projects / GitHub / mt8127 / android_kernel_alcatel_ttab.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
Fix common misspellings
[GitHub/mt8127/android_kernel_alcatel_ttab.git] / fs / ext4 / super.c
CommitLineData
ac27a0ec 1/*
617ba13b 2 * linux/fs/ext4/super.c
ac27a0ec
DK		 3 *
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
8 *
9 * from
10 *
11 * linux/fs/minix/inode.c
12 *
13 * Copyright (C) 1991, 1992 Linus Torvalds
14 *
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
17 */
18
19#include <linux/module.h>
20#include <linux/string.h>
21#include <linux/fs.h>
22#include <linux/time.h>
c5ca7c76 23#include <linux/vmalloc.h>
dab291af 24#include <linux/jbd2.h>
ac27a0ec
DK		 25#include <linux/slab.h>
26#include <linux/init.h>
27#include <linux/blkdev.h>
28#include <linux/parser.h>
ac27a0ec 29#include <linux/buffer_head.h>
a5694255 30#include <linux/exportfs.h>
ac27a0ec
DK		 31#include <linux/vfs.h>
32#include <linux/random.h>
33#include <linux/mount.h>
34#include <linux/namei.h>
35#include <linux/quotaops.h>
36#include <linux/seq_file.h>
9f6200bb 37#include <linux/proc_fs.h>
3197ebdb 38#include <linux/ctype.h>
1330593e 39#include <linux/log2.h>
717d50e4 40#include <linux/crc16.h>
ac27a0ec
DK		 41#include <asm/uaccess.h>
42
bfff6873
LC		 43#include <linux/kthread.h>
44#include <linux/freezer.h>
45
3dcf5451
CH		 46#include "ext4.h"
47#include "ext4_jbd2.h"
ac27a0ec
DK		 48#include "xattr.h"
49#include "acl.h"
3661d286 50#include "mballoc.h"
ac27a0ec 51
9bffad1e
TT		 52#define CREATE_TRACE_POINTS
53#include <trace/events/ext4.h>
54
1f109d5a 55static struct proc_dir_entry *ext4_proc_root;
3197ebdb 56static struct kset *ext4_kset;
0b75a840
LC		 57static struct ext4_lazy_init *ext4_li_info;
58static struct mutex ext4_li_mtx;
59static struct ext4_features *ext4_feat;
9f6200bb 60
617ba13b 61static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
ac27a0ec 62 unsigned long journal_devnum);
e2d67052 63static int ext4_commit_super(struct super_block *sb, int sync);
2b2d6d01
TT		 64static void ext4_mark_recovery_complete(struct super_block *sb,
65 struct ext4_super_block *es);
66static void ext4_clear_journal_err(struct super_block *sb,
67 struct ext4_super_block *es);
617ba13b 68static int ext4_sync_fs(struct super_block *sb, int wait);
2b2d6d01 69static const char *ext4_decode_error(struct super_block *sb, int errno,
ac27a0ec 70 char nbuf[16]);
2b2d6d01
TT		 71static int ext4_remount(struct super_block *sb, int *flags, char *data);
72static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
c4be0c1d 73static int ext4_unfreeze(struct super_block *sb);
2b2d6d01 74static void ext4_write_super(struct super_block *sb);
c4be0c1d 75static int ext4_freeze(struct super_block *sb);
152a0836
AV		 76static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
77 const char *dev_name, void *data);
d39195c3 78static int ext4_feature_set_ok(struct super_block *sb, int readonly);
bfff6873
LC		 79static void ext4_destroy_lazyinit_thread(void);
80static void ext4_unregister_li_request(struct super_block *sb);
8f1f7453 81static void ext4_clear_request_list(void);
ac27a0ec 82
ba69f9ab
JK		 83#if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
84static struct file_system_type ext3_fs_type = {
85 .owner = THIS_MODULE,
86 .name = "ext3",
152a0836 87 .mount = ext4_mount,
ba69f9ab
JK		 88 .kill_sb = kill_block_super,
89 .fs_flags = FS_REQUIRES_DEV,
90};
91#define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
92#else
93#define IS_EXT3_SB(sb) (0)
94#endif
bd81d8ee 95
8fadc143
AR		 96ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
97 struct ext4_group_desc *bg)
bd81d8ee 98{
3a14589c 99 return le32_to_cpu(bg->bg_block_bitmap_lo) |
8fadc143 100 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 101 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 102}
103
8fadc143
AR		 104ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
105 struct ext4_group_desc *bg)
bd81d8ee 106{
5272f837 107 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
8fadc143 108 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 109 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 110}
111
8fadc143
AR		 112ext4_fsblk_t ext4_inode_table(struct super_block *sb,
113 struct ext4_group_desc *bg)
bd81d8ee 114{
5272f837 115 return le32_to_cpu(bg->bg_inode_table_lo) |
8fadc143 116 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 117 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
bd81d8ee
LV		 118}
119
560671a0
AK		 120__u32 ext4_free_blks_count(struct super_block *sb,
121 struct ext4_group_desc *bg)
122{
123 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
124 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 125 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
560671a0
AK		 126}
127
128__u32 ext4_free_inodes_count(struct super_block *sb,
129 struct ext4_group_desc *bg)
130{
131 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
132 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 133 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
560671a0
AK		 134}
135
136__u32 ext4_used_dirs_count(struct super_block *sb,
137 struct ext4_group_desc *bg)
138{
139 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
140 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 141 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
560671a0
AK		 142}
143
144__u32 ext4_itable_unused_count(struct super_block *sb,
145 struct ext4_group_desc *bg)
146{
147 return le16_to_cpu(bg->bg_itable_unused_lo) |
148 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 149 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
560671a0
AK		 150}
151
8fadc143
AR		 152void ext4_block_bitmap_set(struct super_block *sb,
153 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 154{
3a14589c 155 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 156 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
157 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 158}
159
8fadc143
AR		 160void ext4_inode_bitmap_set(struct super_block *sb,
161 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 162{
5272f837 163 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 164 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
165 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 166}
167
8fadc143
AR		 168void ext4_inode_table_set(struct super_block *sb,
169 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 170{
5272f837 171 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 172 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
173 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 174}
175
560671a0
AK		 176void ext4_free_blks_set(struct super_block *sb,
177 struct ext4_group_desc *bg, __u32 count)
178{
179 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
180 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
181 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
182}
183
184void ext4_free_inodes_set(struct super_block *sb,
185 struct ext4_group_desc *bg, __u32 count)
186{
187 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
188 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
189 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
190}
191
192void ext4_used_dirs_set(struct super_block *sb,
193 struct ext4_group_desc *bg, __u32 count)
194{
195 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
196 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
197 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
198}
199
200void ext4_itable_unused_set(struct super_block *sb,
201 struct ext4_group_desc *bg, __u32 count)
202{
203 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
204 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
205 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
206}
207
d3d1faf6
CW		 208
209/* Just increment the non-pointer handle value */
210static handle_t *ext4_get_nojournal(void)
211{
212 handle_t *handle = current->journal_info;
213 unsigned long ref_cnt = (unsigned long)handle;
214
215 BUG_ON(ref_cnt >= EXT4_NOJOURNAL_MAX_REF_COUNT);
216
217 ref_cnt++;
218 handle = (handle_t *)ref_cnt;
219
220 current->journal_info = handle;
221 return handle;
222}
223
224
225/* Decrement the non-pointer handle value */
226static void ext4_put_nojournal(handle_t *handle)
227{
228 unsigned long ref_cnt = (unsigned long)handle;
229
230 BUG_ON(ref_cnt == 0);
231
232 ref_cnt--;
233 handle = (handle_t *)ref_cnt;
234
235 current->journal_info = handle;
236}
237
ac27a0ec 238/*
dab291af 239 * Wrappers for jbd2_journal_start/end.
ac27a0ec
DK		 240 *
241 * The only special thing we need to do here is to make sure that all
242 * journal_end calls result in the superblock being marked dirty, so
243 * that sync() will call the filesystem's write_super callback if
244 * appropriate.
245 */
617ba13b 246handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
ac27a0ec
DK		 247{
248 journal_t *journal;
249
250 if (sb->s_flags & MS_RDONLY)
251 return ERR_PTR(-EROFS);
252
437f88cc 253 vfs_check_frozen(sb, SB_FREEZE_TRANS);
ac27a0ec
DK		 254 /* Special case here: if the journal has aborted behind our
255 * backs (eg. EIO in the commit thread), then we still need to
256 * take the FS itself readonly cleanly. */
617ba13b 257 journal = EXT4_SB(sb)->s_journal;
0390131b
FM		 258 if (journal) {
259 if (is_journal_aborted(journal)) {
c67d859e 260 ext4_abort(sb, "Detected aborted journal");
0390131b
FM		 261 return ERR_PTR(-EROFS);
262 }
263 return jbd2_journal_start(journal, nblocks);
ac27a0ec 264 }
d3d1faf6 265 return ext4_get_nojournal();
ac27a0ec
DK		 266}
267
268/*
269 * The only special thing we need to do here is to make sure that all
dab291af 270 * jbd2_journal_stop calls result in the superblock being marked dirty, so
ac27a0ec
DK		 271 * that sync() will call the filesystem's write_super callback if
272 * appropriate.
273 */
c398eda0 274int __ext4_journal_stop(const char *where, unsigned int line, handle_t *handle)
ac27a0ec
DK		 275{
276 struct super_block *sb;
277 int err;
278 int rc;
279
0390131b 280 if (!ext4_handle_valid(handle)) {
d3d1faf6 281 ext4_put_nojournal(handle);
0390131b
FM		 282 return 0;
283 }
ac27a0ec
DK		 284 sb = handle->h_transaction->t_journal->j_private;
285 err = handle->h_err;
dab291af 286 rc = jbd2_journal_stop(handle);
ac27a0ec
DK		 287
288 if (!err)
289 err = rc;
290 if (err)
c398eda0 291 __ext4_std_error(sb, where, line, err);
ac27a0ec
DK		 292 return err;
293}
294
90c7201b
TT		 295void ext4_journal_abort_handle(const char *caller, unsigned int line,
296 const char *err_fn, struct buffer_head *bh,
297 handle_t *handle, int err)
ac27a0ec
DK		 298{
299 char nbuf[16];
617ba13b 300 const char *errstr = ext4_decode_error(NULL, err, nbuf);
ac27a0ec 301
0390131b
FM		 302 BUG_ON(!ext4_handle_valid(handle));
303
ac27a0ec
DK		 304 if (bh)
305 BUFFER_TRACE(bh, "abort");
306
307 if (!handle->h_err)
308 handle->h_err = err;
309
310 if (is_handle_aborted(handle))
311 return;
312
90c7201b
TT		 313 printk(KERN_ERR "%s:%d: aborting transaction: %s in %s\n",
314 caller, line, errstr, err_fn);
ac27a0ec 315
dab291af 316 jbd2_journal_abort_handle(handle);
ac27a0ec
DK		 317}
318
1c13d5c0
TT		 319static void __save_error_info(struct super_block *sb, const char *func,
320 unsigned int line)
321{
322 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
323
324 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
325 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
326 es->s_last_error_time = cpu_to_le32(get_seconds());
327 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
328 es->s_last_error_line = cpu_to_le32(line);
329 if (!es->s_first_error_time) {
330 es->s_first_error_time = es->s_last_error_time;
331 strncpy(es->s_first_error_func, func,
332 sizeof(es->s_first_error_func));
333 es->s_first_error_line = cpu_to_le32(line);
334 es->s_first_error_ino = es->s_last_error_ino;
335 es->s_first_error_block = es->s_last_error_block;
336 }
66e61a9e
TT		 337 /*
338 * Start the daily error reporting function if it hasn't been
339 * started already
340 */
341 if (!es->s_error_count)
342 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
1c13d5c0
TT		 343 es->s_error_count = cpu_to_le32(le32_to_cpu(es->s_error_count) + 1);
344}
345
346static void save_error_info(struct super_block *sb, const char *func,
347 unsigned int line)
348{
349 __save_error_info(sb, func, line);
350 ext4_commit_super(sb, 1);
351}
352
353
ac27a0ec
DK		 354/* Deal with the reporting of failure conditions on a filesystem such as
355 * inconsistencies detected or read IO failures.
356 *
357 * On ext2, we can store the error state of the filesystem in the
617ba13b 358 * superblock. That is not possible on ext4, because we may have other
ac27a0ec
DK		 359 * write ordering constraints on the superblock which prevent us from
360 * writing it out straight away; and given that the journal is about to
361 * be aborted, we can't rely on the current, or future, transactions to
362 * write out the superblock safely.
363 *
dab291af 364 * We'll just use the jbd2_journal_abort() error code to record an error in
d6b198bc 365 * the journal instead. On recovery, the journal will complain about
ac27a0ec
DK		 366 * that error until we've noted it down and cleared it.
367 */
368
617ba13b 369static void ext4_handle_error(struct super_block *sb)
ac27a0ec 370{
ac27a0ec
DK		 371 if (sb->s_flags & MS_RDONLY)
372 return;
373
2b2d6d01 374 if (!test_opt(sb, ERRORS_CONT)) {
617ba13b 375 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 376
4ab2f15b 377 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
ac27a0ec 378 if (journal)
dab291af 379 jbd2_journal_abort(journal, -EIO);
ac27a0ec 380 }
2b2d6d01 381 if (test_opt(sb, ERRORS_RO)) {
b31e1552 382 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
ac27a0ec
DK		 383 sb->s_flags |= MS_RDONLY;
384 }
ac27a0ec 385 if (test_opt(sb, ERRORS_PANIC))
617ba13b 386 panic("EXT4-fs (device %s): panic forced after error\n",
ac27a0ec
DK		 387 sb->s_id);
388}
389
12062ddd 390void __ext4_error(struct super_block *sb, const char *function,
c398eda0 391 unsigned int line, const char *fmt, ...)
ac27a0ec 392{
0ff2ea7d 393 struct va_format vaf;
ac27a0ec
DK		 394 va_list args;
395
396 va_start(args, fmt);
0ff2ea7d
JP		 397 vaf.fmt = fmt;
398 vaf.va = &args;
399 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
400 sb->s_id, function, line, current->comm, &vaf);
ac27a0ec
DK		 401 va_end(args);
402
617ba13b 403 ext4_handle_error(sb);
ac27a0ec
DK		 404}
405
c398eda0
TT		 406void ext4_error_inode(struct inode *inode, const char *function,
407 unsigned int line, ext4_fsblk_t block,
273df556
FM		 408 const char *fmt, ...)
409{
410 va_list args;
f7c21177 411 struct va_format vaf;
1c13d5c0 412 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
273df556 413
1c13d5c0
TT		 414 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
415 es->s_last_error_block = cpu_to_le64(block);
416 save_error_info(inode->i_sb, function, line);
273df556 417 va_start(args, fmt);
f7c21177
TT		 418 vaf.fmt = fmt;
419 vaf.va = &args;
c398eda0
TT		 420 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: inode #%lu: ",
421 inode->i_sb->s_id, function, line, inode->i_ino);
422 if (block)
f7c21177
TT		 423 printk(KERN_CONT "block %llu: ", block);
424 printk(KERN_CONT "comm %s: %pV\n", current->comm, &vaf);
273df556
FM		 425 va_end(args);
426
427 ext4_handle_error(inode->i_sb);
428}
429
c398eda0 430void ext4_error_file(struct file *file, const char *function,
f7c21177
TT		 431 unsigned int line, ext4_fsblk_t block,
432 const char *fmt, ...)
273df556
FM		 433{
434 va_list args;
f7c21177 435 struct va_format vaf;
1c13d5c0 436 struct ext4_super_block *es;
273df556
FM		 437 struct inode *inode = file->f_dentry->d_inode;
438 char pathname[80], *path;
439
1c13d5c0
TT		 440 es = EXT4_SB(inode->i_sb)->s_es;
441 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
442 save_error_info(inode->i_sb, function, line);
273df556 443 path = d_path(&(file->f_path), pathname, sizeof(pathname));
f9a62d09 444 if (IS_ERR(path))
273df556
FM		 445 path = "(unknown)";
446 printk(KERN_CRIT
f7c21177
TT		 447 "EXT4-fs error (device %s): %s:%d: inode #%lu: ",
448 inode->i_sb->s_id, function, line, inode->i_ino);
449 if (block)
450 printk(KERN_CONT "block %llu: ", block);
451 va_start(args, fmt);
452 vaf.fmt = fmt;
453 vaf.va = &args;
454 printk(KERN_CONT "comm %s: path %s: %pV\n", current->comm, path, &vaf);
273df556
FM		 455 va_end(args);
456
457 ext4_handle_error(inode->i_sb);
458}
459
2b2d6d01 460static const char *ext4_decode_error(struct super_block *sb, int errno,
ac27a0ec
DK		 461 char nbuf[16])
462{
463 char *errstr = NULL;
464
465 switch (errno) {
466 case -EIO:
467 errstr = "IO failure";
468 break;
469 case -ENOMEM:
470 errstr = "Out of memory";
471 break;
472 case -EROFS:
78f1ddbb
TT		 473 if (!sb || (EXT4_SB(sb)->s_journal &&
474 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
ac27a0ec
DK		 475 errstr = "Journal has aborted";
476 else
477 errstr = "Readonly filesystem";
478 break;
479 default:
480 /* If the caller passed in an extra buffer for unknown
481 * errors, textualise them now. Else we just return
482 * NULL. */
483 if (nbuf) {
484 /* Check for truncated error codes... */
485 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
486 errstr = nbuf;
487 }
488 break;
489 }
490
491 return errstr;
492}
493
617ba13b 494/* __ext4_std_error decodes expected errors from journaling functions
ac27a0ec
DK		 495 * automatically and invokes the appropriate error response. */
496
c398eda0
TT		 497void __ext4_std_error(struct super_block *sb, const char *function,
498 unsigned int line, int errno)
ac27a0ec
DK		 499{
500 char nbuf[16];
501 const char *errstr;
502
503 /* Special case: if the error is EROFS, and we're not already
504 * inside a transaction, then there's really no point in logging
505 * an error. */
506 if (errno == -EROFS && journal_current_handle() == NULL &&
507 (sb->s_flags & MS_RDONLY))
508 return;
509
617ba13b 510 errstr = ext4_decode_error(sb, errno, nbuf);
c398eda0
TT		 511 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
512 sb->s_id, function, line, errstr);
1c13d5c0 513 save_error_info(sb, function, line);
ac27a0ec 514
617ba13b 515 ext4_handle_error(sb);
ac27a0ec
DK		 516}
517
518/*
617ba13b 519 * ext4_abort is a much stronger failure handler than ext4_error. The
ac27a0ec
DK		 520 * abort function may be used to deal with unrecoverable failures such
521 * as journal IO errors or ENOMEM at a critical moment in log management.
522 *
523 * We unconditionally force the filesystem into an ABORT|READONLY state,
524 * unless the error response on the fs has been set to panic in which
525 * case we take the easy way out and panic immediately.
526 */
527
c67d859e 528void __ext4_abort(struct super_block *sb, const char *function,
c398eda0 529 unsigned int line, const char *fmt, ...)
ac27a0ec
DK		 530{
531 va_list args;
532
1c13d5c0 533 save_error_info(sb, function, line);
ac27a0ec 534 va_start(args, fmt);
c398eda0
TT		 535 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
536 function, line);
ac27a0ec
DK		 537 vprintk(fmt, args);
538 printk("\n");
539 va_end(args);
540
1c13d5c0
TT		 541 if ((sb->s_flags & MS_RDONLY) == 0) {
542 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
543 sb->s_flags |= MS_RDONLY;
544 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
545 if (EXT4_SB(sb)->s_journal)
546 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
547 save_error_info(sb, function, line);
548 }
ac27a0ec 549 if (test_opt(sb, ERRORS_PANIC))
617ba13b 550 panic("EXT4-fs panic from previous error\n");
ac27a0ec
DK		 551}
552
0ff2ea7d 553void ext4_msg(struct super_block *sb, const char *prefix, const char *fmt, ...)
b31e1552 554{
0ff2ea7d 555 struct va_format vaf;
b31e1552
ES		 556 va_list args;
557
558 va_start(args, fmt);
0ff2ea7d
JP		 559 vaf.fmt = fmt;
560 vaf.va = &args;
561 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
b31e1552
ES		 562 va_end(args);
563}
564
12062ddd 565void __ext4_warning(struct super_block *sb, const char *function,
c398eda0 566 unsigned int line, const char *fmt, ...)
ac27a0ec 567{
0ff2ea7d 568 struct va_format vaf;
ac27a0ec
DK		 569 va_list args;
570
571 va_start(args, fmt);
0ff2ea7d
JP		 572 vaf.fmt = fmt;
573 vaf.va = &args;
574 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
575 sb->s_id, function, line, &vaf);
ac27a0ec
DK		 576 va_end(args);
577}
578
e29136f8
TT		 579void __ext4_grp_locked_error(const char *function, unsigned int line,
580 struct super_block *sb, ext4_group_t grp,
581 unsigned long ino, ext4_fsblk_t block,
582 const char *fmt, ...)
5d1b1b3f
AK		 583__releases(bitlock)
584__acquires(bitlock)
585{
0ff2ea7d 586 struct va_format vaf;
5d1b1b3f
AK		 587 va_list args;
588 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
589
1c13d5c0
TT		 590 es->s_last_error_ino = cpu_to_le32(ino);
591 es->s_last_error_block = cpu_to_le64(block);
592 __save_error_info(sb, function, line);
0ff2ea7d 593
5d1b1b3f 594 va_start(args, fmt);
0ff2ea7d
JP		 595
596 vaf.fmt = fmt;
597 vaf.va = &args;
21149d61 598 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
e29136f8
TT		 599 sb->s_id, function, line, grp);
600 if (ino)
0ff2ea7d 601 printk(KERN_CONT "inode %lu: ", ino);
e29136f8 602 if (block)
0ff2ea7d
JP		 603 printk(KERN_CONT "block %llu:", (unsigned long long) block);
604 printk(KERN_CONT "%pV\n", &vaf);
5d1b1b3f
AK		 605 va_end(args);
606
607 if (test_opt(sb, ERRORS_CONT)) {
e2d67052 608 ext4_commit_super(sb, 0);
5d1b1b3f
AK		 609 return;
610 }
1c13d5c0 611
5d1b1b3f
AK		 612 ext4_unlock_group(sb, grp);
613 ext4_handle_error(sb);
614 /*
615 * We only get here in the ERRORS_RO case; relocking the group
616 * may be dangerous, but nothing bad will happen since the
617 * filesystem will have already been marked read/only and the
618 * journal has been aborted. We return 1 as a hint to callers
619 * who might what to use the return value from
25985edc 620 * ext4_grp_locked_error() to distinguish between the
5d1b1b3f
AK		 621 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
622 * aggressively from the ext4 function in question, with a
623 * more appropriate error code.
624 */
625 ext4_lock_group(sb, grp);
626 return;
627}
628
617ba13b 629void ext4_update_dynamic_rev(struct super_block *sb)
ac27a0ec 630{
617ba13b 631 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 632
617ba13b 633 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
ac27a0ec
DK		 634 return;
635
12062ddd 636 ext4_warning(sb,
ac27a0ec
DK		 637 "updating to rev %d because of new feature flag, "
638 "running e2fsck is recommended",
617ba13b 639 EXT4_DYNAMIC_REV);
ac27a0ec 640
617ba13b
MC		 641 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
642 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
643 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
ac27a0ec
DK		 644 /* leave es->s_feature_*compat flags alone */
645 /* es->s_uuid will be set by e2fsck if empty */
646
647 /*
648 * The rest of the superblock fields should be zero, and if not it
649 * means they are likely already in use, so leave them alone. We
650 * can leave it up to e2fsck to clean up any inconsistencies there.
651 */
652}
653
654/*
655 * Open the external journal device
656 */
b31e1552 657static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
ac27a0ec
DK		 658{
659 struct block_device *bdev;
660 char b[BDEVNAME_SIZE];
661
d4d77629 662 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
ac27a0ec
DK		 663 if (IS_ERR(bdev))
664 goto fail;
665 return bdev;
666
667fail:
b31e1552 668 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
ac27a0ec
DK		 669 __bdevname(dev, b), PTR_ERR(bdev));
670 return NULL;
671}
672
673/*
674 * Release the journal device
675 */
617ba13b 676static int ext4_blkdev_put(struct block_device *bdev)
ac27a0ec 677{
e525fd89 678 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
ac27a0ec
DK		 679}
680
617ba13b 681static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
ac27a0ec
DK		 682{
683 struct block_device *bdev;
684 int ret = -ENODEV;
685
686 bdev = sbi->journal_bdev;
687 if (bdev) {
617ba13b 688 ret = ext4_blkdev_put(bdev);
ac27a0ec
DK		 689 sbi->journal_bdev = NULL;
690 }
691 return ret;
692}
693
694static inline struct inode *orphan_list_entry(struct list_head *l)
695{
617ba13b 696 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
ac27a0ec
DK		 697}
698
617ba13b 699static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
ac27a0ec
DK		 700{
701 struct list_head *l;
702
b31e1552
ES		 703 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
704 le32_to_cpu(sbi->s_es->s_last_orphan));
ac27a0ec
DK		 705
706 printk(KERN_ERR "sb_info orphan list:\n");
707 list_for_each(l, &sbi->s_orphan) {
708 struct inode *inode = orphan_list_entry(l);
709 printk(KERN_ERR " "
710 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
711 inode->i_sb->s_id, inode->i_ino, inode,
712 inode->i_mode, inode->i_nlink,
713 NEXT_ORPHAN(inode));
714 }
715}
716
2b2d6d01 717static void ext4_put_super(struct super_block *sb)
ac27a0ec 718{
617ba13b
MC		 719 struct ext4_sb_info *sbi = EXT4_SB(sb);
720 struct ext4_super_block *es = sbi->s_es;
ef2cabf7 721 int i, err;
ac27a0ec 722
857ac889 723 ext4_unregister_li_request(sb);
e0ccfd95
CH		 724 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
725
4c0425ff
MC		 726 flush_workqueue(sbi->dio_unwritten_wq);
727 destroy_workqueue(sbi->dio_unwritten_wq);
728
a9e220f8 729 lock_super(sb);
8c85e125 730 if (sb->s_dirt)
ebc1ac16 731 ext4_commit_super(sb, 1);
8c85e125 732
0390131b
FM		 733 if (sbi->s_journal) {
734 err = jbd2_journal_destroy(sbi->s_journal);
735 sbi->s_journal = NULL;
736 if (err < 0)
c67d859e 737 ext4_abort(sb, "Couldn't clean up the journal");
0390131b 738 }
d4edac31 739
a1c6c569 740 del_timer(&sbi->s_err_report);
d4edac31
JB		 741 ext4_release_system_zone(sb);
742 ext4_mb_release(sb);
743 ext4_ext_release(sb);
744 ext4_xattr_put_super(sb);
745
ac27a0ec 746 if (!(sb->s_flags & MS_RDONLY)) {
617ba13b 747 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 748 es->s_state = cpu_to_le16(sbi->s_mount_state);
e2d67052 749 ext4_commit_super(sb, 1);
ac27a0ec 750 }
240799cd 751 if (sbi->s_proc) {
9f6200bb 752 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 753 }
3197ebdb 754 kobject_del(&sbi->s_kobj);
ac27a0ec
DK		 755
756 for (i = 0; i < sbi->s_gdb_count; i++)
757 brelse(sbi->s_group_desc[i]);
758 kfree(sbi->s_group_desc);
c5ca7c76
TT		 759 if (is_vmalloc_addr(sbi->s_flex_groups))
760 vfree(sbi->s_flex_groups);
761 else
762 kfree(sbi->s_flex_groups);
ac27a0ec
DK		 763 percpu_counter_destroy(&sbi->s_freeblocks_counter);
764 percpu_counter_destroy(&sbi->s_freeinodes_counter);
765 percpu_counter_destroy(&sbi->s_dirs_counter);
6bc6e63f 766 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
ac27a0ec
DK		 767 brelse(sbi->s_sbh);
768#ifdef CONFIG_QUOTA
769 for (i = 0; i < MAXQUOTAS; i++)
770 kfree(sbi->s_qf_names[i]);
771#endif
772
773 /* Debugging code just in case the in-memory inode orphan list
774 * isn't empty. The on-disk one can be non-empty if we've
775 * detected an error and taken the fs readonly, but the
776 * in-memory list had better be clean by this point. */
777 if (!list_empty(&sbi->s_orphan))
778 dump_orphan_list(sb, sbi);
779 J_ASSERT(list_empty(&sbi->s_orphan));
780
f98393a6 781 invalidate_bdev(sb->s_bdev);
ac27a0ec
DK		 782 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
783 /*
784 * Invalidate the journal device's buffers. We don't want them
785 * floating about in memory - the physical journal device may
786 * hotswapped, and it breaks the `ro-after' testing code.
787 */
788 sync_blockdev(sbi->journal_bdev);
f98393a6 789 invalidate_bdev(sbi->journal_bdev);
617ba13b 790 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 791 }
792 sb->s_fs_info = NULL;
3197ebdb
TT		 793 /*
794 * Now that we are completely done shutting down the
795 * superblock, we need to actually destroy the kobject.
796 */
3197ebdb
TT		 797 unlock_super(sb);
798 kobject_put(&sbi->s_kobj);
799 wait_for_completion(&sbi->s_kobj_unregister);
705895b6 800 kfree(sbi->s_blockgroup_lock);
ac27a0ec 801 kfree(sbi);
ac27a0ec
DK		 802}
803
e18b890b 804static struct kmem_cache *ext4_inode_cachep;
ac27a0ec
DK		 805
806/*
807 * Called inside transaction, so use GFP_NOFS
808 */
617ba13b 809static struct inode *ext4_alloc_inode(struct super_block *sb)
ac27a0ec 810{
617ba13b 811 struct ext4_inode_info *ei;
ac27a0ec 812
e6b4f8da 813 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
ac27a0ec
DK		 814 if (!ei)
815 return NULL;
0b8e58a1 816
ac27a0ec 817 ei->vfs_inode.i_version = 1;
91246c00 818 ei->vfs_inode.i_data.writeback_index = 0;
a86c6181 819 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
c9de560d
AT		 820 INIT_LIST_HEAD(&ei->i_prealloc_list);
821 spin_lock_init(&ei->i_prealloc_lock);
d2a17637
MC		 822 ei->i_reserved_data_blocks = 0;
823 ei->i_reserved_meta_blocks = 0;
824 ei->i_allocated_meta_blocks = 0;
9d0be502 825 ei->i_da_metadata_calc_len = 0;
d2a17637 826 spin_lock_init(&(ei->i_block_reservation_lock));
a9e7f447
DM		 827#ifdef CONFIG_QUOTA
828 ei->i_reserved_quota = 0;
829#endif
8aefcd55 830 ei->jinode = NULL;
c7064ef1 831 INIT_LIST_HEAD(&ei->i_completed_io_list);
744692dc 832 spin_lock_init(&ei->i_completed_io_lock);
8d5d02e6 833 ei->cur_aio_dio = NULL;
b436b9be
JK		 834 ei->i_sync_tid = 0;
835 ei->i_datasync_tid = 0;
f7ad6d2e 836 atomic_set(&ei->i_ioend_count, 0);
e9e3bcec 837 atomic_set(&ei->i_aiodio_unwritten, 0);
0b8e58a1 838
ac27a0ec
DK		 839 return &ei->vfs_inode;
840}
841
7ff9c073
TT		 842static int ext4_drop_inode(struct inode *inode)
843{
844 int drop = generic_drop_inode(inode);
845
846 trace_ext4_drop_inode(inode, drop);
847 return drop;
848}
849
fa0d7e3d
NP		 850static void ext4_i_callback(struct rcu_head *head)
851{
852 struct inode *inode = container_of(head, struct inode, i_rcu);
853 INIT_LIST_HEAD(&inode->i_dentry);
854 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
855}
856
617ba13b 857static void ext4_destroy_inode(struct inode *inode)
ac27a0ec 858{
f7ad6d2e 859 ext4_ioend_wait(inode);
9f7dd93d 860 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
b31e1552
ES		 861 ext4_msg(inode->i_sb, KERN_ERR,
862 "Inode %lu (%p): orphan list check failed!",
863 inode->i_ino, EXT4_I(inode));
9f7dd93d
VA		 864 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
865 EXT4_I(inode), sizeof(struct ext4_inode_info),
866 true);
867 dump_stack();
868 }
fa0d7e3d 869 call_rcu(&inode->i_rcu, ext4_i_callback);
ac27a0ec
DK		 870}
871
51cc5068 872static void init_once(void *foo)
ac27a0ec 873{
617ba13b 874 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
ac27a0ec 875
a35afb83 876 INIT_LIST_HEAD(&ei->i_orphan);
03010a33 877#ifdef CONFIG_EXT4_FS_XATTR
a35afb83 878 init_rwsem(&ei->xattr_sem);
ac27a0ec 879#endif
0e855ac8 880 init_rwsem(&ei->i_data_sem);
a35afb83 881 inode_init_once(&ei->vfs_inode);
ac27a0ec
DK		 882}
883
884static int init_inodecache(void)
885{
617ba13b
MC		 886 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
887 sizeof(struct ext4_inode_info),
ac27a0ec
DK		 888 0, (SLAB_RECLAIM_ACCOUNT|
889 SLAB_MEM_SPREAD),
20c2df83 890 init_once);
617ba13b 891 if (ext4_inode_cachep == NULL)
ac27a0ec
DK		 892 return -ENOMEM;
893 return 0;
894}
895
896static void destroy_inodecache(void)
897{
617ba13b 898 kmem_cache_destroy(ext4_inode_cachep);
ac27a0ec
DK		 899}
900
0930fcc1 901void ext4_clear_inode(struct inode *inode)
ac27a0ec 902{
0930fcc1
AV		 903 invalidate_inode_buffers(inode);
904 end_writeback(inode);
9f754758 905 dquot_drop(inode);
c2ea3fde 906 ext4_discard_preallocations(inode);
8aefcd55
TT		 907 if (EXT4_I(inode)->jinode) {
908 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
909 EXT4_I(inode)->jinode);
910 jbd2_free_inode(EXT4_I(inode)->jinode);
911 EXT4_I(inode)->jinode = NULL;
912 }
ac27a0ec
DK		 913}
914
2b2d6d01
TT		 915static inline void ext4_show_quota_options(struct seq_file *seq,
916 struct super_block *sb)
ac27a0ec
DK		 917{
918#if defined(CONFIG_QUOTA)
617ba13b 919 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 920
5a20bdfc
JK		 921 if (sbi->s_jquota_fmt) {
922 char *fmtname = "";
923
924 switch (sbi->s_jquota_fmt) {
925 case QFMT_VFS_OLD:
926 fmtname = "vfsold";
927 break;
928 case QFMT_VFS_V0:
929 fmtname = "vfsv0";
930 break;
931 case QFMT_VFS_V1:
932 fmtname = "vfsv1";
933 break;
934 }
935 seq_printf(seq, ",jqfmt=%s", fmtname);
936 }
ac27a0ec
DK		 937
938 if (sbi->s_qf_names[USRQUOTA])
939 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
940
941 if (sbi->s_qf_names[GRPQUOTA])
942 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
943
482a7425 944 if (test_opt(sb, USRQUOTA))
ac27a0ec
DK		 945 seq_puts(seq, ",usrquota");
946
482a7425 947 if (test_opt(sb, GRPQUOTA))
ac27a0ec
DK		 948 seq_puts(seq, ",grpquota");
949#endif
950}
951
d9c9bef1
MS		 952/*
953 * Show an option if
954 * - it's set to a non-default value OR
955 * - if the per-sb default is different from the global default
956 */
617ba13b 957static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
ac27a0ec 958{
aa22df2c
AK		 959 int def_errors;
960 unsigned long def_mount_opts;
ac27a0ec 961 struct super_block *sb = vfs->mnt_sb;
d9c9bef1
MS		 962 struct ext4_sb_info *sbi = EXT4_SB(sb);
963 struct ext4_super_block *es = sbi->s_es;
d9c9bef1
MS		 964
965 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
aa22df2c 966 def_errors = le16_to_cpu(es->s_errors);
d9c9bef1
MS		 967
968 if (sbi->s_sb_block != 1)
969 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
970 if (test_opt(sb, MINIX_DF))
971 seq_puts(seq, ",minixdf");
aa22df2c 972 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
d9c9bef1
MS		 973 seq_puts(seq, ",grpid");
974 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
975 seq_puts(seq, ",nogrpid");
976 if (sbi->s_resuid != EXT4_DEF_RESUID ||
977 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
978 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
979 }
980 if (sbi->s_resgid != EXT4_DEF_RESGID ||
981 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
982 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
983 }
bb4f397a 984 if (test_opt(sb, ERRORS_RO)) {
d9c9bef1 985 if (def_errors == EXT4_ERRORS_PANIC ||
bb4f397a
AK		 986 def_errors == EXT4_ERRORS_CONTINUE) {
987 seq_puts(seq, ",errors=remount-ro");
d9c9bef1
MS		 988 }
989 }
aa22df2c 990 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
bb4f397a 991 seq_puts(seq, ",errors=continue");
aa22df2c 992 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
d9c9bef1 993 seq_puts(seq, ",errors=panic");
aa22df2c 994 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
d9c9bef1 995 seq_puts(seq, ",nouid32");
aa22df2c 996 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
d9c9bef1
MS		 997 seq_puts(seq, ",debug");
998 if (test_opt(sb, OLDALLOC))
999 seq_puts(seq, ",oldalloc");
03010a33 1000#ifdef CONFIG_EXT4_FS_XATTR
ea663336 1001 if (test_opt(sb, XATTR_USER))
d9c9bef1 1002 seq_puts(seq, ",user_xattr");
ea663336 1003 if (!test_opt(sb, XATTR_USER))
d9c9bef1 1004 seq_puts(seq, ",nouser_xattr");
d9c9bef1 1005#endif
03010a33 1006#ifdef CONFIG_EXT4_FS_POSIX_ACL
aa22df2c 1007 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
d9c9bef1
MS		 1008 seq_puts(seq, ",acl");
1009 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
1010 seq_puts(seq, ",noacl");
1011#endif
30773840 1012 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
d9c9bef1
MS		 1013 seq_printf(seq, ",commit=%u",
1014 (unsigned) (sbi->s_commit_interval / HZ));
1015 }
30773840
TT		 1016 if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
1017 seq_printf(seq, ",min_batch_time=%u",
1018 (unsigned) sbi->s_min_batch_time);
1019 }
1020 if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
1021 seq_printf(seq, ",max_batch_time=%u",
1022 (unsigned) sbi->s_min_batch_time);
1023 }
1024
571640ca
ES		 1025 /*
1026 * We're changing the default of barrier mount option, so
1027 * let's always display its mount state so it's clear what its
1028 * status is.
1029 */
1030 seq_puts(seq, ",barrier=");
1031 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
cd0b6a39
TT		 1032 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
1033 seq_puts(seq, ",journal_async_commit");
39a4bade
JK		 1034 else if (test_opt(sb, JOURNAL_CHECKSUM))
1035 seq_puts(seq, ",journal_checksum");
25ec56b5
JNC		 1036 if (test_opt(sb, I_VERSION))
1037 seq_puts(seq, ",i_version");
8b67f04a
TT		 1038 if (!test_opt(sb, DELALLOC) &&
1039 !(def_mount_opts & EXT4_DEFM_NODELALLOC))
dd919b98
AK		 1040 seq_puts(seq, ",nodelalloc");
1041
6fd7a467
TT		 1042 if (!test_opt(sb, MBLK_IO_SUBMIT))
1043 seq_puts(seq, ",nomblk_io_submit");
cb45bbe4
MS		 1044 if (sbi->s_stripe)
1045 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
aa22df2c
AK		 1046 /*
1047 * journal mode get enabled in different ways
1048 * So just print the value even if we didn't specify it
1049 */
617ba13b 1050 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
ac27a0ec 1051 seq_puts(seq, ",data=journal");
617ba13b 1052 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
ac27a0ec 1053 seq_puts(seq, ",data=ordered");
617ba13b 1054 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
ac27a0ec
DK		 1055 seq_puts(seq, ",data=writeback");
1056
240799cd
TT		 1057 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1058 seq_printf(seq, ",inode_readahead_blks=%u",
1059 sbi->s_inode_readahead_blks);
1060
5bf5683a
HK		 1061 if (test_opt(sb, DATA_ERR_ABORT))
1062 seq_puts(seq, ",data_err=abort");
1063
afd4672d 1064 if (test_opt(sb, NO_AUTO_DA_ALLOC))
06705bff 1065 seq_puts(seq, ",noauto_da_alloc");
afd4672d 1066
8b67f04a 1067 if (test_opt(sb, DISCARD) && !(def_mount_opts & EXT4_DEFM_DISCARD))
5328e635
ES		 1068 seq_puts(seq, ",discard");
1069
e3bb52ae
ES		 1070 if (test_opt(sb, NOLOAD))
1071 seq_puts(seq, ",norecovery");
1072
744692dc
JZ		 1073 if (test_opt(sb, DIOREAD_NOLOCK))
1074 seq_puts(seq, ",dioread_nolock");
1075
8b67f04a
TT		 1076 if (test_opt(sb, BLOCK_VALIDITY) &&
1077 !(def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY))
1078 seq_puts(seq, ",block_validity");
1079
bfff6873
LC		 1080 if (!test_opt(sb, INIT_INODE_TABLE))
1081 seq_puts(seq, ",noinit_inode_table");
1082 else if (sbi->s_li_wait_mult)
1083 seq_printf(seq, ",init_inode_table=%u",
1084 (unsigned) sbi->s_li_wait_mult);
1085
617ba13b 1086 ext4_show_quota_options(seq, sb);
0b8e58a1 1087
ac27a0ec
DK		 1088 return 0;
1089}
1090
1b961ac0 1091static struct inode *ext4_nfs_get_inode(struct super_block *sb,
0b8e58a1 1092 u64 ino, u32 generation)
ac27a0ec 1093{
ac27a0ec 1094 struct inode *inode;
ac27a0ec 1095
617ba13b 1096 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
ac27a0ec 1097 return ERR_PTR(-ESTALE);
617ba13b 1098 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
ac27a0ec
DK		 1099 return ERR_PTR(-ESTALE);
1100
1101 /* iget isn't really right if the inode is currently unallocated!!
1102 *
617ba13b 1103 * ext4_read_inode will return a bad_inode if the inode had been
ac27a0ec
DK		 1104 * deleted, so we should be safe.
1105 *
1106 * Currently we don't know the generation for parent directory, so
1107 * a generation of 0 means "accept any"
1108 */
1d1fe1ee
DH		 1109 inode = ext4_iget(sb, ino);
1110 if (IS_ERR(inode))
1111 return ERR_CAST(inode);
1112 if (generation && inode->i_generation != generation) {
ac27a0ec
DK		 1113 iput(inode);
1114 return ERR_PTR(-ESTALE);
1115 }
1b961ac0
CH		 1116
1117 return inode;
1118}
1119
1120static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
0b8e58a1 1121 int fh_len, int fh_type)
1b961ac0
CH		 1122{
1123 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1124 ext4_nfs_get_inode);
1125}
1126
1127static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
0b8e58a1 1128 int fh_len, int fh_type)
1b961ac0
CH		 1129{
1130 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1131 ext4_nfs_get_inode);
ac27a0ec
DK		 1132}
1133
c39a7f84
TO		 1134/*
1135 * Try to release metadata pages (indirect blocks, directories) which are
1136 * mapped via the block device. Since these pages could have journal heads
1137 * which would prevent try_to_free_buffers() from freeing them, we must use
1138 * jbd2 layer's try_to_free_buffers() function to release them.
1139 */
0b8e58a1
AD		 1140static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1141 gfp_t wait)
c39a7f84
TO		 1142{
1143 journal_t *journal = EXT4_SB(sb)->s_journal;
1144
1145 WARN_ON(PageChecked(page));
1146 if (!page_has_buffers(page))
1147 return 0;
1148 if (journal)
1149 return jbd2_journal_try_to_free_buffers(journal, page,
1150 wait & ~__GFP_WAIT);
1151 return try_to_free_buffers(page);
1152}
1153
ac27a0ec 1154#ifdef CONFIG_QUOTA
af5bc92d 1155#define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
2b2d6d01 1156#define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
ac27a0ec 1157
617ba13b
MC		 1158static int ext4_write_dquot(struct dquot *dquot);
1159static int ext4_acquire_dquot(struct dquot *dquot);
1160static int ext4_release_dquot(struct dquot *dquot);
1161static int ext4_mark_dquot_dirty(struct dquot *dquot);
1162static int ext4_write_info(struct super_block *sb, int type);
6f28e087 1163static int ext4_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 1164 struct path *path);
ca0e05e4 1165static int ext4_quota_off(struct super_block *sb, int type);
617ba13b
MC		 1166static int ext4_quota_on_mount(struct super_block *sb, int type);
1167static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 1168 size_t len, loff_t off);
617ba13b 1169static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 1170 const char *data, size_t len, loff_t off);
1171
61e225dc 1172static const struct dquot_operations ext4_quota_operations = {
a9e7f447 1173#ifdef CONFIG_QUOTA
60e58e0f 1174 .get_reserved_space = ext4_get_reserved_space,
a9e7f447 1175#endif
617ba13b
MC		 1176 .write_dquot = ext4_write_dquot,
1177 .acquire_dquot = ext4_acquire_dquot,
1178 .release_dquot = ext4_release_dquot,
1179 .mark_dirty = ext4_mark_dquot_dirty,
a5b5ee32
JK		 1180 .write_info = ext4_write_info,
1181 .alloc_dquot = dquot_alloc,
1182 .destroy_dquot = dquot_destroy,
ac27a0ec
DK		 1183};
1184
0d54b217 1185static const struct quotactl_ops ext4_qctl_operations = {
617ba13b 1186 .quota_on = ext4_quota_on,
ca0e05e4 1187 .quota_off = ext4_quota_off,
287a8095
CH		 1188 .quota_sync = dquot_quota_sync,
1189 .get_info = dquot_get_dqinfo,
1190 .set_info = dquot_set_dqinfo,
1191 .get_dqblk = dquot_get_dqblk,
1192 .set_dqblk = dquot_set_dqblk
ac27a0ec
DK		 1193};
1194#endif
1195
ee9b6d61 1196static const struct super_operations ext4_sops = {
617ba13b
MC		 1197 .alloc_inode = ext4_alloc_inode,
1198 .destroy_inode = ext4_destroy_inode,
617ba13b
MC		 1199 .write_inode = ext4_write_inode,
1200 .dirty_inode = ext4_dirty_inode,
7ff9c073 1201 .drop_inode = ext4_drop_inode,
0930fcc1 1202 .evict_inode = ext4_evict_inode,
617ba13b 1203 .put_super = ext4_put_super,
617ba13b 1204 .sync_fs = ext4_sync_fs,
c4be0c1d
TS		 1205 .freeze_fs = ext4_freeze,
1206 .unfreeze_fs = ext4_unfreeze,
617ba13b
MC		 1207 .statfs = ext4_statfs,
1208 .remount_fs = ext4_remount,
617ba13b 1209 .show_options = ext4_show_options,
ac27a0ec 1210#ifdef CONFIG_QUOTA
617ba13b
MC		 1211 .quota_read = ext4_quota_read,
1212 .quota_write = ext4_quota_write,
ac27a0ec 1213#endif
c39a7f84 1214 .bdev_try_to_free_page = bdev_try_to_free_page,
ac27a0ec
DK		 1215};
1216
9ca92389
TT		 1217static const struct super_operations ext4_nojournal_sops = {
1218 .alloc_inode = ext4_alloc_inode,
1219 .destroy_inode = ext4_destroy_inode,
1220 .write_inode = ext4_write_inode,
1221 .dirty_inode = ext4_dirty_inode,
7ff9c073 1222 .drop_inode = ext4_drop_inode,
0930fcc1 1223 .evict_inode = ext4_evict_inode,
9ca92389
TT		 1224 .write_super = ext4_write_super,
1225 .put_super = ext4_put_super,
1226 .statfs = ext4_statfs,
1227 .remount_fs = ext4_remount,
9ca92389
TT		 1228 .show_options = ext4_show_options,
1229#ifdef CONFIG_QUOTA
1230 .quota_read = ext4_quota_read,
1231 .quota_write = ext4_quota_write,
1232#endif
1233 .bdev_try_to_free_page = bdev_try_to_free_page,
1234};
1235
39655164 1236static const struct export_operations ext4_export_ops = {
1b961ac0
CH		 1237 .fh_to_dentry = ext4_fh_to_dentry,
1238 .fh_to_parent = ext4_fh_to_parent,
617ba13b 1239 .get_parent = ext4_get_parent,
ac27a0ec
DK		 1240};
1241
1242enum {
1243 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1244 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
01436ef2 1245 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
ac27a0ec 1246 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
06705bff 1247 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh,
30773840 1248 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
c3191067 1249 Opt_journal_update, Opt_journal_dev,
818d276c 1250 Opt_journal_checksum, Opt_journal_async_commit,
ac27a0ec 1251 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
296c355c 1252 Opt_data_err_abort, Opt_data_err_ignore,
ac27a0ec 1253 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
5a20bdfc
JK		 1254 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1255 Opt_noquota, Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err,
1256 Opt_resize, Opt_usrquota, Opt_grpquota, Opt_i_version,
1449032b
TT		 1257 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1258 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
5328e635 1259 Opt_inode_readahead_blks, Opt_journal_ioprio,
744692dc 1260 Opt_dioread_nolock, Opt_dioread_lock,
5328e635 1261 Opt_discard, Opt_nodiscard,
bfff6873 1262 Opt_init_inode_table, Opt_noinit_inode_table,
ac27a0ec
DK		 1263};
1264
a447c093 1265static const match_table_t tokens = {
ac27a0ec
DK		 1266 {Opt_bsd_df, "bsddf"},
1267 {Opt_minix_df, "minixdf"},
1268 {Opt_grpid, "grpid"},
1269 {Opt_grpid, "bsdgroups"},
1270 {Opt_nogrpid, "nogrpid"},
1271 {Opt_nogrpid, "sysvgroups"},
1272 {Opt_resgid, "resgid=%u"},
1273 {Opt_resuid, "resuid=%u"},
1274 {Opt_sb, "sb=%u"},
1275 {Opt_err_cont, "errors=continue"},
1276 {Opt_err_panic, "errors=panic"},
1277 {Opt_err_ro, "errors=remount-ro"},
1278 {Opt_nouid32, "nouid32"},
ac27a0ec
DK		 1279 {Opt_debug, "debug"},
1280 {Opt_oldalloc, "oldalloc"},
1281 {Opt_orlov, "orlov"},
1282 {Opt_user_xattr, "user_xattr"},
1283 {Opt_nouser_xattr, "nouser_xattr"},
1284 {Opt_acl, "acl"},
1285 {Opt_noacl, "noacl"},
ac27a0ec 1286 {Opt_noload, "noload"},
e3bb52ae 1287 {Opt_noload, "norecovery"},
ac27a0ec
DK		 1288 {Opt_nobh, "nobh"},
1289 {Opt_bh, "bh"},
1290 {Opt_commit, "commit=%u"},
30773840
TT		 1291 {Opt_min_batch_time, "min_batch_time=%u"},
1292 {Opt_max_batch_time, "max_batch_time=%u"},
ac27a0ec 1293 {Opt_journal_update, "journal=update"},
ac27a0ec 1294 {Opt_journal_dev, "journal_dev=%u"},
818d276c
GS		 1295 {Opt_journal_checksum, "journal_checksum"},
1296 {Opt_journal_async_commit, "journal_async_commit"},
ac27a0ec
DK		 1297 {Opt_abort, "abort"},
1298 {Opt_data_journal, "data=journal"},
1299 {Opt_data_ordered, "data=ordered"},
1300 {Opt_data_writeback, "data=writeback"},
5bf5683a
HK		 1301 {Opt_data_err_abort, "data_err=abort"},
1302 {Opt_data_err_ignore, "data_err=ignore"},
ac27a0ec
DK		 1303 {Opt_offusrjquota, "usrjquota="},
1304 {Opt_usrjquota, "usrjquota=%s"},
1305 {Opt_offgrpjquota, "grpjquota="},
1306 {Opt_grpjquota, "grpjquota=%s"},
1307 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1308 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
5a20bdfc 1309 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
ac27a0ec
DK		 1310 {Opt_grpquota, "grpquota"},
1311 {Opt_noquota, "noquota"},
1312 {Opt_quota, "quota"},
1313 {Opt_usrquota, "usrquota"},
1314 {Opt_barrier, "barrier=%u"},
06705bff
TT		 1315 {Opt_barrier, "barrier"},
1316 {Opt_nobarrier, "nobarrier"},
25ec56b5 1317 {Opt_i_version, "i_version"},
c9de560d 1318 {Opt_stripe, "stripe=%u"},
ac27a0ec 1319 {Opt_resize, "resize"},
64769240 1320 {Opt_delalloc, "delalloc"},
dd919b98 1321 {Opt_nodelalloc, "nodelalloc"},
1449032b
TT		 1322 {Opt_mblk_io_submit, "mblk_io_submit"},
1323 {Opt_nomblk_io_submit, "nomblk_io_submit"},
6fd058f7
TT		 1324 {Opt_block_validity, "block_validity"},
1325 {Opt_noblock_validity, "noblock_validity"},
240799cd 1326 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
b3881f74 1327 {Opt_journal_ioprio, "journal_ioprio=%u"},
afd4672d 1328 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
06705bff
TT		 1329 {Opt_auto_da_alloc, "auto_da_alloc"},
1330 {Opt_noauto_da_alloc, "noauto_da_alloc"},
744692dc
JZ		 1331 {Opt_dioread_nolock, "dioread_nolock"},
1332 {Opt_dioread_lock, "dioread_lock"},
5328e635
ES		 1333 {Opt_discard, "discard"},
1334 {Opt_nodiscard, "nodiscard"},
bfff6873
LC		 1335 {Opt_init_inode_table, "init_itable=%u"},
1336 {Opt_init_inode_table, "init_itable"},
1337 {Opt_noinit_inode_table, "noinit_itable"},
f3f12faa 1338 {Opt_err, NULL},
ac27a0ec
DK		 1339};
1340
617ba13b 1341static ext4_fsblk_t get_sb_block(void **data)
ac27a0ec 1342{
617ba13b 1343 ext4_fsblk_t sb_block;
ac27a0ec
DK		 1344 char *options = (char *) *data;
1345
1346 if (!options || strncmp(options, "sb=", 3) != 0)
1347 return 1; /* Default location */
0b8e58a1 1348
ac27a0ec 1349 options += 3;
0b8e58a1 1350 /* TODO: use simple_strtoll with >32bit ext4 */
ac27a0ec
DK		 1351 sb_block = simple_strtoul(options, &options, 0);
1352 if (*options && *options != ',') {
4776004f 1353 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
ac27a0ec
DK		 1354 (char *) *data);
1355 return 1;
1356 }
1357 if (*options == ',')
1358 options++;
1359 *data = (void *) options;
0b8e58a1 1360
ac27a0ec
DK		 1361 return sb_block;
1362}
1363
b3881f74 1364#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
437ca0fd
DM		 1365static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1366 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
b3881f74 1367
56c50f11
DM		 1368#ifdef CONFIG_QUOTA
1369static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1370{
1371 struct ext4_sb_info *sbi = EXT4_SB(sb);
1372 char *qname;
1373
1374 if (sb_any_quota_loaded(sb) &&
1375 !sbi->s_qf_names[qtype]) {
1376 ext4_msg(sb, KERN_ERR,
1377 "Cannot change journaled "
1378 "quota options when quota turned on");
1379 return 0;
1380 }
1381 qname = match_strdup(args);
1382 if (!qname) {
1383 ext4_msg(sb, KERN_ERR,
1384 "Not enough memory for storing quotafile name");
1385 return 0;
1386 }
1387 if (sbi->s_qf_names[qtype] &&
1388 strcmp(sbi->s_qf_names[qtype], qname)) {
1389 ext4_msg(sb, KERN_ERR,
1390 "%s quota file already specified", QTYPE2NAME(qtype));
1391 kfree(qname);
1392 return 0;
1393 }
1394 sbi->s_qf_names[qtype] = qname;
1395 if (strchr(sbi->s_qf_names[qtype], '/')) {
1396 ext4_msg(sb, KERN_ERR,
1397 "quotafile must be on filesystem root");
1398 kfree(sbi->s_qf_names[qtype]);
1399 sbi->s_qf_names[qtype] = NULL;
1400 return 0;
1401 }
fd8c37ec 1402 set_opt(sb, QUOTA);
56c50f11
DM		 1403 return 1;
1404}
1405
1406static int clear_qf_name(struct super_block *sb, int qtype)
1407{
1408
1409 struct ext4_sb_info *sbi = EXT4_SB(sb);
1410
1411 if (sb_any_quota_loaded(sb) &&
1412 sbi->s_qf_names[qtype]) {
1413 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1414 " when quota turned on");
1415 return 0;
1416 }
1417 /*
1418 * The space will be released later when all options are confirmed
1419 * to be correct
1420 */
1421 sbi->s_qf_names[qtype] = NULL;
1422 return 1;
1423}
1424#endif
1425
2b2d6d01 1426static int parse_options(char *options, struct super_block *sb,
c3191067 1427 unsigned long *journal_devnum,
b3881f74 1428 unsigned int *journal_ioprio,
2b2d6d01 1429 ext4_fsblk_t *n_blocks_count, int is_remount)
ac27a0ec 1430{
617ba13b 1431 struct ext4_sb_info *sbi = EXT4_SB(sb);
2b2d6d01 1432 char *p;
ac27a0ec
DK		 1433 substring_t args[MAX_OPT_ARGS];
1434 int data_opt = 0;
1435 int option;
1436#ifdef CONFIG_QUOTA
56c50f11 1437 int qfmt;
ac27a0ec
DK		 1438#endif
1439
1440 if (!options)
1441 return 1;
1442
2b2d6d01 1443 while ((p = strsep(&options, ",")) != NULL) {
ac27a0ec
DK		 1444 int token;
1445 if (!*p)
1446 continue;
1447
15121c18
ES		 1448 /*
1449 * Initialize args struct so we know whether arg was
1450 * found; some options take optional arguments.
1451 */
7dc57615 1452 args[0].to = args[0].from = NULL;
ac27a0ec
DK		 1453 token = match_token(p, tokens, args);
1454 switch (token) {
1455 case Opt_bsd_df:
437ca0fd 1456 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
fd8c37ec 1457 clear_opt(sb, MINIX_DF);
ac27a0ec
DK		 1458 break;
1459 case Opt_minix_df:
437ca0fd 1460 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
fd8c37ec 1461 set_opt(sb, MINIX_DF);
437ca0fd 1462
ac27a0ec
DK		 1463 break;
1464 case Opt_grpid:
437ca0fd 1465 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
fd8c37ec 1466 set_opt(sb, GRPID);
437ca0fd 1467
ac27a0ec
DK		 1468 break;
1469 case Opt_nogrpid:
437ca0fd 1470 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
fd8c37ec 1471 clear_opt(sb, GRPID);
437ca0fd 1472
ac27a0ec
DK		 1473 break;
1474 case Opt_resuid:
1475 if (match_int(&args[0], &option))
1476 return 0;
1477 sbi->s_resuid = option;
1478 break;
1479 case Opt_resgid:
1480 if (match_int(&args[0], &option))
1481 return 0;
1482 sbi->s_resgid = option;
1483 break;
1484 case Opt_sb:
1485 /* handled by get_sb_block() instead of here */
1486 /* *sb_block = match_int(&args[0]); */
1487 break;
1488 case Opt_err_panic:
fd8c37ec
TT		 1489 clear_opt(sb, ERRORS_CONT);
1490 clear_opt(sb, ERRORS_RO);
1491 set_opt(sb, ERRORS_PANIC);
ac27a0ec
DK		 1492 break;
1493 case Opt_err_ro:
fd8c37ec
TT		 1494 clear_opt(sb, ERRORS_CONT);
1495 clear_opt(sb, ERRORS_PANIC);
1496 set_opt(sb, ERRORS_RO);
ac27a0ec
DK		 1497 break;
1498 case Opt_err_cont:
fd8c37ec
TT		 1499 clear_opt(sb, ERRORS_RO);
1500 clear_opt(sb, ERRORS_PANIC);
1501 set_opt(sb, ERRORS_CONT);
ac27a0ec
DK		 1502 break;
1503 case Opt_nouid32:
fd8c37ec 1504 set_opt(sb, NO_UID32);
ac27a0ec 1505 break;
ac27a0ec 1506 case Opt_debug:
fd8c37ec 1507 set_opt(sb, DEBUG);
ac27a0ec
DK		 1508 break;
1509 case Opt_oldalloc:
fd8c37ec 1510 set_opt(sb, OLDALLOC);
ac27a0ec
DK		 1511 break;
1512 case Opt_orlov:
fd8c37ec 1513 clear_opt(sb, OLDALLOC);
ac27a0ec 1514 break;
03010a33 1515#ifdef CONFIG_EXT4_FS_XATTR
ac27a0ec 1516 case Opt_user_xattr:
fd8c37ec 1517 set_opt(sb, XATTR_USER);
ac27a0ec
DK		 1518 break;
1519 case Opt_nouser_xattr:
fd8c37ec 1520 clear_opt(sb, XATTR_USER);
ac27a0ec
DK		 1521 break;
1522#else
1523 case Opt_user_xattr:
1524 case Opt_nouser_xattr:
b31e1552 1525 ext4_msg(sb, KERN_ERR, "(no)user_xattr options not supported");
ac27a0ec
DK		 1526 break;
1527#endif
03010a33 1528#ifdef CONFIG_EXT4_FS_POSIX_ACL
ac27a0ec 1529 case Opt_acl:
fd8c37ec 1530 set_opt(sb, POSIX_ACL);
ac27a0ec
DK		 1531 break;
1532 case Opt_noacl:
fd8c37ec 1533 clear_opt(sb, POSIX_ACL);
ac27a0ec
DK		 1534 break;
1535#else
1536 case Opt_acl:
1537 case Opt_noacl:
b31e1552 1538 ext4_msg(sb, KERN_ERR, "(no)acl options not supported");
ac27a0ec
DK		 1539 break;
1540#endif
ac27a0ec
DK		 1541 case Opt_journal_update:
1542 /* @@@ FIXME */
1543 /* Eventually we will want to be able to create
1544 a journal file here. For now, only allow the
1545 user to specify an existing inode to be the
1546 journal file. */
1547 if (is_remount) {
b31e1552
ES		 1548 ext4_msg(sb, KERN_ERR,
1549 "Cannot specify journal on remount");
ac27a0ec
DK		 1550 return 0;
1551 }
fd8c37ec 1552 set_opt(sb, UPDATE_JOURNAL);
ac27a0ec 1553 break;
ac27a0ec
DK		 1554 case Opt_journal_dev:
1555 if (is_remount) {
b31e1552
ES		 1556 ext4_msg(sb, KERN_ERR,
1557 "Cannot specify journal on remount");
ac27a0ec
DK		 1558 return 0;
1559 }
1560 if (match_int(&args[0], &option))
1561 return 0;
1562 *journal_devnum = option;
1563 break;
818d276c 1564 case Opt_journal_checksum:
fd8c37ec 1565 set_opt(sb, JOURNAL_CHECKSUM);
d4da6c9c 1566 break;
818d276c 1567 case Opt_journal_async_commit:
fd8c37ec
TT		 1568 set_opt(sb, JOURNAL_ASYNC_COMMIT);
1569 set_opt(sb, JOURNAL_CHECKSUM);
818d276c 1570 break;
ac27a0ec 1571 case Opt_noload:
fd8c37ec 1572 set_opt(sb, NOLOAD);
ac27a0ec
DK		 1573 break;
1574 case Opt_commit:
1575 if (match_int(&args[0], &option))
1576 return 0;
1577 if (option < 0)
1578 return 0;
1579 if (option == 0)
cd02ff0b 1580 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
ac27a0ec
DK		 1581 sbi->s_commit_interval = HZ * option;
1582 break;
30773840
TT		 1583 case Opt_max_batch_time:
1584 if (match_int(&args[0], &option))
1585 return 0;
1586 if (option < 0)
1587 return 0;
1588 if (option == 0)
1589 option = EXT4_DEF_MAX_BATCH_TIME;
1590 sbi->s_max_batch_time = option;
1591 break;
1592 case Opt_min_batch_time:
1593 if (match_int(&args[0], &option))
1594 return 0;
1595 if (option < 0)
1596 return 0;
1597 sbi->s_min_batch_time = option;
1598 break;
ac27a0ec 1599 case Opt_data_journal:
617ba13b 1600 data_opt = EXT4_MOUNT_JOURNAL_DATA;
ac27a0ec
DK		 1601 goto datacheck;
1602 case Opt_data_ordered:
617ba13b 1603 data_opt = EXT4_MOUNT_ORDERED_DATA;
ac27a0ec
DK		 1604 goto datacheck;
1605 case Opt_data_writeback:
617ba13b 1606 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
ac27a0ec
DK		 1607 datacheck:
1608 if (is_remount) {
482a7425 1609 if (test_opt(sb, DATA_FLAGS) != data_opt) {
b31e1552
ES		 1610 ext4_msg(sb, KERN_ERR,
1611 "Cannot change data mode on remount");
ac27a0ec
DK		 1612 return 0;
1613 }
1614 } else {
fd8c37ec 1615 clear_opt(sb, DATA_FLAGS);
ac27a0ec
DK		 1616 sbi->s_mount_opt |= data_opt;
1617 }
1618 break;
5bf5683a 1619 case Opt_data_err_abort:
fd8c37ec 1620 set_opt(sb, DATA_ERR_ABORT);
5bf5683a
HK		 1621 break;
1622 case Opt_data_err_ignore:
fd8c37ec 1623 clear_opt(sb, DATA_ERR_ABORT);
5bf5683a 1624 break;
ac27a0ec
DK		 1625#ifdef CONFIG_QUOTA
1626 case Opt_usrjquota:
56c50f11 1627 if (!set_qf_name(sb, USRQUOTA, &args[0]))
ac27a0ec 1628 return 0;
56c50f11
DM		 1629 break;
1630 case Opt_grpjquota:
1631 if (!set_qf_name(sb, GRPQUOTA, &args[0]))
ac27a0ec 1632 return 0;
ac27a0ec
DK		 1633 break;
1634 case Opt_offusrjquota:
56c50f11
DM		 1635 if (!clear_qf_name(sb, USRQUOTA))
1636 return 0;
1637 break;
ac27a0ec 1638 case Opt_offgrpjquota:
56c50f11 1639 if (!clear_qf_name(sb, GRPQUOTA))
ac27a0ec 1640 return 0;
ac27a0ec 1641 break;
56c50f11 1642
ac27a0ec 1643 case Opt_jqfmt_vfsold:
dfc5d03f
JK		 1644 qfmt = QFMT_VFS_OLD;
1645 goto set_qf_format;
ac27a0ec 1646 case Opt_jqfmt_vfsv0:
dfc5d03f 1647 qfmt = QFMT_VFS_V0;
5a20bdfc
JK		 1648 goto set_qf_format;
1649 case Opt_jqfmt_vfsv1:
1650 qfmt = QFMT_VFS_V1;
dfc5d03f 1651set_qf_format:
17bd13b3 1652 if (sb_any_quota_loaded(sb) &&
dfc5d03f 1653 sbi->s_jquota_fmt != qfmt) {
b31e1552 1654 ext4_msg(sb, KERN_ERR, "Cannot change "
dfc5d03f 1655 "journaled quota options when "
b31e1552 1656 "quota turned on");
dfc5d03f
JK		 1657 return 0;
1658 }
1659 sbi->s_jquota_fmt = qfmt;
ac27a0ec
DK		 1660 break;
1661 case Opt_quota:
1662 case Opt_usrquota:
fd8c37ec
TT		 1663 set_opt(sb, QUOTA);
1664 set_opt(sb, USRQUOTA);
ac27a0ec
DK		 1665 break;
1666 case Opt_grpquota:
fd8c37ec
TT		 1667 set_opt(sb, QUOTA);
1668 set_opt(sb, GRPQUOTA);
ac27a0ec
DK		 1669 break;
1670 case Opt_noquota:
17bd13b3 1671 if (sb_any_quota_loaded(sb)) {
b31e1552
ES		 1672 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1673 "options when quota turned on");
ac27a0ec
DK		 1674 return 0;
1675 }
fd8c37ec
TT		 1676 clear_opt(sb, QUOTA);
1677 clear_opt(sb, USRQUOTA);
1678 clear_opt(sb, GRPQUOTA);
ac27a0ec
DK		 1679 break;
1680#else
1681 case Opt_quota:
1682 case Opt_usrquota:
1683 case Opt_grpquota:
b31e1552
ES		 1684 ext4_msg(sb, KERN_ERR,
1685 "quota options not supported");
cd59e7b9 1686 break;
ac27a0ec
DK		 1687 case Opt_usrjquota:
1688 case Opt_grpjquota:
1689 case Opt_offusrjquota:
1690 case Opt_offgrpjquota:
1691 case Opt_jqfmt_vfsold:
1692 case Opt_jqfmt_vfsv0:
5a20bdfc 1693 case Opt_jqfmt_vfsv1:
b31e1552
ES		 1694 ext4_msg(sb, KERN_ERR,
1695 "journaled quota options not supported");
ac27a0ec
DK		 1696 break;
1697 case Opt_noquota:
1698 break;
1699#endif
1700 case Opt_abort:
4ab2f15b 1701 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
ac27a0ec 1702 break;
06705bff 1703 case Opt_nobarrier:
fd8c37ec 1704 clear_opt(sb, BARRIER);
06705bff 1705 break;
ac27a0ec 1706 case Opt_barrier:
15121c18
ES		 1707 if (args[0].from) {
1708 if (match_int(&args[0], &option))
1709 return 0;
1710 } else
1711 option = 1; /* No argument, default to 1 */
ac27a0ec 1712 if (option)
fd8c37ec 1713 set_opt(sb, BARRIER);
ac27a0ec 1714 else
fd8c37ec 1715 clear_opt(sb, BARRIER);
ac27a0ec
DK		 1716 break;
1717 case Opt_ignore:
1718 break;
1719 case Opt_resize:
1720 if (!is_remount) {
b31e1552
ES		 1721 ext4_msg(sb, KERN_ERR,
1722 "resize option only available "
1723 "for remount");
ac27a0ec
DK		 1724 return 0;
1725 }
1726 if (match_int(&args[0], &option) != 0)
1727 return 0;
1728 *n_blocks_count = option;
1729 break;
1730 case Opt_nobh:
206f7ab4
CH		 1731 ext4_msg(sb, KERN_WARNING,
1732 "Ignoring deprecated nobh option");
ac27a0ec
DK		 1733 break;
1734 case Opt_bh:
206f7ab4
CH		 1735 ext4_msg(sb, KERN_WARNING,
1736 "Ignoring deprecated bh option");
ac27a0ec 1737 break;
25ec56b5 1738 case Opt_i_version:
fd8c37ec 1739 set_opt(sb, I_VERSION);
25ec56b5
JNC		 1740 sb->s_flags |= MS_I_VERSION;
1741 break;
dd919b98 1742 case Opt_nodelalloc:
fd8c37ec 1743 clear_opt(sb, DELALLOC);
dd919b98 1744 break;
1449032b 1745 case Opt_mblk_io_submit:
fd8c37ec 1746 set_opt(sb, MBLK_IO_SUBMIT);
1449032b
TT		 1747 break;
1748 case Opt_nomblk_io_submit:
fd8c37ec 1749 clear_opt(sb, MBLK_IO_SUBMIT);
1449032b 1750 break;
c9de560d
AT		 1751 case Opt_stripe:
1752 if (match_int(&args[0], &option))
1753 return 0;
1754 if (option < 0)
1755 return 0;
1756 sbi->s_stripe = option;
1757 break;
64769240 1758 case Opt_delalloc:
fd8c37ec 1759 set_opt(sb, DELALLOC);
64769240 1760 break;
6fd058f7 1761 case Opt_block_validity:
fd8c37ec 1762 set_opt(sb, BLOCK_VALIDITY);
6fd058f7
TT		 1763 break;
1764 case Opt_noblock_validity:
fd8c37ec 1765 clear_opt(sb, BLOCK_VALIDITY);
6fd058f7 1766 break;
240799cd
TT		 1767 case Opt_inode_readahead_blks:
1768 if (match_int(&args[0], &option))
1769 return 0;
1770 if (option < 0 || option > (1 << 30))
1771 return 0;
5dbd571d 1772 if (option && !is_power_of_2(option)) {
b31e1552
ES		 1773 ext4_msg(sb, KERN_ERR,
1774 "EXT4-fs: inode_readahead_blks"
1775 " must be a power of 2");
3197ebdb
TT		 1776 return 0;
1777 }
240799cd
TT		 1778 sbi->s_inode_readahead_blks = option;
1779 break;
b3881f74
TT		 1780 case Opt_journal_ioprio:
1781 if (match_int(&args[0], &option))
1782 return 0;
1783 if (option < 0 || option > 7)
1784 break;
1785 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1786 option);
1787 break;
06705bff 1788 case Opt_noauto_da_alloc:
fd8c37ec 1789 set_opt(sb, NO_AUTO_DA_ALLOC);
06705bff 1790 break;
afd4672d 1791 case Opt_auto_da_alloc:
15121c18
ES		 1792 if (args[0].from) {
1793 if (match_int(&args[0], &option))
1794 return 0;
1795 } else
1796 option = 1; /* No argument, default to 1 */
afd4672d 1797 if (option)
fd8c37ec 1798 clear_opt(sb, NO_AUTO_DA_ALLOC);
afd4672d 1799 else
fd8c37ec 1800 set_opt(sb,NO_AUTO_DA_ALLOC);
afd4672d 1801 break;
5328e635 1802 case Opt_discard:
fd8c37ec 1803 set_opt(sb, DISCARD);
5328e635
ES		 1804 break;
1805 case Opt_nodiscard:
fd8c37ec 1806 clear_opt(sb, DISCARD);
5328e635 1807 break;
744692dc 1808 case Opt_dioread_nolock:
fd8c37ec 1809 set_opt(sb, DIOREAD_NOLOCK);
744692dc
JZ		 1810 break;
1811 case Opt_dioread_lock:
fd8c37ec 1812 clear_opt(sb, DIOREAD_NOLOCK);
744692dc 1813 break;
bfff6873 1814 case Opt_init_inode_table:
fd8c37ec 1815 set_opt(sb, INIT_INODE_TABLE);
bfff6873
LC		 1816 if (args[0].from) {
1817 if (match_int(&args[0], &option))
1818 return 0;
1819 } else
1820 option = EXT4_DEF_LI_WAIT_MULT;
1821 if (option < 0)
1822 return 0;
1823 sbi->s_li_wait_mult = option;
1824 break;
1825 case Opt_noinit_inode_table:
fd8c37ec 1826 clear_opt(sb, INIT_INODE_TABLE);
bfff6873 1827 break;
ac27a0ec 1828 default:
b31e1552
ES		 1829 ext4_msg(sb, KERN_ERR,
1830 "Unrecognized mount option \"%s\" "
1831 "or missing value", p);
ac27a0ec
DK		 1832 return 0;
1833 }
1834 }
1835#ifdef CONFIG_QUOTA
1836 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
482a7425 1837 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
fd8c37ec 1838 clear_opt(sb, USRQUOTA);
ac27a0ec 1839
482a7425 1840 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
fd8c37ec 1841 clear_opt(sb, GRPQUOTA);
ac27a0ec 1842
56c50f11 1843 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
b31e1552
ES		 1844 ext4_msg(sb, KERN_ERR, "old and new quota "
1845 "format mixing");
ac27a0ec
DK		 1846 return 0;
1847 }
1848
1849 if (!sbi->s_jquota_fmt) {
b31e1552
ES		 1850 ext4_msg(sb, KERN_ERR, "journaled quota format "
1851 "not specified");
ac27a0ec
DK		 1852 return 0;
1853 }
1854 } else {
1855 if (sbi->s_jquota_fmt) {
b31e1552 1856 ext4_msg(sb, KERN_ERR, "journaled quota format "
2c8be6b2 1857 "specified with no journaling "
b31e1552 1858 "enabled");
ac27a0ec
DK		 1859 return 0;
1860 }
1861 }
1862#endif
1863 return 1;
1864}
1865
617ba13b 1866static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 1867 int read_only)
1868{
617ba13b 1869 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 1870 int res = 0;
1871
617ba13b 1872 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
b31e1552
ES		 1873 ext4_msg(sb, KERN_ERR, "revision level too high, "
1874 "forcing read-only mode");
ac27a0ec
DK		 1875 res = MS_RDONLY;
1876 }
1877 if (read_only)
1878 return res;
617ba13b 1879 if (!(sbi->s_mount_state & EXT4_VALID_FS))
b31e1552
ES		 1880 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1881 "running e2fsck is recommended");
617ba13b 1882 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
b31e1552
ES		 1883 ext4_msg(sb, KERN_WARNING,
1884 "warning: mounting fs with errors, "
1885 "running e2fsck is recommended");
ac27a0ec
DK		 1886 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1887 le16_to_cpu(es->s_mnt_count) >=
1888 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
b31e1552
ES		 1889 ext4_msg(sb, KERN_WARNING,
1890 "warning: maximal mount count reached, "
1891 "running e2fsck is recommended");
ac27a0ec
DK		 1892 else if (le32_to_cpu(es->s_checkinterval) &&
1893 (le32_to_cpu(es->s_lastcheck) +
1894 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
b31e1552
ES		 1895 ext4_msg(sb, KERN_WARNING,
1896 "warning: checktime reached, "
1897 "running e2fsck is recommended");
0b8e58a1 1898 if (!sbi->s_journal)
0390131b 1899 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec 1900 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 1901 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 1902 le16_add_cpu(&es->s_mnt_count, 1);
ac27a0ec 1903 es->s_mtime = cpu_to_le32(get_seconds());
617ba13b 1904 ext4_update_dynamic_rev(sb);
0390131b
FM		 1905 if (sbi->s_journal)
1906 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 1907
e2d67052 1908 ext4_commit_super(sb, 1);
ac27a0ec 1909 if (test_opt(sb, DEBUG))
a9df9a49 1910 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
a2595b8a 1911 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
ac27a0ec
DK		 1912 sb->s_blocksize,
1913 sbi->s_groups_count,
617ba13b
MC		 1914 EXT4_BLOCKS_PER_GROUP(sb),
1915 EXT4_INODES_PER_GROUP(sb),
a2595b8a 1916 sbi->s_mount_opt, sbi->s_mount_opt2);
ac27a0ec 1917
ac27a0ec
DK		 1918 return res;
1919}
1920
772cb7c8
JS		 1921static int ext4_fill_flex_info(struct super_block *sb)
1922{
1923 struct ext4_sb_info *sbi = EXT4_SB(sb);
1924 struct ext4_group_desc *gdp = NULL;
772cb7c8
JS		 1925 ext4_group_t flex_group_count;
1926 ext4_group_t flex_group;
1927 int groups_per_flex = 0;
c5ca7c76 1928 size_t size;
772cb7c8
JS		 1929 int i;
1930
503358ae
TT		 1931 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1932 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1933
1934 if (groups_per_flex < 2) {
772cb7c8
JS		 1935 sbi->s_log_groups_per_flex = 0;
1936 return 1;
1937 }
1938
c62a11fd
FB		 1939 /* We allocate both existing and potentially added groups */
1940 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
d94e99a6
AK		 1941 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1942 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
c5ca7c76
TT		 1943 size = flex_group_count * sizeof(struct flex_groups);
1944 sbi->s_flex_groups = kzalloc(size, GFP_KERNEL);
1945 if (sbi->s_flex_groups == NULL) {
94de56ab
JP		 1946 sbi->s_flex_groups = vzalloc(size);
1947 if (sbi->s_flex_groups == NULL) {
1948 ext4_msg(sb, KERN_ERR,
1949 "not enough memory for %u flex groups",
1950 flex_group_count);
1951 goto failed;
1952 }
772cb7c8 1953 }
772cb7c8 1954
772cb7c8 1955 for (i = 0; i < sbi->s_groups_count; i++) {
88b6edd1 1956 gdp = ext4_get_group_desc(sb, i, NULL);
772cb7c8
JS		 1957
1958 flex_group = ext4_flex_group(sbi, i);
7ad9bb65
TT		 1959 atomic_add(ext4_free_inodes_count(sb, gdp),
1960 &sbi->s_flex_groups[flex_group].free_inodes);
1961 atomic_add(ext4_free_blks_count(sb, gdp),
1962 &sbi->s_flex_groups[flex_group].free_blocks);
1963 atomic_add(ext4_used_dirs_count(sb, gdp),
1964 &sbi->s_flex_groups[flex_group].used_dirs);
772cb7c8
JS		 1965 }
1966
1967 return 1;
1968failed:
1969 return 0;
1970}
1971
717d50e4
AD		 1972__le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1973 struct ext4_group_desc *gdp)
1974{
1975 __u16 crc = 0;
1976
1977 if (sbi->s_es->s_feature_ro_compat &
1978 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1979 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1980 __le32 le_group = cpu_to_le32(block_group);
1981
1982 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1983 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1984 crc = crc16(crc, (__u8 *)gdp, offset);
1985 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1986 /* for checksum of struct ext4_group_desc do the rest...*/
1987 if ((sbi->s_es->s_feature_incompat &
1988 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1989 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1990 crc = crc16(crc, (__u8 *)gdp + offset,
1991 le16_to_cpu(sbi->s_es->s_desc_size) -
1992 offset);
1993 }
1994
1995 return cpu_to_le16(crc);
1996}
1997
1998int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1999 struct ext4_group_desc *gdp)
2000{
2001 if ((sbi->s_es->s_feature_ro_compat &
2002 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
2003 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
2004 return 0;
2005
2006 return 1;
2007}
2008
ac27a0ec 2009/* Called at mount-time, super-block is locked */
bfff6873
LC		 2010static int ext4_check_descriptors(struct super_block *sb,
2011 ext4_group_t *first_not_zeroed)
ac27a0ec 2012{
617ba13b
MC		 2013 struct ext4_sb_info *sbi = EXT4_SB(sb);
2014 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2015 ext4_fsblk_t last_block;
bd81d8ee
LV		 2016 ext4_fsblk_t block_bitmap;
2017 ext4_fsblk_t inode_bitmap;
2018 ext4_fsblk_t inode_table;
ce421581 2019 int flexbg_flag = 0;
bfff6873 2020 ext4_group_t i, grp = sbi->s_groups_count;
ac27a0ec 2021
ce421581
JS		 2022 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2023 flexbg_flag = 1;
2024
af5bc92d 2025 ext4_debug("Checking group descriptors");
ac27a0ec 2026
197cd65a
AM		 2027 for (i = 0; i < sbi->s_groups_count; i++) {
2028 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2029
ce421581 2030 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 2031 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 2032 else
2033 last_block = first_block +
617ba13b 2034 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 2035
bfff6873
LC		 2036 if ((grp == sbi->s_groups_count) &&
2037 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2038 grp = i;
2039
8fadc143 2040 block_bitmap = ext4_block_bitmap(sb, gdp);
2b2d6d01 2041 if (block_bitmap < first_block || block_bitmap > last_block) {
b31e1552 2042 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2043 "Block bitmap for group %u not in group "
b31e1552 2044 "(block %llu)!", i, block_bitmap);
ac27a0ec
DK		 2045 return 0;
2046 }
8fadc143 2047 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2b2d6d01 2048 if (inode_bitmap < first_block || inode_bitmap > last_block) {
b31e1552 2049 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2050 "Inode bitmap for group %u not in group "
b31e1552 2051 "(block %llu)!", i, inode_bitmap);
ac27a0ec
DK		 2052 return 0;
2053 }
8fadc143 2054 inode_table = ext4_inode_table(sb, gdp);
bd81d8ee 2055 if (inode_table < first_block ||
2b2d6d01 2056 inode_table + sbi->s_itb_per_group - 1 > last_block) {
b31e1552 2057 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2058 "Inode table for group %u not in group "
b31e1552 2059 "(block %llu)!", i, inode_table);
ac27a0ec
DK		 2060 return 0;
2061 }
955ce5f5 2062 ext4_lock_group(sb, i);
717d50e4 2063 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
b31e1552
ES		 2064 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2065 "Checksum for group %u failed (%u!=%u)",
2066 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2067 gdp)), le16_to_cpu(gdp->bg_checksum));
7ee1ec4c 2068 if (!(sb->s_flags & MS_RDONLY)) {
955ce5f5 2069 ext4_unlock_group(sb, i);
8a266467 2070 return 0;
7ee1ec4c 2071 }
717d50e4 2072 }
955ce5f5 2073 ext4_unlock_group(sb, i);
ce421581
JS		 2074 if (!flexbg_flag)
2075 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec 2076 }
bfff6873
LC		 2077 if (NULL != first_not_zeroed)
2078 *first_not_zeroed = grp;
ac27a0ec 2079
bd81d8ee 2080 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
0b8e58a1 2081 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
ac27a0ec
DK		 2082 return 1;
2083}
2084
617ba13b 2085/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 2086 * the superblock) which were deleted from all directories, but held open by
2087 * a process at the time of a crash. We walk the list and try to delete these
2088 * inodes at recovery time (only with a read-write filesystem).
2089 *
2090 * In order to keep the orphan inode chain consistent during traversal (in
2091 * case of crash during recovery), we link each inode into the superblock
2092 * orphan list_head and handle it the same way as an inode deletion during
2093 * normal operation (which journals the operations for us).
2094 *
2095 * We only do an iget() and an iput() on each inode, which is very safe if we
2096 * accidentally point at an in-use or already deleted inode. The worst that
2097 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 2098 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 2099 * e2fsck was run on this filesystem, and it must have already done the orphan
2100 * inode cleanup for us, so we can safely abort without any further action.
2101 */
2b2d6d01
TT		 2102static void ext4_orphan_cleanup(struct super_block *sb,
2103 struct ext4_super_block *es)
ac27a0ec
DK		 2104{
2105 unsigned int s_flags = sb->s_flags;
2106 int nr_orphans = 0, nr_truncates = 0;
2107#ifdef CONFIG_QUOTA
2108 int i;
2109#endif
2110 if (!es->s_last_orphan) {
2111 jbd_debug(4, "no orphan inodes to clean up\n");
2112 return;
2113 }
2114
a8f48a95 2115 if (bdev_read_only(sb->s_bdev)) {
b31e1552
ES		 2116 ext4_msg(sb, KERN_ERR, "write access "
2117 "unavailable, skipping orphan cleanup");
a8f48a95
ES		 2118 return;
2119 }
2120
d39195c3
AG		 2121 /* Check if feature set would not allow a r/w mount */
2122 if (!ext4_feature_set_ok(sb, 0)) {
2123 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2124 "unknown ROCOMPAT features");
2125 return;
2126 }
2127
617ba13b 2128 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
ac27a0ec
DK		 2129 if (es->s_last_orphan)
2130 jbd_debug(1, "Errors on filesystem, "
2131 "clearing orphan list.\n");
2132 es->s_last_orphan = 0;
2133 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2134 return;
2135 }
2136
2137 if (s_flags & MS_RDONLY) {
b31e1552 2138 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
ac27a0ec
DK		 2139 sb->s_flags &= ~MS_RDONLY;
2140 }
2141#ifdef CONFIG_QUOTA
2142 /* Needed for iput() to work correctly and not trash data */
2143 sb->s_flags |= MS_ACTIVE;
2144 /* Turn on quotas so that they are updated correctly */
2145 for (i = 0; i < MAXQUOTAS; i++) {
617ba13b
MC		 2146 if (EXT4_SB(sb)->s_qf_names[i]) {
2147 int ret = ext4_quota_on_mount(sb, i);
ac27a0ec 2148 if (ret < 0)
b31e1552
ES		 2149 ext4_msg(sb, KERN_ERR,
2150 "Cannot turn on journaled "
2151 "quota: error %d", ret);
ac27a0ec
DK		 2152 }
2153 }
2154#endif
2155
2156 while (es->s_last_orphan) {
2157 struct inode *inode;
2158
97bd42b9
JB		 2159 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2160 if (IS_ERR(inode)) {
ac27a0ec
DK		 2161 es->s_last_orphan = 0;
2162 break;
2163 }
2164
617ba13b 2165 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
871a2931 2166 dquot_initialize(inode);
ac27a0ec 2167 if (inode->i_nlink) {
b31e1552
ES		 2168 ext4_msg(sb, KERN_DEBUG,
2169 "%s: truncating inode %lu to %lld bytes",
46e665e9 2170 __func__, inode->i_ino, inode->i_size);
e5f8eab8 2171 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
ac27a0ec 2172 inode->i_ino, inode->i_size);
617ba13b 2173 ext4_truncate(inode);
ac27a0ec
DK		 2174 nr_truncates++;
2175 } else {
b31e1552
ES		 2176 ext4_msg(sb, KERN_DEBUG,
2177 "%s: deleting unreferenced inode %lu",
46e665e9 2178 __func__, inode->i_ino);
ac27a0ec
DK		 2179 jbd_debug(2, "deleting unreferenced inode %lu\n",
2180 inode->i_ino);
2181 nr_orphans++;
2182 }
2183 iput(inode); /* The delete magic happens here! */
2184 }
2185
2b2d6d01 2186#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
ac27a0ec
DK		 2187
2188 if (nr_orphans)
b31e1552
ES		 2189 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2190 PLURAL(nr_orphans));
ac27a0ec 2191 if (nr_truncates)
b31e1552
ES		 2192 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2193 PLURAL(nr_truncates));
ac27a0ec
DK		 2194#ifdef CONFIG_QUOTA
2195 /* Turn quotas off */
2196 for (i = 0; i < MAXQUOTAS; i++) {
2197 if (sb_dqopt(sb)->files[i])
287a8095 2198 dquot_quota_off(sb, i);
ac27a0ec
DK		 2199 }
2200#endif
2201 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2202}
0b8e58a1 2203
cd2291a4
ES		 2204/*
2205 * Maximal extent format file size.
2206 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2207 * extent format containers, within a sector_t, and within i_blocks
2208 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2209 * so that won't be a limiting factor.
2210 *
2211 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2212 */
f287a1a5 2213static loff_t ext4_max_size(int blkbits, int has_huge_files)
cd2291a4
ES		 2214{
2215 loff_t res;
2216 loff_t upper_limit = MAX_LFS_FILESIZE;
2217
2218 /* small i_blocks in vfs inode? */
f287a1a5 2219 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
cd2291a4 2220 /*
90c699a9 2221 * CONFIG_LBDAF is not enabled implies the inode
cd2291a4
ES		 2222 * i_block represent total blocks in 512 bytes
2223 * 32 == size of vfs inode i_blocks * 8
2224 */
2225 upper_limit = (1LL << 32) - 1;
2226
2227 /* total blocks in file system block size */
2228 upper_limit >>= (blkbits - 9);
2229 upper_limit <<= blkbits;
2230 }
2231
2232 /* 32-bit extent-start container, ee_block */
2233 res = 1LL << 32;
2234 res <<= blkbits;
2235 res -= 1;
2236
2237 /* Sanity check against vm- & vfs- imposed limits */
2238 if (res > upper_limit)
2239 res = upper_limit;
2240
2241 return res;
2242}
ac27a0ec 2243
ac27a0ec 2244/*
cd2291a4 2245 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 2246 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2247 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 2248 */
f287a1a5 2249static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
ac27a0ec 2250{
617ba13b 2251 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 2252 int meta_blocks;
2253 loff_t upper_limit;
0b8e58a1
AD		 2254 /* This is calculated to be the largest file size for a dense, block
2255 * mapped file such that the file's total number of 512-byte sectors,
2256 * including data and all indirect blocks, does not exceed (2^48 - 1).
2257 *
2258 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2259 * number of 512-byte sectors of the file.
0fc1b451
AK		 2260 */
2261
f287a1a5 2262 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
0fc1b451 2263 /*
90c699a9 2264 * !has_huge_files or CONFIG_LBDAF not enabled implies that
0b8e58a1
AD		 2265 * the inode i_block field represents total file blocks in
2266 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
0fc1b451
AK		 2267 */
2268 upper_limit = (1LL << 32) - 1;
2269
2270 /* total blocks in file system block size */
2271 upper_limit >>= (bits - 9);
2272
2273 } else {
8180a562
AK		 2274 /*
2275 * We use 48 bit ext4_inode i_blocks
2276 * With EXT4_HUGE_FILE_FL set the i_blocks
2277 * represent total number of blocks in
2278 * file system block size
2279 */
0fc1b451
AK		 2280 upper_limit = (1LL << 48) - 1;
2281
0fc1b451
AK		 2282 }
2283
2284 /* indirect blocks */
2285 meta_blocks = 1;
2286 /* double indirect blocks */
2287 meta_blocks += 1 + (1LL << (bits-2));
2288 /* tripple indirect blocks */
2289 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2290
2291 upper_limit -= meta_blocks;
2292 upper_limit <<= bits;
ac27a0ec
DK		 2293
2294 res += 1LL << (bits-2);
2295 res += 1LL << (2*(bits-2));
2296 res += 1LL << (3*(bits-2));
2297 res <<= bits;
2298 if (res > upper_limit)
2299 res = upper_limit;
0fc1b451
AK		 2300
2301 if (res > MAX_LFS_FILESIZE)
2302 res = MAX_LFS_FILESIZE;
2303
ac27a0ec
DK		 2304 return res;
2305}
2306
617ba13b 2307static ext4_fsblk_t descriptor_loc(struct super_block *sb,
0b8e58a1 2308 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 2309{
617ba13b 2310 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 2311 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 2312 int has_super = 0;
2313
2314 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2315
617ba13b 2316 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
ac27a0ec 2317 nr < first_meta_bg)
70bbb3e0 2318 return logical_sb_block + nr + 1;
ac27a0ec 2319 bg = sbi->s_desc_per_block * nr;
617ba13b 2320 if (ext4_bg_has_super(sb, bg))
ac27a0ec 2321 has_super = 1;
0b8e58a1 2322
617ba13b 2323 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 2324}
2325
c9de560d
AT		 2326/**
2327 * ext4_get_stripe_size: Get the stripe size.
2328 * @sbi: In memory super block info
2329 *
2330 * If we have specified it via mount option, then
2331 * use the mount option value. If the value specified at mount time is
2332 * greater than the blocks per group use the super block value.
2333 * If the super block value is greater than blocks per group return 0.
2334 * Allocator needs it be less than blocks per group.
2335 *
2336 */
2337static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2338{
2339 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2340 unsigned long stripe_width =
2341 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2342
2343 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2344 return sbi->s_stripe;
2345
2346 if (stripe_width <= sbi->s_blocks_per_group)
2347 return stripe_width;
2348
2349 if (stride <= sbi->s_blocks_per_group)
2350 return stride;
2351
2352 return 0;
2353}
ac27a0ec 2354
3197ebdb
TT		 2355/* sysfs supprt */
2356
2357struct ext4_attr {
2358 struct attribute attr;
2359 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
60e6679e 2360 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
3197ebdb
TT		 2361 const char *, size_t);
2362 int offset;
2363};
2364
2365static int parse_strtoul(const char *buf,
2366 unsigned long max, unsigned long *value)
2367{
2368 char *endp;
2369
e7d2860b
AGR		 2370 *value = simple_strtoul(skip_spaces(buf), &endp, 0);
2371 endp = skip_spaces(endp);
3197ebdb
TT		 2372 if (*endp || *value > max)
2373 return -EINVAL;
2374
2375 return 0;
2376}
2377
2378static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2379 struct ext4_sb_info *sbi,
2380 char *buf)
2381{
2382 return snprintf(buf, PAGE_SIZE, "%llu\n",
2383 (s64) percpu_counter_sum(&sbi->s_dirtyblocks_counter));
2384}
2385
2386static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2387 struct ext4_sb_info *sbi, char *buf)
2388{
2389 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2390
f613dfcb
TT		 2391 if (!sb->s_bdev->bd_part)
2392 return snprintf(buf, PAGE_SIZE, "0\n");
3197ebdb
TT		 2393 return snprintf(buf, PAGE_SIZE, "%lu\n",
2394 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2395 sbi->s_sectors_written_start) >> 1);
2396}
2397
2398static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2399 struct ext4_sb_info *sbi, char *buf)
2400{
2401 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2402
f613dfcb
TT		 2403 if (!sb->s_bdev->bd_part)
2404 return snprintf(buf, PAGE_SIZE, "0\n");
3197ebdb 2405 return snprintf(buf, PAGE_SIZE, "%llu\n",
a6b43e38 2406 (unsigned long long)(sbi->s_kbytes_written +
3197ebdb 2407 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
a6b43e38 2408 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
3197ebdb
TT		 2409}
2410
2411static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2412 struct ext4_sb_info *sbi,
2413 const char *buf, size_t count)
2414{
2415 unsigned long t;
2416
2417 if (parse_strtoul(buf, 0x40000000, &t))
2418 return -EINVAL;
2419
5dbd571d 2420 if (t && !is_power_of_2(t))
3197ebdb
TT		 2421 return -EINVAL;
2422
2423 sbi->s_inode_readahead_blks = t;
2424 return count;
2425}
2426
2427static ssize_t sbi_ui_show(struct ext4_attr *a,
0b8e58a1 2428 struct ext4_sb_info *sbi, char *buf)
3197ebdb
TT		 2429{
2430 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2431
2432 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2433}
2434
2435static ssize_t sbi_ui_store(struct ext4_attr *a,
2436 struct ext4_sb_info *sbi,
2437 const char *buf, size_t count)
2438{
2439 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2440 unsigned long t;
2441
2442 if (parse_strtoul(buf, 0xffffffff, &t))
2443 return -EINVAL;
2444 *ui = t;
2445 return count;
2446}
2447
2448#define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2449static struct ext4_attr ext4_attr_##_name = { \
2450 .attr = {.name = __stringify(_name), .mode = _mode }, \
2451 .show = _show, \
2452 .store = _store, \
2453 .offset = offsetof(struct ext4_sb_info, _elname), \
2454}
2455#define EXT4_ATTR(name, mode, show, store) \
2456static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2457
857ac889 2458#define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
3197ebdb
TT		 2459#define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2460#define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2461#define EXT4_RW_ATTR_SBI_UI(name, elname) \
2462 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2463#define ATTR_LIST(name) &ext4_attr_##name.attr
2464
2465EXT4_RO_ATTR(delayed_allocation_blocks);
2466EXT4_RO_ATTR(session_write_kbytes);
2467EXT4_RO_ATTR(lifetime_write_kbytes);
2468EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2469 inode_readahead_blks_store, s_inode_readahead_blks);
11013911 2470EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
3197ebdb
TT		 2471EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2472EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2473EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2474EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2475EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2476EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
55138e0b 2477EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
3197ebdb
TT		 2478
2479static struct attribute *ext4_attrs[] = {
2480 ATTR_LIST(delayed_allocation_blocks),
2481 ATTR_LIST(session_write_kbytes),
2482 ATTR_LIST(lifetime_write_kbytes),
2483 ATTR_LIST(inode_readahead_blks),
11013911 2484 ATTR_LIST(inode_goal),
3197ebdb
TT		 2485 ATTR_LIST(mb_stats),
2486 ATTR_LIST(mb_max_to_scan),
2487 ATTR_LIST(mb_min_to_scan),
2488 ATTR_LIST(mb_order2_req),
2489 ATTR_LIST(mb_stream_req),
2490 ATTR_LIST(mb_group_prealloc),
55138e0b 2491 ATTR_LIST(max_writeback_mb_bump),
3197ebdb
TT		 2492 NULL,
2493};
2494
857ac889
LC		 2495/* Features this copy of ext4 supports */
2496EXT4_INFO_ATTR(lazy_itable_init);
27ee40df 2497EXT4_INFO_ATTR(batched_discard);
857ac889
LC		 2498
2499static struct attribute *ext4_feat_attrs[] = {
2500 ATTR_LIST(lazy_itable_init),
27ee40df 2501 ATTR_LIST(batched_discard),
857ac889
LC		 2502 NULL,
2503};
2504
3197ebdb
TT		 2505static ssize_t ext4_attr_show(struct kobject *kobj,
2506 struct attribute *attr, char *buf)
2507{
2508 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2509 s_kobj);
2510 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2511
2512 return a->show ? a->show(a, sbi, buf) : 0;
2513}
2514
2515static ssize_t ext4_attr_store(struct kobject *kobj,
2516 struct attribute *attr,
2517 const char *buf, size_t len)
2518{
2519 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2520 s_kobj);
2521 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2522
2523 return a->store ? a->store(a, sbi, buf, len) : 0;
2524}
2525
2526static void ext4_sb_release(struct kobject *kobj)
2527{
2528 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2529 s_kobj);
2530 complete(&sbi->s_kobj_unregister);
2531}
2532
52cf25d0 2533static const struct sysfs_ops ext4_attr_ops = {
3197ebdb
TT		 2534 .show = ext4_attr_show,
2535 .store = ext4_attr_store,
2536};
2537
2538static struct kobj_type ext4_ktype = {
2539 .default_attrs = ext4_attrs,
2540 .sysfs_ops = &ext4_attr_ops,
2541 .release = ext4_sb_release,
2542};
2543
857ac889
LC		 2544static void ext4_feat_release(struct kobject *kobj)
2545{
2546 complete(&ext4_feat->f_kobj_unregister);
2547}
2548
2549static struct kobj_type ext4_feat_ktype = {
2550 .default_attrs = ext4_feat_attrs,
2551 .sysfs_ops = &ext4_attr_ops,
2552 .release = ext4_feat_release,
2553};
2554
a13fb1a4
ES		 2555/*
2556 * Check whether this filesystem can be mounted based on
2557 * the features present and the RDONLY/RDWR mount requested.
2558 * Returns 1 if this filesystem can be mounted as requested,
2559 * 0 if it cannot be.
2560 */
2561static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2562{
2563 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2564 ext4_msg(sb, KERN_ERR,
2565 "Couldn't mount because of "
2566 "unsupported optional features (%x)",
2567 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2568 ~EXT4_FEATURE_INCOMPAT_SUPP));
2569 return 0;
2570 }
2571
2572 if (readonly)
2573 return 1;
2574
2575 /* Check that feature set is OK for a read-write mount */
2576 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2577 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2578 "unsupported optional features (%x)",
2579 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2580 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2581 return 0;
2582 }
2583 /*
2584 * Large file size enabled file system can only be mounted
2585 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2586 */
2587 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2588 if (sizeof(blkcnt_t) < sizeof(u64)) {
2589 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2590 "cannot be mounted RDWR without "
2591 "CONFIG_LBDAF");
2592 return 0;
2593 }
2594 }
2595 return 1;
2596}
2597
66e61a9e
TT		 2598/*
2599 * This function is called once a day if we have errors logged
2600 * on the file system
2601 */
2602static void print_daily_error_info(unsigned long arg)
2603{
2604 struct super_block *sb = (struct super_block *) arg;
2605 struct ext4_sb_info *sbi;
2606 struct ext4_super_block *es;
2607
2608 sbi = EXT4_SB(sb);
2609 es = sbi->s_es;
2610
2611 if (es->s_error_count)
2612 ext4_msg(sb, KERN_NOTICE, "error count: %u",
2613 le32_to_cpu(es->s_error_count));
2614 if (es->s_first_error_time) {
2615 printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d",
2616 sb->s_id, le32_to_cpu(es->s_first_error_time),
2617 (int) sizeof(es->s_first_error_func),
2618 es->s_first_error_func,
2619 le32_to_cpu(es->s_first_error_line));
2620 if (es->s_first_error_ino)
2621 printk(": inode %u",
2622 le32_to_cpu(es->s_first_error_ino));
2623 if (es->s_first_error_block)
2624 printk(": block %llu", (unsigned long long)
2625 le64_to_cpu(es->s_first_error_block));
2626 printk("\n");
2627 }
2628 if (es->s_last_error_time) {
2629 printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d",
2630 sb->s_id, le32_to_cpu(es->s_last_error_time),
2631 (int) sizeof(es->s_last_error_func),
2632 es->s_last_error_func,
2633 le32_to_cpu(es->s_last_error_line));
2634 if (es->s_last_error_ino)
2635 printk(": inode %u",
2636 le32_to_cpu(es->s_last_error_ino));
2637 if (es->s_last_error_block)
2638 printk(": block %llu", (unsigned long long)
2639 le64_to_cpu(es->s_last_error_block));
2640 printk("\n");
2641 }
2642 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2643}
2644
bfff6873
LC		 2645static void ext4_lazyinode_timeout(unsigned long data)
2646{
2647 struct task_struct *p = (struct task_struct *)data;
2648 wake_up_process(p);
2649}
2650
2651/* Find next suitable group and run ext4_init_inode_table */
2652static int ext4_run_li_request(struct ext4_li_request *elr)
2653{
2654 struct ext4_group_desc *gdp = NULL;
2655 ext4_group_t group, ngroups;
2656 struct super_block *sb;
2657 unsigned long timeout = 0;
2658 int ret = 0;
2659
2660 sb = elr->lr_super;
2661 ngroups = EXT4_SB(sb)->s_groups_count;
2662
2663 for (group = elr->lr_next_group; group < ngroups; group++) {
2664 gdp = ext4_get_group_desc(sb, group, NULL);
2665 if (!gdp) {
2666 ret = 1;
2667 break;
2668 }
2669
2670 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2671 break;
2672 }
2673
2674 if (group == ngroups)
2675 ret = 1;
2676
2677 if (!ret) {
2678 timeout = jiffies;
2679 ret = ext4_init_inode_table(sb, group,
2680 elr->lr_timeout ? 0 : 1);
2681 if (elr->lr_timeout == 0) {
2682 timeout = jiffies - timeout;
2683 if (elr->lr_sbi->s_li_wait_mult)
2684 timeout *= elr->lr_sbi->s_li_wait_mult;
2685 else
2686 timeout *= 20;
2687 elr->lr_timeout = timeout;
2688 }
2689 elr->lr_next_sched = jiffies + elr->lr_timeout;
2690 elr->lr_next_group = group + 1;
2691 }
2692
2693 return ret;
2694}
2695
2696/*
2697 * Remove lr_request from the list_request and free the
2698 * request tructure. Should be called with li_list_mtx held
2699 */
2700static void ext4_remove_li_request(struct ext4_li_request *elr)
2701{
2702 struct ext4_sb_info *sbi;
2703
2704 if (!elr)
2705 return;
2706
2707 sbi = elr->lr_sbi;
2708
2709 list_del(&elr->lr_request);
2710 sbi->s_li_request = NULL;
2711 kfree(elr);
2712}
2713
2714static void ext4_unregister_li_request(struct super_block *sb)
2715{
2716 struct ext4_li_request *elr = EXT4_SB(sb)->s_li_request;
2717
2718 if (!ext4_li_info)
2719 return;
2720
2721 mutex_lock(&ext4_li_info->li_list_mtx);
2722 ext4_remove_li_request(elr);
2723 mutex_unlock(&ext4_li_info->li_list_mtx);
2724}
2725
8f1f7453
ES		 2726static struct task_struct *ext4_lazyinit_task;
2727
bfff6873
LC		 2728/*
2729 * This is the function where ext4lazyinit thread lives. It walks
2730 * through the request list searching for next scheduled filesystem.
2731 * When such a fs is found, run the lazy initialization request
2732 * (ext4_rn_li_request) and keep track of the time spend in this
2733 * function. Based on that time we compute next schedule time of
2734 * the request. When walking through the list is complete, compute
2735 * next waking time and put itself into sleep.
2736 */
2737static int ext4_lazyinit_thread(void *arg)
2738{
2739 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2740 struct list_head *pos, *n;
2741 struct ext4_li_request *elr;
2742 unsigned long next_wakeup;
2743 DEFINE_WAIT(wait);
bfff6873
LC		 2744
2745 BUG_ON(NULL == eli);
2746
2747 eli->li_timer.data = (unsigned long)current;
2748 eli->li_timer.function = ext4_lazyinode_timeout;
2749
2750 eli->li_task = current;
2751 wake_up(&eli->li_wait_task);
2752
2753cont_thread:
2754 while (true) {
2755 next_wakeup = MAX_JIFFY_OFFSET;
2756
2757 mutex_lock(&eli->li_list_mtx);
2758 if (list_empty(&eli->li_request_list)) {
2759 mutex_unlock(&eli->li_list_mtx);
2760 goto exit_thread;
2761 }
2762
2763 list_for_each_safe(pos, n, &eli->li_request_list) {
2764 elr = list_entry(pos, struct ext4_li_request,
2765 lr_request);
2766
b2c78cd0
TT		 2767 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2768 if (ext4_run_li_request(elr) != 0) {
2769 /* error, remove the lazy_init job */
2770 ext4_remove_li_request(elr);
2771 continue;
2772 }
bfff6873
LC		 2773 }
2774
2775 if (time_before(elr->lr_next_sched, next_wakeup))
2776 next_wakeup = elr->lr_next_sched;
2777 }
2778 mutex_unlock(&eli->li_list_mtx);
2779
2780 if (freezing(current))
2781 refrigerator();
2782
f4245bd4
LC		 2783 if ((time_after_eq(jiffies, next_wakeup)) ||
2784 (MAX_JIFFY_OFFSET == next_wakeup)) {
bfff6873
LC		 2785 cond_resched();
2786 continue;
2787 }
2788
2789 eli->li_timer.expires = next_wakeup;
2790 add_timer(&eli->li_timer);
2791 prepare_to_wait(&eli->li_wait_daemon, &wait,
2792 TASK_INTERRUPTIBLE);
2793 if (time_before(jiffies, next_wakeup))
2794 schedule();
2795 finish_wait(&eli->li_wait_daemon, &wait);
8f1f7453
ES		 2796 if (kthread_should_stop()) {
2797 ext4_clear_request_list();
2798 goto exit_thread;
2799 }
bfff6873
LC		 2800 }
2801
2802exit_thread:
2803 /*
2804 * It looks like the request list is empty, but we need
2805 * to check it under the li_list_mtx lock, to prevent any
2806 * additions into it, and of course we should lock ext4_li_mtx
2807 * to atomically free the list and ext4_li_info, because at
2808 * this point another ext4 filesystem could be registering
2809 * new one.
2810 */
2811 mutex_lock(&ext4_li_mtx);
2812 mutex_lock(&eli->li_list_mtx);
2813 if (!list_empty(&eli->li_request_list)) {
2814 mutex_unlock(&eli->li_list_mtx);
2815 mutex_unlock(&ext4_li_mtx);
2816 goto cont_thread;
2817 }
2818 mutex_unlock(&eli->li_list_mtx);
2819 del_timer_sync(&ext4_li_info->li_timer);
2820 eli->li_task = NULL;
2821 wake_up(&eli->li_wait_task);
2822
2823 kfree(ext4_li_info);
8f1f7453 2824 ext4_lazyinit_task = NULL;
bfff6873
LC		 2825 ext4_li_info = NULL;
2826 mutex_unlock(&ext4_li_mtx);
2827
2828 return 0;
2829}
2830
2831static void ext4_clear_request_list(void)
2832{
2833 struct list_head *pos, *n;
2834 struct ext4_li_request *elr;
2835
2836 mutex_lock(&ext4_li_info->li_list_mtx);
bfff6873
LC		 2837 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2838 elr = list_entry(pos, struct ext4_li_request,
2839 lr_request);
2840 ext4_remove_li_request(elr);
2841 }
2842 mutex_unlock(&ext4_li_info->li_list_mtx);
2843}
2844
2845static int ext4_run_lazyinit_thread(void)
2846{
8f1f7453
ES		 2847 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2848 ext4_li_info, "ext4lazyinit");
2849 if (IS_ERR(ext4_lazyinit_task)) {
2850 int err = PTR_ERR(ext4_lazyinit_task);
bfff6873
LC		 2851 ext4_clear_request_list();
2852 del_timer_sync(&ext4_li_info->li_timer);
2853 kfree(ext4_li_info);
2854 ext4_li_info = NULL;
2855 printk(KERN_CRIT "EXT4: error %d creating inode table "
2856 "initialization thread\n",
2857 err);
2858 return err;
2859 }
2860 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
2861
2862 wait_event(ext4_li_info->li_wait_task, ext4_li_info->li_task != NULL);
2863 return 0;
2864}
2865
2866/*
2867 * Check whether it make sense to run itable init. thread or not.
2868 * If there is at least one uninitialized inode table, return
2869 * corresponding group number, else the loop goes through all
2870 * groups and return total number of groups.
2871 */
2872static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2873{
2874 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2875 struct ext4_group_desc *gdp = NULL;
2876
2877 for (group = 0; group < ngroups; group++) {
2878 gdp = ext4_get_group_desc(sb, group, NULL);
2879 if (!gdp)
2880 continue;
2881
2882 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2883 break;
2884 }
2885
2886 return group;
2887}
2888
2889static int ext4_li_info_new(void)
2890{
2891 struct ext4_lazy_init *eli = NULL;
2892
2893 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2894 if (!eli)
2895 return -ENOMEM;
2896
2897 eli->li_task = NULL;
2898 INIT_LIST_HEAD(&eli->li_request_list);
2899 mutex_init(&eli->li_list_mtx);
2900
2901 init_waitqueue_head(&eli->li_wait_daemon);
2902 init_waitqueue_head(&eli->li_wait_task);
2903 init_timer(&eli->li_timer);
2904 eli->li_state |= EXT4_LAZYINIT_QUIT;
2905
2906 ext4_li_info = eli;
2907
2908 return 0;
2909}
2910
2911static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
2912 ext4_group_t start)
2913{
2914 struct ext4_sb_info *sbi = EXT4_SB(sb);
2915 struct ext4_li_request *elr;
2916 unsigned long rnd;
2917
2918 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
2919 if (!elr)
2920 return NULL;
2921
2922 elr->lr_super = sb;
2923 elr->lr_sbi = sbi;
2924 elr->lr_next_group = start;
2925
2926 /*
2927 * Randomize first schedule time of the request to
2928 * spread the inode table initialization requests
2929 * better.
2930 */
2931 get_random_bytes(&rnd, sizeof(rnd));
2932 elr->lr_next_sched = jiffies + (unsigned long)rnd %
2933 (EXT4_DEF_LI_MAX_START_DELAY * HZ);
2934
2935 return elr;
2936}
2937
2938static int ext4_register_li_request(struct super_block *sb,
2939 ext4_group_t first_not_zeroed)
2940{
2941 struct ext4_sb_info *sbi = EXT4_SB(sb);
2942 struct ext4_li_request *elr;
2943 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
6c5a6cb9 2944 int ret = 0;
bfff6873
LC		 2945
2946 if (sbi->s_li_request != NULL)
beed5ecb 2947 return 0;
bfff6873
LC		 2948
2949 if (first_not_zeroed == ngroups ||
2950 (sb->s_flags & MS_RDONLY) ||
2951 !test_opt(sb, INIT_INODE_TABLE)) {
2952 sbi->s_li_request = NULL;
beed5ecb 2953 return 0;
bfff6873
LC		 2954 }
2955
2956 if (first_not_zeroed == ngroups) {
2957 sbi->s_li_request = NULL;
beed5ecb 2958 return 0;
bfff6873
LC		 2959 }
2960
2961 elr = ext4_li_request_new(sb, first_not_zeroed);
beed5ecb
NK		 2962 if (!elr)
2963 return -ENOMEM;
bfff6873
LC		 2964
2965 mutex_lock(&ext4_li_mtx);
2966
2967 if (NULL == ext4_li_info) {
2968 ret = ext4_li_info_new();
2969 if (ret)
2970 goto out;
2971 }
2972
2973 mutex_lock(&ext4_li_info->li_list_mtx);
2974 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
2975 mutex_unlock(&ext4_li_info->li_list_mtx);
2976
2977 sbi->s_li_request = elr;
2978
2979 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
2980 ret = ext4_run_lazyinit_thread();
2981 if (ret)
2982 goto out;
2983 }
bfff6873 2984out:
beed5ecb
NK		 2985 mutex_unlock(&ext4_li_mtx);
2986 if (ret)
bfff6873 2987 kfree(elr);
bfff6873
LC		 2988 return ret;
2989}
2990
2991/*
2992 * We do not need to lock anything since this is called on
2993 * module unload.
2994 */
2995static void ext4_destroy_lazyinit_thread(void)
2996{
2997 /*
2998 * If thread exited earlier
2999 * there's nothing to be done.
3000 */
8f1f7453 3001 if (!ext4_li_info || !ext4_lazyinit_task)
bfff6873
LC		 3002 return;
3003
8f1f7453 3004 kthread_stop(ext4_lazyinit_task);
bfff6873
LC		 3005}
3006
2b2d6d01 3007static int ext4_fill_super(struct super_block *sb, void *data, int silent)
7477827f
AK		 3008 __releases(kernel_lock)
3009 __acquires(kernel_lock)
ac27a0ec 3010{
d4c402d9 3011 char *orig_data = kstrdup(data, GFP_KERNEL);
2b2d6d01 3012 struct buffer_head *bh;
617ba13b
MC		 3013 struct ext4_super_block *es = NULL;
3014 struct ext4_sb_info *sbi;
3015 ext4_fsblk_t block;
3016 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 3017 ext4_fsblk_t logical_sb_block;
ac27a0ec 3018 unsigned long offset = 0;
ac27a0ec
DK		 3019 unsigned long journal_devnum = 0;
3020 unsigned long def_mount_opts;
3021 struct inode *root;
9f6200bb 3022 char *cp;
0390131b 3023 const char *descr;
dcc7dae3 3024 int ret = -ENOMEM;
ac27a0ec 3025 int blocksize;
4ec11028
TT		 3026 unsigned int db_count;
3027 unsigned int i;
f287a1a5 3028 int needs_recovery, has_huge_files;
bd81d8ee 3029 __u64 blocks_count;
833f4077 3030 int err;
b3881f74 3031 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
bfff6873 3032 ext4_group_t first_not_zeroed;
ac27a0ec
DK		 3033
3034 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3035 if (!sbi)
dcc7dae3 3036 goto out_free_orig;
705895b6
PE		 3037
3038 sbi->s_blockgroup_lock =
3039 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3040 if (!sbi->s_blockgroup_lock) {
3041 kfree(sbi);
dcc7dae3 3042 goto out_free_orig;
705895b6 3043 }
ac27a0ec
DK		 3044 sb->s_fs_info = sbi;
3045 sbi->s_mount_opt = 0;
617ba13b
MC		 3046 sbi->s_resuid = EXT4_DEF_RESUID;
3047 sbi->s_resgid = EXT4_DEF_RESGID;
240799cd 3048 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
d9c9bef1 3049 sbi->s_sb_block = sb_block;
f613dfcb
TT		 3050 if (sb->s_bdev->bd_part)
3051 sbi->s_sectors_written_start =
3052 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
ac27a0ec 3053
9f6200bb
TT		 3054 /* Cleanup superblock name */
3055 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3056 *cp = '!';
3057
dcc7dae3 3058 ret = -EINVAL;
617ba13b 3059 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 3060 if (!blocksize) {
b31e1552 3061 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
ac27a0ec
DK		 3062 goto out_fail;
3063 }
3064
3065 /*
617ba13b 3066 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 3067 * block sizes. We need to calculate the offset from buffer start.
3068 */
617ba13b 3069 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 3070 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3071 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 3072 } else {
70bbb3e0 3073 logical_sb_block = sb_block;
ac27a0ec
DK		 3074 }
3075
70bbb3e0 3076 if (!(bh = sb_bread(sb, logical_sb_block))) {
b31e1552 3077 ext4_msg(sb, KERN_ERR, "unable to read superblock");
ac27a0ec
DK		 3078 goto out_fail;
3079 }
3080 /*
3081 * Note: s_es must be initialized as soon as possible because
617ba13b 3082 * some ext4 macro-instructions depend on its value
ac27a0ec 3083 */
617ba13b 3084 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
ac27a0ec
DK		 3085 sbi->s_es = es;
3086 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 3087 if (sb->s_magic != EXT4_SUPER_MAGIC)
3088 goto cantfind_ext4;
afc32f7e 3089 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
ac27a0ec
DK		 3090
3091 /* Set defaults before we parse the mount options */
3092 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
fd8c37ec 3093 set_opt(sb, INIT_INODE_TABLE);
617ba13b 3094 if (def_mount_opts & EXT4_DEFM_DEBUG)
fd8c37ec 3095 set_opt(sb, DEBUG);
437ca0fd
DM		 3096 if (def_mount_opts & EXT4_DEFM_BSDGROUPS) {
3097 ext4_msg(sb, KERN_WARNING, deprecated_msg, "bsdgroups",
3098 "2.6.38");
fd8c37ec 3099 set_opt(sb, GRPID);
437ca0fd 3100 }
617ba13b 3101 if (def_mount_opts & EXT4_DEFM_UID16)
fd8c37ec 3102 set_opt(sb, NO_UID32);
ea663336 3103 /* xattr user namespace & acls are now defaulted on */
03010a33 3104#ifdef CONFIG_EXT4_FS_XATTR
ea663336 3105 set_opt(sb, XATTR_USER);
2e7842b8 3106#endif
03010a33 3107#ifdef CONFIG_EXT4_FS_POSIX_ACL
ea663336 3108 set_opt(sb, POSIX_ACL);
2e7842b8 3109#endif
6fd7a467 3110 set_opt(sb, MBLK_IO_SUBMIT);
617ba13b 3111 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
fd8c37ec 3112 set_opt(sb, JOURNAL_DATA);
617ba13b 3113 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
fd8c37ec 3114 set_opt(sb, ORDERED_DATA);
617ba13b 3115 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
fd8c37ec 3116 set_opt(sb, WRITEBACK_DATA);
617ba13b
MC		 3117
3118 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
fd8c37ec 3119 set_opt(sb, ERRORS_PANIC);
bb4f397a 3120 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
fd8c37ec 3121 set_opt(sb, ERRORS_CONT);
bb4f397a 3122 else
fd8c37ec 3123 set_opt(sb, ERRORS_RO);
8b67f04a 3124 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
fd8c37ec 3125 set_opt(sb, BLOCK_VALIDITY);
8b67f04a 3126 if (def_mount_opts & EXT4_DEFM_DISCARD)
fd8c37ec 3127 set_opt(sb, DISCARD);
ac27a0ec
DK		 3128
3129 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
3130 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
30773840
TT		 3131 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3132 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3133 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
ac27a0ec 3134
8b67f04a 3135 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
fd8c37ec 3136 set_opt(sb, BARRIER);
ac27a0ec 3137
dd919b98
AK		 3138 /*
3139 * enable delayed allocation by default
3140 * Use -o nodelalloc to turn it off
3141 */
8b67f04a
TT		 3142 if (!IS_EXT3_SB(sb) &&
3143 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
fd8c37ec 3144 set_opt(sb, DELALLOC);
dd919b98 3145
8b67f04a
TT		 3146 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3147 &journal_devnum, &journal_ioprio, NULL, 0)) {
3148 ext4_msg(sb, KERN_WARNING,
3149 "failed to parse options in superblock: %s",
3150 sbi->s_es->s_mount_opts);
3151 }
b3881f74
TT		 3152 if (!parse_options((char *) data, sb, &journal_devnum,
3153 &journal_ioprio, NULL, 0))
ac27a0ec
DK		 3154 goto failed_mount;
3155
3156 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 3157 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec 3158
617ba13b
MC		 3159 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3160 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3161 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3162 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
b31e1552
ES		 3163 ext4_msg(sb, KERN_WARNING,
3164 "feature flags set on rev 0 fs, "
3165 "running e2fsck is recommended");
469108ff 3166
ac27a0ec
DK		 3167 /*
3168 * Check feature flags regardless of the revision level, since we
3169 * previously didn't change the revision level when setting the flags,
3170 * so there is a chance incompat flags are set on a rev 0 filesystem.
3171 */
a13fb1a4 3172 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
ac27a0ec 3173 goto failed_mount;
a13fb1a4 3174
ac27a0ec
DK		 3175 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3176
617ba13b
MC		 3177 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3178 blocksize > EXT4_MAX_BLOCK_SIZE) {
b31e1552
ES		 3179 ext4_msg(sb, KERN_ERR,
3180 "Unsupported filesystem blocksize %d", blocksize);
ac27a0ec
DK		 3181 goto failed_mount;
3182 }
3183
ac27a0ec 3184 if (sb->s_blocksize != blocksize) {
ce40733c
AK		 3185 /* Validate the filesystem blocksize */
3186 if (!sb_set_blocksize(sb, blocksize)) {
b31e1552 3187 ext4_msg(sb, KERN_ERR, "bad block size %d",
ce40733c 3188 blocksize);
ac27a0ec
DK		 3189 goto failed_mount;
3190 }
3191
2b2d6d01 3192 brelse(bh);
70bbb3e0
AM		 3193 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3194 offset = do_div(logical_sb_block, blocksize);
3195 bh = sb_bread(sb, logical_sb_block);
ac27a0ec 3196 if (!bh) {
b31e1552
ES		 3197 ext4_msg(sb, KERN_ERR,
3198 "Can't read superblock on 2nd try");
ac27a0ec
DK		 3199 goto failed_mount;
3200 }
617ba13b 3201 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
ac27a0ec 3202 sbi->s_es = es;
617ba13b 3203 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
b31e1552
ES		 3204 ext4_msg(sb, KERN_ERR,
3205 "Magic mismatch, very weird!");
ac27a0ec
DK		 3206 goto failed_mount;
3207 }
3208 }
3209
a13fb1a4
ES		 3210 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3211 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
f287a1a5
TT		 3212 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3213 has_huge_files);
3214 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
ac27a0ec 3215
617ba13b
MC		 3216 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3217 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3218 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
ac27a0ec
DK		 3219 } else {
3220 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3221 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
617ba13b 3222 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1330593e 3223 (!is_power_of_2(sbi->s_inode_size)) ||
ac27a0ec 3224 (sbi->s_inode_size > blocksize)) {
b31e1552
ES		 3225 ext4_msg(sb, KERN_ERR,
3226 "unsupported inode size: %d",
2b2d6d01 3227 sbi->s_inode_size);
ac27a0ec
DK		 3228 goto failed_mount;
3229 }
ef7f3835
KS		 3230 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3231 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
ac27a0ec 3232 }
0b8e58a1 3233
0d1ee42f
AR		 3234 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3235 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
8fadc143 3236 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 3237 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 3238 !is_power_of_2(sbi->s_desc_size)) {
b31e1552
ES		 3239 ext4_msg(sb, KERN_ERR,
3240 "unsupported descriptor size %lu",
0d1ee42f
AR		 3241 sbi->s_desc_size);
3242 goto failed_mount;
3243 }
3244 } else
3245 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
0b8e58a1 3246
ac27a0ec 3247 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 3248 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
b47b6f38 3249 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
617ba13b 3250 goto cantfind_ext4;
0b8e58a1 3251
617ba13b 3252 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 3253 if (sbi->s_inodes_per_block == 0)
617ba13b 3254 goto cantfind_ext4;
ac27a0ec
DK		 3255 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3256 sbi->s_inodes_per_block;
0d1ee42f 3257 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 3258 sbi->s_sbh = bh;
3259 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 3260 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3261 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
0b8e58a1 3262
2b2d6d01 3263 for (i = 0; i < 4; i++)
ac27a0ec
DK		 3264 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3265 sbi->s_def_hash_version = es->s_def_hash_version;
f99b2589
TT		 3266 i = le32_to_cpu(es->s_flags);
3267 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3268 sbi->s_hash_unsigned = 3;
3269 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3270#ifdef __CHAR_UNSIGNED__
3271 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3272 sbi->s_hash_unsigned = 3;
3273#else
3274 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3275#endif
3276 sb->s_dirt = 1;
3277 }
ac27a0ec
DK		 3278
3279 if (sbi->s_blocks_per_group > blocksize * 8) {
b31e1552
ES		 3280 ext4_msg(sb, KERN_ERR,
3281 "#blocks per group too big: %lu",
2b2d6d01 3282 sbi->s_blocks_per_group);
ac27a0ec
DK		 3283 goto failed_mount;
3284 }
ac27a0ec 3285 if (sbi->s_inodes_per_group > blocksize * 8) {
b31e1552
ES		 3286 ext4_msg(sb, KERN_ERR,
3287 "#inodes per group too big: %lu",
2b2d6d01 3288 sbi->s_inodes_per_group);
ac27a0ec
DK		 3289 goto failed_mount;
3290 }
3291
bf43d84b
ES		 3292 /*
3293 * Test whether we have more sectors than will fit in sector_t,
3294 * and whether the max offset is addressable by the page cache.
3295 */
5a9ae68a 3296 err = generic_check_addressable(sb->s_blocksize_bits,
30ca22c7 3297 ext4_blocks_count(es));
5a9ae68a 3298 if (err) {
b31e1552 3299 ext4_msg(sb, KERN_ERR, "filesystem"
bf43d84b 3300 " too large to mount safely on this system");
ac27a0ec 3301 if (sizeof(sector_t) < 8)
90c699a9 3302 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
5a9ae68a 3303 ret = err;
ac27a0ec
DK		 3304 goto failed_mount;
3305 }
3306
617ba13b
MC		 3307 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3308 goto cantfind_ext4;
e7c95593 3309
0f2ddca6
FTN		 3310 /* check blocks count against device size */
3311 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3312 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
b31e1552
ES		 3313 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3314 "exceeds size of device (%llu blocks)",
0f2ddca6
FTN		 3315 ext4_blocks_count(es), blocks_count);
3316 goto failed_mount;
3317 }
3318
0b8e58a1
AD		 3319 /*
3320 * It makes no sense for the first data block to be beyond the end
3321 * of the filesystem.
3322 */
3323 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
b31e1552
ES		 3324 ext4_msg(sb, KERN_WARNING, "bad geometry: first data"
3325 "block %u is beyond end of filesystem (%llu)",
3326 le32_to_cpu(es->s_first_data_block),
3327 ext4_blocks_count(es));
e7c95593
ES		 3328 goto failed_mount;
3329 }
bd81d8ee
LV		 3330 blocks_count = (ext4_blocks_count(es) -
3331 le32_to_cpu(es->s_first_data_block) +
3332 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3333 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4ec11028 3334 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
b31e1552 3335 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
4ec11028 3336 "(block count %llu, first data block %u, "
b31e1552 3337 "blocks per group %lu)", sbi->s_groups_count,
4ec11028
TT		 3338 ext4_blocks_count(es),
3339 le32_to_cpu(es->s_first_data_block),
3340 EXT4_BLOCKS_PER_GROUP(sb));
3341 goto failed_mount;
3342 }
bd81d8ee 3343 sbi->s_groups_count = blocks_count;
fb0a387d
ES		 3344 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3345 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
617ba13b
MC		 3346 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3347 EXT4_DESC_PER_BLOCK(sb);
2b2d6d01 3348 sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
ac27a0ec
DK		 3349 GFP_KERNEL);
3350 if (sbi->s_group_desc == NULL) {
b31e1552 3351 ext4_msg(sb, KERN_ERR, "not enough memory");
ac27a0ec
DK		 3352 goto failed_mount;
3353 }
3354
3244fcb1 3355#ifdef CONFIG_PROC_FS
9f6200bb
TT		 3356 if (ext4_proc_root)
3357 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
3244fcb1 3358#endif
240799cd 3359
705895b6 3360 bgl_lock_init(sbi->s_blockgroup_lock);
ac27a0ec
DK		 3361
3362 for (i = 0; i < db_count; i++) {
70bbb3e0 3363 block = descriptor_loc(sb, logical_sb_block, i);
ac27a0ec
DK		 3364 sbi->s_group_desc[i] = sb_bread(sb, block);
3365 if (!sbi->s_group_desc[i]) {
b31e1552
ES		 3366 ext4_msg(sb, KERN_ERR,
3367 "can't read group descriptor %d", i);
ac27a0ec
DK		 3368 db_count = i;
3369 goto failed_mount2;
3370 }
3371 }
bfff6873 3372 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
b31e1552 3373 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
ac27a0ec
DK		 3374 goto failed_mount2;
3375 }
772cb7c8
JS		 3376 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3377 if (!ext4_fill_flex_info(sb)) {
b31e1552
ES		 3378 ext4_msg(sb, KERN_ERR,
3379 "unable to initialize "
3380 "flex_bg meta info!");
772cb7c8
JS		 3381 goto failed_mount2;
3382 }
3383
ac27a0ec
DK		 3384 sbi->s_gdb_count = db_count;
3385 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3386 spin_lock_init(&sbi->s_next_gen_lock);
3387
ce7e010a
TT		 3388 err = percpu_counter_init(&sbi->s_freeblocks_counter,
3389 ext4_count_free_blocks(sb));
3390 if (!err) {
3391 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3392 ext4_count_free_inodes(sb));
3393 }
3394 if (!err) {
3395 err = percpu_counter_init(&sbi->s_dirs_counter,
3396 ext4_count_dirs(sb));
3397 }
3398 if (!err) {
3399 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
3400 }
3401 if (err) {
3402 ext4_msg(sb, KERN_ERR, "insufficient memory");
3403 goto failed_mount3;
3404 }
3405
c9de560d 3406 sbi->s_stripe = ext4_get_stripe_size(sbi);
55138e0b 3407 sbi->s_max_writeback_mb_bump = 128;
c9de560d 3408
ac27a0ec
DK		 3409 /*
3410 * set up enough so that it can read an inode
3411 */
9ca92389
TT		 3412 if (!test_opt(sb, NOLOAD) &&
3413 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3414 sb->s_op = &ext4_sops;
3415 else
3416 sb->s_op = &ext4_nojournal_sops;
617ba13b
MC		 3417 sb->s_export_op = &ext4_export_ops;
3418 sb->s_xattr = ext4_xattr_handlers;
ac27a0ec 3419#ifdef CONFIG_QUOTA
617ba13b
MC		 3420 sb->s_qcop = &ext4_qctl_operations;
3421 sb->dq_op = &ext4_quota_operations;
ac27a0ec 3422#endif
f2fa2ffc
AK		 3423 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3424
ac27a0ec 3425 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3b9d4ed2 3426 mutex_init(&sbi->s_orphan_lock);
32ed5058 3427 mutex_init(&sbi->s_resize_lock);
ac27a0ec
DK		 3428
3429 sb->s_root = NULL;
3430
3431 needs_recovery = (es->s_last_orphan != 0 ||
617ba13b
MC		 3432 EXT4_HAS_INCOMPAT_FEATURE(sb,
3433 EXT4_FEATURE_INCOMPAT_RECOVER));
ac27a0ec
DK		 3434
3435 /*
3436 * The first inode we look at is the journal inode. Don't try
3437 * root first: it may be modified in the journal!
3438 */
3439 if (!test_opt(sb, NOLOAD) &&
617ba13b
MC		 3440 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3441 if (ext4_load_journal(sb, es, journal_devnum))
ac27a0ec 3442 goto failed_mount3;
0390131b
FM		 3443 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3444 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
b31e1552
ES		 3445 ext4_msg(sb, KERN_ERR, "required journal recovery "
3446 "suppressed and not mounted read-only");
744692dc 3447 goto failed_mount_wq;
ac27a0ec 3448 } else {
fd8c37ec
TT		 3449 clear_opt(sb, DATA_FLAGS);
3450 set_opt(sb, WRITEBACK_DATA);
0390131b
FM		 3451 sbi->s_journal = NULL;
3452 needs_recovery = 0;
3453 goto no_journal;
ac27a0ec
DK		 3454 }
3455
eb40a09c
JS		 3456 if (ext4_blocks_count(es) > 0xffffffffULL &&
3457 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3458 JBD2_FEATURE_INCOMPAT_64BIT)) {
b31e1552 3459 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
744692dc 3460 goto failed_mount_wq;
eb40a09c
JS		 3461 }
3462
d4da6c9c
LT		 3463 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3464 jbd2_journal_set_features(sbi->s_journal,
3465 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
818d276c 3466 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
d4da6c9c
LT		 3467 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3468 jbd2_journal_set_features(sbi->s_journal,
3469 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
818d276c
GS		 3470 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3471 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
d4da6c9c
LT		 3472 } else {
3473 jbd2_journal_clear_features(sbi->s_journal,
3474 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3475 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3476 }
818d276c 3477
ac27a0ec
DK		 3478 /* We have now updated the journal if required, so we can
3479 * validate the data journaling mode. */
3480 switch (test_opt(sb, DATA_FLAGS)) {
3481 case 0:
3482 /* No mode set, assume a default based on the journal
63f57933
AM		 3483 * capabilities: ORDERED_DATA if the journal can
3484 * cope, else JOURNAL_DATA
3485 */
dab291af
MC		 3486 if (jbd2_journal_check_available_features
3487 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
fd8c37ec 3488 set_opt(sb, ORDERED_DATA);
ac27a0ec 3489 else
fd8c37ec 3490 set_opt(sb, JOURNAL_DATA);
ac27a0ec
DK		 3491 break;
3492
617ba13b
MC		 3493 case EXT4_MOUNT_ORDERED_DATA:
3494 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 3495 if (!jbd2_journal_check_available_features
3496 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
b31e1552
ES		 3497 ext4_msg(sb, KERN_ERR, "Journal does not support "
3498 "requested data journaling mode");
744692dc 3499 goto failed_mount_wq;
ac27a0ec
DK		 3500 }
3501 default:
3502 break;
3503 }
b3881f74 3504 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
ac27a0ec 3505
ce7e010a
TT		 3506 /*
3507 * The journal may have updated the bg summary counts, so we
3508 * need to update the global counters.
3509 */
3510 percpu_counter_set(&sbi->s_freeblocks_counter,
3511 ext4_count_free_blocks(sb));
3512 percpu_counter_set(&sbi->s_freeinodes_counter,
3513 ext4_count_free_inodes(sb));
3514 percpu_counter_set(&sbi->s_dirs_counter,
3515 ext4_count_dirs(sb));
3516 percpu_counter_set(&sbi->s_dirtyblocks_counter, 0);
206f7ab4 3517
ce7e010a 3518no_journal:
fd89d5f2
TH		 3519 /*
3520 * The maximum number of concurrent works can be high and
3521 * concurrency isn't really necessary. Limit it to 1.
3522 */
3523 EXT4_SB(sb)->dio_unwritten_wq =
ae005cbe 3524 alloc_workqueue("ext4-dio-unwritten", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4c0425ff
MC		 3525 if (!EXT4_SB(sb)->dio_unwritten_wq) {
3526 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
3527 goto failed_mount_wq;
3528 }
3529
ac27a0ec 3530 /*
dab291af 3531 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 3532 * so we can safely mount the rest of the filesystem now.
3533 */
3534
1d1fe1ee
DH		 3535 root = ext4_iget(sb, EXT4_ROOT_INO);
3536 if (IS_ERR(root)) {
b31e1552 3537 ext4_msg(sb, KERN_ERR, "get root inode failed");
1d1fe1ee 3538 ret = PTR_ERR(root);
32a9bb57 3539 root = NULL;
ac27a0ec
DK		 3540 goto failed_mount4;
3541 }
3542 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
b31e1552 3543 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
ac27a0ec
DK		 3544 goto failed_mount4;
3545 }
1d1fe1ee
DH		 3546 sb->s_root = d_alloc_root(root);
3547 if (!sb->s_root) {
b31e1552 3548 ext4_msg(sb, KERN_ERR, "get root dentry failed");
1d1fe1ee
DH		 3549 ret = -ENOMEM;
3550 goto failed_mount4;
3551 }
ac27a0ec 3552
2b2d6d01 3553 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
ef7f3835
KS		 3554
3555 /* determine the minimum size of new large inodes, if present */
3556 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3557 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3558 EXT4_GOOD_OLD_INODE_SIZE;
3559 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3560 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
3561 if (sbi->s_want_extra_isize <
3562 le16_to_cpu(es->s_want_extra_isize))
3563 sbi->s_want_extra_isize =
3564 le16_to_cpu(es->s_want_extra_isize);
3565 if (sbi->s_want_extra_isize <
3566 le16_to_cpu(es->s_min_extra_isize))
3567 sbi->s_want_extra_isize =
3568 le16_to_cpu(es->s_min_extra_isize);
3569 }
3570 }
3571 /* Check if enough inode space is available */
3572 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3573 sbi->s_inode_size) {
3574 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3575 EXT4_GOOD_OLD_INODE_SIZE;
b31e1552
ES		 3576 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3577 "available");
ef7f3835
KS		 3578 }
3579
90576c0b
TT		 3580 if (test_opt(sb, DELALLOC) &&
3581 (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)) {
b31e1552
ES		 3582 ext4_msg(sb, KERN_WARNING, "Ignoring delalloc option - "
3583 "requested data journaling mode");
fd8c37ec 3584 clear_opt(sb, DELALLOC);
90576c0b 3585 }
744692dc
JZ		 3586 if (test_opt(sb, DIOREAD_NOLOCK)) {
3587 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3588 ext4_msg(sb, KERN_WARNING, "Ignoring dioread_nolock "
3589 "option - requested data journaling mode");
fd8c37ec 3590 clear_opt(sb, DIOREAD_NOLOCK);
744692dc
JZ		 3591 }
3592 if (sb->s_blocksize < PAGE_SIZE) {
3593 ext4_msg(sb, KERN_WARNING, "Ignoring dioread_nolock "
3594 "option - block size is too small");
fd8c37ec 3595 clear_opt(sb, DIOREAD_NOLOCK);
744692dc
JZ		 3596 }
3597 }
c2774d84 3598
6fd058f7
TT		 3599 err = ext4_setup_system_zone(sb);
3600 if (err) {
b31e1552 3601 ext4_msg(sb, KERN_ERR, "failed to initialize system "
fbe845dd 3602 "zone (%d)", err);
6fd058f7
TT		 3603 goto failed_mount4;
3604 }
3605
c2774d84
AK		 3606 ext4_ext_init(sb);
3607 err = ext4_mb_init(sb, needs_recovery);
3608 if (err) {
421f91d2 3609 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
b31e1552 3610 err);
c2774d84
AK		 3611 goto failed_mount4;
3612 }
3613
bfff6873
LC		 3614 err = ext4_register_li_request(sb, first_not_zeroed);
3615 if (err)
3616 goto failed_mount4;
3617
3197ebdb
TT		 3618 sbi->s_kobj.kset = ext4_kset;
3619 init_completion(&sbi->s_kobj_unregister);
3620 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
3621 "%s", sb->s_id);
3622 if (err) {
3623 ext4_mb_release(sb);
3624 ext4_ext_release(sb);
3625 goto failed_mount4;
3626 };
3627
617ba13b
MC		 3628 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3629 ext4_orphan_cleanup(sb, es);
3630 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
0390131b 3631 if (needs_recovery) {
b31e1552 3632 ext4_msg(sb, KERN_INFO, "recovery complete");
0390131b
FM		 3633 ext4_mark_recovery_complete(sb, es);
3634 }
3635 if (EXT4_SB(sb)->s_journal) {
3636 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3637 descr = " journalled data mode";
3638 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3639 descr = " ordered data mode";
3640 else
3641 descr = " writeback data mode";
3642 } else
3643 descr = "out journal";
3644
d4c402d9 3645 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
8b67f04a
TT		 3646 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
3647 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
ac27a0ec 3648
66e61a9e
TT		 3649 init_timer(&sbi->s_err_report);
3650 sbi->s_err_report.function = print_daily_error_info;
3651 sbi->s_err_report.data = (unsigned long) sb;
3652 if (es->s_error_count)
3653 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
ac27a0ec 3654
d4c402d9 3655 kfree(orig_data);
ac27a0ec
DK		 3656 return 0;
3657
617ba13b 3658cantfind_ext4:
ac27a0ec 3659 if (!silent)
b31e1552 3660 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
ac27a0ec
DK		 3661 goto failed_mount;
3662
3663failed_mount4:
32a9bb57
MK		 3664 iput(root);
3665 sb->s_root = NULL;
b31e1552 3666 ext4_msg(sb, KERN_ERR, "mount failed");
4c0425ff
MC		 3667 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
3668failed_mount_wq:
6fd058f7 3669 ext4_release_system_zone(sb);
0390131b
FM		 3670 if (sbi->s_journal) {
3671 jbd2_journal_destroy(sbi->s_journal);
3672 sbi->s_journal = NULL;
3673 }
ac27a0ec 3674failed_mount3:
c5ca7c76
TT		 3675 if (sbi->s_flex_groups) {
3676 if (is_vmalloc_addr(sbi->s_flex_groups))
3677 vfree(sbi->s_flex_groups);
3678 else
3679 kfree(sbi->s_flex_groups);
3680 }
ce7e010a
TT		 3681 percpu_counter_destroy(&sbi->s_freeblocks_counter);
3682 percpu_counter_destroy(&sbi->s_freeinodes_counter);
3683 percpu_counter_destroy(&sbi->s_dirs_counter);
3684 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
ac27a0ec
DK		 3685failed_mount2:
3686 for (i = 0; i < db_count; i++)
3687 brelse(sbi->s_group_desc[i]);
3688 kfree(sbi->s_group_desc);
3689failed_mount:
240799cd 3690 if (sbi->s_proc) {
9f6200bb 3691 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 3692 }
ac27a0ec
DK		 3693#ifdef CONFIG_QUOTA
3694 for (i = 0; i < MAXQUOTAS; i++)
3695 kfree(sbi->s_qf_names[i]);
3696#endif
617ba13b 3697 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 3698 brelse(bh);
3699out_fail:
3700 sb->s_fs_info = NULL;
f6830165 3701 kfree(sbi->s_blockgroup_lock);
ac27a0ec 3702 kfree(sbi);
dcc7dae3 3703out_free_orig:
d4c402d9 3704 kfree(orig_data);
1d1fe1ee 3705 return ret;
ac27a0ec
DK		 3706}
3707
3708/*
3709 * Setup any per-fs journal parameters now. We'll do this both on
3710 * initial mount, once the journal has been initialised but before we've
3711 * done any recovery; and again on any subsequent remount.
3712 */
617ba13b 3713static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 3714{
617ba13b 3715 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 3716
30773840
TT		 3717 journal->j_commit_interval = sbi->s_commit_interval;
3718 journal->j_min_batch_time = sbi->s_min_batch_time;
3719 journal->j_max_batch_time = sbi->s_max_batch_time;
ac27a0ec 3720
a931da6a 3721 write_lock(&journal->j_state_lock);
ac27a0ec 3722 if (test_opt(sb, BARRIER))
dab291af 3723 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 3724 else
dab291af 3725 journal->j_flags &= ~JBD2_BARRIER;
5bf5683a
HK		 3726 if (test_opt(sb, DATA_ERR_ABORT))
3727 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
3728 else
3729 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
a931da6a 3730 write_unlock(&journal->j_state_lock);
ac27a0ec
DK		 3731}
3732
617ba13b 3733static journal_t *ext4_get_journal(struct super_block *sb,
ac27a0ec
DK		 3734 unsigned int journal_inum)
3735{
3736 struct inode *journal_inode;
3737 journal_t *journal;
3738
0390131b
FM		 3739 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3740
ac27a0ec
DK		 3741 /* First, test for the existence of a valid inode on disk. Bad
3742 * things happen if we iget() an unused inode, as the subsequent
3743 * iput() will try to delete it. */
3744
1d1fe1ee
DH		 3745 journal_inode = ext4_iget(sb, journal_inum);
3746 if (IS_ERR(journal_inode)) {
b31e1552 3747 ext4_msg(sb, KERN_ERR, "no journal found");
ac27a0ec
DK		 3748 return NULL;
3749 }
3750 if (!journal_inode->i_nlink) {
3751 make_bad_inode(journal_inode);
3752 iput(journal_inode);
b31e1552 3753 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
ac27a0ec
DK		 3754 return NULL;
3755 }
3756
e5f8eab8 3757 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
ac27a0ec 3758 journal_inode, journal_inode->i_size);
1d1fe1ee 3759 if (!S_ISREG(journal_inode->i_mode)) {
b31e1552 3760 ext4_msg(sb, KERN_ERR, "invalid journal inode");
ac27a0ec
DK		 3761 iput(journal_inode);
3762 return NULL;
3763 }
3764
dab291af 3765 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 3766 if (!journal) {
b31e1552 3767 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
ac27a0ec
DK		 3768 iput(journal_inode);
3769 return NULL;
3770 }
3771 journal->j_private = sb;
617ba13b 3772 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 3773 return journal;
3774}
3775
617ba13b 3776static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 3777 dev_t j_dev)
3778{
2b2d6d01 3779 struct buffer_head *bh;
ac27a0ec 3780 journal_t *journal;
617ba13b
MC		 3781 ext4_fsblk_t start;
3782 ext4_fsblk_t len;
ac27a0ec 3783 int hblock, blocksize;
617ba13b 3784 ext4_fsblk_t sb_block;
ac27a0ec 3785 unsigned long offset;
2b2d6d01 3786 struct ext4_super_block *es;
ac27a0ec
DK		 3787 struct block_device *bdev;
3788
0390131b
FM		 3789 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3790
b31e1552 3791 bdev = ext4_blkdev_get(j_dev, sb);
ac27a0ec
DK		 3792 if (bdev == NULL)
3793 return NULL;
3794
ac27a0ec 3795 blocksize = sb->s_blocksize;
e1defc4f 3796 hblock = bdev_logical_block_size(bdev);
ac27a0ec 3797 if (blocksize < hblock) {
b31e1552
ES		 3798 ext4_msg(sb, KERN_ERR,
3799 "blocksize too small for journal device");
ac27a0ec
DK		 3800 goto out_bdev;
3801 }
3802
617ba13b
MC		 3803 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3804 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 3805 set_blocksize(bdev, blocksize);
3806 if (!(bh = __bread(bdev, sb_block, blocksize))) {
b31e1552
ES		 3807 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
3808 "external journal");
ac27a0ec
DK		 3809 goto out_bdev;
3810 }
3811
617ba13b
MC		 3812 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3813 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 3814 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b 3815 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
b31e1552
ES		 3816 ext4_msg(sb, KERN_ERR, "external journal has "
3817 "bad superblock");
ac27a0ec
DK		 3818 brelse(bh);
3819 goto out_bdev;
3820 }
3821
617ba13b 3822 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
b31e1552 3823 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
ac27a0ec
DK		 3824 brelse(bh);
3825 goto out_bdev;
3826 }
3827
bd81d8ee 3828 len = ext4_blocks_count(es);
ac27a0ec
DK		 3829 start = sb_block + 1;
3830 brelse(bh); /* we're done with the superblock */
3831
dab291af 3832 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 3833 start, len, blocksize);
3834 if (!journal) {
b31e1552 3835 ext4_msg(sb, KERN_ERR, "failed to create device journal");
ac27a0ec
DK		 3836 goto out_bdev;
3837 }
3838 journal->j_private = sb;
3839 ll_rw_block(READ, 1, &journal->j_sb_buffer);
3840 wait_on_buffer(journal->j_sb_buffer);
3841 if (!buffer_uptodate(journal->j_sb_buffer)) {
b31e1552 3842 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
ac27a0ec
DK		 3843 goto out_journal;
3844 }
3845 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
b31e1552
ES		 3846 ext4_msg(sb, KERN_ERR, "External journal has more than one "
3847 "user (unsupported) - %d",
ac27a0ec
DK		 3848 be32_to_cpu(journal->j_superblock->s_nr_users));
3849 goto out_journal;
3850 }
617ba13b
MC		 3851 EXT4_SB(sb)->journal_bdev = bdev;
3852 ext4_init_journal_params(sb, journal);
ac27a0ec 3853 return journal;
0b8e58a1 3854
ac27a0ec 3855out_journal:
dab291af 3856 jbd2_journal_destroy(journal);
ac27a0ec 3857out_bdev:
617ba13b 3858 ext4_blkdev_put(bdev);
ac27a0ec
DK		 3859 return NULL;
3860}
3861
617ba13b
MC		 3862static int ext4_load_journal(struct super_block *sb,
3863 struct ext4_super_block *es,
ac27a0ec
DK		 3864 unsigned long journal_devnum)
3865{
3866 journal_t *journal;
3867 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
3868 dev_t journal_dev;
3869 int err = 0;
3870 int really_read_only;
3871
0390131b
FM		 3872 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3873
ac27a0ec
DK		 3874 if (journal_devnum &&
3875 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
b31e1552
ES		 3876 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
3877 "numbers have changed");
ac27a0ec
DK		 3878 journal_dev = new_decode_dev(journal_devnum);
3879 } else
3880 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
3881
3882 really_read_only = bdev_read_only(sb->s_bdev);
3883
3884 /*
3885 * Are we loading a blank journal or performing recovery after a
3886 * crash? For recovery, we need to check in advance whether we
3887 * can get read-write access to the device.
3888 */
617ba13b 3889 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
ac27a0ec 3890 if (sb->s_flags & MS_RDONLY) {
b31e1552
ES		 3891 ext4_msg(sb, KERN_INFO, "INFO: recovery "
3892 "required on readonly filesystem");
ac27a0ec 3893 if (really_read_only) {
b31e1552
ES		 3894 ext4_msg(sb, KERN_ERR, "write access "
3895 "unavailable, cannot proceed");
ac27a0ec
DK		 3896 return -EROFS;
3897 }
b31e1552
ES		 3898 ext4_msg(sb, KERN_INFO, "write access will "
3899 "be enabled during recovery");
ac27a0ec
DK		 3900 }
3901 }
3902
3903 if (journal_inum && journal_dev) {
b31e1552
ES		 3904 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
3905 "and inode journals!");
ac27a0ec
DK		 3906 return -EINVAL;
3907 }
3908
3909 if (journal_inum) {
617ba13b 3910 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 3911 return -EINVAL;
3912 } else {
617ba13b 3913 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 3914 return -EINVAL;
3915 }
3916
90576c0b 3917 if (!(journal->j_flags & JBD2_BARRIER))
b31e1552 3918 ext4_msg(sb, KERN_INFO, "barriers disabled");
4776004f 3919
ac27a0ec 3920 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
dab291af 3921 err = jbd2_journal_update_format(journal);
ac27a0ec 3922 if (err) {
b31e1552 3923 ext4_msg(sb, KERN_ERR, "error updating journal");
dab291af 3924 jbd2_journal_destroy(journal);
ac27a0ec
DK		 3925 return err;
3926 }
3927 }
3928
617ba13b 3929 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
dab291af 3930 err = jbd2_journal_wipe(journal, !really_read_only);
1c13d5c0
TT		 3931 if (!err) {
3932 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
3933 if (save)
3934 memcpy(save, ((char *) es) +
3935 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
dab291af 3936 err = jbd2_journal_load(journal);
1c13d5c0
TT		 3937 if (save)
3938 memcpy(((char *) es) + EXT4_S_ERR_START,
3939 save, EXT4_S_ERR_LEN);
3940 kfree(save);
3941 }
ac27a0ec
DK		 3942
3943 if (err) {
b31e1552 3944 ext4_msg(sb, KERN_ERR, "error loading journal");
dab291af 3945 jbd2_journal_destroy(journal);
ac27a0ec
DK		 3946 return err;
3947 }
3948
617ba13b
MC		 3949 EXT4_SB(sb)->s_journal = journal;
3950 ext4_clear_journal_err(sb, es);
ac27a0ec 3951
c41303ce 3952 if (!really_read_only && journal_devnum &&
ac27a0ec
DK		 3953 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3954 es->s_journal_dev = cpu_to_le32(journal_devnum);
ac27a0ec
DK		 3955
3956 /* Make sure we flush the recovery flag to disk. */
e2d67052 3957 ext4_commit_super(sb, 1);
ac27a0ec
DK		 3958 }
3959
3960 return 0;
3961}
3962
e2d67052 3963static int ext4_commit_super(struct super_block *sb, int sync)
ac27a0ec 3964{
e2d67052 3965 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
617ba13b 3966 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
c4be0c1d 3967 int error = 0;
ac27a0ec
DK		 3968
3969 if (!sbh)
c4be0c1d 3970 return error;
914258bf
TT		 3971 if (buffer_write_io_error(sbh)) {
3972 /*
3973 * Oh, dear. A previous attempt to write the
3974 * superblock failed. This could happen because the
3975 * USB device was yanked out. Or it could happen to
3976 * be a transient write error and maybe the block will
3977 * be remapped. Nothing we can do but to retry the
3978 * write and hope for the best.
3979 */
b31e1552
ES		 3980 ext4_msg(sb, KERN_ERR, "previous I/O error to "
3981 "superblock detected");
914258bf
TT		 3982 clear_buffer_write_io_error(sbh);
3983 set_buffer_uptodate(sbh);
3984 }
71290b36
TT		 3985 /*
3986 * If the file system is mounted read-only, don't update the
3987 * superblock write time. This avoids updating the superblock
3988 * write time when we are mounting the root file system
3989 * read/only but we need to replay the journal; at that point,
3990 * for people who are east of GMT and who make their clock
3991 * tick in localtime for Windows bug-for-bug compatibility,
3992 * the clock is set in the future, and this will cause e2fsck
3993 * to complain and force a full file system check.
3994 */
3995 if (!(sb->s_flags & MS_RDONLY))
3996 es->s_wtime = cpu_to_le32(get_seconds());
f613dfcb
TT		 3997 if (sb->s_bdev->bd_part)
3998 es->s_kbytes_written =
3999 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
afc32f7e
TT		 4000 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4001 EXT4_SB(sb)->s_sectors_written_start) >> 1));
f613dfcb
TT		 4002 else
4003 es->s_kbytes_written =
4004 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
ce7e010a
TT		 4005 ext4_free_blocks_count_set(es, percpu_counter_sum_positive(
4006 &EXT4_SB(sb)->s_freeblocks_counter));
4007 es->s_free_inodes_count =
4008 cpu_to_le32(percpu_counter_sum_positive(
4009 &EXT4_SB(sb)->s_freeinodes_counter));
7234ab2a 4010 sb->s_dirt = 0;
ac27a0ec
DK		 4011 BUFFER_TRACE(sbh, "marking dirty");
4012 mark_buffer_dirty(sbh);
914258bf 4013 if (sync) {
c4be0c1d
TS		 4014 error = sync_dirty_buffer(sbh);
4015 if (error)
4016 return error;
4017
4018 error = buffer_write_io_error(sbh);
4019 if (error) {
b31e1552
ES		 4020 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4021 "superblock");
914258bf
TT		 4022 clear_buffer_write_io_error(sbh);
4023 set_buffer_uptodate(sbh);
4024 }
4025 }
c4be0c1d 4026 return error;
ac27a0ec
DK		 4027}
4028
ac27a0ec
DK		 4029/*
4030 * Have we just finished recovery? If so, and if we are mounting (or
4031 * remounting) the filesystem readonly, then we will end up with a
4032 * consistent fs on disk. Record that fact.
4033 */
2b2d6d01
TT		 4034static void ext4_mark_recovery_complete(struct super_block *sb,
4035 struct ext4_super_block *es)
ac27a0ec 4036{
617ba13b 4037 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 4038
0390131b
FM		 4039 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4040 BUG_ON(journal != NULL);
4041 return;
4042 }
dab291af 4043 jbd2_journal_lock_updates(journal);
7ffe1ea8
HK		 4044 if (jbd2_journal_flush(journal) < 0)
4045 goto out;
4046
617ba13b 4047 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
ac27a0ec 4048 sb->s_flags & MS_RDONLY) {
617ba13b 4049 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
e2d67052 4050 ext4_commit_super(sb, 1);
ac27a0ec 4051 }
7ffe1ea8
HK		 4052
4053out:
dab291af 4054 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 4055}
4056
4057/*
4058 * If we are mounting (or read-write remounting) a filesystem whose journal
4059 * has recorded an error from a previous lifetime, move that error to the
4060 * main filesystem now.
4061 */
2b2d6d01
TT		 4062static void ext4_clear_journal_err(struct super_block *sb,
4063 struct ext4_super_block *es)
ac27a0ec
DK		 4064{
4065 journal_t *journal;
4066 int j_errno;
4067 const char *errstr;
4068
0390131b
FM		 4069 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4070
617ba13b 4071 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 4072
4073 /*
4074 * Now check for any error status which may have been recorded in the
617ba13b 4075 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 4076 */
4077
dab291af 4078 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 4079 if (j_errno) {
4080 char nbuf[16];
4081
617ba13b 4082 errstr = ext4_decode_error(sb, j_errno, nbuf);
12062ddd 4083 ext4_warning(sb, "Filesystem error recorded "
ac27a0ec 4084 "from previous mount: %s", errstr);
12062ddd 4085 ext4_warning(sb, "Marking fs in need of filesystem check.");
ac27a0ec 4086
617ba13b
MC		 4087 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4088 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
e2d67052 4089 ext4_commit_super(sb, 1);
ac27a0ec 4090
dab291af 4091 jbd2_journal_clear_err(journal);
ac27a0ec
DK		 4092 }
4093}
4094
4095/*
4096 * Force the running and committing transactions to commit,
4097 * and wait on the commit.
4098 */
617ba13b 4099int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 4100{
4101 journal_t *journal;
0390131b 4102 int ret = 0;
ac27a0ec
DK		 4103
4104 if (sb->s_flags & MS_RDONLY)
4105 return 0;
4106
617ba13b 4107 journal = EXT4_SB(sb)->s_journal;
6b0310fb 4108 if (journal) {
437f88cc 4109 vfs_check_frozen(sb, SB_FREEZE_TRANS);
0390131b 4110 ret = ext4_journal_force_commit(journal);
6b0310fb 4111 }
0390131b 4112
ac27a0ec
DK		 4113 return ret;
4114}
4115
2b2d6d01 4116static void ext4_write_super(struct super_block *sb)
ac27a0ec 4117{
ebc1ac16 4118 lock_super(sb);
9ca92389 4119 ext4_commit_super(sb, 1);
ebc1ac16 4120 unlock_super(sb);
ac27a0ec
DK		 4121}
4122
617ba13b 4123static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec 4124{
14ce0cb4 4125 int ret = 0;
9eddacf9 4126 tid_t target;
8d5d02e6 4127 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4128
9bffad1e 4129 trace_ext4_sync_fs(sb, wait);
8d5d02e6
MC		 4130 flush_workqueue(sbi->dio_unwritten_wq);
4131 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
9ca92389 4132 if (wait)
8d5d02e6 4133 jbd2_log_wait_commit(sbi->s_journal, target);
0390131b 4134 }
14ce0cb4 4135 return ret;
ac27a0ec
DK		 4136}
4137
4138/*
4139 * LVM calls this function before a (read-only) snapshot is created. This
4140 * gives us a chance to flush the journal completely and mark the fs clean.
4141 */
c4be0c1d 4142static int ext4_freeze(struct super_block *sb)
ac27a0ec 4143{
c4be0c1d
TS		 4144 int error = 0;
4145 journal_t *journal;
ac27a0ec 4146
9ca92389
TT		 4147 if (sb->s_flags & MS_RDONLY)
4148 return 0;
ac27a0ec 4149
9ca92389 4150 journal = EXT4_SB(sb)->s_journal;
7ffe1ea8 4151
9ca92389
TT		 4152 /* Now we set up the journal barrier. */
4153 jbd2_journal_lock_updates(journal);
ac27a0ec 4154
9ca92389
TT		 4155 /*
4156 * Don't clear the needs_recovery flag if we failed to flush
4157 * the journal.
4158 */
4159 error = jbd2_journal_flush(journal);
6b0310fb
ES		 4160 if (error < 0)
4161 goto out;
9ca92389
TT		 4162
4163 /* Journal blocked and flushed, clear needs_recovery flag. */
4164 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4165 error = ext4_commit_super(sb, 1);
6b0310fb
ES		 4166out:
4167 /* we rely on s_frozen to stop further updates */
4168 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4169 return error;
ac27a0ec
DK		 4170}
4171
4172/*
4173 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4174 * flag here, even though the filesystem is not technically dirty yet.
4175 */
c4be0c1d 4176static int ext4_unfreeze(struct super_block *sb)
ac27a0ec 4177{
9ca92389
TT		 4178 if (sb->s_flags & MS_RDONLY)
4179 return 0;
4180
4181 lock_super(sb);
4182 /* Reset the needs_recovery flag before the fs is unlocked. */
4183 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4184 ext4_commit_super(sb, 1);
4185 unlock_super(sb);
c4be0c1d 4186 return 0;
ac27a0ec
DK		 4187}
4188
673c6100
TT		 4189/*
4190 * Structure to save mount options for ext4_remount's benefit
4191 */
4192struct ext4_mount_options {
4193 unsigned long s_mount_opt;
a2595b8a 4194 unsigned long s_mount_opt2;
673c6100
TT		 4195 uid_t s_resuid;
4196 gid_t s_resgid;
4197 unsigned long s_commit_interval;
4198 u32 s_min_batch_time, s_max_batch_time;
4199#ifdef CONFIG_QUOTA
4200 int s_jquota_fmt;
4201 char *s_qf_names[MAXQUOTAS];
4202#endif
4203};
4204
2b2d6d01 4205static int ext4_remount(struct super_block *sb, int *flags, char *data)
ac27a0ec 4206{
2b2d6d01 4207 struct ext4_super_block *es;
617ba13b
MC		 4208 struct ext4_sb_info *sbi = EXT4_SB(sb);
4209 ext4_fsblk_t n_blocks_count = 0;
ac27a0ec 4210 unsigned long old_sb_flags;
617ba13b 4211 struct ext4_mount_options old_opts;
c79d967d 4212 int enable_quota = 0;
8a266467 4213 ext4_group_t g;
b3881f74 4214 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
ac27a0ec
DK		 4215 int err;
4216#ifdef CONFIG_QUOTA
4217 int i;
4218#endif
d4c402d9 4219 char *orig_data = kstrdup(data, GFP_KERNEL);
ac27a0ec
DK		 4220
4221 /* Store the original options */
bbd6851a 4222 lock_super(sb);
ac27a0ec
DK		 4223 old_sb_flags = sb->s_flags;
4224 old_opts.s_mount_opt = sbi->s_mount_opt;
a2595b8a 4225 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
ac27a0ec
DK		 4226 old_opts.s_resuid = sbi->s_resuid;
4227 old_opts.s_resgid = sbi->s_resgid;
4228 old_opts.s_commit_interval = sbi->s_commit_interval;
30773840
TT		 4229 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4230 old_opts.s_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 4231#ifdef CONFIG_QUOTA
4232 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4233 for (i = 0; i < MAXQUOTAS; i++)
4234 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
4235#endif
b3881f74
TT		 4236 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4237 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
ac27a0ec
DK		 4238
4239 /*
4240 * Allow the "check" option to be passed as a remount option.
4241 */
b3881f74
TT		 4242 if (!parse_options(data, sb, NULL, &journal_ioprio,
4243 &n_blocks_count, 1)) {
ac27a0ec
DK		 4244 err = -EINVAL;
4245 goto restore_opts;
4246 }
4247
4ab2f15b 4248 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
c67d859e 4249 ext4_abort(sb, "Abort forced by user");
ac27a0ec
DK		 4250
4251 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 4252 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec
DK		 4253
4254 es = sbi->s_es;
4255
b3881f74 4256 if (sbi->s_journal) {
0390131b 4257 ext4_init_journal_params(sb, sbi->s_journal);
b3881f74
TT		 4258 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4259 }
ac27a0ec
DK		 4260
4261 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
bd81d8ee 4262 n_blocks_count > ext4_blocks_count(es)) {
4ab2f15b 4263 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
ac27a0ec
DK		 4264 err = -EROFS;
4265 goto restore_opts;
4266 }
4267
4268 if (*flags & MS_RDONLY) {
0f0dd62f
CH		 4269 err = dquot_suspend(sb, -1);
4270 if (err < 0)
c79d967d 4271 goto restore_opts;
c79d967d 4272
ac27a0ec
DK		 4273 /*
4274 * First of all, the unconditional stuff we have to do
4275 * to disable replay of the journal when we next remount
4276 */
4277 sb->s_flags |= MS_RDONLY;
4278
4279 /*
4280 * OK, test if we are remounting a valid rw partition
4281 * readonly, and if so set the rdonly flag and then
4282 * mark the partition as valid again.
4283 */
617ba13b
MC		 4284 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4285 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 4286 es->s_state = cpu_to_le16(sbi->s_mount_state);
4287
a63c9eb2 4288 if (sbi->s_journal)
0390131b 4289 ext4_mark_recovery_complete(sb, es);
ac27a0ec 4290 } else {
a13fb1a4
ES		 4291 /* Make sure we can mount this feature set readwrite */
4292 if (!ext4_feature_set_ok(sb, 0)) {
ac27a0ec
DK		 4293 err = -EROFS;
4294 goto restore_opts;
4295 }
8a266467
TT		 4296 /*
4297 * Make sure the group descriptor checksums
0b8e58a1 4298 * are sane. If they aren't, refuse to remount r/w.
8a266467
TT		 4299 */
4300 for (g = 0; g < sbi->s_groups_count; g++) {
4301 struct ext4_group_desc *gdp =
4302 ext4_get_group_desc(sb, g, NULL);
4303
4304 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
b31e1552
ES		 4305 ext4_msg(sb, KERN_ERR,
4306 "ext4_remount: Checksum for group %u failed (%u!=%u)",
8a266467
TT		 4307 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4308 le16_to_cpu(gdp->bg_checksum));
4309 err = -EINVAL;
4310 goto restore_opts;
4311 }
4312 }
4313
ead6596b
ES		 4314 /*
4315 * If we have an unprocessed orphan list hanging
4316 * around from a previously readonly bdev mount,
4317 * require a full umount/remount for now.
4318 */
4319 if (es->s_last_orphan) {
b31e1552 4320 ext4_msg(sb, KERN_WARNING, "Couldn't "
ead6596b
ES		 4321 "remount RDWR because of unprocessed "
4322 "orphan inode list. Please "
b31e1552 4323 "umount/remount instead");
ead6596b
ES		 4324 err = -EINVAL;
4325 goto restore_opts;
4326 }
4327
ac27a0ec
DK		 4328 /*
4329 * Mounting a RDONLY partition read-write, so reread
4330 * and store the current valid flag. (It may have
4331 * been changed by e2fsck since we originally mounted
4332 * the partition.)
4333 */
0390131b
FM		 4334 if (sbi->s_journal)
4335 ext4_clear_journal_err(sb, es);
ac27a0ec 4336 sbi->s_mount_state = le16_to_cpu(es->s_state);
617ba13b 4337 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
ac27a0ec 4338 goto restore_opts;
2b2d6d01 4339 if (!ext4_setup_super(sb, es, 0))
ac27a0ec 4340 sb->s_flags &= ~MS_RDONLY;
c79d967d 4341 enable_quota = 1;
ac27a0ec
DK		 4342 }
4343 }
bfff6873
LC		 4344
4345 /*
4346 * Reinitialize lazy itable initialization thread based on
4347 * current settings
4348 */
4349 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4350 ext4_unregister_li_request(sb);
4351 else {
4352 ext4_group_t first_not_zeroed;
4353 first_not_zeroed = ext4_has_uninit_itable(sb);
4354 ext4_register_li_request(sb, first_not_zeroed);
4355 }
4356
6fd058f7 4357 ext4_setup_system_zone(sb);
0390131b 4358 if (sbi->s_journal == NULL)
e2d67052 4359 ext4_commit_super(sb, 1);
0390131b 4360
ac27a0ec
DK		 4361#ifdef CONFIG_QUOTA
4362 /* Release old quota file names */
4363 for (i = 0; i < MAXQUOTAS; i++)
4364 if (old_opts.s_qf_names[i] &&
4365 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4366 kfree(old_opts.s_qf_names[i]);
4367#endif
bbd6851a 4368 unlock_super(sb);
c79d967d 4369 if (enable_quota)
0f0dd62f 4370 dquot_resume(sb, -1);
d4c402d9
CW		 4371
4372 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4373 kfree(orig_data);
ac27a0ec 4374 return 0;
0b8e58a1 4375
ac27a0ec
DK		 4376restore_opts:
4377 sb->s_flags = old_sb_flags;
4378 sbi->s_mount_opt = old_opts.s_mount_opt;
a2595b8a 4379 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
ac27a0ec
DK		 4380 sbi->s_resuid = old_opts.s_resuid;
4381 sbi->s_resgid = old_opts.s_resgid;
4382 sbi->s_commit_interval = old_opts.s_commit_interval;
30773840
TT		 4383 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4384 sbi->s_max_batch_time = old_opts.s_max_batch_time;
ac27a0ec
DK		 4385#ifdef CONFIG_QUOTA
4386 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4387 for (i = 0; i < MAXQUOTAS; i++) {
4388 if (sbi->s_qf_names[i] &&
4389 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4390 kfree(sbi->s_qf_names[i]);
4391 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4392 }
4393#endif
bbd6851a 4394 unlock_super(sb);
d4c402d9 4395 kfree(orig_data);
ac27a0ec
DK		 4396 return err;
4397}
4398
2b2d6d01 4399static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
ac27a0ec
DK		 4400{
4401 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 4402 struct ext4_sb_info *sbi = EXT4_SB(sb);
4403 struct ext4_super_block *es = sbi->s_es;
960cc398 4404 u64 fsid;
ac27a0ec 4405
5e70030d
BP		 4406 if (test_opt(sb, MINIX_DF)) {
4407 sbi->s_overhead_last = 0;
6bc9feff 4408 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
8df9675f 4409 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
5e70030d 4410 ext4_fsblk_t overhead = 0;
ac27a0ec
DK		 4411
4412 /*
5e70030d
BP		 4413 * Compute the overhead (FS structures). This is constant
4414 * for a given filesystem unless the number of block groups
4415 * changes so we cache the previous value until it does.
ac27a0ec
DK		 4416 */
4417
4418 /*
4419 * All of the blocks before first_data_block are
4420 * overhead
4421 */
4422 overhead = le32_to_cpu(es->s_first_data_block);
4423
4424 /*
4425 * Add the overhead attributed to the superblock and
4426 * block group descriptors. If the sparse superblocks
4427 * feature is turned on, then not all groups have this.
4428 */
4429 for (i = 0; i < ngroups; i++) {
617ba13b
MC		 4430 overhead += ext4_bg_has_super(sb, i) +
4431 ext4_bg_num_gdb(sb, i);
ac27a0ec
DK		 4432 cond_resched();
4433 }
4434
4435 /*
4436 * Every block group has an inode bitmap, a block
4437 * bitmap, and an inode table.
4438 */
5e70030d
BP		 4439 overhead += ngroups * (2 + sbi->s_itb_per_group);
4440 sbi->s_overhead_last = overhead;
4441 smp_wmb();
6bc9feff 4442 sbi->s_blocks_last = ext4_blocks_count(es);
ac27a0ec
DK		 4443 }
4444
617ba13b 4445 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 4446 buf->f_bsize = sb->s_blocksize;
5e70030d 4447 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
6bc6e63f
AK		 4448 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
4449 percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
bd81d8ee
LV		 4450 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
4451 if (buf->f_bfree < ext4_r_blocks_count(es))
ac27a0ec
DK		 4452 buf->f_bavail = 0;
4453 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 4454 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
617ba13b 4455 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 4456 fsid = le64_to_cpup((void *)es->s_uuid) ^
4457 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4458 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4459 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
0b8e58a1 4460
ac27a0ec
DK		 4461 return 0;
4462}
4463
0b8e58a1
AD		 4464/* Helper function for writing quotas on sync - we need to start transaction
4465 * before quota file is locked for write. Otherwise the are possible deadlocks:
ac27a0ec 4466 * Process 1 Process 2
617ba13b 4467 * ext4_create() quota_sync()
a269eb18 4468 * jbd2_journal_start() write_dquot()
871a2931 4469 * dquot_initialize() down(dqio_mutex)
dab291af 4470 * down(dqio_mutex) jbd2_journal_start()
ac27a0ec
DK		 4471 *
4472 */
4473
4474#ifdef CONFIG_QUOTA
4475
4476static inline struct inode *dquot_to_inode(struct dquot *dquot)
4477{
4478 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
4479}
4480
617ba13b 4481static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 4482{
4483 int ret, err;
4484 handle_t *handle;
4485 struct inode *inode;
4486
4487 inode = dquot_to_inode(dquot);
617ba13b 4488 handle = ext4_journal_start(inode,
0b8e58a1 4489 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 4490 if (IS_ERR(handle))
4491 return PTR_ERR(handle);
4492 ret = dquot_commit(dquot);
617ba13b 4493 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4494 if (!ret)
4495 ret = err;
4496 return ret;
4497}
4498
617ba13b 4499static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 4500{
4501 int ret, err;
4502 handle_t *handle;
4503
617ba13b 4504 handle = ext4_journal_start(dquot_to_inode(dquot),
0b8e58a1 4505 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 4506 if (IS_ERR(handle))
4507 return PTR_ERR(handle);
4508 ret = dquot_acquire(dquot);
617ba13b 4509 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4510 if (!ret)
4511 ret = err;
4512 return ret;
4513}
4514
617ba13b 4515static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 4516{
4517 int ret, err;
4518 handle_t *handle;
4519
617ba13b 4520 handle = ext4_journal_start(dquot_to_inode(dquot),
0b8e58a1 4521 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 4522 if (IS_ERR(handle)) {
4523 /* Release dquot anyway to avoid endless cycle in dqput() */
4524 dquot_release(dquot);
ac27a0ec 4525 return PTR_ERR(handle);
9c3013e9 4526 }
ac27a0ec 4527 ret = dquot_release(dquot);
617ba13b 4528 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4529 if (!ret)
4530 ret = err;
4531 return ret;
4532}
4533
617ba13b 4534static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec 4535{
2c8be6b2 4536 /* Are we journaling quotas? */
617ba13b
MC		 4537 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
4538 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
ac27a0ec 4539 dquot_mark_dquot_dirty(dquot);
617ba13b 4540 return ext4_write_dquot(dquot);
ac27a0ec
DK		 4541 } else {
4542 return dquot_mark_dquot_dirty(dquot);
4543 }
4544}
4545
617ba13b 4546static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 4547{
4548 int ret, err;
4549 handle_t *handle;
4550
4551 /* Data block + inode block */
617ba13b 4552 handle = ext4_journal_start(sb->s_root->d_inode, 2);
ac27a0ec
DK		 4553 if (IS_ERR(handle))
4554 return PTR_ERR(handle);
4555 ret = dquot_commit_info(sb, type);
617ba13b 4556 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4557 if (!ret)
4558 ret = err;
4559 return ret;
4560}
4561
4562/*
4563 * Turn on quotas during mount time - we need to find
4564 * the quota file and such...
4565 */
617ba13b 4566static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 4567{
287a8095
CH		 4568 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4569 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 4570}
4571
4572/*
4573 * Standard function to be called on quota_on
4574 */
617ba13b 4575static int ext4_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 4576 struct path *path)
ac27a0ec
DK		 4577{
4578 int err;
ac27a0ec
DK		 4579
4580 if (!test_opt(sb, QUOTA))
4581 return -EINVAL;
0623543b 4582
ac27a0ec 4583 /* Quotafile not on the same filesystem? */
f00c9e44 4584 if (path->mnt->mnt_sb != sb)
ac27a0ec 4585 return -EXDEV;
0623543b
JK		 4586 /* Journaling quota? */
4587 if (EXT4_SB(sb)->s_qf_names[type]) {
2b2d6d01 4588 /* Quotafile not in fs root? */
f00c9e44 4589 if (path->dentry->d_parent != sb->s_root)
b31e1552
ES		 4590 ext4_msg(sb, KERN_WARNING,
4591 "Quota file not on filesystem root. "
4592 "Journaled quota will not work");
2b2d6d01 4593 }
0623543b
JK		 4594
4595 /*
4596 * When we journal data on quota file, we have to flush journal to see
4597 * all updates to the file when we bypass pagecache...
4598 */
0390131b 4599 if (EXT4_SB(sb)->s_journal &&
f00c9e44 4600 ext4_should_journal_data(path->dentry->d_inode)) {
0623543b
JK		 4601 /*
4602 * We don't need to lock updates but journal_flush() could
4603 * otherwise be livelocked...
4604 */
4605 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 4606 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
0623543b 4607 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
f00c9e44 4608 if (err)
7ffe1ea8 4609 return err;
0623543b
JK		 4610 }
4611
f00c9e44 4612 return dquot_quota_on(sb, type, format_id, path);
ac27a0ec
DK		 4613}
4614
ca0e05e4
DM		 4615static int ext4_quota_off(struct super_block *sb, int type)
4616{
87009d86
DM		 4617 /* Force all delayed allocation blocks to be allocated.
4618 * Caller already holds s_umount sem */
4619 if (test_opt(sb, DELALLOC))
ca0e05e4 4620 sync_filesystem(sb);
ca0e05e4
DM		 4621
4622 return dquot_quota_off(sb, type);
4623}
4624
ac27a0ec
DK		 4625/* Read data from quotafile - avoid pagecache and such because we cannot afford
4626 * acquiring the locks... As quota files are never truncated and quota code
25985edc 4627 * itself serializes the operations (and no one else should touch the files)
ac27a0ec 4628 * we don't have to be afraid of races */
617ba13b 4629static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 4630 size_t len, loff_t off)
4631{
4632 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 4633 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 4634 int err = 0;
4635 int offset = off & (sb->s_blocksize - 1);
4636 int tocopy;
4637 size_t toread;
4638 struct buffer_head *bh;
4639 loff_t i_size = i_size_read(inode);
4640
4641 if (off > i_size)
4642 return 0;
4643 if (off+len > i_size)
4644 len = i_size-off;
4645 toread = len;
4646 while (toread > 0) {
4647 tocopy = sb->s_blocksize - offset < toread ?
4648 sb->s_blocksize - offset : toread;
617ba13b 4649 bh = ext4_bread(NULL, inode, blk, 0, &err);
ac27a0ec
DK		 4650 if (err)
4651 return err;
4652 if (!bh) /* A hole? */
4653 memset(data, 0, tocopy);
4654 else
4655 memcpy(data, bh->b_data+offset, tocopy);
4656 brelse(bh);
4657 offset = 0;
4658 toread -= tocopy;
4659 data += tocopy;
4660 blk++;
4661 }
4662 return len;
4663}
4664
4665/* Write to quotafile (we know the transaction is already started and has
4666 * enough credits) */
617ba13b 4667static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 4668 const char *data, size_t len, loff_t off)
4669{
4670 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 4671 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 4672 int err = 0;
4673 int offset = off & (sb->s_blocksize - 1);
ac27a0ec
DK		 4674 struct buffer_head *bh;
4675 handle_t *handle = journal_current_handle();
4676
0390131b 4677 if (EXT4_SB(sb)->s_journal && !handle) {
b31e1552
ES		 4678 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4679 " cancelled because transaction is not started",
9c3013e9
JK		 4680 (unsigned long long)off, (unsigned long long)len);
4681 return -EIO;
4682 }
67eeb568
DM		 4683 /*
4684 * Since we account only one data block in transaction credits,
4685 * then it is impossible to cross a block boundary.
4686 */
4687 if (sb->s_blocksize - offset < len) {
4688 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4689 " cancelled because not block aligned",
4690 (unsigned long long)off, (unsigned long long)len);
4691 return -EIO;
4692 }
4693
ac27a0ec 4694 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
67eeb568
DM		 4695 bh = ext4_bread(handle, inode, blk, 1, &err);
4696 if (!bh)
4697 goto out;
62d2b5f2
JK		 4698 err = ext4_journal_get_write_access(handle, bh);
4699 if (err) {
4700 brelse(bh);
4701 goto out;
ac27a0ec 4702 }
67eeb568
DM		 4703 lock_buffer(bh);
4704 memcpy(bh->b_data+offset, data, len);
4705 flush_dcache_page(bh->b_page);
4706 unlock_buffer(bh);
62d2b5f2 4707 err = ext4_handle_dirty_metadata(handle, NULL, bh);
67eeb568 4708 brelse(bh);
ac27a0ec 4709out:
67eeb568 4710 if (err) {
4d04e4fb 4711 mutex_unlock(&inode->i_mutex);
ac27a0ec 4712 return err;
4d04e4fb 4713 }
67eeb568
DM		 4714 if (inode->i_size < off + len) {
4715 i_size_write(inode, off + len);
617ba13b 4716 EXT4_I(inode)->i_disksize = inode->i_size;
ac27a0ec 4717 }
ac27a0ec 4718 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
617ba13b 4719 ext4_mark_inode_dirty(handle, inode);
ac27a0ec 4720 mutex_unlock(&inode->i_mutex);
67eeb568 4721 return len;
ac27a0ec
DK		 4722}
4723
4724#endif
4725
152a0836
AV		 4726static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
4727 const char *dev_name, void *data)
ac27a0ec 4728{
152a0836 4729 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
ac27a0ec
DK		 4730}
4731
37f328eb 4732#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
24b58424
TT		 4733static struct file_system_type ext2_fs_type = {
4734 .owner = THIS_MODULE,
4735 .name = "ext2",
152a0836 4736 .mount = ext4_mount,
24b58424
TT		 4737 .kill_sb = kill_block_super,
4738 .fs_flags = FS_REQUIRES_DEV,
4739};
4740
4741static inline void register_as_ext2(void)
4742{
4743 int err = register_filesystem(&ext2_fs_type);
4744 if (err)
4745 printk(KERN_WARNING
4746 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
4747}
4748
4749static inline void unregister_as_ext2(void)
4750{
4751 unregister_filesystem(&ext2_fs_type);
4752}
51b7e3c9 4753MODULE_ALIAS("ext2");
24b58424
TT		 4754#else
4755static inline void register_as_ext2(void) { }
4756static inline void unregister_as_ext2(void) { }
4757#endif
4758
37f328eb 4759#if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
24b58424
TT		 4760static inline void register_as_ext3(void)
4761{
4762 int err = register_filesystem(&ext3_fs_type);
4763 if (err)
4764 printk(KERN_WARNING
4765 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
4766}
4767
4768static inline void unregister_as_ext3(void)
4769{
4770 unregister_filesystem(&ext3_fs_type);
4771}
51b7e3c9 4772MODULE_ALIAS("ext3");
24b58424
TT		 4773#else
4774static inline void register_as_ext3(void) { }
4775static inline void unregister_as_ext3(void) { }
4776#endif
4777
03010a33
TT		 4778static struct file_system_type ext4_fs_type = {
4779 .owner = THIS_MODULE,
4780 .name = "ext4",
152a0836 4781 .mount = ext4_mount,
03010a33
TT		 4782 .kill_sb = kill_block_super,
4783 .fs_flags = FS_REQUIRES_DEV,
4784};
4785
8f021222 4786static int __init ext4_init_feat_adverts(void)
857ac889
LC		 4787{
4788 struct ext4_features *ef;
4789 int ret = -ENOMEM;
4790
4791 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
4792 if (!ef)
4793 goto out;
4794
4795 ef->f_kobj.kset = ext4_kset;
4796 init_completion(&ef->f_kobj_unregister);
4797 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
4798 "features");
4799 if (ret) {
4800 kfree(ef);
4801 goto out;
4802 }
4803
4804 ext4_feat = ef;
4805 ret = 0;
4806out:
4807 return ret;
4808}
4809
8f021222
LC		 4810static void ext4_exit_feat_adverts(void)
4811{
4812 kobject_put(&ext4_feat->f_kobj);
4813 wait_for_completion(&ext4_feat->f_kobj_unregister);
4814 kfree(ext4_feat);
4815}
4816
e9e3bcec
ES		 4817/* Shared across all ext4 file systems */
4818wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
4819struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
4820
5dabfc78 4821static int __init ext4_init_fs(void)
ac27a0ec 4822{
e9e3bcec 4823 int i, err;
c9de560d 4824
12e9b892 4825 ext4_check_flag_values();
e9e3bcec
ES		 4826
4827 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
4828 mutex_init(&ext4__aio_mutex[i]);
4829 init_waitqueue_head(&ext4__ioend_wq[i]);
4830 }
4831
5dabfc78 4832 err = ext4_init_pageio();
6fd058f7
TT		 4833 if (err)
4834 return err;
5dabfc78 4835 err = ext4_init_system_zone();
bd2d0210 4836 if (err)
dd68314c 4837 goto out7;
3197ebdb
TT		 4838 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
4839 if (!ext4_kset)
dd68314c 4840 goto out6;
9f6200bb 4841 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
dd68314c
TT		 4842 if (!ext4_proc_root)
4843 goto out5;
857ac889
LC		 4844
4845 err = ext4_init_feat_adverts();
dd68314c
TT		 4846 if (err)
4847 goto out4;
857ac889 4848
5dabfc78 4849 err = ext4_init_mballoc();
ac27a0ec 4850 if (err)
6fd058f7 4851 goto out3;
c9de560d 4852
5dabfc78 4853 err = ext4_init_xattr();
c9de560d
AT		 4854 if (err)
4855 goto out2;
ac27a0ec
DK		 4856 err = init_inodecache();
4857 if (err)
4858 goto out1;
24b58424
TT		 4859 register_as_ext2();
4860 register_as_ext3();
03010a33 4861 err = register_filesystem(&ext4_fs_type);
ac27a0ec
DK		 4862 if (err)
4863 goto out;
bfff6873
LC		 4864
4865 ext4_li_info = NULL;
4866 mutex_init(&ext4_li_mtx);
ac27a0ec
DK		 4867 return 0;
4868out:
24b58424
TT		 4869 unregister_as_ext2();
4870 unregister_as_ext3();
ac27a0ec
DK		 4871 destroy_inodecache();
4872out1:
5dabfc78 4873 ext4_exit_xattr();
c9de560d 4874out2:
5dabfc78 4875 ext4_exit_mballoc();
6fd058f7 4876out3:
8f021222 4877 ext4_exit_feat_adverts();
dd68314c 4878out4:
6fd058f7 4879 remove_proc_entry("fs/ext4", NULL);
dd68314c 4880out5:
6fd058f7 4881 kset_unregister(ext4_kset);
dd68314c 4882out6:
5dabfc78 4883 ext4_exit_system_zone();
dd68314c 4884out7:
5dabfc78 4885 ext4_exit_pageio();
ac27a0ec
DK		 4886 return err;
4887}
4888
5dabfc78 4889static void __exit ext4_exit_fs(void)
ac27a0ec 4890{
bfff6873 4891 ext4_destroy_lazyinit_thread();
24b58424
TT		 4892 unregister_as_ext2();
4893 unregister_as_ext3();
03010a33 4894 unregister_filesystem(&ext4_fs_type);
ac27a0ec 4895 destroy_inodecache();
5dabfc78
TT		 4896 ext4_exit_xattr();
4897 ext4_exit_mballoc();
8f021222 4898 ext4_exit_feat_adverts();
9f6200bb 4899 remove_proc_entry("fs/ext4", NULL);
3197ebdb 4900 kset_unregister(ext4_kset);
5dabfc78
TT		 4901 ext4_exit_system_zone();
4902 ext4_exit_pageio();
ac27a0ec
DK		 4903}
4904
4905MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
83982b6f 4906MODULE_DESCRIPTION("Fourth Extended Filesystem");
ac27a0ec 4907MODULE_LICENSE("GPL");
5dabfc78
TT		 4908module_init(ext4_init_fs)
4909module_exit(ext4_exit_fs)
kernel source for telekom puls (alcatel/mediatek)