]> bbs.cooldavid.org Git - net-next-2.6.git/blame - fs/ext4/super.c
git://bbs.cooldavid.org / net-next-2.6.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
ext4: remove failed journal checksum check
[net-next-2.6.git] / fs / ext4 / super.c
CommitLineData
ac27a0ec 1/*
617ba13b 2 * linux/fs/ext4/super.c
ac27a0ec
DK		 3 *
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
8 *
9 * from
10 *
11 * linux/fs/minix/inode.c
12 *
13 * Copyright (C) 1991, 1992 Linus Torvalds
14 *
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
17 */
18
19#include <linux/module.h>
20#include <linux/string.h>
21#include <linux/fs.h>
22#include <linux/time.h>
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
9f6200bb 53struct proc_dir_entry *ext4_proc_root;
3197ebdb 54static struct kset *ext4_kset;
9f6200bb 55
617ba13b 56static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
ac27a0ec 57 unsigned long journal_devnum);
e2d67052 58static int ext4_commit_super(struct super_block *sb, int sync);
2b2d6d01
TT		 59static void ext4_mark_recovery_complete(struct super_block *sb,
60 struct ext4_super_block *es);
61static void ext4_clear_journal_err(struct super_block *sb,
62 struct ext4_super_block *es);
617ba13b 63static int ext4_sync_fs(struct super_block *sb, int wait);
2b2d6d01 64static const char *ext4_decode_error(struct super_block *sb, int errno,
ac27a0ec 65 char nbuf[16]);
2b2d6d01
TT		 66static int ext4_remount(struct super_block *sb, int *flags, char *data);
67static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
c4be0c1d 68static int ext4_unfreeze(struct super_block *sb);
2b2d6d01 69static void ext4_write_super(struct super_block *sb);
c4be0c1d 70static int ext4_freeze(struct super_block *sb);
ac27a0ec 71
bd81d8ee 72
8fadc143
AR		 73ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
74 struct ext4_group_desc *bg)
bd81d8ee 75{
3a14589c 76 return le32_to_cpu(bg->bg_block_bitmap_lo) |
8fadc143 77 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 78 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 79}
80
8fadc143
AR		 81ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
82 struct ext4_group_desc *bg)
bd81d8ee 83{
5272f837 84 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
8fadc143 85 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 86 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 87}
88
8fadc143
AR		 89ext4_fsblk_t ext4_inode_table(struct super_block *sb,
90 struct ext4_group_desc *bg)
bd81d8ee 91{
5272f837 92 return le32_to_cpu(bg->bg_inode_table_lo) |
8fadc143 93 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 94 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
bd81d8ee
LV		 95}
96
560671a0
AK		 97__u32 ext4_free_blks_count(struct super_block *sb,
98 struct ext4_group_desc *bg)
99{
100 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
101 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 102 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
560671a0
AK		 103}
104
105__u32 ext4_free_inodes_count(struct super_block *sb,
106 struct ext4_group_desc *bg)
107{
108 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
109 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 110 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
560671a0
AK		 111}
112
113__u32 ext4_used_dirs_count(struct super_block *sb,
114 struct ext4_group_desc *bg)
115{
116 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
117 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 118 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
560671a0
AK		 119}
120
121__u32 ext4_itable_unused_count(struct super_block *sb,
122 struct ext4_group_desc *bg)
123{
124 return le16_to_cpu(bg->bg_itable_unused_lo) |
125 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 126 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
560671a0
AK		 127}
128
8fadc143
AR		 129void ext4_block_bitmap_set(struct super_block *sb,
130 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 131{
3a14589c 132 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 133 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
134 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 135}
136
8fadc143
AR		 137void ext4_inode_bitmap_set(struct super_block *sb,
138 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 139{
5272f837 140 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 141 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
142 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 143}
144
8fadc143
AR		 145void ext4_inode_table_set(struct super_block *sb,
146 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 147{
5272f837 148 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 149 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
150 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 151}
152
560671a0
AK		 153void ext4_free_blks_set(struct super_block *sb,
154 struct ext4_group_desc *bg, __u32 count)
155{
156 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
157 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
158 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
159}
160
161void ext4_free_inodes_set(struct super_block *sb,
162 struct ext4_group_desc *bg, __u32 count)
163{
164 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
165 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
166 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
167}
168
169void ext4_used_dirs_set(struct super_block *sb,
170 struct ext4_group_desc *bg, __u32 count)
171{
172 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
173 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
174 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
175}
176
177void ext4_itable_unused_set(struct super_block *sb,
178 struct ext4_group_desc *bg, __u32 count)
179{
180 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
181 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
182 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
183}
184
d3d1faf6
CW		 185
186/* Just increment the non-pointer handle value */
187static handle_t *ext4_get_nojournal(void)
188{
189 handle_t *handle = current->journal_info;
190 unsigned long ref_cnt = (unsigned long)handle;
191
192 BUG_ON(ref_cnt >= EXT4_NOJOURNAL_MAX_REF_COUNT);
193
194 ref_cnt++;
195 handle = (handle_t *)ref_cnt;
196
197 current->journal_info = handle;
198 return handle;
199}
200
201
202/* Decrement the non-pointer handle value */
203static void ext4_put_nojournal(handle_t *handle)
204{
205 unsigned long ref_cnt = (unsigned long)handle;
206
207 BUG_ON(ref_cnt == 0);
208
209 ref_cnt--;
210 handle = (handle_t *)ref_cnt;
211
212 current->journal_info = handle;
213}
214
ac27a0ec 215/*
dab291af 216 * Wrappers for jbd2_journal_start/end.
ac27a0ec
DK		 217 *
218 * The only special thing we need to do here is to make sure that all
219 * journal_end calls result in the superblock being marked dirty, so
220 * that sync() will call the filesystem's write_super callback if
221 * appropriate.
222 */
617ba13b 223handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
ac27a0ec
DK		 224{
225 journal_t *journal;
226
227 if (sb->s_flags & MS_RDONLY)
228 return ERR_PTR(-EROFS);
229
230 /* Special case here: if the journal has aborted behind our
231 * backs (eg. EIO in the commit thread), then we still need to
232 * take the FS itself readonly cleanly. */
617ba13b 233 journal = EXT4_SB(sb)->s_journal;
0390131b
FM		 234 if (journal) {
235 if (is_journal_aborted(journal)) {
0b8e58a1 236 ext4_abort(sb, __func__, "Detected aborted journal");
0390131b
FM		 237 return ERR_PTR(-EROFS);
238 }
239 return jbd2_journal_start(journal, nblocks);
ac27a0ec 240 }
d3d1faf6 241 return ext4_get_nojournal();
ac27a0ec
DK		 242}
243
244/*
245 * The only special thing we need to do here is to make sure that all
dab291af 246 * jbd2_journal_stop calls result in the superblock being marked dirty, so
ac27a0ec
DK		 247 * that sync() will call the filesystem's write_super callback if
248 * appropriate.
249 */
617ba13b 250int __ext4_journal_stop(const char *where, handle_t *handle)
ac27a0ec
DK		 251{
252 struct super_block *sb;
253 int err;
254 int rc;
255
0390131b 256 if (!ext4_handle_valid(handle)) {
d3d1faf6 257 ext4_put_nojournal(handle);
0390131b
FM		 258 return 0;
259 }
ac27a0ec
DK		 260 sb = handle->h_transaction->t_journal->j_private;
261 err = handle->h_err;
dab291af 262 rc = jbd2_journal_stop(handle);
ac27a0ec
DK		 263
264 if (!err)
265 err = rc;
266 if (err)
617ba13b 267 __ext4_std_error(sb, where, err);
ac27a0ec
DK		 268 return err;
269}
270
617ba13b 271void ext4_journal_abort_handle(const char *caller, const char *err_fn,
ac27a0ec
DK		 272 struct buffer_head *bh, handle_t *handle, int err)
273{
274 char nbuf[16];
617ba13b 275 const char *errstr = ext4_decode_error(NULL, err, nbuf);
ac27a0ec 276
0390131b
FM		 277 BUG_ON(!ext4_handle_valid(handle));
278
ac27a0ec
DK		 279 if (bh)
280 BUFFER_TRACE(bh, "abort");
281
282 if (!handle->h_err)
283 handle->h_err = err;
284
285 if (is_handle_aborted(handle))
286 return;
287
288 printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
289 caller, errstr, err_fn);
290
dab291af 291 jbd2_journal_abort_handle(handle);
ac27a0ec
DK		 292}
293
294/* Deal with the reporting of failure conditions on a filesystem such as
295 * inconsistencies detected or read IO failures.
296 *
297 * On ext2, we can store the error state of the filesystem in the
617ba13b 298 * superblock. That is not possible on ext4, because we may have other
ac27a0ec
DK		 299 * write ordering constraints on the superblock which prevent us from
300 * writing it out straight away; and given that the journal is about to
301 * be aborted, we can't rely on the current, or future, transactions to
302 * write out the superblock safely.
303 *
dab291af 304 * We'll just use the jbd2_journal_abort() error code to record an error in
ac27a0ec
DK		 305 * the journal instead. On recovery, the journal will compain about
306 * that error until we've noted it down and cleared it.
307 */
308
617ba13b 309static void ext4_handle_error(struct super_block *sb)
ac27a0ec 310{
617ba13b 311 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 312
617ba13b
MC		 313 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
314 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
ac27a0ec
DK		 315
316 if (sb->s_flags & MS_RDONLY)
317 return;
318
2b2d6d01 319 if (!test_opt(sb, ERRORS_CONT)) {
617ba13b 320 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 321
4ab2f15b 322 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
ac27a0ec 323 if (journal)
dab291af 324 jbd2_journal_abort(journal, -EIO);
ac27a0ec 325 }
2b2d6d01 326 if (test_opt(sb, ERRORS_RO)) {
b31e1552 327 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
ac27a0ec
DK		 328 sb->s_flags |= MS_RDONLY;
329 }
e2d67052 330 ext4_commit_super(sb, 1);
ac27a0ec 331 if (test_opt(sb, ERRORS_PANIC))
617ba13b 332 panic("EXT4-fs (device %s): panic forced after error\n",
ac27a0ec
DK		 333 sb->s_id);
334}
335
2b2d6d01
TT		 336void ext4_error(struct super_block *sb, const char *function,
337 const char *fmt, ...)
ac27a0ec
DK		 338{
339 va_list args;
340
341 va_start(args, fmt);
2b2d6d01 342 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
ac27a0ec
DK		 343 vprintk(fmt, args);
344 printk("\n");
345 va_end(args);
346
617ba13b 347 ext4_handle_error(sb);
ac27a0ec
DK		 348}
349
2b2d6d01 350static const char *ext4_decode_error(struct super_block *sb, int errno,
ac27a0ec
DK		 351 char nbuf[16])
352{
353 char *errstr = NULL;
354
355 switch (errno) {
356 case -EIO:
357 errstr = "IO failure";
358 break;
359 case -ENOMEM:
360 errstr = "Out of memory";
361 break;
362 case -EROFS:
78f1ddbb
TT		 363 if (!sb || (EXT4_SB(sb)->s_journal &&
364 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
ac27a0ec
DK		 365 errstr = "Journal has aborted";
366 else
367 errstr = "Readonly filesystem";
368 break;
369 default:
370 /* If the caller passed in an extra buffer for unknown
371 * errors, textualise them now. Else we just return
372 * NULL. */
373 if (nbuf) {
374 /* Check for truncated error codes... */
375 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
376 errstr = nbuf;
377 }
378 break;
379 }
380
381 return errstr;
382}
383
617ba13b 384/* __ext4_std_error decodes expected errors from journaling functions
ac27a0ec
DK		 385 * automatically and invokes the appropriate error response. */
386
2b2d6d01 387void __ext4_std_error(struct super_block *sb, const char *function, int errno)
ac27a0ec
DK		 388{
389 char nbuf[16];
390 const char *errstr;
391
392 /* Special case: if the error is EROFS, and we're not already
393 * inside a transaction, then there's really no point in logging
394 * an error. */
395 if (errno == -EROFS && journal_current_handle() == NULL &&
396 (sb->s_flags & MS_RDONLY))
397 return;
398
617ba13b 399 errstr = ext4_decode_error(sb, errno, nbuf);
2b2d6d01
TT		 400 printk(KERN_CRIT "EXT4-fs error (device %s) in %s: %s\n",
401 sb->s_id, function, errstr);
ac27a0ec 402
617ba13b 403 ext4_handle_error(sb);
ac27a0ec
DK		 404}
405
406/*
617ba13b 407 * ext4_abort is a much stronger failure handler than ext4_error. The
ac27a0ec
DK		 408 * abort function may be used to deal with unrecoverable failures such
409 * as journal IO errors or ENOMEM at a critical moment in log management.
410 *
411 * We unconditionally force the filesystem into an ABORT|READONLY state,
412 * unless the error response on the fs has been set to panic in which
413 * case we take the easy way out and panic immediately.
414 */
415
2b2d6d01
TT		 416void ext4_abort(struct super_block *sb, const char *function,
417 const char *fmt, ...)
ac27a0ec
DK		 418{
419 va_list args;
420
ac27a0ec 421 va_start(args, fmt);
2b2d6d01 422 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
ac27a0ec
DK		 423 vprintk(fmt, args);
424 printk("\n");
425 va_end(args);
426
427 if (test_opt(sb, ERRORS_PANIC))
617ba13b 428 panic("EXT4-fs panic from previous error\n");
ac27a0ec
DK		 429
430 if (sb->s_flags & MS_RDONLY)
431 return;
432
b31e1552 433 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
617ba13b 434 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
ac27a0ec 435 sb->s_flags |= MS_RDONLY;
4ab2f15b 436 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
ef2cabf7
HK		 437 if (EXT4_SB(sb)->s_journal)
438 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
ac27a0ec
DK		 439}
440
b31e1552
ES		 441void ext4_msg (struct super_block * sb, const char *prefix,
442 const char *fmt, ...)
443{
444 va_list args;
445
446 va_start(args, fmt);
447 printk("%sEXT4-fs (%s): ", prefix, sb->s_id);
448 vprintk(fmt, args);
449 printk("\n");
450 va_end(args);
451}
452
2b2d6d01
TT		 453void ext4_warning(struct super_block *sb, const char *function,
454 const char *fmt, ...)
ac27a0ec
DK		 455{
456 va_list args;
457
458 va_start(args, fmt);
617ba13b 459 printk(KERN_WARNING "EXT4-fs warning (device %s): %s: ",
ac27a0ec
DK		 460 sb->s_id, function);
461 vprintk(fmt, args);
462 printk("\n");
463 va_end(args);
464}
465
5d1b1b3f 466void ext4_grp_locked_error(struct super_block *sb, ext4_group_t grp,
0b8e58a1 467 const char *function, const char *fmt, ...)
5d1b1b3f
AK		 468__releases(bitlock)
469__acquires(bitlock)
470{
471 va_list args;
472 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
473
474 va_start(args, fmt);
475 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
476 vprintk(fmt, args);
477 printk("\n");
478 va_end(args);
479
480 if (test_opt(sb, ERRORS_CONT)) {
481 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
482 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
e2d67052 483 ext4_commit_super(sb, 0);
5d1b1b3f
AK		 484 return;
485 }
486 ext4_unlock_group(sb, grp);
487 ext4_handle_error(sb);
488 /*
489 * We only get here in the ERRORS_RO case; relocking the group
490 * may be dangerous, but nothing bad will happen since the
491 * filesystem will have already been marked read/only and the
492 * journal has been aborted. We return 1 as a hint to callers
493 * who might what to use the return value from
494 * ext4_grp_locked_error() to distinguish beween the
495 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
496 * aggressively from the ext4 function in question, with a
497 * more appropriate error code.
498 */
499 ext4_lock_group(sb, grp);
500 return;
501}
502
617ba13b 503void ext4_update_dynamic_rev(struct super_block *sb)
ac27a0ec 504{
617ba13b 505 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 506
617ba13b 507 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
ac27a0ec
DK		 508 return;
509
46e665e9 510 ext4_warning(sb, __func__,
ac27a0ec
DK		 511 "updating to rev %d because of new feature flag, "
512 "running e2fsck is recommended",
617ba13b 513 EXT4_DYNAMIC_REV);
ac27a0ec 514
617ba13b
MC		 515 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
516 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
517 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
ac27a0ec
DK		 518 /* leave es->s_feature_*compat flags alone */
519 /* es->s_uuid will be set by e2fsck if empty */
520
521 /*
522 * The rest of the superblock fields should be zero, and if not it
523 * means they are likely already in use, so leave them alone. We
524 * can leave it up to e2fsck to clean up any inconsistencies there.
525 */
526}
527
528/*
529 * Open the external journal device
530 */
b31e1552 531static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
ac27a0ec
DK		 532{
533 struct block_device *bdev;
534 char b[BDEVNAME_SIZE];
535
536 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
537 if (IS_ERR(bdev))
538 goto fail;
539 return bdev;
540
541fail:
b31e1552 542 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
ac27a0ec
DK		 543 __bdevname(dev, b), PTR_ERR(bdev));
544 return NULL;
545}
546
547/*
548 * Release the journal device
549 */
617ba13b 550static int ext4_blkdev_put(struct block_device *bdev)
ac27a0ec
DK		 551{
552 bd_release(bdev);
9a1c3542 553 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
ac27a0ec
DK		 554}
555
617ba13b 556static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
ac27a0ec
DK		 557{
558 struct block_device *bdev;
559 int ret = -ENODEV;
560
561 bdev = sbi->journal_bdev;
562 if (bdev) {
617ba13b 563 ret = ext4_blkdev_put(bdev);
ac27a0ec
DK		 564 sbi->journal_bdev = NULL;
565 }
566 return ret;
567}
568
569static inline struct inode *orphan_list_entry(struct list_head *l)
570{
617ba13b 571 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
ac27a0ec
DK		 572}
573
617ba13b 574static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
ac27a0ec
DK		 575{
576 struct list_head *l;
577
b31e1552
ES		 578 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
579 le32_to_cpu(sbi->s_es->s_last_orphan));
ac27a0ec
DK		 580
581 printk(KERN_ERR "sb_info orphan list:\n");
582 list_for_each(l, &sbi->s_orphan) {
583 struct inode *inode = orphan_list_entry(l);
584 printk(KERN_ERR " "
585 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
586 inode->i_sb->s_id, inode->i_ino, inode,
587 inode->i_mode, inode->i_nlink,
588 NEXT_ORPHAN(inode));
589 }
590}
591
2b2d6d01 592static void ext4_put_super(struct super_block *sb)
ac27a0ec 593{
617ba13b
MC		 594 struct ext4_sb_info *sbi = EXT4_SB(sb);
595 struct ext4_super_block *es = sbi->s_es;
ef2cabf7 596 int i, err;
ac27a0ec 597
4c0425ff
MC		 598 flush_workqueue(sbi->dio_unwritten_wq);
599 destroy_workqueue(sbi->dio_unwritten_wq);
600
a9e220f8 601 lock_super(sb);
6cfd0148 602 lock_kernel();
8c85e125 603 if (sb->s_dirt)
ebc1ac16 604 ext4_commit_super(sb, 1);
8c85e125 605
6fd058f7 606 ext4_release_system_zone(sb);
c9de560d 607 ext4_mb_release(sb);
a86c6181 608 ext4_ext_release(sb);
617ba13b 609 ext4_xattr_put_super(sb);
0390131b
FM		 610 if (sbi->s_journal) {
611 err = jbd2_journal_destroy(sbi->s_journal);
612 sbi->s_journal = NULL;
613 if (err < 0)
614 ext4_abort(sb, __func__,
615 "Couldn't clean up the journal");
616 }
ac27a0ec 617 if (!(sb->s_flags & MS_RDONLY)) {
617ba13b 618 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 619 es->s_state = cpu_to_le16(sbi->s_mount_state);
e2d67052 620 ext4_commit_super(sb, 1);
ac27a0ec 621 }
240799cd 622 if (sbi->s_proc) {
9f6200bb 623 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 624 }
3197ebdb 625 kobject_del(&sbi->s_kobj);
ac27a0ec
DK		 626
627 for (i = 0; i < sbi->s_gdb_count; i++)
628 brelse(sbi->s_group_desc[i]);
629 kfree(sbi->s_group_desc);
c5ca7c76
TT		 630 if (is_vmalloc_addr(sbi->s_flex_groups))
631 vfree(sbi->s_flex_groups);
632 else
633 kfree(sbi->s_flex_groups);
ac27a0ec
DK		 634 percpu_counter_destroy(&sbi->s_freeblocks_counter);
635 percpu_counter_destroy(&sbi->s_freeinodes_counter);
636 percpu_counter_destroy(&sbi->s_dirs_counter);
6bc6e63f 637 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
ac27a0ec
DK		 638 brelse(sbi->s_sbh);
639#ifdef CONFIG_QUOTA
640 for (i = 0; i < MAXQUOTAS; i++)
641 kfree(sbi->s_qf_names[i]);
642#endif
643
644 /* Debugging code just in case the in-memory inode orphan list
645 * isn't empty. The on-disk one can be non-empty if we've
646 * detected an error and taken the fs readonly, but the
647 * in-memory list had better be clean by this point. */
648 if (!list_empty(&sbi->s_orphan))
649 dump_orphan_list(sb, sbi);
650 J_ASSERT(list_empty(&sbi->s_orphan));
651
f98393a6 652 invalidate_bdev(sb->s_bdev);
ac27a0ec
DK		 653 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
654 /*
655 * Invalidate the journal device's buffers. We don't want them
656 * floating about in memory - the physical journal device may
657 * hotswapped, and it breaks the `ro-after' testing code.
658 */
659 sync_blockdev(sbi->journal_bdev);
f98393a6 660 invalidate_bdev(sbi->journal_bdev);
617ba13b 661 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 662 }
663 sb->s_fs_info = NULL;
3197ebdb
TT		 664 /*
665 * Now that we are completely done shutting down the
666 * superblock, we need to actually destroy the kobject.
667 */
668 unlock_kernel();
669 unlock_super(sb);
670 kobject_put(&sbi->s_kobj);
671 wait_for_completion(&sbi->s_kobj_unregister);
705895b6 672 kfree(sbi->s_blockgroup_lock);
ac27a0ec 673 kfree(sbi);
ac27a0ec
DK		 674}
675
e18b890b 676static struct kmem_cache *ext4_inode_cachep;
ac27a0ec
DK		 677
678/*
679 * Called inside transaction, so use GFP_NOFS
680 */
617ba13b 681static struct inode *ext4_alloc_inode(struct super_block *sb)
ac27a0ec 682{
617ba13b 683 struct ext4_inode_info *ei;
ac27a0ec 684
e6b4f8da 685 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
ac27a0ec
DK		 686 if (!ei)
687 return NULL;
0b8e58a1 688
ac27a0ec 689 ei->vfs_inode.i_version = 1;
91246c00 690 ei->vfs_inode.i_data.writeback_index = 0;
a86c6181 691 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
c9de560d
AT		 692 INIT_LIST_HEAD(&ei->i_prealloc_list);
693 spin_lock_init(&ei->i_prealloc_lock);
0390131b
FM		 694 /*
695 * Note: We can be called before EXT4_SB(sb)->s_journal is set,
696 * therefore it can be null here. Don't check it, just initialize
697 * jinode.
698 */
678aaf48 699 jbd2_journal_init_jbd_inode(&ei->jinode, &ei->vfs_inode);
d2a17637
MC		 700 ei->i_reserved_data_blocks = 0;
701 ei->i_reserved_meta_blocks = 0;
702 ei->i_allocated_meta_blocks = 0;
703 ei->i_delalloc_reserved_flag = 0;
704 spin_lock_init(&(ei->i_block_reservation_lock));
8d5d02e6
MC		 705 INIT_LIST_HEAD(&ei->i_aio_dio_complete_list);
706 ei->cur_aio_dio = NULL;
0b8e58a1 707
ac27a0ec
DK		 708 return &ei->vfs_inode;
709}
710
617ba13b 711static void ext4_destroy_inode(struct inode *inode)
ac27a0ec 712{
9f7dd93d 713 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
b31e1552
ES		 714 ext4_msg(inode->i_sb, KERN_ERR,
715 "Inode %lu (%p): orphan list check failed!",
716 inode->i_ino, EXT4_I(inode));
9f7dd93d
VA		 717 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
718 EXT4_I(inode), sizeof(struct ext4_inode_info),
719 true);
720 dump_stack();
721 }
617ba13b 722 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
ac27a0ec
DK		 723}
724
51cc5068 725static void init_once(void *foo)
ac27a0ec 726{
617ba13b 727 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
ac27a0ec 728
a35afb83 729 INIT_LIST_HEAD(&ei->i_orphan);
03010a33 730#ifdef CONFIG_EXT4_FS_XATTR
a35afb83 731 init_rwsem(&ei->xattr_sem);
ac27a0ec 732#endif
0e855ac8 733 init_rwsem(&ei->i_data_sem);
a35afb83 734 inode_init_once(&ei->vfs_inode);
ac27a0ec
DK		 735}
736
737static int init_inodecache(void)
738{
617ba13b
MC		 739 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
740 sizeof(struct ext4_inode_info),
ac27a0ec
DK		 741 0, (SLAB_RECLAIM_ACCOUNT|
742 SLAB_MEM_SPREAD),
20c2df83 743 init_once);
617ba13b 744 if (ext4_inode_cachep == NULL)
ac27a0ec
DK		 745 return -ENOMEM;
746 return 0;
747}
748
749static void destroy_inodecache(void)
750{
617ba13b 751 kmem_cache_destroy(ext4_inode_cachep);
ac27a0ec
DK		 752}
753
617ba13b 754static void ext4_clear_inode(struct inode *inode)
ac27a0ec 755{
c2ea3fde 756 ext4_discard_preallocations(inode);
0390131b
FM		 757 if (EXT4_JOURNAL(inode))
758 jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal,
678aaf48 759 &EXT4_I(inode)->jinode);
ac27a0ec
DK		 760}
761
2b2d6d01
TT		 762static inline void ext4_show_quota_options(struct seq_file *seq,
763 struct super_block *sb)
ac27a0ec
DK		 764{
765#if defined(CONFIG_QUOTA)
617ba13b 766 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 767
768 if (sbi->s_jquota_fmt)
769 seq_printf(seq, ",jqfmt=%s",
af5bc92d 770 (sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold" : "vfsv0");
ac27a0ec
DK		 771
772 if (sbi->s_qf_names[USRQUOTA])
773 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
774
775 if (sbi->s_qf_names[GRPQUOTA])
776 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
777
617ba13b 778 if (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA)
ac27a0ec
DK		 779 seq_puts(seq, ",usrquota");
780
617ba13b 781 if (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)
ac27a0ec
DK		 782 seq_puts(seq, ",grpquota");
783#endif
784}
785
d9c9bef1
MS		 786/*
787 * Show an option if
788 * - it's set to a non-default value OR
789 * - if the per-sb default is different from the global default
790 */
617ba13b 791static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
ac27a0ec 792{
aa22df2c
AK		 793 int def_errors;
794 unsigned long def_mount_opts;
ac27a0ec 795 struct super_block *sb = vfs->mnt_sb;
d9c9bef1
MS		 796 struct ext4_sb_info *sbi = EXT4_SB(sb);
797 struct ext4_super_block *es = sbi->s_es;
d9c9bef1
MS		 798
799 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
aa22df2c 800 def_errors = le16_to_cpu(es->s_errors);
d9c9bef1
MS		 801
802 if (sbi->s_sb_block != 1)
803 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
804 if (test_opt(sb, MINIX_DF))
805 seq_puts(seq, ",minixdf");
aa22df2c 806 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
d9c9bef1
MS		 807 seq_puts(seq, ",grpid");
808 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
809 seq_puts(seq, ",nogrpid");
810 if (sbi->s_resuid != EXT4_DEF_RESUID ||
811 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
812 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
813 }
814 if (sbi->s_resgid != EXT4_DEF_RESGID ||
815 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
816 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
817 }
bb4f397a 818 if (test_opt(sb, ERRORS_RO)) {
d9c9bef1 819 if (def_errors == EXT4_ERRORS_PANIC ||
bb4f397a
AK		 820 def_errors == EXT4_ERRORS_CONTINUE) {
821 seq_puts(seq, ",errors=remount-ro");
d9c9bef1
MS		 822 }
823 }
aa22df2c 824 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
bb4f397a 825 seq_puts(seq, ",errors=continue");
aa22df2c 826 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
d9c9bef1 827 seq_puts(seq, ",errors=panic");
aa22df2c 828 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
d9c9bef1 829 seq_puts(seq, ",nouid32");
aa22df2c 830 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
d9c9bef1
MS		 831 seq_puts(seq, ",debug");
832 if (test_opt(sb, OLDALLOC))
833 seq_puts(seq, ",oldalloc");
03010a33 834#ifdef CONFIG_EXT4_FS_XATTR
aa22df2c
AK		 835 if (test_opt(sb, XATTR_USER) &&
836 !(def_mount_opts & EXT4_DEFM_XATTR_USER))
d9c9bef1
MS		 837 seq_puts(seq, ",user_xattr");
838 if (!test_opt(sb, XATTR_USER) &&
839 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
840 seq_puts(seq, ",nouser_xattr");
841 }
842#endif
03010a33 843#ifdef CONFIG_EXT4_FS_POSIX_ACL
aa22df2c 844 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
d9c9bef1
MS		 845 seq_puts(seq, ",acl");
846 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
847 seq_puts(seq, ",noacl");
848#endif
30773840 849 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
d9c9bef1
MS		 850 seq_printf(seq, ",commit=%u",
851 (unsigned) (sbi->s_commit_interval / HZ));
852 }
30773840
TT		 853 if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
854 seq_printf(seq, ",min_batch_time=%u",
855 (unsigned) sbi->s_min_batch_time);
856 }
857 if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
858 seq_printf(seq, ",max_batch_time=%u",
859 (unsigned) sbi->s_min_batch_time);
860 }
861
571640ca
ES		 862 /*
863 * We're changing the default of barrier mount option, so
864 * let's always display its mount state so it's clear what its
865 * status is.
866 */
867 seq_puts(seq, ",barrier=");
868 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
cd0b6a39
TT		 869 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
870 seq_puts(seq, ",journal_async_commit");
d9c9bef1
MS		 871 if (test_opt(sb, NOBH))
872 seq_puts(seq, ",nobh");
25ec56b5
JNC		 873 if (test_opt(sb, I_VERSION))
874 seq_puts(seq, ",i_version");
dd919b98
AK		 875 if (!test_opt(sb, DELALLOC))
876 seq_puts(seq, ",nodelalloc");
877
ac27a0ec 878
cb45bbe4
MS		 879 if (sbi->s_stripe)
880 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
aa22df2c
AK		 881 /*
882 * journal mode get enabled in different ways
883 * So just print the value even if we didn't specify it
884 */
617ba13b 885 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
ac27a0ec 886 seq_puts(seq, ",data=journal");
617ba13b 887 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
ac27a0ec 888 seq_puts(seq, ",data=ordered");
617ba13b 889 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
ac27a0ec
DK		 890 seq_puts(seq, ",data=writeback");
891
240799cd
TT		 892 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
893 seq_printf(seq, ",inode_readahead_blks=%u",
894 sbi->s_inode_readahead_blks);
895
5bf5683a
HK		 896 if (test_opt(sb, DATA_ERR_ABORT))
897 seq_puts(seq, ",data_err=abort");
898
afd4672d 899 if (test_opt(sb, NO_AUTO_DA_ALLOC))
06705bff 900 seq_puts(seq, ",noauto_da_alloc");
afd4672d 901
617ba13b 902 ext4_show_quota_options(seq, sb);
0b8e58a1 903
ac27a0ec
DK		 904 return 0;
905}
906
1b961ac0 907static struct inode *ext4_nfs_get_inode(struct super_block *sb,
0b8e58a1 908 u64 ino, u32 generation)
ac27a0ec 909{
ac27a0ec 910 struct inode *inode;
ac27a0ec 911
617ba13b 912 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
ac27a0ec 913 return ERR_PTR(-ESTALE);
617ba13b 914 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
ac27a0ec
DK		 915 return ERR_PTR(-ESTALE);
916
917 /* iget isn't really right if the inode is currently unallocated!!
918 *
617ba13b 919 * ext4_read_inode will return a bad_inode if the inode had been
ac27a0ec
DK		 920 * deleted, so we should be safe.
921 *
922 * Currently we don't know the generation for parent directory, so
923 * a generation of 0 means "accept any"
924 */
1d1fe1ee
DH		 925 inode = ext4_iget(sb, ino);
926 if (IS_ERR(inode))
927 return ERR_CAST(inode);
928 if (generation && inode->i_generation != generation) {
ac27a0ec
DK		 929 iput(inode);
930 return ERR_PTR(-ESTALE);
931 }
1b961ac0
CH		 932
933 return inode;
934}
935
936static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
0b8e58a1 937 int fh_len, int fh_type)
1b961ac0
CH		 938{
939 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
940 ext4_nfs_get_inode);
941}
942
943static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
0b8e58a1 944 int fh_len, int fh_type)
1b961ac0
CH		 945{
946 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
947 ext4_nfs_get_inode);
ac27a0ec
DK		 948}
949
c39a7f84
TO		 950/*
951 * Try to release metadata pages (indirect blocks, directories) which are
952 * mapped via the block device. Since these pages could have journal heads
953 * which would prevent try_to_free_buffers() from freeing them, we must use
954 * jbd2 layer's try_to_free_buffers() function to release them.
955 */
0b8e58a1
AD		 956static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
957 gfp_t wait)
c39a7f84
TO		 958{
959 journal_t *journal = EXT4_SB(sb)->s_journal;
960
961 WARN_ON(PageChecked(page));
962 if (!page_has_buffers(page))
963 return 0;
964 if (journal)
965 return jbd2_journal_try_to_free_buffers(journal, page,
966 wait & ~__GFP_WAIT);
967 return try_to_free_buffers(page);
968}
969
ac27a0ec 970#ifdef CONFIG_QUOTA
af5bc92d 971#define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
2b2d6d01 972#define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
ac27a0ec 973
617ba13b
MC		 974static int ext4_write_dquot(struct dquot *dquot);
975static int ext4_acquire_dquot(struct dquot *dquot);
976static int ext4_release_dquot(struct dquot *dquot);
977static int ext4_mark_dquot_dirty(struct dquot *dquot);
978static int ext4_write_info(struct super_block *sb, int type);
6f28e087
JK		 979static int ext4_quota_on(struct super_block *sb, int type, int format_id,
980 char *path, int remount);
617ba13b
MC		 981static int ext4_quota_on_mount(struct super_block *sb, int type);
982static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 983 size_t len, loff_t off);
617ba13b 984static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 985 const char *data, size_t len, loff_t off);
986
61e225dc 987static const struct dquot_operations ext4_quota_operations = {
edf72453
JK		 988 .initialize = dquot_initialize,
989 .drop = dquot_drop,
ac27a0ec 990 .alloc_space = dquot_alloc_space,
60e58e0f
MC		 991 .reserve_space = dquot_reserve_space,
992 .claim_space = dquot_claim_space,
993 .release_rsv = dquot_release_reserved_space,
994 .get_reserved_space = ext4_get_reserved_space,
ac27a0ec
DK		 995 .alloc_inode = dquot_alloc_inode,
996 .free_space = dquot_free_space,
997 .free_inode = dquot_free_inode,
998 .transfer = dquot_transfer,
617ba13b
MC		 999 .write_dquot = ext4_write_dquot,
1000 .acquire_dquot = ext4_acquire_dquot,
1001 .release_dquot = ext4_release_dquot,
1002 .mark_dirty = ext4_mark_dquot_dirty,
a5b5ee32
JK		 1003 .write_info = ext4_write_info,
1004 .alloc_dquot = dquot_alloc,
1005 .destroy_dquot = dquot_destroy,
ac27a0ec
DK		 1006};
1007
0d54b217 1008static const struct quotactl_ops ext4_qctl_operations = {
617ba13b 1009 .quota_on = ext4_quota_on,
ac27a0ec
DK		 1010 .quota_off = vfs_quota_off,
1011 .quota_sync = vfs_quota_sync,
1012 .get_info = vfs_get_dqinfo,
1013 .set_info = vfs_set_dqinfo,
1014 .get_dqblk = vfs_get_dqblk,
1015 .set_dqblk = vfs_set_dqblk
1016};
1017#endif
1018
ee9b6d61 1019static const struct super_operations ext4_sops = {
617ba13b
MC		 1020 .alloc_inode = ext4_alloc_inode,
1021 .destroy_inode = ext4_destroy_inode,
617ba13b
MC		 1022 .write_inode = ext4_write_inode,
1023 .dirty_inode = ext4_dirty_inode,
1024 .delete_inode = ext4_delete_inode,
1025 .put_super = ext4_put_super,
617ba13b 1026 .sync_fs = ext4_sync_fs,
c4be0c1d
TS		 1027 .freeze_fs = ext4_freeze,
1028 .unfreeze_fs = ext4_unfreeze,
617ba13b
MC		 1029 .statfs = ext4_statfs,
1030 .remount_fs = ext4_remount,
1031 .clear_inode = ext4_clear_inode,
1032 .show_options = ext4_show_options,
ac27a0ec 1033#ifdef CONFIG_QUOTA
617ba13b
MC		 1034 .quota_read = ext4_quota_read,
1035 .quota_write = ext4_quota_write,
ac27a0ec 1036#endif
c39a7f84 1037 .bdev_try_to_free_page = bdev_try_to_free_page,
ac27a0ec
DK		 1038};
1039
9ca92389
TT		 1040static const struct super_operations ext4_nojournal_sops = {
1041 .alloc_inode = ext4_alloc_inode,
1042 .destroy_inode = ext4_destroy_inode,
1043 .write_inode = ext4_write_inode,
1044 .dirty_inode = ext4_dirty_inode,
1045 .delete_inode = ext4_delete_inode,
1046 .write_super = ext4_write_super,
1047 .put_super = ext4_put_super,
1048 .statfs = ext4_statfs,
1049 .remount_fs = ext4_remount,
1050 .clear_inode = ext4_clear_inode,
1051 .show_options = ext4_show_options,
1052#ifdef CONFIG_QUOTA
1053 .quota_read = ext4_quota_read,
1054 .quota_write = ext4_quota_write,
1055#endif
1056 .bdev_try_to_free_page = bdev_try_to_free_page,
1057};
1058
39655164 1059static const struct export_operations ext4_export_ops = {
1b961ac0
CH		 1060 .fh_to_dentry = ext4_fh_to_dentry,
1061 .fh_to_parent = ext4_fh_to_parent,
617ba13b 1062 .get_parent = ext4_get_parent,
ac27a0ec
DK		 1063};
1064
1065enum {
1066 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1067 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
01436ef2 1068 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
ac27a0ec 1069 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
06705bff 1070 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh,
30773840 1071 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
c3191067 1072 Opt_journal_update, Opt_journal_dev,
818d276c 1073 Opt_journal_checksum, Opt_journal_async_commit,
ac27a0ec 1074 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
296c355c 1075 Opt_data_err_abort, Opt_data_err_ignore,
ac27a0ec
DK		 1076 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1077 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
06705bff
TT		 1078 Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err, Opt_resize,
1079 Opt_usrquota, Opt_grpquota, Opt_i_version,
01436ef2 1080 Opt_stripe, Opt_delalloc, Opt_nodelalloc,
6fd058f7 1081 Opt_block_validity, Opt_noblock_validity,
b3881f74 1082 Opt_inode_readahead_blks, Opt_journal_ioprio
ac27a0ec
DK		 1083};
1084
a447c093 1085static const match_table_t tokens = {
ac27a0ec
DK		 1086 {Opt_bsd_df, "bsddf"},
1087 {Opt_minix_df, "minixdf"},
1088 {Opt_grpid, "grpid"},
1089 {Opt_grpid, "bsdgroups"},
1090 {Opt_nogrpid, "nogrpid"},
1091 {Opt_nogrpid, "sysvgroups"},
1092 {Opt_resgid, "resgid=%u"},
1093 {Opt_resuid, "resuid=%u"},
1094 {Opt_sb, "sb=%u"},
1095 {Opt_err_cont, "errors=continue"},
1096 {Opt_err_panic, "errors=panic"},
1097 {Opt_err_ro, "errors=remount-ro"},
1098 {Opt_nouid32, "nouid32"},
ac27a0ec
DK		 1099 {Opt_debug, "debug"},
1100 {Opt_oldalloc, "oldalloc"},
1101 {Opt_orlov, "orlov"},
1102 {Opt_user_xattr, "user_xattr"},
1103 {Opt_nouser_xattr, "nouser_xattr"},
1104 {Opt_acl, "acl"},
1105 {Opt_noacl, "noacl"},
ac27a0ec
DK		 1106 {Opt_noload, "noload"},
1107 {Opt_nobh, "nobh"},
1108 {Opt_bh, "bh"},
1109 {Opt_commit, "commit=%u"},
30773840
TT		 1110 {Opt_min_batch_time, "min_batch_time=%u"},
1111 {Opt_max_batch_time, "max_batch_time=%u"},
ac27a0ec 1112 {Opt_journal_update, "journal=update"},
ac27a0ec 1113 {Opt_journal_dev, "journal_dev=%u"},
818d276c
GS		 1114 {Opt_journal_checksum, "journal_checksum"},
1115 {Opt_journal_async_commit, "journal_async_commit"},
ac27a0ec
DK		 1116 {Opt_abort, "abort"},
1117 {Opt_data_journal, "data=journal"},
1118 {Opt_data_ordered, "data=ordered"},
1119 {Opt_data_writeback, "data=writeback"},
5bf5683a
HK		 1120 {Opt_data_err_abort, "data_err=abort"},
1121 {Opt_data_err_ignore, "data_err=ignore"},
ac27a0ec
DK		 1122 {Opt_offusrjquota, "usrjquota="},
1123 {Opt_usrjquota, "usrjquota=%s"},
1124 {Opt_offgrpjquota, "grpjquota="},
1125 {Opt_grpjquota, "grpjquota=%s"},
1126 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1127 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1128 {Opt_grpquota, "grpquota"},
1129 {Opt_noquota, "noquota"},
1130 {Opt_quota, "quota"},
1131 {Opt_usrquota, "usrquota"},
1132 {Opt_barrier, "barrier=%u"},
06705bff
TT		 1133 {Opt_barrier, "barrier"},
1134 {Opt_nobarrier, "nobarrier"},
25ec56b5 1135 {Opt_i_version, "i_version"},
c9de560d 1136 {Opt_stripe, "stripe=%u"},
ac27a0ec 1137 {Opt_resize, "resize"},
64769240 1138 {Opt_delalloc, "delalloc"},
dd919b98 1139 {Opt_nodelalloc, "nodelalloc"},
6fd058f7
TT		 1140 {Opt_block_validity, "block_validity"},
1141 {Opt_noblock_validity, "noblock_validity"},
240799cd 1142 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
b3881f74 1143 {Opt_journal_ioprio, "journal_ioprio=%u"},
afd4672d 1144 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
06705bff
TT		 1145 {Opt_auto_da_alloc, "auto_da_alloc"},
1146 {Opt_noauto_da_alloc, "noauto_da_alloc"},
f3f12faa 1147 {Opt_err, NULL},
ac27a0ec
DK		 1148};
1149
617ba13b 1150static ext4_fsblk_t get_sb_block(void **data)
ac27a0ec 1151{
617ba13b 1152 ext4_fsblk_t sb_block;
ac27a0ec
DK		 1153 char *options = (char *) *data;
1154
1155 if (!options || strncmp(options, "sb=", 3) != 0)
1156 return 1; /* Default location */
0b8e58a1 1157
ac27a0ec 1158 options += 3;
0b8e58a1 1159 /* TODO: use simple_strtoll with >32bit ext4 */
ac27a0ec
DK		 1160 sb_block = simple_strtoul(options, &options, 0);
1161 if (*options && *options != ',') {
4776004f 1162 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
ac27a0ec
DK		 1163 (char *) *data);
1164 return 1;
1165 }
1166 if (*options == ',')
1167 options++;
1168 *data = (void *) options;
0b8e58a1 1169
ac27a0ec
DK		 1170 return sb_block;
1171}
1172
b3881f74
TT		 1173#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1174
2b2d6d01 1175static int parse_options(char *options, struct super_block *sb,
c3191067 1176 unsigned long *journal_devnum,
b3881f74 1177 unsigned int *journal_ioprio,
2b2d6d01 1178 ext4_fsblk_t *n_blocks_count, int is_remount)
ac27a0ec 1179{
617ba13b 1180 struct ext4_sb_info *sbi = EXT4_SB(sb);
2b2d6d01 1181 char *p;
ac27a0ec
DK		 1182 substring_t args[MAX_OPT_ARGS];
1183 int data_opt = 0;
1184 int option;
1185#ifdef CONFIG_QUOTA
dfc5d03f 1186 int qtype, qfmt;
ac27a0ec
DK		 1187 char *qname;
1188#endif
1189
1190 if (!options)
1191 return 1;
1192
2b2d6d01 1193 while ((p = strsep(&options, ",")) != NULL) {
ac27a0ec
DK		 1194 int token;
1195 if (!*p)
1196 continue;
1197
1198 token = match_token(p, tokens, args);
1199 switch (token) {
1200 case Opt_bsd_df:
2b2d6d01 1201 clear_opt(sbi->s_mount_opt, MINIX_DF);
ac27a0ec
DK		 1202 break;
1203 case Opt_minix_df:
2b2d6d01 1204 set_opt(sbi->s_mount_opt, MINIX_DF);
ac27a0ec
DK		 1205 break;
1206 case Opt_grpid:
2b2d6d01 1207 set_opt(sbi->s_mount_opt, GRPID);
ac27a0ec
DK		 1208 break;
1209 case Opt_nogrpid:
2b2d6d01 1210 clear_opt(sbi->s_mount_opt, GRPID);
ac27a0ec
DK		 1211 break;
1212 case Opt_resuid:
1213 if (match_int(&args[0], &option))
1214 return 0;
1215 sbi->s_resuid = option;
1216 break;
1217 case Opt_resgid:
1218 if (match_int(&args[0], &option))
1219 return 0;
1220 sbi->s_resgid = option;
1221 break;
1222 case Opt_sb:
1223 /* handled by get_sb_block() instead of here */
1224 /* *sb_block = match_int(&args[0]); */
1225 break;
1226 case Opt_err_panic:
2b2d6d01
TT		 1227 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1228 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1229 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
ac27a0ec
DK		 1230 break;
1231 case Opt_err_ro:
2b2d6d01
TT		 1232 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1233 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1234 set_opt(sbi->s_mount_opt, ERRORS_RO);
ac27a0ec
DK		 1235 break;
1236 case Opt_err_cont:
2b2d6d01
TT		 1237 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1238 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1239 set_opt(sbi->s_mount_opt, ERRORS_CONT);
ac27a0ec
DK		 1240 break;
1241 case Opt_nouid32:
2b2d6d01 1242 set_opt(sbi->s_mount_opt, NO_UID32);
ac27a0ec 1243 break;
ac27a0ec 1244 case Opt_debug:
2b2d6d01 1245 set_opt(sbi->s_mount_opt, DEBUG);
ac27a0ec
DK		 1246 break;
1247 case Opt_oldalloc:
2b2d6d01 1248 set_opt(sbi->s_mount_opt, OLDALLOC);
ac27a0ec
DK		 1249 break;
1250 case Opt_orlov:
2b2d6d01 1251 clear_opt(sbi->s_mount_opt, OLDALLOC);
ac27a0ec 1252 break;
03010a33 1253#ifdef CONFIG_EXT4_FS_XATTR
ac27a0ec 1254 case Opt_user_xattr:
2b2d6d01 1255 set_opt(sbi->s_mount_opt, XATTR_USER);
ac27a0ec
DK		 1256 break;
1257 case Opt_nouser_xattr:
2b2d6d01 1258 clear_opt(sbi->s_mount_opt, XATTR_USER);
ac27a0ec
DK		 1259 break;
1260#else
1261 case Opt_user_xattr:
1262 case Opt_nouser_xattr:
b31e1552 1263 ext4_msg(sb, KERN_ERR, "(no)user_xattr options not supported");
ac27a0ec
DK		 1264 break;
1265#endif
03010a33 1266#ifdef CONFIG_EXT4_FS_POSIX_ACL
ac27a0ec
DK		 1267 case Opt_acl:
1268 set_opt(sbi->s_mount_opt, POSIX_ACL);
1269 break;
1270 case Opt_noacl:
1271 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1272 break;
1273#else
1274 case Opt_acl:
1275 case Opt_noacl:
b31e1552 1276 ext4_msg(sb, KERN_ERR, "(no)acl options not supported");
ac27a0ec
DK		 1277 break;
1278#endif
ac27a0ec
DK		 1279 case Opt_journal_update:
1280 /* @@@ FIXME */
1281 /* Eventually we will want to be able to create
1282 a journal file here. For now, only allow the
1283 user to specify an existing inode to be the
1284 journal file. */
1285 if (is_remount) {
b31e1552
ES		 1286 ext4_msg(sb, KERN_ERR,
1287 "Cannot specify journal on remount");
ac27a0ec
DK		 1288 return 0;
1289 }
2b2d6d01 1290 set_opt(sbi->s_mount_opt, UPDATE_JOURNAL);
ac27a0ec 1291 break;
ac27a0ec
DK		 1292 case Opt_journal_dev:
1293 if (is_remount) {
b31e1552
ES		 1294 ext4_msg(sb, KERN_ERR,
1295 "Cannot specify journal on remount");
ac27a0ec
DK		 1296 return 0;
1297 }
1298 if (match_int(&args[0], &option))
1299 return 0;
1300 *journal_devnum = option;
1301 break;
818d276c 1302 case Opt_journal_checksum:
d4da6c9c
LT		 1303 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1304 break;
818d276c
GS		 1305 case Opt_journal_async_commit:
1306 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
d4da6c9c 1307 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
818d276c 1308 break;
ac27a0ec 1309 case Opt_noload:
2b2d6d01 1310 set_opt(sbi->s_mount_opt, NOLOAD);
ac27a0ec
DK		 1311 break;
1312 case Opt_commit:
1313 if (match_int(&args[0], &option))
1314 return 0;
1315 if (option < 0)
1316 return 0;
1317 if (option == 0)
cd02ff0b 1318 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
ac27a0ec
DK		 1319 sbi->s_commit_interval = HZ * option;
1320 break;
30773840
TT		 1321 case Opt_max_batch_time:
1322 if (match_int(&args[0], &option))
1323 return 0;
1324 if (option < 0)
1325 return 0;
1326 if (option == 0)
1327 option = EXT4_DEF_MAX_BATCH_TIME;
1328 sbi->s_max_batch_time = option;
1329 break;
1330 case Opt_min_batch_time:
1331 if (match_int(&args[0], &option))
1332 return 0;
1333 if (option < 0)
1334 return 0;
1335 sbi->s_min_batch_time = option;
1336 break;
ac27a0ec 1337 case Opt_data_journal:
617ba13b 1338 data_opt = EXT4_MOUNT_JOURNAL_DATA;
ac27a0ec
DK		 1339 goto datacheck;
1340 case Opt_data_ordered:
617ba13b 1341 data_opt = EXT4_MOUNT_ORDERED_DATA;
ac27a0ec
DK		 1342 goto datacheck;
1343 case Opt_data_writeback:
617ba13b 1344 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
ac27a0ec
DK		 1345 datacheck:
1346 if (is_remount) {
617ba13b 1347 if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
ac27a0ec 1348 != data_opt) {
b31e1552
ES		 1349 ext4_msg(sb, KERN_ERR,
1350 "Cannot change data mode on remount");
ac27a0ec
DK		 1351 return 0;
1352 }
1353 } else {
617ba13b 1354 sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
ac27a0ec
DK		 1355 sbi->s_mount_opt |= data_opt;
1356 }
1357 break;
5bf5683a
HK		 1358 case Opt_data_err_abort:
1359 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1360 break;
1361 case Opt_data_err_ignore:
1362 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1363 break;
ac27a0ec
DK		 1364#ifdef CONFIG_QUOTA
1365 case Opt_usrjquota:
1366 qtype = USRQUOTA;
1367 goto set_qf_name;
1368 case Opt_grpjquota:
1369 qtype = GRPQUOTA;
1370set_qf_name:
17bd13b3 1371 if (sb_any_quota_loaded(sb) &&
dfc5d03f 1372 !sbi->s_qf_names[qtype]) {
b31e1552
ES		 1373 ext4_msg(sb, KERN_ERR,
1374 "Cannot change journaled "
1375 "quota options when quota turned on");
ac27a0ec
DK		 1376 return 0;
1377 }
1378 qname = match_strdup(&args[0]);
1379 if (!qname) {
b31e1552
ES		 1380 ext4_msg(sb, KERN_ERR,
1381 "Not enough memory for "
1382 "storing quotafile name");
ac27a0ec
DK		 1383 return 0;
1384 }
1385 if (sbi->s_qf_names[qtype] &&
1386 strcmp(sbi->s_qf_names[qtype], qname)) {
b31e1552
ES		 1387 ext4_msg(sb, KERN_ERR,
1388 "%s quota file already "
1389 "specified", QTYPE2NAME(qtype));
ac27a0ec
DK		 1390 kfree(qname);
1391 return 0;
1392 }
1393 sbi->s_qf_names[qtype] = qname;
1394 if (strchr(sbi->s_qf_names[qtype], '/')) {
b31e1552
ES		 1395 ext4_msg(sb, KERN_ERR,
1396 "quotafile must be on "
1397 "filesystem root");
ac27a0ec
DK		 1398 kfree(sbi->s_qf_names[qtype]);
1399 sbi->s_qf_names[qtype] = NULL;
1400 return 0;
1401 }
1402 set_opt(sbi->s_mount_opt, QUOTA);
1403 break;
1404 case Opt_offusrjquota:
1405 qtype = USRQUOTA;
1406 goto clear_qf_name;
1407 case Opt_offgrpjquota:
1408 qtype = GRPQUOTA;
1409clear_qf_name:
17bd13b3 1410 if (sb_any_quota_loaded(sb) &&
dfc5d03f 1411 sbi->s_qf_names[qtype]) {
b31e1552 1412 ext4_msg(sb, KERN_ERR, "Cannot change "
2c8be6b2 1413 "journaled quota options when "
b31e1552 1414 "quota turned on");
ac27a0ec
DK		 1415 return 0;
1416 }
1417 /*
1418 * The space will be released later when all options
1419 * are confirmed to be correct
1420 */
1421 sbi->s_qf_names[qtype] = NULL;
1422 break;
1423 case Opt_jqfmt_vfsold:
dfc5d03f
JK		 1424 qfmt = QFMT_VFS_OLD;
1425 goto set_qf_format;
ac27a0ec 1426 case Opt_jqfmt_vfsv0:
dfc5d03f
JK		 1427 qfmt = QFMT_VFS_V0;
1428set_qf_format:
17bd13b3 1429 if (sb_any_quota_loaded(sb) &&
dfc5d03f 1430 sbi->s_jquota_fmt != qfmt) {
b31e1552 1431 ext4_msg(sb, KERN_ERR, "Cannot change "
dfc5d03f 1432 "journaled quota options when "
b31e1552 1433 "quota turned on");
dfc5d03f
JK		 1434 return 0;
1435 }
1436 sbi->s_jquota_fmt = qfmt;
ac27a0ec
DK		 1437 break;
1438 case Opt_quota:
1439 case Opt_usrquota:
1440 set_opt(sbi->s_mount_opt, QUOTA);
1441 set_opt(sbi->s_mount_opt, USRQUOTA);
1442 break;
1443 case Opt_grpquota:
1444 set_opt(sbi->s_mount_opt, QUOTA);
1445 set_opt(sbi->s_mount_opt, GRPQUOTA);
1446 break;
1447 case Opt_noquota:
17bd13b3 1448 if (sb_any_quota_loaded(sb)) {
b31e1552
ES		 1449 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1450 "options when quota turned on");
ac27a0ec
DK		 1451 return 0;
1452 }
1453 clear_opt(sbi->s_mount_opt, QUOTA);
1454 clear_opt(sbi->s_mount_opt, USRQUOTA);
1455 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1456 break;
1457#else
1458 case Opt_quota:
1459 case Opt_usrquota:
1460 case Opt_grpquota:
b31e1552
ES		 1461 ext4_msg(sb, KERN_ERR,
1462 "quota options not supported");
cd59e7b9 1463 break;
ac27a0ec
DK		 1464 case Opt_usrjquota:
1465 case Opt_grpjquota:
1466 case Opt_offusrjquota:
1467 case Opt_offgrpjquota:
1468 case Opt_jqfmt_vfsold:
1469 case Opt_jqfmt_vfsv0:
b31e1552
ES		 1470 ext4_msg(sb, KERN_ERR,
1471 "journaled quota options not supported");
ac27a0ec
DK		 1472 break;
1473 case Opt_noquota:
1474 break;
1475#endif
1476 case Opt_abort:
4ab2f15b 1477 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
ac27a0ec 1478 break;
06705bff
TT		 1479 case Opt_nobarrier:
1480 clear_opt(sbi->s_mount_opt, BARRIER);
1481 break;
ac27a0ec 1482 case Opt_barrier:
06705bff
TT		 1483 if (match_int(&args[0], &option)) {
1484 set_opt(sbi->s_mount_opt, BARRIER);
1485 break;
1486 }
ac27a0ec
DK		 1487 if (option)
1488 set_opt(sbi->s_mount_opt, BARRIER);
1489 else
1490 clear_opt(sbi->s_mount_opt, BARRIER);
1491 break;
1492 case Opt_ignore:
1493 break;
1494 case Opt_resize:
1495 if (!is_remount) {
b31e1552
ES		 1496 ext4_msg(sb, KERN_ERR,
1497 "resize option only available "
1498 "for remount");
ac27a0ec
DK		 1499 return 0;
1500 }
1501 if (match_int(&args[0], &option) != 0)
1502 return 0;
1503 *n_blocks_count = option;
1504 break;
1505 case Opt_nobh:
1506 set_opt(sbi->s_mount_opt, NOBH);
1507 break;
1508 case Opt_bh:
1509 clear_opt(sbi->s_mount_opt, NOBH);
1510 break;
25ec56b5
JNC		 1511 case Opt_i_version:
1512 set_opt(sbi->s_mount_opt, I_VERSION);
1513 sb->s_flags |= MS_I_VERSION;
1514 break;
dd919b98
AK		 1515 case Opt_nodelalloc:
1516 clear_opt(sbi->s_mount_opt, DELALLOC);
1517 break;
c9de560d
AT		 1518 case Opt_stripe:
1519 if (match_int(&args[0], &option))
1520 return 0;
1521 if (option < 0)
1522 return 0;
1523 sbi->s_stripe = option;
1524 break;
64769240
AT		 1525 case Opt_delalloc:
1526 set_opt(sbi->s_mount_opt, DELALLOC);
1527 break;
6fd058f7
TT		 1528 case Opt_block_validity:
1529 set_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1530 break;
1531 case Opt_noblock_validity:
1532 clear_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1533 break;
240799cd
TT		 1534 case Opt_inode_readahead_blks:
1535 if (match_int(&args[0], &option))
1536 return 0;
1537 if (option < 0 || option > (1 << 30))
1538 return 0;
f7c43950 1539 if (!is_power_of_2(option)) {
b31e1552
ES		 1540 ext4_msg(sb, KERN_ERR,
1541 "EXT4-fs: inode_readahead_blks"
1542 " must be a power of 2");
3197ebdb
TT		 1543 return 0;
1544 }
240799cd
TT		 1545 sbi->s_inode_readahead_blks = option;
1546 break;
b3881f74
TT		 1547 case Opt_journal_ioprio:
1548 if (match_int(&args[0], &option))
1549 return 0;
1550 if (option < 0 || option > 7)
1551 break;
1552 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1553 option);
1554 break;
06705bff
TT		 1555 case Opt_noauto_da_alloc:
1556 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1557 break;
afd4672d 1558 case Opt_auto_da_alloc:
06705bff
TT		 1559 if (match_int(&args[0], &option)) {
1560 clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1561 break;
1562 }
afd4672d
TT		 1563 if (option)
1564 clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1565 else
1566 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1567 break;
ac27a0ec 1568 default:
b31e1552
ES		 1569 ext4_msg(sb, KERN_ERR,
1570 "Unrecognized mount option \"%s\" "
1571 "or missing value", p);
ac27a0ec
DK		 1572 return 0;
1573 }
1574 }
1575#ifdef CONFIG_QUOTA
1576 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
617ba13b 1577 if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
ac27a0ec
DK		 1578 sbi->s_qf_names[USRQUOTA])
1579 clear_opt(sbi->s_mount_opt, USRQUOTA);
1580
617ba13b 1581 if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
ac27a0ec
DK		 1582 sbi->s_qf_names[GRPQUOTA])
1583 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1584
1585 if ((sbi->s_qf_names[USRQUOTA] &&
617ba13b 1586 (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
ac27a0ec 1587 (sbi->s_qf_names[GRPQUOTA] &&
617ba13b 1588 (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
b31e1552
ES		 1589 ext4_msg(sb, KERN_ERR, "old and new quota "
1590 "format mixing");
ac27a0ec
DK		 1591 return 0;
1592 }
1593
1594 if (!sbi->s_jquota_fmt) {
b31e1552
ES		 1595 ext4_msg(sb, KERN_ERR, "journaled quota format "
1596 "not specified");
ac27a0ec
DK		 1597 return 0;
1598 }
1599 } else {
1600 if (sbi->s_jquota_fmt) {
b31e1552 1601 ext4_msg(sb, KERN_ERR, "journaled quota format "
2c8be6b2 1602 "specified with no journaling "
b31e1552 1603 "enabled");
ac27a0ec
DK		 1604 return 0;
1605 }
1606 }
1607#endif
1608 return 1;
1609}
1610
617ba13b 1611static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 1612 int read_only)
1613{
617ba13b 1614 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 1615 int res = 0;
1616
617ba13b 1617 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
b31e1552
ES		 1618 ext4_msg(sb, KERN_ERR, "revision level too high, "
1619 "forcing read-only mode");
ac27a0ec
DK		 1620 res = MS_RDONLY;
1621 }
1622 if (read_only)
1623 return res;
617ba13b 1624 if (!(sbi->s_mount_state & EXT4_VALID_FS))
b31e1552
ES		 1625 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1626 "running e2fsck is recommended");
617ba13b 1627 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
b31e1552
ES		 1628 ext4_msg(sb, KERN_WARNING,
1629 "warning: mounting fs with errors, "
1630 "running e2fsck is recommended");
ac27a0ec
DK		 1631 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1632 le16_to_cpu(es->s_mnt_count) >=
1633 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
b31e1552
ES		 1634 ext4_msg(sb, KERN_WARNING,
1635 "warning: maximal mount count reached, "
1636 "running e2fsck is recommended");
ac27a0ec
DK		 1637 else if (le32_to_cpu(es->s_checkinterval) &&
1638 (le32_to_cpu(es->s_lastcheck) +
1639 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
b31e1552
ES		 1640 ext4_msg(sb, KERN_WARNING,
1641 "warning: checktime reached, "
1642 "running e2fsck is recommended");
0b8e58a1 1643 if (!sbi->s_journal)
0390131b 1644 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec 1645 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 1646 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 1647 le16_add_cpu(&es->s_mnt_count, 1);
ac27a0ec 1648 es->s_mtime = cpu_to_le32(get_seconds());
617ba13b 1649 ext4_update_dynamic_rev(sb);
0390131b
FM		 1650 if (sbi->s_journal)
1651 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 1652
e2d67052 1653 ext4_commit_super(sb, 1);
ac27a0ec 1654 if (test_opt(sb, DEBUG))
a9df9a49 1655 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
7f4520cc 1656 "bpg=%lu, ipg=%lu, mo=%04x]\n",
ac27a0ec
DK		 1657 sb->s_blocksize,
1658 sbi->s_groups_count,
617ba13b
MC		 1659 EXT4_BLOCKS_PER_GROUP(sb),
1660 EXT4_INODES_PER_GROUP(sb),
ac27a0ec
DK		 1661 sbi->s_mount_opt);
1662
ac27a0ec
DK		 1663 return res;
1664}
1665
772cb7c8
JS		 1666static int ext4_fill_flex_info(struct super_block *sb)
1667{
1668 struct ext4_sb_info *sbi = EXT4_SB(sb);
1669 struct ext4_group_desc *gdp = NULL;
772cb7c8
JS		 1670 ext4_group_t flex_group_count;
1671 ext4_group_t flex_group;
1672 int groups_per_flex = 0;
c5ca7c76 1673 size_t size;
772cb7c8
JS		 1674 int i;
1675
503358ae
TT		 1676 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1677 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1678
1679 if (groups_per_flex < 2) {
772cb7c8
JS		 1680 sbi->s_log_groups_per_flex = 0;
1681 return 1;
1682 }
1683
c62a11fd
FB		 1684 /* We allocate both existing and potentially added groups */
1685 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
d94e99a6
AK		 1686 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1687 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
c5ca7c76
TT		 1688 size = flex_group_count * sizeof(struct flex_groups);
1689 sbi->s_flex_groups = kzalloc(size, GFP_KERNEL);
1690 if (sbi->s_flex_groups == NULL) {
1691 sbi->s_flex_groups = vmalloc(size);
1692 if (sbi->s_flex_groups)
1693 memset(sbi->s_flex_groups, 0, size);
1694 }
772cb7c8 1695 if (sbi->s_flex_groups == NULL) {
b31e1552
ES		 1696 ext4_msg(sb, KERN_ERR, "not enough memory for "
1697 "%u flex groups", flex_group_count);
772cb7c8
JS		 1698 goto failed;
1699 }
772cb7c8 1700
772cb7c8 1701 for (i = 0; i < sbi->s_groups_count; i++) {
88b6edd1 1702 gdp = ext4_get_group_desc(sb, i, NULL);
772cb7c8
JS		 1703
1704 flex_group = ext4_flex_group(sbi, i);
7ad9bb65
TT		 1705 atomic_add(ext4_free_inodes_count(sb, gdp),
1706 &sbi->s_flex_groups[flex_group].free_inodes);
1707 atomic_add(ext4_free_blks_count(sb, gdp),
1708 &sbi->s_flex_groups[flex_group].free_blocks);
1709 atomic_add(ext4_used_dirs_count(sb, gdp),
1710 &sbi->s_flex_groups[flex_group].used_dirs);
772cb7c8
JS		 1711 }
1712
1713 return 1;
1714failed:
1715 return 0;
1716}
1717
717d50e4
AD		 1718__le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1719 struct ext4_group_desc *gdp)
1720{
1721 __u16 crc = 0;
1722
1723 if (sbi->s_es->s_feature_ro_compat &
1724 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1725 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1726 __le32 le_group = cpu_to_le32(block_group);
1727
1728 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1729 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1730 crc = crc16(crc, (__u8 *)gdp, offset);
1731 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1732 /* for checksum of struct ext4_group_desc do the rest...*/
1733 if ((sbi->s_es->s_feature_incompat &
1734 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1735 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1736 crc = crc16(crc, (__u8 *)gdp + offset,
1737 le16_to_cpu(sbi->s_es->s_desc_size) -
1738 offset);
1739 }
1740
1741 return cpu_to_le16(crc);
1742}
1743
1744int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1745 struct ext4_group_desc *gdp)
1746{
1747 if ((sbi->s_es->s_feature_ro_compat &
1748 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1749 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1750 return 0;
1751
1752 return 1;
1753}
1754
ac27a0ec 1755/* Called at mount-time, super-block is locked */
197cd65a 1756static int ext4_check_descriptors(struct super_block *sb)
ac27a0ec 1757{
617ba13b
MC		 1758 struct ext4_sb_info *sbi = EXT4_SB(sb);
1759 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1760 ext4_fsblk_t last_block;
bd81d8ee
LV		 1761 ext4_fsblk_t block_bitmap;
1762 ext4_fsblk_t inode_bitmap;
1763 ext4_fsblk_t inode_table;
ce421581 1764 int flexbg_flag = 0;
fd2d4291 1765 ext4_group_t i;
ac27a0ec 1766
ce421581
JS		 1767 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1768 flexbg_flag = 1;
1769
af5bc92d 1770 ext4_debug("Checking group descriptors");
ac27a0ec 1771
197cd65a
AM		 1772 for (i = 0; i < sbi->s_groups_count; i++) {
1773 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1774
ce421581 1775 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 1776 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 1777 else
1778 last_block = first_block +
617ba13b 1779 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 1780
8fadc143 1781 block_bitmap = ext4_block_bitmap(sb, gdp);
2b2d6d01 1782 if (block_bitmap < first_block || block_bitmap > last_block) {
b31e1552 1783 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 1784 "Block bitmap for group %u not in group "
b31e1552 1785 "(block %llu)!", i, block_bitmap);
ac27a0ec
DK		 1786 return 0;
1787 }
8fadc143 1788 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2b2d6d01 1789 if (inode_bitmap < first_block || inode_bitmap > last_block) {
b31e1552 1790 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 1791 "Inode bitmap for group %u not in group "
b31e1552 1792 "(block %llu)!", i, inode_bitmap);
ac27a0ec
DK		 1793 return 0;
1794 }
8fadc143 1795 inode_table = ext4_inode_table(sb, gdp);
bd81d8ee 1796 if (inode_table < first_block ||
2b2d6d01 1797 inode_table + sbi->s_itb_per_group - 1 > last_block) {
b31e1552 1798 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 1799 "Inode table for group %u not in group "
b31e1552 1800 "(block %llu)!", i, inode_table);
ac27a0ec
DK		 1801 return 0;
1802 }
955ce5f5 1803 ext4_lock_group(sb, i);
717d50e4 1804 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
b31e1552
ES		 1805 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1806 "Checksum for group %u failed (%u!=%u)",
1807 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1808 gdp)), le16_to_cpu(gdp->bg_checksum));
7ee1ec4c 1809 if (!(sb->s_flags & MS_RDONLY)) {
955ce5f5 1810 ext4_unlock_group(sb, i);
8a266467 1811 return 0;
7ee1ec4c 1812 }
717d50e4 1813 }
955ce5f5 1814 ext4_unlock_group(sb, i);
ce421581
JS		 1815 if (!flexbg_flag)
1816 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec
DK		 1817 }
1818
bd81d8ee 1819 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
0b8e58a1 1820 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
ac27a0ec
DK		 1821 return 1;
1822}
1823
617ba13b 1824/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 1825 * the superblock) which were deleted from all directories, but held open by
1826 * a process at the time of a crash. We walk the list and try to delete these
1827 * inodes at recovery time (only with a read-write filesystem).
1828 *
1829 * In order to keep the orphan inode chain consistent during traversal (in
1830 * case of crash during recovery), we link each inode into the superblock
1831 * orphan list_head and handle it the same way as an inode deletion during
1832 * normal operation (which journals the operations for us).
1833 *
1834 * We only do an iget() and an iput() on each inode, which is very safe if we
1835 * accidentally point at an in-use or already deleted inode. The worst that
1836 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 1837 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 1838 * e2fsck was run on this filesystem, and it must have already done the orphan
1839 * inode cleanup for us, so we can safely abort without any further action.
1840 */
2b2d6d01
TT		 1841static void ext4_orphan_cleanup(struct super_block *sb,
1842 struct ext4_super_block *es)
ac27a0ec
DK		 1843{
1844 unsigned int s_flags = sb->s_flags;
1845 int nr_orphans = 0, nr_truncates = 0;
1846#ifdef CONFIG_QUOTA
1847 int i;
1848#endif
1849 if (!es->s_last_orphan) {
1850 jbd_debug(4, "no orphan inodes to clean up\n");
1851 return;
1852 }
1853
a8f48a95 1854 if (bdev_read_only(sb->s_bdev)) {
b31e1552
ES		 1855 ext4_msg(sb, KERN_ERR, "write access "
1856 "unavailable, skipping orphan cleanup");
a8f48a95
ES		 1857 return;
1858 }
1859
617ba13b 1860 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
ac27a0ec
DK		 1861 if (es->s_last_orphan)
1862 jbd_debug(1, "Errors on filesystem, "
1863 "clearing orphan list.\n");
1864 es->s_last_orphan = 0;
1865 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1866 return;
1867 }
1868
1869 if (s_flags & MS_RDONLY) {
b31e1552 1870 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
ac27a0ec
DK		 1871 sb->s_flags &= ~MS_RDONLY;
1872 }
1873#ifdef CONFIG_QUOTA
1874 /* Needed for iput() to work correctly and not trash data */
1875 sb->s_flags |= MS_ACTIVE;
1876 /* Turn on quotas so that they are updated correctly */
1877 for (i = 0; i < MAXQUOTAS; i++) {
617ba13b
MC		 1878 if (EXT4_SB(sb)->s_qf_names[i]) {
1879 int ret = ext4_quota_on_mount(sb, i);
ac27a0ec 1880 if (ret < 0)
b31e1552
ES		 1881 ext4_msg(sb, KERN_ERR,
1882 "Cannot turn on journaled "
1883 "quota: error %d", ret);
ac27a0ec
DK		 1884 }
1885 }
1886#endif
1887
1888 while (es->s_last_orphan) {
1889 struct inode *inode;
1890
97bd42b9
JB		 1891 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1892 if (IS_ERR(inode)) {
ac27a0ec
DK		 1893 es->s_last_orphan = 0;
1894 break;
1895 }
1896
617ba13b 1897 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
a269eb18 1898 vfs_dq_init(inode);
ac27a0ec 1899 if (inode->i_nlink) {
b31e1552
ES		 1900 ext4_msg(sb, KERN_DEBUG,
1901 "%s: truncating inode %lu to %lld bytes",
46e665e9 1902 __func__, inode->i_ino, inode->i_size);
e5f8eab8 1903 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
ac27a0ec 1904 inode->i_ino, inode->i_size);
617ba13b 1905 ext4_truncate(inode);
ac27a0ec
DK		 1906 nr_truncates++;
1907 } else {
b31e1552
ES		 1908 ext4_msg(sb, KERN_DEBUG,
1909 "%s: deleting unreferenced inode %lu",
46e665e9 1910 __func__, inode->i_ino);
ac27a0ec
DK		 1911 jbd_debug(2, "deleting unreferenced inode %lu\n",
1912 inode->i_ino);
1913 nr_orphans++;
1914 }
1915 iput(inode); /* The delete magic happens here! */
1916 }
1917
2b2d6d01 1918#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
ac27a0ec
DK		 1919
1920 if (nr_orphans)
b31e1552
ES		 1921 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
1922 PLURAL(nr_orphans));
ac27a0ec 1923 if (nr_truncates)
b31e1552
ES		 1924 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
1925 PLURAL(nr_truncates));
ac27a0ec
DK		 1926#ifdef CONFIG_QUOTA
1927 /* Turn quotas off */
1928 for (i = 0; i < MAXQUOTAS; i++) {
1929 if (sb_dqopt(sb)->files[i])
6f28e087 1930 vfs_quota_off(sb, i, 0);
ac27a0ec
DK		 1931 }
1932#endif
1933 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1934}
0b8e58a1 1935
cd2291a4
ES		 1936/*
1937 * Maximal extent format file size.
1938 * Resulting logical blkno at s_maxbytes must fit in our on-disk
1939 * extent format containers, within a sector_t, and within i_blocks
1940 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
1941 * so that won't be a limiting factor.
1942 *
1943 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1944 */
f287a1a5 1945static loff_t ext4_max_size(int blkbits, int has_huge_files)
cd2291a4
ES		 1946{
1947 loff_t res;
1948 loff_t upper_limit = MAX_LFS_FILESIZE;
1949
1950 /* small i_blocks in vfs inode? */
f287a1a5 1951 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
cd2291a4 1952 /*
90c699a9 1953 * CONFIG_LBDAF is not enabled implies the inode
cd2291a4
ES		 1954 * i_block represent total blocks in 512 bytes
1955 * 32 == size of vfs inode i_blocks * 8
1956 */
1957 upper_limit = (1LL << 32) - 1;
1958
1959 /* total blocks in file system block size */
1960 upper_limit >>= (blkbits - 9);
1961 upper_limit <<= blkbits;
1962 }
1963
1964 /* 32-bit extent-start container, ee_block */
1965 res = 1LL << 32;
1966 res <<= blkbits;
1967 res -= 1;
1968
1969 /* Sanity check against vm- & vfs- imposed limits */
1970 if (res > upper_limit)
1971 res = upper_limit;
1972
1973 return res;
1974}
ac27a0ec 1975
ac27a0ec 1976/*
cd2291a4 1977 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 1978 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
1979 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 1980 */
f287a1a5 1981static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
ac27a0ec 1982{
617ba13b 1983 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 1984 int meta_blocks;
1985 loff_t upper_limit;
0b8e58a1
AD		 1986 /* This is calculated to be the largest file size for a dense, block
1987 * mapped file such that the file's total number of 512-byte sectors,
1988 * including data and all indirect blocks, does not exceed (2^48 - 1).
1989 *
1990 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
1991 * number of 512-byte sectors of the file.
0fc1b451
AK		 1992 */
1993
f287a1a5 1994 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
0fc1b451 1995 /*
90c699a9 1996 * !has_huge_files or CONFIG_LBDAF not enabled implies that
0b8e58a1
AD		 1997 * the inode i_block field represents total file blocks in
1998 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
0fc1b451
AK		 1999 */
2000 upper_limit = (1LL << 32) - 1;
2001
2002 /* total blocks in file system block size */
2003 upper_limit >>= (bits - 9);
2004
2005 } else {
8180a562
AK		 2006 /*
2007 * We use 48 bit ext4_inode i_blocks
2008 * With EXT4_HUGE_FILE_FL set the i_blocks
2009 * represent total number of blocks in
2010 * file system block size
2011 */
0fc1b451
AK		 2012 upper_limit = (1LL << 48) - 1;
2013
0fc1b451
AK		 2014 }
2015
2016 /* indirect blocks */
2017 meta_blocks = 1;
2018 /* double indirect blocks */
2019 meta_blocks += 1 + (1LL << (bits-2));
2020 /* tripple indirect blocks */
2021 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2022
2023 upper_limit -= meta_blocks;
2024 upper_limit <<= bits;
ac27a0ec
DK		 2025
2026 res += 1LL << (bits-2);
2027 res += 1LL << (2*(bits-2));
2028 res += 1LL << (3*(bits-2));
2029 res <<= bits;
2030 if (res > upper_limit)
2031 res = upper_limit;
0fc1b451
AK		 2032
2033 if (res > MAX_LFS_FILESIZE)
2034 res = MAX_LFS_FILESIZE;
2035
ac27a0ec
DK		 2036 return res;
2037}
2038
617ba13b 2039static ext4_fsblk_t descriptor_loc(struct super_block *sb,
0b8e58a1 2040 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 2041{
617ba13b 2042 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 2043 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 2044 int has_super = 0;
2045
2046 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2047
617ba13b 2048 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
ac27a0ec 2049 nr < first_meta_bg)
70bbb3e0 2050 return logical_sb_block + nr + 1;
ac27a0ec 2051 bg = sbi->s_desc_per_block * nr;
617ba13b 2052 if (ext4_bg_has_super(sb, bg))
ac27a0ec 2053 has_super = 1;
0b8e58a1 2054
617ba13b 2055 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 2056}
2057
c9de560d
AT		 2058/**
2059 * ext4_get_stripe_size: Get the stripe size.
2060 * @sbi: In memory super block info
2061 *
2062 * If we have specified it via mount option, then
2063 * use the mount option value. If the value specified at mount time is
2064 * greater than the blocks per group use the super block value.
2065 * If the super block value is greater than blocks per group return 0.
2066 * Allocator needs it be less than blocks per group.
2067 *
2068 */
2069static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2070{
2071 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2072 unsigned long stripe_width =
2073 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2074
2075 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2076 return sbi->s_stripe;
2077
2078 if (stripe_width <= sbi->s_blocks_per_group)
2079 return stripe_width;
2080
2081 if (stride <= sbi->s_blocks_per_group)
2082 return stride;
2083
2084 return 0;
2085}
ac27a0ec 2086
3197ebdb
TT		 2087/* sysfs supprt */
2088
2089struct ext4_attr {
2090 struct attribute attr;
2091 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2092 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2093 const char *, size_t);
2094 int offset;
2095};
2096
2097static int parse_strtoul(const char *buf,
2098 unsigned long max, unsigned long *value)
2099{
2100 char *endp;
2101
2102 while (*buf && isspace(*buf))
2103 buf++;
2104 *value = simple_strtoul(buf, &endp, 0);
2105 while (*endp && isspace(*endp))
2106 endp++;
2107 if (*endp || *value > max)
2108 return -EINVAL;
2109
2110 return 0;
2111}
2112
2113static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2114 struct ext4_sb_info *sbi,
2115 char *buf)
2116{
2117 return snprintf(buf, PAGE_SIZE, "%llu\n",
2118 (s64) percpu_counter_sum(&sbi->s_dirtyblocks_counter));
2119}
2120
2121static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2122 struct ext4_sb_info *sbi, char *buf)
2123{
2124 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2125
2126 return snprintf(buf, PAGE_SIZE, "%lu\n",
2127 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2128 sbi->s_sectors_written_start) >> 1);
2129}
2130
2131static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2132 struct ext4_sb_info *sbi, char *buf)
2133{
2134 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2135
2136 return snprintf(buf, PAGE_SIZE, "%llu\n",
2137 sbi->s_kbytes_written +
2138 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2139 EXT4_SB(sb)->s_sectors_written_start) >> 1));
2140}
2141
2142static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2143 struct ext4_sb_info *sbi,
2144 const char *buf, size_t count)
2145{
2146 unsigned long t;
2147
2148 if (parse_strtoul(buf, 0x40000000, &t))
2149 return -EINVAL;
2150
f7c43950 2151 if (!is_power_of_2(t))
3197ebdb
TT		 2152 return -EINVAL;
2153
2154 sbi->s_inode_readahead_blks = t;
2155 return count;
2156}
2157
2158static ssize_t sbi_ui_show(struct ext4_attr *a,
0b8e58a1 2159 struct ext4_sb_info *sbi, char *buf)
3197ebdb
TT		 2160{
2161 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2162
2163 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2164}
2165
2166static ssize_t sbi_ui_store(struct ext4_attr *a,
2167 struct ext4_sb_info *sbi,
2168 const char *buf, size_t count)
2169{
2170 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2171 unsigned long t;
2172
2173 if (parse_strtoul(buf, 0xffffffff, &t))
2174 return -EINVAL;
2175 *ui = t;
2176 return count;
2177}
2178
2179#define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2180static struct ext4_attr ext4_attr_##_name = { \
2181 .attr = {.name = __stringify(_name), .mode = _mode }, \
2182 .show = _show, \
2183 .store = _store, \
2184 .offset = offsetof(struct ext4_sb_info, _elname), \
2185}
2186#define EXT4_ATTR(name, mode, show, store) \
2187static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2188
2189#define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2190#define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2191#define EXT4_RW_ATTR_SBI_UI(name, elname) \
2192 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2193#define ATTR_LIST(name) &ext4_attr_##name.attr
2194
2195EXT4_RO_ATTR(delayed_allocation_blocks);
2196EXT4_RO_ATTR(session_write_kbytes);
2197EXT4_RO_ATTR(lifetime_write_kbytes);
2198EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2199 inode_readahead_blks_store, s_inode_readahead_blks);
11013911 2200EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
3197ebdb
TT		 2201EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2202EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2203EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2204EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2205EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2206EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
55138e0b 2207EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
3197ebdb
TT		 2208
2209static struct attribute *ext4_attrs[] = {
2210 ATTR_LIST(delayed_allocation_blocks),
2211 ATTR_LIST(session_write_kbytes),
2212 ATTR_LIST(lifetime_write_kbytes),
2213 ATTR_LIST(inode_readahead_blks),
11013911 2214 ATTR_LIST(inode_goal),
3197ebdb
TT		 2215 ATTR_LIST(mb_stats),
2216 ATTR_LIST(mb_max_to_scan),
2217 ATTR_LIST(mb_min_to_scan),
2218 ATTR_LIST(mb_order2_req),
2219 ATTR_LIST(mb_stream_req),
2220 ATTR_LIST(mb_group_prealloc),
55138e0b 2221 ATTR_LIST(max_writeback_mb_bump),
3197ebdb
TT		 2222 NULL,
2223};
2224
2225static ssize_t ext4_attr_show(struct kobject *kobj,
2226 struct attribute *attr, char *buf)
2227{
2228 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2229 s_kobj);
2230 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2231
2232 return a->show ? a->show(a, sbi, buf) : 0;
2233}
2234
2235static ssize_t ext4_attr_store(struct kobject *kobj,
2236 struct attribute *attr,
2237 const char *buf, size_t len)
2238{
2239 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2240 s_kobj);
2241 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2242
2243 return a->store ? a->store(a, sbi, buf, len) : 0;
2244}
2245
2246static void ext4_sb_release(struct kobject *kobj)
2247{
2248 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2249 s_kobj);
2250 complete(&sbi->s_kobj_unregister);
2251}
2252
2253
2254static struct sysfs_ops ext4_attr_ops = {
2255 .show = ext4_attr_show,
2256 .store = ext4_attr_store,
2257};
2258
2259static struct kobj_type ext4_ktype = {
2260 .default_attrs = ext4_attrs,
2261 .sysfs_ops = &ext4_attr_ops,
2262 .release = ext4_sb_release,
2263};
2264
a13fb1a4
ES		 2265/*
2266 * Check whether this filesystem can be mounted based on
2267 * the features present and the RDONLY/RDWR mount requested.
2268 * Returns 1 if this filesystem can be mounted as requested,
2269 * 0 if it cannot be.
2270 */
2271static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2272{
2273 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2274 ext4_msg(sb, KERN_ERR,
2275 "Couldn't mount because of "
2276 "unsupported optional features (%x)",
2277 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2278 ~EXT4_FEATURE_INCOMPAT_SUPP));
2279 return 0;
2280 }
2281
2282 if (readonly)
2283 return 1;
2284
2285 /* Check that feature set is OK for a read-write mount */
2286 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2287 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2288 "unsupported optional features (%x)",
2289 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2290 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2291 return 0;
2292 }
2293 /*
2294 * Large file size enabled file system can only be mounted
2295 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2296 */
2297 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2298 if (sizeof(blkcnt_t) < sizeof(u64)) {
2299 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2300 "cannot be mounted RDWR without "
2301 "CONFIG_LBDAF");
2302 return 0;
2303 }
2304 }
2305 return 1;
2306}
2307
2b2d6d01 2308static int ext4_fill_super(struct super_block *sb, void *data, int silent)
7477827f
AK		 2309 __releases(kernel_lock)
2310 __acquires(kernel_lock)
ac27a0ec 2311{
2b2d6d01 2312 struct buffer_head *bh;
617ba13b
MC		 2313 struct ext4_super_block *es = NULL;
2314 struct ext4_sb_info *sbi;
2315 ext4_fsblk_t block;
2316 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 2317 ext4_fsblk_t logical_sb_block;
ac27a0ec 2318 unsigned long offset = 0;
ac27a0ec
DK		 2319 unsigned long journal_devnum = 0;
2320 unsigned long def_mount_opts;
2321 struct inode *root;
9f6200bb 2322 char *cp;
0390131b 2323 const char *descr;
1d1fe1ee 2324 int ret = -EINVAL;
ac27a0ec 2325 int blocksize;
4ec11028
TT		 2326 unsigned int db_count;
2327 unsigned int i;
f287a1a5 2328 int needs_recovery, has_huge_files;
bd81d8ee 2329 __u64 blocks_count;
833f4077 2330 int err;
b3881f74 2331 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
ac27a0ec
DK		 2332
2333 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
2334 if (!sbi)
2335 return -ENOMEM;
705895b6
PE		 2336
2337 sbi->s_blockgroup_lock =
2338 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
2339 if (!sbi->s_blockgroup_lock) {
2340 kfree(sbi);
2341 return -ENOMEM;
2342 }
ac27a0ec
DK		 2343 sb->s_fs_info = sbi;
2344 sbi->s_mount_opt = 0;
617ba13b
MC		 2345 sbi->s_resuid = EXT4_DEF_RESUID;
2346 sbi->s_resgid = EXT4_DEF_RESGID;
240799cd 2347 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
d9c9bef1 2348 sbi->s_sb_block = sb_block;
afc32f7e
TT		 2349 sbi->s_sectors_written_start = part_stat_read(sb->s_bdev->bd_part,
2350 sectors[1]);
ac27a0ec
DK		 2351
2352 unlock_kernel();
2353
9f6200bb
TT		 2354 /* Cleanup superblock name */
2355 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
2356 *cp = '!';
2357
617ba13b 2358 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 2359 if (!blocksize) {
b31e1552 2360 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
ac27a0ec
DK		 2361 goto out_fail;
2362 }
2363
2364 /*
617ba13b 2365 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 2366 * block sizes. We need to calculate the offset from buffer start.
2367 */
617ba13b 2368 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 2369 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2370 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 2371 } else {
70bbb3e0 2372 logical_sb_block = sb_block;
ac27a0ec
DK		 2373 }
2374
70bbb3e0 2375 if (!(bh = sb_bread(sb, logical_sb_block))) {
b31e1552 2376 ext4_msg(sb, KERN_ERR, "unable to read superblock");
ac27a0ec
DK		 2377 goto out_fail;
2378 }
2379 /*
2380 * Note: s_es must be initialized as soon as possible because
617ba13b 2381 * some ext4 macro-instructions depend on its value
ac27a0ec 2382 */
617ba13b 2383 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
ac27a0ec
DK		 2384 sbi->s_es = es;
2385 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 2386 if (sb->s_magic != EXT4_SUPER_MAGIC)
2387 goto cantfind_ext4;
afc32f7e 2388 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
ac27a0ec
DK		 2389
2390 /* Set defaults before we parse the mount options */
2391 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
617ba13b 2392 if (def_mount_opts & EXT4_DEFM_DEBUG)
ac27a0ec 2393 set_opt(sbi->s_mount_opt, DEBUG);
617ba13b 2394 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
ac27a0ec 2395 set_opt(sbi->s_mount_opt, GRPID);
617ba13b 2396 if (def_mount_opts & EXT4_DEFM_UID16)
ac27a0ec 2397 set_opt(sbi->s_mount_opt, NO_UID32);
03010a33 2398#ifdef CONFIG_EXT4_FS_XATTR
617ba13b 2399 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
ac27a0ec 2400 set_opt(sbi->s_mount_opt, XATTR_USER);
2e7842b8 2401#endif
03010a33 2402#ifdef CONFIG_EXT4_FS_POSIX_ACL
617ba13b 2403 if (def_mount_opts & EXT4_DEFM_ACL)
ac27a0ec 2404 set_opt(sbi->s_mount_opt, POSIX_ACL);
2e7842b8 2405#endif
617ba13b
MC		 2406 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
2407 sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
2408 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
2409 sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
2410 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
2411 sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
2412
2413 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
ac27a0ec 2414 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
bb4f397a 2415 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
ceea16bf 2416 set_opt(sbi->s_mount_opt, ERRORS_CONT);
bb4f397a
AK		 2417 else
2418 set_opt(sbi->s_mount_opt, ERRORS_RO);
ac27a0ec
DK		 2419
2420 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
2421 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
30773840
TT		 2422 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
2423 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
2424 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
ac27a0ec 2425
571640ca 2426 set_opt(sbi->s_mount_opt, BARRIER);
ac27a0ec 2427
dd919b98
AK		 2428 /*
2429 * enable delayed allocation by default
2430 * Use -o nodelalloc to turn it off
2431 */
2432 set_opt(sbi->s_mount_opt, DELALLOC);
2433
b3881f74
TT		 2434 if (!parse_options((char *) data, sb, &journal_devnum,
2435 &journal_ioprio, NULL, 0))
ac27a0ec
DK		 2436 goto failed_mount;
2437
2438 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
617ba13b 2439 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec 2440
617ba13b
MC		 2441 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
2442 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
2443 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
2444 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
b31e1552
ES		 2445 ext4_msg(sb, KERN_WARNING,
2446 "feature flags set on rev 0 fs, "
2447 "running e2fsck is recommended");
469108ff 2448
ac27a0ec
DK		 2449 /*
2450 * Check feature flags regardless of the revision level, since we
2451 * previously didn't change the revision level when setting the flags,
2452 * so there is a chance incompat flags are set on a rev 0 filesystem.
2453 */
a13fb1a4 2454 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
ac27a0ec 2455 goto failed_mount;
a13fb1a4 2456
ac27a0ec
DK		 2457 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
2458
617ba13b
MC		 2459 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
2460 blocksize > EXT4_MAX_BLOCK_SIZE) {
b31e1552
ES		 2461 ext4_msg(sb, KERN_ERR,
2462 "Unsupported filesystem blocksize %d", blocksize);
ac27a0ec
DK		 2463 goto failed_mount;
2464 }
2465
ac27a0ec 2466 if (sb->s_blocksize != blocksize) {
ce40733c
AK		 2467 /* Validate the filesystem blocksize */
2468 if (!sb_set_blocksize(sb, blocksize)) {
b31e1552 2469 ext4_msg(sb, KERN_ERR, "bad block size %d",
ce40733c 2470 blocksize);
ac27a0ec
DK		 2471 goto failed_mount;
2472 }
2473
2b2d6d01 2474 brelse(bh);
70bbb3e0
AM		 2475 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2476 offset = do_div(logical_sb_block, blocksize);
2477 bh = sb_bread(sb, logical_sb_block);
ac27a0ec 2478 if (!bh) {
b31e1552
ES		 2479 ext4_msg(sb, KERN_ERR,
2480 "Can't read superblock on 2nd try");
ac27a0ec
DK		 2481 goto failed_mount;
2482 }
617ba13b 2483 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
ac27a0ec 2484 sbi->s_es = es;
617ba13b 2485 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
b31e1552
ES		 2486 ext4_msg(sb, KERN_ERR,
2487 "Magic mismatch, very weird!");
ac27a0ec
DK		 2488 goto failed_mount;
2489 }
2490 }
2491
a13fb1a4
ES		 2492 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2493 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
f287a1a5
TT		 2494 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
2495 has_huge_files);
2496 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
ac27a0ec 2497
617ba13b
MC		 2498 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2499 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2500 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
ac27a0ec
DK		 2501 } else {
2502 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2503 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
617ba13b 2504 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1330593e 2505 (!is_power_of_2(sbi->s_inode_size)) ||
ac27a0ec 2506 (sbi->s_inode_size > blocksize)) {
b31e1552
ES		 2507 ext4_msg(sb, KERN_ERR,
2508 "unsupported inode size: %d",
2b2d6d01 2509 sbi->s_inode_size);
ac27a0ec
DK		 2510 goto failed_mount;
2511 }
ef7f3835
KS		 2512 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2513 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
ac27a0ec 2514 }
0b8e58a1 2515
0d1ee42f
AR		 2516 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2517 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
8fadc143 2518 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 2519 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 2520 !is_power_of_2(sbi->s_desc_size)) {
b31e1552
ES		 2521 ext4_msg(sb, KERN_ERR,
2522 "unsupported descriptor size %lu",
0d1ee42f
AR		 2523 sbi->s_desc_size);
2524 goto failed_mount;
2525 }
2526 } else
2527 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
0b8e58a1 2528
ac27a0ec 2529 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 2530 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
b47b6f38 2531 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
617ba13b 2532 goto cantfind_ext4;
0b8e58a1 2533
617ba13b 2534 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 2535 if (sbi->s_inodes_per_block == 0)
617ba13b 2536 goto cantfind_ext4;
ac27a0ec
DK		 2537 sbi->s_itb_per_group = sbi->s_inodes_per_group /
2538 sbi->s_inodes_per_block;
0d1ee42f 2539 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 2540 sbi->s_sbh = bh;
2541 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 2542 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2543 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
0b8e58a1 2544
2b2d6d01 2545 for (i = 0; i < 4; i++)
ac27a0ec
DK		 2546 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2547 sbi->s_def_hash_version = es->s_def_hash_version;
f99b2589
TT		 2548 i = le32_to_cpu(es->s_flags);
2549 if (i & EXT2_FLAGS_UNSIGNED_HASH)
2550 sbi->s_hash_unsigned = 3;
2551 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
2552#ifdef __CHAR_UNSIGNED__
2553 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
2554 sbi->s_hash_unsigned = 3;
2555#else
2556 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
2557#endif
2558 sb->s_dirt = 1;
2559 }
ac27a0ec
DK		 2560
2561 if (sbi->s_blocks_per_group > blocksize * 8) {
b31e1552
ES		 2562 ext4_msg(sb, KERN_ERR,
2563 "#blocks per group too big: %lu",
2b2d6d01 2564 sbi->s_blocks_per_group);
ac27a0ec
DK		 2565 goto failed_mount;
2566 }
ac27a0ec 2567 if (sbi->s_inodes_per_group > blocksize * 8) {
b31e1552
ES		 2568 ext4_msg(sb, KERN_ERR,
2569 "#inodes per group too big: %lu",
2b2d6d01 2570 sbi->s_inodes_per_group);
ac27a0ec
DK		 2571 goto failed_mount;
2572 }
2573
bf43d84b
ES		 2574 /*
2575 * Test whether we have more sectors than will fit in sector_t,
2576 * and whether the max offset is addressable by the page cache.
2577 */
2578 if ((ext4_blocks_count(es) >
2579 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) ||
2580 (ext4_blocks_count(es) >
2581 (pgoff_t)(~0ULL) >> (PAGE_CACHE_SHIFT - sb->s_blocksize_bits))) {
b31e1552 2582 ext4_msg(sb, KERN_ERR, "filesystem"
bf43d84b 2583 " too large to mount safely on this system");
ac27a0ec 2584 if (sizeof(sector_t) < 8)
90c699a9 2585 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
bf43d84b 2586 ret = -EFBIG;
ac27a0ec
DK		 2587 goto failed_mount;
2588 }
2589
617ba13b
MC		 2590 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2591 goto cantfind_ext4;
e7c95593 2592
0f2ddca6
FTN		 2593 /* check blocks count against device size */
2594 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
2595 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
b31e1552
ES		 2596 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
2597 "exceeds size of device (%llu blocks)",
0f2ddca6
FTN		 2598 ext4_blocks_count(es), blocks_count);
2599 goto failed_mount;
2600 }
2601
0b8e58a1
AD		 2602 /*
2603 * It makes no sense for the first data block to be beyond the end
2604 * of the filesystem.
2605 */
2606 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
b31e1552
ES		 2607 ext4_msg(sb, KERN_WARNING, "bad geometry: first data"
2608 "block %u is beyond end of filesystem (%llu)",
2609 le32_to_cpu(es->s_first_data_block),
2610 ext4_blocks_count(es));
e7c95593
ES		 2611 goto failed_mount;
2612 }
bd81d8ee
LV		 2613 blocks_count = (ext4_blocks_count(es) -
2614 le32_to_cpu(es->s_first_data_block) +
2615 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2616 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4ec11028 2617 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
b31e1552 2618 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
4ec11028 2619 "(block count %llu, first data block %u, "
b31e1552 2620 "blocks per group %lu)", sbi->s_groups_count,
4ec11028
TT		 2621 ext4_blocks_count(es),
2622 le32_to_cpu(es->s_first_data_block),
2623 EXT4_BLOCKS_PER_GROUP(sb));
2624 goto failed_mount;
2625 }
bd81d8ee 2626 sbi->s_groups_count = blocks_count;
fb0a387d
ES		 2627 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
2628 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
617ba13b
MC		 2629 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2630 EXT4_DESC_PER_BLOCK(sb);
2b2d6d01 2631 sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
ac27a0ec
DK		 2632 GFP_KERNEL);
2633 if (sbi->s_group_desc == NULL) {
b31e1552 2634 ext4_msg(sb, KERN_ERR, "not enough memory");
ac27a0ec
DK		 2635 goto failed_mount;
2636 }
2637
3244fcb1 2638#ifdef CONFIG_PROC_FS
9f6200bb
TT		 2639 if (ext4_proc_root)
2640 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
3244fcb1 2641#endif
240799cd 2642
705895b6 2643 bgl_lock_init(sbi->s_blockgroup_lock);
ac27a0ec
DK		 2644
2645 for (i = 0; i < db_count; i++) {
70bbb3e0 2646 block = descriptor_loc(sb, logical_sb_block, i);
ac27a0ec
DK		 2647 sbi->s_group_desc[i] = sb_bread(sb, block);
2648 if (!sbi->s_group_desc[i]) {
b31e1552
ES		 2649 ext4_msg(sb, KERN_ERR,
2650 "can't read group descriptor %d", i);
ac27a0ec
DK		 2651 db_count = i;
2652 goto failed_mount2;
2653 }
2654 }
2b2d6d01 2655 if (!ext4_check_descriptors(sb)) {
b31e1552 2656 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
ac27a0ec
DK		 2657 goto failed_mount2;
2658 }
772cb7c8
JS		 2659 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2660 if (!ext4_fill_flex_info(sb)) {
b31e1552
ES		 2661 ext4_msg(sb, KERN_ERR,
2662 "unable to initialize "
2663 "flex_bg meta info!");
772cb7c8
JS		 2664 goto failed_mount2;
2665 }
2666
ac27a0ec
DK		 2667 sbi->s_gdb_count = db_count;
2668 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2669 spin_lock_init(&sbi->s_next_gen_lock);
2670
833f4077
PZ		 2671 err = percpu_counter_init(&sbi->s_freeblocks_counter,
2672 ext4_count_free_blocks(sb));
2673 if (!err) {
2674 err = percpu_counter_init(&sbi->s_freeinodes_counter,
2675 ext4_count_free_inodes(sb));
2676 }
2677 if (!err) {
2678 err = percpu_counter_init(&sbi->s_dirs_counter,
2679 ext4_count_dirs(sb));
2680 }
6bc6e63f
AK		 2681 if (!err) {
2682 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
2683 }
833f4077 2684 if (err) {
b31e1552 2685 ext4_msg(sb, KERN_ERR, "insufficient memory");
833f4077
PZ		 2686 goto failed_mount3;
2687 }
ac27a0ec 2688
c9de560d 2689 sbi->s_stripe = ext4_get_stripe_size(sbi);
55138e0b 2690 sbi->s_max_writeback_mb_bump = 128;
c9de560d 2691
ac27a0ec
DK		 2692 /*
2693 * set up enough so that it can read an inode
2694 */
9ca92389
TT		 2695 if (!test_opt(sb, NOLOAD) &&
2696 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
2697 sb->s_op = &ext4_sops;
2698 else
2699 sb->s_op = &ext4_nojournal_sops;
617ba13b
MC		 2700 sb->s_export_op = &ext4_export_ops;
2701 sb->s_xattr = ext4_xattr_handlers;
ac27a0ec 2702#ifdef CONFIG_QUOTA
617ba13b
MC		 2703 sb->s_qcop = &ext4_qctl_operations;
2704 sb->dq_op = &ext4_quota_operations;
ac27a0ec
DK		 2705#endif
2706 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3b9d4ed2 2707 mutex_init(&sbi->s_orphan_lock);
32ed5058 2708 mutex_init(&sbi->s_resize_lock);
ac27a0ec
DK		 2709
2710 sb->s_root = NULL;
2711
2712 needs_recovery = (es->s_last_orphan != 0 ||
617ba13b
MC		 2713 EXT4_HAS_INCOMPAT_FEATURE(sb,
2714 EXT4_FEATURE_INCOMPAT_RECOVER));
ac27a0ec
DK		 2715
2716 /*
2717 * The first inode we look at is the journal inode. Don't try
2718 * root first: it may be modified in the journal!
2719 */
2720 if (!test_opt(sb, NOLOAD) &&
617ba13b
MC		 2721 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2722 if (ext4_load_journal(sb, es, journal_devnum))
ac27a0ec 2723 goto failed_mount3;
0390131b
FM		 2724 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
2725 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
b31e1552
ES		 2726 ext4_msg(sb, KERN_ERR, "required journal recovery "
2727 "suppressed and not mounted read-only");
0390131b 2728 goto failed_mount4;
ac27a0ec 2729 } else {
0390131b
FM		 2730 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
2731 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
2732 sbi->s_journal = NULL;
2733 needs_recovery = 0;
2734 goto no_journal;
ac27a0ec
DK		 2735 }
2736
eb40a09c
JS		 2737 if (ext4_blocks_count(es) > 0xffffffffULL &&
2738 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2739 JBD2_FEATURE_INCOMPAT_64BIT)) {
b31e1552 2740 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
eb40a09c
JS		 2741 goto failed_mount4;
2742 }
2743
d4da6c9c
LT		 2744 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2745 jbd2_journal_set_features(sbi->s_journal,
2746 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
818d276c 2747 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
d4da6c9c
LT		 2748 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2749 jbd2_journal_set_features(sbi->s_journal,
2750 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
818d276c
GS		 2751 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2752 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
d4da6c9c
LT		 2753 } else {
2754 jbd2_journal_clear_features(sbi->s_journal,
2755 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2756 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2757 }
818d276c 2758
ac27a0ec
DK		 2759 /* We have now updated the journal if required, so we can
2760 * validate the data journaling mode. */
2761 switch (test_opt(sb, DATA_FLAGS)) {
2762 case 0:
2763 /* No mode set, assume a default based on the journal
63f57933
AM		 2764 * capabilities: ORDERED_DATA if the journal can
2765 * cope, else JOURNAL_DATA
2766 */
dab291af
MC		 2767 if (jbd2_journal_check_available_features
2768 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
ac27a0ec
DK		 2769 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2770 else
2771 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2772 break;
2773
617ba13b
MC		 2774 case EXT4_MOUNT_ORDERED_DATA:
2775 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 2776 if (!jbd2_journal_check_available_features
2777 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
b31e1552
ES		 2778 ext4_msg(sb, KERN_ERR, "Journal does not support "
2779 "requested data journaling mode");
ac27a0ec
DK		 2780 goto failed_mount4;
2781 }
2782 default:
2783 break;
2784 }
b3881f74 2785 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
ac27a0ec 2786
0390131b 2787no_journal:
ac27a0ec
DK		 2788
2789 if (test_opt(sb, NOBH)) {
617ba13b 2790 if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
b31e1552
ES		 2791 ext4_msg(sb, KERN_WARNING, "Ignoring nobh option - "
2792 "its supported only with writeback mode");
ac27a0ec
DK		 2793 clear_opt(sbi->s_mount_opt, NOBH);
2794 }
2795 }
4c0425ff
MC		 2796 EXT4_SB(sb)->dio_unwritten_wq = create_workqueue("ext4-dio-unwritten");
2797 if (!EXT4_SB(sb)->dio_unwritten_wq) {
2798 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
2799 goto failed_mount_wq;
2800 }
2801
ac27a0ec 2802 /*
dab291af 2803 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 2804 * so we can safely mount the rest of the filesystem now.
2805 */
2806
1d1fe1ee
DH		 2807 root = ext4_iget(sb, EXT4_ROOT_INO);
2808 if (IS_ERR(root)) {
b31e1552 2809 ext4_msg(sb, KERN_ERR, "get root inode failed");
1d1fe1ee 2810 ret = PTR_ERR(root);
ac27a0ec
DK		 2811 goto failed_mount4;
2812 }
2813 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1d1fe1ee 2814 iput(root);
b31e1552 2815 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
ac27a0ec
DK		 2816 goto failed_mount4;
2817 }
1d1fe1ee
DH		 2818 sb->s_root = d_alloc_root(root);
2819 if (!sb->s_root) {
b31e1552 2820 ext4_msg(sb, KERN_ERR, "get root dentry failed");
1d1fe1ee
DH		 2821 iput(root);
2822 ret = -ENOMEM;
2823 goto failed_mount4;
2824 }
ac27a0ec 2825
2b2d6d01 2826 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
ef7f3835
KS		 2827
2828 /* determine the minimum size of new large inodes, if present */
2829 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2830 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2831 EXT4_GOOD_OLD_INODE_SIZE;
2832 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2833 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2834 if (sbi->s_want_extra_isize <
2835 le16_to_cpu(es->s_want_extra_isize))
2836 sbi->s_want_extra_isize =
2837 le16_to_cpu(es->s_want_extra_isize);
2838 if (sbi->s_want_extra_isize <
2839 le16_to_cpu(es->s_min_extra_isize))
2840 sbi->s_want_extra_isize =
2841 le16_to_cpu(es->s_min_extra_isize);
2842 }
2843 }
2844 /* Check if enough inode space is available */
2845 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2846 sbi->s_inode_size) {
2847 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2848 EXT4_GOOD_OLD_INODE_SIZE;
b31e1552
ES		 2849 ext4_msg(sb, KERN_INFO, "required extra inode space not"
2850 "available");
ef7f3835
KS		 2851 }
2852
90576c0b
TT		 2853 if (test_opt(sb, DELALLOC) &&
2854 (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)) {
b31e1552
ES		 2855 ext4_msg(sb, KERN_WARNING, "Ignoring delalloc option - "
2856 "requested data journaling mode");
c2774d84 2857 clear_opt(sbi->s_mount_opt, DELALLOC);
90576c0b 2858 }
c2774d84 2859
6fd058f7
TT		 2860 err = ext4_setup_system_zone(sb);
2861 if (err) {
b31e1552
ES		 2862 ext4_msg(sb, KERN_ERR, "failed to initialize system "
2863 "zone (%d)\n", err);
6fd058f7
TT		 2864 goto failed_mount4;
2865 }
2866
c2774d84
AK		 2867 ext4_ext_init(sb);
2868 err = ext4_mb_init(sb, needs_recovery);
2869 if (err) {
b31e1552
ES		 2870 ext4_msg(sb, KERN_ERR, "failed to initalize mballoc (%d)",
2871 err);
c2774d84
AK		 2872 goto failed_mount4;
2873 }
2874
3197ebdb
TT		 2875 sbi->s_kobj.kset = ext4_kset;
2876 init_completion(&sbi->s_kobj_unregister);
2877 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
2878 "%s", sb->s_id);
2879 if (err) {
2880 ext4_mb_release(sb);
2881 ext4_ext_release(sb);
2882 goto failed_mount4;
2883 };
2884
617ba13b
MC		 2885 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
2886 ext4_orphan_cleanup(sb, es);
2887 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
0390131b 2888 if (needs_recovery) {
b31e1552 2889 ext4_msg(sb, KERN_INFO, "recovery complete");
0390131b
FM		 2890 ext4_mark_recovery_complete(sb, es);
2891 }
2892 if (EXT4_SB(sb)->s_journal) {
2893 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2894 descr = " journalled data mode";
2895 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2896 descr = " ordered data mode";
2897 else
2898 descr = " writeback data mode";
2899 } else
2900 descr = "out journal";
2901
b31e1552 2902 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s", descr);
ac27a0ec
DK		 2903
2904 lock_kernel();
2905 return 0;
2906
617ba13b 2907cantfind_ext4:
ac27a0ec 2908 if (!silent)
b31e1552 2909 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
ac27a0ec
DK		 2910 goto failed_mount;
2911
2912failed_mount4:
b31e1552 2913 ext4_msg(sb, KERN_ERR, "mount failed");
4c0425ff
MC		 2914 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
2915failed_mount_wq:
6fd058f7 2916 ext4_release_system_zone(sb);
0390131b
FM		 2917 if (sbi->s_journal) {
2918 jbd2_journal_destroy(sbi->s_journal);
2919 sbi->s_journal = NULL;
2920 }
ac27a0ec 2921failed_mount3:
c5ca7c76
TT		 2922 if (sbi->s_flex_groups) {
2923 if (is_vmalloc_addr(sbi->s_flex_groups))
2924 vfree(sbi->s_flex_groups);
2925 else
2926 kfree(sbi->s_flex_groups);
2927 }
ac27a0ec
DK		 2928 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2929 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2930 percpu_counter_destroy(&sbi->s_dirs_counter);
6bc6e63f 2931 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
ac27a0ec
DK		 2932failed_mount2:
2933 for (i = 0; i < db_count; i++)
2934 brelse(sbi->s_group_desc[i]);
2935 kfree(sbi->s_group_desc);
2936failed_mount:
240799cd 2937 if (sbi->s_proc) {
9f6200bb 2938 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 2939 }
ac27a0ec
DK		 2940#ifdef CONFIG_QUOTA
2941 for (i = 0; i < MAXQUOTAS; i++)
2942 kfree(sbi->s_qf_names[i]);
2943#endif
617ba13b 2944 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 2945 brelse(bh);
2946out_fail:
2947 sb->s_fs_info = NULL;
f6830165 2948 kfree(sbi->s_blockgroup_lock);
ac27a0ec
DK		 2949 kfree(sbi);
2950 lock_kernel();
1d1fe1ee 2951 return ret;
ac27a0ec
DK		 2952}
2953
2954/*
2955 * Setup any per-fs journal parameters now. We'll do this both on
2956 * initial mount, once the journal has been initialised but before we've
2957 * done any recovery; and again on any subsequent remount.
2958 */
617ba13b 2959static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 2960{
617ba13b 2961 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 2962
30773840
TT		 2963 journal->j_commit_interval = sbi->s_commit_interval;
2964 journal->j_min_batch_time = sbi->s_min_batch_time;
2965 journal->j_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 2966
2967 spin_lock(&journal->j_state_lock);
2968 if (test_opt(sb, BARRIER))
dab291af 2969 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 2970 else
dab291af 2971 journal->j_flags &= ~JBD2_BARRIER;
5bf5683a
HK		 2972 if (test_opt(sb, DATA_ERR_ABORT))
2973 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
2974 else
2975 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
ac27a0ec
DK		 2976 spin_unlock(&journal->j_state_lock);
2977}
2978
617ba13b 2979static journal_t *ext4_get_journal(struct super_block *sb,
ac27a0ec
DK		 2980 unsigned int journal_inum)
2981{
2982 struct inode *journal_inode;
2983 journal_t *journal;
2984
0390131b
FM		 2985 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
2986
ac27a0ec
DK		 2987 /* First, test for the existence of a valid inode on disk. Bad
2988 * things happen if we iget() an unused inode, as the subsequent
2989 * iput() will try to delete it. */
2990
1d1fe1ee
DH		 2991 journal_inode = ext4_iget(sb, journal_inum);
2992 if (IS_ERR(journal_inode)) {
b31e1552 2993 ext4_msg(sb, KERN_ERR, "no journal found");
ac27a0ec
DK		 2994 return NULL;
2995 }
2996 if (!journal_inode->i_nlink) {
2997 make_bad_inode(journal_inode);
2998 iput(journal_inode);
b31e1552 2999 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
ac27a0ec
DK		 3000 return NULL;
3001 }
3002
e5f8eab8 3003 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
ac27a0ec 3004 journal_inode, journal_inode->i_size);
1d1fe1ee 3005 if (!S_ISREG(journal_inode->i_mode)) {
b31e1552 3006 ext4_msg(sb, KERN_ERR, "invalid journal inode");
ac27a0ec
DK		 3007 iput(journal_inode);
3008 return NULL;
3009 }
3010
dab291af 3011 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 3012 if (!journal) {
b31e1552 3013 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
ac27a0ec
DK		 3014 iput(journal_inode);
3015 return NULL;
3016 }
3017 journal->j_private = sb;
617ba13b 3018 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 3019 return journal;
3020}
3021
617ba13b 3022static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 3023 dev_t j_dev)
3024{
2b2d6d01 3025 struct buffer_head *bh;
ac27a0ec 3026 journal_t *journal;
617ba13b
MC		 3027 ext4_fsblk_t start;
3028 ext4_fsblk_t len;
ac27a0ec 3029 int hblock, blocksize;
617ba13b 3030 ext4_fsblk_t sb_block;
ac27a0ec 3031 unsigned long offset;
2b2d6d01 3032 struct ext4_super_block *es;
ac27a0ec
DK		 3033 struct block_device *bdev;
3034
0390131b
FM		 3035 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3036
b31e1552 3037 bdev = ext4_blkdev_get(j_dev, sb);
ac27a0ec
DK		 3038 if (bdev == NULL)
3039 return NULL;
3040
3041 if (bd_claim(bdev, sb)) {
b31e1552
ES		 3042 ext4_msg(sb, KERN_ERR,
3043 "failed to claim external journal device");
9a1c3542 3044 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
ac27a0ec
DK		 3045 return NULL;
3046 }
3047
3048 blocksize = sb->s_blocksize;
e1defc4f 3049 hblock = bdev_logical_block_size(bdev);
ac27a0ec 3050 if (blocksize < hblock) {
b31e1552
ES		 3051 ext4_msg(sb, KERN_ERR,
3052 "blocksize too small for journal device");
ac27a0ec
DK		 3053 goto out_bdev;
3054 }
3055
617ba13b
MC		 3056 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3057 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 3058 set_blocksize(bdev, blocksize);
3059 if (!(bh = __bread(bdev, sb_block, blocksize))) {
b31e1552
ES		 3060 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
3061 "external journal");
ac27a0ec
DK		 3062 goto out_bdev;
3063 }
3064
617ba13b
MC		 3065 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3066 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 3067 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b 3068 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
b31e1552
ES		 3069 ext4_msg(sb, KERN_ERR, "external journal has "
3070 "bad superblock");
ac27a0ec
DK		 3071 brelse(bh);
3072 goto out_bdev;
3073 }
3074
617ba13b 3075 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
b31e1552 3076 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
ac27a0ec
DK		 3077 brelse(bh);
3078 goto out_bdev;
3079 }
3080
bd81d8ee 3081 len = ext4_blocks_count(es);
ac27a0ec
DK		 3082 start = sb_block + 1;
3083 brelse(bh); /* we're done with the superblock */
3084
dab291af 3085 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 3086 start, len, blocksize);
3087 if (!journal) {
b31e1552 3088 ext4_msg(sb, KERN_ERR, "failed to create device journal");
ac27a0ec
DK		 3089 goto out_bdev;
3090 }
3091 journal->j_private = sb;
3092 ll_rw_block(READ, 1, &journal->j_sb_buffer);
3093 wait_on_buffer(journal->j_sb_buffer);
3094 if (!buffer_uptodate(journal->j_sb_buffer)) {
b31e1552 3095 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
ac27a0ec
DK		 3096 goto out_journal;
3097 }
3098 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
b31e1552
ES		 3099 ext4_msg(sb, KERN_ERR, "External journal has more than one "
3100 "user (unsupported) - %d",
ac27a0ec
DK		 3101 be32_to_cpu(journal->j_superblock->s_nr_users));
3102 goto out_journal;
3103 }
617ba13b
MC		 3104 EXT4_SB(sb)->journal_bdev = bdev;
3105 ext4_init_journal_params(sb, journal);
ac27a0ec 3106 return journal;
0b8e58a1 3107
ac27a0ec 3108out_journal:
dab291af 3109 jbd2_journal_destroy(journal);
ac27a0ec 3110out_bdev:
617ba13b 3111 ext4_blkdev_put(bdev);
ac27a0ec
DK		 3112 return NULL;
3113}
3114
617ba13b
MC		 3115static int ext4_load_journal(struct super_block *sb,
3116 struct ext4_super_block *es,
ac27a0ec
DK		 3117 unsigned long journal_devnum)
3118{
3119 journal_t *journal;
3120 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
3121 dev_t journal_dev;
3122 int err = 0;
3123 int really_read_only;
3124
0390131b
FM		 3125 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3126
ac27a0ec
DK		 3127 if (journal_devnum &&
3128 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
b31e1552
ES		 3129 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
3130 "numbers have changed");
ac27a0ec
DK		 3131 journal_dev = new_decode_dev(journal_devnum);
3132 } else
3133 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
3134
3135 really_read_only = bdev_read_only(sb->s_bdev);
3136
3137 /*
3138 * Are we loading a blank journal or performing recovery after a
3139 * crash? For recovery, we need to check in advance whether we
3140 * can get read-write access to the device.
3141 */
617ba13b 3142 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
ac27a0ec 3143 if (sb->s_flags & MS_RDONLY) {
b31e1552
ES		 3144 ext4_msg(sb, KERN_INFO, "INFO: recovery "
3145 "required on readonly filesystem");
ac27a0ec 3146 if (really_read_only) {
b31e1552
ES		 3147 ext4_msg(sb, KERN_ERR, "write access "
3148 "unavailable, cannot proceed");
ac27a0ec
DK		 3149 return -EROFS;
3150 }
b31e1552
ES		 3151 ext4_msg(sb, KERN_INFO, "write access will "
3152 "be enabled during recovery");
ac27a0ec
DK		 3153 }
3154 }
3155
3156 if (journal_inum && journal_dev) {
b31e1552
ES		 3157 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
3158 "and inode journals!");
ac27a0ec
DK		 3159 return -EINVAL;
3160 }
3161
3162 if (journal_inum) {
617ba13b 3163 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 3164 return -EINVAL;
3165 } else {
617ba13b 3166 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 3167 return -EINVAL;
3168 }
3169
90576c0b 3170 if (!(journal->j_flags & JBD2_BARRIER))
b31e1552 3171 ext4_msg(sb, KERN_INFO, "barriers disabled");
4776004f 3172
ac27a0ec 3173 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
dab291af 3174 err = jbd2_journal_update_format(journal);
ac27a0ec 3175 if (err) {
b31e1552 3176 ext4_msg(sb, KERN_ERR, "error updating journal");
dab291af 3177 jbd2_journal_destroy(journal);
ac27a0ec
DK		 3178 return err;
3179 }
3180 }
3181
617ba13b 3182 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
dab291af 3183 err = jbd2_journal_wipe(journal, !really_read_only);
ac27a0ec 3184 if (!err)
dab291af 3185 err = jbd2_journal_load(journal);
ac27a0ec
DK		 3186
3187 if (err) {
b31e1552 3188 ext4_msg(sb, KERN_ERR, "error loading journal");
dab291af 3189 jbd2_journal_destroy(journal);
ac27a0ec
DK		 3190 return err;
3191 }
3192
617ba13b
MC		 3193 EXT4_SB(sb)->s_journal = journal;
3194 ext4_clear_journal_err(sb, es);
ac27a0ec
DK		 3195
3196 if (journal_devnum &&
3197 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3198 es->s_journal_dev = cpu_to_le32(journal_devnum);
ac27a0ec
DK		 3199
3200 /* Make sure we flush the recovery flag to disk. */
e2d67052 3201 ext4_commit_super(sb, 1);
ac27a0ec
DK		 3202 }
3203
3204 return 0;
3205}
3206
e2d67052 3207static int ext4_commit_super(struct super_block *sb, int sync)
ac27a0ec 3208{
e2d67052 3209 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
617ba13b 3210 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
c4be0c1d 3211 int error = 0;
ac27a0ec
DK		 3212
3213 if (!sbh)
c4be0c1d 3214 return error;
914258bf
TT		 3215 if (buffer_write_io_error(sbh)) {
3216 /*
3217 * Oh, dear. A previous attempt to write the
3218 * superblock failed. This could happen because the
3219 * USB device was yanked out. Or it could happen to
3220 * be a transient write error and maybe the block will
3221 * be remapped. Nothing we can do but to retry the
3222 * write and hope for the best.
3223 */
b31e1552
ES		 3224 ext4_msg(sb, KERN_ERR, "previous I/O error to "
3225 "superblock detected");
914258bf
TT		 3226 clear_buffer_write_io_error(sbh);
3227 set_buffer_uptodate(sbh);
3228 }
71290b36
TT		 3229 /*
3230 * If the file system is mounted read-only, don't update the
3231 * superblock write time. This avoids updating the superblock
3232 * write time when we are mounting the root file system
3233 * read/only but we need to replay the journal; at that point,
3234 * for people who are east of GMT and who make their clock
3235 * tick in localtime for Windows bug-for-bug compatibility,
3236 * the clock is set in the future, and this will cause e2fsck
3237 * to complain and force a full file system check.
3238 */
3239 if (!(sb->s_flags & MS_RDONLY))
3240 es->s_wtime = cpu_to_le32(get_seconds());
afc32f7e
TT		 3241 es->s_kbytes_written =
3242 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
3243 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
3244 EXT4_SB(sb)->s_sectors_written_start) >> 1));
5d1b1b3f
AK		 3245 ext4_free_blocks_count_set(es, percpu_counter_sum_positive(
3246 &EXT4_SB(sb)->s_freeblocks_counter));
3247 es->s_free_inodes_count = cpu_to_le32(percpu_counter_sum_positive(
3248 &EXT4_SB(sb)->s_freeinodes_counter));
7234ab2a 3249 sb->s_dirt = 0;
ac27a0ec
DK		 3250 BUFFER_TRACE(sbh, "marking dirty");
3251 mark_buffer_dirty(sbh);
914258bf 3252 if (sync) {
c4be0c1d
TS		 3253 error = sync_dirty_buffer(sbh);
3254 if (error)
3255 return error;
3256
3257 error = buffer_write_io_error(sbh);
3258 if (error) {
b31e1552
ES		 3259 ext4_msg(sb, KERN_ERR, "I/O error while writing "
3260 "superblock");
914258bf
TT		 3261 clear_buffer_write_io_error(sbh);
3262 set_buffer_uptodate(sbh);
3263 }
3264 }
c4be0c1d 3265 return error;
ac27a0ec
DK		 3266}
3267
ac27a0ec
DK		 3268/*
3269 * Have we just finished recovery? If so, and if we are mounting (or
3270 * remounting) the filesystem readonly, then we will end up with a
3271 * consistent fs on disk. Record that fact.
3272 */
2b2d6d01
TT		 3273static void ext4_mark_recovery_complete(struct super_block *sb,
3274 struct ext4_super_block *es)
ac27a0ec 3275{
617ba13b 3276 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 3277
0390131b
FM		 3278 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3279 BUG_ON(journal != NULL);
3280 return;
3281 }
dab291af 3282 jbd2_journal_lock_updates(journal);
7ffe1ea8
HK		 3283 if (jbd2_journal_flush(journal) < 0)
3284 goto out;
3285
617ba13b 3286 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
ac27a0ec 3287 sb->s_flags & MS_RDONLY) {
617ba13b 3288 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
e2d67052 3289 ext4_commit_super(sb, 1);
ac27a0ec 3290 }
7ffe1ea8
HK		 3291
3292out:
dab291af 3293 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 3294}
3295
3296/*
3297 * If we are mounting (or read-write remounting) a filesystem whose journal
3298 * has recorded an error from a previous lifetime, move that error to the
3299 * main filesystem now.
3300 */
2b2d6d01
TT		 3301static void ext4_clear_journal_err(struct super_block *sb,
3302 struct ext4_super_block *es)
ac27a0ec
DK		 3303{
3304 journal_t *journal;
3305 int j_errno;
3306 const char *errstr;
3307
0390131b
FM		 3308 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3309
617ba13b 3310 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 3311
3312 /*
3313 * Now check for any error status which may have been recorded in the
617ba13b 3314 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 3315 */
3316
dab291af 3317 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 3318 if (j_errno) {
3319 char nbuf[16];
3320
617ba13b 3321 errstr = ext4_decode_error(sb, j_errno, nbuf);
46e665e9 3322 ext4_warning(sb, __func__, "Filesystem error recorded "
ac27a0ec 3323 "from previous mount: %s", errstr);
46e665e9 3324 ext4_warning(sb, __func__, "Marking fs in need of "
ac27a0ec
DK		 3325 "filesystem check.");
3326
617ba13b
MC		 3327 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
3328 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
e2d67052 3329 ext4_commit_super(sb, 1);
ac27a0ec 3330
dab291af 3331 jbd2_journal_clear_err(journal);
ac27a0ec
DK		 3332 }
3333}
3334
3335/*
3336 * Force the running and committing transactions to commit,
3337 * and wait on the commit.
3338 */
617ba13b 3339int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 3340{
3341 journal_t *journal;
0390131b 3342 int ret = 0;
ac27a0ec
DK		 3343
3344 if (sb->s_flags & MS_RDONLY)
3345 return 0;
3346
617ba13b 3347 journal = EXT4_SB(sb)->s_journal;
7234ab2a 3348 if (journal)
0390131b 3349 ret = ext4_journal_force_commit(journal);
0390131b 3350
ac27a0ec
DK		 3351 return ret;
3352}
3353
2b2d6d01 3354static void ext4_write_super(struct super_block *sb)
ac27a0ec 3355{
ebc1ac16 3356 lock_super(sb);
9ca92389 3357 ext4_commit_super(sb, 1);
ebc1ac16 3358 unlock_super(sb);
ac27a0ec
DK		 3359}
3360
617ba13b 3361static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec 3362{
14ce0cb4 3363 int ret = 0;
9eddacf9 3364 tid_t target;
8d5d02e6 3365 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 3366
9bffad1e 3367 trace_ext4_sync_fs(sb, wait);
8d5d02e6
MC		 3368 flush_workqueue(sbi->dio_unwritten_wq);
3369 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
9ca92389 3370 if (wait)
8d5d02e6 3371 jbd2_log_wait_commit(sbi->s_journal, target);
0390131b 3372 }
14ce0cb4 3373 return ret;
ac27a0ec
DK		 3374}
3375
3376/*
3377 * LVM calls this function before a (read-only) snapshot is created. This
3378 * gives us a chance to flush the journal completely and mark the fs clean.
3379 */
c4be0c1d 3380static int ext4_freeze(struct super_block *sb)
ac27a0ec 3381{
c4be0c1d
TS		 3382 int error = 0;
3383 journal_t *journal;
ac27a0ec 3384
9ca92389
TT		 3385 if (sb->s_flags & MS_RDONLY)
3386 return 0;
ac27a0ec 3387
9ca92389 3388 journal = EXT4_SB(sb)->s_journal;
7ffe1ea8 3389
9ca92389
TT		 3390 /* Now we set up the journal barrier. */
3391 jbd2_journal_lock_updates(journal);
ac27a0ec 3392
9ca92389
TT		 3393 /*
3394 * Don't clear the needs_recovery flag if we failed to flush
3395 * the journal.
3396 */
3397 error = jbd2_journal_flush(journal);
3398 if (error < 0) {
3399 out:
3400 jbd2_journal_unlock_updates(journal);
3401 return error;
ac27a0ec 3402 }
9ca92389
TT		 3403
3404 /* Journal blocked and flushed, clear needs_recovery flag. */
3405 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3406 error = ext4_commit_super(sb, 1);
3407 if (error)
3408 goto out;
c4be0c1d 3409 return 0;
ac27a0ec
DK		 3410}
3411
3412/*
3413 * Called by LVM after the snapshot is done. We need to reset the RECOVER
3414 * flag here, even though the filesystem is not technically dirty yet.
3415 */
c4be0c1d 3416static int ext4_unfreeze(struct super_block *sb)
ac27a0ec 3417{
9ca92389
TT		 3418 if (sb->s_flags & MS_RDONLY)
3419 return 0;
3420
3421 lock_super(sb);
3422 /* Reset the needs_recovery flag before the fs is unlocked. */
3423 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3424 ext4_commit_super(sb, 1);
3425 unlock_super(sb);
3426 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
c4be0c1d 3427 return 0;
ac27a0ec
DK		 3428}
3429
2b2d6d01 3430static int ext4_remount(struct super_block *sb, int *flags, char *data)
ac27a0ec 3431{
2b2d6d01 3432 struct ext4_super_block *es;
617ba13b
MC		 3433 struct ext4_sb_info *sbi = EXT4_SB(sb);
3434 ext4_fsblk_t n_blocks_count = 0;
ac27a0ec 3435 unsigned long old_sb_flags;
617ba13b 3436 struct ext4_mount_options old_opts;
8a266467 3437 ext4_group_t g;
b3881f74 3438 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
ac27a0ec
DK		 3439 int err;
3440#ifdef CONFIG_QUOTA
3441 int i;
3442#endif
3443
337eb00a
AIB		 3444 lock_kernel();
3445
ac27a0ec 3446 /* Store the original options */
bbd6851a 3447 lock_super(sb);
ac27a0ec
DK		 3448 old_sb_flags = sb->s_flags;
3449 old_opts.s_mount_opt = sbi->s_mount_opt;
3450 old_opts.s_resuid = sbi->s_resuid;
3451 old_opts.s_resgid = sbi->s_resgid;
3452 old_opts.s_commit_interval = sbi->s_commit_interval;
30773840
TT		 3453 old_opts.s_min_batch_time = sbi->s_min_batch_time;
3454 old_opts.s_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 3455#ifdef CONFIG_QUOTA
3456 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
3457 for (i = 0; i < MAXQUOTAS; i++)
3458 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
3459#endif
b3881f74
TT		 3460 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
3461 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
ac27a0ec
DK		 3462
3463 /*
3464 * Allow the "check" option to be passed as a remount option.
3465 */
b3881f74
TT		 3466 if (!parse_options(data, sb, NULL, &journal_ioprio,
3467 &n_blocks_count, 1)) {
ac27a0ec
DK		 3468 err = -EINVAL;
3469 goto restore_opts;
3470 }
3471
4ab2f15b 3472 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
46e665e9 3473 ext4_abort(sb, __func__, "Abort forced by user");
ac27a0ec
DK		 3474
3475 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
617ba13b 3476 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec
DK		 3477
3478 es = sbi->s_es;
3479
b3881f74 3480 if (sbi->s_journal) {
0390131b 3481 ext4_init_journal_params(sb, sbi->s_journal);
b3881f74
TT		 3482 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3483 }
ac27a0ec
DK		 3484
3485 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
bd81d8ee 3486 n_blocks_count > ext4_blocks_count(es)) {
4ab2f15b 3487 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
ac27a0ec
DK		 3488 err = -EROFS;
3489 goto restore_opts;
3490 }
3491
3492 if (*flags & MS_RDONLY) {
3493 /*
3494 * First of all, the unconditional stuff we have to do
3495 * to disable replay of the journal when we next remount
3496 */
3497 sb->s_flags |= MS_RDONLY;
3498
3499 /*
3500 * OK, test if we are remounting a valid rw partition
3501 * readonly, and if so set the rdonly flag and then
3502 * mark the partition as valid again.
3503 */
617ba13b
MC		 3504 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
3505 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 3506 es->s_state = cpu_to_le16(sbi->s_mount_state);
3507
a63c9eb2 3508 if (sbi->s_journal)
0390131b 3509 ext4_mark_recovery_complete(sb, es);
ac27a0ec 3510 } else {
a13fb1a4
ES		 3511 /* Make sure we can mount this feature set readwrite */
3512 if (!ext4_feature_set_ok(sb, 0)) {
ac27a0ec
DK		 3513 err = -EROFS;
3514 goto restore_opts;
3515 }
8a266467
TT		 3516 /*
3517 * Make sure the group descriptor checksums
0b8e58a1 3518 * are sane. If they aren't, refuse to remount r/w.
8a266467
TT		 3519 */
3520 for (g = 0; g < sbi->s_groups_count; g++) {
3521 struct ext4_group_desc *gdp =
3522 ext4_get_group_desc(sb, g, NULL);
3523
3524 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
b31e1552
ES		 3525 ext4_msg(sb, KERN_ERR,
3526 "ext4_remount: Checksum for group %u failed (%u!=%u)",
8a266467
TT		 3527 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
3528 le16_to_cpu(gdp->bg_checksum));
3529 err = -EINVAL;
3530 goto restore_opts;
3531 }
3532 }
3533
ead6596b
ES		 3534 /*
3535 * If we have an unprocessed orphan list hanging
3536 * around from a previously readonly bdev mount,
3537 * require a full umount/remount for now.
3538 */
3539 if (es->s_last_orphan) {
b31e1552 3540 ext4_msg(sb, KERN_WARNING, "Couldn't "
ead6596b
ES		 3541 "remount RDWR because of unprocessed "
3542 "orphan inode list. Please "
b31e1552 3543 "umount/remount instead");
ead6596b
ES		 3544 err = -EINVAL;
3545 goto restore_opts;
3546 }
3547
ac27a0ec
DK		 3548 /*
3549 * Mounting a RDONLY partition read-write, so reread
3550 * and store the current valid flag. (It may have
3551 * been changed by e2fsck since we originally mounted
3552 * the partition.)
3553 */
0390131b
FM		 3554 if (sbi->s_journal)
3555 ext4_clear_journal_err(sb, es);
ac27a0ec 3556 sbi->s_mount_state = le16_to_cpu(es->s_state);
617ba13b 3557 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
ac27a0ec 3558 goto restore_opts;
2b2d6d01 3559 if (!ext4_setup_super(sb, es, 0))
ac27a0ec
DK		 3560 sb->s_flags &= ~MS_RDONLY;
3561 }
3562 }
6fd058f7 3563 ext4_setup_system_zone(sb);
0390131b 3564 if (sbi->s_journal == NULL)
e2d67052 3565 ext4_commit_super(sb, 1);
0390131b 3566
ac27a0ec
DK		 3567#ifdef CONFIG_QUOTA
3568 /* Release old quota file names */
3569 for (i = 0; i < MAXQUOTAS; i++)
3570 if (old_opts.s_qf_names[i] &&
3571 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3572 kfree(old_opts.s_qf_names[i]);
3573#endif
bbd6851a 3574 unlock_super(sb);
337eb00a 3575 unlock_kernel();
ac27a0ec 3576 return 0;
0b8e58a1 3577
ac27a0ec
DK		 3578restore_opts:
3579 sb->s_flags = old_sb_flags;
3580 sbi->s_mount_opt = old_opts.s_mount_opt;
3581 sbi->s_resuid = old_opts.s_resuid;
3582 sbi->s_resgid = old_opts.s_resgid;
3583 sbi->s_commit_interval = old_opts.s_commit_interval;
30773840
TT		 3584 sbi->s_min_batch_time = old_opts.s_min_batch_time;
3585 sbi->s_max_batch_time = old_opts.s_max_batch_time;
ac27a0ec
DK		 3586#ifdef CONFIG_QUOTA
3587 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
3588 for (i = 0; i < MAXQUOTAS; i++) {
3589 if (sbi->s_qf_names[i] &&
3590 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3591 kfree(sbi->s_qf_names[i]);
3592 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
3593 }
3594#endif
bbd6851a 3595 unlock_super(sb);
337eb00a 3596 unlock_kernel();
ac27a0ec
DK		 3597 return err;
3598}
3599
2b2d6d01 3600static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
ac27a0ec
DK		 3601{
3602 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 3603 struct ext4_sb_info *sbi = EXT4_SB(sb);
3604 struct ext4_super_block *es = sbi->s_es;
960cc398 3605 u64 fsid;
ac27a0ec 3606
5e70030d
BP		 3607 if (test_opt(sb, MINIX_DF)) {
3608 sbi->s_overhead_last = 0;
6bc9feff 3609 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
8df9675f 3610 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
5e70030d 3611 ext4_fsblk_t overhead = 0;
ac27a0ec
DK		 3612
3613 /*
5e70030d
BP		 3614 * Compute the overhead (FS structures). This is constant
3615 * for a given filesystem unless the number of block groups
3616 * changes so we cache the previous value until it does.
ac27a0ec
DK		 3617 */
3618
3619 /*
3620 * All of the blocks before first_data_block are
3621 * overhead
3622 */
3623 overhead = le32_to_cpu(es->s_first_data_block);
3624
3625 /*
3626 * Add the overhead attributed to the superblock and
3627 * block group descriptors. If the sparse superblocks
3628 * feature is turned on, then not all groups have this.
3629 */
3630 for (i = 0; i < ngroups; i++) {
617ba13b
MC		 3631 overhead += ext4_bg_has_super(sb, i) +
3632 ext4_bg_num_gdb(sb, i);
ac27a0ec
DK		 3633 cond_resched();
3634 }
3635
3636 /*
3637 * Every block group has an inode bitmap, a block
3638 * bitmap, and an inode table.
3639 */
5e70030d
BP		 3640 overhead += ngroups * (2 + sbi->s_itb_per_group);
3641 sbi->s_overhead_last = overhead;
3642 smp_wmb();
6bc9feff 3643 sbi->s_blocks_last = ext4_blocks_count(es);
ac27a0ec
DK		 3644 }
3645
617ba13b 3646 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 3647 buf->f_bsize = sb->s_blocksize;
5e70030d 3648 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
6bc6e63f
AK		 3649 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
3650 percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
308ba3ec 3651 ext4_free_blocks_count_set(es, buf->f_bfree);
bd81d8ee
LV		 3652 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3653 if (buf->f_bfree < ext4_r_blocks_count(es))
ac27a0ec
DK		 3654 buf->f_bavail = 0;
3655 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 3656 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
5e70030d 3657 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
617ba13b 3658 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 3659 fsid = le64_to_cpup((void *)es->s_uuid) ^
3660 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3661 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3662 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
0b8e58a1 3663
ac27a0ec
DK		 3664 return 0;
3665}
3666
0b8e58a1
AD		 3667/* Helper function for writing quotas on sync - we need to start transaction
3668 * before quota file is locked for write. Otherwise the are possible deadlocks:
ac27a0ec 3669 * Process 1 Process 2
617ba13b 3670 * ext4_create() quota_sync()
a269eb18
JK		 3671 * jbd2_journal_start() write_dquot()
3672 * vfs_dq_init() down(dqio_mutex)
dab291af 3673 * down(dqio_mutex) jbd2_journal_start()
ac27a0ec
DK		 3674 *
3675 */
3676
3677#ifdef CONFIG_QUOTA
3678
3679static inline struct inode *dquot_to_inode(struct dquot *dquot)
3680{
3681 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3682}
3683
617ba13b 3684static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 3685{
3686 int ret, err;
3687 handle_t *handle;
3688 struct inode *inode;
3689
3690 inode = dquot_to_inode(dquot);
617ba13b 3691 handle = ext4_journal_start(inode,
0b8e58a1 3692 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 3693 if (IS_ERR(handle))
3694 return PTR_ERR(handle);
3695 ret = dquot_commit(dquot);
617ba13b 3696 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3697 if (!ret)
3698 ret = err;
3699 return ret;
3700}
3701
617ba13b 3702static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 3703{
3704 int ret, err;
3705 handle_t *handle;
3706
617ba13b 3707 handle = ext4_journal_start(dquot_to_inode(dquot),
0b8e58a1 3708 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 3709 if (IS_ERR(handle))
3710 return PTR_ERR(handle);
3711 ret = dquot_acquire(dquot);
617ba13b 3712 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3713 if (!ret)
3714 ret = err;
3715 return ret;
3716}
3717
617ba13b 3718static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 3719{
3720 int ret, err;
3721 handle_t *handle;
3722
617ba13b 3723 handle = ext4_journal_start(dquot_to_inode(dquot),
0b8e58a1 3724 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 3725 if (IS_ERR(handle)) {
3726 /* Release dquot anyway to avoid endless cycle in dqput() */
3727 dquot_release(dquot);
ac27a0ec 3728 return PTR_ERR(handle);
9c3013e9 3729 }
ac27a0ec 3730 ret = dquot_release(dquot);
617ba13b 3731 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3732 if (!ret)
3733 ret = err;
3734 return ret;
3735}
3736
617ba13b 3737static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec 3738{
2c8be6b2 3739 /* Are we journaling quotas? */
617ba13b
MC		 3740 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3741 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
ac27a0ec 3742 dquot_mark_dquot_dirty(dquot);
617ba13b 3743 return ext4_write_dquot(dquot);
ac27a0ec
DK		 3744 } else {
3745 return dquot_mark_dquot_dirty(dquot);
3746 }
3747}
3748
617ba13b 3749static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 3750{
3751 int ret, err;
3752 handle_t *handle;
3753
3754 /* Data block + inode block */
617ba13b 3755 handle = ext4_journal_start(sb->s_root->d_inode, 2);
ac27a0ec
DK		 3756 if (IS_ERR(handle))
3757 return PTR_ERR(handle);
3758 ret = dquot_commit_info(sb, type);
617ba13b 3759 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3760 if (!ret)
3761 ret = err;
3762 return ret;
3763}
3764
3765/*
3766 * Turn on quotas during mount time - we need to find
3767 * the quota file and such...
3768 */
617ba13b 3769static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 3770{
617ba13b 3771 return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
0b8e58a1 3772 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 3773}
3774
3775/*
3776 * Standard function to be called on quota_on
3777 */
617ba13b 3778static int ext4_quota_on(struct super_block *sb, int type, int format_id,
8264613d 3779 char *name, int remount)
ac27a0ec
DK		 3780{
3781 int err;
8264613d 3782 struct path path;
ac27a0ec
DK		 3783
3784 if (!test_opt(sb, QUOTA))
3785 return -EINVAL;
8264613d 3786 /* When remounting, no checks are needed and in fact, name is NULL */
0623543b 3787 if (remount)
8264613d 3788 return vfs_quota_on(sb, type, format_id, name, remount);
0623543b 3789
8264613d 3790 err = kern_path(name, LOOKUP_FOLLOW, &path);
ac27a0ec
DK		 3791 if (err)
3792 return err;
0623543b 3793
ac27a0ec 3794 /* Quotafile not on the same filesystem? */
8264613d
AV		 3795 if (path.mnt->mnt_sb != sb) {
3796 path_put(&path);
ac27a0ec
DK		 3797 return -EXDEV;
3798 }
0623543b
JK		 3799 /* Journaling quota? */
3800 if (EXT4_SB(sb)->s_qf_names[type]) {
2b2d6d01 3801 /* Quotafile not in fs root? */
8264613d 3802 if (path.dentry->d_parent != sb->s_root)
b31e1552
ES		 3803 ext4_msg(sb, KERN_WARNING,
3804 "Quota file not on filesystem root. "
3805 "Journaled quota will not work");
2b2d6d01 3806 }
0623543b
JK		 3807
3808 /*
3809 * When we journal data on quota file, we have to flush journal to see
3810 * all updates to the file when we bypass pagecache...
3811 */
0390131b
FM		 3812 if (EXT4_SB(sb)->s_journal &&
3813 ext4_should_journal_data(path.dentry->d_inode)) {
0623543b
JK		 3814 /*
3815 * We don't need to lock updates but journal_flush() could
3816 * otherwise be livelocked...
3817 */
3818 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 3819 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
0623543b 3820 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 3821 if (err) {
8264613d 3822 path_put(&path);
7ffe1ea8
HK		 3823 return err;
3824 }
0623543b
JK		 3825 }
3826
8264613d
AV		 3827 err = vfs_quota_on_path(sb, type, format_id, &path);
3828 path_put(&path);
77e69dac 3829 return err;
ac27a0ec
DK		 3830}
3831
3832/* Read data from quotafile - avoid pagecache and such because we cannot afford
3833 * acquiring the locks... As quota files are never truncated and quota code
3834 * itself serializes the operations (and noone else should touch the files)
3835 * we don't have to be afraid of races */
617ba13b 3836static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 3837 size_t len, loff_t off)
3838{
3839 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 3840 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 3841 int err = 0;
3842 int offset = off & (sb->s_blocksize - 1);
3843 int tocopy;
3844 size_t toread;
3845 struct buffer_head *bh;
3846 loff_t i_size = i_size_read(inode);
3847
3848 if (off > i_size)
3849 return 0;
3850 if (off+len > i_size)
3851 len = i_size-off;
3852 toread = len;
3853 while (toread > 0) {
3854 tocopy = sb->s_blocksize - offset < toread ?
3855 sb->s_blocksize - offset : toread;
617ba13b 3856 bh = ext4_bread(NULL, inode, blk, 0, &err);
ac27a0ec
DK		 3857 if (err)
3858 return err;
3859 if (!bh) /* A hole? */
3860 memset(data, 0, tocopy);
3861 else
3862 memcpy(data, bh->b_data+offset, tocopy);
3863 brelse(bh);
3864 offset = 0;
3865 toread -= tocopy;
3866 data += tocopy;
3867 blk++;
3868 }
3869 return len;
3870}
3871
3872/* Write to quotafile (we know the transaction is already started and has
3873 * enough credits) */
617ba13b 3874static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 3875 const char *data, size_t len, loff_t off)
3876{
3877 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 3878 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 3879 int err = 0;
3880 int offset = off & (sb->s_blocksize - 1);
3881 int tocopy;
617ba13b 3882 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
ac27a0ec
DK		 3883 size_t towrite = len;
3884 struct buffer_head *bh;
3885 handle_t *handle = journal_current_handle();
3886
0390131b 3887 if (EXT4_SB(sb)->s_journal && !handle) {
b31e1552
ES		 3888 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
3889 " cancelled because transaction is not started",
9c3013e9
JK		 3890 (unsigned long long)off, (unsigned long long)len);
3891 return -EIO;
3892 }
ac27a0ec
DK		 3893 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
3894 while (towrite > 0) {
3895 tocopy = sb->s_blocksize - offset < towrite ?
3896 sb->s_blocksize - offset : towrite;
617ba13b 3897 bh = ext4_bread(handle, inode, blk, 1, &err);
ac27a0ec
DK		 3898 if (!bh)
3899 goto out;
3900 if (journal_quota) {
617ba13b 3901 err = ext4_journal_get_write_access(handle, bh);
ac27a0ec
DK		 3902 if (err) {
3903 brelse(bh);
3904 goto out;
3905 }
3906 }
3907 lock_buffer(bh);
3908 memcpy(bh->b_data+offset, data, tocopy);
3909 flush_dcache_page(bh->b_page);
3910 unlock_buffer(bh);
3911 if (journal_quota)
0390131b 3912 err = ext4_handle_dirty_metadata(handle, NULL, bh);
ac27a0ec
DK		 3913 else {
3914 /* Always do at least ordered writes for quotas */
678aaf48 3915 err = ext4_jbd2_file_inode(handle, inode);
ac27a0ec
DK		 3916 mark_buffer_dirty(bh);
3917 }
3918 brelse(bh);
3919 if (err)
3920 goto out;
3921 offset = 0;
3922 towrite -= tocopy;
3923 data += tocopy;
3924 blk++;
3925 }
3926out:
4d04e4fb
JK		 3927 if (len == towrite) {
3928 mutex_unlock(&inode->i_mutex);
ac27a0ec 3929 return err;
4d04e4fb 3930 }
ac27a0ec
DK		 3931 if (inode->i_size < off+len-towrite) {
3932 i_size_write(inode, off+len-towrite);
617ba13b 3933 EXT4_I(inode)->i_disksize = inode->i_size;
ac27a0ec 3934 }
ac27a0ec 3935 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
617ba13b 3936 ext4_mark_inode_dirty(handle, inode);
ac27a0ec
DK		 3937 mutex_unlock(&inode->i_mutex);
3938 return len - towrite;
3939}
3940
3941#endif
3942
0b8e58a1
AD		 3943static int ext4_get_sb(struct file_system_type *fs_type, int flags,
3944 const char *dev_name, void *data, struct vfsmount *mnt)
ac27a0ec 3945{
0b8e58a1 3946 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super,mnt);
ac27a0ec
DK		 3947}
3948
03010a33
TT		 3949static struct file_system_type ext4_fs_type = {
3950 .owner = THIS_MODULE,
3951 .name = "ext4",
3952 .get_sb = ext4_get_sb,
3953 .kill_sb = kill_block_super,
3954 .fs_flags = FS_REQUIRES_DEV,
3955};
3956
617ba13b 3957static int __init init_ext4_fs(void)
ac27a0ec 3958{
c9de560d
AT		 3959 int err;
3960
6fd058f7
TT		 3961 err = init_ext4_system_zone();
3962 if (err)
3963 return err;
3197ebdb
TT		 3964 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
3965 if (!ext4_kset)
6fd058f7 3966 goto out4;
9f6200bb 3967 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
c9de560d 3968 err = init_ext4_mballoc();
ac27a0ec 3969 if (err)
6fd058f7 3970 goto out3;
c9de560d
AT		 3971
3972 err = init_ext4_xattr();
3973 if (err)
3974 goto out2;
ac27a0ec
DK		 3975 err = init_inodecache();
3976 if (err)
3977 goto out1;
03010a33 3978 err = register_filesystem(&ext4_fs_type);
ac27a0ec
DK		 3979 if (err)
3980 goto out;
3981 return 0;
3982out:
3983 destroy_inodecache();
3984out1:
617ba13b 3985 exit_ext4_xattr();
c9de560d
AT		 3986out2:
3987 exit_ext4_mballoc();
6fd058f7
TT		 3988out3:
3989 remove_proc_entry("fs/ext4", NULL);
3990 kset_unregister(ext4_kset);
3991out4:
3992 exit_ext4_system_zone();
ac27a0ec
DK		 3993 return err;
3994}
3995
617ba13b 3996static void __exit exit_ext4_fs(void)
ac27a0ec 3997{
03010a33 3998 unregister_filesystem(&ext4_fs_type);
ac27a0ec 3999 destroy_inodecache();
617ba13b 4000 exit_ext4_xattr();
c9de560d 4001 exit_ext4_mballoc();
9f6200bb 4002 remove_proc_entry("fs/ext4", NULL);
3197ebdb 4003 kset_unregister(ext4_kset);
6fd058f7 4004 exit_ext4_system_zone();
ac27a0ec
DK		 4005}
4006
4007MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
83982b6f 4008MODULE_DESCRIPTION("Fourth Extended Filesystem");
ac27a0ec 4009MODULE_LICENSE("GPL");
617ba13b
MC		 4010module_init(init_ext4_fs)
4011module_exit(exit_ext4_fs)
Clone of davem's net-next-2.6 tree