projects / GitHub / LineageOS / android_kernel_motorola_exynos9610.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
ext4: force block allocation on quota_off
[GitHub/LineageOS/android_kernel_motorola_exynos9610.git] / fs / ext4 / super.c
CommitLineData
ac27a0ec 1/*
617ba13b 2 * linux/fs/ext4/super.c
ac27a0ec
DK		 3 *
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
8 *
9 * from
10 *
11 * linux/fs/minix/inode.c
12 *
13 * Copyright (C) 1991, 1992 Linus Torvalds
14 *
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
17 */
18
19#include <linux/module.h>
20#include <linux/string.h>
21#include <linux/fs.h>
22#include <linux/time.h>
c5ca7c76 23#include <linux/vmalloc.h>
dab291af 24#include <linux/jbd2.h>
ac27a0ec
DK		 25#include <linux/slab.h>
26#include <linux/init.h>
27#include <linux/blkdev.h>
28#include <linux/parser.h>
29#include <linux/smp_lock.h>
30#include <linux/buffer_head.h>
a5694255 31#include <linux/exportfs.h>
ac27a0ec
DK		 32#include <linux/vfs.h>
33#include <linux/random.h>
34#include <linux/mount.h>
35#include <linux/namei.h>
36#include <linux/quotaops.h>
37#include <linux/seq_file.h>
9f6200bb 38#include <linux/proc_fs.h>
3197ebdb 39#include <linux/ctype.h>
1330593e 40#include <linux/log2.h>
717d50e4 41#include <linux/crc16.h>
ac27a0ec
DK		 42#include <asm/uaccess.h>
43
3dcf5451
CH		 44#include "ext4.h"
45#include "ext4_jbd2.h"
ac27a0ec
DK		 46#include "xattr.h"
47#include "acl.h"
3661d286 48#include "mballoc.h"
ac27a0ec 49
9bffad1e
TT		 50#define CREATE_TRACE_POINTS
51#include <trace/events/ext4.h>
52
9f6200bb 53struct proc_dir_entry *ext4_proc_root;
3197ebdb 54static struct kset *ext4_kset;
9f6200bb 55
617ba13b 56static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
ac27a0ec 57 unsigned long journal_devnum);
e2d67052 58static int ext4_commit_super(struct super_block *sb, int sync);
2b2d6d01
TT		 59static void ext4_mark_recovery_complete(struct super_block *sb,
60 struct ext4_super_block *es);
61static void ext4_clear_journal_err(struct super_block *sb,
62 struct ext4_super_block *es);
617ba13b 63static int ext4_sync_fs(struct super_block *sb, int wait);
2b2d6d01 64static const char *ext4_decode_error(struct super_block *sb, int errno,
ac27a0ec 65 char nbuf[16]);
2b2d6d01
TT		 66static int ext4_remount(struct super_block *sb, int *flags, char *data);
67static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
c4be0c1d 68static int ext4_unfreeze(struct super_block *sb);
2b2d6d01 69static void ext4_write_super(struct super_block *sb);
c4be0c1d 70static int ext4_freeze(struct super_block *sb);
ba69f9ab
JK		 71static int ext4_get_sb(struct file_system_type *fs_type, int flags,
72 const char *dev_name, void *data, struct vfsmount *mnt);
ac27a0ec 73
ba69f9ab
JK		 74#if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
75static struct file_system_type ext3_fs_type = {
76 .owner = THIS_MODULE,
77 .name = "ext3",
78 .get_sb = ext4_get_sb,
79 .kill_sb = kill_block_super,
80 .fs_flags = FS_REQUIRES_DEV,
81};
82#define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
83#else
84#define IS_EXT3_SB(sb) (0)
85#endif
bd81d8ee 86
8fadc143
AR		 87ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
88 struct ext4_group_desc *bg)
bd81d8ee 89{
3a14589c 90 return le32_to_cpu(bg->bg_block_bitmap_lo) |
8fadc143 91 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 92 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 93}
94
8fadc143
AR		 95ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
96 struct ext4_group_desc *bg)
bd81d8ee 97{
5272f837 98 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
8fadc143 99 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 100 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 101}
102
8fadc143
AR		 103ext4_fsblk_t ext4_inode_table(struct super_block *sb,
104 struct ext4_group_desc *bg)
bd81d8ee 105{
5272f837 106 return le32_to_cpu(bg->bg_inode_table_lo) |
8fadc143 107 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 108 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
bd81d8ee
LV		 109}
110
560671a0
AK		 111__u32 ext4_free_blks_count(struct super_block *sb,
112 struct ext4_group_desc *bg)
113{
114 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
115 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 116 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
560671a0
AK		 117}
118
119__u32 ext4_free_inodes_count(struct super_block *sb,
120 struct ext4_group_desc *bg)
121{
122 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
123 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 124 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
560671a0
AK		 125}
126
127__u32 ext4_used_dirs_count(struct super_block *sb,
128 struct ext4_group_desc *bg)
129{
130 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
131 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 132 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
560671a0
AK		 133}
134
135__u32 ext4_itable_unused_count(struct super_block *sb,
136 struct ext4_group_desc *bg)
137{
138 return le16_to_cpu(bg->bg_itable_unused_lo) |
139 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 140 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
560671a0
AK		 141}
142
8fadc143
AR		 143void ext4_block_bitmap_set(struct super_block *sb,
144 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 145{
3a14589c 146 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 147 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
148 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 149}
150
8fadc143
AR		 151void ext4_inode_bitmap_set(struct super_block *sb,
152 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 153{
5272f837 154 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 155 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
156 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 157}
158
8fadc143
AR		 159void ext4_inode_table_set(struct super_block *sb,
160 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 161{
5272f837 162 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 163 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
164 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 165}
166
560671a0
AK		 167void ext4_free_blks_set(struct super_block *sb,
168 struct ext4_group_desc *bg, __u32 count)
169{
170 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
171 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
172 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
173}
174
175void ext4_free_inodes_set(struct super_block *sb,
176 struct ext4_group_desc *bg, __u32 count)
177{
178 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
179 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
180 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
181}
182
183void ext4_used_dirs_set(struct super_block *sb,
184 struct ext4_group_desc *bg, __u32 count)
185{
186 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
187 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
188 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
189}
190
191void ext4_itable_unused_set(struct super_block *sb,
192 struct ext4_group_desc *bg, __u32 count)
193{
194 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
195 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
196 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
197}
198
d3d1faf6
CW		 199
200/* Just increment the non-pointer handle value */
201static handle_t *ext4_get_nojournal(void)
202{
203 handle_t *handle = current->journal_info;
204 unsigned long ref_cnt = (unsigned long)handle;
205
206 BUG_ON(ref_cnt >= EXT4_NOJOURNAL_MAX_REF_COUNT);
207
208 ref_cnt++;
209 handle = (handle_t *)ref_cnt;
210
211 current->journal_info = handle;
212 return handle;
213}
214
215
216/* Decrement the non-pointer handle value */
217static void ext4_put_nojournal(handle_t *handle)
218{
219 unsigned long ref_cnt = (unsigned long)handle;
220
221 BUG_ON(ref_cnt == 0);
222
223 ref_cnt--;
224 handle = (handle_t *)ref_cnt;
225
226 current->journal_info = handle;
227}
228
ac27a0ec 229/*
dab291af 230 * Wrappers for jbd2_journal_start/end.
ac27a0ec
DK		 231 *
232 * The only special thing we need to do here is to make sure that all
233 * journal_end calls result in the superblock being marked dirty, so
234 * that sync() will call the filesystem's write_super callback if
235 * appropriate.
236 */
617ba13b 237handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
ac27a0ec
DK		 238{
239 journal_t *journal;
240
241 if (sb->s_flags & MS_RDONLY)
242 return ERR_PTR(-EROFS);
243
437f88cc 244 vfs_check_frozen(sb, SB_FREEZE_TRANS);
ac27a0ec
DK		 245 /* Special case here: if the journal has aborted behind our
246 * backs (eg. EIO in the commit thread), then we still need to
247 * take the FS itself readonly cleanly. */
617ba13b 248 journal = EXT4_SB(sb)->s_journal;
0390131b
FM		 249 if (journal) {
250 if (is_journal_aborted(journal)) {
c67d859e 251 ext4_abort(sb, "Detected aborted journal");
0390131b
FM		 252 return ERR_PTR(-EROFS);
253 }
254 return jbd2_journal_start(journal, nblocks);
ac27a0ec 255 }
d3d1faf6 256 return ext4_get_nojournal();
ac27a0ec
DK		 257}
258
259/*
260 * The only special thing we need to do here is to make sure that all
dab291af 261 * jbd2_journal_stop calls result in the superblock being marked dirty, so
ac27a0ec
DK		 262 * that sync() will call the filesystem's write_super callback if
263 * appropriate.
264 */
c398eda0 265int __ext4_journal_stop(const char *where, unsigned int line, handle_t *handle)
ac27a0ec
DK		 266{
267 struct super_block *sb;
268 int err;
269 int rc;
270
0390131b 271 if (!ext4_handle_valid(handle)) {
d3d1faf6 272 ext4_put_nojournal(handle);
0390131b
FM		 273 return 0;
274 }
ac27a0ec
DK		 275 sb = handle->h_transaction->t_journal->j_private;
276 err = handle->h_err;
dab291af 277 rc = jbd2_journal_stop(handle);
ac27a0ec
DK		 278
279 if (!err)
280 err = rc;
281 if (err)
c398eda0 282 __ext4_std_error(sb, where, line, err);
ac27a0ec
DK		 283 return err;
284}
285
90c7201b
TT		 286void ext4_journal_abort_handle(const char *caller, unsigned int line,
287 const char *err_fn, struct buffer_head *bh,
288 handle_t *handle, int err)
ac27a0ec
DK		 289{
290 char nbuf[16];
617ba13b 291 const char *errstr = ext4_decode_error(NULL, err, nbuf);
ac27a0ec 292
0390131b
FM		 293 BUG_ON(!ext4_handle_valid(handle));
294
ac27a0ec
DK		 295 if (bh)
296 BUFFER_TRACE(bh, "abort");
297
298 if (!handle->h_err)
299 handle->h_err = err;
300
301 if (is_handle_aborted(handle))
302 return;
303
90c7201b
TT		 304 printk(KERN_ERR "%s:%d: aborting transaction: %s in %s\n",
305 caller, line, errstr, err_fn);
ac27a0ec 306
dab291af 307 jbd2_journal_abort_handle(handle);
ac27a0ec
DK		 308}
309
1c13d5c0
TT		 310static void __save_error_info(struct super_block *sb, const char *func,
311 unsigned int line)
312{
313 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
314
315 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
316 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
317 es->s_last_error_time = cpu_to_le32(get_seconds());
318 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
319 es->s_last_error_line = cpu_to_le32(line);
320 if (!es->s_first_error_time) {
321 es->s_first_error_time = es->s_last_error_time;
322 strncpy(es->s_first_error_func, func,
323 sizeof(es->s_first_error_func));
324 es->s_first_error_line = cpu_to_le32(line);
325 es->s_first_error_ino = es->s_last_error_ino;
326 es->s_first_error_block = es->s_last_error_block;
327 }
66e61a9e
TT		 328 /*
329 * Start the daily error reporting function if it hasn't been
330 * started already
331 */
332 if (!es->s_error_count)
333 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
1c13d5c0
TT		 334 es->s_error_count = cpu_to_le32(le32_to_cpu(es->s_error_count) + 1);
335}
336
337static void save_error_info(struct super_block *sb, const char *func,
338 unsigned int line)
339{
340 __save_error_info(sb, func, line);
341 ext4_commit_super(sb, 1);
342}
343
344
ac27a0ec
DK		 345/* Deal with the reporting of failure conditions on a filesystem such as
346 * inconsistencies detected or read IO failures.
347 *
348 * On ext2, we can store the error state of the filesystem in the
617ba13b 349 * superblock. That is not possible on ext4, because we may have other
ac27a0ec
DK		 350 * write ordering constraints on the superblock which prevent us from
351 * writing it out straight away; and given that the journal is about to
352 * be aborted, we can't rely on the current, or future, transactions to
353 * write out the superblock safely.
354 *
dab291af 355 * We'll just use the jbd2_journal_abort() error code to record an error in
d6b198bc 356 * the journal instead. On recovery, the journal will complain about
ac27a0ec
DK		 357 * that error until we've noted it down and cleared it.
358 */
359
617ba13b 360static void ext4_handle_error(struct super_block *sb)
ac27a0ec 361{
ac27a0ec
DK		 362 if (sb->s_flags & MS_RDONLY)
363 return;
364
2b2d6d01 365 if (!test_opt(sb, ERRORS_CONT)) {
617ba13b 366 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 367
4ab2f15b 368 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
ac27a0ec 369 if (journal)
dab291af 370 jbd2_journal_abort(journal, -EIO);
ac27a0ec 371 }
2b2d6d01 372 if (test_opt(sb, ERRORS_RO)) {
b31e1552 373 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
ac27a0ec
DK		 374 sb->s_flags |= MS_RDONLY;
375 }
ac27a0ec 376 if (test_opt(sb, ERRORS_PANIC))
617ba13b 377 panic("EXT4-fs (device %s): panic forced after error\n",
ac27a0ec
DK		 378 sb->s_id);
379}
380
12062ddd 381void __ext4_error(struct super_block *sb, const char *function,
c398eda0 382 unsigned int line, const char *fmt, ...)
ac27a0ec
DK		 383{
384 va_list args;
385
386 va_start(args, fmt);
c398eda0
TT		 387 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: comm %s: ",
388 sb->s_id, function, line, current->comm);
ac27a0ec
DK		 389 vprintk(fmt, args);
390 printk("\n");
391 va_end(args);
392
617ba13b 393 ext4_handle_error(sb);
ac27a0ec
DK		 394}
395
c398eda0
TT		 396void ext4_error_inode(struct inode *inode, const char *function,
397 unsigned int line, ext4_fsblk_t block,
273df556
FM		 398 const char *fmt, ...)
399{
400 va_list args;
1c13d5c0 401 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
273df556 402
1c13d5c0
TT		 403 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
404 es->s_last_error_block = cpu_to_le64(block);
405 save_error_info(inode->i_sb, function, line);
273df556 406 va_start(args, fmt);
c398eda0
TT		 407 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: inode #%lu: ",
408 inode->i_sb->s_id, function, line, inode->i_ino);
409 if (block)
410 printk("block %llu: ", block);
411 printk("comm %s: ", current->comm);
273df556
FM		 412 vprintk(fmt, args);
413 printk("\n");
414 va_end(args);
415
416 ext4_handle_error(inode->i_sb);
417}
418
c398eda0
TT		 419void ext4_error_file(struct file *file, const char *function,
420 unsigned int line, const char *fmt, ...)
273df556
FM		 421{
422 va_list args;
1c13d5c0 423 struct ext4_super_block *es;
273df556
FM		 424 struct inode *inode = file->f_dentry->d_inode;
425 char pathname[80], *path;
426
1c13d5c0
TT		 427 es = EXT4_SB(inode->i_sb)->s_es;
428 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
429 save_error_info(inode->i_sb, function, line);
273df556
FM		 430 va_start(args, fmt);
431 path = d_path(&(file->f_path), pathname, sizeof(pathname));
432 if (!path)
433 path = "(unknown)";
434 printk(KERN_CRIT
c398eda0
TT		 435 "EXT4-fs error (device %s): %s:%d: inode #%lu "
436 "(comm %s path %s): ",
437 inode->i_sb->s_id, function, line, inode->i_ino,
438 current->comm, path);
273df556
FM		 439 vprintk(fmt, args);
440 printk("\n");
441 va_end(args);
442
443 ext4_handle_error(inode->i_sb);
444}
445
2b2d6d01 446static const char *ext4_decode_error(struct super_block *sb, int errno,
ac27a0ec
DK		 447 char nbuf[16])
448{
449 char *errstr = NULL;
450
451 switch (errno) {
452 case -EIO:
453 errstr = "IO failure";
454 break;
455 case -ENOMEM:
456 errstr = "Out of memory";
457 break;
458 case -EROFS:
78f1ddbb
TT		 459 if (!sb || (EXT4_SB(sb)->s_journal &&
460 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
ac27a0ec
DK		 461 errstr = "Journal has aborted";
462 else
463 errstr = "Readonly filesystem";
464 break;
465 default:
466 /* If the caller passed in an extra buffer for unknown
467 * errors, textualise them now. Else we just return
468 * NULL. */
469 if (nbuf) {
470 /* Check for truncated error codes... */
471 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
472 errstr = nbuf;
473 }
474 break;
475 }
476
477 return errstr;
478}
479
617ba13b 480/* __ext4_std_error decodes expected errors from journaling functions
ac27a0ec
DK		 481 * automatically and invokes the appropriate error response. */
482
c398eda0
TT		 483void __ext4_std_error(struct super_block *sb, const char *function,
484 unsigned int line, int errno)
ac27a0ec
DK		 485{
486 char nbuf[16];
487 const char *errstr;
488
489 /* Special case: if the error is EROFS, and we're not already
490 * inside a transaction, then there's really no point in logging
491 * an error. */
492 if (errno == -EROFS && journal_current_handle() == NULL &&
493 (sb->s_flags & MS_RDONLY))
494 return;
495
617ba13b 496 errstr = ext4_decode_error(sb, errno, nbuf);
c398eda0
TT		 497 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
498 sb->s_id, function, line, errstr);
1c13d5c0 499 save_error_info(sb, function, line);
ac27a0ec 500
617ba13b 501 ext4_handle_error(sb);
ac27a0ec
DK		 502}
503
504/*
617ba13b 505 * ext4_abort is a much stronger failure handler than ext4_error. The
ac27a0ec
DK		 506 * abort function may be used to deal with unrecoverable failures such
507 * as journal IO errors or ENOMEM at a critical moment in log management.
508 *
509 * We unconditionally force the filesystem into an ABORT|READONLY state,
510 * unless the error response on the fs has been set to panic in which
511 * case we take the easy way out and panic immediately.
512 */
513
c67d859e 514void __ext4_abort(struct super_block *sb, const char *function,
c398eda0 515 unsigned int line, const char *fmt, ...)
ac27a0ec
DK		 516{
517 va_list args;
518
1c13d5c0 519 save_error_info(sb, function, line);
ac27a0ec 520 va_start(args, fmt);
c398eda0
TT		 521 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
522 function, line);
ac27a0ec
DK		 523 vprintk(fmt, args);
524 printk("\n");
525 va_end(args);
526
1c13d5c0
TT		 527 if ((sb->s_flags & MS_RDONLY) == 0) {
528 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
529 sb->s_flags |= MS_RDONLY;
530 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
531 if (EXT4_SB(sb)->s_journal)
532 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
533 save_error_info(sb, function, line);
534 }
ac27a0ec 535 if (test_opt(sb, ERRORS_PANIC))
617ba13b 536 panic("EXT4-fs panic from previous error\n");
ac27a0ec
DK		 537}
538
b31e1552
ES		 539void ext4_msg (struct super_block * sb, const char *prefix,
540 const char *fmt, ...)
541{
542 va_list args;
543
544 va_start(args, fmt);
545 printk("%sEXT4-fs (%s): ", prefix, sb->s_id);
546 vprintk(fmt, args);
547 printk("\n");
548 va_end(args);
549}
550
12062ddd 551void __ext4_warning(struct super_block *sb, const char *function,
c398eda0 552 unsigned int line, const char *fmt, ...)
ac27a0ec
DK		 553{
554 va_list args;
555
556 va_start(args, fmt);
c398eda0
TT		 557 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: ",
558 sb->s_id, function, line);
ac27a0ec
DK		 559 vprintk(fmt, args);
560 printk("\n");
561 va_end(args);
562}
563
e29136f8
TT		 564void __ext4_grp_locked_error(const char *function, unsigned int line,
565 struct super_block *sb, ext4_group_t grp,
566 unsigned long ino, ext4_fsblk_t block,
567 const char *fmt, ...)
5d1b1b3f
AK		 568__releases(bitlock)
569__acquires(bitlock)
570{
571 va_list args;
572 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
573
1c13d5c0
TT		 574 es->s_last_error_ino = cpu_to_le32(ino);
575 es->s_last_error_block = cpu_to_le64(block);
576 __save_error_info(sb, function, line);
5d1b1b3f 577 va_start(args, fmt);
e29136f8
TT		 578 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u",
579 sb->s_id, function, line, grp);
580 if (ino)
581 printk("inode %lu: ", ino);
582 if (block)
583 printk("block %llu:", (unsigned long long) block);
5d1b1b3f
AK		 584 vprintk(fmt, args);
585 printk("\n");
586 va_end(args);
587
588 if (test_opt(sb, ERRORS_CONT)) {
e2d67052 589 ext4_commit_super(sb, 0);
5d1b1b3f
AK		 590 return;
591 }
1c13d5c0 592
5d1b1b3f
AK		 593 ext4_unlock_group(sb, grp);
594 ext4_handle_error(sb);
595 /*
596 * We only get here in the ERRORS_RO case; relocking the group
597 * may be dangerous, but nothing bad will happen since the
598 * filesystem will have already been marked read/only and the
599 * journal has been aborted. We return 1 as a hint to callers
600 * who might what to use the return value from
601 * ext4_grp_locked_error() to distinguish beween the
602 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
603 * aggressively from the ext4 function in question, with a
604 * more appropriate error code.
605 */
606 ext4_lock_group(sb, grp);
607 return;
608}
609
617ba13b 610void ext4_update_dynamic_rev(struct super_block *sb)
ac27a0ec 611{
617ba13b 612 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 613
617ba13b 614 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
ac27a0ec
DK		 615 return;
616
12062ddd 617 ext4_warning(sb,
ac27a0ec
DK		 618 "updating to rev %d because of new feature flag, "
619 "running e2fsck is recommended",
617ba13b 620 EXT4_DYNAMIC_REV);
ac27a0ec 621
617ba13b
MC		 622 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
623 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
624 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
ac27a0ec
DK		 625 /* leave es->s_feature_*compat flags alone */
626 /* es->s_uuid will be set by e2fsck if empty */
627
628 /*
629 * The rest of the superblock fields should be zero, and if not it
630 * means they are likely already in use, so leave them alone. We
631 * can leave it up to e2fsck to clean up any inconsistencies there.
632 */
633}
634
635/*
636 * Open the external journal device
637 */
b31e1552 638static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
ac27a0ec
DK		 639{
640 struct block_device *bdev;
641 char b[BDEVNAME_SIZE];
642
643 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
644 if (IS_ERR(bdev))
645 goto fail;
646 return bdev;
647
648fail:
b31e1552 649 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
ac27a0ec
DK		 650 __bdevname(dev, b), PTR_ERR(bdev));
651 return NULL;
652}
653
654/*
655 * Release the journal device
656 */
617ba13b 657static int ext4_blkdev_put(struct block_device *bdev)
ac27a0ec
DK		 658{
659 bd_release(bdev);
9a1c3542 660 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
ac27a0ec
DK		 661}
662
617ba13b 663static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
ac27a0ec
DK		 664{
665 struct block_device *bdev;
666 int ret = -ENODEV;
667
668 bdev = sbi->journal_bdev;
669 if (bdev) {
617ba13b 670 ret = ext4_blkdev_put(bdev);
ac27a0ec
DK		 671 sbi->journal_bdev = NULL;
672 }
673 return ret;
674}
675
676static inline struct inode *orphan_list_entry(struct list_head *l)
677{
617ba13b 678 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
ac27a0ec
DK		 679}
680
617ba13b 681static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
ac27a0ec
DK		 682{
683 struct list_head *l;
684
b31e1552
ES		 685 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
686 le32_to_cpu(sbi->s_es->s_last_orphan));
ac27a0ec
DK		 687
688 printk(KERN_ERR "sb_info orphan list:\n");
689 list_for_each(l, &sbi->s_orphan) {
690 struct inode *inode = orphan_list_entry(l);
691 printk(KERN_ERR " "
692 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
693 inode->i_sb->s_id, inode->i_ino, inode,
694 inode->i_mode, inode->i_nlink,
695 NEXT_ORPHAN(inode));
696 }
697}
698
2b2d6d01 699static void ext4_put_super(struct super_block *sb)
ac27a0ec 700{
617ba13b
MC		 701 struct ext4_sb_info *sbi = EXT4_SB(sb);
702 struct ext4_super_block *es = sbi->s_es;
ef2cabf7 703 int i, err;
ac27a0ec 704
e0ccfd95
CH		 705 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
706
4c0425ff
MC		 707 flush_workqueue(sbi->dio_unwritten_wq);
708 destroy_workqueue(sbi->dio_unwritten_wq);
709
a9e220f8 710 lock_super(sb);
6cfd0148 711 lock_kernel();
8c85e125 712 if (sb->s_dirt)
ebc1ac16 713 ext4_commit_super(sb, 1);
8c85e125 714
0390131b
FM		 715 if (sbi->s_journal) {
716 err = jbd2_journal_destroy(sbi->s_journal);
717 sbi->s_journal = NULL;
718 if (err < 0)
c67d859e 719 ext4_abort(sb, "Couldn't clean up the journal");
0390131b 720 }
d4edac31
JB		 721
722 ext4_release_system_zone(sb);
723 ext4_mb_release(sb);
724 ext4_ext_release(sb);
725 ext4_xattr_put_super(sb);
726
ac27a0ec 727 if (!(sb->s_flags & MS_RDONLY)) {
617ba13b 728 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 729 es->s_state = cpu_to_le16(sbi->s_mount_state);
e2d67052 730 ext4_commit_super(sb, 1);
ac27a0ec 731 }
240799cd 732 if (sbi->s_proc) {
9f6200bb 733 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 734 }
3197ebdb 735 kobject_del(&sbi->s_kobj);
ac27a0ec
DK		 736
737 for (i = 0; i < sbi->s_gdb_count; i++)
738 brelse(sbi->s_group_desc[i]);
739 kfree(sbi->s_group_desc);
c5ca7c76
TT		 740 if (is_vmalloc_addr(sbi->s_flex_groups))
741 vfree(sbi->s_flex_groups);
742 else
743 kfree(sbi->s_flex_groups);
ac27a0ec
DK		 744 percpu_counter_destroy(&sbi->s_freeblocks_counter);
745 percpu_counter_destroy(&sbi->s_freeinodes_counter);
746 percpu_counter_destroy(&sbi->s_dirs_counter);
6bc6e63f 747 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
ac27a0ec
DK		 748 brelse(sbi->s_sbh);
749#ifdef CONFIG_QUOTA
750 for (i = 0; i < MAXQUOTAS; i++)
751 kfree(sbi->s_qf_names[i]);
752#endif
753
754 /* Debugging code just in case the in-memory inode orphan list
755 * isn't empty. The on-disk one can be non-empty if we've
756 * detected an error and taken the fs readonly, but the
757 * in-memory list had better be clean by this point. */
758 if (!list_empty(&sbi->s_orphan))
759 dump_orphan_list(sb, sbi);
760 J_ASSERT(list_empty(&sbi->s_orphan));
761
f98393a6 762 invalidate_bdev(sb->s_bdev);
ac27a0ec
DK		 763 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
764 /*
765 * Invalidate the journal device's buffers. We don't want them
766 * floating about in memory - the physical journal device may
767 * hotswapped, and it breaks the `ro-after' testing code.
768 */
769 sync_blockdev(sbi->journal_bdev);
f98393a6 770 invalidate_bdev(sbi->journal_bdev);
617ba13b 771 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 772 }
773 sb->s_fs_info = NULL;
3197ebdb
TT		 774 /*
775 * Now that we are completely done shutting down the
776 * superblock, we need to actually destroy the kobject.
777 */
778 unlock_kernel();
779 unlock_super(sb);
780 kobject_put(&sbi->s_kobj);
781 wait_for_completion(&sbi->s_kobj_unregister);
705895b6 782 kfree(sbi->s_blockgroup_lock);
ac27a0ec 783 kfree(sbi);
ac27a0ec
DK		 784}
785
e18b890b 786static struct kmem_cache *ext4_inode_cachep;
ac27a0ec
DK		 787
788/*
789 * Called inside transaction, so use GFP_NOFS
790 */
617ba13b 791static struct inode *ext4_alloc_inode(struct super_block *sb)
ac27a0ec 792{
617ba13b 793 struct ext4_inode_info *ei;
ac27a0ec 794
e6b4f8da 795 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
ac27a0ec
DK		 796 if (!ei)
797 return NULL;
0b8e58a1 798
ac27a0ec 799 ei->vfs_inode.i_version = 1;
91246c00 800 ei->vfs_inode.i_data.writeback_index = 0;
a86c6181 801 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
c9de560d
AT		 802 INIT_LIST_HEAD(&ei->i_prealloc_list);
803 spin_lock_init(&ei->i_prealloc_lock);
0390131b
FM		 804 /*
805 * Note: We can be called before EXT4_SB(sb)->s_journal is set,
806 * therefore it can be null here. Don't check it, just initialize
807 * jinode.
808 */
678aaf48 809 jbd2_journal_init_jbd_inode(&ei->jinode, &ei->vfs_inode);
d2a17637
MC		 810 ei->i_reserved_data_blocks = 0;
811 ei->i_reserved_meta_blocks = 0;
812 ei->i_allocated_meta_blocks = 0;
9d0be502 813 ei->i_da_metadata_calc_len = 0;
d2a17637
MC		 814 ei->i_delalloc_reserved_flag = 0;
815 spin_lock_init(&(ei->i_block_reservation_lock));
a9e7f447
DM		 816#ifdef CONFIG_QUOTA
817 ei->i_reserved_quota = 0;
818#endif
c7064ef1 819 INIT_LIST_HEAD(&ei->i_completed_io_list);
744692dc 820 spin_lock_init(&ei->i_completed_io_lock);
8d5d02e6 821 ei->cur_aio_dio = NULL;
b436b9be
JK		 822 ei->i_sync_tid = 0;
823 ei->i_datasync_tid = 0;
0b8e58a1 824
ac27a0ec
DK		 825 return &ei->vfs_inode;
826}
827
617ba13b 828static void ext4_destroy_inode(struct inode *inode)
ac27a0ec 829{
9f7dd93d 830 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
b31e1552
ES		 831 ext4_msg(inode->i_sb, KERN_ERR,
832 "Inode %lu (%p): orphan list check failed!",
833 inode->i_ino, EXT4_I(inode));
9f7dd93d
VA		 834 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
835 EXT4_I(inode), sizeof(struct ext4_inode_info),
836 true);
837 dump_stack();
838 }
617ba13b 839 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
ac27a0ec
DK		 840}
841
51cc5068 842static void init_once(void *foo)
ac27a0ec 843{
617ba13b 844 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
ac27a0ec 845
a35afb83 846 INIT_LIST_HEAD(&ei->i_orphan);
03010a33 847#ifdef CONFIG_EXT4_FS_XATTR
a35afb83 848 init_rwsem(&ei->xattr_sem);
ac27a0ec 849#endif
0e855ac8 850 init_rwsem(&ei->i_data_sem);
a35afb83 851 inode_init_once(&ei->vfs_inode);
ac27a0ec
DK		 852}
853
854static int init_inodecache(void)
855{
617ba13b
MC		 856 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
857 sizeof(struct ext4_inode_info),
ac27a0ec
DK		 858 0, (SLAB_RECLAIM_ACCOUNT|
859 SLAB_MEM_SPREAD),
20c2df83 860 init_once);
617ba13b 861 if (ext4_inode_cachep == NULL)
ac27a0ec
DK		 862 return -ENOMEM;
863 return 0;
864}
865
866static void destroy_inodecache(void)
867{
617ba13b 868 kmem_cache_destroy(ext4_inode_cachep);
ac27a0ec
DK		 869}
870
617ba13b 871static void ext4_clear_inode(struct inode *inode)
ac27a0ec 872{
9f754758 873 dquot_drop(inode);
c2ea3fde 874 ext4_discard_preallocations(inode);
0390131b
FM		 875 if (EXT4_JOURNAL(inode))
876 jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal,
678aaf48 877 &EXT4_I(inode)->jinode);
ac27a0ec
DK		 878}
879
2b2d6d01
TT		 880static inline void ext4_show_quota_options(struct seq_file *seq,
881 struct super_block *sb)
ac27a0ec
DK		 882{
883#if defined(CONFIG_QUOTA)
617ba13b 884 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 885
5a20bdfc
JK		 886 if (sbi->s_jquota_fmt) {
887 char *fmtname = "";
888
889 switch (sbi->s_jquota_fmt) {
890 case QFMT_VFS_OLD:
891 fmtname = "vfsold";
892 break;
893 case QFMT_VFS_V0:
894 fmtname = "vfsv0";
895 break;
896 case QFMT_VFS_V1:
897 fmtname = "vfsv1";
898 break;
899 }
900 seq_printf(seq, ",jqfmt=%s", fmtname);
901 }
ac27a0ec
DK		 902
903 if (sbi->s_qf_names[USRQUOTA])
904 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
905
906 if (sbi->s_qf_names[GRPQUOTA])
907 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
908
482a7425 909 if (test_opt(sb, USRQUOTA))
ac27a0ec
DK		 910 seq_puts(seq, ",usrquota");
911
482a7425 912 if (test_opt(sb, GRPQUOTA))
ac27a0ec
DK		 913 seq_puts(seq, ",grpquota");
914#endif
915}
916
d9c9bef1
MS		 917/*
918 * Show an option if
919 * - it's set to a non-default value OR
920 * - if the per-sb default is different from the global default
921 */
617ba13b 922static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
ac27a0ec 923{
aa22df2c
AK		 924 int def_errors;
925 unsigned long def_mount_opts;
ac27a0ec 926 struct super_block *sb = vfs->mnt_sb;
d9c9bef1
MS		 927 struct ext4_sb_info *sbi = EXT4_SB(sb);
928 struct ext4_super_block *es = sbi->s_es;
d9c9bef1
MS		 929
930 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
aa22df2c 931 def_errors = le16_to_cpu(es->s_errors);
d9c9bef1
MS		 932
933 if (sbi->s_sb_block != 1)
934 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
935 if (test_opt(sb, MINIX_DF))
936 seq_puts(seq, ",minixdf");
aa22df2c 937 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
d9c9bef1
MS		 938 seq_puts(seq, ",grpid");
939 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
940 seq_puts(seq, ",nogrpid");
941 if (sbi->s_resuid != EXT4_DEF_RESUID ||
942 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
943 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
944 }
945 if (sbi->s_resgid != EXT4_DEF_RESGID ||
946 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
947 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
948 }
bb4f397a 949 if (test_opt(sb, ERRORS_RO)) {
d9c9bef1 950 if (def_errors == EXT4_ERRORS_PANIC ||
bb4f397a
AK		 951 def_errors == EXT4_ERRORS_CONTINUE) {
952 seq_puts(seq, ",errors=remount-ro");
d9c9bef1
MS		 953 }
954 }
aa22df2c 955 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
bb4f397a 956 seq_puts(seq, ",errors=continue");
aa22df2c 957 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
d9c9bef1 958 seq_puts(seq, ",errors=panic");
aa22df2c 959 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
d9c9bef1 960 seq_puts(seq, ",nouid32");
aa22df2c 961 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
d9c9bef1
MS		 962 seq_puts(seq, ",debug");
963 if (test_opt(sb, OLDALLOC))
964 seq_puts(seq, ",oldalloc");
03010a33 965#ifdef CONFIG_EXT4_FS_XATTR
aa22df2c
AK		 966 if (test_opt(sb, XATTR_USER) &&
967 !(def_mount_opts & EXT4_DEFM_XATTR_USER))
d9c9bef1
MS		 968 seq_puts(seq, ",user_xattr");
969 if (!test_opt(sb, XATTR_USER) &&
970 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
971 seq_puts(seq, ",nouser_xattr");
972 }
973#endif
03010a33 974#ifdef CONFIG_EXT4_FS_POSIX_ACL
aa22df2c 975 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
d9c9bef1
MS		 976 seq_puts(seq, ",acl");
977 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
978 seq_puts(seq, ",noacl");
979#endif
30773840 980 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
d9c9bef1
MS		 981 seq_printf(seq, ",commit=%u",
982 (unsigned) (sbi->s_commit_interval / HZ));
983 }
30773840
TT		 984 if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
985 seq_printf(seq, ",min_batch_time=%u",
986 (unsigned) sbi->s_min_batch_time);
987 }
988 if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
989 seq_printf(seq, ",max_batch_time=%u",
990 (unsigned) sbi->s_min_batch_time);
991 }
992
571640ca
ES		 993 /*
994 * We're changing the default of barrier mount option, so
995 * let's always display its mount state so it's clear what its
996 * status is.
997 */
998 seq_puts(seq, ",barrier=");
999 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
cd0b6a39
TT		 1000 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
1001 seq_puts(seq, ",journal_async_commit");
39a4bade
JK		 1002 else if (test_opt(sb, JOURNAL_CHECKSUM))
1003 seq_puts(seq, ",journal_checksum");
25ec56b5
JNC		 1004 if (test_opt(sb, I_VERSION))
1005 seq_puts(seq, ",i_version");
dd919b98
AK		 1006 if (!test_opt(sb, DELALLOC))
1007 seq_puts(seq, ",nodelalloc");
1008
ac27a0ec 1009
cb45bbe4
MS		 1010 if (sbi->s_stripe)
1011 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
aa22df2c
AK		 1012 /*
1013 * journal mode get enabled in different ways
1014 * So just print the value even if we didn't specify it
1015 */
617ba13b 1016 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
ac27a0ec 1017 seq_puts(seq, ",data=journal");
617ba13b 1018 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
ac27a0ec 1019 seq_puts(seq, ",data=ordered");
617ba13b 1020 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
ac27a0ec
DK		 1021 seq_puts(seq, ",data=writeback");
1022
240799cd
TT		 1023 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1024 seq_printf(seq, ",inode_readahead_blks=%u",
1025 sbi->s_inode_readahead_blks);
1026
5bf5683a
HK		 1027 if (test_opt(sb, DATA_ERR_ABORT))
1028 seq_puts(seq, ",data_err=abort");
1029
afd4672d 1030 if (test_opt(sb, NO_AUTO_DA_ALLOC))
06705bff 1031 seq_puts(seq, ",noauto_da_alloc");
afd4672d 1032
5328e635
ES		 1033 if (test_opt(sb, DISCARD))
1034 seq_puts(seq, ",discard");
1035
e3bb52ae
ES		 1036 if (test_opt(sb, NOLOAD))
1037 seq_puts(seq, ",norecovery");
1038
744692dc
JZ		 1039 if (test_opt(sb, DIOREAD_NOLOCK))
1040 seq_puts(seq, ",dioread_nolock");
1041
617ba13b 1042 ext4_show_quota_options(seq, sb);
0b8e58a1 1043
ac27a0ec
DK		 1044 return 0;
1045}
1046
1b961ac0 1047static struct inode *ext4_nfs_get_inode(struct super_block *sb,
0b8e58a1 1048 u64 ino, u32 generation)
ac27a0ec 1049{
ac27a0ec 1050 struct inode *inode;
ac27a0ec 1051
617ba13b 1052 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
ac27a0ec 1053 return ERR_PTR(-ESTALE);
617ba13b 1054 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
ac27a0ec
DK		 1055 return ERR_PTR(-ESTALE);
1056
1057 /* iget isn't really right if the inode is currently unallocated!!
1058 *
617ba13b 1059 * ext4_read_inode will return a bad_inode if the inode had been
ac27a0ec
DK		 1060 * deleted, so we should be safe.
1061 *
1062 * Currently we don't know the generation for parent directory, so
1063 * a generation of 0 means "accept any"
1064 */
1d1fe1ee
DH		 1065 inode = ext4_iget(sb, ino);
1066 if (IS_ERR(inode))
1067 return ERR_CAST(inode);
1068 if (generation && inode->i_generation != generation) {
ac27a0ec
DK		 1069 iput(inode);
1070 return ERR_PTR(-ESTALE);
1071 }
1b961ac0
CH		 1072
1073 return inode;
1074}
1075
1076static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
0b8e58a1 1077 int fh_len, int fh_type)
1b961ac0
CH		 1078{
1079 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1080 ext4_nfs_get_inode);
1081}
1082
1083static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
0b8e58a1 1084 int fh_len, int fh_type)
1b961ac0
CH		 1085{
1086 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1087 ext4_nfs_get_inode);
ac27a0ec
DK		 1088}
1089
c39a7f84
TO		 1090/*
1091 * Try to release metadata pages (indirect blocks, directories) which are
1092 * mapped via the block device. Since these pages could have journal heads
1093 * which would prevent try_to_free_buffers() from freeing them, we must use
1094 * jbd2 layer's try_to_free_buffers() function to release them.
1095 */
0b8e58a1
AD		 1096static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1097 gfp_t wait)
c39a7f84
TO		 1098{
1099 journal_t *journal = EXT4_SB(sb)->s_journal;
1100
1101 WARN_ON(PageChecked(page));
1102 if (!page_has_buffers(page))
1103 return 0;
1104 if (journal)
1105 return jbd2_journal_try_to_free_buffers(journal, page,
1106 wait & ~__GFP_WAIT);
1107 return try_to_free_buffers(page);
1108}
1109
ac27a0ec 1110#ifdef CONFIG_QUOTA
af5bc92d 1111#define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
2b2d6d01 1112#define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
ac27a0ec 1113
617ba13b
MC		 1114static int ext4_write_dquot(struct dquot *dquot);
1115static int ext4_acquire_dquot(struct dquot *dquot);
1116static int ext4_release_dquot(struct dquot *dquot);
1117static int ext4_mark_dquot_dirty(struct dquot *dquot);
1118static int ext4_write_info(struct super_block *sb, int type);
6f28e087 1119static int ext4_quota_on(struct super_block *sb, int type, int format_id,
307ae18a 1120 char *path);
ca0e05e4 1121static int ext4_quota_off(struct super_block *sb, int type);
617ba13b
MC		 1122static int ext4_quota_on_mount(struct super_block *sb, int type);
1123static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 1124 size_t len, loff_t off);
617ba13b 1125static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 1126 const char *data, size_t len, loff_t off);
1127
61e225dc 1128static const struct dquot_operations ext4_quota_operations = {
a9e7f447 1129#ifdef CONFIG_QUOTA
60e58e0f 1130 .get_reserved_space = ext4_get_reserved_space,
a9e7f447 1131#endif
617ba13b
MC		 1132 .write_dquot = ext4_write_dquot,
1133 .acquire_dquot = ext4_acquire_dquot,
1134 .release_dquot = ext4_release_dquot,
1135 .mark_dirty = ext4_mark_dquot_dirty,
a5b5ee32
JK		 1136 .write_info = ext4_write_info,
1137 .alloc_dquot = dquot_alloc,
1138 .destroy_dquot = dquot_destroy,
ac27a0ec
DK		 1139};
1140
0d54b217 1141static const struct quotactl_ops ext4_qctl_operations = {
617ba13b 1142 .quota_on = ext4_quota_on,
ca0e05e4 1143 .quota_off = ext4_quota_off,
287a8095
CH		 1144 .quota_sync = dquot_quota_sync,
1145 .get_info = dquot_get_dqinfo,
1146 .set_info = dquot_set_dqinfo,
1147 .get_dqblk = dquot_get_dqblk,
1148 .set_dqblk = dquot_set_dqblk
ac27a0ec
DK		 1149};
1150#endif
1151
ee9b6d61 1152static const struct super_operations ext4_sops = {
617ba13b
MC		 1153 .alloc_inode = ext4_alloc_inode,
1154 .destroy_inode = ext4_destroy_inode,
617ba13b
MC		 1155 .write_inode = ext4_write_inode,
1156 .dirty_inode = ext4_dirty_inode,
1157 .delete_inode = ext4_delete_inode,
1158 .put_super = ext4_put_super,
617ba13b 1159 .sync_fs = ext4_sync_fs,
c4be0c1d
TS		 1160 .freeze_fs = ext4_freeze,
1161 .unfreeze_fs = ext4_unfreeze,
617ba13b
MC		 1162 .statfs = ext4_statfs,
1163 .remount_fs = ext4_remount,
1164 .clear_inode = ext4_clear_inode,
1165 .show_options = ext4_show_options,
ac27a0ec 1166#ifdef CONFIG_QUOTA
617ba13b
MC		 1167 .quota_read = ext4_quota_read,
1168 .quota_write = ext4_quota_write,
ac27a0ec 1169#endif
c39a7f84 1170 .bdev_try_to_free_page = bdev_try_to_free_page,
ac27a0ec
DK		 1171};
1172
9ca92389
TT		 1173static const struct super_operations ext4_nojournal_sops = {
1174 .alloc_inode = ext4_alloc_inode,
1175 .destroy_inode = ext4_destroy_inode,
1176 .write_inode = ext4_write_inode,
1177 .dirty_inode = ext4_dirty_inode,
1178 .delete_inode = ext4_delete_inode,
1179 .write_super = ext4_write_super,
1180 .put_super = ext4_put_super,
1181 .statfs = ext4_statfs,
1182 .remount_fs = ext4_remount,
1183 .clear_inode = ext4_clear_inode,
1184 .show_options = ext4_show_options,
1185#ifdef CONFIG_QUOTA
1186 .quota_read = ext4_quota_read,
1187 .quota_write = ext4_quota_write,
1188#endif
1189 .bdev_try_to_free_page = bdev_try_to_free_page,
1190};
1191
39655164 1192static const struct export_operations ext4_export_ops = {
1b961ac0
CH		 1193 .fh_to_dentry = ext4_fh_to_dentry,
1194 .fh_to_parent = ext4_fh_to_parent,
617ba13b 1195 .get_parent = ext4_get_parent,
ac27a0ec
DK		 1196};
1197
1198enum {
1199 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1200 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
01436ef2 1201 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
ac27a0ec 1202 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
06705bff 1203 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh,
30773840 1204 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
c3191067 1205 Opt_journal_update, Opt_journal_dev,
818d276c 1206 Opt_journal_checksum, Opt_journal_async_commit,
ac27a0ec 1207 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
296c355c 1208 Opt_data_err_abort, Opt_data_err_ignore,
ac27a0ec 1209 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
5a20bdfc
JK		 1210 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1211 Opt_noquota, Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err,
1212 Opt_resize, Opt_usrquota, Opt_grpquota, Opt_i_version,
01436ef2 1213 Opt_stripe, Opt_delalloc, Opt_nodelalloc,
6fd058f7 1214 Opt_block_validity, Opt_noblock_validity,
5328e635 1215 Opt_inode_readahead_blks, Opt_journal_ioprio,
744692dc 1216 Opt_dioread_nolock, Opt_dioread_lock,
5328e635 1217 Opt_discard, Opt_nodiscard,
ac27a0ec
DK		 1218};
1219
a447c093 1220static const match_table_t tokens = {
ac27a0ec
DK		 1221 {Opt_bsd_df, "bsddf"},
1222 {Opt_minix_df, "minixdf"},
1223 {Opt_grpid, "grpid"},
1224 {Opt_grpid, "bsdgroups"},
1225 {Opt_nogrpid, "nogrpid"},
1226 {Opt_nogrpid, "sysvgroups"},
1227 {Opt_resgid, "resgid=%u"},
1228 {Opt_resuid, "resuid=%u"},
1229 {Opt_sb, "sb=%u"},
1230 {Opt_err_cont, "errors=continue"},
1231 {Opt_err_panic, "errors=panic"},
1232 {Opt_err_ro, "errors=remount-ro"},
1233 {Opt_nouid32, "nouid32"},
ac27a0ec
DK		 1234 {Opt_debug, "debug"},
1235 {Opt_oldalloc, "oldalloc"},
1236 {Opt_orlov, "orlov"},
1237 {Opt_user_xattr, "user_xattr"},
1238 {Opt_nouser_xattr, "nouser_xattr"},
1239 {Opt_acl, "acl"},
1240 {Opt_noacl, "noacl"},
ac27a0ec 1241 {Opt_noload, "noload"},
e3bb52ae 1242 {Opt_noload, "norecovery"},
ac27a0ec
DK		 1243 {Opt_nobh, "nobh"},
1244 {Opt_bh, "bh"},
1245 {Opt_commit, "commit=%u"},
30773840
TT		 1246 {Opt_min_batch_time, "min_batch_time=%u"},
1247 {Opt_max_batch_time, "max_batch_time=%u"},
ac27a0ec 1248 {Opt_journal_update, "journal=update"},
ac27a0ec 1249 {Opt_journal_dev, "journal_dev=%u"},
818d276c
GS		 1250 {Opt_journal_checksum, "journal_checksum"},
1251 {Opt_journal_async_commit, "journal_async_commit"},
ac27a0ec
DK		 1252 {Opt_abort, "abort"},
1253 {Opt_data_journal, "data=journal"},
1254 {Opt_data_ordered, "data=ordered"},
1255 {Opt_data_writeback, "data=writeback"},
5bf5683a
HK		 1256 {Opt_data_err_abort, "data_err=abort"},
1257 {Opt_data_err_ignore, "data_err=ignore"},
ac27a0ec
DK		 1258 {Opt_offusrjquota, "usrjquota="},
1259 {Opt_usrjquota, "usrjquota=%s"},
1260 {Opt_offgrpjquota, "grpjquota="},
1261 {Opt_grpjquota, "grpjquota=%s"},
1262 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1263 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
5a20bdfc 1264 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
ac27a0ec
DK		 1265 {Opt_grpquota, "grpquota"},
1266 {Opt_noquota, "noquota"},
1267 {Opt_quota, "quota"},
1268 {Opt_usrquota, "usrquota"},
1269 {Opt_barrier, "barrier=%u"},
06705bff
TT		 1270 {Opt_barrier, "barrier"},
1271 {Opt_nobarrier, "nobarrier"},
25ec56b5 1272 {Opt_i_version, "i_version"},
c9de560d 1273 {Opt_stripe, "stripe=%u"},
ac27a0ec 1274 {Opt_resize, "resize"},
64769240 1275 {Opt_delalloc, "delalloc"},
dd919b98 1276 {Opt_nodelalloc, "nodelalloc"},
6fd058f7
TT		 1277 {Opt_block_validity, "block_validity"},
1278 {Opt_noblock_validity, "noblock_validity"},
240799cd 1279 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
b3881f74 1280 {Opt_journal_ioprio, "journal_ioprio=%u"},
afd4672d 1281 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
06705bff
TT		 1282 {Opt_auto_da_alloc, "auto_da_alloc"},
1283 {Opt_noauto_da_alloc, "noauto_da_alloc"},
744692dc
JZ		 1284 {Opt_dioread_nolock, "dioread_nolock"},
1285 {Opt_dioread_lock, "dioread_lock"},
5328e635
ES		 1286 {Opt_discard, "discard"},
1287 {Opt_nodiscard, "nodiscard"},
f3f12faa 1288 {Opt_err, NULL},
ac27a0ec
DK		 1289};
1290
617ba13b 1291static ext4_fsblk_t get_sb_block(void **data)
ac27a0ec 1292{
617ba13b 1293 ext4_fsblk_t sb_block;
ac27a0ec
DK		 1294 char *options = (char *) *data;
1295
1296 if (!options || strncmp(options, "sb=", 3) != 0)
1297 return 1; /* Default location */
0b8e58a1 1298
ac27a0ec 1299 options += 3;
0b8e58a1 1300 /* TODO: use simple_strtoll with >32bit ext4 */
ac27a0ec
DK		 1301 sb_block = simple_strtoul(options, &options, 0);
1302 if (*options && *options != ',') {
4776004f 1303 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
ac27a0ec
DK		 1304 (char *) *data);
1305 return 1;
1306 }
1307 if (*options == ',')
1308 options++;
1309 *data = (void *) options;
0b8e58a1 1310
ac27a0ec
DK		 1311 return sb_block;
1312}
1313
b3881f74 1314#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
437ca0fd
DM		 1315static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1316 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
b3881f74 1317
56c50f11
DM		 1318#ifdef CONFIG_QUOTA
1319static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1320{
1321 struct ext4_sb_info *sbi = EXT4_SB(sb);
1322 char *qname;
1323
1324 if (sb_any_quota_loaded(sb) &&
1325 !sbi->s_qf_names[qtype]) {
1326 ext4_msg(sb, KERN_ERR,
1327 "Cannot change journaled "
1328 "quota options when quota turned on");
1329 return 0;
1330 }
1331 qname = match_strdup(args);
1332 if (!qname) {
1333 ext4_msg(sb, KERN_ERR,
1334 "Not enough memory for storing quotafile name");
1335 return 0;
1336 }
1337 if (sbi->s_qf_names[qtype] &&
1338 strcmp(sbi->s_qf_names[qtype], qname)) {
1339 ext4_msg(sb, KERN_ERR,
1340 "%s quota file already specified", QTYPE2NAME(qtype));
1341 kfree(qname);
1342 return 0;
1343 }
1344 sbi->s_qf_names[qtype] = qname;
1345 if (strchr(sbi->s_qf_names[qtype], '/')) {
1346 ext4_msg(sb, KERN_ERR,
1347 "quotafile must be on filesystem root");
1348 kfree(sbi->s_qf_names[qtype]);
1349 sbi->s_qf_names[qtype] = NULL;
1350 return 0;
1351 }
1352 set_opt(sbi->s_mount_opt, QUOTA);
1353 return 1;
1354}
1355
1356static int clear_qf_name(struct super_block *sb, int qtype)
1357{
1358
1359 struct ext4_sb_info *sbi = EXT4_SB(sb);
1360
1361 if (sb_any_quota_loaded(sb) &&
1362 sbi->s_qf_names[qtype]) {
1363 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1364 " when quota turned on");
1365 return 0;
1366 }
1367 /*
1368 * The space will be released later when all options are confirmed
1369 * to be correct
1370 */
1371 sbi->s_qf_names[qtype] = NULL;
1372 return 1;
1373}
1374#endif
1375
2b2d6d01 1376static int parse_options(char *options, struct super_block *sb,
c3191067 1377 unsigned long *journal_devnum,
b3881f74 1378 unsigned int *journal_ioprio,
2b2d6d01 1379 ext4_fsblk_t *n_blocks_count, int is_remount)
ac27a0ec 1380{
617ba13b 1381 struct ext4_sb_info *sbi = EXT4_SB(sb);
2b2d6d01 1382 char *p;
ac27a0ec
DK		 1383 substring_t args[MAX_OPT_ARGS];
1384 int data_opt = 0;
1385 int option;
1386#ifdef CONFIG_QUOTA
56c50f11 1387 int qfmt;
ac27a0ec
DK		 1388#endif
1389
1390 if (!options)
1391 return 1;
1392
2b2d6d01 1393 while ((p = strsep(&options, ",")) != NULL) {
ac27a0ec
DK		 1394 int token;
1395 if (!*p)
1396 continue;
1397
15121c18
ES		 1398 /*
1399 * Initialize args struct so we know whether arg was
1400 * found; some options take optional arguments.
1401 */
1402 args[0].to = args[0].from = 0;
ac27a0ec
DK		 1403 token = match_token(p, tokens, args);
1404 switch (token) {
1405 case Opt_bsd_df:
437ca0fd 1406 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
2b2d6d01 1407 clear_opt(sbi->s_mount_opt, MINIX_DF);
ac27a0ec
DK		 1408 break;
1409 case Opt_minix_df:
437ca0fd 1410 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
2b2d6d01 1411 set_opt(sbi->s_mount_opt, MINIX_DF);
437ca0fd 1412
ac27a0ec
DK		 1413 break;
1414 case Opt_grpid:
437ca0fd 1415 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
2b2d6d01 1416 set_opt(sbi->s_mount_opt, GRPID);
437ca0fd 1417
ac27a0ec
DK		 1418 break;
1419 case Opt_nogrpid:
437ca0fd 1420 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
2b2d6d01 1421 clear_opt(sbi->s_mount_opt, GRPID);
437ca0fd 1422
ac27a0ec
DK		 1423 break;
1424 case Opt_resuid:
1425 if (match_int(&args[0], &option))
1426 return 0;
1427 sbi->s_resuid = option;
1428 break;
1429 case Opt_resgid:
1430 if (match_int(&args[0], &option))
1431 return 0;
1432 sbi->s_resgid = option;
1433 break;
1434 case Opt_sb:
1435 /* handled by get_sb_block() instead of here */
1436 /* *sb_block = match_int(&args[0]); */
1437 break;
1438 case Opt_err_panic:
2b2d6d01
TT		 1439 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1440 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1441 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
ac27a0ec
DK		 1442 break;
1443 case Opt_err_ro:
2b2d6d01
TT		 1444 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1445 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1446 set_opt(sbi->s_mount_opt, ERRORS_RO);
ac27a0ec
DK		 1447 break;
1448 case Opt_err_cont:
2b2d6d01
TT		 1449 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1450 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1451 set_opt(sbi->s_mount_opt, ERRORS_CONT);
ac27a0ec
DK		 1452 break;
1453 case Opt_nouid32:
2b2d6d01 1454 set_opt(sbi->s_mount_opt, NO_UID32);
ac27a0ec 1455 break;
ac27a0ec 1456 case Opt_debug:
2b2d6d01 1457 set_opt(sbi->s_mount_opt, DEBUG);
ac27a0ec
DK		 1458 break;
1459 case Opt_oldalloc:
2b2d6d01 1460 set_opt(sbi->s_mount_opt, OLDALLOC);
ac27a0ec
DK		 1461 break;
1462 case Opt_orlov:
2b2d6d01 1463 clear_opt(sbi->s_mount_opt, OLDALLOC);
ac27a0ec 1464 break;
03010a33 1465#ifdef CONFIG_EXT4_FS_XATTR
ac27a0ec 1466 case Opt_user_xattr:
2b2d6d01 1467 set_opt(sbi->s_mount_opt, XATTR_USER);
ac27a0ec
DK		 1468 break;
1469 case Opt_nouser_xattr:
2b2d6d01 1470 clear_opt(sbi->s_mount_opt, XATTR_USER);
ac27a0ec
DK		 1471 break;
1472#else
1473 case Opt_user_xattr:
1474 case Opt_nouser_xattr:
b31e1552 1475 ext4_msg(sb, KERN_ERR, "(no)user_xattr options not supported");
ac27a0ec
DK		 1476 break;
1477#endif
03010a33 1478#ifdef CONFIG_EXT4_FS_POSIX_ACL
ac27a0ec
DK		 1479 case Opt_acl:
1480 set_opt(sbi->s_mount_opt, POSIX_ACL);
1481 break;
1482 case Opt_noacl:
1483 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1484 break;
1485#else
1486 case Opt_acl:
1487 case Opt_noacl:
b31e1552 1488 ext4_msg(sb, KERN_ERR, "(no)acl options not supported");
ac27a0ec
DK		 1489 break;
1490#endif
ac27a0ec
DK		 1491 case Opt_journal_update:
1492 /* @@@ FIXME */
1493 /* Eventually we will want to be able to create
1494 a journal file here. For now, only allow the
1495 user to specify an existing inode to be the
1496 journal file. */
1497 if (is_remount) {
b31e1552
ES		 1498 ext4_msg(sb, KERN_ERR,
1499 "Cannot specify journal on remount");
ac27a0ec
DK		 1500 return 0;
1501 }
2b2d6d01 1502 set_opt(sbi->s_mount_opt, UPDATE_JOURNAL);
ac27a0ec 1503 break;
ac27a0ec
DK		 1504 case Opt_journal_dev:
1505 if (is_remount) {
b31e1552
ES		 1506 ext4_msg(sb, KERN_ERR,
1507 "Cannot specify journal on remount");
ac27a0ec
DK		 1508 return 0;
1509 }
1510 if (match_int(&args[0], &option))
1511 return 0;
1512 *journal_devnum = option;
1513 break;
818d276c 1514 case Opt_journal_checksum:
d4da6c9c
LT		 1515 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1516 break;
818d276c
GS		 1517 case Opt_journal_async_commit:
1518 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
d4da6c9c 1519 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
818d276c 1520 break;
ac27a0ec 1521 case Opt_noload:
2b2d6d01 1522 set_opt(sbi->s_mount_opt, NOLOAD);
ac27a0ec
DK		 1523 break;
1524 case Opt_commit:
1525 if (match_int(&args[0], &option))
1526 return 0;
1527 if (option < 0)
1528 return 0;
1529 if (option == 0)
cd02ff0b 1530 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
ac27a0ec
DK		 1531 sbi->s_commit_interval = HZ * option;
1532 break;
30773840
TT		 1533 case Opt_max_batch_time:
1534 if (match_int(&args[0], &option))
1535 return 0;
1536 if (option < 0)
1537 return 0;
1538 if (option == 0)
1539 option = EXT4_DEF_MAX_BATCH_TIME;
1540 sbi->s_max_batch_time = option;
1541 break;
1542 case Opt_min_batch_time:
1543 if (match_int(&args[0], &option))
1544 return 0;
1545 if (option < 0)
1546 return 0;
1547 sbi->s_min_batch_time = option;
1548 break;
ac27a0ec 1549 case Opt_data_journal:
617ba13b 1550 data_opt = EXT4_MOUNT_JOURNAL_DATA;
ac27a0ec
DK		 1551 goto datacheck;
1552 case Opt_data_ordered:
617ba13b 1553 data_opt = EXT4_MOUNT_ORDERED_DATA;
ac27a0ec
DK		 1554 goto datacheck;
1555 case Opt_data_writeback:
617ba13b 1556 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
ac27a0ec
DK		 1557 datacheck:
1558 if (is_remount) {
482a7425 1559 if (test_opt(sb, DATA_FLAGS) != data_opt) {
b31e1552
ES		 1560 ext4_msg(sb, KERN_ERR,
1561 "Cannot change data mode on remount");
ac27a0ec
DK		 1562 return 0;
1563 }
1564 } else {
482a7425 1565 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
ac27a0ec
DK		 1566 sbi->s_mount_opt |= data_opt;
1567 }
1568 break;
5bf5683a
HK		 1569 case Opt_data_err_abort:
1570 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1571 break;
1572 case Opt_data_err_ignore:
1573 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1574 break;
ac27a0ec
DK		 1575#ifdef CONFIG_QUOTA
1576 case Opt_usrjquota:
56c50f11 1577 if (!set_qf_name(sb, USRQUOTA, &args[0]))
ac27a0ec 1578 return 0;
56c50f11
DM		 1579 break;
1580 case Opt_grpjquota:
1581 if (!set_qf_name(sb, GRPQUOTA, &args[0]))
ac27a0ec 1582 return 0;
ac27a0ec
DK		 1583 break;
1584 case Opt_offusrjquota:
56c50f11
DM		 1585 if (!clear_qf_name(sb, USRQUOTA))
1586 return 0;
1587 break;
ac27a0ec 1588 case Opt_offgrpjquota:
56c50f11 1589 if (!clear_qf_name(sb, GRPQUOTA))
ac27a0ec 1590 return 0;
ac27a0ec 1591 break;
56c50f11 1592
ac27a0ec 1593 case Opt_jqfmt_vfsold:
dfc5d03f
JK		 1594 qfmt = QFMT_VFS_OLD;
1595 goto set_qf_format;
ac27a0ec 1596 case Opt_jqfmt_vfsv0:
dfc5d03f 1597 qfmt = QFMT_VFS_V0;
5a20bdfc
JK		 1598 goto set_qf_format;
1599 case Opt_jqfmt_vfsv1:
1600 qfmt = QFMT_VFS_V1;
dfc5d03f 1601set_qf_format:
17bd13b3 1602 if (sb_any_quota_loaded(sb) &&
dfc5d03f 1603 sbi->s_jquota_fmt != qfmt) {
b31e1552 1604 ext4_msg(sb, KERN_ERR, "Cannot change "
dfc5d03f 1605 "journaled quota options when "
b31e1552 1606 "quota turned on");
dfc5d03f
JK		 1607 return 0;
1608 }
1609 sbi->s_jquota_fmt = qfmt;
ac27a0ec
DK		 1610 break;
1611 case Opt_quota:
1612 case Opt_usrquota:
1613 set_opt(sbi->s_mount_opt, QUOTA);
1614 set_opt(sbi->s_mount_opt, USRQUOTA);
1615 break;
1616 case Opt_grpquota:
1617 set_opt(sbi->s_mount_opt, QUOTA);
1618 set_opt(sbi->s_mount_opt, GRPQUOTA);
1619 break;
1620 case Opt_noquota:
17bd13b3 1621 if (sb_any_quota_loaded(sb)) {
b31e1552
ES		 1622 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1623 "options when quota turned on");
ac27a0ec
DK		 1624 return 0;
1625 }
1626 clear_opt(sbi->s_mount_opt, QUOTA);
1627 clear_opt(sbi->s_mount_opt, USRQUOTA);
1628 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1629 break;
1630#else
1631 case Opt_quota:
1632 case Opt_usrquota:
1633 case Opt_grpquota:
b31e1552
ES		 1634 ext4_msg(sb, KERN_ERR,
1635 "quota options not supported");
cd59e7b9 1636 break;
ac27a0ec
DK		 1637 case Opt_usrjquota:
1638 case Opt_grpjquota:
1639 case Opt_offusrjquota:
1640 case Opt_offgrpjquota:
1641 case Opt_jqfmt_vfsold:
1642 case Opt_jqfmt_vfsv0:
5a20bdfc 1643 case Opt_jqfmt_vfsv1:
b31e1552
ES		 1644 ext4_msg(sb, KERN_ERR,
1645 "journaled quota options not supported");
ac27a0ec
DK		 1646 break;
1647 case Opt_noquota:
1648 break;
1649#endif
1650 case Opt_abort:
4ab2f15b 1651 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
ac27a0ec 1652 break;
06705bff
TT		 1653 case Opt_nobarrier:
1654 clear_opt(sbi->s_mount_opt, BARRIER);
1655 break;
ac27a0ec 1656 case Opt_barrier:
15121c18
ES		 1657 if (args[0].from) {
1658 if (match_int(&args[0], &option))
1659 return 0;
1660 } else
1661 option = 1; /* No argument, default to 1 */
ac27a0ec
DK		 1662 if (option)
1663 set_opt(sbi->s_mount_opt, BARRIER);
1664 else
1665 clear_opt(sbi->s_mount_opt, BARRIER);
1666 break;
1667 case Opt_ignore:
1668 break;
1669 case Opt_resize:
1670 if (!is_remount) {
b31e1552
ES		 1671 ext4_msg(sb, KERN_ERR,
1672 "resize option only available "
1673 "for remount");
ac27a0ec
DK		 1674 return 0;
1675 }
1676 if (match_int(&args[0], &option) != 0)
1677 return 0;
1678 *n_blocks_count = option;
1679 break;
1680 case Opt_nobh:
206f7ab4
CH		 1681 ext4_msg(sb, KERN_WARNING,
1682 "Ignoring deprecated nobh option");
ac27a0ec
DK		 1683 break;
1684 case Opt_bh:
206f7ab4
CH		 1685 ext4_msg(sb, KERN_WARNING,
1686 "Ignoring deprecated bh option");
ac27a0ec 1687 break;
25ec56b5
JNC		 1688 case Opt_i_version:
1689 set_opt(sbi->s_mount_opt, I_VERSION);
1690 sb->s_flags |= MS_I_VERSION;
1691 break;
dd919b98
AK		 1692 case Opt_nodelalloc:
1693 clear_opt(sbi->s_mount_opt, DELALLOC);
1694 break;
c9de560d
AT		 1695 case Opt_stripe:
1696 if (match_int(&args[0], &option))
1697 return 0;
1698 if (option < 0)
1699 return 0;
1700 sbi->s_stripe = option;
1701 break;
64769240
AT		 1702 case Opt_delalloc:
1703 set_opt(sbi->s_mount_opt, DELALLOC);
1704 break;
6fd058f7
TT		 1705 case Opt_block_validity:
1706 set_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1707 break;
1708 case Opt_noblock_validity:
1709 clear_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1710 break;
240799cd
TT		 1711 case Opt_inode_readahead_blks:
1712 if (match_int(&args[0], &option))
1713 return 0;
1714 if (option < 0 || option > (1 << 30))
1715 return 0;
f7c43950 1716 if (!is_power_of_2(option)) {
b31e1552
ES		 1717 ext4_msg(sb, KERN_ERR,
1718 "EXT4-fs: inode_readahead_blks"
1719 " must be a power of 2");
3197ebdb
TT		 1720 return 0;
1721 }
240799cd
TT		 1722 sbi->s_inode_readahead_blks = option;
1723 break;
b3881f74
TT		 1724 case Opt_journal_ioprio:
1725 if (match_int(&args[0], &option))
1726 return 0;
1727 if (option < 0 || option > 7)
1728 break;
1729 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1730 option);
1731 break;
06705bff
TT		 1732 case Opt_noauto_da_alloc:
1733 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1734 break;
afd4672d 1735 case Opt_auto_da_alloc:
15121c18
ES		 1736 if (args[0].from) {
1737 if (match_int(&args[0], &option))
1738 return 0;
1739 } else
1740 option = 1; /* No argument, default to 1 */
afd4672d
TT		 1741 if (option)
1742 clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1743 else
1744 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1745 break;
5328e635
ES		 1746 case Opt_discard:
1747 set_opt(sbi->s_mount_opt, DISCARD);
1748 break;
1749 case Opt_nodiscard:
1750 clear_opt(sbi->s_mount_opt, DISCARD);
1751 break;
744692dc
JZ		 1752 case Opt_dioread_nolock:
1753 set_opt(sbi->s_mount_opt, DIOREAD_NOLOCK);
1754 break;
1755 case Opt_dioread_lock:
1756 clear_opt(sbi->s_mount_opt, DIOREAD_NOLOCK);
1757 break;
ac27a0ec 1758 default:
b31e1552
ES		 1759 ext4_msg(sb, KERN_ERR,
1760 "Unrecognized mount option \"%s\" "
1761 "or missing value", p);
ac27a0ec
DK		 1762 return 0;
1763 }
1764 }
1765#ifdef CONFIG_QUOTA
1766 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
482a7425 1767 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
ac27a0ec
DK		 1768 clear_opt(sbi->s_mount_opt, USRQUOTA);
1769
482a7425 1770 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
ac27a0ec
DK		 1771 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1772
56c50f11 1773 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
b31e1552
ES		 1774 ext4_msg(sb, KERN_ERR, "old and new quota "
1775 "format mixing");
ac27a0ec
DK		 1776 return 0;
1777 }
1778
1779 if (!sbi->s_jquota_fmt) {
b31e1552
ES		 1780 ext4_msg(sb, KERN_ERR, "journaled quota format "
1781 "not specified");
ac27a0ec
DK		 1782 return 0;
1783 }
1784 } else {
1785 if (sbi->s_jquota_fmt) {
b31e1552 1786 ext4_msg(sb, KERN_ERR, "journaled quota format "
2c8be6b2 1787 "specified with no journaling "
b31e1552 1788 "enabled");
ac27a0ec
DK		 1789 return 0;
1790 }
1791 }
1792#endif
1793 return 1;
1794}
1795
617ba13b 1796static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 1797 int read_only)
1798{
617ba13b 1799 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 1800 int res = 0;
1801
617ba13b 1802 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
b31e1552
ES		 1803 ext4_msg(sb, KERN_ERR, "revision level too high, "
1804 "forcing read-only mode");
ac27a0ec
DK		 1805 res = MS_RDONLY;
1806 }
1807 if (read_only)
1808 return res;
617ba13b 1809 if (!(sbi->s_mount_state & EXT4_VALID_FS))
b31e1552
ES		 1810 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1811 "running e2fsck is recommended");
617ba13b 1812 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
b31e1552
ES		 1813 ext4_msg(sb, KERN_WARNING,
1814 "warning: mounting fs with errors, "
1815 "running e2fsck is recommended");
ac27a0ec
DK		 1816 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1817 le16_to_cpu(es->s_mnt_count) >=
1818 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
b31e1552
ES		 1819 ext4_msg(sb, KERN_WARNING,
1820 "warning: maximal mount count reached, "
1821 "running e2fsck is recommended");
ac27a0ec
DK		 1822 else if (le32_to_cpu(es->s_checkinterval) &&
1823 (le32_to_cpu(es->s_lastcheck) +
1824 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
b31e1552
ES		 1825 ext4_msg(sb, KERN_WARNING,
1826 "warning: checktime reached, "
1827 "running e2fsck is recommended");
0b8e58a1 1828 if (!sbi->s_journal)
0390131b 1829 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec 1830 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 1831 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 1832 le16_add_cpu(&es->s_mnt_count, 1);
ac27a0ec 1833 es->s_mtime = cpu_to_le32(get_seconds());
617ba13b 1834 ext4_update_dynamic_rev(sb);
0390131b
FM		 1835 if (sbi->s_journal)
1836 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 1837
e2d67052 1838 ext4_commit_super(sb, 1);
ac27a0ec 1839 if (test_opt(sb, DEBUG))
a9df9a49 1840 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
7f4520cc 1841 "bpg=%lu, ipg=%lu, mo=%04x]\n",
ac27a0ec
DK		 1842 sb->s_blocksize,
1843 sbi->s_groups_count,
617ba13b
MC		 1844 EXT4_BLOCKS_PER_GROUP(sb),
1845 EXT4_INODES_PER_GROUP(sb),
ac27a0ec
DK		 1846 sbi->s_mount_opt);
1847
ac27a0ec
DK		 1848 return res;
1849}
1850
772cb7c8
JS		 1851static int ext4_fill_flex_info(struct super_block *sb)
1852{
1853 struct ext4_sb_info *sbi = EXT4_SB(sb);
1854 struct ext4_group_desc *gdp = NULL;
772cb7c8
JS		 1855 ext4_group_t flex_group_count;
1856 ext4_group_t flex_group;
1857 int groups_per_flex = 0;
c5ca7c76 1858 size_t size;
772cb7c8
JS		 1859 int i;
1860
503358ae
TT		 1861 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1862 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1863
1864 if (groups_per_flex < 2) {
772cb7c8
JS		 1865 sbi->s_log_groups_per_flex = 0;
1866 return 1;
1867 }
1868
c62a11fd
FB		 1869 /* We allocate both existing and potentially added groups */
1870 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
d94e99a6
AK		 1871 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1872 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
c5ca7c76
TT		 1873 size = flex_group_count * sizeof(struct flex_groups);
1874 sbi->s_flex_groups = kzalloc(size, GFP_KERNEL);
1875 if (sbi->s_flex_groups == NULL) {
1876 sbi->s_flex_groups = vmalloc(size);
1877 if (sbi->s_flex_groups)
1878 memset(sbi->s_flex_groups, 0, size);
1879 }
772cb7c8 1880 if (sbi->s_flex_groups == NULL) {
b31e1552
ES		 1881 ext4_msg(sb, KERN_ERR, "not enough memory for "
1882 "%u flex groups", flex_group_count);
772cb7c8
JS		 1883 goto failed;
1884 }
772cb7c8 1885
772cb7c8 1886 for (i = 0; i < sbi->s_groups_count; i++) {
88b6edd1 1887 gdp = ext4_get_group_desc(sb, i, NULL);
772cb7c8
JS		 1888
1889 flex_group = ext4_flex_group(sbi, i);
7ad9bb65
TT		 1890 atomic_add(ext4_free_inodes_count(sb, gdp),
1891 &sbi->s_flex_groups[flex_group].free_inodes);
1892 atomic_add(ext4_free_blks_count(sb, gdp),
1893 &sbi->s_flex_groups[flex_group].free_blocks);
1894 atomic_add(ext4_used_dirs_count(sb, gdp),
1895 &sbi->s_flex_groups[flex_group].used_dirs);
772cb7c8
JS		 1896 }
1897
1898 return 1;
1899failed:
1900 return 0;
1901}
1902
717d50e4
AD		 1903__le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1904 struct ext4_group_desc *gdp)
1905{
1906 __u16 crc = 0;
1907
1908 if (sbi->s_es->s_feature_ro_compat &
1909 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1910 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1911 __le32 le_group = cpu_to_le32(block_group);
1912
1913 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1914 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1915 crc = crc16(crc, (__u8 *)gdp, offset);
1916 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1917 /* for checksum of struct ext4_group_desc do the rest...*/
1918 if ((sbi->s_es->s_feature_incompat &
1919 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1920 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1921 crc = crc16(crc, (__u8 *)gdp + offset,
1922 le16_to_cpu(sbi->s_es->s_desc_size) -
1923 offset);
1924 }
1925
1926 return cpu_to_le16(crc);
1927}
1928
1929int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1930 struct ext4_group_desc *gdp)
1931{
1932 if ((sbi->s_es->s_feature_ro_compat &
1933 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1934 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1935 return 0;
1936
1937 return 1;
1938}
1939
ac27a0ec 1940/* Called at mount-time, super-block is locked */
197cd65a 1941static int ext4_check_descriptors(struct super_block *sb)
ac27a0ec 1942{
617ba13b
MC		 1943 struct ext4_sb_info *sbi = EXT4_SB(sb);
1944 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1945 ext4_fsblk_t last_block;
bd81d8ee
LV		 1946 ext4_fsblk_t block_bitmap;
1947 ext4_fsblk_t inode_bitmap;
1948 ext4_fsblk_t inode_table;
ce421581 1949 int flexbg_flag = 0;
fd2d4291 1950 ext4_group_t i;
ac27a0ec 1951
ce421581
JS		 1952 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1953 flexbg_flag = 1;
1954
af5bc92d 1955 ext4_debug("Checking group descriptors");
ac27a0ec 1956
197cd65a
AM		 1957 for (i = 0; i < sbi->s_groups_count; i++) {
1958 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1959
ce421581 1960 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 1961 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 1962 else
1963 last_block = first_block +
617ba13b 1964 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 1965
8fadc143 1966 block_bitmap = ext4_block_bitmap(sb, gdp);
2b2d6d01 1967 if (block_bitmap < first_block || block_bitmap > last_block) {
b31e1552 1968 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 1969 "Block bitmap for group %u not in group "
b31e1552 1970 "(block %llu)!", i, block_bitmap);
ac27a0ec
DK		 1971 return 0;
1972 }
8fadc143 1973 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2b2d6d01 1974 if (inode_bitmap < first_block || inode_bitmap > last_block) {
b31e1552 1975 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 1976 "Inode bitmap for group %u not in group "
b31e1552 1977 "(block %llu)!", i, inode_bitmap);
ac27a0ec
DK		 1978 return 0;
1979 }
8fadc143 1980 inode_table = ext4_inode_table(sb, gdp);
bd81d8ee 1981 if (inode_table < first_block ||
2b2d6d01 1982 inode_table + sbi->s_itb_per_group - 1 > last_block) {
b31e1552 1983 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 1984 "Inode table for group %u not in group "
b31e1552 1985 "(block %llu)!", i, inode_table);
ac27a0ec
DK		 1986 return 0;
1987 }
955ce5f5 1988 ext4_lock_group(sb, i);
717d50e4 1989 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
b31e1552
ES		 1990 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1991 "Checksum for group %u failed (%u!=%u)",
1992 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1993 gdp)), le16_to_cpu(gdp->bg_checksum));
7ee1ec4c 1994 if (!(sb->s_flags & MS_RDONLY)) {
955ce5f5 1995 ext4_unlock_group(sb, i);
8a266467 1996 return 0;
7ee1ec4c 1997 }
717d50e4 1998 }
955ce5f5 1999 ext4_unlock_group(sb, i);
ce421581
JS		 2000 if (!flexbg_flag)
2001 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec
DK		 2002 }
2003
bd81d8ee 2004 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
0b8e58a1 2005 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
ac27a0ec
DK		 2006 return 1;
2007}
2008
617ba13b 2009/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 2010 * the superblock) which were deleted from all directories, but held open by
2011 * a process at the time of a crash. We walk the list and try to delete these
2012 * inodes at recovery time (only with a read-write filesystem).
2013 *
2014 * In order to keep the orphan inode chain consistent during traversal (in
2015 * case of crash during recovery), we link each inode into the superblock
2016 * orphan list_head and handle it the same way as an inode deletion during
2017 * normal operation (which journals the operations for us).
2018 *
2019 * We only do an iget() and an iput() on each inode, which is very safe if we
2020 * accidentally point at an in-use or already deleted inode. The worst that
2021 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 2022 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 2023 * e2fsck was run on this filesystem, and it must have already done the orphan
2024 * inode cleanup for us, so we can safely abort without any further action.
2025 */
2b2d6d01
TT		 2026static void ext4_orphan_cleanup(struct super_block *sb,
2027 struct ext4_super_block *es)
ac27a0ec
DK		 2028{
2029 unsigned int s_flags = sb->s_flags;
2030 int nr_orphans = 0, nr_truncates = 0;
2031#ifdef CONFIG_QUOTA
2032 int i;
2033#endif
2034 if (!es->s_last_orphan) {
2035 jbd_debug(4, "no orphan inodes to clean up\n");
2036 return;
2037 }
2038
a8f48a95 2039 if (bdev_read_only(sb->s_bdev)) {
b31e1552
ES		 2040 ext4_msg(sb, KERN_ERR, "write access "
2041 "unavailable, skipping orphan cleanup");
a8f48a95
ES		 2042 return;
2043 }
2044
617ba13b 2045 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
ac27a0ec
DK		 2046 if (es->s_last_orphan)
2047 jbd_debug(1, "Errors on filesystem, "
2048 "clearing orphan list.\n");
2049 es->s_last_orphan = 0;
2050 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2051 return;
2052 }
2053
2054 if (s_flags & MS_RDONLY) {
b31e1552 2055 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
ac27a0ec
DK		 2056 sb->s_flags &= ~MS_RDONLY;
2057 }
2058#ifdef CONFIG_QUOTA
2059 /* Needed for iput() to work correctly and not trash data */
2060 sb->s_flags |= MS_ACTIVE;
2061 /* Turn on quotas so that they are updated correctly */
2062 for (i = 0; i < MAXQUOTAS; i++) {
617ba13b
MC		 2063 if (EXT4_SB(sb)->s_qf_names[i]) {
2064 int ret = ext4_quota_on_mount(sb, i);
ac27a0ec 2065 if (ret < 0)
b31e1552
ES		 2066 ext4_msg(sb, KERN_ERR,
2067 "Cannot turn on journaled "
2068 "quota: error %d", ret);
ac27a0ec
DK		 2069 }
2070 }
2071#endif
2072
2073 while (es->s_last_orphan) {
2074 struct inode *inode;
2075
97bd42b9
JB		 2076 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2077 if (IS_ERR(inode)) {
ac27a0ec
DK		 2078 es->s_last_orphan = 0;
2079 break;
2080 }
2081
617ba13b 2082 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
871a2931 2083 dquot_initialize(inode);
ac27a0ec 2084 if (inode->i_nlink) {
b31e1552
ES		 2085 ext4_msg(sb, KERN_DEBUG,
2086 "%s: truncating inode %lu to %lld bytes",
46e665e9 2087 __func__, inode->i_ino, inode->i_size);
e5f8eab8 2088 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
ac27a0ec 2089 inode->i_ino, inode->i_size);
617ba13b 2090 ext4_truncate(inode);
ac27a0ec
DK		 2091 nr_truncates++;
2092 } else {
b31e1552
ES		 2093 ext4_msg(sb, KERN_DEBUG,
2094 "%s: deleting unreferenced inode %lu",
46e665e9 2095 __func__, inode->i_ino);
ac27a0ec
DK		 2096 jbd_debug(2, "deleting unreferenced inode %lu\n",
2097 inode->i_ino);
2098 nr_orphans++;
2099 }
2100 iput(inode); /* The delete magic happens here! */
2101 }
2102
2b2d6d01 2103#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
ac27a0ec
DK		 2104
2105 if (nr_orphans)
b31e1552
ES		 2106 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2107 PLURAL(nr_orphans));
ac27a0ec 2108 if (nr_truncates)
b31e1552
ES		 2109 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2110 PLURAL(nr_truncates));
ac27a0ec
DK		 2111#ifdef CONFIG_QUOTA
2112 /* Turn quotas off */
2113 for (i = 0; i < MAXQUOTAS; i++) {
2114 if (sb_dqopt(sb)->files[i])
287a8095 2115 dquot_quota_off(sb, i);
ac27a0ec
DK		 2116 }
2117#endif
2118 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2119}
0b8e58a1 2120
cd2291a4
ES		 2121/*
2122 * Maximal extent format file size.
2123 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2124 * extent format containers, within a sector_t, and within i_blocks
2125 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2126 * so that won't be a limiting factor.
2127 *
2128 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2129 */
f287a1a5 2130static loff_t ext4_max_size(int blkbits, int has_huge_files)
cd2291a4
ES		 2131{
2132 loff_t res;
2133 loff_t upper_limit = MAX_LFS_FILESIZE;
2134
2135 /* small i_blocks in vfs inode? */
f287a1a5 2136 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
cd2291a4 2137 /*
90c699a9 2138 * CONFIG_LBDAF is not enabled implies the inode
cd2291a4
ES		 2139 * i_block represent total blocks in 512 bytes
2140 * 32 == size of vfs inode i_blocks * 8
2141 */
2142 upper_limit = (1LL << 32) - 1;
2143
2144 /* total blocks in file system block size */
2145 upper_limit >>= (blkbits - 9);
2146 upper_limit <<= blkbits;
2147 }
2148
2149 /* 32-bit extent-start container, ee_block */
2150 res = 1LL << 32;
2151 res <<= blkbits;
2152 res -= 1;
2153
2154 /* Sanity check against vm- & vfs- imposed limits */
2155 if (res > upper_limit)
2156 res = upper_limit;
2157
2158 return res;
2159}
ac27a0ec 2160
ac27a0ec 2161/*
cd2291a4 2162 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 2163 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2164 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 2165 */
f287a1a5 2166static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
ac27a0ec 2167{
617ba13b 2168 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 2169 int meta_blocks;
2170 loff_t upper_limit;
0b8e58a1
AD		 2171 /* This is calculated to be the largest file size for a dense, block
2172 * mapped file such that the file's total number of 512-byte sectors,
2173 * including data and all indirect blocks, does not exceed (2^48 - 1).
2174 *
2175 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2176 * number of 512-byte sectors of the file.
0fc1b451
AK		 2177 */
2178
f287a1a5 2179 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
0fc1b451 2180 /*
90c699a9 2181 * !has_huge_files or CONFIG_LBDAF not enabled implies that
0b8e58a1
AD		 2182 * the inode i_block field represents total file blocks in
2183 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
0fc1b451
AK		 2184 */
2185 upper_limit = (1LL << 32) - 1;
2186
2187 /* total blocks in file system block size */
2188 upper_limit >>= (bits - 9);
2189
2190 } else {
8180a562
AK		 2191 /*
2192 * We use 48 bit ext4_inode i_blocks
2193 * With EXT4_HUGE_FILE_FL set the i_blocks
2194 * represent total number of blocks in
2195 * file system block size
2196 */
0fc1b451
AK		 2197 upper_limit = (1LL << 48) - 1;
2198
0fc1b451
AK		 2199 }
2200
2201 /* indirect blocks */
2202 meta_blocks = 1;
2203 /* double indirect blocks */
2204 meta_blocks += 1 + (1LL << (bits-2));
2205 /* tripple indirect blocks */
2206 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2207
2208 upper_limit -= meta_blocks;
2209 upper_limit <<= bits;
ac27a0ec
DK		 2210
2211 res += 1LL << (bits-2);
2212 res += 1LL << (2*(bits-2));
2213 res += 1LL << (3*(bits-2));
2214 res <<= bits;
2215 if (res > upper_limit)
2216 res = upper_limit;
0fc1b451
AK		 2217
2218 if (res > MAX_LFS_FILESIZE)
2219 res = MAX_LFS_FILESIZE;
2220
ac27a0ec
DK		 2221 return res;
2222}
2223
617ba13b 2224static ext4_fsblk_t descriptor_loc(struct super_block *sb,
0b8e58a1 2225 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 2226{
617ba13b 2227 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 2228 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 2229 int has_super = 0;
2230
2231 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2232
617ba13b 2233 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
ac27a0ec 2234 nr < first_meta_bg)
70bbb3e0 2235 return logical_sb_block + nr + 1;
ac27a0ec 2236 bg = sbi->s_desc_per_block * nr;
617ba13b 2237 if (ext4_bg_has_super(sb, bg))
ac27a0ec 2238 has_super = 1;
0b8e58a1 2239
617ba13b 2240 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 2241}
2242
c9de560d
AT		 2243/**
2244 * ext4_get_stripe_size: Get the stripe size.
2245 * @sbi: In memory super block info
2246 *
2247 * If we have specified it via mount option, then
2248 * use the mount option value. If the value specified at mount time is
2249 * greater than the blocks per group use the super block value.
2250 * If the super block value is greater than blocks per group return 0.
2251 * Allocator needs it be less than blocks per group.
2252 *
2253 */
2254static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2255{
2256 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2257 unsigned long stripe_width =
2258 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2259
2260 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2261 return sbi->s_stripe;
2262
2263 if (stripe_width <= sbi->s_blocks_per_group)
2264 return stripe_width;
2265
2266 if (stride <= sbi->s_blocks_per_group)
2267 return stride;
2268
2269 return 0;
2270}
ac27a0ec 2271
3197ebdb
TT		 2272/* sysfs supprt */
2273
2274struct ext4_attr {
2275 struct attribute attr;
2276 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
60e6679e 2277 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
3197ebdb
TT		 2278 const char *, size_t);
2279 int offset;
2280};
2281
2282static int parse_strtoul(const char *buf,
2283 unsigned long max, unsigned long *value)
2284{
2285 char *endp;
2286
e7d2860b
AGR		 2287 *value = simple_strtoul(skip_spaces(buf), &endp, 0);
2288 endp = skip_spaces(endp);
3197ebdb
TT		 2289 if (*endp || *value > max)
2290 return -EINVAL;
2291
2292 return 0;
2293}
2294
2295static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2296 struct ext4_sb_info *sbi,
2297 char *buf)
2298{
2299 return snprintf(buf, PAGE_SIZE, "%llu\n",
2300 (s64) percpu_counter_sum(&sbi->s_dirtyblocks_counter));
2301}
2302
2303static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2304 struct ext4_sb_info *sbi, char *buf)
2305{
2306 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2307
f613dfcb
TT		 2308 if (!sb->s_bdev->bd_part)
2309 return snprintf(buf, PAGE_SIZE, "0\n");
3197ebdb
TT		 2310 return snprintf(buf, PAGE_SIZE, "%lu\n",
2311 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2312 sbi->s_sectors_written_start) >> 1);
2313}
2314
2315static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2316 struct ext4_sb_info *sbi, char *buf)
2317{
2318 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2319
f613dfcb
TT		 2320 if (!sb->s_bdev->bd_part)
2321 return snprintf(buf, PAGE_SIZE, "0\n");
3197ebdb 2322 return snprintf(buf, PAGE_SIZE, "%llu\n",
a6b43e38 2323 (unsigned long long)(sbi->s_kbytes_written +
3197ebdb 2324 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
a6b43e38 2325 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
3197ebdb
TT		 2326}
2327
2328static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2329 struct ext4_sb_info *sbi,
2330 const char *buf, size_t count)
2331{
2332 unsigned long t;
2333
2334 if (parse_strtoul(buf, 0x40000000, &t))
2335 return -EINVAL;
2336
f7c43950 2337 if (!is_power_of_2(t))
3197ebdb
TT		 2338 return -EINVAL;
2339
2340 sbi->s_inode_readahead_blks = t;
2341 return count;
2342}
2343
2344static ssize_t sbi_ui_show(struct ext4_attr *a,
0b8e58a1 2345 struct ext4_sb_info *sbi, char *buf)
3197ebdb
TT		 2346{
2347 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2348
2349 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2350}
2351
2352static ssize_t sbi_ui_store(struct ext4_attr *a,
2353 struct ext4_sb_info *sbi,
2354 const char *buf, size_t count)
2355{
2356 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2357 unsigned long t;
2358
2359 if (parse_strtoul(buf, 0xffffffff, &t))
2360 return -EINVAL;
2361 *ui = t;
2362 return count;
2363}
2364
2365#define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2366static struct ext4_attr ext4_attr_##_name = { \
2367 .attr = {.name = __stringify(_name), .mode = _mode }, \
2368 .show = _show, \
2369 .store = _store, \
2370 .offset = offsetof(struct ext4_sb_info, _elname), \
2371}
2372#define EXT4_ATTR(name, mode, show, store) \
2373static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2374
2375#define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2376#define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2377#define EXT4_RW_ATTR_SBI_UI(name, elname) \
2378 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2379#define ATTR_LIST(name) &ext4_attr_##name.attr
2380
2381EXT4_RO_ATTR(delayed_allocation_blocks);
2382EXT4_RO_ATTR(session_write_kbytes);
2383EXT4_RO_ATTR(lifetime_write_kbytes);
2384EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2385 inode_readahead_blks_store, s_inode_readahead_blks);
11013911 2386EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
3197ebdb
TT		 2387EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2388EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2389EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2390EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2391EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2392EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
55138e0b 2393EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
3197ebdb
TT		 2394
2395static struct attribute *ext4_attrs[] = {
2396 ATTR_LIST(delayed_allocation_blocks),
2397 ATTR_LIST(session_write_kbytes),
2398 ATTR_LIST(lifetime_write_kbytes),
2399 ATTR_LIST(inode_readahead_blks),
11013911 2400 ATTR_LIST(inode_goal),
3197ebdb
TT		 2401 ATTR_LIST(mb_stats),
2402 ATTR_LIST(mb_max_to_scan),
2403 ATTR_LIST(mb_min_to_scan),
2404 ATTR_LIST(mb_order2_req),
2405 ATTR_LIST(mb_stream_req),
2406 ATTR_LIST(mb_group_prealloc),
55138e0b 2407 ATTR_LIST(max_writeback_mb_bump),
3197ebdb
TT		 2408 NULL,
2409};
2410
2411static ssize_t ext4_attr_show(struct kobject *kobj,
2412 struct attribute *attr, char *buf)
2413{
2414 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2415 s_kobj);
2416 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2417
2418 return a->show ? a->show(a, sbi, buf) : 0;
2419}
2420
2421static ssize_t ext4_attr_store(struct kobject *kobj,
2422 struct attribute *attr,
2423 const char *buf, size_t len)
2424{
2425 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2426 s_kobj);
2427 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2428
2429 return a->store ? a->store(a, sbi, buf, len) : 0;
2430}
2431
2432static void ext4_sb_release(struct kobject *kobj)
2433{
2434 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2435 s_kobj);
2436 complete(&sbi->s_kobj_unregister);
2437}
2438
2439
52cf25d0 2440static const struct sysfs_ops ext4_attr_ops = {
3197ebdb
TT		 2441 .show = ext4_attr_show,
2442 .store = ext4_attr_store,
2443};
2444
2445static struct kobj_type ext4_ktype = {
2446 .default_attrs = ext4_attrs,
2447 .sysfs_ops = &ext4_attr_ops,
2448 .release = ext4_sb_release,
2449};
2450
a13fb1a4
ES		 2451/*
2452 * Check whether this filesystem can be mounted based on
2453 * the features present and the RDONLY/RDWR mount requested.
2454 * Returns 1 if this filesystem can be mounted as requested,
2455 * 0 if it cannot be.
2456 */
2457static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2458{
2459 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2460 ext4_msg(sb, KERN_ERR,
2461 "Couldn't mount because of "
2462 "unsupported optional features (%x)",
2463 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2464 ~EXT4_FEATURE_INCOMPAT_SUPP));
2465 return 0;
2466 }
2467
2468 if (readonly)
2469 return 1;
2470
2471 /* Check that feature set is OK for a read-write mount */
2472 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2473 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2474 "unsupported optional features (%x)",
2475 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2476 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2477 return 0;
2478 }
2479 /*
2480 * Large file size enabled file system can only be mounted
2481 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2482 */
2483 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2484 if (sizeof(blkcnt_t) < sizeof(u64)) {
2485 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2486 "cannot be mounted RDWR without "
2487 "CONFIG_LBDAF");
2488 return 0;
2489 }
2490 }
2491 return 1;
2492}
2493
66e61a9e
TT		 2494/*
2495 * This function is called once a day if we have errors logged
2496 * on the file system
2497 */
2498static void print_daily_error_info(unsigned long arg)
2499{
2500 struct super_block *sb = (struct super_block *) arg;
2501 struct ext4_sb_info *sbi;
2502 struct ext4_super_block *es;
2503
2504 sbi = EXT4_SB(sb);
2505 es = sbi->s_es;
2506
2507 if (es->s_error_count)
2508 ext4_msg(sb, KERN_NOTICE, "error count: %u",
2509 le32_to_cpu(es->s_error_count));
2510 if (es->s_first_error_time) {
2511 printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d",
2512 sb->s_id, le32_to_cpu(es->s_first_error_time),
2513 (int) sizeof(es->s_first_error_func),
2514 es->s_first_error_func,
2515 le32_to_cpu(es->s_first_error_line));
2516 if (es->s_first_error_ino)
2517 printk(": inode %u",
2518 le32_to_cpu(es->s_first_error_ino));
2519 if (es->s_first_error_block)
2520 printk(": block %llu", (unsigned long long)
2521 le64_to_cpu(es->s_first_error_block));
2522 printk("\n");
2523 }
2524 if (es->s_last_error_time) {
2525 printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d",
2526 sb->s_id, le32_to_cpu(es->s_last_error_time),
2527 (int) sizeof(es->s_last_error_func),
2528 es->s_last_error_func,
2529 le32_to_cpu(es->s_last_error_line));
2530 if (es->s_last_error_ino)
2531 printk(": inode %u",
2532 le32_to_cpu(es->s_last_error_ino));
2533 if (es->s_last_error_block)
2534 printk(": block %llu", (unsigned long long)
2535 le64_to_cpu(es->s_last_error_block));
2536 printk("\n");
2537 }
2538 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2539}
2540
2b2d6d01 2541static int ext4_fill_super(struct super_block *sb, void *data, int silent)
7477827f
AK		 2542 __releases(kernel_lock)
2543 __acquires(kernel_lock)
ac27a0ec 2544{
d4c402d9 2545 char *orig_data = kstrdup(data, GFP_KERNEL);
2b2d6d01 2546 struct buffer_head *bh;
617ba13b
MC		 2547 struct ext4_super_block *es = NULL;
2548 struct ext4_sb_info *sbi;
2549 ext4_fsblk_t block;
2550 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 2551 ext4_fsblk_t logical_sb_block;
ac27a0ec 2552 unsigned long offset = 0;
ac27a0ec
DK		 2553 unsigned long journal_devnum = 0;
2554 unsigned long def_mount_opts;
2555 struct inode *root;
9f6200bb 2556 char *cp;
0390131b 2557 const char *descr;
dcc7dae3 2558 int ret = -ENOMEM;
ac27a0ec 2559 int blocksize;
4ec11028
TT		 2560 unsigned int db_count;
2561 unsigned int i;
f287a1a5 2562 int needs_recovery, has_huge_files;
bd81d8ee 2563 __u64 blocks_count;
833f4077 2564 int err;
b3881f74 2565 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
ac27a0ec
DK		 2566
2567 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
2568 if (!sbi)
dcc7dae3 2569 goto out_free_orig;
705895b6
PE		 2570
2571 sbi->s_blockgroup_lock =
2572 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
2573 if (!sbi->s_blockgroup_lock) {
2574 kfree(sbi);
dcc7dae3 2575 goto out_free_orig;
705895b6 2576 }
ac27a0ec
DK		 2577 sb->s_fs_info = sbi;
2578 sbi->s_mount_opt = 0;
617ba13b
MC		 2579 sbi->s_resuid = EXT4_DEF_RESUID;
2580 sbi->s_resgid = EXT4_DEF_RESGID;
240799cd 2581 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
d9c9bef1 2582 sbi->s_sb_block = sb_block;
f613dfcb
TT		 2583 if (sb->s_bdev->bd_part)
2584 sbi->s_sectors_written_start =
2585 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
ac27a0ec
DK		 2586
2587 unlock_kernel();
2588
9f6200bb
TT		 2589 /* Cleanup superblock name */
2590 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
2591 *cp = '!';
2592
dcc7dae3 2593 ret = -EINVAL;
617ba13b 2594 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 2595 if (!blocksize) {
b31e1552 2596 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
ac27a0ec
DK		 2597 goto out_fail;
2598 }
2599
2600 /*
617ba13b 2601 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 2602 * block sizes. We need to calculate the offset from buffer start.
2603 */
617ba13b 2604 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 2605 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2606 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 2607 } else {
70bbb3e0 2608 logical_sb_block = sb_block;
ac27a0ec
DK		 2609 }
2610
70bbb3e0 2611 if (!(bh = sb_bread(sb, logical_sb_block))) {
b31e1552 2612 ext4_msg(sb, KERN_ERR, "unable to read superblock");
ac27a0ec
DK		 2613 goto out_fail;
2614 }
2615 /*
2616 * Note: s_es must be initialized as soon as possible because
617ba13b 2617 * some ext4 macro-instructions depend on its value
ac27a0ec 2618 */
617ba13b 2619 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
ac27a0ec
DK		 2620 sbi->s_es = es;
2621 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 2622 if (sb->s_magic != EXT4_SUPER_MAGIC)
2623 goto cantfind_ext4;
afc32f7e 2624 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
ac27a0ec
DK		 2625
2626 /* Set defaults before we parse the mount options */
2627 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
617ba13b 2628 if (def_mount_opts & EXT4_DEFM_DEBUG)
ac27a0ec 2629 set_opt(sbi->s_mount_opt, DEBUG);
437ca0fd
DM		 2630 if (def_mount_opts & EXT4_DEFM_BSDGROUPS) {
2631 ext4_msg(sb, KERN_WARNING, deprecated_msg, "bsdgroups",
2632 "2.6.38");
ac27a0ec 2633 set_opt(sbi->s_mount_opt, GRPID);
437ca0fd 2634 }
617ba13b 2635 if (def_mount_opts & EXT4_DEFM_UID16)
ac27a0ec 2636 set_opt(sbi->s_mount_opt, NO_UID32);
03010a33 2637#ifdef CONFIG_EXT4_FS_XATTR
617ba13b 2638 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
ac27a0ec 2639 set_opt(sbi->s_mount_opt, XATTR_USER);
2e7842b8 2640#endif
03010a33 2641#ifdef CONFIG_EXT4_FS_POSIX_ACL
617ba13b 2642 if (def_mount_opts & EXT4_DEFM_ACL)
ac27a0ec 2643 set_opt(sbi->s_mount_opt, POSIX_ACL);
2e7842b8 2644#endif
617ba13b 2645 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
482a7425 2646 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
617ba13b 2647 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
482a7425 2648 set_opt(sbi->s_mount_opt, ORDERED_DATA);
617ba13b 2649 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
482a7425 2650 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
617ba13b
MC		 2651
2652 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
ac27a0ec 2653 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
bb4f397a 2654 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
ceea16bf 2655 set_opt(sbi->s_mount_opt, ERRORS_CONT);
bb4f397a
AK		 2656 else
2657 set_opt(sbi->s_mount_opt, ERRORS_RO);
ac27a0ec
DK		 2658
2659 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
2660 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
30773840
TT		 2661 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
2662 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
2663 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
ac27a0ec 2664
571640ca 2665 set_opt(sbi->s_mount_opt, BARRIER);
ac27a0ec 2666
dd919b98
AK		 2667 /*
2668 * enable delayed allocation by default
2669 * Use -o nodelalloc to turn it off
2670 */
ba69f9ab
JK		 2671 if (!IS_EXT3_SB(sb))
2672 set_opt(sbi->s_mount_opt, DELALLOC);
dd919b98 2673
b3881f74
TT		 2674 if (!parse_options((char *) data, sb, &journal_devnum,
2675 &journal_ioprio, NULL, 0))
ac27a0ec
DK		 2676 goto failed_mount;
2677
2678 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 2679 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec 2680
617ba13b
MC		 2681 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
2682 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
2683 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
2684 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
b31e1552
ES		 2685 ext4_msg(sb, KERN_WARNING,
2686 "feature flags set on rev 0 fs, "
2687 "running e2fsck is recommended");
469108ff 2688
ac27a0ec
DK		 2689 /*
2690 * Check feature flags regardless of the revision level, since we
2691 * previously didn't change the revision level when setting the flags,
2692 * so there is a chance incompat flags are set on a rev 0 filesystem.
2693 */
a13fb1a4 2694 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
ac27a0ec 2695 goto failed_mount;
a13fb1a4 2696
ac27a0ec
DK		 2697 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
2698
617ba13b
MC		 2699 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
2700 blocksize > EXT4_MAX_BLOCK_SIZE) {
b31e1552
ES		 2701 ext4_msg(sb, KERN_ERR,
2702 "Unsupported filesystem blocksize %d", blocksize);
ac27a0ec
DK		 2703 goto failed_mount;
2704 }
2705
ac27a0ec 2706 if (sb->s_blocksize != blocksize) {
ce40733c
AK		 2707 /* Validate the filesystem blocksize */
2708 if (!sb_set_blocksize(sb, blocksize)) {
b31e1552 2709 ext4_msg(sb, KERN_ERR, "bad block size %d",
ce40733c 2710 blocksize);
ac27a0ec
DK		 2711 goto failed_mount;
2712 }
2713
2b2d6d01 2714 brelse(bh);
70bbb3e0
AM		 2715 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2716 offset = do_div(logical_sb_block, blocksize);
2717 bh = sb_bread(sb, logical_sb_block);
ac27a0ec 2718 if (!bh) {
b31e1552
ES		 2719 ext4_msg(sb, KERN_ERR,
2720 "Can't read superblock on 2nd try");
ac27a0ec
DK		 2721 goto failed_mount;
2722 }
617ba13b 2723 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
ac27a0ec 2724 sbi->s_es = es;
617ba13b 2725 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
b31e1552
ES		 2726 ext4_msg(sb, KERN_ERR,
2727 "Magic mismatch, very weird!");
ac27a0ec
DK		 2728 goto failed_mount;
2729 }
2730 }
2731
a13fb1a4
ES		 2732 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2733 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
f287a1a5
TT		 2734 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
2735 has_huge_files);
2736 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
ac27a0ec 2737
617ba13b
MC		 2738 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2739 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2740 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
ac27a0ec
DK		 2741 } else {
2742 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2743 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
617ba13b 2744 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1330593e 2745 (!is_power_of_2(sbi->s_inode_size)) ||
ac27a0ec 2746 (sbi->s_inode_size > blocksize)) {
b31e1552
ES		 2747 ext4_msg(sb, KERN_ERR,
2748 "unsupported inode size: %d",
2b2d6d01 2749 sbi->s_inode_size);
ac27a0ec
DK		 2750 goto failed_mount;
2751 }
ef7f3835
KS		 2752 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2753 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
ac27a0ec 2754 }
0b8e58a1 2755
0d1ee42f
AR		 2756 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2757 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
8fadc143 2758 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 2759 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 2760 !is_power_of_2(sbi->s_desc_size)) {
b31e1552
ES		 2761 ext4_msg(sb, KERN_ERR,
2762 "unsupported descriptor size %lu",
0d1ee42f
AR		 2763 sbi->s_desc_size);
2764 goto failed_mount;
2765 }
2766 } else
2767 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
0b8e58a1 2768
ac27a0ec 2769 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 2770 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
b47b6f38 2771 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
617ba13b 2772 goto cantfind_ext4;
0b8e58a1 2773
617ba13b 2774 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 2775 if (sbi->s_inodes_per_block == 0)
617ba13b 2776 goto cantfind_ext4;
ac27a0ec
DK		 2777 sbi->s_itb_per_group = sbi->s_inodes_per_group /
2778 sbi->s_inodes_per_block;
0d1ee42f 2779 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 2780 sbi->s_sbh = bh;
2781 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 2782 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2783 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
0b8e58a1 2784
2b2d6d01 2785 for (i = 0; i < 4; i++)
ac27a0ec
DK		 2786 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2787 sbi->s_def_hash_version = es->s_def_hash_version;
f99b2589
TT		 2788 i = le32_to_cpu(es->s_flags);
2789 if (i & EXT2_FLAGS_UNSIGNED_HASH)
2790 sbi->s_hash_unsigned = 3;
2791 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
2792#ifdef __CHAR_UNSIGNED__
2793 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
2794 sbi->s_hash_unsigned = 3;
2795#else
2796 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
2797#endif
2798 sb->s_dirt = 1;
2799 }
ac27a0ec
DK		 2800
2801 if (sbi->s_blocks_per_group > blocksize * 8) {
b31e1552
ES		 2802 ext4_msg(sb, KERN_ERR,
2803 "#blocks per group too big: %lu",
2b2d6d01 2804 sbi->s_blocks_per_group);
ac27a0ec
DK		 2805 goto failed_mount;
2806 }
ac27a0ec 2807 if (sbi->s_inodes_per_group > blocksize * 8) {
b31e1552
ES		 2808 ext4_msg(sb, KERN_ERR,
2809 "#inodes per group too big: %lu",
2b2d6d01 2810 sbi->s_inodes_per_group);
ac27a0ec
DK		 2811 goto failed_mount;
2812 }
2813
bf43d84b
ES		 2814 /*
2815 * Test whether we have more sectors than will fit in sector_t,
2816 * and whether the max offset is addressable by the page cache.
2817 */
2818 if ((ext4_blocks_count(es) >
2819 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) ||
2820 (ext4_blocks_count(es) >
2821 (pgoff_t)(~0ULL) >> (PAGE_CACHE_SHIFT - sb->s_blocksize_bits))) {
b31e1552 2822 ext4_msg(sb, KERN_ERR, "filesystem"
bf43d84b 2823 " too large to mount safely on this system");
ac27a0ec 2824 if (sizeof(sector_t) < 8)
90c699a9 2825 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
bf43d84b 2826 ret = -EFBIG;
ac27a0ec
DK		 2827 goto failed_mount;
2828 }
2829
617ba13b
MC		 2830 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2831 goto cantfind_ext4;
e7c95593 2832
0f2ddca6
FTN		 2833 /* check blocks count against device size */
2834 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
2835 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
b31e1552
ES		 2836 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
2837 "exceeds size of device (%llu blocks)",
0f2ddca6
FTN		 2838 ext4_blocks_count(es), blocks_count);
2839 goto failed_mount;
2840 }
2841
0b8e58a1
AD		 2842 /*
2843 * It makes no sense for the first data block to be beyond the end
2844 * of the filesystem.
2845 */
2846 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
b31e1552
ES		 2847 ext4_msg(sb, KERN_WARNING, "bad geometry: first data"
2848 "block %u is beyond end of filesystem (%llu)",
2849 le32_to_cpu(es->s_first_data_block),
2850 ext4_blocks_count(es));
e7c95593
ES		 2851 goto failed_mount;
2852 }
bd81d8ee
LV		 2853 blocks_count = (ext4_blocks_count(es) -
2854 le32_to_cpu(es->s_first_data_block) +
2855 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2856 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4ec11028 2857 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
b31e1552 2858 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
4ec11028 2859 "(block count %llu, first data block %u, "
b31e1552 2860 "blocks per group %lu)", sbi->s_groups_count,
4ec11028
TT		 2861 ext4_blocks_count(es),
2862 le32_to_cpu(es->s_first_data_block),
2863 EXT4_BLOCKS_PER_GROUP(sb));
2864 goto failed_mount;
2865 }
bd81d8ee 2866 sbi->s_groups_count = blocks_count;
fb0a387d
ES		 2867 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
2868 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
617ba13b
MC		 2869 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2870 EXT4_DESC_PER_BLOCK(sb);
2b2d6d01 2871 sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
ac27a0ec
DK		 2872 GFP_KERNEL);
2873 if (sbi->s_group_desc == NULL) {
b31e1552 2874 ext4_msg(sb, KERN_ERR, "not enough memory");
ac27a0ec
DK		 2875 goto failed_mount;
2876 }
2877
3244fcb1 2878#ifdef CONFIG_PROC_FS
9f6200bb
TT		 2879 if (ext4_proc_root)
2880 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
3244fcb1 2881#endif
240799cd 2882
705895b6 2883 bgl_lock_init(sbi->s_blockgroup_lock);
ac27a0ec
DK		 2884
2885 for (i = 0; i < db_count; i++) {
70bbb3e0 2886 block = descriptor_loc(sb, logical_sb_block, i);
ac27a0ec
DK		 2887 sbi->s_group_desc[i] = sb_bread(sb, block);
2888 if (!sbi->s_group_desc[i]) {
b31e1552
ES		 2889 ext4_msg(sb, KERN_ERR,
2890 "can't read group descriptor %d", i);
ac27a0ec
DK		 2891 db_count = i;
2892 goto failed_mount2;
2893 }
2894 }
2b2d6d01 2895 if (!ext4_check_descriptors(sb)) {
b31e1552 2896 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
ac27a0ec
DK		 2897 goto failed_mount2;
2898 }
772cb7c8
JS		 2899 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2900 if (!ext4_fill_flex_info(sb)) {
b31e1552
ES		 2901 ext4_msg(sb, KERN_ERR,
2902 "unable to initialize "
2903 "flex_bg meta info!");
772cb7c8
JS		 2904 goto failed_mount2;
2905 }
2906
ac27a0ec
DK		 2907 sbi->s_gdb_count = db_count;
2908 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2909 spin_lock_init(&sbi->s_next_gen_lock);
2910
c9de560d 2911 sbi->s_stripe = ext4_get_stripe_size(sbi);
55138e0b 2912 sbi->s_max_writeback_mb_bump = 128;
c9de560d 2913
ac27a0ec
DK		 2914 /*
2915 * set up enough so that it can read an inode
2916 */
9ca92389
TT		 2917 if (!test_opt(sb, NOLOAD) &&
2918 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
2919 sb->s_op = &ext4_sops;
2920 else
2921 sb->s_op = &ext4_nojournal_sops;
617ba13b
MC		 2922 sb->s_export_op = &ext4_export_ops;
2923 sb->s_xattr = ext4_xattr_handlers;
ac27a0ec 2924#ifdef CONFIG_QUOTA
617ba13b
MC		 2925 sb->s_qcop = &ext4_qctl_operations;
2926 sb->dq_op = &ext4_quota_operations;
ac27a0ec
DK		 2927#endif
2928 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3b9d4ed2 2929 mutex_init(&sbi->s_orphan_lock);
32ed5058 2930 mutex_init(&sbi->s_resize_lock);
ac27a0ec
DK		 2931
2932 sb->s_root = NULL;
2933
2934 needs_recovery = (es->s_last_orphan != 0 ||
617ba13b
MC		 2935 EXT4_HAS_INCOMPAT_FEATURE(sb,
2936 EXT4_FEATURE_INCOMPAT_RECOVER));
ac27a0ec
DK		 2937
2938 /*
2939 * The first inode we look at is the journal inode. Don't try
2940 * root first: it may be modified in the journal!
2941 */
2942 if (!test_opt(sb, NOLOAD) &&
617ba13b
MC		 2943 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2944 if (ext4_load_journal(sb, es, journal_devnum))
ac27a0ec 2945 goto failed_mount3;
0390131b
FM		 2946 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
2947 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
b31e1552
ES		 2948 ext4_msg(sb, KERN_ERR, "required journal recovery "
2949 "suppressed and not mounted read-only");
744692dc 2950 goto failed_mount_wq;
ac27a0ec 2951 } else {
0390131b
FM		 2952 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
2953 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
2954 sbi->s_journal = NULL;
2955 needs_recovery = 0;
2956 goto no_journal;
ac27a0ec
DK		 2957 }
2958
eb40a09c
JS		 2959 if (ext4_blocks_count(es) > 0xffffffffULL &&
2960 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2961 JBD2_FEATURE_INCOMPAT_64BIT)) {
b31e1552 2962 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
744692dc 2963 goto failed_mount_wq;
eb40a09c
JS		 2964 }
2965
d4da6c9c
LT		 2966 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2967 jbd2_journal_set_features(sbi->s_journal,
2968 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
818d276c 2969 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
d4da6c9c
LT		 2970 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2971 jbd2_journal_set_features(sbi->s_journal,
2972 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
818d276c
GS		 2973 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2974 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
d4da6c9c
LT		 2975 } else {
2976 jbd2_journal_clear_features(sbi->s_journal,
2977 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2978 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2979 }
818d276c 2980
ac27a0ec
DK		 2981 /* We have now updated the journal if required, so we can
2982 * validate the data journaling mode. */
2983 switch (test_opt(sb, DATA_FLAGS)) {
2984 case 0:
2985 /* No mode set, assume a default based on the journal
63f57933
AM		 2986 * capabilities: ORDERED_DATA if the journal can
2987 * cope, else JOURNAL_DATA
2988 */
dab291af
MC		 2989 if (jbd2_journal_check_available_features
2990 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
ac27a0ec
DK		 2991 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2992 else
2993 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2994 break;
2995
617ba13b
MC		 2996 case EXT4_MOUNT_ORDERED_DATA:
2997 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 2998 if (!jbd2_journal_check_available_features
2999 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
b31e1552
ES		 3000 ext4_msg(sb, KERN_ERR, "Journal does not support "
3001 "requested data journaling mode");
744692dc 3002 goto failed_mount_wq;
ac27a0ec
DK		 3003 }
3004 default:
3005 break;
3006 }
b3881f74 3007 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
ac27a0ec 3008
0390131b 3009no_journal:
84061e07
DM		 3010 err = percpu_counter_init(&sbi->s_freeblocks_counter,
3011 ext4_count_free_blocks(sb));
3012 if (!err)
3013 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3014 ext4_count_free_inodes(sb));
3015 if (!err)
3016 err = percpu_counter_init(&sbi->s_dirs_counter,
3017 ext4_count_dirs(sb));
3018 if (!err)
3019 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
3020 if (err) {
3021 ext4_msg(sb, KERN_ERR, "insufficient memory");
3022 goto failed_mount_wq;
3023 }
206f7ab4 3024
4c0425ff
MC		 3025 EXT4_SB(sb)->dio_unwritten_wq = create_workqueue("ext4-dio-unwritten");
3026 if (!EXT4_SB(sb)->dio_unwritten_wq) {
3027 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
3028 goto failed_mount_wq;
3029 }
3030
ac27a0ec 3031 /*
dab291af 3032 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 3033 * so we can safely mount the rest of the filesystem now.
3034 */
3035
1d1fe1ee
DH		 3036 root = ext4_iget(sb, EXT4_ROOT_INO);
3037 if (IS_ERR(root)) {
b31e1552 3038 ext4_msg(sb, KERN_ERR, "get root inode failed");
1d1fe1ee 3039 ret = PTR_ERR(root);
ac27a0ec
DK		 3040 goto failed_mount4;
3041 }
3042 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1d1fe1ee 3043 iput(root);
b31e1552 3044 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
ac27a0ec
DK		 3045 goto failed_mount4;
3046 }
1d1fe1ee
DH		 3047 sb->s_root = d_alloc_root(root);
3048 if (!sb->s_root) {
b31e1552 3049 ext4_msg(sb, KERN_ERR, "get root dentry failed");
1d1fe1ee
DH		 3050 iput(root);
3051 ret = -ENOMEM;
3052 goto failed_mount4;
3053 }
ac27a0ec 3054
2b2d6d01 3055 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
ef7f3835
KS		 3056
3057 /* determine the minimum size of new large inodes, if present */
3058 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3059 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3060 EXT4_GOOD_OLD_INODE_SIZE;
3061 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3062 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
3063 if (sbi->s_want_extra_isize <
3064 le16_to_cpu(es->s_want_extra_isize))
3065 sbi->s_want_extra_isize =
3066 le16_to_cpu(es->s_want_extra_isize);
3067 if (sbi->s_want_extra_isize <
3068 le16_to_cpu(es->s_min_extra_isize))
3069 sbi->s_want_extra_isize =
3070 le16_to_cpu(es->s_min_extra_isize);
3071 }
3072 }
3073 /* Check if enough inode space is available */
3074 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3075 sbi->s_inode_size) {
3076 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3077 EXT4_GOOD_OLD_INODE_SIZE;
b31e1552
ES		 3078 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3079 "available");
ef7f3835
KS		 3080 }
3081
90576c0b
TT		 3082 if (test_opt(sb, DELALLOC) &&
3083 (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)) {
b31e1552
ES		 3084 ext4_msg(sb, KERN_WARNING, "Ignoring delalloc option - "
3085 "requested data journaling mode");
c2774d84 3086 clear_opt(sbi->s_mount_opt, DELALLOC);
90576c0b 3087 }
744692dc
JZ		 3088 if (test_opt(sb, DIOREAD_NOLOCK)) {
3089 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3090 ext4_msg(sb, KERN_WARNING, "Ignoring dioread_nolock "
3091 "option - requested data journaling mode");
3092 clear_opt(sbi->s_mount_opt, DIOREAD_NOLOCK);
3093 }
3094 if (sb->s_blocksize < PAGE_SIZE) {
3095 ext4_msg(sb, KERN_WARNING, "Ignoring dioread_nolock "
3096 "option - block size is too small");
3097 clear_opt(sbi->s_mount_opt, DIOREAD_NOLOCK);
3098 }
3099 }
c2774d84 3100
6fd058f7
TT		 3101 err = ext4_setup_system_zone(sb);
3102 if (err) {
b31e1552 3103 ext4_msg(sb, KERN_ERR, "failed to initialize system "
fbe845dd 3104 "zone (%d)", err);
6fd058f7
TT		 3105 goto failed_mount4;
3106 }
3107
c2774d84
AK		 3108 ext4_ext_init(sb);
3109 err = ext4_mb_init(sb, needs_recovery);
3110 if (err) {
b31e1552
ES		 3111 ext4_msg(sb, KERN_ERR, "failed to initalize mballoc (%d)",
3112 err);
c2774d84
AK		 3113 goto failed_mount4;
3114 }
3115
3197ebdb
TT		 3116 sbi->s_kobj.kset = ext4_kset;
3117 init_completion(&sbi->s_kobj_unregister);
3118 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
3119 "%s", sb->s_id);
3120 if (err) {
3121 ext4_mb_release(sb);
3122 ext4_ext_release(sb);
3123 goto failed_mount4;
3124 };
3125
617ba13b
MC		 3126 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3127 ext4_orphan_cleanup(sb, es);
3128 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
0390131b 3129 if (needs_recovery) {
b31e1552 3130 ext4_msg(sb, KERN_INFO, "recovery complete");
0390131b
FM		 3131 ext4_mark_recovery_complete(sb, es);
3132 }
3133 if (EXT4_SB(sb)->s_journal) {
3134 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3135 descr = " journalled data mode";
3136 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3137 descr = " ordered data mode";
3138 else
3139 descr = " writeback data mode";
3140 } else
3141 descr = "out journal";
3142
d4c402d9
CW		 3143 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
3144 "Opts: %s", descr, orig_data);
ac27a0ec 3145
66e61a9e
TT		 3146 init_timer(&sbi->s_err_report);
3147 sbi->s_err_report.function = print_daily_error_info;
3148 sbi->s_err_report.data = (unsigned long) sb;
3149 if (es->s_error_count)
3150 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
3151
ac27a0ec 3152 lock_kernel();
d4c402d9 3153 kfree(orig_data);
ac27a0ec
DK		 3154 return 0;
3155
617ba13b 3156cantfind_ext4:
ac27a0ec 3157 if (!silent)
b31e1552 3158 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
ac27a0ec
DK		 3159 goto failed_mount;
3160
3161failed_mount4:
b31e1552 3162 ext4_msg(sb, KERN_ERR, "mount failed");
4c0425ff
MC		 3163 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
3164failed_mount_wq:
6fd058f7 3165 ext4_release_system_zone(sb);
0390131b
FM		 3166 if (sbi->s_journal) {
3167 jbd2_journal_destroy(sbi->s_journal);
3168 sbi->s_journal = NULL;
3169 }
84061e07
DM		 3170 percpu_counter_destroy(&sbi->s_freeblocks_counter);
3171 percpu_counter_destroy(&sbi->s_freeinodes_counter);
3172 percpu_counter_destroy(&sbi->s_dirs_counter);
3173 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
ac27a0ec 3174failed_mount3:
c5ca7c76
TT		 3175 if (sbi->s_flex_groups) {
3176 if (is_vmalloc_addr(sbi->s_flex_groups))
3177 vfree(sbi->s_flex_groups);
3178 else
3179 kfree(sbi->s_flex_groups);
3180 }
ac27a0ec
DK		 3181failed_mount2:
3182 for (i = 0; i < db_count; i++)
3183 brelse(sbi->s_group_desc[i]);
3184 kfree(sbi->s_group_desc);
3185failed_mount:
240799cd 3186 if (sbi->s_proc) {
9f6200bb 3187 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 3188 }
ac27a0ec
DK		 3189#ifdef CONFIG_QUOTA
3190 for (i = 0; i < MAXQUOTAS; i++)
3191 kfree(sbi->s_qf_names[i]);
3192#endif
617ba13b 3193 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 3194 brelse(bh);
3195out_fail:
3196 sb->s_fs_info = NULL;
f6830165 3197 kfree(sbi->s_blockgroup_lock);
ac27a0ec
DK		 3198 kfree(sbi);
3199 lock_kernel();
dcc7dae3 3200out_free_orig:
d4c402d9 3201 kfree(orig_data);
1d1fe1ee 3202 return ret;
ac27a0ec
DK		 3203}
3204
3205/*
3206 * Setup any per-fs journal parameters now. We'll do this both on
3207 * initial mount, once the journal has been initialised but before we've
3208 * done any recovery; and again on any subsequent remount.
3209 */
617ba13b 3210static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 3211{
617ba13b 3212 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 3213
30773840
TT		 3214 journal->j_commit_interval = sbi->s_commit_interval;
3215 journal->j_min_batch_time = sbi->s_min_batch_time;
3216 journal->j_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 3217
3218 spin_lock(&journal->j_state_lock);
3219 if (test_opt(sb, BARRIER))
dab291af 3220 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 3221 else
dab291af 3222 journal->j_flags &= ~JBD2_BARRIER;
5bf5683a
HK		 3223 if (test_opt(sb, DATA_ERR_ABORT))
3224 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
3225 else
3226 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
ac27a0ec
DK		 3227 spin_unlock(&journal->j_state_lock);
3228}
3229
617ba13b 3230static journal_t *ext4_get_journal(struct super_block *sb,
ac27a0ec
DK		 3231 unsigned int journal_inum)
3232{
3233 struct inode *journal_inode;
3234 journal_t *journal;
3235
0390131b
FM		 3236 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3237
ac27a0ec
DK		 3238 /* First, test for the existence of a valid inode on disk. Bad
3239 * things happen if we iget() an unused inode, as the subsequent
3240 * iput() will try to delete it. */
3241
1d1fe1ee
DH		 3242 journal_inode = ext4_iget(sb, journal_inum);
3243 if (IS_ERR(journal_inode)) {
b31e1552 3244 ext4_msg(sb, KERN_ERR, "no journal found");
ac27a0ec
DK		 3245 return NULL;
3246 }
3247 if (!journal_inode->i_nlink) {
3248 make_bad_inode(journal_inode);
3249 iput(journal_inode);
b31e1552 3250 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
ac27a0ec
DK		 3251 return NULL;
3252 }
3253
e5f8eab8 3254 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
ac27a0ec 3255 journal_inode, journal_inode->i_size);
1d1fe1ee 3256 if (!S_ISREG(journal_inode->i_mode)) {
b31e1552 3257 ext4_msg(sb, KERN_ERR, "invalid journal inode");
ac27a0ec
DK		 3258 iput(journal_inode);
3259 return NULL;
3260 }
3261
dab291af 3262 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 3263 if (!journal) {
b31e1552 3264 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
ac27a0ec
DK		 3265 iput(journal_inode);
3266 return NULL;
3267 }
3268 journal->j_private = sb;
617ba13b 3269 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 3270 return journal;
3271}
3272
617ba13b 3273static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 3274 dev_t j_dev)
3275{
2b2d6d01 3276 struct buffer_head *bh;
ac27a0ec 3277 journal_t *journal;
617ba13b
MC		 3278 ext4_fsblk_t start;
3279 ext4_fsblk_t len;
ac27a0ec 3280 int hblock, blocksize;
617ba13b 3281 ext4_fsblk_t sb_block;
ac27a0ec 3282 unsigned long offset;
2b2d6d01 3283 struct ext4_super_block *es;
ac27a0ec
DK		 3284 struct block_device *bdev;
3285
0390131b
FM		 3286 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3287
b31e1552 3288 bdev = ext4_blkdev_get(j_dev, sb);
ac27a0ec
DK		 3289 if (bdev == NULL)
3290 return NULL;
3291
3292 if (bd_claim(bdev, sb)) {
b31e1552
ES		 3293 ext4_msg(sb, KERN_ERR,
3294 "failed to claim external journal device");
9a1c3542 3295 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
ac27a0ec
DK		 3296 return NULL;
3297 }
3298
3299 blocksize = sb->s_blocksize;
e1defc4f 3300 hblock = bdev_logical_block_size(bdev);
ac27a0ec 3301 if (blocksize < hblock) {
b31e1552
ES		 3302 ext4_msg(sb, KERN_ERR,
3303 "blocksize too small for journal device");
ac27a0ec
DK		 3304 goto out_bdev;
3305 }
3306
617ba13b
MC		 3307 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3308 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 3309 set_blocksize(bdev, blocksize);
3310 if (!(bh = __bread(bdev, sb_block, blocksize))) {
b31e1552
ES		 3311 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
3312 "external journal");
ac27a0ec
DK		 3313 goto out_bdev;
3314 }
3315
617ba13b
MC		 3316 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3317 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 3318 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b 3319 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
b31e1552
ES		 3320 ext4_msg(sb, KERN_ERR, "external journal has "
3321 "bad superblock");
ac27a0ec
DK		 3322 brelse(bh);
3323 goto out_bdev;
3324 }
3325
617ba13b 3326 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
b31e1552 3327 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
ac27a0ec
DK		 3328 brelse(bh);
3329 goto out_bdev;
3330 }
3331
bd81d8ee 3332 len = ext4_blocks_count(es);
ac27a0ec
DK		 3333 start = sb_block + 1;
3334 brelse(bh); /* we're done with the superblock */
3335
dab291af 3336 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 3337 start, len, blocksize);
3338 if (!journal) {
b31e1552 3339 ext4_msg(sb, KERN_ERR, "failed to create device journal");
ac27a0ec
DK		 3340 goto out_bdev;
3341 }
3342 journal->j_private = sb;
3343 ll_rw_block(READ, 1, &journal->j_sb_buffer);
3344 wait_on_buffer(journal->j_sb_buffer);
3345 if (!buffer_uptodate(journal->j_sb_buffer)) {
b31e1552 3346 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
ac27a0ec
DK		 3347 goto out_journal;
3348 }
3349 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
b31e1552
ES		 3350 ext4_msg(sb, KERN_ERR, "External journal has more than one "
3351 "user (unsupported) - %d",
ac27a0ec
DK		 3352 be32_to_cpu(journal->j_superblock->s_nr_users));
3353 goto out_journal;
3354 }
617ba13b
MC		 3355 EXT4_SB(sb)->journal_bdev = bdev;
3356 ext4_init_journal_params(sb, journal);
ac27a0ec 3357 return journal;
0b8e58a1 3358
ac27a0ec 3359out_journal:
dab291af 3360 jbd2_journal_destroy(journal);
ac27a0ec 3361out_bdev:
617ba13b 3362 ext4_blkdev_put(bdev);
ac27a0ec
DK		 3363 return NULL;
3364}
3365
617ba13b
MC		 3366static int ext4_load_journal(struct super_block *sb,
3367 struct ext4_super_block *es,
ac27a0ec
DK		 3368 unsigned long journal_devnum)
3369{
3370 journal_t *journal;
3371 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
3372 dev_t journal_dev;
3373 int err = 0;
3374 int really_read_only;
3375
0390131b
FM		 3376 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3377
ac27a0ec
DK		 3378 if (journal_devnum &&
3379 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
b31e1552
ES		 3380 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
3381 "numbers have changed");
ac27a0ec
DK		 3382 journal_dev = new_decode_dev(journal_devnum);
3383 } else
3384 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
3385
3386 really_read_only = bdev_read_only(sb->s_bdev);
3387
3388 /*
3389 * Are we loading a blank journal or performing recovery after a
3390 * crash? For recovery, we need to check in advance whether we
3391 * can get read-write access to the device.
3392 */
617ba13b 3393 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
ac27a0ec 3394 if (sb->s_flags & MS_RDONLY) {
b31e1552
ES		 3395 ext4_msg(sb, KERN_INFO, "INFO: recovery "
3396 "required on readonly filesystem");
ac27a0ec 3397 if (really_read_only) {
b31e1552
ES		 3398 ext4_msg(sb, KERN_ERR, "write access "
3399 "unavailable, cannot proceed");
ac27a0ec
DK		 3400 return -EROFS;
3401 }
b31e1552
ES		 3402 ext4_msg(sb, KERN_INFO, "write access will "
3403 "be enabled during recovery");
ac27a0ec
DK		 3404 }
3405 }
3406
3407 if (journal_inum && journal_dev) {
b31e1552
ES		 3408 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
3409 "and inode journals!");
ac27a0ec
DK		 3410 return -EINVAL;
3411 }
3412
3413 if (journal_inum) {
617ba13b 3414 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 3415 return -EINVAL;
3416 } else {
617ba13b 3417 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 3418 return -EINVAL;
3419 }
3420
90576c0b 3421 if (!(journal->j_flags & JBD2_BARRIER))
b31e1552 3422 ext4_msg(sb, KERN_INFO, "barriers disabled");
4776004f 3423
ac27a0ec 3424 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
dab291af 3425 err = jbd2_journal_update_format(journal);
ac27a0ec 3426 if (err) {
b31e1552 3427 ext4_msg(sb, KERN_ERR, "error updating journal");
dab291af 3428 jbd2_journal_destroy(journal);
ac27a0ec
DK		 3429 return err;
3430 }
3431 }
3432
617ba13b 3433 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
dab291af 3434 err = jbd2_journal_wipe(journal, !really_read_only);
1c13d5c0
TT		 3435 if (!err) {
3436 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
3437 if (save)
3438 memcpy(save, ((char *) es) +
3439 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
dab291af 3440 err = jbd2_journal_load(journal);
1c13d5c0
TT		 3441 if (save)
3442 memcpy(((char *) es) + EXT4_S_ERR_START,
3443 save, EXT4_S_ERR_LEN);
3444 kfree(save);
3445 }
ac27a0ec
DK		 3446
3447 if (err) {
b31e1552 3448 ext4_msg(sb, KERN_ERR, "error loading journal");
dab291af 3449 jbd2_journal_destroy(journal);
ac27a0ec
DK		 3450 return err;
3451 }
3452
617ba13b
MC		 3453 EXT4_SB(sb)->s_journal = journal;
3454 ext4_clear_journal_err(sb, es);
ac27a0ec
DK		 3455
3456 if (journal_devnum &&
3457 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3458 es->s_journal_dev = cpu_to_le32(journal_devnum);
ac27a0ec
DK		 3459
3460 /* Make sure we flush the recovery flag to disk. */
e2d67052 3461 ext4_commit_super(sb, 1);
ac27a0ec
DK		 3462 }
3463
3464 return 0;
3465}
3466
e2d67052 3467static int ext4_commit_super(struct super_block *sb, int sync)
ac27a0ec 3468{
e2d67052 3469 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
617ba13b 3470 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
c4be0c1d 3471 int error = 0;
ac27a0ec
DK		 3472
3473 if (!sbh)
c4be0c1d 3474 return error;
914258bf
TT		 3475 if (buffer_write_io_error(sbh)) {
3476 /*
3477 * Oh, dear. A previous attempt to write the
3478 * superblock failed. This could happen because the
3479 * USB device was yanked out. Or it could happen to
3480 * be a transient write error and maybe the block will
3481 * be remapped. Nothing we can do but to retry the
3482 * write and hope for the best.
3483 */
b31e1552
ES		 3484 ext4_msg(sb, KERN_ERR, "previous I/O error to "
3485 "superblock detected");
914258bf
TT		 3486 clear_buffer_write_io_error(sbh);
3487 set_buffer_uptodate(sbh);
3488 }
71290b36
TT		 3489 /*
3490 * If the file system is mounted read-only, don't update the
3491 * superblock write time. This avoids updating the superblock
3492 * write time when we are mounting the root file system
3493 * read/only but we need to replay the journal; at that point,
3494 * for people who are east of GMT and who make their clock
3495 * tick in localtime for Windows bug-for-bug compatibility,
3496 * the clock is set in the future, and this will cause e2fsck
3497 * to complain and force a full file system check.
3498 */
3499 if (!(sb->s_flags & MS_RDONLY))
3500 es->s_wtime = cpu_to_le32(get_seconds());
f613dfcb
TT		 3501 if (sb->s_bdev->bd_part)
3502 es->s_kbytes_written =
3503 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
afc32f7e
TT		 3504 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
3505 EXT4_SB(sb)->s_sectors_written_start) >> 1));
f613dfcb
TT		 3506 else
3507 es->s_kbytes_written =
3508 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
5d1b1b3f
AK		 3509 ext4_free_blocks_count_set(es, percpu_counter_sum_positive(
3510 &EXT4_SB(sb)->s_freeblocks_counter));
3511 es->s_free_inodes_count = cpu_to_le32(percpu_counter_sum_positive(
3512 &EXT4_SB(sb)->s_freeinodes_counter));
7234ab2a 3513 sb->s_dirt = 0;
ac27a0ec
DK		 3514 BUFFER_TRACE(sbh, "marking dirty");
3515 mark_buffer_dirty(sbh);
914258bf 3516 if (sync) {
c4be0c1d
TS		 3517 error = sync_dirty_buffer(sbh);
3518 if (error)
3519 return error;
3520
3521 error = buffer_write_io_error(sbh);
3522 if (error) {
b31e1552
ES		 3523 ext4_msg(sb, KERN_ERR, "I/O error while writing "
3524 "superblock");
914258bf
TT		 3525 clear_buffer_write_io_error(sbh);
3526 set_buffer_uptodate(sbh);
3527 }
3528 }
c4be0c1d 3529 return error;
ac27a0ec
DK		 3530}
3531
ac27a0ec
DK		 3532/*
3533 * Have we just finished recovery? If so, and if we are mounting (or
3534 * remounting) the filesystem readonly, then we will end up with a
3535 * consistent fs on disk. Record that fact.
3536 */
2b2d6d01
TT		 3537static void ext4_mark_recovery_complete(struct super_block *sb,
3538 struct ext4_super_block *es)
ac27a0ec 3539{
617ba13b 3540 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 3541
0390131b
FM		 3542 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3543 BUG_ON(journal != NULL);
3544 return;
3545 }
dab291af 3546 jbd2_journal_lock_updates(journal);
7ffe1ea8
HK		 3547 if (jbd2_journal_flush(journal) < 0)
3548 goto out;
3549
617ba13b 3550 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
ac27a0ec 3551 sb->s_flags & MS_RDONLY) {
617ba13b 3552 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
e2d67052 3553 ext4_commit_super(sb, 1);
ac27a0ec 3554 }
7ffe1ea8
HK		 3555
3556out:
dab291af 3557 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 3558}
3559
3560/*
3561 * If we are mounting (or read-write remounting) a filesystem whose journal
3562 * has recorded an error from a previous lifetime, move that error to the
3563 * main filesystem now.
3564 */
2b2d6d01
TT		 3565static void ext4_clear_journal_err(struct super_block *sb,
3566 struct ext4_super_block *es)
ac27a0ec
DK		 3567{
3568 journal_t *journal;
3569 int j_errno;
3570 const char *errstr;
3571
0390131b
FM		 3572 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3573
617ba13b 3574 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 3575
3576 /*
3577 * Now check for any error status which may have been recorded in the
617ba13b 3578 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 3579 */
3580
dab291af 3581 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 3582 if (j_errno) {
3583 char nbuf[16];
3584
617ba13b 3585 errstr = ext4_decode_error(sb, j_errno, nbuf);
12062ddd 3586 ext4_warning(sb, "Filesystem error recorded "
ac27a0ec 3587 "from previous mount: %s", errstr);
12062ddd 3588 ext4_warning(sb, "Marking fs in need of filesystem check.");
ac27a0ec 3589
617ba13b
MC		 3590 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
3591 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
e2d67052 3592 ext4_commit_super(sb, 1);
ac27a0ec 3593
dab291af 3594 jbd2_journal_clear_err(journal);
ac27a0ec
DK		 3595 }
3596}
3597
3598/*
3599 * Force the running and committing transactions to commit,
3600 * and wait on the commit.
3601 */
617ba13b 3602int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 3603{
3604 journal_t *journal;
0390131b 3605 int ret = 0;
ac27a0ec
DK		 3606
3607 if (sb->s_flags & MS_RDONLY)
3608 return 0;
3609
617ba13b 3610 journal = EXT4_SB(sb)->s_journal;
6b0310fb 3611 if (journal) {
437f88cc 3612 vfs_check_frozen(sb, SB_FREEZE_TRANS);
0390131b 3613 ret = ext4_journal_force_commit(journal);
6b0310fb 3614 }
0390131b 3615
ac27a0ec
DK		 3616 return ret;
3617}
3618
2b2d6d01 3619static void ext4_write_super(struct super_block *sb)
ac27a0ec 3620{
ebc1ac16 3621 lock_super(sb);
9ca92389 3622 ext4_commit_super(sb, 1);
ebc1ac16 3623 unlock_super(sb);
ac27a0ec
DK		 3624}
3625
617ba13b 3626static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec 3627{
14ce0cb4 3628 int ret = 0;
9eddacf9 3629 tid_t target;
8d5d02e6 3630 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 3631
9bffad1e 3632 trace_ext4_sync_fs(sb, wait);
8d5d02e6
MC		 3633 flush_workqueue(sbi->dio_unwritten_wq);
3634 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
9ca92389 3635 if (wait)
8d5d02e6 3636 jbd2_log_wait_commit(sbi->s_journal, target);
0390131b 3637 }
14ce0cb4 3638 return ret;
ac27a0ec
DK		 3639}
3640
3641/*
3642 * LVM calls this function before a (read-only) snapshot is created. This
3643 * gives us a chance to flush the journal completely and mark the fs clean.
3644 */
c4be0c1d 3645static int ext4_freeze(struct super_block *sb)
ac27a0ec 3646{
c4be0c1d
TS		 3647 int error = 0;
3648 journal_t *journal;
ac27a0ec 3649
9ca92389
TT		 3650 if (sb->s_flags & MS_RDONLY)
3651 return 0;
ac27a0ec 3652
9ca92389 3653 journal = EXT4_SB(sb)->s_journal;
7ffe1ea8 3654
9ca92389
TT		 3655 /* Now we set up the journal barrier. */
3656 jbd2_journal_lock_updates(journal);
ac27a0ec 3657
9ca92389
TT		 3658 /*
3659 * Don't clear the needs_recovery flag if we failed to flush
3660 * the journal.
3661 */
3662 error = jbd2_journal_flush(journal);
6b0310fb
ES		 3663 if (error < 0)
3664 goto out;
9ca92389
TT		 3665
3666 /* Journal blocked and flushed, clear needs_recovery flag. */
3667 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3668 error = ext4_commit_super(sb, 1);
6b0310fb
ES		 3669out:
3670 /* we rely on s_frozen to stop further updates */
3671 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3672 return error;
ac27a0ec
DK		 3673}
3674
3675/*
3676 * Called by LVM after the snapshot is done. We need to reset the RECOVER
3677 * flag here, even though the filesystem is not technically dirty yet.
3678 */
c4be0c1d 3679static int ext4_unfreeze(struct super_block *sb)
ac27a0ec 3680{
9ca92389
TT		 3681 if (sb->s_flags & MS_RDONLY)
3682 return 0;
3683
3684 lock_super(sb);
3685 /* Reset the needs_recovery flag before the fs is unlocked. */
3686 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3687 ext4_commit_super(sb, 1);
3688 unlock_super(sb);
c4be0c1d 3689 return 0;
ac27a0ec
DK		 3690}
3691
2b2d6d01 3692static int ext4_remount(struct super_block *sb, int *flags, char *data)
ac27a0ec 3693{
2b2d6d01 3694 struct ext4_super_block *es;
617ba13b
MC		 3695 struct ext4_sb_info *sbi = EXT4_SB(sb);
3696 ext4_fsblk_t n_blocks_count = 0;
ac27a0ec 3697 unsigned long old_sb_flags;
617ba13b 3698 struct ext4_mount_options old_opts;
c79d967d 3699 int enable_quota = 0;
8a266467 3700 ext4_group_t g;
b3881f74 3701 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
ac27a0ec
DK		 3702 int err;
3703#ifdef CONFIG_QUOTA
3704 int i;
3705#endif
d4c402d9 3706 char *orig_data = kstrdup(data, GFP_KERNEL);
ac27a0ec 3707
337eb00a
AIB		 3708 lock_kernel();
3709
ac27a0ec 3710 /* Store the original options */
bbd6851a 3711 lock_super(sb);
ac27a0ec
DK		 3712 old_sb_flags = sb->s_flags;
3713 old_opts.s_mount_opt = sbi->s_mount_opt;
3714 old_opts.s_resuid = sbi->s_resuid;
3715 old_opts.s_resgid = sbi->s_resgid;
3716 old_opts.s_commit_interval = sbi->s_commit_interval;
30773840
TT		 3717 old_opts.s_min_batch_time = sbi->s_min_batch_time;
3718 old_opts.s_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 3719#ifdef CONFIG_QUOTA
3720 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
3721 for (i = 0; i < MAXQUOTAS; i++)
3722 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
3723#endif
b3881f74
TT		 3724 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
3725 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
ac27a0ec
DK		 3726
3727 /*
3728 * Allow the "check" option to be passed as a remount option.
3729 */
b3881f74
TT		 3730 if (!parse_options(data, sb, NULL, &journal_ioprio,
3731 &n_blocks_count, 1)) {
ac27a0ec
DK		 3732 err = -EINVAL;
3733 goto restore_opts;
3734 }
3735
4ab2f15b 3736 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
c67d859e 3737 ext4_abort(sb, "Abort forced by user");
ac27a0ec
DK		 3738
3739 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 3740 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec
DK		 3741
3742 es = sbi->s_es;
3743
b3881f74 3744 if (sbi->s_journal) {
0390131b 3745 ext4_init_journal_params(sb, sbi->s_journal);
b3881f74
TT		 3746 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3747 }
ac27a0ec
DK		 3748
3749 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
bd81d8ee 3750 n_blocks_count > ext4_blocks_count(es)) {
4ab2f15b 3751 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
ac27a0ec
DK		 3752 err = -EROFS;
3753 goto restore_opts;
3754 }
3755
3756 if (*flags & MS_RDONLY) {
0f0dd62f
CH		 3757 err = dquot_suspend(sb, -1);
3758 if (err < 0)
c79d967d 3759 goto restore_opts;
c79d967d 3760
ac27a0ec
DK		 3761 /*
3762 * First of all, the unconditional stuff we have to do
3763 * to disable replay of the journal when we next remount
3764 */
3765 sb->s_flags |= MS_RDONLY;
3766
3767 /*
3768 * OK, test if we are remounting a valid rw partition
3769 * readonly, and if so set the rdonly flag and then
3770 * mark the partition as valid again.
3771 */
617ba13b
MC		 3772 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
3773 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 3774 es->s_state = cpu_to_le16(sbi->s_mount_state);
3775
a63c9eb2 3776 if (sbi->s_journal)
0390131b 3777 ext4_mark_recovery_complete(sb, es);
ac27a0ec 3778 } else {
a13fb1a4
ES		 3779 /* Make sure we can mount this feature set readwrite */
3780 if (!ext4_feature_set_ok(sb, 0)) {
ac27a0ec
DK		 3781 err = -EROFS;
3782 goto restore_opts;
3783 }
8a266467
TT		 3784 /*
3785 * Make sure the group descriptor checksums
0b8e58a1 3786 * are sane. If they aren't, refuse to remount r/w.
8a266467
TT		 3787 */
3788 for (g = 0; g < sbi->s_groups_count; g++) {
3789 struct ext4_group_desc *gdp =
3790 ext4_get_group_desc(sb, g, NULL);
3791
3792 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
b31e1552
ES		 3793 ext4_msg(sb, KERN_ERR,
3794 "ext4_remount: Checksum for group %u failed (%u!=%u)",
8a266467
TT		 3795 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
3796 le16_to_cpu(gdp->bg_checksum));
3797 err = -EINVAL;
3798 goto restore_opts;
3799 }
3800 }
3801
ead6596b
ES		 3802 /*
3803 * If we have an unprocessed orphan list hanging
3804 * around from a previously readonly bdev mount,
3805 * require a full umount/remount for now.
3806 */
3807 if (es->s_last_orphan) {
b31e1552 3808 ext4_msg(sb, KERN_WARNING, "Couldn't "
ead6596b
ES		 3809 "remount RDWR because of unprocessed "
3810 "orphan inode list. Please "
b31e1552 3811 "umount/remount instead");
ead6596b
ES		 3812 err = -EINVAL;
3813 goto restore_opts;
3814 }
3815
ac27a0ec
DK		 3816 /*
3817 * Mounting a RDONLY partition read-write, so reread
3818 * and store the current valid flag. (It may have
3819 * been changed by e2fsck since we originally mounted
3820 * the partition.)
3821 */
0390131b
FM		 3822 if (sbi->s_journal)
3823 ext4_clear_journal_err(sb, es);
ac27a0ec 3824 sbi->s_mount_state = le16_to_cpu(es->s_state);
617ba13b 3825 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
ac27a0ec 3826 goto restore_opts;
2b2d6d01 3827 if (!ext4_setup_super(sb, es, 0))
ac27a0ec 3828 sb->s_flags &= ~MS_RDONLY;
c79d967d 3829 enable_quota = 1;
ac27a0ec
DK		 3830 }
3831 }
6fd058f7 3832 ext4_setup_system_zone(sb);
0390131b 3833 if (sbi->s_journal == NULL)
e2d67052 3834 ext4_commit_super(sb, 1);
0390131b 3835
ac27a0ec
DK		 3836#ifdef CONFIG_QUOTA
3837 /* Release old quota file names */
3838 for (i = 0; i < MAXQUOTAS; i++)
3839 if (old_opts.s_qf_names[i] &&
3840 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3841 kfree(old_opts.s_qf_names[i]);
3842#endif
bbd6851a 3843 unlock_super(sb);
337eb00a 3844 unlock_kernel();
c79d967d 3845 if (enable_quota)
0f0dd62f 3846 dquot_resume(sb, -1);
d4c402d9
CW		 3847
3848 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
3849 kfree(orig_data);
ac27a0ec 3850 return 0;
0b8e58a1 3851
ac27a0ec
DK		 3852restore_opts:
3853 sb->s_flags = old_sb_flags;
3854 sbi->s_mount_opt = old_opts.s_mount_opt;
3855 sbi->s_resuid = old_opts.s_resuid;
3856 sbi->s_resgid = old_opts.s_resgid;
3857 sbi->s_commit_interval = old_opts.s_commit_interval;
30773840
TT		 3858 sbi->s_min_batch_time = old_opts.s_min_batch_time;
3859 sbi->s_max_batch_time = old_opts.s_max_batch_time;
ac27a0ec
DK		 3860#ifdef CONFIG_QUOTA
3861 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
3862 for (i = 0; i < MAXQUOTAS; i++) {
3863 if (sbi->s_qf_names[i] &&
3864 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3865 kfree(sbi->s_qf_names[i]);
3866 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
3867 }
3868#endif
bbd6851a 3869 unlock_super(sb);
337eb00a 3870 unlock_kernel();
d4c402d9 3871 kfree(orig_data);
ac27a0ec
DK		 3872 return err;
3873}
3874
2b2d6d01 3875static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
ac27a0ec
DK		 3876{
3877 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 3878 struct ext4_sb_info *sbi = EXT4_SB(sb);
3879 struct ext4_super_block *es = sbi->s_es;
960cc398 3880 u64 fsid;
ac27a0ec 3881
5e70030d
BP		 3882 if (test_opt(sb, MINIX_DF)) {
3883 sbi->s_overhead_last = 0;
6bc9feff 3884 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
8df9675f 3885 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
5e70030d 3886 ext4_fsblk_t overhead = 0;
ac27a0ec
DK		 3887
3888 /*
5e70030d
BP		 3889 * Compute the overhead (FS structures). This is constant
3890 * for a given filesystem unless the number of block groups
3891 * changes so we cache the previous value until it does.
ac27a0ec
DK		 3892 */
3893
3894 /*
3895 * All of the blocks before first_data_block are
3896 * overhead
3897 */
3898 overhead = le32_to_cpu(es->s_first_data_block);
3899
3900 /*
3901 * Add the overhead attributed to the superblock and
3902 * block group descriptors. If the sparse superblocks
3903 * feature is turned on, then not all groups have this.
3904 */
3905 for (i = 0; i < ngroups; i++) {
617ba13b
MC		 3906 overhead += ext4_bg_has_super(sb, i) +
3907 ext4_bg_num_gdb(sb, i);
ac27a0ec
DK		 3908 cond_resched();
3909 }
3910
3911 /*
3912 * Every block group has an inode bitmap, a block
3913 * bitmap, and an inode table.
3914 */
5e70030d
BP		 3915 overhead += ngroups * (2 + sbi->s_itb_per_group);
3916 sbi->s_overhead_last = overhead;
3917 smp_wmb();
6bc9feff 3918 sbi->s_blocks_last = ext4_blocks_count(es);
ac27a0ec
DK		 3919 }
3920
617ba13b 3921 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 3922 buf->f_bsize = sb->s_blocksize;
5e70030d 3923 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
6bc6e63f
AK		 3924 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
3925 percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
bd81d8ee
LV		 3926 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3927 if (buf->f_bfree < ext4_r_blocks_count(es))
ac27a0ec
DK		 3928 buf->f_bavail = 0;
3929 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 3930 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
617ba13b 3931 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 3932 fsid = le64_to_cpup((void *)es->s_uuid) ^
3933 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3934 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3935 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
0b8e58a1 3936
ac27a0ec
DK		 3937 return 0;
3938}
3939
0b8e58a1
AD		 3940/* Helper function for writing quotas on sync - we need to start transaction
3941 * before quota file is locked for write. Otherwise the are possible deadlocks:
ac27a0ec 3942 * Process 1 Process 2
617ba13b 3943 * ext4_create() quota_sync()
a269eb18 3944 * jbd2_journal_start() write_dquot()
871a2931 3945 * dquot_initialize() down(dqio_mutex)
dab291af 3946 * down(dqio_mutex) jbd2_journal_start()
ac27a0ec
DK		 3947 *
3948 */
3949
3950#ifdef CONFIG_QUOTA
3951
3952static inline struct inode *dquot_to_inode(struct dquot *dquot)
3953{
3954 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3955}
3956
617ba13b 3957static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 3958{
3959 int ret, err;
3960 handle_t *handle;
3961 struct inode *inode;
3962
3963 inode = dquot_to_inode(dquot);
617ba13b 3964 handle = ext4_journal_start(inode,
0b8e58a1 3965 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 3966 if (IS_ERR(handle))
3967 return PTR_ERR(handle);
3968 ret = dquot_commit(dquot);
617ba13b 3969 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3970 if (!ret)
3971 ret = err;
3972 return ret;
3973}
3974
617ba13b 3975static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 3976{
3977 int ret, err;
3978 handle_t *handle;
3979
617ba13b 3980 handle = ext4_journal_start(dquot_to_inode(dquot),
0b8e58a1 3981 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 3982 if (IS_ERR(handle))
3983 return PTR_ERR(handle);
3984 ret = dquot_acquire(dquot);
617ba13b 3985 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3986 if (!ret)
3987 ret = err;
3988 return ret;
3989}
3990
617ba13b 3991static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 3992{
3993 int ret, err;
3994 handle_t *handle;
3995
617ba13b 3996 handle = ext4_journal_start(dquot_to_inode(dquot),
0b8e58a1 3997 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 3998 if (IS_ERR(handle)) {
3999 /* Release dquot anyway to avoid endless cycle in dqput() */
4000 dquot_release(dquot);
ac27a0ec 4001 return PTR_ERR(handle);
9c3013e9 4002 }
ac27a0ec 4003 ret = dquot_release(dquot);
617ba13b 4004 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4005 if (!ret)
4006 ret = err;
4007 return ret;
4008}
4009
617ba13b 4010static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec 4011{
2c8be6b2 4012 /* Are we journaling quotas? */
617ba13b
MC		 4013 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
4014 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
ac27a0ec 4015 dquot_mark_dquot_dirty(dquot);
617ba13b 4016 return ext4_write_dquot(dquot);
ac27a0ec
DK		 4017 } else {
4018 return dquot_mark_dquot_dirty(dquot);
4019 }
4020}
4021
617ba13b 4022static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 4023{
4024 int ret, err;
4025 handle_t *handle;
4026
4027 /* Data block + inode block */
617ba13b 4028 handle = ext4_journal_start(sb->s_root->d_inode, 2);
ac27a0ec
DK		 4029 if (IS_ERR(handle))
4030 return PTR_ERR(handle);
4031 ret = dquot_commit_info(sb, type);
617ba13b 4032 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4033 if (!ret)
4034 ret = err;
4035 return ret;
4036}
4037
4038/*
4039 * Turn on quotas during mount time - we need to find
4040 * the quota file and such...
4041 */
617ba13b 4042static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 4043{
287a8095
CH		 4044 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4045 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 4046}
4047
4048/*
4049 * Standard function to be called on quota_on
4050 */
617ba13b 4051static int ext4_quota_on(struct super_block *sb, int type, int format_id,
307ae18a 4052 char *name)
ac27a0ec
DK		 4053{
4054 int err;
8264613d 4055 struct path path;
ac27a0ec
DK		 4056
4057 if (!test_opt(sb, QUOTA))
4058 return -EINVAL;
0623543b 4059
8264613d 4060 err = kern_path(name, LOOKUP_FOLLOW, &path);
ac27a0ec
DK		 4061 if (err)
4062 return err;
0623543b 4063
ac27a0ec 4064 /* Quotafile not on the same filesystem? */
8264613d
AV		 4065 if (path.mnt->mnt_sb != sb) {
4066 path_put(&path);
ac27a0ec
DK		 4067 return -EXDEV;
4068 }
0623543b
JK		 4069 /* Journaling quota? */
4070 if (EXT4_SB(sb)->s_qf_names[type]) {
2b2d6d01 4071 /* Quotafile not in fs root? */
8264613d 4072 if (path.dentry->d_parent != sb->s_root)
b31e1552
ES		 4073 ext4_msg(sb, KERN_WARNING,
4074 "Quota file not on filesystem root. "
4075 "Journaled quota will not work");
2b2d6d01 4076 }
0623543b
JK		 4077
4078 /*
4079 * When we journal data on quota file, we have to flush journal to see
4080 * all updates to the file when we bypass pagecache...
4081 */
0390131b
FM		 4082 if (EXT4_SB(sb)->s_journal &&
4083 ext4_should_journal_data(path.dentry->d_inode)) {
0623543b
JK		 4084 /*
4085 * We don't need to lock updates but journal_flush() could
4086 * otherwise be livelocked...
4087 */
4088 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 4089 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
0623543b 4090 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 4091 if (err) {
8264613d 4092 path_put(&path);
7ffe1ea8
HK		 4093 return err;
4094 }
0623543b
JK		 4095 }
4096
287a8095 4097 err = dquot_quota_on_path(sb, type, format_id, &path);
8264613d 4098 path_put(&path);
77e69dac 4099 return err;
ac27a0ec
DK		 4100}
4101
ca0e05e4
DM		 4102static int ext4_quota_off(struct super_block *sb, int type)
4103{
4104 /* Force all delayed allocation blocks to be allocated */
4105 if (test_opt(sb, DELALLOC)) {
4106 down_read(&sb->s_umount);
4107 sync_filesystem(sb);
4108 up_read(&sb->s_umount);
4109 }
4110
4111 return dquot_quota_off(sb, type);
4112}
4113
ac27a0ec
DK		 4114/* Read data from quotafile - avoid pagecache and such because we cannot afford
4115 * acquiring the locks... As quota files are never truncated and quota code
4116 * itself serializes the operations (and noone else should touch the files)
4117 * we don't have to be afraid of races */
617ba13b 4118static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 4119 size_t len, loff_t off)
4120{
4121 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 4122 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 4123 int err = 0;
4124 int offset = off & (sb->s_blocksize - 1);
4125 int tocopy;
4126 size_t toread;
4127 struct buffer_head *bh;
4128 loff_t i_size = i_size_read(inode);
4129
4130 if (off > i_size)
4131 return 0;
4132 if (off+len > i_size)
4133 len = i_size-off;
4134 toread = len;
4135 while (toread > 0) {
4136 tocopy = sb->s_blocksize - offset < toread ?
4137 sb->s_blocksize - offset : toread;
617ba13b 4138 bh = ext4_bread(NULL, inode, blk, 0, &err);
ac27a0ec
DK		 4139 if (err)
4140 return err;
4141 if (!bh) /* A hole? */
4142 memset(data, 0, tocopy);
4143 else
4144 memcpy(data, bh->b_data+offset, tocopy);
4145 brelse(bh);
4146 offset = 0;
4147 toread -= tocopy;
4148 data += tocopy;
4149 blk++;
4150 }
4151 return len;
4152}
4153
4154/* Write to quotafile (we know the transaction is already started and has
4155 * enough credits) */
617ba13b 4156static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 4157 const char *data, size_t len, loff_t off)
4158{
4159 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 4160 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 4161 int err = 0;
4162 int offset = off & (sb->s_blocksize - 1);
ac27a0ec
DK		 4163 struct buffer_head *bh;
4164 handle_t *handle = journal_current_handle();
4165
0390131b 4166 if (EXT4_SB(sb)->s_journal && !handle) {
b31e1552
ES		 4167 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4168 " cancelled because transaction is not started",
9c3013e9
JK		 4169 (unsigned long long)off, (unsigned long long)len);
4170 return -EIO;
4171 }
67eeb568
DM		 4172 /*
4173 * Since we account only one data block in transaction credits,
4174 * then it is impossible to cross a block boundary.
4175 */
4176 if (sb->s_blocksize - offset < len) {
4177 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4178 " cancelled because not block aligned",
4179 (unsigned long long)off, (unsigned long long)len);
4180 return -EIO;
4181 }
4182
ac27a0ec 4183 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
67eeb568
DM		 4184 bh = ext4_bread(handle, inode, blk, 1, &err);
4185 if (!bh)
4186 goto out;
62d2b5f2
JK		 4187 err = ext4_journal_get_write_access(handle, bh);
4188 if (err) {
4189 brelse(bh);
4190 goto out;
ac27a0ec 4191 }
67eeb568
DM		 4192 lock_buffer(bh);
4193 memcpy(bh->b_data+offset, data, len);
4194 flush_dcache_page(bh->b_page);
4195 unlock_buffer(bh);
62d2b5f2 4196 err = ext4_handle_dirty_metadata(handle, NULL, bh);
67eeb568 4197 brelse(bh);
ac27a0ec 4198out:
67eeb568 4199 if (err) {
4d04e4fb 4200 mutex_unlock(&inode->i_mutex);
ac27a0ec 4201 return err;
4d04e4fb 4202 }
67eeb568
DM		 4203 if (inode->i_size < off + len) {
4204 i_size_write(inode, off + len);
617ba13b 4205 EXT4_I(inode)->i_disksize = inode->i_size;
ac27a0ec 4206 }
ac27a0ec 4207 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
617ba13b 4208 ext4_mark_inode_dirty(handle, inode);
ac27a0ec 4209 mutex_unlock(&inode->i_mutex);
67eeb568 4210 return len;
ac27a0ec
DK		 4211}
4212
4213#endif
4214
0b8e58a1
AD		 4215static int ext4_get_sb(struct file_system_type *fs_type, int flags,
4216 const char *dev_name, void *data, struct vfsmount *mnt)
ac27a0ec 4217{
0b8e58a1 4218 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super,mnt);
ac27a0ec
DK		 4219}
4220
37f328eb 4221#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
24b58424
TT		 4222static struct file_system_type ext2_fs_type = {
4223 .owner = THIS_MODULE,
4224 .name = "ext2",
4225 .get_sb = ext4_get_sb,
4226 .kill_sb = kill_block_super,
4227 .fs_flags = FS_REQUIRES_DEV,
4228};
4229
4230static inline void register_as_ext2(void)
4231{
4232 int err = register_filesystem(&ext2_fs_type);
4233 if (err)
4234 printk(KERN_WARNING
4235 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
4236}
4237
4238static inline void unregister_as_ext2(void)
4239{
4240 unregister_filesystem(&ext2_fs_type);
4241}
51b7e3c9 4242MODULE_ALIAS("ext2");
24b58424
TT		 4243#else
4244static inline void register_as_ext2(void) { }
4245static inline void unregister_as_ext2(void) { }
4246#endif
4247
37f328eb 4248#if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
24b58424
TT		 4249static inline void register_as_ext3(void)
4250{
4251 int err = register_filesystem(&ext3_fs_type);
4252 if (err)
4253 printk(KERN_WARNING
4254 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
4255}
4256
4257static inline void unregister_as_ext3(void)
4258{
4259 unregister_filesystem(&ext3_fs_type);
4260}
51b7e3c9 4261MODULE_ALIAS("ext3");
24b58424
TT		 4262#else
4263static inline void register_as_ext3(void) { }
4264static inline void unregister_as_ext3(void) { }
4265#endif
4266
03010a33
TT		 4267static struct file_system_type ext4_fs_type = {
4268 .owner = THIS_MODULE,
4269 .name = "ext4",
4270 .get_sb = ext4_get_sb,
4271 .kill_sb = kill_block_super,
4272 .fs_flags = FS_REQUIRES_DEV,
4273};
4274
617ba13b 4275static int __init init_ext4_fs(void)
ac27a0ec 4276{
c9de560d
AT		 4277 int err;
4278
12e9b892 4279 ext4_check_flag_values();
6fd058f7
TT		 4280 err = init_ext4_system_zone();
4281 if (err)
4282 return err;
3197ebdb
TT		 4283 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
4284 if (!ext4_kset)
6fd058f7 4285 goto out4;
9f6200bb 4286 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
c9de560d 4287 err = init_ext4_mballoc();
ac27a0ec 4288 if (err)
6fd058f7 4289 goto out3;
c9de560d
AT		 4290
4291 err = init_ext4_xattr();
4292 if (err)
4293 goto out2;
ac27a0ec
DK		 4294 err = init_inodecache();
4295 if (err)
4296 goto out1;
24b58424
TT		 4297 register_as_ext2();
4298 register_as_ext3();
03010a33 4299 err = register_filesystem(&ext4_fs_type);
ac27a0ec
DK		 4300 if (err)
4301 goto out;
4302 return 0;
4303out:
24b58424
TT		 4304 unregister_as_ext2();
4305 unregister_as_ext3();
ac27a0ec
DK		 4306 destroy_inodecache();
4307out1:
617ba13b 4308 exit_ext4_xattr();
c9de560d
AT		 4309out2:
4310 exit_ext4_mballoc();
6fd058f7
TT		 4311out3:
4312 remove_proc_entry("fs/ext4", NULL);
4313 kset_unregister(ext4_kset);
4314out4:
4315 exit_ext4_system_zone();
ac27a0ec
DK		 4316 return err;
4317}
4318
617ba13b 4319static void __exit exit_ext4_fs(void)
ac27a0ec 4320{
24b58424
TT		 4321 unregister_as_ext2();
4322 unregister_as_ext3();
03010a33 4323 unregister_filesystem(&ext4_fs_type);
ac27a0ec 4324 destroy_inodecache();
617ba13b 4325 exit_ext4_xattr();
c9de560d 4326 exit_ext4_mballoc();
9f6200bb 4327 remove_proc_entry("fs/ext4", NULL);
3197ebdb 4328 kset_unregister(ext4_kset);
6fd058f7 4329 exit_ext4_system_zone();
ac27a0ec
DK		 4330}
4331
4332MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
83982b6f 4333MODULE_DESCRIPTION("Fourth Extended Filesystem");
ac27a0ec 4334MODULE_LICENSE("GPL");
617ba13b
MC		 4335module_init(init_ext4_fs)
4336module_exit(exit_ext4_fs)