13b 3134 bh = ext4_bread(NULL, inode, blk, 0, &err);
ac27a0ec
DK		 3135 if (err)
3136 return err;
3137 if (!bh) /* A hole? */
3138 memset(data, 0, tocopy);
3139 else
3140 memcpy(data, bh->b_data+offset, tocopy);
3141 brelse(bh);
3142 offset = 0;
3143 toread -= tocopy;
3144 data += tocopy;
3145 blk++;
3146 }
3147 return len;
3148}
3149
3150/* Write to quotafile (we know the transaction is already started and has
3151 * enough credits) */
617ba13b 3152static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 3153 const char *data, size_t len, loff_t off)
3154{
3155 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 3156 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 3157 int err = 0;
3158 int offset = off & (sb->s_blocksize - 1);
3159 int tocopy;
617ba13b 3160 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
ac27a0ec
DK		 3161 size_t towrite = len;
3162 struct buffer_head *bh;
3163 handle_t *handle = journal_current_handle();
3164
9c3013e9
JK		 3165 if (!handle) {
3166 printk(KERN_WARNING "EXT4-fs: Quota write (off=%Lu, len=%Lu)"
3167 " cancelled because transaction is not started.\n",
3168 (unsigned long long)off, (unsigned long long)len);
3169 return -EIO;
3170 }
ac27a0ec
DK		 3171 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
3172 while (towrite > 0) {
3173 tocopy = sb->s_blocksize - offset < towrite ?
3174 sb->s_blocksize - offset : towrite;
617ba13b 3175 bh = ext4_bread(handle, inode, blk, 1, &err);
ac27a0ec
DK		 3176 if (!bh)
3177 goto out;
3178 if (journal_quota) {
617ba13b 3179 err = ext4_journal_get_write_access(handle, bh);
ac27a0ec
DK		 3180 if (err) {
3181 brelse(bh);
3182 goto out;
3183 }
3184 }
3185 lock_buffer(bh);
3186 memcpy(bh->b_data+offset, data, tocopy);
3187 flush_dcache_page(bh->b_page);
3188 unlock_buffer(bh);
3189 if (journal_quota)
617ba13b 3190 err = ext4_journal_dirty_metadata(handle, bh);
ac27a0ec
DK		 3191 else {
3192 /* Always do at least ordered writes for quotas */
617ba13b 3193 err = ext4_journal_dirty_data(handle, bh);
ac27a0ec
DK		 3194 mark_buffer_dirty(bh);
3195 }
3196 brelse(bh);
3197 if (err)
3198 goto out;
3199 offset = 0;
3200 towrite -= tocopy;
3201 data += tocopy;
3202 blk++;
3203 }
3204out:
3205 if (len == towrite)
3206 return err;
3207 if (inode->i_size < off+len-towrite) {
3208 i_size_write(inode, off+len-towrite);
617ba13b 3209 EXT4_I(inode)->i_disksize = inode->i_size;
ac27a0ec 3210 }
ac27a0ec 3211 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
617ba13b 3212 ext4_mark_inode_dirty(handle, inode);
ac27a0ec
DK		 3213 mutex_unlock(&inode->i_mutex);
3214 return len - towrite;
3215}
3216
3217#endif
3218
617ba13b 3219static int ext4_get_sb(struct file_system_type *fs_type,
ac27a0ec
DK		 3220 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
3221{
617ba13b 3222 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
ac27a0ec
DK		 3223}
3224
617ba13b 3225static struct file_system_type ext4dev_fs_type = {
ac27a0ec 3226 .owner = THIS_MODULE,
617ba13b
MC		 3227 .name = "ext4dev",
3228 .get_sb = ext4_get_sb,
ac27a0ec
DK		 3229 .kill_sb = kill_block_super,
3230 .fs_flags = FS_REQUIRES_DEV,
3231};
3232
617ba13b 3233static int __init init_ext4_fs(void)
ac27a0ec 3234{
617ba13b 3235 int err = init_ext4_xattr();
ac27a0ec
DK		 3236 if (err)
3237 return err;
3238 err = init_inodecache();
3239 if (err)
3240 goto out1;
63f57933 3241 err = register_filesystem(&ext4dev_fs_type);
ac27a0ec
DK		 3242 if (err)
3243 goto out;
3244 return 0;
3245out:
3246 destroy_inodecache();
3247out1:
617ba13b 3248 exit_ext4_xattr();
ac27a0ec
DK		 3249 return err;
3250}
3251
617ba13b 3252static void __exit exit_ext4_fs(void)
ac27a0ec 3253{
617ba13b 3254 unregister_filesystem(&ext4dev_fs_type);
ac27a0ec 3255 destroy_inodecache();
617ba13b 3256 exit_ext4_xattr();
ac27a0ec
DK		 3257}
3258
3259MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
617ba13b 3260MODULE_DESCRIPTION("Fourth Extended Filesystem with extents");
ac27a0ec 3261MODULE_LICENSE("GPL");
617ba13b
MC		 3262module_init(init_ext4_fs)
3263module_exit(exit_ext4_fs)
Ubuntu Artful kernel mirror