]> git.ipfire.org Git - thirdparty/kernel/linux.git/blame - fs/ext4/super.c
projects / thirdparty / kernel / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
ext4: Handle nested ext4_journal_start/stop calls without a journal
[thirdparty/kernel/linux.git] / fs / ext4 / super.c
CommitLineData
ac27a0ec 1/*
617ba13b 2 * linux/fs/ext4/super.c
ac27a0ec
DK		 3 *
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
8 *
9 * from
10 *
11 * linux/fs/minix/inode.c
12 *
13 * Copyright (C) 1991, 1992 Linus Torvalds
14 *
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
17 */
18
19#include <linux/module.h>
20#include <linux/string.h>
21#include <linux/fs.h>
22#include <linux/time.h>
c5ca7c76 23#include <linux/vmalloc.h>
dab291af 24#include <linux/jbd2.h>
ac27a0ec
DK		 25#include <linux/slab.h>
26#include <linux/init.h>
27#include <linux/blkdev.h>
28#include <linux/parser.h>
29#include <linux/smp_lock.h>
30#include <linux/buffer_head.h>
a5694255 31#include <linux/exportfs.h>
ac27a0ec
DK		 32#include <linux/vfs.h>
33#include <linux/random.h>
34#include <linux/mount.h>
35#include <linux/namei.h>
36#include <linux/quotaops.h>
37#include <linux/seq_file.h>
9f6200bb 38#include <linux/proc_fs.h>
3197ebdb 39#include <linux/ctype.h>
1330593e 40#include <linux/log2.h>
717d50e4 41#include <linux/crc16.h>
ac27a0ec
DK		 42#include <asm/uaccess.h>
43
3dcf5451
CH		 44#include "ext4.h"
45#include "ext4_jbd2.h"
ac27a0ec
DK		 46#include "xattr.h"
47#include "acl.h"
3661d286 48#include "mballoc.h"
ac27a0ec 49
9bffad1e
TT		 50#define CREATE_TRACE_POINTS
51#include <trace/events/ext4.h>
52
f4033903
CW		 53static int default_mb_history_length = 1000;
54
55module_param_named(default_mb_history_length, default_mb_history_length,
56 int, 0644);
57MODULE_PARM_DESC(default_mb_history_length,
58 "Default number of entries saved for mb_history");
59
9f6200bb 60struct proc_dir_entry *ext4_proc_root;
3197ebdb 61static struct kset *ext4_kset;
9f6200bb 62
617ba13b 63static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
ac27a0ec 64 unsigned long journal_devnum);
e2d67052 65static int ext4_commit_super(struct super_block *sb, int sync);
2b2d6d01
TT		 66static void ext4_mark_recovery_complete(struct super_block *sb,
67 struct ext4_super_block *es);
68static void ext4_clear_journal_err(struct super_block *sb,
69 struct ext4_super_block *es);
617ba13b 70static int ext4_sync_fs(struct super_block *sb, int wait);
2b2d6d01 71static const char *ext4_decode_error(struct super_block *sb, int errno,
ac27a0ec 72 char nbuf[16]);
2b2d6d01
TT		 73static int ext4_remount(struct super_block *sb, int *flags, char *data);
74static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
c4be0c1d 75static int ext4_unfreeze(struct super_block *sb);
2b2d6d01 76static void ext4_write_super(struct super_block *sb);
c4be0c1d 77static int ext4_freeze(struct super_block *sb);
ac27a0ec 78
bd81d8ee 79
8fadc143
AR		 80ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
81 struct ext4_group_desc *bg)
bd81d8ee 82{
3a14589c 83 return le32_to_cpu(bg->bg_block_bitmap_lo) |
8fadc143 84 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 85 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 86}
87
8fadc143
AR		 88ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
89 struct ext4_group_desc *bg)
bd81d8ee 90{
5272f837 91 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
8fadc143 92 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 93 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 94}
95
8fadc143
AR		 96ext4_fsblk_t ext4_inode_table(struct super_block *sb,
97 struct ext4_group_desc *bg)
bd81d8ee 98{
5272f837 99 return le32_to_cpu(bg->bg_inode_table_lo) |
8fadc143 100 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 101 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
bd81d8ee
LV		 102}
103
560671a0
AK		 104__u32 ext4_free_blks_count(struct super_block *sb,
105 struct ext4_group_desc *bg)
106{
107 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
108 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 109 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
560671a0
AK		 110}
111
112__u32 ext4_free_inodes_count(struct super_block *sb,
113 struct ext4_group_desc *bg)
114{
115 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
116 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 117 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
560671a0
AK		 118}
119
120__u32 ext4_used_dirs_count(struct super_block *sb,
121 struct ext4_group_desc *bg)
122{
123 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
124 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 125 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
560671a0
AK		 126}
127
128__u32 ext4_itable_unused_count(struct super_block *sb,
129 struct ext4_group_desc *bg)
130{
131 return le16_to_cpu(bg->bg_itable_unused_lo) |
132 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 133 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
560671a0
AK		 134}
135
8fadc143
AR		 136void ext4_block_bitmap_set(struct super_block *sb,
137 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 138{
3a14589c 139 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 140 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
141 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 142}
143
8fadc143
AR		 144void ext4_inode_bitmap_set(struct super_block *sb,
145 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 146{
5272f837 147 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 148 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
149 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 150}
151
8fadc143
AR		 152void ext4_inode_table_set(struct super_block *sb,
153 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 154{
5272f837 155 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 156 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
157 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 158}
159
560671a0
AK		 160void ext4_free_blks_set(struct super_block *sb,
161 struct ext4_group_desc *bg, __u32 count)
162{
163 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
164 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
165 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
166}
167
168void ext4_free_inodes_set(struct super_block *sb,
169 struct ext4_group_desc *bg, __u32 count)
170{
171 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
172 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
173 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
174}
175
176void ext4_used_dirs_set(struct super_block *sb,
177 struct ext4_group_desc *bg, __u32 count)
178{
179 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
180 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
181 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
182}
183
184void ext4_itable_unused_set(struct super_block *sb,
185 struct ext4_group_desc *bg, __u32 count)
186{
187 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
188 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
189 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
190}
191
d3d1faf6
CW		 192
193/* Just increment the non-pointer handle value */
194static handle_t *ext4_get_nojournal(void)
195{
196 handle_t *handle = current->journal_info;
197 unsigned long ref_cnt = (unsigned long)handle;
198
199 BUG_ON(ref_cnt >= EXT4_NOJOURNAL_MAX_REF_COUNT);
200
201 ref_cnt++;
202 handle = (handle_t *)ref_cnt;
203
204 current->journal_info = handle;
205 return handle;
206}
207
208
209/* Decrement the non-pointer handle value */
210static void ext4_put_nojournal(handle_t *handle)
211{
212 unsigned long ref_cnt = (unsigned long)handle;
213
214 BUG_ON(ref_cnt == 0);
215
216 ref_cnt--;
217 handle = (handle_t *)ref_cnt;
218
219 current->journal_info = handle;
220}
221
ac27a0ec 222/*
dab291af 223 * Wrappers for jbd2_journal_start/end.
ac27a0ec
DK		 224 *
225 * The only special thing we need to do here is to make sure that all
226 * journal_end calls result in the superblock being marked dirty, so
227 * that sync() will call the filesystem's write_super callback if
228 * appropriate.
229 */
617ba13b 230handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
ac27a0ec
DK		 231{
232 journal_t *journal;
233
234 if (sb->s_flags & MS_RDONLY)
235 return ERR_PTR(-EROFS);
236
237 /* Special case here: if the journal has aborted behind our
238 * backs (eg. EIO in the commit thread), then we still need to
239 * take the FS itself readonly cleanly. */
617ba13b 240 journal = EXT4_SB(sb)->s_journal;
0390131b
FM		 241 if (journal) {
242 if (is_journal_aborted(journal)) {
0b8e58a1 243 ext4_abort(sb, __func__, "Detected aborted journal");
0390131b
FM		 244 return ERR_PTR(-EROFS);
245 }
246 return jbd2_journal_start(journal, nblocks);
ac27a0ec 247 }
d3d1faf6 248 return ext4_get_nojournal();
ac27a0ec
DK		 249}
250
251/*
252 * The only special thing we need to do here is to make sure that all
dab291af 253 * jbd2_journal_stop calls result in the superblock being marked dirty, so
ac27a0ec
DK		 254 * that sync() will call the filesystem's write_super callback if
255 * appropriate.
256 */
617ba13b 257int __ext4_journal_stop(const char *where, handle_t *handle)
ac27a0ec
DK		 258{
259 struct super_block *sb;
260 int err;
261 int rc;
262
0390131b 263 if (!ext4_handle_valid(handle)) {
d3d1faf6 264 ext4_put_nojournal(handle);
0390131b
FM		 265 return 0;
266 }
ac27a0ec
DK		 267 sb = handle->h_transaction->t_journal->j_private;
268 err = handle->h_err;
dab291af 269 rc = jbd2_journal_stop(handle);
ac27a0ec
DK		 270
271 if (!err)
272 err = rc;
273 if (err)
617ba13b 274 __ext4_std_error(sb, where, err);
ac27a0ec
DK		 275 return err;
276}
277
617ba13b 278void ext4_journal_abort_handle(const char *caller, const char *err_fn,
ac27a0ec
DK		 279 struct buffer_head *bh, handle_t *handle, int err)
280{
281 char nbuf[16];
617ba13b 282 const char *errstr = ext4_decode_error(NULL, err, nbuf);
ac27a0ec 283
0390131b
FM		 284 BUG_ON(!ext4_handle_valid(handle));
285
ac27a0ec
DK		 286 if (bh)
287 BUFFER_TRACE(bh, "abort");
288
289 if (!handle->h_err)
290 handle->h_err = err;
291
292 if (is_handle_aborted(handle))
293 return;
294
295 printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
296 caller, errstr, err_fn);
297
dab291af 298 jbd2_journal_abort_handle(handle);
ac27a0ec
DK		 299}
300
301/* Deal with the reporting of failure conditions on a filesystem such as
302 * inconsistencies detected or read IO failures.
303 *
304 * On ext2, we can store the error state of the filesystem in the
617ba13b 305 * superblock. That is not possible on ext4, because we may have other
ac27a0ec
DK		 306 * write ordering constraints on the superblock which prevent us from
307 * writing it out straight away; and given that the journal is about to
308 * be aborted, we can't rely on the current, or future, transactions to
309 * write out the superblock safely.
310 *
dab291af 311 * We'll just use the jbd2_journal_abort() error code to record an error in
ac27a0ec
DK		 312 * the journal instead. On recovery, the journal will compain about
313 * that error until we've noted it down and cleared it.
314 */
315
617ba13b 316static void ext4_handle_error(struct super_block *sb)
ac27a0ec 317{
617ba13b 318 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 319
617ba13b
MC		 320 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
321 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
ac27a0ec
DK		 322
323 if (sb->s_flags & MS_RDONLY)
324 return;
325
2b2d6d01 326 if (!test_opt(sb, ERRORS_CONT)) {
617ba13b 327 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 328
4ab2f15b 329 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
ac27a0ec 330 if (journal)
dab291af 331 jbd2_journal_abort(journal, -EIO);
ac27a0ec 332 }
2b2d6d01 333 if (test_opt(sb, ERRORS_RO)) {
b31e1552 334 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
ac27a0ec
DK		 335 sb->s_flags |= MS_RDONLY;
336 }
e2d67052 337 ext4_commit_super(sb, 1);
ac27a0ec 338 if (test_opt(sb, ERRORS_PANIC))
617ba13b 339 panic("EXT4-fs (device %s): panic forced after error\n",
ac27a0ec
DK		 340 sb->s_id);
341}
342
2b2d6d01
TT		 343void ext4_error(struct super_block *sb, const char *function,
344 const char *fmt, ...)
ac27a0ec
DK		 345{
346 va_list args;
347
348 va_start(args, fmt);
2b2d6d01 349 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
ac27a0ec
DK		 350 vprintk(fmt, args);
351 printk("\n");
352 va_end(args);
353
617ba13b 354 ext4_handle_error(sb);
ac27a0ec
DK		 355}
356
2b2d6d01 357static const char *ext4_decode_error(struct super_block *sb, int errno,
ac27a0ec
DK		 358 char nbuf[16])
359{
360 char *errstr = NULL;
361
362 switch (errno) {
363 case -EIO:
364 errstr = "IO failure";
365 break;
366 case -ENOMEM:
367 errstr = "Out of memory";
368 break;
369 case -EROFS:
78f1ddbb
TT		 370 if (!sb || (EXT4_SB(sb)->s_journal &&
371 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
ac27a0ec
DK		 372 errstr = "Journal has aborted";
373 else
374 errstr = "Readonly filesystem";
375 break;
376 default:
377 /* If the caller passed in an extra buffer for unknown
378 * errors, textualise them now. Else we just return
379 * NULL. */
380 if (nbuf) {
381 /* Check for truncated error codes... */
382 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
383 errstr = nbuf;
384 }
385 break;
386 }
387
388 return errstr;
389}
390
617ba13b 391/* __ext4_std_error decodes expected errors from journaling functions
ac27a0ec
DK		 392 * automatically and invokes the appropriate error response. */
393
2b2d6d01 394void __ext4_std_error(struct super_block *sb, const char *function, int errno)
ac27a0ec
DK		 395{
396 char nbuf[16];
397 const char *errstr;
398
399 /* Special case: if the error is EROFS, and we're not already
400 * inside a transaction, then there's really no point in logging
401 * an error. */
402 if (errno == -EROFS && journal_current_handle() == NULL &&
403 (sb->s_flags & MS_RDONLY))
404 return;
405
617ba13b 406 errstr = ext4_decode_error(sb, errno, nbuf);
2b2d6d01
TT		 407 printk(KERN_CRIT "EXT4-fs error (device %s) in %s: %s\n",
408 sb->s_id, function, errstr);
ac27a0ec 409
617ba13b 410 ext4_handle_error(sb);
ac27a0ec
DK		 411}
412
413/*
617ba13b 414 * ext4_abort is a much stronger failure handler than ext4_error. The
ac27a0ec
DK		 415 * abort function may be used to deal with unrecoverable failures such
416 * as journal IO errors or ENOMEM at a critical moment in log management.
417 *
418 * We unconditionally force the filesystem into an ABORT|READONLY state,
419 * unless the error response on the fs has been set to panic in which
420 * case we take the easy way out and panic immediately.
421 */
422
2b2d6d01
TT		 423void ext4_abort(struct super_block *sb, const char *function,
424 const char *fmt, ...)
ac27a0ec
DK		 425{
426 va_list args;
427
ac27a0ec 428 va_start(args, fmt);
2b2d6d01 429 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
ac27a0ec
DK		 430 vprintk(fmt, args);
431 printk("\n");
432 va_end(args);
433
434 if (test_opt(sb, ERRORS_PANIC))
617ba13b 435 panic("EXT4-fs panic from previous error\n");
ac27a0ec
DK		 436
437 if (sb->s_flags & MS_RDONLY)
438 return;
439
b31e1552 440 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
617ba13b 441 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
ac27a0ec 442 sb->s_flags |= MS_RDONLY;
4ab2f15b 443 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
ef2cabf7
HK		 444 if (EXT4_SB(sb)->s_journal)
445 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
ac27a0ec
DK		 446}
447
b31e1552
ES		 448void ext4_msg (struct super_block * sb, const char *prefix,
449 const char *fmt, ...)
450{
451 va_list args;
452
453 va_start(args, fmt);
454 printk("%sEXT4-fs (%s): ", prefix, sb->s_id);
455 vprintk(fmt, args);
456 printk("\n");
457 va_end(args);
458}
459
2b2d6d01
TT		 460void ext4_warning(struct super_block *sb, const char *function,
461 const char *fmt, ...)
ac27a0ec
DK		 462{
463 va_list args;
464
465 va_start(args, fmt);
617ba13b 466 printk(KERN_WARNING "EXT4-fs warning (device %s): %s: ",
ac27a0ec
DK		 467 sb->s_id, function);
468 vprintk(fmt, args);
469 printk("\n");
470 va_end(args);
471}
472
5d1b1b3f 473void ext4_grp_locked_error(struct super_block *sb, ext4_group_t grp,
0b8e58a1 474 const char *function, const char *fmt, ...)
5d1b1b3f
AK		 475__releases(bitlock)
476__acquires(bitlock)
477{
478 va_list args;
479 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
480
481 va_start(args, fmt);
482 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
483 vprintk(fmt, args);
484 printk("\n");
485 va_end(args);
486
487 if (test_opt(sb, ERRORS_CONT)) {
488 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
489 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
e2d67052 490 ext4_commit_super(sb, 0);
5d1b1b3f
AK		 491 return;
492 }
493 ext4_unlock_group(sb, grp);
494 ext4_handle_error(sb);
495 /*
496 * We only get here in the ERRORS_RO case; relocking the group
497 * may be dangerous, but nothing bad will happen since the
498 * filesystem will have already been marked read/only and the
499 * journal has been aborted. We return 1 as a hint to callers
500 * who might what to use the return value from
501 * ext4_grp_locked_error() to distinguish beween the
502 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
503 * aggressively from the ext4 function in question, with a
504 * more appropriate error code.
505 */
506 ext4_lock_group(sb, grp);
507 return;
508}
509
617ba13b 510void ext4_update_dynamic_rev(struct super_block *sb)
ac27a0ec 511{
617ba13b 512 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 513
617ba13b 514 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
ac27a0ec
DK		 515 return;
516
46e665e9 517 ext4_warning(sb, __func__,
ac27a0ec
DK		 518 "updating to rev %d because of new feature flag, "
519 "running e2fsck is recommended",
617ba13b 520 EXT4_DYNAMIC_REV);
ac27a0ec 521
617ba13b
MC		 522 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
523 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
524 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
ac27a0ec
DK		 525 /* leave es->s_feature_*compat flags alone */
526 /* es->s_uuid will be set by e2fsck if empty */
527
528 /*
529 * The rest of the superblock fields should be zero, and if not it
530 * means they are likely already in use, so leave them alone. We
531 * can leave it up to e2fsck to clean up any inconsistencies there.
532 */
533}
534
535/*
536 * Open the external journal device
537 */
b31e1552 538static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
ac27a0ec
DK		 539{
540 struct block_device *bdev;
541 char b[BDEVNAME_SIZE];
542
543 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
544 if (IS_ERR(bdev))
545 goto fail;
546 return bdev;
547
548fail:
b31e1552 549 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
ac27a0ec
DK		 550 __bdevname(dev, b), PTR_ERR(bdev));
551 return NULL;
552}
553
554/*
555 * Release the journal device
556 */
617ba13b 557static int ext4_blkdev_put(struct block_device *bdev)
ac27a0ec
DK		 558{
559 bd_release(bdev);
9a1c3542 560 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
ac27a0ec
DK		 561}
562
617ba13b 563static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
ac27a0ec
DK		 564{
565 struct block_device *bdev;
566 int ret = -ENODEV;
567
568 bdev = sbi->journal_bdev;
569 if (bdev) {
617ba13b 570 ret = ext4_blkdev_put(bdev);
ac27a0ec
DK		 571 sbi->journal_bdev = NULL;
572 }
573 return ret;
574}
575
576static inline struct inode *orphan_list_entry(struct list_head *l)
577{
617ba13b 578 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
ac27a0ec
DK		 579}
580
617ba13b 581static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
ac27a0ec
DK		 582{
583 struct list_head *l;
584
b31e1552
ES		 585 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
586 le32_to_cpu(sbi->s_es->s_last_orphan));
ac27a0ec
DK		 587
588 printk(KERN_ERR "sb_info orphan list:\n");
589 list_for_each(l, &sbi->s_orphan) {
590 struct inode *inode = orphan_list_entry(l);
591 printk(KERN_ERR " "
592 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
593 inode->i_sb->s_id, inode->i_ino, inode,
594 inode->i_mode, inode->i_nlink,
595 NEXT_ORPHAN(inode));
596 }
597}
598
2b2d6d01 599static void ext4_put_super(struct super_block *sb)
ac27a0ec 600{
617ba13b
MC		 601 struct ext4_sb_info *sbi = EXT4_SB(sb);
602 struct ext4_super_block *es = sbi->s_es;
ef2cabf7 603 int i, err;
ac27a0ec 604
4c0425ff
MC		 605 flush_workqueue(sbi->dio_unwritten_wq);
606 destroy_workqueue(sbi->dio_unwritten_wq);
607
a9e220f8 608 lock_super(sb);
6cfd0148 609 lock_kernel();
8c85e125 610 if (sb->s_dirt)
ebc1ac16 611 ext4_commit_super(sb, 1);
8c85e125 612
6fd058f7 613 ext4_release_system_zone(sb);
c9de560d 614 ext4_mb_release(sb);
a86c6181 615 ext4_ext_release(sb);
617ba13b 616 ext4_xattr_put_super(sb);
0390131b
FM		 617 if (sbi->s_journal) {
618 err = jbd2_journal_destroy(sbi->s_journal);
619 sbi->s_journal = NULL;
620 if (err < 0)
621 ext4_abort(sb, __func__,
622 "Couldn't clean up the journal");
623 }
ac27a0ec 624 if (!(sb->s_flags & MS_RDONLY)) {
617ba13b 625 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 626 es->s_state = cpu_to_le16(sbi->s_mount_state);
e2d67052 627 ext4_commit_super(sb, 1);
ac27a0ec 628 }
240799cd 629 if (sbi->s_proc) {
9f6200bb 630 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 631 }
3197ebdb 632 kobject_del(&sbi->s_kobj);
ac27a0ec
DK		 633
634 for (i = 0; i < sbi->s_gdb_count; i++)
635 brelse(sbi->s_group_desc[i]);
636 kfree(sbi->s_group_desc);
c5ca7c76
TT		 637 if (is_vmalloc_addr(sbi->s_flex_groups))
638 vfree(sbi->s_flex_groups);
639 else
640 kfree(sbi->s_flex_groups);
ac27a0ec
DK		 641 percpu_counter_destroy(&sbi->s_freeblocks_counter);
642 percpu_counter_destroy(&sbi->s_freeinodes_counter);
643 percpu_counter_destroy(&sbi->s_dirs_counter);
6bc6e63f 644 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
ac27a0ec
DK		 645 brelse(sbi->s_sbh);
646#ifdef CONFIG_QUOTA
647 for (i = 0; i < MAXQUOTAS; i++)
648 kfree(sbi->s_qf_names[i]);
649#endif
650
651 /* Debugging code just in case the in-memory inode orphan list
652 * isn't empty. The on-disk one can be non-empty if we've
653 * detected an error and taken the fs readonly, but the
654 * in-memory list had better be clean by this point. */
655 if (!list_empty(&sbi->s_orphan))
656 dump_orphan_list(sb, sbi);
657 J_ASSERT(list_empty(&sbi->s_orphan));
658
f98393a6 659 invalidate_bdev(sb->s_bdev);
ac27a0ec
DK		 660 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
661 /*
662 * Invalidate the journal device's buffers. We don't want them
663 * floating about in memory - the physical journal device may
664 * hotswapped, and it breaks the `ro-after' testing code.
665 */
666 sync_blockdev(sbi->journal_bdev);
f98393a6 667 invalidate_bdev(sbi->journal_bdev);
617ba13b 668 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 669 }
670 sb->s_fs_info = NULL;
3197ebdb
TT		 671 /*
672 * Now that we are completely done shutting down the
673 * superblock, we need to actually destroy the kobject.
674 */
675 unlock_kernel();
676 unlock_super(sb);
677 kobject_put(&sbi->s_kobj);
678 wait_for_completion(&sbi->s_kobj_unregister);
705895b6 679 kfree(sbi->s_blockgroup_lock);
ac27a0ec 680 kfree(sbi);
ac27a0ec
DK		 681}
682
e18b890b 683static struct kmem_cache *ext4_inode_cachep;
ac27a0ec
DK		 684
685/*
686 * Called inside transaction, so use GFP_NOFS
687 */
617ba13b 688static struct inode *ext4_alloc_inode(struct super_block *sb)
ac27a0ec 689{
617ba13b 690 struct ext4_inode_info *ei;
ac27a0ec 691
e6b4f8da 692 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
ac27a0ec
DK		 693 if (!ei)
694 return NULL;
0b8e58a1 695
ac27a0ec 696 ei->vfs_inode.i_version = 1;
91246c00 697 ei->vfs_inode.i_data.writeback_index = 0;
a86c6181 698 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
c9de560d
AT		 699 INIT_LIST_HEAD(&ei->i_prealloc_list);
700 spin_lock_init(&ei->i_prealloc_lock);
0390131b
FM		 701 /*
702 * Note: We can be called before EXT4_SB(sb)->s_journal is set,
703 * therefore it can be null here. Don't check it, just initialize
704 * jinode.
705 */
678aaf48 706 jbd2_journal_init_jbd_inode(&ei->jinode, &ei->vfs_inode);
d2a17637
MC		 707 ei->i_reserved_data_blocks = 0;
708 ei->i_reserved_meta_blocks = 0;
709 ei->i_allocated_meta_blocks = 0;
710 ei->i_delalloc_reserved_flag = 0;
711 spin_lock_init(&(ei->i_block_reservation_lock));
8d5d02e6
MC		 712 INIT_LIST_HEAD(&ei->i_aio_dio_complete_list);
713 ei->cur_aio_dio = NULL;
0b8e58a1 714
ac27a0ec
DK		 715 return &ei->vfs_inode;
716}
717
617ba13b 718static void ext4_destroy_inode(struct inode *inode)
ac27a0ec 719{
9f7dd93d 720 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
b31e1552
ES		 721 ext4_msg(inode->i_sb, KERN_ERR,
722 "Inode %lu (%p): orphan list check failed!",
723 inode->i_ino, EXT4_I(inode));
9f7dd93d
VA		 724 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
725 EXT4_I(inode), sizeof(struct ext4_inode_info),
726 true);
727 dump_stack();
728 }
617ba13b 729 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
ac27a0ec
DK		 730}
731
51cc5068 732static void init_once(void *foo)
ac27a0ec 733{
617ba13b 734 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
ac27a0ec 735
a35afb83 736 INIT_LIST_HEAD(&ei->i_orphan);
03010a33 737#ifdef CONFIG_EXT4_FS_XATTR
a35afb83 738 init_rwsem(&ei->xattr_sem);
ac27a0ec 739#endif
0e855ac8 740 init_rwsem(&ei->i_data_sem);
a35afb83 741 inode_init_once(&ei->vfs_inode);
ac27a0ec
DK		 742}
743
744static int init_inodecache(void)
745{
617ba13b
MC		 746 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
747 sizeof(struct ext4_inode_info),
ac27a0ec
DK		 748 0, (SLAB_RECLAIM_ACCOUNT|
749 SLAB_MEM_SPREAD),
20c2df83 750 init_once);
617ba13b 751 if (ext4_inode_cachep == NULL)
ac27a0ec
DK		 752 return -ENOMEM;
753 return 0;
754}
755
756static void destroy_inodecache(void)
757{
617ba13b 758 kmem_cache_destroy(ext4_inode_cachep);
ac27a0ec
DK		 759}
760
617ba13b 761static void ext4_clear_inode(struct inode *inode)
ac27a0ec 762{
c2ea3fde 763 ext4_discard_preallocations(inode);
0390131b
FM		 764 if (EXT4_JOURNAL(inode))
765 jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal,
678aaf48 766 &EXT4_I(inode)->jinode);
ac27a0ec
DK		 767}
768
2b2d6d01
TT		 769static inline void ext4_show_quota_options(struct seq_file *seq,
770 struct super_block *sb)
ac27a0ec
DK		 771{
772#if defined(CONFIG_QUOTA)
617ba13b 773 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 774
775 if (sbi->s_jquota_fmt)
776 seq_printf(seq, ",jqfmt=%s",
af5bc92d 777 (sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold" : "vfsv0");
ac27a0ec
DK		 778
779 if (sbi->s_qf_names[USRQUOTA])
780 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
781
782 if (sbi->s_qf_names[GRPQUOTA])
783 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
784
617ba13b 785 if (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA)
ac27a0ec
DK		 786 seq_puts(seq, ",usrquota");
787
617ba13b 788 if (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)
ac27a0ec
DK		 789 seq_puts(seq, ",grpquota");
790#endif
791}
792
d9c9bef1
MS		 793/*
794 * Show an option if
795 * - it's set to a non-default value OR
796 * - if the per-sb default is different from the global default
797 */
617ba13b 798static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
ac27a0ec 799{
aa22df2c
AK		 800 int def_errors;
801 unsigned long def_mount_opts;
ac27a0ec 802 struct super_block *sb = vfs->mnt_sb;
d9c9bef1
MS		 803 struct ext4_sb_info *sbi = EXT4_SB(sb);
804 struct ext4_super_block *es = sbi->s_es;
d9c9bef1
MS		 805
806 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
aa22df2c 807 def_errors = le16_to_cpu(es->s_errors);
d9c9bef1
MS		 808
809 if (sbi->s_sb_block != 1)
810 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
811 if (test_opt(sb, MINIX_DF))
812 seq_puts(seq, ",minixdf");
aa22df2c 813 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
d9c9bef1
MS		 814 seq_puts(seq, ",grpid");
815 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
816 seq_puts(seq, ",nogrpid");
817 if (sbi->s_resuid != EXT4_DEF_RESUID ||
818 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
819 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
820 }
821 if (sbi->s_resgid != EXT4_DEF_RESGID ||
822 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
823 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
824 }
bb4f397a 825 if (test_opt(sb, ERRORS_RO)) {
d9c9bef1 826 if (def_errors == EXT4_ERRORS_PANIC ||
bb4f397a
AK		 827 def_errors == EXT4_ERRORS_CONTINUE) {
828 seq_puts(seq, ",errors=remount-ro");
d9c9bef1
MS		 829 }
830 }
aa22df2c 831 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
bb4f397a 832 seq_puts(seq, ",errors=continue");
aa22df2c 833 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
d9c9bef1 834 seq_puts(seq, ",errors=panic");
aa22df2c 835 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
d9c9bef1 836 seq_puts(seq, ",nouid32");
aa22df2c 837 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
d9c9bef1
MS		 838 seq_puts(seq, ",debug");
839 if (test_opt(sb, OLDALLOC))
840 seq_puts(seq, ",oldalloc");
03010a33 841#ifdef CONFIG_EXT4_FS_XATTR
aa22df2c
AK		 842 if (test_opt(sb, XATTR_USER) &&
843 !(def_mount_opts & EXT4_DEFM_XATTR_USER))
d9c9bef1
MS		 844 seq_puts(seq, ",user_xattr");
845 if (!test_opt(sb, XATTR_USER) &&
846 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
847 seq_puts(seq, ",nouser_xattr");
848 }
849#endif
03010a33 850#ifdef CONFIG_EXT4_FS_POSIX_ACL
aa22df2c 851 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
d9c9bef1
MS		 852 seq_puts(seq, ",acl");
853 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
854 seq_puts(seq, ",noacl");
855#endif
30773840 856 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
d9c9bef1
MS		 857 seq_printf(seq, ",commit=%u",
858 (unsigned) (sbi->s_commit_interval / HZ));
859 }
30773840
TT		 860 if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
861 seq_printf(seq, ",min_batch_time=%u",
862 (unsigned) sbi->s_min_batch_time);
863 }
864 if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
865 seq_printf(seq, ",max_batch_time=%u",
866 (unsigned) sbi->s_min_batch_time);
867 }
868
571640ca
ES		 869 /*
870 * We're changing the default of barrier mount option, so
871 * let's always display its mount state so it's clear what its
872 * status is.
873 */
874 seq_puts(seq, ",barrier=");
875 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
cd0b6a39
TT		 876 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
877 seq_puts(seq, ",journal_async_commit");
d9c9bef1
MS		 878 if (test_opt(sb, NOBH))
879 seq_puts(seq, ",nobh");
25ec56b5
JNC		 880 if (test_opt(sb, I_VERSION))
881 seq_puts(seq, ",i_version");
dd919b98
AK		 882 if (!test_opt(sb, DELALLOC))
883 seq_puts(seq, ",nodelalloc");
884
ac27a0ec 885
cb45bbe4
MS		 886 if (sbi->s_stripe)
887 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
aa22df2c
AK		 888 /*
889 * journal mode get enabled in different ways
890 * So just print the value even if we didn't specify it
891 */
617ba13b 892 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
ac27a0ec 893 seq_puts(seq, ",data=journal");
617ba13b 894 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
ac27a0ec 895 seq_puts(seq, ",data=ordered");
617ba13b 896 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
ac27a0ec
DK		 897 seq_puts(seq, ",data=writeback");
898
240799cd
TT		 899 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
900 seq_printf(seq, ",inode_readahead_blks=%u",
901 sbi->s_inode_readahead_blks);
902
5bf5683a
HK		 903 if (test_opt(sb, DATA_ERR_ABORT))
904 seq_puts(seq, ",data_err=abort");
905
afd4672d 906 if (test_opt(sb, NO_AUTO_DA_ALLOC))
06705bff 907 seq_puts(seq, ",noauto_da_alloc");
afd4672d 908
617ba13b 909 ext4_show_quota_options(seq, sb);
0b8e58a1 910
ac27a0ec
DK		 911 return 0;
912}
913
1b961ac0 914static struct inode *ext4_nfs_get_inode(struct super_block *sb,
0b8e58a1 915 u64 ino, u32 generation)
ac27a0ec 916{
ac27a0ec 917 struct inode *inode;
ac27a0ec 918
617ba13b 919 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
ac27a0ec 920 return ERR_PTR(-ESTALE);
617ba13b 921 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
ac27a0ec
DK		 922 return ERR_PTR(-ESTALE);
923
924 /* iget isn't really right if the inode is currently unallocated!!
925 *
617ba13b 926 * ext4_read_inode will return a bad_inode if the inode had been
ac27a0ec
DK		 927 * deleted, so we should be safe.
928 *
929 * Currently we don't know the generation for parent directory, so
930 * a generation of 0 means "accept any"
931 */
1d1fe1ee
DH		 932 inode = ext4_iget(sb, ino);
933 if (IS_ERR(inode))
934 return ERR_CAST(inode);
935 if (generation && inode->i_generation != generation) {
ac27a0ec
DK		 936 iput(inode);
937 return ERR_PTR(-ESTALE);
938 }
1b961ac0
CH		 939
940 return inode;
941}
942
943static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
0b8e58a1 944 int fh_len, int fh_type)
1b961ac0
CH		 945{
946 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
947 ext4_nfs_get_inode);
948}
949
950static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
0b8e58a1 951 int fh_len, int fh_type)
1b961ac0
CH		 952{
953 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
954 ext4_nfs_get_inode);
ac27a0ec
DK		 955}
956
c39a7f84
TO		 957/*
958 * Try to release metadata pages (indirect blocks, directories) which are
959 * mapped via the block device. Since these pages could have journal heads
960 * which would prevent try_to_free_buffers() from freeing them, we must use
961 * jbd2 layer's try_to_free_buffers() function to release them.
962 */
0b8e58a1
AD		 963static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
964 gfp_t wait)
c39a7f84
TO		 965{
966 journal_t *journal = EXT4_SB(sb)->s_journal;
967
968 WARN_ON(PageChecked(page));
969 if (!page_has_buffers(page))
970 return 0;
971 if (journal)
972 return jbd2_journal_try_to_free_buffers(journal, page,
973 wait & ~__GFP_WAIT);
974 return try_to_free_buffers(page);
975}
976
ac27a0ec 977#ifdef CONFIG_QUOTA
af5bc92d 978#define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
2b2d6d01 979#define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
ac27a0ec 980
617ba13b
MC		 981static int ext4_write_dquot(struct dquot *dquot);
982static int ext4_acquire_dquot(struct dquot *dquot);
983static int ext4_release_dquot(struct dquot *dquot);
984static int ext4_mark_dquot_dirty(struct dquot *dquot);
985static int ext4_write_info(struct super_block *sb, int type);
6f28e087
JK		 986static int ext4_quota_on(struct super_block *sb, int type, int format_id,
987 char *path, int remount);
617ba13b
MC		 988static int ext4_quota_on_mount(struct super_block *sb, int type);
989static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 990 size_t len, loff_t off);
617ba13b 991static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 992 const char *data, size_t len, loff_t off);
993
61e225dc 994static const struct dquot_operations ext4_quota_operations = {
edf72453
JK		 995 .initialize = dquot_initialize,
996 .drop = dquot_drop,
ac27a0ec 997 .alloc_space = dquot_alloc_space,
60e58e0f
MC		 998 .reserve_space = dquot_reserve_space,
999 .claim_space = dquot_claim_space,
1000 .release_rsv = dquot_release_reserved_space,
1001 .get_reserved_space = ext4_get_reserved_space,
ac27a0ec
DK		 1002 .alloc_inode = dquot_alloc_inode,
1003 .free_space = dquot_free_space,
1004 .free_inode = dquot_free_inode,
1005 .transfer = dquot_transfer,
617ba13b
MC		 1006 .write_dquot = ext4_write_dquot,
1007 .acquire_dquot = ext4_acquire_dquot,
1008 .release_dquot = ext4_release_dquot,
1009 .mark_dirty = ext4_mark_dquot_dirty,
a5b5ee32
JK		 1010 .write_info = ext4_write_info,
1011 .alloc_dquot = dquot_alloc,
1012 .destroy_dquot = dquot_destroy,
ac27a0ec
DK		 1013};
1014
0d54b217 1015static const struct quotactl_ops ext4_qctl_operations = {
617ba13b 1016 .quota_on = ext4_quota_on,
ac27a0ec
DK		 1017 .quota_off = vfs_quota_off,
1018 .quota_sync = vfs_quota_sync,
1019 .get_info = vfs_get_dqinfo,
1020 .set_info = vfs_set_dqinfo,
1021 .get_dqblk = vfs_get_dqblk,
1022 .set_dqblk = vfs_set_dqblk
1023};
1024#endif
1025
ee9b6d61 1026static const struct super_operations ext4_sops = {
617ba13b
MC		 1027 .alloc_inode = ext4_alloc_inode,
1028 .destroy_inode = ext4_destroy_inode,
617ba13b
MC		 1029 .write_inode = ext4_write_inode,
1030 .dirty_inode = ext4_dirty_inode,
1031 .delete_inode = ext4_delete_inode,
1032 .put_super = ext4_put_super,
617ba13b 1033 .sync_fs = ext4_sync_fs,
c4be0c1d
TS		 1034 .freeze_fs = ext4_freeze,
1035 .unfreeze_fs = ext4_unfreeze,
617ba13b
MC		 1036 .statfs = ext4_statfs,
1037 .remount_fs = ext4_remount,
1038 .clear_inode = ext4_clear_inode,
1039 .show_options = ext4_show_options,
ac27a0ec 1040#ifdef CONFIG_QUOTA
617ba13b
MC		 1041 .quota_read = ext4_quota_read,
1042 .quota_write = ext4_quota_write,
ac27a0ec 1043#endif
c39a7f84 1044 .bdev_try_to_free_page = bdev_try_to_free_page,
ac27a0ec
DK		 1045};
1046
9ca92389
TT		 1047static const struct super_operations ext4_nojournal_sops = {
1048 .alloc_inode = ext4_alloc_inode,
1049 .destroy_inode = ext4_destroy_inode,
1050 .write_inode = ext4_write_inode,
1051 .dirty_inode = ext4_dirty_inode,
1052 .delete_inode = ext4_delete_inode,
1053 .write_super = ext4_write_super,
1054 .put_super = ext4_put_super,
1055 .statfs = ext4_statfs,
1056 .remount_fs = ext4_remount,
1057 .clear_inode = ext4_clear_inode,
1058 .show_options = ext4_show_options,
1059#ifdef CONFIG_QUOTA
1060 .quota_read = ext4_quota_read,
1061 .quota_write = ext4_quota_write,
1062#endif
1063 .bdev_try_to_free_page = bdev_try_to_free_page,
1064};
1065
39655164 1066static const struct export_operations ext4_export_ops = {
1b961ac0
CH		 1067 .fh_to_dentry = ext4_fh_to_dentry,
1068 .fh_to_parent = ext4_fh_to_parent,
617ba13b 1069 .get_parent = ext4_get_parent,
ac27a0ec
DK		 1070};
1071
1072enum {
1073 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1074 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
01436ef2 1075 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
ac27a0ec 1076 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
06705bff 1077 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh,
30773840 1078 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
c3191067 1079 Opt_journal_update, Opt_journal_dev,
818d276c 1080 Opt_journal_checksum, Opt_journal_async_commit,
ac27a0ec 1081 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
f4033903 1082 Opt_data_err_abort, Opt_data_err_ignore, Opt_mb_history_length,
ac27a0ec
DK		 1083 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1084 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
06705bff
TT		 1085 Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err, Opt_resize,
1086 Opt_usrquota, Opt_grpquota, Opt_i_version,
01436ef2 1087 Opt_stripe, Opt_delalloc, Opt_nodelalloc,
6fd058f7 1088 Opt_block_validity, Opt_noblock_validity,
b3881f74 1089 Opt_inode_readahead_blks, Opt_journal_ioprio
ac27a0ec
DK		 1090};
1091
a447c093 1092static const match_table_t tokens = {
ac27a0ec
DK		 1093 {Opt_bsd_df, "bsddf"},
1094 {Opt_minix_df, "minixdf"},
1095 {Opt_grpid, "grpid"},
1096 {Opt_grpid, "bsdgroups"},
1097 {Opt_nogrpid, "nogrpid"},
1098 {Opt_nogrpid, "sysvgroups"},
1099 {Opt_resgid, "resgid=%u"},
1100 {Opt_resuid, "resuid=%u"},
1101 {Opt_sb, "sb=%u"},
1102 {Opt_err_cont, "errors=continue"},
1103 {Opt_err_panic, "errors=panic"},
1104 {Opt_err_ro, "errors=remount-ro"},
1105 {Opt_nouid32, "nouid32"},
ac27a0ec
DK		 1106 {Opt_debug, "debug"},
1107 {Opt_oldalloc, "oldalloc"},
1108 {Opt_orlov, "orlov"},
1109 {Opt_user_xattr, "user_xattr"},
1110 {Opt_nouser_xattr, "nouser_xattr"},
1111 {Opt_acl, "acl"},
1112 {Opt_noacl, "noacl"},
ac27a0ec
DK		 1113 {Opt_noload, "noload"},
1114 {Opt_nobh, "nobh"},
1115 {Opt_bh, "bh"},
1116 {Opt_commit, "commit=%u"},
30773840
TT		 1117 {Opt_min_batch_time, "min_batch_time=%u"},
1118 {Opt_max_batch_time, "max_batch_time=%u"},
ac27a0ec 1119 {Opt_journal_update, "journal=update"},
ac27a0ec 1120 {Opt_journal_dev, "journal_dev=%u"},
818d276c
GS		 1121 {Opt_journal_checksum, "journal_checksum"},
1122 {Opt_journal_async_commit, "journal_async_commit"},
ac27a0ec
DK		 1123 {Opt_abort, "abort"},
1124 {Opt_data_journal, "data=journal"},
1125 {Opt_data_ordered, "data=ordered"},
1126 {Opt_data_writeback, "data=writeback"},
5bf5683a
HK		 1127 {Opt_data_err_abort, "data_err=abort"},
1128 {Opt_data_err_ignore, "data_err=ignore"},
f4033903 1129 {Opt_mb_history_length, "mb_history_length=%u"},
ac27a0ec
DK		 1130 {Opt_offusrjquota, "usrjquota="},
1131 {Opt_usrjquota, "usrjquota=%s"},
1132 {Opt_offgrpjquota, "grpjquota="},
1133 {Opt_grpjquota, "grpjquota=%s"},
1134 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1135 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1136 {Opt_grpquota, "grpquota"},
1137 {Opt_noquota, "noquota"},
1138 {Opt_quota, "quota"},
1139 {Opt_usrquota, "usrquota"},
1140 {Opt_barrier, "barrier=%u"},
06705bff
TT		 1141 {Opt_barrier, "barrier"},
1142 {Opt_nobarrier, "nobarrier"},
25ec56b5 1143 {Opt_i_version, "i_version"},
c9de560d 1144 {Opt_stripe, "stripe=%u"},
ac27a0ec 1145 {Opt_resize, "resize"},
64769240 1146 {Opt_delalloc, "delalloc"},
dd919b98 1147 {Opt_nodelalloc, "nodelalloc"},
6fd058f7
TT		 1148 {Opt_block_validity, "block_validity"},
1149 {Opt_noblock_validity, "noblock_validity"},
240799cd 1150 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
b3881f74 1151 {Opt_journal_ioprio, "journal_ioprio=%u"},
afd4672d 1152 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
06705bff
TT		 1153 {Opt_auto_da_alloc, "auto_da_alloc"},
1154 {Opt_noauto_da_alloc, "noauto_da_alloc"},
f3f12faa 1155 {Opt_err, NULL},
ac27a0ec
DK		 1156};
1157
617ba13b 1158static ext4_fsblk_t get_sb_block(void **data)
ac27a0ec 1159{
617ba13b 1160 ext4_fsblk_t sb_block;
ac27a0ec
DK		 1161 char *options = (char *) *data;
1162
1163 if (!options || strncmp(options, "sb=", 3) != 0)
1164 return 1; /* Default location */
0b8e58a1 1165
ac27a0ec 1166 options += 3;
0b8e58a1 1167 /* TODO: use simple_strtoll with >32bit ext4 */
ac27a0ec
DK		 1168 sb_block = simple_strtoul(options, &options, 0);
1169 if (*options && *options != ',') {
4776004f 1170 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
ac27a0ec
DK		 1171 (char *) *data);
1172 return 1;
1173 }
1174 if (*options == ',')
1175 options++;
1176 *data = (void *) options;
0b8e58a1 1177
ac27a0ec
DK		 1178 return sb_block;
1179}
1180
b3881f74
TT		 1181#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1182
2b2d6d01 1183static int parse_options(char *options, struct super_block *sb,
c3191067 1184 unsigned long *journal_devnum,
b3881f74 1185 unsigned int *journal_ioprio,
2b2d6d01 1186 ext4_fsblk_t *n_blocks_count, int is_remount)
ac27a0ec 1187{
617ba13b 1188 struct ext4_sb_info *sbi = EXT4_SB(sb);
2b2d6d01 1189 char *p;
ac27a0ec
DK		 1190 substring_t args[MAX_OPT_ARGS];
1191 int data_opt = 0;
1192 int option;
1193#ifdef CONFIG_QUOTA
dfc5d03f 1194 int qtype, qfmt;
ac27a0ec
DK		 1195 char *qname;
1196#endif
1197
1198 if (!options)
1199 return 1;
1200
2b2d6d01 1201 while ((p = strsep(&options, ",")) != NULL) {
ac27a0ec
DK		 1202 int token;
1203 if (!*p)
1204 continue;
1205
1206 token = match_token(p, tokens, args);
1207 switch (token) {
1208 case Opt_bsd_df:
2b2d6d01 1209 clear_opt(sbi->s_mount_opt, MINIX_DF);
ac27a0ec
DK		 1210 break;
1211 case Opt_minix_df:
2b2d6d01 1212 set_opt(sbi->s_mount_opt, MINIX_DF);
ac27a0ec
DK		 1213 break;
1214 case Opt_grpid:
2b2d6d01 1215 set_opt(sbi->s_mount_opt, GRPID);
ac27a0ec
DK		 1216 break;
1217 case Opt_nogrpid:
2b2d6d01 1218 clear_opt(sbi->s_mount_opt, GRPID);
ac27a0ec
DK		 1219 break;
1220 case Opt_resuid:
1221 if (match_int(&args[0], &option))
1222 return 0;
1223 sbi->s_resuid = option;
1224 break;
1225 case Opt_resgid:
1226 if (match_int(&args[0], &option))
1227 return 0;
1228 sbi->s_resgid = option;
1229 break;
1230 case Opt_sb:
1231 /* handled by get_sb_block() instead of here */
1232 /* *sb_block = match_int(&args[0]); */
1233 break;
1234 case Opt_err_panic:
2b2d6d01
TT		 1235 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1236 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1237 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
ac27a0ec
DK		 1238 break;
1239 case Opt_err_ro:
2b2d6d01
TT		 1240 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1241 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1242 set_opt(sbi->s_mount_opt, ERRORS_RO);
ac27a0ec
DK		 1243 break;
1244 case Opt_err_cont:
2b2d6d01
TT		 1245 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1246 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1247 set_opt(sbi->s_mount_opt, ERRORS_CONT);
ac27a0ec
DK		 1248 break;
1249 case Opt_nouid32:
2b2d6d01 1250 set_opt(sbi->s_mount_opt, NO_UID32);
ac27a0ec 1251 break;
ac27a0ec 1252 case Opt_debug:
2b2d6d01 1253 set_opt(sbi->s_mount_opt, DEBUG);
ac27a0ec
DK		 1254 break;
1255 case Opt_oldalloc:
2b2d6d01 1256 set_opt(sbi->s_mount_opt, OLDALLOC);
ac27a0ec
DK		 1257 break;
1258 case Opt_orlov:
2b2d6d01 1259 clear_opt(sbi->s_mount_opt, OLDALLOC);
ac27a0ec 1260 break;
03010a33 1261#ifdef CONFIG_EXT4_FS_XATTR
ac27a0ec 1262 case Opt_user_xattr:
2b2d6d01 1263 set_opt(sbi->s_mount_opt, XATTR_USER);
ac27a0ec
DK		 1264 break;
1265 case Opt_nouser_xattr:
2b2d6d01 1266 clear_opt(sbi->s_mount_opt, XATTR_USER);
ac27a0ec
DK		 1267 break;
1268#else
1269 case Opt_user_xattr:
1270 case Opt_nouser_xattr:
b31e1552 1271 ext4_msg(sb, KERN_ERR, "(no)user_xattr options not supported");
ac27a0ec
DK		 1272 break;
1273#endif
03010a33 1274#ifdef CONFIG_EXT4_FS_POSIX_ACL
ac27a0ec
DK		 1275 case Opt_acl:
1276 set_opt(sbi->s_mount_opt, POSIX_ACL);
1277 break;
1278 case Opt_noacl:
1279 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1280 break;
1281#else
1282 case Opt_acl:
1283 case Opt_noacl:
b31e1552 1284 ext4_msg(sb, KERN_ERR, "(no)acl options not supported");
ac27a0ec
DK		 1285 break;
1286#endif
ac27a0ec
DK		 1287 case Opt_journal_update:
1288 /* @@@ FIXME */
1289 /* Eventually we will want to be able to create
1290 a journal file here. For now, only allow the
1291 user to specify an existing inode to be the
1292 journal file. */
1293 if (is_remount) {
b31e1552
ES		 1294 ext4_msg(sb, KERN_ERR,
1295 "Cannot specify journal on remount");
ac27a0ec
DK		 1296 return 0;
1297 }
2b2d6d01 1298 set_opt(sbi->s_mount_opt, UPDATE_JOURNAL);
ac27a0ec 1299 break;
ac27a0ec
DK		 1300 case Opt_journal_dev:
1301 if (is_remount) {
b31e1552
ES		 1302 ext4_msg(sb, KERN_ERR,
1303 "Cannot specify journal on remount");
ac27a0ec
DK		 1304 return 0;
1305 }
1306 if (match_int(&args[0], &option))
1307 return 0;
1308 *journal_devnum = option;
1309 break;
818d276c 1310 case Opt_journal_checksum:
d0646f7b 1311 break; /* Kept for backwards compatibility */
818d276c
GS		 1312 case Opt_journal_async_commit:
1313 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
818d276c 1314 break;
ac27a0ec 1315 case Opt_noload:
2b2d6d01 1316 set_opt(sbi->s_mount_opt, NOLOAD);
ac27a0ec
DK		 1317 break;
1318 case Opt_commit:
1319 if (match_int(&args[0], &option))
1320 return 0;
1321 if (option < 0)
1322 return 0;
1323 if (option == 0)
cd02ff0b 1324 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
ac27a0ec
DK		 1325 sbi->s_commit_interval = HZ * option;
1326 break;
30773840
TT		 1327 case Opt_max_batch_time:
1328 if (match_int(&args[0], &option))
1329 return 0;
1330 if (option < 0)
1331 return 0;
1332 if (option == 0)
1333 option = EXT4_DEF_MAX_BATCH_TIME;
1334 sbi->s_max_batch_time = option;
1335 break;
1336 case Opt_min_batch_time:
1337 if (match_int(&args[0], &option))
1338 return 0;
1339 if (option < 0)
1340 return 0;
1341 sbi->s_min_batch_time = option;
1342 break;
ac27a0ec 1343 case Opt_data_journal:
617ba13b 1344 data_opt = EXT4_MOUNT_JOURNAL_DATA;
ac27a0ec
DK		 1345 goto datacheck;
1346 case Opt_data_ordered:
617ba13b 1347 data_opt = EXT4_MOUNT_ORDERED_DATA;
ac27a0ec
DK		 1348 goto datacheck;
1349 case Opt_data_writeback:
617ba13b 1350 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
ac27a0ec
DK		 1351 datacheck:
1352 if (is_remount) {
617ba13b 1353 if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
ac27a0ec 1354 != data_opt) {
b31e1552
ES		 1355 ext4_msg(sb, KERN_ERR,
1356 "Cannot change data mode on remount");
ac27a0ec
DK		 1357 return 0;
1358 }
1359 } else {
617ba13b 1360 sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
ac27a0ec
DK		 1361 sbi->s_mount_opt |= data_opt;
1362 }
1363 break;
5bf5683a
HK		 1364 case Opt_data_err_abort:
1365 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1366 break;
1367 case Opt_data_err_ignore:
1368 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1369 break;
f4033903
CW		 1370 case Opt_mb_history_length:
1371 if (match_int(&args[0], &option))
1372 return 0;
1373 if (option < 0)
1374 return 0;
1375 sbi->s_mb_history_max = option;
1376 break;
ac27a0ec
DK		 1377#ifdef CONFIG_QUOTA
1378 case Opt_usrjquota:
1379 qtype = USRQUOTA;
1380 goto set_qf_name;
1381 case Opt_grpjquota:
1382 qtype = GRPQUOTA;
1383set_qf_name:
17bd13b3 1384 if (sb_any_quota_loaded(sb) &&
dfc5d03f 1385 !sbi->s_qf_names[qtype]) {
b31e1552
ES		 1386 ext4_msg(sb, KERN_ERR,
1387 "Cannot change journaled "
1388 "quota options when quota turned on");
ac27a0ec
DK		 1389 return 0;
1390 }
1391 qname = match_strdup(&args[0]);
1392 if (!qname) {
b31e1552
ES		 1393 ext4_msg(sb, KERN_ERR,
1394 "Not enough memory for "
1395 "storing quotafile name");
ac27a0ec
DK		 1396 return 0;
1397 }
1398 if (sbi->s_qf_names[qtype] &&
1399 strcmp(sbi->s_qf_names[qtype], qname)) {
b31e1552
ES		 1400 ext4_msg(sb, KERN_ERR,
1401 "%s quota file already "
1402 "specified", QTYPE2NAME(qtype));
ac27a0ec
DK		 1403 kfree(qname);
1404 return 0;
1405 }
1406 sbi->s_qf_names[qtype] = qname;
1407 if (strchr(sbi->s_qf_names[qtype], '/')) {
b31e1552
ES		 1408 ext4_msg(sb, KERN_ERR,
1409 "quotafile must be on "
1410 "filesystem root");
ac27a0ec
DK		 1411 kfree(sbi->s_qf_names[qtype]);
1412 sbi->s_qf_names[qtype] = NULL;
1413 return 0;
1414 }
1415 set_opt(sbi->s_mount_opt, QUOTA);
1416 break;
1417 case Opt_offusrjquota:
1418 qtype = USRQUOTA;
1419 goto clear_qf_name;
1420 case Opt_offgrpjquota:
1421 qtype = GRPQUOTA;
1422clear_qf_name:
17bd13b3 1423 if (sb_any_quota_loaded(sb) &&
dfc5d03f 1424 sbi->s_qf_names[qtype]) {
b31e1552 1425 ext4_msg(sb, KERN_ERR, "Cannot change "
2c8be6b2 1426 "journaled quota options when "
b31e1552 1427 "quota turned on");
ac27a0ec
DK		 1428 return 0;
1429 }
1430 /*
1431 * The space will be released later when all options
1432 * are confirmed to be correct
1433 */
1434 sbi->s_qf_names[qtype] = NULL;
1435 break;
1436 case Opt_jqfmt_vfsold:
dfc5d03f
JK		 1437 qfmt = QFMT_VFS_OLD;
1438 goto set_qf_format;
ac27a0ec 1439 case Opt_jqfmt_vfsv0:
dfc5d03f
JK		 1440 qfmt = QFMT_VFS_V0;
1441set_qf_format:
17bd13b3 1442 if (sb_any_quota_loaded(sb) &&
dfc5d03f 1443 sbi->s_jquota_fmt != qfmt) {
b31e1552 1444 ext4_msg(sb, KERN_ERR, "Cannot change "
dfc5d03f 1445 "journaled quota options when "
b31e1552 1446 "quota turned on");
dfc5d03f
JK		 1447 return 0;
1448 }
1449 sbi->s_jquota_fmt = qfmt;
ac27a0ec
DK		 1450 break;
1451 case Opt_quota:
1452 case Opt_usrquota:
1453 set_opt(sbi->s_mount_opt, QUOTA);
1454 set_opt(sbi->s_mount_opt, USRQUOTA);
1455 break;
1456 case Opt_grpquota:
1457 set_opt(sbi->s_mount_opt, QUOTA);
1458 set_opt(sbi->s_mount_opt, GRPQUOTA);
1459 break;
1460 case Opt_noquota:
17bd13b3 1461 if (sb_any_quota_loaded(sb)) {
b31e1552
ES		 1462 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1463 "options when quota turned on");
ac27a0ec
DK		 1464 return 0;
1465 }
1466 clear_opt(sbi->s_mount_opt, QUOTA);
1467 clear_opt(sbi->s_mount_opt, USRQUOTA);
1468 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1469 break;
1470#else
1471 case Opt_quota:
1472 case Opt_usrquota:
1473 case Opt_grpquota:
b31e1552
ES		 1474 ext4_msg(sb, KERN_ERR,
1475 "quota options not supported");
cd59e7b9 1476 break;
ac27a0ec
DK		 1477 case Opt_usrjquota:
1478 case Opt_grpjquota:
1479 case Opt_offusrjquota:
1480 case Opt_offgrpjquota:
1481 case Opt_jqfmt_vfsold:
1482 case Opt_jqfmt_vfsv0:
b31e1552
ES		 1483 ext4_msg(sb, KERN_ERR,
1484 "journaled quota options not supported");
ac27a0ec
DK		 1485 break;
1486 case Opt_noquota:
1487 break;
1488#endif
1489 case Opt_abort:
4ab2f15b 1490 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
ac27a0ec 1491 break;
06705bff
TT		 1492 case Opt_nobarrier:
1493 clear_opt(sbi->s_mount_opt, BARRIER);
1494 break;
ac27a0ec 1495 case Opt_barrier:
06705bff
TT		 1496 if (match_int(&args[0], &option)) {
1497 set_opt(sbi->s_mount_opt, BARRIER);
1498 break;
1499 }
ac27a0ec
DK		 1500 if (option)
1501 set_opt(sbi->s_mount_opt, BARRIER);
1502 else
1503 clear_opt(sbi->s_mount_opt, BARRIER);
1504 break;
1505 case Opt_ignore:
1506 break;
1507 case Opt_resize:
1508 if (!is_remount) {
b31e1552
ES		 1509 ext4_msg(sb, KERN_ERR,
1510 "resize option only available "
1511 "for remount");
ac27a0ec
DK		 1512 return 0;
1513 }
1514 if (match_int(&args[0], &option) != 0)
1515 return 0;
1516 *n_blocks_count = option;
1517 break;
1518 case Opt_nobh:
1519 set_opt(sbi->s_mount_opt, NOBH);
1520 break;
1521 case Opt_bh:
1522 clear_opt(sbi->s_mount_opt, NOBH);
1523 break;
25ec56b5
JNC		 1524 case Opt_i_version:
1525 set_opt(sbi->s_mount_opt, I_VERSION);
1526 sb->s_flags |= MS_I_VERSION;
1527 break;
dd919b98
AK		 1528 case Opt_nodelalloc:
1529 clear_opt(sbi->s_mount_opt, DELALLOC);
1530 break;
c9de560d
AT		 1531 case Opt_stripe:
1532 if (match_int(&args[0], &option))
1533 return 0;
1534 if (option < 0)
1535 return 0;
1536 sbi->s_stripe = option;
1537 break;
64769240
AT		 1538 case Opt_delalloc:
1539 set_opt(sbi->s_mount_opt, DELALLOC);
1540 break;
6fd058f7
TT		 1541 case Opt_block_validity:
1542 set_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1543 break;
1544 case Opt_noblock_validity:
1545 clear_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1546 break;
240799cd
TT		 1547 case Opt_inode_readahead_blks:
1548 if (match_int(&args[0], &option))
1549 return 0;
1550 if (option < 0 || option > (1 << 30))
1551 return 0;
f7c43950 1552 if (!is_power_of_2(option)) {
b31e1552
ES		 1553 ext4_msg(sb, KERN_ERR,
1554 "EXT4-fs: inode_readahead_blks"
1555 " must be a power of 2");
3197ebdb
TT		 1556 return 0;
1557 }
240799cd
TT		 1558 sbi->s_inode_readahead_blks = option;
1559 break;
b3881f74
TT		 1560 case Opt_journal_ioprio:
1561 if (match_int(&args[0], &option))
1562 return 0;
1563 if (option < 0 || option > 7)
1564 break;
1565 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1566 option);
1567 break;
06705bff
TT		 1568 case Opt_noauto_da_alloc:
1569 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1570 break;
afd4672d 1571 case Opt_auto_da_alloc:
06705bff
TT		 1572 if (match_int(&args[0], &option)) {
1573 clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1574 break;
1575 }
afd4672d
TT		 1576 if (option)
1577 clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1578 else
1579 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1580 break;
ac27a0ec 1581 default:
b31e1552
ES		 1582 ext4_msg(sb, KERN_ERR,
1583 "Unrecognized mount option \"%s\" "
1584 "or missing value", p);
ac27a0ec
DK		 1585 return 0;
1586 }
1587 }
1588#ifdef CONFIG_QUOTA
1589 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
617ba13b 1590 if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
ac27a0ec
DK		 1591 sbi->s_qf_names[USRQUOTA])
1592 clear_opt(sbi->s_mount_opt, USRQUOTA);
1593
617ba13b 1594 if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
ac27a0ec
DK		 1595 sbi->s_qf_names[GRPQUOTA])
1596 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1597
1598 if ((sbi->s_qf_names[USRQUOTA] &&
617ba13b 1599 (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
ac27a0ec 1600 (sbi->s_qf_names[GRPQUOTA] &&
617ba13b 1601 (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
b31e1552
ES		 1602 ext4_msg(sb, KERN_ERR, "old and new quota "
1603 "format mixing");
ac27a0ec
DK		 1604 return 0;
1605 }
1606
1607 if (!sbi->s_jquota_fmt) {
b31e1552
ES		 1608 ext4_msg(sb, KERN_ERR, "journaled quota format "
1609 "not specified");
ac27a0ec
DK		 1610 return 0;
1611 }
1612 } else {
1613 if (sbi->s_jquota_fmt) {
b31e1552 1614 ext4_msg(sb, KERN_ERR, "journaled quota format "
2c8be6b2 1615 "specified with no journaling "
b31e1552 1616 "enabled");
ac27a0ec
DK		 1617 return 0;
1618 }
1619 }
1620#endif
1621 return 1;
1622}
1623
617ba13b 1624static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 1625 int read_only)
1626{
617ba13b 1627 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 1628 int res = 0;
1629
617ba13b 1630 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
b31e1552
ES		 1631 ext4_msg(sb, KERN_ERR, "revision level too high, "
1632 "forcing read-only mode");
ac27a0ec
DK		 1633 res = MS_RDONLY;
1634 }
1635 if (read_only)
1636 return res;
617ba13b 1637 if (!(sbi->s_mount_state & EXT4_VALID_FS))
b31e1552
ES		 1638 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1639 "running e2fsck is recommended");
617ba13b 1640 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
b31e1552
ES		 1641 ext4_msg(sb, KERN_WARNING,
1642 "warning: mounting fs with errors, "
1643 "running e2fsck is recommended");
ac27a0ec
DK		 1644 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1645 le16_to_cpu(es->s_mnt_count) >=
1646 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
b31e1552
ES		 1647 ext4_msg(sb, KERN_WARNING,
1648 "warning: maximal mount count reached, "
1649 "running e2fsck is recommended");
ac27a0ec
DK		 1650 else if (le32_to_cpu(es->s_checkinterval) &&
1651 (le32_to_cpu(es->s_lastcheck) +
1652 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
b31e1552
ES		 1653 ext4_msg(sb, KERN_WARNING,
1654 "warning: checktime reached, "
1655 "running e2fsck is recommended");
0b8e58a1 1656 if (!sbi->s_journal)
0390131b 1657 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec 1658 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 1659 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 1660 le16_add_cpu(&es->s_mnt_count, 1);
ac27a0ec 1661 es->s_mtime = cpu_to_le32(get_seconds());
617ba13b 1662 ext4_update_dynamic_rev(sb);
0390131b
FM		 1663 if (sbi->s_journal)
1664 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 1665
e2d67052 1666 ext4_commit_super(sb, 1);
ac27a0ec 1667 if (test_opt(sb, DEBUG))
a9df9a49 1668 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
7f4520cc 1669 "bpg=%lu, ipg=%lu, mo=%04x]\n",
ac27a0ec
DK		 1670 sb->s_blocksize,
1671 sbi->s_groups_count,
617ba13b
MC		 1672 EXT4_BLOCKS_PER_GROUP(sb),
1673 EXT4_INODES_PER_GROUP(sb),
ac27a0ec
DK		 1674 sbi->s_mount_opt);
1675
0390131b 1676 if (EXT4_SB(sb)->s_journal) {
b31e1552
ES		 1677 ext4_msg(sb, KERN_INFO, "%s journal on %s",
1678 EXT4_SB(sb)->s_journal->j_inode ? "internal" :
0390131b
FM		 1679 "external", EXT4_SB(sb)->s_journal->j_devname);
1680 } else {
b31e1552 1681 ext4_msg(sb, KERN_INFO, "no journal");
0390131b 1682 }
ac27a0ec
DK		 1683 return res;
1684}
1685
772cb7c8
JS		 1686static int ext4_fill_flex_info(struct super_block *sb)
1687{
1688 struct ext4_sb_info *sbi = EXT4_SB(sb);
1689 struct ext4_group_desc *gdp = NULL;
772cb7c8
JS		 1690 ext4_group_t flex_group_count;
1691 ext4_group_t flex_group;
1692 int groups_per_flex = 0;
c5ca7c76 1693 size_t size;
772cb7c8
JS		 1694 int i;
1695
1696 if (!sbi->s_es->s_log_groups_per_flex) {
1697 sbi->s_log_groups_per_flex = 0;
1698 return 1;
1699 }
1700
1701 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1702 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1703
c62a11fd
FB		 1704 /* We allocate both existing and potentially added groups */
1705 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
d94e99a6
AK		 1706 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1707 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
c5ca7c76
TT		 1708 size = flex_group_count * sizeof(struct flex_groups);
1709 sbi->s_flex_groups = kzalloc(size, GFP_KERNEL);
1710 if (sbi->s_flex_groups == NULL) {
1711 sbi->s_flex_groups = vmalloc(size);
1712 if (sbi->s_flex_groups)
1713 memset(sbi->s_flex_groups, 0, size);
1714 }
772cb7c8 1715 if (sbi->s_flex_groups == NULL) {
b31e1552
ES		 1716 ext4_msg(sb, KERN_ERR, "not enough memory for "
1717 "%u flex groups", flex_group_count);
772cb7c8
JS		 1718 goto failed;
1719 }
772cb7c8 1720
772cb7c8 1721 for (i = 0; i < sbi->s_groups_count; i++) {
88b6edd1 1722 gdp = ext4_get_group_desc(sb, i, NULL);
772cb7c8
JS		 1723
1724 flex_group = ext4_flex_group(sbi, i);
7ad9bb65
TT		 1725 atomic_add(ext4_free_inodes_count(sb, gdp),
1726 &sbi->s_flex_groups[flex_group].free_inodes);
1727 atomic_add(ext4_free_blks_count(sb, gdp),
1728 &sbi->s_flex_groups[flex_group].free_blocks);
1729 atomic_add(ext4_used_dirs_count(sb, gdp),
1730 &sbi->s_flex_groups[flex_group].used_dirs);
772cb7c8
JS		 1731 }
1732
1733 return 1;
1734failed:
1735 return 0;
1736}
1737
717d50e4
AD		 1738__le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1739 struct ext4_group_desc *gdp)
1740{
1741 __u16 crc = 0;
1742
1743 if (sbi->s_es->s_feature_ro_compat &
1744 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1745 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1746 __le32 le_group = cpu_to_le32(block_group);
1747
1748 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1749 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1750 crc = crc16(crc, (__u8 *)gdp, offset);
1751 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1752 /* for checksum of struct ext4_group_desc do the rest...*/
1753 if ((sbi->s_es->s_feature_incompat &
1754 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1755 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1756 crc = crc16(crc, (__u8 *)gdp + offset,
1757 le16_to_cpu(sbi->s_es->s_desc_size) -
1758 offset);
1759 }
1760
1761 return cpu_to_le16(crc);
1762}
1763
1764int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1765 struct ext4_group_desc *gdp)
1766{
1767 if ((sbi->s_es->s_feature_ro_compat &
1768 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1769 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1770 return 0;
1771
1772 return 1;
1773}
1774
ac27a0ec 1775/* Called at mount-time, super-block is locked */
197cd65a 1776static int ext4_check_descriptors(struct super_block *sb)
ac27a0ec 1777{
617ba13b
MC		 1778 struct ext4_sb_info *sbi = EXT4_SB(sb);
1779 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1780 ext4_fsblk_t last_block;
bd81d8ee
LV		 1781 ext4_fsblk_t block_bitmap;
1782 ext4_fsblk_t inode_bitmap;
1783 ext4_fsblk_t inode_table;
ce421581 1784 int flexbg_flag = 0;
fd2d4291 1785 ext4_group_t i;
ac27a0ec 1786
ce421581
JS		 1787 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1788 flexbg_flag = 1;
1789
af5bc92d 1790 ext4_debug("Checking group descriptors");
ac27a0ec 1791
197cd65a
AM		 1792 for (i = 0; i < sbi->s_groups_count; i++) {
1793 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1794
ce421581 1795 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 1796 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 1797 else
1798 last_block = first_block +
617ba13b 1799 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 1800
8fadc143 1801 block_bitmap = ext4_block_bitmap(sb, gdp);
2b2d6d01 1802 if (block_bitmap < first_block || block_bitmap > last_block) {
b31e1552 1803 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 1804 "Block bitmap for group %u not in group "
b31e1552 1805 "(block %llu)!", i, block_bitmap);
ac27a0ec
DK		 1806 return 0;
1807 }
8fadc143 1808 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2b2d6d01 1809 if (inode_bitmap < first_block || inode_bitmap > last_block) {
b31e1552 1810 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 1811 "Inode bitmap for group %u not in group "
b31e1552 1812 "(block %llu)!", i, inode_bitmap);
ac27a0ec
DK		 1813 return 0;
1814 }
8fadc143 1815 inode_table = ext4_inode_table(sb, gdp);
bd81d8ee 1816 if (inode_table < first_block ||
2b2d6d01 1817 inode_table + sbi->s_itb_per_group - 1 > last_block) {
b31e1552 1818 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 1819 "Inode table for group %u not in group "
b31e1552 1820 "(block %llu)!", i, inode_table);
ac27a0ec
DK		 1821 return 0;
1822 }
955ce5f5 1823 ext4_lock_group(sb, i);
717d50e4 1824 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
b31e1552
ES		 1825 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1826 "Checksum for group %u failed (%u!=%u)",
1827 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1828 gdp)), le16_to_cpu(gdp->bg_checksum));
7ee1ec4c 1829 if (!(sb->s_flags & MS_RDONLY)) {
955ce5f5 1830 ext4_unlock_group(sb, i);
8a266467 1831 return 0;
7ee1ec4c 1832 }
717d50e4 1833 }
955ce5f5 1834 ext4_unlock_group(sb, i);
ce421581
JS		 1835 if (!flexbg_flag)
1836 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec
DK		 1837 }
1838
bd81d8ee 1839 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
0b8e58a1 1840 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
ac27a0ec
DK		 1841 return 1;
1842}
1843
617ba13b 1844/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 1845 * the superblock) which were deleted from all directories, but held open by
1846 * a process at the time of a crash. We walk the list and try to delete these
1847 * inodes at recovery time (only with a read-write filesystem).
1848 *
1849 * In order to keep the orphan inode chain consistent during traversal (in
1850 * case of crash during recovery), we link each inode into the superblock
1851 * orphan list_head and handle it the same way as an inode deletion during
1852 * normal operation (which journals the operations for us).
1853 *
1854 * We only do an iget() and an iput() on each inode, which is very safe if we
1855 * accidentally point at an in-use or already deleted inode. The worst that
1856 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 1857 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 1858 * e2fsck was run on this filesystem, and it must have already done the orphan
1859 * inode cleanup for us, so we can safely abort without any further action.
1860 */
2b2d6d01
TT		 1861static void ext4_orphan_cleanup(struct super_block *sb,
1862 struct ext4_super_block *es)
ac27a0ec
DK		 1863{
1864 unsigned int s_flags = sb->s_flags;
1865 int nr_orphans = 0, nr_truncates = 0;
1866#ifdef CONFIG_QUOTA
1867 int i;
1868#endif
1869 if (!es->s_last_orphan) {
1870 jbd_debug(4, "no orphan inodes to clean up\n");
1871 return;
1872 }
1873
a8f48a95 1874 if (bdev_read_only(sb->s_bdev)) {
b31e1552
ES		 1875 ext4_msg(sb, KERN_ERR, "write access "
1876 "unavailable, skipping orphan cleanup");
a8f48a95
ES		 1877 return;
1878 }
1879
617ba13b 1880 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
ac27a0ec
DK		 1881 if (es->s_last_orphan)
1882 jbd_debug(1, "Errors on filesystem, "
1883 "clearing orphan list.\n");
1884 es->s_last_orphan = 0;
1885 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1886 return;
1887 }
1888
1889 if (s_flags & MS_RDONLY) {
b31e1552 1890 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
ac27a0ec
DK		 1891 sb->s_flags &= ~MS_RDONLY;
1892 }
1893#ifdef CONFIG_QUOTA
1894 /* Needed for iput() to work correctly and not trash data */
1895 sb->s_flags |= MS_ACTIVE;
1896 /* Turn on quotas so that they are updated correctly */
1897 for (i = 0; i < MAXQUOTAS; i++) {
617ba13b
MC		 1898 if (EXT4_SB(sb)->s_qf_names[i]) {
1899 int ret = ext4_quota_on_mount(sb, i);
ac27a0ec 1900 if (ret < 0)
b31e1552
ES		 1901 ext4_msg(sb, KERN_ERR,
1902 "Cannot turn on journaled "
1903 "quota: error %d", ret);
ac27a0ec
DK		 1904 }
1905 }
1906#endif
1907
1908 while (es->s_last_orphan) {
1909 struct inode *inode;
1910
97bd42b9
JB		 1911 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1912 if (IS_ERR(inode)) {
ac27a0ec
DK		 1913 es->s_last_orphan = 0;
1914 break;
1915 }
1916
617ba13b 1917 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
a269eb18 1918 vfs_dq_init(inode);
ac27a0ec 1919 if (inode->i_nlink) {
b31e1552
ES		 1920 ext4_msg(sb, KERN_DEBUG,
1921 "%s: truncating inode %lu to %lld bytes",
46e665e9 1922 __func__, inode->i_ino, inode->i_size);
e5f8eab8 1923 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
ac27a0ec 1924 inode->i_ino, inode->i_size);
617ba13b 1925 ext4_truncate(inode);
ac27a0ec
DK		 1926 nr_truncates++;
1927 } else {
b31e1552
ES		 1928 ext4_msg(sb, KERN_DEBUG,
1929 "%s: deleting unreferenced inode %lu",
46e665e9 1930 __func__, inode->i_ino);
ac27a0ec
DK		 1931 jbd_debug(2, "deleting unreferenced inode %lu\n",
1932 inode->i_ino);
1933 nr_orphans++;
1934 }
1935 iput(inode); /* The delete magic happens here! */
1936 }
1937
2b2d6d01 1938#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
ac27a0ec
DK		 1939
1940 if (nr_orphans)
b31e1552
ES		 1941 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
1942 PLURAL(nr_orphans));
ac27a0ec 1943 if (nr_truncates)
b31e1552
ES		 1944 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
1945 PLURAL(nr_truncates));
ac27a0ec
DK		 1946#ifdef CONFIG_QUOTA
1947 /* Turn quotas off */
1948 for (i = 0; i < MAXQUOTAS; i++) {
1949 if (sb_dqopt(sb)->files[i])
6f28e087 1950 vfs_quota_off(sb, i, 0);
ac27a0ec
DK		 1951 }
1952#endif
1953 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1954}
0b8e58a1 1955
cd2291a4
ES		 1956/*
1957 * Maximal extent format file size.
1958 * Resulting logical blkno at s_maxbytes must fit in our on-disk
1959 * extent format containers, within a sector_t, and within i_blocks
1960 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
1961 * so that won't be a limiting factor.
1962 *
1963 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1964 */
f287a1a5 1965static loff_t ext4_max_size(int blkbits, int has_huge_files)
cd2291a4
ES		 1966{
1967 loff_t res;
1968 loff_t upper_limit = MAX_LFS_FILESIZE;
1969
1970 /* small i_blocks in vfs inode? */
f287a1a5 1971 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
cd2291a4 1972 /*
90c699a9 1973 * CONFIG_LBDAF is not enabled implies the inode
cd2291a4
ES		 1974 * i_block represent total blocks in 512 bytes
1975 * 32 == size of vfs inode i_blocks * 8
1976 */
1977 upper_limit = (1LL << 32) - 1;
1978
1979 /* total blocks in file system block size */
1980 upper_limit >>= (blkbits - 9);
1981 upper_limit <<= blkbits;
1982 }
1983
1984 /* 32-bit extent-start container, ee_block */
1985 res = 1LL << 32;
1986 res <<= blkbits;
1987 res -= 1;
1988
1989 /* Sanity check against vm- & vfs- imposed limits */
1990 if (res > upper_limit)
1991 res = upper_limit;
1992
1993 return res;
1994}
ac27a0ec 1995
ac27a0ec 1996/*
cd2291a4 1997 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 1998 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
1999 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 2000 */
f287a1a5 2001static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
ac27a0ec 2002{
617ba13b 2003 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 2004 int meta_blocks;
2005 loff_t upper_limit;
0b8e58a1
AD		 2006 /* This is calculated to be the largest file size for a dense, block
2007 * mapped file such that the file's total number of 512-byte sectors,
2008 * including data and all indirect blocks, does not exceed (2^48 - 1).
2009 *
2010 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2011 * number of 512-byte sectors of the file.
0fc1b451
AK		 2012 */
2013
f287a1a5 2014 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
0fc1b451 2015 /*
90c699a9 2016 * !has_huge_files or CONFIG_LBDAF not enabled implies that
0b8e58a1
AD		 2017 * the inode i_block field represents total file blocks in
2018 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
0fc1b451
AK		 2019 */
2020 upper_limit = (1LL << 32) - 1;
2021
2022 /* total blocks in file system block size */
2023 upper_limit >>= (bits - 9);
2024
2025 } else {
8180a562
AK		 2026 /*
2027 * We use 48 bit ext4_inode i_blocks
2028 * With EXT4_HUGE_FILE_FL set the i_blocks
2029 * represent total number of blocks in
2030 * file system block size
2031 */
0fc1b451
AK		 2032 upper_limit = (1LL << 48) - 1;
2033
0fc1b451
AK		 2034 }
2035
2036 /* indirect blocks */
2037 meta_blocks = 1;
2038 /* double indirect blocks */
2039 meta_blocks += 1 + (1LL << (bits-2));
2040 /* tripple indirect blocks */
2041 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2042
2043 upper_limit -= meta_blocks;
2044 upper_limit <<= bits;
ac27a0ec
DK		 2045
2046 res += 1LL << (bits-2);
2047 res += 1LL << (2*(bits-2));
2048 res += 1LL << (3*(bits-2));
2049 res <<= bits;
2050 if (res > upper_limit)
2051 res = upper_limit;
0fc1b451
AK		 2052
2053 if (res > MAX_LFS_FILESIZE)
2054 res = MAX_LFS_FILESIZE;
2055
ac27a0ec
DK		 2056 return res;
2057}
2058
617ba13b 2059static ext4_fsblk_t descriptor_loc(struct super_block *sb,
0b8e58a1 2060 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 2061{
617ba13b 2062 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 2063 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 2064 int has_super = 0;
2065
2066 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2067
617ba13b 2068 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
ac27a0ec 2069 nr < first_meta_bg)
70bbb3e0 2070 return logical_sb_block + nr + 1;
ac27a0ec 2071 bg = sbi->s_desc_per_block * nr;
617ba13b 2072 if (ext4_bg_has_super(sb, bg))
ac27a0ec 2073 has_super = 1;
0b8e58a1 2074
617ba13b 2075 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 2076}
2077
c9de560d
AT		 2078/**
2079 * ext4_get_stripe_size: Get the stripe size.
2080 * @sbi: In memory super block info
2081 *
2082 * If we have specified it via mount option, then
2083 * use the mount option value. If the value specified at mount time is
2084 * greater than the blocks per group use the super block value.
2085 * If the super block value is greater than blocks per group return 0.
2086 * Allocator needs it be less than blocks per group.
2087 *
2088 */
2089static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2090{
2091 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2092 unsigned long stripe_width =
2093 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2094
2095 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2096 return sbi->s_stripe;
2097
2098 if (stripe_width <= sbi->s_blocks_per_group)
2099 return stripe_width;
2100
2101 if (stride <= sbi->s_blocks_per_group)
2102 return stride;
2103
2104 return 0;
2105}
ac27a0ec 2106
3197ebdb
TT		 2107/* sysfs supprt */
2108
2109struct ext4_attr {
2110 struct attribute attr;
2111 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2112 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2113 const char *, size_t);
2114 int offset;
2115};
2116
2117static int parse_strtoul(const char *buf,
2118 unsigned long max, unsigned long *value)
2119{
2120 char *endp;
2121
2122 while (*buf && isspace(*buf))
2123 buf++;
2124 *value = simple_strtoul(buf, &endp, 0);
2125 while (*endp && isspace(*endp))
2126 endp++;
2127 if (*endp || *value > max)
2128 return -EINVAL;
2129
2130 return 0;
2131}
2132
2133static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2134 struct ext4_sb_info *sbi,
2135 char *buf)
2136{
2137 return snprintf(buf, PAGE_SIZE, "%llu\n",
2138 (s64) percpu_counter_sum(&sbi->s_dirtyblocks_counter));
2139}
2140
2141static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2142 struct ext4_sb_info *sbi, char *buf)
2143{
2144 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2145
2146 return snprintf(buf, PAGE_SIZE, "%lu\n",
2147 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2148 sbi->s_sectors_written_start) >> 1);
2149}
2150
2151static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2152 struct ext4_sb_info *sbi, char *buf)
2153{
2154 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2155
2156 return snprintf(buf, PAGE_SIZE, "%llu\n",
2157 sbi->s_kbytes_written +
2158 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2159 EXT4_SB(sb)->s_sectors_written_start) >> 1));
2160}
2161
2162static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2163 struct ext4_sb_info *sbi,
2164 const char *buf, size_t count)
2165{
2166 unsigned long t;
2167
2168 if (parse_strtoul(buf, 0x40000000, &t))
2169 return -EINVAL;
2170
f7c43950 2171 if (!is_power_of_2(t))
3197ebdb
TT		 2172 return -EINVAL;
2173
2174 sbi->s_inode_readahead_blks = t;
2175 return count;
2176}
2177
2178static ssize_t sbi_ui_show(struct ext4_attr *a,
0b8e58a1 2179 struct ext4_sb_info *sbi, char *buf)
3197ebdb
TT		 2180{
2181 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2182
2183 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2184}
2185
2186static ssize_t sbi_ui_store(struct ext4_attr *a,
2187 struct ext4_sb_info *sbi,
2188 const char *buf, size_t count)
2189{
2190 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2191 unsigned long t;
2192
2193 if (parse_strtoul(buf, 0xffffffff, &t))
2194 return -EINVAL;
2195 *ui = t;
2196 return count;
2197}
2198
2199#define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2200static struct ext4_attr ext4_attr_##_name = { \
2201 .attr = {.name = __stringify(_name), .mode = _mode }, \
2202 .show = _show, \
2203 .store = _store, \
2204 .offset = offsetof(struct ext4_sb_info, _elname), \
2205}
2206#define EXT4_ATTR(name, mode, show, store) \
2207static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2208
2209#define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2210#define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2211#define EXT4_RW_ATTR_SBI_UI(name, elname) \
2212 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2213#define ATTR_LIST(name) &ext4_attr_##name.attr
2214
2215EXT4_RO_ATTR(delayed_allocation_blocks);
2216EXT4_RO_ATTR(session_write_kbytes);
2217EXT4_RO_ATTR(lifetime_write_kbytes);
2218EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2219 inode_readahead_blks_store, s_inode_readahead_blks);
11013911 2220EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
3197ebdb
TT		 2221EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2222EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2223EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2224EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2225EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2226EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
55138e0b 2227EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
3197ebdb
TT		 2228
2229static struct attribute *ext4_attrs[] = {
2230 ATTR_LIST(delayed_allocation_blocks),
2231 ATTR_LIST(session_write_kbytes),
2232 ATTR_LIST(lifetime_write_kbytes),
2233 ATTR_LIST(inode_readahead_blks),
11013911 2234 ATTR_LIST(inode_goal),
3197ebdb
TT		 2235 ATTR_LIST(mb_stats),
2236 ATTR_LIST(mb_max_to_scan),
2237 ATTR_LIST(mb_min_to_scan),
2238 ATTR_LIST(mb_order2_req),
2239 ATTR_LIST(mb_stream_req),
2240 ATTR_LIST(mb_group_prealloc),
55138e0b 2241 ATTR_LIST(max_writeback_mb_bump),
3197ebdb
TT		 2242 NULL,
2243};
2244
2245static ssize_t ext4_attr_show(struct kobject *kobj,
2246 struct attribute *attr, char *buf)
2247{
2248 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2249 s_kobj);
2250 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2251
2252 return a->show ? a->show(a, sbi, buf) : 0;
2253}
2254
2255static ssize_t ext4_attr_store(struct kobject *kobj,
2256 struct attribute *attr,
2257 const char *buf, size_t len)
2258{
2259 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2260 s_kobj);
2261 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2262
2263 return a->store ? a->store(a, sbi, buf, len) : 0;
2264}
2265
2266static void ext4_sb_release(struct kobject *kobj)
2267{
2268 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2269 s_kobj);
2270 complete(&sbi->s_kobj_unregister);
2271}
2272
2273
2274static struct sysfs_ops ext4_attr_ops = {
2275 .show = ext4_attr_show,
2276 .store = ext4_attr_store,
2277};
2278
2279static struct kobj_type ext4_ktype = {
2280 .default_attrs = ext4_attrs,
2281 .sysfs_ops = &ext4_attr_ops,
2282 .release = ext4_sb_release,
2283};
2284
a13fb1a4
ES		 2285/*
2286 * Check whether this filesystem can be mounted based on
2287 * the features present and the RDONLY/RDWR mount requested.
2288 * Returns 1 if this filesystem can be mounted as requested,
2289 * 0 if it cannot be.
2290 */
2291static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2292{
2293 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2294 ext4_msg(sb, KERN_ERR,
2295 "Couldn't mount because of "
2296 "unsupported optional features (%x)",
2297 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2298 ~EXT4_FEATURE_INCOMPAT_SUPP));
2299 return 0;
2300 }
2301
2302 if (readonly)
2303 return 1;
2304
2305 /* Check that feature set is OK for a read-write mount */
2306 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2307 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2308 "unsupported optional features (%x)",
2309 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2310 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2311 return 0;
2312 }
2313 /*
2314 * Large file size enabled file system can only be mounted
2315 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2316 */
2317 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2318 if (sizeof(blkcnt_t) < sizeof(u64)) {
2319 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2320 "cannot be mounted RDWR without "
2321 "CONFIG_LBDAF");
2322 return 0;
2323 }
2324 }
2325 return 1;
2326}
2327
2b2d6d01 2328static int ext4_fill_super(struct super_block *sb, void *data, int silent)
7477827f
AK		 2329 __releases(kernel_lock)
2330 __acquires(kernel_lock)
ac27a0ec 2331{
2b2d6d01 2332 struct buffer_head *bh;
617ba13b
MC		 2333 struct ext4_super_block *es = NULL;
2334 struct ext4_sb_info *sbi;
2335 ext4_fsblk_t block;
2336 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 2337 ext4_fsblk_t logical_sb_block;
ac27a0ec 2338 unsigned long offset = 0;
ac27a0ec
DK		 2339 unsigned long journal_devnum = 0;
2340 unsigned long def_mount_opts;
2341 struct inode *root;
9f6200bb 2342 char *cp;
0390131b 2343 const char *descr;
1d1fe1ee 2344 int ret = -EINVAL;
ac27a0ec 2345 int blocksize;
4ec11028
TT		 2346 unsigned int db_count;
2347 unsigned int i;
f287a1a5 2348 int needs_recovery, has_huge_files;
bd81d8ee 2349 __u64 blocks_count;
833f4077 2350 int err;
b3881f74 2351 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
ac27a0ec
DK		 2352
2353 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
2354 if (!sbi)
2355 return -ENOMEM;
705895b6
PE		 2356
2357 sbi->s_blockgroup_lock =
2358 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
2359 if (!sbi->s_blockgroup_lock) {
2360 kfree(sbi);
2361 return -ENOMEM;
2362 }
ac27a0ec
DK		 2363 sb->s_fs_info = sbi;
2364 sbi->s_mount_opt = 0;
617ba13b
MC		 2365 sbi->s_resuid = EXT4_DEF_RESUID;
2366 sbi->s_resgid = EXT4_DEF_RESGID;
240799cd 2367 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
d9c9bef1 2368 sbi->s_sb_block = sb_block;
afc32f7e
TT		 2369 sbi->s_sectors_written_start = part_stat_read(sb->s_bdev->bd_part,
2370 sectors[1]);
ac27a0ec
DK		 2371
2372 unlock_kernel();
2373
9f6200bb
TT		 2374 /* Cleanup superblock name */
2375 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
2376 *cp = '!';
2377
617ba13b 2378 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 2379 if (!blocksize) {
b31e1552 2380 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
ac27a0ec
DK		 2381 goto out_fail;
2382 }
2383
2384 /*
617ba13b 2385 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 2386 * block sizes. We need to calculate the offset from buffer start.
2387 */
617ba13b 2388 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 2389 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2390 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 2391 } else {
70bbb3e0 2392 logical_sb_block = sb_block;
ac27a0ec
DK		 2393 }
2394
70bbb3e0 2395 if (!(bh = sb_bread(sb, logical_sb_block))) {
b31e1552 2396 ext4_msg(sb, KERN_ERR, "unable to read superblock");
ac27a0ec
DK		 2397 goto out_fail;
2398 }
2399 /*
2400 * Note: s_es must be initialized as soon as possible because
617ba13b 2401 * some ext4 macro-instructions depend on its value
ac27a0ec 2402 */
617ba13b 2403 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
ac27a0ec
DK		 2404 sbi->s_es = es;
2405 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 2406 if (sb->s_magic != EXT4_SUPER_MAGIC)
2407 goto cantfind_ext4;
afc32f7e 2408 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
ac27a0ec
DK		 2409
2410 /* Set defaults before we parse the mount options */
2411 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
617ba13b 2412 if (def_mount_opts & EXT4_DEFM_DEBUG)
ac27a0ec 2413 set_opt(sbi->s_mount_opt, DEBUG);
617ba13b 2414 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
ac27a0ec 2415 set_opt(sbi->s_mount_opt, GRPID);
617ba13b 2416 if (def_mount_opts & EXT4_DEFM_UID16)
ac27a0ec 2417 set_opt(sbi->s_mount_opt, NO_UID32);
03010a33 2418#ifdef CONFIG_EXT4_FS_XATTR
617ba13b 2419 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
ac27a0ec 2420 set_opt(sbi->s_mount_opt, XATTR_USER);
2e7842b8 2421#endif
03010a33 2422#ifdef CONFIG_EXT4_FS_POSIX_ACL
617ba13b 2423 if (def_mount_opts & EXT4_DEFM_ACL)
ac27a0ec 2424 set_opt(sbi->s_mount_opt, POSIX_ACL);
2e7842b8 2425#endif
617ba13b
MC		 2426 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
2427 sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
2428 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
2429 sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
2430 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
2431 sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
2432
2433 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
ac27a0ec 2434 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
bb4f397a 2435 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
ceea16bf 2436 set_opt(sbi->s_mount_opt, ERRORS_CONT);
bb4f397a
AK		 2437 else
2438 set_opt(sbi->s_mount_opt, ERRORS_RO);
ac27a0ec
DK		 2439
2440 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
2441 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
30773840
TT		 2442 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
2443 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
2444 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
f4033903 2445 sbi->s_mb_history_max = default_mb_history_length;
ac27a0ec 2446
571640ca 2447 set_opt(sbi->s_mount_opt, BARRIER);
ac27a0ec 2448
dd919b98
AK		 2449 /*
2450 * enable delayed allocation by default
2451 * Use -o nodelalloc to turn it off
2452 */
2453 set_opt(sbi->s_mount_opt, DELALLOC);
2454
b3881f74
TT		 2455 if (!parse_options((char *) data, sb, &journal_devnum,
2456 &journal_ioprio, NULL, 0))
ac27a0ec
DK		 2457 goto failed_mount;
2458
2459 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
617ba13b 2460 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec 2461
617ba13b
MC		 2462 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
2463 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
2464 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
2465 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
b31e1552
ES		 2466 ext4_msg(sb, KERN_WARNING,
2467 "feature flags set on rev 0 fs, "
2468 "running e2fsck is recommended");
469108ff 2469
ac27a0ec
DK		 2470 /*
2471 * Check feature flags regardless of the revision level, since we
2472 * previously didn't change the revision level when setting the flags,
2473 * so there is a chance incompat flags are set on a rev 0 filesystem.
2474 */
a13fb1a4 2475 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
ac27a0ec 2476 goto failed_mount;
a13fb1a4 2477
ac27a0ec
DK		 2478 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
2479
617ba13b
MC		 2480 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
2481 blocksize > EXT4_MAX_BLOCK_SIZE) {
b31e1552
ES		 2482 ext4_msg(sb, KERN_ERR,
2483 "Unsupported filesystem blocksize %d", blocksize);
ac27a0ec
DK		 2484 goto failed_mount;
2485 }
2486
ac27a0ec 2487 if (sb->s_blocksize != blocksize) {
ce40733c
AK		 2488 /* Validate the filesystem blocksize */
2489 if (!sb_set_blocksize(sb, blocksize)) {
b31e1552 2490 ext4_msg(sb, KERN_ERR, "bad block size %d",
ce40733c 2491 blocksize);
ac27a0ec
DK		 2492 goto failed_mount;
2493 }
2494
2b2d6d01 2495 brelse(bh);
70bbb3e0
AM		 2496 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2497 offset = do_div(logical_sb_block, blocksize);
2498 bh = sb_bread(sb, logical_sb_block);
ac27a0ec 2499 if (!bh) {
b31e1552
ES		 2500 ext4_msg(sb, KERN_ERR,
2501 "Can't read superblock on 2nd try");
ac27a0ec
DK		 2502 goto failed_mount;
2503 }
617ba13b 2504 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
ac27a0ec 2505 sbi->s_es = es;
617ba13b 2506 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
b31e1552
ES		 2507 ext4_msg(sb, KERN_ERR,
2508 "Magic mismatch, very weird!");
ac27a0ec
DK		 2509 goto failed_mount;
2510 }
2511 }
2512
a13fb1a4
ES		 2513 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2514 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
f287a1a5
TT		 2515 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
2516 has_huge_files);
2517 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
ac27a0ec 2518
617ba13b
MC		 2519 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2520 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2521 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
ac27a0ec
DK		 2522 } else {
2523 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2524 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
617ba13b 2525 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1330593e 2526 (!is_power_of_2(sbi->s_inode_size)) ||
ac27a0ec 2527 (sbi->s_inode_size > blocksize)) {
b31e1552
ES		 2528 ext4_msg(sb, KERN_ERR,
2529 "unsupported inode size: %d",
2b2d6d01 2530 sbi->s_inode_size);
ac27a0ec
DK		 2531 goto failed_mount;
2532 }
ef7f3835
KS		 2533 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2534 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
ac27a0ec 2535 }
0b8e58a1 2536
0d1ee42f
AR		 2537 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2538 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
8fadc143 2539 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 2540 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 2541 !is_power_of_2(sbi->s_desc_size)) {
b31e1552
ES		 2542 ext4_msg(sb, KERN_ERR,
2543 "unsupported descriptor size %lu",
0d1ee42f
AR		 2544 sbi->s_desc_size);
2545 goto failed_mount;
2546 }
2547 } else
2548 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
0b8e58a1 2549
ac27a0ec 2550 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 2551 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
b47b6f38 2552 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
617ba13b 2553 goto cantfind_ext4;
0b8e58a1 2554
617ba13b 2555 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 2556 if (sbi->s_inodes_per_block == 0)
617ba13b 2557 goto cantfind_ext4;
ac27a0ec
DK		 2558 sbi->s_itb_per_group = sbi->s_inodes_per_group /
2559 sbi->s_inodes_per_block;
0d1ee42f 2560 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 2561 sbi->s_sbh = bh;
2562 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 2563 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2564 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
0b8e58a1 2565
2b2d6d01 2566 for (i = 0; i < 4; i++)
ac27a0ec
DK		 2567 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2568 sbi->s_def_hash_version = es->s_def_hash_version;
f99b2589
TT		 2569 i = le32_to_cpu(es->s_flags);
2570 if (i & EXT2_FLAGS_UNSIGNED_HASH)
2571 sbi->s_hash_unsigned = 3;
2572 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
2573#ifdef __CHAR_UNSIGNED__
2574 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
2575 sbi->s_hash_unsigned = 3;
2576#else
2577 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
2578#endif
2579 sb->s_dirt = 1;
2580 }
ac27a0ec
DK		 2581
2582 if (sbi->s_blocks_per_group > blocksize * 8) {
b31e1552
ES		 2583 ext4_msg(sb, KERN_ERR,
2584 "#blocks per group too big: %lu",
2b2d6d01 2585 sbi->s_blocks_per_group);
ac27a0ec
DK		 2586 goto failed_mount;
2587 }
ac27a0ec 2588 if (sbi->s_inodes_per_group > blocksize * 8) {
b31e1552
ES		 2589 ext4_msg(sb, KERN_ERR,
2590 "#inodes per group too big: %lu",
2b2d6d01 2591 sbi->s_inodes_per_group);
ac27a0ec
DK		 2592 goto failed_mount;
2593 }
2594
bf43d84b
ES		 2595 /*
2596 * Test whether we have more sectors than will fit in sector_t,
2597 * and whether the max offset is addressable by the page cache.
2598 */
2599 if ((ext4_blocks_count(es) >
2600 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) ||
2601 (ext4_blocks_count(es) >
2602 (pgoff_t)(~0ULL) >> (PAGE_CACHE_SHIFT - sb->s_blocksize_bits))) {
b31e1552 2603 ext4_msg(sb, KERN_ERR, "filesystem"
bf43d84b 2604 " too large to mount safely on this system");
ac27a0ec 2605 if (sizeof(sector_t) < 8)
90c699a9 2606 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
bf43d84b 2607 ret = -EFBIG;
ac27a0ec
DK		 2608 goto failed_mount;
2609 }
2610
617ba13b
MC		 2611 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2612 goto cantfind_ext4;
e7c95593 2613
0f2ddca6
FTN		 2614 /* check blocks count against device size */
2615 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
2616 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
b31e1552
ES		 2617 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
2618 "exceeds size of device (%llu blocks)",
0f2ddca6
FTN		 2619 ext4_blocks_count(es), blocks_count);
2620 goto failed_mount;
2621 }
2622
0b8e58a1
AD		 2623 /*
2624 * It makes no sense for the first data block to be beyond the end
2625 * of the filesystem.
2626 */
2627 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
b31e1552
ES		 2628 ext4_msg(sb, KERN_WARNING, "bad geometry: first data"
2629 "block %u is beyond end of filesystem (%llu)",
2630 le32_to_cpu(es->s_first_data_block),
2631 ext4_blocks_count(es));
e7c95593
ES		 2632 goto failed_mount;
2633 }
bd81d8ee
LV		 2634 blocks_count = (ext4_blocks_count(es) -
2635 le32_to_cpu(es->s_first_data_block) +
2636 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2637 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4ec11028 2638 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
b31e1552 2639 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
4ec11028 2640 "(block count %llu, first data block %u, "
b31e1552 2641 "blocks per group %lu)", sbi->s_groups_count,
4ec11028
TT		 2642 ext4_blocks_count(es),
2643 le32_to_cpu(es->s_first_data_block),
2644 EXT4_BLOCKS_PER_GROUP(sb));
2645 goto failed_mount;
2646 }
bd81d8ee 2647 sbi->s_groups_count = blocks_count;
fb0a387d
ES		 2648 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
2649 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
617ba13b
MC		 2650 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2651 EXT4_DESC_PER_BLOCK(sb);
2b2d6d01 2652 sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
ac27a0ec
DK		 2653 GFP_KERNEL);
2654 if (sbi->s_group_desc == NULL) {
b31e1552 2655 ext4_msg(sb, KERN_ERR, "not enough memory");
ac27a0ec
DK		 2656 goto failed_mount;
2657 }
2658
3244fcb1 2659#ifdef CONFIG_PROC_FS
9f6200bb
TT		 2660 if (ext4_proc_root)
2661 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
3244fcb1 2662#endif
240799cd 2663
705895b6 2664 bgl_lock_init(sbi->s_blockgroup_lock);
ac27a0ec
DK		 2665
2666 for (i = 0; i < db_count; i++) {
70bbb3e0 2667 block = descriptor_loc(sb, logical_sb_block, i);
ac27a0ec
DK		 2668 sbi->s_group_desc[i] = sb_bread(sb, block);
2669 if (!sbi->s_group_desc[i]) {
b31e1552
ES		 2670 ext4_msg(sb, KERN_ERR,
2671 "can't read group descriptor %d", i);
ac27a0ec
DK		 2672 db_count = i;
2673 goto failed_mount2;
2674 }
2675 }
2b2d6d01 2676 if (!ext4_check_descriptors(sb)) {
b31e1552 2677 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
ac27a0ec
DK		 2678 goto failed_mount2;
2679 }
772cb7c8
JS		 2680 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2681 if (!ext4_fill_flex_info(sb)) {
b31e1552
ES		 2682 ext4_msg(sb, KERN_ERR,
2683 "unable to initialize "
2684 "flex_bg meta info!");
772cb7c8
JS		 2685 goto failed_mount2;
2686 }
2687
ac27a0ec
DK		 2688 sbi->s_gdb_count = db_count;
2689 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2690 spin_lock_init(&sbi->s_next_gen_lock);
2691
833f4077
PZ		 2692 err = percpu_counter_init(&sbi->s_freeblocks_counter,
2693 ext4_count_free_blocks(sb));
2694 if (!err) {
2695 err = percpu_counter_init(&sbi->s_freeinodes_counter,
2696 ext4_count_free_inodes(sb));
2697 }
2698 if (!err) {
2699 err = percpu_counter_init(&sbi->s_dirs_counter,
2700 ext4_count_dirs(sb));
2701 }
6bc6e63f
AK		 2702 if (!err) {
2703 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
2704 }
833f4077 2705 if (err) {
b31e1552 2706 ext4_msg(sb, KERN_ERR, "insufficient memory");
833f4077
PZ		 2707 goto failed_mount3;
2708 }
ac27a0ec 2709
c9de560d 2710 sbi->s_stripe = ext4_get_stripe_size(sbi);
55138e0b 2711 sbi->s_max_writeback_mb_bump = 128;
c9de560d 2712
ac27a0ec
DK		 2713 /*
2714 * set up enough so that it can read an inode
2715 */
9ca92389
TT		 2716 if (!test_opt(sb, NOLOAD) &&
2717 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
2718 sb->s_op = &ext4_sops;
2719 else
2720 sb->s_op = &ext4_nojournal_sops;
617ba13b
MC		 2721 sb->s_export_op = &ext4_export_ops;
2722 sb->s_xattr = ext4_xattr_handlers;
ac27a0ec 2723#ifdef CONFIG_QUOTA
617ba13b
MC		 2724 sb->s_qcop = &ext4_qctl_operations;
2725 sb->dq_op = &ext4_quota_operations;
ac27a0ec
DK		 2726#endif
2727 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3b9d4ed2 2728 mutex_init(&sbi->s_orphan_lock);
32ed5058 2729 mutex_init(&sbi->s_resize_lock);
ac27a0ec
DK		 2730
2731 sb->s_root = NULL;
2732
2733 needs_recovery = (es->s_last_orphan != 0 ||
617ba13b
MC		 2734 EXT4_HAS_INCOMPAT_FEATURE(sb,
2735 EXT4_FEATURE_INCOMPAT_RECOVER));
ac27a0ec
DK		 2736
2737 /*
2738 * The first inode we look at is the journal inode. Don't try
2739 * root first: it may be modified in the journal!
2740 */
2741 if (!test_opt(sb, NOLOAD) &&
617ba13b
MC		 2742 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2743 if (ext4_load_journal(sb, es, journal_devnum))
ac27a0ec 2744 goto failed_mount3;
624080ed
TT		 2745 if (!(sb->s_flags & MS_RDONLY) &&
2746 EXT4_SB(sb)->s_journal->j_failed_commit) {
b31e1552 2747 ext4_msg(sb, KERN_CRIT, "error: "
624080ed 2748 "ext4_fill_super: Journal transaction "
b31e1552 2749 "%u is corrupt",
624080ed 2750 EXT4_SB(sb)->s_journal->j_failed_commit);
2b2d6d01 2751 if (test_opt(sb, ERRORS_RO)) {
b31e1552
ES		 2752 ext4_msg(sb, KERN_CRIT,
2753 "Mounting filesystem read-only");
624080ed
TT		 2754 sb->s_flags |= MS_RDONLY;
2755 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2756 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2757 }
2758 if (test_opt(sb, ERRORS_PANIC)) {
2759 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2760 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
e2d67052 2761 ext4_commit_super(sb, 1);
624080ed
TT		 2762 goto failed_mount4;
2763 }
2764 }
0390131b
FM		 2765 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
2766 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
b31e1552
ES		 2767 ext4_msg(sb, KERN_ERR, "required journal recovery "
2768 "suppressed and not mounted read-only");
0390131b 2769 goto failed_mount4;
ac27a0ec 2770 } else {
0390131b
FM		 2771 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
2772 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
2773 sbi->s_journal = NULL;
2774 needs_recovery = 0;
2775 goto no_journal;
ac27a0ec
DK		 2776 }
2777
eb40a09c
JS		 2778 if (ext4_blocks_count(es) > 0xffffffffULL &&
2779 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2780 JBD2_FEATURE_INCOMPAT_64BIT)) {
b31e1552 2781 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
eb40a09c
JS		 2782 goto failed_mount4;
2783 }
2784
d0646f7b
TT		 2785 jbd2_journal_set_features(sbi->s_journal,
2786 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2787 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
2788 jbd2_journal_set_features(sbi->s_journal, 0, 0,
818d276c 2789 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
d0646f7b 2790 else
818d276c
GS		 2791 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2792 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
818d276c 2793
ac27a0ec
DK		 2794 /* We have now updated the journal if required, so we can
2795 * validate the data journaling mode. */
2796 switch (test_opt(sb, DATA_FLAGS)) {
2797 case 0:
2798 /* No mode set, assume a default based on the journal
63f57933
AM		 2799 * capabilities: ORDERED_DATA if the journal can
2800 * cope, else JOURNAL_DATA
2801 */
dab291af
MC		 2802 if (jbd2_journal_check_available_features
2803 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
ac27a0ec
DK		 2804 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2805 else
2806 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2807 break;
2808
617ba13b
MC		 2809 case EXT4_MOUNT_ORDERED_DATA:
2810 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 2811 if (!jbd2_journal_check_available_features
2812 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
b31e1552
ES		 2813 ext4_msg(sb, KERN_ERR, "Journal does not support "
2814 "requested data journaling mode");
ac27a0ec
DK		 2815 goto failed_mount4;
2816 }
2817 default:
2818 break;
2819 }
b3881f74 2820 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
ac27a0ec 2821
0390131b 2822no_journal:
ac27a0ec
DK		 2823
2824 if (test_opt(sb, NOBH)) {
617ba13b 2825 if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
b31e1552
ES		 2826 ext4_msg(sb, KERN_WARNING, "Ignoring nobh option - "
2827 "its supported only with writeback mode");
ac27a0ec
DK		 2828 clear_opt(sbi->s_mount_opt, NOBH);
2829 }
2830 }
4c0425ff
MC		 2831 EXT4_SB(sb)->dio_unwritten_wq = create_workqueue("ext4-dio-unwritten");
2832 if (!EXT4_SB(sb)->dio_unwritten_wq) {
2833 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
2834 goto failed_mount_wq;
2835 }
2836
ac27a0ec 2837 /*
dab291af 2838 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 2839 * so we can safely mount the rest of the filesystem now.
2840 */
2841
1d1fe1ee
DH		 2842 root = ext4_iget(sb, EXT4_ROOT_INO);
2843 if (IS_ERR(root)) {
b31e1552 2844 ext4_msg(sb, KERN_ERR, "get root inode failed");
1d1fe1ee 2845 ret = PTR_ERR(root);
ac27a0ec
DK		 2846 goto failed_mount4;
2847 }
2848 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1d1fe1ee 2849 iput(root);
b31e1552 2850 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
ac27a0ec
DK		 2851 goto failed_mount4;
2852 }
1d1fe1ee
DH		 2853 sb->s_root = d_alloc_root(root);
2854 if (!sb->s_root) {
b31e1552 2855 ext4_msg(sb, KERN_ERR, "get root dentry failed");
1d1fe1ee
DH		 2856 iput(root);
2857 ret = -ENOMEM;
2858 goto failed_mount4;
2859 }
ac27a0ec 2860
2b2d6d01 2861 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
ef7f3835
KS		 2862
2863 /* determine the minimum size of new large inodes, if present */
2864 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2865 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2866 EXT4_GOOD_OLD_INODE_SIZE;
2867 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2868 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2869 if (sbi->s_want_extra_isize <
2870 le16_to_cpu(es->s_want_extra_isize))
2871 sbi->s_want_extra_isize =
2872 le16_to_cpu(es->s_want_extra_isize);
2873 if (sbi->s_want_extra_isize <
2874 le16_to_cpu(es->s_min_extra_isize))
2875 sbi->s_want_extra_isize =
2876 le16_to_cpu(es->s_min_extra_isize);
2877 }
2878 }
2879 /* Check if enough inode space is available */
2880 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2881 sbi->s_inode_size) {
2882 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2883 EXT4_GOOD_OLD_INODE_SIZE;
b31e1552
ES		 2884 ext4_msg(sb, KERN_INFO, "required extra inode space not"
2885 "available");
ef7f3835
KS		 2886 }
2887
c2774d84 2888 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
b31e1552
ES		 2889 ext4_msg(sb, KERN_WARNING, "Ignoring delalloc option - "
2890 "requested data journaling mode");
c2774d84
AK		 2891 clear_opt(sbi->s_mount_opt, DELALLOC);
2892 } else if (test_opt(sb, DELALLOC))
b31e1552 2893 ext4_msg(sb, KERN_INFO, "delayed allocation enabled");
c2774d84 2894
6fd058f7
TT		 2895 err = ext4_setup_system_zone(sb);
2896 if (err) {
b31e1552
ES		 2897 ext4_msg(sb, KERN_ERR, "failed to initialize system "
2898 "zone (%d)\n", err);
6fd058f7
TT		 2899 goto failed_mount4;
2900 }
2901
c2774d84
AK		 2902 ext4_ext_init(sb);
2903 err = ext4_mb_init(sb, needs_recovery);
2904 if (err) {
b31e1552
ES		 2905 ext4_msg(sb, KERN_ERR, "failed to initalize mballoc (%d)",
2906 err);
c2774d84
AK		 2907 goto failed_mount4;
2908 }
2909
3197ebdb
TT		 2910 sbi->s_kobj.kset = ext4_kset;
2911 init_completion(&sbi->s_kobj_unregister);
2912 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
2913 "%s", sb->s_id);
2914 if (err) {
2915 ext4_mb_release(sb);
2916 ext4_ext_release(sb);
2917 goto failed_mount4;
2918 };
2919
617ba13b
MC		 2920 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
2921 ext4_orphan_cleanup(sb, es);
2922 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
0390131b 2923 if (needs_recovery) {
b31e1552 2924 ext4_msg(sb, KERN_INFO, "recovery complete");
0390131b
FM		 2925 ext4_mark_recovery_complete(sb, es);
2926 }
2927 if (EXT4_SB(sb)->s_journal) {
2928 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2929 descr = " journalled data mode";
2930 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2931 descr = " ordered data mode";
2932 else
2933 descr = " writeback data mode";
2934 } else
2935 descr = "out journal";
2936
b31e1552 2937 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s", descr);
ac27a0ec
DK		 2938
2939 lock_kernel();
2940 return 0;
2941
617ba13b 2942cantfind_ext4:
ac27a0ec 2943 if (!silent)
b31e1552 2944 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
ac27a0ec
DK		 2945 goto failed_mount;
2946
2947failed_mount4:
b31e1552 2948 ext4_msg(sb, KERN_ERR, "mount failed");
4c0425ff
MC		 2949 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
2950failed_mount_wq:
6fd058f7 2951 ext4_release_system_zone(sb);
0390131b
FM		 2952 if (sbi->s_journal) {
2953 jbd2_journal_destroy(sbi->s_journal);
2954 sbi->s_journal = NULL;
2955 }
ac27a0ec 2956failed_mount3:
c5ca7c76
TT		 2957 if (sbi->s_flex_groups) {
2958 if (is_vmalloc_addr(sbi->s_flex_groups))
2959 vfree(sbi->s_flex_groups);
2960 else
2961 kfree(sbi->s_flex_groups);
2962 }
ac27a0ec
DK		 2963 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2964 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2965 percpu_counter_destroy(&sbi->s_dirs_counter);
6bc6e63f 2966 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
ac27a0ec
DK		 2967failed_mount2:
2968 for (i = 0; i < db_count; i++)
2969 brelse(sbi->s_group_desc[i]);
2970 kfree(sbi->s_group_desc);
2971failed_mount:
240799cd 2972 if (sbi->s_proc) {
9f6200bb 2973 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 2974 }
ac27a0ec
DK		 2975#ifdef CONFIG_QUOTA
2976 for (i = 0; i < MAXQUOTAS; i++)
2977 kfree(sbi->s_qf_names[i]);
2978#endif
617ba13b 2979 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 2980 brelse(bh);
2981out_fail:
2982 sb->s_fs_info = NULL;
f6830165 2983 kfree(sbi->s_blockgroup_lock);
ac27a0ec
DK		 2984 kfree(sbi);
2985 lock_kernel();
1d1fe1ee 2986 return ret;
ac27a0ec
DK		 2987}
2988
2989/*
2990 * Setup any per-fs journal parameters now. We'll do this both on
2991 * initial mount, once the journal has been initialised but before we've
2992 * done any recovery; and again on any subsequent remount.
2993 */
617ba13b 2994static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 2995{
617ba13b 2996 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 2997
30773840
TT		 2998 journal->j_commit_interval = sbi->s_commit_interval;
2999 journal->j_min_batch_time = sbi->s_min_batch_time;
3000 journal->j_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 3001
3002 spin_lock(&journal->j_state_lock);
3003 if (test_opt(sb, BARRIER))
dab291af 3004 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 3005 else
dab291af 3006 journal->j_flags &= ~JBD2_BARRIER;
5bf5683a
HK		 3007 if (test_opt(sb, DATA_ERR_ABORT))
3008 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
3009 else
3010 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
ac27a0ec
DK		 3011 spin_unlock(&journal->j_state_lock);
3012}
3013
617ba13b 3014static journal_t *ext4_get_journal(struct super_block *sb,
ac27a0ec
DK		 3015 unsigned int journal_inum)
3016{
3017 struct inode *journal_inode;
3018 journal_t *journal;
3019
0390131b
FM		 3020 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3021
ac27a0ec
DK		 3022 /* First, test for the existence of a valid inode on disk. Bad
3023 * things happen if we iget() an unused inode, as the subsequent
3024 * iput() will try to delete it. */
3025
1d1fe1ee
DH		 3026 journal_inode = ext4_iget(sb, journal_inum);
3027 if (IS_ERR(journal_inode)) {
b31e1552 3028 ext4_msg(sb, KERN_ERR, "no journal found");
ac27a0ec
DK		 3029 return NULL;
3030 }
3031 if (!journal_inode->i_nlink) {
3032 make_bad_inode(journal_inode);
3033 iput(journal_inode);
b31e1552 3034 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
ac27a0ec
DK		 3035 return NULL;
3036 }
3037
e5f8eab8 3038 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
ac27a0ec 3039 journal_inode, journal_inode->i_size);
1d1fe1ee 3040 if (!S_ISREG(journal_inode->i_mode)) {
b31e1552 3041 ext4_msg(sb, KERN_ERR, "invalid journal inode");
ac27a0ec
DK		 3042 iput(journal_inode);
3043 return NULL;
3044 }
3045
dab291af 3046 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 3047 if (!journal) {
b31e1552 3048 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
ac27a0ec
DK		 3049 iput(journal_inode);
3050 return NULL;
3051 }
3052 journal->j_private = sb;
617ba13b 3053 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 3054 return journal;
3055}
3056
617ba13b 3057static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 3058 dev_t j_dev)
3059{
2b2d6d01 3060 struct buffer_head *bh;
ac27a0ec 3061 journal_t *journal;
617ba13b
MC		 3062 ext4_fsblk_t start;
3063 ext4_fsblk_t len;
ac27a0ec 3064 int hblock, blocksize;
617ba13b 3065 ext4_fsblk_t sb_block;
ac27a0ec 3066 unsigned long offset;
2b2d6d01 3067 struct ext4_super_block *es;
ac27a0ec
DK		 3068 struct block_device *bdev;
3069
0390131b
FM		 3070 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3071
b31e1552 3072 bdev = ext4_blkdev_get(j_dev, sb);
ac27a0ec
DK		 3073 if (bdev == NULL)
3074 return NULL;
3075
3076 if (bd_claim(bdev, sb)) {
b31e1552
ES		 3077 ext4_msg(sb, KERN_ERR,
3078 "failed to claim external journal device");
9a1c3542 3079 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
ac27a0ec
DK		 3080 return NULL;
3081 }
3082
3083 blocksize = sb->s_blocksize;
e1defc4f 3084 hblock = bdev_logical_block_size(bdev);
ac27a0ec 3085 if (blocksize < hblock) {
b31e1552
ES		 3086 ext4_msg(sb, KERN_ERR,
3087 "blocksize too small for journal device");
ac27a0ec
DK		 3088 goto out_bdev;
3089 }
3090
617ba13b
MC		 3091 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3092 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 3093 set_blocksize(bdev, blocksize);
3094 if (!(bh = __bread(bdev, sb_block, blocksize))) {
b31e1552
ES		 3095 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
3096 "external journal");
ac27a0ec
DK		 3097 goto out_bdev;
3098 }
3099
617ba13b
MC		 3100 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3101 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 3102 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b 3103 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
b31e1552
ES		 3104 ext4_msg(sb, KERN_ERR, "external journal has "
3105 "bad superblock");
ac27a0ec
DK		 3106 brelse(bh);
3107 goto out_bdev;
3108 }
3109
617ba13b 3110 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
b31e1552 3111 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
ac27a0ec
DK		 3112 brelse(bh);
3113 goto out_bdev;
3114 }
3115
bd81d8ee 3116 len = ext4_blocks_count(es);
ac27a0ec
DK		 3117 start = sb_block + 1;
3118 brelse(bh); /* we're done with the superblock */
3119
dab291af 3120 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 3121 start, len, blocksize);
3122 if (!journal) {
b31e1552 3123 ext4_msg(sb, KERN_ERR, "failed to create device journal");
ac27a0ec
DK		 3124 goto out_bdev;
3125 }
3126 journal->j_private = sb;
3127 ll_rw_block(READ, 1, &journal->j_sb_buffer);
3128 wait_on_buffer(journal->j_sb_buffer);
3129 if (!buffer_uptodate(journal->j_sb_buffer)) {
b31e1552 3130 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
ac27a0ec
DK		 3131 goto out_journal;
3132 }
3133 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
b31e1552
ES		 3134 ext4_msg(sb, KERN_ERR, "External journal has more than one "
3135 "user (unsupported) - %d",
ac27a0ec
DK		 3136 be32_to_cpu(journal->j_superblock->s_nr_users));
3137 goto out_journal;
3138 }
617ba13b
MC		 3139 EXT4_SB(sb)->journal_bdev = bdev;
3140 ext4_init_journal_params(sb, journal);
ac27a0ec 3141 return journal;
0b8e58a1 3142
ac27a0ec 3143out_journal:
dab291af 3144 jbd2_journal_destroy(journal);
ac27a0ec 3145out_bdev:
617ba13b 3146 ext4_blkdev_put(bdev);
ac27a0ec
DK		 3147 return NULL;
3148}
3149
617ba13b
MC		 3150static int ext4_load_journal(struct super_block *sb,
3151 struct ext4_super_block *es,
ac27a0ec
DK		 3152 unsigned long journal_devnum)
3153{
3154 journal_t *journal;
3155 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
3156 dev_t journal_dev;
3157 int err = 0;
3158 int really_read_only;
3159
0390131b
FM		 3160 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3161
ac27a0ec
DK		 3162 if (journal_devnum &&
3163 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
b31e1552
ES		 3164 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
3165 "numbers have changed");
ac27a0ec
DK		 3166 journal_dev = new_decode_dev(journal_devnum);
3167 } else
3168 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
3169
3170 really_read_only = bdev_read_only(sb->s_bdev);
3171
3172 /*
3173 * Are we loading a blank journal or performing recovery after a
3174 * crash? For recovery, we need to check in advance whether we
3175 * can get read-write access to the device.
3176 */
617ba13b 3177 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
ac27a0ec 3178 if (sb->s_flags & MS_RDONLY) {
b31e1552
ES		 3179 ext4_msg(sb, KERN_INFO, "INFO: recovery "
3180 "required on readonly filesystem");
ac27a0ec 3181 if (really_read_only) {
b31e1552
ES		 3182 ext4_msg(sb, KERN_ERR, "write access "
3183 "unavailable, cannot proceed");
ac27a0ec
DK		 3184 return -EROFS;
3185 }
b31e1552
ES		 3186 ext4_msg(sb, KERN_INFO, "write access will "
3187 "be enabled during recovery");
ac27a0ec
DK		 3188 }
3189 }
3190
3191 if (journal_inum && journal_dev) {
b31e1552
ES		 3192 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
3193 "and inode journals!");
ac27a0ec
DK		 3194 return -EINVAL;
3195 }
3196
3197 if (journal_inum) {
617ba13b 3198 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 3199 return -EINVAL;
3200 } else {
617ba13b 3201 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 3202 return -EINVAL;
3203 }
3204
4776004f 3205 if (journal->j_flags & JBD2_BARRIER)
b31e1552 3206 ext4_msg(sb, KERN_INFO, "barriers enabled");
4776004f 3207 else
b31e1552 3208 ext4_msg(sb, KERN_INFO, "barriers disabled");
4776004f 3209
ac27a0ec 3210 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
dab291af 3211 err = jbd2_journal_update_format(journal);
ac27a0ec 3212 if (err) {
b31e1552 3213 ext4_msg(sb, KERN_ERR, "error updating journal");
dab291af 3214 jbd2_journal_destroy(journal);
ac27a0ec
DK		 3215 return err;
3216 }
3217 }
3218
617ba13b 3219 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
dab291af 3220 err = jbd2_journal_wipe(journal, !really_read_only);
ac27a0ec 3221 if (!err)
dab291af 3222 err = jbd2_journal_load(journal);
ac27a0ec
DK		 3223
3224 if (err) {
b31e1552 3225 ext4_msg(sb, KERN_ERR, "error loading journal");
dab291af 3226 jbd2_journal_destroy(journal);
ac27a0ec
DK		 3227 return err;
3228 }
3229
617ba13b
MC		 3230 EXT4_SB(sb)->s_journal = journal;
3231 ext4_clear_journal_err(sb, es);
ac27a0ec
DK		 3232
3233 if (journal_devnum &&
3234 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3235 es->s_journal_dev = cpu_to_le32(journal_devnum);
ac27a0ec
DK		 3236
3237 /* Make sure we flush the recovery flag to disk. */
e2d67052 3238 ext4_commit_super(sb, 1);
ac27a0ec
DK		 3239 }
3240
3241 return 0;
3242}
3243
e2d67052 3244static int ext4_commit_super(struct super_block *sb, int sync)
ac27a0ec 3245{
e2d67052 3246 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
617ba13b 3247 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
c4be0c1d 3248 int error = 0;
ac27a0ec
DK		 3249
3250 if (!sbh)
c4be0c1d 3251 return error;
914258bf
TT		 3252 if (buffer_write_io_error(sbh)) {
3253 /*
3254 * Oh, dear. A previous attempt to write the
3255 * superblock failed. This could happen because the
3256 * USB device was yanked out. Or it could happen to
3257 * be a transient write error and maybe the block will
3258 * be remapped. Nothing we can do but to retry the
3259 * write and hope for the best.
3260 */
b31e1552
ES		 3261 ext4_msg(sb, KERN_ERR, "previous I/O error to "
3262 "superblock detected");
914258bf
TT		 3263 clear_buffer_write_io_error(sbh);
3264 set_buffer_uptodate(sbh);
3265 }
71290b36
TT		 3266 /*
3267 * If the file system is mounted read-only, don't update the
3268 * superblock write time. This avoids updating the superblock
3269 * write time when we are mounting the root file system
3270 * read/only but we need to replay the journal; at that point,
3271 * for people who are east of GMT and who make their clock
3272 * tick in localtime for Windows bug-for-bug compatibility,
3273 * the clock is set in the future, and this will cause e2fsck
3274 * to complain and force a full file system check.
3275 */
3276 if (!(sb->s_flags & MS_RDONLY))
3277 es->s_wtime = cpu_to_le32(get_seconds());
afc32f7e
TT		 3278 es->s_kbytes_written =
3279 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
3280 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
3281 EXT4_SB(sb)->s_sectors_written_start) >> 1));
5d1b1b3f
AK		 3282 ext4_free_blocks_count_set(es, percpu_counter_sum_positive(
3283 &EXT4_SB(sb)->s_freeblocks_counter));
3284 es->s_free_inodes_count = cpu_to_le32(percpu_counter_sum_positive(
3285 &EXT4_SB(sb)->s_freeinodes_counter));
7234ab2a 3286 sb->s_dirt = 0;
ac27a0ec
DK		 3287 BUFFER_TRACE(sbh, "marking dirty");
3288 mark_buffer_dirty(sbh);
914258bf 3289 if (sync) {
c4be0c1d
TS		 3290 error = sync_dirty_buffer(sbh);
3291 if (error)
3292 return error;
3293
3294 error = buffer_write_io_error(sbh);
3295 if (error) {
b31e1552
ES		 3296 ext4_msg(sb, KERN_ERR, "I/O error while writing "
3297 "superblock");
914258bf
TT		 3298 clear_buffer_write_io_error(sbh);
3299 set_buffer_uptodate(sbh);
3300 }
3301 }
c4be0c1d 3302 return error;
ac27a0ec
DK		 3303}
3304
ac27a0ec
DK		 3305/*
3306 * Have we just finished recovery? If so, and if we are mounting (or
3307 * remounting) the filesystem readonly, then we will end up with a
3308 * consistent fs on disk. Record that fact.
3309 */
2b2d6d01
TT		 3310static void ext4_mark_recovery_complete(struct super_block *sb,
3311 struct ext4_super_block *es)
ac27a0ec 3312{
617ba13b 3313 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 3314
0390131b
FM		 3315 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3316 BUG_ON(journal != NULL);
3317 return;
3318 }
dab291af 3319 jbd2_journal_lock_updates(journal);
7ffe1ea8
HK		 3320 if (jbd2_journal_flush(journal) < 0)
3321 goto out;
3322
617ba13b 3323 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
ac27a0ec 3324 sb->s_flags & MS_RDONLY) {
617ba13b 3325 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
e2d67052 3326 ext4_commit_super(sb, 1);
ac27a0ec 3327 }
7ffe1ea8
HK		 3328
3329out:
dab291af 3330 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 3331}
3332
3333/*
3334 * If we are mounting (or read-write remounting) a filesystem whose journal
3335 * has recorded an error from a previous lifetime, move that error to the
3336 * main filesystem now.
3337 */
2b2d6d01
TT		 3338static void ext4_clear_journal_err(struct super_block *sb,
3339 struct ext4_super_block *es)
ac27a0ec
DK		 3340{
3341 journal_t *journal;
3342 int j_errno;
3343 const char *errstr;
3344
0390131b
FM		 3345 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3346
617ba13b 3347 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 3348
3349 /*
3350 * Now check for any error status which may have been recorded in the
617ba13b 3351 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 3352 */
3353
dab291af 3354 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 3355 if (j_errno) {
3356 char nbuf[16];
3357
617ba13b 3358 errstr = ext4_decode_error(sb, j_errno, nbuf);
46e665e9 3359 ext4_warning(sb, __func__, "Filesystem error recorded "
ac27a0ec 3360 "from previous mount: %s", errstr);
46e665e9 3361 ext4_warning(sb, __func__, "Marking fs in need of "
ac27a0ec
DK		 3362 "filesystem check.");
3363
617ba13b
MC		 3364 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
3365 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
e2d67052 3366 ext4_commit_super(sb, 1);
ac27a0ec 3367
dab291af 3368 jbd2_journal_clear_err(journal);
ac27a0ec
DK		 3369 }
3370}
3371
3372/*
3373 * Force the running and committing transactions to commit,
3374 * and wait on the commit.
3375 */
617ba13b 3376int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 3377{
3378 journal_t *journal;
0390131b 3379 int ret = 0;
ac27a0ec
DK		 3380
3381 if (sb->s_flags & MS_RDONLY)
3382 return 0;
3383
617ba13b 3384 journal = EXT4_SB(sb)->s_journal;
7234ab2a 3385 if (journal)
0390131b 3386 ret = ext4_journal_force_commit(journal);
0390131b 3387
ac27a0ec
DK		 3388 return ret;
3389}
3390
2b2d6d01 3391static void ext4_write_super(struct super_block *sb)
ac27a0ec 3392{
ebc1ac16 3393 lock_super(sb);
9ca92389 3394 ext4_commit_super(sb, 1);
ebc1ac16 3395 unlock_super(sb);
ac27a0ec
DK		 3396}
3397
617ba13b 3398static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec 3399{
14ce0cb4 3400 int ret = 0;
9eddacf9 3401 tid_t target;
8d5d02e6 3402 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 3403
9bffad1e 3404 trace_ext4_sync_fs(sb, wait);
8d5d02e6
MC		 3405 flush_workqueue(sbi->dio_unwritten_wq);
3406 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
9ca92389 3407 if (wait)
8d5d02e6 3408 jbd2_log_wait_commit(sbi->s_journal, target);
0390131b 3409 }
14ce0cb4 3410 return ret;
ac27a0ec
DK		 3411}
3412
3413/*
3414 * LVM calls this function before a (read-only) snapshot is created. This
3415 * gives us a chance to flush the journal completely and mark the fs clean.
3416 */
c4be0c1d 3417static int ext4_freeze(struct super_block *sb)
ac27a0ec 3418{
c4be0c1d
TS		 3419 int error = 0;
3420 journal_t *journal;
ac27a0ec 3421
9ca92389
TT		 3422 if (sb->s_flags & MS_RDONLY)
3423 return 0;
ac27a0ec 3424
9ca92389 3425 journal = EXT4_SB(sb)->s_journal;
7ffe1ea8 3426
9ca92389
TT		 3427 /* Now we set up the journal barrier. */
3428 jbd2_journal_lock_updates(journal);
ac27a0ec 3429
9ca92389
TT		 3430 /*
3431 * Don't clear the needs_recovery flag if we failed to flush
3432 * the journal.
3433 */
3434 error = jbd2_journal_flush(journal);
3435 if (error < 0) {
3436 out:
3437 jbd2_journal_unlock_updates(journal);
3438 return error;
ac27a0ec 3439 }
9ca92389
TT		 3440
3441 /* Journal blocked and flushed, clear needs_recovery flag. */
3442 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3443 error = ext4_commit_super(sb, 1);
3444 if (error)
3445 goto out;
c4be0c1d 3446 return 0;
ac27a0ec
DK		 3447}
3448
3449/*
3450 * Called by LVM after the snapshot is done. We need to reset the RECOVER
3451 * flag here, even though the filesystem is not technically dirty yet.
3452 */
c4be0c1d 3453static int ext4_unfreeze(struct super_block *sb)
ac27a0ec 3454{
9ca92389
TT		 3455 if (sb->s_flags & MS_RDONLY)
3456 return 0;
3457
3458 lock_super(sb);
3459 /* Reset the needs_recovery flag before the fs is unlocked. */
3460 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3461 ext4_commit_super(sb, 1);
3462 unlock_super(sb);
3463 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
c4be0c1d 3464 return 0;
ac27a0ec
DK		 3465}
3466
2b2d6d01 3467static int ext4_remount(struct super_block *sb, int *flags, char *data)
ac27a0ec 3468{
2b2d6d01 3469 struct ext4_super_block *es;
617ba13b
MC		 3470 struct ext4_sb_info *sbi = EXT4_SB(sb);
3471 ext4_fsblk_t n_blocks_count = 0;
ac27a0ec 3472 unsigned long old_sb_flags;
617ba13b 3473 struct ext4_mount_options old_opts;
8a266467 3474 ext4_group_t g;
b3881f74 3475 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
ac27a0ec
DK		 3476 int err;
3477#ifdef CONFIG_QUOTA
3478 int i;
3479#endif
3480
337eb00a
AIB		 3481 lock_kernel();
3482
ac27a0ec 3483 /* Store the original options */
bbd6851a 3484 lock_super(sb);
ac27a0ec
DK		 3485 old_sb_flags = sb->s_flags;
3486 old_opts.s_mount_opt = sbi->s_mount_opt;
3487 old_opts.s_resuid = sbi->s_resuid;
3488 old_opts.s_resgid = sbi->s_resgid;
3489 old_opts.s_commit_interval = sbi->s_commit_interval;
30773840
TT		 3490 old_opts.s_min_batch_time = sbi->s_min_batch_time;
3491 old_opts.s_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 3492#ifdef CONFIG_QUOTA
3493 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
3494 for (i = 0; i < MAXQUOTAS; i++)
3495 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
3496#endif
b3881f74
TT		 3497 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
3498 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
ac27a0ec
DK		 3499
3500 /*
3501 * Allow the "check" option to be passed as a remount option.
3502 */
b3881f74
TT		 3503 if (!parse_options(data, sb, NULL, &journal_ioprio,
3504 &n_blocks_count, 1)) {
ac27a0ec
DK		 3505 err = -EINVAL;
3506 goto restore_opts;
3507 }
3508
4ab2f15b 3509 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
46e665e9 3510 ext4_abort(sb, __func__, "Abort forced by user");
ac27a0ec
DK		 3511
3512 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
617ba13b 3513 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec
DK		 3514
3515 es = sbi->s_es;
3516
b3881f74 3517 if (sbi->s_journal) {
0390131b 3518 ext4_init_journal_params(sb, sbi->s_journal);
b3881f74
TT		 3519 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3520 }
ac27a0ec
DK		 3521
3522 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
bd81d8ee 3523 n_blocks_count > ext4_blocks_count(es)) {
4ab2f15b 3524 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
ac27a0ec
DK		 3525 err = -EROFS;
3526 goto restore_opts;
3527 }
3528
3529 if (*flags & MS_RDONLY) {
3530 /*
3531 * First of all, the unconditional stuff we have to do
3532 * to disable replay of the journal when we next remount
3533 */
3534 sb->s_flags |= MS_RDONLY;
3535
3536 /*
3537 * OK, test if we are remounting a valid rw partition
3538 * readonly, and if so set the rdonly flag and then
3539 * mark the partition as valid again.
3540 */
617ba13b
MC		 3541 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
3542 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 3543 es->s_state = cpu_to_le16(sbi->s_mount_state);
3544
a63c9eb2 3545 if (sbi->s_journal)
0390131b 3546 ext4_mark_recovery_complete(sb, es);
ac27a0ec 3547 } else {
a13fb1a4
ES		 3548 /* Make sure we can mount this feature set readwrite */
3549 if (!ext4_feature_set_ok(sb, 0)) {
ac27a0ec
DK		 3550 err = -EROFS;
3551 goto restore_opts;
3552 }
8a266467
TT		 3553 /*
3554 * Make sure the group descriptor checksums
0b8e58a1 3555 * are sane. If they aren't, refuse to remount r/w.
8a266467
TT		 3556 */
3557 for (g = 0; g < sbi->s_groups_count; g++) {
3558 struct ext4_group_desc *gdp =
3559 ext4_get_group_desc(sb, g, NULL);
3560
3561 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
b31e1552
ES		 3562 ext4_msg(sb, KERN_ERR,
3563 "ext4_remount: Checksum for group %u failed (%u!=%u)",
8a266467
TT		 3564 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
3565 le16_to_cpu(gdp->bg_checksum));
3566 err = -EINVAL;
3567 goto restore_opts;
3568 }
3569 }
3570
ead6596b
ES		 3571 /*
3572 * If we have an unprocessed orphan list hanging
3573 * around from a previously readonly bdev mount,
3574 * require a full umount/remount for now.
3575 */
3576 if (es->s_last_orphan) {
b31e1552 3577 ext4_msg(sb, KERN_WARNING, "Couldn't "
ead6596b
ES		 3578 "remount RDWR because of unprocessed "
3579 "orphan inode list. Please "
b31e1552 3580 "umount/remount instead");
ead6596b
ES		 3581 err = -EINVAL;
3582 goto restore_opts;
3583 }
3584
ac27a0ec
DK		 3585 /*
3586 * Mounting a RDONLY partition read-write, so reread
3587 * and store the current valid flag. (It may have
3588 * been changed by e2fsck since we originally mounted
3589 * the partition.)
3590 */
0390131b
FM		 3591 if (sbi->s_journal)
3592 ext4_clear_journal_err(sb, es);
ac27a0ec 3593 sbi->s_mount_state = le16_to_cpu(es->s_state);
617ba13b 3594 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
ac27a0ec 3595 goto restore_opts;
2b2d6d01 3596 if (!ext4_setup_super(sb, es, 0))
ac27a0ec
DK		 3597 sb->s_flags &= ~MS_RDONLY;
3598 }
3599 }
6fd058f7 3600 ext4_setup_system_zone(sb);
0390131b 3601 if (sbi->s_journal == NULL)
e2d67052 3602 ext4_commit_super(sb, 1);
0390131b 3603
ac27a0ec
DK		 3604#ifdef CONFIG_QUOTA
3605 /* Release old quota file names */
3606 for (i = 0; i < MAXQUOTAS; i++)
3607 if (old_opts.s_qf_names[i] &&
3608 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3609 kfree(old_opts.s_qf_names[i]);
3610#endif
bbd6851a 3611 unlock_super(sb);
337eb00a 3612 unlock_kernel();
ac27a0ec 3613 return 0;
0b8e58a1 3614
ac27a0ec
DK		 3615restore_opts:
3616 sb->s_flags = old_sb_flags;
3617 sbi->s_mount_opt = old_opts.s_mount_opt;
3618 sbi->s_resuid = old_opts.s_resuid;
3619 sbi->s_resgid = old_opts.s_resgid;
3620 sbi->s_commit_interval = old_opts.s_commit_interval;
30773840
TT		 3621 sbi->s_min_batch_time = old_opts.s_min_batch_time;
3622 sbi->s_max_batch_time = old_opts.s_max_batch_time;
ac27a0ec
DK		 3623#ifdef CONFIG_QUOTA
3624 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
3625 for (i = 0; i < MAXQUOTAS; i++) {
3626 if (sbi->s_qf_names[i] &&
3627 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3628 kfree(sbi->s_qf_names[i]);
3629 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
3630 }
3631#endif
bbd6851a 3632 unlock_super(sb);
337eb00a 3633 unlock_kernel();
ac27a0ec
DK		 3634 return err;
3635}
3636
2b2d6d01 3637static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
ac27a0ec
DK		 3638{
3639 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 3640 struct ext4_sb_info *sbi = EXT4_SB(sb);
3641 struct ext4_super_block *es = sbi->s_es;
960cc398 3642 u64 fsid;
ac27a0ec 3643
5e70030d
BP		 3644 if (test_opt(sb, MINIX_DF)) {
3645 sbi->s_overhead_last = 0;
6bc9feff 3646 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
8df9675f 3647 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
5e70030d 3648 ext4_fsblk_t overhead = 0;
ac27a0ec
DK		 3649
3650 /*
5e70030d
BP		 3651 * Compute the overhead (FS structures). This is constant
3652 * for a given filesystem unless the number of block groups
3653 * changes so we cache the previous value until it does.
ac27a0ec
DK		 3654 */
3655
3656 /*
3657 * All of the blocks before first_data_block are
3658 * overhead
3659 */
3660 overhead = le32_to_cpu(es->s_first_data_block);
3661
3662 /*
3663 * Add the overhead attributed to the superblock and
3664 * block group descriptors. If the sparse superblocks
3665 * feature is turned on, then not all groups have this.
3666 */
3667 for (i = 0; i < ngroups; i++) {
617ba13b
MC		 3668 overhead += ext4_bg_has_super(sb, i) +
3669 ext4_bg_num_gdb(sb, i);
ac27a0ec
DK		 3670 cond_resched();
3671 }
3672
3673 /*
3674 * Every block group has an inode bitmap, a block
3675 * bitmap, and an inode table.
3676 */
5e70030d
BP		 3677 overhead += ngroups * (2 + sbi->s_itb_per_group);
3678 sbi->s_overhead_last = overhead;
3679 smp_wmb();
6bc9feff 3680 sbi->s_blocks_last = ext4_blocks_count(es);
ac27a0ec
DK		 3681 }
3682
617ba13b 3683 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 3684 buf->f_bsize = sb->s_blocksize;
5e70030d 3685 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
6bc6e63f
AK		 3686 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
3687 percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
308ba3ec 3688 ext4_free_blocks_count_set(es, buf->f_bfree);
bd81d8ee
LV		 3689 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3690 if (buf->f_bfree < ext4_r_blocks_count(es))
ac27a0ec
DK		 3691 buf->f_bavail = 0;
3692 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 3693 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
5e70030d 3694 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
617ba13b 3695 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 3696 fsid = le64_to_cpup((void *)es->s_uuid) ^
3697 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3698 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3699 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
0b8e58a1 3700
ac27a0ec
DK		 3701 return 0;
3702}
3703
0b8e58a1
AD		 3704/* Helper function for writing quotas on sync - we need to start transaction
3705 * before quota file is locked for write. Otherwise the are possible deadlocks:
ac27a0ec 3706 * Process 1 Process 2
617ba13b 3707 * ext4_create() quota_sync()
a269eb18
JK		 3708 * jbd2_journal_start() write_dquot()
3709 * vfs_dq_init() down(dqio_mutex)
dab291af 3710 * down(dqio_mutex) jbd2_journal_start()
ac27a0ec
DK		 3711 *
3712 */
3713
3714#ifdef CONFIG_QUOTA
3715
3716static inline struct inode *dquot_to_inode(struct dquot *dquot)
3717{
3718 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3719}
3720
617ba13b 3721static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 3722{
3723 int ret, err;
3724 handle_t *handle;
3725 struct inode *inode;
3726
3727 inode = dquot_to_inode(dquot);
617ba13b 3728 handle = ext4_journal_start(inode,
0b8e58a1 3729 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 3730 if (IS_ERR(handle))
3731 return PTR_ERR(handle);
3732 ret = dquot_commit(dquot);
617ba13b 3733 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3734 if (!ret)
3735 ret = err;
3736 return ret;
3737}
3738
617ba13b 3739static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 3740{
3741 int ret, err;
3742 handle_t *handle;
3743
617ba13b 3744 handle = ext4_journal_start(dquot_to_inode(dquot),
0b8e58a1 3745 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 3746 if (IS_ERR(handle))
3747 return PTR_ERR(handle);
3748 ret = dquot_acquire(dquot);
617ba13b 3749 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3750 if (!ret)
3751 ret = err;
3752 return ret;
3753}
3754
617ba13b 3755static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 3756{
3757 int ret, err;
3758 handle_t *handle;
3759
617ba13b 3760 handle = ext4_journal_start(dquot_to_inode(dquot),
0b8e58a1 3761 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 3762 if (IS_ERR(handle)) {
3763 /* Release dquot anyway to avoid endless cycle in dqput() */
3764 dquot_release(dquot);
ac27a0ec 3765 return PTR_ERR(handle);
9c3013e9 3766 }
ac27a0ec 3767 ret = dquot_release(dquot);
617ba13b 3768 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3769 if (!ret)
3770 ret = err;
3771 return ret;
3772}
3773
617ba13b 3774static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec 3775{
2c8be6b2 3776 /* Are we journaling quotas? */
617ba13b
MC		 3777 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3778 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
ac27a0ec 3779 dquot_mark_dquot_dirty(dquot);
617ba13b 3780 return ext4_write_dquot(dquot);
ac27a0ec
DK		 3781 } else {
3782 return dquot_mark_dquot_dirty(dquot);
3783 }
3784}
3785
617ba13b 3786static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 3787{
3788 int ret, err;
3789 handle_t *handle;
3790
3791 /* Data block + inode block */
617ba13b 3792 handle = ext4_journal_start(sb->s_root->d_inode, 2);
ac27a0ec
DK		 3793 if (IS_ERR(handle))
3794 return PTR_ERR(handle);
3795 ret = dquot_commit_info(sb, type);
617ba13b 3796 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3797 if (!ret)
3798 ret = err;
3799 return ret;
3800}
3801
3802/*
3803 * Turn on quotas during mount time - we need to find
3804 * the quota file and such...
3805 */
617ba13b 3806static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 3807{
617ba13b 3808 return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
0b8e58a1 3809 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 3810}
3811
3812/*
3813 * Standard function to be called on quota_on
3814 */
617ba13b 3815static int ext4_quota_on(struct super_block *sb, int type, int format_id,
8264613d 3816 char *name, int remount)
ac27a0ec
DK		 3817{
3818 int err;
8264613d 3819 struct path path;
ac27a0ec
DK		 3820
3821 if (!test_opt(sb, QUOTA))
3822 return -EINVAL;
8264613d 3823 /* When remounting, no checks are needed and in fact, name is NULL */
0623543b 3824 if (remount)
8264613d 3825 return vfs_quota_on(sb, type, format_id, name, remount);
0623543b 3826
8264613d 3827 err = kern_path(name, LOOKUP_FOLLOW, &path);
ac27a0ec
DK		 3828 if (err)
3829 return err;
0623543b 3830
ac27a0ec 3831 /* Quotafile not on the same filesystem? */
8264613d
AV		 3832 if (path.mnt->mnt_sb != sb) {
3833 path_put(&path);
ac27a0ec
DK		 3834 return -EXDEV;
3835 }
0623543b
JK		 3836 /* Journaling quota? */
3837 if (EXT4_SB(sb)->s_qf_names[type]) {
2b2d6d01 3838 /* Quotafile not in fs root? */
8264613d 3839 if (path.dentry->d_parent != sb->s_root)
b31e1552
ES		 3840 ext4_msg(sb, KERN_WARNING,
3841 "Quota file not on filesystem root. "
3842 "Journaled quota will not work");
2b2d6d01 3843 }
0623543b
JK		 3844
3845 /*
3846 * When we journal data on quota file, we have to flush journal to see
3847 * all updates to the file when we bypass pagecache...
3848 */
0390131b
FM		 3849 if (EXT4_SB(sb)->s_journal &&
3850 ext4_should_journal_data(path.dentry->d_inode)) {
0623543b
JK		 3851 /*
3852 * We don't need to lock updates but journal_flush() could
3853 * otherwise be livelocked...
3854 */
3855 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 3856 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
0623543b 3857 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 3858 if (err) {
8264613d 3859 path_put(&path);
7ffe1ea8
HK		 3860 return err;
3861 }
0623543b
JK		 3862 }
3863
8264613d
AV		 3864 err = vfs_quota_on_path(sb, type, format_id, &path);
3865 path_put(&path);
77e69dac 3866 return err;
ac27a0ec
DK		 3867}
3868
3869/* Read data from quotafile - avoid pagecache and such because we cannot afford
3870 * acquiring the locks... As quota files are never truncated and quota code
3871 * itself serializes the operations (and noone else should touch the files)
3872 * we don't have to be afraid of races */
617ba13b 3873static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 3874 size_t len, loff_t off)
3875{
3876 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 3877 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 3878 int err = 0;
3879 int offset = off & (sb->s_blocksize - 1);
3880 int tocopy;
3881 size_t toread;
3882 struct buffer_head *bh;
3883 loff_t i_size = i_size_read(inode);
3884
3885 if (off > i_size)
3886 return 0;
3887 if (off+len > i_size)
3888 len = i_size-off;
3889 toread = len;
3890 while (toread > 0) {
3891 tocopy = sb->s_blocksize - offset < toread ?
3892 sb->s_blocksize - offset : toread;
617ba13b 3893 bh = ext4_bread(NULL, inode, blk, 0, &err);
ac27a0ec
DK		 3894 if (err)
3895 return err;
3896 if (!bh) /* A hole? */
3897 memset(data, 0, tocopy);
3898 else
3899 memcpy(data, bh->b_data+offset, tocopy);
3900 brelse(bh);
3901 offset = 0;
3902 toread -= tocopy;
3903 data += tocopy;
3904 blk++;
3905 }
3906 return len;
3907}
3908
3909/* Write to quotafile (we know the transaction is already started and has
3910 * enough credits) */
617ba13b 3911static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 3912 const char *data, size_t len, loff_t off)
3913{
3914 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 3915 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 3916 int err = 0;
3917 int offset = off & (sb->s_blocksize - 1);
3918 int tocopy;
617ba13b 3919 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
ac27a0ec
DK		 3920 size_t towrite = len;
3921 struct buffer_head *bh;
3922 handle_t *handle = journal_current_handle();
3923
0390131b 3924 if (EXT4_SB(sb)->s_journal && !handle) {
b31e1552
ES		 3925 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
3926 " cancelled because transaction is not started",
9c3013e9
JK		 3927 (unsigned long long)off, (unsigned long long)len);
3928 return -EIO;
3929 }
ac27a0ec
DK		 3930 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
3931 while (towrite > 0) {
3932 tocopy = sb->s_blocksize - offset < towrite ?
3933 sb->s_blocksize - offset : towrite;
617ba13b 3934 bh = ext4_bread(handle, inode, blk, 1, &err);
ac27a0ec
DK		 3935 if (!bh)
3936 goto out;
3937 if (journal_quota) {
617ba13b 3938 err = ext4_journal_get_write_access(handle, bh);
ac27a0ec
DK		 3939 if (err) {
3940 brelse(bh);
3941 goto out;
3942 }
3943 }
3944 lock_buffer(bh);
3945 memcpy(bh->b_data+offset, data, tocopy);
3946 flush_dcache_page(bh->b_page);
3947 unlock_buffer(bh);
3948 if (journal_quota)
0390131b 3949 err = ext4_handle_dirty_metadata(handle, NULL, bh);
ac27a0ec
DK		 3950 else {
3951 /* Always do at least ordered writes for quotas */
678aaf48 3952 err = ext4_jbd2_file_inode(handle, inode);
ac27a0ec
DK		 3953 mark_buffer_dirty(bh);
3954 }
3955 brelse(bh);
3956 if (err)
3957 goto out;
3958 offset = 0;
3959 towrite -= tocopy;
3960 data += tocopy;
3961 blk++;
3962 }
3963out:
4d04e4fb
JK		 3964 if (len == towrite) {
3965 mutex_unlock(&inode->i_mutex);
ac27a0ec 3966 return err;
4d04e4fb 3967 }
ac27a0ec
DK		 3968 if (inode->i_size < off+len-towrite) {
3969 i_size_write(inode, off+len-towrite);
617ba13b 3970 EXT4_I(inode)->i_disksize = inode->i_size;
ac27a0ec 3971 }
ac27a0ec 3972 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
617ba13b 3973 ext4_mark_inode_dirty(handle, inode);
ac27a0ec
DK		 3974 mutex_unlock(&inode->i_mutex);
3975 return len - towrite;
3976}
3977
3978#endif
3979
0b8e58a1
AD		 3980static int ext4_get_sb(struct file_system_type *fs_type, int flags,
3981 const char *dev_name, void *data, struct vfsmount *mnt)
ac27a0ec 3982{
0b8e58a1 3983 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super,mnt);
ac27a0ec
DK		 3984}
3985
03010a33
TT		 3986static struct file_system_type ext4_fs_type = {
3987 .owner = THIS_MODULE,
3988 .name = "ext4",
3989 .get_sb = ext4_get_sb,
3990 .kill_sb = kill_block_super,
3991 .fs_flags = FS_REQUIRES_DEV,
3992};
3993
3994#ifdef CONFIG_EXT4DEV_COMPAT
0b8e58a1
AD		 3995static int ext4dev_get_sb(struct file_system_type *fs_type, int flags,
3996 const char *dev_name, void *data,struct vfsmount *mnt)
03010a33 3997{
b31e1552
ES		 3998 printk(KERN_WARNING "EXT4-fs (%s): Update your userspace programs "
3999 "to mount using ext4\n", dev_name);
4000 printk(KERN_WARNING "EXT4-fs (%s): ext4dev backwards compatibility "
4001 "will go away by 2.6.31\n", dev_name);
0b8e58a1 4002 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super,mnt);
03010a33
TT		 4003}
4004
617ba13b 4005static struct file_system_type ext4dev_fs_type = {
ac27a0ec 4006 .owner = THIS_MODULE,
617ba13b 4007 .name = "ext4dev",
03010a33 4008 .get_sb = ext4dev_get_sb,
ac27a0ec
DK		 4009 .kill_sb = kill_block_super,
4010 .fs_flags = FS_REQUIRES_DEV,
4011};
03010a33
TT		 4012MODULE_ALIAS("ext4dev");
4013#endif
ac27a0ec 4014
617ba13b 4015static int __init init_ext4_fs(void)
ac27a0ec 4016{
c9de560d
AT		 4017 int err;
4018
6fd058f7
TT		 4019 err = init_ext4_system_zone();
4020 if (err)
4021 return err;
3197ebdb
TT		 4022 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
4023 if (!ext4_kset)
6fd058f7 4024 goto out4;
9f6200bb 4025 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
c9de560d 4026 err = init_ext4_mballoc();
ac27a0ec 4027 if (err)
6fd058f7 4028 goto out3;
c9de560d
AT		 4029
4030 err = init_ext4_xattr();
4031 if (err)
4032 goto out2;
ac27a0ec
DK		 4033 err = init_inodecache();
4034 if (err)
4035 goto out1;
03010a33 4036 err = register_filesystem(&ext4_fs_type);
ac27a0ec
DK		 4037 if (err)
4038 goto out;
03010a33
TT		 4039#ifdef CONFIG_EXT4DEV_COMPAT
4040 err = register_filesystem(&ext4dev_fs_type);
4041 if (err) {
4042 unregister_filesystem(&ext4_fs_type);
4043 goto out;
4044 }
4045#endif
ac27a0ec
DK		 4046 return 0;
4047out:
4048 destroy_inodecache();
4049out1:
617ba13b 4050 exit_ext4_xattr();
c9de560d
AT		 4051out2:
4052 exit_ext4_mballoc();
6fd058f7
TT		 4053out3:
4054 remove_proc_entry("fs/ext4", NULL);
4055 kset_unregister(ext4_kset);
4056out4:
4057 exit_ext4_system_zone();
ac27a0ec
DK		 4058 return err;
4059}
4060
617ba13b 4061static void __exit exit_ext4_fs(void)
ac27a0ec 4062{
03010a33
TT		 4063 unregister_filesystem(&ext4_fs_type);
4064#ifdef CONFIG_EXT4DEV_COMPAT
617ba13b 4065 unregister_filesystem(&ext4dev_fs_type);
03010a33 4066#endif
ac27a0ec 4067 destroy_inodecache();
617ba13b 4068 exit_ext4_xattr();
c9de560d 4069 exit_ext4_mballoc();
9f6200bb 4070 remove_proc_entry("fs/ext4", NULL);
3197ebdb 4071 kset_unregister(ext4_kset);
6fd058f7 4072 exit_ext4_system_zone();
ac27a0ec
DK		 4073}
4074
4075MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
83982b6f 4076MODULE_DESCRIPTION("Fourth Extended Filesystem");
ac27a0ec 4077MODULE_LICENSE("GPL");
617ba13b
MC		 4078module_init(init_ext4_fs)
4079module_exit(exit_ext4_fs)
A mirror of Linus' kernel repository