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