]> Git Repo - linux.git/blame - fs/ext4/super.c
projects / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
Merge branch 'for-linus' of git://git.kernel.dk/linux-2.6-block
[linux.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 ([email protected])
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 ([email protected]), 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;
bfff6873
LC		 57struct ext4_lazy_init *ext4_li_info;
58struct mutex ext4_li_mtx;
857ac889 59struct 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);
bfff6873
LC		 78static void ext4_destroy_lazyinit_thread(void);
79static void ext4_unregister_li_request(struct super_block *sb);
8f1f7453 80static void ext4_clear_request_list(void);
ac27a0ec 81
ba69f9ab
JK		 82#if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
83static struct file_system_type ext3_fs_type = {
84 .owner = THIS_MODULE,
85 .name = "ext3",
152a0836 86 .mount = ext4_mount,
ba69f9ab
JK		 87 .kill_sb = kill_block_super,
88 .fs_flags = FS_REQUIRES_DEV,
89};
90#define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
91#else
92#define IS_EXT3_SB(sb) (0)
93#endif
bd81d8ee 94
8fadc143
AR		 95ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
96 struct ext4_group_desc *bg)
bd81d8ee 97{
3a14589c 98 return le32_to_cpu(bg->bg_block_bitmap_lo) |
8fadc143 99 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 100 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 101}
102
8fadc143
AR		 103ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
104 struct ext4_group_desc *bg)
bd81d8ee 105{
5272f837 106 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
8fadc143 107 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 108 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 109}
110
8fadc143
AR		 111ext4_fsblk_t ext4_inode_table(struct super_block *sb,
112 struct ext4_group_desc *bg)
bd81d8ee 113{
5272f837 114 return le32_to_cpu(bg->bg_inode_table_lo) |
8fadc143 115 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 116 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
bd81d8ee
LV		 117}
118
560671a0
AK		 119__u32 ext4_free_blks_count(struct super_block *sb,
120 struct ext4_group_desc *bg)
121{
122 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
123 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 124 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
560671a0
AK		 125}
126
127__u32 ext4_free_inodes_count(struct super_block *sb,
128 struct ext4_group_desc *bg)
129{
130 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
131 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 132 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
560671a0
AK		 133}
134
135__u32 ext4_used_dirs_count(struct super_block *sb,
136 struct ext4_group_desc *bg)
137{
138 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
139 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 140 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
560671a0
AK		 141}
142
143__u32 ext4_itable_unused_count(struct super_block *sb,
144 struct ext4_group_desc *bg)
145{
146 return le16_to_cpu(bg->bg_itable_unused_lo) |
147 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 148 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
560671a0
AK		 149}
150
8fadc143
AR		 151void ext4_block_bitmap_set(struct super_block *sb,
152 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 153{
3a14589c 154 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 155 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
156 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 157}
158
8fadc143
AR		 159void ext4_inode_bitmap_set(struct super_block *sb,
160 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 161{
5272f837 162 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 163 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
164 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 165}
166
8fadc143
AR		 167void ext4_inode_table_set(struct super_block *sb,
168 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 169{
5272f837 170 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 171 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
172 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 173}
174
560671a0
AK		 175void ext4_free_blks_set(struct super_block *sb,
176 struct ext4_group_desc *bg, __u32 count)
177{
178 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
179 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
180 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
181}
182
183void ext4_free_inodes_set(struct super_block *sb,
184 struct ext4_group_desc *bg, __u32 count)
185{
186 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
187 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
188 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
189}
190
191void ext4_used_dirs_set(struct super_block *sb,
192 struct ext4_group_desc *bg, __u32 count)
193{
194 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
195 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
196 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
197}
198
199void ext4_itable_unused_set(struct super_block *sb,
200 struct ext4_group_desc *bg, __u32 count)
201{
202 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
203 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
204 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
205}
206
d3d1faf6
CW		 207
208/* Just increment the non-pointer handle value */
209static handle_t *ext4_get_nojournal(void)
210{
211 handle_t *handle = current->journal_info;
212 unsigned long ref_cnt = (unsigned long)handle;
213
214 BUG_ON(ref_cnt >= EXT4_NOJOURNAL_MAX_REF_COUNT);
215
216 ref_cnt++;
217 handle = (handle_t *)ref_cnt;
218
219 current->journal_info = handle;
220 return handle;
221}
222
223
224/* Decrement the non-pointer handle value */
225static void ext4_put_nojournal(handle_t *handle)
226{
227 unsigned long ref_cnt = (unsigned long)handle;
228
229 BUG_ON(ref_cnt == 0);
230
231 ref_cnt--;
232 handle = (handle_t *)ref_cnt;
233
234 current->journal_info = handle;
235}
236
ac27a0ec 237/*
dab291af 238 * Wrappers for jbd2_journal_start/end.
ac27a0ec
DK		 239 *
240 * The only special thing we need to do here is to make sure that all
241 * journal_end calls result in the superblock being marked dirty, so
242 * that sync() will call the filesystem's write_super callback if
243 * appropriate.
244 */
617ba13b 245handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
ac27a0ec
DK		 246{
247 journal_t *journal;
248
249 if (sb->s_flags & MS_RDONLY)
250 return ERR_PTR(-EROFS);
251
437f88cc 252 vfs_check_frozen(sb, SB_FREEZE_TRANS);
ac27a0ec
DK		 253 /* Special case here: if the journal has aborted behind our
254 * backs (eg. EIO in the commit thread), then we still need to
255 * take the FS itself readonly cleanly. */
617ba13b 256 journal = EXT4_SB(sb)->s_journal;
0390131b
FM		 257 if (journal) {
258 if (is_journal_aborted(journal)) {
c67d859e 259 ext4_abort(sb, "Detected aborted journal");
0390131b
FM		 260 return ERR_PTR(-EROFS);
261 }
262 return jbd2_journal_start(journal, nblocks);
ac27a0ec 263 }
d3d1faf6 264 return ext4_get_nojournal();
ac27a0ec
DK		 265}
266
267/*
268 * The only special thing we need to do here is to make sure that all
dab291af 269 * jbd2_journal_stop calls result in the superblock being marked dirty, so
ac27a0ec
DK		 270 * that sync() will call the filesystem's write_super callback if
271 * appropriate.
272 */
c398eda0 273int __ext4_journal_stop(const char *where, unsigned int line, handle_t *handle)
ac27a0ec
DK		 274{
275 struct super_block *sb;
276 int err;
277 int rc;
278
0390131b 279 if (!ext4_handle_valid(handle)) {
d3d1faf6 280 ext4_put_nojournal(handle);
0390131b
FM		 281 return 0;
282 }
ac27a0ec
DK		 283 sb = handle->h_transaction->t_journal->j_private;
284 err = handle->h_err;
dab291af 285 rc = jbd2_journal_stop(handle);
ac27a0ec
DK		 286
287 if (!err)
288 err = rc;
289 if (err)
c398eda0 290 __ext4_std_error(sb, where, line, err);
ac27a0ec
DK		 291 return err;
292}
293
90c7201b
TT		 294void ext4_journal_abort_handle(const char *caller, unsigned int line,
295 const char *err_fn, struct buffer_head *bh,
296 handle_t *handle, int err)
ac27a0ec
DK		 297{
298 char nbuf[16];
617ba13b 299 const char *errstr = ext4_decode_error(NULL, err, nbuf);
ac27a0ec 300
0390131b
FM		 301 BUG_ON(!ext4_handle_valid(handle));
302
ac27a0ec
DK		 303 if (bh)
304 BUFFER_TRACE(bh, "abort");
305
306 if (!handle->h_err)
307 handle->h_err = err;
308
309 if (is_handle_aborted(handle))
310 return;
311
90c7201b
TT		 312 printk(KERN_ERR "%s:%d: aborting transaction: %s in %s\n",
313 caller, line, errstr, err_fn);
ac27a0ec 314
dab291af 315 jbd2_journal_abort_handle(handle);
ac27a0ec
DK		 316}
317
1c13d5c0
TT		 318static void __save_error_info(struct super_block *sb, const char *func,
319 unsigned int line)
320{
321 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
322
323 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
324 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
325 es->s_last_error_time = cpu_to_le32(get_seconds());
326 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
327 es->s_last_error_line = cpu_to_le32(line);
328 if (!es->s_first_error_time) {
329 es->s_first_error_time = es->s_last_error_time;
330 strncpy(es->s_first_error_func, func,
331 sizeof(es->s_first_error_func));
332 es->s_first_error_line = cpu_to_le32(line);
333 es->s_first_error_ino = es->s_last_error_ino;
334 es->s_first_error_block = es->s_last_error_block;
335 }
66e61a9e
TT		 336 /*
337 * Start the daily error reporting function if it hasn't been
338 * started already
339 */
340 if (!es->s_error_count)
341 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
1c13d5c0
TT		 342 es->s_error_count = cpu_to_le32(le32_to_cpu(es->s_error_count) + 1);
343}
344
345static void save_error_info(struct super_block *sb, const char *func,
346 unsigned int line)
347{
348 __save_error_info(sb, func, line);
349 ext4_commit_super(sb, 1);
350}
351
352
ac27a0ec
DK		 353/* Deal with the reporting of failure conditions on a filesystem such as
354 * inconsistencies detected or read IO failures.
355 *
356 * On ext2, we can store the error state of the filesystem in the
617ba13b 357 * superblock. That is not possible on ext4, because we may have other
ac27a0ec
DK		 358 * write ordering constraints on the superblock which prevent us from
359 * writing it out straight away; and given that the journal is about to
360 * be aborted, we can't rely on the current, or future, transactions to
361 * write out the superblock safely.
362 *
dab291af 363 * We'll just use the jbd2_journal_abort() error code to record an error in
d6b198bc 364 * the journal instead. On recovery, the journal will complain about
ac27a0ec
DK		 365 * that error until we've noted it down and cleared it.
366 */
367
617ba13b 368static void ext4_handle_error(struct super_block *sb)
ac27a0ec 369{
ac27a0ec
DK		 370 if (sb->s_flags & MS_RDONLY)
371 return;
372
2b2d6d01 373 if (!test_opt(sb, ERRORS_CONT)) {
617ba13b 374 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 375
4ab2f15b 376 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
ac27a0ec 377 if (journal)
dab291af 378 jbd2_journal_abort(journal, -EIO);
ac27a0ec 379 }
2b2d6d01 380 if (test_opt(sb, ERRORS_RO)) {
b31e1552 381 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
ac27a0ec
DK		 382 sb->s_flags |= MS_RDONLY;
383 }
ac27a0ec 384 if (test_opt(sb, ERRORS_PANIC))
617ba13b 385 panic("EXT4-fs (device %s): panic forced after error\n",
ac27a0ec
DK		 386 sb->s_id);
387}
388
12062ddd 389void __ext4_error(struct super_block *sb, const char *function,
c398eda0 390 unsigned int line, const char *fmt, ...)
ac27a0ec 391{
0ff2ea7d 392 struct va_format vaf;
ac27a0ec
DK		 393 va_list args;
394
395 va_start(args, fmt);
0ff2ea7d
JP		 396 vaf.fmt = fmt;
397 vaf.va = &args;
398 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
399 sb->s_id, function, line, current->comm, &vaf);
ac27a0ec
DK		 400 va_end(args);
401
617ba13b 402 ext4_handle_error(sb);
ac27a0ec
DK		 403}
404
c398eda0
TT		 405void ext4_error_inode(struct inode *inode, const char *function,
406 unsigned int line, ext4_fsblk_t block,
273df556
FM		 407 const char *fmt, ...)
408{
409 va_list args;
f7c21177 410 struct va_format vaf;
1c13d5c0 411 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
273df556 412
1c13d5c0
TT		 413 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
414 es->s_last_error_block = cpu_to_le64(block);
415 save_error_info(inode->i_sb, function, line);
273df556 416 va_start(args, fmt);
f7c21177
TT		 417 vaf.fmt = fmt;
418 vaf.va = &args;
c398eda0
TT		 419 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: inode #%lu: ",
420 inode->i_sb->s_id, function, line, inode->i_ino);
421 if (block)
f7c21177
TT		 422 printk(KERN_CONT "block %llu: ", block);
423 printk(KERN_CONT "comm %s: %pV\n", current->comm, &vaf);
273df556
FM		 424 va_end(args);
425
426 ext4_handle_error(inode->i_sb);
427}
428
c398eda0 429void ext4_error_file(struct file *file, const char *function,
f7c21177
TT		 430 unsigned int line, ext4_fsblk_t block,
431 const char *fmt, ...)
273df556
FM		 432{
433 va_list args;
f7c21177 434 struct va_format vaf;
1c13d5c0 435 struct ext4_super_block *es;
273df556
FM		 436 struct inode *inode = file->f_dentry->d_inode;
437 char pathname[80], *path;
438
1c13d5c0
TT		 439 es = EXT4_SB(inode->i_sb)->s_es;
440 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
441 save_error_info(inode->i_sb, function, line);
273df556 442 path = d_path(&(file->f_path), pathname, sizeof(pathname));
f9a62d09 443 if (IS_ERR(path))
273df556
FM		 444 path = "(unknown)";
445 printk(KERN_CRIT
f7c21177
TT		 446 "EXT4-fs error (device %s): %s:%d: inode #%lu: ",
447 inode->i_sb->s_id, function, line, inode->i_ino);
448 if (block)
449 printk(KERN_CONT "block %llu: ", block);
450 va_start(args, fmt);
451 vaf.fmt = fmt;
452 vaf.va = &args;
453 printk(KERN_CONT "comm %s: path %s: %pV\n", current->comm, path, &vaf);
273df556
FM		 454 va_end(args);
455
456 ext4_handle_error(inode->i_sb);
457}
458
2b2d6d01 459static const char *ext4_decode_error(struct super_block *sb, int errno,
ac27a0ec
DK		 460 char nbuf[16])
461{
462 char *errstr = NULL;
463
464 switch (errno) {
465 case -EIO:
466 errstr = "IO failure";
467 break;
468 case -ENOMEM:
469 errstr = "Out of memory";
470 break;
471 case -EROFS:
78f1ddbb
TT		 472 if (!sb || (EXT4_SB(sb)->s_journal &&
473 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
ac27a0ec
DK		 474 errstr = "Journal has aborted";
475 else
476 errstr = "Readonly filesystem";
477 break;
478 default:
479 /* If the caller passed in an extra buffer for unknown
480 * errors, textualise them now. Else we just return
481 * NULL. */
482 if (nbuf) {
483 /* Check for truncated error codes... */
484 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
485 errstr = nbuf;
486 }
487 break;
488 }
489
490 return errstr;
491}
492
617ba13b 493/* __ext4_std_error decodes expected errors from journaling functions
ac27a0ec
DK		 494 * automatically and invokes the appropriate error response. */
495
c398eda0
TT		 496void __ext4_std_error(struct super_block *sb, const char *function,
497 unsigned int line, int errno)
ac27a0ec
DK		 498{
499 char nbuf[16];
500 const char *errstr;
501
502 /* Special case: if the error is EROFS, and we're not already
503 * inside a transaction, then there's really no point in logging
504 * an error. */
505 if (errno == -EROFS && journal_current_handle() == NULL &&
506 (sb->s_flags & MS_RDONLY))
507 return;
508
617ba13b 509 errstr = ext4_decode_error(sb, errno, nbuf);
c398eda0
TT		 510 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
511 sb->s_id, function, line, errstr);
1c13d5c0 512 save_error_info(sb, function, line);
ac27a0ec 513
617ba13b 514 ext4_handle_error(sb);
ac27a0ec
DK		 515}
516
517/*
617ba13b 518 * ext4_abort is a much stronger failure handler than ext4_error. The
ac27a0ec
DK		 519 * abort function may be used to deal with unrecoverable failures such
520 * as journal IO errors or ENOMEM at a critical moment in log management.
521 *
522 * We unconditionally force the filesystem into an ABORT|READONLY state,
523 * unless the error response on the fs has been set to panic in which
524 * case we take the easy way out and panic immediately.
525 */
526
c67d859e 527void __ext4_abort(struct super_block *sb, const char *function,
c398eda0 528 unsigned int line, const char *fmt, ...)
ac27a0ec
DK		 529{
530 va_list args;
531
1c13d5c0 532 save_error_info(sb, function, line);
ac27a0ec 533 va_start(args, fmt);
c398eda0
TT		 534 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
535 function, line);
ac27a0ec
DK		 536 vprintk(fmt, args);
537 printk("\n");
538 va_end(args);
539
1c13d5c0
TT		 540 if ((sb->s_flags & MS_RDONLY) == 0) {
541 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
542 sb->s_flags |= MS_RDONLY;
543 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
544 if (EXT4_SB(sb)->s_journal)
545 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
546 save_error_info(sb, function, line);
547 }
ac27a0ec 548 if (test_opt(sb, ERRORS_PANIC))
617ba13b 549 panic("EXT4-fs panic from previous error\n");
ac27a0ec
DK		 550}
551
0ff2ea7d 552void ext4_msg(struct super_block *sb, const char *prefix, const char *fmt, ...)
b31e1552 553{
0ff2ea7d 554 struct va_format vaf;
b31e1552
ES		 555 va_list args;
556
557 va_start(args, fmt);
0ff2ea7d
JP		 558 vaf.fmt = fmt;
559 vaf.va = &args;
560 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
b31e1552
ES		 561 va_end(args);
562}
563
12062ddd 564void __ext4_warning(struct super_block *sb, const char *function,
c398eda0 565 unsigned int line, const char *fmt, ...)
ac27a0ec 566{
0ff2ea7d 567 struct va_format vaf;
ac27a0ec
DK		 568 va_list args;
569
570 va_start(args, fmt);
0ff2ea7d
JP		 571 vaf.fmt = fmt;
572 vaf.va = &args;
573 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
574 sb->s_id, function, line, &vaf);
ac27a0ec
DK		 575 va_end(args);
576}
577
e29136f8
TT		 578void __ext4_grp_locked_error(const char *function, unsigned int line,
579 struct super_block *sb, ext4_group_t grp,
580 unsigned long ino, ext4_fsblk_t block,
581 const char *fmt, ...)
5d1b1b3f
AK		 582__releases(bitlock)
583__acquires(bitlock)
584{
0ff2ea7d 585 struct va_format vaf;
5d1b1b3f
AK		 586 va_list args;
587 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
588
1c13d5c0
TT		 589 es->s_last_error_ino = cpu_to_le32(ino);
590 es->s_last_error_block = cpu_to_le64(block);
591 __save_error_info(sb, function, line);
0ff2ea7d 592
5d1b1b3f 593 va_start(args, fmt);
0ff2ea7d
JP		 594
595 vaf.fmt = fmt;
596 vaf.va = &args;
e29136f8
TT		 597 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u",
598 sb->s_id, function, line, grp);
599 if (ino)
0ff2ea7d 600 printk(KERN_CONT "inode %lu: ", ino);
e29136f8 601 if (block)
0ff2ea7d
JP		 602 printk(KERN_CONT "block %llu:", (unsigned long long) block);
603 printk(KERN_CONT "%pV\n", &vaf);
5d1b1b3f
AK		 604 va_end(args);
605
606 if (test_opt(sb, ERRORS_CONT)) {
e2d67052 607 ext4_commit_super(sb, 0);
5d1b1b3f
AK		 608 return;
609 }
1c13d5c0 610
5d1b1b3f
AK		 611 ext4_unlock_group(sb, grp);
612 ext4_handle_error(sb);
613 /*
614 * We only get here in the ERRORS_RO case; relocking the group
615 * may be dangerous, but nothing bad will happen since the
616 * filesystem will have already been marked read/only and the
617 * journal has been aborted. We return 1 as a hint to callers
618 * who might what to use the return value from
619 * ext4_grp_locked_error() to distinguish beween the
620 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
621 * aggressively from the ext4 function in question, with a
622 * more appropriate error code.
623 */
624 ext4_lock_group(sb, grp);
625 return;
626}
627
617ba13b 628void ext4_update_dynamic_rev(struct super_block *sb)
ac27a0ec 629{
617ba13b 630 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 631
617ba13b 632 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
ac27a0ec
DK		 633 return;
634
12062ddd 635 ext4_warning(sb,
ac27a0ec
DK		 636 "updating to rev %d because of new feature flag, "
637 "running e2fsck is recommended",
617ba13b 638 EXT4_DYNAMIC_REV);
ac27a0ec 639
617ba13b
MC		 640 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
641 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
642 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
ac27a0ec
DK		 643 /* leave es->s_feature_*compat flags alone */
644 /* es->s_uuid will be set by e2fsck if empty */
645
646 /*
647 * The rest of the superblock fields should be zero, and if not it
648 * means they are likely already in use, so leave them alone. We
649 * can leave it up to e2fsck to clean up any inconsistencies there.
650 */
651}
652
653/*
654 * Open the external journal device
655 */
b31e1552 656static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
ac27a0ec
DK		 657{
658 struct block_device *bdev;
659 char b[BDEVNAME_SIZE];
660
d4d77629 661 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
ac27a0ec
DK		 662 if (IS_ERR(bdev))
663 goto fail;
664 return bdev;
665
666fail:
b31e1552 667 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
ac27a0ec
DK		 668 __bdevname(dev, b), PTR_ERR(bdev));
669 return NULL;
670}
671
672/*
673 * Release the journal device
674 */
617ba13b 675static int ext4_blkdev_put(struct block_device *bdev)
ac27a0ec 676{
e525fd89 677 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
ac27a0ec
DK		 678}
679
617ba13b 680static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
ac27a0ec
DK		 681{
682 struct block_device *bdev;
683 int ret = -ENODEV;
684
685 bdev = sbi->journal_bdev;
686 if (bdev) {
617ba13b 687 ret = ext4_blkdev_put(bdev);
ac27a0ec
DK		 688 sbi->journal_bdev = NULL;
689 }
690 return ret;
691}
692
693static inline struct inode *orphan_list_entry(struct list_head *l)
694{
617ba13b 695 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
ac27a0ec
DK		 696}
697
617ba13b 698static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
ac27a0ec
DK		 699{
700 struct list_head *l;
701
b31e1552
ES		 702 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
703 le32_to_cpu(sbi->s_es->s_last_orphan));
ac27a0ec
DK		 704
705 printk(KERN_ERR "sb_info orphan list:\n");
706 list_for_each(l, &sbi->s_orphan) {
707 struct inode *inode = orphan_list_entry(l);
708 printk(KERN_ERR " "
709 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
710 inode->i_sb->s_id, inode->i_ino, inode,
711 inode->i_mode, inode->i_nlink,
712 NEXT_ORPHAN(inode));
713 }
714}
715
2b2d6d01 716static void ext4_put_super(struct super_block *sb)
ac27a0ec 717{
617ba13b
MC		 718 struct ext4_sb_info *sbi = EXT4_SB(sb);
719 struct ext4_super_block *es = sbi->s_es;
ef2cabf7 720 int i, err;
ac27a0ec 721
857ac889 722 ext4_unregister_li_request(sb);
e0ccfd95
CH		 723 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
724
4c0425ff
MC		 725 flush_workqueue(sbi->dio_unwritten_wq);
726 destroy_workqueue(sbi->dio_unwritten_wq);
727
a9e220f8 728 lock_super(sb);
8c85e125 729 if (sb->s_dirt)
ebc1ac16 730 ext4_commit_super(sb, 1);
8c85e125 731
0390131b
FM		 732 if (sbi->s_journal) {
733 err = jbd2_journal_destroy(sbi->s_journal);
734 sbi->s_journal = NULL;
735 if (err < 0)
c67d859e 736 ext4_abort(sb, "Couldn't clean up the journal");
0390131b 737 }
d4edac31 738
a1c6c569 739 del_timer(&sbi->s_err_report);
d4edac31
JB		 740 ext4_release_system_zone(sb);
741 ext4_mb_release(sb);
742 ext4_ext_release(sb);
743 ext4_xattr_put_super(sb);
744
ac27a0ec 745 if (!(sb->s_flags & MS_RDONLY)) {
617ba13b 746 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 747 es->s_state = cpu_to_le16(sbi->s_mount_state);
e2d67052 748 ext4_commit_super(sb, 1);
ac27a0ec 749 }
240799cd 750 if (sbi->s_proc) {
9f6200bb 751 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 752 }
3197ebdb 753 kobject_del(&sbi->s_kobj);
ac27a0ec
DK		 754
755 for (i = 0; i < sbi->s_gdb_count; i++)
756 brelse(sbi->s_group_desc[i]);
757 kfree(sbi->s_group_desc);
c5ca7c76
TT		 758 if (is_vmalloc_addr(sbi->s_flex_groups))
759 vfree(sbi->s_flex_groups);
760 else
761 kfree(sbi->s_flex_groups);
ac27a0ec
DK		 762 percpu_counter_destroy(&sbi->s_freeblocks_counter);
763 percpu_counter_destroy(&sbi->s_freeinodes_counter);
764 percpu_counter_destroy(&sbi->s_dirs_counter);
6bc6e63f 765 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
ac27a0ec
DK		 766 brelse(sbi->s_sbh);
767#ifdef CONFIG_QUOTA
768 for (i = 0; i < MAXQUOTAS; i++)
769 kfree(sbi->s_qf_names[i]);
770#endif
771
772 /* Debugging code just in case the in-memory inode orphan list
773 * isn't empty. The on-disk one can be non-empty if we've
774 * detected an error and taken the fs readonly, but the
775 * in-memory list had better be clean by this point. */
776 if (!list_empty(&sbi->s_orphan))
777 dump_orphan_list(sb, sbi);
778 J_ASSERT(list_empty(&sbi->s_orphan));
779
f98393a6 780 invalidate_bdev(sb->s_bdev);
ac27a0ec
DK		 781 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
782 /*
783 * Invalidate the journal device's buffers. We don't want them
784 * floating about in memory - the physical journal device may
785 * hotswapped, and it breaks the `ro-after' testing code.
786 */
787 sync_blockdev(sbi->journal_bdev);
f98393a6 788 invalidate_bdev(sbi->journal_bdev);
617ba13b 789 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 790 }
791 sb->s_fs_info = NULL;
3197ebdb
TT		 792 /*
793 * Now that we are completely done shutting down the
794 * superblock, we need to actually destroy the kobject.
795 */
3197ebdb
TT		 796 unlock_super(sb);
797 kobject_put(&sbi->s_kobj);
798 wait_for_completion(&sbi->s_kobj_unregister);
705895b6 799 kfree(sbi->s_blockgroup_lock);
ac27a0ec 800 kfree(sbi);
ac27a0ec
DK		 801}
802
e18b890b 803static struct kmem_cache *ext4_inode_cachep;
ac27a0ec
DK		 804
805/*
806 * Called inside transaction, so use GFP_NOFS
807 */
617ba13b 808static struct inode *ext4_alloc_inode(struct super_block *sb)
ac27a0ec 809{
617ba13b 810 struct ext4_inode_info *ei;
ac27a0ec 811
e6b4f8da 812 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
ac27a0ec
DK		 813 if (!ei)
814 return NULL;
0b8e58a1 815
ac27a0ec 816 ei->vfs_inode.i_version = 1;
91246c00 817 ei->vfs_inode.i_data.writeback_index = 0;
a86c6181 818 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
c9de560d
AT		 819 INIT_LIST_HEAD(&ei->i_prealloc_list);
820 spin_lock_init(&ei->i_prealloc_lock);
d2a17637
MC		 821 ei->i_reserved_data_blocks = 0;
822 ei->i_reserved_meta_blocks = 0;
823 ei->i_allocated_meta_blocks = 0;
9d0be502 824 ei->i_da_metadata_calc_len = 0;
d2a17637 825 spin_lock_init(&(ei->i_block_reservation_lock));
a9e7f447
DM		 826#ifdef CONFIG_QUOTA
827 ei->i_reserved_quota = 0;
828#endif
8aefcd55 829 ei->jinode = NULL;
c7064ef1 830 INIT_LIST_HEAD(&ei->i_completed_io_list);
744692dc 831 spin_lock_init(&ei->i_completed_io_lock);
8d5d02e6 832 ei->cur_aio_dio = NULL;
b436b9be
JK		 833 ei->i_sync_tid = 0;
834 ei->i_datasync_tid = 0;
f7ad6d2e 835 atomic_set(&ei->i_ioend_count, 0);
e9e3bcec 836 atomic_set(&ei->i_aiodio_unwritten, 0);
0b8e58a1 837
ac27a0ec
DK		 838 return &ei->vfs_inode;
839}
840
7ff9c073
TT		 841static int ext4_drop_inode(struct inode *inode)
842{
843 int drop = generic_drop_inode(inode);
844
845 trace_ext4_drop_inode(inode, drop);
846 return drop;
847}
848
fa0d7e3d
NP		 849static void ext4_i_callback(struct rcu_head *head)
850{
851 struct inode *inode = container_of(head, struct inode, i_rcu);
852 INIT_LIST_HEAD(&inode->i_dentry);
853 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
854}
855
617ba13b 856static void ext4_destroy_inode(struct inode *inode)
ac27a0ec 857{
f7ad6d2e 858 ext4_ioend_wait(inode);
9f7dd93d 859 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
b31e1552
ES		 860 ext4_msg(inode->i_sb, KERN_ERR,
861 "Inode %lu (%p): orphan list check failed!",
862 inode->i_ino, EXT4_I(inode));
9f7dd93d
VA		 863 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
864 EXT4_I(inode), sizeof(struct ext4_inode_info),
865 true);
866 dump_stack();
867 }
fa0d7e3d 868 call_rcu(&inode->i_rcu, ext4_i_callback);
ac27a0ec
DK		 869}
870
51cc5068 871static void init_once(void *foo)
ac27a0ec 872{
617ba13b 873 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
ac27a0ec 874
a35afb83 875 INIT_LIST_HEAD(&ei->i_orphan);
03010a33 876#ifdef CONFIG_EXT4_FS_XATTR
a35afb83 877 init_rwsem(&ei->xattr_sem);
ac27a0ec 878#endif
0e855ac8 879 init_rwsem(&ei->i_data_sem);
a35afb83 880 inode_init_once(&ei->vfs_inode);
ac27a0ec
DK		 881}
882
883static int init_inodecache(void)
884{
617ba13b
MC		 885 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
886 sizeof(struct ext4_inode_info),
ac27a0ec
DK		 887 0, (SLAB_RECLAIM_ACCOUNT|
888 SLAB_MEM_SPREAD),
20c2df83 889 init_once);
617ba13b 890 if (ext4_inode_cachep == NULL)
ac27a0ec
DK		 891 return -ENOMEM;
892 return 0;
893}
894
895static void destroy_inodecache(void)
896{
617ba13b 897 kmem_cache_destroy(ext4_inode_cachep);
ac27a0ec
DK		 898}
899
0930fcc1 900void ext4_clear_inode(struct inode *inode)
ac27a0ec 901{
0930fcc1
AV		 902 invalidate_inode_buffers(inode);
903 end_writeback(inode);
9f754758 904 dquot_drop(inode);
c2ea3fde 905 ext4_discard_preallocations(inode);
8aefcd55
TT		 906 if (EXT4_I(inode)->jinode) {
907 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
908 EXT4_I(inode)->jinode);
909 jbd2_free_inode(EXT4_I(inode)->jinode);
910 EXT4_I(inode)->jinode = NULL;
911 }
ac27a0ec
DK		 912}
913
2b2d6d01
TT		 914static inline void ext4_show_quota_options(struct seq_file *seq,
915 struct super_block *sb)
ac27a0ec
DK		 916{
917#if defined(CONFIG_QUOTA)
617ba13b 918 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 919
5a20bdfc
JK		 920 if (sbi->s_jquota_fmt) {
921 char *fmtname = "";
922
923 switch (sbi->s_jquota_fmt) {
924 case QFMT_VFS_OLD:
925 fmtname = "vfsold";
926 break;
927 case QFMT_VFS_V0:
928 fmtname = "vfsv0";
929 break;
930 case QFMT_VFS_V1:
931 fmtname = "vfsv1";
932 break;
933 }
934 seq_printf(seq, ",jqfmt=%s", fmtname);
935 }
ac27a0ec
DK		 936
937 if (sbi->s_qf_names[USRQUOTA])
938 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
939
940 if (sbi->s_qf_names[GRPQUOTA])
941 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
942
482a7425 943 if (test_opt(sb, USRQUOTA))
ac27a0ec
DK		 944 seq_puts(seq, ",usrquota");
945
482a7425 946 if (test_opt(sb, GRPQUOTA))
ac27a0ec
DK		 947 seq_puts(seq, ",grpquota");
948#endif
949}
950
d9c9bef1
MS		 951/*
952 * Show an option if
953 * - it's set to a non-default value OR
954 * - if the per-sb default is different from the global default
955 */
617ba13b 956static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
ac27a0ec 957{
aa22df2c
AK		 958 int def_errors;
959 unsigned long def_mount_opts;
ac27a0ec 960 struct super_block *sb = vfs->mnt_sb;
d9c9bef1
MS		 961 struct ext4_sb_info *sbi = EXT4_SB(sb);
962 struct ext4_super_block *es = sbi->s_es;
d9c9bef1
MS		 963
964 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
aa22df2c 965 def_errors = le16_to_cpu(es->s_errors);
d9c9bef1
MS		 966
967 if (sbi->s_sb_block != 1)
968 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
969 if (test_opt(sb, MINIX_DF))
970 seq_puts(seq, ",minixdf");
aa22df2c 971 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
d9c9bef1
MS		 972 seq_puts(seq, ",grpid");
973 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
974 seq_puts(seq, ",nogrpid");
975 if (sbi->s_resuid != EXT4_DEF_RESUID ||
976 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
977 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
978 }
979 if (sbi->s_resgid != EXT4_DEF_RESGID ||
980 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
981 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
982 }
bb4f397a 983 if (test_opt(sb, ERRORS_RO)) {
d9c9bef1 984 if (def_errors == EXT4_ERRORS_PANIC ||
bb4f397a
AK		 985 def_errors == EXT4_ERRORS_CONTINUE) {
986 seq_puts(seq, ",errors=remount-ro");
d9c9bef1
MS		 987 }
988 }
aa22df2c 989 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
bb4f397a 990 seq_puts(seq, ",errors=continue");
aa22df2c 991 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
d9c9bef1 992 seq_puts(seq, ",errors=panic");
aa22df2c 993 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
d9c9bef1 994 seq_puts(seq, ",nouid32");
aa22df2c 995 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
d9c9bef1
MS		 996 seq_puts(seq, ",debug");
997 if (test_opt(sb, OLDALLOC))
998 seq_puts(seq, ",oldalloc");
03010a33 999#ifdef CONFIG_EXT4_FS_XATTR
aa22df2c
AK		 1000 if (test_opt(sb, XATTR_USER) &&
1001 !(def_mount_opts & EXT4_DEFM_XATTR_USER))
d9c9bef1
MS		 1002 seq_puts(seq, ",user_xattr");
1003 if (!test_opt(sb, XATTR_USER) &&
1004 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
1005 seq_puts(seq, ",nouser_xattr");
1006 }
1007#endif
03010a33 1008#ifdef CONFIG_EXT4_FS_POSIX_ACL
aa22df2c 1009 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
d9c9bef1
MS		 1010 seq_puts(seq, ",acl");
1011 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
1012 seq_puts(seq, ",noacl");
1013#endif
30773840 1014 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
d9c9bef1
MS		 1015 seq_printf(seq, ",commit=%u",
1016 (unsigned) (sbi->s_commit_interval / HZ));
1017 }
30773840
TT		 1018 if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
1019 seq_printf(seq, ",min_batch_time=%u",
1020 (unsigned) sbi->s_min_batch_time);
1021 }
1022 if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
1023 seq_printf(seq, ",max_batch_time=%u",
1024 (unsigned) sbi->s_min_batch_time);
1025 }
1026
571640ca
ES		 1027 /*
1028 * We're changing the default of barrier mount option, so
1029 * let's always display its mount state so it's clear what its
1030 * status is.
1031 */
1032 seq_puts(seq, ",barrier=");
1033 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
cd0b6a39
TT		 1034 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
1035 seq_puts(seq, ",journal_async_commit");
39a4bade
JK		 1036 else if (test_opt(sb, JOURNAL_CHECKSUM))
1037 seq_puts(seq, ",journal_checksum");
25ec56b5
JNC		 1038 if (test_opt(sb, I_VERSION))
1039 seq_puts(seq, ",i_version");
8b67f04a
TT		 1040 if (!test_opt(sb, DELALLOC) &&
1041 !(def_mount_opts & EXT4_DEFM_NODELALLOC))
dd919b98
AK		 1042 seq_puts(seq, ",nodelalloc");
1043
1449032b
TT		 1044 if (test_opt(sb, MBLK_IO_SUBMIT))
1045 seq_puts(seq, ",mblk_io_submit");
cb45bbe4
MS		 1046 if (sbi->s_stripe)
1047 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
aa22df2c
AK		 1048 /*
1049 * journal mode get enabled in different ways
1050 * So just print the value even if we didn't specify it
1051 */
617ba13b 1052 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
ac27a0ec 1053 seq_puts(seq, ",data=journal");
617ba13b 1054 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
ac27a0ec 1055 seq_puts(seq, ",data=ordered");
617ba13b 1056 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
ac27a0ec
DK		 1057 seq_puts(seq, ",data=writeback");
1058
240799cd
TT		 1059 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1060 seq_printf(seq, ",inode_readahead_blks=%u",
1061 sbi->s_inode_readahead_blks);
1062
5bf5683a
HK		 1063 if (test_opt(sb, DATA_ERR_ABORT))
1064 seq_puts(seq, ",data_err=abort");
1065
afd4672d 1066 if (test_opt(sb, NO_AUTO_DA_ALLOC))
06705bff 1067 seq_puts(seq, ",noauto_da_alloc");
afd4672d 1068
8b67f04a 1069 if (test_opt(sb, DISCARD) && !(def_mount_opts & EXT4_DEFM_DISCARD))
5328e635
ES		 1070 seq_puts(seq, ",discard");
1071
e3bb52ae
ES		 1072 if (test_opt(sb, NOLOAD))
1073 seq_puts(seq, ",norecovery");
1074
744692dc
JZ		 1075 if (test_opt(sb, DIOREAD_NOLOCK))
1076 seq_puts(seq, ",dioread_nolock");
1077
8b67f04a
TT		 1078 if (test_opt(sb, BLOCK_VALIDITY) &&
1079 !(def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY))
1080 seq_puts(seq, ",block_validity");
1081
bfff6873
LC		 1082 if (!test_opt(sb, INIT_INODE_TABLE))
1083 seq_puts(seq, ",noinit_inode_table");
1084 else if (sbi->s_li_wait_mult)
1085 seq_printf(seq, ",init_inode_table=%u",
1086 (unsigned) sbi->s_li_wait_mult);
1087
617ba13b 1088 ext4_show_quota_options(seq, sb);
0b8e58a1 1089
ac27a0ec
DK		 1090 return 0;
1091}
1092
1b961ac0 1093static struct inode *ext4_nfs_get_inode(struct super_block *sb,
0b8e58a1 1094 u64 ino, u32 generation)
ac27a0ec 1095{
ac27a0ec 1096 struct inode *inode;
ac27a0ec 1097
617ba13b 1098 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
ac27a0ec 1099 return ERR_PTR(-ESTALE);
617ba13b 1100 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
ac27a0ec
DK		 1101 return ERR_PTR(-ESTALE);
1102
1103 /* iget isn't really right if the inode is currently unallocated!!
1104 *
617ba13b 1105 * ext4_read_inode will return a bad_inode if the inode had been
ac27a0ec
DK		 1106 * deleted, so we should be safe.
1107 *
1108 * Currently we don't know the generation for parent directory, so
1109 * a generation of 0 means "accept any"
1110 */
1d1fe1ee
DH		 1111 inode = ext4_iget(sb, ino);
1112 if (IS_ERR(inode))
1113 return ERR_CAST(inode);
1114 if (generation && inode->i_generation != generation) {
ac27a0ec
DK		 1115 iput(inode);
1116 return ERR_PTR(-ESTALE);
1117 }
1b961ac0
CH		 1118
1119 return inode;
1120}
1121
1122static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
0b8e58a1 1123 int fh_len, int fh_type)
1b961ac0
CH		 1124{
1125 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1126 ext4_nfs_get_inode);
1127}
1128
1129static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
0b8e58a1 1130 int fh_len, int fh_type)
1b961ac0
CH		 1131{
1132 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1133 ext4_nfs_get_inode);
ac27a0ec
DK		 1134}
1135
c39a7f84
TO		 1136/*
1137 * Try to release metadata pages (indirect blocks, directories) which are
1138 * mapped via the block device. Since these pages could have journal heads
1139 * which would prevent try_to_free_buffers() from freeing them, we must use
1140 * jbd2 layer's try_to_free_buffers() function to release them.
1141 */
0b8e58a1
AD		 1142static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1143 gfp_t wait)
c39a7f84
TO		 1144{
1145 journal_t *journal = EXT4_SB(sb)->s_journal;
1146
1147 WARN_ON(PageChecked(page));
1148 if (!page_has_buffers(page))
1149 return 0;
1150 if (journal)
1151 return jbd2_journal_try_to_free_buffers(journal, page,
1152 wait & ~__GFP_WAIT);
1153 return try_to_free_buffers(page);
1154}
1155
ac27a0ec 1156#ifdef CONFIG_QUOTA
af5bc92d 1157#define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
2b2d6d01 1158#define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
ac27a0ec 1159
617ba13b
MC		 1160static int ext4_write_dquot(struct dquot *dquot);
1161static int ext4_acquire_dquot(struct dquot *dquot);
1162static int ext4_release_dquot(struct dquot *dquot);
1163static int ext4_mark_dquot_dirty(struct dquot *dquot);
1164static int ext4_write_info(struct super_block *sb, int type);
6f28e087 1165static int ext4_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 1166 struct path *path);
ca0e05e4 1167static int ext4_quota_off(struct super_block *sb, int type);
617ba13b
MC		 1168static int ext4_quota_on_mount(struct super_block *sb, int type);
1169static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 1170 size_t len, loff_t off);
617ba13b 1171static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 1172 const char *data, size_t len, loff_t off);
1173
61e225dc 1174static const struct dquot_operations ext4_quota_operations = {
a9e7f447 1175#ifdef CONFIG_QUOTA
60e58e0f 1176 .get_reserved_space = ext4_get_reserved_space,
a9e7f447 1177#endif
617ba13b
MC		 1178 .write_dquot = ext4_write_dquot,
1179 .acquire_dquot = ext4_acquire_dquot,
1180 .release_dquot = ext4_release_dquot,
1181 .mark_dirty = ext4_mark_dquot_dirty,
a5b5ee32
JK		 1182 .write_info = ext4_write_info,
1183 .alloc_dquot = dquot_alloc,
1184 .destroy_dquot = dquot_destroy,
ac27a0ec
DK		 1185};
1186
0d54b217 1187static const struct quotactl_ops ext4_qctl_operations = {
617ba13b 1188 .quota_on = ext4_quota_on,
ca0e05e4 1189 .quota_off = ext4_quota_off,
287a8095
CH		 1190 .quota_sync = dquot_quota_sync,
1191 .get_info = dquot_get_dqinfo,
1192 .set_info = dquot_set_dqinfo,
1193 .get_dqblk = dquot_get_dqblk,
1194 .set_dqblk = dquot_set_dqblk
ac27a0ec
DK		 1195};
1196#endif
1197
ee9b6d61 1198static const struct super_operations ext4_sops = {
617ba13b
MC		 1199 .alloc_inode = ext4_alloc_inode,
1200 .destroy_inode = ext4_destroy_inode,
617ba13b
MC		 1201 .write_inode = ext4_write_inode,
1202 .dirty_inode = ext4_dirty_inode,
7ff9c073 1203 .drop_inode = ext4_drop_inode,
0930fcc1 1204 .evict_inode = ext4_evict_inode,
617ba13b 1205 .put_super = ext4_put_super,
617ba13b 1206 .sync_fs = ext4_sync_fs,
c4be0c1d
TS		 1207 .freeze_fs = ext4_freeze,
1208 .unfreeze_fs = ext4_unfreeze,
617ba13b
MC		 1209 .statfs = ext4_statfs,
1210 .remount_fs = ext4_remount,
617ba13b 1211 .show_options = ext4_show_options,
ac27a0ec 1212#ifdef CONFIG_QUOTA
617ba13b
MC		 1213 .quota_read = ext4_quota_read,
1214 .quota_write = ext4_quota_write,
ac27a0ec 1215#endif
c39a7f84 1216 .bdev_try_to_free_page = bdev_try_to_free_page,
ac27a0ec
DK		 1217};
1218
9ca92389
TT		 1219static const struct super_operations ext4_nojournal_sops = {
1220 .alloc_inode = ext4_alloc_inode,
1221 .destroy_inode = ext4_destroy_inode,
1222 .write_inode = ext4_write_inode,
1223 .dirty_inode = ext4_dirty_inode,
7ff9c073 1224 .drop_inode = ext4_drop_inode,
0930fcc1 1225 .evict_inode = ext4_evict_inode,
9ca92389
TT		 1226 .write_super = ext4_write_super,
1227 .put_super = ext4_put_super,
1228 .statfs = ext4_statfs,
1229 .remount_fs = ext4_remount,
9ca92389
TT		 1230 .show_options = ext4_show_options,
1231#ifdef CONFIG_QUOTA
1232 .quota_read = ext4_quota_read,
1233 .quota_write = ext4_quota_write,
1234#endif
1235 .bdev_try_to_free_page = bdev_try_to_free_page,
1236};
1237
39655164 1238static const struct export_operations ext4_export_ops = {
1b961ac0
CH		 1239 .fh_to_dentry = ext4_fh_to_dentry,
1240 .fh_to_parent = ext4_fh_to_parent,
617ba13b 1241 .get_parent = ext4_get_parent,
ac27a0ec
DK		 1242};
1243
1244enum {
1245 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1246 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
01436ef2 1247 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
ac27a0ec 1248 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
06705bff 1249 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh,
30773840 1250 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
c3191067 1251 Opt_journal_update, Opt_journal_dev,
818d276c 1252 Opt_journal_checksum, Opt_journal_async_commit,
ac27a0ec 1253 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
296c355c 1254 Opt_data_err_abort, Opt_data_err_ignore,
ac27a0ec 1255 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
5a20bdfc
JK		 1256 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1257 Opt_noquota, Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err,
1258 Opt_resize, Opt_usrquota, Opt_grpquota, Opt_i_version,
1449032b
TT		 1259 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1260 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
5328e635 1261 Opt_inode_readahead_blks, Opt_journal_ioprio,
744692dc 1262 Opt_dioread_nolock, Opt_dioread_lock,
5328e635 1263 Opt_discard, Opt_nodiscard,
bfff6873 1264 Opt_init_inode_table, Opt_noinit_inode_table,
ac27a0ec
DK		 1265};
1266
a447c093 1267static const match_table_t tokens = {
ac27a0ec
DK		 1268 {Opt_bsd_df, "bsddf"},
1269 {Opt_minix_df, "minixdf"},
1270 {Opt_grpid, "grpid"},
1271 {Opt_grpid, "bsdgroups"},
1272 {Opt_nogrpid, "nogrpid"},
1273 {Opt_nogrpid, "sysvgroups"},
1274 {Opt_resgid, "resgid=%u"},
1275 {Opt_resuid, "resuid=%u"},
1276 {Opt_sb, "sb=%u"},
1277 {Opt_err_cont, "errors=continue"},
1278 {Opt_err_panic, "errors=panic"},
1279 {Opt_err_ro, "errors=remount-ro"},
1280 {Opt_nouid32, "nouid32"},
ac27a0ec
DK		 1281 {Opt_debug, "debug"},
1282 {Opt_oldalloc, "oldalloc"},
1283 {Opt_orlov, "orlov"},
1284 {Opt_user_xattr, "user_xattr"},
1285 {Opt_nouser_xattr, "nouser_xattr"},
1286 {Opt_acl, "acl"},
1287 {Opt_noacl, "noacl"},
ac27a0ec 1288 {Opt_noload, "noload"},
e3bb52ae 1289 {Opt_noload, "norecovery"},
ac27a0ec
DK		 1290 {Opt_nobh, "nobh"},
1291 {Opt_bh, "bh"},
1292 {Opt_commit, "commit=%u"},
30773840
TT		 1293 {Opt_min_batch_time, "min_batch_time=%u"},
1294 {Opt_max_batch_time, "max_batch_time=%u"},
ac27a0ec 1295 {Opt_journal_update, "journal=update"},
ac27a0ec 1296 {Opt_journal_dev, "journal_dev=%u"},
818d276c
GS		 1297 {Opt_journal_checksum, "journal_checksum"},
1298 {Opt_journal_async_commit, "journal_async_commit"},
ac27a0ec
DK		 1299 {Opt_abort, "abort"},
1300 {Opt_data_journal, "data=journal"},
1301 {Opt_data_ordered, "data=ordered"},
1302 {Opt_data_writeback, "data=writeback"},
5bf5683a
HK		 1303 {Opt_data_err_abort, "data_err=abort"},
1304 {Opt_data_err_ignore, "data_err=ignore"},
ac27a0ec
DK		 1305 {Opt_offusrjquota, "usrjquota="},
1306 {Opt_usrjquota, "usrjquota=%s"},
1307 {Opt_offgrpjquota, "grpjquota="},
1308 {Opt_grpjquota, "grpjquota=%s"},
1309 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1310 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
5a20bdfc 1311 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
ac27a0ec
DK		 1312 {Opt_grpquota, "grpquota"},
1313 {Opt_noquota, "noquota"},
1314 {Opt_quota, "quota"},
1315 {Opt_usrquota, "usrquota"},
1316 {Opt_barrier, "barrier=%u"},
06705bff
TT		 1317 {Opt_barrier, "barrier"},
1318 {Opt_nobarrier, "nobarrier"},
25ec56b5 1319 {Opt_i_version, "i_version"},
c9de560d 1320 {Opt_stripe, "stripe=%u"},
ac27a0ec 1321 {Opt_resize, "resize"},
64769240 1322 {Opt_delalloc, "delalloc"},
dd919b98 1323 {Opt_nodelalloc, "nodelalloc"},
1449032b
TT		 1324 {Opt_mblk_io_submit, "mblk_io_submit"},
1325 {Opt_nomblk_io_submit, "nomblk_io_submit"},
6fd058f7
TT		 1326 {Opt_block_validity, "block_validity"},
1327 {Opt_noblock_validity, "noblock_validity"},
240799cd 1328 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
b3881f74 1329 {Opt_journal_ioprio, "journal_ioprio=%u"},
afd4672d 1330 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
06705bff
TT		 1331 {Opt_auto_da_alloc, "auto_da_alloc"},
1332 {Opt_noauto_da_alloc, "noauto_da_alloc"},
744692dc
JZ		 1333 {Opt_dioread_nolock, "dioread_nolock"},
1334 {Opt_dioread_lock, "dioread_lock"},
5328e635
ES		 1335 {Opt_discard, "discard"},
1336 {Opt_nodiscard, "nodiscard"},
bfff6873
LC		 1337 {Opt_init_inode_table, "init_itable=%u"},
1338 {Opt_init_inode_table, "init_itable"},
1339 {Opt_noinit_inode_table, "noinit_itable"},
f3f12faa 1340 {Opt_err, NULL},
ac27a0ec
DK		 1341};
1342
617ba13b 1343static ext4_fsblk_t get_sb_block(void **data)
ac27a0ec 1344{
617ba13b 1345 ext4_fsblk_t sb_block;
ac27a0ec
DK		 1346 char *options = (char *) *data;
1347
1348 if (!options || strncmp(options, "sb=", 3) != 0)
1349 return 1; /* Default location */
0b8e58a1 1350
ac27a0ec 1351 options += 3;
0b8e58a1 1352 /* TODO: use simple_strtoll with >32bit ext4 */
ac27a0ec
DK		 1353 sb_block = simple_strtoul(options, &options, 0);
1354 if (*options && *options != ',') {
4776004f 1355 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
ac27a0ec
DK		 1356 (char *) *data);
1357 return 1;
1358 }
1359 if (*options == ',')
1360 options++;
1361 *data = (void *) options;
0b8e58a1 1362
ac27a0ec
DK		 1363 return sb_block;
1364}
1365
b3881f74 1366#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
437ca0fd
DM		 1367static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1368 "Contact [email protected] if you think we should keep it.\n";
b3881f74 1369
56c50f11
DM		 1370#ifdef CONFIG_QUOTA
1371static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1372{
1373 struct ext4_sb_info *sbi = EXT4_SB(sb);
1374 char *qname;
1375
1376 if (sb_any_quota_loaded(sb) &&
1377 !sbi->s_qf_names[qtype]) {
1378 ext4_msg(sb, KERN_ERR,
1379 "Cannot change journaled "
1380 "quota options when quota turned on");
1381 return 0;
1382 }
1383 qname = match_strdup(args);
1384 if (!qname) {
1385 ext4_msg(sb, KERN_ERR,
1386 "Not enough memory for storing quotafile name");
1387 return 0;
1388 }
1389 if (sbi->s_qf_names[qtype] &&
1390 strcmp(sbi->s_qf_names[qtype], qname)) {
1391 ext4_msg(sb, KERN_ERR,
1392 "%s quota file already specified", QTYPE2NAME(qtype));
1393 kfree(qname);
1394 return 0;
1395 }
1396 sbi->s_qf_names[qtype] = qname;
1397 if (strchr(sbi->s_qf_names[qtype], '/')) {
1398 ext4_msg(sb, KERN_ERR,
1399 "quotafile must be on filesystem root");
1400 kfree(sbi->s_qf_names[qtype]);
1401 sbi->s_qf_names[qtype] = NULL;
1402 return 0;
1403 }
fd8c37ec 1404 set_opt(sb, QUOTA);
56c50f11
DM		 1405 return 1;
1406}
1407
1408static int clear_qf_name(struct super_block *sb, int qtype)
1409{
1410
1411 struct ext4_sb_info *sbi = EXT4_SB(sb);
1412
1413 if (sb_any_quota_loaded(sb) &&
1414 sbi->s_qf_names[qtype]) {
1415 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1416 " when quota turned on");
1417 return 0;
1418 }
1419 /*
1420 * The space will be released later when all options are confirmed
1421 * to be correct
1422 */
1423 sbi->s_qf_names[qtype] = NULL;
1424 return 1;
1425}
1426#endif
1427
2b2d6d01 1428static int parse_options(char *options, struct super_block *sb,
c3191067 1429 unsigned long *journal_devnum,
b3881f74 1430 unsigned int *journal_ioprio,
2b2d6d01 1431 ext4_fsblk_t *n_blocks_count, int is_remount)
ac27a0ec 1432{
617ba13b 1433 struct ext4_sb_info *sbi = EXT4_SB(sb);
2b2d6d01 1434 char *p;
ac27a0ec
DK		 1435 substring_t args[MAX_OPT_ARGS];
1436 int data_opt = 0;
1437 int option;
1438#ifdef CONFIG_QUOTA
56c50f11 1439 int qfmt;
ac27a0ec
DK		 1440#endif
1441
1442 if (!options)
1443 return 1;
1444
2b2d6d01 1445 while ((p = strsep(&options, ",")) != NULL) {
ac27a0ec
DK		 1446 int token;
1447 if (!*p)
1448 continue;
1449
15121c18
ES		 1450 /*
1451 * Initialize args struct so we know whether arg was
1452 * found; some options take optional arguments.
1453 */
1454 args[0].to = args[0].from = 0;
ac27a0ec
DK		 1455 token = match_token(p, tokens, args);
1456 switch (token) {
1457 case Opt_bsd_df:
437ca0fd 1458 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
fd8c37ec 1459 clear_opt(sb, MINIX_DF);
ac27a0ec
DK		 1460 break;
1461 case Opt_minix_df:
437ca0fd 1462 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
fd8c37ec 1463 set_opt(sb, MINIX_DF);
437ca0fd 1464
ac27a0ec
DK		 1465 break;
1466 case Opt_grpid:
437ca0fd 1467 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
fd8c37ec 1468 set_opt(sb, GRPID);
437ca0fd 1469
ac27a0ec
DK		 1470 break;
1471 case Opt_nogrpid:
437ca0fd 1472 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
fd8c37ec 1473 clear_opt(sb, GRPID);
437ca0fd 1474
ac27a0ec
DK		 1475 break;
1476 case Opt_resuid:
1477 if (match_int(&args[0], &option))
1478 return 0;
1479 sbi->s_resuid = option;
1480 break;
1481 case Opt_resgid:
1482 if (match_int(&args[0], &option))
1483 return 0;
1484 sbi->s_resgid = option;
1485 break;
1486 case Opt_sb:
1487 /* handled by get_sb_block() instead of here */
1488 /* *sb_block = match_int(&args[0]); */
1489 break;
1490 case Opt_err_panic:
fd8c37ec
TT		 1491 clear_opt(sb, ERRORS_CONT);
1492 clear_opt(sb, ERRORS_RO);
1493 set_opt(sb, ERRORS_PANIC);
ac27a0ec
DK		 1494 break;
1495 case Opt_err_ro:
fd8c37ec
TT		 1496 clear_opt(sb, ERRORS_CONT);
1497 clear_opt(sb, ERRORS_PANIC);
1498 set_opt(sb, ERRORS_RO);
ac27a0ec
DK		 1499 break;
1500 case Opt_err_cont:
fd8c37ec
TT		 1501 clear_opt(sb, ERRORS_RO);
1502 clear_opt(sb, ERRORS_PANIC);
1503 set_opt(sb, ERRORS_CONT);
ac27a0ec
DK		 1504 break;
1505 case Opt_nouid32:
fd8c37ec 1506 set_opt(sb, NO_UID32);
ac27a0ec 1507 break;
ac27a0ec 1508 case Opt_debug:
fd8c37ec 1509 set_opt(sb, DEBUG);
ac27a0ec
DK		 1510 break;
1511 case Opt_oldalloc:
fd8c37ec 1512 set_opt(sb, OLDALLOC);
ac27a0ec
DK		 1513 break;
1514 case Opt_orlov:
fd8c37ec 1515 clear_opt(sb, OLDALLOC);
ac27a0ec 1516 break;
03010a33 1517#ifdef CONFIG_EXT4_FS_XATTR
ac27a0ec 1518 case Opt_user_xattr:
fd8c37ec 1519 set_opt(sb, XATTR_USER);
ac27a0ec
DK		 1520 break;
1521 case Opt_nouser_xattr:
fd8c37ec 1522 clear_opt(sb, XATTR_USER);
ac27a0ec
DK		 1523 break;
1524#else
1525 case Opt_user_xattr:
1526 case Opt_nouser_xattr:
b31e1552 1527 ext4_msg(sb, KERN_ERR, "(no)user_xattr options not supported");
ac27a0ec
DK		 1528 break;
1529#endif
03010a33 1530#ifdef CONFIG_EXT4_FS_POSIX_ACL
ac27a0ec 1531 case Opt_acl:
fd8c37ec 1532 set_opt(sb, POSIX_ACL);
ac27a0ec
DK		 1533 break;
1534 case Opt_noacl:
fd8c37ec 1535 clear_opt(sb, POSIX_ACL);
ac27a0ec
DK		 1536 break;
1537#else
1538 case Opt_acl:
1539 case Opt_noacl:
b31e1552 1540 ext4_msg(sb, KERN_ERR, "(no)acl options not supported");
ac27a0ec
DK		 1541 break;
1542#endif
ac27a0ec
DK		 1543 case Opt_journal_update:
1544 /* @@@ FIXME */
1545 /* Eventually we will want to be able to create
1546 a journal file here. For now, only allow the
1547 user to specify an existing inode to be the
1548 journal file. */
1549 if (is_remount) {
b31e1552
ES		 1550 ext4_msg(sb, KERN_ERR,
1551 "Cannot specify journal on remount");
ac27a0ec
DK		 1552 return 0;
1553 }
fd8c37ec 1554 set_opt(sb, UPDATE_JOURNAL);
ac27a0ec 1555 break;
ac27a0ec
DK		 1556 case Opt_journal_dev:
1557 if (is_remount) {
b31e1552
ES		 1558 ext4_msg(sb, KERN_ERR,
1559 "Cannot specify journal on remount");
ac27a0ec
DK		 1560 return 0;
1561 }
1562 if (match_int(&args[0], &option))
1563 return 0;
1564 *journal_devnum = option;
1565 break;
818d276c 1566 case Opt_journal_checksum:
fd8c37ec 1567 set_opt(sb, JOURNAL_CHECKSUM);
d4da6c9c 1568 break;
818d276c 1569 case Opt_journal_async_commit:
fd8c37ec
TT		 1570 set_opt(sb, JOURNAL_ASYNC_COMMIT);
1571 set_opt(sb, JOURNAL_CHECKSUM);
818d276c 1572 break;
ac27a0ec 1573 case Opt_noload:
fd8c37ec 1574 set_opt(sb, NOLOAD);
ac27a0ec
DK		 1575 break;
1576 case Opt_commit:
1577 if (match_int(&args[0], &option))
1578 return 0;
1579 if (option < 0)
1580 return 0;
1581 if (option == 0)
cd02ff0b 1582 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
ac27a0ec
DK		 1583 sbi->s_commit_interval = HZ * option;
1584 break;
30773840
TT		 1585 case Opt_max_batch_time:
1586 if (match_int(&args[0], &option))
1587 return 0;
1588 if (option < 0)
1589 return 0;
1590 if (option == 0)
1591 option = EXT4_DEF_MAX_BATCH_TIME;
1592 sbi->s_max_batch_time = option;
1593 break;
1594 case Opt_min_batch_time:
1595 if (match_int(&args[0], &option))
1596 return 0;
1597 if (option < 0)
1598 return 0;
1599 sbi->s_min_batch_time = option;
1600 break;
ac27a0ec 1601 case Opt_data_journal:
617ba13b 1602 data_opt = EXT4_MOUNT_JOURNAL_DATA;
ac27a0ec
DK		 1603 goto datacheck;
1604 case Opt_data_ordered:
617ba13b 1605 data_opt = EXT4_MOUNT_ORDERED_DATA;
ac27a0ec
DK		 1606 goto datacheck;
1607 case Opt_data_writeback:
617ba13b 1608 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
ac27a0ec
DK		 1609 datacheck:
1610 if (is_remount) {
482a7425 1611 if (test_opt(sb, DATA_FLAGS) != data_opt) {
b31e1552
ES		 1612 ext4_msg(sb, KERN_ERR,
1613 "Cannot change data mode on remount");
ac27a0ec
DK		 1614 return 0;
1615 }
1616 } else {
fd8c37ec 1617 clear_opt(sb, DATA_FLAGS);
ac27a0ec
DK		 1618 sbi->s_mount_opt |= data_opt;
1619 }
1620 break;
5bf5683a 1621 case Opt_data_err_abort:
fd8c37ec 1622 set_opt(sb, DATA_ERR_ABORT);
5bf5683a
HK		 1623 break;
1624 case Opt_data_err_ignore:
fd8c37ec 1625 clear_opt(sb, DATA_ERR_ABORT);
5bf5683a 1626 break;
ac27a0ec
DK		 1627#ifdef CONFIG_QUOTA
1628 case Opt_usrjquota:
56c50f11 1629 if (!set_qf_name(sb, USRQUOTA, &args[0]))
ac27a0ec 1630 return 0;
56c50f11
DM		 1631 break;
1632 case Opt_grpjquota:
1633 if (!set_qf_name(sb, GRPQUOTA, &args[0]))
ac27a0ec 1634 return 0;
ac27a0ec
DK		 1635 break;
1636 case Opt_offusrjquota:
56c50f11
DM		 1637 if (!clear_qf_name(sb, USRQUOTA))
1638 return 0;
1639 break;
ac27a0ec 1640 case Opt_offgrpjquota:
56c50f11 1641 if (!clear_qf_name(sb, GRPQUOTA))
ac27a0ec 1642 return 0;
ac27a0ec 1643 break;
56c50f11 1644
ac27a0ec 1645 case Opt_jqfmt_vfsold:
dfc5d03f
JK		 1646 qfmt = QFMT_VFS_OLD;
1647 goto set_qf_format;
ac27a0ec 1648 case Opt_jqfmt_vfsv0:
dfc5d03f 1649 qfmt = QFMT_VFS_V0;
5a20bdfc
JK		 1650 goto set_qf_format;
1651 case Opt_jqfmt_vfsv1:
1652 qfmt = QFMT_VFS_V1;
dfc5d03f 1653set_qf_format:
17bd13b3 1654 if (sb_any_quota_loaded(sb) &&
dfc5d03f 1655 sbi->s_jquota_fmt != qfmt) {
b31e1552 1656 ext4_msg(sb, KERN_ERR, "Cannot change "
dfc5d03f 1657 "journaled quota options when "
b31e1552 1658 "quota turned on");
dfc5d03f
JK		 1659 return 0;
1660 }
1661 sbi->s_jquota_fmt = qfmt;
ac27a0ec
DK		 1662 break;
1663 case Opt_quota:
1664 case Opt_usrquota:
fd8c37ec
TT		 1665 set_opt(sb, QUOTA);
1666 set_opt(sb, USRQUOTA);
ac27a0ec
DK		 1667 break;
1668 case Opt_grpquota:
fd8c37ec
TT		 1669 set_opt(sb, QUOTA);
1670 set_opt(sb, GRPQUOTA);
ac27a0ec
DK		 1671 break;
1672 case Opt_noquota:
17bd13b3 1673 if (sb_any_quota_loaded(sb)) {
b31e1552
ES		 1674 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1675 "options when quota turned on");
ac27a0ec
DK		 1676 return 0;
1677 }
fd8c37ec
TT		 1678 clear_opt(sb, QUOTA);
1679 clear_opt(sb, USRQUOTA);
1680 clear_opt(sb, GRPQUOTA);
ac27a0ec
DK		 1681 break;
1682#else
1683 case Opt_quota:
1684 case Opt_usrquota:
1685 case Opt_grpquota:
b31e1552
ES		 1686 ext4_msg(sb, KERN_ERR,
1687 "quota options not supported");
cd59e7b9 1688 break;
ac27a0ec
DK		 1689 case Opt_usrjquota:
1690 case Opt_grpjquota:
1691 case Opt_offusrjquota:
1692 case Opt_offgrpjquota:
1693 case Opt_jqfmt_vfsold:
1694 case Opt_jqfmt_vfsv0:
5a20bdfc 1695 case Opt_jqfmt_vfsv1:
b31e1552
ES		 1696 ext4_msg(sb, KERN_ERR,
1697 "journaled quota options not supported");
ac27a0ec
DK		 1698 break;
1699 case Opt_noquota:
1700 break;
1701#endif
1702 case Opt_abort:
4ab2f15b 1703 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
ac27a0ec 1704 break;
06705bff 1705 case Opt_nobarrier:
fd8c37ec 1706 clear_opt(sb, BARRIER);
06705bff 1707 break;
ac27a0ec 1708 case Opt_barrier:
15121c18
ES		 1709 if (args[0].from) {
1710 if (match_int(&args[0], &option))
1711 return 0;
1712 } else
1713 option = 1; /* No argument, default to 1 */
ac27a0ec 1714 if (option)
fd8c37ec 1715 set_opt(sb, BARRIER);
ac27a0ec 1716 else
fd8c37ec 1717 clear_opt(sb, BARRIER);
ac27a0ec
DK		 1718 break;
1719 case Opt_ignore:
1720 break;
1721 case Opt_resize:
1722 if (!is_remount) {
b31e1552
ES		 1723 ext4_msg(sb, KERN_ERR,
1724 "resize option only available "
1725 "for remount");
ac27a0ec
DK		 1726 return 0;
1727 }
1728 if (match_int(&args[0], &option) != 0)
1729 return 0;
1730 *n_blocks_count = option;
1731 break;
1732 case Opt_nobh:
206f7ab4
CH		 1733 ext4_msg(sb, KERN_WARNING,
1734 "Ignoring deprecated nobh option");
ac27a0ec
DK		 1735 break;
1736 case Opt_bh:
206f7ab4
CH		 1737 ext4_msg(sb, KERN_WARNING,
1738 "Ignoring deprecated bh option");
ac27a0ec 1739 break;
25ec56b5 1740 case Opt_i_version:
fd8c37ec 1741 set_opt(sb, I_VERSION);
25ec56b5
JNC		 1742 sb->s_flags |= MS_I_VERSION;
1743 break;
dd919b98 1744 case Opt_nodelalloc:
fd8c37ec 1745 clear_opt(sb, DELALLOC);
dd919b98 1746 break;
1449032b 1747 case Opt_mblk_io_submit:
fd8c37ec 1748 set_opt(sb, MBLK_IO_SUBMIT);
1449032b
TT		 1749 break;
1750 case Opt_nomblk_io_submit:
fd8c37ec 1751 clear_opt(sb, MBLK_IO_SUBMIT);
1449032b 1752 break;
c9de560d
AT		 1753 case Opt_stripe:
1754 if (match_int(&args[0], &option))
1755 return 0;
1756 if (option < 0)
1757 return 0;
1758 sbi->s_stripe = option;
1759 break;
64769240 1760 case Opt_delalloc:
fd8c37ec 1761 set_opt(sb, DELALLOC);
64769240 1762 break;
6fd058f7 1763 case Opt_block_validity:
fd8c37ec 1764 set_opt(sb, BLOCK_VALIDITY);
6fd058f7
TT		 1765 break;
1766 case Opt_noblock_validity:
fd8c37ec 1767 clear_opt(sb, BLOCK_VALIDITY);
6fd058f7 1768 break;
240799cd
TT		 1769 case Opt_inode_readahead_blks:
1770 if (match_int(&args[0], &option))
1771 return 0;
1772 if (option < 0 || option > (1 << 30))
1773 return 0;
f7c43950 1774 if (!is_power_of_2(option)) {
b31e1552
ES		 1775 ext4_msg(sb, KERN_ERR,
1776 "EXT4-fs: inode_readahead_blks"
1777 " must be a power of 2");
3197ebdb
TT		 1778 return 0;
1779 }
240799cd
TT		 1780 sbi->s_inode_readahead_blks = option;
1781 break;
b3881f74
TT		 1782 case Opt_journal_ioprio:
1783 if (match_int(&args[0], &option))
1784 return 0;
1785 if (option < 0 || option > 7)
1786 break;
1787 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1788 option);
1789 break;
06705bff 1790 case Opt_noauto_da_alloc:
fd8c37ec 1791 set_opt(sb, NO_AUTO_DA_ALLOC);
06705bff 1792 break;
afd4672d 1793 case Opt_auto_da_alloc:
15121c18
ES		 1794 if (args[0].from) {
1795 if (match_int(&args[0], &option))
1796 return 0;
1797 } else
1798 option = 1; /* No argument, default to 1 */
afd4672d 1799 if (option)
fd8c37ec 1800 clear_opt(sb, NO_AUTO_DA_ALLOC);
afd4672d 1801 else
fd8c37ec 1802 set_opt(sb,NO_AUTO_DA_ALLOC);
afd4672d 1803 break;
5328e635 1804 case Opt_discard:
fd8c37ec 1805 set_opt(sb, DISCARD);
5328e635
ES		 1806 break;
1807 case Opt_nodiscard:
fd8c37ec 1808 clear_opt(sb, DISCARD);
5328e635 1809 break;
744692dc 1810 case Opt_dioread_nolock:
fd8c37ec 1811 set_opt(sb, DIOREAD_NOLOCK);
744692dc
JZ		 1812 break;
1813 case Opt_dioread_lock:
fd8c37ec 1814 clear_opt(sb, DIOREAD_NOLOCK);
744692dc 1815 break;
bfff6873 1816 case Opt_init_inode_table:
fd8c37ec 1817 set_opt(sb, INIT_INODE_TABLE);
bfff6873
LC		 1818 if (args[0].from) {
1819 if (match_int(&args[0], &option))
1820 return 0;
1821 } else
1822 option = EXT4_DEF_LI_WAIT_MULT;
1823 if (option < 0)
1824 return 0;
1825 sbi->s_li_wait_mult = option;
1826 break;
1827 case Opt_noinit_inode_table:
fd8c37ec 1828 clear_opt(sb, INIT_INODE_TABLE);
bfff6873 1829 break;
ac27a0ec 1830 default:
b31e1552
ES		 1831 ext4_msg(sb, KERN_ERR,
1832 "Unrecognized mount option \"%s\" "
1833 "or missing value", p);
ac27a0ec
DK		 1834 return 0;
1835 }
1836 }
1837#ifdef CONFIG_QUOTA
1838 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
482a7425 1839 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
fd8c37ec 1840 clear_opt(sb, USRQUOTA);
ac27a0ec 1841
482a7425 1842 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
fd8c37ec 1843 clear_opt(sb, GRPQUOTA);
ac27a0ec 1844
56c50f11 1845 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
b31e1552
ES		 1846 ext4_msg(sb, KERN_ERR, "old and new quota "
1847 "format mixing");
ac27a0ec
DK		 1848 return 0;
1849 }
1850
1851 if (!sbi->s_jquota_fmt) {
b31e1552
ES		 1852 ext4_msg(sb, KERN_ERR, "journaled quota format "
1853 "not specified");
ac27a0ec
DK		 1854 return 0;
1855 }
1856 } else {
1857 if (sbi->s_jquota_fmt) {
b31e1552 1858 ext4_msg(sb, KERN_ERR, "journaled quota format "
2c8be6b2 1859 "specified with no journaling "
b31e1552 1860 "enabled");
ac27a0ec
DK		 1861 return 0;
1862 }
1863 }
1864#endif
1865 return 1;
1866}
1867
617ba13b 1868static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 1869 int read_only)
1870{
617ba13b 1871 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 1872 int res = 0;
1873
617ba13b 1874 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
b31e1552
ES		 1875 ext4_msg(sb, KERN_ERR, "revision level too high, "
1876 "forcing read-only mode");
ac27a0ec
DK		 1877 res = MS_RDONLY;
1878 }
1879 if (read_only)
1880 return res;
617ba13b 1881 if (!(sbi->s_mount_state & EXT4_VALID_FS))
b31e1552
ES		 1882 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1883 "running e2fsck is recommended");
617ba13b 1884 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
b31e1552
ES		 1885 ext4_msg(sb, KERN_WARNING,
1886 "warning: mounting fs with errors, "
1887 "running e2fsck is recommended");
ac27a0ec
DK		 1888 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1889 le16_to_cpu(es->s_mnt_count) >=
1890 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
b31e1552
ES		 1891 ext4_msg(sb, KERN_WARNING,
1892 "warning: maximal mount count reached, "
1893 "running e2fsck is recommended");
ac27a0ec
DK		 1894 else if (le32_to_cpu(es->s_checkinterval) &&
1895 (le32_to_cpu(es->s_lastcheck) +
1896 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
b31e1552
ES		 1897 ext4_msg(sb, KERN_WARNING,
1898 "warning: checktime reached, "
1899 "running e2fsck is recommended");
0b8e58a1 1900 if (!sbi->s_journal)
0390131b 1901 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec 1902 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 1903 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 1904 le16_add_cpu(&es->s_mnt_count, 1);
ac27a0ec 1905 es->s_mtime = cpu_to_le32(get_seconds());
617ba13b 1906 ext4_update_dynamic_rev(sb);
0390131b
FM		 1907 if (sbi->s_journal)
1908 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 1909
e2d67052 1910 ext4_commit_super(sb, 1);
ac27a0ec 1911 if (test_opt(sb, DEBUG))
a9df9a49 1912 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
a2595b8a 1913 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
ac27a0ec
DK		 1914 sb->s_blocksize,
1915 sbi->s_groups_count,
617ba13b
MC		 1916 EXT4_BLOCKS_PER_GROUP(sb),
1917 EXT4_INODES_PER_GROUP(sb),
a2595b8a 1918 sbi->s_mount_opt, sbi->s_mount_opt2);
ac27a0ec 1919
ac27a0ec
DK		 1920 return res;
1921}
1922
772cb7c8
JS		 1923static int ext4_fill_flex_info(struct super_block *sb)
1924{
1925 struct ext4_sb_info *sbi = EXT4_SB(sb);
1926 struct ext4_group_desc *gdp = NULL;
772cb7c8
JS		 1927 ext4_group_t flex_group_count;
1928 ext4_group_t flex_group;
1929 int groups_per_flex = 0;
c5ca7c76 1930 size_t size;
772cb7c8
JS		 1931 int i;
1932
503358ae
TT		 1933 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1934 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1935
1936 if (groups_per_flex < 2) {
772cb7c8
JS		 1937 sbi->s_log_groups_per_flex = 0;
1938 return 1;
1939 }
1940
c62a11fd
FB		 1941 /* We allocate both existing and potentially added groups */
1942 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
d94e99a6
AK		 1943 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1944 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
c5ca7c76
TT		 1945 size = flex_group_count * sizeof(struct flex_groups);
1946 sbi->s_flex_groups = kzalloc(size, GFP_KERNEL);
1947 if (sbi->s_flex_groups == NULL) {
94de56ab
JP		 1948 sbi->s_flex_groups = vzalloc(size);
1949 if (sbi->s_flex_groups == NULL) {
1950 ext4_msg(sb, KERN_ERR,
1951 "not enough memory for %u flex groups",
1952 flex_group_count);
1953 goto failed;
1954 }
772cb7c8 1955 }
772cb7c8 1956
772cb7c8 1957 for (i = 0; i < sbi->s_groups_count; i++) {
88b6edd1 1958 gdp = ext4_get_group_desc(sb, i, NULL);
772cb7c8
JS		 1959
1960 flex_group = ext4_flex_group(sbi, i);
7ad9bb65
TT		 1961 atomic_add(ext4_free_inodes_count(sb, gdp),
1962 &sbi->s_flex_groups[flex_group].free_inodes);
1963 atomic_add(ext4_free_blks_count(sb, gdp),
1964 &sbi->s_flex_groups[flex_group].free_blocks);
1965 atomic_add(ext4_used_dirs_count(sb, gdp),
1966 &sbi->s_flex_groups[flex_group].used_dirs);
772cb7c8
JS		 1967 }
1968
1969 return 1;
1970failed:
1971 return 0;
1972}
1973
717d50e4
AD		 1974__le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1975 struct ext4_group_desc *gdp)
1976{
1977 __u16 crc = 0;
1978
1979 if (sbi->s_es->s_feature_ro_compat &
1980 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1981 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1982 __le32 le_group = cpu_to_le32(block_group);
1983
1984 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1985 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1986 crc = crc16(crc, (__u8 *)gdp, offset);
1987 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1988 /* for checksum of struct ext4_group_desc do the rest...*/
1989 if ((sbi->s_es->s_feature_incompat &
1990 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1991 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1992 crc = crc16(crc, (__u8 *)gdp + offset,
1993 le16_to_cpu(sbi->s_es->s_desc_size) -
1994 offset);
1995 }
1996
1997 return cpu_to_le16(crc);
1998}
1999
2000int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
2001 struct ext4_group_desc *gdp)
2002{
2003 if ((sbi->s_es->s_feature_ro_compat &
2004 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
2005 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
2006 return 0;
2007
2008 return 1;
2009}
2010
ac27a0ec 2011/* Called at mount-time, super-block is locked */
bfff6873
LC		 2012static int ext4_check_descriptors(struct super_block *sb,
2013 ext4_group_t *first_not_zeroed)
ac27a0ec 2014{
617ba13b
MC		 2015 struct ext4_sb_info *sbi = EXT4_SB(sb);
2016 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2017 ext4_fsblk_t last_block;
bd81d8ee
LV		 2018 ext4_fsblk_t block_bitmap;
2019 ext4_fsblk_t inode_bitmap;
2020 ext4_fsblk_t inode_table;
ce421581 2021 int flexbg_flag = 0;
bfff6873 2022 ext4_group_t i, grp = sbi->s_groups_count;
ac27a0ec 2023
ce421581
JS		 2024 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2025 flexbg_flag = 1;
2026
af5bc92d 2027 ext4_debug("Checking group descriptors");
ac27a0ec 2028
197cd65a
AM		 2029 for (i = 0; i < sbi->s_groups_count; i++) {
2030 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2031
ce421581 2032 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 2033 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 2034 else
2035 last_block = first_block +
617ba13b 2036 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 2037
bfff6873
LC		 2038 if ((grp == sbi->s_groups_count) &&
2039 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2040 grp = i;
2041
8fadc143 2042 block_bitmap = ext4_block_bitmap(sb, gdp);
2b2d6d01 2043 if (block_bitmap < first_block || block_bitmap > last_block) {
b31e1552 2044 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2045 "Block bitmap for group %u not in group "
b31e1552 2046 "(block %llu)!", i, block_bitmap);
ac27a0ec
DK		 2047 return 0;
2048 }
8fadc143 2049 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2b2d6d01 2050 if (inode_bitmap < first_block || inode_bitmap > last_block) {
b31e1552 2051 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2052 "Inode bitmap for group %u not in group "
b31e1552 2053 "(block %llu)!", i, inode_bitmap);
ac27a0ec
DK		 2054 return 0;
2055 }
8fadc143 2056 inode_table = ext4_inode_table(sb, gdp);
bd81d8ee 2057 if (inode_table < first_block ||
2b2d6d01 2058 inode_table + sbi->s_itb_per_group - 1 > last_block) {
b31e1552 2059 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2060 "Inode table for group %u not in group "
b31e1552 2061 "(block %llu)!", i, inode_table);
ac27a0ec
DK		 2062 return 0;
2063 }
955ce5f5 2064 ext4_lock_group(sb, i);
717d50e4 2065 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
b31e1552
ES		 2066 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2067 "Checksum for group %u failed (%u!=%u)",
2068 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2069 gdp)), le16_to_cpu(gdp->bg_checksum));
7ee1ec4c 2070 if (!(sb->s_flags & MS_RDONLY)) {
955ce5f5 2071 ext4_unlock_group(sb, i);
8a266467 2072 return 0;
7ee1ec4c 2073 }
717d50e4 2074 }
955ce5f5 2075 ext4_unlock_group(sb, i);
ce421581
JS		 2076 if (!flexbg_flag)
2077 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec 2078 }
bfff6873
LC		 2079 if (NULL != first_not_zeroed)
2080 *first_not_zeroed = grp;
ac27a0ec 2081
bd81d8ee 2082 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
0b8e58a1 2083 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
ac27a0ec
DK		 2084 return 1;
2085}
2086
617ba13b 2087/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 2088 * the superblock) which were deleted from all directories, but held open by
2089 * a process at the time of a crash. We walk the list and try to delete these
2090 * inodes at recovery time (only with a read-write filesystem).
2091 *
2092 * In order to keep the orphan inode chain consistent during traversal (in
2093 * case of crash during recovery), we link each inode into the superblock
2094 * orphan list_head and handle it the same way as an inode deletion during
2095 * normal operation (which journals the operations for us).
2096 *
2097 * We only do an iget() and an iput() on each inode, which is very safe if we
2098 * accidentally point at an in-use or already deleted inode. The worst that
2099 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 2100 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 2101 * e2fsck was run on this filesystem, and it must have already done the orphan
2102 * inode cleanup for us, so we can safely abort without any further action.
2103 */
2b2d6d01
TT		 2104static void ext4_orphan_cleanup(struct super_block *sb,
2105 struct ext4_super_block *es)
ac27a0ec
DK		 2106{
2107 unsigned int s_flags = sb->s_flags;
2108 int nr_orphans = 0, nr_truncates = 0;
2109#ifdef CONFIG_QUOTA
2110 int i;
2111#endif
2112 if (!es->s_last_orphan) {
2113 jbd_debug(4, "no orphan inodes to clean up\n");
2114 return;
2115 }
2116
a8f48a95 2117 if (bdev_read_only(sb->s_bdev)) {
b31e1552
ES		 2118 ext4_msg(sb, KERN_ERR, "write access "
2119 "unavailable, skipping orphan cleanup");
a8f48a95
ES		 2120 return;
2121 }
2122
617ba13b 2123 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
ac27a0ec
DK		 2124 if (es->s_last_orphan)
2125 jbd_debug(1, "Errors on filesystem, "
2126 "clearing orphan list.\n");
2127 es->s_last_orphan = 0;
2128 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2129 return;
2130 }
2131
2132 if (s_flags & MS_RDONLY) {
b31e1552 2133 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
ac27a0ec
DK		 2134 sb->s_flags &= ~MS_RDONLY;
2135 }
2136#ifdef CONFIG_QUOTA
2137 /* Needed for iput() to work correctly and not trash data */
2138 sb->s_flags |= MS_ACTIVE;
2139 /* Turn on quotas so that they are updated correctly */
2140 for (i = 0; i < MAXQUOTAS; i++) {
617ba13b
MC		 2141 if (EXT4_SB(sb)->s_qf_names[i]) {
2142 int ret = ext4_quota_on_mount(sb, i);
ac27a0ec 2143 if (ret < 0)
b31e1552
ES		 2144 ext4_msg(sb, KERN_ERR,
2145 "Cannot turn on journaled "
2146 "quota: error %d", ret);
ac27a0ec
DK		 2147 }
2148 }
2149#endif
2150
2151 while (es->s_last_orphan) {
2152 struct inode *inode;
2153
97bd42b9
JB		 2154 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2155 if (IS_ERR(inode)) {
ac27a0ec
DK		 2156 es->s_last_orphan = 0;
2157 break;
2158 }
2159
617ba13b 2160 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
871a2931 2161 dquot_initialize(inode);
ac27a0ec 2162 if (inode->i_nlink) {
b31e1552
ES		 2163 ext4_msg(sb, KERN_DEBUG,
2164 "%s: truncating inode %lu to %lld bytes",
46e665e9 2165 __func__, inode->i_ino, inode->i_size);
e5f8eab8 2166 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
ac27a0ec 2167 inode->i_ino, inode->i_size);
617ba13b 2168 ext4_truncate(inode);
ac27a0ec
DK		 2169 nr_truncates++;
2170 } else {
b31e1552
ES		 2171 ext4_msg(sb, KERN_DEBUG,
2172 "%s: deleting unreferenced inode %lu",
46e665e9 2173 __func__, inode->i_ino);
ac27a0ec
DK		 2174 jbd_debug(2, "deleting unreferenced inode %lu\n",
2175 inode->i_ino);
2176 nr_orphans++;
2177 }
2178 iput(inode); /* The delete magic happens here! */
2179 }
2180
2b2d6d01 2181#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
ac27a0ec
DK		 2182
2183 if (nr_orphans)
b31e1552
ES		 2184 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2185 PLURAL(nr_orphans));
ac27a0ec 2186 if (nr_truncates)
b31e1552
ES		 2187 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2188 PLURAL(nr_truncates));
ac27a0ec
DK		 2189#ifdef CONFIG_QUOTA
2190 /* Turn quotas off */
2191 for (i = 0; i < MAXQUOTAS; i++) {
2192 if (sb_dqopt(sb)->files[i])
287a8095 2193 dquot_quota_off(sb, i);
ac27a0ec
DK		 2194 }
2195#endif
2196 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2197}
0b8e58a1 2198
cd2291a4
ES		 2199/*
2200 * Maximal extent format file size.
2201 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2202 * extent format containers, within a sector_t, and within i_blocks
2203 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2204 * so that won't be a limiting factor.
2205 *
2206 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2207 */
f287a1a5 2208static loff_t ext4_max_size(int blkbits, int has_huge_files)
cd2291a4
ES		 2209{
2210 loff_t res;
2211 loff_t upper_limit = MAX_LFS_FILESIZE;
2212
2213 /* small i_blocks in vfs inode? */
f287a1a5 2214 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
cd2291a4 2215 /*
90c699a9 2216 * CONFIG_LBDAF is not enabled implies the inode
cd2291a4
ES		 2217 * i_block represent total blocks in 512 bytes
2218 * 32 == size of vfs inode i_blocks * 8
2219 */
2220 upper_limit = (1LL << 32) - 1;
2221
2222 /* total blocks in file system block size */
2223 upper_limit >>= (blkbits - 9);
2224 upper_limit <<= blkbits;
2225 }
2226
2227 /* 32-bit extent-start container, ee_block */
2228 res = 1LL << 32;
2229 res <<= blkbits;
2230 res -= 1;
2231
2232 /* Sanity check against vm- & vfs- imposed limits */
2233 if (res > upper_limit)
2234 res = upper_limit;
2235
2236 return res;
2237}
ac27a0ec 2238
ac27a0ec 2239/*
cd2291a4 2240 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 2241 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2242 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 2243 */
f287a1a5 2244static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
ac27a0ec 2245{
617ba13b 2246 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 2247 int meta_blocks;
2248 loff_t upper_limit;
0b8e58a1
AD		 2249 /* This is calculated to be the largest file size for a dense, block
2250 * mapped file such that the file's total number of 512-byte sectors,
2251 * including data and all indirect blocks, does not exceed (2^48 - 1).
2252 *
2253 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2254 * number of 512-byte sectors of the file.
0fc1b451
AK		 2255 */
2256
f287a1a5 2257 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
0fc1b451 2258 /*
90c699a9 2259 * !has_huge_files or CONFIG_LBDAF not enabled implies that
0b8e58a1
AD		 2260 * the inode i_block field represents total file blocks in
2261 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
0fc1b451
AK		 2262 */
2263 upper_limit = (1LL << 32) - 1;
2264
2265 /* total blocks in file system block size */
2266 upper_limit >>= (bits - 9);
2267
2268 } else {
8180a562
AK		 2269 /*
2270 * We use 48 bit ext4_inode i_blocks
2271 * With EXT4_HUGE_FILE_FL set the i_blocks
2272 * represent total number of blocks in
2273 * file system block size
2274 */
0fc1b451
AK		 2275 upper_limit = (1LL << 48) - 1;
2276
0fc1b451
AK		 2277 }
2278
2279 /* indirect blocks */
2280 meta_blocks = 1;
2281 /* double indirect blocks */
2282 meta_blocks += 1 + (1LL << (bits-2));
2283 /* tripple indirect blocks */
2284 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2285
2286 upper_limit -= meta_blocks;
2287 upper_limit <<= bits;
ac27a0ec
DK		 2288
2289 res += 1LL << (bits-2);
2290 res += 1LL << (2*(bits-2));
2291 res += 1LL << (3*(bits-2));
2292 res <<= bits;
2293 if (res > upper_limit)
2294 res = upper_limit;
0fc1b451
AK		 2295
2296 if (res > MAX_LFS_FILESIZE)
2297 res = MAX_LFS_FILESIZE;
2298
ac27a0ec
DK		 2299 return res;
2300}
2301
617ba13b 2302static ext4_fsblk_t descriptor_loc(struct super_block *sb,
0b8e58a1 2303 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 2304{
617ba13b 2305 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 2306 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 2307 int has_super = 0;
2308
2309 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2310
617ba13b 2311 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
ac27a0ec 2312 nr < first_meta_bg)
70bbb3e0 2313 return logical_sb_block + nr + 1;
ac27a0ec 2314 bg = sbi->s_desc_per_block * nr;
617ba13b 2315 if (ext4_bg_has_super(sb, bg))
ac27a0ec 2316 has_super = 1;
0b8e58a1 2317
617ba13b 2318 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 2319}
2320
c9de560d
AT		 2321/**
2322 * ext4_get_stripe_size: Get the stripe size.
2323 * @sbi: In memory super block info
2324 *
2325 * If we have specified it via mount option, then
2326 * use the mount option value. If the value specified at mount time is
2327 * greater than the blocks per group use the super block value.
2328 * If the super block value is greater than blocks per group return 0.
2329 * Allocator needs it be less than blocks per group.
2330 *
2331 */
2332static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2333{
2334 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2335 unsigned long stripe_width =
2336 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2337
2338 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2339 return sbi->s_stripe;
2340
2341 if (stripe_width <= sbi->s_blocks_per_group)
2342 return stripe_width;
2343
2344 if (stride <= sbi->s_blocks_per_group)
2345 return stride;
2346
2347 return 0;
2348}
ac27a0ec 2349
3197ebdb
TT		 2350/* sysfs supprt */
2351
2352struct ext4_attr {
2353 struct attribute attr;
2354 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
60e6679e 2355 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
3197ebdb
TT		 2356 const char *, size_t);
2357 int offset;
2358};
2359
2360static int parse_strtoul(const char *buf,
2361 unsigned long max, unsigned long *value)
2362{
2363 char *endp;
2364
e7d2860b
AGR		 2365 *value = simple_strtoul(skip_spaces(buf), &endp, 0);
2366 endp = skip_spaces(endp);
3197ebdb
TT		 2367 if (*endp || *value > max)
2368 return -EINVAL;
2369
2370 return 0;
2371}
2372
2373static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2374 struct ext4_sb_info *sbi,
2375 char *buf)
2376{
2377 return snprintf(buf, PAGE_SIZE, "%llu\n",
2378 (s64) percpu_counter_sum(&sbi->s_dirtyblocks_counter));
2379}
2380
2381static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2382 struct ext4_sb_info *sbi, char *buf)
2383{
2384 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2385
f613dfcb
TT		 2386 if (!sb->s_bdev->bd_part)
2387 return snprintf(buf, PAGE_SIZE, "0\n");
3197ebdb
TT		 2388 return snprintf(buf, PAGE_SIZE, "%lu\n",
2389 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2390 sbi->s_sectors_written_start) >> 1);
2391}
2392
2393static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2394 struct ext4_sb_info *sbi, char *buf)
2395{
2396 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2397
f613dfcb
TT		 2398 if (!sb->s_bdev->bd_part)
2399 return snprintf(buf, PAGE_SIZE, "0\n");
3197ebdb 2400 return snprintf(buf, PAGE_SIZE, "%llu\n",
a6b43e38 2401 (unsigned long long)(sbi->s_kbytes_written +
3197ebdb 2402 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
a6b43e38 2403 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
3197ebdb
TT		 2404}
2405
2406static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2407 struct ext4_sb_info *sbi,
2408 const char *buf, size_t count)
2409{
2410 unsigned long t;
2411
2412 if (parse_strtoul(buf, 0x40000000, &t))
2413 return -EINVAL;
2414
f7c43950 2415 if (!is_power_of_2(t))
3197ebdb
TT		 2416 return -EINVAL;
2417
2418 sbi->s_inode_readahead_blks = t;
2419 return count;
2420}
2421
2422static ssize_t sbi_ui_show(struct ext4_attr *a,
0b8e58a1 2423 struct ext4_sb_info *sbi, char *buf)
3197ebdb
TT		 2424{
2425 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2426
2427 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2428}
2429
2430static ssize_t sbi_ui_store(struct ext4_attr *a,
2431 struct ext4_sb_info *sbi,
2432 const char *buf, size_t count)
2433{
2434 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2435 unsigned long t;
2436
2437 if (parse_strtoul(buf, 0xffffffff, &t))
2438 return -EINVAL;
2439 *ui = t;
2440 return count;
2441}
2442
2443#define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2444static struct ext4_attr ext4_attr_##_name = { \
2445 .attr = {.name = __stringify(_name), .mode = _mode }, \
2446 .show = _show, \
2447 .store = _store, \
2448 .offset = offsetof(struct ext4_sb_info, _elname), \
2449}
2450#define EXT4_ATTR(name, mode, show, store) \
2451static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2452
857ac889 2453#define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
3197ebdb
TT		 2454#define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2455#define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2456#define EXT4_RW_ATTR_SBI_UI(name, elname) \
2457 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2458#define ATTR_LIST(name) &ext4_attr_##name.attr
2459
2460EXT4_RO_ATTR(delayed_allocation_blocks);
2461EXT4_RO_ATTR(session_write_kbytes);
2462EXT4_RO_ATTR(lifetime_write_kbytes);
2463EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2464 inode_readahead_blks_store, s_inode_readahead_blks);
11013911 2465EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
3197ebdb
TT		 2466EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2467EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2468EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2469EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2470EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2471EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
55138e0b 2472EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
3197ebdb
TT		 2473
2474static struct attribute *ext4_attrs[] = {
2475 ATTR_LIST(delayed_allocation_blocks),
2476 ATTR_LIST(session_write_kbytes),
2477 ATTR_LIST(lifetime_write_kbytes),
2478 ATTR_LIST(inode_readahead_blks),
11013911 2479 ATTR_LIST(inode_goal),
3197ebdb
TT		 2480 ATTR_LIST(mb_stats),
2481 ATTR_LIST(mb_max_to_scan),
2482 ATTR_LIST(mb_min_to_scan),
2483 ATTR_LIST(mb_order2_req),
2484 ATTR_LIST(mb_stream_req),
2485 ATTR_LIST(mb_group_prealloc),
55138e0b 2486 ATTR_LIST(max_writeback_mb_bump),
3197ebdb
TT		 2487 NULL,
2488};
2489
857ac889
LC		 2490/* Features this copy of ext4 supports */
2491EXT4_INFO_ATTR(lazy_itable_init);
27ee40df 2492EXT4_INFO_ATTR(batched_discard);
857ac889
LC		 2493
2494static struct attribute *ext4_feat_attrs[] = {
2495 ATTR_LIST(lazy_itable_init),
27ee40df 2496 ATTR_LIST(batched_discard),
857ac889
LC		 2497 NULL,
2498};
2499
3197ebdb
TT		 2500static ssize_t ext4_attr_show(struct kobject *kobj,
2501 struct attribute *attr, char *buf)
2502{
2503 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2504 s_kobj);
2505 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2506
2507 return a->show ? a->show(a, sbi, buf) : 0;
2508}
2509
2510static ssize_t ext4_attr_store(struct kobject *kobj,
2511 struct attribute *attr,
2512 const char *buf, size_t len)
2513{
2514 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2515 s_kobj);
2516 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2517
2518 return a->store ? a->store(a, sbi, buf, len) : 0;
2519}
2520
2521static void ext4_sb_release(struct kobject *kobj)
2522{
2523 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2524 s_kobj);
2525 complete(&sbi->s_kobj_unregister);
2526}
2527
52cf25d0 2528static const struct sysfs_ops ext4_attr_ops = {
3197ebdb
TT		 2529 .show = ext4_attr_show,
2530 .store = ext4_attr_store,
2531};
2532
2533static struct kobj_type ext4_ktype = {
2534 .default_attrs = ext4_attrs,
2535 .sysfs_ops = &ext4_attr_ops,
2536 .release = ext4_sb_release,
2537};
2538
857ac889
LC		 2539static void ext4_feat_release(struct kobject *kobj)
2540{
2541 complete(&ext4_feat->f_kobj_unregister);
2542}
2543
2544static struct kobj_type ext4_feat_ktype = {
2545 .default_attrs = ext4_feat_attrs,
2546 .sysfs_ops = &ext4_attr_ops,
2547 .release = ext4_feat_release,
2548};
2549
a13fb1a4
ES		 2550/*
2551 * Check whether this filesystem can be mounted based on
2552 * the features present and the RDONLY/RDWR mount requested.
2553 * Returns 1 if this filesystem can be mounted as requested,
2554 * 0 if it cannot be.
2555 */
2556static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2557{
2558 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2559 ext4_msg(sb, KERN_ERR,
2560 "Couldn't mount because of "
2561 "unsupported optional features (%x)",
2562 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2563 ~EXT4_FEATURE_INCOMPAT_SUPP));
2564 return 0;
2565 }
2566
2567 if (readonly)
2568 return 1;
2569
2570 /* Check that feature set is OK for a read-write mount */
2571 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2572 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2573 "unsupported optional features (%x)",
2574 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2575 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2576 return 0;
2577 }
2578 /*
2579 * Large file size enabled file system can only be mounted
2580 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2581 */
2582 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2583 if (sizeof(blkcnt_t) < sizeof(u64)) {
2584 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2585 "cannot be mounted RDWR without "
2586 "CONFIG_LBDAF");
2587 return 0;
2588 }
2589 }
2590 return 1;
2591}
2592
66e61a9e
TT		 2593/*
2594 * This function is called once a day if we have errors logged
2595 * on the file system
2596 */
2597static void print_daily_error_info(unsigned long arg)
2598{
2599 struct super_block *sb = (struct super_block *) arg;
2600 struct ext4_sb_info *sbi;
2601 struct ext4_super_block *es;
2602
2603 sbi = EXT4_SB(sb);
2604 es = sbi->s_es;
2605
2606 if (es->s_error_count)
2607 ext4_msg(sb, KERN_NOTICE, "error count: %u",
2608 le32_to_cpu(es->s_error_count));
2609 if (es->s_first_error_time) {
2610 printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d",
2611 sb->s_id, le32_to_cpu(es->s_first_error_time),
2612 (int) sizeof(es->s_first_error_func),
2613 es->s_first_error_func,
2614 le32_to_cpu(es->s_first_error_line));
2615 if (es->s_first_error_ino)
2616 printk(": inode %u",
2617 le32_to_cpu(es->s_first_error_ino));
2618 if (es->s_first_error_block)
2619 printk(": block %llu", (unsigned long long)
2620 le64_to_cpu(es->s_first_error_block));
2621 printk("\n");
2622 }
2623 if (es->s_last_error_time) {
2624 printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d",
2625 sb->s_id, le32_to_cpu(es->s_last_error_time),
2626 (int) sizeof(es->s_last_error_func),
2627 es->s_last_error_func,
2628 le32_to_cpu(es->s_last_error_line));
2629 if (es->s_last_error_ino)
2630 printk(": inode %u",
2631 le32_to_cpu(es->s_last_error_ino));
2632 if (es->s_last_error_block)
2633 printk(": block %llu", (unsigned long long)
2634 le64_to_cpu(es->s_last_error_block));
2635 printk("\n");
2636 }
2637 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2638}
2639
bfff6873
LC		 2640static void ext4_lazyinode_timeout(unsigned long data)
2641{
2642 struct task_struct *p = (struct task_struct *)data;
2643 wake_up_process(p);
2644}
2645
2646/* Find next suitable group and run ext4_init_inode_table */
2647static int ext4_run_li_request(struct ext4_li_request *elr)
2648{
2649 struct ext4_group_desc *gdp = NULL;
2650 ext4_group_t group, ngroups;
2651 struct super_block *sb;
2652 unsigned long timeout = 0;
2653 int ret = 0;
2654
2655 sb = elr->lr_super;
2656 ngroups = EXT4_SB(sb)->s_groups_count;
2657
2658 for (group = elr->lr_next_group; group < ngroups; group++) {
2659 gdp = ext4_get_group_desc(sb, group, NULL);
2660 if (!gdp) {
2661 ret = 1;
2662 break;
2663 }
2664
2665 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2666 break;
2667 }
2668
2669 if (group == ngroups)
2670 ret = 1;
2671
2672 if (!ret) {
2673 timeout = jiffies;
2674 ret = ext4_init_inode_table(sb, group,
2675 elr->lr_timeout ? 0 : 1);
2676 if (elr->lr_timeout == 0) {
2677 timeout = jiffies - timeout;
2678 if (elr->lr_sbi->s_li_wait_mult)
2679 timeout *= elr->lr_sbi->s_li_wait_mult;
2680 else
2681 timeout *= 20;
2682 elr->lr_timeout = timeout;
2683 }
2684 elr->lr_next_sched = jiffies + elr->lr_timeout;
2685 elr->lr_next_group = group + 1;
2686 }
2687
2688 return ret;
2689}
2690
2691/*
2692 * Remove lr_request from the list_request and free the
2693 * request tructure. Should be called with li_list_mtx held
2694 */
2695static void ext4_remove_li_request(struct ext4_li_request *elr)
2696{
2697 struct ext4_sb_info *sbi;
2698
2699 if (!elr)
2700 return;
2701
2702 sbi = elr->lr_sbi;
2703
2704 list_del(&elr->lr_request);
2705 sbi->s_li_request = NULL;
2706 kfree(elr);
2707}
2708
2709static void ext4_unregister_li_request(struct super_block *sb)
2710{
2711 struct ext4_li_request *elr = EXT4_SB(sb)->s_li_request;
2712
2713 if (!ext4_li_info)
2714 return;
2715
2716 mutex_lock(&ext4_li_info->li_list_mtx);
2717 ext4_remove_li_request(elr);
2718 mutex_unlock(&ext4_li_info->li_list_mtx);
2719}
2720
8f1f7453
ES		 2721static struct task_struct *ext4_lazyinit_task;
2722
bfff6873
LC		 2723/*
2724 * This is the function where ext4lazyinit thread lives. It walks
2725 * through the request list searching for next scheduled filesystem.
2726 * When such a fs is found, run the lazy initialization request
2727 * (ext4_rn_li_request) and keep track of the time spend in this
2728 * function. Based on that time we compute next schedule time of
2729 * the request. When walking through the list is complete, compute
2730 * next waking time and put itself into sleep.
2731 */
2732static int ext4_lazyinit_thread(void *arg)
2733{
2734 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2735 struct list_head *pos, *n;
2736 struct ext4_li_request *elr;
2737 unsigned long next_wakeup;
2738 DEFINE_WAIT(wait);
bfff6873
LC		 2739
2740 BUG_ON(NULL == eli);
2741
2742 eli->li_timer.data = (unsigned long)current;
2743 eli->li_timer.function = ext4_lazyinode_timeout;
2744
2745 eli->li_task = current;
2746 wake_up(&eli->li_wait_task);
2747
2748cont_thread:
2749 while (true) {
2750 next_wakeup = MAX_JIFFY_OFFSET;
2751
2752 mutex_lock(&eli->li_list_mtx);
2753 if (list_empty(&eli->li_request_list)) {
2754 mutex_unlock(&eli->li_list_mtx);
2755 goto exit_thread;
2756 }
2757
2758 list_for_each_safe(pos, n, &eli->li_request_list) {
2759 elr = list_entry(pos, struct ext4_li_request,
2760 lr_request);
2761
b2c78cd0
TT		 2762 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2763 if (ext4_run_li_request(elr) != 0) {
2764 /* error, remove the lazy_init job */
2765 ext4_remove_li_request(elr);
2766 continue;
2767 }
bfff6873
LC		 2768 }
2769
2770 if (time_before(elr->lr_next_sched, next_wakeup))
2771 next_wakeup = elr->lr_next_sched;
2772 }
2773 mutex_unlock(&eli->li_list_mtx);
2774
2775 if (freezing(current))
2776 refrigerator();
2777
f4245bd4
LC		 2778 if ((time_after_eq(jiffies, next_wakeup)) ||
2779 (MAX_JIFFY_OFFSET == next_wakeup)) {
bfff6873
LC		 2780 cond_resched();
2781 continue;
2782 }
2783
2784 eli->li_timer.expires = next_wakeup;
2785 add_timer(&eli->li_timer);
2786 prepare_to_wait(&eli->li_wait_daemon, &wait,
2787 TASK_INTERRUPTIBLE);
2788 if (time_before(jiffies, next_wakeup))
2789 schedule();
2790 finish_wait(&eli->li_wait_daemon, &wait);
8f1f7453
ES		 2791 if (kthread_should_stop()) {
2792 ext4_clear_request_list();
2793 goto exit_thread;
2794 }
bfff6873
LC		 2795 }
2796
2797exit_thread:
2798 /*
2799 * It looks like the request list is empty, but we need
2800 * to check it under the li_list_mtx lock, to prevent any
2801 * additions into it, and of course we should lock ext4_li_mtx
2802 * to atomically free the list and ext4_li_info, because at
2803 * this point another ext4 filesystem could be registering
2804 * new one.
2805 */
2806 mutex_lock(&ext4_li_mtx);
2807 mutex_lock(&eli->li_list_mtx);
2808 if (!list_empty(&eli->li_request_list)) {
2809 mutex_unlock(&eli->li_list_mtx);
2810 mutex_unlock(&ext4_li_mtx);
2811 goto cont_thread;
2812 }
2813 mutex_unlock(&eli->li_list_mtx);
2814 del_timer_sync(&ext4_li_info->li_timer);
2815 eli->li_task = NULL;
2816 wake_up(&eli->li_wait_task);
2817
2818 kfree(ext4_li_info);
8f1f7453 2819 ext4_lazyinit_task = NULL;
bfff6873
LC		 2820 ext4_li_info = NULL;
2821 mutex_unlock(&ext4_li_mtx);
2822
2823 return 0;
2824}
2825
2826static void ext4_clear_request_list(void)
2827{
2828 struct list_head *pos, *n;
2829 struct ext4_li_request *elr;
2830
2831 mutex_lock(&ext4_li_info->li_list_mtx);
bfff6873
LC		 2832 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2833 elr = list_entry(pos, struct ext4_li_request,
2834 lr_request);
2835 ext4_remove_li_request(elr);
2836 }
2837 mutex_unlock(&ext4_li_info->li_list_mtx);
2838}
2839
2840static int ext4_run_lazyinit_thread(void)
2841{
8f1f7453
ES		 2842 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2843 ext4_li_info, "ext4lazyinit");
2844 if (IS_ERR(ext4_lazyinit_task)) {
2845 int err = PTR_ERR(ext4_lazyinit_task);
bfff6873
LC		 2846 ext4_clear_request_list();
2847 del_timer_sync(&ext4_li_info->li_timer);
2848 kfree(ext4_li_info);
2849 ext4_li_info = NULL;
2850 printk(KERN_CRIT "EXT4: error %d creating inode table "
2851 "initialization thread\n",
2852 err);
2853 return err;
2854 }
2855 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
2856
2857 wait_event(ext4_li_info->li_wait_task, ext4_li_info->li_task != NULL);
2858 return 0;
2859}
2860
2861/*
2862 * Check whether it make sense to run itable init. thread or not.
2863 * If there is at least one uninitialized inode table, return
2864 * corresponding group number, else the loop goes through all
2865 * groups and return total number of groups.
2866 */
2867static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2868{
2869 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2870 struct ext4_group_desc *gdp = NULL;
2871
2872 for (group = 0; group < ngroups; group++) {
2873 gdp = ext4_get_group_desc(sb, group, NULL);
2874 if (!gdp)
2875 continue;
2876
2877 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2878 break;
2879 }
2880
2881 return group;
2882}
2883
2884static int ext4_li_info_new(void)
2885{
2886 struct ext4_lazy_init *eli = NULL;
2887
2888 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2889 if (!eli)
2890 return -ENOMEM;
2891
2892 eli->li_task = NULL;
2893 INIT_LIST_HEAD(&eli->li_request_list);
2894 mutex_init(&eli->li_list_mtx);
2895
2896 init_waitqueue_head(&eli->li_wait_daemon);
2897 init_waitqueue_head(&eli->li_wait_task);
2898 init_timer(&eli->li_timer);
2899 eli->li_state |= EXT4_LAZYINIT_QUIT;
2900
2901 ext4_li_info = eli;
2902
2903 return 0;
2904}
2905
2906static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
2907 ext4_group_t start)
2908{
2909 struct ext4_sb_info *sbi = EXT4_SB(sb);
2910 struct ext4_li_request *elr;
2911 unsigned long rnd;
2912
2913 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
2914 if (!elr)
2915 return NULL;
2916
2917 elr->lr_super = sb;
2918 elr->lr_sbi = sbi;
2919 elr->lr_next_group = start;
2920
2921 /*
2922 * Randomize first schedule time of the request to
2923 * spread the inode table initialization requests
2924 * better.
2925 */
2926 get_random_bytes(&rnd, sizeof(rnd));
2927 elr->lr_next_sched = jiffies + (unsigned long)rnd %
2928 (EXT4_DEF_LI_MAX_START_DELAY * HZ);
2929
2930 return elr;
2931}
2932
2933static int ext4_register_li_request(struct super_block *sb,
2934 ext4_group_t first_not_zeroed)
2935{
2936 struct ext4_sb_info *sbi = EXT4_SB(sb);
2937 struct ext4_li_request *elr;
2938 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
6c5a6cb9 2939 int ret = 0;
bfff6873
LC		 2940
2941 if (sbi->s_li_request != NULL)
beed5ecb 2942 return 0;
bfff6873
LC		 2943
2944 if (first_not_zeroed == ngroups ||
2945 (sb->s_flags & MS_RDONLY) ||
2946 !test_opt(sb, INIT_INODE_TABLE)) {
2947 sbi->s_li_request = NULL;
beed5ecb 2948 return 0;
bfff6873
LC		 2949 }
2950
2951 if (first_not_zeroed == ngroups) {
2952 sbi->s_li_request = NULL;
beed5ecb 2953 return 0;
bfff6873
LC		 2954 }
2955
2956 elr = ext4_li_request_new(sb, first_not_zeroed);
beed5ecb
NK		 2957 if (!elr)
2958 return -ENOMEM;
bfff6873
LC		 2959
2960 mutex_lock(&ext4_li_mtx);
2961
2962 if (NULL == ext4_li_info) {
2963 ret = ext4_li_info_new();
2964 if (ret)
2965 goto out;
2966 }
2967
2968 mutex_lock(&ext4_li_info->li_list_mtx);
2969 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
2970 mutex_unlock(&ext4_li_info->li_list_mtx);
2971
2972 sbi->s_li_request = elr;
2973
2974 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
2975 ret = ext4_run_lazyinit_thread();
2976 if (ret)
2977 goto out;
2978 }
bfff6873 2979out:
beed5ecb
NK		 2980 mutex_unlock(&ext4_li_mtx);
2981 if (ret)
bfff6873 2982 kfree(elr);
bfff6873
LC		 2983 return ret;
2984}
2985
2986/*
2987 * We do not need to lock anything since this is called on
2988 * module unload.
2989 */
2990static void ext4_destroy_lazyinit_thread(void)
2991{
2992 /*
2993 * If thread exited earlier
2994 * there's nothing to be done.
2995 */
8f1f7453 2996 if (!ext4_li_info || !ext4_lazyinit_task)
bfff6873
LC		 2997 return;
2998
8f1f7453 2999 kthread_stop(ext4_lazyinit_task);
bfff6873
LC		 3000}
3001
2b2d6d01 3002static int ext4_fill_super(struct super_block *sb, void *data, int silent)
7477827f
AK		 3003 __releases(kernel_lock)
3004 __acquires(kernel_lock)
ac27a0ec 3005{
d4c402d9 3006 char *orig_data = kstrdup(data, GFP_KERNEL);
2b2d6d01 3007 struct buffer_head *bh;
617ba13b
MC		 3008 struct ext4_super_block *es = NULL;
3009 struct ext4_sb_info *sbi;
3010 ext4_fsblk_t block;
3011 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 3012 ext4_fsblk_t logical_sb_block;
ac27a0ec 3013 unsigned long offset = 0;
ac27a0ec
DK		 3014 unsigned long journal_devnum = 0;
3015 unsigned long def_mount_opts;
3016 struct inode *root;
9f6200bb 3017 char *cp;
0390131b 3018 const char *descr;
dcc7dae3 3019 int ret = -ENOMEM;
ac27a0ec 3020 int blocksize;
4ec11028
TT		 3021 unsigned int db_count;
3022 unsigned int i;
f287a1a5 3023 int needs_recovery, has_huge_files;
bd81d8ee 3024 __u64 blocks_count;
833f4077 3025 int err;
b3881f74 3026 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
bfff6873 3027 ext4_group_t first_not_zeroed;
ac27a0ec
DK		 3028
3029 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3030 if (!sbi)
dcc7dae3 3031 goto out_free_orig;
705895b6
PE		 3032
3033 sbi->s_blockgroup_lock =
3034 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3035 if (!sbi->s_blockgroup_lock) {
3036 kfree(sbi);
dcc7dae3 3037 goto out_free_orig;
705895b6 3038 }
ac27a0ec
DK		 3039 sb->s_fs_info = sbi;
3040 sbi->s_mount_opt = 0;
617ba13b
MC		 3041 sbi->s_resuid = EXT4_DEF_RESUID;
3042 sbi->s_resgid = EXT4_DEF_RESGID;
240799cd 3043 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
d9c9bef1 3044 sbi->s_sb_block = sb_block;
f613dfcb
TT		 3045 if (sb->s_bdev->bd_part)
3046 sbi->s_sectors_written_start =
3047 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
ac27a0ec 3048
9f6200bb
TT		 3049 /* Cleanup superblock name */
3050 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3051 *cp = '!';
3052
dcc7dae3 3053 ret = -EINVAL;
617ba13b 3054 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 3055 if (!blocksize) {
b31e1552 3056 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
ac27a0ec
DK		 3057 goto out_fail;
3058 }
3059
3060 /*
617ba13b 3061 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 3062 * block sizes. We need to calculate the offset from buffer start.
3063 */
617ba13b 3064 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 3065 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3066 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 3067 } else {
70bbb3e0 3068 logical_sb_block = sb_block;
ac27a0ec
DK		 3069 }
3070
70bbb3e0 3071 if (!(bh = sb_bread(sb, logical_sb_block))) {
b31e1552 3072 ext4_msg(sb, KERN_ERR, "unable to read superblock");
ac27a0ec
DK		 3073 goto out_fail;
3074 }
3075 /*
3076 * Note: s_es must be initialized as soon as possible because
617ba13b 3077 * some ext4 macro-instructions depend on its value
ac27a0ec 3078 */
617ba13b 3079 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
ac27a0ec
DK		 3080 sbi->s_es = es;
3081 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 3082 if (sb->s_magic != EXT4_SUPER_MAGIC)
3083 goto cantfind_ext4;
afc32f7e 3084 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
ac27a0ec
DK		 3085
3086 /* Set defaults before we parse the mount options */
3087 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
fd8c37ec 3088 set_opt(sb, INIT_INODE_TABLE);
617ba13b 3089 if (def_mount_opts & EXT4_DEFM_DEBUG)
fd8c37ec 3090 set_opt(sb, DEBUG);
437ca0fd
DM		 3091 if (def_mount_opts & EXT4_DEFM_BSDGROUPS) {
3092 ext4_msg(sb, KERN_WARNING, deprecated_msg, "bsdgroups",
3093 "2.6.38");
fd8c37ec 3094 set_opt(sb, GRPID);
437ca0fd 3095 }
617ba13b 3096 if (def_mount_opts & EXT4_DEFM_UID16)
fd8c37ec 3097 set_opt(sb, NO_UID32);
03010a33 3098#ifdef CONFIG_EXT4_FS_XATTR
617ba13b 3099 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
fd8c37ec 3100 set_opt(sb, XATTR_USER);
2e7842b8 3101#endif
03010a33 3102#ifdef CONFIG_EXT4_FS_POSIX_ACL
617ba13b 3103 if (def_mount_opts & EXT4_DEFM_ACL)
fd8c37ec 3104 set_opt(sb, POSIX_ACL);
2e7842b8 3105#endif
617ba13b 3106 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
fd8c37ec 3107 set_opt(sb, JOURNAL_DATA);
617ba13b 3108 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
fd8c37ec 3109 set_opt(sb, ORDERED_DATA);
617ba13b 3110 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
fd8c37ec 3111 set_opt(sb, WRITEBACK_DATA);
617ba13b
MC		 3112
3113 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
fd8c37ec 3114 set_opt(sb, ERRORS_PANIC);
bb4f397a 3115 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
fd8c37ec 3116 set_opt(sb, ERRORS_CONT);
bb4f397a 3117 else
fd8c37ec 3118 set_opt(sb, ERRORS_RO);
8b67f04a 3119 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
fd8c37ec 3120 set_opt(sb, BLOCK_VALIDITY);
8b67f04a 3121 if (def_mount_opts & EXT4_DEFM_DISCARD)
fd8c37ec 3122 set_opt(sb, DISCARD);
ac27a0ec
DK		 3123
3124 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
3125 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
30773840
TT		 3126 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3127 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3128 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
ac27a0ec 3129
8b67f04a 3130 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
fd8c37ec 3131 set_opt(sb, BARRIER);
ac27a0ec 3132
dd919b98
AK		 3133 /*
3134 * enable delayed allocation by default
3135 * Use -o nodelalloc to turn it off
3136 */
8b67f04a
TT		 3137 if (!IS_EXT3_SB(sb) &&
3138 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
fd8c37ec 3139 set_opt(sb, DELALLOC);
dd919b98 3140
8b67f04a
TT		 3141 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3142 &journal_devnum, &journal_ioprio, NULL, 0)) {
3143 ext4_msg(sb, KERN_WARNING,
3144 "failed to parse options in superblock: %s",
3145 sbi->s_es->s_mount_opts);
3146 }
b3881f74
TT		 3147 if (!parse_options((char *) data, sb, &journal_devnum,
3148 &journal_ioprio, NULL, 0))
ac27a0ec
DK		 3149 goto failed_mount;
3150
3151 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 3152 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec 3153
617ba13b
MC		 3154 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3155 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3156 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3157 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
b31e1552
ES		 3158 ext4_msg(sb, KERN_WARNING,
3159 "feature flags set on rev 0 fs, "
3160 "running e2fsck is recommended");
469108ff 3161
ac27a0ec
DK		 3162 /*
3163 * Check feature flags regardless of the revision level, since we
3164 * previously didn't change the revision level when setting the flags,
3165 * so there is a chance incompat flags are set on a rev 0 filesystem.
3166 */
a13fb1a4 3167 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
ac27a0ec 3168 goto failed_mount;
a13fb1a4 3169
ac27a0ec
DK		 3170 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3171
617ba13b
MC		 3172 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3173 blocksize > EXT4_MAX_BLOCK_SIZE) {
b31e1552
ES		 3174 ext4_msg(sb, KERN_ERR,
3175 "Unsupported filesystem blocksize %d", blocksize);
ac27a0ec
DK		 3176 goto failed_mount;
3177 }
3178
ac27a0ec 3179 if (sb->s_blocksize != blocksize) {
ce40733c
AK		 3180 /* Validate the filesystem blocksize */
3181 if (!sb_set_blocksize(sb, blocksize)) {
b31e1552 3182 ext4_msg(sb, KERN_ERR, "bad block size %d",
ce40733c 3183 blocksize);
ac27a0ec
DK		 3184 goto failed_mount;
3185 }
3186
2b2d6d01 3187 brelse(bh);
70bbb3e0
AM		 3188 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3189 offset = do_div(logical_sb_block, blocksize);
3190 bh = sb_bread(sb, logical_sb_block);
ac27a0ec 3191 if (!bh) {
b31e1552
ES		 3192 ext4_msg(sb, KERN_ERR,
3193 "Can't read superblock on 2nd try");
ac27a0ec
DK		 3194 goto failed_mount;
3195 }
617ba13b 3196 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
ac27a0ec 3197 sbi->s_es = es;
617ba13b 3198 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
b31e1552
ES		 3199 ext4_msg(sb, KERN_ERR,
3200 "Magic mismatch, very weird!");
ac27a0ec
DK		 3201 goto failed_mount;
3202 }
3203 }
3204
a13fb1a4
ES		 3205 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3206 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
f287a1a5
TT		 3207 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3208 has_huge_files);
3209 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
ac27a0ec 3210
617ba13b
MC		 3211 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3212 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3213 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
ac27a0ec
DK		 3214 } else {
3215 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3216 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
617ba13b 3217 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1330593e 3218 (!is_power_of_2(sbi->s_inode_size)) ||
ac27a0ec 3219 (sbi->s_inode_size > blocksize)) {
b31e1552
ES		 3220 ext4_msg(sb, KERN_ERR,
3221 "unsupported inode size: %d",
2b2d6d01 3222 sbi->s_inode_size);
ac27a0ec
DK		 3223 goto failed_mount;
3224 }
ef7f3835
KS		 3225 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3226 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
ac27a0ec 3227 }
0b8e58a1 3228
0d1ee42f
AR		 3229 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3230 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
8fadc143 3231 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 3232 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 3233 !is_power_of_2(sbi->s_desc_size)) {
b31e1552
ES		 3234 ext4_msg(sb, KERN_ERR,
3235 "unsupported descriptor size %lu",
0d1ee42f
AR		 3236 sbi->s_desc_size);
3237 goto failed_mount;
3238 }
3239 } else
3240 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
0b8e58a1 3241
ac27a0ec 3242 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 3243 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
b47b6f38 3244 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
617ba13b 3245 goto cantfind_ext4;
0b8e58a1 3246
617ba13b 3247 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 3248 if (sbi->s_inodes_per_block == 0)
617ba13b 3249 goto cantfind_ext4;
ac27a0ec
DK		 3250 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3251 sbi->s_inodes_per_block;
0d1ee42f 3252 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 3253 sbi->s_sbh = bh;
3254 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 3255 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3256 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
0b8e58a1 3257
2b2d6d01 3258 for (i = 0; i < 4; i++)
ac27a0ec
DK		 3259 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3260 sbi->s_def_hash_version = es->s_def_hash_version;
f99b2589
TT		 3261 i = le32_to_cpu(es->s_flags);
3262 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3263 sbi->s_hash_unsigned = 3;
3264 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3265#ifdef __CHAR_UNSIGNED__
3266 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3267 sbi->s_hash_unsigned = 3;
3268#else
3269 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3270#endif
3271 sb->s_dirt = 1;
3272 }
ac27a0ec
DK		 3273
3274 if (sbi->s_blocks_per_group > blocksize * 8) {
b31e1552
ES		 3275 ext4_msg(sb, KERN_ERR,
3276 "#blocks per group too big: %lu",
2b2d6d01 3277 sbi->s_blocks_per_group);
ac27a0ec
DK		 3278 goto failed_mount;
3279 }
ac27a0ec 3280 if (sbi->s_inodes_per_group > blocksize * 8) {
b31e1552
ES		 3281 ext4_msg(sb, KERN_ERR,
3282 "#inodes per group too big: %lu",
2b2d6d01 3283 sbi->s_inodes_per_group);
ac27a0ec
DK		 3284 goto failed_mount;
3285 }
3286
bf43d84b
ES		 3287 /*
3288 * Test whether we have more sectors than will fit in sector_t,
3289 * and whether the max offset is addressable by the page cache.
3290 */
5a9ae68a 3291 err = generic_check_addressable(sb->s_blocksize_bits,
30ca22c7 3292 ext4_blocks_count(es));
5a9ae68a 3293 if (err) {
b31e1552 3294 ext4_msg(sb, KERN_ERR, "filesystem"
bf43d84b 3295 " too large to mount safely on this system");
ac27a0ec 3296 if (sizeof(sector_t) < 8)
90c699a9 3297 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
5a9ae68a 3298 ret = err;
ac27a0ec
DK		 3299 goto failed_mount;
3300 }
3301
617ba13b
MC		 3302 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3303 goto cantfind_ext4;
e7c95593 3304
0f2ddca6
FTN		 3305 /* check blocks count against device size */
3306 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3307 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
b31e1552
ES		 3308 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3309 "exceeds size of device (%llu blocks)",
0f2ddca6
FTN		 3310 ext4_blocks_count(es), blocks_count);
3311 goto failed_mount;
3312 }
3313
0b8e58a1
AD		 3314 /*
3315 * It makes no sense for the first data block to be beyond the end
3316 * of the filesystem.
3317 */
3318 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
b31e1552
ES		 3319 ext4_msg(sb, KERN_WARNING, "bad geometry: first data"
3320 "block %u is beyond end of filesystem (%llu)",
3321 le32_to_cpu(es->s_first_data_block),
3322 ext4_blocks_count(es));
e7c95593
ES		 3323 goto failed_mount;
3324 }
bd81d8ee
LV		 3325 blocks_count = (ext4_blocks_count(es) -
3326 le32_to_cpu(es->s_first_data_block) +
3327 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3328 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4ec11028 3329 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
b31e1552 3330 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
4ec11028 3331 "(block count %llu, first data block %u, "
b31e1552 3332 "blocks per group %lu)", sbi->s_groups_count,
4ec11028
TT		 3333 ext4_blocks_count(es),
3334 le32_to_cpu(es->s_first_data_block),
3335 EXT4_BLOCKS_PER_GROUP(sb));
3336 goto failed_mount;
3337 }
bd81d8ee 3338 sbi->s_groups_count = blocks_count;
fb0a387d
ES		 3339 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3340 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
617ba13b
MC		 3341 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3342 EXT4_DESC_PER_BLOCK(sb);
2b2d6d01 3343 sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
ac27a0ec
DK		 3344 GFP_KERNEL);
3345 if (sbi->s_group_desc == NULL) {
b31e1552 3346 ext4_msg(sb, KERN_ERR, "not enough memory");
ac27a0ec
DK		 3347 goto failed_mount;
3348 }
3349
3244fcb1 3350#ifdef CONFIG_PROC_FS
9f6200bb
TT		 3351 if (ext4_proc_root)
3352 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
3244fcb1 3353#endif
240799cd 3354
705895b6 3355 bgl_lock_init(sbi->s_blockgroup_lock);
ac27a0ec
DK		 3356
3357 for (i = 0; i < db_count; i++) {
70bbb3e0 3358 block = descriptor_loc(sb, logical_sb_block, i);
ac27a0ec
DK		 3359 sbi->s_group_desc[i] = sb_bread(sb, block);
3360 if (!sbi->s_group_desc[i]) {
b31e1552
ES		 3361 ext4_msg(sb, KERN_ERR,
3362 "can't read group descriptor %d", i);
ac27a0ec
DK		 3363 db_count = i;
3364 goto failed_mount2;
3365 }
3366 }
bfff6873 3367 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
b31e1552 3368 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
ac27a0ec
DK		 3369 goto failed_mount2;
3370 }
772cb7c8
JS		 3371 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3372 if (!ext4_fill_flex_info(sb)) {
b31e1552
ES		 3373 ext4_msg(sb, KERN_ERR,
3374 "unable to initialize "
3375 "flex_bg meta info!");
772cb7c8
JS		 3376 goto failed_mount2;
3377 }
3378
ac27a0ec
DK		 3379 sbi->s_gdb_count = db_count;
3380 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3381 spin_lock_init(&sbi->s_next_gen_lock);
3382
ce7e010a
TT		 3383 err = percpu_counter_init(&sbi->s_freeblocks_counter,
3384 ext4_count_free_blocks(sb));
3385 if (!err) {
3386 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3387 ext4_count_free_inodes(sb));
3388 }
3389 if (!err) {
3390 err = percpu_counter_init(&sbi->s_dirs_counter,
3391 ext4_count_dirs(sb));
3392 }
3393 if (!err) {
3394 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
3395 }
3396 if (err) {
3397 ext4_msg(sb, KERN_ERR, "insufficient memory");
3398 goto failed_mount3;
3399 }
3400
c9de560d 3401 sbi->s_stripe = ext4_get_stripe_size(sbi);
55138e0b 3402 sbi->s_max_writeback_mb_bump = 128;
c9de560d 3403
ac27a0ec
DK		 3404 /*
3405 * set up enough so that it can read an inode
3406 */
9ca92389
TT		 3407 if (!test_opt(sb, NOLOAD) &&
3408 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3409 sb->s_op = &ext4_sops;
3410 else
3411 sb->s_op = &ext4_nojournal_sops;
617ba13b
MC		 3412 sb->s_export_op = &ext4_export_ops;
3413 sb->s_xattr = ext4_xattr_handlers;
ac27a0ec 3414#ifdef CONFIG_QUOTA
617ba13b
MC		 3415 sb->s_qcop = &ext4_qctl_operations;
3416 sb->dq_op = &ext4_quota_operations;
ac27a0ec 3417#endif
f2fa2ffc
AK		 3418 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3419
ac27a0ec 3420 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3b9d4ed2 3421 mutex_init(&sbi->s_orphan_lock);
32ed5058 3422 mutex_init(&sbi->s_resize_lock);
ac27a0ec
DK		 3423
3424 sb->s_root = NULL;
3425
3426 needs_recovery = (es->s_last_orphan != 0 ||
617ba13b
MC		 3427 EXT4_HAS_INCOMPAT_FEATURE(sb,
3428 EXT4_FEATURE_INCOMPAT_RECOVER));
ac27a0ec
DK		 3429
3430 /*
3431 * The first inode we look at is the journal inode. Don't try
3432 * root first: it may be modified in the journal!
3433 */
3434 if (!test_opt(sb, NOLOAD) &&
617ba13b
MC		 3435 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3436 if (ext4_load_journal(sb, es, journal_devnum))
ac27a0ec 3437 goto failed_mount3;
0390131b
FM		 3438 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3439 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
b31e1552
ES		 3440 ext4_msg(sb, KERN_ERR, "required journal recovery "
3441 "suppressed and not mounted read-only");
744692dc 3442 goto failed_mount_wq;
ac27a0ec 3443 } else {
fd8c37ec
TT		 3444 clear_opt(sb, DATA_FLAGS);
3445 set_opt(sb, WRITEBACK_DATA);
0390131b
FM		 3446 sbi->s_journal = NULL;
3447 needs_recovery = 0;
3448 goto no_journal;
ac27a0ec
DK		 3449 }
3450
eb40a09c
JS		 3451 if (ext4_blocks_count(es) > 0xffffffffULL &&
3452 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3453 JBD2_FEATURE_INCOMPAT_64BIT)) {
b31e1552 3454 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
744692dc 3455 goto failed_mount_wq;
eb40a09c
JS		 3456 }
3457
d4da6c9c
LT		 3458 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3459 jbd2_journal_set_features(sbi->s_journal,
3460 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
818d276c 3461 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
d4da6c9c
LT		 3462 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3463 jbd2_journal_set_features(sbi->s_journal,
3464 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
818d276c
GS		 3465 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3466 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
d4da6c9c
LT		 3467 } else {
3468 jbd2_journal_clear_features(sbi->s_journal,
3469 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3470 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3471 }
818d276c 3472
ac27a0ec
DK		 3473 /* We have now updated the journal if required, so we can
3474 * validate the data journaling mode. */
3475 switch (test_opt(sb, DATA_FLAGS)) {
3476 case 0:
3477 /* No mode set, assume a default based on the journal
63f57933
AM		 3478 * capabilities: ORDERED_DATA if the journal can
3479 * cope, else JOURNAL_DATA
3480 */
dab291af
MC		 3481 if (jbd2_journal_check_available_features
3482 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
fd8c37ec 3483 set_opt(sb, ORDERED_DATA);
ac27a0ec 3484 else
fd8c37ec 3485 set_opt(sb, JOURNAL_DATA);
ac27a0ec
DK		 3486 break;
3487
617ba13b
MC		 3488 case EXT4_MOUNT_ORDERED_DATA:
3489 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 3490 if (!jbd2_journal_check_available_features
3491 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
b31e1552
ES		 3492 ext4_msg(sb, KERN_ERR, "Journal does not support "
3493 "requested data journaling mode");
744692dc 3494 goto failed_mount_wq;
ac27a0ec
DK		 3495 }
3496 default:
3497 break;
3498 }
b3881f74 3499 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
ac27a0ec 3500
ce7e010a
TT		 3501 /*
3502 * The journal may have updated the bg summary counts, so we
3503 * need to update the global counters.
3504 */
3505 percpu_counter_set(&sbi->s_freeblocks_counter,
3506 ext4_count_free_blocks(sb));
3507 percpu_counter_set(&sbi->s_freeinodes_counter,
3508 ext4_count_free_inodes(sb));
3509 percpu_counter_set(&sbi->s_dirs_counter,
3510 ext4_count_dirs(sb));
3511 percpu_counter_set(&sbi->s_dirtyblocks_counter, 0);
206f7ab4 3512
ce7e010a 3513no_journal:
fd89d5f2
TH		 3514 /*
3515 * The maximum number of concurrent works can be high and
3516 * concurrency isn't really necessary. Limit it to 1.
3517 */
3518 EXT4_SB(sb)->dio_unwritten_wq =
3519 alloc_workqueue("ext4-dio-unwritten", WQ_MEM_RECLAIM, 1);
4c0425ff
MC		 3520 if (!EXT4_SB(sb)->dio_unwritten_wq) {
3521 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
3522 goto failed_mount_wq;
3523 }
3524
ac27a0ec 3525 /*
dab291af 3526 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 3527 * so we can safely mount the rest of the filesystem now.
3528 */
3529
1d1fe1ee
DH		 3530 root = ext4_iget(sb, EXT4_ROOT_INO);
3531 if (IS_ERR(root)) {
b31e1552 3532 ext4_msg(sb, KERN_ERR, "get root inode failed");
1d1fe1ee 3533 ret = PTR_ERR(root);
ac27a0ec
DK		 3534 goto failed_mount4;
3535 }
3536 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1d1fe1ee 3537 iput(root);
b31e1552 3538 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
ac27a0ec
DK		 3539 goto failed_mount4;
3540 }
1d1fe1ee
DH		 3541 sb->s_root = d_alloc_root(root);
3542 if (!sb->s_root) {
b31e1552 3543 ext4_msg(sb, KERN_ERR, "get root dentry failed");
1d1fe1ee
DH		 3544 iput(root);
3545 ret = -ENOMEM;
3546 goto failed_mount4;
3547 }
ac27a0ec 3548
2b2d6d01 3549 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
ef7f3835
KS		 3550
3551 /* determine the minimum size of new large inodes, if present */
3552 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3553 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3554 EXT4_GOOD_OLD_INODE_SIZE;
3555 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3556 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
3557 if (sbi->s_want_extra_isize <
3558 le16_to_cpu(es->s_want_extra_isize))
3559 sbi->s_want_extra_isize =
3560 le16_to_cpu(es->s_want_extra_isize);
3561 if (sbi->s_want_extra_isize <
3562 le16_to_cpu(es->s_min_extra_isize))
3563 sbi->s_want_extra_isize =
3564 le16_to_cpu(es->s_min_extra_isize);
3565 }
3566 }
3567 /* Check if enough inode space is available */
3568 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3569 sbi->s_inode_size) {
3570 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3571 EXT4_GOOD_OLD_INODE_SIZE;
b31e1552
ES		 3572 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3573 "available");
ef7f3835
KS		 3574 }
3575
90576c0b
TT		 3576 if (test_opt(sb, DELALLOC) &&
3577 (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)) {
b31e1552
ES		 3578 ext4_msg(sb, KERN_WARNING, "Ignoring delalloc option - "
3579 "requested data journaling mode");
fd8c37ec 3580 clear_opt(sb, DELALLOC);
90576c0b 3581 }
744692dc
JZ		 3582 if (test_opt(sb, DIOREAD_NOLOCK)) {
3583 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3584 ext4_msg(sb, KERN_WARNING, "Ignoring dioread_nolock "
3585 "option - requested data journaling mode");
fd8c37ec 3586 clear_opt(sb, DIOREAD_NOLOCK);
744692dc
JZ		 3587 }
3588 if (sb->s_blocksize < PAGE_SIZE) {
3589 ext4_msg(sb, KERN_WARNING, "Ignoring dioread_nolock "
3590 "option - block size is too small");
fd8c37ec 3591 clear_opt(sb, DIOREAD_NOLOCK);
744692dc
JZ		 3592 }
3593 }
c2774d84 3594
6fd058f7
TT		 3595 err = ext4_setup_system_zone(sb);
3596 if (err) {
b31e1552 3597 ext4_msg(sb, KERN_ERR, "failed to initialize system "
fbe845dd 3598 "zone (%d)", err);
6fd058f7
TT		 3599 goto failed_mount4;
3600 }
3601
c2774d84
AK		 3602 ext4_ext_init(sb);
3603 err = ext4_mb_init(sb, needs_recovery);
3604 if (err) {
421f91d2 3605 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
b31e1552 3606 err);
c2774d84
AK		 3607 goto failed_mount4;
3608 }
3609
bfff6873
LC		 3610 err = ext4_register_li_request(sb, first_not_zeroed);
3611 if (err)
3612 goto failed_mount4;
3613
3197ebdb
TT		 3614 sbi->s_kobj.kset = ext4_kset;
3615 init_completion(&sbi->s_kobj_unregister);
3616 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
3617 "%s", sb->s_id);
3618 if (err) {
3619 ext4_mb_release(sb);
3620 ext4_ext_release(sb);
3621 goto failed_mount4;
3622 };
3623
617ba13b
MC		 3624 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3625 ext4_orphan_cleanup(sb, es);
3626 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
0390131b 3627 if (needs_recovery) {
b31e1552 3628 ext4_msg(sb, KERN_INFO, "recovery complete");
0390131b
FM		 3629 ext4_mark_recovery_complete(sb, es);
3630 }
3631 if (EXT4_SB(sb)->s_journal) {
3632 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3633 descr = " journalled data mode";
3634 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3635 descr = " ordered data mode";
3636 else
3637 descr = " writeback data mode";
3638 } else
3639 descr = "out journal";
3640
d4c402d9 3641 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
8b67f04a
TT		 3642 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
3643 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
ac27a0ec 3644
66e61a9e
TT		 3645 init_timer(&sbi->s_err_report);
3646 sbi->s_err_report.function = print_daily_error_info;
3647 sbi->s_err_report.data = (unsigned long) sb;
3648 if (es->s_error_count)
3649 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
ac27a0ec 3650
d4c402d9 3651 kfree(orig_data);
ac27a0ec
DK		 3652 return 0;
3653
617ba13b 3654cantfind_ext4:
ac27a0ec 3655 if (!silent)
b31e1552 3656 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
ac27a0ec
DK		 3657 goto failed_mount;
3658
3659failed_mount4:
b31e1552 3660 ext4_msg(sb, KERN_ERR, "mount failed");
4c0425ff
MC		 3661 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
3662failed_mount_wq:
6fd058f7 3663 ext4_release_system_zone(sb);
0390131b
FM		 3664 if (sbi->s_journal) {
3665 jbd2_journal_destroy(sbi->s_journal);
3666 sbi->s_journal = NULL;
3667 }
ac27a0ec 3668failed_mount3:
c5ca7c76
TT		 3669 if (sbi->s_flex_groups) {
3670 if (is_vmalloc_addr(sbi->s_flex_groups))
3671 vfree(sbi->s_flex_groups);
3672 else
3673 kfree(sbi->s_flex_groups);
3674 }
ce7e010a
TT		 3675 percpu_counter_destroy(&sbi->s_freeblocks_counter);
3676 percpu_counter_destroy(&sbi->s_freeinodes_counter);
3677 percpu_counter_destroy(&sbi->s_dirs_counter);
3678 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
ac27a0ec
DK		 3679failed_mount2:
3680 for (i = 0; i < db_count; i++)
3681 brelse(sbi->s_group_desc[i]);
3682 kfree(sbi->s_group_desc);
3683failed_mount:
240799cd 3684 if (sbi->s_proc) {
9f6200bb 3685 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 3686 }
ac27a0ec
DK		 3687#ifdef CONFIG_QUOTA
3688 for (i = 0; i < MAXQUOTAS; i++)
3689 kfree(sbi->s_qf_names[i]);
3690#endif
617ba13b 3691 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 3692 brelse(bh);
3693out_fail:
3694 sb->s_fs_info = NULL;
f6830165 3695 kfree(sbi->s_blockgroup_lock);
ac27a0ec 3696 kfree(sbi);
dcc7dae3 3697out_free_orig:
d4c402d9 3698 kfree(orig_data);
1d1fe1ee 3699 return ret;
ac27a0ec
DK		 3700}
3701
3702/*
3703 * Setup any per-fs journal parameters now. We'll do this both on
3704 * initial mount, once the journal has been initialised but before we've
3705 * done any recovery; and again on any subsequent remount.
3706 */
617ba13b 3707static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 3708{
617ba13b 3709 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 3710
30773840
TT		 3711 journal->j_commit_interval = sbi->s_commit_interval;
3712 journal->j_min_batch_time = sbi->s_min_batch_time;
3713 journal->j_max_batch_time = sbi->s_max_batch_time;
ac27a0ec 3714
a931da6a 3715 write_lock(&journal->j_state_lock);
ac27a0ec 3716 if (test_opt(sb, BARRIER))
dab291af 3717 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 3718 else
dab291af 3719 journal->j_flags &= ~JBD2_BARRIER;
5bf5683a
HK		 3720 if (test_opt(sb, DATA_ERR_ABORT))
3721 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
3722 else
3723 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
a931da6a 3724 write_unlock(&journal->j_state_lock);
ac27a0ec
DK		 3725}
3726
617ba13b 3727static journal_t *ext4_get_journal(struct super_block *sb,
ac27a0ec
DK		 3728 unsigned int journal_inum)
3729{
3730 struct inode *journal_inode;
3731 journal_t *journal;
3732
0390131b
FM		 3733 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3734
ac27a0ec
DK		 3735 /* First, test for the existence of a valid inode on disk. Bad
3736 * things happen if we iget() an unused inode, as the subsequent
3737 * iput() will try to delete it. */
3738
1d1fe1ee
DH		 3739 journal_inode = ext4_iget(sb, journal_inum);
3740 if (IS_ERR(journal_inode)) {
b31e1552 3741 ext4_msg(sb, KERN_ERR, "no journal found");
ac27a0ec
DK		 3742 return NULL;
3743 }
3744 if (!journal_inode->i_nlink) {
3745 make_bad_inode(journal_inode);
3746 iput(journal_inode);
b31e1552 3747 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
ac27a0ec
DK		 3748 return NULL;
3749 }
3750
e5f8eab8 3751 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
ac27a0ec 3752 journal_inode, journal_inode->i_size);
1d1fe1ee 3753 if (!S_ISREG(journal_inode->i_mode)) {
b31e1552 3754 ext4_msg(sb, KERN_ERR, "invalid journal inode");
ac27a0ec
DK		 3755 iput(journal_inode);
3756 return NULL;
3757 }
3758
dab291af 3759 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 3760 if (!journal) {
b31e1552 3761 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
ac27a0ec
DK		 3762 iput(journal_inode);
3763 return NULL;
3764 }
3765 journal->j_private = sb;
617ba13b 3766 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 3767 return journal;
3768}
3769
617ba13b 3770static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 3771 dev_t j_dev)
3772{
2b2d6d01 3773 struct buffer_head *bh;
ac27a0ec 3774 journal_t *journal;
617ba13b
MC		 3775 ext4_fsblk_t start;
3776 ext4_fsblk_t len;
ac27a0ec 3777 int hblock, blocksize;
617ba13b 3778 ext4_fsblk_t sb_block;
ac27a0ec 3779 unsigned long offset;
2b2d6d01 3780 struct ext4_super_block *es;
ac27a0ec
DK		 3781 struct block_device *bdev;
3782
0390131b
FM		 3783 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3784
b31e1552 3785 bdev = ext4_blkdev_get(j_dev, sb);
ac27a0ec
DK		 3786 if (bdev == NULL)
3787 return NULL;
3788
ac27a0ec 3789 blocksize = sb->s_blocksize;
e1defc4f 3790 hblock = bdev_logical_block_size(bdev);
ac27a0ec 3791 if (blocksize < hblock) {
b31e1552
ES		 3792 ext4_msg(sb, KERN_ERR,
3793 "blocksize too small for journal device");
ac27a0ec
DK		 3794 goto out_bdev;
3795 }
3796
617ba13b
MC		 3797 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3798 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 3799 set_blocksize(bdev, blocksize);
3800 if (!(bh = __bread(bdev, sb_block, blocksize))) {
b31e1552
ES		 3801 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
3802 "external journal");
ac27a0ec
DK		 3803 goto out_bdev;
3804 }
3805
617ba13b
MC		 3806 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3807 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 3808 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b 3809 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
b31e1552
ES		 3810 ext4_msg(sb, KERN_ERR, "external journal has "
3811 "bad superblock");
ac27a0ec
DK		 3812 brelse(bh);
3813 goto out_bdev;
3814 }
3815
617ba13b 3816 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
b31e1552 3817 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
ac27a0ec
DK		 3818 brelse(bh);
3819 goto out_bdev;
3820 }
3821
bd81d8ee 3822 len = ext4_blocks_count(es);
ac27a0ec
DK		 3823 start = sb_block + 1;
3824 brelse(bh); /* we're done with the superblock */
3825
dab291af 3826 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 3827 start, len, blocksize);
3828 if (!journal) {
b31e1552 3829 ext4_msg(sb, KERN_ERR, "failed to create device journal");
ac27a0ec
DK		 3830 goto out_bdev;
3831 }
3832 journal->j_private = sb;
3833 ll_rw_block(READ, 1, &journal->j_sb_buffer);
3834 wait_on_buffer(journal->j_sb_buffer);
3835 if (!buffer_uptodate(journal->j_sb_buffer)) {
b31e1552 3836 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
ac27a0ec
DK		 3837 goto out_journal;
3838 }
3839 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
b31e1552
ES		 3840 ext4_msg(sb, KERN_ERR, "External journal has more than one "
3841 "user (unsupported) - %d",
ac27a0ec
DK		 3842 be32_to_cpu(journal->j_superblock->s_nr_users));
3843 goto out_journal;
3844 }
617ba13b
MC		 3845 EXT4_SB(sb)->journal_bdev = bdev;
3846 ext4_init_journal_params(sb, journal);
ac27a0ec 3847 return journal;
0b8e58a1 3848
ac27a0ec 3849out_journal:
dab291af 3850 jbd2_journal_destroy(journal);
ac27a0ec 3851out_bdev:
617ba13b 3852 ext4_blkdev_put(bdev);
ac27a0ec
DK		 3853 return NULL;
3854}
3855
617ba13b
MC		 3856static int ext4_load_journal(struct super_block *sb,
3857 struct ext4_super_block *es,
ac27a0ec
DK		 3858 unsigned long journal_devnum)
3859{
3860 journal_t *journal;
3861 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
3862 dev_t journal_dev;
3863 int err = 0;
3864 int really_read_only;
3865
0390131b
FM		 3866 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3867
ac27a0ec
DK		 3868 if (journal_devnum &&
3869 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
b31e1552
ES		 3870 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
3871 "numbers have changed");
ac27a0ec
DK		 3872 journal_dev = new_decode_dev(journal_devnum);
3873 } else
3874 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
3875
3876 really_read_only = bdev_read_only(sb->s_bdev);
3877
3878 /*
3879 * Are we loading a blank journal or performing recovery after a
3880 * crash? For recovery, we need to check in advance whether we
3881 * can get read-write access to the device.
3882 */
617ba13b 3883 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
ac27a0ec 3884 if (sb->s_flags & MS_RDONLY) {
b31e1552
ES		 3885 ext4_msg(sb, KERN_INFO, "INFO: recovery "
3886 "required on readonly filesystem");
ac27a0ec 3887 if (really_read_only) {
b31e1552
ES		 3888 ext4_msg(sb, KERN_ERR, "write access "
3889 "unavailable, cannot proceed");
ac27a0ec
DK		 3890 return -EROFS;
3891 }
b31e1552
ES		 3892 ext4_msg(sb, KERN_INFO, "write access will "
3893 "be enabled during recovery");
ac27a0ec
DK		 3894 }
3895 }
3896
3897 if (journal_inum && journal_dev) {
b31e1552
ES		 3898 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
3899 "and inode journals!");
ac27a0ec
DK		 3900 return -EINVAL;
3901 }
3902
3903 if (journal_inum) {
617ba13b 3904 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 3905 return -EINVAL;
3906 } else {
617ba13b 3907 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 3908 return -EINVAL;
3909 }
3910
90576c0b 3911 if (!(journal->j_flags & JBD2_BARRIER))
b31e1552 3912 ext4_msg(sb, KERN_INFO, "barriers disabled");
4776004f 3913
ac27a0ec 3914 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
dab291af 3915 err = jbd2_journal_update_format(journal);
ac27a0ec 3916 if (err) {
b31e1552 3917 ext4_msg(sb, KERN_ERR, "error updating journal");
dab291af 3918 jbd2_journal_destroy(journal);
ac27a0ec
DK		 3919 return err;
3920 }
3921 }
3922
617ba13b 3923 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
dab291af 3924 err = jbd2_journal_wipe(journal, !really_read_only);
1c13d5c0
TT		 3925 if (!err) {
3926 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
3927 if (save)
3928 memcpy(save, ((char *) es) +
3929 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
dab291af 3930 err = jbd2_journal_load(journal);
1c13d5c0
TT		 3931 if (save)
3932 memcpy(((char *) es) + EXT4_S_ERR_START,
3933 save, EXT4_S_ERR_LEN);
3934 kfree(save);
3935 }
ac27a0ec
DK		 3936
3937 if (err) {
b31e1552 3938 ext4_msg(sb, KERN_ERR, "error loading journal");
dab291af 3939 jbd2_journal_destroy(journal);
ac27a0ec
DK		 3940 return err;
3941 }
3942
617ba13b
MC		 3943 EXT4_SB(sb)->s_journal = journal;
3944 ext4_clear_journal_err(sb, es);
ac27a0ec 3945
c41303ce 3946 if (!really_read_only && journal_devnum &&
ac27a0ec
DK		 3947 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3948 es->s_journal_dev = cpu_to_le32(journal_devnum);
ac27a0ec
DK		 3949
3950 /* Make sure we flush the recovery flag to disk. */
e2d67052 3951 ext4_commit_super(sb, 1);
ac27a0ec
DK		 3952 }
3953
3954 return 0;
3955}
3956
e2d67052 3957static int ext4_commit_super(struct super_block *sb, int sync)
ac27a0ec 3958{
e2d67052 3959 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
617ba13b 3960 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
c4be0c1d 3961 int error = 0;
ac27a0ec
DK		 3962
3963 if (!sbh)
c4be0c1d 3964 return error;
914258bf
TT		 3965 if (buffer_write_io_error(sbh)) {
3966 /*
3967 * Oh, dear. A previous attempt to write the
3968 * superblock failed. This could happen because the
3969 * USB device was yanked out. Or it could happen to
3970 * be a transient write error and maybe the block will
3971 * be remapped. Nothing we can do but to retry the
3972 * write and hope for the best.
3973 */
b31e1552
ES		 3974 ext4_msg(sb, KERN_ERR, "previous I/O error to "
3975 "superblock detected");
914258bf
TT		 3976 clear_buffer_write_io_error(sbh);
3977 set_buffer_uptodate(sbh);
3978 }
71290b36
TT		 3979 /*
3980 * If the file system is mounted read-only, don't update the
3981 * superblock write time. This avoids updating the superblock
3982 * write time when we are mounting the root file system
3983 * read/only but we need to replay the journal; at that point,
3984 * for people who are east of GMT and who make their clock
3985 * tick in localtime for Windows bug-for-bug compatibility,
3986 * the clock is set in the future, and this will cause e2fsck
3987 * to complain and force a full file system check.
3988 */
3989 if (!(sb->s_flags & MS_RDONLY))
3990 es->s_wtime = cpu_to_le32(get_seconds());
f613dfcb
TT		 3991 if (sb->s_bdev->bd_part)
3992 es->s_kbytes_written =
3993 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
afc32f7e
TT		 3994 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
3995 EXT4_SB(sb)->s_sectors_written_start) >> 1));
f613dfcb
TT		 3996 else
3997 es->s_kbytes_written =
3998 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
ce7e010a
TT		 3999 ext4_free_blocks_count_set(es, percpu_counter_sum_positive(
4000 &EXT4_SB(sb)->s_freeblocks_counter));
4001 es->s_free_inodes_count =
4002 cpu_to_le32(percpu_counter_sum_positive(
4003 &EXT4_SB(sb)->s_freeinodes_counter));
7234ab2a 4004 sb->s_dirt = 0;
ac27a0ec
DK		 4005 BUFFER_TRACE(sbh, "marking dirty");
4006 mark_buffer_dirty(sbh);
914258bf 4007 if (sync) {
c4be0c1d
TS		 4008 error = sync_dirty_buffer(sbh);
4009 if (error)
4010 return error;
4011
4012 error = buffer_write_io_error(sbh);
4013 if (error) {
b31e1552
ES		 4014 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4015 "superblock");
914258bf
TT		 4016 clear_buffer_write_io_error(sbh);
4017 set_buffer_uptodate(sbh);
4018 }
4019 }
c4be0c1d 4020 return error;
ac27a0ec
DK		 4021}
4022
ac27a0ec
DK		 4023/*
4024 * Have we just finished recovery? If so, and if we are mounting (or
4025 * remounting) the filesystem readonly, then we will end up with a
4026 * consistent fs on disk. Record that fact.
4027 */
2b2d6d01
TT		 4028static void ext4_mark_recovery_complete(struct super_block *sb,
4029 struct ext4_super_block *es)
ac27a0ec 4030{
617ba13b 4031 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 4032
0390131b
FM		 4033 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4034 BUG_ON(journal != NULL);
4035 return;
4036 }
dab291af 4037 jbd2_journal_lock_updates(journal);
7ffe1ea8
HK		 4038 if (jbd2_journal_flush(journal) < 0)
4039 goto out;
4040
617ba13b 4041 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
ac27a0ec 4042 sb->s_flags & MS_RDONLY) {
617ba13b 4043 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
e2d67052 4044 ext4_commit_super(sb, 1);
ac27a0ec 4045 }
7ffe1ea8
HK		 4046
4047out:
dab291af 4048 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 4049}
4050
4051/*
4052 * If we are mounting (or read-write remounting) a filesystem whose journal
4053 * has recorded an error from a previous lifetime, move that error to the
4054 * main filesystem now.
4055 */
2b2d6d01
TT		 4056static void ext4_clear_journal_err(struct super_block *sb,
4057 struct ext4_super_block *es)
ac27a0ec
DK		 4058{
4059 journal_t *journal;
4060 int j_errno;
4061 const char *errstr;
4062
0390131b
FM		 4063 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4064
617ba13b 4065 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 4066
4067 /*
4068 * Now check for any error status which may have been recorded in the
617ba13b 4069 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 4070 */
4071
dab291af 4072 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 4073 if (j_errno) {
4074 char nbuf[16];
4075
617ba13b 4076 errstr = ext4_decode_error(sb, j_errno, nbuf);
12062ddd 4077 ext4_warning(sb, "Filesystem error recorded "
ac27a0ec 4078 "from previous mount: %s", errstr);
12062ddd 4079 ext4_warning(sb, "Marking fs in need of filesystem check.");
ac27a0ec 4080
617ba13b
MC		 4081 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4082 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
e2d67052 4083 ext4_commit_super(sb, 1);
ac27a0ec 4084
dab291af 4085 jbd2_journal_clear_err(journal);
ac27a0ec
DK		 4086 }
4087}
4088
4089/*
4090 * Force the running and committing transactions to commit,
4091 * and wait on the commit.
4092 */
617ba13b 4093int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 4094{
4095 journal_t *journal;
0390131b 4096 int ret = 0;
ac27a0ec
DK		 4097
4098 if (sb->s_flags & MS_RDONLY)
4099 return 0;
4100
617ba13b 4101 journal = EXT4_SB(sb)->s_journal;
6b0310fb 4102 if (journal) {
437f88cc 4103 vfs_check_frozen(sb, SB_FREEZE_TRANS);
0390131b 4104 ret = ext4_journal_force_commit(journal);
6b0310fb 4105 }
0390131b 4106
ac27a0ec
DK		 4107 return ret;
4108}
4109
2b2d6d01 4110static void ext4_write_super(struct super_block *sb)
ac27a0ec 4111{
ebc1ac16 4112 lock_super(sb);
9ca92389 4113 ext4_commit_super(sb, 1);
ebc1ac16 4114 unlock_super(sb);
ac27a0ec
DK		 4115}
4116
617ba13b 4117static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec 4118{
14ce0cb4 4119 int ret = 0;
9eddacf9 4120 tid_t target;
8d5d02e6 4121 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4122
9bffad1e 4123 trace_ext4_sync_fs(sb, wait);
8d5d02e6
MC		 4124 flush_workqueue(sbi->dio_unwritten_wq);
4125 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
9ca92389 4126 if (wait)
8d5d02e6 4127 jbd2_log_wait_commit(sbi->s_journal, target);
0390131b 4128 }
14ce0cb4 4129 return ret;
ac27a0ec
DK		 4130}
4131
4132/*
4133 * LVM calls this function before a (read-only) snapshot is created. This
4134 * gives us a chance to flush the journal completely and mark the fs clean.
4135 */
c4be0c1d 4136static int ext4_freeze(struct super_block *sb)
ac27a0ec 4137{
c4be0c1d
TS		 4138 int error = 0;
4139 journal_t *journal;
ac27a0ec 4140
9ca92389
TT		 4141 if (sb->s_flags & MS_RDONLY)
4142 return 0;
ac27a0ec 4143
9ca92389 4144 journal = EXT4_SB(sb)->s_journal;
7ffe1ea8 4145
9ca92389
TT		 4146 /* Now we set up the journal barrier. */
4147 jbd2_journal_lock_updates(journal);
ac27a0ec 4148
9ca92389
TT		 4149 /*
4150 * Don't clear the needs_recovery flag if we failed to flush
4151 * the journal.
4152 */
4153 error = jbd2_journal_flush(journal);
6b0310fb
ES		 4154 if (error < 0)
4155 goto out;
9ca92389
TT		 4156
4157 /* Journal blocked and flushed, clear needs_recovery flag. */
4158 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4159 error = ext4_commit_super(sb, 1);
6b0310fb
ES		 4160out:
4161 /* we rely on s_frozen to stop further updates */
4162 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4163 return error;
ac27a0ec
DK		 4164}
4165
4166/*
4167 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4168 * flag here, even though the filesystem is not technically dirty yet.
4169 */
c4be0c1d 4170static int ext4_unfreeze(struct super_block *sb)
ac27a0ec 4171{
9ca92389
TT		 4172 if (sb->s_flags & MS_RDONLY)
4173 return 0;
4174
4175 lock_super(sb);
4176 /* Reset the needs_recovery flag before the fs is unlocked. */
4177 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4178 ext4_commit_super(sb, 1);
4179 unlock_super(sb);
c4be0c1d 4180 return 0;
ac27a0ec
DK		 4181}
4182
673c6100
TT		 4183/*
4184 * Structure to save mount options for ext4_remount's benefit
4185 */
4186struct ext4_mount_options {
4187 unsigned long s_mount_opt;
a2595b8a 4188 unsigned long s_mount_opt2;
673c6100
TT		 4189 uid_t s_resuid;
4190 gid_t s_resgid;
4191 unsigned long s_commit_interval;
4192 u32 s_min_batch_time, s_max_batch_time;
4193#ifdef CONFIG_QUOTA
4194 int s_jquota_fmt;
4195 char *s_qf_names[MAXQUOTAS];
4196#endif
4197};
4198
2b2d6d01 4199static int ext4_remount(struct super_block *sb, int *flags, char *data)
ac27a0ec 4200{
2b2d6d01 4201 struct ext4_super_block *es;
617ba13b
MC		 4202 struct ext4_sb_info *sbi = EXT4_SB(sb);
4203 ext4_fsblk_t n_blocks_count = 0;
ac27a0ec 4204 unsigned long old_sb_flags;
617ba13b 4205 struct ext4_mount_options old_opts;
c79d967d 4206 int enable_quota = 0;
8a266467 4207 ext4_group_t g;
b3881f74 4208 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
ac27a0ec
DK		 4209 int err;
4210#ifdef CONFIG_QUOTA
4211 int i;
4212#endif
d4c402d9 4213 char *orig_data = kstrdup(data, GFP_KERNEL);
ac27a0ec
DK		 4214
4215 /* Store the original options */
bbd6851a 4216 lock_super(sb);
ac27a0ec
DK		 4217 old_sb_flags = sb->s_flags;
4218 old_opts.s_mount_opt = sbi->s_mount_opt;
a2595b8a 4219 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
ac27a0ec
DK		 4220 old_opts.s_resuid = sbi->s_resuid;
4221 old_opts.s_resgid = sbi->s_resgid;
4222 old_opts.s_commit_interval = sbi->s_commit_interval;
30773840
TT		 4223 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4224 old_opts.s_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 4225#ifdef CONFIG_QUOTA
4226 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4227 for (i = 0; i < MAXQUOTAS; i++)
4228 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
4229#endif
b3881f74
TT		 4230 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4231 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
ac27a0ec
DK		 4232
4233 /*
4234 * Allow the "check" option to be passed as a remount option.
4235 */
b3881f74
TT		 4236 if (!parse_options(data, sb, NULL, &journal_ioprio,
4237 &n_blocks_count, 1)) {
ac27a0ec
DK		 4238 err = -EINVAL;
4239 goto restore_opts;
4240 }
4241
4ab2f15b 4242 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
c67d859e 4243 ext4_abort(sb, "Abort forced by user");
ac27a0ec
DK		 4244
4245 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 4246 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec
DK		 4247
4248 es = sbi->s_es;
4249
b3881f74 4250 if (sbi->s_journal) {
0390131b 4251 ext4_init_journal_params(sb, sbi->s_journal);
b3881f74
TT		 4252 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4253 }
ac27a0ec
DK		 4254
4255 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
bd81d8ee 4256 n_blocks_count > ext4_blocks_count(es)) {
4ab2f15b 4257 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
ac27a0ec
DK		 4258 err = -EROFS;
4259 goto restore_opts;
4260 }
4261
4262 if (*flags & MS_RDONLY) {
0f0dd62f
CH		 4263 err = dquot_suspend(sb, -1);
4264 if (err < 0)
c79d967d 4265 goto restore_opts;
c79d967d 4266
ac27a0ec
DK		 4267 /*
4268 * First of all, the unconditional stuff we have to do
4269 * to disable replay of the journal when we next remount
4270 */
4271 sb->s_flags |= MS_RDONLY;
4272
4273 /*
4274 * OK, test if we are remounting a valid rw partition
4275 * readonly, and if so set the rdonly flag and then
4276 * mark the partition as valid again.
4277 */
617ba13b
MC		 4278 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4279 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 4280 es->s_state = cpu_to_le16(sbi->s_mount_state);
4281
a63c9eb2 4282 if (sbi->s_journal)
0390131b 4283 ext4_mark_recovery_complete(sb, es);
ac27a0ec 4284 } else {
a13fb1a4
ES		 4285 /* Make sure we can mount this feature set readwrite */
4286 if (!ext4_feature_set_ok(sb, 0)) {
ac27a0ec
DK		 4287 err = -EROFS;
4288 goto restore_opts;
4289 }
8a266467
TT		 4290 /*
4291 * Make sure the group descriptor checksums
0b8e58a1 4292 * are sane. If they aren't, refuse to remount r/w.
8a266467
TT		 4293 */
4294 for (g = 0; g < sbi->s_groups_count; g++) {
4295 struct ext4_group_desc *gdp =
4296 ext4_get_group_desc(sb, g, NULL);
4297
4298 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
b31e1552
ES		 4299 ext4_msg(sb, KERN_ERR,
4300 "ext4_remount: Checksum for group %u failed (%u!=%u)",
8a266467
TT		 4301 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4302 le16_to_cpu(gdp->bg_checksum));
4303 err = -EINVAL;
4304 goto restore_opts;
4305 }
4306 }
4307
ead6596b
ES		 4308 /*
4309 * If we have an unprocessed orphan list hanging
4310 * around from a previously readonly bdev mount,
4311 * require a full umount/remount for now.
4312 */
4313 if (es->s_last_orphan) {
b31e1552 4314 ext4_msg(sb, KERN_WARNING, "Couldn't "
ead6596b
ES		 4315 "remount RDWR because of unprocessed "
4316 "orphan inode list. Please "
b31e1552 4317 "umount/remount instead");
ead6596b
ES		 4318 err = -EINVAL;
4319 goto restore_opts;
4320 }
4321
ac27a0ec
DK		 4322 /*
4323 * Mounting a RDONLY partition read-write, so reread
4324 * and store the current valid flag. (It may have
4325 * been changed by e2fsck since we originally mounted
4326 * the partition.)
4327 */
0390131b
FM		 4328 if (sbi->s_journal)
4329 ext4_clear_journal_err(sb, es);
ac27a0ec 4330 sbi->s_mount_state = le16_to_cpu(es->s_state);
617ba13b 4331 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
ac27a0ec 4332 goto restore_opts;
2b2d6d01 4333 if (!ext4_setup_super(sb, es, 0))
ac27a0ec 4334 sb->s_flags &= ~MS_RDONLY;
c79d967d 4335 enable_quota = 1;
ac27a0ec
DK		 4336 }
4337 }
bfff6873
LC		 4338
4339 /*
4340 * Reinitialize lazy itable initialization thread based on
4341 * current settings
4342 */
4343 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4344 ext4_unregister_li_request(sb);
4345 else {
4346 ext4_group_t first_not_zeroed;
4347 first_not_zeroed = ext4_has_uninit_itable(sb);
4348 ext4_register_li_request(sb, first_not_zeroed);
4349 }
4350
6fd058f7 4351 ext4_setup_system_zone(sb);
0390131b 4352 if (sbi->s_journal == NULL)
e2d67052 4353 ext4_commit_super(sb, 1);
0390131b 4354
ac27a0ec
DK		 4355#ifdef CONFIG_QUOTA
4356 /* Release old quota file names */
4357 for (i = 0; i < MAXQUOTAS; i++)
4358 if (old_opts.s_qf_names[i] &&
4359 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4360 kfree(old_opts.s_qf_names[i]);
4361#endif
bbd6851a 4362 unlock_super(sb);
c79d967d 4363 if (enable_quota)
0f0dd62f 4364 dquot_resume(sb, -1);
d4c402d9
CW		 4365
4366 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4367 kfree(orig_data);
ac27a0ec 4368 return 0;
0b8e58a1 4369
ac27a0ec
DK		 4370restore_opts:
4371 sb->s_flags = old_sb_flags;
4372 sbi->s_mount_opt = old_opts.s_mount_opt;
a2595b8a 4373 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
ac27a0ec
DK		 4374 sbi->s_resuid = old_opts.s_resuid;
4375 sbi->s_resgid = old_opts.s_resgid;
4376 sbi->s_commit_interval = old_opts.s_commit_interval;
30773840
TT		 4377 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4378 sbi->s_max_batch_time = old_opts.s_max_batch_time;
ac27a0ec
DK		 4379#ifdef CONFIG_QUOTA
4380 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4381 for (i = 0; i < MAXQUOTAS; i++) {
4382 if (sbi->s_qf_names[i] &&
4383 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4384 kfree(sbi->s_qf_names[i]);
4385 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4386 }
4387#endif
bbd6851a 4388 unlock_super(sb);
d4c402d9 4389 kfree(orig_data);
ac27a0ec
DK		 4390 return err;
4391}
4392
2b2d6d01 4393static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
ac27a0ec
DK		 4394{
4395 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 4396 struct ext4_sb_info *sbi = EXT4_SB(sb);
4397 struct ext4_super_block *es = sbi->s_es;
960cc398 4398 u64 fsid;
ac27a0ec 4399
5e70030d
BP		 4400 if (test_opt(sb, MINIX_DF)) {
4401 sbi->s_overhead_last = 0;
6bc9feff 4402 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
8df9675f 4403 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
5e70030d 4404 ext4_fsblk_t overhead = 0;
ac27a0ec
DK		 4405
4406 /*
5e70030d
BP		 4407 * Compute the overhead (FS structures). This is constant
4408 * for a given filesystem unless the number of block groups
4409 * changes so we cache the previous value until it does.
ac27a0ec
DK		 4410 */
4411
4412 /*
4413 * All of the blocks before first_data_block are
4414 * overhead
4415 */
4416 overhead = le32_to_cpu(es->s_first_data_block);
4417
4418 /*
4419 * Add the overhead attributed to the superblock and
4420 * block group descriptors. If the sparse superblocks
4421 * feature is turned on, then not all groups have this.
4422 */
4423 for (i = 0; i < ngroups; i++) {
617ba13b
MC		 4424 overhead += ext4_bg_has_super(sb, i) +
4425 ext4_bg_num_gdb(sb, i);
ac27a0ec
DK		 4426 cond_resched();
4427 }
4428
4429 /*
4430 * Every block group has an inode bitmap, a block
4431 * bitmap, and an inode table.
4432 */
5e70030d
BP		 4433 overhead += ngroups * (2 + sbi->s_itb_per_group);
4434 sbi->s_overhead_last = overhead;
4435 smp_wmb();
6bc9feff 4436 sbi->s_blocks_last = ext4_blocks_count(es);
ac27a0ec
DK		 4437 }
4438
617ba13b 4439 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 4440 buf->f_bsize = sb->s_blocksize;
5e70030d 4441 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
6bc6e63f
AK		 4442 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
4443 percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
bd81d8ee
LV		 4444 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
4445 if (buf->f_bfree < ext4_r_blocks_count(es))
ac27a0ec
DK		 4446 buf->f_bavail = 0;
4447 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 4448 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
617ba13b 4449 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 4450 fsid = le64_to_cpup((void *)es->s_uuid) ^
4451 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4452 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4453 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
0b8e58a1 4454
ac27a0ec
DK		 4455 return 0;
4456}
4457
0b8e58a1
AD		 4458/* Helper function for writing quotas on sync - we need to start transaction
4459 * before quota file is locked for write. Otherwise the are possible deadlocks:
ac27a0ec 4460 * Process 1 Process 2
617ba13b 4461 * ext4_create() quota_sync()
a269eb18 4462 * jbd2_journal_start() write_dquot()
871a2931 4463 * dquot_initialize() down(dqio_mutex)
dab291af 4464 * down(dqio_mutex) jbd2_journal_start()
ac27a0ec
DK		 4465 *
4466 */
4467
4468#ifdef CONFIG_QUOTA
4469
4470static inline struct inode *dquot_to_inode(struct dquot *dquot)
4471{
4472 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
4473}
4474
617ba13b 4475static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 4476{
4477 int ret, err;
4478 handle_t *handle;
4479 struct inode *inode;
4480
4481 inode = dquot_to_inode(dquot);
617ba13b 4482 handle = ext4_journal_start(inode,
0b8e58a1 4483 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 4484 if (IS_ERR(handle))
4485 return PTR_ERR(handle);
4486 ret = dquot_commit(dquot);
617ba13b 4487 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4488 if (!ret)
4489 ret = err;
4490 return ret;
4491}
4492
617ba13b 4493static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 4494{
4495 int ret, err;
4496 handle_t *handle;
4497
617ba13b 4498 handle = ext4_journal_start(dquot_to_inode(dquot),
0b8e58a1 4499 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 4500 if (IS_ERR(handle))
4501 return PTR_ERR(handle);
4502 ret = dquot_acquire(dquot);
617ba13b 4503 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4504 if (!ret)
4505 ret = err;
4506 return ret;
4507}
4508
617ba13b 4509static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 4510{
4511 int ret, err;
4512 handle_t *handle;
4513
617ba13b 4514 handle = ext4_journal_start(dquot_to_inode(dquot),
0b8e58a1 4515 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 4516 if (IS_ERR(handle)) {
4517 /* Release dquot anyway to avoid endless cycle in dqput() */
4518 dquot_release(dquot);
ac27a0ec 4519 return PTR_ERR(handle);
9c3013e9 4520 }
ac27a0ec 4521 ret = dquot_release(dquot);
617ba13b 4522 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4523 if (!ret)
4524 ret = err;
4525 return ret;
4526}
4527
617ba13b 4528static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec 4529{
2c8be6b2 4530 /* Are we journaling quotas? */
617ba13b
MC		 4531 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
4532 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
ac27a0ec 4533 dquot_mark_dquot_dirty(dquot);
617ba13b 4534 return ext4_write_dquot(dquot);
ac27a0ec
DK		 4535 } else {
4536 return dquot_mark_dquot_dirty(dquot);
4537 }
4538}
4539
617ba13b 4540static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 4541{
4542 int ret, err;
4543 handle_t *handle;
4544
4545 /* Data block + inode block */
617ba13b 4546 handle = ext4_journal_start(sb->s_root->d_inode, 2);
ac27a0ec
DK		 4547 if (IS_ERR(handle))
4548 return PTR_ERR(handle);
4549 ret = dquot_commit_info(sb, type);
617ba13b 4550 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4551 if (!ret)
4552 ret = err;
4553 return ret;
4554}
4555
4556/*
4557 * Turn on quotas during mount time - we need to find
4558 * the quota file and such...
4559 */
617ba13b 4560static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 4561{
287a8095
CH		 4562 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4563 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 4564}
4565
4566/*
4567 * Standard function to be called on quota_on
4568 */
617ba13b 4569static int ext4_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 4570 struct path *path)
ac27a0ec
DK		 4571{
4572 int err;
ac27a0ec
DK		 4573
4574 if (!test_opt(sb, QUOTA))
4575 return -EINVAL;
0623543b 4576
ac27a0ec 4577 /* Quotafile not on the same filesystem? */
f00c9e44 4578 if (path->mnt->mnt_sb != sb)
ac27a0ec 4579 return -EXDEV;
0623543b
JK		 4580 /* Journaling quota? */
4581 if (EXT4_SB(sb)->s_qf_names[type]) {
2b2d6d01 4582 /* Quotafile not in fs root? */
f00c9e44 4583 if (path->dentry->d_parent != sb->s_root)
b31e1552
ES		 4584 ext4_msg(sb, KERN_WARNING,
4585 "Quota file not on filesystem root. "
4586 "Journaled quota will not work");
2b2d6d01 4587 }
0623543b
JK		 4588
4589 /*
4590 * When we journal data on quota file, we have to flush journal to see
4591 * all updates to the file when we bypass pagecache...
4592 */
0390131b 4593 if (EXT4_SB(sb)->s_journal &&
f00c9e44 4594 ext4_should_journal_data(path->dentry->d_inode)) {
0623543b
JK		 4595 /*
4596 * We don't need to lock updates but journal_flush() could
4597 * otherwise be livelocked...
4598 */
4599 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 4600 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
0623543b 4601 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
f00c9e44 4602 if (err)
7ffe1ea8 4603 return err;
0623543b
JK		 4604 }
4605
f00c9e44 4606 return dquot_quota_on(sb, type, format_id, path);
ac27a0ec
DK		 4607}
4608
ca0e05e4
DM		 4609static int ext4_quota_off(struct super_block *sb, int type)
4610{
87009d86
DM		 4611 /* Force all delayed allocation blocks to be allocated.
4612 * Caller already holds s_umount sem */
4613 if (test_opt(sb, DELALLOC))
ca0e05e4 4614 sync_filesystem(sb);
ca0e05e4
DM		 4615
4616 return dquot_quota_off(sb, type);
4617}
4618
ac27a0ec
DK		 4619/* Read data from quotafile - avoid pagecache and such because we cannot afford
4620 * acquiring the locks... As quota files are never truncated and quota code
4621 * itself serializes the operations (and noone else should touch the files)
4622 * we don't have to be afraid of races */
617ba13b 4623static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 4624 size_t len, loff_t off)
4625{
4626 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 4627 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 4628 int err = 0;
4629 int offset = off & (sb->s_blocksize - 1);
4630 int tocopy;
4631 size_t toread;
4632 struct buffer_head *bh;
4633 loff_t i_size = i_size_read(inode);
4634
4635 if (off > i_size)
4636 return 0;
4637 if (off+len > i_size)
4638 len = i_size-off;
4639 toread = len;
4640 while (toread > 0) {
4641 tocopy = sb->s_blocksize - offset < toread ?
4642 sb->s_blocksize - offset : toread;
617ba13b 4643 bh = ext4_bread(NULL, inode, blk, 0, &err);
ac27a0ec
DK		 4644 if (err)
4645 return err;
4646 if (!bh) /* A hole? */
4647 memset(data, 0, tocopy);
4648 else
4649 memcpy(data, bh->b_data+offset, tocopy);
4650 brelse(bh);
4651 offset = 0;
4652 toread -= tocopy;
4653 data += tocopy;
4654 blk++;
4655 }
4656 return len;
4657}
4658
4659/* Write to quotafile (we know the transaction is already started and has
4660 * enough credits) */
617ba13b 4661static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 4662 const char *data, size_t len, loff_t off)
4663{
4664 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 4665 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 4666 int err = 0;
4667 int offset = off & (sb->s_blocksize - 1);
ac27a0ec
DK		 4668 struct buffer_head *bh;
4669 handle_t *handle = journal_current_handle();
4670
0390131b 4671 if (EXT4_SB(sb)->s_journal && !handle) {
b31e1552
ES		 4672 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4673 " cancelled because transaction is not started",
9c3013e9
JK		 4674 (unsigned long long)off, (unsigned long long)len);
4675 return -EIO;
4676 }
67eeb568
DM		 4677 /*
4678 * Since we account only one data block in transaction credits,
4679 * then it is impossible to cross a block boundary.
4680 */
4681 if (sb->s_blocksize - offset < len) {
4682 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4683 " cancelled because not block aligned",
4684 (unsigned long long)off, (unsigned long long)len);
4685 return -EIO;
4686 }
4687
ac27a0ec 4688 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
67eeb568
DM		 4689 bh = ext4_bread(handle, inode, blk, 1, &err);
4690 if (!bh)
4691 goto out;
62d2b5f2
JK		 4692 err = ext4_journal_get_write_access(handle, bh);
4693 if (err) {
4694 brelse(bh);
4695 goto out;
ac27a0ec 4696 }
67eeb568
DM		 4697 lock_buffer(bh);
4698 memcpy(bh->b_data+offset, data, len);
4699 flush_dcache_page(bh->b_page);
4700 unlock_buffer(bh);
62d2b5f2 4701 err = ext4_handle_dirty_metadata(handle, NULL, bh);
67eeb568 4702 brelse(bh);
ac27a0ec 4703out:
67eeb568 4704 if (err) {
4d04e4fb 4705 mutex_unlock(&inode->i_mutex);
ac27a0ec 4706 return err;
4d04e4fb 4707 }
67eeb568
DM		 4708 if (inode->i_size < off + len) {
4709 i_size_write(inode, off + len);
617ba13b 4710 EXT4_I(inode)->i_disksize = inode->i_size;
ac27a0ec 4711 }
ac27a0ec 4712 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
617ba13b 4713 ext4_mark_inode_dirty(handle, inode);
ac27a0ec 4714 mutex_unlock(&inode->i_mutex);
67eeb568 4715 return len;
ac27a0ec
DK		 4716}
4717
4718#endif
4719
152a0836
AV		 4720static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
4721 const char *dev_name, void *data)
ac27a0ec 4722{
152a0836 4723 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
ac27a0ec
DK		 4724}
4725
37f328eb 4726#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
24b58424
TT		 4727static struct file_system_type ext2_fs_type = {
4728 .owner = THIS_MODULE,
4729 .name = "ext2",
152a0836 4730 .mount = ext4_mount,
24b58424
TT		 4731 .kill_sb = kill_block_super,
4732 .fs_flags = FS_REQUIRES_DEV,
4733};
4734
4735static inline void register_as_ext2(void)
4736{
4737 int err = register_filesystem(&ext2_fs_type);
4738 if (err)
4739 printk(KERN_WARNING
4740 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
4741}
4742
4743static inline void unregister_as_ext2(void)
4744{
4745 unregister_filesystem(&ext2_fs_type);
4746}
51b7e3c9 4747MODULE_ALIAS("ext2");
24b58424
TT		 4748#else
4749static inline void register_as_ext2(void) { }
4750static inline void unregister_as_ext2(void) { }
4751#endif
4752
37f328eb 4753#if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
24b58424
TT		 4754static inline void register_as_ext3(void)
4755{
4756 int err = register_filesystem(&ext3_fs_type);
4757 if (err)
4758 printk(KERN_WARNING
4759 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
4760}
4761
4762static inline void unregister_as_ext3(void)
4763{
4764 unregister_filesystem(&ext3_fs_type);
4765}
51b7e3c9 4766MODULE_ALIAS("ext3");
24b58424
TT		 4767#else
4768static inline void register_as_ext3(void) { }
4769static inline void unregister_as_ext3(void) { }
4770#endif
4771
03010a33
TT		 4772static struct file_system_type ext4_fs_type = {
4773 .owner = THIS_MODULE,
4774 .name = "ext4",
152a0836 4775 .mount = ext4_mount,
03010a33
TT		 4776 .kill_sb = kill_block_super,
4777 .fs_flags = FS_REQUIRES_DEV,
4778};
4779
8f021222 4780static int __init ext4_init_feat_adverts(void)
857ac889
LC		 4781{
4782 struct ext4_features *ef;
4783 int ret = -ENOMEM;
4784
4785 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
4786 if (!ef)
4787 goto out;
4788
4789 ef->f_kobj.kset = ext4_kset;
4790 init_completion(&ef->f_kobj_unregister);
4791 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
4792 "features");
4793 if (ret) {
4794 kfree(ef);
4795 goto out;
4796 }
4797
4798 ext4_feat = ef;
4799 ret = 0;
4800out:
4801 return ret;
4802}
4803
8f021222
LC		 4804static void ext4_exit_feat_adverts(void)
4805{
4806 kobject_put(&ext4_feat->f_kobj);
4807 wait_for_completion(&ext4_feat->f_kobj_unregister);
4808 kfree(ext4_feat);
4809}
4810
e9e3bcec
ES		 4811/* Shared across all ext4 file systems */
4812wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
4813struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
4814
5dabfc78 4815static int __init ext4_init_fs(void)
ac27a0ec 4816{
e9e3bcec 4817 int i, err;
c9de560d 4818
12e9b892 4819 ext4_check_flag_values();
e9e3bcec
ES		 4820
4821 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
4822 mutex_init(&ext4__aio_mutex[i]);
4823 init_waitqueue_head(&ext4__ioend_wq[i]);
4824 }
4825
5dabfc78 4826 err = ext4_init_pageio();
6fd058f7
TT		 4827 if (err)
4828 return err;
5dabfc78 4829 err = ext4_init_system_zone();
bd2d0210 4830 if (err)
dd68314c 4831 goto out7;
3197ebdb
TT		 4832 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
4833 if (!ext4_kset)
dd68314c 4834 goto out6;
9f6200bb 4835 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
dd68314c
TT		 4836 if (!ext4_proc_root)
4837 goto out5;
857ac889
LC		 4838
4839 err = ext4_init_feat_adverts();
dd68314c
TT		 4840 if (err)
4841 goto out4;
857ac889 4842
5dabfc78 4843 err = ext4_init_mballoc();
ac27a0ec 4844 if (err)
6fd058f7 4845 goto out3;
c9de560d 4846
5dabfc78 4847 err = ext4_init_xattr();
c9de560d
AT		 4848 if (err)
4849 goto out2;
ac27a0ec
DK		 4850 err = init_inodecache();
4851 if (err)
4852 goto out1;
24b58424
TT		 4853 register_as_ext2();
4854 register_as_ext3();
03010a33 4855 err = register_filesystem(&ext4_fs_type);
ac27a0ec
DK		 4856 if (err)
4857 goto out;
bfff6873
LC		 4858
4859 ext4_li_info = NULL;
4860 mutex_init(&ext4_li_mtx);
ac27a0ec
DK		 4861 return 0;
4862out:
24b58424
TT		 4863 unregister_as_ext2();
4864 unregister_as_ext3();
ac27a0ec
DK		 4865 destroy_inodecache();
4866out1:
5dabfc78 4867 ext4_exit_xattr();
c9de560d 4868out2:
5dabfc78 4869 ext4_exit_mballoc();
6fd058f7 4870out3:
8f021222 4871 ext4_exit_feat_adverts();
dd68314c 4872out4:
6fd058f7 4873 remove_proc_entry("fs/ext4", NULL);
dd68314c 4874out5:
6fd058f7 4875 kset_unregister(ext4_kset);
dd68314c 4876out6:
5dabfc78 4877 ext4_exit_system_zone();
dd68314c 4878out7:
5dabfc78 4879 ext4_exit_pageio();
ac27a0ec
DK		 4880 return err;
4881}
4882
5dabfc78 4883static void __exit ext4_exit_fs(void)
ac27a0ec 4884{
bfff6873 4885 ext4_destroy_lazyinit_thread();
24b58424
TT		 4886 unregister_as_ext2();
4887 unregister_as_ext3();
03010a33 4888 unregister_filesystem(&ext4_fs_type);
ac27a0ec 4889 destroy_inodecache();
5dabfc78
TT		 4890 ext4_exit_xattr();
4891 ext4_exit_mballoc();
8f021222 4892 ext4_exit_feat_adverts();
9f6200bb 4893 remove_proc_entry("fs/ext4", NULL);
3197ebdb 4894 kset_unregister(ext4_kset);
5dabfc78
TT		 4895 ext4_exit_system_zone();
4896 ext4_exit_pageio();
ac27a0ec
DK		 4897}
4898
4899MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
83982b6f 4900MODULE_DESCRIPTION("Fourth Extended Filesystem");
ac27a0ec 4901MODULE_LICENSE("GPL");
5dabfc78
TT		 4902module_init(ext4_init_fs)
4903module_exit(ext4_exit_fs)
Empty description
This page took 1.458297 seconds and 4 git commands to generate.