]> bbs.cooldavid.org Git - net-next-2.6.git/blame - fs/ext4/super.c
git://bbs.cooldavid.org / net-next-2.6.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
Shrink struct task_struct::oomkilladj
[net-next-2.6.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>
ac27a0ec
DK		 40
41#include <asm/uaccess.h>
42
43#include "xattr.h"
44#include "acl.h"
45#include "namei.h"
46
617ba13b 47static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
ac27a0ec 48 unsigned long journal_devnum);
617ba13b 49static int ext4_create_journal(struct super_block *, struct ext4_super_block *,
ac27a0ec 50 unsigned int);
617ba13b
MC		 51static void ext4_commit_super (struct super_block * sb,
52 struct ext4_super_block * es,
ac27a0ec 53 int sync);
617ba13b
MC		 54static void ext4_mark_recovery_complete(struct super_block * sb,
55 struct ext4_super_block * es);
56static void ext4_clear_journal_err(struct super_block * sb,
57 struct ext4_super_block * es);
58static int ext4_sync_fs(struct super_block *sb, int wait);
59static const char *ext4_decode_error(struct super_block * sb, int errno,
ac27a0ec 60 char nbuf[16]);
617ba13b
MC		 61static int ext4_remount (struct super_block * sb, int * flags, char * data);
62static int ext4_statfs (struct dentry * dentry, struct kstatfs * buf);
63static void ext4_unlockfs(struct super_block *sb);
64static void ext4_write_super (struct super_block * sb);
65static void ext4_write_super_lockfs(struct super_block *sb);
ac27a0ec 66
bd81d8ee 67
8fadc143
AR		 68ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
69 struct ext4_group_desc *bg)
bd81d8ee
LV		 70{
71 return le32_to_cpu(bg->bg_block_bitmap) |
8fadc143
AR		 72 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
73 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 74}
75
8fadc143
AR		 76ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
77 struct ext4_group_desc *bg)
bd81d8ee
LV		 78{
79 return le32_to_cpu(bg->bg_inode_bitmap) |
8fadc143
AR		 80 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
81 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 82}
83
8fadc143
AR		 84ext4_fsblk_t ext4_inode_table(struct super_block *sb,
85 struct ext4_group_desc *bg)
bd81d8ee
LV		 86{
87 return le32_to_cpu(bg->bg_inode_table) |
8fadc143
AR		 88 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
89 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
bd81d8ee
LV		 90}
91
8fadc143
AR		 92void ext4_block_bitmap_set(struct super_block *sb,
93 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee
LV		 94{
95 bg->bg_block_bitmap = cpu_to_le32((u32)blk);
8fadc143
AR		 96 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
97 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 98}
99
8fadc143
AR		 100void ext4_inode_bitmap_set(struct super_block *sb,
101 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee
LV		 102{
103 bg->bg_inode_bitmap = cpu_to_le32((u32)blk);
8fadc143
AR		 104 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
105 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 106}
107
8fadc143
AR		 108void ext4_inode_table_set(struct super_block *sb,
109 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee
LV		 110{
111 bg->bg_inode_table = cpu_to_le32((u32)blk);
8fadc143
AR		 112 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
113 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 114}
115
ac27a0ec 116/*
dab291af 117 * Wrappers for jbd2_journal_start/end.
ac27a0ec
DK		 118 *
119 * The only special thing we need to do here is to make sure that all
120 * journal_end calls result in the superblock being marked dirty, so
121 * that sync() will call the filesystem's write_super callback if
122 * appropriate.
123 */
617ba13b 124handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
ac27a0ec
DK		 125{
126 journal_t *journal;
127
128 if (sb->s_flags & MS_RDONLY)
129 return ERR_PTR(-EROFS);
130
131 /* Special case here: if the journal has aborted behind our
132 * backs (eg. EIO in the commit thread), then we still need to
133 * take the FS itself readonly cleanly. */
617ba13b 134 journal = EXT4_SB(sb)->s_journal;
ac27a0ec 135 if (is_journal_aborted(journal)) {
617ba13b 136 ext4_abort(sb, __FUNCTION__,
ac27a0ec
DK		 137 "Detected aborted journal");
138 return ERR_PTR(-EROFS);
139 }
140
dab291af 141 return jbd2_journal_start(journal, nblocks);
ac27a0ec
DK		 142}
143
144/*
145 * The only special thing we need to do here is to make sure that all
dab291af 146 * jbd2_journal_stop calls result in the superblock being marked dirty, so
ac27a0ec
DK		 147 * that sync() will call the filesystem's write_super callback if
148 * appropriate.
149 */
617ba13b 150int __ext4_journal_stop(const char *where, handle_t *handle)
ac27a0ec
DK		 151{
152 struct super_block *sb;
153 int err;
154 int rc;
155
156 sb = handle->h_transaction->t_journal->j_private;
157 err = handle->h_err;
dab291af 158 rc = jbd2_journal_stop(handle);
ac27a0ec
DK		 159
160 if (!err)
161 err = rc;
162 if (err)
617ba13b 163 __ext4_std_error(sb, where, err);
ac27a0ec
DK		 164 return err;
165}
166
617ba13b 167void ext4_journal_abort_handle(const char *caller, const char *err_fn,
ac27a0ec
DK		 168 struct buffer_head *bh, handle_t *handle, int err)
169{
170 char nbuf[16];
617ba13b 171 const char *errstr = ext4_decode_error(NULL, err, nbuf);
ac27a0ec
DK		 172
173 if (bh)
174 BUFFER_TRACE(bh, "abort");
175
176 if (!handle->h_err)
177 handle->h_err = err;
178
179 if (is_handle_aborted(handle))
180 return;
181
182 printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
183 caller, errstr, err_fn);
184
dab291af 185 jbd2_journal_abort_handle(handle);
ac27a0ec
DK		 186}
187
188/* Deal with the reporting of failure conditions on a filesystem such as
189 * inconsistencies detected or read IO failures.
190 *
191 * On ext2, we can store the error state of the filesystem in the
617ba13b 192 * superblock. That is not possible on ext4, because we may have other
ac27a0ec
DK		 193 * write ordering constraints on the superblock which prevent us from
194 * writing it out straight away; and given that the journal is about to
195 * be aborted, we can't rely on the current, or future, transactions to
196 * write out the superblock safely.
197 *
dab291af 198 * We'll just use the jbd2_journal_abort() error code to record an error in
ac27a0ec
DK		 199 * the journal instead. On recovery, the journal will compain about
200 * that error until we've noted it down and cleared it.
201 */
202
617ba13b 203static void ext4_handle_error(struct super_block *sb)
ac27a0ec 204{
617ba13b 205 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 206
617ba13b
MC		 207 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
208 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
ac27a0ec
DK		 209
210 if (sb->s_flags & MS_RDONLY)
211 return;
212
213 if (!test_opt (sb, ERRORS_CONT)) {
617ba13b 214 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 215
617ba13b 216 EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
ac27a0ec 217 if (journal)
dab291af 218 jbd2_journal_abort(journal, -EIO);
ac27a0ec
DK		 219 }
220 if (test_opt (sb, ERRORS_RO)) {
221 printk (KERN_CRIT "Remounting filesystem read-only\n");
222 sb->s_flags |= MS_RDONLY;
223 }
617ba13b 224 ext4_commit_super(sb, es, 1);
ac27a0ec 225 if (test_opt(sb, ERRORS_PANIC))
617ba13b 226 panic("EXT4-fs (device %s): panic forced after error\n",
ac27a0ec
DK		 227 sb->s_id);
228}
229
617ba13b 230void ext4_error (struct super_block * sb, const char * function,
ac27a0ec
DK		 231 const char * fmt, ...)
232{
233 va_list args;
234
235 va_start(args, fmt);
617ba13b 236 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ",sb->s_id, function);
ac27a0ec
DK		 237 vprintk(fmt, args);
238 printk("\n");
239 va_end(args);
240
617ba13b 241 ext4_handle_error(sb);
ac27a0ec
DK		 242}
243
617ba13b 244static const char *ext4_decode_error(struct super_block * sb, int errno,
ac27a0ec
DK		 245 char nbuf[16])
246{
247 char *errstr = NULL;
248
249 switch (errno) {
250 case -EIO:
251 errstr = "IO failure";
252 break;
253 case -ENOMEM:
254 errstr = "Out of memory";
255 break;
256 case -EROFS:
dab291af 257 if (!sb || EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT)
ac27a0ec
DK		 258 errstr = "Journal has aborted";
259 else
260 errstr = "Readonly filesystem";
261 break;
262 default:
263 /* If the caller passed in an extra buffer for unknown
264 * errors, textualise them now. Else we just return
265 * NULL. */
266 if (nbuf) {
267 /* Check for truncated error codes... */
268 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
269 errstr = nbuf;
270 }
271 break;
272 }
273
274 return errstr;
275}
276
617ba13b 277/* __ext4_std_error decodes expected errors from journaling functions
ac27a0ec
DK		 278 * automatically and invokes the appropriate error response. */
279
617ba13b 280void __ext4_std_error (struct super_block * sb, const char * function,
ac27a0ec
DK		 281 int errno)
282{
283 char nbuf[16];
284 const char *errstr;
285
286 /* Special case: if the error is EROFS, and we're not already
287 * inside a transaction, then there's really no point in logging
288 * an error. */
289 if (errno == -EROFS && journal_current_handle() == NULL &&
290 (sb->s_flags & MS_RDONLY))
291 return;
292
617ba13b
MC		 293 errstr = ext4_decode_error(sb, errno, nbuf);
294 printk (KERN_CRIT "EXT4-fs error (device %s) in %s: %s\n",
ac27a0ec
DK		 295 sb->s_id, function, errstr);
296
617ba13b 297 ext4_handle_error(sb);
ac27a0ec
DK		 298}
299
300/*
617ba13b 301 * ext4_abort is a much stronger failure handler than ext4_error. The
ac27a0ec
DK		 302 * abort function may be used to deal with unrecoverable failures such
303 * as journal IO errors or ENOMEM at a critical moment in log management.
304 *
305 * We unconditionally force the filesystem into an ABORT|READONLY state,
306 * unless the error response on the fs has been set to panic in which
307 * case we take the easy way out and panic immediately.
308 */
309
617ba13b 310void ext4_abort (struct super_block * sb, const char * function,
ac27a0ec
DK		 311 const char * fmt, ...)
312{
313 va_list args;
314
617ba13b 315 printk (KERN_CRIT "ext4_abort called.\n");
ac27a0ec
DK		 316
317 va_start(args, fmt);
617ba13b 318 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ",sb->s_id, function);
ac27a0ec
DK		 319 vprintk(fmt, args);
320 printk("\n");
321 va_end(args);
322
323 if (test_opt(sb, ERRORS_PANIC))
617ba13b 324 panic("EXT4-fs panic from previous error\n");
ac27a0ec
DK		 325
326 if (sb->s_flags & MS_RDONLY)
327 return;
328
329 printk(KERN_CRIT "Remounting filesystem read-only\n");
617ba13b 330 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
ac27a0ec 331 sb->s_flags |= MS_RDONLY;
617ba13b 332 EXT4_SB(sb)->s_mount_opt |= EXT4_MOUNT_ABORT;
dab291af 333 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
ac27a0ec
DK		 334}
335
617ba13b 336void ext4_warning (struct super_block * sb, const char * function,
ac27a0ec
DK		 337 const char * fmt, ...)
338{
339 va_list args;
340
341 va_start(args, fmt);
617ba13b 342 printk(KERN_WARNING "EXT4-fs warning (device %s): %s: ",
ac27a0ec
DK		 343 sb->s_id, function);
344 vprintk(fmt, args);
345 printk("\n");
346 va_end(args);
347}
348
617ba13b 349void ext4_update_dynamic_rev(struct super_block *sb)
ac27a0ec 350{
617ba13b 351 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 352
617ba13b 353 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
ac27a0ec
DK		 354 return;
355
617ba13b 356 ext4_warning(sb, __FUNCTION__,
ac27a0ec
DK		 357 "updating to rev %d because of new feature flag, "
358 "running e2fsck is recommended",
617ba13b 359 EXT4_DYNAMIC_REV);
ac27a0ec 360
617ba13b
MC		 361 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
362 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
363 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
ac27a0ec
DK		 364 /* leave es->s_feature_*compat flags alone */
365 /* es->s_uuid will be set by e2fsck if empty */
366
367 /*
368 * The rest of the superblock fields should be zero, and if not it
369 * means they are likely already in use, so leave them alone. We
370 * can leave it up to e2fsck to clean up any inconsistencies there.
371 */
372}
373
374/*
375 * Open the external journal device
376 */
617ba13b 377static struct block_device *ext4_blkdev_get(dev_t dev)
ac27a0ec
DK		 378{
379 struct block_device *bdev;
380 char b[BDEVNAME_SIZE];
381
382 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
383 if (IS_ERR(bdev))
384 goto fail;
385 return bdev;
386
387fail:
617ba13b 388 printk(KERN_ERR "EXT4: failed to open journal device %s: %ld\n",
ac27a0ec
DK		 389 __bdevname(dev, b), PTR_ERR(bdev));
390 return NULL;
391}
392
393/*
394 * Release the journal device
395 */
617ba13b 396static int ext4_blkdev_put(struct block_device *bdev)
ac27a0ec
DK		 397{
398 bd_release(bdev);
399 return blkdev_put(bdev);
400}
401
617ba13b 402static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
ac27a0ec
DK		 403{
404 struct block_device *bdev;
405 int ret = -ENODEV;
406
407 bdev = sbi->journal_bdev;
408 if (bdev) {
617ba13b 409 ret = ext4_blkdev_put(bdev);
ac27a0ec
DK		 410 sbi->journal_bdev = NULL;
411 }
412 return ret;
413}
414
415static inline struct inode *orphan_list_entry(struct list_head *l)
416{
617ba13b 417 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
ac27a0ec
DK		 418}
419
617ba13b 420static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
ac27a0ec
DK		 421{
422 struct list_head *l;
423
424 printk(KERN_ERR "sb orphan head is %d\n",
425 le32_to_cpu(sbi->s_es->s_last_orphan));
426
427 printk(KERN_ERR "sb_info orphan list:\n");
428 list_for_each(l, &sbi->s_orphan) {
429 struct inode *inode = orphan_list_entry(l);
430 printk(KERN_ERR " "
431 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
432 inode->i_sb->s_id, inode->i_ino, inode,
433 inode->i_mode, inode->i_nlink,
434 NEXT_ORPHAN(inode));
435 }
436}
437
617ba13b 438static void ext4_put_super (struct super_block * sb)
ac27a0ec 439{
617ba13b
MC		 440 struct ext4_sb_info *sbi = EXT4_SB(sb);
441 struct ext4_super_block *es = sbi->s_es;
ac27a0ec
DK		 442 int i;
443
a86c6181 444 ext4_ext_release(sb);
617ba13b 445 ext4_xattr_put_super(sb);
dab291af 446 jbd2_journal_destroy(sbi->s_journal);
ac27a0ec 447 if (!(sb->s_flags & MS_RDONLY)) {
617ba13b 448 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec
DK		 449 es->s_state = cpu_to_le16(sbi->s_mount_state);
450 BUFFER_TRACE(sbi->s_sbh, "marking dirty");
451 mark_buffer_dirty(sbi->s_sbh);
617ba13b 452 ext4_commit_super(sb, es, 1);
ac27a0ec
DK		 453 }
454
455 for (i = 0; i < sbi->s_gdb_count; i++)
456 brelse(sbi->s_group_desc[i]);
457 kfree(sbi->s_group_desc);
458 percpu_counter_destroy(&sbi->s_freeblocks_counter);
459 percpu_counter_destroy(&sbi->s_freeinodes_counter);
460 percpu_counter_destroy(&sbi->s_dirs_counter);
461 brelse(sbi->s_sbh);
462#ifdef CONFIG_QUOTA
463 for (i = 0; i < MAXQUOTAS; i++)
464 kfree(sbi->s_qf_names[i]);
465#endif
466
467 /* Debugging code just in case the in-memory inode orphan list
468 * isn't empty. The on-disk one can be non-empty if we've
469 * detected an error and taken the fs readonly, but the
470 * in-memory list had better be clean by this point. */
471 if (!list_empty(&sbi->s_orphan))
472 dump_orphan_list(sb, sbi);
473 J_ASSERT(list_empty(&sbi->s_orphan));
474
f98393a6 475 invalidate_bdev(sb->s_bdev);
ac27a0ec
DK		 476 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
477 /*
478 * Invalidate the journal device's buffers. We don't want them
479 * floating about in memory - the physical journal device may
480 * hotswapped, and it breaks the `ro-after' testing code.
481 */
482 sync_blockdev(sbi->journal_bdev);
f98393a6 483 invalidate_bdev(sbi->journal_bdev);
617ba13b 484 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 485 }
486 sb->s_fs_info = NULL;
487 kfree(sbi);
488 return;
489}
490
e18b890b 491static struct kmem_cache *ext4_inode_cachep;
ac27a0ec
DK		 492
493/*
494 * Called inside transaction, so use GFP_NOFS
495 */
617ba13b 496static struct inode *ext4_alloc_inode(struct super_block *sb)
ac27a0ec 497{
617ba13b 498 struct ext4_inode_info *ei;
ac27a0ec 499
e6b4f8da 500 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
ac27a0ec
DK		 501 if (!ei)
502 return NULL;
617ba13b
MC		 503#ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
504 ei->i_acl = EXT4_ACL_NOT_CACHED;
505 ei->i_default_acl = EXT4_ACL_NOT_CACHED;
ac27a0ec
DK		 506#endif
507 ei->i_block_alloc_info = NULL;
508 ei->vfs_inode.i_version = 1;
a86c6181 509 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
ac27a0ec
DK		 510 return &ei->vfs_inode;
511}
512
617ba13b 513static void ext4_destroy_inode(struct inode *inode)
ac27a0ec 514{
9f7dd93d
VA		 515 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
516 printk("EXT4 Inode %p: orphan list check failed!\n",
517 EXT4_I(inode));
518 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
519 EXT4_I(inode), sizeof(struct ext4_inode_info),
520 true);
521 dump_stack();
522 }
617ba13b 523 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
ac27a0ec
DK		 524}
525
4ba9b9d0 526static void init_once(struct kmem_cache *cachep, void *foo)
ac27a0ec 527{
617ba13b 528 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
ac27a0ec 529
a35afb83 530 INIT_LIST_HEAD(&ei->i_orphan);
617ba13b 531#ifdef CONFIG_EXT4DEV_FS_XATTR
a35afb83 532 init_rwsem(&ei->xattr_sem);
ac27a0ec 533#endif
a35afb83
CL		 534 mutex_init(&ei->truncate_mutex);
535 inode_init_once(&ei->vfs_inode);
ac27a0ec
DK		 536}
537
538static int init_inodecache(void)
539{
617ba13b
MC		 540 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
541 sizeof(struct ext4_inode_info),
ac27a0ec
DK		 542 0, (SLAB_RECLAIM_ACCOUNT|
543 SLAB_MEM_SPREAD),
20c2df83 544 init_once);
617ba13b 545 if (ext4_inode_cachep == NULL)
ac27a0ec
DK		 546 return -ENOMEM;
547 return 0;
548}
549
550static void destroy_inodecache(void)
551{
617ba13b 552 kmem_cache_destroy(ext4_inode_cachep);
ac27a0ec
DK		 553}
554
617ba13b 555static void ext4_clear_inode(struct inode *inode)
ac27a0ec 556{
617ba13b
MC		 557 struct ext4_block_alloc_info *rsv = EXT4_I(inode)->i_block_alloc_info;
558#ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
559 if (EXT4_I(inode)->i_acl &&
560 EXT4_I(inode)->i_acl != EXT4_ACL_NOT_CACHED) {
561 posix_acl_release(EXT4_I(inode)->i_acl);
562 EXT4_I(inode)->i_acl = EXT4_ACL_NOT_CACHED;
563 }
564 if (EXT4_I(inode)->i_default_acl &&
565 EXT4_I(inode)->i_default_acl != EXT4_ACL_NOT_CACHED) {
566 posix_acl_release(EXT4_I(inode)->i_default_acl);
567 EXT4_I(inode)->i_default_acl = EXT4_ACL_NOT_CACHED;
ac27a0ec
DK		 568 }
569#endif
617ba13b
MC		 570 ext4_discard_reservation(inode);
571 EXT4_I(inode)->i_block_alloc_info = NULL;
ac27a0ec
DK		 572 if (unlikely(rsv))
573 kfree(rsv);
574}
575
617ba13b 576static inline void ext4_show_quota_options(struct seq_file *seq, struct super_block *sb)
ac27a0ec
DK		 577{
578#if defined(CONFIG_QUOTA)
617ba13b 579 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 580
581 if (sbi->s_jquota_fmt)
582 seq_printf(seq, ",jqfmt=%s",
583 (sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold": "vfsv0");
584
585 if (sbi->s_qf_names[USRQUOTA])
586 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
587
588 if (sbi->s_qf_names[GRPQUOTA])
589 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
590
617ba13b 591 if (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA)
ac27a0ec
DK		 592 seq_puts(seq, ",usrquota");
593
617ba13b 594 if (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)
ac27a0ec
DK		 595 seq_puts(seq, ",grpquota");
596#endif
597}
598
d9c9bef1
MS		 599/*
600 * Show an option if
601 * - it's set to a non-default value OR
602 * - if the per-sb default is different from the global default
603 */
617ba13b 604static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
ac27a0ec
DK		 605{
606 struct super_block *sb = vfs->mnt_sb;
d9c9bef1
MS		 607 struct ext4_sb_info *sbi = EXT4_SB(sb);
608 struct ext4_super_block *es = sbi->s_es;
609 unsigned long def_mount_opts;
610
611 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
612
613 if (sbi->s_sb_block != 1)
614 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
615 if (test_opt(sb, MINIX_DF))
616 seq_puts(seq, ",minixdf");
617 if (test_opt(sb, GRPID))
618 seq_puts(seq, ",grpid");
619 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
620 seq_puts(seq, ",nogrpid");
621 if (sbi->s_resuid != EXT4_DEF_RESUID ||
622 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
623 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
624 }
625 if (sbi->s_resgid != EXT4_DEF_RESGID ||
626 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
627 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
628 }
629 if (test_opt(sb, ERRORS_CONT)) {
630 int def_errors = le16_to_cpu(es->s_errors);
631
632 if (def_errors == EXT4_ERRORS_PANIC ||
633 def_errors == EXT4_ERRORS_RO) {
634 seq_puts(seq, ",errors=continue");
635 }
636 }
637 if (test_opt(sb, ERRORS_RO))
638 seq_puts(seq, ",errors=remount-ro");
639 if (test_opt(sb, ERRORS_PANIC))
640 seq_puts(seq, ",errors=panic");
641 if (test_opt(sb, NO_UID32))
642 seq_puts(seq, ",nouid32");
643 if (test_opt(sb, DEBUG))
644 seq_puts(seq, ",debug");
645 if (test_opt(sb, OLDALLOC))
646 seq_puts(seq, ",oldalloc");
647#ifdef CONFIG_EXT4_FS_XATTR
648 if (test_opt(sb, XATTR_USER))
649 seq_puts(seq, ",user_xattr");
650 if (!test_opt(sb, XATTR_USER) &&
651 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
652 seq_puts(seq, ",nouser_xattr");
653 }
654#endif
655#ifdef CONFIG_EXT4_FS_POSIX_ACL
656 if (test_opt(sb, POSIX_ACL))
657 seq_puts(seq, ",acl");
658 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
659 seq_puts(seq, ",noacl");
660#endif
661 if (!test_opt(sb, RESERVATION))
662 seq_puts(seq, ",noreservation");
663 if (sbi->s_commit_interval) {
664 seq_printf(seq, ",commit=%u",
665 (unsigned) (sbi->s_commit_interval / HZ));
666 }
667 if (test_opt(sb, BARRIER))
668 seq_puts(seq, ",barrier=1");
669 if (test_opt(sb, NOBH))
670 seq_puts(seq, ",nobh");
671 if (!test_opt(sb, EXTENTS))
672 seq_puts(seq, ",noextents");
ac27a0ec 673
617ba13b 674 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
ac27a0ec 675 seq_puts(seq, ",data=journal");
617ba13b 676 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
ac27a0ec 677 seq_puts(seq, ",data=ordered");
617ba13b 678 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
ac27a0ec
DK		 679 seq_puts(seq, ",data=writeback");
680
617ba13b 681 ext4_show_quota_options(seq, sb);
ac27a0ec
DK		 682
683 return 0;
684}
685
686
617ba13b 687static struct dentry *ext4_get_dentry(struct super_block *sb, void *vobjp)
ac27a0ec
DK		 688{
689 __u32 *objp = vobjp;
690 unsigned long ino = objp[0];
691 __u32 generation = objp[1];
692 struct inode *inode;
693 struct dentry *result;
694
617ba13b 695 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
ac27a0ec 696 return ERR_PTR(-ESTALE);
617ba13b 697 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
ac27a0ec
DK		 698 return ERR_PTR(-ESTALE);
699
700 /* iget isn't really right if the inode is currently unallocated!!
701 *
617ba13b 702 * ext4_read_inode will return a bad_inode if the inode had been
ac27a0ec
DK		 703 * deleted, so we should be safe.
704 *
705 * Currently we don't know the generation for parent directory, so
706 * a generation of 0 means "accept any"
707 */
708 inode = iget(sb, ino);
709 if (inode == NULL)
710 return ERR_PTR(-ENOMEM);
711 if (is_bad_inode(inode) ||
712 (generation && inode->i_generation != generation)) {
713 iput(inode);
714 return ERR_PTR(-ESTALE);
715 }
716 /* now to find a dentry.
717 * If possible, get a well-connected one
718 */
719 result = d_alloc_anon(inode);
720 if (!result) {
721 iput(inode);
722 return ERR_PTR(-ENOMEM);
723 }
724 return result;
725}
726
727#ifdef CONFIG_QUOTA
728#define QTYPE2NAME(t) ((t)==USRQUOTA?"user":"group")
729#define QTYPE2MOPT(on, t) ((t)==USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
730
617ba13b
MC		 731static int ext4_dquot_initialize(struct inode *inode, int type);
732static int ext4_dquot_drop(struct inode *inode);
733static int ext4_write_dquot(struct dquot *dquot);
734static int ext4_acquire_dquot(struct dquot *dquot);
735static int ext4_release_dquot(struct dquot *dquot);
736static int ext4_mark_dquot_dirty(struct dquot *dquot);
737static int ext4_write_info(struct super_block *sb, int type);
738static int ext4_quota_on(struct super_block *sb, int type, int format_id, char *path);
739static int ext4_quota_on_mount(struct super_block *sb, int type);
740static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 741 size_t len, loff_t off);
617ba13b 742static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 743 const char *data, size_t len, loff_t off);
744
617ba13b
MC		 745static struct dquot_operations ext4_quota_operations = {
746 .initialize = ext4_dquot_initialize,
747 .drop = ext4_dquot_drop,
ac27a0ec
DK		 748 .alloc_space = dquot_alloc_space,
749 .alloc_inode = dquot_alloc_inode,
750 .free_space = dquot_free_space,
751 .free_inode = dquot_free_inode,
752 .transfer = dquot_transfer,
617ba13b
MC		 753 .write_dquot = ext4_write_dquot,
754 .acquire_dquot = ext4_acquire_dquot,
755 .release_dquot = ext4_release_dquot,
756 .mark_dirty = ext4_mark_dquot_dirty,
757 .write_info = ext4_write_info
ac27a0ec
DK		 758};
759
617ba13b
MC		 760static struct quotactl_ops ext4_qctl_operations = {
761 .quota_on = ext4_quota_on,
ac27a0ec
DK		 762 .quota_off = vfs_quota_off,
763 .quota_sync = vfs_quota_sync,
764 .get_info = vfs_get_dqinfo,
765 .set_info = vfs_set_dqinfo,
766 .get_dqblk = vfs_get_dqblk,
767 .set_dqblk = vfs_set_dqblk
768};
769#endif
770
ee9b6d61 771static const struct super_operations ext4_sops = {
617ba13b
MC		 772 .alloc_inode = ext4_alloc_inode,
773 .destroy_inode = ext4_destroy_inode,
774 .read_inode = ext4_read_inode,
775 .write_inode = ext4_write_inode,
776 .dirty_inode = ext4_dirty_inode,
777 .delete_inode = ext4_delete_inode,
778 .put_super = ext4_put_super,
779 .write_super = ext4_write_super,
780 .sync_fs = ext4_sync_fs,
781 .write_super_lockfs = ext4_write_super_lockfs,
782 .unlockfs = ext4_unlockfs,
783 .statfs = ext4_statfs,
784 .remount_fs = ext4_remount,
785 .clear_inode = ext4_clear_inode,
786 .show_options = ext4_show_options,
ac27a0ec 787#ifdef CONFIG_QUOTA
617ba13b
MC		 788 .quota_read = ext4_quota_read,
789 .quota_write = ext4_quota_write,
ac27a0ec
DK		 790#endif
791};
792
617ba13b
MC		 793static struct export_operations ext4_export_ops = {
794 .get_parent = ext4_get_parent,
795 .get_dentry = ext4_get_dentry,
ac27a0ec
DK		 796};
797
798enum {
799 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
800 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
801 Opt_nouid32, Opt_nocheck, Opt_debug, Opt_oldalloc, Opt_orlov,
802 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
803 Opt_reservation, Opt_noreservation, Opt_noload, Opt_nobh, Opt_bh,
804 Opt_commit, Opt_journal_update, Opt_journal_inum, Opt_journal_dev,
805 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
806 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
807 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
808 Opt_ignore, Opt_barrier, Opt_err, Opt_resize, Opt_usrquota,
1e2462f9 809 Opt_grpquota, Opt_extents, Opt_noextents,
ac27a0ec
DK		 810};
811
812static match_table_t tokens = {
813 {Opt_bsd_df, "bsddf"},
814 {Opt_minix_df, "minixdf"},
815 {Opt_grpid, "grpid"},
816 {Opt_grpid, "bsdgroups"},
817 {Opt_nogrpid, "nogrpid"},
818 {Opt_nogrpid, "sysvgroups"},
819 {Opt_resgid, "resgid=%u"},
820 {Opt_resuid, "resuid=%u"},
821 {Opt_sb, "sb=%u"},
822 {Opt_err_cont, "errors=continue"},
823 {Opt_err_panic, "errors=panic"},
824 {Opt_err_ro, "errors=remount-ro"},
825 {Opt_nouid32, "nouid32"},
826 {Opt_nocheck, "nocheck"},
827 {Opt_nocheck, "check=none"},
828 {Opt_debug, "debug"},
829 {Opt_oldalloc, "oldalloc"},
830 {Opt_orlov, "orlov"},
831 {Opt_user_xattr, "user_xattr"},
832 {Opt_nouser_xattr, "nouser_xattr"},
833 {Opt_acl, "acl"},
834 {Opt_noacl, "noacl"},
835 {Opt_reservation, "reservation"},
836 {Opt_noreservation, "noreservation"},
837 {Opt_noload, "noload"},
838 {Opt_nobh, "nobh"},
839 {Opt_bh, "bh"},
840 {Opt_commit, "commit=%u"},
841 {Opt_journal_update, "journal=update"},
842 {Opt_journal_inum, "journal=%u"},
843 {Opt_journal_dev, "journal_dev=%u"},
844 {Opt_abort, "abort"},
845 {Opt_data_journal, "data=journal"},
846 {Opt_data_ordered, "data=ordered"},
847 {Opt_data_writeback, "data=writeback"},
848 {Opt_offusrjquota, "usrjquota="},
849 {Opt_usrjquota, "usrjquota=%s"},
850 {Opt_offgrpjquota, "grpjquota="},
851 {Opt_grpjquota, "grpjquota=%s"},
852 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
853 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
854 {Opt_grpquota, "grpquota"},
855 {Opt_noquota, "noquota"},
856 {Opt_quota, "quota"},
857 {Opt_usrquota, "usrquota"},
858 {Opt_barrier, "barrier=%u"},
a86c6181 859 {Opt_extents, "extents"},
1e2462f9 860 {Opt_noextents, "noextents"},
ac27a0ec
DK		 861 {Opt_err, NULL},
862 {Opt_resize, "resize"},
863};
864
617ba13b 865static ext4_fsblk_t get_sb_block(void **data)
ac27a0ec 866{
617ba13b 867 ext4_fsblk_t sb_block;
ac27a0ec
DK		 868 char *options = (char *) *data;
869
870 if (!options || strncmp(options, "sb=", 3) != 0)
871 return 1; /* Default location */
872 options += 3;
617ba13b 873 /*todo: use simple_strtoll with >32bit ext4 */
ac27a0ec
DK		 874 sb_block = simple_strtoul(options, &options, 0);
875 if (*options && *options != ',') {
617ba13b 876 printk("EXT4-fs: Invalid sb specification: %s\n",
ac27a0ec
DK		 877 (char *) *data);
878 return 1;
879 }
880 if (*options == ',')
881 options++;
882 *data = (void *) options;
883 return sb_block;
884}
885
886static int parse_options (char *options, struct super_block *sb,
887 unsigned int *inum, unsigned long *journal_devnum,
617ba13b 888 ext4_fsblk_t *n_blocks_count, int is_remount)
ac27a0ec 889{
617ba13b 890 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 891 char * p;
892 substring_t args[MAX_OPT_ARGS];
893 int data_opt = 0;
894 int option;
895#ifdef CONFIG_QUOTA
896 int qtype;
897 char *qname;
898#endif
899
900 if (!options)
901 return 1;
902
903 while ((p = strsep (&options, ",")) != NULL) {
904 int token;
905 if (!*p)
906 continue;
907
908 token = match_token(p, tokens, args);
909 switch (token) {
910 case Opt_bsd_df:
911 clear_opt (sbi->s_mount_opt, MINIX_DF);
912 break;
913 case Opt_minix_df:
914 set_opt (sbi->s_mount_opt, MINIX_DF);
915 break;
916 case Opt_grpid:
917 set_opt (sbi->s_mount_opt, GRPID);
918 break;
919 case Opt_nogrpid:
920 clear_opt (sbi->s_mount_opt, GRPID);
921 break;
922 case Opt_resuid:
923 if (match_int(&args[0], &option))
924 return 0;
925 sbi->s_resuid = option;
926 break;
927 case Opt_resgid:
928 if (match_int(&args[0], &option))
929 return 0;
930 sbi->s_resgid = option;
931 break;
932 case Opt_sb:
933 /* handled by get_sb_block() instead of here */
934 /* *sb_block = match_int(&args[0]); */
935 break;
936 case Opt_err_panic:
937 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
938 clear_opt (sbi->s_mount_opt, ERRORS_RO);
939 set_opt (sbi->s_mount_opt, ERRORS_PANIC);
940 break;
941 case Opt_err_ro:
942 clear_opt (sbi->s_mount_opt, ERRORS_CONT);
943 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
944 set_opt (sbi->s_mount_opt, ERRORS_RO);
945 break;
946 case Opt_err_cont:
947 clear_opt (sbi->s_mount_opt, ERRORS_RO);
948 clear_opt (sbi->s_mount_opt, ERRORS_PANIC);
949 set_opt (sbi->s_mount_opt, ERRORS_CONT);
950 break;
951 case Opt_nouid32:
952 set_opt (sbi->s_mount_opt, NO_UID32);
953 break;
954 case Opt_nocheck:
955 clear_opt (sbi->s_mount_opt, CHECK);
956 break;
957 case Opt_debug:
958 set_opt (sbi->s_mount_opt, DEBUG);
959 break;
960 case Opt_oldalloc:
961 set_opt (sbi->s_mount_opt, OLDALLOC);
962 break;
963 case Opt_orlov:
964 clear_opt (sbi->s_mount_opt, OLDALLOC);
965 break;
617ba13b 966#ifdef CONFIG_EXT4DEV_FS_XATTR
ac27a0ec
DK		 967 case Opt_user_xattr:
968 set_opt (sbi->s_mount_opt, XATTR_USER);
969 break;
970 case Opt_nouser_xattr:
971 clear_opt (sbi->s_mount_opt, XATTR_USER);
972 break;
973#else
974 case Opt_user_xattr:
975 case Opt_nouser_xattr:
617ba13b 976 printk("EXT4 (no)user_xattr options not supported\n");
ac27a0ec
DK		 977 break;
978#endif
617ba13b 979#ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
ac27a0ec
DK		 980 case Opt_acl:
981 set_opt(sbi->s_mount_opt, POSIX_ACL);
982 break;
983 case Opt_noacl:
984 clear_opt(sbi->s_mount_opt, POSIX_ACL);
985 break;
986#else
987 case Opt_acl:
988 case Opt_noacl:
617ba13b 989 printk("EXT4 (no)acl options not supported\n");
ac27a0ec
DK		 990 break;
991#endif
992 case Opt_reservation:
993 set_opt(sbi->s_mount_opt, RESERVATION);
994 break;
995 case Opt_noreservation:
996 clear_opt(sbi->s_mount_opt, RESERVATION);
997 break;
998 case Opt_journal_update:
999 /* @@@ FIXME */
1000 /* Eventually we will want to be able to create
1001 a journal file here. For now, only allow the
1002 user to specify an existing inode to be the
1003 journal file. */
1004 if (is_remount) {
617ba13b 1005 printk(KERN_ERR "EXT4-fs: cannot specify "
ac27a0ec
DK		 1006 "journal on remount\n");
1007 return 0;
1008 }
1009 set_opt (sbi->s_mount_opt, UPDATE_JOURNAL);
1010 break;
1011 case Opt_journal_inum:
1012 if (is_remount) {
617ba13b 1013 printk(KERN_ERR "EXT4-fs: cannot specify "
ac27a0ec
DK		 1014 "journal on remount\n");
1015 return 0;
1016 }
1017 if (match_int(&args[0], &option))
1018 return 0;
1019 *inum = option;
1020 break;
1021 case Opt_journal_dev:
1022 if (is_remount) {
617ba13b 1023 printk(KERN_ERR "EXT4-fs: cannot specify "
ac27a0ec
DK		 1024 "journal on remount\n");
1025 return 0;
1026 }
1027 if (match_int(&args[0], &option))
1028 return 0;
1029 *journal_devnum = option;
1030 break;
1031 case Opt_noload:
1032 set_opt (sbi->s_mount_opt, NOLOAD);
1033 break;
1034 case Opt_commit:
1035 if (match_int(&args[0], &option))
1036 return 0;
1037 if (option < 0)
1038 return 0;
1039 if (option == 0)
1040 option = JBD_DEFAULT_MAX_COMMIT_AGE;
1041 sbi->s_commit_interval = HZ * option;
1042 break;
1043 case Opt_data_journal:
617ba13b 1044 data_opt = EXT4_MOUNT_JOURNAL_DATA;
ac27a0ec
DK		 1045 goto datacheck;
1046 case Opt_data_ordered:
617ba13b 1047 data_opt = EXT4_MOUNT_ORDERED_DATA;
ac27a0ec
DK		 1048 goto datacheck;
1049 case Opt_data_writeback:
617ba13b 1050 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
ac27a0ec
DK		 1051 datacheck:
1052 if (is_remount) {
617ba13b 1053 if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
ac27a0ec
DK		 1054 != data_opt) {
1055 printk(KERN_ERR
617ba13b 1056 "EXT4-fs: cannot change data "
ac27a0ec
DK		 1057 "mode on remount\n");
1058 return 0;
1059 }
1060 } else {
617ba13b 1061 sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
ac27a0ec
DK		 1062 sbi->s_mount_opt |= data_opt;
1063 }
1064 break;
1065#ifdef CONFIG_QUOTA
1066 case Opt_usrjquota:
1067 qtype = USRQUOTA;
1068 goto set_qf_name;
1069 case Opt_grpjquota:
1070 qtype = GRPQUOTA;
1071set_qf_name:
1072 if (sb_any_quota_enabled(sb)) {
1073 printk(KERN_ERR
617ba13b 1074 "EXT4-fs: Cannot change journalled "
ac27a0ec
DK		 1075 "quota options when quota turned on.\n");
1076 return 0;
1077 }
1078 qname = match_strdup(&args[0]);
1079 if (!qname) {
1080 printk(KERN_ERR
617ba13b 1081 "EXT4-fs: not enough memory for "
ac27a0ec
DK		 1082 "storing quotafile name.\n");
1083 return 0;
1084 }
1085 if (sbi->s_qf_names[qtype] &&
1086 strcmp(sbi->s_qf_names[qtype], qname)) {
1087 printk(KERN_ERR
617ba13b 1088 "EXT4-fs: %s quota file already "
ac27a0ec
DK		 1089 "specified.\n", QTYPE2NAME(qtype));
1090 kfree(qname);
1091 return 0;
1092 }
1093 sbi->s_qf_names[qtype] = qname;
1094 if (strchr(sbi->s_qf_names[qtype], '/')) {
1095 printk(KERN_ERR
617ba13b 1096 "EXT4-fs: quotafile must be on "
ac27a0ec
DK		 1097 "filesystem root.\n");
1098 kfree(sbi->s_qf_names[qtype]);
1099 sbi->s_qf_names[qtype] = NULL;
1100 return 0;
1101 }
1102 set_opt(sbi->s_mount_opt, QUOTA);
1103 break;
1104 case Opt_offusrjquota:
1105 qtype = USRQUOTA;
1106 goto clear_qf_name;
1107 case Opt_offgrpjquota:
1108 qtype = GRPQUOTA;
1109clear_qf_name:
1110 if (sb_any_quota_enabled(sb)) {
617ba13b 1111 printk(KERN_ERR "EXT4-fs: Cannot change "
ac27a0ec
DK		 1112 "journalled quota options when "
1113 "quota turned on.\n");
1114 return 0;
1115 }
1116 /*
1117 * The space will be released later when all options
1118 * are confirmed to be correct
1119 */
1120 sbi->s_qf_names[qtype] = NULL;
1121 break;
1122 case Opt_jqfmt_vfsold:
1123 sbi->s_jquota_fmt = QFMT_VFS_OLD;
1124 break;
1125 case Opt_jqfmt_vfsv0:
1126 sbi->s_jquota_fmt = QFMT_VFS_V0;
1127 break;
1128 case Opt_quota:
1129 case Opt_usrquota:
1130 set_opt(sbi->s_mount_opt, QUOTA);
1131 set_opt(sbi->s_mount_opt, USRQUOTA);
1132 break;
1133 case Opt_grpquota:
1134 set_opt(sbi->s_mount_opt, QUOTA);
1135 set_opt(sbi->s_mount_opt, GRPQUOTA);
1136 break;
1137 case Opt_noquota:
1138 if (sb_any_quota_enabled(sb)) {
617ba13b 1139 printk(KERN_ERR "EXT4-fs: Cannot change quota "
ac27a0ec
DK		 1140 "options when quota turned on.\n");
1141 return 0;
1142 }
1143 clear_opt(sbi->s_mount_opt, QUOTA);
1144 clear_opt(sbi->s_mount_opt, USRQUOTA);
1145 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1146 break;
1147#else
1148 case Opt_quota:
1149 case Opt_usrquota:
1150 case Opt_grpquota:
1151 case Opt_usrjquota:
1152 case Opt_grpjquota:
1153 case Opt_offusrjquota:
1154 case Opt_offgrpjquota:
1155 case Opt_jqfmt_vfsold:
1156 case Opt_jqfmt_vfsv0:
1157 printk(KERN_ERR
617ba13b 1158 "EXT4-fs: journalled quota options not "
ac27a0ec
DK		 1159 "supported.\n");
1160 break;
1161 case Opt_noquota:
1162 break;
1163#endif
1164 case Opt_abort:
1165 set_opt(sbi->s_mount_opt, ABORT);
1166 break;
1167 case Opt_barrier:
1168 if (match_int(&args[0], &option))
1169 return 0;
1170 if (option)
1171 set_opt(sbi->s_mount_opt, BARRIER);
1172 else
1173 clear_opt(sbi->s_mount_opt, BARRIER);
1174 break;
1175 case Opt_ignore:
1176 break;
1177 case Opt_resize:
1178 if (!is_remount) {
617ba13b 1179 printk("EXT4-fs: resize option only available "
ac27a0ec
DK		 1180 "for remount\n");
1181 return 0;
1182 }
1183 if (match_int(&args[0], &option) != 0)
1184 return 0;
1185 *n_blocks_count = option;
1186 break;
1187 case Opt_nobh:
1188 set_opt(sbi->s_mount_opt, NOBH);
1189 break;
1190 case Opt_bh:
1191 clear_opt(sbi->s_mount_opt, NOBH);
1192 break;
a86c6181
AT		 1193 case Opt_extents:
1194 set_opt (sbi->s_mount_opt, EXTENTS);
1195 break;
1e2462f9
MC		 1196 case Opt_noextents:
1197 clear_opt (sbi->s_mount_opt, EXTENTS);
1198 break;
ac27a0ec
DK		 1199 default:
1200 printk (KERN_ERR
617ba13b 1201 "EXT4-fs: Unrecognized mount option \"%s\" "
ac27a0ec
DK		 1202 "or missing value\n", p);
1203 return 0;
1204 }
1205 }
1206#ifdef CONFIG_QUOTA
1207 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
617ba13b 1208 if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
ac27a0ec
DK		 1209 sbi->s_qf_names[USRQUOTA])
1210 clear_opt(sbi->s_mount_opt, USRQUOTA);
1211
617ba13b 1212 if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
ac27a0ec
DK		 1213 sbi->s_qf_names[GRPQUOTA])
1214 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1215
1216 if ((sbi->s_qf_names[USRQUOTA] &&
617ba13b 1217 (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
ac27a0ec 1218 (sbi->s_qf_names[GRPQUOTA] &&
617ba13b
MC		 1219 (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
1220 printk(KERN_ERR "EXT4-fs: old and new quota "
ac27a0ec
DK		 1221 "format mixing.\n");
1222 return 0;
1223 }
1224
1225 if (!sbi->s_jquota_fmt) {
617ba13b 1226 printk(KERN_ERR "EXT4-fs: journalled quota format "
ac27a0ec
DK		 1227 "not specified.\n");
1228 return 0;
1229 }
1230 } else {
1231 if (sbi->s_jquota_fmt) {
617ba13b 1232 printk(KERN_ERR "EXT4-fs: journalled quota format "
ac27a0ec
DK		 1233 "specified with no journalling "
1234 "enabled.\n");
1235 return 0;
1236 }
1237 }
1238#endif
1239 return 1;
1240}
1241
617ba13b 1242static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 1243 int read_only)
1244{
617ba13b 1245 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 1246 int res = 0;
1247
617ba13b
MC		 1248 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1249 printk (KERN_ERR "EXT4-fs warning: revision level too high, "
ac27a0ec
DK		 1250 "forcing read-only mode\n");
1251 res = MS_RDONLY;
1252 }
1253 if (read_only)
1254 return res;
617ba13b
MC		 1255 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1256 printk (KERN_WARNING "EXT4-fs warning: mounting unchecked fs, "
ac27a0ec 1257 "running e2fsck is recommended\n");
617ba13b 1258 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
ac27a0ec 1259 printk (KERN_WARNING
617ba13b 1260 "EXT4-fs warning: mounting fs with errors, "
ac27a0ec
DK		 1261 "running e2fsck is recommended\n");
1262 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1263 le16_to_cpu(es->s_mnt_count) >=
1264 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1265 printk (KERN_WARNING
617ba13b 1266 "EXT4-fs warning: maximal mount count reached, "
ac27a0ec
DK		 1267 "running e2fsck is recommended\n");
1268 else if (le32_to_cpu(es->s_checkinterval) &&
1269 (le32_to_cpu(es->s_lastcheck) +
1270 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1271 printk (KERN_WARNING
617ba13b 1272 "EXT4-fs warning: checktime reached, "
ac27a0ec
DK		 1273 "running e2fsck is recommended\n");
1274#if 0
1275 /* @@@ We _will_ want to clear the valid bit if we find
63f57933
AM		 1276 * inconsistencies, to force a fsck at reboot. But for
1277 * a plain journaled filesystem we can keep it set as
1278 * valid forever! :)
1279 */
617ba13b 1280 es->s_state = cpu_to_le16(le16_to_cpu(es->s_state) & ~EXT4_VALID_FS);
ac27a0ec
DK		 1281#endif
1282 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 1283 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
ac27a0ec
DK		 1284 es->s_mnt_count=cpu_to_le16(le16_to_cpu(es->s_mnt_count) + 1);
1285 es->s_mtime = cpu_to_le32(get_seconds());
617ba13b
MC		 1286 ext4_update_dynamic_rev(sb);
1287 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 1288
617ba13b 1289 ext4_commit_super(sb, es, 1);
ac27a0ec 1290 if (test_opt(sb, DEBUG))
617ba13b 1291 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%lu, "
ac27a0ec
DK		 1292 "bpg=%lu, ipg=%lu, mo=%04lx]\n",
1293 sb->s_blocksize,
1294 sbi->s_groups_count,
617ba13b
MC		 1295 EXT4_BLOCKS_PER_GROUP(sb),
1296 EXT4_INODES_PER_GROUP(sb),
ac27a0ec
DK		 1297 sbi->s_mount_opt);
1298
617ba13b
MC		 1299 printk(KERN_INFO "EXT4 FS on %s, ", sb->s_id);
1300 if (EXT4_SB(sb)->s_journal->j_inode == NULL) {
ac27a0ec
DK		 1301 char b[BDEVNAME_SIZE];
1302
1303 printk("external journal on %s\n",
617ba13b 1304 bdevname(EXT4_SB(sb)->s_journal->j_dev, b));
ac27a0ec
DK		 1305 } else {
1306 printk("internal journal\n");
1307 }
1308 return res;
1309}
1310
1311/* Called at mount-time, super-block is locked */
617ba13b 1312static int ext4_check_descriptors (struct super_block * sb)
ac27a0ec 1313{
617ba13b
MC		 1314 struct ext4_sb_info *sbi = EXT4_SB(sb);
1315 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1316 ext4_fsblk_t last_block;
bd81d8ee
LV		 1317 ext4_fsblk_t block_bitmap;
1318 ext4_fsblk_t inode_bitmap;
1319 ext4_fsblk_t inode_table;
617ba13b 1320 struct ext4_group_desc * gdp = NULL;
ac27a0ec
DK		 1321 int desc_block = 0;
1322 int i;
1323
617ba13b 1324 ext4_debug ("Checking group descriptors");
ac27a0ec
DK		 1325
1326 for (i = 0; i < sbi->s_groups_count; i++)
1327 {
1328 if (i == sbi->s_groups_count - 1)
bd81d8ee 1329 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 1330 else
1331 last_block = first_block +
617ba13b 1332 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 1333
617ba13b
MC		 1334 if ((i % EXT4_DESC_PER_BLOCK(sb)) == 0)
1335 gdp = (struct ext4_group_desc *)
ac27a0ec 1336 sbi->s_group_desc[desc_block++]->b_data;
8fadc143 1337 block_bitmap = ext4_block_bitmap(sb, gdp);
bd81d8ee 1338 if (block_bitmap < first_block || block_bitmap > last_block)
ac27a0ec 1339 {
617ba13b 1340 ext4_error (sb, "ext4_check_descriptors",
ac27a0ec 1341 "Block bitmap for group %d"
2ae02107 1342 " not in group (block %llu)!",
bd81d8ee 1343 i, block_bitmap);
ac27a0ec
DK		 1344 return 0;
1345 }
8fadc143 1346 inode_bitmap = ext4_inode_bitmap(sb, gdp);
bd81d8ee 1347 if (inode_bitmap < first_block || inode_bitmap > last_block)
ac27a0ec 1348 {
617ba13b 1349 ext4_error (sb, "ext4_check_descriptors",
ac27a0ec 1350 "Inode bitmap for group %d"
2ae02107 1351 " not in group (block %llu)!",
bd81d8ee 1352 i, inode_bitmap);
ac27a0ec
DK		 1353 return 0;
1354 }
8fadc143 1355 inode_table = ext4_inode_table(sb, gdp);
bd81d8ee 1356 if (inode_table < first_block ||
780dcdb2 1357 inode_table + sbi->s_itb_per_group - 1 > last_block)
ac27a0ec 1358 {
617ba13b 1359 ext4_error (sb, "ext4_check_descriptors",
ac27a0ec 1360 "Inode table for group %d"
2ae02107 1361 " not in group (block %llu)!",
bd81d8ee 1362 i, inode_table);
ac27a0ec
DK		 1363 return 0;
1364 }
617ba13b 1365 first_block += EXT4_BLOCKS_PER_GROUP(sb);
0d1ee42f
AR		 1366 gdp = (struct ext4_group_desc *)
1367 ((__u8 *)gdp + EXT4_DESC_SIZE(sb));
ac27a0ec
DK		 1368 }
1369
bd81d8ee 1370 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
617ba13b 1371 sbi->s_es->s_free_inodes_count=cpu_to_le32(ext4_count_free_inodes(sb));
ac27a0ec
DK		 1372 return 1;
1373}
1374
1375
617ba13b 1376/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 1377 * the superblock) which were deleted from all directories, but held open by
1378 * a process at the time of a crash. We walk the list and try to delete these
1379 * inodes at recovery time (only with a read-write filesystem).
1380 *
1381 * In order to keep the orphan inode chain consistent during traversal (in
1382 * case of crash during recovery), we link each inode into the superblock
1383 * orphan list_head and handle it the same way as an inode deletion during
1384 * normal operation (which journals the operations for us).
1385 *
1386 * We only do an iget() and an iput() on each inode, which is very safe if we
1387 * accidentally point at an in-use or already deleted inode. The worst that
1388 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 1389 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 1390 * e2fsck was run on this filesystem, and it must have already done the orphan
1391 * inode cleanup for us, so we can safely abort without any further action.
1392 */
617ba13b
MC		 1393static void ext4_orphan_cleanup (struct super_block * sb,
1394 struct ext4_super_block * es)
ac27a0ec
DK		 1395{
1396 unsigned int s_flags = sb->s_flags;
1397 int nr_orphans = 0, nr_truncates = 0;
1398#ifdef CONFIG_QUOTA
1399 int i;
1400#endif
1401 if (!es->s_last_orphan) {
1402 jbd_debug(4, "no orphan inodes to clean up\n");
1403 return;
1404 }
1405
a8f48a95
ES		 1406 if (bdev_read_only(sb->s_bdev)) {
1407 printk(KERN_ERR "EXT4-fs: write access "
1408 "unavailable, skipping orphan cleanup.\n");
1409 return;
1410 }
1411
617ba13b 1412 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
ac27a0ec
DK		 1413 if (es->s_last_orphan)
1414 jbd_debug(1, "Errors on filesystem, "
1415 "clearing orphan list.\n");
1416 es->s_last_orphan = 0;
1417 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1418 return;
1419 }
1420
1421 if (s_flags & MS_RDONLY) {
617ba13b 1422 printk(KERN_INFO "EXT4-fs: %s: orphan cleanup on readonly fs\n",
ac27a0ec
DK		 1423 sb->s_id);
1424 sb->s_flags &= ~MS_RDONLY;
1425 }
1426#ifdef CONFIG_QUOTA
1427 /* Needed for iput() to work correctly and not trash data */
1428 sb->s_flags |= MS_ACTIVE;
1429 /* Turn on quotas so that they are updated correctly */
1430 for (i = 0; i < MAXQUOTAS; i++) {
617ba13b
MC		 1431 if (EXT4_SB(sb)->s_qf_names[i]) {
1432 int ret = ext4_quota_on_mount(sb, i);
ac27a0ec
DK		 1433 if (ret < 0)
1434 printk(KERN_ERR
617ba13b 1435 "EXT4-fs: Cannot turn on journalled "
ac27a0ec
DK		 1436 "quota: error %d\n", ret);
1437 }
1438 }
1439#endif
1440
1441 while (es->s_last_orphan) {
1442 struct inode *inode;
1443
1444 if (!(inode =
617ba13b 1445 ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan)))) {
ac27a0ec
DK		 1446 es->s_last_orphan = 0;
1447 break;
1448 }
1449
617ba13b 1450 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
ac27a0ec
DK		 1451 DQUOT_INIT(inode);
1452 if (inode->i_nlink) {
1453 printk(KERN_DEBUG
1454 "%s: truncating inode %lu to %Ld bytes\n",
1455 __FUNCTION__, inode->i_ino, inode->i_size);
1456 jbd_debug(2, "truncating inode %lu to %Ld bytes\n",
1457 inode->i_ino, inode->i_size);
617ba13b 1458 ext4_truncate(inode);
ac27a0ec
DK		 1459 nr_truncates++;
1460 } else {
1461 printk(KERN_DEBUG
1462 "%s: deleting unreferenced inode %lu\n",
1463 __FUNCTION__, inode->i_ino);
1464 jbd_debug(2, "deleting unreferenced inode %lu\n",
1465 inode->i_ino);
1466 nr_orphans++;
1467 }
1468 iput(inode); /* The delete magic happens here! */
1469 }
1470
1471#define PLURAL(x) (x), ((x)==1) ? "" : "s"
1472
1473 if (nr_orphans)
617ba13b 1474 printk(KERN_INFO "EXT4-fs: %s: %d orphan inode%s deleted\n",
ac27a0ec
DK		 1475 sb->s_id, PLURAL(nr_orphans));
1476 if (nr_truncates)
617ba13b 1477 printk(KERN_INFO "EXT4-fs: %s: %d truncate%s cleaned up\n",
ac27a0ec
DK		 1478 sb->s_id, PLURAL(nr_truncates));
1479#ifdef CONFIG_QUOTA
1480 /* Turn quotas off */
1481 for (i = 0; i < MAXQUOTAS; i++) {
1482 if (sb_dqopt(sb)->files[i])
1483 vfs_quota_off(sb, i);
1484 }
1485#endif
1486 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1487}
1488
ac27a0ec
DK		 1489/*
1490 * Maximal file size. There is a direct, and {,double-,triple-}indirect
1491 * block limit, and also a limit of (2^32 - 1) 512-byte sectors in i_blocks.
1492 * We need to be 1 filesystem block less than the 2^32 sector limit.
1493 */
617ba13b 1494static loff_t ext4_max_size(int bits)
ac27a0ec 1495{
617ba13b 1496 loff_t res = EXT4_NDIR_BLOCKS;
ac27a0ec
DK		 1497 /* This constant is calculated to be the largest file size for a
1498 * dense, 4k-blocksize file such that the total number of
1499 * sectors in the file, including data and all indirect blocks,
1500 * does not exceed 2^32. */
1501 const loff_t upper_limit = 0x1ff7fffd000LL;
1502
1503 res += 1LL << (bits-2);
1504 res += 1LL << (2*(bits-2));
1505 res += 1LL << (3*(bits-2));
1506 res <<= bits;
1507 if (res > upper_limit)
1508 res = upper_limit;
1509 return res;
1510}
1511
617ba13b 1512static ext4_fsblk_t descriptor_loc(struct super_block *sb,
70bbb3e0 1513 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 1514{
617ba13b 1515 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 1516 unsigned long bg, first_meta_bg;
1517 int has_super = 0;
1518
1519 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
1520
617ba13b 1521 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
ac27a0ec 1522 nr < first_meta_bg)
70bbb3e0 1523 return logical_sb_block + nr + 1;
ac27a0ec 1524 bg = sbi->s_desc_per_block * nr;
617ba13b 1525 if (ext4_bg_has_super(sb, bg))
ac27a0ec 1526 has_super = 1;
617ba13b 1527 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 1528}
1529
1530
617ba13b 1531static int ext4_fill_super (struct super_block *sb, void *data, int silent)
ac27a0ec
DK		 1532{
1533 struct buffer_head * bh;
617ba13b
MC		 1534 struct ext4_super_block *es = NULL;
1535 struct ext4_sb_info *sbi;
1536 ext4_fsblk_t block;
1537 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 1538 ext4_fsblk_t logical_sb_block;
ac27a0ec
DK		 1539 unsigned long offset = 0;
1540 unsigned int journal_inum = 0;
1541 unsigned long journal_devnum = 0;
1542 unsigned long def_mount_opts;
1543 struct inode *root;
1544 int blocksize;
1545 int hblock;
1546 int db_count;
1547 int i;
1548 int needs_recovery;
1549 __le32 features;
bd81d8ee 1550 __u64 blocks_count;
833f4077 1551 int err;
ac27a0ec
DK		 1552
1553 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
1554 if (!sbi)
1555 return -ENOMEM;
1556 sb->s_fs_info = sbi;
1557 sbi->s_mount_opt = 0;
617ba13b
MC		 1558 sbi->s_resuid = EXT4_DEF_RESUID;
1559 sbi->s_resgid = EXT4_DEF_RESGID;
d9c9bef1 1560 sbi->s_sb_block = sb_block;
ac27a0ec
DK		 1561
1562 unlock_kernel();
1563
617ba13b 1564 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 1565 if (!blocksize) {
617ba13b 1566 printk(KERN_ERR "EXT4-fs: unable to set blocksize\n");
ac27a0ec
DK		 1567 goto out_fail;
1568 }
1569
1570 /*
617ba13b 1571 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 1572 * block sizes. We need to calculate the offset from buffer start.
1573 */
617ba13b 1574 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 1575 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
1576 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 1577 } else {
70bbb3e0 1578 logical_sb_block = sb_block;
ac27a0ec
DK		 1579 }
1580
70bbb3e0 1581 if (!(bh = sb_bread(sb, logical_sb_block))) {
617ba13b 1582 printk (KERN_ERR "EXT4-fs: unable to read superblock\n");
ac27a0ec
DK		 1583 goto out_fail;
1584 }
1585 /*
1586 * Note: s_es must be initialized as soon as possible because
617ba13b 1587 * some ext4 macro-instructions depend on its value
ac27a0ec 1588 */
617ba13b 1589 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
ac27a0ec
DK		 1590 sbi->s_es = es;
1591 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 1592 if (sb->s_magic != EXT4_SUPER_MAGIC)
1593 goto cantfind_ext4;
ac27a0ec
DK		 1594
1595 /* Set defaults before we parse the mount options */
1596 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
617ba13b 1597 if (def_mount_opts & EXT4_DEFM_DEBUG)
ac27a0ec 1598 set_opt(sbi->s_mount_opt, DEBUG);
617ba13b 1599 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
ac27a0ec 1600 set_opt(sbi->s_mount_opt, GRPID);
617ba13b 1601 if (def_mount_opts & EXT4_DEFM_UID16)
ac27a0ec 1602 set_opt(sbi->s_mount_opt, NO_UID32);
2e7842b8 1603#ifdef CONFIG_EXT4DEV_FS_XATTR
617ba13b 1604 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
ac27a0ec 1605 set_opt(sbi->s_mount_opt, XATTR_USER);
2e7842b8
HD		 1606#endif
1607#ifdef CONFIG_EXT4DEV_FS_POSIX_ACL
617ba13b 1608 if (def_mount_opts & EXT4_DEFM_ACL)
ac27a0ec 1609 set_opt(sbi->s_mount_opt, POSIX_ACL);
2e7842b8 1610#endif
617ba13b
MC		 1611 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
1612 sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
1613 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
1614 sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
1615 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
1616 sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
1617
1618 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
ac27a0ec 1619 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
617ba13b 1620 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_RO)
ac27a0ec 1621 set_opt(sbi->s_mount_opt, ERRORS_RO);
ceea16bf
DM		 1622 else
1623 set_opt(sbi->s_mount_opt, ERRORS_CONT);
ac27a0ec
DK		 1624
1625 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
1626 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
1627
1628 set_opt(sbi->s_mount_opt, RESERVATION);
1629
1e2462f9
MC		 1630 /*
1631 * turn on extents feature by default in ext4 filesystem
1632 * User -o noextents to turn it off
1633 */
1634 set_opt(sbi->s_mount_opt, EXTENTS);
1635
ac27a0ec
DK		 1636 if (!parse_options ((char *) data, sb, &journal_inum, &journal_devnum,
1637 NULL, 0))
1638 goto failed_mount;
1639
1640 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
617ba13b 1641 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec 1642
617ba13b
MC		 1643 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
1644 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
1645 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
1646 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
ac27a0ec 1647 printk(KERN_WARNING
617ba13b 1648 "EXT4-fs warning: feature flags set on rev 0 fs, "
ac27a0ec
DK		 1649 "running e2fsck is recommended\n");
1650 /*
1651 * Check feature flags regardless of the revision level, since we
1652 * previously didn't change the revision level when setting the flags,
1653 * so there is a chance incompat flags are set on a rev 0 filesystem.
1654 */
617ba13b 1655 features = EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP);
ac27a0ec 1656 if (features) {
617ba13b 1657 printk(KERN_ERR "EXT4-fs: %s: couldn't mount because of "
ac27a0ec
DK		 1658 "unsupported optional features (%x).\n",
1659 sb->s_id, le32_to_cpu(features));
1660 goto failed_mount;
1661 }
617ba13b 1662 features = EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP);
ac27a0ec 1663 if (!(sb->s_flags & MS_RDONLY) && features) {
617ba13b 1664 printk(KERN_ERR "EXT4-fs: %s: couldn't mount RDWR because of "
ac27a0ec
DK		 1665 "unsupported optional features (%x).\n",
1666 sb->s_id, le32_to_cpu(features));
1667 goto failed_mount;
1668 }
1669 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
1670
617ba13b
MC		 1671 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
1672 blocksize > EXT4_MAX_BLOCK_SIZE) {
ac27a0ec 1673 printk(KERN_ERR
617ba13b 1674 "EXT4-fs: Unsupported filesystem blocksize %d on %s.\n",
ac27a0ec
DK		 1675 blocksize, sb->s_id);
1676 goto failed_mount;
1677 }
1678
1679 hblock = bdev_hardsect_size(sb->s_bdev);
1680 if (sb->s_blocksize != blocksize) {
1681 /*
1682 * Make sure the blocksize for the filesystem is larger
1683 * than the hardware sectorsize for the machine.
1684 */
1685 if (blocksize < hblock) {
617ba13b 1686 printk(KERN_ERR "EXT4-fs: blocksize %d too small for "
ac27a0ec
DK		 1687 "device blocksize %d.\n", blocksize, hblock);
1688 goto failed_mount;
1689 }
1690
1691 brelse (bh);
1692 sb_set_blocksize(sb, blocksize);
70bbb3e0
AM		 1693 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
1694 offset = do_div(logical_sb_block, blocksize);
1695 bh = sb_bread(sb, logical_sb_block);
ac27a0ec
DK		 1696 if (!bh) {
1697 printk(KERN_ERR
617ba13b 1698 "EXT4-fs: Can't read superblock on 2nd try.\n");
ac27a0ec
DK		 1699 goto failed_mount;
1700 }
617ba13b 1701 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
ac27a0ec 1702 sbi->s_es = es;
617ba13b 1703 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
ac27a0ec 1704 printk (KERN_ERR
617ba13b 1705 "EXT4-fs: Magic mismatch, very weird !\n");
ac27a0ec
DK		 1706 goto failed_mount;
1707 }
1708 }
1709
617ba13b 1710 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits);
ac27a0ec 1711
617ba13b
MC		 1712 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
1713 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
1714 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
ac27a0ec
DK		 1715 } else {
1716 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
1717 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
617ba13b 1718 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1330593e 1719 (!is_power_of_2(sbi->s_inode_size)) ||
ac27a0ec
DK		 1720 (sbi->s_inode_size > blocksize)) {
1721 printk (KERN_ERR
617ba13b 1722 "EXT4-fs: unsupported inode size: %d\n",
ac27a0ec
DK		 1723 sbi->s_inode_size);
1724 goto failed_mount;
1725 }
ef7f3835
KS		 1726 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
1727 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
ac27a0ec 1728 }
617ba13b 1729 sbi->s_frag_size = EXT4_MIN_FRAG_SIZE <<
ac27a0ec
DK		 1730 le32_to_cpu(es->s_log_frag_size);
1731 if (blocksize != sbi->s_frag_size) {
1732 printk(KERN_ERR
617ba13b 1733 "EXT4-fs: fragsize %lu != blocksize %u (unsupported)\n",
ac27a0ec
DK		 1734 sbi->s_frag_size, blocksize);
1735 goto failed_mount;
1736 }
0d1ee42f
AR		 1737 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
1738 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
8fadc143 1739 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f
AR		 1740 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
1741 sbi->s_desc_size & (sbi->s_desc_size - 1)) {
1742 printk(KERN_ERR
8fadc143 1743 "EXT4-fs: unsupported descriptor size %lu\n",
0d1ee42f
AR		 1744 sbi->s_desc_size);
1745 goto failed_mount;
1746 }
1747 } else
1748 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
ac27a0ec
DK		 1749 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
1750 sbi->s_frags_per_group = le32_to_cpu(es->s_frags_per_group);
1751 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
617ba13b
MC		 1752 if (EXT4_INODE_SIZE(sb) == 0)
1753 goto cantfind_ext4;
1754 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 1755 if (sbi->s_inodes_per_block == 0)
617ba13b 1756 goto cantfind_ext4;
ac27a0ec
DK		 1757 sbi->s_itb_per_group = sbi->s_inodes_per_group /
1758 sbi->s_inodes_per_block;
0d1ee42f 1759 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 1760 sbi->s_sbh = bh;
1761 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 1762 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
1763 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
ac27a0ec
DK		 1764 for (i=0; i < 4; i++)
1765 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
1766 sbi->s_def_hash_version = es->s_def_hash_version;
1767
1768 if (sbi->s_blocks_per_group > blocksize * 8) {
1769 printk (KERN_ERR
617ba13b 1770 "EXT4-fs: #blocks per group too big: %lu\n",
ac27a0ec
DK		 1771 sbi->s_blocks_per_group);
1772 goto failed_mount;
1773 }
1774 if (sbi->s_frags_per_group > blocksize * 8) {
1775 printk (KERN_ERR
617ba13b 1776 "EXT4-fs: #fragments per group too big: %lu\n",
ac27a0ec
DK		 1777 sbi->s_frags_per_group);
1778 goto failed_mount;
1779 }
1780 if (sbi->s_inodes_per_group > blocksize * 8) {
1781 printk (KERN_ERR
617ba13b 1782 "EXT4-fs: #inodes per group too big: %lu\n",
ac27a0ec
DK		 1783 sbi->s_inodes_per_group);
1784 goto failed_mount;
1785 }
1786
bd81d8ee 1787 if (ext4_blocks_count(es) >
ac27a0ec 1788 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) {
617ba13b 1789 printk(KERN_ERR "EXT4-fs: filesystem on %s:"
ac27a0ec
DK		 1790 " too large to mount safely\n", sb->s_id);
1791 if (sizeof(sector_t) < 8)
617ba13b 1792 printk(KERN_WARNING "EXT4-fs: CONFIG_LBD not "
ac27a0ec
DK		 1793 "enabled\n");
1794 goto failed_mount;
1795 }
1796
617ba13b
MC		 1797 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
1798 goto cantfind_ext4;
bd81d8ee
LV		 1799 blocks_count = (ext4_blocks_count(es) -
1800 le32_to_cpu(es->s_first_data_block) +
1801 EXT4_BLOCKS_PER_GROUP(sb) - 1);
1802 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
1803 sbi->s_groups_count = blocks_count;
617ba13b
MC		 1804 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
1805 EXT4_DESC_PER_BLOCK(sb);
ac27a0ec
DK		 1806 sbi->s_group_desc = kmalloc(db_count * sizeof (struct buffer_head *),
1807 GFP_KERNEL);
1808 if (sbi->s_group_desc == NULL) {
617ba13b 1809 printk (KERN_ERR "EXT4-fs: not enough memory\n");
ac27a0ec
DK		 1810 goto failed_mount;
1811 }
1812
1813 bgl_lock_init(&sbi->s_blockgroup_lock);
1814
1815 for (i = 0; i < db_count; i++) {
70bbb3e0 1816 block = descriptor_loc(sb, logical_sb_block, i);
ac27a0ec
DK		 1817 sbi->s_group_desc[i] = sb_bread(sb, block);
1818 if (!sbi->s_group_desc[i]) {
617ba13b 1819 printk (KERN_ERR "EXT4-fs: "
ac27a0ec
DK		 1820 "can't read group descriptor %d\n", i);
1821 db_count = i;
1822 goto failed_mount2;
1823 }
1824 }
617ba13b
MC		 1825 if (!ext4_check_descriptors (sb)) {
1826 printk(KERN_ERR "EXT4-fs: group descriptors corrupted!\n");
ac27a0ec
DK		 1827 goto failed_mount2;
1828 }
1829 sbi->s_gdb_count = db_count;
1830 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
1831 spin_lock_init(&sbi->s_next_gen_lock);
1832
833f4077
PZ		 1833 err = percpu_counter_init(&sbi->s_freeblocks_counter,
1834 ext4_count_free_blocks(sb));
1835 if (!err) {
1836 err = percpu_counter_init(&sbi->s_freeinodes_counter,
1837 ext4_count_free_inodes(sb));
1838 }
1839 if (!err) {
1840 err = percpu_counter_init(&sbi->s_dirs_counter,
1841 ext4_count_dirs(sb));
1842 }
1843 if (err) {
1844 printk(KERN_ERR "EXT4-fs: insufficient memory\n");
1845 goto failed_mount3;
1846 }
ac27a0ec
DK		 1847
1848 /* per fileystem reservation list head & lock */
1849 spin_lock_init(&sbi->s_rsv_window_lock);
1850 sbi->s_rsv_window_root = RB_ROOT;
1851 /* Add a single, static dummy reservation to the start of the
1852 * reservation window list --- it gives us a placeholder for
1853 * append-at-start-of-list which makes the allocation logic
1854 * _much_ simpler. */
617ba13b
MC		 1855 sbi->s_rsv_window_head.rsv_start = EXT4_RESERVE_WINDOW_NOT_ALLOCATED;
1856 sbi->s_rsv_window_head.rsv_end = EXT4_RESERVE_WINDOW_NOT_ALLOCATED;
ac27a0ec
DK		 1857 sbi->s_rsv_window_head.rsv_alloc_hit = 0;
1858 sbi->s_rsv_window_head.rsv_goal_size = 0;
617ba13b 1859 ext4_rsv_window_add(sb, &sbi->s_rsv_window_head);
ac27a0ec
DK		 1860
1861 /*
1862 * set up enough so that it can read an inode
1863 */
617ba13b
MC		 1864 sb->s_op = &ext4_sops;
1865 sb->s_export_op = &ext4_export_ops;
1866 sb->s_xattr = ext4_xattr_handlers;
ac27a0ec 1867#ifdef CONFIG_QUOTA
617ba13b
MC		 1868 sb->s_qcop = &ext4_qctl_operations;
1869 sb->dq_op = &ext4_quota_operations;
ac27a0ec
DK		 1870#endif
1871 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
1872
1873 sb->s_root = NULL;
1874
1875 needs_recovery = (es->s_last_orphan != 0 ||
617ba13b
MC		 1876 EXT4_HAS_INCOMPAT_FEATURE(sb,
1877 EXT4_FEATURE_INCOMPAT_RECOVER));
ac27a0ec
DK		 1878
1879 /*
1880 * The first inode we look at is the journal inode. Don't try
1881 * root first: it may be modified in the journal!
1882 */
1883 if (!test_opt(sb, NOLOAD) &&
617ba13b
MC		 1884 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
1885 if (ext4_load_journal(sb, es, journal_devnum))
ac27a0ec
DK		 1886 goto failed_mount3;
1887 } else if (journal_inum) {
617ba13b 1888 if (ext4_create_journal(sb, es, journal_inum))
ac27a0ec
DK		 1889 goto failed_mount3;
1890 } else {
1891 if (!silent)
1892 printk (KERN_ERR
617ba13b 1893 "ext4: No journal on filesystem on %s\n",
ac27a0ec
DK		 1894 sb->s_id);
1895 goto failed_mount3;
1896 }
1897
eb40a09c
JS		 1898 if (ext4_blocks_count(es) > 0xffffffffULL &&
1899 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
1900 JBD2_FEATURE_INCOMPAT_64BIT)) {
1901 printk(KERN_ERR "ext4: Failed to set 64-bit journal feature\n");
1902 goto failed_mount4;
1903 }
1904
ac27a0ec
DK		 1905 /* We have now updated the journal if required, so we can
1906 * validate the data journaling mode. */
1907 switch (test_opt(sb, DATA_FLAGS)) {
1908 case 0:
1909 /* No mode set, assume a default based on the journal
63f57933
AM		 1910 * capabilities: ORDERED_DATA if the journal can
1911 * cope, else JOURNAL_DATA
1912 */
dab291af
MC		 1913 if (jbd2_journal_check_available_features
1914 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
ac27a0ec
DK		 1915 set_opt(sbi->s_mount_opt, ORDERED_DATA);
1916 else
1917 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
1918 break;
1919
617ba13b
MC		 1920 case EXT4_MOUNT_ORDERED_DATA:
1921 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 1922 if (!jbd2_journal_check_available_features
1923 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
617ba13b 1924 printk(KERN_ERR "EXT4-fs: Journal does not support "
ac27a0ec
DK		 1925 "requested data journaling mode\n");
1926 goto failed_mount4;
1927 }
1928 default:
1929 break;
1930 }
1931
1932 if (test_opt(sb, NOBH)) {
617ba13b
MC		 1933 if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
1934 printk(KERN_WARNING "EXT4-fs: Ignoring nobh option - "
ac27a0ec
DK		 1935 "its supported only with writeback mode\n");
1936 clear_opt(sbi->s_mount_opt, NOBH);
1937 }
1938 }
1939 /*
dab291af 1940 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 1941 * so we can safely mount the rest of the filesystem now.
1942 */
1943
617ba13b 1944 root = iget(sb, EXT4_ROOT_INO);
ac27a0ec
DK		 1945 sb->s_root = d_alloc_root(root);
1946 if (!sb->s_root) {
617ba13b 1947 printk(KERN_ERR "EXT4-fs: get root inode failed\n");
ac27a0ec
DK		 1948 iput(root);
1949 goto failed_mount4;
1950 }
1951 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1952 dput(sb->s_root);
1953 sb->s_root = NULL;
617ba13b 1954 printk(KERN_ERR "EXT4-fs: corrupt root inode, run e2fsck\n");
ac27a0ec
DK		 1955 goto failed_mount4;
1956 }
1957
617ba13b 1958 ext4_setup_super (sb, es, sb->s_flags & MS_RDONLY);
ef7f3835
KS		 1959
1960 /* determine the minimum size of new large inodes, if present */
1961 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
1962 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
1963 EXT4_GOOD_OLD_INODE_SIZE;
1964 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
1965 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
1966 if (sbi->s_want_extra_isize <
1967 le16_to_cpu(es->s_want_extra_isize))
1968 sbi->s_want_extra_isize =
1969 le16_to_cpu(es->s_want_extra_isize);
1970 if (sbi->s_want_extra_isize <
1971 le16_to_cpu(es->s_min_extra_isize))
1972 sbi->s_want_extra_isize =
1973 le16_to_cpu(es->s_min_extra_isize);
1974 }
1975 }
1976 /* Check if enough inode space is available */
1977 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
1978 sbi->s_inode_size) {
1979 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
1980 EXT4_GOOD_OLD_INODE_SIZE;
1981 printk(KERN_INFO "EXT4-fs: required extra inode space not"
1982 "available.\n");
1983 }
1984
ac27a0ec
DK		 1985 /*
1986 * akpm: core read_super() calls in here with the superblock locked.
1987 * That deadlocks, because orphan cleanup needs to lock the superblock
1988 * in numerous places. Here we just pop the lock - it's relatively
1989 * harmless, because we are now ready to accept write_super() requests,
1990 * and aviro says that's the only reason for hanging onto the
1991 * superblock lock.
1992 */
617ba13b
MC		 1993 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
1994 ext4_orphan_cleanup(sb, es);
1995 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
ac27a0ec 1996 if (needs_recovery)
617ba13b
MC		 1997 printk (KERN_INFO "EXT4-fs: recovery complete.\n");
1998 ext4_mark_recovery_complete(sb, es);
1999 printk (KERN_INFO "EXT4-fs: mounted filesystem with %s data mode.\n",
2000 test_opt(sb,DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA ? "journal":
2001 test_opt(sb,DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA ? "ordered":
ac27a0ec
DK		 2002 "writeback");
2003
a86c6181
AT		 2004 ext4_ext_init(sb);
2005
ac27a0ec
DK		 2006 lock_kernel();
2007 return 0;
2008
617ba13b 2009cantfind_ext4:
ac27a0ec 2010 if (!silent)
617ba13b 2011 printk(KERN_ERR "VFS: Can't find ext4 filesystem on dev %s.\n",
ac27a0ec
DK		 2012 sb->s_id);
2013 goto failed_mount;
2014
2015failed_mount4:
dab291af 2016 jbd2_journal_destroy(sbi->s_journal);
ac27a0ec
DK		 2017failed_mount3:
2018 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2019 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2020 percpu_counter_destroy(&sbi->s_dirs_counter);
2021failed_mount2:
2022 for (i = 0; i < db_count; i++)
2023 brelse(sbi->s_group_desc[i]);
2024 kfree(sbi->s_group_desc);
2025failed_mount:
2026#ifdef CONFIG_QUOTA
2027 for (i = 0; i < MAXQUOTAS; i++)
2028 kfree(sbi->s_qf_names[i]);
2029#endif
617ba13b 2030 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 2031 brelse(bh);
2032out_fail:
2033 sb->s_fs_info = NULL;
2034 kfree(sbi);
2035 lock_kernel();
2036 return -EINVAL;
2037}
2038
2039/*
2040 * Setup any per-fs journal parameters now. We'll do this both on
2041 * initial mount, once the journal has been initialised but before we've
2042 * done any recovery; and again on any subsequent remount.
2043 */
617ba13b 2044static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 2045{
617ba13b 2046 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 2047
2048 if (sbi->s_commit_interval)
2049 journal->j_commit_interval = sbi->s_commit_interval;
617ba13b 2050 /* We could also set up an ext4-specific default for the commit
ac27a0ec
DK		 2051 * interval here, but for now we'll just fall back to the jbd
2052 * default. */
2053
2054 spin_lock(&journal->j_state_lock);
2055 if (test_opt(sb, BARRIER))
dab291af 2056 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 2057 else
dab291af 2058 journal->j_flags &= ~JBD2_BARRIER;
ac27a0ec
DK		 2059 spin_unlock(&journal->j_state_lock);
2060}
2061
617ba13b 2062static journal_t *ext4_get_journal(struct super_block *sb,
ac27a0ec
DK		 2063 unsigned int journal_inum)
2064{
2065 struct inode *journal_inode;
2066 journal_t *journal;
2067
2068 /* First, test for the existence of a valid inode on disk. Bad
2069 * things happen if we iget() an unused inode, as the subsequent
2070 * iput() will try to delete it. */
2071
2072 journal_inode = iget(sb, journal_inum);
2073 if (!journal_inode) {
617ba13b 2074 printk(KERN_ERR "EXT4-fs: no journal found.\n");
ac27a0ec
DK		 2075 return NULL;
2076 }
2077 if (!journal_inode->i_nlink) {
2078 make_bad_inode(journal_inode);
2079 iput(journal_inode);
617ba13b 2080 printk(KERN_ERR "EXT4-fs: journal inode is deleted.\n");
ac27a0ec
DK		 2081 return NULL;
2082 }
2083
2084 jbd_debug(2, "Journal inode found at %p: %Ld bytes\n",
2085 journal_inode, journal_inode->i_size);
2086 if (is_bad_inode(journal_inode) || !S_ISREG(journal_inode->i_mode)) {
617ba13b 2087 printk(KERN_ERR "EXT4-fs: invalid journal inode.\n");
ac27a0ec
DK		 2088 iput(journal_inode);
2089 return NULL;
2090 }
2091
dab291af 2092 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 2093 if (!journal) {
617ba13b 2094 printk(KERN_ERR "EXT4-fs: Could not load journal inode\n");
ac27a0ec
DK		 2095 iput(journal_inode);
2096 return NULL;
2097 }
2098 journal->j_private = sb;
617ba13b 2099 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 2100 return journal;
2101}
2102
617ba13b 2103static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 2104 dev_t j_dev)
2105{
2106 struct buffer_head * bh;
2107 journal_t *journal;
617ba13b
MC		 2108 ext4_fsblk_t start;
2109 ext4_fsblk_t len;
ac27a0ec 2110 int hblock, blocksize;
617ba13b 2111 ext4_fsblk_t sb_block;
ac27a0ec 2112 unsigned long offset;
617ba13b 2113 struct ext4_super_block * es;
ac27a0ec
DK		 2114 struct block_device *bdev;
2115
617ba13b 2116 bdev = ext4_blkdev_get(j_dev);
ac27a0ec
DK		 2117 if (bdev == NULL)
2118 return NULL;
2119
2120 if (bd_claim(bdev, sb)) {
2121 printk(KERN_ERR
8c55e204 2122 "EXT4: failed to claim external journal device.\n");
ac27a0ec
DK		 2123 blkdev_put(bdev);
2124 return NULL;
2125 }
2126
2127 blocksize = sb->s_blocksize;
2128 hblock = bdev_hardsect_size(bdev);
2129 if (blocksize < hblock) {
2130 printk(KERN_ERR
617ba13b 2131 "EXT4-fs: blocksize too small for journal device.\n");
ac27a0ec
DK		 2132 goto out_bdev;
2133 }
2134
617ba13b
MC		 2135 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
2136 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 2137 set_blocksize(bdev, blocksize);
2138 if (!(bh = __bread(bdev, sb_block, blocksize))) {
617ba13b 2139 printk(KERN_ERR "EXT4-fs: couldn't read superblock of "
ac27a0ec
DK		 2140 "external journal\n");
2141 goto out_bdev;
2142 }
2143
617ba13b
MC		 2144 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
2145 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 2146 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b
MC		 2147 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
2148 printk(KERN_ERR "EXT4-fs: external journal has "
ac27a0ec
DK		 2149 "bad superblock\n");
2150 brelse(bh);
2151 goto out_bdev;
2152 }
2153
617ba13b
MC		 2154 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
2155 printk(KERN_ERR "EXT4-fs: journal UUID does not match\n");
ac27a0ec
DK		 2156 brelse(bh);
2157 goto out_bdev;
2158 }
2159
bd81d8ee 2160 len = ext4_blocks_count(es);
ac27a0ec
DK		 2161 start = sb_block + 1;
2162 brelse(bh); /* we're done with the superblock */
2163
dab291af 2164 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 2165 start, len, blocksize);
2166 if (!journal) {
617ba13b 2167 printk(KERN_ERR "EXT4-fs: failed to create device journal\n");
ac27a0ec
DK		 2168 goto out_bdev;
2169 }
2170 journal->j_private = sb;
2171 ll_rw_block(READ, 1, &journal->j_sb_buffer);
2172 wait_on_buffer(journal->j_sb_buffer);
2173 if (!buffer_uptodate(journal->j_sb_buffer)) {
617ba13b 2174 printk(KERN_ERR "EXT4-fs: I/O error on journal device\n");
ac27a0ec
DK		 2175 goto out_journal;
2176 }
2177 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
617ba13b 2178 printk(KERN_ERR "EXT4-fs: External journal has more than one "
ac27a0ec
DK		 2179 "user (unsupported) - %d\n",
2180 be32_to_cpu(journal->j_superblock->s_nr_users));
2181 goto out_journal;
2182 }
617ba13b
MC		 2183 EXT4_SB(sb)->journal_bdev = bdev;
2184 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 2185 return journal;
2186out_journal:
dab291af 2187 jbd2_journal_destroy(journal);
ac27a0ec 2188out_bdev:
617ba13b 2189 ext4_blkdev_put(bdev);
ac27a0ec
DK		 2190 return NULL;
2191}
2192
617ba13b
MC		 2193static int ext4_load_journal(struct super_block *sb,
2194 struct ext4_super_block *es,
ac27a0ec
DK		 2195 unsigned long journal_devnum)
2196{
2197 journal_t *journal;
2198 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
2199 dev_t journal_dev;
2200 int err = 0;
2201 int really_read_only;
2202
2203 if (journal_devnum &&
2204 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
617ba13b 2205 printk(KERN_INFO "EXT4-fs: external journal device major/minor "
ac27a0ec
DK		 2206 "numbers have changed\n");
2207 journal_dev = new_decode_dev(journal_devnum);
2208 } else
2209 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
2210
2211 really_read_only = bdev_read_only(sb->s_bdev);
2212
2213 /*
2214 * Are we loading a blank journal or performing recovery after a
2215 * crash? For recovery, we need to check in advance whether we
2216 * can get read-write access to the device.
2217 */
2218
617ba13b 2219 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
ac27a0ec 2220 if (sb->s_flags & MS_RDONLY) {
617ba13b 2221 printk(KERN_INFO "EXT4-fs: INFO: recovery "
ac27a0ec
DK		 2222 "required on readonly filesystem.\n");
2223 if (really_read_only) {
617ba13b 2224 printk(KERN_ERR "EXT4-fs: write access "
ac27a0ec
DK		 2225 "unavailable, cannot proceed.\n");
2226 return -EROFS;
2227 }
617ba13b 2228 printk (KERN_INFO "EXT4-fs: write access will "
ac27a0ec
DK		 2229 "be enabled during recovery.\n");
2230 }
2231 }
2232
2233 if (journal_inum && journal_dev) {
617ba13b 2234 printk(KERN_ERR "EXT4-fs: filesystem has both journal "
ac27a0ec
DK		 2235 "and inode journals!\n");
2236 return -EINVAL;
2237 }
2238
2239 if (journal_inum) {
617ba13b 2240 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 2241 return -EINVAL;
2242 } else {
617ba13b 2243 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 2244 return -EINVAL;
2245 }
2246
2247 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
dab291af 2248 err = jbd2_journal_update_format(journal);
ac27a0ec 2249 if (err) {
617ba13b 2250 printk(KERN_ERR "EXT4-fs: error updating journal.\n");
dab291af 2251 jbd2_journal_destroy(journal);
ac27a0ec
DK		 2252 return err;
2253 }
2254 }
2255
617ba13b 2256 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
dab291af 2257 err = jbd2_journal_wipe(journal, !really_read_only);
ac27a0ec 2258 if (!err)
dab291af 2259 err = jbd2_journal_load(journal);
ac27a0ec
DK		 2260
2261 if (err) {
617ba13b 2262 printk(KERN_ERR "EXT4-fs: error loading journal.\n");
dab291af 2263 jbd2_journal_destroy(journal);
ac27a0ec
DK		 2264 return err;
2265 }
2266
617ba13b
MC		 2267 EXT4_SB(sb)->s_journal = journal;
2268 ext4_clear_journal_err(sb, es);
ac27a0ec
DK		 2269
2270 if (journal_devnum &&
2271 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
2272 es->s_journal_dev = cpu_to_le32(journal_devnum);
2273 sb->s_dirt = 1;
2274
2275 /* Make sure we flush the recovery flag to disk. */
617ba13b 2276 ext4_commit_super(sb, es, 1);
ac27a0ec
DK		 2277 }
2278
2279 return 0;
2280}
2281
617ba13b
MC		 2282static int ext4_create_journal(struct super_block * sb,
2283 struct ext4_super_block * es,
ac27a0ec
DK		 2284 unsigned int journal_inum)
2285{
2286 journal_t *journal;
6c675bd4 2287 int err;
ac27a0ec
DK		 2288
2289 if (sb->s_flags & MS_RDONLY) {
617ba13b 2290 printk(KERN_ERR "EXT4-fs: readonly filesystem when trying to "
ac27a0ec
DK		 2291 "create journal.\n");
2292 return -EROFS;
2293 }
2294
6c675bd4
BP		 2295 journal = ext4_get_journal(sb, journal_inum);
2296 if (!journal)
ac27a0ec
DK		 2297 return -EINVAL;
2298
617ba13b 2299 printk(KERN_INFO "EXT4-fs: creating new journal on inode %u\n",
ac27a0ec
DK		 2300 journal_inum);
2301
6c675bd4
BP		 2302 err = jbd2_journal_create(journal);
2303 if (err) {
617ba13b 2304 printk(KERN_ERR "EXT4-fs: error creating journal.\n");
dab291af 2305 jbd2_journal_destroy(journal);
ac27a0ec
DK		 2306 return -EIO;
2307 }
2308
617ba13b 2309 EXT4_SB(sb)->s_journal = journal;
ac27a0ec 2310
617ba13b
MC		 2311 ext4_update_dynamic_rev(sb);
2312 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2313 EXT4_SET_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL);
ac27a0ec
DK		 2314
2315 es->s_journal_inum = cpu_to_le32(journal_inum);
2316 sb->s_dirt = 1;
2317
2318 /* Make sure we flush the recovery flag to disk. */
617ba13b 2319 ext4_commit_super(sb, es, 1);
ac27a0ec
DK		 2320
2321 return 0;
2322}
2323
617ba13b
MC		 2324static void ext4_commit_super (struct super_block * sb,
2325 struct ext4_super_block * es,
ac27a0ec
DK		 2326 int sync)
2327{
617ba13b 2328 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
ac27a0ec
DK		 2329
2330 if (!sbh)
2331 return;
2332 es->s_wtime = cpu_to_le32(get_seconds());
bd81d8ee 2333 ext4_free_blocks_count_set(es, ext4_count_free_blocks(sb));
617ba13b 2334 es->s_free_inodes_count = cpu_to_le32(ext4_count_free_inodes(sb));
ac27a0ec
DK		 2335 BUFFER_TRACE(sbh, "marking dirty");
2336 mark_buffer_dirty(sbh);
2337 if (sync)
2338 sync_dirty_buffer(sbh);
2339}
2340
2341
2342/*
2343 * Have we just finished recovery? If so, and if we are mounting (or
2344 * remounting) the filesystem readonly, then we will end up with a
2345 * consistent fs on disk. Record that fact.
2346 */
617ba13b
MC		 2347static void ext4_mark_recovery_complete(struct super_block * sb,
2348 struct ext4_super_block * es)
ac27a0ec 2349{
617ba13b 2350 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 2351
dab291af
MC		 2352 jbd2_journal_lock_updates(journal);
2353 jbd2_journal_flush(journal);
32c37730 2354 lock_super(sb);
617ba13b 2355 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
ac27a0ec 2356 sb->s_flags & MS_RDONLY) {
617ba13b 2357 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 2358 sb->s_dirt = 0;
617ba13b 2359 ext4_commit_super(sb, es, 1);
ac27a0ec 2360 }
32c37730 2361 unlock_super(sb);
dab291af 2362 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 2363}
2364
2365/*
2366 * If we are mounting (or read-write remounting) a filesystem whose journal
2367 * has recorded an error from a previous lifetime, move that error to the
2368 * main filesystem now.
2369 */
617ba13b
MC		 2370static void ext4_clear_journal_err(struct super_block * sb,
2371 struct ext4_super_block * es)
ac27a0ec
DK		 2372{
2373 journal_t *journal;
2374 int j_errno;
2375 const char *errstr;
2376
617ba13b 2377 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 2378
2379 /*
2380 * Now check for any error status which may have been recorded in the
617ba13b 2381 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 2382 */
2383
dab291af 2384 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 2385 if (j_errno) {
2386 char nbuf[16];
2387
617ba13b
MC		 2388 errstr = ext4_decode_error(sb, j_errno, nbuf);
2389 ext4_warning(sb, __FUNCTION__, "Filesystem error recorded "
ac27a0ec 2390 "from previous mount: %s", errstr);
617ba13b 2391 ext4_warning(sb, __FUNCTION__, "Marking fs in need of "
ac27a0ec
DK		 2392 "filesystem check.");
2393
617ba13b
MC		 2394 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2395 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2396 ext4_commit_super (sb, es, 1);
ac27a0ec 2397
dab291af 2398 jbd2_journal_clear_err(journal);
ac27a0ec
DK		 2399 }
2400}
2401
2402/*
2403 * Force the running and committing transactions to commit,
2404 * and wait on the commit.
2405 */
617ba13b 2406int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 2407{
2408 journal_t *journal;
2409 int ret;
2410
2411 if (sb->s_flags & MS_RDONLY)
2412 return 0;
2413
617ba13b 2414 journal = EXT4_SB(sb)->s_journal;
ac27a0ec 2415 sb->s_dirt = 0;
617ba13b 2416 ret = ext4_journal_force_commit(journal);
ac27a0ec
DK		 2417 return ret;
2418}
2419
2420/*
617ba13b 2421 * Ext4 always journals updates to the superblock itself, so we don't
ac27a0ec
DK		 2422 * have to propagate any other updates to the superblock on disk at this
2423 * point. Just start an async writeback to get the buffers on their way
2424 * to the disk.
2425 *
2426 * This implicitly triggers the writebehind on sync().
2427 */
2428
617ba13b 2429static void ext4_write_super (struct super_block * sb)
ac27a0ec
DK		 2430{
2431 if (mutex_trylock(&sb->s_lock) != 0)
2432 BUG();
2433 sb->s_dirt = 0;
2434}
2435
617ba13b 2436static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec
DK		 2437{
2438 tid_t target;
2439
2440 sb->s_dirt = 0;
dab291af 2441 if (jbd2_journal_start_commit(EXT4_SB(sb)->s_journal, &target)) {
ac27a0ec 2442 if (wait)
dab291af 2443 jbd2_log_wait_commit(EXT4_SB(sb)->s_journal, target);
ac27a0ec
DK		 2444 }
2445 return 0;
2446}
2447
2448/*
2449 * LVM calls this function before a (read-only) snapshot is created. This
2450 * gives us a chance to flush the journal completely and mark the fs clean.
2451 */
617ba13b 2452static void ext4_write_super_lockfs(struct super_block *sb)
ac27a0ec
DK		 2453{
2454 sb->s_dirt = 0;
2455
2456 if (!(sb->s_flags & MS_RDONLY)) {
617ba13b 2457 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 2458
2459 /* Now we set up the journal barrier. */
dab291af
MC		 2460 jbd2_journal_lock_updates(journal);
2461 jbd2_journal_flush(journal);
ac27a0ec
DK		 2462
2463 /* Journal blocked and flushed, clear needs_recovery flag. */
617ba13b
MC		 2464 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2465 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
ac27a0ec
DK		 2466 }
2467}
2468
2469/*
2470 * Called by LVM after the snapshot is done. We need to reset the RECOVER
2471 * flag here, even though the filesystem is not technically dirty yet.
2472 */
617ba13b 2473static void ext4_unlockfs(struct super_block *sb)
ac27a0ec
DK		 2474{
2475 if (!(sb->s_flags & MS_RDONLY)) {
2476 lock_super(sb);
2477 /* Reser the needs_recovery flag before the fs is unlocked. */
617ba13b
MC		 2478 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
2479 ext4_commit_super(sb, EXT4_SB(sb)->s_es, 1);
ac27a0ec 2480 unlock_super(sb);
dab291af 2481 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
ac27a0ec
DK		 2482 }
2483}
2484
617ba13b 2485static int ext4_remount (struct super_block * sb, int * flags, char * data)
ac27a0ec 2486{
617ba13b
MC		 2487 struct ext4_super_block * es;
2488 struct ext4_sb_info *sbi = EXT4_SB(sb);
2489 ext4_fsblk_t n_blocks_count = 0;
ac27a0ec 2490 unsigned long old_sb_flags;
617ba13b 2491 struct ext4_mount_options old_opts;
ac27a0ec
DK		 2492 int err;
2493#ifdef CONFIG_QUOTA
2494 int i;
2495#endif
2496
2497 /* Store the original options */
2498 old_sb_flags = sb->s_flags;
2499 old_opts.s_mount_opt = sbi->s_mount_opt;
2500 old_opts.s_resuid = sbi->s_resuid;
2501 old_opts.s_resgid = sbi->s_resgid;
2502 old_opts.s_commit_interval = sbi->s_commit_interval;
2503#ifdef CONFIG_QUOTA
2504 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
2505 for (i = 0; i < MAXQUOTAS; i++)
2506 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
2507#endif
2508
2509 /*
2510 * Allow the "check" option to be passed as a remount option.
2511 */
2512 if (!parse_options(data, sb, NULL, NULL, &n_blocks_count, 1)) {
2513 err = -EINVAL;
2514 goto restore_opts;
2515 }
2516
617ba13b
MC		 2517 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT)
2518 ext4_abort(sb, __FUNCTION__, "Abort forced by user");
ac27a0ec
DK		 2519
2520 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
617ba13b 2521 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec
DK		 2522
2523 es = sbi->s_es;
2524
617ba13b 2525 ext4_init_journal_params(sb, sbi->s_journal);
ac27a0ec
DK		 2526
2527 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
bd81d8ee 2528 n_blocks_count > ext4_blocks_count(es)) {
617ba13b 2529 if (sbi->s_mount_opt & EXT4_MOUNT_ABORT) {
ac27a0ec
DK		 2530 err = -EROFS;
2531 goto restore_opts;
2532 }
2533
2534 if (*flags & MS_RDONLY) {
2535 /*
2536 * First of all, the unconditional stuff we have to do
2537 * to disable replay of the journal when we next remount
2538 */
2539 sb->s_flags |= MS_RDONLY;
2540
2541 /*
2542 * OK, test if we are remounting a valid rw partition
2543 * readonly, and if so set the rdonly flag and then
2544 * mark the partition as valid again.
2545 */
617ba13b
MC		 2546 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
2547 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 2548 es->s_state = cpu_to_le16(sbi->s_mount_state);
2549
32c37730
JK		 2550 /*
2551 * We have to unlock super so that we can wait for
2552 * transactions.
2553 */
2554 unlock_super(sb);
617ba13b 2555 ext4_mark_recovery_complete(sb, es);
32c37730 2556 lock_super(sb);
ac27a0ec
DK		 2557 } else {
2558 __le32 ret;
617ba13b
MC		 2559 if ((ret = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2560 ~EXT4_FEATURE_RO_COMPAT_SUPP))) {
2561 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
ac27a0ec
DK		 2562 "remount RDWR because of unsupported "
2563 "optional features (%x).\n",
2564 sb->s_id, le32_to_cpu(ret));
2565 err = -EROFS;
2566 goto restore_opts;
2567 }
ead6596b
ES		 2568
2569 /*
2570 * If we have an unprocessed orphan list hanging
2571 * around from a previously readonly bdev mount,
2572 * require a full umount/remount for now.
2573 */
2574 if (es->s_last_orphan) {
2575 printk(KERN_WARNING "EXT4-fs: %s: couldn't "
2576 "remount RDWR because of unprocessed "
2577 "orphan inode list. Please "
2578 "umount/remount instead.\n",
2579 sb->s_id);
2580 err = -EINVAL;
2581 goto restore_opts;
2582 }
2583
ac27a0ec
DK		 2584 /*
2585 * Mounting a RDONLY partition read-write, so reread
2586 * and store the current valid flag. (It may have
2587 * been changed by e2fsck since we originally mounted
2588 * the partition.)
2589 */
617ba13b 2590 ext4_clear_journal_err(sb, es);
ac27a0ec 2591 sbi->s_mount_state = le16_to_cpu(es->s_state);
617ba13b 2592 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
ac27a0ec 2593 goto restore_opts;
617ba13b 2594 if (!ext4_setup_super (sb, es, 0))
ac27a0ec
DK		 2595 sb->s_flags &= ~MS_RDONLY;
2596 }
2597 }
2598#ifdef CONFIG_QUOTA
2599 /* Release old quota file names */
2600 for (i = 0; i < MAXQUOTAS; i++)
2601 if (old_opts.s_qf_names[i] &&
2602 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2603 kfree(old_opts.s_qf_names[i]);
2604#endif
2605 return 0;
2606restore_opts:
2607 sb->s_flags = old_sb_flags;
2608 sbi->s_mount_opt = old_opts.s_mount_opt;
2609 sbi->s_resuid = old_opts.s_resuid;
2610 sbi->s_resgid = old_opts.s_resgid;
2611 sbi->s_commit_interval = old_opts.s_commit_interval;
2612#ifdef CONFIG_QUOTA
2613 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
2614 for (i = 0; i < MAXQUOTAS; i++) {
2615 if (sbi->s_qf_names[i] &&
2616 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
2617 kfree(sbi->s_qf_names[i]);
2618 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
2619 }
2620#endif
2621 return err;
2622}
2623
617ba13b 2624static int ext4_statfs (struct dentry * dentry, struct kstatfs * buf)
ac27a0ec
DK		 2625{
2626 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 2627 struct ext4_sb_info *sbi = EXT4_SB(sb);
2628 struct ext4_super_block *es = sbi->s_es;
960cc398 2629 u64 fsid;
ac27a0ec 2630
5e70030d
BP		 2631 if (test_opt(sb, MINIX_DF)) {
2632 sbi->s_overhead_last = 0;
2633 } else if (sbi->s_blocks_last != le32_to_cpu(es->s_blocks_count)) {
2634 unsigned long ngroups = sbi->s_groups_count, i;
2635 ext4_fsblk_t overhead = 0;
ac27a0ec
DK		 2636 smp_rmb();
2637
2638 /*
5e70030d
BP		 2639 * Compute the overhead (FS structures). This is constant
2640 * for a given filesystem unless the number of block groups
2641 * changes so we cache the previous value until it does.
ac27a0ec
DK		 2642 */
2643
2644 /*
2645 * All of the blocks before first_data_block are
2646 * overhead
2647 */
2648 overhead = le32_to_cpu(es->s_first_data_block);
2649
2650 /*
2651 * Add the overhead attributed to the superblock and
2652 * block group descriptors. If the sparse superblocks
2653 * feature is turned on, then not all groups have this.
2654 */
2655 for (i = 0; i < ngroups; i++) {
617ba13b
MC		 2656 overhead += ext4_bg_has_super(sb, i) +
2657 ext4_bg_num_gdb(sb, i);
ac27a0ec
DK		 2658 cond_resched();
2659 }
2660
2661 /*
2662 * Every block group has an inode bitmap, a block
2663 * bitmap, and an inode table.
2664 */
5e70030d
BP		 2665 overhead += ngroups * (2 + sbi->s_itb_per_group);
2666 sbi->s_overhead_last = overhead;
2667 smp_wmb();
2668 sbi->s_blocks_last = le32_to_cpu(es->s_blocks_count);
ac27a0ec
DK		 2669 }
2670
617ba13b 2671 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 2672 buf->f_bsize = sb->s_blocksize;
5e70030d 2673 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
52d9f3b4 2674 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter);
5e70030d 2675 es->s_free_blocks_count = cpu_to_le32(buf->f_bfree);
bd81d8ee
LV		 2676 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
2677 if (buf->f_bfree < ext4_r_blocks_count(es))
ac27a0ec
DK		 2678 buf->f_bavail = 0;
2679 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 2680 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
5e70030d 2681 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
617ba13b 2682 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 2683 fsid = le64_to_cpup((void *)es->s_uuid) ^
2684 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
2685 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
2686 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
ac27a0ec
DK		 2687 return 0;
2688}
2689
2690/* Helper function for writing quotas on sync - we need to start transaction before quota file
2691 * is locked for write. Otherwise the are possible deadlocks:
2692 * Process 1 Process 2
617ba13b 2693 * ext4_create() quota_sync()
dab291af 2694 * jbd2_journal_start() write_dquot()
ac27a0ec 2695 * DQUOT_INIT() down(dqio_mutex)
dab291af 2696 * down(dqio_mutex) jbd2_journal_start()
ac27a0ec
DK		 2697 *
2698 */
2699
2700#ifdef CONFIG_QUOTA
2701
2702static inline struct inode *dquot_to_inode(struct dquot *dquot)
2703{
2704 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
2705}
2706
617ba13b 2707static int ext4_dquot_initialize(struct inode *inode, int type)
ac27a0ec
DK		 2708{
2709 handle_t *handle;
2710 int ret, err;
2711
2712 /* We may create quota structure so we need to reserve enough blocks */
617ba13b 2713 handle = ext4_journal_start(inode, 2*EXT4_QUOTA_INIT_BLOCKS(inode->i_sb));
ac27a0ec
DK		 2714 if (IS_ERR(handle))
2715 return PTR_ERR(handle);
2716 ret = dquot_initialize(inode, type);
617ba13b 2717 err = ext4_journal_stop(handle);
ac27a0ec
DK		 2718 if (!ret)
2719 ret = err;
2720 return ret;
2721}
2722
617ba13b 2723static int ext4_dquot_drop(struct inode *inode)
ac27a0ec
DK		 2724{
2725 handle_t *handle;
2726 int ret, err;
2727
2728 /* We may delete quota structure so we need to reserve enough blocks */
617ba13b 2729 handle = ext4_journal_start(inode, 2*EXT4_QUOTA_DEL_BLOCKS(inode->i_sb));
ac27a0ec
DK		 2730 if (IS_ERR(handle))
2731 return PTR_ERR(handle);
2732 ret = dquot_drop(inode);
617ba13b 2733 err = ext4_journal_stop(handle);
ac27a0ec
DK		 2734 if (!ret)
2735 ret = err;
2736 return ret;
2737}
2738
617ba13b 2739static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 2740{
2741 int ret, err;
2742 handle_t *handle;
2743 struct inode *inode;
2744
2745 inode = dquot_to_inode(dquot);
617ba13b
MC		 2746 handle = ext4_journal_start(inode,
2747 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 2748 if (IS_ERR(handle))
2749 return PTR_ERR(handle);
2750 ret = dquot_commit(dquot);
617ba13b 2751 err = ext4_journal_stop(handle);
ac27a0ec
DK		 2752 if (!ret)
2753 ret = err;
2754 return ret;
2755}
2756
617ba13b 2757static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 2758{
2759 int ret, err;
2760 handle_t *handle;
2761
617ba13b
MC		 2762 handle = ext4_journal_start(dquot_to_inode(dquot),
2763 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 2764 if (IS_ERR(handle))
2765 return PTR_ERR(handle);
2766 ret = dquot_acquire(dquot);
617ba13b 2767 err = ext4_journal_stop(handle);
ac27a0ec
DK		 2768 if (!ret)
2769 ret = err;
2770 return ret;
2771}
2772
617ba13b 2773static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 2774{
2775 int ret, err;
2776 handle_t *handle;
2777
617ba13b
MC		 2778 handle = ext4_journal_start(dquot_to_inode(dquot),
2779 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 2780 if (IS_ERR(handle)) {
2781 /* Release dquot anyway to avoid endless cycle in dqput() */
2782 dquot_release(dquot);
ac27a0ec 2783 return PTR_ERR(handle);
9c3013e9 2784 }
ac27a0ec 2785 ret = dquot_release(dquot);
617ba13b 2786 err = ext4_journal_stop(handle);
ac27a0ec
DK		 2787 if (!ret)
2788 ret = err;
2789 return ret;
2790}
2791
617ba13b 2792static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec
DK		 2793{
2794 /* Are we journalling quotas? */
617ba13b
MC		 2795 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
2796 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
ac27a0ec 2797 dquot_mark_dquot_dirty(dquot);
617ba13b 2798 return ext4_write_dquot(dquot);
ac27a0ec
DK		 2799 } else {
2800 return dquot_mark_dquot_dirty(dquot);
2801 }
2802}
2803
617ba13b 2804static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 2805{
2806 int ret, err;
2807 handle_t *handle;
2808
2809 /* Data block + inode block */
617ba13b 2810 handle = ext4_journal_start(sb->s_root->d_inode, 2);
ac27a0ec
DK		 2811 if (IS_ERR(handle))
2812 return PTR_ERR(handle);
2813 ret = dquot_commit_info(sb, type);
617ba13b 2814 err = ext4_journal_stop(handle);
ac27a0ec
DK		 2815 if (!ret)
2816 ret = err;
2817 return ret;
2818}
2819
2820/*
2821 * Turn on quotas during mount time - we need to find
2822 * the quota file and such...
2823 */
617ba13b 2824static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 2825{
617ba13b
MC		 2826 return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
2827 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 2828}
2829
2830/*
2831 * Standard function to be called on quota_on
2832 */
617ba13b 2833static int ext4_quota_on(struct super_block *sb, int type, int format_id,
ac27a0ec
DK		 2834 char *path)
2835{
2836 int err;
2837 struct nameidata nd;
2838
2839 if (!test_opt(sb, QUOTA))
2840 return -EINVAL;
2841 /* Not journalling quota? */
617ba13b
MC		 2842 if (!EXT4_SB(sb)->s_qf_names[USRQUOTA] &&
2843 !EXT4_SB(sb)->s_qf_names[GRPQUOTA])
ac27a0ec
DK		 2844 return vfs_quota_on(sb, type, format_id, path);
2845 err = path_lookup(path, LOOKUP_FOLLOW, &nd);
2846 if (err)
2847 return err;
2848 /* Quotafile not on the same filesystem? */
2849 if (nd.mnt->mnt_sb != sb) {
2850 path_release(&nd);
2851 return -EXDEV;
2852 }
2853 /* Quotafile not of fs root? */
2854 if (nd.dentry->d_parent->d_inode != sb->s_root->d_inode)
2855 printk(KERN_WARNING
617ba13b 2856 "EXT4-fs: Quota file not on filesystem root. "
ac27a0ec
DK		 2857 "Journalled quota will not work.\n");
2858 path_release(&nd);
2859 return vfs_quota_on(sb, type, format_id, path);
2860}
2861
2862/* Read data from quotafile - avoid pagecache and such because we cannot afford
2863 * acquiring the locks... As quota files are never truncated and quota code
2864 * itself serializes the operations (and noone else should touch the files)
2865 * we don't have to be afraid of races */
617ba13b 2866static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 2867 size_t len, loff_t off)
2868{
2869 struct inode *inode = sb_dqopt(sb)->files[type];
617ba13b 2870 sector_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 2871 int err = 0;
2872 int offset = off & (sb->s_blocksize - 1);
2873 int tocopy;
2874 size_t toread;
2875 struct buffer_head *bh;
2876 loff_t i_size = i_size_read(inode);
2877
2878 if (off > i_size)
2879 return 0;
2880 if (off+len > i_size)
2881 len = i_size-off;
2882 toread = len;
2883 while (toread > 0) {
2884 tocopy = sb->s_blocksize - offset < toread ?
2885 sb->s_blocksize - offset : toread;
617ba13b 2886 bh = ext4_bread(NULL, inode, blk, 0, &err);
ac27a0ec
DK		 2887 if (err)
2888 return err;
2889 if (!bh) /* A hole? */
2890 memset(data, 0, tocopy);
2891 else
2892 memcpy(data, bh->b_data+offset, tocopy);
2893 brelse(bh);
2894 offset = 0;
2895 toread -= tocopy;
2896 data += tocopy;
2897 blk++;
2898 }
2899 return len;
2900}
2901
2902/* Write to quotafile (we know the transaction is already started and has
2903 * enough credits) */
617ba13b 2904static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 2905 const char *data, size_t len, loff_t off)
2906{
2907 struct inode *inode = sb_dqopt(sb)->files[type];
617ba13b 2908 sector_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 2909 int err = 0;
2910 int offset = off & (sb->s_blocksize - 1);
2911 int tocopy;
617ba13b 2912 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
ac27a0ec
DK		 2913 size_t towrite = len;
2914 struct buffer_head *bh;
2915 handle_t *handle = journal_current_handle();
2916
9c3013e9
JK		 2917 if (!handle) {
2918 printk(KERN_WARNING "EXT4-fs: Quota write (off=%Lu, len=%Lu)"
2919 " cancelled because transaction is not started.\n",
2920 (unsigned long long)off, (unsigned long long)len);
2921 return -EIO;
2922 }
ac27a0ec
DK		 2923 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
2924 while (towrite > 0) {
2925 tocopy = sb->s_blocksize - offset < towrite ?
2926 sb->s_blocksize - offset : towrite;
617ba13b 2927 bh = ext4_bread(handle, inode, blk, 1, &err);
ac27a0ec
DK		 2928 if (!bh)
2929 goto out;
2930 if (journal_quota) {
617ba13b 2931 err = ext4_journal_get_write_access(handle, bh);
ac27a0ec
DK		 2932 if (err) {
2933 brelse(bh);
2934 goto out;
2935 }
2936 }
2937 lock_buffer(bh);
2938 memcpy(bh->b_data+offset, data, tocopy);
2939 flush_dcache_page(bh->b_page);
2940 unlock_buffer(bh);
2941 if (journal_quota)
617ba13b 2942 err = ext4_journal_dirty_metadata(handle, bh);
ac27a0ec
DK		 2943 else {
2944 /* Always do at least ordered writes for quotas */
617ba13b 2945 err = ext4_journal_dirty_data(handle, bh);
ac27a0ec
DK		 2946 mark_buffer_dirty(bh);
2947 }
2948 brelse(bh);
2949 if (err)
2950 goto out;
2951 offset = 0;
2952 towrite -= tocopy;
2953 data += tocopy;
2954 blk++;
2955 }
2956out:
2957 if (len == towrite)
2958 return err;
2959 if (inode->i_size < off+len-towrite) {
2960 i_size_write(inode, off+len-towrite);
617ba13b 2961 EXT4_I(inode)->i_disksize = inode->i_size;
ac27a0ec
DK		 2962 }
2963 inode->i_version++;
2964 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
617ba13b 2965 ext4_mark_inode_dirty(handle, inode);
ac27a0ec
DK		 2966 mutex_unlock(&inode->i_mutex);
2967 return len - towrite;
2968}
2969
2970#endif
2971
617ba13b 2972static int ext4_get_sb(struct file_system_type *fs_type,
ac27a0ec
DK		 2973 int flags, const char *dev_name, void *data, struct vfsmount *mnt)
2974{
617ba13b 2975 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super, mnt);
ac27a0ec
DK		 2976}
2977
617ba13b 2978static struct file_system_type ext4dev_fs_type = {
ac27a0ec 2979 .owner = THIS_MODULE,
617ba13b
MC		 2980 .name = "ext4dev",
2981 .get_sb = ext4_get_sb,
ac27a0ec
DK		 2982 .kill_sb = kill_block_super,
2983 .fs_flags = FS_REQUIRES_DEV,
2984};
2985
617ba13b 2986static int __init init_ext4_fs(void)
ac27a0ec 2987{
617ba13b 2988 int err = init_ext4_xattr();
ac27a0ec
DK		 2989 if (err)
2990 return err;
2991 err = init_inodecache();
2992 if (err)
2993 goto out1;
63f57933 2994 err = register_filesystem(&ext4dev_fs_type);
ac27a0ec
DK		 2995 if (err)
2996 goto out;
2997 return 0;
2998out:
2999 destroy_inodecache();
3000out1:
617ba13b 3001 exit_ext4_xattr();
ac27a0ec
DK		 3002 return err;
3003}
3004
617ba13b 3005static void __exit exit_ext4_fs(void)
ac27a0ec 3006{
617ba13b 3007 unregister_filesystem(&ext4dev_fs_type);
ac27a0ec 3008 destroy_inodecache();
617ba13b 3009 exit_ext4_xattr();
ac27a0ec
DK		 3010}
3011
3012MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
617ba13b 3013MODULE_DESCRIPTION("Fourth Extended Filesystem with extents");
ac27a0ec 3014MODULE_LICENSE("GPL");
617ba13b
MC		 3015module_init(init_ext4_fs)
3016module_exit(exit_ext4_fs)
Clone of davem's net-next-2.6 tree