]> bbs.cooldavid.org Git - net-next-2.6.git/blame - fs/ext4/super.c
git://bbs.cooldavid.org / net-next-2.6.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
quota: decouple fs reserved space from quota reservation
[net-next-2.6.git] / fs / ext4 / super.c
CommitLineData
ac27a0ec 1/*
617ba13b 2 * linux/fs/ext4/super.c
ac27a0ec
DK		 3 *
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
8 *
9 * from
10 *
11 * linux/fs/minix/inode.c
12 *
13 * Copyright (C) 1991, 1992 Linus Torvalds
14 *
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
17 */
18
19#include <linux/module.h>
20#include <linux/string.h>
21#include <linux/fs.h>
22#include <linux/time.h>
c5ca7c76 23#include <linux/vmalloc.h>
dab291af 24#include <linux/jbd2.h>
ac27a0ec
DK		 25#include <linux/slab.h>
26#include <linux/init.h>
27#include <linux/blkdev.h>
28#include <linux/parser.h>
29#include <linux/smp_lock.h>
30#include <linux/buffer_head.h>
a5694255 31#include <linux/exportfs.h>
ac27a0ec
DK		 32#include <linux/vfs.h>
33#include <linux/random.h>
34#include <linux/mount.h>
35#include <linux/namei.h>
36#include <linux/quotaops.h>
37#include <linux/seq_file.h>
9f6200bb 38#include <linux/proc_fs.h>
3197ebdb 39#include <linux/ctype.h>
1330593e 40#include <linux/log2.h>
717d50e4 41#include <linux/crc16.h>
ac27a0ec
DK		 42#include <asm/uaccess.h>
43
3dcf5451
CH		 44#include "ext4.h"
45#include "ext4_jbd2.h"
ac27a0ec
DK		 46#include "xattr.h"
47#include "acl.h"
3661d286 48#include "mballoc.h"
ac27a0ec 49
9bffad1e
TT		 50#define CREATE_TRACE_POINTS
51#include <trace/events/ext4.h>
52
9f6200bb 53struct proc_dir_entry *ext4_proc_root;
3197ebdb 54static struct kset *ext4_kset;
9f6200bb 55
617ba13b 56static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
ac27a0ec 57 unsigned long journal_devnum);
e2d67052 58static int ext4_commit_super(struct super_block *sb, int sync);
2b2d6d01
TT		 59static void ext4_mark_recovery_complete(struct super_block *sb,
60 struct ext4_super_block *es);
61static void ext4_clear_journal_err(struct super_block *sb,
62 struct ext4_super_block *es);
617ba13b 63static int ext4_sync_fs(struct super_block *sb, int wait);
2b2d6d01 64static const char *ext4_decode_error(struct super_block *sb, int errno,
ac27a0ec 65 char nbuf[16]);
2b2d6d01
TT		 66static int ext4_remount(struct super_block *sb, int *flags, char *data);
67static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
c4be0c1d 68static int ext4_unfreeze(struct super_block *sb);
2b2d6d01 69static void ext4_write_super(struct super_block *sb);
c4be0c1d 70static int ext4_freeze(struct super_block *sb);
ac27a0ec 71
bd81d8ee 72
8fadc143
AR		 73ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
74 struct ext4_group_desc *bg)
bd81d8ee 75{
3a14589c 76 return le32_to_cpu(bg->bg_block_bitmap_lo) |
8fadc143 77 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 78 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 79}
80
8fadc143
AR		 81ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
82 struct ext4_group_desc *bg)
bd81d8ee 83{
5272f837 84 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
8fadc143 85 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 86 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 87}
88
8fadc143
AR		 89ext4_fsblk_t ext4_inode_table(struct super_block *sb,
90 struct ext4_group_desc *bg)
bd81d8ee 91{
5272f837 92 return le32_to_cpu(bg->bg_inode_table_lo) |
8fadc143 93 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 94 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
bd81d8ee
LV		 95}
96
560671a0
AK		 97__u32 ext4_free_blks_count(struct super_block *sb,
98 struct ext4_group_desc *bg)
99{
100 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
101 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 102 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
560671a0
AK		 103}
104
105__u32 ext4_free_inodes_count(struct super_block *sb,
106 struct ext4_group_desc *bg)
107{
108 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
109 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 110 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
560671a0
AK		 111}
112
113__u32 ext4_used_dirs_count(struct super_block *sb,
114 struct ext4_group_desc *bg)
115{
116 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
117 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 118 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
560671a0
AK		 119}
120
121__u32 ext4_itable_unused_count(struct super_block *sb,
122 struct ext4_group_desc *bg)
123{
124 return le16_to_cpu(bg->bg_itable_unused_lo) |
125 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 126 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
560671a0
AK		 127}
128
8fadc143
AR		 129void ext4_block_bitmap_set(struct super_block *sb,
130 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 131{
3a14589c 132 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 133 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
134 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 135}
136
8fadc143
AR		 137void ext4_inode_bitmap_set(struct super_block *sb,
138 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 139{
5272f837 140 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 141 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
142 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 143}
144
8fadc143
AR		 145void ext4_inode_table_set(struct super_block *sb,
146 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 147{
5272f837 148 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 149 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
150 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 151}
152
560671a0
AK		 153void ext4_free_blks_set(struct super_block *sb,
154 struct ext4_group_desc *bg, __u32 count)
155{
156 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
157 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
158 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
159}
160
161void ext4_free_inodes_set(struct super_block *sb,
162 struct ext4_group_desc *bg, __u32 count)
163{
164 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
165 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
166 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
167}
168
169void ext4_used_dirs_set(struct super_block *sb,
170 struct ext4_group_desc *bg, __u32 count)
171{
172 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
173 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
174 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
175}
176
177void ext4_itable_unused_set(struct super_block *sb,
178 struct ext4_group_desc *bg, __u32 count)
179{
180 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
181 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
182 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
183}
184
d3d1faf6
CW		 185
186/* Just increment the non-pointer handle value */
187static handle_t *ext4_get_nojournal(void)
188{
189 handle_t *handle = current->journal_info;
190 unsigned long ref_cnt = (unsigned long)handle;
191
192 BUG_ON(ref_cnt >= EXT4_NOJOURNAL_MAX_REF_COUNT);
193
194 ref_cnt++;
195 handle = (handle_t *)ref_cnt;
196
197 current->journal_info = handle;
198 return handle;
199}
200
201
202/* Decrement the non-pointer handle value */
203static void ext4_put_nojournal(handle_t *handle)
204{
205 unsigned long ref_cnt = (unsigned long)handle;
206
207 BUG_ON(ref_cnt == 0);
208
209 ref_cnt--;
210 handle = (handle_t *)ref_cnt;
211
212 current->journal_info = handle;
213}
214
ac27a0ec 215/*
dab291af 216 * Wrappers for jbd2_journal_start/end.
ac27a0ec
DK		 217 *
218 * The only special thing we need to do here is to make sure that all
219 * journal_end calls result in the superblock being marked dirty, so
220 * that sync() will call the filesystem's write_super callback if
221 * appropriate.
222 */
617ba13b 223handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
ac27a0ec
DK		 224{
225 journal_t *journal;
226
227 if (sb->s_flags & MS_RDONLY)
228 return ERR_PTR(-EROFS);
229
230 /* Special case here: if the journal has aborted behind our
231 * backs (eg. EIO in the commit thread), then we still need to
232 * take the FS itself readonly cleanly. */
617ba13b 233 journal = EXT4_SB(sb)->s_journal;
0390131b
FM		 234 if (journal) {
235 if (is_journal_aborted(journal)) {
0b8e58a1 236 ext4_abort(sb, __func__, "Detected aborted journal");
0390131b
FM		 237 return ERR_PTR(-EROFS);
238 }
239 return jbd2_journal_start(journal, nblocks);
ac27a0ec 240 }
d3d1faf6 241 return ext4_get_nojournal();
ac27a0ec
DK		 242}
243
244/*
245 * The only special thing we need to do here is to make sure that all
dab291af 246 * jbd2_journal_stop calls result in the superblock being marked dirty, so
ac27a0ec
DK		 247 * that sync() will call the filesystem's write_super callback if
248 * appropriate.
249 */
617ba13b 250int __ext4_journal_stop(const char *where, handle_t *handle)
ac27a0ec
DK		 251{
252 struct super_block *sb;
253 int err;
254 int rc;
255
0390131b 256 if (!ext4_handle_valid(handle)) {
d3d1faf6 257 ext4_put_nojournal(handle);
0390131b
FM		 258 return 0;
259 }
ac27a0ec
DK		 260 sb = handle->h_transaction->t_journal->j_private;
261 err = handle->h_err;
dab291af 262 rc = jbd2_journal_stop(handle);
ac27a0ec
DK		 263
264 if (!err)
265 err = rc;
266 if (err)
617ba13b 267 __ext4_std_error(sb, where, err);
ac27a0ec
DK		 268 return err;
269}
270
617ba13b 271void ext4_journal_abort_handle(const char *caller, const char *err_fn,
ac27a0ec
DK		 272 struct buffer_head *bh, handle_t *handle, int err)
273{
274 char nbuf[16];
617ba13b 275 const char *errstr = ext4_decode_error(NULL, err, nbuf);
ac27a0ec 276
0390131b
FM		 277 BUG_ON(!ext4_handle_valid(handle));
278
ac27a0ec
DK		 279 if (bh)
280 BUFFER_TRACE(bh, "abort");
281
282 if (!handle->h_err)
283 handle->h_err = err;
284
285 if (is_handle_aborted(handle))
286 return;
287
288 printk(KERN_ERR "%s: aborting transaction: %s in %s\n",
289 caller, errstr, err_fn);
290
dab291af 291 jbd2_journal_abort_handle(handle);
ac27a0ec
DK		 292}
293
294/* Deal with the reporting of failure conditions on a filesystem such as
295 * inconsistencies detected or read IO failures.
296 *
297 * On ext2, we can store the error state of the filesystem in the
617ba13b 298 * superblock. That is not possible on ext4, because we may have other
ac27a0ec
DK		 299 * write ordering constraints on the superblock which prevent us from
300 * writing it out straight away; and given that the journal is about to
301 * be aborted, we can't rely on the current, or future, transactions to
302 * write out the superblock safely.
303 *
dab291af 304 * We'll just use the jbd2_journal_abort() error code to record an error in
ac27a0ec
DK		 305 * the journal instead. On recovery, the journal will compain about
306 * that error until we've noted it down and cleared it.
307 */
308
617ba13b 309static void ext4_handle_error(struct super_block *sb)
ac27a0ec 310{
617ba13b 311 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 312
617ba13b
MC		 313 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
314 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
ac27a0ec
DK		 315
316 if (sb->s_flags & MS_RDONLY)
317 return;
318
2b2d6d01 319 if (!test_opt(sb, ERRORS_CONT)) {
617ba13b 320 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 321
4ab2f15b 322 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
ac27a0ec 323 if (journal)
dab291af 324 jbd2_journal_abort(journal, -EIO);
ac27a0ec 325 }
2b2d6d01 326 if (test_opt(sb, ERRORS_RO)) {
b31e1552 327 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
ac27a0ec
DK		 328 sb->s_flags |= MS_RDONLY;
329 }
e2d67052 330 ext4_commit_super(sb, 1);
ac27a0ec 331 if (test_opt(sb, ERRORS_PANIC))
617ba13b 332 panic("EXT4-fs (device %s): panic forced after error\n",
ac27a0ec
DK		 333 sb->s_id);
334}
335
2b2d6d01
TT		 336void ext4_error(struct super_block *sb, const char *function,
337 const char *fmt, ...)
ac27a0ec
DK		 338{
339 va_list args;
340
341 va_start(args, fmt);
2b2d6d01 342 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
ac27a0ec
DK		 343 vprintk(fmt, args);
344 printk("\n");
345 va_end(args);
346
617ba13b 347 ext4_handle_error(sb);
ac27a0ec
DK		 348}
349
2b2d6d01 350static const char *ext4_decode_error(struct super_block *sb, int errno,
ac27a0ec
DK		 351 char nbuf[16])
352{
353 char *errstr = NULL;
354
355 switch (errno) {
356 case -EIO:
357 errstr = "IO failure";
358 break;
359 case -ENOMEM:
360 errstr = "Out of memory";
361 break;
362 case -EROFS:
78f1ddbb
TT		 363 if (!sb || (EXT4_SB(sb)->s_journal &&
364 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
ac27a0ec
DK		 365 errstr = "Journal has aborted";
366 else
367 errstr = "Readonly filesystem";
368 break;
369 default:
370 /* If the caller passed in an extra buffer for unknown
371 * errors, textualise them now. Else we just return
372 * NULL. */
373 if (nbuf) {
374 /* Check for truncated error codes... */
375 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
376 errstr = nbuf;
377 }
378 break;
379 }
380
381 return errstr;
382}
383
617ba13b 384/* __ext4_std_error decodes expected errors from journaling functions
ac27a0ec
DK		 385 * automatically and invokes the appropriate error response. */
386
2b2d6d01 387void __ext4_std_error(struct super_block *sb, const char *function, int errno)
ac27a0ec
DK		 388{
389 char nbuf[16];
390 const char *errstr;
391
392 /* Special case: if the error is EROFS, and we're not already
393 * inside a transaction, then there's really no point in logging
394 * an error. */
395 if (errno == -EROFS && journal_current_handle() == NULL &&
396 (sb->s_flags & MS_RDONLY))
397 return;
398
617ba13b 399 errstr = ext4_decode_error(sb, errno, nbuf);
2b2d6d01
TT		 400 printk(KERN_CRIT "EXT4-fs error (device %s) in %s: %s\n",
401 sb->s_id, function, errstr);
ac27a0ec 402
617ba13b 403 ext4_handle_error(sb);
ac27a0ec
DK		 404}
405
406/*
617ba13b 407 * ext4_abort is a much stronger failure handler than ext4_error. The
ac27a0ec
DK		 408 * abort function may be used to deal with unrecoverable failures such
409 * as journal IO errors or ENOMEM at a critical moment in log management.
410 *
411 * We unconditionally force the filesystem into an ABORT|READONLY state,
412 * unless the error response on the fs has been set to panic in which
413 * case we take the easy way out and panic immediately.
414 */
415
2b2d6d01
TT		 416void ext4_abort(struct super_block *sb, const char *function,
417 const char *fmt, ...)
ac27a0ec
DK		 418{
419 va_list args;
420
ac27a0ec 421 va_start(args, fmt);
2b2d6d01 422 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
ac27a0ec
DK		 423 vprintk(fmt, args);
424 printk("\n");
425 va_end(args);
426
427 if (test_opt(sb, ERRORS_PANIC))
617ba13b 428 panic("EXT4-fs panic from previous error\n");
ac27a0ec
DK		 429
430 if (sb->s_flags & MS_RDONLY)
431 return;
432
b31e1552 433 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
617ba13b 434 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
ac27a0ec 435 sb->s_flags |= MS_RDONLY;
4ab2f15b 436 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
ef2cabf7
HK		 437 if (EXT4_SB(sb)->s_journal)
438 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
ac27a0ec
DK		 439}
440
b31e1552
ES		 441void ext4_msg (struct super_block * sb, const char *prefix,
442 const char *fmt, ...)
443{
444 va_list args;
445
446 va_start(args, fmt);
447 printk("%sEXT4-fs (%s): ", prefix, sb->s_id);
448 vprintk(fmt, args);
449 printk("\n");
450 va_end(args);
451}
452
2b2d6d01
TT		 453void ext4_warning(struct super_block *sb, const char *function,
454 const char *fmt, ...)
ac27a0ec
DK		 455{
456 va_list args;
457
458 va_start(args, fmt);
617ba13b 459 printk(KERN_WARNING "EXT4-fs warning (device %s): %s: ",
ac27a0ec
DK		 460 sb->s_id, function);
461 vprintk(fmt, args);
462 printk("\n");
463 va_end(args);
464}
465
5d1b1b3f 466void ext4_grp_locked_error(struct super_block *sb, ext4_group_t grp,
0b8e58a1 467 const char *function, const char *fmt, ...)
5d1b1b3f
AK		 468__releases(bitlock)
469__acquires(bitlock)
470{
471 va_list args;
472 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
473
474 va_start(args, fmt);
475 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
476 vprintk(fmt, args);
477 printk("\n");
478 va_end(args);
479
480 if (test_opt(sb, ERRORS_CONT)) {
481 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
482 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
e2d67052 483 ext4_commit_super(sb, 0);
5d1b1b3f
AK		 484 return;
485 }
486 ext4_unlock_group(sb, grp);
487 ext4_handle_error(sb);
488 /*
489 * We only get here in the ERRORS_RO case; relocking the group
490 * may be dangerous, but nothing bad will happen since the
491 * filesystem will have already been marked read/only and the
492 * journal has been aborted. We return 1 as a hint to callers
493 * who might what to use the return value from
494 * ext4_grp_locked_error() to distinguish beween the
495 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
496 * aggressively from the ext4 function in question, with a
497 * more appropriate error code.
498 */
499 ext4_lock_group(sb, grp);
500 return;
501}
502
617ba13b 503void ext4_update_dynamic_rev(struct super_block *sb)
ac27a0ec 504{
617ba13b 505 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 506
617ba13b 507 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
ac27a0ec
DK		 508 return;
509
46e665e9 510 ext4_warning(sb, __func__,
ac27a0ec
DK		 511 "updating to rev %d because of new feature flag, "
512 "running e2fsck is recommended",
617ba13b 513 EXT4_DYNAMIC_REV);
ac27a0ec 514
617ba13b
MC		 515 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
516 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
517 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
ac27a0ec
DK		 518 /* leave es->s_feature_*compat flags alone */
519 /* es->s_uuid will be set by e2fsck if empty */
520
521 /*
522 * The rest of the superblock fields should be zero, and if not it
523 * means they are likely already in use, so leave them alone. We
524 * can leave it up to e2fsck to clean up any inconsistencies there.
525 */
526}
527
528/*
529 * Open the external journal device
530 */
b31e1552 531static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
ac27a0ec
DK		 532{
533 struct block_device *bdev;
534 char b[BDEVNAME_SIZE];
535
536 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
537 if (IS_ERR(bdev))
538 goto fail;
539 return bdev;
540
541fail:
b31e1552 542 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
ac27a0ec
DK		 543 __bdevname(dev, b), PTR_ERR(bdev));
544 return NULL;
545}
546
547/*
548 * Release the journal device
549 */
617ba13b 550static int ext4_blkdev_put(struct block_device *bdev)
ac27a0ec
DK		 551{
552 bd_release(bdev);
9a1c3542 553 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
ac27a0ec
DK		 554}
555
617ba13b 556static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
ac27a0ec
DK		 557{
558 struct block_device *bdev;
559 int ret = -ENODEV;
560
561 bdev = sbi->journal_bdev;
562 if (bdev) {
617ba13b 563 ret = ext4_blkdev_put(bdev);
ac27a0ec
DK		 564 sbi->journal_bdev = NULL;
565 }
566 return ret;
567}
568
569static inline struct inode *orphan_list_entry(struct list_head *l)
570{
617ba13b 571 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
ac27a0ec
DK		 572}
573
617ba13b 574static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
ac27a0ec
DK		 575{
576 struct list_head *l;
577
b31e1552
ES		 578 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
579 le32_to_cpu(sbi->s_es->s_last_orphan));
ac27a0ec
DK		 580
581 printk(KERN_ERR "sb_info orphan list:\n");
582 list_for_each(l, &sbi->s_orphan) {
583 struct inode *inode = orphan_list_entry(l);
584 printk(KERN_ERR " "
585 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
586 inode->i_sb->s_id, inode->i_ino, inode,
587 inode->i_mode, inode->i_nlink,
588 NEXT_ORPHAN(inode));
589 }
590}
591
2b2d6d01 592static void ext4_put_super(struct super_block *sb)
ac27a0ec 593{
617ba13b
MC		 594 struct ext4_sb_info *sbi = EXT4_SB(sb);
595 struct ext4_super_block *es = sbi->s_es;
ef2cabf7 596 int i, err;
ac27a0ec 597
4c0425ff
MC		 598 flush_workqueue(sbi->dio_unwritten_wq);
599 destroy_workqueue(sbi->dio_unwritten_wq);
600
a9e220f8 601 lock_super(sb);
6cfd0148 602 lock_kernel();
8c85e125 603 if (sb->s_dirt)
ebc1ac16 604 ext4_commit_super(sb, 1);
8c85e125 605
0390131b
FM		 606 if (sbi->s_journal) {
607 err = jbd2_journal_destroy(sbi->s_journal);
608 sbi->s_journal = NULL;
609 if (err < 0)
610 ext4_abort(sb, __func__,
611 "Couldn't clean up the journal");
612 }
d4edac31
JB		 613
614 ext4_release_system_zone(sb);
615 ext4_mb_release(sb);
616 ext4_ext_release(sb);
617 ext4_xattr_put_super(sb);
618
ac27a0ec 619 if (!(sb->s_flags & MS_RDONLY)) {
617ba13b 620 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 621 es->s_state = cpu_to_le16(sbi->s_mount_state);
e2d67052 622 ext4_commit_super(sb, 1);
ac27a0ec 623 }
240799cd 624 if (sbi->s_proc) {
9f6200bb 625 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 626 }
3197ebdb 627 kobject_del(&sbi->s_kobj);
ac27a0ec
DK		 628
629 for (i = 0; i < sbi->s_gdb_count; i++)
630 brelse(sbi->s_group_desc[i]);
631 kfree(sbi->s_group_desc);
c5ca7c76
TT		 632 if (is_vmalloc_addr(sbi->s_flex_groups))
633 vfree(sbi->s_flex_groups);
634 else
635 kfree(sbi->s_flex_groups);
ac27a0ec
DK		 636 percpu_counter_destroy(&sbi->s_freeblocks_counter);
637 percpu_counter_destroy(&sbi->s_freeinodes_counter);
638 percpu_counter_destroy(&sbi->s_dirs_counter);
6bc6e63f 639 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
ac27a0ec
DK		 640 brelse(sbi->s_sbh);
641#ifdef CONFIG_QUOTA
642 for (i = 0; i < MAXQUOTAS; i++)
643 kfree(sbi->s_qf_names[i]);
644#endif
645
646 /* Debugging code just in case the in-memory inode orphan list
647 * isn't empty. The on-disk one can be non-empty if we've
648 * detected an error and taken the fs readonly, but the
649 * in-memory list had better be clean by this point. */
650 if (!list_empty(&sbi->s_orphan))
651 dump_orphan_list(sb, sbi);
652 J_ASSERT(list_empty(&sbi->s_orphan));
653
f98393a6 654 invalidate_bdev(sb->s_bdev);
ac27a0ec
DK		 655 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
656 /*
657 * Invalidate the journal device's buffers. We don't want them
658 * floating about in memory - the physical journal device may
659 * hotswapped, and it breaks the `ro-after' testing code.
660 */
661 sync_blockdev(sbi->journal_bdev);
f98393a6 662 invalidate_bdev(sbi->journal_bdev);
617ba13b 663 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 664 }
665 sb->s_fs_info = NULL;
3197ebdb
TT		 666 /*
667 * Now that we are completely done shutting down the
668 * superblock, we need to actually destroy the kobject.
669 */
670 unlock_kernel();
671 unlock_super(sb);
672 kobject_put(&sbi->s_kobj);
673 wait_for_completion(&sbi->s_kobj_unregister);
705895b6 674 kfree(sbi->s_blockgroup_lock);
ac27a0ec 675 kfree(sbi);
ac27a0ec
DK		 676}
677
e18b890b 678static struct kmem_cache *ext4_inode_cachep;
ac27a0ec
DK		 679
680/*
681 * Called inside transaction, so use GFP_NOFS
682 */
617ba13b 683static struct inode *ext4_alloc_inode(struct super_block *sb)
ac27a0ec 684{
617ba13b 685 struct ext4_inode_info *ei;
ac27a0ec 686
e6b4f8da 687 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
ac27a0ec
DK		 688 if (!ei)
689 return NULL;
0b8e58a1 690
ac27a0ec 691 ei->vfs_inode.i_version = 1;
91246c00 692 ei->vfs_inode.i_data.writeback_index = 0;
a86c6181 693 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
c9de560d
AT		 694 INIT_LIST_HEAD(&ei->i_prealloc_list);
695 spin_lock_init(&ei->i_prealloc_lock);
0390131b
FM		 696 /*
697 * Note: We can be called before EXT4_SB(sb)->s_journal is set,
698 * therefore it can be null here. Don't check it, just initialize
699 * jinode.
700 */
678aaf48 701 jbd2_journal_init_jbd_inode(&ei->jinode, &ei->vfs_inode);
d2a17637
MC		 702 ei->i_reserved_data_blocks = 0;
703 ei->i_reserved_meta_blocks = 0;
704 ei->i_allocated_meta_blocks = 0;
705 ei->i_delalloc_reserved_flag = 0;
706 spin_lock_init(&(ei->i_block_reservation_lock));
8d5d02e6
MC		 707 INIT_LIST_HEAD(&ei->i_aio_dio_complete_list);
708 ei->cur_aio_dio = NULL;
b436b9be
JK		 709 ei->i_sync_tid = 0;
710 ei->i_datasync_tid = 0;
0b8e58a1 711
ac27a0ec
DK		 712 return &ei->vfs_inode;
713}
714
617ba13b 715static void ext4_destroy_inode(struct inode *inode)
ac27a0ec 716{
9f7dd93d 717 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
b31e1552
ES		 718 ext4_msg(inode->i_sb, KERN_ERR,
719 "Inode %lu (%p): orphan list check failed!",
720 inode->i_ino, EXT4_I(inode));
9f7dd93d
VA		 721 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
722 EXT4_I(inode), sizeof(struct ext4_inode_info),
723 true);
724 dump_stack();
725 }
617ba13b 726 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
ac27a0ec
DK		 727}
728
51cc5068 729static void init_once(void *foo)
ac27a0ec 730{
617ba13b 731 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
ac27a0ec 732
a35afb83 733 INIT_LIST_HEAD(&ei->i_orphan);
03010a33 734#ifdef CONFIG_EXT4_FS_XATTR
a35afb83 735 init_rwsem(&ei->xattr_sem);
ac27a0ec 736#endif
0e855ac8 737 init_rwsem(&ei->i_data_sem);
a35afb83 738 inode_init_once(&ei->vfs_inode);
ac27a0ec
DK		 739}
740
741static int init_inodecache(void)
742{
617ba13b
MC		 743 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
744 sizeof(struct ext4_inode_info),
ac27a0ec
DK		 745 0, (SLAB_RECLAIM_ACCOUNT|
746 SLAB_MEM_SPREAD),
20c2df83 747 init_once);
617ba13b 748 if (ext4_inode_cachep == NULL)
ac27a0ec
DK		 749 return -ENOMEM;
750 return 0;
751}
752
753static void destroy_inodecache(void)
754{
617ba13b 755 kmem_cache_destroy(ext4_inode_cachep);
ac27a0ec
DK		 756}
757
617ba13b 758static void ext4_clear_inode(struct inode *inode)
ac27a0ec 759{
c2ea3fde 760 ext4_discard_preallocations(inode);
0390131b
FM		 761 if (EXT4_JOURNAL(inode))
762 jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal,
678aaf48 763 &EXT4_I(inode)->jinode);
ac27a0ec
DK		 764}
765
2b2d6d01
TT		 766static inline void ext4_show_quota_options(struct seq_file *seq,
767 struct super_block *sb)
ac27a0ec
DK		 768{
769#if defined(CONFIG_QUOTA)
617ba13b 770 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 771
5a20bdfc
JK		 772 if (sbi->s_jquota_fmt) {
773 char *fmtname = "";
774
775 switch (sbi->s_jquota_fmt) {
776 case QFMT_VFS_OLD:
777 fmtname = "vfsold";
778 break;
779 case QFMT_VFS_V0:
780 fmtname = "vfsv0";
781 break;
782 case QFMT_VFS_V1:
783 fmtname = "vfsv1";
784 break;
785 }
786 seq_printf(seq, ",jqfmt=%s", fmtname);
787 }
ac27a0ec
DK		 788
789 if (sbi->s_qf_names[USRQUOTA])
790 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
791
792 if (sbi->s_qf_names[GRPQUOTA])
793 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
794
617ba13b 795 if (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA)
ac27a0ec
DK		 796 seq_puts(seq, ",usrquota");
797
617ba13b 798 if (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)
ac27a0ec
DK		 799 seq_puts(seq, ",grpquota");
800#endif
801}
802
d9c9bef1
MS		 803/*
804 * Show an option if
805 * - it's set to a non-default value OR
806 * - if the per-sb default is different from the global default
807 */
617ba13b 808static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
ac27a0ec 809{
aa22df2c
AK		 810 int def_errors;
811 unsigned long def_mount_opts;
ac27a0ec 812 struct super_block *sb = vfs->mnt_sb;
d9c9bef1
MS		 813 struct ext4_sb_info *sbi = EXT4_SB(sb);
814 struct ext4_super_block *es = sbi->s_es;
d9c9bef1
MS		 815
816 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
aa22df2c 817 def_errors = le16_to_cpu(es->s_errors);
d9c9bef1
MS		 818
819 if (sbi->s_sb_block != 1)
820 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
821 if (test_opt(sb, MINIX_DF))
822 seq_puts(seq, ",minixdf");
aa22df2c 823 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
d9c9bef1
MS		 824 seq_puts(seq, ",grpid");
825 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
826 seq_puts(seq, ",nogrpid");
827 if (sbi->s_resuid != EXT4_DEF_RESUID ||
828 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
829 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
830 }
831 if (sbi->s_resgid != EXT4_DEF_RESGID ||
832 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
833 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
834 }
bb4f397a 835 if (test_opt(sb, ERRORS_RO)) {
d9c9bef1 836 if (def_errors == EXT4_ERRORS_PANIC ||
bb4f397a
AK		 837 def_errors == EXT4_ERRORS_CONTINUE) {
838 seq_puts(seq, ",errors=remount-ro");
d9c9bef1
MS		 839 }
840 }
aa22df2c 841 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
bb4f397a 842 seq_puts(seq, ",errors=continue");
aa22df2c 843 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
d9c9bef1 844 seq_puts(seq, ",errors=panic");
aa22df2c 845 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
d9c9bef1 846 seq_puts(seq, ",nouid32");
aa22df2c 847 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
d9c9bef1
MS		 848 seq_puts(seq, ",debug");
849 if (test_opt(sb, OLDALLOC))
850 seq_puts(seq, ",oldalloc");
03010a33 851#ifdef CONFIG_EXT4_FS_XATTR
aa22df2c
AK		 852 if (test_opt(sb, XATTR_USER) &&
853 !(def_mount_opts & EXT4_DEFM_XATTR_USER))
d9c9bef1
MS		 854 seq_puts(seq, ",user_xattr");
855 if (!test_opt(sb, XATTR_USER) &&
856 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
857 seq_puts(seq, ",nouser_xattr");
858 }
859#endif
03010a33 860#ifdef CONFIG_EXT4_FS_POSIX_ACL
aa22df2c 861 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
d9c9bef1
MS		 862 seq_puts(seq, ",acl");
863 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
864 seq_puts(seq, ",noacl");
865#endif
30773840 866 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
d9c9bef1
MS		 867 seq_printf(seq, ",commit=%u",
868 (unsigned) (sbi->s_commit_interval / HZ));
869 }
30773840
TT		 870 if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
871 seq_printf(seq, ",min_batch_time=%u",
872 (unsigned) sbi->s_min_batch_time);
873 }
874 if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
875 seq_printf(seq, ",max_batch_time=%u",
876 (unsigned) sbi->s_min_batch_time);
877 }
878
571640ca
ES		 879 /*
880 * We're changing the default of barrier mount option, so
881 * let's always display its mount state so it's clear what its
882 * status is.
883 */
884 seq_puts(seq, ",barrier=");
885 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
cd0b6a39
TT		 886 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
887 seq_puts(seq, ",journal_async_commit");
d9c9bef1
MS		 888 if (test_opt(sb, NOBH))
889 seq_puts(seq, ",nobh");
25ec56b5
JNC		 890 if (test_opt(sb, I_VERSION))
891 seq_puts(seq, ",i_version");
dd919b98
AK		 892 if (!test_opt(sb, DELALLOC))
893 seq_puts(seq, ",nodelalloc");
894
ac27a0ec 895
cb45bbe4
MS		 896 if (sbi->s_stripe)
897 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
aa22df2c
AK		 898 /*
899 * journal mode get enabled in different ways
900 * So just print the value even if we didn't specify it
901 */
617ba13b 902 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
ac27a0ec 903 seq_puts(seq, ",data=journal");
617ba13b 904 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
ac27a0ec 905 seq_puts(seq, ",data=ordered");
617ba13b 906 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
ac27a0ec
DK		 907 seq_puts(seq, ",data=writeback");
908
240799cd
TT		 909 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
910 seq_printf(seq, ",inode_readahead_blks=%u",
911 sbi->s_inode_readahead_blks);
912
5bf5683a
HK		 913 if (test_opt(sb, DATA_ERR_ABORT))
914 seq_puts(seq, ",data_err=abort");
915
afd4672d 916 if (test_opt(sb, NO_AUTO_DA_ALLOC))
06705bff 917 seq_puts(seq, ",noauto_da_alloc");
afd4672d 918
5328e635
ES		 919 if (test_opt(sb, DISCARD))
920 seq_puts(seq, ",discard");
921
e3bb52ae
ES		 922 if (test_opt(sb, NOLOAD))
923 seq_puts(seq, ",norecovery");
924
617ba13b 925 ext4_show_quota_options(seq, sb);
0b8e58a1 926
ac27a0ec
DK		 927 return 0;
928}
929
1b961ac0 930static struct inode *ext4_nfs_get_inode(struct super_block *sb,
0b8e58a1 931 u64 ino, u32 generation)
ac27a0ec 932{
ac27a0ec 933 struct inode *inode;
ac27a0ec 934
617ba13b 935 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
ac27a0ec 936 return ERR_PTR(-ESTALE);
617ba13b 937 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
ac27a0ec
DK		 938 return ERR_PTR(-ESTALE);
939
940 /* iget isn't really right if the inode is currently unallocated!!
941 *
617ba13b 942 * ext4_read_inode will return a bad_inode if the inode had been
ac27a0ec
DK		 943 * deleted, so we should be safe.
944 *
945 * Currently we don't know the generation for parent directory, so
946 * a generation of 0 means "accept any"
947 */
1d1fe1ee
DH		 948 inode = ext4_iget(sb, ino);
949 if (IS_ERR(inode))
950 return ERR_CAST(inode);
951 if (generation && inode->i_generation != generation) {
ac27a0ec
DK		 952 iput(inode);
953 return ERR_PTR(-ESTALE);
954 }
1b961ac0
CH		 955
956 return inode;
957}
958
959static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
0b8e58a1 960 int fh_len, int fh_type)
1b961ac0
CH		 961{
962 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
963 ext4_nfs_get_inode);
964}
965
966static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
0b8e58a1 967 int fh_len, int fh_type)
1b961ac0
CH		 968{
969 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
970 ext4_nfs_get_inode);
ac27a0ec
DK		 971}
972
c39a7f84
TO		 973/*
974 * Try to release metadata pages (indirect blocks, directories) which are
975 * mapped via the block device. Since these pages could have journal heads
976 * which would prevent try_to_free_buffers() from freeing them, we must use
977 * jbd2 layer's try_to_free_buffers() function to release them.
978 */
0b8e58a1
AD		 979static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
980 gfp_t wait)
c39a7f84
TO		 981{
982 journal_t *journal = EXT4_SB(sb)->s_journal;
983
984 WARN_ON(PageChecked(page));
985 if (!page_has_buffers(page))
986 return 0;
987 if (journal)
988 return jbd2_journal_try_to_free_buffers(journal, page,
989 wait & ~__GFP_WAIT);
990 return try_to_free_buffers(page);
991}
992
ac27a0ec 993#ifdef CONFIG_QUOTA
af5bc92d 994#define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
2b2d6d01 995#define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
ac27a0ec 996
617ba13b
MC		 997static int ext4_write_dquot(struct dquot *dquot);
998static int ext4_acquire_dquot(struct dquot *dquot);
999static int ext4_release_dquot(struct dquot *dquot);
1000static int ext4_mark_dquot_dirty(struct dquot *dquot);
1001static int ext4_write_info(struct super_block *sb, int type);
6f28e087
JK		 1002static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1003 char *path, int remount);
617ba13b
MC		 1004static int ext4_quota_on_mount(struct super_block *sb, int type);
1005static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 1006 size_t len, loff_t off);
617ba13b 1007static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 1008 const char *data, size_t len, loff_t off);
1009
61e225dc 1010static const struct dquot_operations ext4_quota_operations = {
edf72453
JK		 1011 .initialize = dquot_initialize,
1012 .drop = dquot_drop,
ac27a0ec 1013 .alloc_space = dquot_alloc_space,
60e58e0f
MC		 1014 .reserve_space = dquot_reserve_space,
1015 .claim_space = dquot_claim_space,
1016 .release_rsv = dquot_release_reserved_space,
1017 .get_reserved_space = ext4_get_reserved_space,
ac27a0ec
DK		 1018 .alloc_inode = dquot_alloc_inode,
1019 .free_space = dquot_free_space,
1020 .free_inode = dquot_free_inode,
1021 .transfer = dquot_transfer,
617ba13b
MC		 1022 .write_dquot = ext4_write_dquot,
1023 .acquire_dquot = ext4_acquire_dquot,
1024 .release_dquot = ext4_release_dquot,
1025 .mark_dirty = ext4_mark_dquot_dirty,
a5b5ee32
JK		 1026 .write_info = ext4_write_info,
1027 .alloc_dquot = dquot_alloc,
1028 .destroy_dquot = dquot_destroy,
ac27a0ec
DK		 1029};
1030
0d54b217 1031static const struct quotactl_ops ext4_qctl_operations = {
617ba13b 1032 .quota_on = ext4_quota_on,
ac27a0ec
DK		 1033 .quota_off = vfs_quota_off,
1034 .quota_sync = vfs_quota_sync,
1035 .get_info = vfs_get_dqinfo,
1036 .set_info = vfs_set_dqinfo,
1037 .get_dqblk = vfs_get_dqblk,
1038 .set_dqblk = vfs_set_dqblk
1039};
1040#endif
1041
ee9b6d61 1042static const struct super_operations ext4_sops = {
617ba13b
MC		 1043 .alloc_inode = ext4_alloc_inode,
1044 .destroy_inode = ext4_destroy_inode,
617ba13b
MC		 1045 .write_inode = ext4_write_inode,
1046 .dirty_inode = ext4_dirty_inode,
1047 .delete_inode = ext4_delete_inode,
1048 .put_super = ext4_put_super,
617ba13b 1049 .sync_fs = ext4_sync_fs,
c4be0c1d
TS		 1050 .freeze_fs = ext4_freeze,
1051 .unfreeze_fs = ext4_unfreeze,
617ba13b
MC		 1052 .statfs = ext4_statfs,
1053 .remount_fs = ext4_remount,
1054 .clear_inode = ext4_clear_inode,
1055 .show_options = ext4_show_options,
ac27a0ec 1056#ifdef CONFIG_QUOTA
617ba13b
MC		 1057 .quota_read = ext4_quota_read,
1058 .quota_write = ext4_quota_write,
ac27a0ec 1059#endif
c39a7f84 1060 .bdev_try_to_free_page = bdev_try_to_free_page,
ac27a0ec
DK		 1061};
1062
9ca92389
TT		 1063static const struct super_operations ext4_nojournal_sops = {
1064 .alloc_inode = ext4_alloc_inode,
1065 .destroy_inode = ext4_destroy_inode,
1066 .write_inode = ext4_write_inode,
1067 .dirty_inode = ext4_dirty_inode,
1068 .delete_inode = ext4_delete_inode,
1069 .write_super = ext4_write_super,
1070 .put_super = ext4_put_super,
1071 .statfs = ext4_statfs,
1072 .remount_fs = ext4_remount,
1073 .clear_inode = ext4_clear_inode,
1074 .show_options = ext4_show_options,
1075#ifdef CONFIG_QUOTA
1076 .quota_read = ext4_quota_read,
1077 .quota_write = ext4_quota_write,
1078#endif
1079 .bdev_try_to_free_page = bdev_try_to_free_page,
1080};
1081
39655164 1082static const struct export_operations ext4_export_ops = {
1b961ac0
CH		 1083 .fh_to_dentry = ext4_fh_to_dentry,
1084 .fh_to_parent = ext4_fh_to_parent,
617ba13b 1085 .get_parent = ext4_get_parent,
ac27a0ec
DK		 1086};
1087
1088enum {
1089 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1090 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
01436ef2 1091 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
ac27a0ec 1092 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
06705bff 1093 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh,
30773840 1094 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
c3191067 1095 Opt_journal_update, Opt_journal_dev,
818d276c 1096 Opt_journal_checksum, Opt_journal_async_commit,
ac27a0ec 1097 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
296c355c 1098 Opt_data_err_abort, Opt_data_err_ignore,
ac27a0ec 1099 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
5a20bdfc
JK		 1100 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1101 Opt_noquota, Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err,
1102 Opt_resize, Opt_usrquota, Opt_grpquota, Opt_i_version,
01436ef2 1103 Opt_stripe, Opt_delalloc, Opt_nodelalloc,
6fd058f7 1104 Opt_block_validity, Opt_noblock_validity,
5328e635
ES		 1105 Opt_inode_readahead_blks, Opt_journal_ioprio,
1106 Opt_discard, Opt_nodiscard,
ac27a0ec
DK		 1107};
1108
a447c093 1109static const match_table_t tokens = {
ac27a0ec
DK		 1110 {Opt_bsd_df, "bsddf"},
1111 {Opt_minix_df, "minixdf"},
1112 {Opt_grpid, "grpid"},
1113 {Opt_grpid, "bsdgroups"},
1114 {Opt_nogrpid, "nogrpid"},
1115 {Opt_nogrpid, "sysvgroups"},
1116 {Opt_resgid, "resgid=%u"},
1117 {Opt_resuid, "resuid=%u"},
1118 {Opt_sb, "sb=%u"},
1119 {Opt_err_cont, "errors=continue"},
1120 {Opt_err_panic, "errors=panic"},
1121 {Opt_err_ro, "errors=remount-ro"},
1122 {Opt_nouid32, "nouid32"},
ac27a0ec
DK		 1123 {Opt_debug, "debug"},
1124 {Opt_oldalloc, "oldalloc"},
1125 {Opt_orlov, "orlov"},
1126 {Opt_user_xattr, "user_xattr"},
1127 {Opt_nouser_xattr, "nouser_xattr"},
1128 {Opt_acl, "acl"},
1129 {Opt_noacl, "noacl"},
ac27a0ec 1130 {Opt_noload, "noload"},
e3bb52ae 1131 {Opt_noload, "norecovery"},
ac27a0ec
DK		 1132 {Opt_nobh, "nobh"},
1133 {Opt_bh, "bh"},
1134 {Opt_commit, "commit=%u"},
30773840
TT		 1135 {Opt_min_batch_time, "min_batch_time=%u"},
1136 {Opt_max_batch_time, "max_batch_time=%u"},
ac27a0ec 1137 {Opt_journal_update, "journal=update"},
ac27a0ec 1138 {Opt_journal_dev, "journal_dev=%u"},
818d276c
GS		 1139 {Opt_journal_checksum, "journal_checksum"},
1140 {Opt_journal_async_commit, "journal_async_commit"},
ac27a0ec
DK		 1141 {Opt_abort, "abort"},
1142 {Opt_data_journal, "data=journal"},
1143 {Opt_data_ordered, "data=ordered"},
1144 {Opt_data_writeback, "data=writeback"},
5bf5683a
HK		 1145 {Opt_data_err_abort, "data_err=abort"},
1146 {Opt_data_err_ignore, "data_err=ignore"},
ac27a0ec
DK		 1147 {Opt_offusrjquota, "usrjquota="},
1148 {Opt_usrjquota, "usrjquota=%s"},
1149 {Opt_offgrpjquota, "grpjquota="},
1150 {Opt_grpjquota, "grpjquota=%s"},
1151 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1152 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
5a20bdfc 1153 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
ac27a0ec
DK		 1154 {Opt_grpquota, "grpquota"},
1155 {Opt_noquota, "noquota"},
1156 {Opt_quota, "quota"},
1157 {Opt_usrquota, "usrquota"},
1158 {Opt_barrier, "barrier=%u"},
06705bff
TT		 1159 {Opt_barrier, "barrier"},
1160 {Opt_nobarrier, "nobarrier"},
25ec56b5 1161 {Opt_i_version, "i_version"},
c9de560d 1162 {Opt_stripe, "stripe=%u"},
ac27a0ec 1163 {Opt_resize, "resize"},
64769240 1164 {Opt_delalloc, "delalloc"},
dd919b98 1165 {Opt_nodelalloc, "nodelalloc"},
6fd058f7
TT		 1166 {Opt_block_validity, "block_validity"},
1167 {Opt_noblock_validity, "noblock_validity"},
240799cd 1168 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
b3881f74 1169 {Opt_journal_ioprio, "journal_ioprio=%u"},
afd4672d 1170 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
06705bff
TT		 1171 {Opt_auto_da_alloc, "auto_da_alloc"},
1172 {Opt_noauto_da_alloc, "noauto_da_alloc"},
5328e635
ES		 1173 {Opt_discard, "discard"},
1174 {Opt_nodiscard, "nodiscard"},
f3f12faa 1175 {Opt_err, NULL},
ac27a0ec
DK		 1176};
1177
617ba13b 1178static ext4_fsblk_t get_sb_block(void **data)
ac27a0ec 1179{
617ba13b 1180 ext4_fsblk_t sb_block;
ac27a0ec
DK		 1181 char *options = (char *) *data;
1182
1183 if (!options || strncmp(options, "sb=", 3) != 0)
1184 return 1; /* Default location */
0b8e58a1 1185
ac27a0ec 1186 options += 3;
0b8e58a1 1187 /* TODO: use simple_strtoll with >32bit ext4 */
ac27a0ec
DK		 1188 sb_block = simple_strtoul(options, &options, 0);
1189 if (*options && *options != ',') {
4776004f 1190 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
ac27a0ec
DK		 1191 (char *) *data);
1192 return 1;
1193 }
1194 if (*options == ',')
1195 options++;
1196 *data = (void *) options;
0b8e58a1 1197
ac27a0ec
DK		 1198 return sb_block;
1199}
1200
b3881f74
TT		 1201#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1202
2b2d6d01 1203static int parse_options(char *options, struct super_block *sb,
c3191067 1204 unsigned long *journal_devnum,
b3881f74 1205 unsigned int *journal_ioprio,
2b2d6d01 1206 ext4_fsblk_t *n_blocks_count, int is_remount)
ac27a0ec 1207{
617ba13b 1208 struct ext4_sb_info *sbi = EXT4_SB(sb);
2b2d6d01 1209 char *p;
ac27a0ec
DK		 1210 substring_t args[MAX_OPT_ARGS];
1211 int data_opt = 0;
1212 int option;
1213#ifdef CONFIG_QUOTA
dfc5d03f 1214 int qtype, qfmt;
ac27a0ec
DK		 1215 char *qname;
1216#endif
1217
1218 if (!options)
1219 return 1;
1220
2b2d6d01 1221 while ((p = strsep(&options, ",")) != NULL) {
ac27a0ec
DK		 1222 int token;
1223 if (!*p)
1224 continue;
1225
1226 token = match_token(p, tokens, args);
1227 switch (token) {
1228 case Opt_bsd_df:
2b2d6d01 1229 clear_opt(sbi->s_mount_opt, MINIX_DF);
ac27a0ec
DK		 1230 break;
1231 case Opt_minix_df:
2b2d6d01 1232 set_opt(sbi->s_mount_opt, MINIX_DF);
ac27a0ec
DK		 1233 break;
1234 case Opt_grpid:
2b2d6d01 1235 set_opt(sbi->s_mount_opt, GRPID);
ac27a0ec
DK		 1236 break;
1237 case Opt_nogrpid:
2b2d6d01 1238 clear_opt(sbi->s_mount_opt, GRPID);
ac27a0ec
DK		 1239 break;
1240 case Opt_resuid:
1241 if (match_int(&args[0], &option))
1242 return 0;
1243 sbi->s_resuid = option;
1244 break;
1245 case Opt_resgid:
1246 if (match_int(&args[0], &option))
1247 return 0;
1248 sbi->s_resgid = option;
1249 break;
1250 case Opt_sb:
1251 /* handled by get_sb_block() instead of here */
1252 /* *sb_block = match_int(&args[0]); */
1253 break;
1254 case Opt_err_panic:
2b2d6d01
TT		 1255 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1256 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1257 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
ac27a0ec
DK		 1258 break;
1259 case Opt_err_ro:
2b2d6d01
TT		 1260 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1261 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1262 set_opt(sbi->s_mount_opt, ERRORS_RO);
ac27a0ec
DK		 1263 break;
1264 case Opt_err_cont:
2b2d6d01
TT		 1265 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1266 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1267 set_opt(sbi->s_mount_opt, ERRORS_CONT);
ac27a0ec
DK		 1268 break;
1269 case Opt_nouid32:
2b2d6d01 1270 set_opt(sbi->s_mount_opt, NO_UID32);
ac27a0ec 1271 break;
ac27a0ec 1272 case Opt_debug:
2b2d6d01 1273 set_opt(sbi->s_mount_opt, DEBUG);
ac27a0ec
DK		 1274 break;
1275 case Opt_oldalloc:
2b2d6d01 1276 set_opt(sbi->s_mount_opt, OLDALLOC);
ac27a0ec
DK		 1277 break;
1278 case Opt_orlov:
2b2d6d01 1279 clear_opt(sbi->s_mount_opt, OLDALLOC);
ac27a0ec 1280 break;
03010a33 1281#ifdef CONFIG_EXT4_FS_XATTR
ac27a0ec 1282 case Opt_user_xattr:
2b2d6d01 1283 set_opt(sbi->s_mount_opt, XATTR_USER);
ac27a0ec
DK		 1284 break;
1285 case Opt_nouser_xattr:
2b2d6d01 1286 clear_opt(sbi->s_mount_opt, XATTR_USER);
ac27a0ec
DK		 1287 break;
1288#else
1289 case Opt_user_xattr:
1290 case Opt_nouser_xattr:
b31e1552 1291 ext4_msg(sb, KERN_ERR, "(no)user_xattr options not supported");
ac27a0ec
DK		 1292 break;
1293#endif
03010a33 1294#ifdef CONFIG_EXT4_FS_POSIX_ACL
ac27a0ec
DK		 1295 case Opt_acl:
1296 set_opt(sbi->s_mount_opt, POSIX_ACL);
1297 break;
1298 case Opt_noacl:
1299 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1300 break;
1301#else
1302 case Opt_acl:
1303 case Opt_noacl:
b31e1552 1304 ext4_msg(sb, KERN_ERR, "(no)acl options not supported");
ac27a0ec
DK		 1305 break;
1306#endif
ac27a0ec
DK		 1307 case Opt_journal_update:
1308 /* @@@ FIXME */
1309 /* Eventually we will want to be able to create
1310 a journal file here. For now, only allow the
1311 user to specify an existing inode to be the
1312 journal file. */
1313 if (is_remount) {
b31e1552
ES		 1314 ext4_msg(sb, KERN_ERR,
1315 "Cannot specify journal on remount");
ac27a0ec
DK		 1316 return 0;
1317 }
2b2d6d01 1318 set_opt(sbi->s_mount_opt, UPDATE_JOURNAL);
ac27a0ec 1319 break;
ac27a0ec
DK		 1320 case Opt_journal_dev:
1321 if (is_remount) {
b31e1552
ES		 1322 ext4_msg(sb, KERN_ERR,
1323 "Cannot specify journal on remount");
ac27a0ec
DK		 1324 return 0;
1325 }
1326 if (match_int(&args[0], &option))
1327 return 0;
1328 *journal_devnum = option;
1329 break;
818d276c 1330 case Opt_journal_checksum:
d4da6c9c
LT		 1331 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1332 break;
818d276c
GS		 1333 case Opt_journal_async_commit:
1334 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
d4da6c9c 1335 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
818d276c 1336 break;
ac27a0ec 1337 case Opt_noload:
2b2d6d01 1338 set_opt(sbi->s_mount_opt, NOLOAD);
ac27a0ec
DK		 1339 break;
1340 case Opt_commit:
1341 if (match_int(&args[0], &option))
1342 return 0;
1343 if (option < 0)
1344 return 0;
1345 if (option == 0)
cd02ff0b 1346 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
ac27a0ec
DK		 1347 sbi->s_commit_interval = HZ * option;
1348 break;
30773840
TT		 1349 case Opt_max_batch_time:
1350 if (match_int(&args[0], &option))
1351 return 0;
1352 if (option < 0)
1353 return 0;
1354 if (option == 0)
1355 option = EXT4_DEF_MAX_BATCH_TIME;
1356 sbi->s_max_batch_time = option;
1357 break;
1358 case Opt_min_batch_time:
1359 if (match_int(&args[0], &option))
1360 return 0;
1361 if (option < 0)
1362 return 0;
1363 sbi->s_min_batch_time = option;
1364 break;
ac27a0ec 1365 case Opt_data_journal:
617ba13b 1366 data_opt = EXT4_MOUNT_JOURNAL_DATA;
ac27a0ec
DK		 1367 goto datacheck;
1368 case Opt_data_ordered:
617ba13b 1369 data_opt = EXT4_MOUNT_ORDERED_DATA;
ac27a0ec
DK		 1370 goto datacheck;
1371 case Opt_data_writeback:
617ba13b 1372 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
ac27a0ec
DK		 1373 datacheck:
1374 if (is_remount) {
617ba13b 1375 if ((sbi->s_mount_opt & EXT4_MOUNT_DATA_FLAGS)
ac27a0ec 1376 != data_opt) {
b31e1552
ES		 1377 ext4_msg(sb, KERN_ERR,
1378 "Cannot change data mode on remount");
ac27a0ec
DK		 1379 return 0;
1380 }
1381 } else {
617ba13b 1382 sbi->s_mount_opt &= ~EXT4_MOUNT_DATA_FLAGS;
ac27a0ec
DK		 1383 sbi->s_mount_opt |= data_opt;
1384 }
1385 break;
5bf5683a
HK		 1386 case Opt_data_err_abort:
1387 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1388 break;
1389 case Opt_data_err_ignore:
1390 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1391 break;
ac27a0ec
DK		 1392#ifdef CONFIG_QUOTA
1393 case Opt_usrjquota:
1394 qtype = USRQUOTA;
1395 goto set_qf_name;
1396 case Opt_grpjquota:
1397 qtype = GRPQUOTA;
1398set_qf_name:
17bd13b3 1399 if (sb_any_quota_loaded(sb) &&
dfc5d03f 1400 !sbi->s_qf_names[qtype]) {
b31e1552
ES		 1401 ext4_msg(sb, KERN_ERR,
1402 "Cannot change journaled "
1403 "quota options when quota turned on");
ac27a0ec
DK		 1404 return 0;
1405 }
1406 qname = match_strdup(&args[0]);
1407 if (!qname) {
b31e1552
ES		 1408 ext4_msg(sb, KERN_ERR,
1409 "Not enough memory for "
1410 "storing quotafile name");
ac27a0ec
DK		 1411 return 0;
1412 }
1413 if (sbi->s_qf_names[qtype] &&
1414 strcmp(sbi->s_qf_names[qtype], qname)) {
b31e1552
ES		 1415 ext4_msg(sb, KERN_ERR,
1416 "%s quota file already "
1417 "specified", QTYPE2NAME(qtype));
ac27a0ec
DK		 1418 kfree(qname);
1419 return 0;
1420 }
1421 sbi->s_qf_names[qtype] = qname;
1422 if (strchr(sbi->s_qf_names[qtype], '/')) {
b31e1552
ES		 1423 ext4_msg(sb, KERN_ERR,
1424 "quotafile must be on "
1425 "filesystem root");
ac27a0ec
DK		 1426 kfree(sbi->s_qf_names[qtype]);
1427 sbi->s_qf_names[qtype] = NULL;
1428 return 0;
1429 }
1430 set_opt(sbi->s_mount_opt, QUOTA);
1431 break;
1432 case Opt_offusrjquota:
1433 qtype = USRQUOTA;
1434 goto clear_qf_name;
1435 case Opt_offgrpjquota:
1436 qtype = GRPQUOTA;
1437clear_qf_name:
17bd13b3 1438 if (sb_any_quota_loaded(sb) &&
dfc5d03f 1439 sbi->s_qf_names[qtype]) {
b31e1552 1440 ext4_msg(sb, KERN_ERR, "Cannot change "
2c8be6b2 1441 "journaled quota options when "
b31e1552 1442 "quota turned on");
ac27a0ec
DK		 1443 return 0;
1444 }
1445 /*
1446 * The space will be released later when all options
1447 * are confirmed to be correct
1448 */
1449 sbi->s_qf_names[qtype] = NULL;
1450 break;
1451 case Opt_jqfmt_vfsold:
dfc5d03f
JK		 1452 qfmt = QFMT_VFS_OLD;
1453 goto set_qf_format;
ac27a0ec 1454 case Opt_jqfmt_vfsv0:
dfc5d03f 1455 qfmt = QFMT_VFS_V0;
5a20bdfc
JK		 1456 goto set_qf_format;
1457 case Opt_jqfmt_vfsv1:
1458 qfmt = QFMT_VFS_V1;
dfc5d03f 1459set_qf_format:
17bd13b3 1460 if (sb_any_quota_loaded(sb) &&
dfc5d03f 1461 sbi->s_jquota_fmt != qfmt) {
b31e1552 1462 ext4_msg(sb, KERN_ERR, "Cannot change "
dfc5d03f 1463 "journaled quota options when "
b31e1552 1464 "quota turned on");
dfc5d03f
JK		 1465 return 0;
1466 }
1467 sbi->s_jquota_fmt = qfmt;
ac27a0ec
DK		 1468 break;
1469 case Opt_quota:
1470 case Opt_usrquota:
1471 set_opt(sbi->s_mount_opt, QUOTA);
1472 set_opt(sbi->s_mount_opt, USRQUOTA);
1473 break;
1474 case Opt_grpquota:
1475 set_opt(sbi->s_mount_opt, QUOTA);
1476 set_opt(sbi->s_mount_opt, GRPQUOTA);
1477 break;
1478 case Opt_noquota:
17bd13b3 1479 if (sb_any_quota_loaded(sb)) {
b31e1552
ES		 1480 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1481 "options when quota turned on");
ac27a0ec
DK		 1482 return 0;
1483 }
1484 clear_opt(sbi->s_mount_opt, QUOTA);
1485 clear_opt(sbi->s_mount_opt, USRQUOTA);
1486 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1487 break;
1488#else
1489 case Opt_quota:
1490 case Opt_usrquota:
1491 case Opt_grpquota:
b31e1552
ES		 1492 ext4_msg(sb, KERN_ERR,
1493 "quota options not supported");
cd59e7b9 1494 break;
ac27a0ec
DK		 1495 case Opt_usrjquota:
1496 case Opt_grpjquota:
1497 case Opt_offusrjquota:
1498 case Opt_offgrpjquota:
1499 case Opt_jqfmt_vfsold:
1500 case Opt_jqfmt_vfsv0:
5a20bdfc 1501 case Opt_jqfmt_vfsv1:
b31e1552
ES		 1502 ext4_msg(sb, KERN_ERR,
1503 "journaled quota options not supported");
ac27a0ec
DK		 1504 break;
1505 case Opt_noquota:
1506 break;
1507#endif
1508 case Opt_abort:
4ab2f15b 1509 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
ac27a0ec 1510 break;
06705bff
TT		 1511 case Opt_nobarrier:
1512 clear_opt(sbi->s_mount_opt, BARRIER);
1513 break;
ac27a0ec 1514 case Opt_barrier:
06705bff
TT		 1515 if (match_int(&args[0], &option)) {
1516 set_opt(sbi->s_mount_opt, BARRIER);
1517 break;
1518 }
ac27a0ec
DK		 1519 if (option)
1520 set_opt(sbi->s_mount_opt, BARRIER);
1521 else
1522 clear_opt(sbi->s_mount_opt, BARRIER);
1523 break;
1524 case Opt_ignore:
1525 break;
1526 case Opt_resize:
1527 if (!is_remount) {
b31e1552
ES		 1528 ext4_msg(sb, KERN_ERR,
1529 "resize option only available "
1530 "for remount");
ac27a0ec
DK		 1531 return 0;
1532 }
1533 if (match_int(&args[0], &option) != 0)
1534 return 0;
1535 *n_blocks_count = option;
1536 break;
1537 case Opt_nobh:
1538 set_opt(sbi->s_mount_opt, NOBH);
1539 break;
1540 case Opt_bh:
1541 clear_opt(sbi->s_mount_opt, NOBH);
1542 break;
25ec56b5
JNC		 1543 case Opt_i_version:
1544 set_opt(sbi->s_mount_opt, I_VERSION);
1545 sb->s_flags |= MS_I_VERSION;
1546 break;
dd919b98
AK		 1547 case Opt_nodelalloc:
1548 clear_opt(sbi->s_mount_opt, DELALLOC);
1549 break;
c9de560d
AT		 1550 case Opt_stripe:
1551 if (match_int(&args[0], &option))
1552 return 0;
1553 if (option < 0)
1554 return 0;
1555 sbi->s_stripe = option;
1556 break;
64769240
AT		 1557 case Opt_delalloc:
1558 set_opt(sbi->s_mount_opt, DELALLOC);
1559 break;
6fd058f7
TT		 1560 case Opt_block_validity:
1561 set_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1562 break;
1563 case Opt_noblock_validity:
1564 clear_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1565 break;
240799cd
TT		 1566 case Opt_inode_readahead_blks:
1567 if (match_int(&args[0], &option))
1568 return 0;
1569 if (option < 0 || option > (1 << 30))
1570 return 0;
f7c43950 1571 if (!is_power_of_2(option)) {
b31e1552
ES		 1572 ext4_msg(sb, KERN_ERR,
1573 "EXT4-fs: inode_readahead_blks"
1574 " must be a power of 2");
3197ebdb
TT		 1575 return 0;
1576 }
240799cd
TT		 1577 sbi->s_inode_readahead_blks = option;
1578 break;
b3881f74
TT		 1579 case Opt_journal_ioprio:
1580 if (match_int(&args[0], &option))
1581 return 0;
1582 if (option < 0 || option > 7)
1583 break;
1584 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1585 option);
1586 break;
06705bff
TT		 1587 case Opt_noauto_da_alloc:
1588 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1589 break;
afd4672d 1590 case Opt_auto_da_alloc:
06705bff
TT		 1591 if (match_int(&args[0], &option)) {
1592 clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1593 break;
1594 }
afd4672d
TT		 1595 if (option)
1596 clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1597 else
1598 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1599 break;
5328e635
ES		 1600 case Opt_discard:
1601 set_opt(sbi->s_mount_opt, DISCARD);
1602 break;
1603 case Opt_nodiscard:
1604 clear_opt(sbi->s_mount_opt, DISCARD);
1605 break;
ac27a0ec 1606 default:
b31e1552
ES		 1607 ext4_msg(sb, KERN_ERR,
1608 "Unrecognized mount option \"%s\" "
1609 "or missing value", p);
ac27a0ec
DK		 1610 return 0;
1611 }
1612 }
1613#ifdef CONFIG_QUOTA
1614 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
617ba13b 1615 if ((sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA) &&
ac27a0ec
DK		 1616 sbi->s_qf_names[USRQUOTA])
1617 clear_opt(sbi->s_mount_opt, USRQUOTA);
1618
617ba13b 1619 if ((sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA) &&
ac27a0ec
DK		 1620 sbi->s_qf_names[GRPQUOTA])
1621 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1622
1623 if ((sbi->s_qf_names[USRQUOTA] &&
617ba13b 1624 (sbi->s_mount_opt & EXT4_MOUNT_GRPQUOTA)) ||
ac27a0ec 1625 (sbi->s_qf_names[GRPQUOTA] &&
617ba13b 1626 (sbi->s_mount_opt & EXT4_MOUNT_USRQUOTA))) {
b31e1552
ES		 1627 ext4_msg(sb, KERN_ERR, "old and new quota "
1628 "format mixing");
ac27a0ec
DK		 1629 return 0;
1630 }
1631
1632 if (!sbi->s_jquota_fmt) {
b31e1552
ES		 1633 ext4_msg(sb, KERN_ERR, "journaled quota format "
1634 "not specified");
ac27a0ec
DK		 1635 return 0;
1636 }
1637 } else {
1638 if (sbi->s_jquota_fmt) {
b31e1552 1639 ext4_msg(sb, KERN_ERR, "journaled quota format "
2c8be6b2 1640 "specified with no journaling "
b31e1552 1641 "enabled");
ac27a0ec
DK		 1642 return 0;
1643 }
1644 }
1645#endif
1646 return 1;
1647}
1648
617ba13b 1649static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 1650 int read_only)
1651{
617ba13b 1652 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 1653 int res = 0;
1654
617ba13b 1655 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
b31e1552
ES		 1656 ext4_msg(sb, KERN_ERR, "revision level too high, "
1657 "forcing read-only mode");
ac27a0ec
DK		 1658 res = MS_RDONLY;
1659 }
1660 if (read_only)
1661 return res;
617ba13b 1662 if (!(sbi->s_mount_state & EXT4_VALID_FS))
b31e1552
ES		 1663 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1664 "running e2fsck is recommended");
617ba13b 1665 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
b31e1552
ES		 1666 ext4_msg(sb, KERN_WARNING,
1667 "warning: mounting fs with errors, "
1668 "running e2fsck is recommended");
ac27a0ec
DK		 1669 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1670 le16_to_cpu(es->s_mnt_count) >=
1671 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
b31e1552
ES		 1672 ext4_msg(sb, KERN_WARNING,
1673 "warning: maximal mount count reached, "
1674 "running e2fsck is recommended");
ac27a0ec
DK		 1675 else if (le32_to_cpu(es->s_checkinterval) &&
1676 (le32_to_cpu(es->s_lastcheck) +
1677 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
b31e1552
ES		 1678 ext4_msg(sb, KERN_WARNING,
1679 "warning: checktime reached, "
1680 "running e2fsck is recommended");
0b8e58a1 1681 if (!sbi->s_journal)
0390131b 1682 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec 1683 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 1684 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 1685 le16_add_cpu(&es->s_mnt_count, 1);
ac27a0ec 1686 es->s_mtime = cpu_to_le32(get_seconds());
617ba13b 1687 ext4_update_dynamic_rev(sb);
0390131b
FM		 1688 if (sbi->s_journal)
1689 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 1690
e2d67052 1691 ext4_commit_super(sb, 1);
ac27a0ec 1692 if (test_opt(sb, DEBUG))
a9df9a49 1693 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
7f4520cc 1694 "bpg=%lu, ipg=%lu, mo=%04x]\n",
ac27a0ec
DK		 1695 sb->s_blocksize,
1696 sbi->s_groups_count,
617ba13b
MC		 1697 EXT4_BLOCKS_PER_GROUP(sb),
1698 EXT4_INODES_PER_GROUP(sb),
ac27a0ec
DK		 1699 sbi->s_mount_opt);
1700
ac27a0ec
DK		 1701 return res;
1702}
1703
772cb7c8
JS		 1704static int ext4_fill_flex_info(struct super_block *sb)
1705{
1706 struct ext4_sb_info *sbi = EXT4_SB(sb);
1707 struct ext4_group_desc *gdp = NULL;
772cb7c8
JS		 1708 ext4_group_t flex_group_count;
1709 ext4_group_t flex_group;
1710 int groups_per_flex = 0;
c5ca7c76 1711 size_t size;
772cb7c8
JS		 1712 int i;
1713
503358ae
TT		 1714 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1715 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1716
1717 if (groups_per_flex < 2) {
772cb7c8
JS		 1718 sbi->s_log_groups_per_flex = 0;
1719 return 1;
1720 }
1721
c62a11fd
FB		 1722 /* We allocate both existing and potentially added groups */
1723 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
d94e99a6
AK		 1724 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1725 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
c5ca7c76
TT		 1726 size = flex_group_count * sizeof(struct flex_groups);
1727 sbi->s_flex_groups = kzalloc(size, GFP_KERNEL);
1728 if (sbi->s_flex_groups == NULL) {
1729 sbi->s_flex_groups = vmalloc(size);
1730 if (sbi->s_flex_groups)
1731 memset(sbi->s_flex_groups, 0, size);
1732 }
772cb7c8 1733 if (sbi->s_flex_groups == NULL) {
b31e1552
ES		 1734 ext4_msg(sb, KERN_ERR, "not enough memory for "
1735 "%u flex groups", flex_group_count);
772cb7c8
JS		 1736 goto failed;
1737 }
772cb7c8 1738
772cb7c8 1739 for (i = 0; i < sbi->s_groups_count; i++) {
88b6edd1 1740 gdp = ext4_get_group_desc(sb, i, NULL);
772cb7c8
JS		 1741
1742 flex_group = ext4_flex_group(sbi, i);
7ad9bb65
TT		 1743 atomic_add(ext4_free_inodes_count(sb, gdp),
1744 &sbi->s_flex_groups[flex_group].free_inodes);
1745 atomic_add(ext4_free_blks_count(sb, gdp),
1746 &sbi->s_flex_groups[flex_group].free_blocks);
1747 atomic_add(ext4_used_dirs_count(sb, gdp),
1748 &sbi->s_flex_groups[flex_group].used_dirs);
772cb7c8
JS		 1749 }
1750
1751 return 1;
1752failed:
1753 return 0;
1754}
1755
717d50e4
AD		 1756__le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1757 struct ext4_group_desc *gdp)
1758{
1759 __u16 crc = 0;
1760
1761 if (sbi->s_es->s_feature_ro_compat &
1762 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1763 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1764 __le32 le_group = cpu_to_le32(block_group);
1765
1766 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1767 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1768 crc = crc16(crc, (__u8 *)gdp, offset);
1769 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1770 /* for checksum of struct ext4_group_desc do the rest...*/
1771 if ((sbi->s_es->s_feature_incompat &
1772 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1773 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1774 crc = crc16(crc, (__u8 *)gdp + offset,
1775 le16_to_cpu(sbi->s_es->s_desc_size) -
1776 offset);
1777 }
1778
1779 return cpu_to_le16(crc);
1780}
1781
1782int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1783 struct ext4_group_desc *gdp)
1784{
1785 if ((sbi->s_es->s_feature_ro_compat &
1786 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1787 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1788 return 0;
1789
1790 return 1;
1791}
1792
ac27a0ec 1793/* Called at mount-time, super-block is locked */
197cd65a 1794static int ext4_check_descriptors(struct super_block *sb)
ac27a0ec 1795{
617ba13b
MC		 1796 struct ext4_sb_info *sbi = EXT4_SB(sb);
1797 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1798 ext4_fsblk_t last_block;
bd81d8ee
LV		 1799 ext4_fsblk_t block_bitmap;
1800 ext4_fsblk_t inode_bitmap;
1801 ext4_fsblk_t inode_table;
ce421581 1802 int flexbg_flag = 0;
fd2d4291 1803 ext4_group_t i;
ac27a0ec 1804
ce421581
JS		 1805 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1806 flexbg_flag = 1;
1807
af5bc92d 1808 ext4_debug("Checking group descriptors");
ac27a0ec 1809
197cd65a
AM		 1810 for (i = 0; i < sbi->s_groups_count; i++) {
1811 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1812
ce421581 1813 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 1814 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 1815 else
1816 last_block = first_block +
617ba13b 1817 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 1818
8fadc143 1819 block_bitmap = ext4_block_bitmap(sb, gdp);
2b2d6d01 1820 if (block_bitmap < first_block || block_bitmap > last_block) {
b31e1552 1821 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 1822 "Block bitmap for group %u not in group "
b31e1552 1823 "(block %llu)!", i, block_bitmap);
ac27a0ec
DK		 1824 return 0;
1825 }
8fadc143 1826 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2b2d6d01 1827 if (inode_bitmap < first_block || inode_bitmap > last_block) {
b31e1552 1828 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 1829 "Inode bitmap for group %u not in group "
b31e1552 1830 "(block %llu)!", i, inode_bitmap);
ac27a0ec
DK		 1831 return 0;
1832 }
8fadc143 1833 inode_table = ext4_inode_table(sb, gdp);
bd81d8ee 1834 if (inode_table < first_block ||
2b2d6d01 1835 inode_table + sbi->s_itb_per_group - 1 > last_block) {
b31e1552 1836 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 1837 "Inode table for group %u not in group "
b31e1552 1838 "(block %llu)!", i, inode_table);
ac27a0ec
DK		 1839 return 0;
1840 }
955ce5f5 1841 ext4_lock_group(sb, i);
717d50e4 1842 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
b31e1552
ES		 1843 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1844 "Checksum for group %u failed (%u!=%u)",
1845 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1846 gdp)), le16_to_cpu(gdp->bg_checksum));
7ee1ec4c 1847 if (!(sb->s_flags & MS_RDONLY)) {
955ce5f5 1848 ext4_unlock_group(sb, i);
8a266467 1849 return 0;
7ee1ec4c 1850 }
717d50e4 1851 }
955ce5f5 1852 ext4_unlock_group(sb, i);
ce421581
JS		 1853 if (!flexbg_flag)
1854 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec
DK		 1855 }
1856
bd81d8ee 1857 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
0b8e58a1 1858 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
ac27a0ec
DK		 1859 return 1;
1860}
1861
617ba13b 1862/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 1863 * the superblock) which were deleted from all directories, but held open by
1864 * a process at the time of a crash. We walk the list and try to delete these
1865 * inodes at recovery time (only with a read-write filesystem).
1866 *
1867 * In order to keep the orphan inode chain consistent during traversal (in
1868 * case of crash during recovery), we link each inode into the superblock
1869 * orphan list_head and handle it the same way as an inode deletion during
1870 * normal operation (which journals the operations for us).
1871 *
1872 * We only do an iget() and an iput() on each inode, which is very safe if we
1873 * accidentally point at an in-use or already deleted inode. The worst that
1874 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 1875 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 1876 * e2fsck was run on this filesystem, and it must have already done the orphan
1877 * inode cleanup for us, so we can safely abort without any further action.
1878 */
2b2d6d01
TT		 1879static void ext4_orphan_cleanup(struct super_block *sb,
1880 struct ext4_super_block *es)
ac27a0ec
DK		 1881{
1882 unsigned int s_flags = sb->s_flags;
1883 int nr_orphans = 0, nr_truncates = 0;
1884#ifdef CONFIG_QUOTA
1885 int i;
1886#endif
1887 if (!es->s_last_orphan) {
1888 jbd_debug(4, "no orphan inodes to clean up\n");
1889 return;
1890 }
1891
a8f48a95 1892 if (bdev_read_only(sb->s_bdev)) {
b31e1552
ES		 1893 ext4_msg(sb, KERN_ERR, "write access "
1894 "unavailable, skipping orphan cleanup");
a8f48a95
ES		 1895 return;
1896 }
1897
617ba13b 1898 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
ac27a0ec
DK		 1899 if (es->s_last_orphan)
1900 jbd_debug(1, "Errors on filesystem, "
1901 "clearing orphan list.\n");
1902 es->s_last_orphan = 0;
1903 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1904 return;
1905 }
1906
1907 if (s_flags & MS_RDONLY) {
b31e1552 1908 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
ac27a0ec
DK		 1909 sb->s_flags &= ~MS_RDONLY;
1910 }
1911#ifdef CONFIG_QUOTA
1912 /* Needed for iput() to work correctly and not trash data */
1913 sb->s_flags |= MS_ACTIVE;
1914 /* Turn on quotas so that they are updated correctly */
1915 for (i = 0; i < MAXQUOTAS; i++) {
617ba13b
MC		 1916 if (EXT4_SB(sb)->s_qf_names[i]) {
1917 int ret = ext4_quota_on_mount(sb, i);
ac27a0ec 1918 if (ret < 0)
b31e1552
ES		 1919 ext4_msg(sb, KERN_ERR,
1920 "Cannot turn on journaled "
1921 "quota: error %d", ret);
ac27a0ec
DK		 1922 }
1923 }
1924#endif
1925
1926 while (es->s_last_orphan) {
1927 struct inode *inode;
1928
97bd42b9
JB		 1929 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
1930 if (IS_ERR(inode)) {
ac27a0ec
DK		 1931 es->s_last_orphan = 0;
1932 break;
1933 }
1934
617ba13b 1935 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
a269eb18 1936 vfs_dq_init(inode);
ac27a0ec 1937 if (inode->i_nlink) {
b31e1552
ES		 1938 ext4_msg(sb, KERN_DEBUG,
1939 "%s: truncating inode %lu to %lld bytes",
46e665e9 1940 __func__, inode->i_ino, inode->i_size);
e5f8eab8 1941 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
ac27a0ec 1942 inode->i_ino, inode->i_size);
617ba13b 1943 ext4_truncate(inode);
ac27a0ec
DK		 1944 nr_truncates++;
1945 } else {
b31e1552
ES		 1946 ext4_msg(sb, KERN_DEBUG,
1947 "%s: deleting unreferenced inode %lu",
46e665e9 1948 __func__, inode->i_ino);
ac27a0ec
DK		 1949 jbd_debug(2, "deleting unreferenced inode %lu\n",
1950 inode->i_ino);
1951 nr_orphans++;
1952 }
1953 iput(inode); /* The delete magic happens here! */
1954 }
1955
2b2d6d01 1956#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
ac27a0ec
DK		 1957
1958 if (nr_orphans)
b31e1552
ES		 1959 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
1960 PLURAL(nr_orphans));
ac27a0ec 1961 if (nr_truncates)
b31e1552
ES		 1962 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
1963 PLURAL(nr_truncates));
ac27a0ec
DK		 1964#ifdef CONFIG_QUOTA
1965 /* Turn quotas off */
1966 for (i = 0; i < MAXQUOTAS; i++) {
1967 if (sb_dqopt(sb)->files[i])
6f28e087 1968 vfs_quota_off(sb, i, 0);
ac27a0ec
DK		 1969 }
1970#endif
1971 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
1972}
0b8e58a1 1973
cd2291a4
ES		 1974/*
1975 * Maximal extent format file size.
1976 * Resulting logical blkno at s_maxbytes must fit in our on-disk
1977 * extent format containers, within a sector_t, and within i_blocks
1978 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
1979 * so that won't be a limiting factor.
1980 *
1981 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
1982 */
f287a1a5 1983static loff_t ext4_max_size(int blkbits, int has_huge_files)
cd2291a4
ES		 1984{
1985 loff_t res;
1986 loff_t upper_limit = MAX_LFS_FILESIZE;
1987
1988 /* small i_blocks in vfs inode? */
f287a1a5 1989 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
cd2291a4 1990 /*
90c699a9 1991 * CONFIG_LBDAF is not enabled implies the inode
cd2291a4
ES		 1992 * i_block represent total blocks in 512 bytes
1993 * 32 == size of vfs inode i_blocks * 8
1994 */
1995 upper_limit = (1LL << 32) - 1;
1996
1997 /* total blocks in file system block size */
1998 upper_limit >>= (blkbits - 9);
1999 upper_limit <<= blkbits;
2000 }
2001
2002 /* 32-bit extent-start container, ee_block */
2003 res = 1LL << 32;
2004 res <<= blkbits;
2005 res -= 1;
2006
2007 /* Sanity check against vm- & vfs- imposed limits */
2008 if (res > upper_limit)
2009 res = upper_limit;
2010
2011 return res;
2012}
ac27a0ec 2013
ac27a0ec 2014/*
cd2291a4 2015 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 2016 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2017 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 2018 */
f287a1a5 2019static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
ac27a0ec 2020{
617ba13b 2021 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 2022 int meta_blocks;
2023 loff_t upper_limit;
0b8e58a1
AD		 2024 /* This is calculated to be the largest file size for a dense, block
2025 * mapped file such that the file's total number of 512-byte sectors,
2026 * including data and all indirect blocks, does not exceed (2^48 - 1).
2027 *
2028 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2029 * number of 512-byte sectors of the file.
0fc1b451
AK		 2030 */
2031
f287a1a5 2032 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
0fc1b451 2033 /*
90c699a9 2034 * !has_huge_files or CONFIG_LBDAF not enabled implies that
0b8e58a1
AD		 2035 * the inode i_block field represents total file blocks in
2036 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
0fc1b451
AK		 2037 */
2038 upper_limit = (1LL << 32) - 1;
2039
2040 /* total blocks in file system block size */
2041 upper_limit >>= (bits - 9);
2042
2043 } else {
8180a562
AK		 2044 /*
2045 * We use 48 bit ext4_inode i_blocks
2046 * With EXT4_HUGE_FILE_FL set the i_blocks
2047 * represent total number of blocks in
2048 * file system block size
2049 */
0fc1b451
AK		 2050 upper_limit = (1LL << 48) - 1;
2051
0fc1b451
AK		 2052 }
2053
2054 /* indirect blocks */
2055 meta_blocks = 1;
2056 /* double indirect blocks */
2057 meta_blocks += 1 + (1LL << (bits-2));
2058 /* tripple indirect blocks */
2059 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2060
2061 upper_limit -= meta_blocks;
2062 upper_limit <<= bits;
ac27a0ec
DK		 2063
2064 res += 1LL << (bits-2);
2065 res += 1LL << (2*(bits-2));
2066 res += 1LL << (3*(bits-2));
2067 res <<= bits;
2068 if (res > upper_limit)
2069 res = upper_limit;
0fc1b451
AK		 2070
2071 if (res > MAX_LFS_FILESIZE)
2072 res = MAX_LFS_FILESIZE;
2073
ac27a0ec
DK		 2074 return res;
2075}
2076
617ba13b 2077static ext4_fsblk_t descriptor_loc(struct super_block *sb,
0b8e58a1 2078 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 2079{
617ba13b 2080 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 2081 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 2082 int has_super = 0;
2083
2084 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2085
617ba13b 2086 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
ac27a0ec 2087 nr < first_meta_bg)
70bbb3e0 2088 return logical_sb_block + nr + 1;
ac27a0ec 2089 bg = sbi->s_desc_per_block * nr;
617ba13b 2090 if (ext4_bg_has_super(sb, bg))
ac27a0ec 2091 has_super = 1;
0b8e58a1 2092
617ba13b 2093 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 2094}
2095
c9de560d
AT		 2096/**
2097 * ext4_get_stripe_size: Get the stripe size.
2098 * @sbi: In memory super block info
2099 *
2100 * If we have specified it via mount option, then
2101 * use the mount option value. If the value specified at mount time is
2102 * greater than the blocks per group use the super block value.
2103 * If the super block value is greater than blocks per group return 0.
2104 * Allocator needs it be less than blocks per group.
2105 *
2106 */
2107static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2108{
2109 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2110 unsigned long stripe_width =
2111 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2112
2113 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2114 return sbi->s_stripe;
2115
2116 if (stripe_width <= sbi->s_blocks_per_group)
2117 return stripe_width;
2118
2119 if (stride <= sbi->s_blocks_per_group)
2120 return stride;
2121
2122 return 0;
2123}
ac27a0ec 2124
3197ebdb
TT		 2125/* sysfs supprt */
2126
2127struct ext4_attr {
2128 struct attribute attr;
2129 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2130 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2131 const char *, size_t);
2132 int offset;
2133};
2134
2135static int parse_strtoul(const char *buf,
2136 unsigned long max, unsigned long *value)
2137{
2138 char *endp;
2139
e7d2860b
AGR		 2140 *value = simple_strtoul(skip_spaces(buf), &endp, 0);
2141 endp = skip_spaces(endp);
3197ebdb
TT		 2142 if (*endp || *value > max)
2143 return -EINVAL;
2144
2145 return 0;
2146}
2147
2148static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2149 struct ext4_sb_info *sbi,
2150 char *buf)
2151{
2152 return snprintf(buf, PAGE_SIZE, "%llu\n",
2153 (s64) percpu_counter_sum(&sbi->s_dirtyblocks_counter));
2154}
2155
2156static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2157 struct ext4_sb_info *sbi, char *buf)
2158{
2159 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2160
2161 return snprintf(buf, PAGE_SIZE, "%lu\n",
2162 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2163 sbi->s_sectors_written_start) >> 1);
2164}
2165
2166static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2167 struct ext4_sb_info *sbi, char *buf)
2168{
2169 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2170
2171 return snprintf(buf, PAGE_SIZE, "%llu\n",
2172 sbi->s_kbytes_written +
2173 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2174 EXT4_SB(sb)->s_sectors_written_start) >> 1));
2175}
2176
2177static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2178 struct ext4_sb_info *sbi,
2179 const char *buf, size_t count)
2180{
2181 unsigned long t;
2182
2183 if (parse_strtoul(buf, 0x40000000, &t))
2184 return -EINVAL;
2185
f7c43950 2186 if (!is_power_of_2(t))
3197ebdb
TT		 2187 return -EINVAL;
2188
2189 sbi->s_inode_readahead_blks = t;
2190 return count;
2191}
2192
2193static ssize_t sbi_ui_show(struct ext4_attr *a,
0b8e58a1 2194 struct ext4_sb_info *sbi, char *buf)
3197ebdb
TT		 2195{
2196 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2197
2198 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2199}
2200
2201static ssize_t sbi_ui_store(struct ext4_attr *a,
2202 struct ext4_sb_info *sbi,
2203 const char *buf, size_t count)
2204{
2205 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2206 unsigned long t;
2207
2208 if (parse_strtoul(buf, 0xffffffff, &t))
2209 return -EINVAL;
2210 *ui = t;
2211 return count;
2212}
2213
2214#define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2215static struct ext4_attr ext4_attr_##_name = { \
2216 .attr = {.name = __stringify(_name), .mode = _mode }, \
2217 .show = _show, \
2218 .store = _store, \
2219 .offset = offsetof(struct ext4_sb_info, _elname), \
2220}
2221#define EXT4_ATTR(name, mode, show, store) \
2222static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2223
2224#define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2225#define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2226#define EXT4_RW_ATTR_SBI_UI(name, elname) \
2227 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2228#define ATTR_LIST(name) &ext4_attr_##name.attr
2229
2230EXT4_RO_ATTR(delayed_allocation_blocks);
2231EXT4_RO_ATTR(session_write_kbytes);
2232EXT4_RO_ATTR(lifetime_write_kbytes);
2233EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2234 inode_readahead_blks_store, s_inode_readahead_blks);
11013911 2235EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
3197ebdb
TT		 2236EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2237EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2238EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2239EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2240EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2241EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
55138e0b 2242EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
3197ebdb
TT		 2243
2244static struct attribute *ext4_attrs[] = {
2245 ATTR_LIST(delayed_allocation_blocks),
2246 ATTR_LIST(session_write_kbytes),
2247 ATTR_LIST(lifetime_write_kbytes),
2248 ATTR_LIST(inode_readahead_blks),
11013911 2249 ATTR_LIST(inode_goal),
3197ebdb
TT		 2250 ATTR_LIST(mb_stats),
2251 ATTR_LIST(mb_max_to_scan),
2252 ATTR_LIST(mb_min_to_scan),
2253 ATTR_LIST(mb_order2_req),
2254 ATTR_LIST(mb_stream_req),
2255 ATTR_LIST(mb_group_prealloc),
55138e0b 2256 ATTR_LIST(max_writeback_mb_bump),
3197ebdb
TT		 2257 NULL,
2258};
2259
2260static ssize_t ext4_attr_show(struct kobject *kobj,
2261 struct attribute *attr, char *buf)
2262{
2263 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2264 s_kobj);
2265 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2266
2267 return a->show ? a->show(a, sbi, buf) : 0;
2268}
2269
2270static ssize_t ext4_attr_store(struct kobject *kobj,
2271 struct attribute *attr,
2272 const char *buf, size_t len)
2273{
2274 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2275 s_kobj);
2276 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2277
2278 return a->store ? a->store(a, sbi, buf, len) : 0;
2279}
2280
2281static void ext4_sb_release(struct kobject *kobj)
2282{
2283 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2284 s_kobj);
2285 complete(&sbi->s_kobj_unregister);
2286}
2287
2288
2289static struct sysfs_ops ext4_attr_ops = {
2290 .show = ext4_attr_show,
2291 .store = ext4_attr_store,
2292};
2293
2294static struct kobj_type ext4_ktype = {
2295 .default_attrs = ext4_attrs,
2296 .sysfs_ops = &ext4_attr_ops,
2297 .release = ext4_sb_release,
2298};
2299
a13fb1a4
ES		 2300/*
2301 * Check whether this filesystem can be mounted based on
2302 * the features present and the RDONLY/RDWR mount requested.
2303 * Returns 1 if this filesystem can be mounted as requested,
2304 * 0 if it cannot be.
2305 */
2306static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2307{
2308 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2309 ext4_msg(sb, KERN_ERR,
2310 "Couldn't mount because of "
2311 "unsupported optional features (%x)",
2312 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2313 ~EXT4_FEATURE_INCOMPAT_SUPP));
2314 return 0;
2315 }
2316
2317 if (readonly)
2318 return 1;
2319
2320 /* Check that feature set is OK for a read-write mount */
2321 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2322 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2323 "unsupported optional features (%x)",
2324 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2325 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2326 return 0;
2327 }
2328 /*
2329 * Large file size enabled file system can only be mounted
2330 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2331 */
2332 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2333 if (sizeof(blkcnt_t) < sizeof(u64)) {
2334 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2335 "cannot be mounted RDWR without "
2336 "CONFIG_LBDAF");
2337 return 0;
2338 }
2339 }
2340 return 1;
2341}
2342
2b2d6d01 2343static int ext4_fill_super(struct super_block *sb, void *data, int silent)
7477827f
AK		 2344 __releases(kernel_lock)
2345 __acquires(kernel_lock)
ac27a0ec 2346{
2b2d6d01 2347 struct buffer_head *bh;
617ba13b
MC		 2348 struct ext4_super_block *es = NULL;
2349 struct ext4_sb_info *sbi;
2350 ext4_fsblk_t block;
2351 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 2352 ext4_fsblk_t logical_sb_block;
ac27a0ec 2353 unsigned long offset = 0;
ac27a0ec
DK		 2354 unsigned long journal_devnum = 0;
2355 unsigned long def_mount_opts;
2356 struct inode *root;
9f6200bb 2357 char *cp;
0390131b 2358 const char *descr;
1d1fe1ee 2359 int ret = -EINVAL;
ac27a0ec 2360 int blocksize;
4ec11028
TT		 2361 unsigned int db_count;
2362 unsigned int i;
f287a1a5 2363 int needs_recovery, has_huge_files;
bd81d8ee 2364 __u64 blocks_count;
833f4077 2365 int err;
b3881f74 2366 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
ac27a0ec
DK		 2367
2368 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
2369 if (!sbi)
2370 return -ENOMEM;
705895b6
PE		 2371
2372 sbi->s_blockgroup_lock =
2373 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
2374 if (!sbi->s_blockgroup_lock) {
2375 kfree(sbi);
2376 return -ENOMEM;
2377 }
ac27a0ec
DK		 2378 sb->s_fs_info = sbi;
2379 sbi->s_mount_opt = 0;
617ba13b
MC		 2380 sbi->s_resuid = EXT4_DEF_RESUID;
2381 sbi->s_resgid = EXT4_DEF_RESGID;
240799cd 2382 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
d9c9bef1 2383 sbi->s_sb_block = sb_block;
afc32f7e
TT		 2384 sbi->s_sectors_written_start = part_stat_read(sb->s_bdev->bd_part,
2385 sectors[1]);
ac27a0ec
DK		 2386
2387 unlock_kernel();
2388
9f6200bb
TT		 2389 /* Cleanup superblock name */
2390 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
2391 *cp = '!';
2392
617ba13b 2393 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 2394 if (!blocksize) {
b31e1552 2395 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
ac27a0ec
DK		 2396 goto out_fail;
2397 }
2398
2399 /*
617ba13b 2400 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 2401 * block sizes. We need to calculate the offset from buffer start.
2402 */
617ba13b 2403 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 2404 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2405 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 2406 } else {
70bbb3e0 2407 logical_sb_block = sb_block;
ac27a0ec
DK		 2408 }
2409
70bbb3e0 2410 if (!(bh = sb_bread(sb, logical_sb_block))) {
b31e1552 2411 ext4_msg(sb, KERN_ERR, "unable to read superblock");
ac27a0ec
DK		 2412 goto out_fail;
2413 }
2414 /*
2415 * Note: s_es must be initialized as soon as possible because
617ba13b 2416 * some ext4 macro-instructions depend on its value
ac27a0ec 2417 */
617ba13b 2418 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
ac27a0ec
DK		 2419 sbi->s_es = es;
2420 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 2421 if (sb->s_magic != EXT4_SUPER_MAGIC)
2422 goto cantfind_ext4;
afc32f7e 2423 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
ac27a0ec
DK		 2424
2425 /* Set defaults before we parse the mount options */
2426 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
617ba13b 2427 if (def_mount_opts & EXT4_DEFM_DEBUG)
ac27a0ec 2428 set_opt(sbi->s_mount_opt, DEBUG);
617ba13b 2429 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
ac27a0ec 2430 set_opt(sbi->s_mount_opt, GRPID);
617ba13b 2431 if (def_mount_opts & EXT4_DEFM_UID16)
ac27a0ec 2432 set_opt(sbi->s_mount_opt, NO_UID32);
03010a33 2433#ifdef CONFIG_EXT4_FS_XATTR
617ba13b 2434 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
ac27a0ec 2435 set_opt(sbi->s_mount_opt, XATTR_USER);
2e7842b8 2436#endif
03010a33 2437#ifdef CONFIG_EXT4_FS_POSIX_ACL
617ba13b 2438 if (def_mount_opts & EXT4_DEFM_ACL)
ac27a0ec 2439 set_opt(sbi->s_mount_opt, POSIX_ACL);
2e7842b8 2440#endif
617ba13b
MC		 2441 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
2442 sbi->s_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
2443 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
2444 sbi->s_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
2445 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
2446 sbi->s_mount_opt |= EXT4_MOUNT_WRITEBACK_DATA;
2447
2448 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
ac27a0ec 2449 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
bb4f397a 2450 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
ceea16bf 2451 set_opt(sbi->s_mount_opt, ERRORS_CONT);
bb4f397a
AK		 2452 else
2453 set_opt(sbi->s_mount_opt, ERRORS_RO);
ac27a0ec
DK		 2454
2455 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
2456 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
30773840
TT		 2457 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
2458 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
2459 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
ac27a0ec 2460
571640ca 2461 set_opt(sbi->s_mount_opt, BARRIER);
ac27a0ec 2462
dd919b98
AK		 2463 /*
2464 * enable delayed allocation by default
2465 * Use -o nodelalloc to turn it off
2466 */
2467 set_opt(sbi->s_mount_opt, DELALLOC);
2468
b3881f74
TT		 2469 if (!parse_options((char *) data, sb, &journal_devnum,
2470 &journal_ioprio, NULL, 0))
ac27a0ec
DK		 2471 goto failed_mount;
2472
2473 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
617ba13b 2474 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec 2475
617ba13b
MC		 2476 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
2477 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
2478 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
2479 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
b31e1552
ES		 2480 ext4_msg(sb, KERN_WARNING,
2481 "feature flags set on rev 0 fs, "
2482 "running e2fsck is recommended");
469108ff 2483
ac27a0ec
DK		 2484 /*
2485 * Check feature flags regardless of the revision level, since we
2486 * previously didn't change the revision level when setting the flags,
2487 * so there is a chance incompat flags are set on a rev 0 filesystem.
2488 */
a13fb1a4 2489 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
ac27a0ec 2490 goto failed_mount;
a13fb1a4 2491
ac27a0ec
DK		 2492 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
2493
617ba13b
MC		 2494 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
2495 blocksize > EXT4_MAX_BLOCK_SIZE) {
b31e1552
ES		 2496 ext4_msg(sb, KERN_ERR,
2497 "Unsupported filesystem blocksize %d", blocksize);
ac27a0ec
DK		 2498 goto failed_mount;
2499 }
2500
ac27a0ec 2501 if (sb->s_blocksize != blocksize) {
ce40733c
AK		 2502 /* Validate the filesystem blocksize */
2503 if (!sb_set_blocksize(sb, blocksize)) {
b31e1552 2504 ext4_msg(sb, KERN_ERR, "bad block size %d",
ce40733c 2505 blocksize);
ac27a0ec
DK		 2506 goto failed_mount;
2507 }
2508
2b2d6d01 2509 brelse(bh);
70bbb3e0
AM		 2510 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2511 offset = do_div(logical_sb_block, blocksize);
2512 bh = sb_bread(sb, logical_sb_block);
ac27a0ec 2513 if (!bh) {
b31e1552
ES		 2514 ext4_msg(sb, KERN_ERR,
2515 "Can't read superblock on 2nd try");
ac27a0ec
DK		 2516 goto failed_mount;
2517 }
617ba13b 2518 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
ac27a0ec 2519 sbi->s_es = es;
617ba13b 2520 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
b31e1552
ES		 2521 ext4_msg(sb, KERN_ERR,
2522 "Magic mismatch, very weird!");
ac27a0ec
DK		 2523 goto failed_mount;
2524 }
2525 }
2526
a13fb1a4
ES		 2527 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2528 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
f287a1a5
TT		 2529 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
2530 has_huge_files);
2531 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
ac27a0ec 2532
617ba13b
MC		 2533 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2534 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2535 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
ac27a0ec
DK		 2536 } else {
2537 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2538 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
617ba13b 2539 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1330593e 2540 (!is_power_of_2(sbi->s_inode_size)) ||
ac27a0ec 2541 (sbi->s_inode_size > blocksize)) {
b31e1552
ES		 2542 ext4_msg(sb, KERN_ERR,
2543 "unsupported inode size: %d",
2b2d6d01 2544 sbi->s_inode_size);
ac27a0ec
DK		 2545 goto failed_mount;
2546 }
ef7f3835
KS		 2547 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2548 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
ac27a0ec 2549 }
0b8e58a1 2550
0d1ee42f
AR		 2551 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2552 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
8fadc143 2553 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 2554 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 2555 !is_power_of_2(sbi->s_desc_size)) {
b31e1552
ES		 2556 ext4_msg(sb, KERN_ERR,
2557 "unsupported descriptor size %lu",
0d1ee42f
AR		 2558 sbi->s_desc_size);
2559 goto failed_mount;
2560 }
2561 } else
2562 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
0b8e58a1 2563
ac27a0ec 2564 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 2565 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
b47b6f38 2566 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
617ba13b 2567 goto cantfind_ext4;
0b8e58a1 2568
617ba13b 2569 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 2570 if (sbi->s_inodes_per_block == 0)
617ba13b 2571 goto cantfind_ext4;
ac27a0ec
DK		 2572 sbi->s_itb_per_group = sbi->s_inodes_per_group /
2573 sbi->s_inodes_per_block;
0d1ee42f 2574 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 2575 sbi->s_sbh = bh;
2576 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 2577 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2578 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
0b8e58a1 2579
2b2d6d01 2580 for (i = 0; i < 4; i++)
ac27a0ec
DK		 2581 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2582 sbi->s_def_hash_version = es->s_def_hash_version;
f99b2589
TT		 2583 i = le32_to_cpu(es->s_flags);
2584 if (i & EXT2_FLAGS_UNSIGNED_HASH)
2585 sbi->s_hash_unsigned = 3;
2586 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
2587#ifdef __CHAR_UNSIGNED__
2588 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
2589 sbi->s_hash_unsigned = 3;
2590#else
2591 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
2592#endif
2593 sb->s_dirt = 1;
2594 }
ac27a0ec
DK		 2595
2596 if (sbi->s_blocks_per_group > blocksize * 8) {
b31e1552
ES		 2597 ext4_msg(sb, KERN_ERR,
2598 "#blocks per group too big: %lu",
2b2d6d01 2599 sbi->s_blocks_per_group);
ac27a0ec
DK		 2600 goto failed_mount;
2601 }
ac27a0ec 2602 if (sbi->s_inodes_per_group > blocksize * 8) {
b31e1552
ES		 2603 ext4_msg(sb, KERN_ERR,
2604 "#inodes per group too big: %lu",
2b2d6d01 2605 sbi->s_inodes_per_group);
ac27a0ec
DK		 2606 goto failed_mount;
2607 }
2608
bf43d84b
ES		 2609 /*
2610 * Test whether we have more sectors than will fit in sector_t,
2611 * and whether the max offset is addressable by the page cache.
2612 */
2613 if ((ext4_blocks_count(es) >
2614 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) ||
2615 (ext4_blocks_count(es) >
2616 (pgoff_t)(~0ULL) >> (PAGE_CACHE_SHIFT - sb->s_blocksize_bits))) {
b31e1552 2617 ext4_msg(sb, KERN_ERR, "filesystem"
bf43d84b 2618 " too large to mount safely on this system");
ac27a0ec 2619 if (sizeof(sector_t) < 8)
90c699a9 2620 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
bf43d84b 2621 ret = -EFBIG;
ac27a0ec
DK		 2622 goto failed_mount;
2623 }
2624
617ba13b
MC		 2625 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2626 goto cantfind_ext4;
e7c95593 2627
0f2ddca6
FTN		 2628 /* check blocks count against device size */
2629 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
2630 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
b31e1552
ES		 2631 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
2632 "exceeds size of device (%llu blocks)",
0f2ddca6
FTN		 2633 ext4_blocks_count(es), blocks_count);
2634 goto failed_mount;
2635 }
2636
0b8e58a1
AD		 2637 /*
2638 * It makes no sense for the first data block to be beyond the end
2639 * of the filesystem.
2640 */
2641 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
b31e1552
ES		 2642 ext4_msg(sb, KERN_WARNING, "bad geometry: first data"
2643 "block %u is beyond end of filesystem (%llu)",
2644 le32_to_cpu(es->s_first_data_block),
2645 ext4_blocks_count(es));
e7c95593
ES		 2646 goto failed_mount;
2647 }
bd81d8ee
LV		 2648 blocks_count = (ext4_blocks_count(es) -
2649 le32_to_cpu(es->s_first_data_block) +
2650 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2651 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4ec11028 2652 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
b31e1552 2653 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
4ec11028 2654 "(block count %llu, first data block %u, "
b31e1552 2655 "blocks per group %lu)", sbi->s_groups_count,
4ec11028
TT		 2656 ext4_blocks_count(es),
2657 le32_to_cpu(es->s_first_data_block),
2658 EXT4_BLOCKS_PER_GROUP(sb));
2659 goto failed_mount;
2660 }
bd81d8ee 2661 sbi->s_groups_count = blocks_count;
fb0a387d
ES		 2662 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
2663 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
617ba13b
MC		 2664 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2665 EXT4_DESC_PER_BLOCK(sb);
2b2d6d01 2666 sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
ac27a0ec
DK		 2667 GFP_KERNEL);
2668 if (sbi->s_group_desc == NULL) {
b31e1552 2669 ext4_msg(sb, KERN_ERR, "not enough memory");
ac27a0ec
DK		 2670 goto failed_mount;
2671 }
2672
3244fcb1 2673#ifdef CONFIG_PROC_FS
9f6200bb
TT		 2674 if (ext4_proc_root)
2675 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
3244fcb1 2676#endif
240799cd 2677
705895b6 2678 bgl_lock_init(sbi->s_blockgroup_lock);
ac27a0ec
DK		 2679
2680 for (i = 0; i < db_count; i++) {
70bbb3e0 2681 block = descriptor_loc(sb, logical_sb_block, i);
ac27a0ec
DK		 2682 sbi->s_group_desc[i] = sb_bread(sb, block);
2683 if (!sbi->s_group_desc[i]) {
b31e1552
ES		 2684 ext4_msg(sb, KERN_ERR,
2685 "can't read group descriptor %d", i);
ac27a0ec
DK		 2686 db_count = i;
2687 goto failed_mount2;
2688 }
2689 }
2b2d6d01 2690 if (!ext4_check_descriptors(sb)) {
b31e1552 2691 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
ac27a0ec
DK		 2692 goto failed_mount2;
2693 }
772cb7c8
JS		 2694 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2695 if (!ext4_fill_flex_info(sb)) {
b31e1552
ES		 2696 ext4_msg(sb, KERN_ERR,
2697 "unable to initialize "
2698 "flex_bg meta info!");
772cb7c8
JS		 2699 goto failed_mount2;
2700 }
2701
ac27a0ec
DK		 2702 sbi->s_gdb_count = db_count;
2703 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2704 spin_lock_init(&sbi->s_next_gen_lock);
2705
833f4077
PZ		 2706 err = percpu_counter_init(&sbi->s_freeblocks_counter,
2707 ext4_count_free_blocks(sb));
2708 if (!err) {
2709 err = percpu_counter_init(&sbi->s_freeinodes_counter,
2710 ext4_count_free_inodes(sb));
2711 }
2712 if (!err) {
2713 err = percpu_counter_init(&sbi->s_dirs_counter,
2714 ext4_count_dirs(sb));
2715 }
6bc6e63f
AK		 2716 if (!err) {
2717 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
2718 }
833f4077 2719 if (err) {
b31e1552 2720 ext4_msg(sb, KERN_ERR, "insufficient memory");
833f4077
PZ		 2721 goto failed_mount3;
2722 }
ac27a0ec 2723
c9de560d 2724 sbi->s_stripe = ext4_get_stripe_size(sbi);
55138e0b 2725 sbi->s_max_writeback_mb_bump = 128;
c9de560d 2726
ac27a0ec
DK		 2727 /*
2728 * set up enough so that it can read an inode
2729 */
9ca92389
TT		 2730 if (!test_opt(sb, NOLOAD) &&
2731 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
2732 sb->s_op = &ext4_sops;
2733 else
2734 sb->s_op = &ext4_nojournal_sops;
617ba13b
MC		 2735 sb->s_export_op = &ext4_export_ops;
2736 sb->s_xattr = ext4_xattr_handlers;
ac27a0ec 2737#ifdef CONFIG_QUOTA
617ba13b
MC		 2738 sb->s_qcop = &ext4_qctl_operations;
2739 sb->dq_op = &ext4_quota_operations;
ac27a0ec
DK		 2740#endif
2741 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3b9d4ed2 2742 mutex_init(&sbi->s_orphan_lock);
32ed5058 2743 mutex_init(&sbi->s_resize_lock);
ac27a0ec
DK		 2744
2745 sb->s_root = NULL;
2746
2747 needs_recovery = (es->s_last_orphan != 0 ||
617ba13b
MC		 2748 EXT4_HAS_INCOMPAT_FEATURE(sb,
2749 EXT4_FEATURE_INCOMPAT_RECOVER));
ac27a0ec
DK		 2750
2751 /*
2752 * The first inode we look at is the journal inode. Don't try
2753 * root first: it may be modified in the journal!
2754 */
2755 if (!test_opt(sb, NOLOAD) &&
617ba13b
MC		 2756 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2757 if (ext4_load_journal(sb, es, journal_devnum))
ac27a0ec 2758 goto failed_mount3;
0390131b
FM		 2759 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
2760 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
b31e1552
ES		 2761 ext4_msg(sb, KERN_ERR, "required journal recovery "
2762 "suppressed and not mounted read-only");
0390131b 2763 goto failed_mount4;
ac27a0ec 2764 } else {
0390131b
FM		 2765 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
2766 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
2767 sbi->s_journal = NULL;
2768 needs_recovery = 0;
2769 goto no_journal;
ac27a0ec
DK		 2770 }
2771
eb40a09c
JS		 2772 if (ext4_blocks_count(es) > 0xffffffffULL &&
2773 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2774 JBD2_FEATURE_INCOMPAT_64BIT)) {
b31e1552 2775 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
eb40a09c
JS		 2776 goto failed_mount4;
2777 }
2778
d4da6c9c
LT		 2779 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2780 jbd2_journal_set_features(sbi->s_journal,
2781 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
818d276c 2782 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
d4da6c9c
LT		 2783 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2784 jbd2_journal_set_features(sbi->s_journal,
2785 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
818d276c
GS		 2786 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2787 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
d4da6c9c
LT		 2788 } else {
2789 jbd2_journal_clear_features(sbi->s_journal,
2790 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2791 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2792 }
818d276c 2793
ac27a0ec
DK		 2794 /* We have now updated the journal if required, so we can
2795 * validate the data journaling mode. */
2796 switch (test_opt(sb, DATA_FLAGS)) {
2797 case 0:
2798 /* No mode set, assume a default based on the journal
63f57933
AM		 2799 * capabilities: ORDERED_DATA if the journal can
2800 * cope, else JOURNAL_DATA
2801 */
dab291af
MC		 2802 if (jbd2_journal_check_available_features
2803 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
ac27a0ec
DK		 2804 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2805 else
2806 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2807 break;
2808
617ba13b
MC		 2809 case EXT4_MOUNT_ORDERED_DATA:
2810 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 2811 if (!jbd2_journal_check_available_features
2812 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
b31e1552
ES		 2813 ext4_msg(sb, KERN_ERR, "Journal does not support "
2814 "requested data journaling mode");
ac27a0ec
DK		 2815 goto failed_mount4;
2816 }
2817 default:
2818 break;
2819 }
b3881f74 2820 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
ac27a0ec 2821
0390131b 2822no_journal:
ac27a0ec
DK		 2823
2824 if (test_opt(sb, NOBH)) {
617ba13b 2825 if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
b31e1552
ES		 2826 ext4_msg(sb, KERN_WARNING, "Ignoring nobh option - "
2827 "its supported only with writeback mode");
ac27a0ec
DK		 2828 clear_opt(sbi->s_mount_opt, NOBH);
2829 }
2830 }
4c0425ff
MC		 2831 EXT4_SB(sb)->dio_unwritten_wq = create_workqueue("ext4-dio-unwritten");
2832 if (!EXT4_SB(sb)->dio_unwritten_wq) {
2833 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
2834 goto failed_mount_wq;
2835 }
2836
ac27a0ec 2837 /*
dab291af 2838 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 2839 * so we can safely mount the rest of the filesystem now.
2840 */
2841
1d1fe1ee
DH		 2842 root = ext4_iget(sb, EXT4_ROOT_INO);
2843 if (IS_ERR(root)) {
b31e1552 2844 ext4_msg(sb, KERN_ERR, "get root inode failed");
1d1fe1ee 2845 ret = PTR_ERR(root);
ac27a0ec
DK		 2846 goto failed_mount4;
2847 }
2848 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1d1fe1ee 2849 iput(root);
b31e1552 2850 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
ac27a0ec
DK		 2851 goto failed_mount4;
2852 }
1d1fe1ee
DH		 2853 sb->s_root = d_alloc_root(root);
2854 if (!sb->s_root) {
b31e1552 2855 ext4_msg(sb, KERN_ERR, "get root dentry failed");
1d1fe1ee
DH		 2856 iput(root);
2857 ret = -ENOMEM;
2858 goto failed_mount4;
2859 }
ac27a0ec 2860
2b2d6d01 2861 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
ef7f3835
KS		 2862
2863 /* determine the minimum size of new large inodes, if present */
2864 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2865 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2866 EXT4_GOOD_OLD_INODE_SIZE;
2867 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2868 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2869 if (sbi->s_want_extra_isize <
2870 le16_to_cpu(es->s_want_extra_isize))
2871 sbi->s_want_extra_isize =
2872 le16_to_cpu(es->s_want_extra_isize);
2873 if (sbi->s_want_extra_isize <
2874 le16_to_cpu(es->s_min_extra_isize))
2875 sbi->s_want_extra_isize =
2876 le16_to_cpu(es->s_min_extra_isize);
2877 }
2878 }
2879 /* Check if enough inode space is available */
2880 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2881 sbi->s_inode_size) {
2882 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2883 EXT4_GOOD_OLD_INODE_SIZE;
b31e1552
ES		 2884 ext4_msg(sb, KERN_INFO, "required extra inode space not"
2885 "available");
ef7f3835
KS		 2886 }
2887
90576c0b
TT		 2888 if (test_opt(sb, DELALLOC) &&
2889 (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)) {
b31e1552
ES		 2890 ext4_msg(sb, KERN_WARNING, "Ignoring delalloc option - "
2891 "requested data journaling mode");
c2774d84 2892 clear_opt(sbi->s_mount_opt, DELALLOC);
90576c0b 2893 }
c2774d84 2894
6fd058f7
TT		 2895 err = ext4_setup_system_zone(sb);
2896 if (err) {
b31e1552
ES		 2897 ext4_msg(sb, KERN_ERR, "failed to initialize system "
2898 "zone (%d)\n", err);
6fd058f7
TT		 2899 goto failed_mount4;
2900 }
2901
c2774d84
AK		 2902 ext4_ext_init(sb);
2903 err = ext4_mb_init(sb, needs_recovery);
2904 if (err) {
b31e1552
ES		 2905 ext4_msg(sb, KERN_ERR, "failed to initalize mballoc (%d)",
2906 err);
c2774d84
AK		 2907 goto failed_mount4;
2908 }
2909
3197ebdb
TT		 2910 sbi->s_kobj.kset = ext4_kset;
2911 init_completion(&sbi->s_kobj_unregister);
2912 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
2913 "%s", sb->s_id);
2914 if (err) {
2915 ext4_mb_release(sb);
2916 ext4_ext_release(sb);
2917 goto failed_mount4;
2918 };
2919
617ba13b
MC		 2920 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
2921 ext4_orphan_cleanup(sb, es);
2922 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
0390131b 2923 if (needs_recovery) {
b31e1552 2924 ext4_msg(sb, KERN_INFO, "recovery complete");
0390131b
FM		 2925 ext4_mark_recovery_complete(sb, es);
2926 }
2927 if (EXT4_SB(sb)->s_journal) {
2928 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2929 descr = " journalled data mode";
2930 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2931 descr = " ordered data mode";
2932 else
2933 descr = " writeback data mode";
2934 } else
2935 descr = "out journal";
2936
b31e1552 2937 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s", descr);
ac27a0ec
DK		 2938
2939 lock_kernel();
2940 return 0;
2941
617ba13b 2942cantfind_ext4:
ac27a0ec 2943 if (!silent)
b31e1552 2944 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
ac27a0ec
DK		 2945 goto failed_mount;
2946
2947failed_mount4:
b31e1552 2948 ext4_msg(sb, KERN_ERR, "mount failed");
4c0425ff
MC		 2949 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
2950failed_mount_wq:
6fd058f7 2951 ext4_release_system_zone(sb);
0390131b
FM		 2952 if (sbi->s_journal) {
2953 jbd2_journal_destroy(sbi->s_journal);
2954 sbi->s_journal = NULL;
2955 }
ac27a0ec 2956failed_mount3:
c5ca7c76
TT		 2957 if (sbi->s_flex_groups) {
2958 if (is_vmalloc_addr(sbi->s_flex_groups))
2959 vfree(sbi->s_flex_groups);
2960 else
2961 kfree(sbi->s_flex_groups);
2962 }
ac27a0ec
DK		 2963 percpu_counter_destroy(&sbi->s_freeblocks_counter);
2964 percpu_counter_destroy(&sbi->s_freeinodes_counter);
2965 percpu_counter_destroy(&sbi->s_dirs_counter);
6bc6e63f 2966 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
ac27a0ec
DK		 2967failed_mount2:
2968 for (i = 0; i < db_count; i++)
2969 brelse(sbi->s_group_desc[i]);
2970 kfree(sbi->s_group_desc);
2971failed_mount:
240799cd 2972 if (sbi->s_proc) {
9f6200bb 2973 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 2974 }
ac27a0ec
DK		 2975#ifdef CONFIG_QUOTA
2976 for (i = 0; i < MAXQUOTAS; i++)
2977 kfree(sbi->s_qf_names[i]);
2978#endif
617ba13b 2979 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 2980 brelse(bh);
2981out_fail:
2982 sb->s_fs_info = NULL;
f6830165 2983 kfree(sbi->s_blockgroup_lock);
ac27a0ec
DK		 2984 kfree(sbi);
2985 lock_kernel();
1d1fe1ee 2986 return ret;
ac27a0ec
DK		 2987}
2988
2989/*
2990 * Setup any per-fs journal parameters now. We'll do this both on
2991 * initial mount, once the journal has been initialised but before we've
2992 * done any recovery; and again on any subsequent remount.
2993 */
617ba13b 2994static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 2995{
617ba13b 2996 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 2997
30773840
TT		 2998 journal->j_commit_interval = sbi->s_commit_interval;
2999 journal->j_min_batch_time = sbi->s_min_batch_time;
3000 journal->j_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 3001
3002 spin_lock(&journal->j_state_lock);
3003 if (test_opt(sb, BARRIER))
dab291af 3004 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 3005 else
dab291af 3006 journal->j_flags &= ~JBD2_BARRIER;
5bf5683a
HK		 3007 if (test_opt(sb, DATA_ERR_ABORT))
3008 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
3009 else
3010 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
ac27a0ec
DK		 3011 spin_unlock(&journal->j_state_lock);
3012}
3013
617ba13b 3014static journal_t *ext4_get_journal(struct super_block *sb,
ac27a0ec
DK		 3015 unsigned int journal_inum)
3016{
3017 struct inode *journal_inode;
3018 journal_t *journal;
3019
0390131b
FM		 3020 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3021
ac27a0ec
DK		 3022 /* First, test for the existence of a valid inode on disk. Bad
3023 * things happen if we iget() an unused inode, as the subsequent
3024 * iput() will try to delete it. */
3025
1d1fe1ee
DH		 3026 journal_inode = ext4_iget(sb, journal_inum);
3027 if (IS_ERR(journal_inode)) {
b31e1552 3028 ext4_msg(sb, KERN_ERR, "no journal found");
ac27a0ec
DK		 3029 return NULL;
3030 }
3031 if (!journal_inode->i_nlink) {
3032 make_bad_inode(journal_inode);
3033 iput(journal_inode);
b31e1552 3034 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
ac27a0ec
DK		 3035 return NULL;
3036 }
3037
e5f8eab8 3038 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
ac27a0ec 3039 journal_inode, journal_inode->i_size);
1d1fe1ee 3040 if (!S_ISREG(journal_inode->i_mode)) {
b31e1552 3041 ext4_msg(sb, KERN_ERR, "invalid journal inode");
ac27a0ec
DK		 3042 iput(journal_inode);
3043 return NULL;
3044 }
3045
dab291af 3046 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 3047 if (!journal) {
b31e1552 3048 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
ac27a0ec
DK		 3049 iput(journal_inode);
3050 return NULL;
3051 }
3052 journal->j_private = sb;
617ba13b 3053 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 3054 return journal;
3055}
3056
617ba13b 3057static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 3058 dev_t j_dev)
3059{
2b2d6d01 3060 struct buffer_head *bh;
ac27a0ec 3061 journal_t *journal;
617ba13b
MC		 3062 ext4_fsblk_t start;
3063 ext4_fsblk_t len;
ac27a0ec 3064 int hblock, blocksize;
617ba13b 3065 ext4_fsblk_t sb_block;
ac27a0ec 3066 unsigned long offset;
2b2d6d01 3067 struct ext4_super_block *es;
ac27a0ec
DK		 3068 struct block_device *bdev;
3069
0390131b
FM		 3070 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3071
b31e1552 3072 bdev = ext4_blkdev_get(j_dev, sb);
ac27a0ec
DK		 3073 if (bdev == NULL)
3074 return NULL;
3075
3076 if (bd_claim(bdev, sb)) {
b31e1552
ES		 3077 ext4_msg(sb, KERN_ERR,
3078 "failed to claim external journal device");
9a1c3542 3079 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
ac27a0ec
DK		 3080 return NULL;
3081 }
3082
3083 blocksize = sb->s_blocksize;
e1defc4f 3084 hblock = bdev_logical_block_size(bdev);
ac27a0ec 3085 if (blocksize < hblock) {
b31e1552
ES		 3086 ext4_msg(sb, KERN_ERR,
3087 "blocksize too small for journal device");
ac27a0ec
DK		 3088 goto out_bdev;
3089 }
3090
617ba13b
MC		 3091 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3092 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 3093 set_blocksize(bdev, blocksize);
3094 if (!(bh = __bread(bdev, sb_block, blocksize))) {
b31e1552
ES		 3095 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
3096 "external journal");
ac27a0ec
DK		 3097 goto out_bdev;
3098 }
3099
617ba13b
MC		 3100 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3101 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 3102 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b 3103 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
b31e1552
ES		 3104 ext4_msg(sb, KERN_ERR, "external journal has "
3105 "bad superblock");
ac27a0ec
DK		 3106 brelse(bh);
3107 goto out_bdev;
3108 }
3109
617ba13b 3110 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
b31e1552 3111 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
ac27a0ec
DK		 3112 brelse(bh);
3113 goto out_bdev;
3114 }
3115
bd81d8ee 3116 len = ext4_blocks_count(es);
ac27a0ec
DK		 3117 start = sb_block + 1;
3118 brelse(bh); /* we're done with the superblock */
3119
dab291af 3120 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 3121 start, len, blocksize);
3122 if (!journal) {
b31e1552 3123 ext4_msg(sb, KERN_ERR, "failed to create device journal");
ac27a0ec
DK		 3124 goto out_bdev;
3125 }
3126 journal->j_private = sb;
3127 ll_rw_block(READ, 1, &journal->j_sb_buffer);
3128 wait_on_buffer(journal->j_sb_buffer);
3129 if (!buffer_uptodate(journal->j_sb_buffer)) {
b31e1552 3130 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
ac27a0ec
DK		 3131 goto out_journal;
3132 }
3133 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
b31e1552
ES		 3134 ext4_msg(sb, KERN_ERR, "External journal has more than one "
3135 "user (unsupported) - %d",
ac27a0ec
DK		 3136 be32_to_cpu(journal->j_superblock->s_nr_users));
3137 goto out_journal;
3138 }
617ba13b
MC		 3139 EXT4_SB(sb)->journal_bdev = bdev;
3140 ext4_init_journal_params(sb, journal);
ac27a0ec 3141 return journal;
0b8e58a1 3142
ac27a0ec 3143out_journal:
dab291af 3144 jbd2_journal_destroy(journal);
ac27a0ec 3145out_bdev:
617ba13b 3146 ext4_blkdev_put(bdev);
ac27a0ec
DK		 3147 return NULL;
3148}
3149
617ba13b
MC		 3150static int ext4_load_journal(struct super_block *sb,
3151 struct ext4_super_block *es,
ac27a0ec
DK		 3152 unsigned long journal_devnum)
3153{
3154 journal_t *journal;
3155 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
3156 dev_t journal_dev;
3157 int err = 0;
3158 int really_read_only;
3159
0390131b
FM		 3160 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3161
ac27a0ec
DK		 3162 if (journal_devnum &&
3163 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
b31e1552
ES		 3164 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
3165 "numbers have changed");
ac27a0ec
DK		 3166 journal_dev = new_decode_dev(journal_devnum);
3167 } else
3168 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
3169
3170 really_read_only = bdev_read_only(sb->s_bdev);
3171
3172 /*
3173 * Are we loading a blank journal or performing recovery after a
3174 * crash? For recovery, we need to check in advance whether we
3175 * can get read-write access to the device.
3176 */
617ba13b 3177 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
ac27a0ec 3178 if (sb->s_flags & MS_RDONLY) {
b31e1552
ES		 3179 ext4_msg(sb, KERN_INFO, "INFO: recovery "
3180 "required on readonly filesystem");
ac27a0ec 3181 if (really_read_only) {
b31e1552
ES		 3182 ext4_msg(sb, KERN_ERR, "write access "
3183 "unavailable, cannot proceed");
ac27a0ec
DK		 3184 return -EROFS;
3185 }
b31e1552
ES		 3186 ext4_msg(sb, KERN_INFO, "write access will "
3187 "be enabled during recovery");
ac27a0ec
DK		 3188 }
3189 }
3190
3191 if (journal_inum && journal_dev) {
b31e1552
ES		 3192 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
3193 "and inode journals!");
ac27a0ec
DK		 3194 return -EINVAL;
3195 }
3196
3197 if (journal_inum) {
617ba13b 3198 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 3199 return -EINVAL;
3200 } else {
617ba13b 3201 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 3202 return -EINVAL;
3203 }
3204
90576c0b 3205 if (!(journal->j_flags & JBD2_BARRIER))
b31e1552 3206 ext4_msg(sb, KERN_INFO, "barriers disabled");
4776004f 3207
ac27a0ec 3208 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
dab291af 3209 err = jbd2_journal_update_format(journal);
ac27a0ec 3210 if (err) {
b31e1552 3211 ext4_msg(sb, KERN_ERR, "error updating journal");
dab291af 3212 jbd2_journal_destroy(journal);
ac27a0ec
DK		 3213 return err;
3214 }
3215 }
3216
617ba13b 3217 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
dab291af 3218 err = jbd2_journal_wipe(journal, !really_read_only);
ac27a0ec 3219 if (!err)
dab291af 3220 err = jbd2_journal_load(journal);
ac27a0ec
DK		 3221
3222 if (err) {
b31e1552 3223 ext4_msg(sb, KERN_ERR, "error loading journal");
dab291af 3224 jbd2_journal_destroy(journal);
ac27a0ec
DK		 3225 return err;
3226 }
3227
617ba13b
MC		 3228 EXT4_SB(sb)->s_journal = journal;
3229 ext4_clear_journal_err(sb, es);
ac27a0ec
DK		 3230
3231 if (journal_devnum &&
3232 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3233 es->s_journal_dev = cpu_to_le32(journal_devnum);
ac27a0ec
DK		 3234
3235 /* Make sure we flush the recovery flag to disk. */
e2d67052 3236 ext4_commit_super(sb, 1);
ac27a0ec
DK		 3237 }
3238
3239 return 0;
3240}
3241
e2d67052 3242static int ext4_commit_super(struct super_block *sb, int sync)
ac27a0ec 3243{
e2d67052 3244 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
617ba13b 3245 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
c4be0c1d 3246 int error = 0;
ac27a0ec
DK		 3247
3248 if (!sbh)
c4be0c1d 3249 return error;
914258bf
TT		 3250 if (buffer_write_io_error(sbh)) {
3251 /*
3252 * Oh, dear. A previous attempt to write the
3253 * superblock failed. This could happen because the
3254 * USB device was yanked out. Or it could happen to
3255 * be a transient write error and maybe the block will
3256 * be remapped. Nothing we can do but to retry the
3257 * write and hope for the best.
3258 */
b31e1552
ES		 3259 ext4_msg(sb, KERN_ERR, "previous I/O error to "
3260 "superblock detected");
914258bf
TT		 3261 clear_buffer_write_io_error(sbh);
3262 set_buffer_uptodate(sbh);
3263 }
71290b36
TT		 3264 /*
3265 * If the file system is mounted read-only, don't update the
3266 * superblock write time. This avoids updating the superblock
3267 * write time when we are mounting the root file system
3268 * read/only but we need to replay the journal; at that point,
3269 * for people who are east of GMT and who make their clock
3270 * tick in localtime for Windows bug-for-bug compatibility,
3271 * the clock is set in the future, and this will cause e2fsck
3272 * to complain and force a full file system check.
3273 */
3274 if (!(sb->s_flags & MS_RDONLY))
3275 es->s_wtime = cpu_to_le32(get_seconds());
afc32f7e
TT		 3276 es->s_kbytes_written =
3277 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
3278 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
3279 EXT4_SB(sb)->s_sectors_written_start) >> 1));
5d1b1b3f
AK		 3280 ext4_free_blocks_count_set(es, percpu_counter_sum_positive(
3281 &EXT4_SB(sb)->s_freeblocks_counter));
3282 es->s_free_inodes_count = cpu_to_le32(percpu_counter_sum_positive(
3283 &EXT4_SB(sb)->s_freeinodes_counter));
7234ab2a 3284 sb->s_dirt = 0;
ac27a0ec
DK		 3285 BUFFER_TRACE(sbh, "marking dirty");
3286 mark_buffer_dirty(sbh);
914258bf 3287 if (sync) {
c4be0c1d
TS		 3288 error = sync_dirty_buffer(sbh);
3289 if (error)
3290 return error;
3291
3292 error = buffer_write_io_error(sbh);
3293 if (error) {
b31e1552
ES		 3294 ext4_msg(sb, KERN_ERR, "I/O error while writing "
3295 "superblock");
914258bf
TT		 3296 clear_buffer_write_io_error(sbh);
3297 set_buffer_uptodate(sbh);
3298 }
3299 }
c4be0c1d 3300 return error;
ac27a0ec
DK		 3301}
3302
ac27a0ec
DK		 3303/*
3304 * Have we just finished recovery? If so, and if we are mounting (or
3305 * remounting) the filesystem readonly, then we will end up with a
3306 * consistent fs on disk. Record that fact.
3307 */
2b2d6d01
TT		 3308static void ext4_mark_recovery_complete(struct super_block *sb,
3309 struct ext4_super_block *es)
ac27a0ec 3310{
617ba13b 3311 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 3312
0390131b
FM		 3313 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3314 BUG_ON(journal != NULL);
3315 return;
3316 }
dab291af 3317 jbd2_journal_lock_updates(journal);
7ffe1ea8
HK		 3318 if (jbd2_journal_flush(journal) < 0)
3319 goto out;
3320
617ba13b 3321 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
ac27a0ec 3322 sb->s_flags & MS_RDONLY) {
617ba13b 3323 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
e2d67052 3324 ext4_commit_super(sb, 1);
ac27a0ec 3325 }
7ffe1ea8
HK		 3326
3327out:
dab291af 3328 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 3329}
3330
3331/*
3332 * If we are mounting (or read-write remounting) a filesystem whose journal
3333 * has recorded an error from a previous lifetime, move that error to the
3334 * main filesystem now.
3335 */
2b2d6d01
TT		 3336static void ext4_clear_journal_err(struct super_block *sb,
3337 struct ext4_super_block *es)
ac27a0ec
DK		 3338{
3339 journal_t *journal;
3340 int j_errno;
3341 const char *errstr;
3342
0390131b
FM		 3343 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3344
617ba13b 3345 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 3346
3347 /*
3348 * Now check for any error status which may have been recorded in the
617ba13b 3349 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 3350 */
3351
dab291af 3352 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 3353 if (j_errno) {
3354 char nbuf[16];
3355
617ba13b 3356 errstr = ext4_decode_error(sb, j_errno, nbuf);
46e665e9 3357 ext4_warning(sb, __func__, "Filesystem error recorded "
ac27a0ec 3358 "from previous mount: %s", errstr);
46e665e9 3359 ext4_warning(sb, __func__, "Marking fs in need of "
ac27a0ec
DK		 3360 "filesystem check.");
3361
617ba13b
MC		 3362 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
3363 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
e2d67052 3364 ext4_commit_super(sb, 1);
ac27a0ec 3365
dab291af 3366 jbd2_journal_clear_err(journal);
ac27a0ec
DK		 3367 }
3368}
3369
3370/*
3371 * Force the running and committing transactions to commit,
3372 * and wait on the commit.
3373 */
617ba13b 3374int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 3375{
3376 journal_t *journal;
0390131b 3377 int ret = 0;
ac27a0ec
DK		 3378
3379 if (sb->s_flags & MS_RDONLY)
3380 return 0;
3381
617ba13b 3382 journal = EXT4_SB(sb)->s_journal;
7234ab2a 3383 if (journal)
0390131b 3384 ret = ext4_journal_force_commit(journal);
0390131b 3385
ac27a0ec
DK		 3386 return ret;
3387}
3388
2b2d6d01 3389static void ext4_write_super(struct super_block *sb)
ac27a0ec 3390{
ebc1ac16 3391 lock_super(sb);
9ca92389 3392 ext4_commit_super(sb, 1);
ebc1ac16 3393 unlock_super(sb);
ac27a0ec
DK		 3394}
3395
617ba13b 3396static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec 3397{
14ce0cb4 3398 int ret = 0;
9eddacf9 3399 tid_t target;
8d5d02e6 3400 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 3401
9bffad1e 3402 trace_ext4_sync_fs(sb, wait);
8d5d02e6
MC		 3403 flush_workqueue(sbi->dio_unwritten_wq);
3404 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
9ca92389 3405 if (wait)
8d5d02e6 3406 jbd2_log_wait_commit(sbi->s_journal, target);
0390131b 3407 }
14ce0cb4 3408 return ret;
ac27a0ec
DK		 3409}
3410
3411/*
3412 * LVM calls this function before a (read-only) snapshot is created. This
3413 * gives us a chance to flush the journal completely and mark the fs clean.
3414 */
c4be0c1d 3415static int ext4_freeze(struct super_block *sb)
ac27a0ec 3416{
c4be0c1d
TS		 3417 int error = 0;
3418 journal_t *journal;
ac27a0ec 3419
9ca92389
TT		 3420 if (sb->s_flags & MS_RDONLY)
3421 return 0;
ac27a0ec 3422
9ca92389 3423 journal = EXT4_SB(sb)->s_journal;
7ffe1ea8 3424
9ca92389
TT		 3425 /* Now we set up the journal barrier. */
3426 jbd2_journal_lock_updates(journal);
ac27a0ec 3427
9ca92389
TT		 3428 /*
3429 * Don't clear the needs_recovery flag if we failed to flush
3430 * the journal.
3431 */
3432 error = jbd2_journal_flush(journal);
3433 if (error < 0) {
3434 out:
3435 jbd2_journal_unlock_updates(journal);
3436 return error;
ac27a0ec 3437 }
9ca92389
TT		 3438
3439 /* Journal blocked and flushed, clear needs_recovery flag. */
3440 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3441 error = ext4_commit_super(sb, 1);
3442 if (error)
3443 goto out;
c4be0c1d 3444 return 0;
ac27a0ec
DK		 3445}
3446
3447/*
3448 * Called by LVM after the snapshot is done. We need to reset the RECOVER
3449 * flag here, even though the filesystem is not technically dirty yet.
3450 */
c4be0c1d 3451static int ext4_unfreeze(struct super_block *sb)
ac27a0ec 3452{
9ca92389
TT		 3453 if (sb->s_flags & MS_RDONLY)
3454 return 0;
3455
3456 lock_super(sb);
3457 /* Reset the needs_recovery flag before the fs is unlocked. */
3458 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3459 ext4_commit_super(sb, 1);
3460 unlock_super(sb);
3461 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
c4be0c1d 3462 return 0;
ac27a0ec
DK		 3463}
3464
2b2d6d01 3465static int ext4_remount(struct super_block *sb, int *flags, char *data)
ac27a0ec 3466{
2b2d6d01 3467 struct ext4_super_block *es;
617ba13b
MC		 3468 struct ext4_sb_info *sbi = EXT4_SB(sb);
3469 ext4_fsblk_t n_blocks_count = 0;
ac27a0ec 3470 unsigned long old_sb_flags;
617ba13b 3471 struct ext4_mount_options old_opts;
8a266467 3472 ext4_group_t g;
b3881f74 3473 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
ac27a0ec
DK		 3474 int err;
3475#ifdef CONFIG_QUOTA
3476 int i;
3477#endif
3478
337eb00a
AIB		 3479 lock_kernel();
3480
ac27a0ec 3481 /* Store the original options */
bbd6851a 3482 lock_super(sb);
ac27a0ec
DK		 3483 old_sb_flags = sb->s_flags;
3484 old_opts.s_mount_opt = sbi->s_mount_opt;
3485 old_opts.s_resuid = sbi->s_resuid;
3486 old_opts.s_resgid = sbi->s_resgid;
3487 old_opts.s_commit_interval = sbi->s_commit_interval;
30773840
TT		 3488 old_opts.s_min_batch_time = sbi->s_min_batch_time;
3489 old_opts.s_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 3490#ifdef CONFIG_QUOTA
3491 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
3492 for (i = 0; i < MAXQUOTAS; i++)
3493 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
3494#endif
b3881f74
TT		 3495 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
3496 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
ac27a0ec
DK		 3497
3498 /*
3499 * Allow the "check" option to be passed as a remount option.
3500 */
b3881f74
TT		 3501 if (!parse_options(data, sb, NULL, &journal_ioprio,
3502 &n_blocks_count, 1)) {
ac27a0ec
DK		 3503 err = -EINVAL;
3504 goto restore_opts;
3505 }
3506
4ab2f15b 3507 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
46e665e9 3508 ext4_abort(sb, __func__, "Abort forced by user");
ac27a0ec
DK		 3509
3510 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
617ba13b 3511 ((sbi->s_mount_opt & EXT4_MOUNT_POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec
DK		 3512
3513 es = sbi->s_es;
3514
b3881f74 3515 if (sbi->s_journal) {
0390131b 3516 ext4_init_journal_params(sb, sbi->s_journal);
b3881f74
TT		 3517 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3518 }
ac27a0ec
DK		 3519
3520 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
bd81d8ee 3521 n_blocks_count > ext4_blocks_count(es)) {
4ab2f15b 3522 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
ac27a0ec
DK		 3523 err = -EROFS;
3524 goto restore_opts;
3525 }
3526
3527 if (*flags & MS_RDONLY) {
3528 /*
3529 * First of all, the unconditional stuff we have to do
3530 * to disable replay of the journal when we next remount
3531 */
3532 sb->s_flags |= MS_RDONLY;
3533
3534 /*
3535 * OK, test if we are remounting a valid rw partition
3536 * readonly, and if so set the rdonly flag and then
3537 * mark the partition as valid again.
3538 */
617ba13b
MC		 3539 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
3540 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 3541 es->s_state = cpu_to_le16(sbi->s_mount_state);
3542
a63c9eb2 3543 if (sbi->s_journal)
0390131b 3544 ext4_mark_recovery_complete(sb, es);
ac27a0ec 3545 } else {
a13fb1a4
ES		 3546 /* Make sure we can mount this feature set readwrite */
3547 if (!ext4_feature_set_ok(sb, 0)) {
ac27a0ec
DK		 3548 err = -EROFS;
3549 goto restore_opts;
3550 }
8a266467
TT		 3551 /*
3552 * Make sure the group descriptor checksums
0b8e58a1 3553 * are sane. If they aren't, refuse to remount r/w.
8a266467
TT		 3554 */
3555 for (g = 0; g < sbi->s_groups_count; g++) {
3556 struct ext4_group_desc *gdp =
3557 ext4_get_group_desc(sb, g, NULL);
3558
3559 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
b31e1552
ES		 3560 ext4_msg(sb, KERN_ERR,
3561 "ext4_remount: Checksum for group %u failed (%u!=%u)",
8a266467
TT		 3562 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
3563 le16_to_cpu(gdp->bg_checksum));
3564 err = -EINVAL;
3565 goto restore_opts;
3566 }
3567 }
3568
ead6596b
ES		 3569 /*
3570 * If we have an unprocessed orphan list hanging
3571 * around from a previously readonly bdev mount,
3572 * require a full umount/remount for now.
3573 */
3574 if (es->s_last_orphan) {
b31e1552 3575 ext4_msg(sb, KERN_WARNING, "Couldn't "
ead6596b
ES		 3576 "remount RDWR because of unprocessed "
3577 "orphan inode list. Please "
b31e1552 3578 "umount/remount instead");
ead6596b
ES		 3579 err = -EINVAL;
3580 goto restore_opts;
3581 }
3582
ac27a0ec
DK		 3583 /*
3584 * Mounting a RDONLY partition read-write, so reread
3585 * and store the current valid flag. (It may have
3586 * been changed by e2fsck since we originally mounted
3587 * the partition.)
3588 */
0390131b
FM		 3589 if (sbi->s_journal)
3590 ext4_clear_journal_err(sb, es);
ac27a0ec 3591 sbi->s_mount_state = le16_to_cpu(es->s_state);
617ba13b 3592 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
ac27a0ec 3593 goto restore_opts;
2b2d6d01 3594 if (!ext4_setup_super(sb, es, 0))
ac27a0ec
DK		 3595 sb->s_flags &= ~MS_RDONLY;
3596 }
3597 }
6fd058f7 3598 ext4_setup_system_zone(sb);
0390131b 3599 if (sbi->s_journal == NULL)
e2d67052 3600 ext4_commit_super(sb, 1);
0390131b 3601
ac27a0ec
DK		 3602#ifdef CONFIG_QUOTA
3603 /* Release old quota file names */
3604 for (i = 0; i < MAXQUOTAS; i++)
3605 if (old_opts.s_qf_names[i] &&
3606 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3607 kfree(old_opts.s_qf_names[i]);
3608#endif
bbd6851a 3609 unlock_super(sb);
337eb00a 3610 unlock_kernel();
ac27a0ec 3611 return 0;
0b8e58a1 3612
ac27a0ec
DK		 3613restore_opts:
3614 sb->s_flags = old_sb_flags;
3615 sbi->s_mount_opt = old_opts.s_mount_opt;
3616 sbi->s_resuid = old_opts.s_resuid;
3617 sbi->s_resgid = old_opts.s_resgid;
3618 sbi->s_commit_interval = old_opts.s_commit_interval;
30773840
TT		 3619 sbi->s_min_batch_time = old_opts.s_min_batch_time;
3620 sbi->s_max_batch_time = old_opts.s_max_batch_time;
ac27a0ec
DK		 3621#ifdef CONFIG_QUOTA
3622 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
3623 for (i = 0; i < MAXQUOTAS; i++) {
3624 if (sbi->s_qf_names[i] &&
3625 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3626 kfree(sbi->s_qf_names[i]);
3627 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
3628 }
3629#endif
bbd6851a 3630 unlock_super(sb);
337eb00a 3631 unlock_kernel();
ac27a0ec
DK		 3632 return err;
3633}
3634
2b2d6d01 3635static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
ac27a0ec
DK		 3636{
3637 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 3638 struct ext4_sb_info *sbi = EXT4_SB(sb);
3639 struct ext4_super_block *es = sbi->s_es;
960cc398 3640 u64 fsid;
ac27a0ec 3641
5e70030d
BP		 3642 if (test_opt(sb, MINIX_DF)) {
3643 sbi->s_overhead_last = 0;
6bc9feff 3644 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
8df9675f 3645 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
5e70030d 3646 ext4_fsblk_t overhead = 0;
ac27a0ec
DK		 3647
3648 /*
5e70030d
BP		 3649 * Compute the overhead (FS structures). This is constant
3650 * for a given filesystem unless the number of block groups
3651 * changes so we cache the previous value until it does.
ac27a0ec
DK		 3652 */
3653
3654 /*
3655 * All of the blocks before first_data_block are
3656 * overhead
3657 */
3658 overhead = le32_to_cpu(es->s_first_data_block);
3659
3660 /*
3661 * Add the overhead attributed to the superblock and
3662 * block group descriptors. If the sparse superblocks
3663 * feature is turned on, then not all groups have this.
3664 */
3665 for (i = 0; i < ngroups; i++) {
617ba13b
MC		 3666 overhead += ext4_bg_has_super(sb, i) +
3667 ext4_bg_num_gdb(sb, i);
ac27a0ec
DK		 3668 cond_resched();
3669 }
3670
3671 /*
3672 * Every block group has an inode bitmap, a block
3673 * bitmap, and an inode table.
3674 */
5e70030d
BP		 3675 overhead += ngroups * (2 + sbi->s_itb_per_group);
3676 sbi->s_overhead_last = overhead;
3677 smp_wmb();
6bc9feff 3678 sbi->s_blocks_last = ext4_blocks_count(es);
ac27a0ec
DK		 3679 }
3680
617ba13b 3681 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 3682 buf->f_bsize = sb->s_blocksize;
5e70030d 3683 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
6bc6e63f
AK		 3684 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
3685 percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
bd81d8ee
LV		 3686 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3687 if (buf->f_bfree < ext4_r_blocks_count(es))
ac27a0ec
DK		 3688 buf->f_bavail = 0;
3689 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 3690 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
617ba13b 3691 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 3692 fsid = le64_to_cpup((void *)es->s_uuid) ^
3693 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3694 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3695 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
0b8e58a1 3696
ac27a0ec
DK		 3697 return 0;
3698}
3699
0b8e58a1
AD		 3700/* Helper function for writing quotas on sync - we need to start transaction
3701 * before quota file is locked for write. Otherwise the are possible deadlocks:
ac27a0ec 3702 * Process 1 Process 2
617ba13b 3703 * ext4_create() quota_sync()
a269eb18
JK		 3704 * jbd2_journal_start() write_dquot()
3705 * vfs_dq_init() down(dqio_mutex)
dab291af 3706 * down(dqio_mutex) jbd2_journal_start()
ac27a0ec
DK		 3707 *
3708 */
3709
3710#ifdef CONFIG_QUOTA
3711
3712static inline struct inode *dquot_to_inode(struct dquot *dquot)
3713{
3714 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3715}
3716
617ba13b 3717static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 3718{
3719 int ret, err;
3720 handle_t *handle;
3721 struct inode *inode;
3722
3723 inode = dquot_to_inode(dquot);
617ba13b 3724 handle = ext4_journal_start(inode,
0b8e58a1 3725 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 3726 if (IS_ERR(handle))
3727 return PTR_ERR(handle);
3728 ret = dquot_commit(dquot);
617ba13b 3729 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3730 if (!ret)
3731 ret = err;
3732 return ret;
3733}
3734
617ba13b 3735static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 3736{
3737 int ret, err;
3738 handle_t *handle;
3739
617ba13b 3740 handle = ext4_journal_start(dquot_to_inode(dquot),
0b8e58a1 3741 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 3742 if (IS_ERR(handle))
3743 return PTR_ERR(handle);
3744 ret = dquot_acquire(dquot);
617ba13b 3745 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3746 if (!ret)
3747 ret = err;
3748 return ret;
3749}
3750
617ba13b 3751static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 3752{
3753 int ret, err;
3754 handle_t *handle;
3755
617ba13b 3756 handle = ext4_journal_start(dquot_to_inode(dquot),
0b8e58a1 3757 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 3758 if (IS_ERR(handle)) {
3759 /* Release dquot anyway to avoid endless cycle in dqput() */
3760 dquot_release(dquot);
ac27a0ec 3761 return PTR_ERR(handle);
9c3013e9 3762 }
ac27a0ec 3763 ret = dquot_release(dquot);
617ba13b 3764 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3765 if (!ret)
3766 ret = err;
3767 return ret;
3768}
3769
617ba13b 3770static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec 3771{
2c8be6b2 3772 /* Are we journaling quotas? */
617ba13b
MC		 3773 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3774 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
ac27a0ec 3775 dquot_mark_dquot_dirty(dquot);
617ba13b 3776 return ext4_write_dquot(dquot);
ac27a0ec
DK		 3777 } else {
3778 return dquot_mark_dquot_dirty(dquot);
3779 }
3780}
3781
617ba13b 3782static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 3783{
3784 int ret, err;
3785 handle_t *handle;
3786
3787 /* Data block + inode block */
617ba13b 3788 handle = ext4_journal_start(sb->s_root->d_inode, 2);
ac27a0ec
DK		 3789 if (IS_ERR(handle))
3790 return PTR_ERR(handle);
3791 ret = dquot_commit_info(sb, type);
617ba13b 3792 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3793 if (!ret)
3794 ret = err;
3795 return ret;
3796}
3797
3798/*
3799 * Turn on quotas during mount time - we need to find
3800 * the quota file and such...
3801 */
617ba13b 3802static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 3803{
617ba13b 3804 return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
0b8e58a1 3805 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 3806}
3807
3808/*
3809 * Standard function to be called on quota_on
3810 */
617ba13b 3811static int ext4_quota_on(struct super_block *sb, int type, int format_id,
8264613d 3812 char *name, int remount)
ac27a0ec
DK		 3813{
3814 int err;
8264613d 3815 struct path path;
ac27a0ec
DK		 3816
3817 if (!test_opt(sb, QUOTA))
3818 return -EINVAL;
8264613d 3819 /* When remounting, no checks are needed and in fact, name is NULL */
0623543b 3820 if (remount)
8264613d 3821 return vfs_quota_on(sb, type, format_id, name, remount);
0623543b 3822
8264613d 3823 err = kern_path(name, LOOKUP_FOLLOW, &path);
ac27a0ec
DK		 3824 if (err)
3825 return err;
0623543b 3826
ac27a0ec 3827 /* Quotafile not on the same filesystem? */
8264613d
AV		 3828 if (path.mnt->mnt_sb != sb) {
3829 path_put(&path);
ac27a0ec
DK		 3830 return -EXDEV;
3831 }
0623543b
JK		 3832 /* Journaling quota? */
3833 if (EXT4_SB(sb)->s_qf_names[type]) {
2b2d6d01 3834 /* Quotafile not in fs root? */
8264613d 3835 if (path.dentry->d_parent != sb->s_root)
b31e1552
ES		 3836 ext4_msg(sb, KERN_WARNING,
3837 "Quota file not on filesystem root. "
3838 "Journaled quota will not work");
2b2d6d01 3839 }
0623543b
JK		 3840
3841 /*
3842 * When we journal data on quota file, we have to flush journal to see
3843 * all updates to the file when we bypass pagecache...
3844 */
0390131b
FM		 3845 if (EXT4_SB(sb)->s_journal &&
3846 ext4_should_journal_data(path.dentry->d_inode)) {
0623543b
JK		 3847 /*
3848 * We don't need to lock updates but journal_flush() could
3849 * otherwise be livelocked...
3850 */
3851 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 3852 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
0623543b 3853 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 3854 if (err) {
8264613d 3855 path_put(&path);
7ffe1ea8
HK		 3856 return err;
3857 }
0623543b
JK		 3858 }
3859
8264613d
AV		 3860 err = vfs_quota_on_path(sb, type, format_id, &path);
3861 path_put(&path);
77e69dac 3862 return err;
ac27a0ec
DK		 3863}
3864
3865/* Read data from quotafile - avoid pagecache and such because we cannot afford
3866 * acquiring the locks... As quota files are never truncated and quota code
3867 * itself serializes the operations (and noone else should touch the files)
3868 * we don't have to be afraid of races */
617ba13b 3869static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 3870 size_t len, loff_t off)
3871{
3872 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 3873 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 3874 int err = 0;
3875 int offset = off & (sb->s_blocksize - 1);
3876 int tocopy;
3877 size_t toread;
3878 struct buffer_head *bh;
3879 loff_t i_size = i_size_read(inode);
3880
3881 if (off > i_size)
3882 return 0;
3883 if (off+len > i_size)
3884 len = i_size-off;
3885 toread = len;
3886 while (toread > 0) {
3887 tocopy = sb->s_blocksize - offset < toread ?
3888 sb->s_blocksize - offset : toread;
617ba13b 3889 bh = ext4_bread(NULL, inode, blk, 0, &err);
ac27a0ec
DK		 3890 if (err)
3891 return err;
3892 if (!bh) /* A hole? */
3893 memset(data, 0, tocopy);
3894 else
3895 memcpy(data, bh->b_data+offset, tocopy);
3896 brelse(bh);
3897 offset = 0;
3898 toread -= tocopy;
3899 data += tocopy;
3900 blk++;
3901 }
3902 return len;
3903}
3904
3905/* Write to quotafile (we know the transaction is already started and has
3906 * enough credits) */
617ba13b 3907static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 3908 const char *data, size_t len, loff_t off)
3909{
3910 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 3911 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 3912 int err = 0;
3913 int offset = off & (sb->s_blocksize - 1);
3914 int tocopy;
617ba13b 3915 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
ac27a0ec
DK		 3916 size_t towrite = len;
3917 struct buffer_head *bh;
3918 handle_t *handle = journal_current_handle();
3919
0390131b 3920 if (EXT4_SB(sb)->s_journal && !handle) {
b31e1552
ES		 3921 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
3922 " cancelled because transaction is not started",
9c3013e9
JK		 3923 (unsigned long long)off, (unsigned long long)len);
3924 return -EIO;
3925 }
ac27a0ec
DK		 3926 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
3927 while (towrite > 0) {
3928 tocopy = sb->s_blocksize - offset < towrite ?
3929 sb->s_blocksize - offset : towrite;
617ba13b 3930 bh = ext4_bread(handle, inode, blk, 1, &err);
ac27a0ec
DK		 3931 if (!bh)
3932 goto out;
3933 if (journal_quota) {
617ba13b 3934 err = ext4_journal_get_write_access(handle, bh);
ac27a0ec
DK		 3935 if (err) {
3936 brelse(bh);
3937 goto out;
3938 }
3939 }
3940 lock_buffer(bh);
3941 memcpy(bh->b_data+offset, data, tocopy);
3942 flush_dcache_page(bh->b_page);
3943 unlock_buffer(bh);
3944 if (journal_quota)
0390131b 3945 err = ext4_handle_dirty_metadata(handle, NULL, bh);
ac27a0ec
DK		 3946 else {
3947 /* Always do at least ordered writes for quotas */
678aaf48 3948 err = ext4_jbd2_file_inode(handle, inode);
ac27a0ec
DK		 3949 mark_buffer_dirty(bh);
3950 }
3951 brelse(bh);
3952 if (err)
3953 goto out;
3954 offset = 0;
3955 towrite -= tocopy;
3956 data += tocopy;
3957 blk++;
3958 }
3959out:
4d04e4fb
JK		 3960 if (len == towrite) {
3961 mutex_unlock(&inode->i_mutex);
ac27a0ec 3962 return err;
4d04e4fb 3963 }
ac27a0ec
DK		 3964 if (inode->i_size < off+len-towrite) {
3965 i_size_write(inode, off+len-towrite);
617ba13b 3966 EXT4_I(inode)->i_disksize = inode->i_size;
ac27a0ec 3967 }
ac27a0ec 3968 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
617ba13b 3969 ext4_mark_inode_dirty(handle, inode);
ac27a0ec
DK		 3970 mutex_unlock(&inode->i_mutex);
3971 return len - towrite;
3972}
3973
3974#endif
3975
0b8e58a1
AD		 3976static int ext4_get_sb(struct file_system_type *fs_type, int flags,
3977 const char *dev_name, void *data, struct vfsmount *mnt)
ac27a0ec 3978{
0b8e58a1 3979 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super,mnt);
ac27a0ec
DK		 3980}
3981
a214238d 3982#if !defined(CONTIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
24b58424
TT		 3983static struct file_system_type ext2_fs_type = {
3984 .owner = THIS_MODULE,
3985 .name = "ext2",
3986 .get_sb = ext4_get_sb,
3987 .kill_sb = kill_block_super,
3988 .fs_flags = FS_REQUIRES_DEV,
3989};
3990
3991static inline void register_as_ext2(void)
3992{
3993 int err = register_filesystem(&ext2_fs_type);
3994 if (err)
3995 printk(KERN_WARNING
3996 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
3997}
3998
3999static inline void unregister_as_ext2(void)
4000{
4001 unregister_filesystem(&ext2_fs_type);
4002}
4003#else
4004static inline void register_as_ext2(void) { }
4005static inline void unregister_as_ext2(void) { }
4006#endif
4007
a214238d 4008#if !defined(CONTIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
24b58424
TT		 4009static struct file_system_type ext3_fs_type = {
4010 .owner = THIS_MODULE,
4011 .name = "ext3",
4012 .get_sb = ext4_get_sb,
4013 .kill_sb = kill_block_super,
4014 .fs_flags = FS_REQUIRES_DEV,
4015};
4016
4017static inline void register_as_ext3(void)
4018{
4019 int err = register_filesystem(&ext3_fs_type);
4020 if (err)
4021 printk(KERN_WARNING
4022 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
4023}
4024
4025static inline void unregister_as_ext3(void)
4026{
4027 unregister_filesystem(&ext3_fs_type);
4028}
4029#else
4030static inline void register_as_ext3(void) { }
4031static inline void unregister_as_ext3(void) { }
4032#endif
4033
03010a33
TT		 4034static struct file_system_type ext4_fs_type = {
4035 .owner = THIS_MODULE,
4036 .name = "ext4",
4037 .get_sb = ext4_get_sb,
4038 .kill_sb = kill_block_super,
4039 .fs_flags = FS_REQUIRES_DEV,
4040};
4041
617ba13b 4042static int __init init_ext4_fs(void)
ac27a0ec 4043{
c9de560d
AT		 4044 int err;
4045
6fd058f7
TT		 4046 err = init_ext4_system_zone();
4047 if (err)
4048 return err;
3197ebdb
TT		 4049 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
4050 if (!ext4_kset)
6fd058f7 4051 goto out4;
9f6200bb 4052 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
c9de560d 4053 err = init_ext4_mballoc();
ac27a0ec 4054 if (err)
6fd058f7 4055 goto out3;
c9de560d
AT		 4056
4057 err = init_ext4_xattr();
4058 if (err)
4059 goto out2;
ac27a0ec
DK		 4060 err = init_inodecache();
4061 if (err)
4062 goto out1;
24b58424
TT		 4063 register_as_ext2();
4064 register_as_ext3();
03010a33 4065 err = register_filesystem(&ext4_fs_type);
ac27a0ec
DK		 4066 if (err)
4067 goto out;
4068 return 0;
4069out:
24b58424
TT		 4070 unregister_as_ext2();
4071 unregister_as_ext3();
ac27a0ec
DK		 4072 destroy_inodecache();
4073out1:
617ba13b 4074 exit_ext4_xattr();
c9de560d
AT		 4075out2:
4076 exit_ext4_mballoc();
6fd058f7
TT		 4077out3:
4078 remove_proc_entry("fs/ext4", NULL);
4079 kset_unregister(ext4_kset);
4080out4:
4081 exit_ext4_system_zone();
ac27a0ec
DK		 4082 return err;
4083}
4084
617ba13b 4085static void __exit exit_ext4_fs(void)
ac27a0ec 4086{
24b58424
TT		 4087 unregister_as_ext2();
4088 unregister_as_ext3();
03010a33 4089 unregister_filesystem(&ext4_fs_type);
ac27a0ec 4090 destroy_inodecache();
617ba13b 4091 exit_ext4_xattr();
c9de560d 4092 exit_ext4_mballoc();
9f6200bb 4093 remove_proc_entry("fs/ext4", NULL);
3197ebdb 4094 kset_unregister(ext4_kset);
6fd058f7 4095 exit_ext4_system_zone();
ac27a0ec
DK		 4096}
4097
4098MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
83982b6f 4099MODULE_DESCRIPTION("Fourth Extended Filesystem");
ac27a0ec 4100MODULE_LICENSE("GPL");
617ba13b
MC		 4101module_init(init_ext4_fs)
4102module_exit(exit_ext4_fs)
Clone of davem's net-next-2.6 tree