projects / GitHub / mt8127 / android_kernel_alcatel_ttab.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
fs: Add FITRIM ioctl
[GitHub/mt8127/android_kernel_alcatel_ttab.git] / fs / ext4 / super.c
CommitLineData
ac27a0ec 1/*
617ba13b 2 * linux/fs/ext4/super.c
ac27a0ec
DK		 3 *
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
8 *
9 * from
10 *
11 * linux/fs/minix/inode.c
12 *
13 * Copyright (C) 1991, 1992 Linus Torvalds
14 *
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
17 */
18
19#include <linux/module.h>
20#include <linux/string.h>
21#include <linux/fs.h>
22#include <linux/time.h>
c5ca7c76 23#include <linux/vmalloc.h>
dab291af 24#include <linux/jbd2.h>
ac27a0ec
DK		 25#include <linux/slab.h>
26#include <linux/init.h>
27#include <linux/blkdev.h>
28#include <linux/parser.h>
29#include <linux/smp_lock.h>
30#include <linux/buffer_head.h>
a5694255 31#include <linux/exportfs.h>
ac27a0ec
DK		 32#include <linux/vfs.h>
33#include <linux/random.h>
34#include <linux/mount.h>
35#include <linux/namei.h>
36#include <linux/quotaops.h>
37#include <linux/seq_file.h>
9f6200bb 38#include <linux/proc_fs.h>
3197ebdb 39#include <linux/ctype.h>
1330593e 40#include <linux/log2.h>
717d50e4 41#include <linux/crc16.h>
ac27a0ec
DK		 42#include <asm/uaccess.h>
43
bfff6873
LC		 44#include <linux/kthread.h>
45#include <linux/freezer.h>
46
3dcf5451
CH		 47#include "ext4.h"
48#include "ext4_jbd2.h"
ac27a0ec
DK		 49#include "xattr.h"
50#include "acl.h"
3661d286 51#include "mballoc.h"
ac27a0ec 52
9bffad1e
TT		 53#define CREATE_TRACE_POINTS
54#include <trace/events/ext4.h>
55
9f6200bb 56struct proc_dir_entry *ext4_proc_root;
3197ebdb 57static struct kset *ext4_kset;
bfff6873
LC		 58struct ext4_lazy_init *ext4_li_info;
59struct mutex ext4_li_mtx;
857ac889 60struct ext4_features *ext4_feat;
9f6200bb 61
617ba13b 62static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
ac27a0ec 63 unsigned long journal_devnum);
e2d67052 64static int ext4_commit_super(struct super_block *sb, int sync);
2b2d6d01
TT		 65static void ext4_mark_recovery_complete(struct super_block *sb,
66 struct ext4_super_block *es);
67static void ext4_clear_journal_err(struct super_block *sb,
68 struct ext4_super_block *es);
617ba13b 69static int ext4_sync_fs(struct super_block *sb, int wait);
2b2d6d01 70static const char *ext4_decode_error(struct super_block *sb, int errno,
ac27a0ec 71 char nbuf[16]);
2b2d6d01
TT		 72static int ext4_remount(struct super_block *sb, int *flags, char *data);
73static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
c4be0c1d 74static int ext4_unfreeze(struct super_block *sb);
2b2d6d01 75static void ext4_write_super(struct super_block *sb);
c4be0c1d 76static int ext4_freeze(struct super_block *sb);
ba69f9ab
JK		 77static int ext4_get_sb(struct file_system_type *fs_type, int flags,
78 const char *dev_name, void *data, struct vfsmount *mnt);
bfff6873
LC		 79static void ext4_destroy_lazyinit_thread(void);
80static void ext4_unregister_li_request(struct super_block *sb);
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",
86 .get_sb = ext4_get_sb,
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
DK		 391{
392 va_list args;
393
394 va_start(args, fmt);
c398eda0
TT		 395 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: comm %s: ",
396 sb->s_id, function, line, current->comm);
ac27a0ec
DK		 397 vprintk(fmt, args);
398 printk("\n");
399 va_end(args);
400
617ba13b 401 ext4_handle_error(sb);
ac27a0ec
DK		 402}
403
c398eda0
TT		 404void ext4_error_inode(struct inode *inode, const char *function,
405 unsigned int line, ext4_fsblk_t block,
273df556
FM		 406 const char *fmt, ...)
407{
408 va_list args;
1c13d5c0 409 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
273df556 410
1c13d5c0
TT		 411 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
412 es->s_last_error_block = cpu_to_le64(block);
413 save_error_info(inode->i_sb, function, line);
273df556 414 va_start(args, fmt);
c398eda0
TT		 415 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: inode #%lu: ",
416 inode->i_sb->s_id, function, line, inode->i_ino);
417 if (block)
418 printk("block %llu: ", block);
419 printk("comm %s: ", current->comm);
273df556
FM		 420 vprintk(fmt, args);
421 printk("\n");
422 va_end(args);
423
424 ext4_handle_error(inode->i_sb);
425}
426
c398eda0
TT		 427void ext4_error_file(struct file *file, const char *function,
428 unsigned int line, const char *fmt, ...)
273df556
FM		 429{
430 va_list args;
1c13d5c0 431 struct ext4_super_block *es;
273df556
FM		 432 struct inode *inode = file->f_dentry->d_inode;
433 char pathname[80], *path;
434
1c13d5c0
TT		 435 es = EXT4_SB(inode->i_sb)->s_es;
436 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
437 save_error_info(inode->i_sb, function, line);
273df556
FM		 438 va_start(args, fmt);
439 path = d_path(&(file->f_path), pathname, sizeof(pathname));
440 if (!path)
441 path = "(unknown)";
442 printk(KERN_CRIT
c398eda0
TT		 443 "EXT4-fs error (device %s): %s:%d: inode #%lu "
444 "(comm %s path %s): ",
445 inode->i_sb->s_id, function, line, inode->i_ino,
446 current->comm, path);
273df556
FM		 447 vprintk(fmt, args);
448 printk("\n");
449 va_end(args);
450
451 ext4_handle_error(inode->i_sb);
452}
453
2b2d6d01 454static const char *ext4_decode_error(struct super_block *sb, int errno,
ac27a0ec
DK		 455 char nbuf[16])
456{
457 char *errstr = NULL;
458
459 switch (errno) {
460 case -EIO:
461 errstr = "IO failure";
462 break;
463 case -ENOMEM:
464 errstr = "Out of memory";
465 break;
466 case -EROFS:
78f1ddbb
TT		 467 if (!sb || (EXT4_SB(sb)->s_journal &&
468 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
ac27a0ec
DK		 469 errstr = "Journal has aborted";
470 else
471 errstr = "Readonly filesystem";
472 break;
473 default:
474 /* If the caller passed in an extra buffer for unknown
475 * errors, textualise them now. Else we just return
476 * NULL. */
477 if (nbuf) {
478 /* Check for truncated error codes... */
479 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
480 errstr = nbuf;
481 }
482 break;
483 }
484
485 return errstr;
486}
487
617ba13b 488/* __ext4_std_error decodes expected errors from journaling functions
ac27a0ec
DK		 489 * automatically and invokes the appropriate error response. */
490
c398eda0
TT		 491void __ext4_std_error(struct super_block *sb, const char *function,
492 unsigned int line, int errno)
ac27a0ec
DK		 493{
494 char nbuf[16];
495 const char *errstr;
496
497 /* Special case: if the error is EROFS, and we're not already
498 * inside a transaction, then there's really no point in logging
499 * an error. */
500 if (errno == -EROFS && journal_current_handle() == NULL &&
501 (sb->s_flags & MS_RDONLY))
502 return;
503
617ba13b 504 errstr = ext4_decode_error(sb, errno, nbuf);
c398eda0
TT		 505 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
506 sb->s_id, function, line, errstr);
1c13d5c0 507 save_error_info(sb, function, line);
ac27a0ec 508
617ba13b 509 ext4_handle_error(sb);
ac27a0ec
DK		 510}
511
512/*
617ba13b 513 * ext4_abort is a much stronger failure handler than ext4_error. The
ac27a0ec
DK		 514 * abort function may be used to deal with unrecoverable failures such
515 * as journal IO errors or ENOMEM at a critical moment in log management.
516 *
517 * We unconditionally force the filesystem into an ABORT|READONLY state,
518 * unless the error response on the fs has been set to panic in which
519 * case we take the easy way out and panic immediately.
520 */
521
c67d859e 522void __ext4_abort(struct super_block *sb, const char *function,
c398eda0 523 unsigned int line, const char *fmt, ...)
ac27a0ec
DK		 524{
525 va_list args;
526
1c13d5c0 527 save_error_info(sb, function, line);
ac27a0ec 528 va_start(args, fmt);
c398eda0
TT		 529 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
530 function, line);
ac27a0ec
DK		 531 vprintk(fmt, args);
532 printk("\n");
533 va_end(args);
534
1c13d5c0
TT		 535 if ((sb->s_flags & MS_RDONLY) == 0) {
536 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
537 sb->s_flags |= MS_RDONLY;
538 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
539 if (EXT4_SB(sb)->s_journal)
540 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
541 save_error_info(sb, function, line);
542 }
ac27a0ec 543 if (test_opt(sb, ERRORS_PANIC))
617ba13b 544 panic("EXT4-fs panic from previous error\n");
ac27a0ec
DK		 545}
546
b31e1552
ES		 547void ext4_msg (struct super_block * sb, const char *prefix,
548 const char *fmt, ...)
549{
550 va_list args;
551
552 va_start(args, fmt);
553 printk("%sEXT4-fs (%s): ", prefix, sb->s_id);
554 vprintk(fmt, args);
555 printk("\n");
556 va_end(args);
557}
558
12062ddd 559void __ext4_warning(struct super_block *sb, const char *function,
c398eda0 560 unsigned int line, const char *fmt, ...)
ac27a0ec
DK		 561{
562 va_list args;
563
564 va_start(args, fmt);
c398eda0
TT		 565 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: ",
566 sb->s_id, function, line);
ac27a0ec
DK		 567 vprintk(fmt, args);
568 printk("\n");
569 va_end(args);
570}
571
e29136f8
TT		 572void __ext4_grp_locked_error(const char *function, unsigned int line,
573 struct super_block *sb, ext4_group_t grp,
574 unsigned long ino, ext4_fsblk_t block,
575 const char *fmt, ...)
5d1b1b3f
AK		 576__releases(bitlock)
577__acquires(bitlock)
578{
579 va_list args;
580 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
581
1c13d5c0
TT		 582 es->s_last_error_ino = cpu_to_le32(ino);
583 es->s_last_error_block = cpu_to_le64(block);
584 __save_error_info(sb, function, line);
5d1b1b3f 585 va_start(args, fmt);
e29136f8
TT		 586 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u",
587 sb->s_id, function, line, grp);
588 if (ino)
589 printk("inode %lu: ", ino);
590 if (block)
591 printk("block %llu:", (unsigned long long) block);
5d1b1b3f
AK		 592 vprintk(fmt, args);
593 printk("\n");
594 va_end(args);
595
596 if (test_opt(sb, ERRORS_CONT)) {
e2d67052 597 ext4_commit_super(sb, 0);
5d1b1b3f
AK		 598 return;
599 }
1c13d5c0 600
5d1b1b3f
AK		 601 ext4_unlock_group(sb, grp);
602 ext4_handle_error(sb);
603 /*
604 * We only get here in the ERRORS_RO case; relocking the group
605 * may be dangerous, but nothing bad will happen since the
606 * filesystem will have already been marked read/only and the
607 * journal has been aborted. We return 1 as a hint to callers
608 * who might what to use the return value from
609 * ext4_grp_locked_error() to distinguish beween the
610 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
611 * aggressively from the ext4 function in question, with a
612 * more appropriate error code.
613 */
614 ext4_lock_group(sb, grp);
615 return;
616}
617
617ba13b 618void ext4_update_dynamic_rev(struct super_block *sb)
ac27a0ec 619{
617ba13b 620 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 621
617ba13b 622 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
ac27a0ec
DK		 623 return;
624
12062ddd 625 ext4_warning(sb,
ac27a0ec
DK		 626 "updating to rev %d because of new feature flag, "
627 "running e2fsck is recommended",
617ba13b 628 EXT4_DYNAMIC_REV);
ac27a0ec 629
617ba13b
MC		 630 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
631 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
632 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
ac27a0ec
DK		 633 /* leave es->s_feature_*compat flags alone */
634 /* es->s_uuid will be set by e2fsck if empty */
635
636 /*
637 * The rest of the superblock fields should be zero, and if not it
638 * means they are likely already in use, so leave them alone. We
639 * can leave it up to e2fsck to clean up any inconsistencies there.
640 */
641}
642
643/*
644 * Open the external journal device
645 */
b31e1552 646static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
ac27a0ec
DK		 647{
648 struct block_device *bdev;
649 char b[BDEVNAME_SIZE];
650
651 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
652 if (IS_ERR(bdev))
653 goto fail;
654 return bdev;
655
656fail:
b31e1552 657 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
ac27a0ec
DK		 658 __bdevname(dev, b), PTR_ERR(bdev));
659 return NULL;
660}
661
662/*
663 * Release the journal device
664 */
617ba13b 665static int ext4_blkdev_put(struct block_device *bdev)
ac27a0ec
DK		 666{
667 bd_release(bdev);
9a1c3542 668 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
ac27a0ec
DK		 669}
670
617ba13b 671static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
ac27a0ec
DK		 672{
673 struct block_device *bdev;
674 int ret = -ENODEV;
675
676 bdev = sbi->journal_bdev;
677 if (bdev) {
617ba13b 678 ret = ext4_blkdev_put(bdev);
ac27a0ec
DK		 679 sbi->journal_bdev = NULL;
680 }
681 return ret;
682}
683
684static inline struct inode *orphan_list_entry(struct list_head *l)
685{
617ba13b 686 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
ac27a0ec
DK		 687}
688
617ba13b 689static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
ac27a0ec
DK		 690{
691 struct list_head *l;
692
b31e1552
ES		 693 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
694 le32_to_cpu(sbi->s_es->s_last_orphan));
ac27a0ec
DK		 695
696 printk(KERN_ERR "sb_info orphan list:\n");
697 list_for_each(l, &sbi->s_orphan) {
698 struct inode *inode = orphan_list_entry(l);
699 printk(KERN_ERR " "
700 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
701 inode->i_sb->s_id, inode->i_ino, inode,
702 inode->i_mode, inode->i_nlink,
703 NEXT_ORPHAN(inode));
704 }
705}
706
2b2d6d01 707static void ext4_put_super(struct super_block *sb)
ac27a0ec 708{
617ba13b
MC		 709 struct ext4_sb_info *sbi = EXT4_SB(sb);
710 struct ext4_super_block *es = sbi->s_es;
ef2cabf7 711 int i, err;
ac27a0ec 712
857ac889 713 ext4_unregister_li_request(sb);
e0ccfd95
CH		 714 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
715
4c0425ff
MC		 716 flush_workqueue(sbi->dio_unwritten_wq);
717 destroy_workqueue(sbi->dio_unwritten_wq);
718
a9e220f8 719 lock_super(sb);
6cfd0148 720 lock_kernel();
8c85e125 721 if (sb->s_dirt)
ebc1ac16 722 ext4_commit_super(sb, 1);
8c85e125 723
0390131b
FM		 724 if (sbi->s_journal) {
725 err = jbd2_journal_destroy(sbi->s_journal);
726 sbi->s_journal = NULL;
727 if (err < 0)
c67d859e 728 ext4_abort(sb, "Couldn't clean up the journal");
0390131b 729 }
d4edac31 730
a1c6c569 731 del_timer(&sbi->s_err_report);
d4edac31
JB		 732 ext4_release_system_zone(sb);
733 ext4_mb_release(sb);
734 ext4_ext_release(sb);
735 ext4_xattr_put_super(sb);
736
ac27a0ec 737 if (!(sb->s_flags & MS_RDONLY)) {
617ba13b 738 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 739 es->s_state = cpu_to_le16(sbi->s_mount_state);
e2d67052 740 ext4_commit_super(sb, 1);
ac27a0ec 741 }
240799cd 742 if (sbi->s_proc) {
9f6200bb 743 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 744 }
3197ebdb 745 kobject_del(&sbi->s_kobj);
ac27a0ec
DK		 746
747 for (i = 0; i < sbi->s_gdb_count; i++)
748 brelse(sbi->s_group_desc[i]);
749 kfree(sbi->s_group_desc);
c5ca7c76
TT		 750 if (is_vmalloc_addr(sbi->s_flex_groups))
751 vfree(sbi->s_flex_groups);
752 else
753 kfree(sbi->s_flex_groups);
ac27a0ec
DK		 754 percpu_counter_destroy(&sbi->s_freeblocks_counter);
755 percpu_counter_destroy(&sbi->s_freeinodes_counter);
756 percpu_counter_destroy(&sbi->s_dirs_counter);
6bc6e63f 757 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
ac27a0ec
DK		 758 brelse(sbi->s_sbh);
759#ifdef CONFIG_QUOTA
760 for (i = 0; i < MAXQUOTAS; i++)
761 kfree(sbi->s_qf_names[i]);
762#endif
763
764 /* Debugging code just in case the in-memory inode orphan list
765 * isn't empty. The on-disk one can be non-empty if we've
766 * detected an error and taken the fs readonly, but the
767 * in-memory list had better be clean by this point. */
768 if (!list_empty(&sbi->s_orphan))
769 dump_orphan_list(sb, sbi);
770 J_ASSERT(list_empty(&sbi->s_orphan));
771
f98393a6 772 invalidate_bdev(sb->s_bdev);
ac27a0ec
DK		 773 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
774 /*
775 * Invalidate the journal device's buffers. We don't want them
776 * floating about in memory - the physical journal device may
777 * hotswapped, and it breaks the `ro-after' testing code.
778 */
779 sync_blockdev(sbi->journal_bdev);
f98393a6 780 invalidate_bdev(sbi->journal_bdev);
617ba13b 781 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 782 }
783 sb->s_fs_info = NULL;
3197ebdb
TT		 784 /*
785 * Now that we are completely done shutting down the
786 * superblock, we need to actually destroy the kobject.
787 */
788 unlock_kernel();
789 unlock_super(sb);
790 kobject_put(&sbi->s_kobj);
791 wait_for_completion(&sbi->s_kobj_unregister);
705895b6 792 kfree(sbi->s_blockgroup_lock);
ac27a0ec 793 kfree(sbi);
ac27a0ec
DK		 794}
795
e18b890b 796static struct kmem_cache *ext4_inode_cachep;
ac27a0ec
DK		 797
798/*
799 * Called inside transaction, so use GFP_NOFS
800 */
617ba13b 801static struct inode *ext4_alloc_inode(struct super_block *sb)
ac27a0ec 802{
617ba13b 803 struct ext4_inode_info *ei;
ac27a0ec 804
e6b4f8da 805 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
ac27a0ec
DK		 806 if (!ei)
807 return NULL;
0b8e58a1 808
ac27a0ec 809 ei->vfs_inode.i_version = 1;
91246c00 810 ei->vfs_inode.i_data.writeback_index = 0;
a86c6181 811 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
c9de560d
AT		 812 INIT_LIST_HEAD(&ei->i_prealloc_list);
813 spin_lock_init(&ei->i_prealloc_lock);
0390131b
FM		 814 /*
815 * Note: We can be called before EXT4_SB(sb)->s_journal is set,
816 * therefore it can be null here. Don't check it, just initialize
817 * jinode.
818 */
678aaf48 819 jbd2_journal_init_jbd_inode(&ei->jinode, &ei->vfs_inode);
d2a17637
MC		 820 ei->i_reserved_data_blocks = 0;
821 ei->i_reserved_meta_blocks = 0;
822 ei->i_allocated_meta_blocks = 0;
9d0be502 823 ei->i_da_metadata_calc_len = 0;
d2a17637
MC		 824 ei->i_delalloc_reserved_flag = 0;
825 spin_lock_init(&(ei->i_block_reservation_lock));
a9e7f447
DM		 826#ifdef CONFIG_QUOTA
827 ei->i_reserved_quota = 0;
828#endif
c7064ef1 829 INIT_LIST_HEAD(&ei->i_completed_io_list);
744692dc 830 spin_lock_init(&ei->i_completed_io_lock);
8d5d02e6 831 ei->cur_aio_dio = NULL;
b436b9be
JK		 832 ei->i_sync_tid = 0;
833 ei->i_datasync_tid = 0;
0b8e58a1 834
ac27a0ec
DK		 835 return &ei->vfs_inode;
836}
837
617ba13b 838static void ext4_destroy_inode(struct inode *inode)
ac27a0ec 839{
9f7dd93d 840 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
b31e1552
ES		 841 ext4_msg(inode->i_sb, KERN_ERR,
842 "Inode %lu (%p): orphan list check failed!",
843 inode->i_ino, EXT4_I(inode));
9f7dd93d
VA		 844 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
845 EXT4_I(inode), sizeof(struct ext4_inode_info),
846 true);
847 dump_stack();
848 }
617ba13b 849 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
ac27a0ec
DK		 850}
851
51cc5068 852static void init_once(void *foo)
ac27a0ec 853{
617ba13b 854 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
ac27a0ec 855
a35afb83 856 INIT_LIST_HEAD(&ei->i_orphan);
03010a33 857#ifdef CONFIG_EXT4_FS_XATTR
a35afb83 858 init_rwsem(&ei->xattr_sem);
ac27a0ec 859#endif
0e855ac8 860 init_rwsem(&ei->i_data_sem);
a35afb83 861 inode_init_once(&ei->vfs_inode);
ac27a0ec
DK		 862}
863
864static int init_inodecache(void)
865{
617ba13b
MC		 866 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
867 sizeof(struct ext4_inode_info),
ac27a0ec
DK		 868 0, (SLAB_RECLAIM_ACCOUNT|
869 SLAB_MEM_SPREAD),
20c2df83 870 init_once);
617ba13b 871 if (ext4_inode_cachep == NULL)
ac27a0ec
DK		 872 return -ENOMEM;
873 return 0;
874}
875
876static void destroy_inodecache(void)
877{
617ba13b 878 kmem_cache_destroy(ext4_inode_cachep);
ac27a0ec
DK		 879}
880
0930fcc1 881void ext4_clear_inode(struct inode *inode)
ac27a0ec 882{
0930fcc1
AV		 883 invalidate_inode_buffers(inode);
884 end_writeback(inode);
9f754758 885 dquot_drop(inode);
c2ea3fde 886 ext4_discard_preallocations(inode);
0390131b
FM		 887 if (EXT4_JOURNAL(inode))
888 jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal,
678aaf48 889 &EXT4_I(inode)->jinode);
ac27a0ec
DK		 890}
891
2b2d6d01
TT		 892static inline void ext4_show_quota_options(struct seq_file *seq,
893 struct super_block *sb)
ac27a0ec
DK		 894{
895#if defined(CONFIG_QUOTA)
617ba13b 896 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 897
5a20bdfc
JK		 898 if (sbi->s_jquota_fmt) {
899 char *fmtname = "";
900
901 switch (sbi->s_jquota_fmt) {
902 case QFMT_VFS_OLD:
903 fmtname = "vfsold";
904 break;
905 case QFMT_VFS_V0:
906 fmtname = "vfsv0";
907 break;
908 case QFMT_VFS_V1:
909 fmtname = "vfsv1";
910 break;
911 }
912 seq_printf(seq, ",jqfmt=%s", fmtname);
913 }
ac27a0ec
DK		 914
915 if (sbi->s_qf_names[USRQUOTA])
916 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
917
918 if (sbi->s_qf_names[GRPQUOTA])
919 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
920
482a7425 921 if (test_opt(sb, USRQUOTA))
ac27a0ec
DK		 922 seq_puts(seq, ",usrquota");
923
482a7425 924 if (test_opt(sb, GRPQUOTA))
ac27a0ec
DK		 925 seq_puts(seq, ",grpquota");
926#endif
927}
928
d9c9bef1
MS		 929/*
930 * Show an option if
931 * - it's set to a non-default value OR
932 * - if the per-sb default is different from the global default
933 */
617ba13b 934static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
ac27a0ec 935{
aa22df2c
AK		 936 int def_errors;
937 unsigned long def_mount_opts;
ac27a0ec 938 struct super_block *sb = vfs->mnt_sb;
d9c9bef1
MS		 939 struct ext4_sb_info *sbi = EXT4_SB(sb);
940 struct ext4_super_block *es = sbi->s_es;
d9c9bef1
MS		 941
942 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
aa22df2c 943 def_errors = le16_to_cpu(es->s_errors);
d9c9bef1
MS		 944
945 if (sbi->s_sb_block != 1)
946 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
947 if (test_opt(sb, MINIX_DF))
948 seq_puts(seq, ",minixdf");
aa22df2c 949 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
d9c9bef1
MS		 950 seq_puts(seq, ",grpid");
951 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
952 seq_puts(seq, ",nogrpid");
953 if (sbi->s_resuid != EXT4_DEF_RESUID ||
954 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
955 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
956 }
957 if (sbi->s_resgid != EXT4_DEF_RESGID ||
958 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
959 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
960 }
bb4f397a 961 if (test_opt(sb, ERRORS_RO)) {
d9c9bef1 962 if (def_errors == EXT4_ERRORS_PANIC ||
bb4f397a
AK		 963 def_errors == EXT4_ERRORS_CONTINUE) {
964 seq_puts(seq, ",errors=remount-ro");
d9c9bef1
MS		 965 }
966 }
aa22df2c 967 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
bb4f397a 968 seq_puts(seq, ",errors=continue");
aa22df2c 969 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
d9c9bef1 970 seq_puts(seq, ",errors=panic");
aa22df2c 971 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
d9c9bef1 972 seq_puts(seq, ",nouid32");
aa22df2c 973 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
d9c9bef1
MS		 974 seq_puts(seq, ",debug");
975 if (test_opt(sb, OLDALLOC))
976 seq_puts(seq, ",oldalloc");
03010a33 977#ifdef CONFIG_EXT4_FS_XATTR
aa22df2c
AK		 978 if (test_opt(sb, XATTR_USER) &&
979 !(def_mount_opts & EXT4_DEFM_XATTR_USER))
d9c9bef1
MS		 980 seq_puts(seq, ",user_xattr");
981 if (!test_opt(sb, XATTR_USER) &&
982 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
983 seq_puts(seq, ",nouser_xattr");
984 }
985#endif
03010a33 986#ifdef CONFIG_EXT4_FS_POSIX_ACL
aa22df2c 987 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
d9c9bef1
MS		 988 seq_puts(seq, ",acl");
989 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
990 seq_puts(seq, ",noacl");
991#endif
30773840 992 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
d9c9bef1
MS		 993 seq_printf(seq, ",commit=%u",
994 (unsigned) (sbi->s_commit_interval / HZ));
995 }
30773840
TT		 996 if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
997 seq_printf(seq, ",min_batch_time=%u",
998 (unsigned) sbi->s_min_batch_time);
999 }
1000 if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
1001 seq_printf(seq, ",max_batch_time=%u",
1002 (unsigned) sbi->s_min_batch_time);
1003 }
1004
571640ca
ES		 1005 /*
1006 * We're changing the default of barrier mount option, so
1007 * let's always display its mount state so it's clear what its
1008 * status is.
1009 */
1010 seq_puts(seq, ",barrier=");
1011 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
cd0b6a39
TT		 1012 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
1013 seq_puts(seq, ",journal_async_commit");
39a4bade
JK		 1014 else if (test_opt(sb, JOURNAL_CHECKSUM))
1015 seq_puts(seq, ",journal_checksum");
25ec56b5
JNC		 1016 if (test_opt(sb, I_VERSION))
1017 seq_puts(seq, ",i_version");
8b67f04a
TT		 1018 if (!test_opt(sb, DELALLOC) &&
1019 !(def_mount_opts & EXT4_DEFM_NODELALLOC))
dd919b98
AK		 1020 seq_puts(seq, ",nodelalloc");
1021
cb45bbe4
MS		 1022 if (sbi->s_stripe)
1023 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
aa22df2c
AK		 1024 /*
1025 * journal mode get enabled in different ways
1026 * So just print the value even if we didn't specify it
1027 */
617ba13b 1028 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
ac27a0ec 1029 seq_puts(seq, ",data=journal");
617ba13b 1030 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
ac27a0ec 1031 seq_puts(seq, ",data=ordered");
617ba13b 1032 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
ac27a0ec
DK		 1033 seq_puts(seq, ",data=writeback");
1034
240799cd
TT		 1035 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1036 seq_printf(seq, ",inode_readahead_blks=%u",
1037 sbi->s_inode_readahead_blks);
1038
5bf5683a
HK		 1039 if (test_opt(sb, DATA_ERR_ABORT))
1040 seq_puts(seq, ",data_err=abort");
1041
afd4672d 1042 if (test_opt(sb, NO_AUTO_DA_ALLOC))
06705bff 1043 seq_puts(seq, ",noauto_da_alloc");
afd4672d 1044
8b67f04a 1045 if (test_opt(sb, DISCARD) && !(def_mount_opts & EXT4_DEFM_DISCARD))
5328e635
ES		 1046 seq_puts(seq, ",discard");
1047
e3bb52ae
ES		 1048 if (test_opt(sb, NOLOAD))
1049 seq_puts(seq, ",norecovery");
1050
744692dc
JZ		 1051 if (test_opt(sb, DIOREAD_NOLOCK))
1052 seq_puts(seq, ",dioread_nolock");
1053
8b67f04a
TT		 1054 if (test_opt(sb, BLOCK_VALIDITY) &&
1055 !(def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY))
1056 seq_puts(seq, ",block_validity");
1057
bfff6873
LC		 1058 if (!test_opt(sb, INIT_INODE_TABLE))
1059 seq_puts(seq, ",noinit_inode_table");
1060 else if (sbi->s_li_wait_mult)
1061 seq_printf(seq, ",init_inode_table=%u",
1062 (unsigned) sbi->s_li_wait_mult);
1063
617ba13b 1064 ext4_show_quota_options(seq, sb);
0b8e58a1 1065
ac27a0ec
DK		 1066 return 0;
1067}
1068
1b961ac0 1069static struct inode *ext4_nfs_get_inode(struct super_block *sb,
0b8e58a1 1070 u64 ino, u32 generation)
ac27a0ec 1071{
ac27a0ec 1072 struct inode *inode;
ac27a0ec 1073
617ba13b 1074 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
ac27a0ec 1075 return ERR_PTR(-ESTALE);
617ba13b 1076 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
ac27a0ec
DK		 1077 return ERR_PTR(-ESTALE);
1078
1079 /* iget isn't really right if the inode is currently unallocated!!
1080 *
617ba13b 1081 * ext4_read_inode will return a bad_inode if the inode had been
ac27a0ec
DK		 1082 * deleted, so we should be safe.
1083 *
1084 * Currently we don't know the generation for parent directory, so
1085 * a generation of 0 means "accept any"
1086 */
1d1fe1ee
DH		 1087 inode = ext4_iget(sb, ino);
1088 if (IS_ERR(inode))
1089 return ERR_CAST(inode);
1090 if (generation && inode->i_generation != generation) {
ac27a0ec
DK		 1091 iput(inode);
1092 return ERR_PTR(-ESTALE);
1093 }
1b961ac0
CH		 1094
1095 return inode;
1096}
1097
1098static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
0b8e58a1 1099 int fh_len, int fh_type)
1b961ac0
CH		 1100{
1101 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1102 ext4_nfs_get_inode);
1103}
1104
1105static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
0b8e58a1 1106 int fh_len, int fh_type)
1b961ac0
CH		 1107{
1108 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1109 ext4_nfs_get_inode);
ac27a0ec
DK		 1110}
1111
c39a7f84
TO		 1112/*
1113 * Try to release metadata pages (indirect blocks, directories) which are
1114 * mapped via the block device. Since these pages could have journal heads
1115 * which would prevent try_to_free_buffers() from freeing them, we must use
1116 * jbd2 layer's try_to_free_buffers() function to release them.
1117 */
0b8e58a1
AD		 1118static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1119 gfp_t wait)
c39a7f84
TO		 1120{
1121 journal_t *journal = EXT4_SB(sb)->s_journal;
1122
1123 WARN_ON(PageChecked(page));
1124 if (!page_has_buffers(page))
1125 return 0;
1126 if (journal)
1127 return jbd2_journal_try_to_free_buffers(journal, page,
1128 wait & ~__GFP_WAIT);
1129 return try_to_free_buffers(page);
1130}
1131
ac27a0ec 1132#ifdef CONFIG_QUOTA
af5bc92d 1133#define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
2b2d6d01 1134#define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
ac27a0ec 1135
617ba13b
MC		 1136static int ext4_write_dquot(struct dquot *dquot);
1137static int ext4_acquire_dquot(struct dquot *dquot);
1138static int ext4_release_dquot(struct dquot *dquot);
1139static int ext4_mark_dquot_dirty(struct dquot *dquot);
1140static int ext4_write_info(struct super_block *sb, int type);
6f28e087 1141static int ext4_quota_on(struct super_block *sb, int type, int format_id,
307ae18a 1142 char *path);
ca0e05e4 1143static int ext4_quota_off(struct super_block *sb, int type);
617ba13b
MC		 1144static int ext4_quota_on_mount(struct super_block *sb, int type);
1145static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 1146 size_t len, loff_t off);
617ba13b 1147static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 1148 const char *data, size_t len, loff_t off);
1149
61e225dc 1150static const struct dquot_operations ext4_quota_operations = {
a9e7f447 1151#ifdef CONFIG_QUOTA
60e58e0f 1152 .get_reserved_space = ext4_get_reserved_space,
a9e7f447 1153#endif
617ba13b
MC		 1154 .write_dquot = ext4_write_dquot,
1155 .acquire_dquot = ext4_acquire_dquot,
1156 .release_dquot = ext4_release_dquot,
1157 .mark_dirty = ext4_mark_dquot_dirty,
a5b5ee32
JK		 1158 .write_info = ext4_write_info,
1159 .alloc_dquot = dquot_alloc,
1160 .destroy_dquot = dquot_destroy,
ac27a0ec
DK		 1161};
1162
0d54b217 1163static const struct quotactl_ops ext4_qctl_operations = {
617ba13b 1164 .quota_on = ext4_quota_on,
ca0e05e4 1165 .quota_off = ext4_quota_off,
287a8095
CH		 1166 .quota_sync = dquot_quota_sync,
1167 .get_info = dquot_get_dqinfo,
1168 .set_info = dquot_set_dqinfo,
1169 .get_dqblk = dquot_get_dqblk,
1170 .set_dqblk = dquot_set_dqblk
ac27a0ec
DK		 1171};
1172#endif
1173
ee9b6d61 1174static const struct super_operations ext4_sops = {
617ba13b
MC		 1175 .alloc_inode = ext4_alloc_inode,
1176 .destroy_inode = ext4_destroy_inode,
617ba13b
MC		 1177 .write_inode = ext4_write_inode,
1178 .dirty_inode = ext4_dirty_inode,
0930fcc1 1179 .evict_inode = ext4_evict_inode,
617ba13b 1180 .put_super = ext4_put_super,
617ba13b 1181 .sync_fs = ext4_sync_fs,
c4be0c1d
TS		 1182 .freeze_fs = ext4_freeze,
1183 .unfreeze_fs = ext4_unfreeze,
617ba13b
MC		 1184 .statfs = ext4_statfs,
1185 .remount_fs = ext4_remount,
617ba13b 1186 .show_options = ext4_show_options,
ac27a0ec 1187#ifdef CONFIG_QUOTA
617ba13b
MC		 1188 .quota_read = ext4_quota_read,
1189 .quota_write = ext4_quota_write,
ac27a0ec 1190#endif
c39a7f84 1191 .bdev_try_to_free_page = bdev_try_to_free_page,
ac27a0ec
DK		 1192};
1193
9ca92389
TT		 1194static const struct super_operations ext4_nojournal_sops = {
1195 .alloc_inode = ext4_alloc_inode,
1196 .destroy_inode = ext4_destroy_inode,
1197 .write_inode = ext4_write_inode,
1198 .dirty_inode = ext4_dirty_inode,
0930fcc1 1199 .evict_inode = ext4_evict_inode,
9ca92389
TT		 1200 .write_super = ext4_write_super,
1201 .put_super = ext4_put_super,
1202 .statfs = ext4_statfs,
1203 .remount_fs = ext4_remount,
9ca92389
TT		 1204 .show_options = ext4_show_options,
1205#ifdef CONFIG_QUOTA
1206 .quota_read = ext4_quota_read,
1207 .quota_write = ext4_quota_write,
1208#endif
1209 .bdev_try_to_free_page = bdev_try_to_free_page,
1210};
1211
39655164 1212static const struct export_operations ext4_export_ops = {
1b961ac0
CH		 1213 .fh_to_dentry = ext4_fh_to_dentry,
1214 .fh_to_parent = ext4_fh_to_parent,
617ba13b 1215 .get_parent = ext4_get_parent,
ac27a0ec
DK		 1216};
1217
1218enum {
1219 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1220 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
01436ef2 1221 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
ac27a0ec 1222 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
06705bff 1223 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh,
30773840 1224 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
c3191067 1225 Opt_journal_update, Opt_journal_dev,
818d276c 1226 Opt_journal_checksum, Opt_journal_async_commit,
ac27a0ec 1227 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
296c355c 1228 Opt_data_err_abort, Opt_data_err_ignore,
ac27a0ec 1229 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
5a20bdfc
JK		 1230 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1231 Opt_noquota, Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err,
1232 Opt_resize, Opt_usrquota, Opt_grpquota, Opt_i_version,
01436ef2 1233 Opt_stripe, Opt_delalloc, Opt_nodelalloc,
6fd058f7 1234 Opt_block_validity, Opt_noblock_validity,
5328e635 1235 Opt_inode_readahead_blks, Opt_journal_ioprio,
744692dc 1236 Opt_dioread_nolock, Opt_dioread_lock,
5328e635 1237 Opt_discard, Opt_nodiscard,
bfff6873 1238 Opt_init_inode_table, Opt_noinit_inode_table,
ac27a0ec
DK		 1239};
1240
a447c093 1241static const match_table_t tokens = {
ac27a0ec
DK		 1242 {Opt_bsd_df, "bsddf"},
1243 {Opt_minix_df, "minixdf"},
1244 {Opt_grpid, "grpid"},
1245 {Opt_grpid, "bsdgroups"},
1246 {Opt_nogrpid, "nogrpid"},
1247 {Opt_nogrpid, "sysvgroups"},
1248 {Opt_resgid, "resgid=%u"},
1249 {Opt_resuid, "resuid=%u"},
1250 {Opt_sb, "sb=%u"},
1251 {Opt_err_cont, "errors=continue"},
1252 {Opt_err_panic, "errors=panic"},
1253 {Opt_err_ro, "errors=remount-ro"},
1254 {Opt_nouid32, "nouid32"},
ac27a0ec
DK		 1255 {Opt_debug, "debug"},
1256 {Opt_oldalloc, "oldalloc"},
1257 {Opt_orlov, "orlov"},
1258 {Opt_user_xattr, "user_xattr"},
1259 {Opt_nouser_xattr, "nouser_xattr"},
1260 {Opt_acl, "acl"},
1261 {Opt_noacl, "noacl"},
ac27a0ec 1262 {Opt_noload, "noload"},
e3bb52ae 1263 {Opt_noload, "norecovery"},
ac27a0ec
DK		 1264 {Opt_nobh, "nobh"},
1265 {Opt_bh, "bh"},
1266 {Opt_commit, "commit=%u"},
30773840
TT		 1267 {Opt_min_batch_time, "min_batch_time=%u"},
1268 {Opt_max_batch_time, "max_batch_time=%u"},
ac27a0ec 1269 {Opt_journal_update, "journal=update"},
ac27a0ec 1270 {Opt_journal_dev, "journal_dev=%u"},
818d276c
GS		 1271 {Opt_journal_checksum, "journal_checksum"},
1272 {Opt_journal_async_commit, "journal_async_commit"},
ac27a0ec
DK		 1273 {Opt_abort, "abort"},
1274 {Opt_data_journal, "data=journal"},
1275 {Opt_data_ordered, "data=ordered"},
1276 {Opt_data_writeback, "data=writeback"},
5bf5683a
HK		 1277 {Opt_data_err_abort, "data_err=abort"},
1278 {Opt_data_err_ignore, "data_err=ignore"},
ac27a0ec
DK		 1279 {Opt_offusrjquota, "usrjquota="},
1280 {Opt_usrjquota, "usrjquota=%s"},
1281 {Opt_offgrpjquota, "grpjquota="},
1282 {Opt_grpjquota, "grpjquota=%s"},
1283 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1284 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
5a20bdfc 1285 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
ac27a0ec
DK		 1286 {Opt_grpquota, "grpquota"},
1287 {Opt_noquota, "noquota"},
1288 {Opt_quota, "quota"},
1289 {Opt_usrquota, "usrquota"},
1290 {Opt_barrier, "barrier=%u"},
06705bff
TT		 1291 {Opt_barrier, "barrier"},
1292 {Opt_nobarrier, "nobarrier"},
25ec56b5 1293 {Opt_i_version, "i_version"},
c9de560d 1294 {Opt_stripe, "stripe=%u"},
ac27a0ec 1295 {Opt_resize, "resize"},
64769240 1296 {Opt_delalloc, "delalloc"},
dd919b98 1297 {Opt_nodelalloc, "nodelalloc"},
6fd058f7
TT		 1298 {Opt_block_validity, "block_validity"},
1299 {Opt_noblock_validity, "noblock_validity"},
240799cd 1300 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
b3881f74 1301 {Opt_journal_ioprio, "journal_ioprio=%u"},
afd4672d 1302 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
06705bff
TT		 1303 {Opt_auto_da_alloc, "auto_da_alloc"},
1304 {Opt_noauto_da_alloc, "noauto_da_alloc"},
744692dc
JZ		 1305 {Opt_dioread_nolock, "dioread_nolock"},
1306 {Opt_dioread_lock, "dioread_lock"},
5328e635
ES		 1307 {Opt_discard, "discard"},
1308 {Opt_nodiscard, "nodiscard"},
bfff6873
LC		 1309 {Opt_init_inode_table, "init_itable=%u"},
1310 {Opt_init_inode_table, "init_itable"},
1311 {Opt_noinit_inode_table, "noinit_itable"},
f3f12faa 1312 {Opt_err, NULL},
ac27a0ec
DK		 1313};
1314
617ba13b 1315static ext4_fsblk_t get_sb_block(void **data)
ac27a0ec 1316{
617ba13b 1317 ext4_fsblk_t sb_block;
ac27a0ec
DK		 1318 char *options = (char *) *data;
1319
1320 if (!options || strncmp(options, "sb=", 3) != 0)
1321 return 1; /* Default location */
0b8e58a1 1322
ac27a0ec 1323 options += 3;
0b8e58a1 1324 /* TODO: use simple_strtoll with >32bit ext4 */
ac27a0ec
DK		 1325 sb_block = simple_strtoul(options, &options, 0);
1326 if (*options && *options != ',') {
4776004f 1327 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
ac27a0ec
DK		 1328 (char *) *data);
1329 return 1;
1330 }
1331 if (*options == ',')
1332 options++;
1333 *data = (void *) options;
0b8e58a1 1334
ac27a0ec
DK		 1335 return sb_block;
1336}
1337
b3881f74 1338#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
437ca0fd
DM		 1339static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1340 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
b3881f74 1341
56c50f11
DM		 1342#ifdef CONFIG_QUOTA
1343static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1344{
1345 struct ext4_sb_info *sbi = EXT4_SB(sb);
1346 char *qname;
1347
1348 if (sb_any_quota_loaded(sb) &&
1349 !sbi->s_qf_names[qtype]) {
1350 ext4_msg(sb, KERN_ERR,
1351 "Cannot change journaled "
1352 "quota options when quota turned on");
1353 return 0;
1354 }
1355 qname = match_strdup(args);
1356 if (!qname) {
1357 ext4_msg(sb, KERN_ERR,
1358 "Not enough memory for storing quotafile name");
1359 return 0;
1360 }
1361 if (sbi->s_qf_names[qtype] &&
1362 strcmp(sbi->s_qf_names[qtype], qname)) {
1363 ext4_msg(sb, KERN_ERR,
1364 "%s quota file already specified", QTYPE2NAME(qtype));
1365 kfree(qname);
1366 return 0;
1367 }
1368 sbi->s_qf_names[qtype] = qname;
1369 if (strchr(sbi->s_qf_names[qtype], '/')) {
1370 ext4_msg(sb, KERN_ERR,
1371 "quotafile must be on filesystem root");
1372 kfree(sbi->s_qf_names[qtype]);
1373 sbi->s_qf_names[qtype] = NULL;
1374 return 0;
1375 }
1376 set_opt(sbi->s_mount_opt, QUOTA);
1377 return 1;
1378}
1379
1380static int clear_qf_name(struct super_block *sb, int qtype)
1381{
1382
1383 struct ext4_sb_info *sbi = EXT4_SB(sb);
1384
1385 if (sb_any_quota_loaded(sb) &&
1386 sbi->s_qf_names[qtype]) {
1387 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1388 " when quota turned on");
1389 return 0;
1390 }
1391 /*
1392 * The space will be released later when all options are confirmed
1393 * to be correct
1394 */
1395 sbi->s_qf_names[qtype] = NULL;
1396 return 1;
1397}
1398#endif
1399
2b2d6d01 1400static int parse_options(char *options, struct super_block *sb,
c3191067 1401 unsigned long *journal_devnum,
b3881f74 1402 unsigned int *journal_ioprio,
2b2d6d01 1403 ext4_fsblk_t *n_blocks_count, int is_remount)
ac27a0ec 1404{
617ba13b 1405 struct ext4_sb_info *sbi = EXT4_SB(sb);
2b2d6d01 1406 char *p;
ac27a0ec
DK		 1407 substring_t args[MAX_OPT_ARGS];
1408 int data_opt = 0;
1409 int option;
1410#ifdef CONFIG_QUOTA
56c50f11 1411 int qfmt;
ac27a0ec
DK		 1412#endif
1413
1414 if (!options)
1415 return 1;
1416
2b2d6d01 1417 while ((p = strsep(&options, ",")) != NULL) {
ac27a0ec
DK		 1418 int token;
1419 if (!*p)
1420 continue;
1421
15121c18
ES		 1422 /*
1423 * Initialize args struct so we know whether arg was
1424 * found; some options take optional arguments.
1425 */
1426 args[0].to = args[0].from = 0;
ac27a0ec
DK		 1427 token = match_token(p, tokens, args);
1428 switch (token) {
1429 case Opt_bsd_df:
437ca0fd 1430 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
2b2d6d01 1431 clear_opt(sbi->s_mount_opt, MINIX_DF);
ac27a0ec
DK		 1432 break;
1433 case Opt_minix_df:
437ca0fd 1434 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
2b2d6d01 1435 set_opt(sbi->s_mount_opt, MINIX_DF);
437ca0fd 1436
ac27a0ec
DK		 1437 break;
1438 case Opt_grpid:
437ca0fd 1439 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
2b2d6d01 1440 set_opt(sbi->s_mount_opt, GRPID);
437ca0fd 1441
ac27a0ec
DK		 1442 break;
1443 case Opt_nogrpid:
437ca0fd 1444 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
2b2d6d01 1445 clear_opt(sbi->s_mount_opt, GRPID);
437ca0fd 1446
ac27a0ec
DK		 1447 break;
1448 case Opt_resuid:
1449 if (match_int(&args[0], &option))
1450 return 0;
1451 sbi->s_resuid = option;
1452 break;
1453 case Opt_resgid:
1454 if (match_int(&args[0], &option))
1455 return 0;
1456 sbi->s_resgid = option;
1457 break;
1458 case Opt_sb:
1459 /* handled by get_sb_block() instead of here */
1460 /* *sb_block = match_int(&args[0]); */
1461 break;
1462 case Opt_err_panic:
2b2d6d01
TT		 1463 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1464 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1465 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
ac27a0ec
DK		 1466 break;
1467 case Opt_err_ro:
2b2d6d01
TT		 1468 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1469 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1470 set_opt(sbi->s_mount_opt, ERRORS_RO);
ac27a0ec
DK		 1471 break;
1472 case Opt_err_cont:
2b2d6d01
TT		 1473 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1474 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1475 set_opt(sbi->s_mount_opt, ERRORS_CONT);
ac27a0ec
DK		 1476 break;
1477 case Opt_nouid32:
2b2d6d01 1478 set_opt(sbi->s_mount_opt, NO_UID32);
ac27a0ec 1479 break;
ac27a0ec 1480 case Opt_debug:
2b2d6d01 1481 set_opt(sbi->s_mount_opt, DEBUG);
ac27a0ec
DK		 1482 break;
1483 case Opt_oldalloc:
2b2d6d01 1484 set_opt(sbi->s_mount_opt, OLDALLOC);
ac27a0ec
DK		 1485 break;
1486 case Opt_orlov:
2b2d6d01 1487 clear_opt(sbi->s_mount_opt, OLDALLOC);
ac27a0ec 1488 break;
03010a33 1489#ifdef CONFIG_EXT4_FS_XATTR
ac27a0ec 1490 case Opt_user_xattr:
2b2d6d01 1491 set_opt(sbi->s_mount_opt, XATTR_USER);
ac27a0ec
DK		 1492 break;
1493 case Opt_nouser_xattr:
2b2d6d01 1494 clear_opt(sbi->s_mount_opt, XATTR_USER);
ac27a0ec
DK		 1495 break;
1496#else
1497 case Opt_user_xattr:
1498 case Opt_nouser_xattr:
b31e1552 1499 ext4_msg(sb, KERN_ERR, "(no)user_xattr options not supported");
ac27a0ec
DK		 1500 break;
1501#endif
03010a33 1502#ifdef CONFIG_EXT4_FS_POSIX_ACL
ac27a0ec
DK		 1503 case Opt_acl:
1504 set_opt(sbi->s_mount_opt, POSIX_ACL);
1505 break;
1506 case Opt_noacl:
1507 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1508 break;
1509#else
1510 case Opt_acl:
1511 case Opt_noacl:
b31e1552 1512 ext4_msg(sb, KERN_ERR, "(no)acl options not supported");
ac27a0ec
DK		 1513 break;
1514#endif
ac27a0ec
DK		 1515 case Opt_journal_update:
1516 /* @@@ FIXME */
1517 /* Eventually we will want to be able to create
1518 a journal file here. For now, only allow the
1519 user to specify an existing inode to be the
1520 journal file. */
1521 if (is_remount) {
b31e1552
ES		 1522 ext4_msg(sb, KERN_ERR,
1523 "Cannot specify journal on remount");
ac27a0ec
DK		 1524 return 0;
1525 }
2b2d6d01 1526 set_opt(sbi->s_mount_opt, UPDATE_JOURNAL);
ac27a0ec 1527 break;
ac27a0ec
DK		 1528 case Opt_journal_dev:
1529 if (is_remount) {
b31e1552
ES		 1530 ext4_msg(sb, KERN_ERR,
1531 "Cannot specify journal on remount");
ac27a0ec
DK		 1532 return 0;
1533 }
1534 if (match_int(&args[0], &option))
1535 return 0;
1536 *journal_devnum = option;
1537 break;
818d276c 1538 case Opt_journal_checksum:
d4da6c9c
LT		 1539 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1540 break;
818d276c
GS		 1541 case Opt_journal_async_commit:
1542 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
d4da6c9c 1543 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
818d276c 1544 break;
ac27a0ec 1545 case Opt_noload:
2b2d6d01 1546 set_opt(sbi->s_mount_opt, NOLOAD);
ac27a0ec
DK		 1547 break;
1548 case Opt_commit:
1549 if (match_int(&args[0], &option))
1550 return 0;
1551 if (option < 0)
1552 return 0;
1553 if (option == 0)
cd02ff0b 1554 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
ac27a0ec
DK		 1555 sbi->s_commit_interval = HZ * option;
1556 break;
30773840
TT		 1557 case Opt_max_batch_time:
1558 if (match_int(&args[0], &option))
1559 return 0;
1560 if (option < 0)
1561 return 0;
1562 if (option == 0)
1563 option = EXT4_DEF_MAX_BATCH_TIME;
1564 sbi->s_max_batch_time = option;
1565 break;
1566 case Opt_min_batch_time:
1567 if (match_int(&args[0], &option))
1568 return 0;
1569 if (option < 0)
1570 return 0;
1571 sbi->s_min_batch_time = option;
1572 break;
ac27a0ec 1573 case Opt_data_journal:
617ba13b 1574 data_opt = EXT4_MOUNT_JOURNAL_DATA;
ac27a0ec
DK		 1575 goto datacheck;
1576 case Opt_data_ordered:
617ba13b 1577 data_opt = EXT4_MOUNT_ORDERED_DATA;
ac27a0ec
DK		 1578 goto datacheck;
1579 case Opt_data_writeback:
617ba13b 1580 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
ac27a0ec
DK		 1581 datacheck:
1582 if (is_remount) {
482a7425 1583 if (test_opt(sb, DATA_FLAGS) != data_opt) {
b31e1552
ES		 1584 ext4_msg(sb, KERN_ERR,
1585 "Cannot change data mode on remount");
ac27a0ec
DK		 1586 return 0;
1587 }
1588 } else {
482a7425 1589 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
ac27a0ec
DK		 1590 sbi->s_mount_opt |= data_opt;
1591 }
1592 break;
5bf5683a
HK		 1593 case Opt_data_err_abort:
1594 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1595 break;
1596 case Opt_data_err_ignore:
1597 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1598 break;
ac27a0ec
DK		 1599#ifdef CONFIG_QUOTA
1600 case Opt_usrjquota:
56c50f11 1601 if (!set_qf_name(sb, USRQUOTA, &args[0]))
ac27a0ec 1602 return 0;
56c50f11
DM		 1603 break;
1604 case Opt_grpjquota:
1605 if (!set_qf_name(sb, GRPQUOTA, &args[0]))
ac27a0ec 1606 return 0;
ac27a0ec
DK		 1607 break;
1608 case Opt_offusrjquota:
56c50f11
DM		 1609 if (!clear_qf_name(sb, USRQUOTA))
1610 return 0;
1611 break;
ac27a0ec 1612 case Opt_offgrpjquota:
56c50f11 1613 if (!clear_qf_name(sb, GRPQUOTA))
ac27a0ec 1614 return 0;
ac27a0ec 1615 break;
56c50f11 1616
ac27a0ec 1617 case Opt_jqfmt_vfsold:
dfc5d03f
JK		 1618 qfmt = QFMT_VFS_OLD;
1619 goto set_qf_format;
ac27a0ec 1620 case Opt_jqfmt_vfsv0:
dfc5d03f 1621 qfmt = QFMT_VFS_V0;
5a20bdfc
JK		 1622 goto set_qf_format;
1623 case Opt_jqfmt_vfsv1:
1624 qfmt = QFMT_VFS_V1;
dfc5d03f 1625set_qf_format:
17bd13b3 1626 if (sb_any_quota_loaded(sb) &&
dfc5d03f 1627 sbi->s_jquota_fmt != qfmt) {
b31e1552 1628 ext4_msg(sb, KERN_ERR, "Cannot change "
dfc5d03f 1629 "journaled quota options when "
b31e1552 1630 "quota turned on");
dfc5d03f
JK		 1631 return 0;
1632 }
1633 sbi->s_jquota_fmt = qfmt;
ac27a0ec
DK		 1634 break;
1635 case Opt_quota:
1636 case Opt_usrquota:
1637 set_opt(sbi->s_mount_opt, QUOTA);
1638 set_opt(sbi->s_mount_opt, USRQUOTA);
1639 break;
1640 case Opt_grpquota:
1641 set_opt(sbi->s_mount_opt, QUOTA);
1642 set_opt(sbi->s_mount_opt, GRPQUOTA);
1643 break;
1644 case Opt_noquota:
17bd13b3 1645 if (sb_any_quota_loaded(sb)) {
b31e1552
ES		 1646 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1647 "options when quota turned on");
ac27a0ec
DK		 1648 return 0;
1649 }
1650 clear_opt(sbi->s_mount_opt, QUOTA);
1651 clear_opt(sbi->s_mount_opt, USRQUOTA);
1652 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1653 break;
1654#else
1655 case Opt_quota:
1656 case Opt_usrquota:
1657 case Opt_grpquota:
b31e1552
ES		 1658 ext4_msg(sb, KERN_ERR,
1659 "quota options not supported");
cd59e7b9 1660 break;
ac27a0ec
DK		 1661 case Opt_usrjquota:
1662 case Opt_grpjquota:
1663 case Opt_offusrjquota:
1664 case Opt_offgrpjquota:
1665 case Opt_jqfmt_vfsold:
1666 case Opt_jqfmt_vfsv0:
5a20bdfc 1667 case Opt_jqfmt_vfsv1:
b31e1552
ES		 1668 ext4_msg(sb, KERN_ERR,
1669 "journaled quota options not supported");
ac27a0ec
DK		 1670 break;
1671 case Opt_noquota:
1672 break;
1673#endif
1674 case Opt_abort:
4ab2f15b 1675 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
ac27a0ec 1676 break;
06705bff
TT		 1677 case Opt_nobarrier:
1678 clear_opt(sbi->s_mount_opt, BARRIER);
1679 break;
ac27a0ec 1680 case Opt_barrier:
15121c18
ES		 1681 if (args[0].from) {
1682 if (match_int(&args[0], &option))
1683 return 0;
1684 } else
1685 option = 1; /* No argument, default to 1 */
ac27a0ec
DK		 1686 if (option)
1687 set_opt(sbi->s_mount_opt, BARRIER);
1688 else
1689 clear_opt(sbi->s_mount_opt, BARRIER);
1690 break;
1691 case Opt_ignore:
1692 break;
1693 case Opt_resize:
1694 if (!is_remount) {
b31e1552
ES		 1695 ext4_msg(sb, KERN_ERR,
1696 "resize option only available "
1697 "for remount");
ac27a0ec
DK		 1698 return 0;
1699 }
1700 if (match_int(&args[0], &option) != 0)
1701 return 0;
1702 *n_blocks_count = option;
1703 break;
1704 case Opt_nobh:
206f7ab4
CH		 1705 ext4_msg(sb, KERN_WARNING,
1706 "Ignoring deprecated nobh option");
ac27a0ec
DK		 1707 break;
1708 case Opt_bh:
206f7ab4
CH		 1709 ext4_msg(sb, KERN_WARNING,
1710 "Ignoring deprecated bh option");
ac27a0ec 1711 break;
25ec56b5
JNC		 1712 case Opt_i_version:
1713 set_opt(sbi->s_mount_opt, I_VERSION);
1714 sb->s_flags |= MS_I_VERSION;
1715 break;
dd919b98
AK		 1716 case Opt_nodelalloc:
1717 clear_opt(sbi->s_mount_opt, DELALLOC);
1718 break;
c9de560d
AT		 1719 case Opt_stripe:
1720 if (match_int(&args[0], &option))
1721 return 0;
1722 if (option < 0)
1723 return 0;
1724 sbi->s_stripe = option;
1725 break;
64769240
AT		 1726 case Opt_delalloc:
1727 set_opt(sbi->s_mount_opt, DELALLOC);
1728 break;
6fd058f7
TT		 1729 case Opt_block_validity:
1730 set_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1731 break;
1732 case Opt_noblock_validity:
1733 clear_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1734 break;
240799cd
TT		 1735 case Opt_inode_readahead_blks:
1736 if (match_int(&args[0], &option))
1737 return 0;
1738 if (option < 0 || option > (1 << 30))
1739 return 0;
f7c43950 1740 if (!is_power_of_2(option)) {
b31e1552
ES		 1741 ext4_msg(sb, KERN_ERR,
1742 "EXT4-fs: inode_readahead_blks"
1743 " must be a power of 2");
3197ebdb
TT		 1744 return 0;
1745 }
240799cd
TT		 1746 sbi->s_inode_readahead_blks = option;
1747 break;
b3881f74
TT		 1748 case Opt_journal_ioprio:
1749 if (match_int(&args[0], &option))
1750 return 0;
1751 if (option < 0 || option > 7)
1752 break;
1753 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1754 option);
1755 break;
06705bff
TT		 1756 case Opt_noauto_da_alloc:
1757 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1758 break;
afd4672d 1759 case Opt_auto_da_alloc:
15121c18
ES		 1760 if (args[0].from) {
1761 if (match_int(&args[0], &option))
1762 return 0;
1763 } else
1764 option = 1; /* No argument, default to 1 */
afd4672d
TT		 1765 if (option)
1766 clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1767 else
1768 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1769 break;
5328e635
ES		 1770 case Opt_discard:
1771 set_opt(sbi->s_mount_opt, DISCARD);
1772 break;
1773 case Opt_nodiscard:
1774 clear_opt(sbi->s_mount_opt, DISCARD);
1775 break;
744692dc
JZ		 1776 case Opt_dioread_nolock:
1777 set_opt(sbi->s_mount_opt, DIOREAD_NOLOCK);
1778 break;
1779 case Opt_dioread_lock:
1780 clear_opt(sbi->s_mount_opt, DIOREAD_NOLOCK);
1781 break;
bfff6873
LC		 1782 case Opt_init_inode_table:
1783 set_opt(sbi->s_mount_opt, INIT_INODE_TABLE);
1784 if (args[0].from) {
1785 if (match_int(&args[0], &option))
1786 return 0;
1787 } else
1788 option = EXT4_DEF_LI_WAIT_MULT;
1789 if (option < 0)
1790 return 0;
1791 sbi->s_li_wait_mult = option;
1792 break;
1793 case Opt_noinit_inode_table:
1794 clear_opt(sbi->s_mount_opt, INIT_INODE_TABLE);
1795 break;
ac27a0ec 1796 default:
b31e1552
ES		 1797 ext4_msg(sb, KERN_ERR,
1798 "Unrecognized mount option \"%s\" "
1799 "or missing value", p);
ac27a0ec
DK		 1800 return 0;
1801 }
1802 }
1803#ifdef CONFIG_QUOTA
1804 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
482a7425 1805 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
ac27a0ec
DK		 1806 clear_opt(sbi->s_mount_opt, USRQUOTA);
1807
482a7425 1808 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
ac27a0ec
DK		 1809 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1810
56c50f11 1811 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
b31e1552
ES		 1812 ext4_msg(sb, KERN_ERR, "old and new quota "
1813 "format mixing");
ac27a0ec
DK		 1814 return 0;
1815 }
1816
1817 if (!sbi->s_jquota_fmt) {
b31e1552
ES		 1818 ext4_msg(sb, KERN_ERR, "journaled quota format "
1819 "not specified");
ac27a0ec
DK		 1820 return 0;
1821 }
1822 } else {
1823 if (sbi->s_jquota_fmt) {
b31e1552 1824 ext4_msg(sb, KERN_ERR, "journaled quota format "
2c8be6b2 1825 "specified with no journaling "
b31e1552 1826 "enabled");
ac27a0ec
DK		 1827 return 0;
1828 }
1829 }
1830#endif
1831 return 1;
1832}
1833
617ba13b 1834static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 1835 int read_only)
1836{
617ba13b 1837 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 1838 int res = 0;
1839
617ba13b 1840 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
b31e1552
ES		 1841 ext4_msg(sb, KERN_ERR, "revision level too high, "
1842 "forcing read-only mode");
ac27a0ec
DK		 1843 res = MS_RDONLY;
1844 }
1845 if (read_only)
1846 return res;
617ba13b 1847 if (!(sbi->s_mount_state & EXT4_VALID_FS))
b31e1552
ES		 1848 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1849 "running e2fsck is recommended");
617ba13b 1850 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
b31e1552
ES		 1851 ext4_msg(sb, KERN_WARNING,
1852 "warning: mounting fs with errors, "
1853 "running e2fsck is recommended");
ac27a0ec
DK		 1854 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1855 le16_to_cpu(es->s_mnt_count) >=
1856 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
b31e1552
ES		 1857 ext4_msg(sb, KERN_WARNING,
1858 "warning: maximal mount count reached, "
1859 "running e2fsck is recommended");
ac27a0ec
DK		 1860 else if (le32_to_cpu(es->s_checkinterval) &&
1861 (le32_to_cpu(es->s_lastcheck) +
1862 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
b31e1552
ES		 1863 ext4_msg(sb, KERN_WARNING,
1864 "warning: checktime reached, "
1865 "running e2fsck is recommended");
0b8e58a1 1866 if (!sbi->s_journal)
0390131b 1867 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec 1868 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 1869 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 1870 le16_add_cpu(&es->s_mnt_count, 1);
ac27a0ec 1871 es->s_mtime = cpu_to_le32(get_seconds());
617ba13b 1872 ext4_update_dynamic_rev(sb);
0390131b
FM		 1873 if (sbi->s_journal)
1874 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 1875
e2d67052 1876 ext4_commit_super(sb, 1);
ac27a0ec 1877 if (test_opt(sb, DEBUG))
a9df9a49 1878 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
7f4520cc 1879 "bpg=%lu, ipg=%lu, mo=%04x]\n",
ac27a0ec
DK		 1880 sb->s_blocksize,
1881 sbi->s_groups_count,
617ba13b
MC		 1882 EXT4_BLOCKS_PER_GROUP(sb),
1883 EXT4_INODES_PER_GROUP(sb),
ac27a0ec
DK		 1884 sbi->s_mount_opt);
1885
ac27a0ec
DK		 1886 return res;
1887}
1888
772cb7c8
JS		 1889static int ext4_fill_flex_info(struct super_block *sb)
1890{
1891 struct ext4_sb_info *sbi = EXT4_SB(sb);
1892 struct ext4_group_desc *gdp = NULL;
772cb7c8
JS		 1893 ext4_group_t flex_group_count;
1894 ext4_group_t flex_group;
1895 int groups_per_flex = 0;
c5ca7c76 1896 size_t size;
772cb7c8
JS		 1897 int i;
1898
503358ae
TT		 1899 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1900 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1901
1902 if (groups_per_flex < 2) {
772cb7c8
JS		 1903 sbi->s_log_groups_per_flex = 0;
1904 return 1;
1905 }
1906
c62a11fd
FB		 1907 /* We allocate both existing and potentially added groups */
1908 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
d94e99a6
AK		 1909 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1910 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
c5ca7c76
TT		 1911 size = flex_group_count * sizeof(struct flex_groups);
1912 sbi->s_flex_groups = kzalloc(size, GFP_KERNEL);
1913 if (sbi->s_flex_groups == NULL) {
1914 sbi->s_flex_groups = vmalloc(size);
1915 if (sbi->s_flex_groups)
1916 memset(sbi->s_flex_groups, 0, size);
1917 }
772cb7c8 1918 if (sbi->s_flex_groups == NULL) {
b31e1552
ES		 1919 ext4_msg(sb, KERN_ERR, "not enough memory for "
1920 "%u flex groups", flex_group_count);
772cb7c8
JS		 1921 goto failed;
1922 }
772cb7c8 1923
772cb7c8 1924 for (i = 0; i < sbi->s_groups_count; i++) {
88b6edd1 1925 gdp = ext4_get_group_desc(sb, i, NULL);
772cb7c8
JS		 1926
1927 flex_group = ext4_flex_group(sbi, i);
7ad9bb65
TT		 1928 atomic_add(ext4_free_inodes_count(sb, gdp),
1929 &sbi->s_flex_groups[flex_group].free_inodes);
1930 atomic_add(ext4_free_blks_count(sb, gdp),
1931 &sbi->s_flex_groups[flex_group].free_blocks);
1932 atomic_add(ext4_used_dirs_count(sb, gdp),
1933 &sbi->s_flex_groups[flex_group].used_dirs);
772cb7c8
JS		 1934 }
1935
1936 return 1;
1937failed:
1938 return 0;
1939}
1940
717d50e4
AD		 1941__le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1942 struct ext4_group_desc *gdp)
1943{
1944 __u16 crc = 0;
1945
1946 if (sbi->s_es->s_feature_ro_compat &
1947 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1948 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1949 __le32 le_group = cpu_to_le32(block_group);
1950
1951 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1952 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1953 crc = crc16(crc, (__u8 *)gdp, offset);
1954 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1955 /* for checksum of struct ext4_group_desc do the rest...*/
1956 if ((sbi->s_es->s_feature_incompat &
1957 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1958 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1959 crc = crc16(crc, (__u8 *)gdp + offset,
1960 le16_to_cpu(sbi->s_es->s_desc_size) -
1961 offset);
1962 }
1963
1964 return cpu_to_le16(crc);
1965}
1966
1967int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1968 struct ext4_group_desc *gdp)
1969{
1970 if ((sbi->s_es->s_feature_ro_compat &
1971 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1972 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1973 return 0;
1974
1975 return 1;
1976}
1977
ac27a0ec 1978/* Called at mount-time, super-block is locked */
bfff6873
LC		 1979static int ext4_check_descriptors(struct super_block *sb,
1980 ext4_group_t *first_not_zeroed)
ac27a0ec 1981{
617ba13b
MC		 1982 struct ext4_sb_info *sbi = EXT4_SB(sb);
1983 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1984 ext4_fsblk_t last_block;
bd81d8ee
LV		 1985 ext4_fsblk_t block_bitmap;
1986 ext4_fsblk_t inode_bitmap;
1987 ext4_fsblk_t inode_table;
ce421581 1988 int flexbg_flag = 0;
bfff6873 1989 ext4_group_t i, grp = sbi->s_groups_count;
ac27a0ec 1990
ce421581
JS		 1991 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1992 flexbg_flag = 1;
1993
af5bc92d 1994 ext4_debug("Checking group descriptors");
ac27a0ec 1995
197cd65a
AM		 1996 for (i = 0; i < sbi->s_groups_count; i++) {
1997 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1998
ce421581 1999 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 2000 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 2001 else
2002 last_block = first_block +
617ba13b 2003 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 2004
bfff6873
LC		 2005 if ((grp == sbi->s_groups_count) &&
2006 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2007 grp = i;
2008
8fadc143 2009 block_bitmap = ext4_block_bitmap(sb, gdp);
2b2d6d01 2010 if (block_bitmap < first_block || block_bitmap > last_block) {
b31e1552 2011 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2012 "Block bitmap for group %u not in group "
b31e1552 2013 "(block %llu)!", i, block_bitmap);
ac27a0ec
DK		 2014 return 0;
2015 }
8fadc143 2016 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2b2d6d01 2017 if (inode_bitmap < first_block || inode_bitmap > last_block) {
b31e1552 2018 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2019 "Inode bitmap for group %u not in group "
b31e1552 2020 "(block %llu)!", i, inode_bitmap);
ac27a0ec
DK		 2021 return 0;
2022 }
8fadc143 2023 inode_table = ext4_inode_table(sb, gdp);
bd81d8ee 2024 if (inode_table < first_block ||
2b2d6d01 2025 inode_table + sbi->s_itb_per_group - 1 > last_block) {
b31e1552 2026 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2027 "Inode table for group %u not in group "
b31e1552 2028 "(block %llu)!", i, inode_table);
ac27a0ec
DK		 2029 return 0;
2030 }
955ce5f5 2031 ext4_lock_group(sb, i);
717d50e4 2032 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
b31e1552
ES		 2033 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2034 "Checksum for group %u failed (%u!=%u)",
2035 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2036 gdp)), le16_to_cpu(gdp->bg_checksum));
7ee1ec4c 2037 if (!(sb->s_flags & MS_RDONLY)) {
955ce5f5 2038 ext4_unlock_group(sb, i);
8a266467 2039 return 0;
7ee1ec4c 2040 }
717d50e4 2041 }
955ce5f5 2042 ext4_unlock_group(sb, i);
ce421581
JS		 2043 if (!flexbg_flag)
2044 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec 2045 }
bfff6873
LC		 2046 if (NULL != first_not_zeroed)
2047 *first_not_zeroed = grp;
ac27a0ec 2048
bd81d8ee 2049 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
0b8e58a1 2050 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
ac27a0ec
DK		 2051 return 1;
2052}
2053
617ba13b 2054/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 2055 * the superblock) which were deleted from all directories, but held open by
2056 * a process at the time of a crash. We walk the list and try to delete these
2057 * inodes at recovery time (only with a read-write filesystem).
2058 *
2059 * In order to keep the orphan inode chain consistent during traversal (in
2060 * case of crash during recovery), we link each inode into the superblock
2061 * orphan list_head and handle it the same way as an inode deletion during
2062 * normal operation (which journals the operations for us).
2063 *
2064 * We only do an iget() and an iput() on each inode, which is very safe if we
2065 * accidentally point at an in-use or already deleted inode. The worst that
2066 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 2067 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 2068 * e2fsck was run on this filesystem, and it must have already done the orphan
2069 * inode cleanup for us, so we can safely abort without any further action.
2070 */
2b2d6d01
TT		 2071static void ext4_orphan_cleanup(struct super_block *sb,
2072 struct ext4_super_block *es)
ac27a0ec
DK		 2073{
2074 unsigned int s_flags = sb->s_flags;
2075 int nr_orphans = 0, nr_truncates = 0;
2076#ifdef CONFIG_QUOTA
2077 int i;
2078#endif
2079 if (!es->s_last_orphan) {
2080 jbd_debug(4, "no orphan inodes to clean up\n");
2081 return;
2082 }
2083
a8f48a95 2084 if (bdev_read_only(sb->s_bdev)) {
b31e1552
ES		 2085 ext4_msg(sb, KERN_ERR, "write access "
2086 "unavailable, skipping orphan cleanup");
a8f48a95
ES		 2087 return;
2088 }
2089
617ba13b 2090 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
ac27a0ec
DK		 2091 if (es->s_last_orphan)
2092 jbd_debug(1, "Errors on filesystem, "
2093 "clearing orphan list.\n");
2094 es->s_last_orphan = 0;
2095 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2096 return;
2097 }
2098
2099 if (s_flags & MS_RDONLY) {
b31e1552 2100 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
ac27a0ec
DK		 2101 sb->s_flags &= ~MS_RDONLY;
2102 }
2103#ifdef CONFIG_QUOTA
2104 /* Needed for iput() to work correctly and not trash data */
2105 sb->s_flags |= MS_ACTIVE;
2106 /* Turn on quotas so that they are updated correctly */
2107 for (i = 0; i < MAXQUOTAS; i++) {
617ba13b
MC		 2108 if (EXT4_SB(sb)->s_qf_names[i]) {
2109 int ret = ext4_quota_on_mount(sb, i);
ac27a0ec 2110 if (ret < 0)
b31e1552
ES		 2111 ext4_msg(sb, KERN_ERR,
2112 "Cannot turn on journaled "
2113 "quota: error %d", ret);
ac27a0ec
DK		 2114 }
2115 }
2116#endif
2117
2118 while (es->s_last_orphan) {
2119 struct inode *inode;
2120
97bd42b9
JB		 2121 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2122 if (IS_ERR(inode)) {
ac27a0ec
DK		 2123 es->s_last_orphan = 0;
2124 break;
2125 }
2126
617ba13b 2127 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
871a2931 2128 dquot_initialize(inode);
ac27a0ec 2129 if (inode->i_nlink) {
b31e1552
ES		 2130 ext4_msg(sb, KERN_DEBUG,
2131 "%s: truncating inode %lu to %lld bytes",
46e665e9 2132 __func__, inode->i_ino, inode->i_size);
e5f8eab8 2133 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
ac27a0ec 2134 inode->i_ino, inode->i_size);
617ba13b 2135 ext4_truncate(inode);
ac27a0ec
DK		 2136 nr_truncates++;
2137 } else {
b31e1552
ES		 2138 ext4_msg(sb, KERN_DEBUG,
2139 "%s: deleting unreferenced inode %lu",
46e665e9 2140 __func__, inode->i_ino);
ac27a0ec
DK		 2141 jbd_debug(2, "deleting unreferenced inode %lu\n",
2142 inode->i_ino);
2143 nr_orphans++;
2144 }
2145 iput(inode); /* The delete magic happens here! */
2146 }
2147
2b2d6d01 2148#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
ac27a0ec
DK		 2149
2150 if (nr_orphans)
b31e1552
ES		 2151 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2152 PLURAL(nr_orphans));
ac27a0ec 2153 if (nr_truncates)
b31e1552
ES		 2154 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2155 PLURAL(nr_truncates));
ac27a0ec
DK		 2156#ifdef CONFIG_QUOTA
2157 /* Turn quotas off */
2158 for (i = 0; i < MAXQUOTAS; i++) {
2159 if (sb_dqopt(sb)->files[i])
287a8095 2160 dquot_quota_off(sb, i);
ac27a0ec
DK		 2161 }
2162#endif
2163 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2164}
0b8e58a1 2165
cd2291a4
ES		 2166/*
2167 * Maximal extent format file size.
2168 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2169 * extent format containers, within a sector_t, and within i_blocks
2170 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2171 * so that won't be a limiting factor.
2172 *
2173 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2174 */
f287a1a5 2175static loff_t ext4_max_size(int blkbits, int has_huge_files)
cd2291a4
ES		 2176{
2177 loff_t res;
2178 loff_t upper_limit = MAX_LFS_FILESIZE;
2179
2180 /* small i_blocks in vfs inode? */
f287a1a5 2181 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
cd2291a4 2182 /*
90c699a9 2183 * CONFIG_LBDAF is not enabled implies the inode
cd2291a4
ES		 2184 * i_block represent total blocks in 512 bytes
2185 * 32 == size of vfs inode i_blocks * 8
2186 */
2187 upper_limit = (1LL << 32) - 1;
2188
2189 /* total blocks in file system block size */
2190 upper_limit >>= (blkbits - 9);
2191 upper_limit <<= blkbits;
2192 }
2193
2194 /* 32-bit extent-start container, ee_block */
2195 res = 1LL << 32;
2196 res <<= blkbits;
2197 res -= 1;
2198
2199 /* Sanity check against vm- & vfs- imposed limits */
2200 if (res > upper_limit)
2201 res = upper_limit;
2202
2203 return res;
2204}
ac27a0ec 2205
ac27a0ec 2206/*
cd2291a4 2207 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 2208 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2209 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 2210 */
f287a1a5 2211static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
ac27a0ec 2212{
617ba13b 2213 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 2214 int meta_blocks;
2215 loff_t upper_limit;
0b8e58a1
AD		 2216 /* This is calculated to be the largest file size for a dense, block
2217 * mapped file such that the file's total number of 512-byte sectors,
2218 * including data and all indirect blocks, does not exceed (2^48 - 1).
2219 *
2220 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2221 * number of 512-byte sectors of the file.
0fc1b451
AK		 2222 */
2223
f287a1a5 2224 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
0fc1b451 2225 /*
90c699a9 2226 * !has_huge_files or CONFIG_LBDAF not enabled implies that
0b8e58a1
AD		 2227 * the inode i_block field represents total file blocks in
2228 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
0fc1b451
AK		 2229 */
2230 upper_limit = (1LL << 32) - 1;
2231
2232 /* total blocks in file system block size */
2233 upper_limit >>= (bits - 9);
2234
2235 } else {
8180a562
AK		 2236 /*
2237 * We use 48 bit ext4_inode i_blocks
2238 * With EXT4_HUGE_FILE_FL set the i_blocks
2239 * represent total number of blocks in
2240 * file system block size
2241 */
0fc1b451
AK		 2242 upper_limit = (1LL << 48) - 1;
2243
0fc1b451
AK		 2244 }
2245
2246 /* indirect blocks */
2247 meta_blocks = 1;
2248 /* double indirect blocks */
2249 meta_blocks += 1 + (1LL << (bits-2));
2250 /* tripple indirect blocks */
2251 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2252
2253 upper_limit -= meta_blocks;
2254 upper_limit <<= bits;
ac27a0ec
DK		 2255
2256 res += 1LL << (bits-2);
2257 res += 1LL << (2*(bits-2));
2258 res += 1LL << (3*(bits-2));
2259 res <<= bits;
2260 if (res > upper_limit)
2261 res = upper_limit;
0fc1b451
AK		 2262
2263 if (res > MAX_LFS_FILESIZE)
2264 res = MAX_LFS_FILESIZE;
2265
ac27a0ec
DK		 2266 return res;
2267}
2268
617ba13b 2269static ext4_fsblk_t descriptor_loc(struct super_block *sb,
0b8e58a1 2270 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 2271{
617ba13b 2272 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 2273 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 2274 int has_super = 0;
2275
2276 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2277
617ba13b 2278 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
ac27a0ec 2279 nr < first_meta_bg)
70bbb3e0 2280 return logical_sb_block + nr + 1;
ac27a0ec 2281 bg = sbi->s_desc_per_block * nr;
617ba13b 2282 if (ext4_bg_has_super(sb, bg))
ac27a0ec 2283 has_super = 1;
0b8e58a1 2284
617ba13b 2285 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 2286}
2287
c9de560d
AT		 2288/**
2289 * ext4_get_stripe_size: Get the stripe size.
2290 * @sbi: In memory super block info
2291 *
2292 * If we have specified it via mount option, then
2293 * use the mount option value. If the value specified at mount time is
2294 * greater than the blocks per group use the super block value.
2295 * If the super block value is greater than blocks per group return 0.
2296 * Allocator needs it be less than blocks per group.
2297 *
2298 */
2299static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2300{
2301 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2302 unsigned long stripe_width =
2303 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2304
2305 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2306 return sbi->s_stripe;
2307
2308 if (stripe_width <= sbi->s_blocks_per_group)
2309 return stripe_width;
2310
2311 if (stride <= sbi->s_blocks_per_group)
2312 return stride;
2313
2314 return 0;
2315}
ac27a0ec 2316
3197ebdb
TT		 2317/* sysfs supprt */
2318
2319struct ext4_attr {
2320 struct attribute attr;
2321 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
60e6679e 2322 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
3197ebdb
TT		 2323 const char *, size_t);
2324 int offset;
2325};
2326
2327static int parse_strtoul(const char *buf,
2328 unsigned long max, unsigned long *value)
2329{
2330 char *endp;
2331
e7d2860b
AGR		 2332 *value = simple_strtoul(skip_spaces(buf), &endp, 0);
2333 endp = skip_spaces(endp);
3197ebdb
TT		 2334 if (*endp || *value > max)
2335 return -EINVAL;
2336
2337 return 0;
2338}
2339
2340static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2341 struct ext4_sb_info *sbi,
2342 char *buf)
2343{
2344 return snprintf(buf, PAGE_SIZE, "%llu\n",
2345 (s64) percpu_counter_sum(&sbi->s_dirtyblocks_counter));
2346}
2347
2348static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2349 struct ext4_sb_info *sbi, char *buf)
2350{
2351 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2352
f613dfcb
TT		 2353 if (!sb->s_bdev->bd_part)
2354 return snprintf(buf, PAGE_SIZE, "0\n");
3197ebdb
TT		 2355 return snprintf(buf, PAGE_SIZE, "%lu\n",
2356 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2357 sbi->s_sectors_written_start) >> 1);
2358}
2359
2360static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2361 struct ext4_sb_info *sbi, char *buf)
2362{
2363 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2364
f613dfcb
TT		 2365 if (!sb->s_bdev->bd_part)
2366 return snprintf(buf, PAGE_SIZE, "0\n");
3197ebdb 2367 return snprintf(buf, PAGE_SIZE, "%llu\n",
a6b43e38 2368 (unsigned long long)(sbi->s_kbytes_written +
3197ebdb 2369 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
a6b43e38 2370 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
3197ebdb
TT		 2371}
2372
2373static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2374 struct ext4_sb_info *sbi,
2375 const char *buf, size_t count)
2376{
2377 unsigned long t;
2378
2379 if (parse_strtoul(buf, 0x40000000, &t))
2380 return -EINVAL;
2381
f7c43950 2382 if (!is_power_of_2(t))
3197ebdb
TT		 2383 return -EINVAL;
2384
2385 sbi->s_inode_readahead_blks = t;
2386 return count;
2387}
2388
2389static ssize_t sbi_ui_show(struct ext4_attr *a,
0b8e58a1 2390 struct ext4_sb_info *sbi, char *buf)
3197ebdb
TT		 2391{
2392 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2393
2394 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2395}
2396
2397static ssize_t sbi_ui_store(struct ext4_attr *a,
2398 struct ext4_sb_info *sbi,
2399 const char *buf, size_t count)
2400{
2401 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2402 unsigned long t;
2403
2404 if (parse_strtoul(buf, 0xffffffff, &t))
2405 return -EINVAL;
2406 *ui = t;
2407 return count;
2408}
2409
2410#define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2411static struct ext4_attr ext4_attr_##_name = { \
2412 .attr = {.name = __stringify(_name), .mode = _mode }, \
2413 .show = _show, \
2414 .store = _store, \
2415 .offset = offsetof(struct ext4_sb_info, _elname), \
2416}
2417#define EXT4_ATTR(name, mode, show, store) \
2418static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2419
857ac889 2420#define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
3197ebdb
TT		 2421#define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2422#define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2423#define EXT4_RW_ATTR_SBI_UI(name, elname) \
2424 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2425#define ATTR_LIST(name) &ext4_attr_##name.attr
2426
2427EXT4_RO_ATTR(delayed_allocation_blocks);
2428EXT4_RO_ATTR(session_write_kbytes);
2429EXT4_RO_ATTR(lifetime_write_kbytes);
2430EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2431 inode_readahead_blks_store, s_inode_readahead_blks);
11013911 2432EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
3197ebdb
TT		 2433EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2434EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2435EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2436EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2437EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2438EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
55138e0b 2439EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
3197ebdb
TT		 2440
2441static struct attribute *ext4_attrs[] = {
2442 ATTR_LIST(delayed_allocation_blocks),
2443 ATTR_LIST(session_write_kbytes),
2444 ATTR_LIST(lifetime_write_kbytes),
2445 ATTR_LIST(inode_readahead_blks),
11013911 2446 ATTR_LIST(inode_goal),
3197ebdb
TT		 2447 ATTR_LIST(mb_stats),
2448 ATTR_LIST(mb_max_to_scan),
2449 ATTR_LIST(mb_min_to_scan),
2450 ATTR_LIST(mb_order2_req),
2451 ATTR_LIST(mb_stream_req),
2452 ATTR_LIST(mb_group_prealloc),
55138e0b 2453 ATTR_LIST(max_writeback_mb_bump),
3197ebdb
TT		 2454 NULL,
2455};
2456
857ac889
LC		 2457/* Features this copy of ext4 supports */
2458EXT4_INFO_ATTR(lazy_itable_init);
2459
2460static struct attribute *ext4_feat_attrs[] = {
2461 ATTR_LIST(lazy_itable_init),
2462 NULL,
2463};
2464
3197ebdb
TT		 2465static ssize_t ext4_attr_show(struct kobject *kobj,
2466 struct attribute *attr, char *buf)
2467{
2468 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2469 s_kobj);
2470 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2471
2472 return a->show ? a->show(a, sbi, buf) : 0;
2473}
2474
2475static ssize_t ext4_attr_store(struct kobject *kobj,
2476 struct attribute *attr,
2477 const char *buf, size_t len)
2478{
2479 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2480 s_kobj);
2481 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2482
2483 return a->store ? a->store(a, sbi, buf, len) : 0;
2484}
2485
2486static void ext4_sb_release(struct kobject *kobj)
2487{
2488 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2489 s_kobj);
2490 complete(&sbi->s_kobj_unregister);
2491}
2492
52cf25d0 2493static const struct sysfs_ops ext4_attr_ops = {
3197ebdb
TT		 2494 .show = ext4_attr_show,
2495 .store = ext4_attr_store,
2496};
2497
2498static struct kobj_type ext4_ktype = {
2499 .default_attrs = ext4_attrs,
2500 .sysfs_ops = &ext4_attr_ops,
2501 .release = ext4_sb_release,
2502};
2503
857ac889
LC		 2504static void ext4_feat_release(struct kobject *kobj)
2505{
2506 complete(&ext4_feat->f_kobj_unregister);
2507}
2508
2509static struct kobj_type ext4_feat_ktype = {
2510 .default_attrs = ext4_feat_attrs,
2511 .sysfs_ops = &ext4_attr_ops,
2512 .release = ext4_feat_release,
2513};
2514
a13fb1a4
ES		 2515/*
2516 * Check whether this filesystem can be mounted based on
2517 * the features present and the RDONLY/RDWR mount requested.
2518 * Returns 1 if this filesystem can be mounted as requested,
2519 * 0 if it cannot be.
2520 */
2521static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2522{
2523 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2524 ext4_msg(sb, KERN_ERR,
2525 "Couldn't mount because of "
2526 "unsupported optional features (%x)",
2527 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2528 ~EXT4_FEATURE_INCOMPAT_SUPP));
2529 return 0;
2530 }
2531
2532 if (readonly)
2533 return 1;
2534
2535 /* Check that feature set is OK for a read-write mount */
2536 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2537 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2538 "unsupported optional features (%x)",
2539 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2540 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2541 return 0;
2542 }
2543 /*
2544 * Large file size enabled file system can only be mounted
2545 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2546 */
2547 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2548 if (sizeof(blkcnt_t) < sizeof(u64)) {
2549 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2550 "cannot be mounted RDWR without "
2551 "CONFIG_LBDAF");
2552 return 0;
2553 }
2554 }
2555 return 1;
2556}
2557
66e61a9e
TT		 2558/*
2559 * This function is called once a day if we have errors logged
2560 * on the file system
2561 */
2562static void print_daily_error_info(unsigned long arg)
2563{
2564 struct super_block *sb = (struct super_block *) arg;
2565 struct ext4_sb_info *sbi;
2566 struct ext4_super_block *es;
2567
2568 sbi = EXT4_SB(sb);
2569 es = sbi->s_es;
2570
2571 if (es->s_error_count)
2572 ext4_msg(sb, KERN_NOTICE, "error count: %u",
2573 le32_to_cpu(es->s_error_count));
2574 if (es->s_first_error_time) {
2575 printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d",
2576 sb->s_id, le32_to_cpu(es->s_first_error_time),
2577 (int) sizeof(es->s_first_error_func),
2578 es->s_first_error_func,
2579 le32_to_cpu(es->s_first_error_line));
2580 if (es->s_first_error_ino)
2581 printk(": inode %u",
2582 le32_to_cpu(es->s_first_error_ino));
2583 if (es->s_first_error_block)
2584 printk(": block %llu", (unsigned long long)
2585 le64_to_cpu(es->s_first_error_block));
2586 printk("\n");
2587 }
2588 if (es->s_last_error_time) {
2589 printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d",
2590 sb->s_id, le32_to_cpu(es->s_last_error_time),
2591 (int) sizeof(es->s_last_error_func),
2592 es->s_last_error_func,
2593 le32_to_cpu(es->s_last_error_line));
2594 if (es->s_last_error_ino)
2595 printk(": inode %u",
2596 le32_to_cpu(es->s_last_error_ino));
2597 if (es->s_last_error_block)
2598 printk(": block %llu", (unsigned long long)
2599 le64_to_cpu(es->s_last_error_block));
2600 printk("\n");
2601 }
2602 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2603}
2604
bfff6873
LC		 2605static void ext4_lazyinode_timeout(unsigned long data)
2606{
2607 struct task_struct *p = (struct task_struct *)data;
2608 wake_up_process(p);
2609}
2610
2611/* Find next suitable group and run ext4_init_inode_table */
2612static int ext4_run_li_request(struct ext4_li_request *elr)
2613{
2614 struct ext4_group_desc *gdp = NULL;
2615 ext4_group_t group, ngroups;
2616 struct super_block *sb;
2617 unsigned long timeout = 0;
2618 int ret = 0;
2619
2620 sb = elr->lr_super;
2621 ngroups = EXT4_SB(sb)->s_groups_count;
2622
2623 for (group = elr->lr_next_group; group < ngroups; group++) {
2624 gdp = ext4_get_group_desc(sb, group, NULL);
2625 if (!gdp) {
2626 ret = 1;
2627 break;
2628 }
2629
2630 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2631 break;
2632 }
2633
2634 if (group == ngroups)
2635 ret = 1;
2636
2637 if (!ret) {
2638 timeout = jiffies;
2639 ret = ext4_init_inode_table(sb, group,
2640 elr->lr_timeout ? 0 : 1);
2641 if (elr->lr_timeout == 0) {
2642 timeout = jiffies - timeout;
2643 if (elr->lr_sbi->s_li_wait_mult)
2644 timeout *= elr->lr_sbi->s_li_wait_mult;
2645 else
2646 timeout *= 20;
2647 elr->lr_timeout = timeout;
2648 }
2649 elr->lr_next_sched = jiffies + elr->lr_timeout;
2650 elr->lr_next_group = group + 1;
2651 }
2652
2653 return ret;
2654}
2655
2656/*
2657 * Remove lr_request from the list_request and free the
2658 * request tructure. Should be called with li_list_mtx held
2659 */
2660static void ext4_remove_li_request(struct ext4_li_request *elr)
2661{
2662 struct ext4_sb_info *sbi;
2663
2664 if (!elr)
2665 return;
2666
2667 sbi = elr->lr_sbi;
2668
2669 list_del(&elr->lr_request);
2670 sbi->s_li_request = NULL;
2671 kfree(elr);
2672}
2673
2674static void ext4_unregister_li_request(struct super_block *sb)
2675{
2676 struct ext4_li_request *elr = EXT4_SB(sb)->s_li_request;
2677
2678 if (!ext4_li_info)
2679 return;
2680
2681 mutex_lock(&ext4_li_info->li_list_mtx);
2682 ext4_remove_li_request(elr);
2683 mutex_unlock(&ext4_li_info->li_list_mtx);
2684}
2685
2686/*
2687 * This is the function where ext4lazyinit thread lives. It walks
2688 * through the request list searching for next scheduled filesystem.
2689 * When such a fs is found, run the lazy initialization request
2690 * (ext4_rn_li_request) and keep track of the time spend in this
2691 * function. Based on that time we compute next schedule time of
2692 * the request. When walking through the list is complete, compute
2693 * next waking time and put itself into sleep.
2694 */
2695static int ext4_lazyinit_thread(void *arg)
2696{
2697 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2698 struct list_head *pos, *n;
2699 struct ext4_li_request *elr;
2700 unsigned long next_wakeup;
2701 DEFINE_WAIT(wait);
2702 int ret;
2703
2704 BUG_ON(NULL == eli);
2705
2706 eli->li_timer.data = (unsigned long)current;
2707 eli->li_timer.function = ext4_lazyinode_timeout;
2708
2709 eli->li_task = current;
2710 wake_up(&eli->li_wait_task);
2711
2712cont_thread:
2713 while (true) {
2714 next_wakeup = MAX_JIFFY_OFFSET;
2715
2716 mutex_lock(&eli->li_list_mtx);
2717 if (list_empty(&eli->li_request_list)) {
2718 mutex_unlock(&eli->li_list_mtx);
2719 goto exit_thread;
2720 }
2721
2722 list_for_each_safe(pos, n, &eli->li_request_list) {
2723 elr = list_entry(pos, struct ext4_li_request,
2724 lr_request);
2725
2726 if (time_after_eq(jiffies, elr->lr_next_sched))
2727 ret = ext4_run_li_request(elr);
2728
2729 if (ret) {
2730 ret = 0;
2731 ext4_remove_li_request(elr);
2732 continue;
2733 }
2734
2735 if (time_before(elr->lr_next_sched, next_wakeup))
2736 next_wakeup = elr->lr_next_sched;
2737 }
2738 mutex_unlock(&eli->li_list_mtx);
2739
2740 if (freezing(current))
2741 refrigerator();
2742
2743 if (time_after_eq(jiffies, next_wakeup)) {
2744 cond_resched();
2745 continue;
2746 }
2747
2748 eli->li_timer.expires = next_wakeup;
2749 add_timer(&eli->li_timer);
2750 prepare_to_wait(&eli->li_wait_daemon, &wait,
2751 TASK_INTERRUPTIBLE);
2752 if (time_before(jiffies, next_wakeup))
2753 schedule();
2754 finish_wait(&eli->li_wait_daemon, &wait);
2755 }
2756
2757exit_thread:
2758 /*
2759 * It looks like the request list is empty, but we need
2760 * to check it under the li_list_mtx lock, to prevent any
2761 * additions into it, and of course we should lock ext4_li_mtx
2762 * to atomically free the list and ext4_li_info, because at
2763 * this point another ext4 filesystem could be registering
2764 * new one.
2765 */
2766 mutex_lock(&ext4_li_mtx);
2767 mutex_lock(&eli->li_list_mtx);
2768 if (!list_empty(&eli->li_request_list)) {
2769 mutex_unlock(&eli->li_list_mtx);
2770 mutex_unlock(&ext4_li_mtx);
2771 goto cont_thread;
2772 }
2773 mutex_unlock(&eli->li_list_mtx);
2774 del_timer_sync(&ext4_li_info->li_timer);
2775 eli->li_task = NULL;
2776 wake_up(&eli->li_wait_task);
2777
2778 kfree(ext4_li_info);
2779 ext4_li_info = NULL;
2780 mutex_unlock(&ext4_li_mtx);
2781
2782 return 0;
2783}
2784
2785static void ext4_clear_request_list(void)
2786{
2787 struct list_head *pos, *n;
2788 struct ext4_li_request *elr;
2789
2790 mutex_lock(&ext4_li_info->li_list_mtx);
2791 if (list_empty(&ext4_li_info->li_request_list))
2792 return;
2793
2794 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2795 elr = list_entry(pos, struct ext4_li_request,
2796 lr_request);
2797 ext4_remove_li_request(elr);
2798 }
2799 mutex_unlock(&ext4_li_info->li_list_mtx);
2800}
2801
2802static int ext4_run_lazyinit_thread(void)
2803{
2804 struct task_struct *t;
2805
2806 t = kthread_run(ext4_lazyinit_thread, ext4_li_info, "ext4lazyinit");
2807 if (IS_ERR(t)) {
2808 int err = PTR_ERR(t);
2809 ext4_clear_request_list();
2810 del_timer_sync(&ext4_li_info->li_timer);
2811 kfree(ext4_li_info);
2812 ext4_li_info = NULL;
2813 printk(KERN_CRIT "EXT4: error %d creating inode table "
2814 "initialization thread\n",
2815 err);
2816 return err;
2817 }
2818 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
2819
2820 wait_event(ext4_li_info->li_wait_task, ext4_li_info->li_task != NULL);
2821 return 0;
2822}
2823
2824/*
2825 * Check whether it make sense to run itable init. thread or not.
2826 * If there is at least one uninitialized inode table, return
2827 * corresponding group number, else the loop goes through all
2828 * groups and return total number of groups.
2829 */
2830static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2831{
2832 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2833 struct ext4_group_desc *gdp = NULL;
2834
2835 for (group = 0; group < ngroups; group++) {
2836 gdp = ext4_get_group_desc(sb, group, NULL);
2837 if (!gdp)
2838 continue;
2839
2840 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2841 break;
2842 }
2843
2844 return group;
2845}
2846
2847static int ext4_li_info_new(void)
2848{
2849 struct ext4_lazy_init *eli = NULL;
2850
2851 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2852 if (!eli)
2853 return -ENOMEM;
2854
2855 eli->li_task = NULL;
2856 INIT_LIST_HEAD(&eli->li_request_list);
2857 mutex_init(&eli->li_list_mtx);
2858
2859 init_waitqueue_head(&eli->li_wait_daemon);
2860 init_waitqueue_head(&eli->li_wait_task);
2861 init_timer(&eli->li_timer);
2862 eli->li_state |= EXT4_LAZYINIT_QUIT;
2863
2864 ext4_li_info = eli;
2865
2866 return 0;
2867}
2868
2869static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
2870 ext4_group_t start)
2871{
2872 struct ext4_sb_info *sbi = EXT4_SB(sb);
2873 struct ext4_li_request *elr;
2874 unsigned long rnd;
2875
2876 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
2877 if (!elr)
2878 return NULL;
2879
2880 elr->lr_super = sb;
2881 elr->lr_sbi = sbi;
2882 elr->lr_next_group = start;
2883
2884 /*
2885 * Randomize first schedule time of the request to
2886 * spread the inode table initialization requests
2887 * better.
2888 */
2889 get_random_bytes(&rnd, sizeof(rnd));
2890 elr->lr_next_sched = jiffies + (unsigned long)rnd %
2891 (EXT4_DEF_LI_MAX_START_DELAY * HZ);
2892
2893 return elr;
2894}
2895
2896static int ext4_register_li_request(struct super_block *sb,
2897 ext4_group_t first_not_zeroed)
2898{
2899 struct ext4_sb_info *sbi = EXT4_SB(sb);
2900 struct ext4_li_request *elr;
2901 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
2902 int ret = 0;
2903
2904 if (sbi->s_li_request != NULL)
2905 goto out;
2906
2907 if (first_not_zeroed == ngroups ||
2908 (sb->s_flags & MS_RDONLY) ||
2909 !test_opt(sb, INIT_INODE_TABLE)) {
2910 sbi->s_li_request = NULL;
2911 goto out;
2912 }
2913
2914 if (first_not_zeroed == ngroups) {
2915 sbi->s_li_request = NULL;
2916 goto out;
2917 }
2918
2919 elr = ext4_li_request_new(sb, first_not_zeroed);
2920 if (!elr) {
2921 ret = -ENOMEM;
2922 goto out;
2923 }
2924
2925 mutex_lock(&ext4_li_mtx);
2926
2927 if (NULL == ext4_li_info) {
2928 ret = ext4_li_info_new();
2929 if (ret)
2930 goto out;
2931 }
2932
2933 mutex_lock(&ext4_li_info->li_list_mtx);
2934 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
2935 mutex_unlock(&ext4_li_info->li_list_mtx);
2936
2937 sbi->s_li_request = elr;
2938
2939 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
2940 ret = ext4_run_lazyinit_thread();
2941 if (ret)
2942 goto out;
2943 }
2944
2945 mutex_unlock(&ext4_li_mtx);
2946
2947out:
2948 if (ret) {
2949 mutex_unlock(&ext4_li_mtx);
2950 kfree(elr);
2951 }
2952 return ret;
2953}
2954
2955/*
2956 * We do not need to lock anything since this is called on
2957 * module unload.
2958 */
2959static void ext4_destroy_lazyinit_thread(void)
2960{
2961 /*
2962 * If thread exited earlier
2963 * there's nothing to be done.
2964 */
2965 if (!ext4_li_info)
2966 return;
2967
2968 ext4_clear_request_list();
2969
2970 while (ext4_li_info->li_task) {
2971 wake_up(&ext4_li_info->li_wait_daemon);
2972 wait_event(ext4_li_info->li_wait_task,
2973 ext4_li_info->li_task == NULL);
2974 }
2975}
2976
2b2d6d01 2977static int ext4_fill_super(struct super_block *sb, void *data, int silent)
7477827f
AK		 2978 __releases(kernel_lock)
2979 __acquires(kernel_lock)
ac27a0ec 2980{
d4c402d9 2981 char *orig_data = kstrdup(data, GFP_KERNEL);
2b2d6d01 2982 struct buffer_head *bh;
617ba13b
MC		 2983 struct ext4_super_block *es = NULL;
2984 struct ext4_sb_info *sbi;
2985 ext4_fsblk_t block;
2986 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 2987 ext4_fsblk_t logical_sb_block;
ac27a0ec 2988 unsigned long offset = 0;
ac27a0ec
DK		 2989 unsigned long journal_devnum = 0;
2990 unsigned long def_mount_opts;
2991 struct inode *root;
9f6200bb 2992 char *cp;
0390131b 2993 const char *descr;
dcc7dae3 2994 int ret = -ENOMEM;
ac27a0ec 2995 int blocksize;
4ec11028
TT		 2996 unsigned int db_count;
2997 unsigned int i;
f287a1a5 2998 int needs_recovery, has_huge_files;
bd81d8ee 2999 __u64 blocks_count;
833f4077 3000 int err;
b3881f74 3001 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
bfff6873 3002 ext4_group_t first_not_zeroed;
ac27a0ec
DK		 3003
3004 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3005 if (!sbi)
dcc7dae3 3006 goto out_free_orig;
705895b6
PE		 3007
3008 sbi->s_blockgroup_lock =
3009 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3010 if (!sbi->s_blockgroup_lock) {
3011 kfree(sbi);
dcc7dae3 3012 goto out_free_orig;
705895b6 3013 }
ac27a0ec
DK		 3014 sb->s_fs_info = sbi;
3015 sbi->s_mount_opt = 0;
617ba13b
MC		 3016 sbi->s_resuid = EXT4_DEF_RESUID;
3017 sbi->s_resgid = EXT4_DEF_RESGID;
240799cd 3018 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
d9c9bef1 3019 sbi->s_sb_block = sb_block;
f613dfcb
TT		 3020 if (sb->s_bdev->bd_part)
3021 sbi->s_sectors_written_start =
3022 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
ac27a0ec
DK		 3023
3024 unlock_kernel();
3025
9f6200bb
TT		 3026 /* Cleanup superblock name */
3027 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3028 *cp = '!';
3029
dcc7dae3 3030 ret = -EINVAL;
617ba13b 3031 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 3032 if (!blocksize) {
b31e1552 3033 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
ac27a0ec
DK		 3034 goto out_fail;
3035 }
3036
3037 /*
617ba13b 3038 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 3039 * block sizes. We need to calculate the offset from buffer start.
3040 */
617ba13b 3041 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 3042 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3043 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 3044 } else {
70bbb3e0 3045 logical_sb_block = sb_block;
ac27a0ec
DK		 3046 }
3047
70bbb3e0 3048 if (!(bh = sb_bread(sb, logical_sb_block))) {
b31e1552 3049 ext4_msg(sb, KERN_ERR, "unable to read superblock");
ac27a0ec
DK		 3050 goto out_fail;
3051 }
3052 /*
3053 * Note: s_es must be initialized as soon as possible because
617ba13b 3054 * some ext4 macro-instructions depend on its value
ac27a0ec 3055 */
617ba13b 3056 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
ac27a0ec
DK		 3057 sbi->s_es = es;
3058 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 3059 if (sb->s_magic != EXT4_SUPER_MAGIC)
3060 goto cantfind_ext4;
afc32f7e 3061 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
ac27a0ec
DK		 3062
3063 /* Set defaults before we parse the mount options */
3064 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
bfff6873 3065 set_opt(sbi->s_mount_opt, INIT_INODE_TABLE);
617ba13b 3066 if (def_mount_opts & EXT4_DEFM_DEBUG)
ac27a0ec 3067 set_opt(sbi->s_mount_opt, DEBUG);
437ca0fd
DM		 3068 if (def_mount_opts & EXT4_DEFM_BSDGROUPS) {
3069 ext4_msg(sb, KERN_WARNING, deprecated_msg, "bsdgroups",
3070 "2.6.38");
ac27a0ec 3071 set_opt(sbi->s_mount_opt, GRPID);
437ca0fd 3072 }
617ba13b 3073 if (def_mount_opts & EXT4_DEFM_UID16)
ac27a0ec 3074 set_opt(sbi->s_mount_opt, NO_UID32);
03010a33 3075#ifdef CONFIG_EXT4_FS_XATTR
617ba13b 3076 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
ac27a0ec 3077 set_opt(sbi->s_mount_opt, XATTR_USER);
2e7842b8 3078#endif
03010a33 3079#ifdef CONFIG_EXT4_FS_POSIX_ACL
617ba13b 3080 if (def_mount_opts & EXT4_DEFM_ACL)
ac27a0ec 3081 set_opt(sbi->s_mount_opt, POSIX_ACL);
2e7842b8 3082#endif
617ba13b 3083 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
482a7425 3084 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
617ba13b 3085 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
482a7425 3086 set_opt(sbi->s_mount_opt, ORDERED_DATA);
617ba13b 3087 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
482a7425 3088 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
617ba13b
MC		 3089
3090 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
ac27a0ec 3091 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
bb4f397a 3092 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
ceea16bf 3093 set_opt(sbi->s_mount_opt, ERRORS_CONT);
bb4f397a
AK		 3094 else
3095 set_opt(sbi->s_mount_opt, ERRORS_RO);
8b67f04a
TT		 3096 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
3097 set_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
3098 if (def_mount_opts & EXT4_DEFM_DISCARD)
3099 set_opt(sbi->s_mount_opt, DISCARD);
ac27a0ec
DK		 3100
3101 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
3102 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
30773840
TT		 3103 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3104 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3105 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
ac27a0ec 3106
8b67f04a
TT		 3107 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3108 set_opt(sbi->s_mount_opt, BARRIER);
ac27a0ec 3109
dd919b98
AK		 3110 /*
3111 * enable delayed allocation by default
3112 * Use -o nodelalloc to turn it off
3113 */
8b67f04a
TT		 3114 if (!IS_EXT3_SB(sb) &&
3115 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
ba69f9ab 3116 set_opt(sbi->s_mount_opt, DELALLOC);
dd919b98 3117
8b67f04a
TT		 3118 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3119 &journal_devnum, &journal_ioprio, NULL, 0)) {
3120 ext4_msg(sb, KERN_WARNING,
3121 "failed to parse options in superblock: %s",
3122 sbi->s_es->s_mount_opts);
3123 }
b3881f74
TT		 3124 if (!parse_options((char *) data, sb, &journal_devnum,
3125 &journal_ioprio, NULL, 0))
ac27a0ec
DK		 3126 goto failed_mount;
3127
3128 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 3129 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec 3130
617ba13b
MC		 3131 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3132 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3133 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3134 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
b31e1552
ES		 3135 ext4_msg(sb, KERN_WARNING,
3136 "feature flags set on rev 0 fs, "
3137 "running e2fsck is recommended");
469108ff 3138
ac27a0ec
DK		 3139 /*
3140 * Check feature flags regardless of the revision level, since we
3141 * previously didn't change the revision level when setting the flags,
3142 * so there is a chance incompat flags are set on a rev 0 filesystem.
3143 */
a13fb1a4 3144 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
ac27a0ec 3145 goto failed_mount;
a13fb1a4 3146
ac27a0ec
DK		 3147 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3148
617ba13b
MC		 3149 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3150 blocksize > EXT4_MAX_BLOCK_SIZE) {
b31e1552
ES		 3151 ext4_msg(sb, KERN_ERR,
3152 "Unsupported filesystem blocksize %d", blocksize);
ac27a0ec
DK		 3153 goto failed_mount;
3154 }
3155
ac27a0ec 3156 if (sb->s_blocksize != blocksize) {
ce40733c
AK		 3157 /* Validate the filesystem blocksize */
3158 if (!sb_set_blocksize(sb, blocksize)) {
b31e1552 3159 ext4_msg(sb, KERN_ERR, "bad block size %d",
ce40733c 3160 blocksize);
ac27a0ec
DK		 3161 goto failed_mount;
3162 }
3163
2b2d6d01 3164 brelse(bh);
70bbb3e0
AM		 3165 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3166 offset = do_div(logical_sb_block, blocksize);
3167 bh = sb_bread(sb, logical_sb_block);
ac27a0ec 3168 if (!bh) {
b31e1552
ES		 3169 ext4_msg(sb, KERN_ERR,
3170 "Can't read superblock on 2nd try");
ac27a0ec
DK		 3171 goto failed_mount;
3172 }
617ba13b 3173 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
ac27a0ec 3174 sbi->s_es = es;
617ba13b 3175 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
b31e1552
ES		 3176 ext4_msg(sb, KERN_ERR,
3177 "Magic mismatch, very weird!");
ac27a0ec
DK		 3178 goto failed_mount;
3179 }
3180 }
3181
a13fb1a4
ES		 3182 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3183 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
f287a1a5
TT		 3184 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3185 has_huge_files);
3186 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
ac27a0ec 3187
617ba13b
MC		 3188 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3189 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3190 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
ac27a0ec
DK		 3191 } else {
3192 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3193 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
617ba13b 3194 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1330593e 3195 (!is_power_of_2(sbi->s_inode_size)) ||
ac27a0ec 3196 (sbi->s_inode_size > blocksize)) {
b31e1552
ES		 3197 ext4_msg(sb, KERN_ERR,
3198 "unsupported inode size: %d",
2b2d6d01 3199 sbi->s_inode_size);
ac27a0ec
DK		 3200 goto failed_mount;
3201 }
ef7f3835
KS		 3202 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3203 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
ac27a0ec 3204 }
0b8e58a1 3205
0d1ee42f
AR		 3206 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3207 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
8fadc143 3208 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 3209 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 3210 !is_power_of_2(sbi->s_desc_size)) {
b31e1552
ES		 3211 ext4_msg(sb, KERN_ERR,
3212 "unsupported descriptor size %lu",
0d1ee42f
AR		 3213 sbi->s_desc_size);
3214 goto failed_mount;
3215 }
3216 } else
3217 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
0b8e58a1 3218
ac27a0ec 3219 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 3220 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
b47b6f38 3221 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
617ba13b 3222 goto cantfind_ext4;
0b8e58a1 3223
617ba13b 3224 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 3225 if (sbi->s_inodes_per_block == 0)
617ba13b 3226 goto cantfind_ext4;
ac27a0ec
DK		 3227 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3228 sbi->s_inodes_per_block;
0d1ee42f 3229 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 3230 sbi->s_sbh = bh;
3231 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 3232 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3233 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
0b8e58a1 3234
2b2d6d01 3235 for (i = 0; i < 4; i++)
ac27a0ec
DK		 3236 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3237 sbi->s_def_hash_version = es->s_def_hash_version;
f99b2589
TT		 3238 i = le32_to_cpu(es->s_flags);
3239 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3240 sbi->s_hash_unsigned = 3;
3241 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3242#ifdef __CHAR_UNSIGNED__
3243 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3244 sbi->s_hash_unsigned = 3;
3245#else
3246 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3247#endif
3248 sb->s_dirt = 1;
3249 }
ac27a0ec
DK		 3250
3251 if (sbi->s_blocks_per_group > blocksize * 8) {
b31e1552
ES		 3252 ext4_msg(sb, KERN_ERR,
3253 "#blocks per group too big: %lu",
2b2d6d01 3254 sbi->s_blocks_per_group);
ac27a0ec
DK		 3255 goto failed_mount;
3256 }
ac27a0ec 3257 if (sbi->s_inodes_per_group > blocksize * 8) {
b31e1552
ES		 3258 ext4_msg(sb, KERN_ERR,
3259 "#inodes per group too big: %lu",
2b2d6d01 3260 sbi->s_inodes_per_group);
ac27a0ec
DK		 3261 goto failed_mount;
3262 }
3263
bf43d84b
ES		 3264 /*
3265 * Test whether we have more sectors than will fit in sector_t,
3266 * and whether the max offset is addressable by the page cache.
3267 */
3268 if ((ext4_blocks_count(es) >
3269 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) ||
3270 (ext4_blocks_count(es) >
3271 (pgoff_t)(~0ULL) >> (PAGE_CACHE_SHIFT - sb->s_blocksize_bits))) {
b31e1552 3272 ext4_msg(sb, KERN_ERR, "filesystem"
bf43d84b 3273 " too large to mount safely on this system");
ac27a0ec 3274 if (sizeof(sector_t) < 8)
90c699a9 3275 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
bf43d84b 3276 ret = -EFBIG;
ac27a0ec
DK		 3277 goto failed_mount;
3278 }
3279
617ba13b
MC		 3280 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3281 goto cantfind_ext4;
e7c95593 3282
0f2ddca6
FTN		 3283 /* check blocks count against device size */
3284 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3285 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
b31e1552
ES		 3286 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3287 "exceeds size of device (%llu blocks)",
0f2ddca6
FTN		 3288 ext4_blocks_count(es), blocks_count);
3289 goto failed_mount;
3290 }
3291
0b8e58a1
AD		 3292 /*
3293 * It makes no sense for the first data block to be beyond the end
3294 * of the filesystem.
3295 */
3296 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
b31e1552
ES		 3297 ext4_msg(sb, KERN_WARNING, "bad geometry: first data"
3298 "block %u is beyond end of filesystem (%llu)",
3299 le32_to_cpu(es->s_first_data_block),
3300 ext4_blocks_count(es));
e7c95593
ES		 3301 goto failed_mount;
3302 }
bd81d8ee
LV		 3303 blocks_count = (ext4_blocks_count(es) -
3304 le32_to_cpu(es->s_first_data_block) +
3305 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3306 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4ec11028 3307 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
b31e1552 3308 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
4ec11028 3309 "(block count %llu, first data block %u, "
b31e1552 3310 "blocks per group %lu)", sbi->s_groups_count,
4ec11028
TT		 3311 ext4_blocks_count(es),
3312 le32_to_cpu(es->s_first_data_block),
3313 EXT4_BLOCKS_PER_GROUP(sb));
3314 goto failed_mount;
3315 }
bd81d8ee 3316 sbi->s_groups_count = blocks_count;
fb0a387d
ES		 3317 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3318 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
617ba13b
MC		 3319 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3320 EXT4_DESC_PER_BLOCK(sb);
2b2d6d01 3321 sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
ac27a0ec
DK		 3322 GFP_KERNEL);
3323 if (sbi->s_group_desc == NULL) {
b31e1552 3324 ext4_msg(sb, KERN_ERR, "not enough memory");
ac27a0ec
DK		 3325 goto failed_mount;
3326 }
3327
3244fcb1 3328#ifdef CONFIG_PROC_FS
9f6200bb
TT		 3329 if (ext4_proc_root)
3330 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
3244fcb1 3331#endif
240799cd 3332
705895b6 3333 bgl_lock_init(sbi->s_blockgroup_lock);
ac27a0ec
DK		 3334
3335 for (i = 0; i < db_count; i++) {
70bbb3e0 3336 block = descriptor_loc(sb, logical_sb_block, i);
ac27a0ec
DK		 3337 sbi->s_group_desc[i] = sb_bread(sb, block);
3338 if (!sbi->s_group_desc[i]) {
b31e1552
ES		 3339 ext4_msg(sb, KERN_ERR,
3340 "can't read group descriptor %d", i);
ac27a0ec
DK		 3341 db_count = i;
3342 goto failed_mount2;
3343 }
3344 }
bfff6873 3345 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
b31e1552 3346 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
ac27a0ec
DK		 3347 goto failed_mount2;
3348 }
772cb7c8
JS		 3349 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3350 if (!ext4_fill_flex_info(sb)) {
b31e1552
ES		 3351 ext4_msg(sb, KERN_ERR,
3352 "unable to initialize "
3353 "flex_bg meta info!");
772cb7c8
JS		 3354 goto failed_mount2;
3355 }
3356
ac27a0ec
DK		 3357 sbi->s_gdb_count = db_count;
3358 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3359 spin_lock_init(&sbi->s_next_gen_lock);
3360
c9de560d 3361 sbi->s_stripe = ext4_get_stripe_size(sbi);
55138e0b 3362 sbi->s_max_writeback_mb_bump = 128;
c9de560d 3363
ac27a0ec
DK		 3364 /*
3365 * set up enough so that it can read an inode
3366 */
9ca92389
TT		 3367 if (!test_opt(sb, NOLOAD) &&
3368 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3369 sb->s_op = &ext4_sops;
3370 else
3371 sb->s_op = &ext4_nojournal_sops;
617ba13b
MC		 3372 sb->s_export_op = &ext4_export_ops;
3373 sb->s_xattr = ext4_xattr_handlers;
ac27a0ec 3374#ifdef CONFIG_QUOTA
617ba13b
MC		 3375 sb->s_qcop = &ext4_qctl_operations;
3376 sb->dq_op = &ext4_quota_operations;
ac27a0ec
DK		 3377#endif
3378 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3b9d4ed2 3379 mutex_init(&sbi->s_orphan_lock);
32ed5058 3380 mutex_init(&sbi->s_resize_lock);
ac27a0ec
DK		 3381
3382 sb->s_root = NULL;
3383
3384 needs_recovery = (es->s_last_orphan != 0 ||
617ba13b
MC		 3385 EXT4_HAS_INCOMPAT_FEATURE(sb,
3386 EXT4_FEATURE_INCOMPAT_RECOVER));
ac27a0ec
DK		 3387
3388 /*
3389 * The first inode we look at is the journal inode. Don't try
3390 * root first: it may be modified in the journal!
3391 */
3392 if (!test_opt(sb, NOLOAD) &&
617ba13b
MC		 3393 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3394 if (ext4_load_journal(sb, es, journal_devnum))
ac27a0ec 3395 goto failed_mount3;
0390131b
FM		 3396 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3397 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
b31e1552
ES		 3398 ext4_msg(sb, KERN_ERR, "required journal recovery "
3399 "suppressed and not mounted read-only");
744692dc 3400 goto failed_mount_wq;
ac27a0ec 3401 } else {
0390131b
FM		 3402 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
3403 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
3404 sbi->s_journal = NULL;
3405 needs_recovery = 0;
3406 goto no_journal;
ac27a0ec
DK		 3407 }
3408
eb40a09c
JS		 3409 if (ext4_blocks_count(es) > 0xffffffffULL &&
3410 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3411 JBD2_FEATURE_INCOMPAT_64BIT)) {
b31e1552 3412 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
744692dc 3413 goto failed_mount_wq;
eb40a09c
JS		 3414 }
3415
d4da6c9c
LT		 3416 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3417 jbd2_journal_set_features(sbi->s_journal,
3418 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
818d276c 3419 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
d4da6c9c
LT		 3420 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3421 jbd2_journal_set_features(sbi->s_journal,
3422 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
818d276c
GS		 3423 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3424 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
d4da6c9c
LT		 3425 } else {
3426 jbd2_journal_clear_features(sbi->s_journal,
3427 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3428 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3429 }
818d276c 3430
ac27a0ec
DK		 3431 /* We have now updated the journal if required, so we can
3432 * validate the data journaling mode. */
3433 switch (test_opt(sb, DATA_FLAGS)) {
3434 case 0:
3435 /* No mode set, assume a default based on the journal
63f57933
AM		 3436 * capabilities: ORDERED_DATA if the journal can
3437 * cope, else JOURNAL_DATA
3438 */
dab291af
MC		 3439 if (jbd2_journal_check_available_features
3440 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
ac27a0ec
DK		 3441 set_opt(sbi->s_mount_opt, ORDERED_DATA);
3442 else
3443 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
3444 break;
3445
617ba13b
MC		 3446 case EXT4_MOUNT_ORDERED_DATA:
3447 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 3448 if (!jbd2_journal_check_available_features
3449 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
b31e1552
ES		 3450 ext4_msg(sb, KERN_ERR, "Journal does not support "
3451 "requested data journaling mode");
744692dc 3452 goto failed_mount_wq;
ac27a0ec
DK		 3453 }
3454 default:
3455 break;
3456 }
b3881f74 3457 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
ac27a0ec 3458
0390131b 3459no_journal:
84061e07
DM		 3460 err = percpu_counter_init(&sbi->s_freeblocks_counter,
3461 ext4_count_free_blocks(sb));
3462 if (!err)
3463 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3464 ext4_count_free_inodes(sb));
3465 if (!err)
3466 err = percpu_counter_init(&sbi->s_dirs_counter,
3467 ext4_count_dirs(sb));
3468 if (!err)
3469 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
3470 if (err) {
3471 ext4_msg(sb, KERN_ERR, "insufficient memory");
3472 goto failed_mount_wq;
3473 }
206f7ab4 3474
4c0425ff
MC		 3475 EXT4_SB(sb)->dio_unwritten_wq = create_workqueue("ext4-dio-unwritten");
3476 if (!EXT4_SB(sb)->dio_unwritten_wq) {
3477 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
3478 goto failed_mount_wq;
3479 }
3480
ac27a0ec 3481 /*
dab291af 3482 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 3483 * so we can safely mount the rest of the filesystem now.
3484 */
3485
1d1fe1ee
DH		 3486 root = ext4_iget(sb, EXT4_ROOT_INO);
3487 if (IS_ERR(root)) {
b31e1552 3488 ext4_msg(sb, KERN_ERR, "get root inode failed");
1d1fe1ee 3489 ret = PTR_ERR(root);
ac27a0ec
DK		 3490 goto failed_mount4;
3491 }
3492 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1d1fe1ee 3493 iput(root);
b31e1552 3494 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
ac27a0ec
DK		 3495 goto failed_mount4;
3496 }
1d1fe1ee
DH		 3497 sb->s_root = d_alloc_root(root);
3498 if (!sb->s_root) {
b31e1552 3499 ext4_msg(sb, KERN_ERR, "get root dentry failed");
1d1fe1ee
DH		 3500 iput(root);
3501 ret = -ENOMEM;
3502 goto failed_mount4;
3503 }
ac27a0ec 3504
2b2d6d01 3505 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
ef7f3835
KS		 3506
3507 /* determine the minimum size of new large inodes, if present */
3508 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3509 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3510 EXT4_GOOD_OLD_INODE_SIZE;
3511 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3512 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
3513 if (sbi->s_want_extra_isize <
3514 le16_to_cpu(es->s_want_extra_isize))
3515 sbi->s_want_extra_isize =
3516 le16_to_cpu(es->s_want_extra_isize);
3517 if (sbi->s_want_extra_isize <
3518 le16_to_cpu(es->s_min_extra_isize))
3519 sbi->s_want_extra_isize =
3520 le16_to_cpu(es->s_min_extra_isize);
3521 }
3522 }
3523 /* Check if enough inode space is available */
3524 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3525 sbi->s_inode_size) {
3526 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3527 EXT4_GOOD_OLD_INODE_SIZE;
b31e1552
ES		 3528 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3529 "available");
ef7f3835
KS		 3530 }
3531
90576c0b
TT		 3532 if (test_opt(sb, DELALLOC) &&
3533 (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)) {
b31e1552
ES		 3534 ext4_msg(sb, KERN_WARNING, "Ignoring delalloc option - "
3535 "requested data journaling mode");
c2774d84 3536 clear_opt(sbi->s_mount_opt, DELALLOC);
90576c0b 3537 }
744692dc
JZ		 3538 if (test_opt(sb, DIOREAD_NOLOCK)) {
3539 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3540 ext4_msg(sb, KERN_WARNING, "Ignoring dioread_nolock "
3541 "option - requested data journaling mode");
3542 clear_opt(sbi->s_mount_opt, DIOREAD_NOLOCK);
3543 }
3544 if (sb->s_blocksize < PAGE_SIZE) {
3545 ext4_msg(sb, KERN_WARNING, "Ignoring dioread_nolock "
3546 "option - block size is too small");
3547 clear_opt(sbi->s_mount_opt, DIOREAD_NOLOCK);
3548 }
3549 }
c2774d84 3550
6fd058f7
TT		 3551 err = ext4_setup_system_zone(sb);
3552 if (err) {
b31e1552 3553 ext4_msg(sb, KERN_ERR, "failed to initialize system "
fbe845dd 3554 "zone (%d)", err);
6fd058f7
TT		 3555 goto failed_mount4;
3556 }
3557
c2774d84
AK		 3558 ext4_ext_init(sb);
3559 err = ext4_mb_init(sb, needs_recovery);
3560 if (err) {
421f91d2 3561 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
b31e1552 3562 err);
c2774d84
AK		 3563 goto failed_mount4;
3564 }
3565
bfff6873
LC		 3566 err = ext4_register_li_request(sb, first_not_zeroed);
3567 if (err)
3568 goto failed_mount4;
3569
3197ebdb
TT		 3570 sbi->s_kobj.kset = ext4_kset;
3571 init_completion(&sbi->s_kobj_unregister);
3572 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
3573 "%s", sb->s_id);
3574 if (err) {
3575 ext4_mb_release(sb);
3576 ext4_ext_release(sb);
3577 goto failed_mount4;
3578 };
3579
617ba13b
MC		 3580 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3581 ext4_orphan_cleanup(sb, es);
3582 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
0390131b 3583 if (needs_recovery) {
b31e1552 3584 ext4_msg(sb, KERN_INFO, "recovery complete");
0390131b
FM		 3585 ext4_mark_recovery_complete(sb, es);
3586 }
3587 if (EXT4_SB(sb)->s_journal) {
3588 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3589 descr = " journalled data mode";
3590 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3591 descr = " ordered data mode";
3592 else
3593 descr = " writeback data mode";
3594 } else
3595 descr = "out journal";
3596
d4c402d9 3597 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
8b67f04a
TT		 3598 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
3599 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
ac27a0ec 3600
66e61a9e
TT		 3601 init_timer(&sbi->s_err_report);
3602 sbi->s_err_report.function = print_daily_error_info;
3603 sbi->s_err_report.data = (unsigned long) sb;
3604 if (es->s_error_count)
3605 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
ac27a0ec
DK		 3606
3607 lock_kernel();
d4c402d9 3608 kfree(orig_data);
ac27a0ec
DK		 3609 return 0;
3610
617ba13b 3611cantfind_ext4:
ac27a0ec 3612 if (!silent)
b31e1552 3613 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
ac27a0ec
DK		 3614 goto failed_mount;
3615
3616failed_mount4:
b31e1552 3617 ext4_msg(sb, KERN_ERR, "mount failed");
4c0425ff
MC		 3618 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
3619failed_mount_wq:
6fd058f7 3620 ext4_release_system_zone(sb);
0390131b
FM		 3621 if (sbi->s_journal) {
3622 jbd2_journal_destroy(sbi->s_journal);
3623 sbi->s_journal = NULL;
3624 }
84061e07
DM		 3625 percpu_counter_destroy(&sbi->s_freeblocks_counter);
3626 percpu_counter_destroy(&sbi->s_freeinodes_counter);
3627 percpu_counter_destroy(&sbi->s_dirs_counter);
3628 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
ac27a0ec 3629failed_mount3:
c5ca7c76
TT		 3630 if (sbi->s_flex_groups) {
3631 if (is_vmalloc_addr(sbi->s_flex_groups))
3632 vfree(sbi->s_flex_groups);
3633 else
3634 kfree(sbi->s_flex_groups);
3635 }
ac27a0ec
DK		 3636failed_mount2:
3637 for (i = 0; i < db_count; i++)
3638 brelse(sbi->s_group_desc[i]);
3639 kfree(sbi->s_group_desc);
3640failed_mount:
240799cd 3641 if (sbi->s_proc) {
9f6200bb 3642 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 3643 }
ac27a0ec
DK		 3644#ifdef CONFIG_QUOTA
3645 for (i = 0; i < MAXQUOTAS; i++)
3646 kfree(sbi->s_qf_names[i]);
3647#endif
617ba13b 3648 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 3649 brelse(bh);
3650out_fail:
3651 sb->s_fs_info = NULL;
f6830165 3652 kfree(sbi->s_blockgroup_lock);
ac27a0ec
DK		 3653 kfree(sbi);
3654 lock_kernel();
dcc7dae3 3655out_free_orig:
d4c402d9 3656 kfree(orig_data);
1d1fe1ee 3657 return ret;
ac27a0ec
DK		 3658}
3659
3660/*
3661 * Setup any per-fs journal parameters now. We'll do this both on
3662 * initial mount, once the journal has been initialised but before we've
3663 * done any recovery; and again on any subsequent remount.
3664 */
617ba13b 3665static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 3666{
617ba13b 3667 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 3668
30773840
TT		 3669 journal->j_commit_interval = sbi->s_commit_interval;
3670 journal->j_min_batch_time = sbi->s_min_batch_time;
3671 journal->j_max_batch_time = sbi->s_max_batch_time;
ac27a0ec 3672
a931da6a 3673 write_lock(&journal->j_state_lock);
ac27a0ec 3674 if (test_opt(sb, BARRIER))
dab291af 3675 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 3676 else
dab291af 3677 journal->j_flags &= ~JBD2_BARRIER;
5bf5683a
HK		 3678 if (test_opt(sb, DATA_ERR_ABORT))
3679 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
3680 else
3681 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
a931da6a 3682 write_unlock(&journal->j_state_lock);
ac27a0ec
DK		 3683}
3684
617ba13b 3685static journal_t *ext4_get_journal(struct super_block *sb,
ac27a0ec
DK		 3686 unsigned int journal_inum)
3687{
3688 struct inode *journal_inode;
3689 journal_t *journal;
3690
0390131b
FM		 3691 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3692
ac27a0ec
DK		 3693 /* First, test for the existence of a valid inode on disk. Bad
3694 * things happen if we iget() an unused inode, as the subsequent
3695 * iput() will try to delete it. */
3696
1d1fe1ee
DH		 3697 journal_inode = ext4_iget(sb, journal_inum);
3698 if (IS_ERR(journal_inode)) {
b31e1552 3699 ext4_msg(sb, KERN_ERR, "no journal found");
ac27a0ec
DK		 3700 return NULL;
3701 }
3702 if (!journal_inode->i_nlink) {
3703 make_bad_inode(journal_inode);
3704 iput(journal_inode);
b31e1552 3705 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
ac27a0ec
DK		 3706 return NULL;
3707 }
3708
e5f8eab8 3709 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
ac27a0ec 3710 journal_inode, journal_inode->i_size);
1d1fe1ee 3711 if (!S_ISREG(journal_inode->i_mode)) {
b31e1552 3712 ext4_msg(sb, KERN_ERR, "invalid journal inode");
ac27a0ec
DK		 3713 iput(journal_inode);
3714 return NULL;
3715 }
3716
dab291af 3717 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 3718 if (!journal) {
b31e1552 3719 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
ac27a0ec
DK		 3720 iput(journal_inode);
3721 return NULL;
3722 }
3723 journal->j_private = sb;
617ba13b 3724 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 3725 return journal;
3726}
3727
617ba13b 3728static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 3729 dev_t j_dev)
3730{
2b2d6d01 3731 struct buffer_head *bh;
ac27a0ec 3732 journal_t *journal;
617ba13b
MC		 3733 ext4_fsblk_t start;
3734 ext4_fsblk_t len;
ac27a0ec 3735 int hblock, blocksize;
617ba13b 3736 ext4_fsblk_t sb_block;
ac27a0ec 3737 unsigned long offset;
2b2d6d01 3738 struct ext4_super_block *es;
ac27a0ec
DK		 3739 struct block_device *bdev;
3740
0390131b
FM		 3741 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3742
b31e1552 3743 bdev = ext4_blkdev_get(j_dev, sb);
ac27a0ec
DK		 3744 if (bdev == NULL)
3745 return NULL;
3746
3747 if (bd_claim(bdev, sb)) {
b31e1552
ES		 3748 ext4_msg(sb, KERN_ERR,
3749 "failed to claim external journal device");
9a1c3542 3750 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
ac27a0ec
DK		 3751 return NULL;
3752 }
3753
3754 blocksize = sb->s_blocksize;
e1defc4f 3755 hblock = bdev_logical_block_size(bdev);
ac27a0ec 3756 if (blocksize < hblock) {
b31e1552
ES		 3757 ext4_msg(sb, KERN_ERR,
3758 "blocksize too small for journal device");
ac27a0ec
DK		 3759 goto out_bdev;
3760 }
3761
617ba13b
MC		 3762 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3763 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 3764 set_blocksize(bdev, blocksize);
3765 if (!(bh = __bread(bdev, sb_block, blocksize))) {
b31e1552
ES		 3766 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
3767 "external journal");
ac27a0ec
DK		 3768 goto out_bdev;
3769 }
3770
617ba13b
MC		 3771 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3772 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 3773 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b 3774 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
b31e1552
ES		 3775 ext4_msg(sb, KERN_ERR, "external journal has "
3776 "bad superblock");
ac27a0ec
DK		 3777 brelse(bh);
3778 goto out_bdev;
3779 }
3780
617ba13b 3781 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
b31e1552 3782 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
ac27a0ec
DK		 3783 brelse(bh);
3784 goto out_bdev;
3785 }
3786
bd81d8ee 3787 len = ext4_blocks_count(es);
ac27a0ec
DK		 3788 start = sb_block + 1;
3789 brelse(bh); /* we're done with the superblock */
3790
dab291af 3791 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 3792 start, len, blocksize);
3793 if (!journal) {
b31e1552 3794 ext4_msg(sb, KERN_ERR, "failed to create device journal");
ac27a0ec
DK		 3795 goto out_bdev;
3796 }
3797 journal->j_private = sb;
3798 ll_rw_block(READ, 1, &journal->j_sb_buffer);
3799 wait_on_buffer(journal->j_sb_buffer);
3800 if (!buffer_uptodate(journal->j_sb_buffer)) {
b31e1552 3801 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
ac27a0ec
DK		 3802 goto out_journal;
3803 }
3804 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
b31e1552
ES		 3805 ext4_msg(sb, KERN_ERR, "External journal has more than one "
3806 "user (unsupported) - %d",
ac27a0ec
DK		 3807 be32_to_cpu(journal->j_superblock->s_nr_users));
3808 goto out_journal;
3809 }
617ba13b
MC		 3810 EXT4_SB(sb)->journal_bdev = bdev;
3811 ext4_init_journal_params(sb, journal);
ac27a0ec 3812 return journal;
0b8e58a1 3813
ac27a0ec 3814out_journal:
dab291af 3815 jbd2_journal_destroy(journal);
ac27a0ec 3816out_bdev:
617ba13b 3817 ext4_blkdev_put(bdev);
ac27a0ec
DK		 3818 return NULL;
3819}
3820
617ba13b
MC		 3821static int ext4_load_journal(struct super_block *sb,
3822 struct ext4_super_block *es,
ac27a0ec
DK		 3823 unsigned long journal_devnum)
3824{
3825 journal_t *journal;
3826 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
3827 dev_t journal_dev;
3828 int err = 0;
3829 int really_read_only;
3830
0390131b
FM		 3831 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3832
ac27a0ec
DK		 3833 if (journal_devnum &&
3834 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
b31e1552
ES		 3835 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
3836 "numbers have changed");
ac27a0ec
DK		 3837 journal_dev = new_decode_dev(journal_devnum);
3838 } else
3839 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
3840
3841 really_read_only = bdev_read_only(sb->s_bdev);
3842
3843 /*
3844 * Are we loading a blank journal or performing recovery after a
3845 * crash? For recovery, we need to check in advance whether we
3846 * can get read-write access to the device.
3847 */
617ba13b 3848 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
ac27a0ec 3849 if (sb->s_flags & MS_RDONLY) {
b31e1552
ES		 3850 ext4_msg(sb, KERN_INFO, "INFO: recovery "
3851 "required on readonly filesystem");
ac27a0ec 3852 if (really_read_only) {
b31e1552
ES		 3853 ext4_msg(sb, KERN_ERR, "write access "
3854 "unavailable, cannot proceed");
ac27a0ec
DK		 3855 return -EROFS;
3856 }
b31e1552
ES		 3857 ext4_msg(sb, KERN_INFO, "write access will "
3858 "be enabled during recovery");
ac27a0ec
DK		 3859 }
3860 }
3861
3862 if (journal_inum && journal_dev) {
b31e1552
ES		 3863 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
3864 "and inode journals!");
ac27a0ec
DK		 3865 return -EINVAL;
3866 }
3867
3868 if (journal_inum) {
617ba13b 3869 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 3870 return -EINVAL;
3871 } else {
617ba13b 3872 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 3873 return -EINVAL;
3874 }
3875
90576c0b 3876 if (!(journal->j_flags & JBD2_BARRIER))
b31e1552 3877 ext4_msg(sb, KERN_INFO, "barriers disabled");
4776004f 3878
ac27a0ec 3879 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
dab291af 3880 err = jbd2_journal_update_format(journal);
ac27a0ec 3881 if (err) {
b31e1552 3882 ext4_msg(sb, KERN_ERR, "error updating journal");
dab291af 3883 jbd2_journal_destroy(journal);
ac27a0ec
DK		 3884 return err;
3885 }
3886 }
3887
617ba13b 3888 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
dab291af 3889 err = jbd2_journal_wipe(journal, !really_read_only);
1c13d5c0
TT		 3890 if (!err) {
3891 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
3892 if (save)
3893 memcpy(save, ((char *) es) +
3894 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
dab291af 3895 err = jbd2_journal_load(journal);
1c13d5c0
TT		 3896 if (save)
3897 memcpy(((char *) es) + EXT4_S_ERR_START,
3898 save, EXT4_S_ERR_LEN);
3899 kfree(save);
3900 }
ac27a0ec
DK		 3901
3902 if (err) {
b31e1552 3903 ext4_msg(sb, KERN_ERR, "error loading journal");
dab291af 3904 jbd2_journal_destroy(journal);
ac27a0ec
DK		 3905 return err;
3906 }
3907
617ba13b
MC		 3908 EXT4_SB(sb)->s_journal = journal;
3909 ext4_clear_journal_err(sb, es);
ac27a0ec 3910
c41303ce 3911 if (!really_read_only && journal_devnum &&
ac27a0ec
DK		 3912 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3913 es->s_journal_dev = cpu_to_le32(journal_devnum);
ac27a0ec
DK		 3914
3915 /* Make sure we flush the recovery flag to disk. */
e2d67052 3916 ext4_commit_super(sb, 1);
ac27a0ec
DK		 3917 }
3918
3919 return 0;
3920}
3921
e2d67052 3922static int ext4_commit_super(struct super_block *sb, int sync)
ac27a0ec 3923{
e2d67052 3924 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
617ba13b 3925 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
c4be0c1d 3926 int error = 0;
ac27a0ec
DK		 3927
3928 if (!sbh)
c4be0c1d 3929 return error;
914258bf
TT		 3930 if (buffer_write_io_error(sbh)) {
3931 /*
3932 * Oh, dear. A previous attempt to write the
3933 * superblock failed. This could happen because the
3934 * USB device was yanked out. Or it could happen to
3935 * be a transient write error and maybe the block will
3936 * be remapped. Nothing we can do but to retry the
3937 * write and hope for the best.
3938 */
b31e1552
ES		 3939 ext4_msg(sb, KERN_ERR, "previous I/O error to "
3940 "superblock detected");
914258bf
TT		 3941 clear_buffer_write_io_error(sbh);
3942 set_buffer_uptodate(sbh);
3943 }
71290b36
TT		 3944 /*
3945 * If the file system is mounted read-only, don't update the
3946 * superblock write time. This avoids updating the superblock
3947 * write time when we are mounting the root file system
3948 * read/only but we need to replay the journal; at that point,
3949 * for people who are east of GMT and who make their clock
3950 * tick in localtime for Windows bug-for-bug compatibility,
3951 * the clock is set in the future, and this will cause e2fsck
3952 * to complain and force a full file system check.
3953 */
3954 if (!(sb->s_flags & MS_RDONLY))
3955 es->s_wtime = cpu_to_le32(get_seconds());
f613dfcb
TT		 3956 if (sb->s_bdev->bd_part)
3957 es->s_kbytes_written =
3958 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
afc32f7e
TT		 3959 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
3960 EXT4_SB(sb)->s_sectors_written_start) >> 1));
f613dfcb
TT		 3961 else
3962 es->s_kbytes_written =
3963 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
5d1b1b3f
AK		 3964 ext4_free_blocks_count_set(es, percpu_counter_sum_positive(
3965 &EXT4_SB(sb)->s_freeblocks_counter));
3966 es->s_free_inodes_count = cpu_to_le32(percpu_counter_sum_positive(
3967 &EXT4_SB(sb)->s_freeinodes_counter));
7234ab2a 3968 sb->s_dirt = 0;
ac27a0ec
DK		 3969 BUFFER_TRACE(sbh, "marking dirty");
3970 mark_buffer_dirty(sbh);
914258bf 3971 if (sync) {
c4be0c1d
TS		 3972 error = sync_dirty_buffer(sbh);
3973 if (error)
3974 return error;
3975
3976 error = buffer_write_io_error(sbh);
3977 if (error) {
b31e1552
ES		 3978 ext4_msg(sb, KERN_ERR, "I/O error while writing "
3979 "superblock");
914258bf
TT		 3980 clear_buffer_write_io_error(sbh);
3981 set_buffer_uptodate(sbh);
3982 }
3983 }
c4be0c1d 3984 return error;
ac27a0ec
DK		 3985}
3986
ac27a0ec
DK		 3987/*
3988 * Have we just finished recovery? If so, and if we are mounting (or
3989 * remounting) the filesystem readonly, then we will end up with a
3990 * consistent fs on disk. Record that fact.
3991 */
2b2d6d01
TT		 3992static void ext4_mark_recovery_complete(struct super_block *sb,
3993 struct ext4_super_block *es)
ac27a0ec 3994{
617ba13b 3995 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 3996
0390131b
FM		 3997 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3998 BUG_ON(journal != NULL);
3999 return;
4000 }
dab291af 4001 jbd2_journal_lock_updates(journal);
7ffe1ea8
HK		 4002 if (jbd2_journal_flush(journal) < 0)
4003 goto out;
4004
617ba13b 4005 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
ac27a0ec 4006 sb->s_flags & MS_RDONLY) {
617ba13b 4007 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
e2d67052 4008 ext4_commit_super(sb, 1);
ac27a0ec 4009 }
7ffe1ea8
HK		 4010
4011out:
dab291af 4012 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 4013}
4014
4015/*
4016 * If we are mounting (or read-write remounting) a filesystem whose journal
4017 * has recorded an error from a previous lifetime, move that error to the
4018 * main filesystem now.
4019 */
2b2d6d01
TT		 4020static void ext4_clear_journal_err(struct super_block *sb,
4021 struct ext4_super_block *es)
ac27a0ec
DK		 4022{
4023 journal_t *journal;
4024 int j_errno;
4025 const char *errstr;
4026
0390131b
FM		 4027 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4028
617ba13b 4029 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 4030
4031 /*
4032 * Now check for any error status which may have been recorded in the
617ba13b 4033 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 4034 */
4035
dab291af 4036 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 4037 if (j_errno) {
4038 char nbuf[16];
4039
617ba13b 4040 errstr = ext4_decode_error(sb, j_errno, nbuf);
12062ddd 4041 ext4_warning(sb, "Filesystem error recorded "
ac27a0ec 4042 "from previous mount: %s", errstr);
12062ddd 4043 ext4_warning(sb, "Marking fs in need of filesystem check.");
ac27a0ec 4044
617ba13b
MC		 4045 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4046 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
e2d67052 4047 ext4_commit_super(sb, 1);
ac27a0ec 4048
dab291af 4049 jbd2_journal_clear_err(journal);
ac27a0ec
DK		 4050 }
4051}
4052
4053/*
4054 * Force the running and committing transactions to commit,
4055 * and wait on the commit.
4056 */
617ba13b 4057int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 4058{
4059 journal_t *journal;
0390131b 4060 int ret = 0;
ac27a0ec
DK		 4061
4062 if (sb->s_flags & MS_RDONLY)
4063 return 0;
4064
617ba13b 4065 journal = EXT4_SB(sb)->s_journal;
6b0310fb 4066 if (journal) {
437f88cc 4067 vfs_check_frozen(sb, SB_FREEZE_TRANS);
0390131b 4068 ret = ext4_journal_force_commit(journal);
6b0310fb 4069 }
0390131b 4070
ac27a0ec
DK		 4071 return ret;
4072}
4073
2b2d6d01 4074static void ext4_write_super(struct super_block *sb)
ac27a0ec 4075{
ebc1ac16 4076 lock_super(sb);
9ca92389 4077 ext4_commit_super(sb, 1);
ebc1ac16 4078 unlock_super(sb);
ac27a0ec
DK		 4079}
4080
617ba13b 4081static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec 4082{
14ce0cb4 4083 int ret = 0;
9eddacf9 4084 tid_t target;
8d5d02e6 4085 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4086
9bffad1e 4087 trace_ext4_sync_fs(sb, wait);
8d5d02e6
MC		 4088 flush_workqueue(sbi->dio_unwritten_wq);
4089 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
9ca92389 4090 if (wait)
8d5d02e6 4091 jbd2_log_wait_commit(sbi->s_journal, target);
0390131b 4092 }
14ce0cb4 4093 return ret;
ac27a0ec
DK		 4094}
4095
4096/*
4097 * LVM calls this function before a (read-only) snapshot is created. This
4098 * gives us a chance to flush the journal completely and mark the fs clean.
4099 */
c4be0c1d 4100static int ext4_freeze(struct super_block *sb)
ac27a0ec 4101{
c4be0c1d
TS		 4102 int error = 0;
4103 journal_t *journal;
ac27a0ec 4104
9ca92389
TT		 4105 if (sb->s_flags & MS_RDONLY)
4106 return 0;
ac27a0ec 4107
9ca92389 4108 journal = EXT4_SB(sb)->s_journal;
7ffe1ea8 4109
9ca92389
TT		 4110 /* Now we set up the journal barrier. */
4111 jbd2_journal_lock_updates(journal);
ac27a0ec 4112
9ca92389
TT		 4113 /*
4114 * Don't clear the needs_recovery flag if we failed to flush
4115 * the journal.
4116 */
4117 error = jbd2_journal_flush(journal);
6b0310fb
ES		 4118 if (error < 0)
4119 goto out;
9ca92389
TT		 4120
4121 /* Journal blocked and flushed, clear needs_recovery flag. */
4122 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4123 error = ext4_commit_super(sb, 1);
6b0310fb
ES		 4124out:
4125 /* we rely on s_frozen to stop further updates */
4126 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4127 return error;
ac27a0ec
DK		 4128}
4129
4130/*
4131 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4132 * flag here, even though the filesystem is not technically dirty yet.
4133 */
c4be0c1d 4134static int ext4_unfreeze(struct super_block *sb)
ac27a0ec 4135{
9ca92389
TT		 4136 if (sb->s_flags & MS_RDONLY)
4137 return 0;
4138
4139 lock_super(sb);
4140 /* Reset the needs_recovery flag before the fs is unlocked. */
4141 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4142 ext4_commit_super(sb, 1);
4143 unlock_super(sb);
c4be0c1d 4144 return 0;
ac27a0ec
DK		 4145}
4146
2b2d6d01 4147static int ext4_remount(struct super_block *sb, int *flags, char *data)
ac27a0ec 4148{
2b2d6d01 4149 struct ext4_super_block *es;
617ba13b
MC		 4150 struct ext4_sb_info *sbi = EXT4_SB(sb);
4151 ext4_fsblk_t n_blocks_count = 0;
ac27a0ec 4152 unsigned long old_sb_flags;
617ba13b 4153 struct ext4_mount_options old_opts;
c79d967d 4154 int enable_quota = 0;
8a266467 4155 ext4_group_t g;
b3881f74 4156 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
ac27a0ec
DK		 4157 int err;
4158#ifdef CONFIG_QUOTA
4159 int i;
4160#endif
d4c402d9 4161 char *orig_data = kstrdup(data, GFP_KERNEL);
ac27a0ec 4162
337eb00a
AIB		 4163 lock_kernel();
4164
ac27a0ec 4165 /* Store the original options */
bbd6851a 4166 lock_super(sb);
ac27a0ec
DK		 4167 old_sb_flags = sb->s_flags;
4168 old_opts.s_mount_opt = sbi->s_mount_opt;
4169 old_opts.s_resuid = sbi->s_resuid;
4170 old_opts.s_resgid = sbi->s_resgid;
4171 old_opts.s_commit_interval = sbi->s_commit_interval;
30773840
TT		 4172 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4173 old_opts.s_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 4174#ifdef CONFIG_QUOTA
4175 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4176 for (i = 0; i < MAXQUOTAS; i++)
4177 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
4178#endif
b3881f74
TT		 4179 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4180 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
ac27a0ec
DK		 4181
4182 /*
4183 * Allow the "check" option to be passed as a remount option.
4184 */
b3881f74
TT		 4185 if (!parse_options(data, sb, NULL, &journal_ioprio,
4186 &n_blocks_count, 1)) {
ac27a0ec
DK		 4187 err = -EINVAL;
4188 goto restore_opts;
4189 }
4190
4ab2f15b 4191 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
c67d859e 4192 ext4_abort(sb, "Abort forced by user");
ac27a0ec
DK		 4193
4194 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 4195 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec
DK		 4196
4197 es = sbi->s_es;
4198
b3881f74 4199 if (sbi->s_journal) {
0390131b 4200 ext4_init_journal_params(sb, sbi->s_journal);
b3881f74
TT		 4201 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4202 }
ac27a0ec
DK		 4203
4204 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
bd81d8ee 4205 n_blocks_count > ext4_blocks_count(es)) {
4ab2f15b 4206 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
ac27a0ec
DK		 4207 err = -EROFS;
4208 goto restore_opts;
4209 }
4210
4211 if (*flags & MS_RDONLY) {
0f0dd62f
CH		 4212 err = dquot_suspend(sb, -1);
4213 if (err < 0)
c79d967d 4214 goto restore_opts;
c79d967d 4215
ac27a0ec
DK		 4216 /*
4217 * First of all, the unconditional stuff we have to do
4218 * to disable replay of the journal when we next remount
4219 */
4220 sb->s_flags |= MS_RDONLY;
4221
4222 /*
4223 * OK, test if we are remounting a valid rw partition
4224 * readonly, and if so set the rdonly flag and then
4225 * mark the partition as valid again.
4226 */
617ba13b
MC		 4227 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4228 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 4229 es->s_state = cpu_to_le16(sbi->s_mount_state);
4230
a63c9eb2 4231 if (sbi->s_journal)
0390131b 4232 ext4_mark_recovery_complete(sb, es);
ac27a0ec 4233 } else {
a13fb1a4
ES		 4234 /* Make sure we can mount this feature set readwrite */
4235 if (!ext4_feature_set_ok(sb, 0)) {
ac27a0ec
DK		 4236 err = -EROFS;
4237 goto restore_opts;
4238 }
8a266467
TT		 4239 /*
4240 * Make sure the group descriptor checksums
0b8e58a1 4241 * are sane. If they aren't, refuse to remount r/w.
8a266467
TT		 4242 */
4243 for (g = 0; g < sbi->s_groups_count; g++) {
4244 struct ext4_group_desc *gdp =
4245 ext4_get_group_desc(sb, g, NULL);
4246
4247 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
b31e1552
ES		 4248 ext4_msg(sb, KERN_ERR,
4249 "ext4_remount: Checksum for group %u failed (%u!=%u)",
8a266467
TT		 4250 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4251 le16_to_cpu(gdp->bg_checksum));
4252 err = -EINVAL;
4253 goto restore_opts;
4254 }
4255 }
4256
ead6596b
ES		 4257 /*
4258 * If we have an unprocessed orphan list hanging
4259 * around from a previously readonly bdev mount,
4260 * require a full umount/remount for now.
4261 */
4262 if (es->s_last_orphan) {
b31e1552 4263 ext4_msg(sb, KERN_WARNING, "Couldn't "
ead6596b
ES		 4264 "remount RDWR because of unprocessed "
4265 "orphan inode list. Please "
b31e1552 4266 "umount/remount instead");
ead6596b
ES		 4267 err = -EINVAL;
4268 goto restore_opts;
4269 }
4270
ac27a0ec
DK		 4271 /*
4272 * Mounting a RDONLY partition read-write, so reread
4273 * and store the current valid flag. (It may have
4274 * been changed by e2fsck since we originally mounted
4275 * the partition.)
4276 */
0390131b
FM		 4277 if (sbi->s_journal)
4278 ext4_clear_journal_err(sb, es);
ac27a0ec 4279 sbi->s_mount_state = le16_to_cpu(es->s_state);
617ba13b 4280 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
ac27a0ec 4281 goto restore_opts;
2b2d6d01 4282 if (!ext4_setup_super(sb, es, 0))
ac27a0ec 4283 sb->s_flags &= ~MS_RDONLY;
c79d967d 4284 enable_quota = 1;
ac27a0ec
DK		 4285 }
4286 }
bfff6873
LC		 4287
4288 /*
4289 * Reinitialize lazy itable initialization thread based on
4290 * current settings
4291 */
4292 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4293 ext4_unregister_li_request(sb);
4294 else {
4295 ext4_group_t first_not_zeroed;
4296 first_not_zeroed = ext4_has_uninit_itable(sb);
4297 ext4_register_li_request(sb, first_not_zeroed);
4298 }
4299
6fd058f7 4300 ext4_setup_system_zone(sb);
0390131b 4301 if (sbi->s_journal == NULL)
e2d67052 4302 ext4_commit_super(sb, 1);
0390131b 4303
ac27a0ec
DK		 4304#ifdef CONFIG_QUOTA
4305 /* Release old quota file names */
4306 for (i = 0; i < MAXQUOTAS; i++)
4307 if (old_opts.s_qf_names[i] &&
4308 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4309 kfree(old_opts.s_qf_names[i]);
4310#endif
bbd6851a 4311 unlock_super(sb);
337eb00a 4312 unlock_kernel();
c79d967d 4313 if (enable_quota)
0f0dd62f 4314 dquot_resume(sb, -1);
d4c402d9
CW		 4315
4316 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4317 kfree(orig_data);
ac27a0ec 4318 return 0;
0b8e58a1 4319
ac27a0ec
DK		 4320restore_opts:
4321 sb->s_flags = old_sb_flags;
4322 sbi->s_mount_opt = old_opts.s_mount_opt;
4323 sbi->s_resuid = old_opts.s_resuid;
4324 sbi->s_resgid = old_opts.s_resgid;
4325 sbi->s_commit_interval = old_opts.s_commit_interval;
30773840
TT		 4326 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4327 sbi->s_max_batch_time = old_opts.s_max_batch_time;
ac27a0ec
DK		 4328#ifdef CONFIG_QUOTA
4329 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4330 for (i = 0; i < MAXQUOTAS; i++) {
4331 if (sbi->s_qf_names[i] &&
4332 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4333 kfree(sbi->s_qf_names[i]);
4334 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4335 }
4336#endif
bbd6851a 4337 unlock_super(sb);
337eb00a 4338 unlock_kernel();
d4c402d9 4339 kfree(orig_data);
ac27a0ec
DK		 4340 return err;
4341}
4342
2b2d6d01 4343static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
ac27a0ec
DK		 4344{
4345 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 4346 struct ext4_sb_info *sbi = EXT4_SB(sb);
4347 struct ext4_super_block *es = sbi->s_es;
960cc398 4348 u64 fsid;
ac27a0ec 4349
5e70030d
BP		 4350 if (test_opt(sb, MINIX_DF)) {
4351 sbi->s_overhead_last = 0;
6bc9feff 4352 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
8df9675f 4353 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
5e70030d 4354 ext4_fsblk_t overhead = 0;
ac27a0ec
DK		 4355
4356 /*
5e70030d
BP		 4357 * Compute the overhead (FS structures). This is constant
4358 * for a given filesystem unless the number of block groups
4359 * changes so we cache the previous value until it does.
ac27a0ec
DK		 4360 */
4361
4362 /*
4363 * All of the blocks before first_data_block are
4364 * overhead
4365 */
4366 overhead = le32_to_cpu(es->s_first_data_block);
4367
4368 /*
4369 * Add the overhead attributed to the superblock and
4370 * block group descriptors. If the sparse superblocks
4371 * feature is turned on, then not all groups have this.
4372 */
4373 for (i = 0; i < ngroups; i++) {
617ba13b
MC		 4374 overhead += ext4_bg_has_super(sb, i) +
4375 ext4_bg_num_gdb(sb, i);
ac27a0ec
DK		 4376 cond_resched();
4377 }
4378
4379 /*
4380 * Every block group has an inode bitmap, a block
4381 * bitmap, and an inode table.
4382 */
5e70030d
BP		 4383 overhead += ngroups * (2 + sbi->s_itb_per_group);
4384 sbi->s_overhead_last = overhead;
4385 smp_wmb();
6bc9feff 4386 sbi->s_blocks_last = ext4_blocks_count(es);
ac27a0ec
DK		 4387 }
4388
617ba13b 4389 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 4390 buf->f_bsize = sb->s_blocksize;
5e70030d 4391 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
6bc6e63f
AK		 4392 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
4393 percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
bd81d8ee
LV		 4394 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
4395 if (buf->f_bfree < ext4_r_blocks_count(es))
ac27a0ec
DK		 4396 buf->f_bavail = 0;
4397 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 4398 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
617ba13b 4399 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 4400 fsid = le64_to_cpup((void *)es->s_uuid) ^
4401 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4402 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4403 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
0b8e58a1 4404
ac27a0ec
DK		 4405 return 0;
4406}
4407
0b8e58a1
AD		 4408/* Helper function for writing quotas on sync - we need to start transaction
4409 * before quota file is locked for write. Otherwise the are possible deadlocks:
ac27a0ec 4410 * Process 1 Process 2
617ba13b 4411 * ext4_create() quota_sync()
a269eb18 4412 * jbd2_journal_start() write_dquot()
871a2931 4413 * dquot_initialize() down(dqio_mutex)
dab291af 4414 * down(dqio_mutex) jbd2_journal_start()
ac27a0ec
DK		 4415 *
4416 */
4417
4418#ifdef CONFIG_QUOTA
4419
4420static inline struct inode *dquot_to_inode(struct dquot *dquot)
4421{
4422 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
4423}
4424
617ba13b 4425static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 4426{
4427 int ret, err;
4428 handle_t *handle;
4429 struct inode *inode;
4430
4431 inode = dquot_to_inode(dquot);
617ba13b 4432 handle = ext4_journal_start(inode,
0b8e58a1 4433 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 4434 if (IS_ERR(handle))
4435 return PTR_ERR(handle);
4436 ret = dquot_commit(dquot);
617ba13b 4437 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4438 if (!ret)
4439 ret = err;
4440 return ret;
4441}
4442
617ba13b 4443static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 4444{
4445 int ret, err;
4446 handle_t *handle;
4447
617ba13b 4448 handle = ext4_journal_start(dquot_to_inode(dquot),
0b8e58a1 4449 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 4450 if (IS_ERR(handle))
4451 return PTR_ERR(handle);
4452 ret = dquot_acquire(dquot);
617ba13b 4453 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4454 if (!ret)
4455 ret = err;
4456 return ret;
4457}
4458
617ba13b 4459static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 4460{
4461 int ret, err;
4462 handle_t *handle;
4463
617ba13b 4464 handle = ext4_journal_start(dquot_to_inode(dquot),
0b8e58a1 4465 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 4466 if (IS_ERR(handle)) {
4467 /* Release dquot anyway to avoid endless cycle in dqput() */
4468 dquot_release(dquot);
ac27a0ec 4469 return PTR_ERR(handle);
9c3013e9 4470 }
ac27a0ec 4471 ret = dquot_release(dquot);
617ba13b 4472 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4473 if (!ret)
4474 ret = err;
4475 return ret;
4476}
4477
617ba13b 4478static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec 4479{
2c8be6b2 4480 /* Are we journaling quotas? */
617ba13b
MC		 4481 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
4482 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
ac27a0ec 4483 dquot_mark_dquot_dirty(dquot);
617ba13b 4484 return ext4_write_dquot(dquot);
ac27a0ec
DK		 4485 } else {
4486 return dquot_mark_dquot_dirty(dquot);
4487 }
4488}
4489
617ba13b 4490static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 4491{
4492 int ret, err;
4493 handle_t *handle;
4494
4495 /* Data block + inode block */
617ba13b 4496 handle = ext4_journal_start(sb->s_root->d_inode, 2);
ac27a0ec
DK		 4497 if (IS_ERR(handle))
4498 return PTR_ERR(handle);
4499 ret = dquot_commit_info(sb, type);
617ba13b 4500 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4501 if (!ret)
4502 ret = err;
4503 return ret;
4504}
4505
4506/*
4507 * Turn on quotas during mount time - we need to find
4508 * the quota file and such...
4509 */
617ba13b 4510static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 4511{
287a8095
CH		 4512 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4513 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 4514}
4515
4516/*
4517 * Standard function to be called on quota_on
4518 */
617ba13b 4519static int ext4_quota_on(struct super_block *sb, int type, int format_id,
307ae18a 4520 char *name)
ac27a0ec
DK		 4521{
4522 int err;
8264613d 4523 struct path path;
ac27a0ec
DK		 4524
4525 if (!test_opt(sb, QUOTA))
4526 return -EINVAL;
0623543b 4527
8264613d 4528 err = kern_path(name, LOOKUP_FOLLOW, &path);
ac27a0ec
DK		 4529 if (err)
4530 return err;
0623543b 4531
ac27a0ec 4532 /* Quotafile not on the same filesystem? */
8264613d
AV		 4533 if (path.mnt->mnt_sb != sb) {
4534 path_put(&path);
ac27a0ec
DK		 4535 return -EXDEV;
4536 }
0623543b
JK		 4537 /* Journaling quota? */
4538 if (EXT4_SB(sb)->s_qf_names[type]) {
2b2d6d01 4539 /* Quotafile not in fs root? */
8264613d 4540 if (path.dentry->d_parent != sb->s_root)
b31e1552
ES		 4541 ext4_msg(sb, KERN_WARNING,
4542 "Quota file not on filesystem root. "
4543 "Journaled quota will not work");
2b2d6d01 4544 }
0623543b
JK		 4545
4546 /*
4547 * When we journal data on quota file, we have to flush journal to see
4548 * all updates to the file when we bypass pagecache...
4549 */
0390131b
FM		 4550 if (EXT4_SB(sb)->s_journal &&
4551 ext4_should_journal_data(path.dentry->d_inode)) {
0623543b
JK		 4552 /*
4553 * We don't need to lock updates but journal_flush() could
4554 * otherwise be livelocked...
4555 */
4556 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 4557 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
0623543b 4558 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 4559 if (err) {
8264613d 4560 path_put(&path);
7ffe1ea8
HK		 4561 return err;
4562 }
0623543b
JK		 4563 }
4564
287a8095 4565 err = dquot_quota_on_path(sb, type, format_id, &path);
8264613d 4566 path_put(&path);
77e69dac 4567 return err;
ac27a0ec
DK		 4568}
4569
ca0e05e4
DM		 4570static int ext4_quota_off(struct super_block *sb, int type)
4571{
4572 /* Force all delayed allocation blocks to be allocated */
4573 if (test_opt(sb, DELALLOC)) {
4574 down_read(&sb->s_umount);
4575 sync_filesystem(sb);
4576 up_read(&sb->s_umount);
4577 }
4578
4579 return dquot_quota_off(sb, type);
4580}
4581
ac27a0ec
DK		 4582/* Read data from quotafile - avoid pagecache and such because we cannot afford
4583 * acquiring the locks... As quota files are never truncated and quota code
4584 * itself serializes the operations (and noone else should touch the files)
4585 * we don't have to be afraid of races */
617ba13b 4586static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 4587 size_t len, loff_t off)
4588{
4589 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 4590 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 4591 int err = 0;
4592 int offset = off & (sb->s_blocksize - 1);
4593 int tocopy;
4594 size_t toread;
4595 struct buffer_head *bh;
4596 loff_t i_size = i_size_read(inode);
4597
4598 if (off > i_size)
4599 return 0;
4600 if (off+len > i_size)
4601 len = i_size-off;
4602 toread = len;
4603 while (toread > 0) {
4604 tocopy = sb->s_blocksize - offset < toread ?
4605 sb->s_blocksize - offset : toread;
617ba13b 4606 bh = ext4_bread(NULL, inode, blk, 0, &err);
ac27a0ec
DK		 4607 if (err)
4608 return err;
4609 if (!bh) /* A hole? */
4610 memset(data, 0, tocopy);
4611 else
4612 memcpy(data, bh->b_data+offset, tocopy);
4613 brelse(bh);
4614 offset = 0;
4615 toread -= tocopy;
4616 data += tocopy;
4617 blk++;
4618 }
4619 return len;
4620}
4621
4622/* Write to quotafile (we know the transaction is already started and has
4623 * enough credits) */
617ba13b 4624static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 4625 const char *data, size_t len, loff_t off)
4626{
4627 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 4628 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 4629 int err = 0;
4630 int offset = off & (sb->s_blocksize - 1);
ac27a0ec
DK		 4631 struct buffer_head *bh;
4632 handle_t *handle = journal_current_handle();
4633
0390131b 4634 if (EXT4_SB(sb)->s_journal && !handle) {
b31e1552
ES		 4635 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4636 " cancelled because transaction is not started",
9c3013e9
JK		 4637 (unsigned long long)off, (unsigned long long)len);
4638 return -EIO;
4639 }
67eeb568
DM		 4640 /*
4641 * Since we account only one data block in transaction credits,
4642 * then it is impossible to cross a block boundary.
4643 */
4644 if (sb->s_blocksize - offset < len) {
4645 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4646 " cancelled because not block aligned",
4647 (unsigned long long)off, (unsigned long long)len);
4648 return -EIO;
4649 }
4650
ac27a0ec 4651 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
67eeb568
DM		 4652 bh = ext4_bread(handle, inode, blk, 1, &err);
4653 if (!bh)
4654 goto out;
62d2b5f2
JK		 4655 err = ext4_journal_get_write_access(handle, bh);
4656 if (err) {
4657 brelse(bh);
4658 goto out;
ac27a0ec 4659 }
67eeb568
DM		 4660 lock_buffer(bh);
4661 memcpy(bh->b_data+offset, data, len);
4662 flush_dcache_page(bh->b_page);
4663 unlock_buffer(bh);
62d2b5f2 4664 err = ext4_handle_dirty_metadata(handle, NULL, bh);
67eeb568 4665 brelse(bh);
ac27a0ec 4666out:
67eeb568 4667 if (err) {
4d04e4fb 4668 mutex_unlock(&inode->i_mutex);
ac27a0ec 4669 return err;
4d04e4fb 4670 }
67eeb568
DM		 4671 if (inode->i_size < off + len) {
4672 i_size_write(inode, off + len);
617ba13b 4673 EXT4_I(inode)->i_disksize = inode->i_size;
ac27a0ec 4674 }
ac27a0ec 4675 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
617ba13b 4676 ext4_mark_inode_dirty(handle, inode);
ac27a0ec 4677 mutex_unlock(&inode->i_mutex);
67eeb568 4678 return len;
ac27a0ec
DK		 4679}
4680
4681#endif
4682
0b8e58a1
AD		 4683static int ext4_get_sb(struct file_system_type *fs_type, int flags,
4684 const char *dev_name, void *data, struct vfsmount *mnt)
ac27a0ec 4685{
0b8e58a1 4686 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super,mnt);
ac27a0ec
DK		 4687}
4688
37f328eb 4689#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
24b58424
TT		 4690static struct file_system_type ext2_fs_type = {
4691 .owner = THIS_MODULE,
4692 .name = "ext2",
4693 .get_sb = ext4_get_sb,
4694 .kill_sb = kill_block_super,
4695 .fs_flags = FS_REQUIRES_DEV,
4696};
4697
4698static inline void register_as_ext2(void)
4699{
4700 int err = register_filesystem(&ext2_fs_type);
4701 if (err)
4702 printk(KERN_WARNING
4703 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
4704}
4705
4706static inline void unregister_as_ext2(void)
4707{
4708 unregister_filesystem(&ext2_fs_type);
4709}
51b7e3c9 4710MODULE_ALIAS("ext2");
24b58424
TT		 4711#else
4712static inline void register_as_ext2(void) { }
4713static inline void unregister_as_ext2(void) { }
4714#endif
4715
37f328eb 4716#if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
24b58424
TT		 4717static inline void register_as_ext3(void)
4718{
4719 int err = register_filesystem(&ext3_fs_type);
4720 if (err)
4721 printk(KERN_WARNING
4722 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
4723}
4724
4725static inline void unregister_as_ext3(void)
4726{
4727 unregister_filesystem(&ext3_fs_type);
4728}
51b7e3c9 4729MODULE_ALIAS("ext3");
24b58424
TT		 4730#else
4731static inline void register_as_ext3(void) { }
4732static inline void unregister_as_ext3(void) { }
4733#endif
4734
03010a33
TT		 4735static struct file_system_type ext4_fs_type = {
4736 .owner = THIS_MODULE,
4737 .name = "ext4",
4738 .get_sb = ext4_get_sb,
4739 .kill_sb = kill_block_super,
4740 .fs_flags = FS_REQUIRES_DEV,
4741};
4742
857ac889
LC		 4743int __init ext4_init_feat_adverts(void)
4744{
4745 struct ext4_features *ef;
4746 int ret = -ENOMEM;
4747
4748 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
4749 if (!ef)
4750 goto out;
4751
4752 ef->f_kobj.kset = ext4_kset;
4753 init_completion(&ef->f_kobj_unregister);
4754 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
4755 "features");
4756 if (ret) {
4757 kfree(ef);
4758 goto out;
4759 }
4760
4761 ext4_feat = ef;
4762 ret = 0;
4763out:
4764 return ret;
4765}
4766
617ba13b 4767static int __init init_ext4_fs(void)
ac27a0ec 4768{
c9de560d
AT		 4769 int err;
4770
12e9b892 4771 ext4_check_flag_values();
bd2d0210 4772 err = init_ext4_pageio();
6fd058f7
TT		 4773 if (err)
4774 return err;
bd2d0210
TT		 4775 err = init_ext4_system_zone();
4776 if (err)
4777 goto out5;
3197ebdb
TT		 4778 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
4779 if (!ext4_kset)
6fd058f7 4780 goto out4;
9f6200bb 4781 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
857ac889
LC		 4782
4783 err = ext4_init_feat_adverts();
4784
c9de560d 4785 err = init_ext4_mballoc();
ac27a0ec 4786 if (err)
6fd058f7 4787 goto out3;
c9de560d
AT		 4788
4789 err = init_ext4_xattr();
4790 if (err)
4791 goto out2;
ac27a0ec
DK		 4792 err = init_inodecache();
4793 if (err)
4794 goto out1;
24b58424
TT		 4795 register_as_ext2();
4796 register_as_ext3();
03010a33 4797 err = register_filesystem(&ext4_fs_type);
ac27a0ec
DK		 4798 if (err)
4799 goto out;
bfff6873
LC		 4800
4801 ext4_li_info = NULL;
4802 mutex_init(&ext4_li_mtx);
ac27a0ec
DK		 4803 return 0;
4804out:
24b58424
TT		 4805 unregister_as_ext2();
4806 unregister_as_ext3();
ac27a0ec
DK		 4807 destroy_inodecache();
4808out1:
617ba13b 4809 exit_ext4_xattr();
c9de560d
AT		 4810out2:
4811 exit_ext4_mballoc();
6fd058f7 4812out3:
857ac889 4813 kfree(ext4_feat);
6fd058f7
TT		 4814 remove_proc_entry("fs/ext4", NULL);
4815 kset_unregister(ext4_kset);
4816out4:
4817 exit_ext4_system_zone();
bd2d0210
TT		 4818out5:
4819 exit_ext4_pageio();
ac27a0ec
DK		 4820 return err;
4821}
4822
617ba13b 4823static void __exit exit_ext4_fs(void)
ac27a0ec 4824{
bfff6873 4825 ext4_destroy_lazyinit_thread();
24b58424
TT		 4826 unregister_as_ext2();
4827 unregister_as_ext3();
03010a33 4828 unregister_filesystem(&ext4_fs_type);
ac27a0ec 4829 destroy_inodecache();
617ba13b 4830 exit_ext4_xattr();
c9de560d 4831 exit_ext4_mballoc();
9f6200bb 4832 remove_proc_entry("fs/ext4", NULL);
3197ebdb 4833 kset_unregister(ext4_kset);
6fd058f7 4834 exit_ext4_system_zone();
bd2d0210 4835 exit_ext4_pageio();
ac27a0ec
DK		 4836}
4837
4838MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
83982b6f 4839MODULE_DESCRIPTION("Fourth Extended Filesystem");
ac27a0ec 4840MODULE_LICENSE("GPL");
617ba13b
MC		 4841module_init(init_ext4_fs)
4842module_exit(exit_ext4_fs)
kernel source for telekom puls (alcatel/mediatek)