projects / GitHub / mt8127 / android_kernel_alcatel_ttab.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
ext4: Use end_io callback to avoid direct I/O fallback to buffered I/O
[GitHub/mt8127/android_kernel_alcatel_ttab.git] / fs / ext4 / super.c
CommitLineData
ac27a0ec 1/*
617ba13b 2 * linux/fs/ext4/super.c
ac27a0ec
DK		 3 *
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
8 *
9 * from
10 *
11 * linux/fs/minix/inode.c
12 *
13 * Copyright (C) 1991, 1992 Linus Torvalds
14 *
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
17 */
18
19#include <linux/module.h>
20#include <linux/string.h>
21#include <linux/fs.h>
22#include <linux/time.h>
c5ca7c76 23#include <linux/vmalloc.h>
dab291af 24#include <linux/jbd2.h>
ac27a0ec
DK		 25#include <linux/slab.h>
26#include <linux/init.h>
27#include <linux/blkdev.h>
28#include <linux/parser.h>
29#include <linux/smp_lock.h>
30#include <linux/buffer_head.h>
a5694255 31#include <linux/exportfs.h>
ac27a0ec
DK		 32#include <linux/vfs.h>
33#include <linux/random.h>
34#include <linux/mount.h>
35#include <linux/namei.h>
36#include <linux/quotaops.h>
37#include <linux/seq_file.h>
9f6200bb 38#include <linux/proc_fs.h>
3197ebdb 39#include <linux/ctype.h>
1330593e 40#include <linux/log2.h>
717d50e4 41#include <linux/crc16.h>
ac27a0ec
DK		 42#include <asm/uaccess.h>
43
3dcf5451
CH		 44#include "ext4.h"
45#include "ext4_jbd2.h"
ac27a0ec
DK		 46#include "xattr.h"
47#include "acl.h"
3661d286 48#include "mballoc.h"
ac27a0ec 49
9bffad1e
TT		 50#define CREATE_TRACE_POINTS
51#include <trace/events/ext4.h>
52
f4033903
CW		 53static int default_mb_history_length = 1000;
54
55module_param_named(default_mb_history_length, default_mb_history_length,
56 int, 0644);
57MODULE_PARM_DESC(default_mb_history_length,
58 "Default number of entries saved for mb_history");
59
9f6200bb 60struct proc_dir_entry *ext4_proc_root;
3197ebdb 61static struct kset *ext4_kset;
9f6200bb 62
617ba13b 63static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
ac27a0ec 64 unsigned long journal_devnum);
e2d67052 65static int ext4_commit_super(struct super_block *sb, int sync);
2b2d6d01
TT		 66static void ext4_mark_recovery_complete(struct super_block *sb,
67 struct ext4_super_block *es);
68static void ext4_clear_journal_err(struct super_block *sb,
69 struct ext4_super_block *es);
617ba13b 70static int ext4_sync_fs(struct super_block *sb, int wait);
2b2d6d01 71static const char *ext4_decode_error(struct super_block *sb, int errno,
ac27a0ec 72 char nbuf[16]);
2b2d6d01
TT		 73static int ext4_remount(struct super_block *sb, int *flags, char *data);
74static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
c4be0c1d 75static int ext4_unfreeze(struct super_block *sb);
2b2d6d01 76static void ext4_write_super(struct super_block *sb);
c4be0c1d 77static int ext4_freeze(struct super_block *sb);
ac27a0ec 78
bd81d8ee 79
8fadc143
AR		 80ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
81 struct ext4_group_desc *bg)
bd81d8ee 82{
3a14589c 83 return le32_to_cpu(bg->bg_block_bitmap_lo) |
8fadc143 84 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 85 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 86}
87
8fadc143
AR		 88ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
89 struct ext4_group_desc *bg)
bd81d8ee 90{
5272f837 91 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
8fadc143 92 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 93 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 94}
95
8fadc143
AR		 96ext4_fsblk_t ext4_inode_table(struct super_block *sb,
97 struct ext4_group_desc *bg)
bd81d8ee 98{
5272f837 99 return le32_to_cpu(bg->bg_inode_table_lo) |
8fadc143 100 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 101 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
bd81d8ee
LV		 102}
103
560671a0
AK		 104__u32 ext4_free_blks_count(struct super_block *sb,
105 struct ext4_group_desc *bg)
106{
107 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
108 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 109 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
560671a0
AK		 110}
111
112__u32 ext4_free_inodes_count(struct super_block *sb,
113 struct ext4_group_desc *bg)
114{
115 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
116 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 117 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
560671a0
AK		 118}
119
120__u32 ext4_used_dirs_count(struct super_block *sb,
121 struct ext4_group_desc *bg)
122{
123 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
124 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 125 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
560671a0
AK		 126}
127
128__u32 ext4_itable_unused_count(struct super_block *sb,
129 struct ext4_group_desc *bg)
130{
131 return le16_to_cpu(bg->bg_itable_unused_lo) |
132 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 133 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
560671a0
AK		 134}
135
8fadc143
AR		 136void ext4_block_bitmap_set(struct super_block *sb,
137 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 138{
3a14589c 139 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 140 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
141 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 142}
143
8fadc143
AR		 144void ext4_inode_bitmap_set(struct super_block *sb,
145 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 146{
5272f837 147 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 148 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
149 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 150}
151
8fadc143
AR		 152void ext4_inode_table_set(struct super_block *sb,
153 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 154{
5272f837 155 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 156 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
157 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 158}
159
560671a0
AK		 160void ext4_free_blks_set(struct super_block *sb,
161 struct ext4_group_desc *bg, __u32 count)
162{
163 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
164 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
165 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
166}
167
168void ext4_free_inodes_set(struct super_block *sb,
169 struct ext4_group_desc *bg, __u32 count)
170{
171 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
172 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
173 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
174}
175
176void ext4_used_dirs_set(struct super_block *sb,
177 struct ext4_group_desc *bg, __u32 count)
178{
179 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
180 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
181 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
182}
183
184void ext4_itable_unused_set(struct super_block *sb,
185 struct ext4_group_desc *bg, __u32 count)
186{
187 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
188 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
189 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
190}
191
ac27a0ec 192/*
dab291af 193 * Wrappers for jbd2_journal_start/end.
ac27a0ec
DK		 194 *
195 * The only special thing we need to do here is to make sure that all
196 * journal_end calls result in the superblock being marked dirty, so
197 * that sync() will call the filesystem's write_super callback if
198 * appropriate.
199 */
617ba13b 200handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
ac27a0ec
DK		 201{
202 journal_t *journal;
203
204 if (sb->s_flags & MS_RDONLY)
205 return ERR_PTR(-EROFS);
206
207 /* Special case here: if the journal has aborted behind our
208 * backs (eg. EIO in the commit thread), then we still need to
209 * take the FS itself readonly cleanly. */
617ba13b 210 journal = EXT4_SB(sb)->s_journal;
0390131b
FM		 211 if (journal) {
212 if (is_journal_aborted(journal)) {
0b8e58a1 213 ext4_abort(sb, __func__, "Detected aborted journal");
0390131b
FM		 214 return ERR_PTR(-EROFS);
215 }
216 return jbd2_journal_start(journal, nblocks);
ac27a0ec 217 }
0390131b
FM		 218 /*
219 * We're not journaling, return the appropriate indication.
220 */
221 current->journal_info = EXT4_NOJOURNAL_HANDLE;
222 return current->journal_info;
ac27a0ec
DK		 223}
224
225/*
226 * The only special thing we need to do here is to make sure that all
dab291af 227 * jbd2_journal_stop calls result in the superblock being marked dirty, so
ac27a0ec
DK		 228 * that sync() will call the filesystem's write_super callback if
229 * appropriate.
230 */
617ba13b 231int __ext4_journal_stop(const char *where, handle_t *handle)
ac27a0ec
DK		 232{
233 struct super_block *sb;
234 int err;
235 int rc;
236
0390131b
FM		 237 if (!ext4_handle_valid(handle)) {
238 /*
239 * Do this here since we don't call jbd2_journal_stop() in
240 * no-journal mode.
241 */
242 current->journal_info = NULL;
243 return 0;
244 }
ac27a0ec
DK		 245 sb = handle->h_transaction->t_journal->j_private;
246 err = handle->h_err;
dab291af 247 rc = jbd2_journal_stop(handle);
ac27a0ec
DK		 248
249 if (!err)
250 err = rc;
251 if (err)
617ba13b 252 __ext4_std_error(sb, where, err);
ac27a0ec
DK		 253 return err;
254}
255
617ba13b 256void ext4_journal_abort_handle(const char *caller, const char *err_fn,
ac27a0ec
DK		 257 struct buffer_head *bh, handle_t *handle, int err)
258{
259 char nbuf[16];
617ba13b 260 const char *errstr = ext4_decode_error(NULL, err, nbuf);
ac27a0ec 261
0390131b
FM		 262 BUG_ON(!ext4_handle_valid(handle));
263
ac27a0ec
DK		 264 if (bh)
265 BUFFER_TRACE(bh, "abort");
266
267 if (!handle->h_err)
268 handle->h_err = err;
269
270 if (is_handle_aborted(handle))
271 return;
272
273 printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
274 caller, errstr, err_fn);
275
dab291af 276 jbd2_journal_abort_handle(handle);
ac27a0ec
DK		 277}
278
279/* Deal with the reporting of failure conditions on a filesystem such as
280 * inconsistencies detected or read IO failures.
281 *
282 * On ext2, we can store the error state of the filesystem in the
617ba13b 283 * superblock. That is not possible on ext4, because we may have other
ac27a0ec
DK		 284 * write ordering constraints on the superblock which prevent us from
285 * writing it out straight away; and given that the journal is about to
286 * be aborted, we can't rely on the current, or future, transactions to
287 * write out the superblock safely.
288 *
dab291af 289 * We'll just use the jbd2_journal_abort() error code to record an error in
ac27a0ec
DK		 290 * the journal instead. On recovery, the journal will compain about
291 * that error until we've noted it down and cleared it.
292 */
293
617ba13b 294static void ext4_handle_error(struct super_block *sb)
ac27a0ec 295{
617ba13b 296 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 297
617ba13b
MC		 298 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
299 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
ac27a0ec
DK		 300
301 if (sb->s_flags & MS_RDONLY)
302 return;
303
2b2d6d01 304 if (!test_opt(sb, ERRORS_CONT)) {
617ba13b 305 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 306
4ab2f15b 307 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
ac27a0ec 308 if (journal)
dab291af 309 jbd2_journal_abort(journal, -EIO);
ac27a0ec 310 }
2b2d6d01 311 if (test_opt(sb, ERRORS_RO)) {
b31e1552 312 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
ac27a0ec
DK		 313 sb->s_flags |= MS_RDONLY;
314 }
e2d67052 315 ext4_commit_super(sb, 1);
ac27a0ec 316 if (test_opt(sb, ERRORS_PANIC))
617ba13b 317 panic("EXT4-fs (device %s): panic forced after error\n",
ac27a0ec
DK		 318 sb->s_id);
319}
320
2b2d6d01
TT		 321void ext4_error(struct super_block *sb, const char *function,
322 const char *fmt, ...)
ac27a0ec
DK		 323{
324 va_list args;
325
326 va_start(args, fmt);
2b2d6d01 327 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
ac27a0ec
DK		 328 vprintk(fmt, args);
329 printk("\n");
330 va_end(args);
331
617ba13b 332 ext4_handle_error(sb);
ac27a0ec
DK		 333}
334
2b2d6d01 335static const char *ext4_decode_error(struct super_block *sb, int errno,
ac27a0ec
DK		 336 char nbuf[16])
337{
338 char *errstr = NULL;
339
340 switch (errno) {
341 case -EIO:
342 errstr = "IO failure";
343 break;
344 case -ENOMEM:
345 errstr = "Out of memory";
346 break;
347 case -EROFS:
78f1ddbb
TT		 348 if (!sb || (EXT4_SB(sb)->s_journal &&
349 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
ac27a0ec
DK		 350 errstr = "Journal has aborted";
351 else
352 errstr = "Readonly filesystem";
353 break;
354 default:
355 /* If the caller passed in an extra buffer for unknown
356 * errors, textualise them now. Else we just return
357 * NULL. */
358 if (nbuf) {
359 /* Check for truncated error codes... */
360 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
361 errstr = nbuf;
362 }
363 break;
364 }
365
366 return errstr;
367}
368
617ba13b 369/* __ext4_std_error decodes expected errors from journaling functions
ac27a0ec
DK		 370 * automatically and invokes the appropriate error response. */
371
2b2d6d01 372void __ext4_std_error(struct super_block *sb, const char *function, int errno)
ac27a0ec
DK		 373{
374 char nbuf[16];
375 const char *errstr;
376
377 /* Special case: if the error is EROFS, and we're not already
378 * inside a transaction, then there's really no point in logging
379 * an error. */
380 if (errno == -EROFS && journal_current_handle() == NULL &&
381 (sb->s_flags & MS_RDONLY))
382 return;
383
617ba13b 384 errstr = ext4_decode_error(sb, errno, nbuf);
2b2d6d01
TT		 385 printk(KERN_CRIT "EXT4-fs error (device %s) in %s: %s\n",
386 sb->s_id, function, errstr);
ac27a0ec 387
617ba13b 388 ext4_handle_error(sb);
ac27a0ec
DK		 389}
390
391/*
617ba13b 392 * ext4_abort is a much stronger failure handler than ext4_error. The
ac27a0ec
DK		 393 * abort function may be used to deal with unrecoverable failures such
394 * as journal IO errors or ENOMEM at a critical moment in log management.
395 *
396 * We unconditionally force the filesystem into an ABORT|READONLY state,
397 * unless the error response on the fs has been set to panic in which
398 * case we take the easy way out and panic immediately.
399 */
400
2b2d6d01
TT		 401void ext4_abort(struct super_block *sb, const char *function,
402 const char *fmt, ...)
ac27a0ec
DK		 403{
404 va_list args;
405
ac27a0ec 406 va_start(args, fmt);
2b2d6d01 407 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
ac27a0ec
DK		 408 vprintk(fmt, args);
409 printk("\n");
410 va_end(args);
411
412 if (test_opt(sb, ERRORS_PANIC))
617ba13b 413 panic("EXT4-fs panic from previous error\n");
ac27a0ec
DK		 414
415 if (sb->s_flags & MS_RDONLY)
416 return;
417
b31e1552 418 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
617ba13b 419 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
ac27a0ec 420 sb->s_flags |= MS_RDONLY;
4ab2f15b 421 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
ef2cabf7
HK		 422 if (EXT4_SB(sb)->s_journal)
423 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
ac27a0ec
DK		 424}
425
b31e1552
ES		 426void ext4_msg (struct super_block * sb, const char *prefix,
427 const char *fmt, ...)
428{
429 va_list args;
430
431 va_start(args, fmt);
432 printk("%sEXT4-fs (%s): ", prefix, sb->s_id);
433 vprintk(fmt, args);
434 printk("\n");
435 va_end(args);
436}
437
2b2d6d01
TT		 438void ext4_warning(struct super_block *sb, const char *function,
439 const char *fmt, ...)
ac27a0ec
DK		 440{
441 va_list args;
442
443 va_start(args, fmt);
617ba13b 444 printk(KERN_WARNING "EXT4-fs warning (device %s): %s: ",
ac27a0ec
DK		 445 sb->s_id, function);
446 vprintk(fmt, args);
447 printk("\n");
448 va_end(args);
449}
450
5d1b1b3f 451void ext4_grp_locked_error(struct super_block *sb, ext4_group_t grp,
0b8e58a1 452 const char *function, const char *fmt, ...)
5d1b1b3f
AK		 453__releases(bitlock)
454__acquires(bitlock)
455{
456 va_list args;
457 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
458
459 va_start(args, fmt);
460 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
461 vprintk(fmt, args);
462 printk("\n");
463 va_end(args);
464
465 if (test_opt(sb, ERRORS_CONT)) {
466 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
467 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
e2d67052 468 ext4_commit_super(sb, 0);
5d1b1b3f
AK		 469 return;
470 }
471 ext4_unlock_group(sb, grp);
472 ext4_handle_error(sb);
473 /*
474 * We only get here in the ERRORS_RO case; relocking the group
475 * may be dangerous, but nothing bad will happen since the
476 * filesystem will have already been marked read/only and the
477 * journal has been aborted. We return 1 as a hint to callers
478 * who might what to use the return value from
479 * ext4_grp_locked_error() to distinguish beween the
480 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
481 * aggressively from the ext4 function in question, with a
482 * more appropriate error code.
483 */
484 ext4_lock_group(sb, grp);
485 return;
486}
487
617ba13b 488void ext4_update_dynamic_rev(struct super_block *sb)
ac27a0ec 489{
617ba13b 490 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 491
617ba13b 492 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
ac27a0ec
DK		 493 return;
494
46e665e9 495 ext4_warning(sb, __func__,
ac27a0ec
DK		 496 "updating to rev %d because of new feature flag, "
497 "running e2fsck is recommended",
617ba13b 498 EXT4_DYNAMIC_REV);
ac27a0ec 499
617ba13b
MC		 500 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
501 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
502 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
ac27a0ec
DK		 503 /* leave es->s_feature_*compat flags alone */
504 /* es->s_uuid will be set by e2fsck if empty */
505
506 /*
507 * The rest of the superblock fields should be zero, and if not it
508 * means they are likely already in use, so leave them alone. We
509 * can leave it up to e2fsck to clean up any inconsistencies there.
510 */
511}
512
513/*
514 * Open the external journal device
515 */
b31e1552 516static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
ac27a0ec
DK		 517{
518 struct block_device *bdev;
519 char b[BDEVNAME_SIZE];
520
521 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
522 if (IS_ERR(bdev))
523 goto fail;
524 return bdev;
525
526fail:
b31e1552 527 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
ac27a0ec
DK		 528 __bdevname(dev, b), PTR_ERR(bdev));
529 return NULL;
530}
531
532/*
533 * Release the journal device
534 */
617ba13b 535static int ext4_blkdev_put(struct block_device *bdev)
ac27a0ec
DK		 536{
537 bd_release(bdev);
9a1c3542 538 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
ac27a0ec
DK		 539}
540
617ba13b 541static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
ac27a0ec
DK		 542{
543 struct block_device *bdev;
544 int ret = -ENODEV;
545
546 bdev = sbi->journal_bdev;
547 if (bdev) {
617ba13b 548 ret = ext4_blkdev_put(bdev);
ac27a0ec
DK		 549 sbi->journal_bdev = NULL;
550 }
551 return ret;
552}
553
554static inline struct inode *orphan_list_entry(struct list_head *l)
555{
617ba13b 556 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
ac27a0ec
DK		 557}
558
617ba13b 559static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
ac27a0ec
DK		 560{
561 struct list_head *l;
562
b31e1552
ES		 563 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
564 le32_to_cpu(sbi->s_es->s_last_orphan));
ac27a0ec
DK		 565
566 printk(KERN_ERR "sb_info orphan list:\n");
567 list_for_each(l, &sbi->s_orphan) {
568 struct inode *inode = orphan_list_entry(l);
569 printk(KERN_ERR " "
570 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
571 inode->i_sb->s_id, inode->i_ino, inode,
572 inode->i_mode, inode->i_nlink,
573 NEXT_ORPHAN(inode));
574 }
575}
576
2b2d6d01 577static void ext4_put_super(struct super_block *sb)
ac27a0ec 578{
617ba13b
MC		 579 struct ext4_sb_info *sbi = EXT4_SB(sb);
580 struct ext4_super_block *es = sbi->s_es;
ef2cabf7 581 int i, err;
ac27a0ec 582
4c0425ff
MC		 583 flush_workqueue(sbi->dio_unwritten_wq);
584 destroy_workqueue(sbi->dio_unwritten_wq);
585
a9e220f8 586 lock_super(sb);
6cfd0148 587 lock_kernel();
8c85e125 588 if (sb->s_dirt)
ebc1ac16 589 ext4_commit_super(sb, 1);
8c85e125 590
6fd058f7 591 ext4_release_system_zone(sb);
c9de560d 592 ext4_mb_release(sb);
a86c6181 593 ext4_ext_release(sb);
617ba13b 594 ext4_xattr_put_super(sb);
0390131b
FM		 595 if (sbi->s_journal) {
596 err = jbd2_journal_destroy(sbi->s_journal);
597 sbi->s_journal = NULL;
598 if (err < 0)
599 ext4_abort(sb, __func__,
600 "Couldn't clean up the journal");
601 }
ac27a0ec 602 if (!(sb->s_flags & MS_RDONLY)) {
617ba13b 603 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 604 es->s_state = cpu_to_le16(sbi->s_mount_state);
e2d67052 605 ext4_commit_super(sb, 1);
ac27a0ec 606 }
240799cd 607 if (sbi->s_proc) {
9f6200bb 608 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 609 }
3197ebdb 610 kobject_del(&sbi->s_kobj);
ac27a0ec
DK		 611
612 for (i = 0; i < sbi->s_gdb_count; i++)
613 brelse(sbi->s_group_desc[i]);
614 kfree(sbi->s_group_desc);
c5ca7c76
TT		 615 if (is_vmalloc_addr(sbi->s_flex_groups))
616 vfree(sbi->s_flex_groups);
617 else
618 kfree(sbi->s_flex_groups);
ac27a0ec
DK		 619 percpu_counter_destroy(&sbi->s_freeblocks_counter);
620 percpu_counter_destroy(&sbi->s_freeinodes_counter);
621 percpu_counter_destroy(&sbi->s_dirs_counter);
6bc6e63f 622 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
ac27a0ec
DK		 623 brelse(sbi->s_sbh);
624#ifdef CONFIG_QUOTA
625 for (i = 0; i < MAXQUOTAS; i++)
626 kfree(sbi->s_qf_names[i]);
627#endif
628
629 /* Debugging code just in case the in-memory inode orphan list
630 * isn't empty. The on-disk one can be non-empty if we've
631 * detected an error and taken the fs readonly, but the
632 * in-memory list had better be clean by this point. */
633 if (!list_empty(&sbi->s_orphan))
634 dump_orphan_list(sb, sbi);
635 J_ASSERT(list_empty(&sbi->s_orphan));
636
f98393a6 637 invalidate_bdev(sb->s_bdev);
ac27a0ec
DK		 638 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
639 /*
640 * Invalidate the journal device's buffers. We don't want them
641 * floating about in memory - the physical journal device may
642 * hotswapped, and it breaks the `ro-after' testing code.
643 */
644 sync_blockdev(sbi->journal_bdev);
f98393a6 645 invalidate_bdev(sbi->journal_bdev);
617ba13b 646 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 647 }
648 sb->s_fs_info = NULL;
3197ebdb
TT		 649 /*
650 * Now that we are completely done shutting down the
651 * superblock, we need to actually destroy the kobject.
652 */
653 unlock_kernel();
654 unlock_super(sb);
655 kobject_put(&sbi->s_kobj);
656 wait_for_completion(&sbi->s_kobj_unregister);
705895b6 657 kfree(sbi->s_blockgroup_lock);
ac27a0ec 658 kfree(sbi);
ac27a0ec
DK		 659}
660
e18b890b 661static struct kmem_cache *ext4_inode_cachep;
ac27a0ec
DK		 662
663/*
664 * Called inside transaction, so use GFP_NOFS
665 */
617ba13b 666static struct inode *ext4_alloc_inode(struct super_block *sb)
ac27a0ec 667{
617ba13b 668 struct ext4_inode_info *ei;
ac27a0ec 669
e6b4f8da 670 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
ac27a0ec
DK		 671 if (!ei)
672 return NULL;
0b8e58a1 673
ac27a0ec 674 ei->vfs_inode.i_version = 1;
91246c00 675 ei->vfs_inode.i_data.writeback_index = 0;
a86c6181 676 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
c9de560d
AT		 677 INIT_LIST_HEAD(&ei->i_prealloc_list);
678 spin_lock_init(&ei->i_prealloc_lock);
0390131b
FM		 679 /*
680 * Note: We can be called before EXT4_SB(sb)->s_journal is set,
681 * therefore it can be null here. Don't check it, just initialize
682 * jinode.
683 */
678aaf48 684 jbd2_journal_init_jbd_inode(&ei->jinode, &ei->vfs_inode);
d2a17637
MC		 685 ei->i_reserved_data_blocks = 0;
686 ei->i_reserved_meta_blocks = 0;
687 ei->i_allocated_meta_blocks = 0;
688 ei->i_delalloc_reserved_flag = 0;
689 spin_lock_init(&(ei->i_block_reservation_lock));
0b8e58a1 690
ac27a0ec
DK		 691 return &ei->vfs_inode;
692}
693
617ba13b 694static void ext4_destroy_inode(struct inode *inode)
ac27a0ec 695{
9f7dd93d 696 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
b31e1552
ES		 697 ext4_msg(inode->i_sb, KERN_ERR,
698 "Inode %lu (%p): orphan list check failed!",
699 inode->i_ino, EXT4_I(inode));
9f7dd93d
VA		 700 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
701 EXT4_I(inode), sizeof(struct ext4_inode_info),
702 true);
703 dump_stack();
704 }
617ba13b 705 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
ac27a0ec
DK		 706}
707
51cc5068 708static void init_once(void *foo)
ac27a0ec 709{
617ba13b 710 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
ac27a0ec 711
a35afb83 712 INIT_LIST_HEAD(&ei->i_orphan);
03010a33 713#ifdef CONFIG_EXT4_FS_XATTR
a35afb83 714 init_rwsem(&ei->xattr_sem);
ac27a0ec 715#endif
0e855ac8 716 init_rwsem(&ei->i_data_sem);
a35afb83 717 inode_init_once(&ei->vfs_inode);
ac27a0ec
DK		 718}
719
720static int init_inodecache(void)
721{
617ba13b
MC		 722 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
723 sizeof(struct ext4_inode_info),
ac27a0ec
DK		 724 0, (SLAB_RECLAIM_ACCOUNT|
725 SLAB_MEM_SPREAD),
20c2df83 726 init_once);
617ba13b 727 if (ext4_inode_cachep == NULL)
ac27a0ec
DK		 728 return -ENOMEM;
729 return 0;
730}
731
732static void destroy_inodecache(void)
733{
617ba13b 734 kmem_cache_destroy(ext4_inode_cachep);
ac27a0ec
DK		 735}
736
617ba13b 737static void ext4_clear_inode(struct inode *inode)
ac27a0ec 738{
c2ea3fde 739 ext4_discard_preallocations(inode);
0390131b
FM		 740 if (EXT4_JOURNAL(inode))
741 jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal,
678aaf48 742 &EXT4_I(inode)->jinode);
ac27a0ec
DK		 743}
744
2b2d6d01
TT		 745static inline void ext4_show_quota_options(struct seq_file *seq,
746 struct super_block *sb)
ac27a0ec
DK		 747{
748#if defined(CONFIG_QUOTA)
617ba13b 749 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 750
751 if (sbi->s_jquota_fmt)
752 seq_printf(seq, ",jqfmt=%s",
af5bc92d 753 (sbi->s_jquota_fmt == QFMT_VFS_OLD) ? "vfsold" : "vfsv0");
ac27a0ec
DK		 754
755 if (sbi->s_qf_names[USRQUOTA])
756 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
757
758 if (sbi->s_qf_names[GRPQUOTA])
759 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
760
617ba13b 761 if (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA)
ac27a0ec
DK		 762 seq_puts(seq, ",usrquota");
763
617ba13b 764 if (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)
ac27a0ec
DK		 765 seq_puts(seq, ",grpquota");
766#endif
767}
768
d9c9bef1
MS		 769/*
770 * Show an option if
771 * - it's set to a non-default value OR
772 * - if the per-sb default is different from the global default
773 */
617ba13b 774static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
ac27a0ec 775{
aa22df2c
AK		 776 int def_errors;
777 unsigned long def_mount_opts;
ac27a0ec 778 struct super_block *sb = vfs->mnt_sb;
d9c9bef1
MS		 779 struct ext4_sb_info *sbi = EXT4_SB(sb);
780 struct ext4_super_block *es = sbi->s_es;
d9c9bef1
MS		 781
782 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
aa22df2c 783 def_errors = le16_to_cpu(es->s_errors);
d9c9bef1
MS		 784
785 if (sbi->s_sb_block != 1)
786 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
787 if (test_opt(sb, MINIX_DF))
788 seq_puts(seq, ",minixdf");
aa22df2c 789 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
d9c9bef1
MS		 790 seq_puts(seq, ",grpid");
791 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
792 seq_puts(seq, ",nogrpid");
793 if (sbi->s_resuid != EXT4_DEF_RESUID ||
794 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
795 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
796 }
797 if (sbi->s_resgid != EXT4_DEF_RESGID ||
798 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
799 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
800 }
bb4f397a 801 if (test_opt(sb, ERRORS_RO)) {
d9c9bef1 802 if (def_errors == EXT4_ERRORS_PANIC ||
bb4f397a
AK		 803 def_errors == EXT4_ERRORS_CONTINUE) {
804 seq_puts(seq, ",errors=remount-ro");
d9c9bef1
MS		 805 }
806 }
aa22df2c 807 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
bb4f397a 808 seq_puts(seq, ",errors=continue");
aa22df2c 809 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
d9c9bef1 810 seq_puts(seq, ",errors=panic");
aa22df2c 811 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
d9c9bef1 812 seq_puts(seq, ",nouid32");
aa22df2c 813 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
d9c9bef1
MS		 814 seq_puts(seq, ",debug");
815 if (test_opt(sb, OLDALLOC))
816 seq_puts(seq, ",oldalloc");
03010a33 817#ifdef CONFIG_EXT4_FS_XATTR
aa22df2c
AK		 818 if (test_opt(sb, XATTR_USER) &&
819 !(def_mount_opts & EXT4_DEFM_XATTR_USER))
d9c9bef1
MS		 820 seq_puts(seq, ",user_xattr");
821 if (!test_opt(sb, XATTR_USER) &&
822 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
823 seq_puts(seq, ",nouser_xattr");
824 }
825#endif
03010a33 826#ifdef CONFIG_EXT4_FS_POSIX_ACL
aa22df2c 827 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
d9c9bef1
MS		 828 seq_puts(seq, ",acl");
829 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
830 seq_puts(seq, ",noacl");
831#endif
30773840 832 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
d9c9bef1
MS		 833 seq_printf(seq, ",commit=%u",
834 (unsigned) (sbi->s_commit_interval / HZ));
835 }
30773840
TT		 836 if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
837 seq_printf(seq, ",min_batch_time=%u",
838 (unsigned) sbi->s_min_batch_time);
839 }
840 if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
841 seq_printf(seq, ",max_batch_time=%u",
842 (unsigned) sbi->s_min_batch_time);
843 }
844
571640ca
ES		 845 /*
846 * We're changing the default of barrier mount option, so
847 * let's always display its mount state so it's clear what its
848 * status is.
849 */
850 seq_puts(seq, ",barrier=");
851 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
cd0b6a39
TT		 852 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
853 seq_puts(seq, ",journal_async_commit");
d9c9bef1
MS		 854 if (test_opt(sb, NOBH))
855 seq_puts(seq, ",nobh");
25ec56b5
JNC		 856 if (test_opt(sb, I_VERSION))
857 seq_puts(seq, ",i_version");
dd919b98
AK		 858 if (!test_opt(sb, DELALLOC))
859 seq_puts(seq, ",nodelalloc");
860
ac27a0ec 861
cb45bbe4
MS		 862 if (sbi->s_stripe)
863 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
aa22df2c
AK		 864 /*
865 * journal mode get enabled in different ways
866 * So just print the value even if we didn't specify it
867 */
617ba13b 868 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
ac27a0ec 869 seq_puts(seq, ",data=journal");
617ba13b 870 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
ac27a0ec 871 seq_puts(seq, ",data=ordered");
617ba13b 872 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
ac27a0ec
DK		 873 seq_puts(seq, ",data=writeback");
874
240799cd
TT		 875 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
876 seq_printf(seq, ",inode_readahead_blks=%u",
877 sbi->s_inode_readahead_blks);
878
5bf5683a
HK		 879 if (test_opt(sb, DATA_ERR_ABORT))
880 seq_puts(seq, ",data_err=abort");
881
afd4672d 882 if (test_opt(sb, NO_AUTO_DA_ALLOC))
06705bff 883 seq_puts(seq, ",noauto_da_alloc");
afd4672d 884
617ba13b 885 ext4_show_quota_options(seq, sb);
0b8e58a1 886
ac27a0ec
DK		 887 return 0;
888}
889
1b961ac0 890static struct inode *ext4_nfs_get_inode(struct super_block *sb,
0b8e58a1 891 u64 ino, u32 generation)
ac27a0ec 892{
ac27a0ec 893 struct inode *inode;
ac27a0ec 894
617ba13b 895 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
ac27a0ec 896 return ERR_PTR(-ESTALE);
617ba13b 897 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
ac27a0ec
DK		 898 return ERR_PTR(-ESTALE);
899
900 /* iget isn't really right if the inode is currently unallocated!!
901 *
617ba13b 902 * ext4_read_inode will return a bad_inode if the inode had been
ac27a0ec
DK		 903 * deleted, so we should be safe.
904 *
905 * Currently we don't know the generation for parent directory, so
906 * a generation of 0 means "accept any"
907 */
1d1fe1ee
DH		 908 inode = ext4_iget(sb, ino);
909 if (IS_ERR(inode))
910 return ERR_CAST(inode);
911 if (generation && inode->i_generation != generation) {
ac27a0ec
DK		 912 iput(inode);
913 return ERR_PTR(-ESTALE);
914 }
1b961ac0
CH		 915
916 return inode;
917}
918
919static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
0b8e58a1 920 int fh_len, int fh_type)
1b961ac0
CH		 921{
922 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
923 ext4_nfs_get_inode);
924}
925
926static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
0b8e58a1 927 int fh_len, int fh_type)
1b961ac0
CH		 928{
929 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
930 ext4_nfs_get_inode);
ac27a0ec
DK		 931}
932
c39a7f84
TO		 933/*
934 * Try to release metadata pages (indirect blocks, directories) which are
935 * mapped via the block device. Since these pages could have journal heads
936 * which would prevent try_to_free_buffers() from freeing them, we must use
937 * jbd2 layer's try_to_free_buffers() function to release them.
938 */
0b8e58a1
AD		 939static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
940 gfp_t wait)
c39a7f84
TO		 941{
942 journal_t *journal = EXT4_SB(sb)->s_journal;
943
944 WARN_ON(PageChecked(page));
945 if (!page_has_buffers(page))
946 return 0;
947 if (journal)
948 return jbd2_journal_try_to_free_buffers(journal, page,
949 wait & ~__GFP_WAIT);
950 return try_to_free_buffers(page);
951}
952
ac27a0ec 953#ifdef CONFIG_QUOTA
af5bc92d 954#define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
2b2d6d01 955#define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
ac27a0ec 956
617ba13b
MC		 957static int ext4_write_dquot(struct dquot *dquot);
958static int ext4_acquire_dquot(struct dquot *dquot);
959static int ext4_release_dquot(struct dquot *dquot);
960static int ext4_mark_dquot_dirty(struct dquot *dquot);
961static int ext4_write_info(struct super_block *sb, int type);
6f28e087
JK		 962static int ext4_quota_on(struct super_block *sb, int type, int format_id,
963 char *path, int remount);
617ba13b
MC		 964static int ext4_quota_on_mount(struct super_block *sb, int type);
965static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 966 size_t len, loff_t off);
617ba13b 967static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 968 const char *data, size_t len, loff_t off);
969
61e225dc 970static const struct dquot_operations ext4_quota_operations = {
edf72453
JK		 971 .initialize = dquot_initialize,
972 .drop = dquot_drop,
ac27a0ec 973 .alloc_space = dquot_alloc_space,
60e58e0f
MC		 974 .reserve_space = dquot_reserve_space,
975 .claim_space = dquot_claim_space,
976 .release_rsv = dquot_release_reserved_space,
977 .get_reserved_space = ext4_get_reserved_space,
ac27a0ec
DK		 978 .alloc_inode = dquot_alloc_inode,
979 .free_space = dquot_free_space,
980 .free_inode = dquot_free_inode,
981 .transfer = dquot_transfer,
617ba13b
MC		 982 .write_dquot = ext4_write_dquot,
983 .acquire_dquot = ext4_acquire_dquot,
984 .release_dquot = ext4_release_dquot,
985 .mark_dirty = ext4_mark_dquot_dirty,
a5b5ee32
JK		 986 .write_info = ext4_write_info,
987 .alloc_dquot = dquot_alloc,
988 .destroy_dquot = dquot_destroy,
ac27a0ec
DK		 989};
990
0d54b217 991static const struct quotactl_ops ext4_qctl_operations = {
617ba13b 992 .quota_on = ext4_quota_on,
ac27a0ec
DK		 993 .quota_off = vfs_quota_off,
994 .quota_sync = vfs_quota_sync,
995 .get_info = vfs_get_dqinfo,
996 .set_info = vfs_set_dqinfo,
997 .get_dqblk = vfs_get_dqblk,
998 .set_dqblk = vfs_set_dqblk
999};
1000#endif
1001
ee9b6d61 1002static const struct super_operations ext4_sops = {
617ba13b
MC		 1003 .alloc_inode = ext4_alloc_inode,
1004 .destroy_inode = ext4_destroy_inode,
617ba13b
MC		 1005 .write_inode = ext4_write_inode,
1006 .dirty_inode = ext4_dirty_inode,
1007 .delete_inode = ext4_delete_inode,
1008 .put_super = ext4_put_super,
617ba13b 1009 .sync_fs = ext4_sync_fs,
c4be0c1d
TS		 1010 .freeze_fs = ext4_freeze,
1011 .unfreeze_fs = ext4_unfreeze,
617ba13b
MC		 1012 .statfs = ext4_statfs,
1013 .remount_fs = ext4_remount,
1014 .clear_inode = ext4_clear_inode,
1015 .show_options = ext4_show_options,
ac27a0ec 1016#ifdef CONFIG_QUOTA
617ba13b
MC		 1017 .quota_read = ext4_quota_read,
1018 .quota_write = ext4_quota_write,
ac27a0ec 1019#endif
c39a7f84 1020 .bdev_try_to_free_page = bdev_try_to_free_page,
ac27a0ec
DK		 1021};
1022
9ca92389
TT		 1023static const struct super_operations ext4_nojournal_sops = {
1024 .alloc_inode = ext4_alloc_inode,
1025 .destroy_inode = ext4_destroy_inode,
1026 .write_inode = ext4_write_inode,
1027 .dirty_inode = ext4_dirty_inode,
1028 .delete_inode = ext4_delete_inode,
1029 .write_super = ext4_write_super,
1030 .put_super = ext4_put_super,
1031 .statfs = ext4_statfs,
1032 .remount_fs = ext4_remount,
1033 .clear_inode = ext4_clear_inode,
1034 .show_options = ext4_show_options,
1035#ifdef CONFIG_QUOTA
1036 .quota_read = ext4_quota_read,
1037 .quota_write = ext4_quota_write,
1038#endif
1039 .bdev_try_to_free_page = bdev_try_to_free_page,
1040};
1041
39655164 1042static const struct export_operations ext4_export_ops = {
1b961ac0
CH		 1043 .fh_to_dentry = ext4_fh_to_dentry,
1044 .fh_to_parent = ext4_fh_to_parent,
617ba13b 1045 .get_parent = ext4_get_parent,
ac27a0ec
DK		 1046};
1047
1048enum {
1049 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1050 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
01436ef2 1051 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
ac27a0ec 1052 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
06705bff 1053 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh,
30773840 1054 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
c3191067 1055 Opt_journal_update, Opt_journal_dev,
818d276c 1056 Opt_journal_checksum, Opt_journal_async_commit,
ac27a0ec 1057 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
f4033903 1058 Opt_data_err_abort, Opt_data_err_ignore, Opt_mb_history_length,
ac27a0ec
DK		 1059 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1060 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_quota, Opt_noquota,
06705bff
TT		 1061 Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err, Opt_resize,
1062 Opt_usrquota, Opt_grpquota, Opt_i_version,
01436ef2 1063 Opt_stripe, Opt_delalloc, Opt_nodelalloc,
6fd058f7 1064 Opt_block_validity, Opt_noblock_validity,
b3881f74 1065 Opt_inode_readahead_blks, Opt_journal_ioprio
ac27a0ec
DK		 1066};
1067
a447c093 1068static const match_table_t tokens = {
ac27a0ec
DK		 1069 {Opt_bsd_df, "bsddf"},
1070 {Opt_minix_df, "minixdf"},
1071 {Opt_grpid, "grpid"},
1072 {Opt_grpid, "bsdgroups"},
1073 {Opt_nogrpid, "nogrpid"},
1074 {Opt_nogrpid, "sysvgroups"},
1075 {Opt_resgid, "resgid=%u"},
1076 {Opt_resuid, "resuid=%u"},
1077 {Opt_sb, "sb=%u"},
1078 {Opt_err_cont, "errors=continue"},
1079 {Opt_err_panic, "errors=panic"},
1080 {Opt_err_ro, "errors=remount-ro"},
1081 {Opt_nouid32, "nouid32"},
ac27a0ec
DK		 1082 {Opt_debug, "debug"},
1083 {Opt_oldalloc, "oldalloc"},
1084 {Opt_orlov, "orlov"},
1085 {Opt_user_xattr, "user_xattr"},
1086 {Opt_nouser_xattr, "nouser_xattr"},
1087 {Opt_acl, "acl"},
1088 {Opt_noacl, "noacl"},
ac27a0ec
DK		 1089 {Opt_noload, "noload"},
1090 {Opt_nobh, "nobh"},
1091 {Opt_bh, "bh"},
1092 {Opt_commit, "commit=%u"},
30773840
TT		 1093 {Opt_min_batch_time, "min_batch_time=%u"},
1094 {Opt_max_batch_time, "max_batch_time=%u"},
ac27a0ec 1095 {Opt_journal_update, "journal=update"},
ac27a0ec 1096 {Opt_journal_dev, "journal_dev=%u"},
818d276c
GS		 1097 {Opt_journal_checksum, "journal_checksum"},
1098 {Opt_journal_async_commit, "journal_async_commit"},
ac27a0ec
DK		 1099 {Opt_abort, "abort"},
1100 {Opt_data_journal, "data=journal"},
1101 {Opt_data_ordered, "data=ordered"},
1102 {Opt_data_writeback, "data=writeback"},
5bf5683a
HK		 1103 {Opt_data_err_abort, "data_err=abort"},
1104 {Opt_data_err_ignore, "data_err=ignore"},
f4033903 1105 {Opt_mb_history_length, "mb_history_length=%u"},
ac27a0ec
DK		 1106 {Opt_offusrjquota, "usrjquota="},
1107 {Opt_usrjquota, "usrjquota=%s"},
1108 {Opt_offgrpjquota, "grpjquota="},
1109 {Opt_grpjquota, "grpjquota=%s"},
1110 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1111 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1112 {Opt_grpquota, "grpquota"},
1113 {Opt_noquota, "noquota"},
1114 {Opt_quota, "quota"},
1115 {Opt_usrquota, "usrquota"},
1116 {Opt_barrier, "barrier=%u"},
06705bff
TT		 1117 {Opt_barrier, "barrier"},
1118 {Opt_nobarrier, "nobarrier"},
25ec56b5 1119 {Opt_i_version, "i_version"},
c9de560d 1120 {Opt_stripe, "stripe=%u"},
ac27a0ec 1121 {Opt_resize, "resize"},
64769240 1122 {Opt_delalloc, "delalloc"},
dd919b98 1123 {Opt_nodelalloc, "nodelalloc"},
6fd058f7
TT		 1124 {Opt_block_validity, "block_validity"},
1125 {Opt_noblock_validity, "noblock_validity"},
240799cd 1126 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
b3881f74 1127 {Opt_journal_ioprio, "journal_ioprio=%u"},
afd4672d 1128 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
06705bff
TT		 1129 {Opt_auto_da_alloc, "auto_da_alloc"},
1130 {Opt_noauto_da_alloc, "noauto_da_alloc"},
f3f12faa 1131 {Opt_err, NULL},
ac27a0ec
DK		 1132};
1133
617ba13b 1134static ext4_fsblk_t get_sb_block(void **data)
ac27a0ec 1135{
617ba13b 1136 ext4_fsblk_t sb_block;
ac27a0ec
DK		 1137 char *options = (char *) *data;
1138
1139 if (!options || strncmp(options, "sb=", 3) != 0)
1140 return 1; /* Default location */
0b8e58a1 1141
ac27a0ec 1142 options += 3;
0b8e58a1 1143 /* TODO: use simple_strtoll with >32bit ext4 */
ac27a0ec
DK		 1144 sb_block = simple_strtoul(options, &options, 0);
1145 if (*options && *options != ',') {
4776004f 1146 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
ac27a0ec
DK		 1147 (char *) *data);
1148 return 1;
1149 }
1150 if (*options == ',')
1151 options++;
1152 *data = (void *) options;
0b8e58a1 1153
ac27a0ec
DK		 1154 return sb_block;
1155}
1156
b3881f74
TT		 1157#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1158
2b2d6d01 1159static int parse_options(char *options, struct super_block *sb,
c3191067 1160 unsigned long *journal_devnum,
b3881f74 1161 unsigned int *journal_ioprio,
2b2d6d01 1162 ext4_fsblk_t *n_blocks_count, int is_remount)
ac27a0ec 1163{
617ba13b 1164 struct ext4_sb_info *sbi = EXT4_SB(sb);
2b2d6d01 1165 char *p;
ac27a0ec
DK		 1166 substring_t args[MAX_OPT_ARGS];
1167 int data_opt = 0;
1168 int option;
1169#ifdef CONFIG_QUOTA
dfc5d03f 1170 int qtype, qfmt;
ac27a0ec
DK		 1171 char *qname;
1172#endif
1173
1174 if (!options)
1175 return 1;
1176
2b2d6d01 1177 while ((p = strsep(&options, ",")) != NULL) {
ac27a0ec
DK		 1178 int token;
1179 if (!*p)
1180 continue;
1181
1182 token = match_token(p, tokens, args);
1183 switch (token) {
1184 case Opt_bsd_df:
2b2d6d01 1185 clear_opt(sbi->s_mount_opt, MINIX_DF);
ac27a0ec
DK		 1186 break;
1187 case Opt_minix_df:
2b2d6d01 1188 set_opt(sbi->s_mount_opt, MINIX_DF);
ac27a0ec
DK		 1189 break;
1190 case Opt_grpid:
2b2d6d01 1191 set_opt(sbi->s_mount_opt, GRPID);
ac27a0ec
DK		 1192 break;
1193 case Opt_nogrpid:
2b2d6d01 1194 clear_opt(sbi->s_mount_opt, GRPID);
ac27a0ec
DK		 1195 break;
1196 case Opt_resuid:
1197 if (match_int(&args[0], &option))
1198 return 0;
1199 sbi->s_resuid = option;
1200 break;
1201 case Opt_resgid:
1202 if (match_int(&args[0], &option))
1203 return 0;
1204 sbi->s_resgid = option;
1205 break;
1206 case Opt_sb:
1207 /* handled by get_sb_block() instead of here */
1208 /* *sb_block = match_int(&args[0]); */
1209 break;
1210 case Opt_err_panic:
2b2d6d01
TT		 1211 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1212 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1213 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
ac27a0ec
DK		 1214 break;
1215 case Opt_err_ro:
2b2d6d01
TT		 1216 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1217 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1218 set_opt(sbi->s_mount_opt, ERRORS_RO);
ac27a0ec
DK		 1219 break;
1220 case Opt_err_cont:
2b2d6d01
TT		 1221 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1222 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1223 set_opt(sbi->s_mount_opt, ERRORS_CONT);
ac27a0ec
DK		 1224 break;
1225 case Opt_nouid32:
2b2d6d01 1226 set_opt(sbi->s_mount_opt, NO_UID32);
ac27a0ec 1227 break;
ac27a0ec 1228 case Opt_debug:
2b2d6d01 1229 set_opt(sbi->s_mount_opt, DEBUG);
ac27a0ec
DK		 1230 break;
1231 case Opt_oldalloc:
2b2d6d01 1232 set_opt(sbi->s_mount_opt, OLDALLOC);
ac27a0ec
DK		 1233 break;
1234 case Opt_orlov:
2b2d6d01 1235 clear_opt(sbi->s_mount_opt, OLDALLOC);
ac27a0ec 1236 break;
03010a33 1237#ifdef CONFIG_EXT4_FS_XATTR
ac27a0ec 1238 case Opt_user_xattr:
2b2d6d01 1239 set_opt(sbi->s_mount_opt, XATTR_USER);
ac27a0ec
DK		 1240 break;
1241 case Opt_nouser_xattr:
2b2d6d01 1242 clear_opt(sbi->s_mount_opt, XATTR_USER);
ac27a0ec
DK		 1243 break;
1244#else
1245 case Opt_user_xattr:
1246 case Opt_nouser_xattr:
b31e1552 1247 ext4_msg(sb, KERN_ERR, "(no)user_xattr options not supported");
ac27a0ec
DK		 1248 break;
1249#endif
03010a33 1250#ifdef CONFIG_EXT4_FS_POSIX_ACL
ac27a0ec
DK		 1251 case Opt_acl:
1252 set_opt(sbi->s_mount_opt, POSIX_ACL);
1253 break;
1254 case Opt_noacl:
1255 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1256 break;
1257#else
1258 case Opt_acl:
1259 case Opt_noacl:
b31e1552 1260 ext4_msg(sb, KERN_ERR, "(no)acl options not supported");
ac27a0ec
DK		 1261 break;
1262#endif
ac27a0ec
DK		 1263 case Opt_journal_update:
1264 /* @@@ FIXME */
1265 /* Eventually we will want to be able to create
1266 a journal file here. For now, only allow the
1267 user to specify an existing inode to be the
1268 journal file. */
1269 if (is_remount) {
b31e1552
ES		 1270 ext4_msg(sb, KERN_ERR,
1271 "Cannot specify journal on remount");
ac27a0ec
DK		 1272 return 0;
1273 }
2b2d6d01 1274 set_opt(sbi->s_mount_opt, UPDATE_JOURNAL);
ac27a0ec 1275 break;
ac27a0ec
DK		 1276 case Opt_journal_dev:
1277 if (is_remount) {
b31e1552
ES		 1278 ext4_msg(sb, KERN_ERR,
1279 "Cannot specify journal on remount");
ac27a0ec
DK		 1280 return 0;
1281 }
1282 if (match_int(&args[0], &option))
1283 return 0;
1284 *journal_devnum = option;
1285 break;
818d276c 1286 case Opt_journal_checksum:
d0646f7b 1287 break; /* Kept for backwards compatibility */
818d276c
GS		 1288 case Opt_journal_async_commit:
1289 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
818d276c 1290 break;
ac27a0ec 1291 case Opt_noload:
2b2d6d01 1292 set_opt(sbi->s_mount_opt, NOLOAD);
ac27a0ec
DK		 1293 break;
1294 case Opt_commit:
1295 if (match_int(&args[0], &option))
1296 return 0;
1297 if (option < 0)
1298 return 0;
1299 if (option == 0)
cd02ff0b 1300 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
ac27a0ec
DK		 1301 sbi->s_commit_interval = HZ * option;
1302 break;
30773840
TT		 1303 case Opt_max_batch_time:
1304 if (match_int(&args[0], &option))
1305 return 0;
1306 if (option < 0)
1307 return 0;
1308 if (option == 0)
1309 option = EXT4_DEF_MAX_BATCH_TIME;
1310 sbi->s_max_batch_time = option;
1311 break;
1312 case Opt_min_batch_time:
1313 if (match_int(&args[0], &option))
1314 return 0;
1315 if (option < 0)
1316 return 0;
1317 sbi->s_min_batch_time = option;
1318 break;
ac27a0ec 1319 case Opt_data_journal:
617ba13b 1320 data_opt = EXT4_MOUNT_JOURNAL_DATA;
ac27a0ec
DK		 1321 goto datacheck;
1322 case Opt_data_ordered:
617ba13b 1323 data_opt = EXT4_MOUNT_ORDERED_DATA;
ac27a0ec
DK		 1324 goto datacheck;
1325 case Opt_data_writeback:
617ba13b 1326 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
ac27a0ec
DK		 1327 datacheck:
1328 if (is_remount) {
617ba13b 1329 if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
ac27a0ec 1330 != data_opt) {
b31e1552
ES		 1331 ext4_msg(sb, KERN_ERR,
1332 "Cannot change data mode on remount");
ac27a0ec
DK		 1333 return 0;
1334 }
1335 } else {
617ba13b 1336 sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
ac27a0ec
DK		 1337 sbi->s_mount_opt |= data_opt;
1338 }
1339 break;
5bf5683a
HK		 1340 case Opt_data_err_abort:
1341 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1342 break;
1343 case Opt_data_err_ignore:
1344 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1345 break;
f4033903
CW		 1346 case Opt_mb_history_length:
1347 if (match_int(&args[0], &option))
1348 return 0;
1349 if (option < 0)
1350 return 0;
1351 sbi->s_mb_history_max = option;
1352 break;
ac27a0ec
DK		 1353#ifdef CONFIG_QUOTA
1354 case Opt_usrjquota:
1355 qtype = USRQUOTA;
1356 goto set_qf_name;
1357 case Opt_grpjquota:
1358 qtype = GRPQUOTA;
1359set_qf_name:
17bd13b3 1360 if (sb_any_quota_loaded(sb) &&
dfc5d03f 1361 !sbi->s_qf_names[qtype]) {
b31e1552
ES		 1362 ext4_msg(sb, KERN_ERR,
1363 "Cannot change journaled "
1364 "quota options when quota turned on");
ac27a0ec
DK		 1365 return 0;
1366 }
1367 qname = match_strdup(&args[0]);
1368 if (!qname) {
b31e1552
ES		 1369 ext4_msg(sb, KERN_ERR,
1370 "Not enough memory for "
1371 "storing quotafile name");
ac27a0ec
DK		 1372 return 0;
1373 }
1374 if (sbi->s_qf_names[qtype] &&
1375 strcmp(sbi->s_qf_names[qtype], qname)) {
b31e1552
ES		 1376 ext4_msg(sb, KERN_ERR,
1377 "%s quota file already "
1378 "specified", QTYPE2NAME(qtype));
ac27a0ec
DK		 1379 kfree(qname);
1380 return 0;
1381 }
1382 sbi->s_qf_names[qtype] = qname;
1383 if (strchr(sbi->s_qf_names[qtype], '/')) {
b31e1552
ES		 1384 ext4_msg(sb, KERN_ERR,
1385 "quotafile must be on "
1386 "filesystem root");
ac27a0ec
DK		 1387 kfree(sbi->s_qf_names[qtype]);
1388 sbi->s_qf_names[qtype] = NULL;
1389 return 0;
1390 }
1391 set_opt(sbi->s_mount_opt, QUOTA);
1392 break;
1393 case Opt_offusrjquota:
1394 qtype = USRQUOTA;
1395 goto clear_qf_name;
1396 case Opt_offgrpjquota:
1397 qtype = GRPQUOTA;
1398clear_qf_name:
17bd13b3 1399 if (sb_any_quota_loaded(sb) &&
dfc5d03f 1400 sbi->s_qf_names[qtype]) {
b31e1552 1401 ext4_msg(sb, KERN_ERR, "Cannot change "
2c8be6b2 1402 "journaled quota options when "
b31e1552 1403 "quota turned on");
ac27a0ec
DK		 1404 return 0;
1405 }
1406 /*
1407 * The space will be released later when all options
1408 * are confirmed to be correct
1409 */
1410 sbi->s_qf_names[qtype] = NULL;
1411 break;
1412 case Opt_jqfmt_vfsold:
dfc5d03f
JK		 1413 qfmt = QFMT_VFS_OLD;
1414 goto set_qf_format;
ac27a0ec 1415 case Opt_jqfmt_vfsv0:
dfc5d03f
JK		 1416 qfmt = QFMT_VFS_V0;
1417set_qf_format:
17bd13b3 1418 if (sb_any_quota_loaded(sb) &&
dfc5d03f 1419 sbi->s_jquota_fmt != qfmt) {
b31e1552 1420 ext4_msg(sb, KERN_ERR, "Cannot change "
dfc5d03f 1421 "journaled quota options when "
b31e1552 1422 "quota turned on");
dfc5d03f
JK		 1423 return 0;
1424 }
1425 sbi->s_jquota_fmt = qfmt;
ac27a0ec
DK		 1426 break;
1427 case Opt_quota:
1428 case Opt_usrquota:
1429 set_opt(sbi->s_mount_opt, QUOTA);
1430 set_opt(sbi->s_mount_opt, USRQUOTA);
1431 break;
1432 case Opt_grpquota:
1433 set_opt(sbi->s_mount_opt, QUOTA);
1434 set_opt(sbi->s_mount_opt, GRPQUOTA);
1435 break;
1436 case Opt_noquota:
17bd13b3 1437 if (sb_any_quota_loaded(sb)) {
b31e1552
ES		 1438 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1439 "options when quota turned on");
ac27a0ec
DK		 1440 return 0;
1441 }
1442 clear_opt(sbi->s_mount_opt, QUOTA);
1443 clear_opt(sbi->s_mount_opt, USRQUOTA);
1444 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1445 break;
1446#else
1447 case Opt_quota:
1448 case Opt_usrquota:
1449 case Opt_grpquota:
b31e1552
ES		 1450 ext4_msg(sb, KERN_ERR,
1451 "quota options not supported");
cd59e7b9 1452 break;
ac27a0ec
DK		 1453 case Opt_usrjquota:
1454 case Opt_grpjquota:
1455 case Opt_offusrjquota:
1456 case Opt_offgrpjquota:
1457 case Opt_jqfmt_vfsold:
1458 case Opt_jqfmt_vfsv0:
b31e1552
ES		 1459 ext4_msg(sb, KERN_ERR,
1460 "journaled quota options not supported");
ac27a0ec
DK		 1461 break;
1462 case Opt_noquota:
1463 break;
1464#endif
1465 case Opt_abort:
4ab2f15b 1466 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
ac27a0ec 1467 break;
06705bff
TT		 1468 case Opt_nobarrier:
1469 clear_opt(sbi->s_mount_opt, BARRIER);
1470 break;
ac27a0ec 1471 case Opt_barrier:
06705bff
TT		 1472 if (match_int(&args[0], &option)) {
1473 set_opt(sbi->s_mount_opt, BARRIER);
1474 break;
1475 }
ac27a0ec
DK		 1476 if (option)
1477 set_opt(sbi->s_mount_opt, BARRIER);
1478 else
1479 clear_opt(sbi->s_mount_opt, BARRIER);
1480 break;
1481 case Opt_ignore:
1482 break;
1483 case Opt_resize:
1484 if (!is_remount) {
b31e1552
ES		 1485 ext4_msg(sb, KERN_ERR,
1486 "resize option only available "
1487 "for remount");
ac27a0ec
DK		 1488 return 0;
1489 }
1490 if (match_int(&args[0], &option) != 0)
1491 return 0;
1492 *n_blocks_count = option;
1493 break;
1494 case Opt_nobh:
1495 set_opt(sbi->s_mount_opt, NOBH);
1496 break;
1497 case Opt_bh:
1498 clear_opt(sbi->s_mount_opt, NOBH);
1499 break;
25ec56b5
JNC		 1500 case Opt_i_version:
1501 set_opt(sbi->s_mount_opt, I_VERSION);
1502 sb->s_flags |= MS_I_VERSION;
1503 break;
dd919b98
AK		 1504 case Opt_nodelalloc:
1505 clear_opt(sbi->s_mount_opt, DELALLOC);
1506 break;
c9de560d
AT		 1507 case Opt_stripe:
1508 if (match_int(&args[0], &option))
1509 return 0;
1510 if (option < 0)
1511 return 0;
1512 sbi->s_stripe = option;
1513 break;
64769240
AT		 1514 case Opt_delalloc:
1515 set_opt(sbi->s_mount_opt, DELALLOC);
1516 break;
6fd058f7
TT		 1517 case Opt_block_validity:
1518 set_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1519 break;
1520 case Opt_noblock_validity:
1521 clear_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1522 break;
240799cd
TT		 1523 case Opt_inode_readahead_blks:
1524 if (match_int(&args[0], &option))
1525 return 0;
1526 if (option < 0 || option > (1 << 30))
1527 return 0;
f7c43950 1528 if (!is_power_of_2(option)) {
b31e1552
ES		 1529 ext4_msg(sb, KERN_ERR,
1530 "EXT4-fs: inode_readahead_blks"
1531 " must be a power of 2");
3197ebdb
TT		 1532 return 0;
1533 }
240799cd
TT		 1534 sbi->s_inode_readahead_blks = option;
1535 break;
b3881f74
TT		 1536 case Opt_journal_ioprio:
1537 if (match_int(&args[0], &option))
1538 return 0;
1539 if (option < 0 || option > 7)
1540 break;
1541 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1542 option);
1543 break;
06705bff
TT		 1544 case Opt_noauto_da_alloc:
1545 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1546 break;
afd4672d 1547 case Opt_auto_da_alloc:
06705bff
TT		 1548 if (match_int(&args[0], &option)) {
1549 clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1550 break;
1551 }
afd4672d
TT		 1552 if (option)
1553 clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1554 else
1555 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1556 break;
ac27a0ec 1557 default:
b31e1552
ES		 1558 ext4_msg(sb, KERN_ERR,
1559 "Unrecognized mount option \"%s\" "
1560 "or missing value", p);
ac27a0ec
DK		 1561 return 0;
1562 }
1563 }
1564#ifdef CONFIG_QUOTA
1565 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
617ba13b 1566 if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
ac27a0ec
DK		 1567 sbi->s_qf_names[USRQUOTA])
1568 clear_opt(sbi->s_mount_opt, USRQUOTA);
1569
617ba13b 1570 if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
ac27a0ec
DK		 1571 sbi->s_qf_names[GRPQUOTA])
1572 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1573
1574 if ((sbi->s_qf_names[USRQUOTA] &&
617ba13b 1575 (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
ac27a0ec 1576 (sbi->s_qf_names[GRPQUOTA] &&
617ba13b 1577 (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
b31e1552
ES		 1578 ext4_msg(sb, KERN_ERR, "old and new quota "
1579 "format mixing");
ac27a0ec
DK		 1580 return 0;
1581 }
1582
1583 if (!sbi->s_jquota_fmt) {
b31e1552
ES		 1584 ext4_msg(sb, KERN_ERR, "journaled quota format "
1585 "not specified");
ac27a0ec
DK		 1586 return 0;
1587 }
1588 } else {
1589 if (sbi->s_jquota_fmt) {
b31e1552 1590 ext4_msg(sb, KERN_ERR, "journaled quota format "
2c8be6b2 1591 "specified with no journaling "
b31e1552 1592 "enabled");
ac27a0ec
DK		 1593 return 0;
1594 }
1595 }
1596#endif
1597 return 1;
1598}
1599
617ba13b 1600static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 1601 int read_only)
1602{
617ba13b 1603 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 1604 int res = 0;
1605
617ba13b 1606 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
b31e1552
ES		 1607 ext4_msg(sb, KERN_ERR, "revision level too high, "
1608 "forcing read-only mode");
ac27a0ec
DK		 1609 res = MS_RDONLY;
1610 }
1611 if (read_only)
1612 return res;
617ba13b 1613 if (!(sbi->s_mount_state & EXT4_VALID_FS))
b31e1552
ES		 1614 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1615 "running e2fsck is recommended");
617ba13b 1616 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
b31e1552
ES		 1617 ext4_msg(sb, KERN_WARNING,
1618 "warning: mounting fs with errors, "
1619 "running e2fsck is recommended");
ac27a0ec
DK		 1620 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1621 le16_to_cpu(es->s_mnt_count) >=
1622 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
b31e1552
ES		 1623 ext4_msg(sb, KERN_WARNING,
1624 "warning: maximal mount count reached, "
1625 "running e2fsck is recommended");
ac27a0ec
DK		 1626 else if (le32_to_cpu(es->s_checkinterval) &&
1627 (le32_to_cpu(es->s_lastcheck) +
1628 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
b31e1552
ES		 1629 ext4_msg(sb, KERN_WARNING,
1630 "warning: checktime reached, "
1631 "running e2fsck is recommended");
0b8e58a1 1632 if (!sbi->s_journal)
0390131b 1633 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec 1634 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 1635 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 1636 le16_add_cpu(&es->s_mnt_count, 1);
ac27a0ec 1637 es->s_mtime = cpu_to_le32(get_seconds());
617ba13b 1638 ext4_update_dynamic_rev(sb);
0390131b
FM		 1639 if (sbi->s_journal)
1640 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 1641
e2d67052 1642 ext4_commit_super(sb, 1);
ac27a0ec 1643 if (test_opt(sb, DEBUG))
a9df9a49 1644 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
7f4520cc 1645 "bpg=%lu, ipg=%lu, mo=%04x]\n",
ac27a0ec
DK		 1646 sb->s_blocksize,
1647 sbi->s_groups_count,
617ba13b
MC		 1648 EXT4_BLOCKS_PER_GROUP(sb),
1649 EXT4_INODES_PER_GROUP(sb),
ac27a0ec
DK		 1650 sbi->s_mount_opt);
1651
0390131b 1652 if (EXT4_SB(sb)->s_journal) {
b31e1552
ES		 1653 ext4_msg(sb, KERN_INFO, "%s journal on %s",
1654 EXT4_SB(sb)->s_journal->j_inode ? "internal" :
0390131b
FM		 1655 "external", EXT4_SB(sb)->s_journal->j_devname);
1656 } else {
b31e1552 1657 ext4_msg(sb, KERN_INFO, "no journal");
0390131b 1658 }
ac27a0ec
DK		 1659 return res;
1660}
1661
772cb7c8
JS		 1662static int ext4_fill_flex_info(struct super_block *sb)
1663{
1664 struct ext4_sb_info *sbi = EXT4_SB(sb);
1665 struct ext4_group_desc *gdp = NULL;
772cb7c8
JS		 1666 ext4_group_t flex_group_count;
1667 ext4_group_t flex_group;
1668 int groups_per_flex = 0;
c5ca7c76 1669 size_t size;
772cb7c8
JS		 1670 int i;
1671
1672 if (!sbi->s_es->s_log_groups_per_flex) {
1673 sbi->s_log_groups_per_flex = 0;
1674 return 1;
1675 }
1676
1677 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1678 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1679
c62a11fd
FB		 1680 /* We allocate both existing and potentially added groups */
1681 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
d94e99a6
AK		 1682 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1683 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
c5ca7c76
TT		 1684 size = flex_group_count * sizeof(struct flex_groups);
1685 sbi->s_flex_groups = kzalloc(size, GFP_KERNEL);
1686 if (sbi->s_flex_groups == NULL) {
1687 sbi->s_flex_groups = vmalloc(size);
1688 if (sbi->s_flex_groups)
1689 memset(sbi->s_flex_groups, 0, size);
1690 }
772cb7c8 1691 if (sbi->s_flex_groups == NULL) {
b31e1552
ES		 1692 ext4_msg(sb, KERN_ERR, "not enough memory for "
1693 "%u flex groups", flex_group_count);
772cb7c8
JS		 1694 goto failed;
1695 }
772cb7c8 1696
772cb7c8 1697 for (i = 0; i < sbi->s_groups_count; i++) {
88b6edd1 1698 gdp = ext4_get_group_desc(sb, i, NULL);
772cb7c8
JS		 1699
1700 flex_group = ext4_flex_group(sbi, i);
7ad9bb65
TT		 1701 atomic_add(ext4_free_inodes_count(sb, gdp),
1702 &sbi->s_flex_groups[flex_group].free_inodes);
1703 atomic_add(ext4_free_blks_count(sb, gdp),
1704 &sbi->s_flex_groups[flex_group].free_blocks);
1705 atomic_add(ext4_used_dirs_count(sb, gdp),
1706 &sbi->s_flex_groups[flex_group].used_dirs);
772cb7c8
JS		 1707 }
1708
1709 return 1;
1710failed:
1711 return 0;
1712}
1713
717d50e4
AD		 1714__le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1715 struct ext4_group_desc *gdp)
1716{
1717 __u16 crc = 0;
1718
1719 if (sbi->s_es->s_feature_ro_compat &
1720 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1721 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1722 __le32 le_group = cpu_to_le32(block_group);
1723
1724 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1725 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1726 crc = crc16(crc, (__u8 *)gdp, offset);
1727 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1728 /* for checksum of struct ext4_group_desc do the rest...*/
1729 if ((sbi->s_es->s_feature_incompat &
1730 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1731 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1732 crc = crc16(crc, (__u8 *)gdp + offset,
1733 le16_to_cpu(sbi->s_es->s_desc_size) -
1734 offset);
1735 }
1736
1737 return cpu_to_le16(crc);
1738}
1739
1740int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1741 struct ext4_group_desc *gdp)
1742{
1743 if ((sbi->s_es->s_feature_ro_compat &
1744 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1745 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1746 return 0;
1747
1748 return 1;
1749}
1750
ac27a0ec 1751/* Called at mount-time, super-block is locked */
197cd65a 1752static int ext4_check_descriptors(struct super_block *sb)
ac27a0ec 1753{
617ba13b
MC		 1754 struct ext4_sb_info *sbi = EXT4_SB(sb);
1755 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1756 ext4_fsblk_t last_block;
bd81d8ee
LV		 1757 ext4_fsblk_t block_bitmap;
1758 ext4_fsblk_t inode_bitmap;
1759 ext4_fsblk_t inode_table;
ce421581 1760 int flexbg_flag = 0;
fd2d4291 1761 ext4_group_t i;
ac27a0ec 1762
ce421581
JS		 1763 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1764 flexbg_flag = 1;
1765
af5bc92d 1766 ext4_debug("Checking group descriptors");
ac27a0ec 1767
197cd65a
AM		 1768 for (i = 0; i < sbi->s_groups_count; i++) {
1769 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1770
ce421581 1771 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 1772 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 1773 else
1774 last_block = first_block +
617ba13b 1775 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 1776
8fadc143 1777 block_bitmap = ext4_block_bitmap(sb, gdp);
2b2d6d01 1778 if (block_bitmap < first_block || block_bitmap > last_block) {
b31e1552 1779 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 1780 "Block bitmap for group %u not in group "
b31e1552 1781 "(block %llu)!", i, block_bitmap);
ac27a0ec
DK		 1782 return 0;
1783 }
8fadc143 1784 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2b2d6d01 1785 if (inode_bitmap < first_block || inode_bitmap > last_block) {
b31e1552 1786 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 1787 "Inode bitmap for group %u not in group "
b31e1552 1788 "(block %llu)!", i, inode_bitmap);
ac27a0ec
DK		 1789 return 0;
1790 }
8fadc143 1791 inode_table = ext4_inode_table(sb, gdp);
bd81d8ee 1792 if (inode_table < first_block ||
2b2d6d01 1793 inode_table + sbi->s_itb_per_group - 1 > last_block) {
b31e1552 1794 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 1795 "Inode table for group %u not in group "
b31e1552 1796 "(block %llu)!", i, inode_table);
ac27a0ec
DK		 1797 return 0;
1798 }
955ce5f5 1799 ext4_lock_group(sb, i);
717d50e4 1800 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
b31e1552
ES		 1801 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1802 "Checksum for group %u failed (%u!=%u)",
1803 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1804 gdp)), le16_to_cpu(gdp->bg_checksum));
7ee1ec4c 1805 if (!(sb->s_flags & MS_RDONLY)) {
955ce5f5 1806 ext4_unlock_group(sb, i);
8a266467 1807 return 0;
7ee1ec4c 1808 }
717d50e4 1809 }
955ce5f5 1810 ext4_unlock_group(sb, i);
ce421581
JS		 1811 if (!flexbg_flag)
1812 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec
DK		 1813 }
1814
bd81d8ee 1815 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
0b8e58a1 1816 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
ac27a0ec
DK		 1817 return 1;
1818}
1819
617ba13b 1820/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 1821 * the superblock) which were deleted from all directories, but held open by
1822 * a process at the time of a crash. We walk the list and try to delete these
1823 * inodes at recovery time (only with a read-write filesystem).
1824 *
1825 * In order to keep the orphan inode chain consistent during traversal (in
1826 * case of crash during recovery), we link each inode into the superblock
1827 * orphan list_head and handle it the same way as an inode deletion during
1828 * normal operation (which journals the operations for us).
1829 *
1830 * We only do an iget() and an iput() on each inode, which is very safe if we
1831 * accidentally point at an in-use or already deleted inode. The worst that
1832 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 1833 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 1834 * e2fsck was run on this filesystem, and it must have already done the orphan
1835 * inode cleanup for us, so we can safely abort without any further action.
1836 */
2b2d6d01
TT		 1837static void ext4_orphan_cleanup(struct super_block *sb,
1838 struct ext4_super_block *es)
ac27a0ec
DK		 1839{
1840 unsigned int s_flags = sb->s_flags;
1841 int nr_orphans = 0, nr_truncates = 0;
1842#ifdef CONFIG_QUOTA
1843 int i;
1844#endif
1845 if (!es->s_last_orphan) {
1846 jbd_debug(4, "no orphan inodes to clean up\n");
1847 return;
1848 }
1849
a8f48a95 1850 if (bdev_read_only(sb->s_bdev)) {
b31e1552
ES		 1851 ext4_msg(sb, KERN_ERR, "write access "
1852 "unavailable, skipping orphan cleanup");
a8f48a95
ES		 1853 return;
1854 }
1855
617ba13b 1856 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
ac27a0ec
DK		 1857 if (es->s_last_orphan)
1858 jbd_debug(1, "Errors on filesystem, "
1859 "clearing orphan list.\n");
1860 es->s_last_orphan = 0;
1861 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1862 return;
1863 }
1864
1865 if (s_flags & MS_RDONLY) {
b31e1552 1866 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
ac27a0ec
DK		 1867 sb->s_flags &= ~MS_RDONLY;
1868 }
1869#ifdef CONFIG_QUOTA
1870 /* Needed for iput() to work correctly and not trash data */
1871 sb->s_flags |= MS_ACTIVE;
1872 /* Turn on quotas so that they are updated correctly */
1873 for (i = 0; i < MAXQUOTAS; i++) {
617ba13b
MC		 1874 if (EXT4_SB(sb)->s_qf_names[i]) {
1875 int ret = ext4_quota_on_mount(sb, i);
ac27a0ec 1876 if (ret < 0)
b31e1552
ES		 1877 ext4_msg(sb, KERN_ERR,
1878 "Cannot turn on journaled "
1879 "quota: error %d", ret);
ac27a0ec
DK		 1880 }
1881 }
1882#endif
1883
1884 while (es->s_last_orphan) {
1885 struct inode *inode;
1886
97bd42b9
JB		 1887 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1888 if (IS_ERR(inode)) {
ac27a0ec
DK		 1889 es->s_last_orphan = 0;
1890 break;
1891 }
1892
617ba13b 1893 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
a269eb18 1894 vfs_dq_init(inode);
ac27a0ec 1895 if (inode->i_nlink) {
b31e1552
ES		 1896 ext4_msg(sb, KERN_DEBUG,
1897 "%s: truncating inode %lu to %lld bytes",
46e665e9 1898 __func__, inode->i_ino, inode->i_size);
e5f8eab8 1899 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
ac27a0ec 1900 inode->i_ino, inode->i_size);
617ba13b 1901 ext4_truncate(inode);
ac27a0ec
DK		 1902 nr_truncates++;
1903 } else {
b31e1552
ES		 1904 ext4_msg(sb, KERN_DEBUG,
1905 "%s: deleting unreferenced inode %lu",
46e665e9 1906 __func__, inode->i_ino);
ac27a0ec
DK		 1907 jbd_debug(2, "deleting unreferenced inode %lu\n",
1908 inode->i_ino);
1909 nr_orphans++;
1910 }
1911 iput(inode); /* The delete magic happens here! */
1912 }
1913
2b2d6d01 1914#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
ac27a0ec
DK		 1915
1916 if (nr_orphans)
b31e1552
ES		 1917 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
1918 PLURAL(nr_orphans));
ac27a0ec 1919 if (nr_truncates)
b31e1552
ES		 1920 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
1921 PLURAL(nr_truncates));
ac27a0ec
DK		 1922#ifdef CONFIG_QUOTA
1923 /* Turn quotas off */
1924 for (i = 0; i < MAXQUOTAS; i++) {
1925 if (sb_dqopt(sb)->files[i])
6f28e087 1926 vfs_quota_off(sb, i, 0);
ac27a0ec
DK		 1927 }
1928#endif
1929 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1930}
0b8e58a1 1931
cd2291a4
ES		 1932/*
1933 * Maximal extent format file size.
1934 * Resulting logical blkno at s_maxbytes must fit in our on-disk
1935 * extent format containers, within a sector_t, and within i_blocks
1936 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
1937 * so that won't be a limiting factor.
1938 *
1939 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1940 */
f287a1a5 1941static loff_t ext4_max_size(int blkbits, int has_huge_files)
cd2291a4
ES		 1942{
1943 loff_t res;
1944 loff_t upper_limit = MAX_LFS_FILESIZE;
1945
1946 /* small i_blocks in vfs inode? */
f287a1a5 1947 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
cd2291a4 1948 /*
90c699a9 1949 * CONFIG_LBDAF is not enabled implies the inode
cd2291a4
ES		 1950 * i_block represent total blocks in 512 bytes
1951 * 32 == size of vfs inode i_blocks * 8
1952 */
1953 upper_limit = (1LL << 32) - 1;
1954
1955 /* total blocks in file system block size */
1956 upper_limit >>= (blkbits - 9);
1957 upper_limit <<= blkbits;
1958 }
1959
1960 /* 32-bit extent-start container, ee_block */
1961 res = 1LL << 32;
1962 res <<= blkbits;
1963 res -= 1;
1964
1965 /* Sanity check against vm- & vfs- imposed limits */
1966 if (res > upper_limit)
1967 res = upper_limit;
1968
1969 return res;
1970}
ac27a0ec 1971
ac27a0ec 1972/*
cd2291a4 1973 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 1974 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
1975 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 1976 */
f287a1a5 1977static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
ac27a0ec 1978{
617ba13b 1979 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 1980 int meta_blocks;
1981 loff_t upper_limit;
0b8e58a1
AD		 1982 /* This is calculated to be the largest file size for a dense, block
1983 * mapped file such that the file's total number of 512-byte sectors,
1984 * including data and all indirect blocks, does not exceed (2^48 - 1).
1985 *
1986 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
1987 * number of 512-byte sectors of the file.
0fc1b451
AK		 1988 */
1989
f287a1a5 1990 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
0fc1b451 1991 /*
90c699a9 1992 * !has_huge_files or CONFIG_LBDAF not enabled implies that
0b8e58a1
AD		 1993 * the inode i_block field represents total file blocks in
1994 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
0fc1b451
AK		 1995 */
1996 upper_limit = (1LL << 32) - 1;
1997
1998 /* total blocks in file system block size */
1999 upper_limit >>= (bits - 9);
2000
2001 } else {
8180a562
AK		 2002 /*
2003 * We use 48 bit ext4_inode i_blocks
2004 * With EXT4_HUGE_FILE_FL set the i_blocks
2005 * represent total number of blocks in
2006 * file system block size
2007 */
0fc1b451
AK		 2008 upper_limit = (1LL << 48) - 1;
2009
0fc1b451
AK		 2010 }
2011
2012 /* indirect blocks */
2013 meta_blocks = 1;
2014 /* double indirect blocks */
2015 meta_blocks += 1 + (1LL << (bits-2));
2016 /* tripple indirect blocks */
2017 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2018
2019 upper_limit -= meta_blocks;
2020 upper_limit <<= bits;
ac27a0ec
DK		 2021
2022 res += 1LL << (bits-2);
2023 res += 1LL << (2*(bits-2));
2024 res += 1LL << (3*(bits-2));
2025 res <<= bits;
2026 if (res > upper_limit)
2027 res = upper_limit;
0fc1b451
AK		 2028
2029 if (res > MAX_LFS_FILESIZE)
2030 res = MAX_LFS_FILESIZE;
2031
ac27a0ec
DK		 2032 return res;
2033}
2034
617ba13b 2035static ext4_fsblk_t descriptor_loc(struct super_block *sb,
0b8e58a1 2036 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 2037{
617ba13b 2038 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 2039 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 2040 int has_super = 0;
2041
2042 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2043
617ba13b 2044 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
ac27a0ec 2045 nr < first_meta_bg)
70bbb3e0 2046 return logical_sb_block + nr + 1;
ac27a0ec 2047 bg = sbi->s_desc_per_block * nr;
617ba13b 2048 if (ext4_bg_has_super(sb, bg))
ac27a0ec 2049 has_super = 1;
0b8e58a1 2050
617ba13b 2051 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 2052}
2053
c9de560d
AT		 2054/**
2055 * ext4_get_stripe_size: Get the stripe size.
2056 * @sbi: In memory super block info
2057 *
2058 * If we have specified it via mount option, then
2059 * use the mount option value. If the value specified at mount time is
2060 * greater than the blocks per group use the super block value.
2061 * If the super block value is greater than blocks per group return 0.
2062 * Allocator needs it be less than blocks per group.
2063 *
2064 */
2065static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2066{
2067 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2068 unsigned long stripe_width =
2069 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2070
2071 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2072 return sbi->s_stripe;
2073
2074 if (stripe_width <= sbi->s_blocks_per_group)
2075 return stripe_width;
2076
2077 if (stride <= sbi->s_blocks_per_group)
2078 return stride;
2079
2080 return 0;
2081}
ac27a0ec 2082
3197ebdb
TT		 2083/* sysfs supprt */
2084
2085struct ext4_attr {
2086 struct attribute attr;
2087 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2088 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2089 const char *, size_t);
2090 int offset;
2091};
2092
2093static int parse_strtoul(const char *buf,
2094 unsigned long max, unsigned long *value)
2095{
2096 char *endp;
2097
2098 while (*buf && isspace(*buf))
2099 buf++;
2100 *value = simple_strtoul(buf, &endp, 0);
2101 while (*endp && isspace(*endp))
2102 endp++;
2103 if (*endp || *value > max)
2104 return -EINVAL;
2105
2106 return 0;
2107}
2108
2109static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2110 struct ext4_sb_info *sbi,
2111 char *buf)
2112{
2113 return snprintf(buf, PAGE_SIZE, "%llu\n",
2114 (s64) percpu_counter_sum(&sbi->s_dirtyblocks_counter));
2115}
2116
2117static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2118 struct ext4_sb_info *sbi, char *buf)
2119{
2120 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2121
2122 return snprintf(buf, PAGE_SIZE, "%lu\n",
2123 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2124 sbi->s_sectors_written_start) >> 1);
2125}
2126
2127static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2128 struct ext4_sb_info *sbi, char *buf)
2129{
2130 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2131
2132 return snprintf(buf, PAGE_SIZE, "%llu\n",
2133 sbi->s_kbytes_written +
2134 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2135 EXT4_SB(sb)->s_sectors_written_start) >> 1));
2136}
2137
2138static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2139 struct ext4_sb_info *sbi,
2140 const char *buf, size_t count)
2141{
2142 unsigned long t;
2143
2144 if (parse_strtoul(buf, 0x40000000, &t))
2145 return -EINVAL;
2146
f7c43950 2147 if (!is_power_of_2(t))
3197ebdb
TT		 2148 return -EINVAL;
2149
2150 sbi->s_inode_readahead_blks = t;
2151 return count;
2152}
2153
2154static ssize_t sbi_ui_show(struct ext4_attr *a,
0b8e58a1 2155 struct ext4_sb_info *sbi, char *buf)
3197ebdb
TT		 2156{
2157 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2158
2159 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2160}
2161
2162static ssize_t sbi_ui_store(struct ext4_attr *a,
2163 struct ext4_sb_info *sbi,
2164 const char *buf, size_t count)
2165{
2166 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2167 unsigned long t;
2168
2169 if (parse_strtoul(buf, 0xffffffff, &t))
2170 return -EINVAL;
2171 *ui = t;
2172 return count;
2173}
2174
2175#define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2176static struct ext4_attr ext4_attr_##_name = { \
2177 .attr = {.name = __stringify(_name), .mode = _mode }, \
2178 .show = _show, \
2179 .store = _store, \
2180 .offset = offsetof(struct ext4_sb_info, _elname), \
2181}
2182#define EXT4_ATTR(name, mode, show, store) \
2183static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2184
2185#define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2186#define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2187#define EXT4_RW_ATTR_SBI_UI(name, elname) \
2188 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2189#define ATTR_LIST(name) &ext4_attr_##name.attr
2190
2191EXT4_RO_ATTR(delayed_allocation_blocks);
2192EXT4_RO_ATTR(session_write_kbytes);
2193EXT4_RO_ATTR(lifetime_write_kbytes);
2194EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2195 inode_readahead_blks_store, s_inode_readahead_blks);
11013911 2196EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
3197ebdb
TT		 2197EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2198EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2199EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2200EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2201EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2202EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
55138e0b 2203EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
3197ebdb
TT		 2204
2205static struct attribute *ext4_attrs[] = {
2206 ATTR_LIST(delayed_allocation_blocks),
2207 ATTR_LIST(session_write_kbytes),
2208 ATTR_LIST(lifetime_write_kbytes),
2209 ATTR_LIST(inode_readahead_blks),
11013911 2210 ATTR_LIST(inode_goal),
3197ebdb
TT		 2211 ATTR_LIST(mb_stats),
2212 ATTR_LIST(mb_max_to_scan),
2213 ATTR_LIST(mb_min_to_scan),
2214 ATTR_LIST(mb_order2_req),
2215 ATTR_LIST(mb_stream_req),
2216 ATTR_LIST(mb_group_prealloc),
55138e0b 2217 ATTR_LIST(max_writeback_mb_bump),
3197ebdb
TT		 2218 NULL,
2219};
2220
2221static ssize_t ext4_attr_show(struct kobject *kobj,
2222 struct attribute *attr, char *buf)
2223{
2224 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2225 s_kobj);
2226 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2227
2228 return a->show ? a->show(a, sbi, buf) : 0;
2229}
2230
2231static ssize_t ext4_attr_store(struct kobject *kobj,
2232 struct attribute *attr,
2233 const char *buf, size_t len)
2234{
2235 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2236 s_kobj);
2237 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2238
2239 return a->store ? a->store(a, sbi, buf, len) : 0;
2240}
2241
2242static void ext4_sb_release(struct kobject *kobj)
2243{
2244 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2245 s_kobj);
2246 complete(&sbi->s_kobj_unregister);
2247}
2248
2249
2250static struct sysfs_ops ext4_attr_ops = {
2251 .show = ext4_attr_show,
2252 .store = ext4_attr_store,
2253};
2254
2255static struct kobj_type ext4_ktype = {
2256 .default_attrs = ext4_attrs,
2257 .sysfs_ops = &ext4_attr_ops,
2258 .release = ext4_sb_release,
2259};
2260
a13fb1a4
ES		 2261/*
2262 * Check whether this filesystem can be mounted based on
2263 * the features present and the RDONLY/RDWR mount requested.
2264 * Returns 1 if this filesystem can be mounted as requested,
2265 * 0 if it cannot be.
2266 */
2267static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2268{
2269 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2270 ext4_msg(sb, KERN_ERR,
2271 "Couldn't mount because of "
2272 "unsupported optional features (%x)",
2273 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2274 ~EXT4_FEATURE_INCOMPAT_SUPP));
2275 return 0;
2276 }
2277
2278 if (readonly)
2279 return 1;
2280
2281 /* Check that feature set is OK for a read-write mount */
2282 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2283 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2284 "unsupported optional features (%x)",
2285 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2286 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2287 return 0;
2288 }
2289 /*
2290 * Large file size enabled file system can only be mounted
2291 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2292 */
2293 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2294 if (sizeof(blkcnt_t) < sizeof(u64)) {
2295 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2296 "cannot be mounted RDWR without "
2297 "CONFIG_LBDAF");
2298 return 0;
2299 }
2300 }
2301 return 1;
2302}
2303
2b2d6d01 2304static int ext4_fill_super(struct super_block *sb, void *data, int silent)
7477827f
AK		 2305 __releases(kernel_lock)
2306 __acquires(kernel_lock)
ac27a0ec 2307{
2b2d6d01 2308 struct buffer_head *bh;
617ba13b
MC		 2309 struct ext4_super_block *es = NULL;
2310 struct ext4_sb_info *sbi;
2311 ext4_fsblk_t block;
2312 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 2313 ext4_fsblk_t logical_sb_block;
ac27a0ec 2314 unsigned long offset = 0;
ac27a0ec
DK		 2315 unsigned long journal_devnum = 0;
2316 unsigned long def_mount_opts;
2317 struct inode *root;
9f6200bb 2318 char *cp;
0390131b 2319 const char *descr;
1d1fe1ee 2320 int ret = -EINVAL;
ac27a0ec 2321 int blocksize;
4ec11028
TT		 2322 unsigned int db_count;
2323 unsigned int i;
f287a1a5 2324 int needs_recovery, has_huge_files;
bd81d8ee 2325 __u64 blocks_count;
833f4077 2326 int err;
b3881f74 2327 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
ac27a0ec
DK		 2328
2329 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
2330 if (!sbi)
2331 return -ENOMEM;
705895b6
PE		 2332
2333 sbi->s_blockgroup_lock =
2334 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
2335 if (!sbi->s_blockgroup_lock) {
2336 kfree(sbi);
2337 return -ENOMEM;
2338 }
ac27a0ec
DK		 2339 sb->s_fs_info = sbi;
2340 sbi->s_mount_opt = 0;
617ba13b
MC		 2341 sbi->s_resuid = EXT4_DEF_RESUID;
2342 sbi->s_resgid = EXT4_DEF_RESGID;
240799cd 2343 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
d9c9bef1 2344 sbi->s_sb_block = sb_block;
afc32f7e
TT		 2345 sbi->s_sectors_written_start = part_stat_read(sb->s_bdev->bd_part,
2346 sectors[1]);
ac27a0ec
DK		 2347
2348 unlock_kernel();
2349
9f6200bb
TT		 2350 /* Cleanup superblock name */
2351 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
2352 *cp = '!';
2353
617ba13b 2354 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 2355 if (!blocksize) {
b31e1552 2356 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
ac27a0ec
DK		 2357 goto out_fail;
2358 }
2359
2360 /*
617ba13b 2361 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 2362 * block sizes. We need to calculate the offset from buffer start.
2363 */
617ba13b 2364 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 2365 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2366 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 2367 } else {
70bbb3e0 2368 logical_sb_block = sb_block;
ac27a0ec
DK		 2369 }
2370
70bbb3e0 2371 if (!(bh = sb_bread(sb, logical_sb_block))) {
b31e1552 2372 ext4_msg(sb, KERN_ERR, "unable to read superblock");
ac27a0ec
DK		 2373 goto out_fail;
2374 }
2375 /*
2376 * Note: s_es must be initialized as soon as possible because
617ba13b 2377 * some ext4 macro-instructions depend on its value
ac27a0ec 2378 */
617ba13b 2379 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
ac27a0ec
DK		 2380 sbi->s_es = es;
2381 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 2382 if (sb->s_magic != EXT4_SUPER_MAGIC)
2383 goto cantfind_ext4;
afc32f7e 2384 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
ac27a0ec
DK		 2385
2386 /* Set defaults before we parse the mount options */
2387 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
617ba13b 2388 if (def_mount_opts & EXT4_DEFM_DEBUG)
ac27a0ec 2389 set_opt(sbi->s_mount_opt, DEBUG);
617ba13b 2390 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
ac27a0ec 2391 set_opt(sbi->s_mount_opt, GRPID);
617ba13b 2392 if (def_mount_opts & EXT4_DEFM_UID16)
ac27a0ec 2393 set_opt(sbi->s_mount_opt, NO_UID32);
03010a33 2394#ifdef CONFIG_EXT4_FS_XATTR
617ba13b 2395 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
ac27a0ec 2396 set_opt(sbi->s_mount_opt, XATTR_USER);
2e7842b8 2397#endif
03010a33 2398#ifdef CONFIG_EXT4_FS_POSIX_ACL
617ba13b 2399 if (def_mount_opts & EXT4_DEFM_ACL)
ac27a0ec 2400 set_opt(sbi->s_mount_opt, POSIX_ACL);
2e7842b8 2401#endif
617ba13b
MC		 2402 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
2403 sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
2404 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
2405 sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
2406 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
2407 sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
2408
2409 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
ac27a0ec 2410 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
bb4f397a 2411 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
ceea16bf 2412 set_opt(sbi->s_mount_opt, ERRORS_CONT);
bb4f397a
AK		 2413 else
2414 set_opt(sbi->s_mount_opt, ERRORS_RO);
ac27a0ec
DK		 2415
2416 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
2417 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
30773840
TT		 2418 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
2419 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
2420 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
f4033903 2421 sbi->s_mb_history_max = default_mb_history_length;
ac27a0ec 2422
571640ca 2423 set_opt(sbi->s_mount_opt, BARRIER);
ac27a0ec 2424
dd919b98
AK		 2425 /*
2426 * enable delayed allocation by default
2427 * Use -o nodelalloc to turn it off
2428 */
2429 set_opt(sbi->s_mount_opt, DELALLOC);
2430
b3881f74
TT		 2431 if (!parse_options((char *) data, sb, &journal_devnum,
2432 &journal_ioprio, NULL, 0))
ac27a0ec
DK		 2433 goto failed_mount;
2434
2435 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
617ba13b 2436 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec 2437
617ba13b
MC		 2438 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
2439 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
2440 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
2441 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
b31e1552
ES		 2442 ext4_msg(sb, KERN_WARNING,
2443 "feature flags set on rev 0 fs, "
2444 "running e2fsck is recommended");
469108ff 2445
ac27a0ec
DK		 2446 /*
2447 * Check feature flags regardless of the revision level, since we
2448 * previously didn't change the revision level when setting the flags,
2449 * so there is a chance incompat flags are set on a rev 0 filesystem.
2450 */
a13fb1a4 2451 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
ac27a0ec 2452 goto failed_mount;
a13fb1a4 2453
ac27a0ec
DK		 2454 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
2455
617ba13b
MC		 2456 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
2457 blocksize > EXT4_MAX_BLOCK_SIZE) {
b31e1552
ES		 2458 ext4_msg(sb, KERN_ERR,
2459 "Unsupported filesystem blocksize %d", blocksize);
ac27a0ec
DK		 2460 goto failed_mount;
2461 }
2462
ac27a0ec 2463 if (sb->s_blocksize != blocksize) {
ce40733c
AK		 2464 /* Validate the filesystem blocksize */
2465 if (!sb_set_blocksize(sb, blocksize)) {
b31e1552 2466 ext4_msg(sb, KERN_ERR, "bad block size %d",
ce40733c 2467 blocksize);
ac27a0ec
DK		 2468 goto failed_mount;
2469 }
2470
2b2d6d01 2471 brelse(bh);
70bbb3e0
AM		 2472 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2473 offset = do_div(logical_sb_block, blocksize);
2474 bh = sb_bread(sb, logical_sb_block);
ac27a0ec 2475 if (!bh) {
b31e1552
ES		 2476 ext4_msg(sb, KERN_ERR,
2477 "Can't read superblock on 2nd try");
ac27a0ec
DK		 2478 goto failed_mount;
2479 }
617ba13b 2480 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
ac27a0ec 2481 sbi->s_es = es;
617ba13b 2482 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
b31e1552
ES		 2483 ext4_msg(sb, KERN_ERR,
2484 "Magic mismatch, very weird!");
ac27a0ec
DK		 2485 goto failed_mount;
2486 }
2487 }
2488
a13fb1a4
ES		 2489 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2490 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
f287a1a5
TT		 2491 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
2492 has_huge_files);
2493 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
ac27a0ec 2494
617ba13b
MC		 2495 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2496 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2497 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
ac27a0ec
DK		 2498 } else {
2499 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2500 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
617ba13b 2501 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1330593e 2502 (!is_power_of_2(sbi->s_inode_size)) ||
ac27a0ec 2503 (sbi->s_inode_size > blocksize)) {
b31e1552
ES		 2504 ext4_msg(sb, KERN_ERR,
2505 "unsupported inode size: %d",
2b2d6d01 2506 sbi->s_inode_size);
ac27a0ec
DK		 2507 goto failed_mount;
2508 }
ef7f3835
KS		 2509 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2510 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
ac27a0ec 2511 }
0b8e58a1 2512
0d1ee42f
AR		 2513 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2514 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
8fadc143 2515 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 2516 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 2517 !is_power_of_2(sbi->s_desc_size)) {
b31e1552
ES		 2518 ext4_msg(sb, KERN_ERR,
2519 "unsupported descriptor size %lu",
0d1ee42f
AR		 2520 sbi->s_desc_size);
2521 goto failed_mount;
2522 }
2523 } else
2524 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
0b8e58a1 2525
ac27a0ec 2526 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 2527 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
b47b6f38 2528 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
617ba13b 2529 goto cantfind_ext4;
0b8e58a1 2530
617ba13b 2531 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 2532 if (sbi->s_inodes_per_block == 0)
617ba13b 2533 goto cantfind_ext4;
ac27a0ec
DK		 2534 sbi->s_itb_per_group = sbi->s_inodes_per_group /
2535 sbi->s_inodes_per_block;
0d1ee42f 2536 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 2537 sbi->s_sbh = bh;
2538 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 2539 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2540 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
0b8e58a1 2541
2b2d6d01 2542 for (i = 0; i < 4; i++)
ac27a0ec
DK		 2543 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2544 sbi->s_def_hash_version = es->s_def_hash_version;
f99b2589
TT		 2545 i = le32_to_cpu(es->s_flags);
2546 if (i & EXT2_FLAGS_UNSIGNED_HASH)
2547 sbi->s_hash_unsigned = 3;
2548 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
2549#ifdef __CHAR_UNSIGNED__
2550 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
2551 sbi->s_hash_unsigned = 3;
2552#else
2553 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
2554#endif
2555 sb->s_dirt = 1;
2556 }
ac27a0ec
DK		 2557
2558 if (sbi->s_blocks_per_group > blocksize * 8) {
b31e1552
ES		 2559 ext4_msg(sb, KERN_ERR,
2560 "#blocks per group too big: %lu",
2b2d6d01 2561 sbi->s_blocks_per_group);
ac27a0ec
DK		 2562 goto failed_mount;
2563 }
ac27a0ec 2564 if (sbi->s_inodes_per_group > blocksize * 8) {
b31e1552
ES		 2565 ext4_msg(sb, KERN_ERR,
2566 "#inodes per group too big: %lu",
2b2d6d01 2567 sbi->s_inodes_per_group);
ac27a0ec
DK		 2568 goto failed_mount;
2569 }
2570
bf43d84b
ES		 2571 /*
2572 * Test whether we have more sectors than will fit in sector_t,
2573 * and whether the max offset is addressable by the page cache.
2574 */
2575 if ((ext4_blocks_count(es) >
2576 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) ||
2577 (ext4_blocks_count(es) >
2578 (pgoff_t)(~0ULL) >> (PAGE_CACHE_SHIFT - sb->s_blocksize_bits))) {
b31e1552 2579 ext4_msg(sb, KERN_ERR, "filesystem"
bf43d84b 2580 " too large to mount safely on this system");
ac27a0ec 2581 if (sizeof(sector_t) < 8)
90c699a9 2582 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
bf43d84b 2583 ret = -EFBIG;
ac27a0ec
DK		 2584 goto failed_mount;
2585 }
2586
617ba13b
MC		 2587 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2588 goto cantfind_ext4;
e7c95593 2589
0f2ddca6
FTN		 2590 /* check blocks count against device size */
2591 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
2592 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
b31e1552
ES		 2593 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
2594 "exceeds size of device (%llu blocks)",
0f2ddca6
FTN		 2595 ext4_blocks_count(es), blocks_count);
2596 goto failed_mount;
2597 }
2598
0b8e58a1
AD		 2599 /*
2600 * It makes no sense for the first data block to be beyond the end
2601 * of the filesystem.
2602 */
2603 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
b31e1552
ES		 2604 ext4_msg(sb, KERN_WARNING, "bad geometry: first data"
2605 "block %u is beyond end of filesystem (%llu)",
2606 le32_to_cpu(es->s_first_data_block),
2607 ext4_blocks_count(es));
e7c95593
ES		 2608 goto failed_mount;
2609 }
bd81d8ee
LV		 2610 blocks_count = (ext4_blocks_count(es) -
2611 le32_to_cpu(es->s_first_data_block) +
2612 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2613 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4ec11028 2614 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
b31e1552 2615 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
4ec11028 2616 "(block count %llu, first data block %u, "
b31e1552 2617 "blocks per group %lu)", sbi->s_groups_count,
4ec11028
TT		 2618 ext4_blocks_count(es),
2619 le32_to_cpu(es->s_first_data_block),
2620 EXT4_BLOCKS_PER_GROUP(sb));
2621 goto failed_mount;
2622 }
bd81d8ee 2623 sbi->s_groups_count = blocks_count;
fb0a387d
ES		 2624 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
2625 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
617ba13b
MC		 2626 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2627 EXT4_DESC_PER_BLOCK(sb);
2b2d6d01 2628 sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
ac27a0ec
DK		 2629 GFP_KERNEL);
2630 if (sbi->s_group_desc == NULL) {
b31e1552 2631 ext4_msg(sb, KERN_ERR, "not enough memory");
ac27a0ec
DK		 2632 goto failed_mount;
2633 }
2634
3244fcb1 2635#ifdef CONFIG_PROC_FS
9f6200bb
TT		 2636 if (ext4_proc_root)
2637 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
3244fcb1 2638#endif
240799cd 2639
705895b6 2640 bgl_lock_init(sbi->s_blockgroup_lock);
ac27a0ec
DK		 2641
2642 for (i = 0; i < db_count; i++) {
70bbb3e0 2643 block = descriptor_loc(sb, logical_sb_block, i);
ac27a0ec
DK		 2644 sbi->s_group_desc[i] = sb_bread(sb, block);
2645 if (!sbi->s_group_desc[i]) {
b31e1552
ES		 2646 ext4_msg(sb, KERN_ERR,
2647 "can't read group descriptor %d", i);
ac27a0ec
DK		 2648 db_count = i;
2649 goto failed_mount2;
2650 }
2651 }
2b2d6d01 2652 if (!ext4_check_descriptors(sb)) {
b31e1552 2653 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
ac27a0ec
DK		 2654 goto failed_mount2;
2655 }
772cb7c8
JS		 2656 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2657 if (!ext4_fill_flex_info(sb)) {
b31e1552
ES		 2658 ext4_msg(sb, KERN_ERR,
2659 "unable to initialize "
2660 "flex_bg meta info!");
772cb7c8
JS		 2661 goto failed_mount2;
2662 }
2663
ac27a0ec
DK		 2664 sbi->s_gdb_count = db_count;
2665 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2666 spin_lock_init(&sbi->s_next_gen_lock);
2667
833f4077
PZ		 2668 err = percpu_counter_init(&sbi->s_freeblocks_counter,
2669 ext4_count_free_blocks(sb));
2670 if (!err) {
2671 err = percpu_counter_init(&sbi->s_freeinodes_counter,
2672 ext4_count_free_inodes(sb));
2673 }
2674 if (!err) {
2675 err = percpu_counter_init(&sbi->s_dirs_counter,
2676 ext4_count_dirs(sb));
2677 }
6bc6e63f
AK		 2678 if (!err) {
2679 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
2680 }
833f4077 2681 if (err) {
b31e1552 2682 ext4_msg(sb, KERN_ERR, "insufficient memory");
833f4077
PZ		 2683 goto failed_mount3;
2684 }
ac27a0ec 2685
c9de560d 2686 sbi->s_stripe = ext4_get_stripe_size(sbi);
55138e0b 2687 sbi->s_max_writeback_mb_bump = 128;
c9de560d 2688
ac27a0ec
DK		 2689 /*
2690 * set up enough so that it can read an inode
2691 */
9ca92389
TT		 2692 if (!test_opt(sb, NOLOAD) &&
2693 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
2694 sb->s_op = &ext4_sops;
2695 else
2696 sb->s_op = &ext4_nojournal_sops;
617ba13b
MC		 2697 sb->s_export_op = &ext4_export_ops;
2698 sb->s_xattr = ext4_xattr_handlers;
ac27a0ec 2699#ifdef CONFIG_QUOTA
617ba13b
MC		 2700 sb->s_qcop = &ext4_qctl_operations;
2701 sb->dq_op = &ext4_quota_operations;
ac27a0ec
DK		 2702#endif
2703 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3b9d4ed2 2704 mutex_init(&sbi->s_orphan_lock);
32ed5058 2705 mutex_init(&sbi->s_resize_lock);
ac27a0ec
DK		 2706
2707 sb->s_root = NULL;
2708
2709 needs_recovery = (es->s_last_orphan != 0 ||
617ba13b
MC		 2710 EXT4_HAS_INCOMPAT_FEATURE(sb,
2711 EXT4_FEATURE_INCOMPAT_RECOVER));
ac27a0ec
DK		 2712
2713 /*
2714 * The first inode we look at is the journal inode. Don't try
2715 * root first: it may be modified in the journal!
2716 */
2717 if (!test_opt(sb, NOLOAD) &&
617ba13b
MC		 2718 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2719 if (ext4_load_journal(sb, es, journal_devnum))
ac27a0ec 2720 goto failed_mount3;
624080ed
TT		 2721 if (!(sb->s_flags & MS_RDONLY) &&
2722 EXT4_SB(sb)->s_journal->j_failed_commit) {
b31e1552 2723 ext4_msg(sb, KERN_CRIT, "error: "
624080ed 2724 "ext4_fill_super: Journal transaction "
b31e1552 2725 "%u is corrupt",
624080ed 2726 EXT4_SB(sb)->s_journal->j_failed_commit);
2b2d6d01 2727 if (test_opt(sb, ERRORS_RO)) {
b31e1552
ES		 2728 ext4_msg(sb, KERN_CRIT,
2729 "Mounting filesystem read-only");
624080ed
TT		 2730 sb->s_flags |= MS_RDONLY;
2731 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2732 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
2733 }
2734 if (test_opt(sb, ERRORS_PANIC)) {
2735 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
2736 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
e2d67052 2737 ext4_commit_super(sb, 1);
624080ed
TT		 2738 goto failed_mount4;
2739 }
2740 }
0390131b
FM		 2741 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
2742 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
b31e1552
ES		 2743 ext4_msg(sb, KERN_ERR, "required journal recovery "
2744 "suppressed and not mounted read-only");
0390131b 2745 goto failed_mount4;
ac27a0ec 2746 } else {
0390131b
FM		 2747 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
2748 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
2749 sbi->s_journal = NULL;
2750 needs_recovery = 0;
2751 goto no_journal;
ac27a0ec
DK		 2752 }
2753
eb40a09c
JS		 2754 if (ext4_blocks_count(es) > 0xffffffffULL &&
2755 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2756 JBD2_FEATURE_INCOMPAT_64BIT)) {
b31e1552 2757 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
eb40a09c
JS		 2758 goto failed_mount4;
2759 }
2760
d0646f7b
TT		 2761 jbd2_journal_set_features(sbi->s_journal,
2762 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
2763 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
2764 jbd2_journal_set_features(sbi->s_journal, 0, 0,
818d276c 2765 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
d0646f7b 2766 else
818d276c
GS		 2767 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2768 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
818d276c 2769
ac27a0ec
DK		 2770 /* We have now updated the journal if required, so we can
2771 * validate the data journaling mode. */
2772 switch (test_opt(sb, DATA_FLAGS)) {
2773 case 0:
2774 /* No mode set, assume a default based on the journal
63f57933
AM		 2775 * capabilities: ORDERED_DATA if the journal can
2776 * cope, else JOURNAL_DATA
2777 */
dab291af
MC		 2778 if (jbd2_journal_check_available_features
2779 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
ac27a0ec
DK		 2780 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2781 else
2782 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2783 break;
2784
617ba13b
MC		 2785 case EXT4_MOUNT_ORDERED_DATA:
2786 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 2787 if (!jbd2_journal_check_available_features
2788 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
b31e1552
ES		 2789 ext4_msg(sb, KERN_ERR, "Journal does not support "
2790 "requested data journaling mode");
ac27a0ec
DK		 2791 goto failed_mount4;
2792 }
2793 default:
2794 break;
2795 }
b3881f74 2796 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
ac27a0ec 2797
0390131b 2798no_journal:
ac27a0ec
DK		 2799
2800 if (test_opt(sb, NOBH)) {
617ba13b 2801 if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
b31e1552
ES		 2802 ext4_msg(sb, KERN_WARNING, "Ignoring nobh option - "
2803 "its supported only with writeback mode");
ac27a0ec
DK		 2804 clear_opt(sbi->s_mount_opt, NOBH);
2805 }
2806 }
4c0425ff
MC		 2807 EXT4_SB(sb)->dio_unwritten_wq = create_workqueue("ext4-dio-unwritten");
2808 if (!EXT4_SB(sb)->dio_unwritten_wq) {
2809 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
2810 goto failed_mount_wq;
2811 }
2812
ac27a0ec 2813 /*
dab291af 2814 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 2815 * so we can safely mount the rest of the filesystem now.
2816 */
2817
1d1fe1ee
DH		 2818 root = ext4_iget(sb, EXT4_ROOT_INO);
2819 if (IS_ERR(root)) {
b31e1552 2820 ext4_msg(sb, KERN_ERR, "get root inode failed");
1d1fe1ee 2821 ret = PTR_ERR(root);
ac27a0ec
DK		 2822 goto failed_mount4;
2823 }
2824 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1d1fe1ee 2825 iput(root);
b31e1552 2826 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
ac27a0ec
DK		 2827 goto failed_mount4;
2828 }
1d1fe1ee
DH		 2829 sb->s_root = d_alloc_root(root);
2830 if (!sb->s_root) {
b31e1552 2831 ext4_msg(sb, KERN_ERR, "get root dentry failed");
1d1fe1ee
DH		 2832 iput(root);
2833 ret = -ENOMEM;
2834 goto failed_mount4;
2835 }
ac27a0ec 2836
2b2d6d01 2837 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
ef7f3835
KS		 2838
2839 /* determine the minimum size of new large inodes, if present */
2840 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2841 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2842 EXT4_GOOD_OLD_INODE_SIZE;
2843 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2844 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2845 if (sbi->s_want_extra_isize <
2846 le16_to_cpu(es->s_want_extra_isize))
2847 sbi->s_want_extra_isize =
2848 le16_to_cpu(es->s_want_extra_isize);
2849 if (sbi->s_want_extra_isize <
2850 le16_to_cpu(es->s_min_extra_isize))
2851 sbi->s_want_extra_isize =
2852 le16_to_cpu(es->s_min_extra_isize);
2853 }
2854 }
2855 /* Check if enough inode space is available */
2856 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2857 sbi->s_inode_size) {
2858 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2859 EXT4_GOOD_OLD_INODE_SIZE;
b31e1552
ES		 2860 ext4_msg(sb, KERN_INFO, "required extra inode space not"
2861 "available");
ef7f3835
KS		 2862 }
2863
c2774d84 2864 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
b31e1552
ES		 2865 ext4_msg(sb, KERN_WARNING, "Ignoring delalloc option - "
2866 "requested data journaling mode");
c2774d84
AK		 2867 clear_opt(sbi->s_mount_opt, DELALLOC);
2868 } else if (test_opt(sb, DELALLOC))
b31e1552 2869 ext4_msg(sb, KERN_INFO, "delayed allocation enabled");
c2774d84 2870
6fd058f7
TT		 2871 err = ext4_setup_system_zone(sb);
2872 if (err) {
b31e1552
ES		 2873 ext4_msg(sb, KERN_ERR, "failed to initialize system "
2874 "zone (%d)\n", err);
6fd058f7
TT		 2875 goto failed_mount4;
2876 }
2877
c2774d84
AK		 2878 ext4_ext_init(sb);
2879 err = ext4_mb_init(sb, needs_recovery);
2880 if (err) {
b31e1552
ES		 2881 ext4_msg(sb, KERN_ERR, "failed to initalize mballoc (%d)",
2882 err);
c2774d84
AK		 2883 goto failed_mount4;
2884 }
2885
3197ebdb
TT		 2886 sbi->s_kobj.kset = ext4_kset;
2887 init_completion(&sbi->s_kobj_unregister);
2888 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
2889 "%s", sb->s_id);
2890 if (err) {
2891 ext4_mb_release(sb);
2892 ext4_ext_release(sb);
2893 goto failed_mount4;
2894 };
2895
617ba13b
MC		 2896 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
2897 ext4_orphan_cleanup(sb, es);
2898 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
0390131b 2899 if (needs_recovery) {
b31e1552 2900 ext4_msg(sb, KERN_INFO, "recovery complete");
0390131b
FM		 2901 ext4_mark_recovery_complete(sb, es);
2902 }
2903 if (EXT4_SB(sb)->s_journal) {
2904 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2905 descr = " journalled data mode";
2906 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2907 descr = " ordered data mode";
2908 else
2909 descr = " writeback data mode";
2910 } else
2911 descr = "out journal";
2912
b31e1552 2913 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s", descr);
ac27a0ec
DK		 2914
2915 lock_kernel();
2916 return 0;
2917
617ba13b 2918cantfind_ext4:
ac27a0ec 2919 if (!silent)
b31e1552 2920 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
ac27a0ec
DK		 2921 goto failed_mount;
2922
2923failed_mount4:
b31e1552 2924 ext4_msg(sb, KERN_ERR, "mount failed");
4c0425ff
MC		 2925 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
2926failed_mount_wq:
6fd058f7 2927 ext4_release_system_zone(sb);
0390131b
FM		 2928 if (sbi->s_journal) {
2929 jbd2_journal_destroy(sbi->s_journal);
2930 sbi->s_journal = NULL;
2931 }
ac27a0ec 2932failed_mount3:
c5ca7c76
TT		 2933 if (sbi->s_flex_groups) {
2934 if (is_vmalloc_addr(sbi->s_flex_groups))
2935 vfree(sbi->s_flex_groups);
2936 else
2937 kfree(sbi->s_flex_groups);
2938 }
ac27a0ec
DK		 2939 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2940 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2941 percpu_counter_destroy(&sbi->s_dirs_counter);
6bc6e63f 2942 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
ac27a0ec
DK		 2943failed_mount2:
2944 for (i = 0; i < db_count; i++)
2945 brelse(sbi->s_group_desc[i]);
2946 kfree(sbi->s_group_desc);
2947failed_mount:
240799cd 2948 if (sbi->s_proc) {
9f6200bb 2949 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 2950 }
ac27a0ec
DK		 2951#ifdef CONFIG_QUOTA
2952 for (i = 0; i < MAXQUOTAS; i++)
2953 kfree(sbi->s_qf_names[i]);
2954#endif
617ba13b 2955 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 2956 brelse(bh);
2957out_fail:
2958 sb->s_fs_info = NULL;
f6830165 2959 kfree(sbi->s_blockgroup_lock);
ac27a0ec
DK		 2960 kfree(sbi);
2961 lock_kernel();
1d1fe1ee 2962 return ret;
ac27a0ec
DK		 2963}
2964
2965/*
2966 * Setup any per-fs journal parameters now. We'll do this both on
2967 * initial mount, once the journal has been initialised but before we've
2968 * done any recovery; and again on any subsequent remount.
2969 */
617ba13b 2970static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 2971{
617ba13b 2972 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 2973
30773840
TT		 2974 journal->j_commit_interval = sbi->s_commit_interval;
2975 journal->j_min_batch_time = sbi->s_min_batch_time;
2976 journal->j_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 2977
2978 spin_lock(&journal->j_state_lock);
2979 if (test_opt(sb, BARRIER))
dab291af 2980 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 2981 else
dab291af 2982 journal->j_flags &= ~JBD2_BARRIER;
5bf5683a
HK		 2983 if (test_opt(sb, DATA_ERR_ABORT))
2984 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
2985 else
2986 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
ac27a0ec
DK		 2987 spin_unlock(&journal->j_state_lock);
2988}
2989
617ba13b 2990static journal_t *ext4_get_journal(struct super_block *sb,
ac27a0ec
DK		 2991 unsigned int journal_inum)
2992{
2993 struct inode *journal_inode;
2994 journal_t *journal;
2995
0390131b
FM		 2996 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
2997
ac27a0ec
DK		 2998 /* First, test for the existence of a valid inode on disk. Bad
2999 * things happen if we iget() an unused inode, as the subsequent
3000 * iput() will try to delete it. */
3001
1d1fe1ee
DH		 3002 journal_inode = ext4_iget(sb, journal_inum);
3003 if (IS_ERR(journal_inode)) {
b31e1552 3004 ext4_msg(sb, KERN_ERR, "no journal found");
ac27a0ec
DK		 3005 return NULL;
3006 }
3007 if (!journal_inode->i_nlink) {
3008 make_bad_inode(journal_inode);
3009 iput(journal_inode);
b31e1552 3010 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
ac27a0ec
DK		 3011 return NULL;
3012 }
3013
e5f8eab8 3014 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
ac27a0ec 3015 journal_inode, journal_inode->i_size);
1d1fe1ee 3016 if (!S_ISREG(journal_inode->i_mode)) {
b31e1552 3017 ext4_msg(sb, KERN_ERR, "invalid journal inode");
ac27a0ec
DK		 3018 iput(journal_inode);
3019 return NULL;
3020 }
3021
dab291af 3022 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 3023 if (!journal) {
b31e1552 3024 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
ac27a0ec
DK		 3025 iput(journal_inode);
3026 return NULL;
3027 }
3028 journal->j_private = sb;
617ba13b 3029 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 3030 return journal;
3031}
3032
617ba13b 3033static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 3034 dev_t j_dev)
3035{
2b2d6d01 3036 struct buffer_head *bh;
ac27a0ec 3037 journal_t *journal;
617ba13b
MC		 3038 ext4_fsblk_t start;
3039 ext4_fsblk_t len;
ac27a0ec 3040 int hblock, blocksize;
617ba13b 3041 ext4_fsblk_t sb_block;
ac27a0ec 3042 unsigned long offset;
2b2d6d01 3043 struct ext4_super_block *es;
ac27a0ec
DK		 3044 struct block_device *bdev;
3045
0390131b
FM		 3046 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3047
b31e1552 3048 bdev = ext4_blkdev_get(j_dev, sb);
ac27a0ec
DK		 3049 if (bdev == NULL)
3050 return NULL;
3051
3052 if (bd_claim(bdev, sb)) {
b31e1552
ES		 3053 ext4_msg(sb, KERN_ERR,
3054 "failed to claim external journal device");
9a1c3542 3055 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
ac27a0ec
DK		 3056 return NULL;
3057 }
3058
3059 blocksize = sb->s_blocksize;
e1defc4f 3060 hblock = bdev_logical_block_size(bdev);
ac27a0ec 3061 if (blocksize < hblock) {
b31e1552
ES		 3062 ext4_msg(sb, KERN_ERR,
3063 "blocksize too small for journal device");
ac27a0ec
DK		 3064 goto out_bdev;
3065 }
3066
617ba13b
MC		 3067 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3068 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 3069 set_blocksize(bdev, blocksize);
3070 if (!(bh = __bread(bdev, sb_block, blocksize))) {
b31e1552
ES		 3071 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
3072 "external journal");
ac27a0ec
DK		 3073 goto out_bdev;
3074 }
3075
617ba13b
MC		 3076 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3077 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 3078 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b 3079 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
b31e1552
ES		 3080 ext4_msg(sb, KERN_ERR, "external journal has "
3081 "bad superblock");
ac27a0ec
DK		 3082 brelse(bh);
3083 goto out_bdev;
3084 }
3085
617ba13b 3086 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
b31e1552 3087 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
ac27a0ec
DK		 3088 brelse(bh);
3089 goto out_bdev;
3090 }
3091
bd81d8ee 3092 len = ext4_blocks_count(es);
ac27a0ec
DK		 3093 start = sb_block + 1;
3094 brelse(bh); /* we're done with the superblock */
3095
dab291af 3096 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 3097 start, len, blocksize);
3098 if (!journal) {
b31e1552 3099 ext4_msg(sb, KERN_ERR, "failed to create device journal");
ac27a0ec
DK		 3100 goto out_bdev;
3101 }
3102 journal->j_private = sb;
3103 ll_rw_block(READ, 1, &journal->j_sb_buffer);
3104 wait_on_buffer(journal->j_sb_buffer);
3105 if (!buffer_uptodate(journal->j_sb_buffer)) {
b31e1552 3106 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
ac27a0ec
DK		 3107 goto out_journal;
3108 }
3109 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
b31e1552
ES		 3110 ext4_msg(sb, KERN_ERR, "External journal has more than one "
3111 "user (unsupported) - %d",
ac27a0ec
DK		 3112 be32_to_cpu(journal->j_superblock->s_nr_users));
3113 goto out_journal;
3114 }
617ba13b
MC		 3115 EXT4_SB(sb)->journal_bdev = bdev;
3116 ext4_init_journal_params(sb, journal);
ac27a0ec 3117 return journal;
0b8e58a1 3118
ac27a0ec 3119out_journal:
dab291af 3120 jbd2_journal_destroy(journal);
ac27a0ec 3121out_bdev:
617ba13b 3122 ext4_blkdev_put(bdev);
ac27a0ec
DK		 3123 return NULL;
3124}
3125
617ba13b
MC		 3126static int ext4_load_journal(struct super_block *sb,
3127 struct ext4_super_block *es,
ac27a0ec
DK		 3128 unsigned long journal_devnum)
3129{
3130 journal_t *journal;
3131 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
3132 dev_t journal_dev;
3133 int err = 0;
3134 int really_read_only;
3135
0390131b
FM		 3136 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3137
ac27a0ec
DK		 3138 if (journal_devnum &&
3139 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
b31e1552
ES		 3140 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
3141 "numbers have changed");
ac27a0ec
DK		 3142 journal_dev = new_decode_dev(journal_devnum);
3143 } else
3144 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
3145
3146 really_read_only = bdev_read_only(sb->s_bdev);
3147
3148 /*
3149 * Are we loading a blank journal or performing recovery after a
3150 * crash? For recovery, we need to check in advance whether we
3151 * can get read-write access to the device.
3152 */
617ba13b 3153 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
ac27a0ec 3154 if (sb->s_flags & MS_RDONLY) {
b31e1552
ES		 3155 ext4_msg(sb, KERN_INFO, "INFO: recovery "
3156 "required on readonly filesystem");
ac27a0ec 3157 if (really_read_only) {
b31e1552
ES		 3158 ext4_msg(sb, KERN_ERR, "write access "
3159 "unavailable, cannot proceed");
ac27a0ec
DK		 3160 return -EROFS;
3161 }
b31e1552
ES		 3162 ext4_msg(sb, KERN_INFO, "write access will "
3163 "be enabled during recovery");
ac27a0ec
DK		 3164 }
3165 }
3166
3167 if (journal_inum && journal_dev) {
b31e1552
ES		 3168 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
3169 "and inode journals!");
ac27a0ec
DK		 3170 return -EINVAL;
3171 }
3172
3173 if (journal_inum) {
617ba13b 3174 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 3175 return -EINVAL;
3176 } else {
617ba13b 3177 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 3178 return -EINVAL;
3179 }
3180
4776004f 3181 if (journal->j_flags & JBD2_BARRIER)
b31e1552 3182 ext4_msg(sb, KERN_INFO, "barriers enabled");
4776004f 3183 else
b31e1552 3184 ext4_msg(sb, KERN_INFO, "barriers disabled");
4776004f 3185
ac27a0ec 3186 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
dab291af 3187 err = jbd2_journal_update_format(journal);
ac27a0ec 3188 if (err) {
b31e1552 3189 ext4_msg(sb, KERN_ERR, "error updating journal");
dab291af 3190 jbd2_journal_destroy(journal);
ac27a0ec
DK		 3191 return err;
3192 }
3193 }
3194
617ba13b 3195 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
dab291af 3196 err = jbd2_journal_wipe(journal, !really_read_only);
ac27a0ec 3197 if (!err)
dab291af 3198 err = jbd2_journal_load(journal);
ac27a0ec
DK		 3199
3200 if (err) {
b31e1552 3201 ext4_msg(sb, KERN_ERR, "error loading journal");
dab291af 3202 jbd2_journal_destroy(journal);
ac27a0ec
DK		 3203 return err;
3204 }
3205
617ba13b
MC		 3206 EXT4_SB(sb)->s_journal = journal;
3207 ext4_clear_journal_err(sb, es);
ac27a0ec
DK		 3208
3209 if (journal_devnum &&
3210 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3211 es->s_journal_dev = cpu_to_le32(journal_devnum);
ac27a0ec
DK		 3212
3213 /* Make sure we flush the recovery flag to disk. */
e2d67052 3214 ext4_commit_super(sb, 1);
ac27a0ec
DK		 3215 }
3216
3217 return 0;
3218}
3219
e2d67052 3220static int ext4_commit_super(struct super_block *sb, int sync)
ac27a0ec 3221{
e2d67052 3222 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
617ba13b 3223 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
c4be0c1d 3224 int error = 0;
ac27a0ec
DK		 3225
3226 if (!sbh)
c4be0c1d 3227 return error;
914258bf
TT		 3228 if (buffer_write_io_error(sbh)) {
3229 /*
3230 * Oh, dear. A previous attempt to write the
3231 * superblock failed. This could happen because the
3232 * USB device was yanked out. Or it could happen to
3233 * be a transient write error and maybe the block will
3234 * be remapped. Nothing we can do but to retry the
3235 * write and hope for the best.
3236 */
b31e1552
ES		 3237 ext4_msg(sb, KERN_ERR, "previous I/O error to "
3238 "superblock detected");
914258bf
TT		 3239 clear_buffer_write_io_error(sbh);
3240 set_buffer_uptodate(sbh);
3241 }
71290b36
TT		 3242 /*
3243 * If the file system is mounted read-only, don't update the
3244 * superblock write time. This avoids updating the superblock
3245 * write time when we are mounting the root file system
3246 * read/only but we need to replay the journal; at that point,
3247 * for people who are east of GMT and who make their clock
3248 * tick in localtime for Windows bug-for-bug compatibility,
3249 * the clock is set in the future, and this will cause e2fsck
3250 * to complain and force a full file system check.
3251 */
3252 if (!(sb->s_flags & MS_RDONLY))
3253 es->s_wtime = cpu_to_le32(get_seconds());
afc32f7e
TT		 3254 es->s_kbytes_written =
3255 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
3256 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
3257 EXT4_SB(sb)->s_sectors_written_start) >> 1));
5d1b1b3f
AK		 3258 ext4_free_blocks_count_set(es, percpu_counter_sum_positive(
3259 &EXT4_SB(sb)->s_freeblocks_counter));
3260 es->s_free_inodes_count = cpu_to_le32(percpu_counter_sum_positive(
3261 &EXT4_SB(sb)->s_freeinodes_counter));
7234ab2a 3262 sb->s_dirt = 0;
ac27a0ec
DK		 3263 BUFFER_TRACE(sbh, "marking dirty");
3264 mark_buffer_dirty(sbh);
914258bf 3265 if (sync) {
c4be0c1d
TS		 3266 error = sync_dirty_buffer(sbh);
3267 if (error)
3268 return error;
3269
3270 error = buffer_write_io_error(sbh);
3271 if (error) {
b31e1552
ES		 3272 ext4_msg(sb, KERN_ERR, "I/O error while writing "
3273 "superblock");
914258bf
TT		 3274 clear_buffer_write_io_error(sbh);
3275 set_buffer_uptodate(sbh);
3276 }
3277 }
c4be0c1d 3278 return error;
ac27a0ec
DK		 3279}
3280
ac27a0ec
DK		 3281/*
3282 * Have we just finished recovery? If so, and if we are mounting (or
3283 * remounting) the filesystem readonly, then we will end up with a
3284 * consistent fs on disk. Record that fact.
3285 */
2b2d6d01
TT		 3286static void ext4_mark_recovery_complete(struct super_block *sb,
3287 struct ext4_super_block *es)
ac27a0ec 3288{
617ba13b 3289 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 3290
0390131b
FM		 3291 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3292 BUG_ON(journal != NULL);
3293 return;
3294 }
dab291af 3295 jbd2_journal_lock_updates(journal);
7ffe1ea8
HK		 3296 if (jbd2_journal_flush(journal) < 0)
3297 goto out;
3298
617ba13b 3299 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
ac27a0ec 3300 sb->s_flags & MS_RDONLY) {
617ba13b 3301 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
e2d67052 3302 ext4_commit_super(sb, 1);
ac27a0ec 3303 }
7ffe1ea8
HK		 3304
3305out:
dab291af 3306 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 3307}
3308
3309/*
3310 * If we are mounting (or read-write remounting) a filesystem whose journal
3311 * has recorded an error from a previous lifetime, move that error to the
3312 * main filesystem now.
3313 */
2b2d6d01
TT		 3314static void ext4_clear_journal_err(struct super_block *sb,
3315 struct ext4_super_block *es)
ac27a0ec
DK		 3316{
3317 journal_t *journal;
3318 int j_errno;
3319 const char *errstr;
3320
0390131b
FM		 3321 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3322
617ba13b 3323 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 3324
3325 /*
3326 * Now check for any error status which may have been recorded in the
617ba13b 3327 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 3328 */
3329
dab291af 3330 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 3331 if (j_errno) {
3332 char nbuf[16];
3333
617ba13b 3334 errstr = ext4_decode_error(sb, j_errno, nbuf);
46e665e9 3335 ext4_warning(sb, __func__, "Filesystem error recorded "
ac27a0ec 3336 "from previous mount: %s", errstr);
46e665e9 3337 ext4_warning(sb, __func__, "Marking fs in need of "
ac27a0ec
DK		 3338 "filesystem check.");
3339
617ba13b
MC		 3340 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
3341 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
e2d67052 3342 ext4_commit_super(sb, 1);
ac27a0ec 3343
dab291af 3344 jbd2_journal_clear_err(journal);
ac27a0ec
DK		 3345 }
3346}
3347
3348/*
3349 * Force the running and committing transactions to commit,
3350 * and wait on the commit.
3351 */
617ba13b 3352int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 3353{
3354 journal_t *journal;
0390131b 3355 int ret = 0;
ac27a0ec
DK		 3356
3357 if (sb->s_flags & MS_RDONLY)
3358 return 0;
3359
617ba13b 3360 journal = EXT4_SB(sb)->s_journal;
7234ab2a 3361 if (journal)
0390131b 3362 ret = ext4_journal_force_commit(journal);
0390131b 3363
ac27a0ec
DK		 3364 return ret;
3365}
3366
2b2d6d01 3367static void ext4_write_super(struct super_block *sb)
ac27a0ec 3368{
ebc1ac16 3369 lock_super(sb);
9ca92389 3370 ext4_commit_super(sb, 1);
ebc1ac16 3371 unlock_super(sb);
ac27a0ec
DK		 3372}
3373
617ba13b 3374static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec 3375{
14ce0cb4 3376 int ret = 0;
9eddacf9 3377 tid_t target;
ac27a0ec 3378
9bffad1e 3379 trace_ext4_sync_fs(sb, wait);
9ca92389
TT		 3380 if (jbd2_journal_start_commit(EXT4_SB(sb)->s_journal, &target)) {
3381 if (wait)
3382 jbd2_log_wait_commit(EXT4_SB(sb)->s_journal, target);
0390131b 3383 }
14ce0cb4 3384 return ret;
ac27a0ec
DK		 3385}
3386
3387/*
3388 * LVM calls this function before a (read-only) snapshot is created. This
3389 * gives us a chance to flush the journal completely and mark the fs clean.
3390 */
c4be0c1d 3391static int ext4_freeze(struct super_block *sb)
ac27a0ec 3392{
c4be0c1d
TS		 3393 int error = 0;
3394 journal_t *journal;
ac27a0ec 3395
9ca92389
TT		 3396 if (sb->s_flags & MS_RDONLY)
3397 return 0;
ac27a0ec 3398
9ca92389 3399 journal = EXT4_SB(sb)->s_journal;
7ffe1ea8 3400
9ca92389
TT		 3401 /* Now we set up the journal barrier. */
3402 jbd2_journal_lock_updates(journal);
ac27a0ec 3403
9ca92389
TT		 3404 /*
3405 * Don't clear the needs_recovery flag if we failed to flush
3406 * the journal.
3407 */
3408 error = jbd2_journal_flush(journal);
3409 if (error < 0) {
3410 out:
3411 jbd2_journal_unlock_updates(journal);
3412 return error;
ac27a0ec 3413 }
9ca92389
TT		 3414
3415 /* Journal blocked and flushed, clear needs_recovery flag. */
3416 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3417 error = ext4_commit_super(sb, 1);
3418 if (error)
3419 goto out;
c4be0c1d 3420 return 0;
ac27a0ec
DK		 3421}
3422
3423/*
3424 * Called by LVM after the snapshot is done. We need to reset the RECOVER
3425 * flag here, even though the filesystem is not technically dirty yet.
3426 */
c4be0c1d 3427static int ext4_unfreeze(struct super_block *sb)
ac27a0ec 3428{
9ca92389
TT		 3429 if (sb->s_flags & MS_RDONLY)
3430 return 0;
3431
3432 lock_super(sb);
3433 /* Reset the needs_recovery flag before the fs is unlocked. */
3434 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3435 ext4_commit_super(sb, 1);
3436 unlock_super(sb);
3437 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
c4be0c1d 3438 return 0;
ac27a0ec
DK		 3439}
3440
2b2d6d01 3441static int ext4_remount(struct super_block *sb, int *flags, char *data)
ac27a0ec 3442{
2b2d6d01 3443 struct ext4_super_block *es;
617ba13b
MC		 3444 struct ext4_sb_info *sbi = EXT4_SB(sb);
3445 ext4_fsblk_t n_blocks_count = 0;
ac27a0ec 3446 unsigned long old_sb_flags;
617ba13b 3447 struct ext4_mount_options old_opts;
8a266467 3448 ext4_group_t g;
b3881f74 3449 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
ac27a0ec
DK		 3450 int err;
3451#ifdef CONFIG_QUOTA
3452 int i;
3453#endif
3454
337eb00a
AIB		 3455 lock_kernel();
3456
ac27a0ec 3457 /* Store the original options */
bbd6851a 3458 lock_super(sb);
ac27a0ec
DK		 3459 old_sb_flags = sb->s_flags;
3460 old_opts.s_mount_opt = sbi->s_mount_opt;
3461 old_opts.s_resuid = sbi->s_resuid;
3462 old_opts.s_resgid = sbi->s_resgid;
3463 old_opts.s_commit_interval = sbi->s_commit_interval;
30773840
TT		 3464 old_opts.s_min_batch_time = sbi->s_min_batch_time;
3465 old_opts.s_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 3466#ifdef CONFIG_QUOTA
3467 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
3468 for (i = 0; i < MAXQUOTAS; i++)
3469 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
3470#endif
b3881f74
TT		 3471 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
3472 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
ac27a0ec
DK		 3473
3474 /*
3475 * Allow the "check" option to be passed as a remount option.
3476 */
b3881f74
TT		 3477 if (!parse_options(data, sb, NULL, &journal_ioprio,
3478 &n_blocks_count, 1)) {
ac27a0ec
DK		 3479 err = -EINVAL;
3480 goto restore_opts;
3481 }
3482
4ab2f15b 3483 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
46e665e9 3484 ext4_abort(sb, __func__, "Abort forced by user");
ac27a0ec
DK		 3485
3486 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
617ba13b 3487 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec
DK		 3488
3489 es = sbi->s_es;
3490
b3881f74 3491 if (sbi->s_journal) {
0390131b 3492 ext4_init_journal_params(sb, sbi->s_journal);
b3881f74
TT		 3493 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3494 }
ac27a0ec
DK		 3495
3496 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
bd81d8ee 3497 n_blocks_count > ext4_blocks_count(es)) {
4ab2f15b 3498 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
ac27a0ec
DK		 3499 err = -EROFS;
3500 goto restore_opts;
3501 }
3502
3503 if (*flags & MS_RDONLY) {
3504 /*
3505 * First of all, the unconditional stuff we have to do
3506 * to disable replay of the journal when we next remount
3507 */
3508 sb->s_flags |= MS_RDONLY;
3509
3510 /*
3511 * OK, test if we are remounting a valid rw partition
3512 * readonly, and if so set the rdonly flag and then
3513 * mark the partition as valid again.
3514 */
617ba13b
MC		 3515 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
3516 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 3517 es->s_state = cpu_to_le16(sbi->s_mount_state);
3518
a63c9eb2 3519 if (sbi->s_journal)
0390131b 3520 ext4_mark_recovery_complete(sb, es);
ac27a0ec 3521 } else {
a13fb1a4
ES		 3522 /* Make sure we can mount this feature set readwrite */
3523 if (!ext4_feature_set_ok(sb, 0)) {
ac27a0ec
DK		 3524 err = -EROFS;
3525 goto restore_opts;
3526 }
8a266467
TT		 3527 /*
3528 * Make sure the group descriptor checksums
0b8e58a1 3529 * are sane. If they aren't, refuse to remount r/w.
8a266467
TT		 3530 */
3531 for (g = 0; g < sbi->s_groups_count; g++) {
3532 struct ext4_group_desc *gdp =
3533 ext4_get_group_desc(sb, g, NULL);
3534
3535 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
b31e1552
ES		 3536 ext4_msg(sb, KERN_ERR,
3537 "ext4_remount: Checksum for group %u failed (%u!=%u)",
8a266467
TT		 3538 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
3539 le16_to_cpu(gdp->bg_checksum));
3540 err = -EINVAL;
3541 goto restore_opts;
3542 }
3543 }
3544
ead6596b
ES		 3545 /*
3546 * If we have an unprocessed orphan list hanging
3547 * around from a previously readonly bdev mount,
3548 * require a full umount/remount for now.
3549 */
3550 if (es->s_last_orphan) {
b31e1552 3551 ext4_msg(sb, KERN_WARNING, "Couldn't "
ead6596b
ES		 3552 "remount RDWR because of unprocessed "
3553 "orphan inode list. Please "
b31e1552 3554 "umount/remount instead");
ead6596b
ES		 3555 err = -EINVAL;
3556 goto restore_opts;
3557 }
3558
ac27a0ec
DK		 3559 /*
3560 * Mounting a RDONLY partition read-write, so reread
3561 * and store the current valid flag. (It may have
3562 * been changed by e2fsck since we originally mounted
3563 * the partition.)
3564 */
0390131b
FM		 3565 if (sbi->s_journal)
3566 ext4_clear_journal_err(sb, es);
ac27a0ec 3567 sbi->s_mount_state = le16_to_cpu(es->s_state);
617ba13b 3568 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
ac27a0ec 3569 goto restore_opts;
2b2d6d01 3570 if (!ext4_setup_super(sb, es, 0))
ac27a0ec
DK		 3571 sb->s_flags &= ~MS_RDONLY;
3572 }
3573 }
6fd058f7 3574 ext4_setup_system_zone(sb);
0390131b 3575 if (sbi->s_journal == NULL)
e2d67052 3576 ext4_commit_super(sb, 1);
0390131b 3577
ac27a0ec
DK		 3578#ifdef CONFIG_QUOTA
3579 /* Release old quota file names */
3580 for (i = 0; i < MAXQUOTAS; i++)
3581 if (old_opts.s_qf_names[i] &&
3582 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3583 kfree(old_opts.s_qf_names[i]);
3584#endif
bbd6851a 3585 unlock_super(sb);
337eb00a 3586 unlock_kernel();
ac27a0ec 3587 return 0;
0b8e58a1 3588
ac27a0ec
DK		 3589restore_opts:
3590 sb->s_flags = old_sb_flags;
3591 sbi->s_mount_opt = old_opts.s_mount_opt;
3592 sbi->s_resuid = old_opts.s_resuid;
3593 sbi->s_resgid = old_opts.s_resgid;
3594 sbi->s_commit_interval = old_opts.s_commit_interval;
30773840
TT		 3595 sbi->s_min_batch_time = old_opts.s_min_batch_time;
3596 sbi->s_max_batch_time = old_opts.s_max_batch_time;
ac27a0ec
DK		 3597#ifdef CONFIG_QUOTA
3598 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
3599 for (i = 0; i < MAXQUOTAS; i++) {
3600 if (sbi->s_qf_names[i] &&
3601 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3602 kfree(sbi->s_qf_names[i]);
3603 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
3604 }
3605#endif
bbd6851a 3606 unlock_super(sb);
337eb00a 3607 unlock_kernel();
ac27a0ec
DK		 3608 return err;
3609}
3610
2b2d6d01 3611static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
ac27a0ec
DK		 3612{
3613 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 3614 struct ext4_sb_info *sbi = EXT4_SB(sb);
3615 struct ext4_super_block *es = sbi->s_es;
960cc398 3616 u64 fsid;
ac27a0ec 3617
5e70030d
BP		 3618 if (test_opt(sb, MINIX_DF)) {
3619 sbi->s_overhead_last = 0;
6bc9feff 3620 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
8df9675f 3621 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
5e70030d 3622 ext4_fsblk_t overhead = 0;
ac27a0ec
DK		 3623
3624 /*
5e70030d
BP		 3625 * Compute the overhead (FS structures). This is constant
3626 * for a given filesystem unless the number of block groups
3627 * changes so we cache the previous value until it does.
ac27a0ec
DK		 3628 */
3629
3630 /*
3631 * All of the blocks before first_data_block are
3632 * overhead
3633 */
3634 overhead = le32_to_cpu(es->s_first_data_block);
3635
3636 /*
3637 * Add the overhead attributed to the superblock and
3638 * block group descriptors. If the sparse superblocks
3639 * feature is turned on, then not all groups have this.
3640 */
3641 for (i = 0; i < ngroups; i++) {
617ba13b
MC		 3642 overhead += ext4_bg_has_super(sb, i) +
3643 ext4_bg_num_gdb(sb, i);
ac27a0ec
DK		 3644 cond_resched();
3645 }
3646
3647 /*
3648 * Every block group has an inode bitmap, a block
3649 * bitmap, and an inode table.
3650 */
5e70030d
BP		 3651 overhead += ngroups * (2 + sbi->s_itb_per_group);
3652 sbi->s_overhead_last = overhead;
3653 smp_wmb();
6bc9feff 3654 sbi->s_blocks_last = ext4_blocks_count(es);
ac27a0ec
DK		 3655 }
3656
617ba13b 3657 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 3658 buf->f_bsize = sb->s_blocksize;
5e70030d 3659 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
6bc6e63f
AK		 3660 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
3661 percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
308ba3ec 3662 ext4_free_blocks_count_set(es, buf->f_bfree);
bd81d8ee
LV		 3663 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3664 if (buf->f_bfree < ext4_r_blocks_count(es))
ac27a0ec
DK		 3665 buf->f_bavail = 0;
3666 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 3667 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
5e70030d 3668 es->s_free_inodes_count = cpu_to_le32(buf->f_ffree);
617ba13b 3669 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 3670 fsid = le64_to_cpup((void *)es->s_uuid) ^
3671 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3672 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3673 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
0b8e58a1 3674
ac27a0ec
DK		 3675 return 0;
3676}
3677
0b8e58a1
AD		 3678/* Helper function for writing quotas on sync - we need to start transaction
3679 * before quota file is locked for write. Otherwise the are possible deadlocks:
ac27a0ec 3680 * Process 1 Process 2
617ba13b 3681 * ext4_create() quota_sync()
a269eb18
JK		 3682 * jbd2_journal_start() write_dquot()
3683 * vfs_dq_init() down(dqio_mutex)
dab291af 3684 * down(dqio_mutex) jbd2_journal_start()
ac27a0ec
DK		 3685 *
3686 */
3687
3688#ifdef CONFIG_QUOTA
3689
3690static inline struct inode *dquot_to_inode(struct dquot *dquot)
3691{
3692 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3693}
3694
617ba13b 3695static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 3696{
3697 int ret, err;
3698 handle_t *handle;
3699 struct inode *inode;
3700
3701 inode = dquot_to_inode(dquot);
617ba13b 3702 handle = ext4_journal_start(inode,
0b8e58a1 3703 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 3704 if (IS_ERR(handle))
3705 return PTR_ERR(handle);
3706 ret = dquot_commit(dquot);
617ba13b 3707 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3708 if (!ret)
3709 ret = err;
3710 return ret;
3711}
3712
617ba13b 3713static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 3714{
3715 int ret, err;
3716 handle_t *handle;
3717
617ba13b 3718 handle = ext4_journal_start(dquot_to_inode(dquot),
0b8e58a1 3719 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 3720 if (IS_ERR(handle))
3721 return PTR_ERR(handle);
3722 ret = dquot_acquire(dquot);
617ba13b 3723 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3724 if (!ret)
3725 ret = err;
3726 return ret;
3727}
3728
617ba13b 3729static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 3730{
3731 int ret, err;
3732 handle_t *handle;
3733
617ba13b 3734 handle = ext4_journal_start(dquot_to_inode(dquot),
0b8e58a1 3735 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 3736 if (IS_ERR(handle)) {
3737 /* Release dquot anyway to avoid endless cycle in dqput() */
3738 dquot_release(dquot);
ac27a0ec 3739 return PTR_ERR(handle);
9c3013e9 3740 }
ac27a0ec 3741 ret = dquot_release(dquot);
617ba13b 3742 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3743 if (!ret)
3744 ret = err;
3745 return ret;
3746}
3747
617ba13b 3748static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec 3749{
2c8be6b2 3750 /* Are we journaling quotas? */
617ba13b
MC		 3751 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3752 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
ac27a0ec 3753 dquot_mark_dquot_dirty(dquot);
617ba13b 3754 return ext4_write_dquot(dquot);
ac27a0ec
DK		 3755 } else {
3756 return dquot_mark_dquot_dirty(dquot);
3757 }
3758}
3759
617ba13b 3760static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 3761{
3762 int ret, err;
3763 handle_t *handle;
3764
3765 /* Data block + inode block */
617ba13b 3766 handle = ext4_journal_start(sb->s_root->d_inode, 2);
ac27a0ec
DK		 3767 if (IS_ERR(handle))
3768 return PTR_ERR(handle);
3769 ret = dquot_commit_info(sb, type);
617ba13b 3770 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3771 if (!ret)
3772 ret = err;
3773 return ret;
3774}
3775
3776/*
3777 * Turn on quotas during mount time - we need to find
3778 * the quota file and such...
3779 */
617ba13b 3780static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 3781{
617ba13b 3782 return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
0b8e58a1 3783 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 3784}
3785
3786/*
3787 * Standard function to be called on quota_on
3788 */
617ba13b 3789static int ext4_quota_on(struct super_block *sb, int type, int format_id,
8264613d 3790 char *name, int remount)
ac27a0ec
DK		 3791{
3792 int err;
8264613d 3793 struct path path;
ac27a0ec
DK		 3794
3795 if (!test_opt(sb, QUOTA))
3796 return -EINVAL;
8264613d 3797 /* When remounting, no checks are needed and in fact, name is NULL */
0623543b 3798 if (remount)
8264613d 3799 return vfs_quota_on(sb, type, format_id, name, remount);
0623543b 3800
8264613d 3801 err = kern_path(name, LOOKUP_FOLLOW, &path);
ac27a0ec
DK		 3802 if (err)
3803 return err;
0623543b 3804
ac27a0ec 3805 /* Quotafile not on the same filesystem? */
8264613d
AV		 3806 if (path.mnt->mnt_sb != sb) {
3807 path_put(&path);
ac27a0ec
DK		 3808 return -EXDEV;
3809 }
0623543b
JK		 3810 /* Journaling quota? */
3811 if (EXT4_SB(sb)->s_qf_names[type]) {
2b2d6d01 3812 /* Quotafile not in fs root? */
8264613d 3813 if (path.dentry->d_parent != sb->s_root)
b31e1552
ES		 3814 ext4_msg(sb, KERN_WARNING,
3815 "Quota file not on filesystem root. "
3816 "Journaled quota will not work");
2b2d6d01 3817 }
0623543b
JK		 3818
3819 /*
3820 * When we journal data on quota file, we have to flush journal to see
3821 * all updates to the file when we bypass pagecache...
3822 */
0390131b
FM		 3823 if (EXT4_SB(sb)->s_journal &&
3824 ext4_should_journal_data(path.dentry->d_inode)) {
0623543b
JK		 3825 /*
3826 * We don't need to lock updates but journal_flush() could
3827 * otherwise be livelocked...
3828 */
3829 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 3830 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
0623543b 3831 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 3832 if (err) {
8264613d 3833 path_put(&path);
7ffe1ea8
HK		 3834 return err;
3835 }
0623543b
JK		 3836 }
3837
8264613d
AV		 3838 err = vfs_quota_on_path(sb, type, format_id, &path);
3839 path_put(&path);
77e69dac 3840 return err;
ac27a0ec
DK		 3841}
3842
3843/* Read data from quotafile - avoid pagecache and such because we cannot afford
3844 * acquiring the locks... As quota files are never truncated and quota code
3845 * itself serializes the operations (and noone else should touch the files)
3846 * we don't have to be afraid of races */
617ba13b 3847static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 3848 size_t len, loff_t off)
3849{
3850 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 3851 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 3852 int err = 0;
3853 int offset = off & (sb->s_blocksize - 1);
3854 int tocopy;
3855 size_t toread;
3856 struct buffer_head *bh;
3857 loff_t i_size = i_size_read(inode);
3858
3859 if (off > i_size)
3860 return 0;
3861 if (off+len > i_size)
3862 len = i_size-off;
3863 toread = len;
3864 while (toread > 0) {
3865 tocopy = sb->s_blocksize - offset < toread ?
3866 sb->s_blocksize - offset : toread;
617ba13b 3867 bh = ext4_bread(NULL, inode, blk, 0, &err);
ac27a0ec
DK		 3868 if (err)
3869 return err;
3870 if (!bh) /* A hole? */
3871 memset(data, 0, tocopy);
3872 else
3873 memcpy(data, bh->b_data+offset, tocopy);
3874 brelse(bh);
3875 offset = 0;
3876 toread -= tocopy;
3877 data += tocopy;
3878 blk++;
3879 }
3880 return len;
3881}
3882
3883/* Write to quotafile (we know the transaction is already started and has
3884 * enough credits) */
617ba13b 3885static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 3886 const char *data, size_t len, loff_t off)
3887{
3888 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 3889 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 3890 int err = 0;
3891 int offset = off & (sb->s_blocksize - 1);
3892 int tocopy;
617ba13b 3893 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
ac27a0ec
DK		 3894 size_t towrite = len;
3895 struct buffer_head *bh;
3896 handle_t *handle = journal_current_handle();
3897
0390131b 3898 if (EXT4_SB(sb)->s_journal && !handle) {
b31e1552
ES		 3899 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
3900 " cancelled because transaction is not started",
9c3013e9
JK		 3901 (unsigned long long)off, (unsigned long long)len);
3902 return -EIO;
3903 }
ac27a0ec
DK		 3904 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
3905 while (towrite > 0) {
3906 tocopy = sb->s_blocksize - offset < towrite ?
3907 sb->s_blocksize - offset : towrite;
617ba13b 3908 bh = ext4_bread(handle, inode, blk, 1, &err);
ac27a0ec
DK		 3909 if (!bh)
3910 goto out;
3911 if (journal_quota) {
617ba13b 3912 err = ext4_journal_get_write_access(handle, bh);
ac27a0ec
DK		 3913 if (err) {
3914 brelse(bh);
3915 goto out;
3916 }
3917 }
3918 lock_buffer(bh);
3919 memcpy(bh->b_data+offset, data, tocopy);
3920 flush_dcache_page(bh->b_page);
3921 unlock_buffer(bh);
3922 if (journal_quota)
0390131b 3923 err = ext4_handle_dirty_metadata(handle, NULL, bh);
ac27a0ec
DK		 3924 else {
3925 /* Always do at least ordered writes for quotas */
678aaf48 3926 err = ext4_jbd2_file_inode(handle, inode);
ac27a0ec
DK		 3927 mark_buffer_dirty(bh);
3928 }
3929 brelse(bh);
3930 if (err)
3931 goto out;
3932 offset = 0;
3933 towrite -= tocopy;
3934 data += tocopy;
3935 blk++;
3936 }
3937out:
4d04e4fb
JK		 3938 if (len == towrite) {
3939 mutex_unlock(&inode->i_mutex);
ac27a0ec 3940 return err;
4d04e4fb 3941 }
ac27a0ec
DK		 3942 if (inode->i_size < off+len-towrite) {
3943 i_size_write(inode, off+len-towrite);
617ba13b 3944 EXT4_I(inode)->i_disksize = inode->i_size;
ac27a0ec 3945 }
ac27a0ec 3946 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
617ba13b 3947 ext4_mark_inode_dirty(handle, inode);
ac27a0ec
DK		 3948 mutex_unlock(&inode->i_mutex);
3949 return len - towrite;
3950}
3951
3952#endif
3953
0b8e58a1
AD		 3954static int ext4_get_sb(struct file_system_type *fs_type, int flags,
3955 const char *dev_name, void *data, struct vfsmount *mnt)
ac27a0ec 3956{
0b8e58a1 3957 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super,mnt);
ac27a0ec
DK		 3958}
3959
03010a33
TT		 3960static struct file_system_type ext4_fs_type = {
3961 .owner = THIS_MODULE,
3962 .name = "ext4",
3963 .get_sb = ext4_get_sb,
3964 .kill_sb = kill_block_super,
3965 .fs_flags = FS_REQUIRES_DEV,
3966};
3967
3968#ifdef CONFIG_EXT4DEV_COMPAT
0b8e58a1
AD		 3969static int ext4dev_get_sb(struct file_system_type *fs_type, int flags,
3970 const char *dev_name, void *data,struct vfsmount *mnt)
03010a33 3971{
b31e1552
ES		 3972 printk(KERN_WARNING "EXT4-fs (%s): Update your userspace programs "
3973 "to mount using ext4\n", dev_name);
3974 printk(KERN_WARNING "EXT4-fs (%s): ext4dev backwards compatibility "
3975 "will go away by 2.6.31\n", dev_name);
0b8e58a1 3976 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super,mnt);
03010a33
TT		 3977}
3978
617ba13b 3979static struct file_system_type ext4dev_fs_type = {
ac27a0ec 3980 .owner = THIS_MODULE,
617ba13b 3981 .name = "ext4dev",
03010a33 3982 .get_sb = ext4dev_get_sb,
ac27a0ec
DK		 3983 .kill_sb = kill_block_super,
3984 .fs_flags = FS_REQUIRES_DEV,
3985};
03010a33
TT		 3986MODULE_ALIAS("ext4dev");
3987#endif
ac27a0ec 3988
617ba13b 3989static int __init init_ext4_fs(void)
ac27a0ec 3990{
c9de560d
AT		 3991 int err;
3992
6fd058f7
TT		 3993 err = init_ext4_system_zone();
3994 if (err)
3995 return err;
3197ebdb
TT		 3996 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
3997 if (!ext4_kset)
6fd058f7 3998 goto out4;
9f6200bb 3999 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
c9de560d 4000 err = init_ext4_mballoc();
ac27a0ec 4001 if (err)
6fd058f7 4002 goto out3;
c9de560d
AT		 4003
4004 err = init_ext4_xattr();
4005 if (err)
4006 goto out2;
ac27a0ec
DK		 4007 err = init_inodecache();
4008 if (err)
4009 goto out1;
03010a33 4010 err = register_filesystem(&ext4_fs_type);
ac27a0ec
DK		 4011 if (err)
4012 goto out;
03010a33
TT		 4013#ifdef CONFIG_EXT4DEV_COMPAT
4014 err = register_filesystem(&ext4dev_fs_type);
4015 if (err) {
4016 unregister_filesystem(&ext4_fs_type);
4017 goto out;
4018 }
4019#endif
ac27a0ec
DK		 4020 return 0;
4021out:
4022 destroy_inodecache();
4023out1:
617ba13b 4024 exit_ext4_xattr();
c9de560d
AT		 4025out2:
4026 exit_ext4_mballoc();
6fd058f7
TT		 4027out3:
4028 remove_proc_entry("fs/ext4", NULL);
4029 kset_unregister(ext4_kset);
4030out4:
4031 exit_ext4_system_zone();
ac27a0ec
DK		 4032 return err;
4033}
4034
617ba13b 4035static void __exit exit_ext4_fs(void)
ac27a0ec 4036{
03010a33
TT		 4037 unregister_filesystem(&ext4_fs_type);
4038#ifdef CONFIG_EXT4DEV_COMPAT
617ba13b 4039 unregister_filesystem(&ext4dev_fs_type);
03010a33 4040#endif
ac27a0ec 4041 destroy_inodecache();
617ba13b 4042 exit_ext4_xattr();
c9de560d 4043 exit_ext4_mballoc();
9f6200bb 4044 remove_proc_entry("fs/ext4", NULL);
3197ebdb 4045 kset_unregister(ext4_kset);
6fd058f7 4046 exit_ext4_system_zone();
ac27a0ec
DK		 4047}
4048
4049MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
83982b6f 4050MODULE_DESCRIPTION("Fourth Extended Filesystem");
ac27a0ec 4051MODULE_LICENSE("GPL");
617ba13b
MC		 4052module_init(init_ext4_fs)
4053module_exit(exit_ext4_fs)
kernel source for telekom puls (alcatel/mediatek)