projects / deliverable / linux.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
ext4: Convert BUG_ON checks to use ext4_error() instead
[deliverable/linux.git] / fs / ext4 / super.c
CommitLineData
ac27a0ec 1/*
617ba13b 2 * linux/fs/ext4/super.c
ac27a0ec
DK		 3 *
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (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
12062ddd 336void __ext4_error(struct super_block *sb, const char *function,
2b2d6d01 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
273df556
FM		 350void ext4_error_inode(const char *function, struct inode *inode,
351 const char *fmt, ...)
352{
353 va_list args;
354
355 va_start(args, fmt);
356 printk(KERN_CRIT "EXT4-fs error (device %s): %s: inode #%lu: (comm %s) ",
357 inode->i_sb->s_id, function, inode->i_ino, current->comm);
358 vprintk(fmt, args);
359 printk("\n");
360 va_end(args);
361
362 ext4_handle_error(inode->i_sb);
363}
364
365void ext4_error_file(const char *function, struct file *file,
366 const char *fmt, ...)
367{
368 va_list args;
369 struct inode *inode = file->f_dentry->d_inode;
370 char pathname[80], *path;
371
372 va_start(args, fmt);
373 path = d_path(&(file->f_path), pathname, sizeof(pathname));
374 if (!path)
375 path = "(unknown)";
376 printk(KERN_CRIT
377 "EXT4-fs error (device %s): %s: inode #%lu (comm %s path %s): ",
378 inode->i_sb->s_id, function, inode->i_ino, current->comm, path);
379 vprintk(fmt, args);
380 printk("\n");
381 va_end(args);
382
383 ext4_handle_error(inode->i_sb);
384}
385
2b2d6d01 386static const char *ext4_decode_error(struct super_block *sb, int errno,
ac27a0ec
DK		 387 char nbuf[16])
388{
389 char *errstr = NULL;
390
391 switch (errno) {
392 case -EIO:
393 errstr = "IO failure";
394 break;
395 case -ENOMEM:
396 errstr = "Out of memory";
397 break;
398 case -EROFS:
78f1ddbb
TT		 399 if (!sb || (EXT4_SB(sb)->s_journal &&
400 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
ac27a0ec
DK		 401 errstr = "Journal has aborted";
402 else
403 errstr = "Readonly filesystem";
404 break;
405 default:
406 /* If the caller passed in an extra buffer for unknown
407 * errors, textualise them now. Else we just return
408 * NULL. */
409 if (nbuf) {
410 /* Check for truncated error codes... */
411 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
412 errstr = nbuf;
413 }
414 break;
415 }
416
417 return errstr;
418}
419
617ba13b 420/* __ext4_std_error decodes expected errors from journaling functions
ac27a0ec
DK		 421 * automatically and invokes the appropriate error response. */
422
2b2d6d01 423void __ext4_std_error(struct super_block *sb, const char *function, int errno)
ac27a0ec
DK		 424{
425 char nbuf[16];
426 const char *errstr;
427
428 /* Special case: if the error is EROFS, and we're not already
429 * inside a transaction, then there's really no point in logging
430 * an error. */
431 if (errno == -EROFS && journal_current_handle() == NULL &&
432 (sb->s_flags & MS_RDONLY))
433 return;
434
617ba13b 435 errstr = ext4_decode_error(sb, errno, nbuf);
2b2d6d01
TT		 436 printk(KERN_CRIT "EXT4-fs error (device %s) in %s: %s\n",
437 sb->s_id, function, errstr);
ac27a0ec 438
617ba13b 439 ext4_handle_error(sb);
ac27a0ec
DK		 440}
441
442/*
617ba13b 443 * ext4_abort is a much stronger failure handler than ext4_error. The
ac27a0ec
DK		 444 * abort function may be used to deal with unrecoverable failures such
445 * as journal IO errors or ENOMEM at a critical moment in log management.
446 *
447 * We unconditionally force the filesystem into an ABORT|READONLY state,
448 * unless the error response on the fs has been set to panic in which
449 * case we take the easy way out and panic immediately.
450 */
451
2b2d6d01
TT		 452void ext4_abort(struct super_block *sb, const char *function,
453 const char *fmt, ...)
ac27a0ec
DK		 454{
455 va_list args;
456
ac27a0ec 457 va_start(args, fmt);
2b2d6d01 458 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
ac27a0ec
DK		 459 vprintk(fmt, args);
460 printk("\n");
461 va_end(args);
462
463 if (test_opt(sb, ERRORS_PANIC))
617ba13b 464 panic("EXT4-fs panic from previous error\n");
ac27a0ec
DK		 465
466 if (sb->s_flags & MS_RDONLY)
467 return;
468
b31e1552 469 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
617ba13b 470 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
ac27a0ec 471 sb->s_flags |= MS_RDONLY;
4ab2f15b 472 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
ef2cabf7
HK		 473 if (EXT4_SB(sb)->s_journal)
474 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
ac27a0ec
DK		 475}
476
b31e1552
ES		 477void ext4_msg (struct super_block * sb, const char *prefix,
478 const char *fmt, ...)
479{
480 va_list args;
481
482 va_start(args, fmt);
483 printk("%sEXT4-fs (%s): ", prefix, sb->s_id);
484 vprintk(fmt, args);
485 printk("\n");
486 va_end(args);
487}
488
12062ddd 489void __ext4_warning(struct super_block *sb, const char *function,
2b2d6d01 490 const char *fmt, ...)
ac27a0ec
DK		 491{
492 va_list args;
493
494 va_start(args, fmt);
617ba13b 495 printk(KERN_WARNING "EXT4-fs warning (device %s): %s: ",
ac27a0ec
DK		 496 sb->s_id, function);
497 vprintk(fmt, args);
498 printk("\n");
499 va_end(args);
500}
501
5d1b1b3f 502void ext4_grp_locked_error(struct super_block *sb, ext4_group_t grp,
0b8e58a1 503 const char *function, const char *fmt, ...)
5d1b1b3f
AK		 504__releases(bitlock)
505__acquires(bitlock)
506{
507 va_list args;
508 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
509
510 va_start(args, fmt);
511 printk(KERN_CRIT "EXT4-fs error (device %s): %s: ", sb->s_id, function);
512 vprintk(fmt, args);
513 printk("\n");
514 va_end(args);
515
516 if (test_opt(sb, ERRORS_CONT)) {
517 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
518 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
e2d67052 519 ext4_commit_super(sb, 0);
5d1b1b3f
AK		 520 return;
521 }
522 ext4_unlock_group(sb, grp);
523 ext4_handle_error(sb);
524 /*
525 * We only get here in the ERRORS_RO case; relocking the group
526 * may be dangerous, but nothing bad will happen since the
527 * filesystem will have already been marked read/only and the
528 * journal has been aborted. We return 1 as a hint to callers
529 * who might what to use the return value from
530 * ext4_grp_locked_error() to distinguish beween the
531 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
532 * aggressively from the ext4 function in question, with a
533 * more appropriate error code.
534 */
535 ext4_lock_group(sb, grp);
536 return;
537}
538
617ba13b 539void ext4_update_dynamic_rev(struct super_block *sb)
ac27a0ec 540{
617ba13b 541 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 542
617ba13b 543 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
ac27a0ec
DK		 544 return;
545
12062ddd 546 ext4_warning(sb,
ac27a0ec
DK		 547 "updating to rev %d because of new feature flag, "
548 "running e2fsck is recommended",
617ba13b 549 EXT4_DYNAMIC_REV);
ac27a0ec 550
617ba13b
MC		 551 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
552 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
553 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
ac27a0ec
DK		 554 /* leave es->s_feature_*compat flags alone */
555 /* es->s_uuid will be set by e2fsck if empty */
556
557 /*
558 * The rest of the superblock fields should be zero, and if not it
559 * means they are likely already in use, so leave them alone. We
560 * can leave it up to e2fsck to clean up any inconsistencies there.
561 */
562}
563
564/*
565 * Open the external journal device
566 */
b31e1552 567static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
ac27a0ec
DK		 568{
569 struct block_device *bdev;
570 char b[BDEVNAME_SIZE];
571
572 bdev = open_by_devnum(dev, FMODE_READ|FMODE_WRITE);
573 if (IS_ERR(bdev))
574 goto fail;
575 return bdev;
576
577fail:
b31e1552 578 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
ac27a0ec
DK		 579 __bdevname(dev, b), PTR_ERR(bdev));
580 return NULL;
581}
582
583/*
584 * Release the journal device
585 */
617ba13b 586static int ext4_blkdev_put(struct block_device *bdev)
ac27a0ec
DK		 587{
588 bd_release(bdev);
9a1c3542 589 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
ac27a0ec
DK		 590}
591
617ba13b 592static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
ac27a0ec
DK		 593{
594 struct block_device *bdev;
595 int ret = -ENODEV;
596
597 bdev = sbi->journal_bdev;
598 if (bdev) {
617ba13b 599 ret = ext4_blkdev_put(bdev);
ac27a0ec
DK		 600 sbi->journal_bdev = NULL;
601 }
602 return ret;
603}
604
605static inline struct inode *orphan_list_entry(struct list_head *l)
606{
617ba13b 607 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
ac27a0ec
DK		 608}
609
617ba13b 610static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
ac27a0ec
DK		 611{
612 struct list_head *l;
613
b31e1552
ES		 614 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
615 le32_to_cpu(sbi->s_es->s_last_orphan));
ac27a0ec
DK		 616
617 printk(KERN_ERR "sb_info orphan list:\n");
618 list_for_each(l, &sbi->s_orphan) {
619 struct inode *inode = orphan_list_entry(l);
620 printk(KERN_ERR " "
621 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
622 inode->i_sb->s_id, inode->i_ino, inode,
623 inode->i_mode, inode->i_nlink,
624 NEXT_ORPHAN(inode));
625 }
626}
627
2b2d6d01 628static void ext4_put_super(struct super_block *sb)
ac27a0ec 629{
617ba13b
MC		 630 struct ext4_sb_info *sbi = EXT4_SB(sb);
631 struct ext4_super_block *es = sbi->s_es;
ef2cabf7 632 int i, err;
ac27a0ec 633
4c0425ff
MC		 634 flush_workqueue(sbi->dio_unwritten_wq);
635 destroy_workqueue(sbi->dio_unwritten_wq);
636
a9e220f8 637 lock_super(sb);
6cfd0148 638 lock_kernel();
8c85e125 639 if (sb->s_dirt)
ebc1ac16 640 ext4_commit_super(sb, 1);
8c85e125 641
0390131b
FM		 642 if (sbi->s_journal) {
643 err = jbd2_journal_destroy(sbi->s_journal);
644 sbi->s_journal = NULL;
645 if (err < 0)
646 ext4_abort(sb, __func__,
647 "Couldn't clean up the journal");
648 }
d4edac31
JB		 649
650 ext4_release_system_zone(sb);
651 ext4_mb_release(sb);
652 ext4_ext_release(sb);
653 ext4_xattr_put_super(sb);
654
ac27a0ec 655 if (!(sb->s_flags & MS_RDONLY)) {
617ba13b 656 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 657 es->s_state = cpu_to_le16(sbi->s_mount_state);
e2d67052 658 ext4_commit_super(sb, 1);
ac27a0ec 659 }
240799cd 660 if (sbi->s_proc) {
9f6200bb 661 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 662 }
3197ebdb 663 kobject_del(&sbi->s_kobj);
ac27a0ec
DK		 664
665 for (i = 0; i < sbi->s_gdb_count; i++)
666 brelse(sbi->s_group_desc[i]);
667 kfree(sbi->s_group_desc);
c5ca7c76
TT		 668 if (is_vmalloc_addr(sbi->s_flex_groups))
669 vfree(sbi->s_flex_groups);
670 else
671 kfree(sbi->s_flex_groups);
ac27a0ec
DK		 672 percpu_counter_destroy(&sbi->s_freeblocks_counter);
673 percpu_counter_destroy(&sbi->s_freeinodes_counter);
674 percpu_counter_destroy(&sbi->s_dirs_counter);
6bc6e63f 675 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
ac27a0ec
DK		 676 brelse(sbi->s_sbh);
677#ifdef CONFIG_QUOTA
678 for (i = 0; i < MAXQUOTAS; i++)
679 kfree(sbi->s_qf_names[i]);
680#endif
681
682 /* Debugging code just in case the in-memory inode orphan list
683 * isn't empty. The on-disk one can be non-empty if we've
684 * detected an error and taken the fs readonly, but the
685 * in-memory list had better be clean by this point. */
686 if (!list_empty(&sbi->s_orphan))
687 dump_orphan_list(sb, sbi);
688 J_ASSERT(list_empty(&sbi->s_orphan));
689
f98393a6 690 invalidate_bdev(sb->s_bdev);
ac27a0ec
DK		 691 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
692 /*
693 * Invalidate the journal device's buffers. We don't want them
694 * floating about in memory - the physical journal device may
695 * hotswapped, and it breaks the `ro-after' testing code.
696 */
697 sync_blockdev(sbi->journal_bdev);
f98393a6 698 invalidate_bdev(sbi->journal_bdev);
617ba13b 699 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 700 }
701 sb->s_fs_info = NULL;
3197ebdb
TT		 702 /*
703 * Now that we are completely done shutting down the
704 * superblock, we need to actually destroy the kobject.
705 */
706 unlock_kernel();
707 unlock_super(sb);
708 kobject_put(&sbi->s_kobj);
709 wait_for_completion(&sbi->s_kobj_unregister);
705895b6 710 kfree(sbi->s_blockgroup_lock);
ac27a0ec 711 kfree(sbi);
ac27a0ec
DK		 712}
713
e18b890b 714static struct kmem_cache *ext4_inode_cachep;
ac27a0ec
DK		 715
716/*
717 * Called inside transaction, so use GFP_NOFS
718 */
617ba13b 719static struct inode *ext4_alloc_inode(struct super_block *sb)
ac27a0ec 720{
617ba13b 721 struct ext4_inode_info *ei;
ac27a0ec 722
e6b4f8da 723 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
ac27a0ec
DK		 724 if (!ei)
725 return NULL;
0b8e58a1 726
ac27a0ec 727 ei->vfs_inode.i_version = 1;
91246c00 728 ei->vfs_inode.i_data.writeback_index = 0;
a86c6181 729 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
c9de560d
AT		 730 INIT_LIST_HEAD(&ei->i_prealloc_list);
731 spin_lock_init(&ei->i_prealloc_lock);
0390131b
FM		 732 /*
733 * Note: We can be called before EXT4_SB(sb)->s_journal is set,
734 * therefore it can be null here. Don't check it, just initialize
735 * jinode.
736 */
678aaf48 737 jbd2_journal_init_jbd_inode(&ei->jinode, &ei->vfs_inode);
d2a17637
MC		 738 ei->i_reserved_data_blocks = 0;
739 ei->i_reserved_meta_blocks = 0;
740 ei->i_allocated_meta_blocks = 0;
9d0be502 741 ei->i_da_metadata_calc_len = 0;
d2a17637
MC		 742 ei->i_delalloc_reserved_flag = 0;
743 spin_lock_init(&(ei->i_block_reservation_lock));
a9e7f447
DM		 744#ifdef CONFIG_QUOTA
745 ei->i_reserved_quota = 0;
746#endif
c7064ef1 747 INIT_LIST_HEAD(&ei->i_completed_io_list);
744692dc 748 spin_lock_init(&ei->i_completed_io_lock);
8d5d02e6 749 ei->cur_aio_dio = NULL;
b436b9be
JK		 750 ei->i_sync_tid = 0;
751 ei->i_datasync_tid = 0;
0b8e58a1 752
ac27a0ec
DK		 753 return &ei->vfs_inode;
754}
755
617ba13b 756static void ext4_destroy_inode(struct inode *inode)
ac27a0ec 757{
9f7dd93d 758 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
b31e1552
ES		 759 ext4_msg(inode->i_sb, KERN_ERR,
760 "Inode %lu (%p): orphan list check failed!",
761 inode->i_ino, EXT4_I(inode));
9f7dd93d
VA		 762 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
763 EXT4_I(inode), sizeof(struct ext4_inode_info),
764 true);
765 dump_stack();
766 }
617ba13b 767 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
ac27a0ec
DK		 768}
769
51cc5068 770static void init_once(void *foo)
ac27a0ec 771{
617ba13b 772 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
ac27a0ec 773
a35afb83 774 INIT_LIST_HEAD(&ei->i_orphan);
03010a33 775#ifdef CONFIG_EXT4_FS_XATTR
a35afb83 776 init_rwsem(&ei->xattr_sem);
ac27a0ec 777#endif
0e855ac8 778 init_rwsem(&ei->i_data_sem);
a35afb83 779 inode_init_once(&ei->vfs_inode);
ac27a0ec
DK		 780}
781
782static int init_inodecache(void)
783{
617ba13b
MC		 784 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
785 sizeof(struct ext4_inode_info),
ac27a0ec
DK		 786 0, (SLAB_RECLAIM_ACCOUNT|
787 SLAB_MEM_SPREAD),
20c2df83 788 init_once);
617ba13b 789 if (ext4_inode_cachep == NULL)
ac27a0ec
DK		 790 return -ENOMEM;
791 return 0;
792}
793
794static void destroy_inodecache(void)
795{
617ba13b 796 kmem_cache_destroy(ext4_inode_cachep);
ac27a0ec
DK		 797}
798
617ba13b 799static void ext4_clear_inode(struct inode *inode)
ac27a0ec 800{
c2ea3fde 801 ext4_discard_preallocations(inode);
0390131b
FM		 802 if (EXT4_JOURNAL(inode))
803 jbd2_journal_release_jbd_inode(EXT4_SB(inode->i_sb)->s_journal,
678aaf48 804 &EXT4_I(inode)->jinode);
ac27a0ec
DK		 805}
806
2b2d6d01
TT		 807static inline void ext4_show_quota_options(struct seq_file *seq,
808 struct super_block *sb)
ac27a0ec
DK		 809{
810#if defined(CONFIG_QUOTA)
617ba13b 811 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 812
5a20bdfc
JK		 813 if (sbi->s_jquota_fmt) {
814 char *fmtname = "";
815
816 switch (sbi->s_jquota_fmt) {
817 case QFMT_VFS_OLD:
818 fmtname = "vfsold";
819 break;
820 case QFMT_VFS_V0:
821 fmtname = "vfsv0";
822 break;
823 case QFMT_VFS_V1:
824 fmtname = "vfsv1";
825 break;
826 }
827 seq_printf(seq, ",jqfmt=%s", fmtname);
828 }
ac27a0ec
DK		 829
830 if (sbi->s_qf_names[USRQUOTA])
831 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
832
833 if (sbi->s_qf_names[GRPQUOTA])
834 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
835
482a7425 836 if (test_opt(sb, USRQUOTA))
ac27a0ec
DK		 837 seq_puts(seq, ",usrquota");
838
482a7425 839 if (test_opt(sb, GRPQUOTA))
ac27a0ec
DK		 840 seq_puts(seq, ",grpquota");
841#endif
842}
843
d9c9bef1
MS		 844/*
845 * Show an option if
846 * - it's set to a non-default value OR
847 * - if the per-sb default is different from the global default
848 */
617ba13b 849static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
ac27a0ec 850{
aa22df2c
AK		 851 int def_errors;
852 unsigned long def_mount_opts;
ac27a0ec 853 struct super_block *sb = vfs->mnt_sb;
d9c9bef1
MS		 854 struct ext4_sb_info *sbi = EXT4_SB(sb);
855 struct ext4_super_block *es = sbi->s_es;
d9c9bef1
MS		 856
857 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
aa22df2c 858 def_errors = le16_to_cpu(es->s_errors);
d9c9bef1
MS		 859
860 if (sbi->s_sb_block != 1)
861 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
862 if (test_opt(sb, MINIX_DF))
863 seq_puts(seq, ",minixdf");
aa22df2c 864 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
d9c9bef1
MS		 865 seq_puts(seq, ",grpid");
866 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
867 seq_puts(seq, ",nogrpid");
868 if (sbi->s_resuid != EXT4_DEF_RESUID ||
869 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
870 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
871 }
872 if (sbi->s_resgid != EXT4_DEF_RESGID ||
873 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
874 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
875 }
bb4f397a 876 if (test_opt(sb, ERRORS_RO)) {
d9c9bef1 877 if (def_errors == EXT4_ERRORS_PANIC ||
bb4f397a
AK		 878 def_errors == EXT4_ERRORS_CONTINUE) {
879 seq_puts(seq, ",errors=remount-ro");
d9c9bef1
MS		 880 }
881 }
aa22df2c 882 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
bb4f397a 883 seq_puts(seq, ",errors=continue");
aa22df2c 884 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
d9c9bef1 885 seq_puts(seq, ",errors=panic");
aa22df2c 886 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
d9c9bef1 887 seq_puts(seq, ",nouid32");
aa22df2c 888 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
d9c9bef1
MS		 889 seq_puts(seq, ",debug");
890 if (test_opt(sb, OLDALLOC))
891 seq_puts(seq, ",oldalloc");
03010a33 892#ifdef CONFIG_EXT4_FS_XATTR
aa22df2c
AK		 893 if (test_opt(sb, XATTR_USER) &&
894 !(def_mount_opts & EXT4_DEFM_XATTR_USER))
d9c9bef1
MS		 895 seq_puts(seq, ",user_xattr");
896 if (!test_opt(sb, XATTR_USER) &&
897 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
898 seq_puts(seq, ",nouser_xattr");
899 }
900#endif
03010a33 901#ifdef CONFIG_EXT4_FS_POSIX_ACL
aa22df2c 902 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
d9c9bef1
MS		 903 seq_puts(seq, ",acl");
904 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
905 seq_puts(seq, ",noacl");
906#endif
30773840 907 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
d9c9bef1
MS		 908 seq_printf(seq, ",commit=%u",
909 (unsigned) (sbi->s_commit_interval / HZ));
910 }
30773840
TT		 911 if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
912 seq_printf(seq, ",min_batch_time=%u",
913 (unsigned) sbi->s_min_batch_time);
914 }
915 if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
916 seq_printf(seq, ",max_batch_time=%u",
917 (unsigned) sbi->s_min_batch_time);
918 }
919
571640ca
ES		 920 /*
921 * We're changing the default of barrier mount option, so
922 * let's always display its mount state so it's clear what its
923 * status is.
924 */
925 seq_puts(seq, ",barrier=");
926 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
cd0b6a39
TT		 927 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
928 seq_puts(seq, ",journal_async_commit");
d9c9bef1
MS		 929 if (test_opt(sb, NOBH))
930 seq_puts(seq, ",nobh");
25ec56b5
JNC		 931 if (test_opt(sb, I_VERSION))
932 seq_puts(seq, ",i_version");
dd919b98
AK		 933 if (!test_opt(sb, DELALLOC))
934 seq_puts(seq, ",nodelalloc");
935
ac27a0ec 936
cb45bbe4
MS		 937 if (sbi->s_stripe)
938 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
aa22df2c
AK		 939 /*
940 * journal mode get enabled in different ways
941 * So just print the value even if we didn't specify it
942 */
617ba13b 943 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
ac27a0ec 944 seq_puts(seq, ",data=journal");
617ba13b 945 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
ac27a0ec 946 seq_puts(seq, ",data=ordered");
617ba13b 947 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
ac27a0ec
DK		 948 seq_puts(seq, ",data=writeback");
949
240799cd
TT		 950 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
951 seq_printf(seq, ",inode_readahead_blks=%u",
952 sbi->s_inode_readahead_blks);
953
5bf5683a
HK		 954 if (test_opt(sb, DATA_ERR_ABORT))
955 seq_puts(seq, ",data_err=abort");
956
afd4672d 957 if (test_opt(sb, NO_AUTO_DA_ALLOC))
06705bff 958 seq_puts(seq, ",noauto_da_alloc");
afd4672d 959
5328e635
ES		 960 if (test_opt(sb, DISCARD))
961 seq_puts(seq, ",discard");
962
e3bb52ae
ES		 963 if (test_opt(sb, NOLOAD))
964 seq_puts(seq, ",norecovery");
965
744692dc
JZ		 966 if (test_opt(sb, DIOREAD_NOLOCK))
967 seq_puts(seq, ",dioread_nolock");
968
617ba13b 969 ext4_show_quota_options(seq, sb);
0b8e58a1 970
ac27a0ec
DK		 971 return 0;
972}
973
1b961ac0 974static struct inode *ext4_nfs_get_inode(struct super_block *sb,
0b8e58a1 975 u64 ino, u32 generation)
ac27a0ec 976{
ac27a0ec 977 struct inode *inode;
ac27a0ec 978
617ba13b 979 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
ac27a0ec 980 return ERR_PTR(-ESTALE);
617ba13b 981 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
ac27a0ec
DK		 982 return ERR_PTR(-ESTALE);
983
984 /* iget isn't really right if the inode is currently unallocated!!
985 *
617ba13b 986 * ext4_read_inode will return a bad_inode if the inode had been
ac27a0ec
DK		 987 * deleted, so we should be safe.
988 *
989 * Currently we don't know the generation for parent directory, so
990 * a generation of 0 means "accept any"
991 */
1d1fe1ee
DH		 992 inode = ext4_iget(sb, ino);
993 if (IS_ERR(inode))
994 return ERR_CAST(inode);
995 if (generation && inode->i_generation != generation) {
ac27a0ec
DK		 996 iput(inode);
997 return ERR_PTR(-ESTALE);
998 }
1b961ac0
CH		 999
1000 return inode;
1001}
1002
1003static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
0b8e58a1 1004 int fh_len, int fh_type)
1b961ac0
CH		 1005{
1006 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1007 ext4_nfs_get_inode);
1008}
1009
1010static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
0b8e58a1 1011 int fh_len, int fh_type)
1b961ac0
CH		 1012{
1013 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1014 ext4_nfs_get_inode);
ac27a0ec
DK		 1015}
1016
c39a7f84
TO		 1017/*
1018 * Try to release metadata pages (indirect blocks, directories) which are
1019 * mapped via the block device. Since these pages could have journal heads
1020 * which would prevent try_to_free_buffers() from freeing them, we must use
1021 * jbd2 layer's try_to_free_buffers() function to release them.
1022 */
0b8e58a1
AD		 1023static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1024 gfp_t wait)
c39a7f84
TO		 1025{
1026 journal_t *journal = EXT4_SB(sb)->s_journal;
1027
1028 WARN_ON(PageChecked(page));
1029 if (!page_has_buffers(page))
1030 return 0;
1031 if (journal)
1032 return jbd2_journal_try_to_free_buffers(journal, page,
1033 wait & ~__GFP_WAIT);
1034 return try_to_free_buffers(page);
1035}
1036
ac27a0ec 1037#ifdef CONFIG_QUOTA
af5bc92d 1038#define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
2b2d6d01 1039#define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
ac27a0ec 1040
617ba13b
MC		 1041static int ext4_write_dquot(struct dquot *dquot);
1042static int ext4_acquire_dquot(struct dquot *dquot);
1043static int ext4_release_dquot(struct dquot *dquot);
1044static int ext4_mark_dquot_dirty(struct dquot *dquot);
1045static int ext4_write_info(struct super_block *sb, int type);
6f28e087
JK		 1046static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1047 char *path, int remount);
617ba13b
MC		 1048static int ext4_quota_on_mount(struct super_block *sb, int type);
1049static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 1050 size_t len, loff_t off);
617ba13b 1051static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 1052 const char *data, size_t len, loff_t off);
1053
61e225dc 1054static const struct dquot_operations ext4_quota_operations = {
edf72453
JK		 1055 .initialize = dquot_initialize,
1056 .drop = dquot_drop,
ac27a0ec 1057 .alloc_space = dquot_alloc_space,
60e58e0f
MC		 1058 .reserve_space = dquot_reserve_space,
1059 .claim_space = dquot_claim_space,
1060 .release_rsv = dquot_release_reserved_space,
a9e7f447 1061#ifdef CONFIG_QUOTA
60e58e0f 1062 .get_reserved_space = ext4_get_reserved_space,
a9e7f447 1063#endif
ac27a0ec
DK		 1064 .alloc_inode = dquot_alloc_inode,
1065 .free_space = dquot_free_space,
1066 .free_inode = dquot_free_inode,
1067 .transfer = dquot_transfer,
617ba13b
MC		 1068 .write_dquot = ext4_write_dquot,
1069 .acquire_dquot = ext4_acquire_dquot,
1070 .release_dquot = ext4_release_dquot,
1071 .mark_dirty = ext4_mark_dquot_dirty,
a5b5ee32
JK		 1072 .write_info = ext4_write_info,
1073 .alloc_dquot = dquot_alloc,
1074 .destroy_dquot = dquot_destroy,
ac27a0ec
DK		 1075};
1076
0d54b217 1077static const struct quotactl_ops ext4_qctl_operations = {
617ba13b 1078 .quota_on = ext4_quota_on,
ac27a0ec
DK		 1079 .quota_off = vfs_quota_off,
1080 .quota_sync = vfs_quota_sync,
1081 .get_info = vfs_get_dqinfo,
1082 .set_info = vfs_set_dqinfo,
1083 .get_dqblk = vfs_get_dqblk,
1084 .set_dqblk = vfs_set_dqblk
1085};
1086#endif
1087
ee9b6d61 1088static const struct super_operations ext4_sops = {
617ba13b
MC		 1089 .alloc_inode = ext4_alloc_inode,
1090 .destroy_inode = ext4_destroy_inode,
617ba13b
MC		 1091 .write_inode = ext4_write_inode,
1092 .dirty_inode = ext4_dirty_inode,
1093 .delete_inode = ext4_delete_inode,
1094 .put_super = ext4_put_super,
617ba13b 1095 .sync_fs = ext4_sync_fs,
c4be0c1d
TS		 1096 .freeze_fs = ext4_freeze,
1097 .unfreeze_fs = ext4_unfreeze,
617ba13b
MC		 1098 .statfs = ext4_statfs,
1099 .remount_fs = ext4_remount,
1100 .clear_inode = ext4_clear_inode,
1101 .show_options = ext4_show_options,
ac27a0ec 1102#ifdef CONFIG_QUOTA
617ba13b
MC		 1103 .quota_read = ext4_quota_read,
1104 .quota_write = ext4_quota_write,
ac27a0ec 1105#endif
c39a7f84 1106 .bdev_try_to_free_page = bdev_try_to_free_page,
ac27a0ec
DK		 1107};
1108
9ca92389
TT		 1109static const struct super_operations ext4_nojournal_sops = {
1110 .alloc_inode = ext4_alloc_inode,
1111 .destroy_inode = ext4_destroy_inode,
1112 .write_inode = ext4_write_inode,
1113 .dirty_inode = ext4_dirty_inode,
1114 .delete_inode = ext4_delete_inode,
1115 .write_super = ext4_write_super,
1116 .put_super = ext4_put_super,
1117 .statfs = ext4_statfs,
1118 .remount_fs = ext4_remount,
1119 .clear_inode = ext4_clear_inode,
1120 .show_options = ext4_show_options,
1121#ifdef CONFIG_QUOTA
1122 .quota_read = ext4_quota_read,
1123 .quota_write = ext4_quota_write,
1124#endif
1125 .bdev_try_to_free_page = bdev_try_to_free_page,
1126};
1127
39655164 1128static const struct export_operations ext4_export_ops = {
1b961ac0
CH		 1129 .fh_to_dentry = ext4_fh_to_dentry,
1130 .fh_to_parent = ext4_fh_to_parent,
617ba13b 1131 .get_parent = ext4_get_parent,
ac27a0ec
DK		 1132};
1133
1134enum {
1135 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1136 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
01436ef2 1137 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
ac27a0ec 1138 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
06705bff 1139 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh,
30773840 1140 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
c3191067 1141 Opt_journal_update, Opt_journal_dev,
818d276c 1142 Opt_journal_checksum, Opt_journal_async_commit,
ac27a0ec 1143 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
296c355c 1144 Opt_data_err_abort, Opt_data_err_ignore,
ac27a0ec 1145 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
5a20bdfc
JK		 1146 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1147 Opt_noquota, Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err,
1148 Opt_resize, Opt_usrquota, Opt_grpquota, Opt_i_version,
01436ef2 1149 Opt_stripe, Opt_delalloc, Opt_nodelalloc,
6fd058f7 1150 Opt_block_validity, Opt_noblock_validity,
5328e635 1151 Opt_inode_readahead_blks, Opt_journal_ioprio,
744692dc 1152 Opt_dioread_nolock, Opt_dioread_lock,
5328e635 1153 Opt_discard, Opt_nodiscard,
ac27a0ec
DK		 1154};
1155
a447c093 1156static const match_table_t tokens = {
ac27a0ec
DK		 1157 {Opt_bsd_df, "bsddf"},
1158 {Opt_minix_df, "minixdf"},
1159 {Opt_grpid, "grpid"},
1160 {Opt_grpid, "bsdgroups"},
1161 {Opt_nogrpid, "nogrpid"},
1162 {Opt_nogrpid, "sysvgroups"},
1163 {Opt_resgid, "resgid=%u"},
1164 {Opt_resuid, "resuid=%u"},
1165 {Opt_sb, "sb=%u"},
1166 {Opt_err_cont, "errors=continue"},
1167 {Opt_err_panic, "errors=panic"},
1168 {Opt_err_ro, "errors=remount-ro"},
1169 {Opt_nouid32, "nouid32"},
ac27a0ec
DK		 1170 {Opt_debug, "debug"},
1171 {Opt_oldalloc, "oldalloc"},
1172 {Opt_orlov, "orlov"},
1173 {Opt_user_xattr, "user_xattr"},
1174 {Opt_nouser_xattr, "nouser_xattr"},
1175 {Opt_acl, "acl"},
1176 {Opt_noacl, "noacl"},
ac27a0ec 1177 {Opt_noload, "noload"},
e3bb52ae 1178 {Opt_noload, "norecovery"},
ac27a0ec
DK		 1179 {Opt_nobh, "nobh"},
1180 {Opt_bh, "bh"},
1181 {Opt_commit, "commit=%u"},
30773840
TT		 1182 {Opt_min_batch_time, "min_batch_time=%u"},
1183 {Opt_max_batch_time, "max_batch_time=%u"},
ac27a0ec 1184 {Opt_journal_update, "journal=update"},
ac27a0ec 1185 {Opt_journal_dev, "journal_dev=%u"},
818d276c
GS		 1186 {Opt_journal_checksum, "journal_checksum"},
1187 {Opt_journal_async_commit, "journal_async_commit"},
ac27a0ec
DK		 1188 {Opt_abort, "abort"},
1189 {Opt_data_journal, "data=journal"},
1190 {Opt_data_ordered, "data=ordered"},
1191 {Opt_data_writeback, "data=writeback"},
5bf5683a
HK		 1192 {Opt_data_err_abort, "data_err=abort"},
1193 {Opt_data_err_ignore, "data_err=ignore"},
ac27a0ec
DK		 1194 {Opt_offusrjquota, "usrjquota="},
1195 {Opt_usrjquota, "usrjquota=%s"},
1196 {Opt_offgrpjquota, "grpjquota="},
1197 {Opt_grpjquota, "grpjquota=%s"},
1198 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1199 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
5a20bdfc 1200 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
ac27a0ec
DK		 1201 {Opt_grpquota, "grpquota"},
1202 {Opt_noquota, "noquota"},
1203 {Opt_quota, "quota"},
1204 {Opt_usrquota, "usrquota"},
1205 {Opt_barrier, "barrier=%u"},
06705bff
TT		 1206 {Opt_barrier, "barrier"},
1207 {Opt_nobarrier, "nobarrier"},
25ec56b5 1208 {Opt_i_version, "i_version"},
c9de560d 1209 {Opt_stripe, "stripe=%u"},
ac27a0ec 1210 {Opt_resize, "resize"},
64769240 1211 {Opt_delalloc, "delalloc"},
dd919b98 1212 {Opt_nodelalloc, "nodelalloc"},
6fd058f7
TT		 1213 {Opt_block_validity, "block_validity"},
1214 {Opt_noblock_validity, "noblock_validity"},
240799cd 1215 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
b3881f74 1216 {Opt_journal_ioprio, "journal_ioprio=%u"},
afd4672d 1217 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
06705bff
TT		 1218 {Opt_auto_da_alloc, "auto_da_alloc"},
1219 {Opt_noauto_da_alloc, "noauto_da_alloc"},
744692dc
JZ		 1220 {Opt_dioread_nolock, "dioread_nolock"},
1221 {Opt_dioread_lock, "dioread_lock"},
5328e635
ES		 1222 {Opt_discard, "discard"},
1223 {Opt_nodiscard, "nodiscard"},
f3f12faa 1224 {Opt_err, NULL},
ac27a0ec
DK		 1225};
1226
617ba13b 1227static ext4_fsblk_t get_sb_block(void **data)
ac27a0ec 1228{
617ba13b 1229 ext4_fsblk_t sb_block;
ac27a0ec
DK		 1230 char *options = (char *) *data;
1231
1232 if (!options || strncmp(options, "sb=", 3) != 0)
1233 return 1; /* Default location */
0b8e58a1 1234
ac27a0ec 1235 options += 3;
0b8e58a1 1236 /* TODO: use simple_strtoll with >32bit ext4 */
ac27a0ec
DK		 1237 sb_block = simple_strtoul(options, &options, 0);
1238 if (*options && *options != ',') {
4776004f 1239 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
ac27a0ec
DK		 1240 (char *) *data);
1241 return 1;
1242 }
1243 if (*options == ',')
1244 options++;
1245 *data = (void *) options;
0b8e58a1 1246
ac27a0ec
DK		 1247 return sb_block;
1248}
1249
b3881f74 1250#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
437ca0fd
DM		 1251static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1252 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
b3881f74 1253
56c50f11
DM		 1254#ifdef CONFIG_QUOTA
1255static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1256{
1257 struct ext4_sb_info *sbi = EXT4_SB(sb);
1258 char *qname;
1259
1260 if (sb_any_quota_loaded(sb) &&
1261 !sbi->s_qf_names[qtype]) {
1262 ext4_msg(sb, KERN_ERR,
1263 "Cannot change journaled "
1264 "quota options when quota turned on");
1265 return 0;
1266 }
1267 qname = match_strdup(args);
1268 if (!qname) {
1269 ext4_msg(sb, KERN_ERR,
1270 "Not enough memory for storing quotafile name");
1271 return 0;
1272 }
1273 if (sbi->s_qf_names[qtype] &&
1274 strcmp(sbi->s_qf_names[qtype], qname)) {
1275 ext4_msg(sb, KERN_ERR,
1276 "%s quota file already specified", QTYPE2NAME(qtype));
1277 kfree(qname);
1278 return 0;
1279 }
1280 sbi->s_qf_names[qtype] = qname;
1281 if (strchr(sbi->s_qf_names[qtype], '/')) {
1282 ext4_msg(sb, KERN_ERR,
1283 "quotafile must be on filesystem root");
1284 kfree(sbi->s_qf_names[qtype]);
1285 sbi->s_qf_names[qtype] = NULL;
1286 return 0;
1287 }
1288 set_opt(sbi->s_mount_opt, QUOTA);
1289 return 1;
1290}
1291
1292static int clear_qf_name(struct super_block *sb, int qtype)
1293{
1294
1295 struct ext4_sb_info *sbi = EXT4_SB(sb);
1296
1297 if (sb_any_quota_loaded(sb) &&
1298 sbi->s_qf_names[qtype]) {
1299 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1300 " when quota turned on");
1301 return 0;
1302 }
1303 /*
1304 * The space will be released later when all options are confirmed
1305 * to be correct
1306 */
1307 sbi->s_qf_names[qtype] = NULL;
1308 return 1;
1309}
1310#endif
1311
2b2d6d01 1312static int parse_options(char *options, struct super_block *sb,
c3191067 1313 unsigned long *journal_devnum,
b3881f74 1314 unsigned int *journal_ioprio,
2b2d6d01 1315 ext4_fsblk_t *n_blocks_count, int is_remount)
ac27a0ec 1316{
617ba13b 1317 struct ext4_sb_info *sbi = EXT4_SB(sb);
2b2d6d01 1318 char *p;
ac27a0ec
DK		 1319 substring_t args[MAX_OPT_ARGS];
1320 int data_opt = 0;
1321 int option;
1322#ifdef CONFIG_QUOTA
56c50f11 1323 int qfmt;
ac27a0ec
DK		 1324#endif
1325
1326 if (!options)
1327 return 1;
1328
2b2d6d01 1329 while ((p = strsep(&options, ",")) != NULL) {
ac27a0ec
DK		 1330 int token;
1331 if (!*p)
1332 continue;
1333
15121c18
ES		 1334 /*
1335 * Initialize args struct so we know whether arg was
1336 * found; some options take optional arguments.
1337 */
1338 args[0].to = args[0].from = 0;
ac27a0ec
DK		 1339 token = match_token(p, tokens, args);
1340 switch (token) {
1341 case Opt_bsd_df:
437ca0fd 1342 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
2b2d6d01 1343 clear_opt(sbi->s_mount_opt, MINIX_DF);
ac27a0ec
DK		 1344 break;
1345 case Opt_minix_df:
437ca0fd 1346 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
2b2d6d01 1347 set_opt(sbi->s_mount_opt, MINIX_DF);
437ca0fd 1348
ac27a0ec
DK		 1349 break;
1350 case Opt_grpid:
437ca0fd 1351 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
2b2d6d01 1352 set_opt(sbi->s_mount_opt, GRPID);
437ca0fd 1353
ac27a0ec
DK		 1354 break;
1355 case Opt_nogrpid:
437ca0fd 1356 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
2b2d6d01 1357 clear_opt(sbi->s_mount_opt, GRPID);
437ca0fd 1358
ac27a0ec
DK		 1359 break;
1360 case Opt_resuid:
1361 if (match_int(&args[0], &option))
1362 return 0;
1363 sbi->s_resuid = option;
1364 break;
1365 case Opt_resgid:
1366 if (match_int(&args[0], &option))
1367 return 0;
1368 sbi->s_resgid = option;
1369 break;
1370 case Opt_sb:
1371 /* handled by get_sb_block() instead of here */
1372 /* *sb_block = match_int(&args[0]); */
1373 break;
1374 case Opt_err_panic:
2b2d6d01
TT		 1375 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1376 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1377 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
ac27a0ec
DK		 1378 break;
1379 case Opt_err_ro:
2b2d6d01
TT		 1380 clear_opt(sbi->s_mount_opt, ERRORS_CONT);
1381 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1382 set_opt(sbi->s_mount_opt, ERRORS_RO);
ac27a0ec
DK		 1383 break;
1384 case Opt_err_cont:
2b2d6d01
TT		 1385 clear_opt(sbi->s_mount_opt, ERRORS_RO);
1386 clear_opt(sbi->s_mount_opt, ERRORS_PANIC);
1387 set_opt(sbi->s_mount_opt, ERRORS_CONT);
ac27a0ec
DK		 1388 break;
1389 case Opt_nouid32:
2b2d6d01 1390 set_opt(sbi->s_mount_opt, NO_UID32);
ac27a0ec 1391 break;
ac27a0ec 1392 case Opt_debug:
2b2d6d01 1393 set_opt(sbi->s_mount_opt, DEBUG);
ac27a0ec
DK		 1394 break;
1395 case Opt_oldalloc:
2b2d6d01 1396 set_opt(sbi->s_mount_opt, OLDALLOC);
ac27a0ec
DK		 1397 break;
1398 case Opt_orlov:
2b2d6d01 1399 clear_opt(sbi->s_mount_opt, OLDALLOC);
ac27a0ec 1400 break;
03010a33 1401#ifdef CONFIG_EXT4_FS_XATTR
ac27a0ec 1402 case Opt_user_xattr:
2b2d6d01 1403 set_opt(sbi->s_mount_opt, XATTR_USER);
ac27a0ec
DK		 1404 break;
1405 case Opt_nouser_xattr:
2b2d6d01 1406 clear_opt(sbi->s_mount_opt, XATTR_USER);
ac27a0ec
DK		 1407 break;
1408#else
1409 case Opt_user_xattr:
1410 case Opt_nouser_xattr:
b31e1552 1411 ext4_msg(sb, KERN_ERR, "(no)user_xattr options not supported");
ac27a0ec
DK		 1412 break;
1413#endif
03010a33 1414#ifdef CONFIG_EXT4_FS_POSIX_ACL
ac27a0ec
DK		 1415 case Opt_acl:
1416 set_opt(sbi->s_mount_opt, POSIX_ACL);
1417 break;
1418 case Opt_noacl:
1419 clear_opt(sbi->s_mount_opt, POSIX_ACL);
1420 break;
1421#else
1422 case Opt_acl:
1423 case Opt_noacl:
b31e1552 1424 ext4_msg(sb, KERN_ERR, "(no)acl options not supported");
ac27a0ec
DK		 1425 break;
1426#endif
ac27a0ec
DK		 1427 case Opt_journal_update:
1428 /* @@@ FIXME */
1429 /* Eventually we will want to be able to create
1430 a journal file here. For now, only allow the
1431 user to specify an existing inode to be the
1432 journal file. */
1433 if (is_remount) {
b31e1552
ES		 1434 ext4_msg(sb, KERN_ERR,
1435 "Cannot specify journal on remount");
ac27a0ec
DK		 1436 return 0;
1437 }
2b2d6d01 1438 set_opt(sbi->s_mount_opt, UPDATE_JOURNAL);
ac27a0ec 1439 break;
ac27a0ec
DK		 1440 case Opt_journal_dev:
1441 if (is_remount) {
b31e1552
ES		 1442 ext4_msg(sb, KERN_ERR,
1443 "Cannot specify journal on remount");
ac27a0ec
DK		 1444 return 0;
1445 }
1446 if (match_int(&args[0], &option))
1447 return 0;
1448 *journal_devnum = option;
1449 break;
818d276c 1450 case Opt_journal_checksum:
d4da6c9c
LT		 1451 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
1452 break;
818d276c
GS		 1453 case Opt_journal_async_commit:
1454 set_opt(sbi->s_mount_opt, JOURNAL_ASYNC_COMMIT);
d4da6c9c 1455 set_opt(sbi->s_mount_opt, JOURNAL_CHECKSUM);
818d276c 1456 break;
ac27a0ec 1457 case Opt_noload:
2b2d6d01 1458 set_opt(sbi->s_mount_opt, NOLOAD);
ac27a0ec
DK		 1459 break;
1460 case Opt_commit:
1461 if (match_int(&args[0], &option))
1462 return 0;
1463 if (option < 0)
1464 return 0;
1465 if (option == 0)
cd02ff0b 1466 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
ac27a0ec
DK		 1467 sbi->s_commit_interval = HZ * option;
1468 break;
30773840
TT		 1469 case Opt_max_batch_time:
1470 if (match_int(&args[0], &option))
1471 return 0;
1472 if (option < 0)
1473 return 0;
1474 if (option == 0)
1475 option = EXT4_DEF_MAX_BATCH_TIME;
1476 sbi->s_max_batch_time = option;
1477 break;
1478 case Opt_min_batch_time:
1479 if (match_int(&args[0], &option))
1480 return 0;
1481 if (option < 0)
1482 return 0;
1483 sbi->s_min_batch_time = option;
1484 break;
ac27a0ec 1485 case Opt_data_journal:
617ba13b 1486 data_opt = EXT4_MOUNT_JOURNAL_DATA;
ac27a0ec
DK		 1487 goto datacheck;
1488 case Opt_data_ordered:
617ba13b 1489 data_opt = EXT4_MOUNT_ORDERED_DATA;
ac27a0ec
DK		 1490 goto datacheck;
1491 case Opt_data_writeback:
617ba13b 1492 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
ac27a0ec
DK		 1493 datacheck:
1494 if (is_remount) {
482a7425 1495 if (test_opt(sb, DATA_FLAGS) != data_opt) {
b31e1552
ES		 1496 ext4_msg(sb, KERN_ERR,
1497 "Cannot change data mode on remount");
ac27a0ec
DK		 1498 return 0;
1499 }
1500 } else {
482a7425 1501 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
ac27a0ec
DK		 1502 sbi->s_mount_opt |= data_opt;
1503 }
1504 break;
5bf5683a
HK		 1505 case Opt_data_err_abort:
1506 set_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1507 break;
1508 case Opt_data_err_ignore:
1509 clear_opt(sbi->s_mount_opt, DATA_ERR_ABORT);
1510 break;
ac27a0ec
DK		 1511#ifdef CONFIG_QUOTA
1512 case Opt_usrjquota:
56c50f11 1513 if (!set_qf_name(sb, USRQUOTA, &args[0]))
ac27a0ec 1514 return 0;
56c50f11
DM		 1515 break;
1516 case Opt_grpjquota:
1517 if (!set_qf_name(sb, GRPQUOTA, &args[0]))
ac27a0ec 1518 return 0;
ac27a0ec
DK		 1519 break;
1520 case Opt_offusrjquota:
56c50f11
DM		 1521 if (!clear_qf_name(sb, USRQUOTA))
1522 return 0;
1523 break;
ac27a0ec 1524 case Opt_offgrpjquota:
56c50f11 1525 if (!clear_qf_name(sb, GRPQUOTA))
ac27a0ec 1526 return 0;
ac27a0ec 1527 break;
56c50f11 1528
ac27a0ec 1529 case Opt_jqfmt_vfsold:
dfc5d03f
JK		 1530 qfmt = QFMT_VFS_OLD;
1531 goto set_qf_format;
ac27a0ec 1532 case Opt_jqfmt_vfsv0:
dfc5d03f 1533 qfmt = QFMT_VFS_V0;
5a20bdfc
JK		 1534 goto set_qf_format;
1535 case Opt_jqfmt_vfsv1:
1536 qfmt = QFMT_VFS_V1;
dfc5d03f 1537set_qf_format:
17bd13b3 1538 if (sb_any_quota_loaded(sb) &&
dfc5d03f 1539 sbi->s_jquota_fmt != qfmt) {
b31e1552 1540 ext4_msg(sb, KERN_ERR, "Cannot change "
dfc5d03f 1541 "journaled quota options when "
b31e1552 1542 "quota turned on");
dfc5d03f
JK		 1543 return 0;
1544 }
1545 sbi->s_jquota_fmt = qfmt;
ac27a0ec
DK		 1546 break;
1547 case Opt_quota:
1548 case Opt_usrquota:
1549 set_opt(sbi->s_mount_opt, QUOTA);
1550 set_opt(sbi->s_mount_opt, USRQUOTA);
1551 break;
1552 case Opt_grpquota:
1553 set_opt(sbi->s_mount_opt, QUOTA);
1554 set_opt(sbi->s_mount_opt, GRPQUOTA);
1555 break;
1556 case Opt_noquota:
17bd13b3 1557 if (sb_any_quota_loaded(sb)) {
b31e1552
ES		 1558 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1559 "options when quota turned on");
ac27a0ec
DK		 1560 return 0;
1561 }
1562 clear_opt(sbi->s_mount_opt, QUOTA);
1563 clear_opt(sbi->s_mount_opt, USRQUOTA);
1564 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1565 break;
1566#else
1567 case Opt_quota:
1568 case Opt_usrquota:
1569 case Opt_grpquota:
b31e1552
ES		 1570 ext4_msg(sb, KERN_ERR,
1571 "quota options not supported");
cd59e7b9 1572 break;
ac27a0ec
DK		 1573 case Opt_usrjquota:
1574 case Opt_grpjquota:
1575 case Opt_offusrjquota:
1576 case Opt_offgrpjquota:
1577 case Opt_jqfmt_vfsold:
1578 case Opt_jqfmt_vfsv0:
5a20bdfc 1579 case Opt_jqfmt_vfsv1:
b31e1552
ES		 1580 ext4_msg(sb, KERN_ERR,
1581 "journaled quota options not supported");
ac27a0ec
DK		 1582 break;
1583 case Opt_noquota:
1584 break;
1585#endif
1586 case Opt_abort:
4ab2f15b 1587 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
ac27a0ec 1588 break;
06705bff
TT		 1589 case Opt_nobarrier:
1590 clear_opt(sbi->s_mount_opt, BARRIER);
1591 break;
ac27a0ec 1592 case Opt_barrier:
15121c18
ES		 1593 if (args[0].from) {
1594 if (match_int(&args[0], &option))
1595 return 0;
1596 } else
1597 option = 1; /* No argument, default to 1 */
ac27a0ec
DK		 1598 if (option)
1599 set_opt(sbi->s_mount_opt, BARRIER);
1600 else
1601 clear_opt(sbi->s_mount_opt, BARRIER);
1602 break;
1603 case Opt_ignore:
1604 break;
1605 case Opt_resize:
1606 if (!is_remount) {
b31e1552
ES		 1607 ext4_msg(sb, KERN_ERR,
1608 "resize option only available "
1609 "for remount");
ac27a0ec
DK		 1610 return 0;
1611 }
1612 if (match_int(&args[0], &option) != 0)
1613 return 0;
1614 *n_blocks_count = option;
1615 break;
1616 case Opt_nobh:
1617 set_opt(sbi->s_mount_opt, NOBH);
1618 break;
1619 case Opt_bh:
1620 clear_opt(sbi->s_mount_opt, NOBH);
1621 break;
25ec56b5
JNC		 1622 case Opt_i_version:
1623 set_opt(sbi->s_mount_opt, I_VERSION);
1624 sb->s_flags |= MS_I_VERSION;
1625 break;
dd919b98
AK		 1626 case Opt_nodelalloc:
1627 clear_opt(sbi->s_mount_opt, DELALLOC);
1628 break;
c9de560d
AT		 1629 case Opt_stripe:
1630 if (match_int(&args[0], &option))
1631 return 0;
1632 if (option < 0)
1633 return 0;
1634 sbi->s_stripe = option;
1635 break;
64769240
AT		 1636 case Opt_delalloc:
1637 set_opt(sbi->s_mount_opt, DELALLOC);
1638 break;
6fd058f7
TT		 1639 case Opt_block_validity:
1640 set_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1641 break;
1642 case Opt_noblock_validity:
1643 clear_opt(sbi->s_mount_opt, BLOCK_VALIDITY);
1644 break;
240799cd
TT		 1645 case Opt_inode_readahead_blks:
1646 if (match_int(&args[0], &option))
1647 return 0;
1648 if (option < 0 || option > (1 << 30))
1649 return 0;
f7c43950 1650 if (!is_power_of_2(option)) {
b31e1552
ES		 1651 ext4_msg(sb, KERN_ERR,
1652 "EXT4-fs: inode_readahead_blks"
1653 " must be a power of 2");
3197ebdb
TT		 1654 return 0;
1655 }
240799cd
TT		 1656 sbi->s_inode_readahead_blks = option;
1657 break;
b3881f74
TT		 1658 case Opt_journal_ioprio:
1659 if (match_int(&args[0], &option))
1660 return 0;
1661 if (option < 0 || option > 7)
1662 break;
1663 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1664 option);
1665 break;
06705bff
TT		 1666 case Opt_noauto_da_alloc:
1667 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1668 break;
afd4672d 1669 case Opt_auto_da_alloc:
15121c18
ES		 1670 if (args[0].from) {
1671 if (match_int(&args[0], &option))
1672 return 0;
1673 } else
1674 option = 1; /* No argument, default to 1 */
afd4672d
TT		 1675 if (option)
1676 clear_opt(sbi->s_mount_opt, NO_AUTO_DA_ALLOC);
1677 else
1678 set_opt(sbi->s_mount_opt,NO_AUTO_DA_ALLOC);
1679 break;
5328e635
ES		 1680 case Opt_discard:
1681 set_opt(sbi->s_mount_opt, DISCARD);
1682 break;
1683 case Opt_nodiscard:
1684 clear_opt(sbi->s_mount_opt, DISCARD);
1685 break;
744692dc
JZ		 1686 case Opt_dioread_nolock:
1687 set_opt(sbi->s_mount_opt, DIOREAD_NOLOCK);
1688 break;
1689 case Opt_dioread_lock:
1690 clear_opt(sbi->s_mount_opt, DIOREAD_NOLOCK);
1691 break;
ac27a0ec 1692 default:
b31e1552
ES		 1693 ext4_msg(sb, KERN_ERR,
1694 "Unrecognized mount option \"%s\" "
1695 "or missing value", p);
ac27a0ec
DK		 1696 return 0;
1697 }
1698 }
1699#ifdef CONFIG_QUOTA
1700 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
482a7425 1701 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
ac27a0ec
DK		 1702 clear_opt(sbi->s_mount_opt, USRQUOTA);
1703
482a7425 1704 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
ac27a0ec
DK		 1705 clear_opt(sbi->s_mount_opt, GRPQUOTA);
1706
56c50f11 1707 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
b31e1552
ES		 1708 ext4_msg(sb, KERN_ERR, "old and new quota "
1709 "format mixing");
ac27a0ec
DK		 1710 return 0;
1711 }
1712
1713 if (!sbi->s_jquota_fmt) {
b31e1552
ES		 1714 ext4_msg(sb, KERN_ERR, "journaled quota format "
1715 "not specified");
ac27a0ec
DK		 1716 return 0;
1717 }
1718 } else {
1719 if (sbi->s_jquota_fmt) {
b31e1552 1720 ext4_msg(sb, KERN_ERR, "journaled quota format "
2c8be6b2 1721 "specified with no journaling "
b31e1552 1722 "enabled");
ac27a0ec
DK		 1723 return 0;
1724 }
1725 }
1726#endif
1727 return 1;
1728}
1729
617ba13b 1730static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 1731 int read_only)
1732{
617ba13b 1733 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 1734 int res = 0;
1735
617ba13b 1736 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
b31e1552
ES		 1737 ext4_msg(sb, KERN_ERR, "revision level too high, "
1738 "forcing read-only mode");
ac27a0ec
DK		 1739 res = MS_RDONLY;
1740 }
1741 if (read_only)
1742 return res;
617ba13b 1743 if (!(sbi->s_mount_state & EXT4_VALID_FS))
b31e1552
ES		 1744 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1745 "running e2fsck is recommended");
617ba13b 1746 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
b31e1552
ES		 1747 ext4_msg(sb, KERN_WARNING,
1748 "warning: mounting fs with errors, "
1749 "running e2fsck is recommended");
ac27a0ec
DK		 1750 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1751 le16_to_cpu(es->s_mnt_count) >=
1752 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
b31e1552
ES		 1753 ext4_msg(sb, KERN_WARNING,
1754 "warning: maximal mount count reached, "
1755 "running e2fsck is recommended");
ac27a0ec
DK		 1756 else if (le32_to_cpu(es->s_checkinterval) &&
1757 (le32_to_cpu(es->s_lastcheck) +
1758 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
b31e1552
ES		 1759 ext4_msg(sb, KERN_WARNING,
1760 "warning: checktime reached, "
1761 "running e2fsck is recommended");
0b8e58a1 1762 if (!sbi->s_journal)
0390131b 1763 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec 1764 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 1765 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 1766 le16_add_cpu(&es->s_mnt_count, 1);
ac27a0ec 1767 es->s_mtime = cpu_to_le32(get_seconds());
617ba13b 1768 ext4_update_dynamic_rev(sb);
0390131b
FM		 1769 if (sbi->s_journal)
1770 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 1771
e2d67052 1772 ext4_commit_super(sb, 1);
ac27a0ec 1773 if (test_opt(sb, DEBUG))
a9df9a49 1774 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
7f4520cc 1775 "bpg=%lu, ipg=%lu, mo=%04x]\n",
ac27a0ec
DK		 1776 sb->s_blocksize,
1777 sbi->s_groups_count,
617ba13b
MC		 1778 EXT4_BLOCKS_PER_GROUP(sb),
1779 EXT4_INODES_PER_GROUP(sb),
ac27a0ec
DK		 1780 sbi->s_mount_opt);
1781
ac27a0ec
DK		 1782 return res;
1783}
1784
772cb7c8
JS		 1785static int ext4_fill_flex_info(struct super_block *sb)
1786{
1787 struct ext4_sb_info *sbi = EXT4_SB(sb);
1788 struct ext4_group_desc *gdp = NULL;
772cb7c8
JS		 1789 ext4_group_t flex_group_count;
1790 ext4_group_t flex_group;
1791 int groups_per_flex = 0;
c5ca7c76 1792 size_t size;
772cb7c8
JS		 1793 int i;
1794
503358ae
TT		 1795 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1796 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1797
1798 if (groups_per_flex < 2) {
772cb7c8
JS		 1799 sbi->s_log_groups_per_flex = 0;
1800 return 1;
1801 }
1802
c62a11fd
FB		 1803 /* We allocate both existing and potentially added groups */
1804 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
d94e99a6
AK		 1805 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1806 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
c5ca7c76
TT		 1807 size = flex_group_count * sizeof(struct flex_groups);
1808 sbi->s_flex_groups = kzalloc(size, GFP_KERNEL);
1809 if (sbi->s_flex_groups == NULL) {
1810 sbi->s_flex_groups = vmalloc(size);
1811 if (sbi->s_flex_groups)
1812 memset(sbi->s_flex_groups, 0, size);
1813 }
772cb7c8 1814 if (sbi->s_flex_groups == NULL) {
b31e1552
ES		 1815 ext4_msg(sb, KERN_ERR, "not enough memory for "
1816 "%u flex groups", flex_group_count);
772cb7c8
JS		 1817 goto failed;
1818 }
772cb7c8 1819
772cb7c8 1820 for (i = 0; i < sbi->s_groups_count; i++) {
88b6edd1 1821 gdp = ext4_get_group_desc(sb, i, NULL);
772cb7c8
JS		 1822
1823 flex_group = ext4_flex_group(sbi, i);
7ad9bb65
TT		 1824 atomic_add(ext4_free_inodes_count(sb, gdp),
1825 &sbi->s_flex_groups[flex_group].free_inodes);
1826 atomic_add(ext4_free_blks_count(sb, gdp),
1827 &sbi->s_flex_groups[flex_group].free_blocks);
1828 atomic_add(ext4_used_dirs_count(sb, gdp),
1829 &sbi->s_flex_groups[flex_group].used_dirs);
772cb7c8
JS		 1830 }
1831
1832 return 1;
1833failed:
1834 return 0;
1835}
1836
717d50e4
AD		 1837__le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1838 struct ext4_group_desc *gdp)
1839{
1840 __u16 crc = 0;
1841
1842 if (sbi->s_es->s_feature_ro_compat &
1843 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1844 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1845 __le32 le_group = cpu_to_le32(block_group);
1846
1847 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1848 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1849 crc = crc16(crc, (__u8 *)gdp, offset);
1850 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1851 /* for checksum of struct ext4_group_desc do the rest...*/
1852 if ((sbi->s_es->s_feature_incompat &
1853 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1854 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1855 crc = crc16(crc, (__u8 *)gdp + offset,
1856 le16_to_cpu(sbi->s_es->s_desc_size) -
1857 offset);
1858 }
1859
1860 return cpu_to_le16(crc);
1861}
1862
1863int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1864 struct ext4_group_desc *gdp)
1865{
1866 if ((sbi->s_es->s_feature_ro_compat &
1867 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
1868 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
1869 return 0;
1870
1871 return 1;
1872}
1873
ac27a0ec 1874/* Called at mount-time, super-block is locked */
197cd65a 1875static int ext4_check_descriptors(struct super_block *sb)
ac27a0ec 1876{
617ba13b
MC		 1877 struct ext4_sb_info *sbi = EXT4_SB(sb);
1878 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
1879 ext4_fsblk_t last_block;
bd81d8ee
LV		 1880 ext4_fsblk_t block_bitmap;
1881 ext4_fsblk_t inode_bitmap;
1882 ext4_fsblk_t inode_table;
ce421581 1883 int flexbg_flag = 0;
fd2d4291 1884 ext4_group_t i;
ac27a0ec 1885
ce421581
JS		 1886 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
1887 flexbg_flag = 1;
1888
af5bc92d 1889 ext4_debug("Checking group descriptors");
ac27a0ec 1890
197cd65a
AM		 1891 for (i = 0; i < sbi->s_groups_count; i++) {
1892 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
1893
ce421581 1894 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 1895 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 1896 else
1897 last_block = first_block +
617ba13b 1898 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 1899
8fadc143 1900 block_bitmap = ext4_block_bitmap(sb, gdp);
2b2d6d01 1901 if (block_bitmap < first_block || block_bitmap > last_block) {
b31e1552 1902 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 1903 "Block bitmap for group %u not in group "
b31e1552 1904 "(block %llu)!", i, block_bitmap);
ac27a0ec
DK		 1905 return 0;
1906 }
8fadc143 1907 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2b2d6d01 1908 if (inode_bitmap < first_block || inode_bitmap > last_block) {
b31e1552 1909 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 1910 "Inode bitmap for group %u not in group "
b31e1552 1911 "(block %llu)!", i, inode_bitmap);
ac27a0ec
DK		 1912 return 0;
1913 }
8fadc143 1914 inode_table = ext4_inode_table(sb, gdp);
bd81d8ee 1915 if (inode_table < first_block ||
2b2d6d01 1916 inode_table + sbi->s_itb_per_group - 1 > last_block) {
b31e1552 1917 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 1918 "Inode table for group %u not in group "
b31e1552 1919 "(block %llu)!", i, inode_table);
ac27a0ec
DK		 1920 return 0;
1921 }
955ce5f5 1922 ext4_lock_group(sb, i);
717d50e4 1923 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
b31e1552
ES		 1924 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
1925 "Checksum for group %u failed (%u!=%u)",
1926 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
1927 gdp)), le16_to_cpu(gdp->bg_checksum));
7ee1ec4c 1928 if (!(sb->s_flags & MS_RDONLY)) {
955ce5f5 1929 ext4_unlock_group(sb, i);
8a266467 1930 return 0;
7ee1ec4c 1931 }
717d50e4 1932 }
955ce5f5 1933 ext4_unlock_group(sb, i);
ce421581
JS		 1934 if (!flexbg_flag)
1935 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec
DK		 1936 }
1937
bd81d8ee 1938 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
0b8e58a1 1939 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
ac27a0ec
DK		 1940 return 1;
1941}
1942
617ba13b 1943/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 1944 * the superblock) which were deleted from all directories, but held open by
1945 * a process at the time of a crash. We walk the list and try to delete these
1946 * inodes at recovery time (only with a read-write filesystem).
1947 *
1948 * In order to keep the orphan inode chain consistent during traversal (in
1949 * case of crash during recovery), we link each inode into the superblock
1950 * orphan list_head and handle it the same way as an inode deletion during
1951 * normal operation (which journals the operations for us).
1952 *
1953 * We only do an iget() and an iput() on each inode, which is very safe if we
1954 * accidentally point at an in-use or already deleted inode. The worst that
1955 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 1956 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 1957 * e2fsck was run on this filesystem, and it must have already done the orphan
1958 * inode cleanup for us, so we can safely abort without any further action.
1959 */
2b2d6d01
TT		 1960static void ext4_orphan_cleanup(struct super_block *sb,
1961 struct ext4_super_block *es)
ac27a0ec
DK		 1962{
1963 unsigned int s_flags = sb->s_flags;
1964 int nr_orphans = 0, nr_truncates = 0;
1965#ifdef CONFIG_QUOTA
1966 int i;
1967#endif
1968 if (!es->s_last_orphan) {
1969 jbd_debug(4, "no orphan inodes to clean up\n");
1970 return;
1971 }
1972
a8f48a95 1973 if (bdev_read_only(sb->s_bdev)) {
b31e1552
ES		 1974 ext4_msg(sb, KERN_ERR, "write access "
1975 "unavailable, skipping orphan cleanup");
a8f48a95
ES		 1976 return;
1977 }
1978
617ba13b 1979 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
ac27a0ec
DK		 1980 if (es->s_last_orphan)
1981 jbd_debug(1, "Errors on filesystem, "
1982 "clearing orphan list.\n");
1983 es->s_last_orphan = 0;
1984 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
1985 return;
1986 }
1987
1988 if (s_flags & MS_RDONLY) {
b31e1552 1989 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
ac27a0ec
DK		 1990 sb->s_flags &= ~MS_RDONLY;
1991 }
1992#ifdef CONFIG_QUOTA
1993 /* Needed for iput() to work correctly and not trash data */
1994 sb->s_flags |= MS_ACTIVE;
1995 /* Turn on quotas so that they are updated correctly */
1996 for (i = 0; i < MAXQUOTAS; i++) {
617ba13b
MC		 1997 if (EXT4_SB(sb)->s_qf_names[i]) {
1998 int ret = ext4_quota_on_mount(sb, i);
ac27a0ec 1999 if (ret < 0)
b31e1552
ES		 2000 ext4_msg(sb, KERN_ERR,
2001 "Cannot turn on journaled "
2002 "quota: error %d", ret);
ac27a0ec
DK		 2003 }
2004 }
2005#endif
2006
2007 while (es->s_last_orphan) {
2008 struct inode *inode;
2009
97bd42b9
JB		 2010 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2011 if (IS_ERR(inode)) {
ac27a0ec
DK		 2012 es->s_last_orphan = 0;
2013 break;
2014 }
2015
617ba13b 2016 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
a269eb18 2017 vfs_dq_init(inode);
ac27a0ec 2018 if (inode->i_nlink) {
b31e1552
ES		 2019 ext4_msg(sb, KERN_DEBUG,
2020 "%s: truncating inode %lu to %lld bytes",
46e665e9 2021 __func__, inode->i_ino, inode->i_size);
e5f8eab8 2022 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
ac27a0ec 2023 inode->i_ino, inode->i_size);
617ba13b 2024 ext4_truncate(inode);
ac27a0ec
DK		 2025 nr_truncates++;
2026 } else {
b31e1552
ES		 2027 ext4_msg(sb, KERN_DEBUG,
2028 "%s: deleting unreferenced inode %lu",
46e665e9 2029 __func__, inode->i_ino);
ac27a0ec
DK		 2030 jbd_debug(2, "deleting unreferenced inode %lu\n",
2031 inode->i_ino);
2032 nr_orphans++;
2033 }
2034 iput(inode); /* The delete magic happens here! */
2035 }
2036
2b2d6d01 2037#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
ac27a0ec
DK		 2038
2039 if (nr_orphans)
b31e1552
ES		 2040 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2041 PLURAL(nr_orphans));
ac27a0ec 2042 if (nr_truncates)
b31e1552
ES		 2043 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2044 PLURAL(nr_truncates));
ac27a0ec
DK		 2045#ifdef CONFIG_QUOTA
2046 /* Turn quotas off */
2047 for (i = 0; i < MAXQUOTAS; i++) {
2048 if (sb_dqopt(sb)->files[i])
6f28e087 2049 vfs_quota_off(sb, i, 0);
ac27a0ec
DK		 2050 }
2051#endif
2052 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2053}
0b8e58a1 2054
cd2291a4
ES		 2055/*
2056 * Maximal extent format file size.
2057 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2058 * extent format containers, within a sector_t, and within i_blocks
2059 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2060 * so that won't be a limiting factor.
2061 *
2062 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2063 */
f287a1a5 2064static loff_t ext4_max_size(int blkbits, int has_huge_files)
cd2291a4
ES		 2065{
2066 loff_t res;
2067 loff_t upper_limit = MAX_LFS_FILESIZE;
2068
2069 /* small i_blocks in vfs inode? */
f287a1a5 2070 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
cd2291a4 2071 /*
90c699a9 2072 * CONFIG_LBDAF is not enabled implies the inode
cd2291a4
ES		 2073 * i_block represent total blocks in 512 bytes
2074 * 32 == size of vfs inode i_blocks * 8
2075 */
2076 upper_limit = (1LL << 32) - 1;
2077
2078 /* total blocks in file system block size */
2079 upper_limit >>= (blkbits - 9);
2080 upper_limit <<= blkbits;
2081 }
2082
2083 /* 32-bit extent-start container, ee_block */
2084 res = 1LL << 32;
2085 res <<= blkbits;
2086 res -= 1;
2087
2088 /* Sanity check against vm- & vfs- imposed limits */
2089 if (res > upper_limit)
2090 res = upper_limit;
2091
2092 return res;
2093}
ac27a0ec 2094
ac27a0ec 2095/*
cd2291a4 2096 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 2097 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2098 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 2099 */
f287a1a5 2100static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
ac27a0ec 2101{
617ba13b 2102 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 2103 int meta_blocks;
2104 loff_t upper_limit;
0b8e58a1
AD		 2105 /* This is calculated to be the largest file size for a dense, block
2106 * mapped file such that the file's total number of 512-byte sectors,
2107 * including data and all indirect blocks, does not exceed (2^48 - 1).
2108 *
2109 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2110 * number of 512-byte sectors of the file.
0fc1b451
AK		 2111 */
2112
f287a1a5 2113 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
0fc1b451 2114 /*
90c699a9 2115 * !has_huge_files or CONFIG_LBDAF not enabled implies that
0b8e58a1
AD		 2116 * the inode i_block field represents total file blocks in
2117 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
0fc1b451
AK		 2118 */
2119 upper_limit = (1LL << 32) - 1;
2120
2121 /* total blocks in file system block size */
2122 upper_limit >>= (bits - 9);
2123
2124 } else {
8180a562
AK		 2125 /*
2126 * We use 48 bit ext4_inode i_blocks
2127 * With EXT4_HUGE_FILE_FL set the i_blocks
2128 * represent total number of blocks in
2129 * file system block size
2130 */
0fc1b451
AK		 2131 upper_limit = (1LL << 48) - 1;
2132
0fc1b451
AK		 2133 }
2134
2135 /* indirect blocks */
2136 meta_blocks = 1;
2137 /* double indirect blocks */
2138 meta_blocks += 1 + (1LL << (bits-2));
2139 /* tripple indirect blocks */
2140 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2141
2142 upper_limit -= meta_blocks;
2143 upper_limit <<= bits;
ac27a0ec
DK		 2144
2145 res += 1LL << (bits-2);
2146 res += 1LL << (2*(bits-2));
2147 res += 1LL << (3*(bits-2));
2148 res <<= bits;
2149 if (res > upper_limit)
2150 res = upper_limit;
0fc1b451
AK		 2151
2152 if (res > MAX_LFS_FILESIZE)
2153 res = MAX_LFS_FILESIZE;
2154
ac27a0ec
DK		 2155 return res;
2156}
2157
617ba13b 2158static ext4_fsblk_t descriptor_loc(struct super_block *sb,
0b8e58a1 2159 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 2160{
617ba13b 2161 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 2162 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 2163 int has_super = 0;
2164
2165 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2166
617ba13b 2167 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
ac27a0ec 2168 nr < first_meta_bg)
70bbb3e0 2169 return logical_sb_block + nr + 1;
ac27a0ec 2170 bg = sbi->s_desc_per_block * nr;
617ba13b 2171 if (ext4_bg_has_super(sb, bg))
ac27a0ec 2172 has_super = 1;
0b8e58a1 2173
617ba13b 2174 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 2175}
2176
c9de560d
AT		 2177/**
2178 * ext4_get_stripe_size: Get the stripe size.
2179 * @sbi: In memory super block info
2180 *
2181 * If we have specified it via mount option, then
2182 * use the mount option value. If the value specified at mount time is
2183 * greater than the blocks per group use the super block value.
2184 * If the super block value is greater than blocks per group return 0.
2185 * Allocator needs it be less than blocks per group.
2186 *
2187 */
2188static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2189{
2190 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2191 unsigned long stripe_width =
2192 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2193
2194 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2195 return sbi->s_stripe;
2196
2197 if (stripe_width <= sbi->s_blocks_per_group)
2198 return stripe_width;
2199
2200 if (stride <= sbi->s_blocks_per_group)
2201 return stride;
2202
2203 return 0;
2204}
ac27a0ec 2205
3197ebdb
TT		 2206/* sysfs supprt */
2207
2208struct ext4_attr {
2209 struct attribute attr;
2210 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2211 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2212 const char *, size_t);
2213 int offset;
2214};
2215
2216static int parse_strtoul(const char *buf,
2217 unsigned long max, unsigned long *value)
2218{
2219 char *endp;
2220
e7d2860b
AGR		 2221 *value = simple_strtoul(skip_spaces(buf), &endp, 0);
2222 endp = skip_spaces(endp);
3197ebdb
TT		 2223 if (*endp || *value > max)
2224 return -EINVAL;
2225
2226 return 0;
2227}
2228
2229static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2230 struct ext4_sb_info *sbi,
2231 char *buf)
2232{
2233 return snprintf(buf, PAGE_SIZE, "%llu\n",
2234 (s64) percpu_counter_sum(&sbi->s_dirtyblocks_counter));
2235}
2236
2237static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2238 struct ext4_sb_info *sbi, char *buf)
2239{
2240 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2241
2242 return snprintf(buf, PAGE_SIZE, "%lu\n",
2243 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2244 sbi->s_sectors_written_start) >> 1);
2245}
2246
2247static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2248 struct ext4_sb_info *sbi, char *buf)
2249{
2250 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2251
2252 return snprintf(buf, PAGE_SIZE, "%llu\n",
a6b43e38 2253 (unsigned long long)(sbi->s_kbytes_written +
3197ebdb 2254 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
a6b43e38 2255 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
3197ebdb
TT		 2256}
2257
2258static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2259 struct ext4_sb_info *sbi,
2260 const char *buf, size_t count)
2261{
2262 unsigned long t;
2263
2264 if (parse_strtoul(buf, 0x40000000, &t))
2265 return -EINVAL;
2266
f7c43950 2267 if (!is_power_of_2(t))
3197ebdb
TT		 2268 return -EINVAL;
2269
2270 sbi->s_inode_readahead_blks = t;
2271 return count;
2272}
2273
2274static ssize_t sbi_ui_show(struct ext4_attr *a,
0b8e58a1 2275 struct ext4_sb_info *sbi, char *buf)
3197ebdb
TT		 2276{
2277 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2278
2279 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2280}
2281
2282static ssize_t sbi_ui_store(struct ext4_attr *a,
2283 struct ext4_sb_info *sbi,
2284 const char *buf, size_t count)
2285{
2286 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2287 unsigned long t;
2288
2289 if (parse_strtoul(buf, 0xffffffff, &t))
2290 return -EINVAL;
2291 *ui = t;
2292 return count;
2293}
2294
2295#define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2296static struct ext4_attr ext4_attr_##_name = { \
2297 .attr = {.name = __stringify(_name), .mode = _mode }, \
2298 .show = _show, \
2299 .store = _store, \
2300 .offset = offsetof(struct ext4_sb_info, _elname), \
2301}
2302#define EXT4_ATTR(name, mode, show, store) \
2303static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2304
2305#define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2306#define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2307#define EXT4_RW_ATTR_SBI_UI(name, elname) \
2308 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2309#define ATTR_LIST(name) &ext4_attr_##name.attr
2310
2311EXT4_RO_ATTR(delayed_allocation_blocks);
2312EXT4_RO_ATTR(session_write_kbytes);
2313EXT4_RO_ATTR(lifetime_write_kbytes);
2314EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2315 inode_readahead_blks_store, s_inode_readahead_blks);
11013911 2316EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
3197ebdb
TT		 2317EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2318EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2319EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2320EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2321EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2322EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
55138e0b 2323EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
3197ebdb
TT		 2324
2325static struct attribute *ext4_attrs[] = {
2326 ATTR_LIST(delayed_allocation_blocks),
2327 ATTR_LIST(session_write_kbytes),
2328 ATTR_LIST(lifetime_write_kbytes),
2329 ATTR_LIST(inode_readahead_blks),
11013911 2330 ATTR_LIST(inode_goal),
3197ebdb
TT		 2331 ATTR_LIST(mb_stats),
2332 ATTR_LIST(mb_max_to_scan),
2333 ATTR_LIST(mb_min_to_scan),
2334 ATTR_LIST(mb_order2_req),
2335 ATTR_LIST(mb_stream_req),
2336 ATTR_LIST(mb_group_prealloc),
55138e0b 2337 ATTR_LIST(max_writeback_mb_bump),
3197ebdb
TT		 2338 NULL,
2339};
2340
2341static ssize_t ext4_attr_show(struct kobject *kobj,
2342 struct attribute *attr, char *buf)
2343{
2344 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2345 s_kobj);
2346 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2347
2348 return a->show ? a->show(a, sbi, buf) : 0;
2349}
2350
2351static ssize_t ext4_attr_store(struct kobject *kobj,
2352 struct attribute *attr,
2353 const char *buf, size_t len)
2354{
2355 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2356 s_kobj);
2357 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2358
2359 return a->store ? a->store(a, sbi, buf, len) : 0;
2360}
2361
2362static void ext4_sb_release(struct kobject *kobj)
2363{
2364 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2365 s_kobj);
2366 complete(&sbi->s_kobj_unregister);
2367}
2368
2369
2370static struct sysfs_ops ext4_attr_ops = {
2371 .show = ext4_attr_show,
2372 .store = ext4_attr_store,
2373};
2374
2375static struct kobj_type ext4_ktype = {
2376 .default_attrs = ext4_attrs,
2377 .sysfs_ops = &ext4_attr_ops,
2378 .release = ext4_sb_release,
2379};
2380
a13fb1a4
ES		 2381/*
2382 * Check whether this filesystem can be mounted based on
2383 * the features present and the RDONLY/RDWR mount requested.
2384 * Returns 1 if this filesystem can be mounted as requested,
2385 * 0 if it cannot be.
2386 */
2387static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2388{
2389 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2390 ext4_msg(sb, KERN_ERR,
2391 "Couldn't mount because of "
2392 "unsupported optional features (%x)",
2393 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2394 ~EXT4_FEATURE_INCOMPAT_SUPP));
2395 return 0;
2396 }
2397
2398 if (readonly)
2399 return 1;
2400
2401 /* Check that feature set is OK for a read-write mount */
2402 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2403 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2404 "unsupported optional features (%x)",
2405 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2406 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2407 return 0;
2408 }
2409 /*
2410 * Large file size enabled file system can only be mounted
2411 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2412 */
2413 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2414 if (sizeof(blkcnt_t) < sizeof(u64)) {
2415 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2416 "cannot be mounted RDWR without "
2417 "CONFIG_LBDAF");
2418 return 0;
2419 }
2420 }
2421 return 1;
2422}
2423
2b2d6d01 2424static int ext4_fill_super(struct super_block *sb, void *data, int silent)
7477827f
AK		 2425 __releases(kernel_lock)
2426 __acquires(kernel_lock)
ac27a0ec 2427{
2b2d6d01 2428 struct buffer_head *bh;
617ba13b
MC		 2429 struct ext4_super_block *es = NULL;
2430 struct ext4_sb_info *sbi;
2431 ext4_fsblk_t block;
2432 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 2433 ext4_fsblk_t logical_sb_block;
ac27a0ec 2434 unsigned long offset = 0;
ac27a0ec
DK		 2435 unsigned long journal_devnum = 0;
2436 unsigned long def_mount_opts;
2437 struct inode *root;
9f6200bb 2438 char *cp;
0390131b 2439 const char *descr;
1d1fe1ee 2440 int ret = -EINVAL;
ac27a0ec 2441 int blocksize;
4ec11028
TT		 2442 unsigned int db_count;
2443 unsigned int i;
f287a1a5 2444 int needs_recovery, has_huge_files;
bd81d8ee 2445 __u64 blocks_count;
833f4077 2446 int err;
b3881f74 2447 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
ac27a0ec
DK		 2448
2449 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
2450 if (!sbi)
2451 return -ENOMEM;
705895b6
PE		 2452
2453 sbi->s_blockgroup_lock =
2454 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
2455 if (!sbi->s_blockgroup_lock) {
2456 kfree(sbi);
2457 return -ENOMEM;
2458 }
ac27a0ec
DK		 2459 sb->s_fs_info = sbi;
2460 sbi->s_mount_opt = 0;
617ba13b
MC		 2461 sbi->s_resuid = EXT4_DEF_RESUID;
2462 sbi->s_resgid = EXT4_DEF_RESGID;
240799cd 2463 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
d9c9bef1 2464 sbi->s_sb_block = sb_block;
afc32f7e
TT		 2465 sbi->s_sectors_written_start = part_stat_read(sb->s_bdev->bd_part,
2466 sectors[1]);
ac27a0ec
DK		 2467
2468 unlock_kernel();
2469
9f6200bb
TT		 2470 /* Cleanup superblock name */
2471 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
2472 *cp = '!';
2473
617ba13b 2474 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 2475 if (!blocksize) {
b31e1552 2476 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
ac27a0ec
DK		 2477 goto out_fail;
2478 }
2479
2480 /*
617ba13b 2481 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 2482 * block sizes. We need to calculate the offset from buffer start.
2483 */
617ba13b 2484 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 2485 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2486 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 2487 } else {
70bbb3e0 2488 logical_sb_block = sb_block;
ac27a0ec
DK		 2489 }
2490
70bbb3e0 2491 if (!(bh = sb_bread(sb, logical_sb_block))) {
b31e1552 2492 ext4_msg(sb, KERN_ERR, "unable to read superblock");
ac27a0ec
DK		 2493 goto out_fail;
2494 }
2495 /*
2496 * Note: s_es must be initialized as soon as possible because
617ba13b 2497 * some ext4 macro-instructions depend on its value
ac27a0ec 2498 */
617ba13b 2499 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
ac27a0ec
DK		 2500 sbi->s_es = es;
2501 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 2502 if (sb->s_magic != EXT4_SUPER_MAGIC)
2503 goto cantfind_ext4;
afc32f7e 2504 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
ac27a0ec
DK		 2505
2506 /* Set defaults before we parse the mount options */
2507 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
617ba13b 2508 if (def_mount_opts & EXT4_DEFM_DEBUG)
ac27a0ec 2509 set_opt(sbi->s_mount_opt, DEBUG);
437ca0fd
DM		 2510 if (def_mount_opts & EXT4_DEFM_BSDGROUPS) {
2511 ext4_msg(sb, KERN_WARNING, deprecated_msg, "bsdgroups",
2512 "2.6.38");
ac27a0ec 2513 set_opt(sbi->s_mount_opt, GRPID);
437ca0fd 2514 }
617ba13b 2515 if (def_mount_opts & EXT4_DEFM_UID16)
ac27a0ec 2516 set_opt(sbi->s_mount_opt, NO_UID32);
03010a33 2517#ifdef CONFIG_EXT4_FS_XATTR
617ba13b 2518 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
ac27a0ec 2519 set_opt(sbi->s_mount_opt, XATTR_USER);
2e7842b8 2520#endif
03010a33 2521#ifdef CONFIG_EXT4_FS_POSIX_ACL
617ba13b 2522 if (def_mount_opts & EXT4_DEFM_ACL)
ac27a0ec 2523 set_opt(sbi->s_mount_opt, POSIX_ACL);
2e7842b8 2524#endif
617ba13b 2525 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
482a7425 2526 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
617ba13b 2527 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
482a7425 2528 set_opt(sbi->s_mount_opt, ORDERED_DATA);
617ba13b 2529 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
482a7425 2530 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
617ba13b
MC		 2531
2532 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
ac27a0ec 2533 set_opt(sbi->s_mount_opt, ERRORS_PANIC);
bb4f397a 2534 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
ceea16bf 2535 set_opt(sbi->s_mount_opt, ERRORS_CONT);
bb4f397a
AK		 2536 else
2537 set_opt(sbi->s_mount_opt, ERRORS_RO);
ac27a0ec
DK		 2538
2539 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
2540 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
30773840
TT		 2541 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
2542 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
2543 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
ac27a0ec 2544
571640ca 2545 set_opt(sbi->s_mount_opt, BARRIER);
ac27a0ec 2546
dd919b98
AK		 2547 /*
2548 * enable delayed allocation by default
2549 * Use -o nodelalloc to turn it off
2550 */
2551 set_opt(sbi->s_mount_opt, DELALLOC);
2552
b3881f74
TT		 2553 if (!parse_options((char *) data, sb, &journal_devnum,
2554 &journal_ioprio, NULL, 0))
ac27a0ec
DK		 2555 goto failed_mount;
2556
2557 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 2558 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec 2559
617ba13b
MC		 2560 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
2561 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
2562 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
2563 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
b31e1552
ES		 2564 ext4_msg(sb, KERN_WARNING,
2565 "feature flags set on rev 0 fs, "
2566 "running e2fsck is recommended");
469108ff 2567
ac27a0ec
DK		 2568 /*
2569 * Check feature flags regardless of the revision level, since we
2570 * previously didn't change the revision level when setting the flags,
2571 * so there is a chance incompat flags are set on a rev 0 filesystem.
2572 */
a13fb1a4 2573 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
ac27a0ec 2574 goto failed_mount;
a13fb1a4 2575
ac27a0ec
DK		 2576 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
2577
617ba13b
MC		 2578 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
2579 blocksize > EXT4_MAX_BLOCK_SIZE) {
b31e1552
ES		 2580 ext4_msg(sb, KERN_ERR,
2581 "Unsupported filesystem blocksize %d", blocksize);
ac27a0ec
DK		 2582 goto failed_mount;
2583 }
2584
ac27a0ec 2585 if (sb->s_blocksize != blocksize) {
ce40733c
AK		 2586 /* Validate the filesystem blocksize */
2587 if (!sb_set_blocksize(sb, blocksize)) {
b31e1552 2588 ext4_msg(sb, KERN_ERR, "bad block size %d",
ce40733c 2589 blocksize);
ac27a0ec
DK		 2590 goto failed_mount;
2591 }
2592
2b2d6d01 2593 brelse(bh);
70bbb3e0
AM		 2594 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
2595 offset = do_div(logical_sb_block, blocksize);
2596 bh = sb_bread(sb, logical_sb_block);
ac27a0ec 2597 if (!bh) {
b31e1552
ES		 2598 ext4_msg(sb, KERN_ERR,
2599 "Can't read superblock on 2nd try");
ac27a0ec
DK		 2600 goto failed_mount;
2601 }
617ba13b 2602 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
ac27a0ec 2603 sbi->s_es = es;
617ba13b 2604 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
b31e1552
ES		 2605 ext4_msg(sb, KERN_ERR,
2606 "Magic mismatch, very weird!");
ac27a0ec
DK		 2607 goto failed_mount;
2608 }
2609 }
2610
a13fb1a4
ES		 2611 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
2612 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
f287a1a5
TT		 2613 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
2614 has_huge_files);
2615 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
ac27a0ec 2616
617ba13b
MC		 2617 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
2618 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
2619 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
ac27a0ec
DK		 2620 } else {
2621 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
2622 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
617ba13b 2623 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1330593e 2624 (!is_power_of_2(sbi->s_inode_size)) ||
ac27a0ec 2625 (sbi->s_inode_size > blocksize)) {
b31e1552
ES		 2626 ext4_msg(sb, KERN_ERR,
2627 "unsupported inode size: %d",
2b2d6d01 2628 sbi->s_inode_size);
ac27a0ec
DK		 2629 goto failed_mount;
2630 }
ef7f3835
KS		 2631 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
2632 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
ac27a0ec 2633 }
0b8e58a1 2634
0d1ee42f
AR		 2635 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
2636 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
8fadc143 2637 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 2638 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 2639 !is_power_of_2(sbi->s_desc_size)) {
b31e1552
ES		 2640 ext4_msg(sb, KERN_ERR,
2641 "unsupported descriptor size %lu",
0d1ee42f
AR		 2642 sbi->s_desc_size);
2643 goto failed_mount;
2644 }
2645 } else
2646 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
0b8e58a1 2647
ac27a0ec 2648 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 2649 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
b47b6f38 2650 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
617ba13b 2651 goto cantfind_ext4;
0b8e58a1 2652
617ba13b 2653 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 2654 if (sbi->s_inodes_per_block == 0)
617ba13b 2655 goto cantfind_ext4;
ac27a0ec
DK		 2656 sbi->s_itb_per_group = sbi->s_inodes_per_group /
2657 sbi->s_inodes_per_block;
0d1ee42f 2658 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 2659 sbi->s_sbh = bh;
2660 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 2661 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
2662 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
0b8e58a1 2663
2b2d6d01 2664 for (i = 0; i < 4; i++)
ac27a0ec
DK		 2665 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
2666 sbi->s_def_hash_version = es->s_def_hash_version;
f99b2589
TT		 2667 i = le32_to_cpu(es->s_flags);
2668 if (i & EXT2_FLAGS_UNSIGNED_HASH)
2669 sbi->s_hash_unsigned = 3;
2670 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
2671#ifdef __CHAR_UNSIGNED__
2672 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
2673 sbi->s_hash_unsigned = 3;
2674#else
2675 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
2676#endif
2677 sb->s_dirt = 1;
2678 }
ac27a0ec
DK		 2679
2680 if (sbi->s_blocks_per_group > blocksize * 8) {
b31e1552
ES		 2681 ext4_msg(sb, KERN_ERR,
2682 "#blocks per group too big: %lu",
2b2d6d01 2683 sbi->s_blocks_per_group);
ac27a0ec
DK		 2684 goto failed_mount;
2685 }
ac27a0ec 2686 if (sbi->s_inodes_per_group > blocksize * 8) {
b31e1552
ES		 2687 ext4_msg(sb, KERN_ERR,
2688 "#inodes per group too big: %lu",
2b2d6d01 2689 sbi->s_inodes_per_group);
ac27a0ec
DK		 2690 goto failed_mount;
2691 }
2692
bf43d84b
ES		 2693 /*
2694 * Test whether we have more sectors than will fit in sector_t,
2695 * and whether the max offset is addressable by the page cache.
2696 */
2697 if ((ext4_blocks_count(es) >
2698 (sector_t)(~0ULL) >> (sb->s_blocksize_bits - 9)) ||
2699 (ext4_blocks_count(es) >
2700 (pgoff_t)(~0ULL) >> (PAGE_CACHE_SHIFT - sb->s_blocksize_bits))) {
b31e1552 2701 ext4_msg(sb, KERN_ERR, "filesystem"
bf43d84b 2702 " too large to mount safely on this system");
ac27a0ec 2703 if (sizeof(sector_t) < 8)
90c699a9 2704 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
bf43d84b 2705 ret = -EFBIG;
ac27a0ec
DK		 2706 goto failed_mount;
2707 }
2708
617ba13b
MC		 2709 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
2710 goto cantfind_ext4;
e7c95593 2711
0f2ddca6
FTN		 2712 /* check blocks count against device size */
2713 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
2714 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
b31e1552
ES		 2715 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
2716 "exceeds size of device (%llu blocks)",
0f2ddca6
FTN		 2717 ext4_blocks_count(es), blocks_count);
2718 goto failed_mount;
2719 }
2720
0b8e58a1
AD		 2721 /*
2722 * It makes no sense for the first data block to be beyond the end
2723 * of the filesystem.
2724 */
2725 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
b31e1552
ES		 2726 ext4_msg(sb, KERN_WARNING, "bad geometry: first data"
2727 "block %u is beyond end of filesystem (%llu)",
2728 le32_to_cpu(es->s_first_data_block),
2729 ext4_blocks_count(es));
e7c95593
ES		 2730 goto failed_mount;
2731 }
bd81d8ee
LV		 2732 blocks_count = (ext4_blocks_count(es) -
2733 le32_to_cpu(es->s_first_data_block) +
2734 EXT4_BLOCKS_PER_GROUP(sb) - 1);
2735 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4ec11028 2736 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
b31e1552 2737 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
4ec11028 2738 "(block count %llu, first data block %u, "
b31e1552 2739 "blocks per group %lu)", sbi->s_groups_count,
4ec11028
TT		 2740 ext4_blocks_count(es),
2741 le32_to_cpu(es->s_first_data_block),
2742 EXT4_BLOCKS_PER_GROUP(sb));
2743 goto failed_mount;
2744 }
bd81d8ee 2745 sbi->s_groups_count = blocks_count;
fb0a387d
ES		 2746 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
2747 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
617ba13b
MC		 2748 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
2749 EXT4_DESC_PER_BLOCK(sb);
2b2d6d01 2750 sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
ac27a0ec
DK		 2751 GFP_KERNEL);
2752 if (sbi->s_group_desc == NULL) {
b31e1552 2753 ext4_msg(sb, KERN_ERR, "not enough memory");
ac27a0ec
DK		 2754 goto failed_mount;
2755 }
2756
3244fcb1 2757#ifdef CONFIG_PROC_FS
9f6200bb
TT		 2758 if (ext4_proc_root)
2759 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
3244fcb1 2760#endif
240799cd 2761
705895b6 2762 bgl_lock_init(sbi->s_blockgroup_lock);
ac27a0ec
DK		 2763
2764 for (i = 0; i < db_count; i++) {
70bbb3e0 2765 block = descriptor_loc(sb, logical_sb_block, i);
ac27a0ec
DK		 2766 sbi->s_group_desc[i] = sb_bread(sb, block);
2767 if (!sbi->s_group_desc[i]) {
b31e1552
ES		 2768 ext4_msg(sb, KERN_ERR,
2769 "can't read group descriptor %d", i);
ac27a0ec
DK		 2770 db_count = i;
2771 goto failed_mount2;
2772 }
2773 }
2b2d6d01 2774 if (!ext4_check_descriptors(sb)) {
b31e1552 2775 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
ac27a0ec
DK		 2776 goto failed_mount2;
2777 }
772cb7c8
JS		 2778 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2779 if (!ext4_fill_flex_info(sb)) {
b31e1552
ES		 2780 ext4_msg(sb, KERN_ERR,
2781 "unable to initialize "
2782 "flex_bg meta info!");
772cb7c8
JS		 2783 goto failed_mount2;
2784 }
2785
ac27a0ec
DK		 2786 sbi->s_gdb_count = db_count;
2787 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
2788 spin_lock_init(&sbi->s_next_gen_lock);
2789
833f4077
PZ		 2790 err = percpu_counter_init(&sbi->s_freeblocks_counter,
2791 ext4_count_free_blocks(sb));
2792 if (!err) {
2793 err = percpu_counter_init(&sbi->s_freeinodes_counter,
2794 ext4_count_free_inodes(sb));
2795 }
2796 if (!err) {
2797 err = percpu_counter_init(&sbi->s_dirs_counter,
2798 ext4_count_dirs(sb));
2799 }
6bc6e63f
AK		 2800 if (!err) {
2801 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
2802 }
833f4077 2803 if (err) {
b31e1552 2804 ext4_msg(sb, KERN_ERR, "insufficient memory");
833f4077
PZ		 2805 goto failed_mount3;
2806 }
ac27a0ec 2807
c9de560d 2808 sbi->s_stripe = ext4_get_stripe_size(sbi);
55138e0b 2809 sbi->s_max_writeback_mb_bump = 128;
c9de560d 2810
ac27a0ec
DK		 2811 /*
2812 * set up enough so that it can read an inode
2813 */
9ca92389
TT		 2814 if (!test_opt(sb, NOLOAD) &&
2815 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
2816 sb->s_op = &ext4_sops;
2817 else
2818 sb->s_op = &ext4_nojournal_sops;
617ba13b
MC		 2819 sb->s_export_op = &ext4_export_ops;
2820 sb->s_xattr = ext4_xattr_handlers;
ac27a0ec 2821#ifdef CONFIG_QUOTA
617ba13b
MC		 2822 sb->s_qcop = &ext4_qctl_operations;
2823 sb->dq_op = &ext4_quota_operations;
ac27a0ec
DK		 2824#endif
2825 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3b9d4ed2 2826 mutex_init(&sbi->s_orphan_lock);
32ed5058 2827 mutex_init(&sbi->s_resize_lock);
ac27a0ec
DK		 2828
2829 sb->s_root = NULL;
2830
2831 needs_recovery = (es->s_last_orphan != 0 ||
617ba13b
MC		 2832 EXT4_HAS_INCOMPAT_FEATURE(sb,
2833 EXT4_FEATURE_INCOMPAT_RECOVER));
ac27a0ec
DK		 2834
2835 /*
2836 * The first inode we look at is the journal inode. Don't try
2837 * root first: it may be modified in the journal!
2838 */
2839 if (!test_opt(sb, NOLOAD) &&
617ba13b
MC		 2840 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
2841 if (ext4_load_journal(sb, es, journal_devnum))
ac27a0ec 2842 goto failed_mount3;
0390131b
FM		 2843 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
2844 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
b31e1552
ES		 2845 ext4_msg(sb, KERN_ERR, "required journal recovery "
2846 "suppressed and not mounted read-only");
744692dc 2847 goto failed_mount_wq;
ac27a0ec 2848 } else {
0390131b
FM		 2849 clear_opt(sbi->s_mount_opt, DATA_FLAGS);
2850 set_opt(sbi->s_mount_opt, WRITEBACK_DATA);
2851 sbi->s_journal = NULL;
2852 needs_recovery = 0;
2853 goto no_journal;
ac27a0ec
DK		 2854 }
2855
eb40a09c
JS		 2856 if (ext4_blocks_count(es) > 0xffffffffULL &&
2857 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
2858 JBD2_FEATURE_INCOMPAT_64BIT)) {
b31e1552 2859 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
744692dc 2860 goto failed_mount_wq;
eb40a09c
JS		 2861 }
2862
d4da6c9c
LT		 2863 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
2864 jbd2_journal_set_features(sbi->s_journal,
2865 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
818d276c 2866 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
d4da6c9c
LT		 2867 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
2868 jbd2_journal_set_features(sbi->s_journal,
2869 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
818d276c
GS		 2870 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
2871 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
d4da6c9c
LT		 2872 } else {
2873 jbd2_journal_clear_features(sbi->s_journal,
2874 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
2875 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
2876 }
818d276c 2877
ac27a0ec
DK		 2878 /* We have now updated the journal if required, so we can
2879 * validate the data journaling mode. */
2880 switch (test_opt(sb, DATA_FLAGS)) {
2881 case 0:
2882 /* No mode set, assume a default based on the journal
63f57933
AM		 2883 * capabilities: ORDERED_DATA if the journal can
2884 * cope, else JOURNAL_DATA
2885 */
dab291af
MC		 2886 if (jbd2_journal_check_available_features
2887 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
ac27a0ec
DK		 2888 set_opt(sbi->s_mount_opt, ORDERED_DATA);
2889 else
2890 set_opt(sbi->s_mount_opt, JOURNAL_DATA);
2891 break;
2892
617ba13b
MC		 2893 case EXT4_MOUNT_ORDERED_DATA:
2894 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 2895 if (!jbd2_journal_check_available_features
2896 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
b31e1552
ES		 2897 ext4_msg(sb, KERN_ERR, "Journal does not support "
2898 "requested data journaling mode");
744692dc 2899 goto failed_mount_wq;
ac27a0ec
DK		 2900 }
2901 default:
2902 break;
2903 }
b3881f74 2904 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
ac27a0ec 2905
0390131b 2906no_journal:
ac27a0ec 2907 if (test_opt(sb, NOBH)) {
617ba13b 2908 if (!(test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)) {
b31e1552
ES		 2909 ext4_msg(sb, KERN_WARNING, "Ignoring nobh option - "
2910 "its supported only with writeback mode");
ac27a0ec
DK		 2911 clear_opt(sbi->s_mount_opt, NOBH);
2912 }
744692dc
JZ		 2913 if (test_opt(sb, DIOREAD_NOLOCK)) {
2914 ext4_msg(sb, KERN_WARNING, "dioread_nolock option is "
2915 "not supported with nobh mode");
2916 goto failed_mount_wq;
2917 }
ac27a0ec 2918 }
4c0425ff
MC		 2919 EXT4_SB(sb)->dio_unwritten_wq = create_workqueue("ext4-dio-unwritten");
2920 if (!EXT4_SB(sb)->dio_unwritten_wq) {
2921 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
2922 goto failed_mount_wq;
2923 }
2924
ac27a0ec 2925 /*
dab291af 2926 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 2927 * so we can safely mount the rest of the filesystem now.
2928 */
2929
1d1fe1ee
DH		 2930 root = ext4_iget(sb, EXT4_ROOT_INO);
2931 if (IS_ERR(root)) {
b31e1552 2932 ext4_msg(sb, KERN_ERR, "get root inode failed");
1d1fe1ee 2933 ret = PTR_ERR(root);
ac27a0ec
DK		 2934 goto failed_mount4;
2935 }
2936 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1d1fe1ee 2937 iput(root);
b31e1552 2938 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
ac27a0ec
DK		 2939 goto failed_mount4;
2940 }
1d1fe1ee
DH		 2941 sb->s_root = d_alloc_root(root);
2942 if (!sb->s_root) {
b31e1552 2943 ext4_msg(sb, KERN_ERR, "get root dentry failed");
1d1fe1ee
DH		 2944 iput(root);
2945 ret = -ENOMEM;
2946 goto failed_mount4;
2947 }
ac27a0ec 2948
2b2d6d01 2949 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
ef7f3835
KS		 2950
2951 /* determine the minimum size of new large inodes, if present */
2952 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
2953 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2954 EXT4_GOOD_OLD_INODE_SIZE;
2955 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
2956 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
2957 if (sbi->s_want_extra_isize <
2958 le16_to_cpu(es->s_want_extra_isize))
2959 sbi->s_want_extra_isize =
2960 le16_to_cpu(es->s_want_extra_isize);
2961 if (sbi->s_want_extra_isize <
2962 le16_to_cpu(es->s_min_extra_isize))
2963 sbi->s_want_extra_isize =
2964 le16_to_cpu(es->s_min_extra_isize);
2965 }
2966 }
2967 /* Check if enough inode space is available */
2968 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
2969 sbi->s_inode_size) {
2970 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
2971 EXT4_GOOD_OLD_INODE_SIZE;
b31e1552
ES		 2972 ext4_msg(sb, KERN_INFO, "required extra inode space not"
2973 "available");
ef7f3835
KS		 2974 }
2975
90576c0b
TT		 2976 if (test_opt(sb, DELALLOC) &&
2977 (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)) {
b31e1552
ES		 2978 ext4_msg(sb, KERN_WARNING, "Ignoring delalloc option - "
2979 "requested data journaling mode");
c2774d84 2980 clear_opt(sbi->s_mount_opt, DELALLOC);
90576c0b 2981 }
744692dc
JZ		 2982 if (test_opt(sb, DIOREAD_NOLOCK)) {
2983 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
2984 ext4_msg(sb, KERN_WARNING, "Ignoring dioread_nolock "
2985 "option - requested data journaling mode");
2986 clear_opt(sbi->s_mount_opt, DIOREAD_NOLOCK);
2987 }
2988 if (sb->s_blocksize < PAGE_SIZE) {
2989 ext4_msg(sb, KERN_WARNING, "Ignoring dioread_nolock "
2990 "option - block size is too small");
2991 clear_opt(sbi->s_mount_opt, DIOREAD_NOLOCK);
2992 }
2993 }
c2774d84 2994
6fd058f7
TT		 2995 err = ext4_setup_system_zone(sb);
2996 if (err) {
b31e1552
ES		 2997 ext4_msg(sb, KERN_ERR, "failed to initialize system "
2998 "zone (%d)\n", err);
6fd058f7
TT		 2999 goto failed_mount4;
3000 }
3001
c2774d84
AK		 3002 ext4_ext_init(sb);
3003 err = ext4_mb_init(sb, needs_recovery);
3004 if (err) {
b31e1552
ES		 3005 ext4_msg(sb, KERN_ERR, "failed to initalize mballoc (%d)",
3006 err);
c2774d84
AK		 3007 goto failed_mount4;
3008 }
3009
3197ebdb
TT		 3010 sbi->s_kobj.kset = ext4_kset;
3011 init_completion(&sbi->s_kobj_unregister);
3012 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
3013 "%s", sb->s_id);
3014 if (err) {
3015 ext4_mb_release(sb);
3016 ext4_ext_release(sb);
3017 goto failed_mount4;
3018 };
3019
617ba13b
MC		 3020 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3021 ext4_orphan_cleanup(sb, es);
3022 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
0390131b 3023 if (needs_recovery) {
b31e1552 3024 ext4_msg(sb, KERN_INFO, "recovery complete");
0390131b
FM		 3025 ext4_mark_recovery_complete(sb, es);
3026 }
3027 if (EXT4_SB(sb)->s_journal) {
3028 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3029 descr = " journalled data mode";
3030 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3031 descr = " ordered data mode";
3032 else
3033 descr = " writeback data mode";
3034 } else
3035 descr = "out journal";
3036
b31e1552 3037 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s", descr);
ac27a0ec
DK		 3038
3039 lock_kernel();
3040 return 0;
3041
617ba13b 3042cantfind_ext4:
ac27a0ec 3043 if (!silent)
b31e1552 3044 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
ac27a0ec
DK		 3045 goto failed_mount;
3046
3047failed_mount4:
b31e1552 3048 ext4_msg(sb, KERN_ERR, "mount failed");
4c0425ff
MC		 3049 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
3050failed_mount_wq:
6fd058f7 3051 ext4_release_system_zone(sb);
0390131b
FM		 3052 if (sbi->s_journal) {
3053 jbd2_journal_destroy(sbi->s_journal);
3054 sbi->s_journal = NULL;
3055 }
ac27a0ec 3056failed_mount3:
c5ca7c76
TT		 3057 if (sbi->s_flex_groups) {
3058 if (is_vmalloc_addr(sbi->s_flex_groups))
3059 vfree(sbi->s_flex_groups);
3060 else
3061 kfree(sbi->s_flex_groups);
3062 }
ac27a0ec
DK		 3063 percpu_counter_destroy(&sbi->s_freeblocks_counter);
3064 percpu_counter_destroy(&sbi->s_freeinodes_counter);
3065 percpu_counter_destroy(&sbi->s_dirs_counter);
6bc6e63f 3066 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
ac27a0ec
DK		 3067failed_mount2:
3068 for (i = 0; i < db_count; i++)
3069 brelse(sbi->s_group_desc[i]);
3070 kfree(sbi->s_group_desc);
3071failed_mount:
240799cd 3072 if (sbi->s_proc) {
9f6200bb 3073 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 3074 }
ac27a0ec
DK		 3075#ifdef CONFIG_QUOTA
3076 for (i = 0; i < MAXQUOTAS; i++)
3077 kfree(sbi->s_qf_names[i]);
3078#endif
617ba13b 3079 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 3080 brelse(bh);
3081out_fail:
3082 sb->s_fs_info = NULL;
f6830165 3083 kfree(sbi->s_blockgroup_lock);
ac27a0ec
DK		 3084 kfree(sbi);
3085 lock_kernel();
1d1fe1ee 3086 return ret;
ac27a0ec
DK		 3087}
3088
3089/*
3090 * Setup any per-fs journal parameters now. We'll do this both on
3091 * initial mount, once the journal has been initialised but before we've
3092 * done any recovery; and again on any subsequent remount.
3093 */
617ba13b 3094static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 3095{
617ba13b 3096 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 3097
30773840
TT		 3098 journal->j_commit_interval = sbi->s_commit_interval;
3099 journal->j_min_batch_time = sbi->s_min_batch_time;
3100 journal->j_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 3101
3102 spin_lock(&journal->j_state_lock);
3103 if (test_opt(sb, BARRIER))
dab291af 3104 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 3105 else
dab291af 3106 journal->j_flags &= ~JBD2_BARRIER;
5bf5683a
HK		 3107 if (test_opt(sb, DATA_ERR_ABORT))
3108 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
3109 else
3110 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
ac27a0ec
DK		 3111 spin_unlock(&journal->j_state_lock);
3112}
3113
617ba13b 3114static journal_t *ext4_get_journal(struct super_block *sb,
ac27a0ec
DK		 3115 unsigned int journal_inum)
3116{
3117 struct inode *journal_inode;
3118 journal_t *journal;
3119
0390131b
FM		 3120 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3121
ac27a0ec
DK		 3122 /* First, test for the existence of a valid inode on disk. Bad
3123 * things happen if we iget() an unused inode, as the subsequent
3124 * iput() will try to delete it. */
3125
1d1fe1ee
DH		 3126 journal_inode = ext4_iget(sb, journal_inum);
3127 if (IS_ERR(journal_inode)) {
b31e1552 3128 ext4_msg(sb, KERN_ERR, "no journal found");
ac27a0ec
DK		 3129 return NULL;
3130 }
3131 if (!journal_inode->i_nlink) {
3132 make_bad_inode(journal_inode);
3133 iput(journal_inode);
b31e1552 3134 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
ac27a0ec
DK		 3135 return NULL;
3136 }
3137
e5f8eab8 3138 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
ac27a0ec 3139 journal_inode, journal_inode->i_size);
1d1fe1ee 3140 if (!S_ISREG(journal_inode->i_mode)) {
b31e1552 3141 ext4_msg(sb, KERN_ERR, "invalid journal inode");
ac27a0ec
DK		 3142 iput(journal_inode);
3143 return NULL;
3144 }
3145
dab291af 3146 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 3147 if (!journal) {
b31e1552 3148 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
ac27a0ec
DK		 3149 iput(journal_inode);
3150 return NULL;
3151 }
3152 journal->j_private = sb;
617ba13b 3153 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 3154 return journal;
3155}
3156
617ba13b 3157static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 3158 dev_t j_dev)
3159{
2b2d6d01 3160 struct buffer_head *bh;
ac27a0ec 3161 journal_t *journal;
617ba13b
MC		 3162 ext4_fsblk_t start;
3163 ext4_fsblk_t len;
ac27a0ec 3164 int hblock, blocksize;
617ba13b 3165 ext4_fsblk_t sb_block;
ac27a0ec 3166 unsigned long offset;
2b2d6d01 3167 struct ext4_super_block *es;
ac27a0ec
DK		 3168 struct block_device *bdev;
3169
0390131b
FM		 3170 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3171
b31e1552 3172 bdev = ext4_blkdev_get(j_dev, sb);
ac27a0ec
DK		 3173 if (bdev == NULL)
3174 return NULL;
3175
3176 if (bd_claim(bdev, sb)) {
b31e1552
ES		 3177 ext4_msg(sb, KERN_ERR,
3178 "failed to claim external journal device");
9a1c3542 3179 blkdev_put(bdev, FMODE_READ|FMODE_WRITE);
ac27a0ec
DK		 3180 return NULL;
3181 }
3182
3183 blocksize = sb->s_blocksize;
e1defc4f 3184 hblock = bdev_logical_block_size(bdev);
ac27a0ec 3185 if (blocksize < hblock) {
b31e1552
ES		 3186 ext4_msg(sb, KERN_ERR,
3187 "blocksize too small for journal device");
ac27a0ec
DK		 3188 goto out_bdev;
3189 }
3190
617ba13b
MC		 3191 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3192 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 3193 set_blocksize(bdev, blocksize);
3194 if (!(bh = __bread(bdev, sb_block, blocksize))) {
b31e1552
ES		 3195 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
3196 "external journal");
ac27a0ec
DK		 3197 goto out_bdev;
3198 }
3199
617ba13b
MC		 3200 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3201 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 3202 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b 3203 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
b31e1552
ES		 3204 ext4_msg(sb, KERN_ERR, "external journal has "
3205 "bad superblock");
ac27a0ec
DK		 3206 brelse(bh);
3207 goto out_bdev;
3208 }
3209
617ba13b 3210 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
b31e1552 3211 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
ac27a0ec
DK		 3212 brelse(bh);
3213 goto out_bdev;
3214 }
3215
bd81d8ee 3216 len = ext4_blocks_count(es);
ac27a0ec
DK		 3217 start = sb_block + 1;
3218 brelse(bh); /* we're done with the superblock */
3219
dab291af 3220 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 3221 start, len, blocksize);
3222 if (!journal) {
b31e1552 3223 ext4_msg(sb, KERN_ERR, "failed to create device journal");
ac27a0ec
DK		 3224 goto out_bdev;
3225 }
3226 journal->j_private = sb;
3227 ll_rw_block(READ, 1, &journal->j_sb_buffer);
3228 wait_on_buffer(journal->j_sb_buffer);
3229 if (!buffer_uptodate(journal->j_sb_buffer)) {
b31e1552 3230 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
ac27a0ec
DK		 3231 goto out_journal;
3232 }
3233 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
b31e1552
ES		 3234 ext4_msg(sb, KERN_ERR, "External journal has more than one "
3235 "user (unsupported) - %d",
ac27a0ec
DK		 3236 be32_to_cpu(journal->j_superblock->s_nr_users));
3237 goto out_journal;
3238 }
617ba13b
MC		 3239 EXT4_SB(sb)->journal_bdev = bdev;
3240 ext4_init_journal_params(sb, journal);
ac27a0ec 3241 return journal;
0b8e58a1 3242
ac27a0ec 3243out_journal:
dab291af 3244 jbd2_journal_destroy(journal);
ac27a0ec 3245out_bdev:
617ba13b 3246 ext4_blkdev_put(bdev);
ac27a0ec
DK		 3247 return NULL;
3248}
3249
617ba13b
MC		 3250static int ext4_load_journal(struct super_block *sb,
3251 struct ext4_super_block *es,
ac27a0ec
DK		 3252 unsigned long journal_devnum)
3253{
3254 journal_t *journal;
3255 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
3256 dev_t journal_dev;
3257 int err = 0;
3258 int really_read_only;
3259
0390131b
FM		 3260 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3261
ac27a0ec
DK		 3262 if (journal_devnum &&
3263 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
b31e1552
ES		 3264 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
3265 "numbers have changed");
ac27a0ec
DK		 3266 journal_dev = new_decode_dev(journal_devnum);
3267 } else
3268 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
3269
3270 really_read_only = bdev_read_only(sb->s_bdev);
3271
3272 /*
3273 * Are we loading a blank journal or performing recovery after a
3274 * crash? For recovery, we need to check in advance whether we
3275 * can get read-write access to the device.
3276 */
617ba13b 3277 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
ac27a0ec 3278 if (sb->s_flags & MS_RDONLY) {
b31e1552
ES		 3279 ext4_msg(sb, KERN_INFO, "INFO: recovery "
3280 "required on readonly filesystem");
ac27a0ec 3281 if (really_read_only) {
b31e1552
ES		 3282 ext4_msg(sb, KERN_ERR, "write access "
3283 "unavailable, cannot proceed");
ac27a0ec
DK		 3284 return -EROFS;
3285 }
b31e1552
ES		 3286 ext4_msg(sb, KERN_INFO, "write access will "
3287 "be enabled during recovery");
ac27a0ec
DK		 3288 }
3289 }
3290
3291 if (journal_inum && journal_dev) {
b31e1552
ES		 3292 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
3293 "and inode journals!");
ac27a0ec
DK		 3294 return -EINVAL;
3295 }
3296
3297 if (journal_inum) {
617ba13b 3298 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 3299 return -EINVAL;
3300 } else {
617ba13b 3301 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 3302 return -EINVAL;
3303 }
3304
90576c0b 3305 if (!(journal->j_flags & JBD2_BARRIER))
b31e1552 3306 ext4_msg(sb, KERN_INFO, "barriers disabled");
4776004f 3307
ac27a0ec 3308 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
dab291af 3309 err = jbd2_journal_update_format(journal);
ac27a0ec 3310 if (err) {
b31e1552 3311 ext4_msg(sb, KERN_ERR, "error updating journal");
dab291af 3312 jbd2_journal_destroy(journal);
ac27a0ec
DK		 3313 return err;
3314 }
3315 }
3316
617ba13b 3317 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
dab291af 3318 err = jbd2_journal_wipe(journal, !really_read_only);
ac27a0ec 3319 if (!err)
dab291af 3320 err = jbd2_journal_load(journal);
ac27a0ec
DK		 3321
3322 if (err) {
b31e1552 3323 ext4_msg(sb, KERN_ERR, "error loading journal");
dab291af 3324 jbd2_journal_destroy(journal);
ac27a0ec
DK		 3325 return err;
3326 }
3327
617ba13b
MC		 3328 EXT4_SB(sb)->s_journal = journal;
3329 ext4_clear_journal_err(sb, es);
ac27a0ec
DK		 3330
3331 if (journal_devnum &&
3332 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3333 es->s_journal_dev = cpu_to_le32(journal_devnum);
ac27a0ec
DK		 3334
3335 /* Make sure we flush the recovery flag to disk. */
e2d67052 3336 ext4_commit_super(sb, 1);
ac27a0ec
DK		 3337 }
3338
3339 return 0;
3340}
3341
e2d67052 3342static int ext4_commit_super(struct super_block *sb, int sync)
ac27a0ec 3343{
e2d67052 3344 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
617ba13b 3345 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
c4be0c1d 3346 int error = 0;
ac27a0ec
DK		 3347
3348 if (!sbh)
c4be0c1d 3349 return error;
914258bf
TT		 3350 if (buffer_write_io_error(sbh)) {
3351 /*
3352 * Oh, dear. A previous attempt to write the
3353 * superblock failed. This could happen because the
3354 * USB device was yanked out. Or it could happen to
3355 * be a transient write error and maybe the block will
3356 * be remapped. Nothing we can do but to retry the
3357 * write and hope for the best.
3358 */
b31e1552
ES		 3359 ext4_msg(sb, KERN_ERR, "previous I/O error to "
3360 "superblock detected");
914258bf
TT		 3361 clear_buffer_write_io_error(sbh);
3362 set_buffer_uptodate(sbh);
3363 }
71290b36
TT		 3364 /*
3365 * If the file system is mounted read-only, don't update the
3366 * superblock write time. This avoids updating the superblock
3367 * write time when we are mounting the root file system
3368 * read/only but we need to replay the journal; at that point,
3369 * for people who are east of GMT and who make their clock
3370 * tick in localtime for Windows bug-for-bug compatibility,
3371 * the clock is set in the future, and this will cause e2fsck
3372 * to complain and force a full file system check.
3373 */
3374 if (!(sb->s_flags & MS_RDONLY))
3375 es->s_wtime = cpu_to_le32(get_seconds());
afc32f7e
TT		 3376 es->s_kbytes_written =
3377 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
3378 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
3379 EXT4_SB(sb)->s_sectors_written_start) >> 1));
5d1b1b3f
AK		 3380 ext4_free_blocks_count_set(es, percpu_counter_sum_positive(
3381 &EXT4_SB(sb)->s_freeblocks_counter));
3382 es->s_free_inodes_count = cpu_to_le32(percpu_counter_sum_positive(
3383 &EXT4_SB(sb)->s_freeinodes_counter));
7234ab2a 3384 sb->s_dirt = 0;
ac27a0ec
DK		 3385 BUFFER_TRACE(sbh, "marking dirty");
3386 mark_buffer_dirty(sbh);
914258bf 3387 if (sync) {
c4be0c1d
TS		 3388 error = sync_dirty_buffer(sbh);
3389 if (error)
3390 return error;
3391
3392 error = buffer_write_io_error(sbh);
3393 if (error) {
b31e1552
ES		 3394 ext4_msg(sb, KERN_ERR, "I/O error while writing "
3395 "superblock");
914258bf
TT		 3396 clear_buffer_write_io_error(sbh);
3397 set_buffer_uptodate(sbh);
3398 }
3399 }
c4be0c1d 3400 return error;
ac27a0ec
DK		 3401}
3402
ac27a0ec
DK		 3403/*
3404 * Have we just finished recovery? If so, and if we are mounting (or
3405 * remounting) the filesystem readonly, then we will end up with a
3406 * consistent fs on disk. Record that fact.
3407 */
2b2d6d01
TT		 3408static void ext4_mark_recovery_complete(struct super_block *sb,
3409 struct ext4_super_block *es)
ac27a0ec 3410{
617ba13b 3411 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 3412
0390131b
FM		 3413 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3414 BUG_ON(journal != NULL);
3415 return;
3416 }
dab291af 3417 jbd2_journal_lock_updates(journal);
7ffe1ea8
HK		 3418 if (jbd2_journal_flush(journal) < 0)
3419 goto out;
3420
617ba13b 3421 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
ac27a0ec 3422 sb->s_flags & MS_RDONLY) {
617ba13b 3423 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
e2d67052 3424 ext4_commit_super(sb, 1);
ac27a0ec 3425 }
7ffe1ea8
HK		 3426
3427out:
dab291af 3428 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 3429}
3430
3431/*
3432 * If we are mounting (or read-write remounting) a filesystem whose journal
3433 * has recorded an error from a previous lifetime, move that error to the
3434 * main filesystem now.
3435 */
2b2d6d01
TT		 3436static void ext4_clear_journal_err(struct super_block *sb,
3437 struct ext4_super_block *es)
ac27a0ec
DK		 3438{
3439 journal_t *journal;
3440 int j_errno;
3441 const char *errstr;
3442
0390131b
FM		 3443 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3444
617ba13b 3445 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 3446
3447 /*
3448 * Now check for any error status which may have been recorded in the
617ba13b 3449 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 3450 */
3451
dab291af 3452 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 3453 if (j_errno) {
3454 char nbuf[16];
3455
617ba13b 3456 errstr = ext4_decode_error(sb, j_errno, nbuf);
12062ddd 3457 ext4_warning(sb, "Filesystem error recorded "
ac27a0ec 3458 "from previous mount: %s", errstr);
12062ddd 3459 ext4_warning(sb, "Marking fs in need of filesystem check.");
ac27a0ec 3460
617ba13b
MC		 3461 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
3462 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
e2d67052 3463 ext4_commit_super(sb, 1);
ac27a0ec 3464
dab291af 3465 jbd2_journal_clear_err(journal);
ac27a0ec
DK		 3466 }
3467}
3468
3469/*
3470 * Force the running and committing transactions to commit,
3471 * and wait on the commit.
3472 */
617ba13b 3473int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 3474{
3475 journal_t *journal;
0390131b 3476 int ret = 0;
ac27a0ec
DK		 3477
3478 if (sb->s_flags & MS_RDONLY)
3479 return 0;
3480
617ba13b 3481 journal = EXT4_SB(sb)->s_journal;
7234ab2a 3482 if (journal)
0390131b 3483 ret = ext4_journal_force_commit(journal);
0390131b 3484
ac27a0ec
DK		 3485 return ret;
3486}
3487
2b2d6d01 3488static void ext4_write_super(struct super_block *sb)
ac27a0ec 3489{
ebc1ac16 3490 lock_super(sb);
9ca92389 3491 ext4_commit_super(sb, 1);
ebc1ac16 3492 unlock_super(sb);
ac27a0ec
DK		 3493}
3494
617ba13b 3495static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec 3496{
14ce0cb4 3497 int ret = 0;
9eddacf9 3498 tid_t target;
8d5d02e6 3499 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 3500
9bffad1e 3501 trace_ext4_sync_fs(sb, wait);
8d5d02e6
MC		 3502 flush_workqueue(sbi->dio_unwritten_wq);
3503 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
9ca92389 3504 if (wait)
8d5d02e6 3505 jbd2_log_wait_commit(sbi->s_journal, target);
0390131b 3506 }
14ce0cb4 3507 return ret;
ac27a0ec
DK		 3508}
3509
3510/*
3511 * LVM calls this function before a (read-only) snapshot is created. This
3512 * gives us a chance to flush the journal completely and mark the fs clean.
3513 */
c4be0c1d 3514static int ext4_freeze(struct super_block *sb)
ac27a0ec 3515{
c4be0c1d
TS		 3516 int error = 0;
3517 journal_t *journal;
ac27a0ec 3518
9ca92389
TT		 3519 if (sb->s_flags & MS_RDONLY)
3520 return 0;
ac27a0ec 3521
9ca92389 3522 journal = EXT4_SB(sb)->s_journal;
7ffe1ea8 3523
9ca92389
TT		 3524 /* Now we set up the journal barrier. */
3525 jbd2_journal_lock_updates(journal);
ac27a0ec 3526
9ca92389
TT		 3527 /*
3528 * Don't clear the needs_recovery flag if we failed to flush
3529 * the journal.
3530 */
3531 error = jbd2_journal_flush(journal);
3532 if (error < 0) {
3533 out:
3534 jbd2_journal_unlock_updates(journal);
3535 return error;
ac27a0ec 3536 }
9ca92389
TT		 3537
3538 /* Journal blocked and flushed, clear needs_recovery flag. */
3539 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3540 error = ext4_commit_super(sb, 1);
3541 if (error)
3542 goto out;
c4be0c1d 3543 return 0;
ac27a0ec
DK		 3544}
3545
3546/*
3547 * Called by LVM after the snapshot is done. We need to reset the RECOVER
3548 * flag here, even though the filesystem is not technically dirty yet.
3549 */
c4be0c1d 3550static int ext4_unfreeze(struct super_block *sb)
ac27a0ec 3551{
9ca92389
TT		 3552 if (sb->s_flags & MS_RDONLY)
3553 return 0;
3554
3555 lock_super(sb);
3556 /* Reset the needs_recovery flag before the fs is unlocked. */
3557 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
3558 ext4_commit_super(sb, 1);
3559 unlock_super(sb);
3560 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
c4be0c1d 3561 return 0;
ac27a0ec
DK		 3562}
3563
2b2d6d01 3564static int ext4_remount(struct super_block *sb, int *flags, char *data)
ac27a0ec 3565{
2b2d6d01 3566 struct ext4_super_block *es;
617ba13b
MC		 3567 struct ext4_sb_info *sbi = EXT4_SB(sb);
3568 ext4_fsblk_t n_blocks_count = 0;
ac27a0ec 3569 unsigned long old_sb_flags;
617ba13b 3570 struct ext4_mount_options old_opts;
8a266467 3571 ext4_group_t g;
b3881f74 3572 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
ac27a0ec
DK		 3573 int err;
3574#ifdef CONFIG_QUOTA
3575 int i;
3576#endif
3577
337eb00a
AIB		 3578 lock_kernel();
3579
ac27a0ec 3580 /* Store the original options */
bbd6851a 3581 lock_super(sb);
ac27a0ec
DK		 3582 old_sb_flags = sb->s_flags;
3583 old_opts.s_mount_opt = sbi->s_mount_opt;
3584 old_opts.s_resuid = sbi->s_resuid;
3585 old_opts.s_resgid = sbi->s_resgid;
3586 old_opts.s_commit_interval = sbi->s_commit_interval;
30773840
TT		 3587 old_opts.s_min_batch_time = sbi->s_min_batch_time;
3588 old_opts.s_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 3589#ifdef CONFIG_QUOTA
3590 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
3591 for (i = 0; i < MAXQUOTAS; i++)
3592 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
3593#endif
b3881f74
TT		 3594 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
3595 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
ac27a0ec
DK		 3596
3597 /*
3598 * Allow the "check" option to be passed as a remount option.
3599 */
b3881f74
TT		 3600 if (!parse_options(data, sb, NULL, &journal_ioprio,
3601 &n_blocks_count, 1)) {
ac27a0ec
DK		 3602 err = -EINVAL;
3603 goto restore_opts;
3604 }
3605
4ab2f15b 3606 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
46e665e9 3607 ext4_abort(sb, __func__, "Abort forced by user");
ac27a0ec
DK		 3608
3609 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 3610 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec
DK		 3611
3612 es = sbi->s_es;
3613
b3881f74 3614 if (sbi->s_journal) {
0390131b 3615 ext4_init_journal_params(sb, sbi->s_journal);
b3881f74
TT		 3616 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3617 }
ac27a0ec
DK		 3618
3619 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
bd81d8ee 3620 n_blocks_count > ext4_blocks_count(es)) {
4ab2f15b 3621 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
ac27a0ec
DK		 3622 err = -EROFS;
3623 goto restore_opts;
3624 }
3625
3626 if (*flags & MS_RDONLY) {
3627 /*
3628 * First of all, the unconditional stuff we have to do
3629 * to disable replay of the journal when we next remount
3630 */
3631 sb->s_flags |= MS_RDONLY;
3632
3633 /*
3634 * OK, test if we are remounting a valid rw partition
3635 * readonly, and if so set the rdonly flag and then
3636 * mark the partition as valid again.
3637 */
617ba13b
MC		 3638 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
3639 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 3640 es->s_state = cpu_to_le16(sbi->s_mount_state);
3641
a63c9eb2 3642 if (sbi->s_journal)
0390131b 3643 ext4_mark_recovery_complete(sb, es);
ac27a0ec 3644 } else {
a13fb1a4
ES		 3645 /* Make sure we can mount this feature set readwrite */
3646 if (!ext4_feature_set_ok(sb, 0)) {
ac27a0ec
DK		 3647 err = -EROFS;
3648 goto restore_opts;
3649 }
8a266467
TT		 3650 /*
3651 * Make sure the group descriptor checksums
0b8e58a1 3652 * are sane. If they aren't, refuse to remount r/w.
8a266467
TT		 3653 */
3654 for (g = 0; g < sbi->s_groups_count; g++) {
3655 struct ext4_group_desc *gdp =
3656 ext4_get_group_desc(sb, g, NULL);
3657
3658 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
b31e1552
ES		 3659 ext4_msg(sb, KERN_ERR,
3660 "ext4_remount: Checksum for group %u failed (%u!=%u)",
8a266467
TT		 3661 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
3662 le16_to_cpu(gdp->bg_checksum));
3663 err = -EINVAL;
3664 goto restore_opts;
3665 }
3666 }
3667
ead6596b
ES		 3668 /*
3669 * If we have an unprocessed orphan list hanging
3670 * around from a previously readonly bdev mount,
3671 * require a full umount/remount for now.
3672 */
3673 if (es->s_last_orphan) {
b31e1552 3674 ext4_msg(sb, KERN_WARNING, "Couldn't "
ead6596b
ES		 3675 "remount RDWR because of unprocessed "
3676 "orphan inode list. Please "
b31e1552 3677 "umount/remount instead");
ead6596b
ES		 3678 err = -EINVAL;
3679 goto restore_opts;
3680 }
3681
ac27a0ec
DK		 3682 /*
3683 * Mounting a RDONLY partition read-write, so reread
3684 * and store the current valid flag. (It may have
3685 * been changed by e2fsck since we originally mounted
3686 * the partition.)
3687 */
0390131b
FM		 3688 if (sbi->s_journal)
3689 ext4_clear_journal_err(sb, es);
ac27a0ec 3690 sbi->s_mount_state = le16_to_cpu(es->s_state);
617ba13b 3691 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
ac27a0ec 3692 goto restore_opts;
2b2d6d01 3693 if (!ext4_setup_super(sb, es, 0))
ac27a0ec
DK		 3694 sb->s_flags &= ~MS_RDONLY;
3695 }
3696 }
6fd058f7 3697 ext4_setup_system_zone(sb);
0390131b 3698 if (sbi->s_journal == NULL)
e2d67052 3699 ext4_commit_super(sb, 1);
0390131b 3700
ac27a0ec
DK		 3701#ifdef CONFIG_QUOTA
3702 /* Release old quota file names */
3703 for (i = 0; i < MAXQUOTAS; i++)
3704 if (old_opts.s_qf_names[i] &&
3705 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3706 kfree(old_opts.s_qf_names[i]);
3707#endif
bbd6851a 3708 unlock_super(sb);
337eb00a 3709 unlock_kernel();
ac27a0ec 3710 return 0;
0b8e58a1 3711
ac27a0ec
DK		 3712restore_opts:
3713 sb->s_flags = old_sb_flags;
3714 sbi->s_mount_opt = old_opts.s_mount_opt;
3715 sbi->s_resuid = old_opts.s_resuid;
3716 sbi->s_resgid = old_opts.s_resgid;
3717 sbi->s_commit_interval = old_opts.s_commit_interval;
30773840
TT		 3718 sbi->s_min_batch_time = old_opts.s_min_batch_time;
3719 sbi->s_max_batch_time = old_opts.s_max_batch_time;
ac27a0ec
DK		 3720#ifdef CONFIG_QUOTA
3721 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
3722 for (i = 0; i < MAXQUOTAS; i++) {
3723 if (sbi->s_qf_names[i] &&
3724 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
3725 kfree(sbi->s_qf_names[i]);
3726 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
3727 }
3728#endif
bbd6851a 3729 unlock_super(sb);
337eb00a 3730 unlock_kernel();
ac27a0ec
DK		 3731 return err;
3732}
3733
2b2d6d01 3734static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
ac27a0ec
DK		 3735{
3736 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 3737 struct ext4_sb_info *sbi = EXT4_SB(sb);
3738 struct ext4_super_block *es = sbi->s_es;
960cc398 3739 u64 fsid;
ac27a0ec 3740
5e70030d
BP		 3741 if (test_opt(sb, MINIX_DF)) {
3742 sbi->s_overhead_last = 0;
6bc9feff 3743 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
8df9675f 3744 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
5e70030d 3745 ext4_fsblk_t overhead = 0;
ac27a0ec
DK		 3746
3747 /*
5e70030d
BP		 3748 * Compute the overhead (FS structures). This is constant
3749 * for a given filesystem unless the number of block groups
3750 * changes so we cache the previous value until it does.
ac27a0ec
DK		 3751 */
3752
3753 /*
3754 * All of the blocks before first_data_block are
3755 * overhead
3756 */
3757 overhead = le32_to_cpu(es->s_first_data_block);
3758
3759 /*
3760 * Add the overhead attributed to the superblock and
3761 * block group descriptors. If the sparse superblocks
3762 * feature is turned on, then not all groups have this.
3763 */
3764 for (i = 0; i < ngroups; i++) {
617ba13b
MC		 3765 overhead += ext4_bg_has_super(sb, i) +
3766 ext4_bg_num_gdb(sb, i);
ac27a0ec
DK		 3767 cond_resched();
3768 }
3769
3770 /*
3771 * Every block group has an inode bitmap, a block
3772 * bitmap, and an inode table.
3773 */
5e70030d
BP		 3774 overhead += ngroups * (2 + sbi->s_itb_per_group);
3775 sbi->s_overhead_last = overhead;
3776 smp_wmb();
6bc9feff 3777 sbi->s_blocks_last = ext4_blocks_count(es);
ac27a0ec
DK		 3778 }
3779
617ba13b 3780 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 3781 buf->f_bsize = sb->s_blocksize;
5e70030d 3782 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
6bc6e63f
AK		 3783 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
3784 percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
bd81d8ee
LV		 3785 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
3786 if (buf->f_bfree < ext4_r_blocks_count(es))
ac27a0ec
DK		 3787 buf->f_bavail = 0;
3788 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 3789 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
617ba13b 3790 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 3791 fsid = le64_to_cpup((void *)es->s_uuid) ^
3792 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
3793 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
3794 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
0b8e58a1 3795
ac27a0ec
DK		 3796 return 0;
3797}
3798
0b8e58a1
AD		 3799/* Helper function for writing quotas on sync - we need to start transaction
3800 * before quota file is locked for write. Otherwise the are possible deadlocks:
ac27a0ec 3801 * Process 1 Process 2
617ba13b 3802 * ext4_create() quota_sync()
a269eb18
JK		 3803 * jbd2_journal_start() write_dquot()
3804 * vfs_dq_init() down(dqio_mutex)
dab291af 3805 * down(dqio_mutex) jbd2_journal_start()
ac27a0ec
DK		 3806 *
3807 */
3808
3809#ifdef CONFIG_QUOTA
3810
3811static inline struct inode *dquot_to_inode(struct dquot *dquot)
3812{
3813 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
3814}
3815
617ba13b 3816static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 3817{
3818 int ret, err;
3819 handle_t *handle;
3820 struct inode *inode;
3821
3822 inode = dquot_to_inode(dquot);
617ba13b 3823 handle = ext4_journal_start(inode,
0b8e58a1 3824 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 3825 if (IS_ERR(handle))
3826 return PTR_ERR(handle);
3827 ret = dquot_commit(dquot);
617ba13b 3828 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3829 if (!ret)
3830 ret = err;
3831 return ret;
3832}
3833
617ba13b 3834static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 3835{
3836 int ret, err;
3837 handle_t *handle;
3838
617ba13b 3839 handle = ext4_journal_start(dquot_to_inode(dquot),
0b8e58a1 3840 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 3841 if (IS_ERR(handle))
3842 return PTR_ERR(handle);
3843 ret = dquot_acquire(dquot);
617ba13b 3844 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3845 if (!ret)
3846 ret = err;
3847 return ret;
3848}
3849
617ba13b 3850static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 3851{
3852 int ret, err;
3853 handle_t *handle;
3854
617ba13b 3855 handle = ext4_journal_start(dquot_to_inode(dquot),
0b8e58a1 3856 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 3857 if (IS_ERR(handle)) {
3858 /* Release dquot anyway to avoid endless cycle in dqput() */
3859 dquot_release(dquot);
ac27a0ec 3860 return PTR_ERR(handle);
9c3013e9 3861 }
ac27a0ec 3862 ret = dquot_release(dquot);
617ba13b 3863 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3864 if (!ret)
3865 ret = err;
3866 return ret;
3867}
3868
617ba13b 3869static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec 3870{
2c8be6b2 3871 /* Are we journaling quotas? */
617ba13b
MC		 3872 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
3873 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
ac27a0ec 3874 dquot_mark_dquot_dirty(dquot);
617ba13b 3875 return ext4_write_dquot(dquot);
ac27a0ec
DK		 3876 } else {
3877 return dquot_mark_dquot_dirty(dquot);
3878 }
3879}
3880
617ba13b 3881static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 3882{
3883 int ret, err;
3884 handle_t *handle;
3885
3886 /* Data block + inode block */
617ba13b 3887 handle = ext4_journal_start(sb->s_root->d_inode, 2);
ac27a0ec
DK		 3888 if (IS_ERR(handle))
3889 return PTR_ERR(handle);
3890 ret = dquot_commit_info(sb, type);
617ba13b 3891 err = ext4_journal_stop(handle);
ac27a0ec
DK		 3892 if (!ret)
3893 ret = err;
3894 return ret;
3895}
3896
3897/*
3898 * Turn on quotas during mount time - we need to find
3899 * the quota file and such...
3900 */
617ba13b 3901static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 3902{
617ba13b 3903 return vfs_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
0b8e58a1 3904 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 3905}
3906
3907/*
3908 * Standard function to be called on quota_on
3909 */
617ba13b 3910static int ext4_quota_on(struct super_block *sb, int type, int format_id,
8264613d 3911 char *name, int remount)
ac27a0ec
DK		 3912{
3913 int err;
8264613d 3914 struct path path;
ac27a0ec
DK		 3915
3916 if (!test_opt(sb, QUOTA))
3917 return -EINVAL;
8264613d 3918 /* When remounting, no checks are needed and in fact, name is NULL */
0623543b 3919 if (remount)
8264613d 3920 return vfs_quota_on(sb, type, format_id, name, remount);
0623543b 3921
8264613d 3922 err = kern_path(name, LOOKUP_FOLLOW, &path);
ac27a0ec
DK		 3923 if (err)
3924 return err;
0623543b 3925
ac27a0ec 3926 /* Quotafile not on the same filesystem? */
8264613d
AV		 3927 if (path.mnt->mnt_sb != sb) {
3928 path_put(&path);
ac27a0ec
DK		 3929 return -EXDEV;
3930 }
0623543b
JK		 3931 /* Journaling quota? */
3932 if (EXT4_SB(sb)->s_qf_names[type]) {
2b2d6d01 3933 /* Quotafile not in fs root? */
8264613d 3934 if (path.dentry->d_parent != sb->s_root)
b31e1552
ES		 3935 ext4_msg(sb, KERN_WARNING,
3936 "Quota file not on filesystem root. "
3937 "Journaled quota will not work");
2b2d6d01 3938 }
0623543b
JK		 3939
3940 /*
3941 * When we journal data on quota file, we have to flush journal to see
3942 * all updates to the file when we bypass pagecache...
3943 */
0390131b
FM		 3944 if (EXT4_SB(sb)->s_journal &&
3945 ext4_should_journal_data(path.dentry->d_inode)) {
0623543b
JK		 3946 /*
3947 * We don't need to lock updates but journal_flush() could
3948 * otherwise be livelocked...
3949 */
3950 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 3951 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
0623543b 3952 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 3953 if (err) {
8264613d 3954 path_put(&path);
7ffe1ea8
HK		 3955 return err;
3956 }
0623543b
JK		 3957 }
3958
8264613d
AV		 3959 err = vfs_quota_on_path(sb, type, format_id, &path);
3960 path_put(&path);
77e69dac 3961 return err;
ac27a0ec
DK		 3962}
3963
3964/* Read data from quotafile - avoid pagecache and such because we cannot afford
3965 * acquiring the locks... As quota files are never truncated and quota code
3966 * itself serializes the operations (and noone else should touch the files)
3967 * we don't have to be afraid of races */
617ba13b 3968static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 3969 size_t len, loff_t off)
3970{
3971 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 3972 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 3973 int err = 0;
3974 int offset = off & (sb->s_blocksize - 1);
3975 int tocopy;
3976 size_t toread;
3977 struct buffer_head *bh;
3978 loff_t i_size = i_size_read(inode);
3979
3980 if (off > i_size)
3981 return 0;
3982 if (off+len > i_size)
3983 len = i_size-off;
3984 toread = len;
3985 while (toread > 0) {
3986 tocopy = sb->s_blocksize - offset < toread ?
3987 sb->s_blocksize - offset : toread;
617ba13b 3988 bh = ext4_bread(NULL, inode, blk, 0, &err);
ac27a0ec
DK		 3989 if (err)
3990 return err;
3991 if (!bh) /* A hole? */
3992 memset(data, 0, tocopy);
3993 else
3994 memcpy(data, bh->b_data+offset, tocopy);
3995 brelse(bh);
3996 offset = 0;
3997 toread -= tocopy;
3998 data += tocopy;
3999 blk++;
4000 }
4001 return len;
4002}
4003
4004/* Write to quotafile (we know the transaction is already started and has
4005 * enough credits) */
617ba13b 4006static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 4007 const char *data, size_t len, loff_t off)
4008{
4009 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 4010 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 4011 int err = 0;
4012 int offset = off & (sb->s_blocksize - 1);
4013 int tocopy;
617ba13b 4014 int journal_quota = EXT4_SB(sb)->s_qf_names[type] != NULL;
ac27a0ec
DK		 4015 size_t towrite = len;
4016 struct buffer_head *bh;
4017 handle_t *handle = journal_current_handle();
4018
0390131b 4019 if (EXT4_SB(sb)->s_journal && !handle) {
b31e1552
ES		 4020 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4021 " cancelled because transaction is not started",
9c3013e9
JK		 4022 (unsigned long long)off, (unsigned long long)len);
4023 return -EIO;
4024 }
ac27a0ec
DK		 4025 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
4026 while (towrite > 0) {
4027 tocopy = sb->s_blocksize - offset < towrite ?
4028 sb->s_blocksize - offset : towrite;
617ba13b 4029 bh = ext4_bread(handle, inode, blk, 1, &err);
ac27a0ec
DK		 4030 if (!bh)
4031 goto out;
4032 if (journal_quota) {
617ba13b 4033 err = ext4_journal_get_write_access(handle, bh);
ac27a0ec
DK		 4034 if (err) {
4035 brelse(bh);
4036 goto out;
4037 }
4038 }
4039 lock_buffer(bh);
4040 memcpy(bh->b_data+offset, data, tocopy);
4041 flush_dcache_page(bh->b_page);
4042 unlock_buffer(bh);
4043 if (journal_quota)
0390131b 4044 err = ext4_handle_dirty_metadata(handle, NULL, bh);
ac27a0ec
DK		 4045 else {
4046 /* Always do at least ordered writes for quotas */
678aaf48 4047 err = ext4_jbd2_file_inode(handle, inode);
ac27a0ec
DK		 4048 mark_buffer_dirty(bh);
4049 }
4050 brelse(bh);
4051 if (err)
4052 goto out;
4053 offset = 0;
4054 towrite -= tocopy;
4055 data += tocopy;
4056 blk++;
4057 }
4058out:
4d04e4fb
JK		 4059 if (len == towrite) {
4060 mutex_unlock(&inode->i_mutex);
ac27a0ec 4061 return err;
4d04e4fb 4062 }
ac27a0ec
DK		 4063 if (inode->i_size < off+len-towrite) {
4064 i_size_write(inode, off+len-towrite);
617ba13b 4065 EXT4_I(inode)->i_disksize = inode->i_size;
ac27a0ec 4066 }
ac27a0ec 4067 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
617ba13b 4068 ext4_mark_inode_dirty(handle, inode);
ac27a0ec
DK		 4069 mutex_unlock(&inode->i_mutex);
4070 return len - towrite;
4071}
4072
4073#endif
4074
0b8e58a1
AD		 4075static int ext4_get_sb(struct file_system_type *fs_type, int flags,
4076 const char *dev_name, void *data, struct vfsmount *mnt)
ac27a0ec 4077{
0b8e58a1 4078 return get_sb_bdev(fs_type, flags, dev_name, data, ext4_fill_super,mnt);
ac27a0ec
DK		 4079}
4080
a214238d 4081#if !defined(CONTIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
24b58424
TT		 4082static struct file_system_type ext2_fs_type = {
4083 .owner = THIS_MODULE,
4084 .name = "ext2",
4085 .get_sb = ext4_get_sb,
4086 .kill_sb = kill_block_super,
4087 .fs_flags = FS_REQUIRES_DEV,
4088};
4089
4090static inline void register_as_ext2(void)
4091{
4092 int err = register_filesystem(&ext2_fs_type);
4093 if (err)
4094 printk(KERN_WARNING
4095 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
4096}
4097
4098static inline void unregister_as_ext2(void)
4099{
4100 unregister_filesystem(&ext2_fs_type);
4101}
51b7e3c9 4102MODULE_ALIAS("ext2");
24b58424
TT		 4103#else
4104static inline void register_as_ext2(void) { }
4105static inline void unregister_as_ext2(void) { }
4106#endif
4107
a214238d 4108#if !defined(CONTIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
24b58424
TT		 4109static struct file_system_type ext3_fs_type = {
4110 .owner = THIS_MODULE,
4111 .name = "ext3",
4112 .get_sb = ext4_get_sb,
4113 .kill_sb = kill_block_super,
4114 .fs_flags = FS_REQUIRES_DEV,
4115};
4116
4117static inline void register_as_ext3(void)
4118{
4119 int err = register_filesystem(&ext3_fs_type);
4120 if (err)
4121 printk(KERN_WARNING
4122 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
4123}
4124
4125static inline void unregister_as_ext3(void)
4126{
4127 unregister_filesystem(&ext3_fs_type);
4128}
51b7e3c9 4129MODULE_ALIAS("ext3");
24b58424
TT		 4130#else
4131static inline void register_as_ext3(void) { }
4132static inline void unregister_as_ext3(void) { }
4133#endif
4134
03010a33
TT		 4135static struct file_system_type ext4_fs_type = {
4136 .owner = THIS_MODULE,
4137 .name = "ext4",
4138 .get_sb = ext4_get_sb,
4139 .kill_sb = kill_block_super,
4140 .fs_flags = FS_REQUIRES_DEV,
4141};
4142
617ba13b 4143static int __init init_ext4_fs(void)
ac27a0ec 4144{
c9de560d
AT		 4145 int err;
4146
6fd058f7
TT		 4147 err = init_ext4_system_zone();
4148 if (err)
4149 return err;
3197ebdb
TT		 4150 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
4151 if (!ext4_kset)
6fd058f7 4152 goto out4;
9f6200bb 4153 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
c9de560d 4154 err = init_ext4_mballoc();
ac27a0ec 4155 if (err)
6fd058f7 4156 goto out3;
c9de560d
AT		 4157
4158 err = init_ext4_xattr();
4159 if (err)
4160 goto out2;
ac27a0ec
DK		 4161 err = init_inodecache();
4162 if (err)
4163 goto out1;
24b58424
TT		 4164 register_as_ext2();
4165 register_as_ext3();
03010a33 4166 err = register_filesystem(&ext4_fs_type);
ac27a0ec
DK		 4167 if (err)
4168 goto out;
4169 return 0;
4170out:
24b58424
TT		 4171 unregister_as_ext2();
4172 unregister_as_ext3();
ac27a0ec
DK		 4173 destroy_inodecache();
4174out1:
617ba13b 4175 exit_ext4_xattr();
c9de560d
AT		 4176out2:
4177 exit_ext4_mballoc();
6fd058f7
TT		 4178out3:
4179 remove_proc_entry("fs/ext4", NULL);
4180 kset_unregister(ext4_kset);
4181out4:
4182 exit_ext4_system_zone();
ac27a0ec
DK		 4183 return err;
4184}
4185
617ba13b 4186static void __exit exit_ext4_fs(void)
ac27a0ec 4187{
24b58424
TT		 4188 unregister_as_ext2();
4189 unregister_as_ext3();
03010a33 4190 unregister_filesystem(&ext4_fs_type);
ac27a0ec 4191 destroy_inodecache();
617ba13b 4192 exit_ext4_xattr();
c9de560d 4193 exit_ext4_mballoc();
9f6200bb 4194 remove_proc_entry("fs/ext4", NULL);
3197ebdb 4195 kset_unregister(ext4_kset);
6fd058f7 4196 exit_ext4_system_zone();
ac27a0ec
DK		 4197}
4198
4199MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
83982b6f 4200MODULE_DESCRIPTION("Fourth Extended Filesystem");
ac27a0ec 4201MODULE_LICENSE("GPL");
617ba13b
MC		 4202module_init(init_ext4_fs)
4203module_exit(exit_ext4_fs)
Linux kernel - Deliverables
This page took 0.697224 seconds and 5 git commands to generate.