]> bbs.cooldavid.org Git - net-next-2.6.git/blame - fs/ext4/super.c
git://bbs.cooldavid.org / net-next-2.6.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
ext4: Don't error out the fs if the user tries to make a file too big
[net-next-2.6.git] / fs / ext4 / super.c
CommitLineData
ac27a0ec 1/*
617ba13b 2 * linux/fs/ext4/super.c
ac27a0ec
DK		 3 *
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
8 *
9 * from
10 *
11 * linux/fs/minix/inode.c
12 *
13 * Copyright (C) 1991, 1992 Linus Torvalds
14 *
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
17 */
18
19#include <linux/module.h>
20#include <linux/string.h>
21#include <linux/fs.h>
22#include <linux/time.h>
c5ca7c76 23#include <linux/vmalloc.h>
dab291af 24#include <linux/jbd2.h>
ac27a0ec
DK		 25#include <linux/slab.h>
26#include <linux/init.h>
27#include <linux/blkdev.h>
28#include <linux/parser.h>
29#include <linux/smp_lock.h>
30#include <linux/buffer_head.h>
a5694255 31#include <linux/exportfs.h>
ac27a0ec
DK		 32#include <linux/vfs.h>
33#include <linux/random.h>
34#include <linux/mount.h>
35#include <linux/namei.h>
36#include <linux/quotaops.h>
37#include <linux/seq_file.h>
9f6200bb 38#include <linux/proc_fs.h>
3197ebdb 39#include <linux/ctype.h>
1330593e 40#include <linux/log2.h>
717d50e4 41#include <linux/crc16.h>
ac27a0ec
DK		 42#include <asm/uaccess.h>
43
3dcf5451
CH		 44#include "ext4.h"
45#include "ext4_jbd2.h"
ac27a0ec
DK		 46#include "xattr.h"
47#include "acl.h"
3661d286 48#include "mballoc.h"
ac27a0ec 49
9bffad1e
TT		 50#define CREATE_TRACE_POINTS
51#include <trace/events/ext4.h>
52
9f6200bb 53struct proc_dir_entry *ext4_proc_root;
3197ebdb 54static struct kset *ext4_kset;
9f6200bb 55
617ba13b 56static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
ac27a0ec 57 unsigned long journal_devnum);
e2d67052 58static int ext4_commit_super(struct super_block *sb, int sync);
2b2d6d01
TT		 59static void ext4_mark_recovery_complete(struct super_block *sb,
60 struct ext4_super_block *es);
61static void ext4_clear_journal_err(struct super_block *sb,
62 struct ext4_super_block *es);
617ba13b 63static int ext4_sync_fs(struct super_block *sb, int wait);
2b2d6d01 64static const char *ext4_decode_error(struct super_block *sb, int errno,
ac27a0ec 65 char nbuf[16]);
2b2d6d01
TT		 66static int ext4_remount(struct super_block *sb, int *flags, char *data);
67static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
c4be0c1d 68static int ext4_unfreeze(struct super_block *sb);
2b2d6d01 69static void ext4_write_super(struct super_block *sb);
c4be0c1d 70static int ext4_freeze(struct super_block *sb);
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
6b0310fb 244 vfs_check_frozen(sb, SB_FREEZE_WRITE);
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);
617ba13b
MC		 1121static int ext4_quota_on_mount(struct super_block *sb, int type);
1122static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 1123 size_t len, loff_t off);
617ba13b 1124static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 1125 const char *data, size_t len, loff_t off);
1126
61e225dc 1127static const struct dquot_operations ext4_quota_operations = {
a9e7f447 1128#ifdef CONFIG_QUOTA
60e58e0f 1129 .get_reserved_space = ext4_get_reserved_space,
a9e7f447 1130#endif
617ba13b
MC		 1131 .write_dquot = ext4_write_dquot,
1132 .acquire_dquot = ext4_acquire_dquot,
1133 .release_dquot = ext4_release_dquot,
1134 .mark_dirty = ext4_mark_dquot_dirty,
a5b5ee32
JK		 1135 .write_info = ext4_write_info,
1136 .alloc_dquot = dquot_alloc,
1137 .destroy_dquot = dquot_destroy,
ac27a0ec
DK		 1138};
1139
0d54b217 1140static const struct quotactl_ops ext4_qctl_operations = {
617ba13b 1141 .quota_on = ext4_quota_on,
287a8095
CH		 1142 .quota_off = dquot_quota_off,
1143 .quota_sync = dquot_quota_sync,
1144 .get_info = dquot_get_dqinfo,
1145 .set_info = dquot_set_dqinfo,
1146 .get_dqblk = dquot_get_dqblk,
1147 .set_dqblk = dquot_set_dqblk
ac27a0ec
DK		 1148};
1149#endif
1150
ee9b6d61 1151static const struct super_operations ext4_sops = {
617ba13b
MC		 1152 .alloc_inode = ext4_alloc_inode,
1153 .destroy_inode = ext4_destroy_inode,
617ba13b
MC		 1154 .write_inode = ext4_write_inode,
1155 .dirty_inode = ext4_dirty_inode,
1156 .delete_inode = ext4_delete_inode,
1157 .put_super = ext4_put_super,
617ba13b 1158 .sync_fs = ext4_sync_fs,
c4be0c1d
TS		 1159 .freeze_fs = ext4_freeze,
1160 .unfreeze_fs = ext4_unfreeze,
617ba13b
MC		 1161 .statfs = ext4_statfs,
1162 .remount_fs = ext4_remount,
1163 .clear_inode = ext4_clear_inode,
1164 .show_options = ext4_show_options,
ac27a0ec 1165#ifdef CONFIG_QUOTA
617ba13b
MC		 1166 .quota_read = ext4_quota_read,
1167 .quota_write = ext4_quota_write,
ac27a0ec 1168#endif
c39a7f84 1169 .bdev_try_to_free_page = bdev_try_to_free_page,
ac27a0ec
DK		 1170};
1171
9ca92389
TT		 1172static const struct super_operations ext4_nojournal_sops = {
1173 .alloc_inode = ext4_alloc_inode,
1174 .destroy_inode = ext4_destroy_inode,
1175 .write_inode = ext4_write_inode,
1176 .dirty_inode = ext4_dirty_inode,
1177 .delete_inode = ext4_delete_inode,
1178 .write_super = ext4_write_super,
1179 .put_super = ext4_put_super,
1180 .statfs = ext4_statfs,
1181 .remount_fs = ext4_remount,
1182 .clear_inode = ext4_clear_inode,
1183 .show_options = ext4_show_options,
1184#ifdef CONFIG_QUOTA
1185 .quota_read = ext4_quota_read,
1186 .quota_write = ext4_quota_write,
1187#endif
1188 .bdev_try_to_free_page = bdev_try_to_free_page,
1189};
1190
39655164 1191static const struct export_operations ext4_export_ops = {
1b961ac0
CH		 1192 .fh_to_dentry = ext4_fh_to_dentry,
1193 .fh_to_parent = ext4_fh_to_parent,
617ba13b 1194 .get_parent = ext4_get_parent,
ac27a0ec
DK		 1195};
1196
1197enum {
1198 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1199 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
01436ef2 1200 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
ac27a0ec 1201 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
06705bff 1202 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh,
30773840 1203 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
c3191067 1204 Opt_journal_update, Opt_journal_dev,
818d276c 1205 Opt_journal_checksum, Opt_journal_async_commit,
ac27a0ec 1206 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
296c355c 1207 Opt_data_err_abort, Opt_data_err_ignore,
ac27a0ec 1208 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
5a20bdfc
JK		 1209 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1210 Opt_noquota, Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err,
1211 Opt_resize, Opt_usrquota, Opt_grpquota, Opt_i_version,
01436ef2 1212 Opt_stripe, Opt_delalloc, Opt_nodelalloc,
6fd058f7 1213 Opt_block_validity, Opt_noblock_validity,
5328e635 1214 Opt_inode_readahead_blks, Opt_journal_ioprio,
744692dc 1215 Opt_dioread_nolock, Opt_dioread_lock,
5328e635 1216 Opt_discard, Opt_nodiscard,
ac27a0ec
DK		 1217};
1218
a447c093 1219static const match_table_t tokens = {
ac27a0ec
DK		 1220 {Opt_bsd_df, "bsddf"},
1221 {Opt_minix_df, "minixdf"},
1222 {Opt_grpid, "grpid"},
1223 {Opt_grpid, "bsdgroups"},
1224 {Opt_nogrpid, "nogrpid"},
1225 {Opt_nogrpid, "sysvgroups"},
1226 {Opt_resgid, "resgid=%u"},
1227 {Opt_resuid, "resuid=%u"},
1228 {Opt_sb, "sb=%u"},
1229 {Opt_err_cont, "errors=continue"},
1230 {Opt_err_panic, "errors=panic"},
1231 {Opt_err_ro, "errors=remount-ro"},
1232 {Opt_nouid32, "nouid32"},
ac27a0ec
DK		 1233 {Opt_debug, "debug"},
1234 {Opt_oldalloc, "oldalloc"},
1235 {Opt_orlov, "orlov"},
1236 {Opt_user_xattr, "user_xattr"},
1237 {Opt_nouser_xattr, "nouser_xattr"},
1238 {Opt_acl, "acl"},
1239 {Opt_noacl, "noacl"},
ac27a0ec 1240 {Opt_noload, "noload"},
e3bb52ae 1241 {Opt_noload, "norecovery"},
ac27a0ec
DK		 1242 {Opt_nobh, "nobh"},
1243 {Opt_bh, "bh"},
1244 {Opt_commit, "commit=%u"},
30773840
TT		 1245 {Opt_min_batch_time, "min_batch_time=%u"},
1246 {Opt_max_batch_time, "max_batch_time=%u"},
ac27a0ec 1247 {Opt_journal_update, "journal=update"},
ac27a0ec 1248 {Opt_journal_dev, "journal_dev=%u"},
818d276c
GS		 1249 {Opt_journal_checksum, "journal_checksum"},
1250 {Opt_journal_async_commit, "journal_async_commit"},
ac27a0ec
DK		 1251 {Opt_abort, "abort"},
1252 {Opt_data_journal, "data=journal"},
1253 {Opt_data_ordered, "data=ordered"},
1254 {Opt_data_writeback, "data=writeback"},
5bf5683a
HK		 1255 {Opt_data_err_abort, "data_err=abort"},
1256 {Opt_data_err_ignore, "data_err=ignore"},
ac27a0ec
DK		 1257 {Opt_offusrjquota, "usrjquota="},
1258 {Opt_usrjquota, "usrjquota=%s"},
1259 {Opt_offgrpjquota, "grpjquota="},
1260 {Opt_grpjquota, "grpjquota=%s"},
1261 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1262 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
5a20bdfc 1263 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
ac27a0ec
DK		 1264 {Opt_grpquota, "grpquota"},
1265 {Opt_noquota, "noquota"},
1266 {Opt_quota, "quota"},
1267 {Opt_usrquota, "usrquota"},
1268 {Opt_barrier, "barrier=%u"},
06705bff
TT		 1269 {Opt_barrier, "barrier"},
1270 {Opt_nobarrier, "nobarrier"},
25ec56b5 1271 {Opt_i_version, "i_version"},
c9de560d 1272 {Opt_stripe, "stripe=%u"},
ac27a0ec 1273 {Opt_resize, "resize"},
64769240 1274 {Opt_delalloc, "delalloc"},
dd919b98 1275 {Opt_nodelalloc, "nodelalloc"},
6fd058f7
TT		 1276 {Opt_block_validity, "block_validity"},
1277 {Opt_noblock_validity, "noblock_validity"},
240799cd 1278 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
b3881f74 1279 {Opt_journal_ioprio, "journal_ioprio=%u"},
afd4672d 1280 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
06705bff
TT		 1281 {Opt_auto_da_alloc, "auto_da_alloc"},
1282 {Opt_noauto_da_alloc, "noauto_da_alloc"},
744692dc
JZ		 1283 {Opt_dioread_nolock, "dioread_nolock"},
1284 {Opt_dioread_lock, "dioread_lock"},
5328e635
ES		 1285 {Opt_discard, "discard"},
1286 {Opt_nodiscard, "nodiscard"},
f3f12faa 1287 {Opt_err, NULL},
ac27a0ec
DK		 1288};
1289
617ba13b 1290static ext4_fsblk_t get_sb_block(void **data)
ac27a0ec 1291{
617ba13b 1292 ext4_fsblk_t sb_block;
ac27a0ec
DK		 1293 char *options = (char *) *data;
1294
1295 if (!options || strncmp(options, "sb=", 3) != 0)
1296 return 1; /* Default location */
0b8e58a1 1297
ac27a0ec 1298 options += 3;
0b8e58a1 1299 /* TODO: use simple_strtoll with >32bit ext4 */
ac27a0ec
DK		 1300 sb_block = simple_strtoul(options, &options, 0);
1301 if (*options && *options != ',') {
4776004f 1302 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
ac27a0ec
DK		 1303 (char *) *data);
1304 return 1;
1305 }
1306 if (*options == ',')
1307 options++;
1308 *data = (void *) options;
0b8e58a1 1309
ac27a0ec
DK		 1310 return sb_block;
1311}
1312
b3881f74 1313#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
437ca0fd
DM		 1314static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1315 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
b3881f74 1316
56c50f11
DM		 1317#ifdef CONFIG_QUOTA
1318static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1319{
1320 struct ext4_sb_info *sbi = EXT4_SB(sb);
1321 char *qname;
1322
1323 if (sb_any_quota_loaded(sb) &&
1324 !sbi->s_qf_names[qtype]) {
1325 ext4_msg(sb, KERN_ERR,
1326 "Cannot change journaled "
1327 "quota options when quota turned on");
1328 return 0;
1329 }
1330 qname = match_strdup(args);
1331 if (!qname) {
1332 ext4_msg(sb, KERN_ERR,
1333 "Not enough memory for storing quotafile name");
1334 return 0;
1335 }
1336 if (sbi->s_qf_names[qtype] &&
1337 strcmp(sbi->s_qf_names[qtype], qname)) {
1338 ext4_msg(sb, KERN_ERR,
1339 "%s quota file already specified", QTYPE2NAME(qtype));
1340 kfree(qname);
1341 return 0;
1342 }
1343 sbi->s_qf_names[qtype] = qname;
1344 if (strchr(sbi->s_qf_names[qtype], '/')) {
1345 ext4_msg(sb, KERN_ERR,
1346 "quotafile must be on filesystem root");
1347 kfree(sbi->s_qf_names[qtype]);
1348 sbi->s_qf_names[qtype] = NULL;
1349 return 0;
1350 }
1351 set_opt(sbi->s_mount_opt, QUOTA);
1352 return 1;
1353}
1354
1355static int clear_qf_name(struct super_block *sb, int qtype)
1356{
1357
1358 struct ext4_sb_info *sbi = EXT4_SB(sb);
1359
1360 if (sb_any_quota_loaded(sb) &&
1361 sbi->s_qf_names[qtype]) {
1362 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1363 " when quota turned on");
1364 return 0;
1365 }
1366 /*
1367 * The space will be released later when all options are confirmed
1368 * to be correct
1369 */
1370 sbi->s_qf_names[qtype] = NULL;
1371 return 1;
1372}
1373#endif
1374
2b2d6d01 1375static int parse_options(char *options, struct super_block *sb,
c3191067 1376 unsigned long *journal_devnum,
b3881f74 1377 unsigned int *journal_ioprio,
2b2d6d01 1378 ext4_fsblk_t *n_blocks_count, int is_remount)
ac27a0ec 1379{
617ba13b 1380 struct ext4_sb_info *sbi = EXT4_SB(sb);
2b2d6d01 1381 char *p;
ac27a0ec
DK		 1382 substring_t args[MAX_OPT_ARGS];
1383 int data_opt = 0;
1384 int option;
1385#ifdef CONFIG_QUOTA
56c50f11 1386 int qfmt;
ac27a0ec
DK		 1387#endif
1388
1389 if (!options)
1390 return 1;
1391
2b2d6d01 1392 while ((p = strsep(&options, ",")) != NULL) {
ac27a0ec
DK		 1393 int token;
1394 if (!*p)
1395 continue;
1396
15121c18
ES		 1397 /*
1398 * Initialize args struct so we know whether arg was
1399 * found; some options take optional arguments.
1400 */
1401 args[0].to = args[0].from = 0;
ac27a0ec
DK		 1402 token = match_token(p, tokens, args);
1403 switch (token) {
1404 case Opt_bsd_df:
437ca0fd 1405 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
2b2d6d01 1406 clear_opt(sbi->s_mount_opt, MINIX_DF);
ac27a0ec
DK		 1407 break;
1408 case Opt_minix_df:
437ca0fd 1409 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
2b2d6d01 1410 set_opt(sbi->s_mount_opt, MINIX_DF);
437ca0fd 1411
ac27a0ec
DK		 1412 break;
1413 case Opt_grpid:
437ca0fd 1414 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
2b2d6d01 1415 set_opt(sbi->s_mount_opt, GRPID);
437ca0fd 1416
ac27a0ec
DK		 1417 break;
1418 case Opt_nogrpid:
437ca0fd 1419 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
2b2d6d01 1420 clear_opt(sbi->s_mount_opt, GRPID);
437ca0fd 1421
ac27a0ec
DK		 1422 break;
1423 case Opt_resuid:
1424 if (match_int(&args[0], &option))
1425 return 0;
1426 sbi->s_resuid = option;
1427 break;
1428 case Opt_resgid:
1429 if (match_int(&args[0], &option))
1430 return 0;
1431 sbi->s_resgid = option;
1432 break;
1433 case Opt_sb:
1434 /* handled by get_sb_block() instead of here */
1435 /* *sb_block = match_int(&args[0]); */
1436 break;
1437 case Opt_err_panic:
2b2d6d01
TT		 1438 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1439 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1440 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
ac27a0ec
DK		 1441 break;
1442 case Opt_err_ro:
2b2d6d01
TT		 1443 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1444 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1445 set_opt(sbi->s_mount_opt, ERRORS_RO);
ac27a0ec
DK		 1446 break;
1447 case Opt_err_cont:
2b2d6d01
TT		 1448 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1449 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1450 set_opt(sbi->s_mount_opt, ERRORS_CONT);
ac27a0ec
DK		 1451 break;
1452 case Opt_nouid32:
2b2d6d01 1453 set_opt(sbi->s_mount_opt, NO_UID32);
ac27a0ec 1454 break;
ac27a0ec 1455 case Opt_debug:
2b2d6d01 1456 set_opt(sbi->s_mount_opt, DEBUG);
ac27a0ec
DK		 1457 break;
1458 case Opt_oldalloc:
2b2d6d01 1459 set_opt(sbi->s_mount_opt, OLDALLOC);
ac27a0ec
DK		 1460 break;
1461 case Opt_orlov:
2b2d6d01 1462 clear_opt(sbi->s_mount_opt, OLDALLOC);
ac27a0ec 1463 break;
03010a33 1464#ifdef CONFIG_EXT4_FS_XATTR
ac27a0ec 1465 case Opt_user_xattr:
2b2d6d01 1466 set_opt(sbi->s_mount_opt, XATTR_USER);
ac27a0ec
DK		 1467 break;
1468 case Opt_nouser_xattr:
2b2d6d01 1469 clear_opt(sbi->s_mount_opt, XATTR_USER);
ac27a0ec
DK		 1470 break;
1471#else
1472 case Opt_user_xattr:
1473 case Opt_nouser_xattr:
b31e1552 1474 ext4_msg(sb, KERN_ERR, "(no)user_xattr options not supported");
ac27a0ec
DK		 1475 break;
1476#endif
03010a33 1477#ifdef CONFIG_EXT4_FS_POSIX_ACL
ac27a0ec
DK		 1478 case Opt_acl:
1479 set_opt(sbi->s_mount_opt, POSIX_ACL);
1480 break;
1481 case Opt_noacl:
1482 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1483 break;
1484#else
1485 case Opt_acl:
1486 case Opt_noacl:
b31e1552 1487 ext4_msg(sb, KERN_ERR, "(no)acl options not supported");
ac27a0ec
DK		 1488 break;
1489#endif
ac27a0ec
DK		 1490 case Opt_journal_update:
1491 /* @@@ FIXME */
1492 /* Eventually we will want to be able to create
1493 a journal file here. For now, only allow the
1494 user to specify an existing inode to be the
1495 journal file. */
1496 if (is_remount) {
b31e1552
ES		 1497 ext4_msg(sb, KERN_ERR,
1498 "Cannot specify journal on remount");
ac27a0ec
DK		 1499 return 0;
1500 }
2b2d6d01 1501 set_opt(sbi->s_mount_opt, UPDATE_JOURNAL);
ac27a0ec 1502 break;
ac27a0ec
DK		 1503 case Opt_journal_dev:
1504 if (is_remount) {
b31e1552
ES		 1505 ext4_msg(sb, KERN_ERR,
1506 "Cannot specify journal on remount");
ac27a0ec
DK		 1507 return 0;
1508 }
1509 if (match_int(&args[0], &option))
1510 return 0;
1511 *journal_devnum = option;
1512 break;
818d276c 1513 case Opt_journal_checksum:
d4da6c9c
LT		 1514 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1515 break;
818d276c
GS		 1516 case Opt_journal_async_commit:
1517 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
d4da6c9c 1518 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
818d276c 1519 break;
ac27a0ec 1520 case Opt_noload:
2b2d6d01 1521 set_opt(sbi->s_mount_opt, NOLOAD);
ac27a0ec
DK		 1522 break;
1523 case Opt_commit:
1524 if (match_int(&args[0], &option))
1525 return 0;
1526 if (option < 0)
1527 return 0;
1528 if (option == 0)
cd02ff0b 1529 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
ac27a0ec
DK		 1530 sbi->s_commit_interval = HZ * option;
1531 break;
30773840
TT		 1532 case Opt_max_batch_time:
1533 if (match_int(&args[0], &option))
1534 return 0;
1535 if (option < 0)
1536 return 0;
1537 if (option == 0)
1538 option = EXT4_DEF_MAX_BATCH_TIME;
1539 sbi->s_max_batch_time = option;
1540 break;
1541 case Opt_min_batch_time:
1542 if (match_int(&args[0], &option))
1543 return 0;
1544 if (option < 0)
1545 return 0;
1546 sbi->s_min_batch_time = option;
1547 break;
ac27a0ec 1548 case Opt_data_journal:
617ba13b 1549 data_opt = EXT4_MOUNT_JOURNAL_DATA;
ac27a0ec
DK		 1550 goto datacheck;
1551 case Opt_data_ordered:
617ba13b 1552 data_opt = EXT4_MOUNT_ORDERED_DATA;
ac27a0ec
DK		 1553 goto datacheck;
1554 case Opt_data_writeback:
617ba13b 1555 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
ac27a0ec
DK		 1556 datacheck:
1557 if (is_remount) {
482a7425 1558 if (test_opt(sb, DATA_FLAGS) != data_opt) {
b31e1552
ES		 1559 ext4_msg(sb, KERN_ERR,
1560 "Cannot change data mode on remount");
ac27a0ec
DK		 1561 return 0;
1562 }
1563 } else {
482a7425 1564 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
ac27a0ec
DK		 1565 sbi->s_mount_opt |= data_opt;
1566 }
1567 break;
5bf5683a
HK		 1568 case Opt_data_err_abort:
1569 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1570 break;
1571 case Opt_data_err_ignore:
1572 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1573 break;
ac27a0ec
DK		 1574#ifdef CONFIG_QUOTA
1575 case Opt_usrjquota:
56c50f11 1576 if (!set_qf_name(sb, USRQUOTA, &args[0]))
ac27a0ec 1577 return 0;
56c50f11
DM		 1578 break;
1579 case Opt_grpjquota:
1580 if (!set_qf_name(sb, GRPQUOTA, &args[0]))
ac27a0ec 1581 return 0;
ac27a0ec
DK		 1582 break;
1583 case Opt_offusrjquota:
56c50f11
DM		 1584 if (!clear_qf_name(sb, USRQUOTA))
1585 return 0;
1586 break;
ac27a0ec 1587 case Opt_offgrpjquota:
56c50f11 1588 if (!clear_qf_name(sb, GRPQUOTA))
ac27a0ec 1589 return 0;
ac27a0ec 1590 break;
56c50f11 1591
ac27a0ec 1592 case Opt_jqfmt_vfsold:
dfc5d03f
JK		 1593 qfmt = QFMT_VFS_OLD;
1594 goto set_qf_format;
ac27a0ec 1595 case Opt_jqfmt_vfsv0:
dfc5d03f 1596 qfmt = QFMT_VFS_V0;
5a20bdfc
JK		 1597 goto set_qf_format;
1598 case Opt_jqfmt_vfsv1:
1599 qfmt = QFMT_VFS_V1;
dfc5d03f 1600set_qf_format:
17bd13b3 1601 if (sb_any_quota_loaded(sb) &&
dfc5d03f 1602 sbi->s_jquota_fmt != qfmt) {
b31e1552 1603 ext4_msg(sb, KERN_ERR, "Cannot change "
dfc5d03f 1604 "journaled quota options when "
b31e1552 1605 "quota turned on");
dfc5d03f
JK		 1606 return 0;
1607 }
1608 sbi->s_jquota_fmt = qfmt;
ac27a0ec
DK		 1609 break;
1610 case Opt_quota:
1611 case Opt_usrquota:
1612 set_opt(sbi->s_mount_opt, QUOTA);
1613 set_opt(sbi->s_mount_opt, USRQUOTA);
1614 break;
1615 case Opt_grpquota:
1616 set_opt(sbi->s_mount_opt, QUOTA);
1617 set_opt(sbi->s_mount_opt, GRPQUOTA);
1618 break;
1619 case Opt_noquota:
17bd13b3 1620 if (sb_any_quota_loaded(sb)) {
b31e1552
ES		 1621 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1622 "options when quota turned on");
ac27a0ec
DK		 1623 return 0;
1624 }
1625 clear_opt(sbi->s_mount_opt, QUOTA);
1626 clear_opt(sbi->s_mount_opt, USRQUOTA);
1627 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1628 break;
1629#else
1630 case Opt_quota:
1631 case Opt_usrquota:
1632 case Opt_grpquota:
b31e1552
ES		 1633 ext4_msg(sb, KERN_ERR,
1634 "quota options not supported");
cd59e7b9 1635 break;
ac27a0ec
DK		 1636 case Opt_usrjquota:
1637 case Opt_grpjquota:
1638 case Opt_offusrjquota:
1639 case Opt_offgrpjquota:
1640 case Opt_jqfmt_vfsold:
1641 case Opt_jqfmt_vfsv0:
5a20bdfc 1642 case Opt_jqfmt_vfsv1:
b31e1552
ES		 1643 ext4_msg(sb, KERN_ERR,
1644 "journaled quota options not supported");
ac27a0ec
DK		 1645 break;
1646 case Opt_noquota:
1647 break;
1648#endif
1649 case Opt_abort:
4ab2f15b 1650 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
ac27a0ec 1651 break;
06705bff
TT		 1652 case Opt_nobarrier:
1653 clear_opt(sbi->s_mount_opt, BARRIER);
1654 break;
ac27a0ec 1655 case Opt_barrier:
15121c18
ES		 1656 if (args[0].from) {
1657 if (match_int(&args[0], &option))
1658 return 0;
1659 } else
1660 option = 1; /* No argument, default to 1 */
ac27a0ec
DK		 1661 if (option)
1662 set_opt(sbi->s_mount_opt, BARRIER);
1663 else
1664 clear_opt(sbi->s_mount_opt, BARRIER);
1665 break;
1666 case Opt_ignore:
1667 break;
1668 case Opt_resize:
1669 if (!is_remount) {
b31e1552
ES		 1670 ext4_msg(sb, KERN_ERR,
1671 "resize option only available "
1672 "for remount");
ac27a0ec
DK		 1673 return 0;
1674 }
1675 if (match_int(&args[0], &option) != 0)
1676 return 0;
1677 *n_blocks_count = option;
1678 break;
1679 case Opt_nobh:
206f7ab4
CH		 1680 ext4_msg(sb, KERN_WARNING,
1681 "Ignoring deprecated nobh option");
ac27a0ec
DK		 1682 break;
1683 case Opt_bh:
206f7ab4
CH		 1684 ext4_msg(sb, KERN_WARNING,
1685 "Ignoring deprecated bh option");
ac27a0ec 1686 break;
25ec56b5
JNC		 1687 case Opt_i_version:
1688 set_opt(sbi->s_mount_opt, I_VERSION);
1689 sb->s_flags |= MS_I_VERSION;
1690 break;
dd919b98
AK		 1691 case Opt_nodelalloc:
1692 clear_opt(sbi->s_mount_opt, DELALLOC);
1693 break;
c9de560d
AT		 1694 case Opt_stripe:
1695 if (match_int(&args[0], &option))
1696 return 0;
1697 if (option < 0)
1698 return 0;
1699 sbi->s_stripe = option;
1700 break;
64769240
AT		 1701 case Opt_delalloc:
1702 set_opt(sbi->s_mount_opt, DELALLOC);
1703 break;
6fd058f7
TT		 1704 case Opt_block_validity:
1705 set_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1706 break;
1707 case Opt_noblock_validity:
1708 clear_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1709 break;
240799cd
TT		 1710 case Opt_inode_readahead_blks:
1711 if (match_int(&args[0], &option))
1712 return 0;
1713 if (option < 0 || option > (1 << 30))
1714 return 0;
f7c43950 1715 if (!is_power_of_2(option)) {
b31e1552
ES		 1716 ext4_msg(sb, KERN_ERR,
1717 "EXT4-fs: inode_readahead_blks"
1718 " must be a power of 2");
3197ebdb
TT		 1719 return 0;
1720 }
240799cd
TT		 1721 sbi->s_inode_readahead_blks = option;
1722 break;
b3881f74
TT		 1723 case Opt_journal_ioprio:
1724 if (match_int(&args[0], &option))
1725 return 0;
1726 if (option < 0 || option > 7)
1727 break;
1728 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1729 option);
1730 break;
06705bff
TT		 1731 case Opt_noauto_da_alloc:
1732 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1733 break;
afd4672d 1734 case Opt_auto_da_alloc:
15121c18
ES		 1735 if (args[0].from) {
1736 if (match_int(&args[0], &option))
1737 return 0;
1738 } else
1739 option = 1; /* No argument, default to 1 */
afd4672d
TT		 1740 if (option)
1741 clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1742 else
1743 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1744 break;
5328e635
ES		 1745 case Opt_discard:
1746 set_opt(sbi->s_mount_opt, DISCARD);
1747 break;
1748 case Opt_nodiscard:
1749 clear_opt(sbi->s_mount_opt, DISCARD);
1750 break;
744692dc
JZ		 1751 case Opt_dioread_nolock:
1752 set_opt(sbi->s_mount_opt, DIOREAD_NOLOCK);
1753 break;
1754 case Opt_dioread_lock:
1755 clear_opt(sbi->s_mount_opt, DIOREAD_NOLOCK);
1756 break;
ac27a0ec 1757 default:
b31e1552
ES		 1758 ext4_msg(sb, KERN_ERR,
1759 "Unrecognized mount option \"%s\" "
1760 "or missing value", p);
ac27a0ec
DK		 1761 return 0;
1762 }
1763 }
1764#ifdef CONFIG_QUOTA
1765 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
482a7425 1766 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
ac27a0ec
DK		 1767 clear_opt(sbi->s_mount_opt, USRQUOTA);
1768
482a7425 1769 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
ac27a0ec
DK		 1770 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1771
56c50f11 1772 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
b31e1552
ES		 1773 ext4_msg(sb, KERN_ERR, "old and new quota "
1774 "format mixing");
ac27a0ec
DK		 1775 return 0;
1776 }
1777
1778 if (!sbi->s_jquota_fmt) {
b31e1552
ES		 1779 ext4_msg(sb, KERN_ERR, "journaled quota format "
1780 "not specified");
ac27a0ec
DK		 1781 return 0;
1782 }
1783 } else {
1784 if (sbi->s_jquota_fmt) {
b31e1552 1785 ext4_msg(sb, KERN_ERR, "journaled quota format "
2c8be6b2 1786 "specified with no journaling "
b31e1552 1787 "enabled");
ac27a0ec
DK		 1788 return 0;
1789 }
1790 }
1791#endif
1792 return 1;
1793}
1794
617ba13b 1795static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 1796 int read_only)
1797{
617ba13b 1798 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 1799 int res = 0;
1800
617ba13b 1801 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
b31e1552
ES		 1802 ext4_msg(sb, KERN_ERR, "revision level too high, "
1803 "forcing read-only mode");
ac27a0ec
DK		 1804 res = MS_RDONLY;
1805 }
1806 if (read_only)
1807 return res;
617ba13b 1808 if (!(sbi->s_mount_state & EXT4_VALID_FS))
b31e1552
ES		 1809 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1810 "running e2fsck is recommended");
617ba13b 1811 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
b31e1552
ES		 1812 ext4_msg(sb, KERN_WARNING,
1813 "warning: mounting fs with errors, "
1814 "running e2fsck is recommended");
ac27a0ec
DK		 1815 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1816 le16_to_cpu(es->s_mnt_count) >=
1817 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
b31e1552
ES		 1818 ext4_msg(sb, KERN_WARNING,
1819 "warning: maximal mount count reached, "
1820 "running e2fsck is recommended");
ac27a0ec
DK		 1821 else if (le32_to_cpu(es->s_checkinterval) &&
1822 (le32_to_cpu(es->s_lastcheck) +
1823 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
b31e1552
ES		 1824 ext4_msg(sb, KERN_WARNING,
1825 "warning: checktime reached, "
1826 "running e2fsck is recommended");
0b8e58a1 1827 if (!sbi->s_journal)
0390131b 1828 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec 1829 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 1830 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 1831 le16_add_cpu(&es->s_mnt_count, 1);
ac27a0ec 1832 es->s_mtime = cpu_to_le32(get_seconds());
617ba13b 1833 ext4_update_dynamic_rev(sb);
0390131b
FM		 1834 if (sbi->s_journal)
1835 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 1836
e2d67052 1837 ext4_commit_super(sb, 1);
ac27a0ec 1838 if (test_opt(sb, DEBUG))
a9df9a49 1839 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
7f4520cc 1840 "bpg=%lu, ipg=%lu, mo=%04x]\n",
ac27a0ec
DK		 1841 sb->s_blocksize,
1842 sbi->s_groups_count,
617ba13b
MC		 1843 EXT4_BLOCKS_PER_GROUP(sb),
1844 EXT4_INODES_PER_GROUP(sb),
ac27a0ec
DK		 1845 sbi->s_mount_opt);
1846
ac27a0ec
DK		 1847 return res;
1848}
1849
772cb7c8
JS		 1850static int ext4_fill_flex_info(struct super_block *sb)
1851{
1852 struct ext4_sb_info *sbi = EXT4_SB(sb);
1853 struct ext4_group_desc *gdp = NULL;
772cb7c8
JS		 1854 ext4_group_t flex_group_count;
1855 ext4_group_t flex_group;
1856 int groups_per_flex = 0;
c5ca7c76 1857 size_t size;
772cb7c8
JS		 1858 int i;
1859
503358ae
TT		 1860 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1861 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1862
1863 if (groups_per_flex < 2) {
772cb7c8
JS		 1864 sbi->s_log_groups_per_flex = 0;
1865 return 1;
1866 }
1867
c62a11fd
FB		 1868 /* We allocate both existing and potentially added groups */
1869 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
d94e99a6
AK		 1870 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1871 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
c5ca7c76
TT		 1872 size = flex_group_count * sizeof(struct flex_groups);
1873 sbi->s_flex_groups = kzalloc(size, GFP_KERNEL);
1874 if (sbi->s_flex_groups == NULL) {
1875 sbi->s_flex_groups = vmalloc(size);
1876 if (sbi->s_flex_groups)
1877 memset(sbi->s_flex_groups, 0, size);
1878 }
772cb7c8 1879 if (sbi->s_flex_groups == NULL) {
b31e1552
ES		 1880 ext4_msg(sb, KERN_ERR, "not enough memory for "
1881 "%u flex groups", flex_group_count);
772cb7c8
JS		 1882 goto failed;
1883 }
772cb7c8 1884
772cb7c8 1885 for (i = 0; i < sbi->s_groups_count; i++) {
88b6edd1 1886 gdp = ext4_get_group_desc(sb, i, NULL);
772cb7c8
JS		 1887
1888 flex_group = ext4_flex_group(sbi, i);
7ad9bb65
TT		 1889 atomic_add(ext4_free_inodes_count(sb, gdp),
1890 &sbi->s_flex_groups[flex_group].free_inodes);
1891 atomic_add(ext4_free_blks_count(sb, gdp),
1892 &sbi->s_flex_groups[flex_group].free_blocks);
1893 atomic_add(ext4_used_dirs_count(sb, gdp),
1894 &sbi->s_flex_groups[flex_group].used_dirs);
772cb7c8
JS		 1895 }
1896
1897 return 1;
1898failed:
1899 return 0;
1900}
1901
717d50e4
AD		 1902__le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1903 struct ext4_group_desc *gdp)
1904{
1905 __u16 crc = 0;
1906
1907 if (sbi->s_es->s_feature_ro_compat &
1908 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1909 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1910 __le32 le_group = cpu_to_le32(block_group);
1911
1912 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1913 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1914 crc = crc16(crc, (__u8 *)gdp, offset);
1915 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1916 /* for checksum of struct ext4_group_desc do the rest...*/
1917 if ((sbi->s_es->s_feature_incompat &
1918 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1919 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1920 crc = crc16(crc, (__u8 *)gdp + offset,
1921 le16_to_cpu(sbi->s_es->s_desc_size) -
1922 offset);
1923 }
1924
1925 return cpu_to_le16(crc);
1926}
1927
1928int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1929 struct ext4_group_desc *gdp)
1930{
1931 if ((sbi->s_es->s_feature_ro_compat &
1932 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1933 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1934 return 0;
1935
1936 return 1;
1937}
1938
ac27a0ec 1939/* Called at mount-time, super-block is locked */
197cd65a 1940static int ext4_check_descriptors(struct super_block *sb)
ac27a0ec 1941{
617ba13b
MC		 1942 struct ext4_sb_info *sbi = EXT4_SB(sb);
1943 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1944 ext4_fsblk_t last_block;
bd81d8ee
LV		 1945 ext4_fsblk_t block_bitmap;
1946 ext4_fsblk_t inode_bitmap;
1947 ext4_fsblk_t inode_table;
ce421581 1948 int flexbg_flag = 0;
fd2d4291 1949 ext4_group_t i;
ac27a0ec 1950
ce421581
JS		 1951 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1952 flexbg_flag = 1;
1953
af5bc92d 1954 ext4_debug("Checking group descriptors");
ac27a0ec 1955
197cd65a
AM		 1956 for (i = 0; i < sbi->s_groups_count; i++) {
1957 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1958
ce421581 1959 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 1960 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 1961 else
1962 last_block = first_block +
617ba13b 1963 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 1964
8fadc143 1965 block_bitmap = ext4_block_bitmap(sb, gdp);
2b2d6d01 1966 if (block_bitmap < first_block || block_bitmap > last_block) {
b31e1552 1967 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 1968 "Block bitmap for group %u not in group "
b31e1552 1969 "(block %llu)!", i, block_bitmap);
ac27a0ec
DK		 1970 return 0;
1971 }
8fadc143 1972 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2b2d6d01 1973 if (inode_bitmap < first_block || inode_bitmap > last_block) {
b31e1552 1974 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 1975 "Inode bitmap for group %u not in group "
b31e1552 1976 "(block %llu)!", i, inode_bitmap);
ac27a0ec
DK		 1977 return 0;
1978 }
8fadc143 1979 inode_table = ext4_inode_table(sb, gdp);
bd81d8ee 1980 if (inode_table < first_block ||
2b2d6d01 1981 inode_table + sbi->s_itb_per_group - 1 > last_block) {
b31e1552 1982 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 1983 "Inode table for group %u not in group "
b31e1552 1984 "(block %llu)!", i, inode_table);
ac27a0ec
DK		 1985 return 0;
1986 }
955ce5f5 1987 ext4_lock_group(sb, i);
717d50e4 1988 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
b31e1552
ES		 1989 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1990 "Checksum for group %u failed (%u!=%u)",
1991 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1992 gdp)), le16_to_cpu(gdp->bg_checksum));
7ee1ec4c 1993 if (!(sb->s_flags & MS_RDONLY)) {
955ce5f5 1994 ext4_unlock_group(sb, i);
8a266467 1995 return 0;
7ee1ec4c 1996 }
717d50e4 1997 }
955ce5f5 1998 ext4_unlock_group(sb, i);
ce421581
JS		 1999 if (!flexbg_flag)
2000 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec
DK		 2001 }
2002
bd81d8ee 2003 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
0b8e58a1 2004 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
ac27a0ec
DK		 2005 return 1;
2006}
2007
617ba13b 2008/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 2009 * the superblock) which were deleted from all directories, but held open by
2010 * a process at the time of a crash. We walk the list and try to delete these
2011 * inodes at recovery time (only with a read-write filesystem).
2012 *
2013 * In order to keep the orphan inode chain consistent during traversal (in
2014 * case of crash during recovery), we link each inode into the superblock
2015 * orphan list_head and handle it the same way as an inode deletion during
2016 * normal operation (which journals the operations for us).
2017 *
2018 * We only do an iget() and an iput() on each inode, which is very safe if we
2019 * accidentally point at an in-use or already deleted inode. The worst that
2020 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 2021 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 2022 * e2fsck was run on this filesystem, and it must have already done the orphan
2023 * inode cleanup for us, so we can safely abort without any further action.
2024 */
2b2d6d01
TT		 2025static void ext4_orphan_cleanup(struct super_block *sb,
2026 struct ext4_super_block *es)
ac27a0ec
DK		 2027{
2028 unsigned int s_flags = sb->s_flags;
2029 int nr_orphans = 0, nr_truncates = 0;
2030#ifdef CONFIG_QUOTA
2031 int i;
2032#endif
2033 if (!es->s_last_orphan) {
2034 jbd_debug(4, "no orphan inodes to clean up\n");
2035 return;
2036 }
2037
a8f48a95 2038 if (bdev_read_only(sb->s_bdev)) {
b31e1552
ES		 2039 ext4_msg(sb, KERN_ERR, "write access "
2040 "unavailable, skipping orphan cleanup");
a8f48a95
ES		 2041 return;
2042 }
2043
617ba13b 2044 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
ac27a0ec
DK		 2045 if (es->s_last_orphan)
2046 jbd_debug(1, "Errors on filesystem, "
2047 "clearing orphan list.\n");
2048 es->s_last_orphan = 0;
2049 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2050 return;
2051 }
2052
2053 if (s_flags & MS_RDONLY) {
b31e1552 2054 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
ac27a0ec
DK		 2055 sb->s_flags &= ~MS_RDONLY;
2056 }
2057#ifdef CONFIG_QUOTA
2058 /* Needed for iput() to work correctly and not trash data */
2059 sb->s_flags |= MS_ACTIVE;
2060 /* Turn on quotas so that they are updated correctly */
2061 for (i = 0; i < MAXQUOTAS; i++) {
617ba13b
MC		 2062 if (EXT4_SB(sb)->s_qf_names[i]) {
2063 int ret = ext4_quota_on_mount(sb, i);
ac27a0ec 2064 if (ret < 0)
b31e1552
ES		 2065 ext4_msg(sb, KERN_ERR,
2066 "Cannot turn on journaled "
2067 "quota: error %d", ret);
ac27a0ec
DK		 2068 }
2069 }
2070#endif
2071
2072 while (es->s_last_orphan) {
2073 struct inode *inode;
2074
97bd42b9
JB		 2075 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2076 if (IS_ERR(inode)) {
ac27a0ec
DK		 2077 es->s_last_orphan = 0;
2078 break;
2079 }
2080
617ba13b 2081 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
871a2931 2082 dquot_initialize(inode);
ac27a0ec 2083 if (inode->i_nlink) {
b31e1552
ES		 2084 ext4_msg(sb, KERN_DEBUG,
2085 "%s: truncating inode %lu to %lld bytes",
46e665e9 2086 __func__, inode->i_ino, inode->i_size);
e5f8eab8 2087 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
ac27a0ec 2088 inode->i_ino, inode->i_size);
617ba13b 2089 ext4_truncate(inode);
ac27a0ec
DK		 2090 nr_truncates++;
2091 } else {
b31e1552
ES		 2092 ext4_msg(sb, KERN_DEBUG,
2093 "%s: deleting unreferenced inode %lu",
46e665e9 2094 __func__, inode->i_ino);
ac27a0ec
DK		 2095 jbd_debug(2, "deleting unreferenced inode %lu\n",
2096 inode->i_ino);
2097 nr_orphans++;
2098 }
2099 iput(inode); /* The delete magic happens here! */
2100 }
2101
2b2d6d01 2102#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
ac27a0ec
DK		 2103
2104 if (nr_orphans)
b31e1552
ES		 2105 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2106 PLURAL(nr_orphans));
ac27a0ec 2107 if (nr_truncates)
b31e1552
ES		 2108 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2109 PLURAL(nr_truncates));
ac27a0ec
DK		 2110#ifdef CONFIG_QUOTA
2111 /* Turn quotas off */
2112 for (i = 0; i < MAXQUOTAS; i++) {
2113 if (sb_dqopt(sb)->files[i])
287a8095 2114 dquot_quota_off(sb, i);
ac27a0ec
DK		 2115 }
2116#endif
2117 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2118}
0b8e58a1 2119
cd2291a4
ES		 2120/*
2121 * Maximal extent format file size.
2122 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2123 * extent format containers, within a sector_t, and within i_blocks
2124 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2125 * so that won't be a limiting factor.
2126 *
2127 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2128 */
f287a1a5 2129static loff_t ext4_max_size(int blkbits, int has_huge_files)
cd2291a4
ES		 2130{
2131 loff_t res;
2132 loff_t upper_limit = MAX_LFS_FILESIZE;
2133
2134 /* small i_blocks in vfs inode? */
f287a1a5 2135 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
cd2291a4 2136 /*
90c699a9 2137 * CONFIG_LBDAF is not enabled implies the inode
cd2291a4
ES		 2138 * i_block represent total blocks in 512 bytes
2139 * 32 == size of vfs inode i_blocks * 8
2140 */
2141 upper_limit = (1LL << 32) - 1;
2142
2143 /* total blocks in file system block size */
2144 upper_limit >>= (blkbits - 9);
2145 upper_limit <<= blkbits;
2146 }
2147
2148 /* 32-bit extent-start container, ee_block */
2149 res = 1LL << 32;
2150 res <<= blkbits;
2151 res -= 1;
2152
2153 /* Sanity check against vm- & vfs- imposed limits */
2154 if (res > upper_limit)
2155 res = upper_limit;
2156
2157 return res;
2158}
ac27a0ec 2159
ac27a0ec 2160/*
cd2291a4 2161 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 2162 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2163 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 2164 */
f287a1a5 2165static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
ac27a0ec 2166{
617ba13b 2167 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 2168 int meta_blocks;
2169 loff_t upper_limit;
0b8e58a1
AD		 2170 /* This is calculated to be the largest file size for a dense, block
2171 * mapped file such that the file's total number of 512-byte sectors,
2172 * including data and all indirect blocks, does not exceed (2^48 - 1).
2173 *
2174 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2175 * number of 512-byte sectors of the file.
0fc1b451
AK		 2176 */
2177
f287a1a5 2178 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
0fc1b451 2179 /*
90c699a9 2180 * !has_huge_files or CONFIG_LBDAF not enabled implies that
0b8e58a1
AD		 2181 * the inode i_block field represents total file blocks in
2182 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
0fc1b451
AK		 2183 */
2184 upper_limit = (1LL << 32) - 1;
2185
2186 /* total blocks in file system block size */
2187 upper_limit >>= (bits - 9);
2188
2189 } else {
8180a562
AK		 2190 /*
2191 * We use 48 bit ext4_inode i_blocks
2192 * With EXT4_HUGE_FILE_FL set the i_blocks
2193 * represent total number of blocks in
2194 * file system block size
2195 */
0fc1b451
AK		 2196 upper_limit = (1LL << 48) - 1;
2197
0fc1b451
AK		 2198 }
2199
2200 /* indirect blocks */
2201 meta_blocks = 1;
2202 /* double indirect blocks */
2203 meta_blocks += 1 + (1LL << (bits-2));
2204 /* tripple indirect blocks */
2205 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2206
2207 upper_limit -= meta_blocks;
2208 upper_limit <<= bits;
ac27a0ec
DK		 2209
2210 res += 1LL << (bits-2);
2211 res += 1LL << (2*(bits-2));
2212 res += 1LL << (3*(bits-2));
2213 res <<= bits;
2214 if (res > upper_limit)
2215 res = upper_limit;
0fc1b451
AK		 2216
2217 if (res > MAX_LFS_FILESIZE)
2218 res = MAX_LFS_FILESIZE;
2219
ac27a0ec
DK		 2220 return res;
2221}
2222
617ba13b 2223static ext4_fsblk_t descriptor_loc(struct super_block *sb,
0b8e58a1 2224 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 2225{
617ba13b 2226 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 2227 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 2228 int has_super = 0;
2229
2230 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2231
617ba13b 2232 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
ac27a0ec 2233 nr < first_meta_bg)
70bbb3e0 2234 return logical_sb_block + nr + 1;
ac27a0ec 2235 bg = sbi->s_desc_per_block * nr;
617ba13b 2236 if (ext4_bg_has_super(sb, bg))
ac27a0ec 2237 has_super = 1;
0b8e58a1 2238
617ba13b 2239 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 2240}
2241
c9de560d
AT		 2242/**
2243 * ext4_get_stripe_size: Get the stripe size.
2244 * @sbi: In memory super block info
2245 *
2246 * If we have specified it via mount option, then
2247 * use the mount option value. If the value specified at mount time is
2248 * greater than the blocks per group use the super block value.
2249 * If the super block value is greater than blocks per group return 0.
2250 * Allocator needs it be less than blocks per group.
2251 *
2252 */
2253static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2254{
2255 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2256 unsigned long stripe_width =
2257 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2258
2259 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2260 return sbi->s_stripe;
2261
2262 if (stripe_width <= sbi->s_blocks_per_group)
2263 return stripe_width;
2264
2265 if (stride <= sbi->s_blocks_per_group)
2266 return stride;
2267
2268 return 0;
2269}
ac27a0ec 2270
3197ebdb
TT		 2271/* sysfs supprt */
2272
2273struct ext4_attr {
2274 struct attribute attr;
2275 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
60e6679e 2276 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
3197ebdb
TT		 2277 const char *, size_t);
2278 int offset;
2279};
2280
2281static int parse_strtoul(const char *buf,
2282 unsigned long max, unsigned long *value)
2283{
2284 char *endp;
2285
e7d2860b
AGR		 2286 *value = simple_strtoul(skip_spaces(buf), &endp, 0);
2287 endp = skip_spaces(endp);
3197ebdb
TT		 2288 if (*endp || *value > max)
2289 return -EINVAL;
2290
2291 return 0;
2292}
2293
2294static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2295 struct ext4_sb_info *sbi,
2296 char *buf)
2297{
2298 return snprintf(buf, PAGE_SIZE, "%llu\n",
2299 (s64) percpu_counter_sum(&sbi->s_dirtyblocks_counter));
2300}
2301
2302static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2303 struct ext4_sb_info *sbi, char *buf)
2304{
2305 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2306
2307 return snprintf(buf, PAGE_SIZE, "%lu\n",
2308 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2309 sbi->s_sectors_written_start) >> 1);
2310}
2311
2312static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2313 struct ext4_sb_info *sbi, char *buf)
2314{
2315 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2316
2317 return snprintf(buf, PAGE_SIZE, "%llu\n",
a6b43e38 2318 (unsigned long long)(sbi->s_kbytes_written +
3197ebdb 2319 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
a6b43e38 2320 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
3197ebdb
TT		 2321}
2322
2323static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2324 struct ext4_sb_info *sbi,
2325 const char *buf, size_t count)
2326{
2327 unsigned long t;
2328
2329 if (parse_strtoul(buf, 0x40000000, &t))
2330 return -EINVAL;
2331
f7c43950 2332 if (!is_power_of_2(t))
3197ebdb
TT		 2333 return -EINVAL;
2334
2335 sbi->s_inode_readahead_blks = t;
2336 return count;
2337}
2338
2339static ssize_t sbi_ui_show(struct ext4_attr *a,
0b8e58a1 2340 struct ext4_sb_info *sbi, char *buf)
3197ebdb
TT		 2341{
2342 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2343
2344 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2345}
2346
2347static ssize_t sbi_ui_store(struct ext4_attr *a,
2348 struct ext4_sb_info *sbi,
2349 const char *buf, size_t count)
2350{
2351 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2352 unsigned long t;
2353
2354 if (parse_strtoul(buf, 0xffffffff, &t))
2355 return -EINVAL;
2356 *ui = t;
2357 return count;
2358}
2359
2360#define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2361static struct ext4_attr ext4_attr_##_name = { \
2362 .attr = {.name = __stringify(_name), .mode = _mode }, \
2363 .show = _show, \
2364 .store = _store, \
2365 .offset = offsetof(struct ext4_sb_info, _elname), \
2366}
2367#define EXT4_ATTR(name, mode, show, store) \
2368static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2369
2370#define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2371#define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2372#define EXT4_RW_ATTR_SBI_UI(name, elname) \
2373 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2374#define ATTR_LIST(name) &ext4_attr_##name.attr
2375
2376EXT4_RO_ATTR(delayed_allocation_blocks);
2377EXT4_RO_ATTR(session_write_kbytes);
2378EXT4_RO_ATTR(lifetime_write_kbytes);
2379EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2380 inode_readahead_blks_store, s_inode_readahead_blks);
11013911 2381EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
3197ebdb
TT		 2382EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2383EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2384EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2385EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2386EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2387EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
55138e0b 2388EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
3197ebdb
TT		 2389
2390static struct attribute *ext4_attrs[] = {
2391 ATTR_LIST(delayed_allocation_blocks),
2392 ATTR_LIST(session_write_kbytes),
2393 ATTR_LIST(lifetime_write_kbytes),
2394 ATTR_LIST(inode_readahead_blks),
11013911 2395 ATTR_LIST(inode_goal),
3197ebdb
TT		 2396 ATTR_LIST(mb_stats),
2397 ATTR_LIST(mb_max_to_scan),
2398 ATTR_LIST(mb_min_to_scan),
2399 ATTR_LIST(mb_order2_req),
2400 ATTR_LIST(mb_stream_req),
2401 ATTR_LIST(mb_group_prealloc),
55138e0b 2402 ATTR_LIST(max_writeback_mb_bump),
3197ebdb
TT		 2403 NULL,
2404};
2405
2406static ssize_t ext4_attr_show(struct kobject *kobj,
2407 struct attribute *attr, char *buf)
2408{
2409 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2410 s_kobj);
2411 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2412
2413 return a->show ? a->show(a, sbi, buf) : 0;
2414}
2415
2416static ssize_t ext4_attr_store(struct kobject *kobj,
2417 struct attribute *attr,
2418 const char *buf, size_t len)
2419{
2420 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2421 s_kobj);
2422 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2423
2424 return a->store ? a->store(a, sbi, buf, len) : 0;
2425}
2426
2427static void ext4_sb_release(struct kobject *kobj)
2428{
2429 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2430 s_kobj);
2431 complete(&sbi->s_kobj_unregister);
2432}
2433
2434
52cf25d0 2435static const struct sysfs_ops ext4_attr_ops = {
3197ebdb
TT		 2436 .show = ext4_attr_show,
2437 .store = ext4_attr_store,
2438};
2439
2440static struct kobj_type ext4_ktype = {
2441 .default_attrs = ext4_attrs,
2442 .sysfs_ops = &ext4_attr_ops,
2443 .release = ext4_sb_release,
2444};
2445
a13fb1a4
ES		 2446/*
2447 * Check whether this filesystem can be mounted based on
2448 * the features present and the RDONLY/RDWR mount requested.
2449 * Returns 1 if this filesystem can be mounted as requested,
2450 * 0 if it cannot be.
2451 */
2452static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2453{
2454 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2455 ext4_msg(sb, KERN_ERR,
2456 "Couldn't mount because of "
2457 "unsupported optional features (%x)",
2458 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2459 ~EXT4_FEATURE_INCOMPAT_SUPP));
2460 return 0;
2461 }
2462
2463 if (readonly)
2464 return 1;
2465
2466 /* Check that feature set is OK for a read-write mount */
2467 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2468 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2469 "unsupported optional features (%x)",
2470 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2471 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2472 return 0;
2473 }
2474 /*
2475 * Large file size enabled file system can only be mounted
2476 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2477 */
2478 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2479 if (sizeof(blkcnt_t) < sizeof(u64)) {
2480 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2481 "cannot be mounted RDWR without "
2482 "CONFIG_LBDAF");
2483 return 0;
2484 }
2485 }
2486 return 1;
2487}
2488
66e61a9e
TT		 2489/*
2490 * This function is called once a day if we have errors logged
2491 * on the file system
2492 */
2493static void print_daily_error_info(unsigned long arg)
2494{
2495 struct super_block *sb = (struct super_block *) arg;
2496 struct ext4_sb_info *sbi;
2497 struct ext4_super_block *es;
2498
2499 sbi = EXT4_SB(sb);
2500 es = sbi->s_es;
2501
2502 if (es->s_error_count)
2503 ext4_msg(sb, KERN_NOTICE, "error count: %u",
2504 le32_to_cpu(es->s_error_count));
2505 if (es->s_first_error_time) {
2506 printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d",
2507 sb->s_id, le32_to_cpu(es->s_first_error_time),
2508 (int) sizeof(es->s_first_error_func),
2509 es->s_first_error_func,
2510 le32_to_cpu(es->s_first_error_line));
2511 if (es->s_first_error_ino)
2512 printk(": inode %u",
2513 le32_to_cpu(es->s_first_error_ino));
2514 if (es->s_first_error_block)
2515 printk(": block %llu", (unsigned long long)
2516 le64_to_cpu(es->s_first_error_block));
2517 printk("\n");
2518 }
2519 if (es->s_last_error_time) {
2520 printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d",
2521 sb->s_id, le32_to_cpu(es->s_last_error_time),
2522 (int) sizeof(es->s_last_error_func),
2523 es->s_last_error_func,
2524 le32_to_cpu(es->s_last_error_line));
2525 if (es->s_last_error_ino)
2526 printk(": inode %u",
2527 le32_to_cpu(es->s_last_error_ino));
2528 if (es->s_last_error_block)
2529 printk(": block %llu", (unsigned long long)
2530 le64_to_cpu(es->s_last_error_block));
2531 printk("\n");
2532 }
2533 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2534}
2535
2b2d6d01 2536static int ext4_fill_super(struct super_block *sb, void *data, int silent)
7477827f
AK		 2537 __releases(kernel_lock)
2538 __acquires(kernel_lock)
ac27a0ec 2539{
d4c402d9 2540 char *orig_data = kstrdup(data, GFP_KERNEL);
2b2d6d01 2541 struct buffer_head *bh;
617ba13b
MC		 2542 struct ext4_super_block *es = NULL;
2543 struct ext4_sb_info *sbi;
2544 ext4_fsblk_t block;
2545 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 2546 ext4_fsblk_t logical_sb_block;
ac27a0ec 2547 unsigned long offset = 0;
ac27a0ec
DK		 2548 unsigned long journal_devnum = 0;
2549 unsigned long def_mount_opts;
2550 struct inode *root;
9f6200bb 2551 char *cp;
0390131b 2552 const char *descr;
1d1fe1ee 2553 int ret = -EINVAL;
ac27a0ec 2554 int blocksize;
4ec11028
TT		 2555 unsigned int db_count;
2556 unsigned int i;
f287a1a5 2557 int needs_recovery, has_huge_files;
bd81d8ee 2558 __u64 blocks_count;
833f4077 2559 int err;
b3881f74 2560 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
ac27a0ec
DK		 2561
2562 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
2563 if (!sbi)
2564 return -ENOMEM;
705895b6
PE		 2565
2566 sbi->s_blockgroup_lock =
2567 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
2568 if (!sbi->s_blockgroup_lock) {
2569 kfree(sbi);
2570 return -ENOMEM;
2571 }
ac27a0ec
DK		 2572 sb->s_fs_info = sbi;
2573 sbi->s_mount_opt = 0;
617ba13b
MC		 2574 sbi->s_resuid = EXT4_DEF_RESUID;
2575 sbi->s_resgid = EXT4_DEF_RESGID;
240799cd 2576 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
d9c9bef1 2577 sbi->s_sb_block = sb_block;
afc32f7e
TT		 2578 sbi->s_sectors_written_start = part_stat_read(sb->s_bdev->bd_part,
2579 sectors[1]);
ac27a0ec
DK		 2580
2581 unlock_kernel();
2582
9f6200bb
TT		 2583 /* Cleanup superblock name */
2584 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
2585 *cp = '!';
2586
617ba13b 2587 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 2588 if (!blocksize) {
b31e1552 2589 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
ac27a0ec
DK		 2590 goto out_fail;
2591 }
2592
2593 /*
617ba13b 2594 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 2595 * block sizes. We need to calculate the offset from buffer start.
2596 */
617ba13b 2597 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 2598 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2599 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 2600 } else {
70bbb3e0 2601 logical_sb_block = sb_block;
ac27a0ec
DK		 2602 }
2603
70bbb3e0 2604 if (!(bh = sb_bread(sb, logical_sb_block))) {
b31e1552 2605 ext4_msg(sb, KERN_ERR, "unable to read superblock");
ac27a0ec
DK		 2606 goto out_fail;
2607 }
2608 /*
2609 * Note: s_es must be initialized as soon as possible because
617ba13b 2610 * some ext4 macro-instructions depend on its value
ac27a0ec 2611 */
617ba13b 2612 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
ac27a0ec
DK		 2613 sbi->s_es = es;
2614 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 2615 if (sb->s_magic != EXT4_SUPER_MAGIC)
2616 goto cantfind_ext4;
afc32f7e 2617 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
ac27a0ec
DK		 2618
2619 /* Set defaults before we parse the mount options */
2620 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
617ba13b 2621 if (def_mount_opts & EXT4_DEFM_DEBUG)
ac27a0ec 2622 set_opt(sbi->s_mount_opt, DEBUG);
437ca0fd
DM		 2623 if (def_mount_opts & EXT4_DEFM_BSDGROUPS) {
2624 ext4_msg(sb, KERN_WARNING, deprecated_msg, "bsdgroups",
2625 "2.6.38");
ac27a0ec 2626 set_opt(sbi->s_mount_opt, GRPID);
437ca0fd 2627 }
617ba13b 2628 if (def_mount_opts & EXT4_DEFM_UID16)
ac27a0ec 2629 set_opt(sbi->s_mount_opt, NO_UID32);
03010a33 2630#ifdef CONFIG_EXT4_FS_XATTR
617ba13b 2631 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
ac27a0ec 2632 set_opt(sbi->s_mount_opt, XATTR_USER);
2e7842b8 2633#endif
03010a33 2634#ifdef CONFIG_EXT4_FS_POSIX_ACL
617ba13b 2635 if (def_mount_opts & EXT4_DEFM_ACL)
ac27a0ec 2636 set_opt(sbi->s_mount_opt, POSIX_ACL);
2e7842b8 2637#endif
617ba13b 2638 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
482a7425 2639 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
617ba13b 2640 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
482a7425 2641 set_opt(sbi->s_mount_opt, ORDERED_DATA);
617ba13b 2642 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
482a7425 2643 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
617ba13b
MC		 2644
2645 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
ac27a0ec 2646 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
bb4f397a 2647 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
ceea16bf 2648 set_opt(sbi->s_mount_opt, ERRORS_CONT);
bb4f397a
AK		 2649 else
2650 set_opt(sbi->s_mount_opt, ERRORS_RO);
ac27a0ec
DK		 2651
2652 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
2653 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
30773840
TT		 2654 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
2655 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
2656 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
ac27a0ec 2657
571640ca 2658 set_opt(sbi->s_mount_opt, BARRIER);
ac27a0ec 2659
dd919b98
AK		 2660 /*
2661 * enable delayed allocation by default
2662 * Use -o nodelalloc to turn it off
2663 */
ba69f9ab
JK		 2664 if (!IS_EXT3_SB(sb))
2665 set_opt(sbi->s_mount_opt, DELALLOC);
dd919b98 2666
b3881f74
TT		 2667 if (!parse_options((char *) data, sb, &journal_devnum,
2668 &journal_ioprio, NULL, 0))
ac27a0ec
DK		 2669 goto failed_mount;
2670
2671 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 2672 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec 2673
617ba13b
MC		 2674 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
2675 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
2676 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
2677 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
b31e1552
ES		 2678 ext4_msg(sb, KERN_WARNING,
2679 "feature flags set on rev 0 fs, "
2680 "running e2fsck is recommended");
469108ff 2681
ac27a0ec
DK		 2682 /*
2683 * Check feature flags regardless of the revision level, since we
2684 * previously didn't change the revision level when setting the flags,
2685 * so there is a chance incompat flags are set on a rev 0 filesystem.
2686 */
a13fb1a4 2687 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
ac27a0ec 2688 goto failed_mount;
a13fb1a4 2689
ac27a0ec
DK		 2690 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
2691
617ba13b
MC		 2692 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
2693 blocksize > EXT4_MAX_BLOCK_SIZE) {
b31e1552
ES		 2694 ext4_msg(sb, KERN_ERR,
2695 "Unsupported filesystem blocksize %d", blocksize);
ac27a0ec
DK		 2696 goto failed_mount;
2697 }
2698
ac27a0ec 2699 if (sb->s_blocksize != blocksize) {
ce40733c
AK		 2700 /* Validate the filesystem blocksize */
2701 if (!sb_set_blocksize(sb, blocksize)) {
b31e1552 2702 ext4_msg(sb, KERN_ERR, "bad block size %d",
ce40733c 2703 blocksize);
ac27a0ec
DK		 2704 goto failed_mount;
2705 }
2706
2b2d6d01 2707 brelse(bh);
70bbb3e0
AM		 2708 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2709 offset = do_div(logical_sb_block, blocksize);
2710 bh = sb_bread(sb, logical_sb_block);
ac27a0ec 2711 if (!bh) {
b31e1552
ES		 2712 ext4_msg(sb, KERN_ERR,
2713 "Can't read superblock on 2nd try");
ac27a0ec
DK		 2714 goto failed_mount;
2715 }
617ba13b 2716 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
ac27a0ec 2717 sbi->s_es = es;
617ba13b 2718 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
b31e1552
ES		 2719 ext4_msg(sb, KERN_ERR,
2720 "Magic mismatch, very weird!");
ac27a0ec
DK		 2721 goto failed_mount;
2722 }
2723 }
2724
a13fb1a4
ES		 2725 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2726 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
f287a1a5
TT		 2727 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
2728 has_huge_files);
2729 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
ac27a0ec 2730
617ba13b
MC		 2731 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2732 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2733 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
ac27a0ec
DK		 2734 } else {
2735 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2736 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
617ba13b 2737 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1330593e 2738 (!is_power_of_2(sbi->s_inode_size)) ||
ac27a0ec 2739 (sbi->s_inode_size > blocksize)) {
b31e1552
ES		 2740 ext4_msg(sb, KERN_ERR,
2741 "unsupported inode size: %d",
2b2d6d01 2742 sbi->s_inode_size);
ac27a0ec
DK		 2743 goto failed_mount;
2744 }
ef7f3835
KS		 2745 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2746 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
ac27a0ec 2747 }
0b8e58a1 2748
0d1ee42f
AR		 2749 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2750 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
8fadc143 2751 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 2752 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 2753 !is_power_of_2(sbi->s_desc_size)) {
b31e1552
ES		 2754 ext4_msg(sb, KERN_ERR,
2755 "unsupported descriptor size %lu",
0d1ee42f
AR		 2756 sbi->s_desc_size);
2757 goto failed_mount;
2758 }
2759 } else
2760 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
0b8e58a1 2761
ac27a0ec 2762 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 2763 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
b47b6f38 2764 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
617ba13b 2765 goto cantfind_ext4;
0b8e58a1 2766
617ba13b 2767 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 2768 if (sbi->s_inodes_per_block == 0)
617ba13b 2769 goto cantfind_ext4;
ac27a0ec
DK		 2770 sbi->s_itb_per_group = sbi->s_inodes_per_group /
2771 sbi->s_inodes_per_block;
0d1ee42f 2772 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 2773 sbi->s_sbh = bh;
2774 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 2775 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2776 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
0b8e58a1 2777
2b2d6d01 2778 for (i = 0; i < 4; i++)
ac27a0ec
DK		 2779 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2780 sbi->s_def_hash_version = es->s_def_hash_version;
f99b2589
TT		 2781 i = le32_to_cpu(es->s_flags);
2782 if (i & EXT2_FLAGS_UNSIGNED_HASH)
2783 sbi->s_hash_unsigned = 3;
2784 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
2785#ifdef __CHAR_UNSIGNED__
2786 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
2787 sbi->s_hash_unsigned = 3;
2788#else
2789 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
2790#endif
2791 sb->s_dirt = 1;
2792 }
ac27a0ec
DK		 2793
2794 if (sbi->s_blocks_per_group > blocksize * 8) {
b31e1552
ES		 2795 ext4_msg(sb, KERN_ERR,
2796 "#blocks per group too big: %lu",
2b2d6d01 2797 sbi->s_blocks_per_group);
ac27a0ec
DK		 2798 goto failed_mount;
2799 }
ac27a0ec 2800 if (sbi->s_inodes_per_group > blocksize * 8) {
b31e1552
ES		 2801 ext4_msg(sb, KERN_ERR,
2802 "#inodes per group too big: %lu",
2b2d6d01 2803 sbi->s_inodes_per_group);
ac27a0ec
DK		 2804 goto failed_mount;
2805 }
2806
bf43d84b
ES		 2807 /*
2808 * Test whether we have more sectors than will fit in sector_t,
2809 * and whether the max offset is addressable by the page cache.
2810 */
2811 if ((ext4_blocks_count(es) >
2812 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) ||
2813 (ext4_blocks_count(es) >
2814 (pgoff_t)(~0ULL) >> (PAGE_CACHE_SHIFT - sb->s_blocksize_bits))) {
b31e1552 2815 ext4_msg(sb, KERN_ERR, "filesystem"
bf43d84b 2816 " too large to mount safely on this system");
ac27a0ec 2817 if (sizeof(sector_t) < 8)
90c699a9 2818 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
bf43d84b 2819 ret = -EFBIG;
ac27a0ec
DK		 2820 goto failed_mount;
2821 }
2822
617ba13b
MC		 2823 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2824 goto cantfind_ext4;
e7c95593 2825
0f2ddca6
FTN		 2826 /* check blocks count against device size */
2827 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
2828 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
b31e1552
ES		 2829 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
2830 "exceeds size of device (%llu blocks)",
0f2ddca6
FTN		 2831 ext4_blocks_count(es), blocks_count);
2832 goto failed_mount;
2833 }
2834
0b8e58a1
AD		 2835 /*
2836 * It makes no sense for the first data block to be beyond the end
2837 * of the filesystem.
2838 */
2839 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
b31e1552
ES		 2840 ext4_msg(sb, KERN_WARNING, "bad geometry: first data"
2841 "block %u is beyond end of filesystem (%llu)",
2842 le32_to_cpu(es->s_first_data_block),
2843 ext4_blocks_count(es));
e7c95593
ES		 2844 goto failed_mount;
2845 }
bd81d8ee
LV		 2846 blocks_count = (ext4_blocks_count(es) -
2847 le32_to_cpu(es->s_first_data_block) +
2848 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2849 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4ec11028 2850 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
b31e1552 2851 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
4ec11028 2852 "(block count %llu, first data block %u, "
b31e1552 2853 "blocks per group %lu)", sbi->s_groups_count,
4ec11028
TT		 2854 ext4_blocks_count(es),
2855 le32_to_cpu(es->s_first_data_block),
2856 EXT4_BLOCKS_PER_GROUP(sb));
2857 goto failed_mount;
2858 }
bd81d8ee 2859 sbi->s_groups_count = blocks_count;
fb0a387d
ES		 2860 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
2861 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
617ba13b
MC		 2862 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2863 EXT4_DESC_PER_BLOCK(sb);
2b2d6d01 2864 sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
ac27a0ec
DK		 2865 GFP_KERNEL);
2866 if (sbi->s_group_desc == NULL) {
b31e1552 2867 ext4_msg(sb, KERN_ERR, "not enough memory");
ac27a0ec
DK		 2868 goto failed_mount;
2869 }
2870
3244fcb1 2871#ifdef CONFIG_PROC_FS
9f6200bb
TT		 2872 if (ext4_proc_root)
2873 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
3244fcb1 2874#endif
240799cd 2875
705895b6 2876 bgl_lock_init(sbi->s_blockgroup_lock);
ac27a0ec
DK		 2877
2878 for (i = 0; i < db_count; i++) {
70bbb3e0 2879 block = descriptor_loc(sb, logical_sb_block, i);
ac27a0ec
DK		 2880 sbi->s_group_desc[i] = sb_bread(sb, block);
2881 if (!sbi->s_group_desc[i]) {
b31e1552
ES		 2882 ext4_msg(sb, KERN_ERR,
2883 "can't read group descriptor %d", i);
ac27a0ec
DK		 2884 db_count = i;
2885 goto failed_mount2;
2886 }
2887 }
2b2d6d01 2888 if (!ext4_check_descriptors(sb)) {
b31e1552 2889 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
ac27a0ec
DK		 2890 goto failed_mount2;
2891 }
772cb7c8
JS		 2892 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2893 if (!ext4_fill_flex_info(sb)) {
b31e1552
ES		 2894 ext4_msg(sb, KERN_ERR,
2895 "unable to initialize "
2896 "flex_bg meta info!");
772cb7c8
JS		 2897 goto failed_mount2;
2898 }
2899
ac27a0ec
DK		 2900 sbi->s_gdb_count = db_count;
2901 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2902 spin_lock_init(&sbi->s_next_gen_lock);
2903
c9de560d 2904 sbi->s_stripe = ext4_get_stripe_size(sbi);
55138e0b 2905 sbi->s_max_writeback_mb_bump = 128;
c9de560d 2906
ac27a0ec
DK		 2907 /*
2908 * set up enough so that it can read an inode
2909 */
9ca92389
TT		 2910 if (!test_opt(sb, NOLOAD) &&
2911 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
2912 sb->s_op = &ext4_sops;
2913 else
2914 sb->s_op = &ext4_nojournal_sops;
617ba13b
MC		 2915 sb->s_export_op = &ext4_export_ops;
2916 sb->s_xattr = ext4_xattr_handlers;
ac27a0ec 2917#ifdef CONFIG_QUOTA
617ba13b
MC		 2918 sb->s_qcop = &ext4_qctl_operations;
2919 sb->dq_op = &ext4_quota_operations;
ac27a0ec
DK		 2920#endif
2921 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3b9d4ed2 2922 mutex_init(&sbi->s_orphan_lock);
32ed5058 2923 mutex_init(&sbi->s_resize_lock);
ac27a0ec
DK		 2924
2925 sb->s_root = NULL;
2926
2927 needs_recovery = (es->s_last_orphan != 0 ||
617ba13b
MC		 2928 EXT4_HAS_INCOMPAT_FEATURE(sb,
2929 EXT4_FEATURE_INCOMPAT_RECOVER));
ac27a0ec
DK		 2930
2931 /*
2932 * The first inode we look at is the journal inode. Don't try
2933 * root first: it may be modified in the journal!
2934 */
2935 if (!test_opt(sb, NOLOAD) &&
617ba13b
MC		 2936 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2937 if (ext4_load_journal(sb, es, journal_devnum))
ac27a0ec 2938 goto failed_mount3;
0390131b
FM		 2939 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
2940 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
b31e1552
ES		 2941 ext4_msg(sb, KERN_ERR, "required journal recovery "
2942 "suppressed and not mounted read-only");
744692dc 2943 goto failed_mount_wq;
ac27a0ec 2944 } else {
0390131b
FM		 2945 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
2946 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
2947 sbi->s_journal = NULL;
2948 needs_recovery = 0;
2949 goto no_journal;
ac27a0ec
DK		 2950 }
2951
eb40a09c
JS		 2952 if (ext4_blocks_count(es) > 0xffffffffULL &&
2953 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2954 JBD2_FEATURE_INCOMPAT_64BIT)) {
b31e1552 2955 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
744692dc 2956 goto failed_mount_wq;
eb40a09c
JS		 2957 }
2958
d4da6c9c
LT		 2959 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2960 jbd2_journal_set_features(sbi->s_journal,
2961 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
818d276c 2962 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
d4da6c9c
LT		 2963 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2964 jbd2_journal_set_features(sbi->s_journal,
2965 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
818d276c
GS		 2966 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2967 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
d4da6c9c
LT		 2968 } else {
2969 jbd2_journal_clear_features(sbi->s_journal,
2970 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2971 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2972 }
818d276c 2973
ac27a0ec
DK		 2974 /* We have now updated the journal if required, so we can
2975 * validate the data journaling mode. */
2976 switch (test_opt(sb, DATA_FLAGS)) {
2977 case 0:
2978 /* No mode set, assume a default based on the journal
63f57933
AM		 2979 * capabilities: ORDERED_DATA if the journal can
2980 * cope, else JOURNAL_DATA
2981 */
dab291af
MC		 2982 if (jbd2_journal_check_available_features
2983 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
ac27a0ec
DK		 2984 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2985 else
2986 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2987 break;
2988
617ba13b
MC		 2989 case EXT4_MOUNT_ORDERED_DATA:
2990 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 2991 if (!jbd2_journal_check_available_features
2992 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
b31e1552
ES		 2993 ext4_msg(sb, KERN_ERR, "Journal does not support "
2994 "requested data journaling mode");
744692dc 2995 goto failed_mount_wq;
ac27a0ec
DK		 2996 }
2997 default:
2998 break;
2999 }
b3881f74 3000 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
ac27a0ec 3001
0390131b 3002no_journal:
84061e07
DM		 3003 err = percpu_counter_init(&sbi->s_freeblocks_counter,
3004 ext4_count_free_blocks(sb));
3005 if (!err)
3006 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3007 ext4_count_free_inodes(sb));
3008 if (!err)
3009 err = percpu_counter_init(&sbi->s_dirs_counter,
3010 ext4_count_dirs(sb));
3011 if (!err)
3012 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
3013 if (err) {
3014 ext4_msg(sb, KERN_ERR, "insufficient memory");
3015 goto failed_mount_wq;
3016 }
206f7ab4 3017
4c0425ff
MC		 3018 EXT4_SB(sb)->dio_unwritten_wq = create_workqueue("ext4-dio-unwritten");
3019 if (!EXT4_SB(sb)->dio_unwritten_wq) {
3020 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
3021 goto failed_mount_wq;
3022 }
3023
ac27a0ec 3024 /*
dab291af 3025 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 3026 * so we can safely mount the rest of the filesystem now.
3027 */
3028
1d1fe1ee
DH		 3029 root = ext4_iget(sb, EXT4_ROOT_INO);
3030 if (IS_ERR(root)) {
b31e1552 3031 ext4_msg(sb, KERN_ERR, "get root inode failed");
1d1fe1ee 3032 ret = PTR_ERR(root);
ac27a0ec
DK		 3033 goto failed_mount4;
3034 }
3035 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1d1fe1ee 3036 iput(root);
b31e1552 3037 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
ac27a0ec
DK		 3038 goto failed_mount4;
3039 }
1d1fe1ee
DH		 3040 sb->s_root = d_alloc_root(root);
3041 if (!sb->s_root) {
b31e1552 3042 ext4_msg(sb, KERN_ERR, "get root dentry failed");
1d1fe1ee
DH		 3043 iput(root);
3044 ret = -ENOMEM;
3045 goto failed_mount4;
3046 }
ac27a0ec 3047
2b2d6d01 3048 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
ef7f3835
KS		 3049
3050 /* determine the minimum size of new large inodes, if present */
3051 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3052 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3053 EXT4_GOOD_OLD_INODE_SIZE;
3054 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3055 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
3056 if (sbi->s_want_extra_isize <
3057 le16_to_cpu(es->s_want_extra_isize))
3058 sbi->s_want_extra_isize =
3059 le16_to_cpu(es->s_want_extra_isize);
3060 if (sbi->s_want_extra_isize <
3061 le16_to_cpu(es->s_min_extra_isize))
3062 sbi->s_want_extra_isize =
3063 le16_to_cpu(es->s_min_extra_isize);
3064 }
3065 }
3066 /* Check if enough inode space is available */
3067 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3068 sbi->s_inode_size) {
3069 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3070 EXT4_GOOD_OLD_INODE_SIZE;
b31e1552
ES		 3071 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3072 "available");
ef7f3835
KS		 3073 }
3074
90576c0b
TT		 3075 if (test_opt(sb, DELALLOC) &&
3076 (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)) {
b31e1552
ES		 3077 ext4_msg(sb, KERN_WARNING, "Ignoring delalloc option - "
3078 "requested data journaling mode");
c2774d84 3079 clear_opt(sbi->s_mount_opt, DELALLOC);
90576c0b 3080 }
744692dc
JZ		 3081 if (test_opt(sb, DIOREAD_NOLOCK)) {
3082 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3083 ext4_msg(sb, KERN_WARNING, "Ignoring dioread_nolock "
3084 "option - requested data journaling mode");
3085 clear_opt(sbi->s_mount_opt, DIOREAD_NOLOCK);
3086 }
3087 if (sb->s_blocksize < PAGE_SIZE) {
3088 ext4_msg(sb, KERN_WARNING, "Ignoring dioread_nolock "
3089 "option - block size is too small");
3090 clear_opt(sbi->s_mount_opt, DIOREAD_NOLOCK);
3091 }
3092 }
c2774d84 3093
6fd058f7
TT		 3094 err = ext4_setup_system_zone(sb);
3095 if (err) {
b31e1552 3096 ext4_msg(sb, KERN_ERR, "failed to initialize system "
fbe845dd 3097 "zone (%d)", err);
6fd058f7
TT		 3098 goto failed_mount4;
3099 }
3100
c2774d84
AK		 3101 ext4_ext_init(sb);
3102 err = ext4_mb_init(sb, needs_recovery);
3103 if (err) {
b31e1552
ES		 3104 ext4_msg(sb, KERN_ERR, "failed to initalize mballoc (%d)",
3105 err);
c2774d84
AK		 3106 goto failed_mount4;
3107 }
3108
3197ebdb
TT		 3109 sbi->s_kobj.kset = ext4_kset;
3110 init_completion(&sbi->s_kobj_unregister);
3111 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
3112 "%s", sb->s_id);
3113 if (err) {
3114 ext4_mb_release(sb);
3115 ext4_ext_release(sb);
3116 goto failed_mount4;
3117 };
3118
617ba13b
MC		 3119 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3120 ext4_orphan_cleanup(sb, es);
3121 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
0390131b 3122 if (needs_recovery) {
b31e1552 3123 ext4_msg(sb, KERN_INFO, "recovery complete");
0390131b
FM		 3124 ext4_mark_recovery_complete(sb, es);
3125 }
3126 if (EXT4_SB(sb)->s_journal) {
3127 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3128 descr = " journalled data mode";
3129 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3130 descr = " ordered data mode";
3131 else
3132 descr = " writeback data mode";
3133 } else
3134 descr = "out journal";
3135
d4c402d9
CW		 3136 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
3137 "Opts: %s", descr, orig_data);
ac27a0ec 3138
66e61a9e
TT		 3139 init_timer(&sbi->s_err_report);
3140 sbi->s_err_report.function = print_daily_error_info;
3141 sbi->s_err_report.data = (unsigned long) sb;
3142 if (es->s_error_count)
3143 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
3144
ac27a0ec 3145 lock_kernel();
d4c402d9 3146 kfree(orig_data);
ac27a0ec
DK		 3147 return 0;
3148
617ba13b 3149cantfind_ext4:
ac27a0ec 3150 if (!silent)
b31e1552 3151 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
ac27a0ec
DK		 3152 goto failed_mount;
3153
3154failed_mount4:
b31e1552 3155 ext4_msg(sb, KERN_ERR, "mount failed");
4c0425ff
MC		 3156 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
3157failed_mount_wq:
6fd058f7 3158 ext4_release_system_zone(sb);
0390131b
FM		 3159 if (sbi->s_journal) {
3160 jbd2_journal_destroy(sbi->s_journal);
3161 sbi->s_journal = NULL;
3162 }
84061e07
DM		 3163 percpu_counter_destroy(&sbi->s_freeblocks_counter);
3164 percpu_counter_destroy(&sbi->s_freeinodes_counter);
3165 percpu_counter_destroy(&sbi->s_dirs_counter);
3166 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
ac27a0ec 3167failed_mount3:
c5ca7c76
TT		 3168 if (sbi->s_flex_groups) {
3169 if (is_vmalloc_addr(sbi->s_flex_groups))
3170 vfree(sbi->s_flex_groups);
3171 else
3172 kfree(sbi->s_flex_groups);
3173 }
ac27a0ec
DK		 3174failed_mount2:
3175 for (i = 0; i < db_count; i++)
3176 brelse(sbi->s_group_desc[i]);
3177 kfree(sbi->s_group_desc);
3178failed_mount:
240799cd 3179 if (sbi->s_proc) {
9f6200bb 3180 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 3181 }
ac27a0ec
DK		 3182#ifdef CONFIG_QUOTA
3183 for (i = 0; i < MAXQUOTAS; i++)
3184 kfree(sbi->s_qf_names[i]);
3185#endif
617ba13b 3186 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 3187 brelse(bh);
3188out_fail:
3189 sb->s_fs_info = NULL;
f6830165 3190 kfree(sbi->s_blockgroup_lock);
ac27a0ec
DK		 3191 kfree(sbi);
3192 lock_kernel();
d4c402d9 3193 kfree(orig_data);
1d1fe1ee 3194 return ret;
ac27a0ec
DK		 3195}
3196
3197/*
3198 * Setup any per-fs journal parameters now. We'll do this both on
3199 * initial mount, once the journal has been initialised but before we've
3200 * done any recovery; and again on any subsequent remount.
3201 */
617ba13b 3202static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 3203{
617ba13b 3204 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 3205
30773840
TT		 3206 journal->j_commit_interval = sbi->s_commit_interval;
3207 journal->j_min_batch_time = sbi->s_min_batch_time;
3208 journal->j_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 3209
3210 spin_lock(&journal->j_state_lock);
3211 if (test_opt(sb, BARRIER))
dab291af 3212 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 3213 else
dab291af 3214 journal->j_flags &= ~JBD2_BARRIER;
5bf5683a
HK		 3215 if (test_opt(sb, DATA_ERR_ABORT))
3216 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
3217 else
3218 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
ac27a0ec
DK		 3219 spin_unlock(&journal->j_state_lock);
3220}
3221
617ba13b 3222static journal_t *ext4_get_journal(struct super_block *sb,
ac27a0ec
DK		 3223 unsigned int journal_inum)
3224{
3225 struct inode *journal_inode;
3226 journal_t *journal;
3227
0390131b
FM		 3228 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3229
ac27a0ec
DK		 3230 /* First, test for the existence of a valid inode on disk. Bad
3231 * things happen if we iget() an unused inode, as the subsequent
3232 * iput() will try to delete it. */
3233
1d1fe1ee
DH		 3234 journal_inode = ext4_iget(sb, journal_inum);
3235 if (IS_ERR(journal_inode)) {
b31e1552 3236 ext4_msg(sb, KERN_ERR, "no journal found");
ac27a0ec
DK		 3237 return NULL;
3238 }
3239 if (!journal_inode->i_nlink) {
3240 make_bad_inode(journal_inode);
3241 iput(journal_inode);
b31e1552 3242 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
ac27a0ec
DK		 3243 return NULL;
3244 }
3245
e5f8eab8 3246 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
ac27a0ec 3247 journal_inode, journal_inode->i_size);
1d1fe1ee 3248 if (!S_ISREG(journal_inode->i_mode)) {
b31e1552 3249 ext4_msg(sb, KERN_ERR, "invalid journal inode");
ac27a0ec
DK		 3250 iput(journal_inode);
3251 return NULL;
3252 }
3253
dab291af 3254 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 3255 if (!journal) {
b31e1552 3256 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
ac27a0ec
DK		 3257 iput(journal_inode);
3258 return NULL;
3259 }
3260 journal->j_private = sb;
617ba13b 3261 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 3262 return journal;
3263}
3264
617ba13b 3265static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 3266 dev_t j_dev)
3267{
2b2d6d01 3268 struct buffer_head *bh;
ac27a0ec 3269 journal_t *journal;
617ba13b
MC		 3270 ext4_fsblk_t start;
3271 ext4_fsblk_t len;
ac27a0ec 3272 int hblock, blocksize;
617ba13b 3273 ext4_fsblk_t sb_block;
ac27a0ec 3274 unsigned long offset;
2b2d6d01 3275 struct ext4_super_block *es;
ac27a0ec
DK		 3276 struct block_device *bdev;
3277
0390131b
FM		 3278 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3279
b31e1552 3280 bdev = ext4_blkdev_get(j_dev, sb);
ac27a0ec
DK		 3281 if (bdev == NULL)
3282 return NULL;
3283
3284 if (bd_claim(bdev, sb)) {
b31e1552
ES		 3285 ext4_msg(sb, KERN_ERR,
3286 "failed to claim external journal device");
9a1c3542 3287 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
ac27a0ec
DK		 3288 return NULL;
3289 }
3290
3291 blocksize = sb->s_blocksize;
e1defc4f 3292 hblock = bdev_logical_block_size(bdev);
ac27a0ec 3293 if (blocksize < hblock) {
b31e1552
ES		 3294 ext4_msg(sb, KERN_ERR,
3295 "blocksize too small for journal device");
ac27a0ec
DK		 3296 goto out_bdev;
3297 }
3298
617ba13b
MC		 3299 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3300 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 3301 set_blocksize(bdev, blocksize);
3302 if (!(bh = __bread(bdev, sb_block, blocksize))) {
b31e1552
ES		 3303 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
3304 "external journal");
ac27a0ec
DK		 3305 goto out_bdev;
3306 }
3307
617ba13b
MC		 3308 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3309 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 3310 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b 3311 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
b31e1552
ES		 3312 ext4_msg(sb, KERN_ERR, "external journal has "
3313 "bad superblock");
ac27a0ec
DK		 3314 brelse(bh);
3315 goto out_bdev;
3316 }
3317
617ba13b 3318 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
b31e1552 3319 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
ac27a0ec
DK		 3320 brelse(bh);
3321 goto out_bdev;
3322 }
3323
bd81d8ee 3324 len = ext4_blocks_count(es);
ac27a0ec
DK		 3325 start = sb_block + 1;
3326 brelse(bh); /* we're done with the superblock */
3327
dab291af 3328 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 3329 start, len, blocksize);
3330 if (!journal) {
b31e1552 3331 ext4_msg(sb, KERN_ERR, "failed to create device journal");
ac27a0ec
DK		 3332 goto out_bdev;
3333 }
3334 journal->j_private = sb;
3335 ll_rw_block(READ, 1, &journal->j_sb_buffer);
3336 wait_on_buffer(journal->j_sb_buffer);
3337 if (!buffer_uptodate(journal->j_sb_buffer)) {
b31e1552 3338 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
ac27a0ec
DK		 3339 goto out_journal;
3340 }
3341 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
b31e1552
ES		 3342 ext4_msg(sb, KERN_ERR, "External journal has more than one "
3343 "user (unsupported) - %d",
ac27a0ec
DK		 3344 be32_to_cpu(journal->j_superblock->s_nr_users));
3345 goto out_journal;
3346 }
617ba13b
MC		 3347 EXT4_SB(sb)->journal_bdev = bdev;
3348 ext4_init_journal_params(sb, journal);
ac27a0ec 3349 return journal;
0b8e58a1 3350
ac27a0ec 3351out_journal:
dab291af 3352 jbd2_journal_destroy(journal);
ac27a0ec 3353out_bdev:
617ba13b 3354 ext4_blkdev_put(bdev);
ac27a0ec
DK		 3355 return NULL;
3356}
3357
617ba13b
MC		 3358static int ext4_load_journal(struct super_block *sb,
3359 struct ext4_super_block *es,
ac27a0ec
DK		 3360 unsigned long journal_devnum)
3361{
3362 journal_t *journal;
3363 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
3364 dev_t journal_dev;
3365 int err = 0;
3366 int really_read_only;
3367
0390131b
FM		 3368 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3369
ac27a0ec
DK		 3370 if (journal_devnum &&
3371 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
b31e1552
ES		 3372 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
3373 "numbers have changed");
ac27a0ec
DK		 3374 journal_dev = new_decode_dev(journal_devnum);
3375 } else
3376 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
3377
3378 really_read_only = bdev_read_only(sb->s_bdev);
3379
3380 /*
3381 * Are we loading a blank journal or performing recovery after a
3382 * crash? For recovery, we need to check in advance whether we
3383 * can get read-write access to the device.
3384 */
617ba13b 3385 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
ac27a0ec 3386 if (sb->s_flags & MS_RDONLY) {
b31e1552
ES		 3387 ext4_msg(sb, KERN_INFO, "INFO: recovery "
3388 "required on readonly filesystem");
ac27a0ec 3389 if (really_read_only) {
b31e1552
ES		 3390 ext4_msg(sb, KERN_ERR, "write access "
3391 "unavailable, cannot proceed");
ac27a0ec
DK		 3392 return -EROFS;
3393 }
b31e1552
ES		 3394 ext4_msg(sb, KERN_INFO, "write access will "
3395 "be enabled during recovery");
ac27a0ec
DK		 3396 }
3397 }
3398
3399 if (journal_inum && journal_dev) {
b31e1552
ES		 3400 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
3401 "and inode journals!");
ac27a0ec
DK		 3402 return -EINVAL;
3403 }
3404
3405 if (journal_inum) {
617ba13b 3406 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 3407 return -EINVAL;
3408 } else {
617ba13b 3409 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 3410 return -EINVAL;
3411 }
3412
90576c0b 3413 if (!(journal->j_flags & JBD2_BARRIER))
b31e1552 3414 ext4_msg(sb, KERN_INFO, "barriers disabled");
4776004f 3415
ac27a0ec 3416 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
dab291af 3417 err = jbd2_journal_update_format(journal);
ac27a0ec 3418 if (err) {
b31e1552 3419 ext4_msg(sb, KERN_ERR, "error updating journal");
dab291af 3420 jbd2_journal_destroy(journal);
ac27a0ec
DK		 3421 return err;
3422 }
3423 }
3424
617ba13b 3425 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
dab291af 3426 err = jbd2_journal_wipe(journal, !really_read_only);
1c13d5c0
TT		 3427 if (!err) {
3428 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
3429 if (save)
3430 memcpy(save, ((char *) es) +
3431 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
dab291af 3432 err = jbd2_journal_load(journal);
1c13d5c0
TT		 3433 if (save)
3434 memcpy(((char *) es) + EXT4_S_ERR_START,
3435 save, EXT4_S_ERR_LEN);
3436 kfree(save);
3437 }
ac27a0ec
DK		 3438
3439 if (err) {
b31e1552 3440 ext4_msg(sb, KERN_ERR, "error loading journal");
dab291af 3441 jbd2_journal_destroy(journal);
ac27a0ec
DK		 3442 return err;
3443 }
3444
617ba13b
MC		 3445 EXT4_SB(sb)->s_journal = journal;
3446 ext4_clear_journal_err(sb, es);
ac27a0ec
DK		 3447
3448 if (journal_devnum &&
3449 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3450 es->s_journal_dev = cpu_to_le32(journal_devnum);
ac27a0ec
DK		 3451
3452 /* Make sure we flush the recovery flag to disk. */
e2d67052 3453 ext4_commit_super(sb, 1);
ac27a0ec
DK		 3454 }
3455
3456 return 0;
3457}
3458
e2d67052 3459static int ext4_commit_super(struct super_block *sb, int sync)
ac27a0ec 3460{
e2d67052 3461 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
617ba13b 3462 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
c4be0c1d 3463 int error = 0;
ac27a0ec
DK		 3464
3465 if (!sbh)
c4be0c1d 3466 return error;
914258bf
TT		 3467 if (buffer_write_io_error(sbh)) {
3468 /*
3469 * Oh, dear. A previous attempt to write the
3470 * superblock failed. This could happen because the
3471 * USB device was yanked out. Or it could happen to
3472 * be a transient write error and maybe the block will
3473 * be remapped. Nothing we can do but to retry the
3474 * write and hope for the best.
3475 */
b31e1552
ES		 3476 ext4_msg(sb, KERN_ERR, "previous I/O error to "
3477 "superblock detected");
914258bf
TT		 3478 clear_buffer_write_io_error(sbh);
3479 set_buffer_uptodate(sbh);
3480 }
71290b36
TT		 3481 /*
3482 * If the file system is mounted read-only, don't update the
3483 * superblock write time. This avoids updating the superblock
3484 * write time when we are mounting the root file system
3485 * read/only but we need to replay the journal; at that point,
3486 * for people who are east of GMT and who make their clock
3487 * tick in localtime for Windows bug-for-bug compatibility,
3488 * the clock is set in the future, and this will cause e2fsck
3489 * to complain and force a full file system check.
3490 */
3491 if (!(sb->s_flags & MS_RDONLY))
3492 es->s_wtime = cpu_to_le32(get_seconds());
afc32f7e 3493 es->s_kbytes_written =
60e6679e 3494 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
afc32f7e
TT		 3495 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
3496 EXT4_SB(sb)->s_sectors_written_start) >> 1));
5d1b1b3f
AK		 3497 ext4_free_blocks_count_set(es, percpu_counter_sum_positive(
3498 &EXT4_SB(sb)->s_freeblocks_counter));
3499 es->s_free_inodes_count = cpu_to_le32(percpu_counter_sum_positive(
3500 &EXT4_SB(sb)->s_freeinodes_counter));
7234ab2a 3501 sb->s_dirt = 0;
ac27a0ec
DK		 3502 BUFFER_TRACE(sbh, "marking dirty");
3503 mark_buffer_dirty(sbh);
914258bf 3504 if (sync) {
c4be0c1d
TS		 3505 error = sync_dirty_buffer(sbh);
3506 if (error)
3507 return error;
3508
3509 error = buffer_write_io_error(sbh);
3510 if (error) {
b31e1552
ES		 3511 ext4_msg(sb, KERN_ERR, "I/O error while writing "
3512 "superblock");
914258bf
TT		 3513 clear_buffer_write_io_error(sbh);
3514 set_buffer_uptodate(sbh);
3515 }
3516 }
c4be0c1d 3517 return error;
ac27a0ec
DK		 3518}
3519
ac27a0ec
DK		 3520/*
3521 * Have we just finished recovery? If so, and if we are mounting (or
3522 * remounting) the filesystem readonly, then we will end up with a
3523 * consistent fs on disk. Record that fact.
3524 */
2b2d6d01
TT		 3525static void ext4_mark_recovery_complete(struct super_block *sb,
3526 struct ext4_super_block *es)
ac27a0ec 3527{
617ba13b 3528 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 3529
0390131b
FM		 3530 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3531 BUG_ON(journal != NULL);
3532 return;
3533 }
dab291af 3534 jbd2_journal_lock_updates(journal);
7ffe1ea8
HK		 3535 if (jbd2_journal_flush(journal) < 0)
3536 goto out;
3537
617ba13b 3538 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
ac27a0ec 3539 sb->s_flags & MS_RDONLY) {
617ba13b 3540 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
e2d67052 3541 ext4_commit_super(sb, 1);
ac27a0ec 3542 }
7ffe1ea8
HK		 3543
3544out:
dab291af 3545 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 3546}
3547
3548/*
3549 * If we are mounting (or read-write remounting) a filesystem whose journal
3550 * has recorded an error from a previous lifetime, move that error to the
3551 * main filesystem now.
3552 */
2b2d6d01
TT		 3553static void ext4_clear_journal_err(struct super_block *sb,
3554 struct ext4_super_block *es)
ac27a0ec
DK		 3555{
3556 journal_t *journal;
3557 int j_errno;
3558 const char *errstr;
3559
0390131b
FM		 3560 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3561
617ba13b 3562 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 3563
3564 /*
3565 * Now check for any error status which may have been recorded in the
617ba13b 3566 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 3567 */
3568
dab291af 3569 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 3570 if (j_errno) {
3571 char nbuf[16];
3572
617ba13b 3573 errstr = ext4_decode_error(sb, j_errno, nbuf);
12062ddd 3574 ext4_warning(sb, "Filesystem error recorded "
ac27a0ec 3575 "from previous mount: %s", errstr);
12062ddd 3576 ext4_warning(sb, "Marking fs in need of filesystem check.");
ac27a0ec 3577
617ba13b
MC		 3578 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
3579 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
e2d67052 3580 ext4_commit_super(sb, 1);
ac27a0ec 3581
dab291af 3582 jbd2_journal_clear_err(journal);
ac27a0ec
DK		 3583 }
3584}
3585
3586/*
3587 * Force the running and committing transactions to commit,
3588 * and wait on the commit.
3589 */
617ba13b 3590int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 3591{
3592 journal_t *journal;
0390131b 3593 int ret = 0;
ac27a0ec
DK		 3594
3595 if (sb->s_flags & MS_RDONLY)
3596 return 0;
3597
617ba13b 3598 journal = EXT4_SB(sb)->s_journal;
6b0310fb
ES		 3599 if (journal) {
3600 vfs_check_frozen(sb, SB_FREEZE_WRITE);
0390131b 3601 ret = ext4_journal_force_commit(journal);
6b0310fb 3602 }
0390131b 3603
ac27a0ec
DK		 3604 return ret;
3605}
3606
2b2d6d01 3607static void ext4_write_super(struct super_block *sb)
ac27a0ec 3608{
ebc1ac16 3609 lock_super(sb);
9ca92389 3610 ext4_commit_super(sb, 1);
ebc1ac16 3611 unlock_super(sb);
ac27a0ec
DK		 3612}
3613
617ba13b 3614static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec 3615{
14ce0cb4 3616 int ret = 0;
9eddacf9 3617 tid_t target;
8d5d02e6 3618 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 3619
9bffad1e 3620 trace_ext4_sync_fs(sb, wait);
8d5d02e6
MC		 3621 flush_workqueue(sbi->dio_unwritten_wq);
3622 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
9ca92389 3623 if (wait)
8d5d02e6 3624 jbd2_log_wait_commit(sbi->s_journal, target);
0390131b 3625 }
14ce0cb4 3626 return ret;
ac27a0ec
DK		 3627}
3628
3629/*
3630 * LVM calls this function before a (read-only) snapshot is created. This
3631 * gives us a chance to flush the journal completely and mark the fs clean.
3632 */
c4be0c1d 3633static int ext4_freeze(struct super_block *sb)
ac27a0ec 3634{
c4be0c1d
TS		 3635 int error = 0;
3636 journal_t *journal;
ac27a0ec 3637
9ca92389
TT		 3638 if (sb->s_flags & MS_RDONLY)
3639 return 0;
ac27a0ec 3640
9ca92389 3641 journal = EXT4_SB(sb)->s_journal;
7ffe1ea8 3642
9ca92389
TT		 3643 /* Now we set up the journal barrier. */
3644 jbd2_journal_lock_updates(journal);
ac27a0ec 3645
9ca92389
TT		 3646 /*
3647 * Don't clear the needs_recovery flag if we failed to flush
3648 * the journal.
3649 */
3650 error = jbd2_journal_flush(journal);
6b0310fb
ES		 3651 if (error < 0)
3652 goto out;
9ca92389
TT		 3653
3654 /* Journal blocked and flushed, clear needs_recovery flag. */
3655 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3656 error = ext4_commit_super(sb, 1);
6b0310fb
ES		 3657out:
3658 /* we rely on s_frozen to stop further updates */
3659 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
3660 return error;
ac27a0ec
DK		 3661}
3662
3663/*
3664 * Called by LVM after the snapshot is done. We need to reset the RECOVER
3665 * flag here, even though the filesystem is not technically dirty yet.
3666 */
c4be0c1d 3667static int ext4_unfreeze(struct super_block *sb)
ac27a0ec 3668{
9ca92389
TT		 3669 if (sb->s_flags & MS_RDONLY)
3670 return 0;
3671
3672 lock_super(sb);
3673 /* Reset the needs_recovery flag before the fs is unlocked. */
3674 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3675 ext4_commit_super(sb, 1);
3676 unlock_super(sb);
c4be0c1d 3677 return 0;
ac27a0ec
DK		 3678}
3679
2b2d6d01 3680static int ext4_remount(struct super_block *sb, int *flags, char *data)
ac27a0ec 3681{
2b2d6d01 3682 struct ext4_super_block *es;
617ba13b
MC		 3683 struct ext4_sb_info *sbi = EXT4_SB(sb);
3684 ext4_fsblk_t n_blocks_count = 0;
ac27a0ec 3685 unsigned long old_sb_flags;
617ba13b 3686 struct ext4_mount_options old_opts;
c79d967d 3687 int enable_quota = 0;
8a266467 3688 ext4_group_t g;
b3881f74 3689 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
ac27a0ec
DK		 3690 int err;
3691#ifdef CONFIG_QUOTA
3692 int i;
3693#endif
d4c402d9 3694 char *orig_data = kstrdup(data, GFP_KERNEL);
ac27a0ec 3695
337eb00a
AIB		 3696 lock_kernel();
3697
ac27a0ec 3698 /* Store the original options */
bbd6851a 3699 lock_super(sb);
ac27a0ec
DK		 3700 old_sb_flags = sb->s_flags;
3701 old_opts.s_mount_opt = sbi->s_mount_opt;
3702 old_opts.s_resuid = sbi->s_resuid;
3703 old_opts.s_resgid = sbi->s_resgid;
3704 old_opts.s_commit_interval = sbi->s_commit_interval;
30773840
TT		 3705 old_opts.s_min_batch_time = sbi->s_min_batch_time;
3706 old_opts.s_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 3707#ifdef CONFIG_QUOTA
3708 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
3709 for (i = 0; i < MAXQUOTAS; i++)
3710 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
3711#endif
b3881f74
TT		 3712 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
3713 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
ac27a0ec
DK		 3714
3715 /*
3716 * Allow the "check" option to be passed as a remount option.
3717 */
b3881f74
TT		 3718 if (!parse_options(data, sb, NULL, &journal_ioprio,
3719 &n_blocks_count, 1)) {
ac27a0ec
DK		 3720 err = -EINVAL;
3721 goto restore_opts;
3722 }
3723
4ab2f15b 3724 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
c67d859e 3725 ext4_abort(sb, "Abort forced by user");
ac27a0ec
DK		 3726
3727 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 3728 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec
DK		 3729
3730 es = sbi->s_es;
3731
b3881f74 3732 if (sbi->s_journal) {
0390131b 3733 ext4_init_journal_params(sb, sbi->s_journal);
b3881f74
TT		 3734 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3735 }
ac27a0ec
DK		 3736
3737 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
bd81d8ee 3738 n_blocks_count > ext4_blocks_count(es)) {
4ab2f15b 3739 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
ac27a0ec
DK		 3740 err = -EROFS;
3741 goto restore_opts;
3742 }
3743
3744 if (*flags & MS_RDONLY) {
0f0dd62f
CH		 3745 err = dquot_suspend(sb, -1);
3746 if (err < 0)
c79d967d 3747 goto restore_opts;
c79d967d 3748
ac27a0ec
DK		 3749 /*
3750 * First of all, the unconditional stuff we have to do
3751 * to disable replay of the journal when we next remount
3752 */
3753 sb->s_flags |= MS_RDONLY;
3754
3755 /*
3756 * OK, test if we are remounting a valid rw partition
3757 * readonly, and if so set the rdonly flag and then
3758 * mark the partition as valid again.
3759 */
617ba13b
MC		 3760 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
3761 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 3762 es->s_state = cpu_to_le16(sbi->s_mount_state);
3763
a63c9eb2 3764 if (sbi->s_journal)
0390131b 3765 ext4_mark_recovery_complete(sb, es);
ac27a0ec 3766 } else {
a13fb1a4
ES		 3767 /* Make sure we can mount this feature set readwrite */
3768 if (!ext4_feature_set_ok(sb, 0)) {
ac27a0ec
DK		 3769 err = -EROFS;
3770 goto restore_opts;
3771 }
8a266467
TT		 3772 /*
3773 * Make sure the group descriptor checksums
0b8e58a1 3774 * are sane. If they aren't, refuse to remount r/w.
8a266467
TT		 3775 */
3776 for (g = 0; g < sbi->s_groups_count; g++) {
3777 struct ext4_group_desc *gdp =
3778 ext4_get_group_desc(sb, g, NULL);
3779
3780 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
b31e1552
ES		 3781 ext4_msg(sb, KERN_ERR,
3782 "ext4_remount: Checksum for group %u failed (%u!=%u)",
8a266467
TT		 3783 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
3784 le16_to_cpu(gdp->bg_checksum));
3785 err = -EINVAL;
3786 goto restore_opts;
3787 }
3788 }
3789
ead6596b
ES		 3790 /*
3791 * If we have an unprocessed orphan list hanging
3792 * around from a previously readonly bdev mount,
3793 * require a full umount/remount for now.
3794 */
3795 if (es->s_last_orphan) {
b31e1552 3796 ext4_msg(sb, KERN_WARNING, "Couldn't "
ead6596b
ES		 3797 "remount RDWR because of unprocessed "
3798 "orphan inode list. Please "
b31e1552 3799 "umount/remount instead");
ead6596b
ES		 3800 err = -EINVAL;
3801 goto restore_opts;
3802 }
3803
ac27a0ec
DK		 3804 /*
3805 * Mounting a RDONLY partition read-write, so reread
3806 * and store the current valid flag. (It may have
3807 * been changed by e2fsck since we originally mounted
3808 * the partition.)
3809 */
0390131b
FM		 3810 if (sbi->s_journal)
3811 ext4_clear_journal_err(sb, es);
ac27a0ec 3812 sbi->s_mount_state = le16_to_cpu(es->s_state);
617ba13b 3813 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
ac27a0ec 3814 goto restore_opts;
2b2d6d01 3815 if (!ext4_setup_super(sb, es, 0))
ac27a0ec 3816 sb->s_flags &= ~MS_RDONLY;
c79d967d 3817 enable_quota = 1;
ac27a0ec
DK		 3818 }
3819 }
6fd058f7 3820 ext4_setup_system_zone(sb);
0390131b 3821 if (sbi->s_journal == NULL)
e2d67052 3822 ext4_commit_super(sb, 1);
0390131b 3823
ac27a0ec
DK		 3824#ifdef CONFIG_QUOTA
3825 /* Release old quota file names */
3826 for (i = 0; i < MAXQUOTAS; i++)
3827 if (old_opts.s_qf_names[i] &&
3828 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3829 kfree(old_opts.s_qf_names[i]);
3830#endif
bbd6851a 3831 unlock_super(sb);
337eb00a 3832 unlock_kernel();
c79d967d 3833 if (enable_quota)
0f0dd62f 3834 dquot_resume(sb, -1);
d4c402d9
CW		 3835
3836 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
3837 kfree(orig_data);
ac27a0ec 3838 return 0;
0b8e58a1 3839
ac27a0ec
DK		 3840restore_opts:
3841 sb->s_flags = old_sb_flags;
3842 sbi->s_mount_opt = old_opts.s_mount_opt;
3843 sbi->s_resuid = old_opts.s_resuid;
3844 sbi->s_resgid = old_opts.s_resgid;
3845 sbi->s_commit_interval = old_opts.s_commit_interval;
30773840
TT		 3846 sbi->s_min_batch_time = old_opts.s_min_batch_time;
3847 sbi->s_max_batch_time = old_opts.s_max_batch_time;
ac27a0ec
DK		 3848#ifdef CONFIG_QUOTA
3849 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
3850 for (i = 0; i < MAXQUOTAS; i++) {
3851 if (sbi->s_qf_names[i] &&
3852 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3853 kfree(sbi->s_qf_names[i]);
3854 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
3855 }
3856#endif
bbd6851a 3857 unlock_super(sb);
337eb00a 3858 unlock_kernel();
d4c402d9 3859 kfree(orig_data);
ac27a0ec
DK		 3860 return err;
3861}
3862
2b2d6d01 3863static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
ac27a0ec
DK		 3864{
3865 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 3866 struct ext4_sb_info *sbi = EXT4_SB(sb);
3867 struct ext4_super_block *es = sbi->s_es;
960cc398 3868 u64 fsid;
ac27a0ec 3869
5e70030d
BP		 3870 if (test_opt(sb, MINIX_DF)) {
3871 sbi->s_overhead_last = 0;
6bc9feff 3872 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
8df9675f 3873 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
5e70030d 3874 ext4_fsblk_t overhead = 0;
ac27a0ec
DK		 3875
3876 /*
5e70030d
BP		 3877 * Compute the overhead (FS structures). This is constant
3878 * for a given filesystem unless the number of block groups
3879 * changes so we cache the previous value until it does.
ac27a0ec
DK		 3880 */
3881
3882 /*
3883 * All of the blocks before first_data_block are
3884 * overhead
3885 */
3886 overhead = le32_to_cpu(es->s_first_data_block);
3887
3888 /*
3889 * Add the overhead attributed to the superblock and
3890 * block group descriptors. If the sparse superblocks
3891 * feature is turned on, then not all groups have this.
3892 */
3893 for (i = 0; i < ngroups; i++) {
617ba13b
MC		 3894 overhead += ext4_bg_has_super(sb, i) +
3895 ext4_bg_num_gdb(sb, i);
ac27a0ec
DK		 3896 cond_resched();
3897 }
3898
3899 /*
3900 * Every block group has an inode bitmap, a block
3901 * bitmap, and an inode table.
3902 */
5e70030d
BP		 3903 overhead += ngroups * (2 + sbi->s_itb_per_group);
3904 sbi->s_overhead_last = overhead;
3905 smp_wmb();
6bc9feff 3906 sbi->s_blocks_last = ext4_blocks_count(es);
ac27a0ec
DK		 3907 }
3908
617ba13b 3909 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 3910 buf->f_bsize = sb->s_blocksize;
5e70030d 3911 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
6bc6e63f
AK		 3912 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
3913 percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
bd81d8ee
LV		 3914 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3915 if (buf->f_bfree < ext4_r_blocks_count(es))
ac27a0ec
DK		 3916 buf->f_bavail = 0;
3917 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 3918 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
617ba13b 3919 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 3920 fsid = le64_to_cpup((void *)es->s_uuid) ^
3921 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3922 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3923 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
0b8e58a1 3924
ac27a0ec
DK		 3925 return 0;
3926}
3927
0b8e58a1
AD		 3928/* Helper function for writing quotas on sync - we need to start transaction
3929 * before quota file is locked for write. Otherwise the are possible deadlocks:
ac27a0ec 3930 * Process 1 Process 2
617ba13b 3931 * ext4_create() quota_sync()
a269eb18 3932 * jbd2_journal_start() write_dquot()
871a2931 3933 * dquot_initialize() down(dqio_mutex)
dab291af 3934 * down(dqio_mutex) jbd2_journal_start()
ac27a0ec
DK		 3935 *
3936 */
3937
3938#ifdef CONFIG_QUOTA
3939
3940static inline struct inode *dquot_to_inode(struct dquot *dquot)
3941{
3942 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3943}
3944
617ba13b 3945static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 3946{
3947 int ret, err;
3948 handle_t *handle;
3949 struct inode *inode;
3950
3951 inode = dquot_to_inode(dquot);
617ba13b 3952 handle = ext4_journal_start(inode,
0b8e58a1 3953 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 3954 if (IS_ERR(handle))
3955 return PTR_ERR(handle);
3956 ret = dquot_commit(dquot);
617ba13b 3957 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3958 if (!ret)
3959 ret = err;
3960 return ret;
3961}
3962
617ba13b 3963static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 3964{
3965 int ret, err;
3966 handle_t *handle;
3967
617ba13b 3968 handle = ext4_journal_start(dquot_to_inode(dquot),
0b8e58a1 3969 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 3970 if (IS_ERR(handle))
3971 return PTR_ERR(handle);
3972 ret = dquot_acquire(dquot);
617ba13b 3973 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3974 if (!ret)
3975 ret = err;
3976 return ret;
3977}
3978
617ba13b 3979static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 3980{
3981 int ret, err;
3982 handle_t *handle;
3983
617ba13b 3984 handle = ext4_journal_start(dquot_to_inode(dquot),
0b8e58a1 3985 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 3986 if (IS_ERR(handle)) {
3987 /* Release dquot anyway to avoid endless cycle in dqput() */
3988 dquot_release(dquot);
ac27a0ec 3989 return PTR_ERR(handle);
9c3013e9 3990 }
ac27a0ec 3991 ret = dquot_release(dquot);
617ba13b 3992 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3993 if (!ret)
3994 ret = err;
3995 return ret;
3996}
3997
617ba13b 3998static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec 3999{
2c8be6b2 4000 /* Are we journaling quotas? */
617ba13b
MC		 4001 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
4002 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
ac27a0ec 4003 dquot_mark_dquot_dirty(dquot);
617ba13b 4004 return ext4_write_dquot(dquot);
ac27a0ec
DK		 4005 } else {
4006 return dquot_mark_dquot_dirty(dquot);
4007 }
4008}
4009
617ba13b 4010static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 4011{
4012 int ret, err;
4013 handle_t *handle;
4014
4015 /* Data block + inode block */
617ba13b 4016 handle = ext4_journal_start(sb->s_root->d_inode, 2);
ac27a0ec
DK		 4017 if (IS_ERR(handle))
4018 return PTR_ERR(handle);
4019 ret = dquot_commit_info(sb, type);
617ba13b 4020 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4021 if (!ret)
4022 ret = err;
4023 return ret;
4024}
4025
4026/*
4027 * Turn on quotas during mount time - we need to find
4028 * the quota file and such...
4029 */
617ba13b 4030static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 4031{
287a8095
CH		 4032 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4033 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 4034}
4035
4036/*
4037 * Standard function to be called on quota_on
4038 */
617ba13b 4039static int ext4_quota_on(struct super_block *sb, int type, int format_id,
307ae18a 4040 char *name)
ac27a0ec
DK		 4041{
4042 int err;
8264613d 4043 struct path path;
ac27a0ec
DK		 4044
4045 if (!test_opt(sb, QUOTA))
4046 return -EINVAL;
0623543b 4047
8264613d 4048 err = kern_path(name, LOOKUP_FOLLOW, &path);
ac27a0ec
DK		 4049 if (err)
4050 return err;
0623543b 4051
ac27a0ec 4052 /* Quotafile not on the same filesystem? */
8264613d
AV		 4053 if (path.mnt->mnt_sb != sb) {
4054 path_put(&path);
ac27a0ec
DK		 4055 return -EXDEV;
4056 }
0623543b
JK		 4057 /* Journaling quota? */
4058 if (EXT4_SB(sb)->s_qf_names[type]) {
2b2d6d01 4059 /* Quotafile not in fs root? */
8264613d 4060 if (path.dentry->d_parent != sb->s_root)
b31e1552
ES		 4061 ext4_msg(sb, KERN_WARNING,
4062 "Quota file not on filesystem root. "
4063 "Journaled quota will not work");
2b2d6d01 4064 }
0623543b
JK		 4065
4066 /*
4067 * When we journal data on quota file, we have to flush journal to see
4068 * all updates to the file when we bypass pagecache...
4069 */
0390131b
FM		 4070 if (EXT4_SB(sb)->s_journal &&
4071 ext4_should_journal_data(path.dentry->d_inode)) {
0623543b
JK		 4072 /*
4073 * We don't need to lock updates but journal_flush() could
4074 * otherwise be livelocked...
4075 */
4076 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 4077 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
0623543b 4078 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 4079 if (err) {
8264613d 4080 path_put(&path);
7ffe1ea8
HK		 4081 return err;
4082 }
0623543b
JK		 4083 }
4084
287a8095 4085 err = dquot_quota_on_path(sb, type, format_id, &path);
8264613d 4086 path_put(&path);
77e69dac 4087 return err;
ac27a0ec
DK		 4088}
4089
4090/* Read data from quotafile - avoid pagecache and such because we cannot afford
4091 * acquiring the locks... As quota files are never truncated and quota code
4092 * itself serializes the operations (and noone else should touch the files)
4093 * we don't have to be afraid of races */
617ba13b 4094static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 4095 size_t len, loff_t off)
4096{
4097 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 4098 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 4099 int err = 0;
4100 int offset = off & (sb->s_blocksize - 1);
4101 int tocopy;
4102 size_t toread;
4103 struct buffer_head *bh;
4104 loff_t i_size = i_size_read(inode);
4105
4106 if (off > i_size)
4107 return 0;
4108 if (off+len > i_size)
4109 len = i_size-off;
4110 toread = len;
4111 while (toread > 0) {
4112 tocopy = sb->s_blocksize - offset < toread ?
4113 sb->s_blocksize - offset : toread;
617ba13b 4114 bh = ext4_bread(NULL, inode, blk, 0, &err);
ac27a0ec
DK		 4115 if (err)
4116 return err;
4117 if (!bh) /* A hole? */
4118 memset(data, 0, tocopy);
4119 else
4120 memcpy(data, bh->b_data+offset, tocopy);
4121 brelse(bh);
4122 offset = 0;
4123 toread -= tocopy;
4124 data += tocopy;
4125 blk++;
4126 }
4127 return len;
4128}
4129
4130/* Write to quotafile (we know the transaction is already started and has
4131 * enough credits) */
617ba13b 4132static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 4133 const char *data, size_t len, loff_t off)
4134{
4135 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 4136 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 4137 int err = 0;
4138 int offset = off & (sb->s_blocksize - 1);
617ba13b 4139 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
ac27a0ec
DK		 4140 struct buffer_head *bh;
4141 handle_t *handle = journal_current_handle();
4142
0390131b 4143 if (EXT4_SB(sb)->s_journal && !handle) {
b31e1552
ES		 4144 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4145 " cancelled because transaction is not started",
9c3013e9
JK		 4146 (unsigned long long)off, (unsigned long long)len);
4147 return -EIO;
4148 }
67eeb568
DM		 4149 /*
4150 * Since we account only one data block in transaction credits,
4151 * then it is impossible to cross a block boundary.
4152 */
4153 if (sb->s_blocksize - offset < len) {
4154 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4155 " cancelled because not block aligned",
4156 (unsigned long long)off, (unsigned long long)len);
4157 return -EIO;
4158 }
4159
ac27a0ec 4160 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
67eeb568
DM		 4161 bh = ext4_bread(handle, inode, blk, 1, &err);
4162 if (!bh)
4163 goto out;
4164 if (journal_quota) {
4165 err = ext4_journal_get_write_access(handle, bh);
4166 if (err) {
4167 brelse(bh);
ac27a0ec 4168 goto out;
ac27a0ec 4169 }
ac27a0ec 4170 }
67eeb568
DM		 4171 lock_buffer(bh);
4172 memcpy(bh->b_data+offset, data, len);
4173 flush_dcache_page(bh->b_page);
4174 unlock_buffer(bh);
4175 if (journal_quota)
4176 err = ext4_handle_dirty_metadata(handle, NULL, bh);
4177 else {
4178 /* Always do at least ordered writes for quotas */
4179 err = ext4_jbd2_file_inode(handle, inode);
4180 mark_buffer_dirty(bh);
4181 }
4182 brelse(bh);
ac27a0ec 4183out:
67eeb568 4184 if (err) {
4d04e4fb 4185 mutex_unlock(&inode->i_mutex);
ac27a0ec 4186 return err;
4d04e4fb 4187 }
67eeb568
DM		 4188 if (inode->i_size < off + len) {
4189 i_size_write(inode, off + len);
617ba13b 4190 EXT4_I(inode)->i_disksize = inode->i_size;
ac27a0ec 4191 }
ac27a0ec 4192 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
617ba13b 4193 ext4_mark_inode_dirty(handle, inode);
ac27a0ec 4194 mutex_unlock(&inode->i_mutex);
67eeb568 4195 return len;
ac27a0ec
DK		 4196}
4197
4198#endif
4199
0b8e58a1
AD		 4200static int ext4_get_sb(struct file_system_type *fs_type, int flags,
4201 const char *dev_name, void *data, struct vfsmount *mnt)
ac27a0ec 4202{
0b8e58a1 4203 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super,mnt);
ac27a0ec
DK		 4204}
4205
37f328eb 4206#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
24b58424
TT		 4207static struct file_system_type ext2_fs_type = {
4208 .owner = THIS_MODULE,
4209 .name = "ext2",
4210 .get_sb = ext4_get_sb,
4211 .kill_sb = kill_block_super,
4212 .fs_flags = FS_REQUIRES_DEV,
4213};
4214
4215static inline void register_as_ext2(void)
4216{
4217 int err = register_filesystem(&ext2_fs_type);
4218 if (err)
4219 printk(KERN_WARNING
4220 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
4221}
4222
4223static inline void unregister_as_ext2(void)
4224{
4225 unregister_filesystem(&ext2_fs_type);
4226}
51b7e3c9 4227MODULE_ALIAS("ext2");
24b58424
TT		 4228#else
4229static inline void register_as_ext2(void) { }
4230static inline void unregister_as_ext2(void) { }
4231#endif
4232
37f328eb 4233#if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
24b58424
TT		 4234static inline void register_as_ext3(void)
4235{
4236 int err = register_filesystem(&ext3_fs_type);
4237 if (err)
4238 printk(KERN_WARNING
4239 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
4240}
4241
4242static inline void unregister_as_ext3(void)
4243{
4244 unregister_filesystem(&ext3_fs_type);
4245}
51b7e3c9 4246MODULE_ALIAS("ext3");
24b58424
TT		 4247#else
4248static inline void register_as_ext3(void) { }
4249static inline void unregister_as_ext3(void) { }
4250#endif
4251
03010a33
TT		 4252static struct file_system_type ext4_fs_type = {
4253 .owner = THIS_MODULE,
4254 .name = "ext4",
4255 .get_sb = ext4_get_sb,
4256 .kill_sb = kill_block_super,
4257 .fs_flags = FS_REQUIRES_DEV,
4258};
4259
617ba13b 4260static int __init init_ext4_fs(void)
ac27a0ec 4261{
c9de560d
AT		 4262 int err;
4263
12e9b892 4264 ext4_check_flag_values();
6fd058f7
TT		 4265 err = init_ext4_system_zone();
4266 if (err)
4267 return err;
3197ebdb
TT		 4268 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
4269 if (!ext4_kset)
6fd058f7 4270 goto out4;
9f6200bb 4271 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
c9de560d 4272 err = init_ext4_mballoc();
ac27a0ec 4273 if (err)
6fd058f7 4274 goto out3;
c9de560d
AT		 4275
4276 err = init_ext4_xattr();
4277 if (err)
4278 goto out2;
ac27a0ec
DK		 4279 err = init_inodecache();
4280 if (err)
4281 goto out1;
24b58424
TT		 4282 register_as_ext2();
4283 register_as_ext3();
03010a33 4284 err = register_filesystem(&ext4_fs_type);
ac27a0ec
DK		 4285 if (err)
4286 goto out;
4287 return 0;
4288out:
24b58424
TT		 4289 unregister_as_ext2();
4290 unregister_as_ext3();
ac27a0ec
DK		 4291 destroy_inodecache();
4292out1:
617ba13b 4293 exit_ext4_xattr();
c9de560d
AT		 4294out2:
4295 exit_ext4_mballoc();
6fd058f7
TT		 4296out3:
4297 remove_proc_entry("fs/ext4", NULL);
4298 kset_unregister(ext4_kset);
4299out4:
4300 exit_ext4_system_zone();
ac27a0ec
DK		 4301 return err;
4302}
4303
617ba13b 4304static void __exit exit_ext4_fs(void)
ac27a0ec 4305{
24b58424
TT		 4306 unregister_as_ext2();
4307 unregister_as_ext3();
03010a33 4308 unregister_filesystem(&ext4_fs_type);
ac27a0ec 4309 destroy_inodecache();
617ba13b 4310 exit_ext4_xattr();
c9de560d 4311 exit_ext4_mballoc();
9f6200bb 4312 remove_proc_entry("fs/ext4", NULL);
3197ebdb 4313 kset_unregister(ext4_kset);
6fd058f7 4314 exit_ext4_system_zone();
ac27a0ec
DK		 4315}
4316
4317MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
83982b6f 4318MODULE_DESCRIPTION("Fourth Extended Filesystem");
ac27a0ec 4319MODULE_LICENSE("GPL");
617ba13b
MC		 4320module_init(init_ext4_fs)
4321module_exit(exit_ext4_fs)
Clone of davem's net-next-2.6 tree