projects / linux-2.6-block.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
freezer: unexport refrigerator() and update try_to_freeze() slightly
[linux-2.6-block.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>
ac27a0ec 29#include <linux/buffer_head.h>
a5694255 30#include <linux/exportfs.h>
ac27a0ec
DK		 31#include <linux/vfs.h>
32#include <linux/random.h>
33#include <linux/mount.h>
34#include <linux/namei.h>
35#include <linux/quotaops.h>
36#include <linux/seq_file.h>
9f6200bb 37#include <linux/proc_fs.h>
3197ebdb 38#include <linux/ctype.h>
1330593e 39#include <linux/log2.h>
717d50e4 40#include <linux/crc16.h>
7abc52c2 41#include <linux/cleancache.h>
ac27a0ec
DK		 42#include <asm/uaccess.h>
43
bfff6873
LC		 44#include <linux/kthread.h>
45#include <linux/freezer.h>
46
3dcf5451 47#include "ext4.h"
6f91bc5f 48#include "ext4_extents.h"
3dcf5451 49#include "ext4_jbd2.h"
ac27a0ec
DK		 50#include "xattr.h"
51#include "acl.h"
3661d286 52#include "mballoc.h"
ac27a0ec 53
9bffad1e
TT		 54#define CREATE_TRACE_POINTS
55#include <trace/events/ext4.h>
56
1f109d5a 57static struct proc_dir_entry *ext4_proc_root;
3197ebdb 58static struct kset *ext4_kset;
0b75a840
LC		 59static struct ext4_lazy_init *ext4_li_info;
60static struct mutex ext4_li_mtx;
61static struct ext4_features *ext4_feat;
9f6200bb 62
617ba13b 63static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
ac27a0ec 64 unsigned long journal_devnum);
e2d67052 65static int ext4_commit_super(struct super_block *sb, int sync);
2b2d6d01
TT		 66static void ext4_mark_recovery_complete(struct super_block *sb,
67 struct ext4_super_block *es);
68static void ext4_clear_journal_err(struct super_block *sb,
69 struct ext4_super_block *es);
617ba13b 70static int ext4_sync_fs(struct super_block *sb, int wait);
2b2d6d01 71static const char *ext4_decode_error(struct super_block *sb, int errno,
ac27a0ec 72 char nbuf[16]);
2b2d6d01
TT		 73static int ext4_remount(struct super_block *sb, int *flags, char *data);
74static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
c4be0c1d 75static int ext4_unfreeze(struct super_block *sb);
2b2d6d01 76static void ext4_write_super(struct super_block *sb);
c4be0c1d 77static int ext4_freeze(struct super_block *sb);
152a0836
AV		 78static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
79 const char *dev_name, void *data);
2035e776
TT		 80static inline int ext2_feature_set_ok(struct super_block *sb);
81static inline int ext3_feature_set_ok(struct super_block *sb);
d39195c3 82static int ext4_feature_set_ok(struct super_block *sb, int readonly);
bfff6873
LC		 83static void ext4_destroy_lazyinit_thread(void);
84static void ext4_unregister_li_request(struct super_block *sb);
8f1f7453 85static void ext4_clear_request_list(void);
ac27a0ec 86
2035e776
TT		 87#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
88static struct file_system_type ext2_fs_type = {
89 .owner = THIS_MODULE,
90 .name = "ext2",
91 .mount = ext4_mount,
92 .kill_sb = kill_block_super,
93 .fs_flags = FS_REQUIRES_DEV,
94};
95#define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
96#else
97#define IS_EXT2_SB(sb) (0)
98#endif
99
100
ba69f9ab
JK		 101#if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
102static struct file_system_type ext3_fs_type = {
103 .owner = THIS_MODULE,
104 .name = "ext3",
152a0836 105 .mount = ext4_mount,
ba69f9ab
JK		 106 .kill_sb = kill_block_super,
107 .fs_flags = FS_REQUIRES_DEV,
108};
109#define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
110#else
111#define IS_EXT3_SB(sb) (0)
112#endif
bd81d8ee 113
9933fc0a
TT		 114void *ext4_kvmalloc(size_t size, gfp_t flags)
115{
116 void *ret;
117
118 ret = kmalloc(size, flags);
119 if (!ret)
120 ret = __vmalloc(size, flags, PAGE_KERNEL);
121 return ret;
122}
123
124void *ext4_kvzalloc(size_t size, gfp_t flags)
125{
126 void *ret;
127
db9481c0 128 ret = kzalloc(size, flags);
9933fc0a
TT		 129 if (!ret)
130 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
131 return ret;
132}
133
134void ext4_kvfree(void *ptr)
135{
136 if (is_vmalloc_addr(ptr))
137 vfree(ptr);
138 else
139 kfree(ptr);
140
141}
142
8fadc143
AR		 143ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
144 struct ext4_group_desc *bg)
bd81d8ee 145{
3a14589c 146 return le32_to_cpu(bg->bg_block_bitmap_lo) |
8fadc143 147 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 148 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 149}
150
8fadc143
AR		 151ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
152 struct ext4_group_desc *bg)
bd81d8ee 153{
5272f837 154 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
8fadc143 155 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 156 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 157}
158
8fadc143
AR		 159ext4_fsblk_t ext4_inode_table(struct super_block *sb,
160 struct ext4_group_desc *bg)
bd81d8ee 161{
5272f837 162 return le32_to_cpu(bg->bg_inode_table_lo) |
8fadc143 163 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 164 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
bd81d8ee
LV		 165}
166
021b65bb
TT		 167__u32 ext4_free_group_clusters(struct super_block *sb,
168 struct ext4_group_desc *bg)
560671a0
AK		 169{
170 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
171 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 172 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
560671a0
AK		 173}
174
175__u32 ext4_free_inodes_count(struct super_block *sb,
176 struct ext4_group_desc *bg)
177{
178 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
179 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 180 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
560671a0
AK		 181}
182
183__u32 ext4_used_dirs_count(struct super_block *sb,
184 struct ext4_group_desc *bg)
185{
186 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
187 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 188 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
560671a0
AK		 189}
190
191__u32 ext4_itable_unused_count(struct super_block *sb,
192 struct ext4_group_desc *bg)
193{
194 return le16_to_cpu(bg->bg_itable_unused_lo) |
195 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 196 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
560671a0
AK		 197}
198
8fadc143
AR		 199void ext4_block_bitmap_set(struct super_block *sb,
200 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 201{
3a14589c 202 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 203 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
204 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 205}
206
8fadc143
AR		 207void ext4_inode_bitmap_set(struct super_block *sb,
208 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 209{
5272f837 210 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 211 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
212 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 213}
214
8fadc143
AR		 215void ext4_inode_table_set(struct super_block *sb,
216 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 217{
5272f837 218 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 219 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
220 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 221}
222
021b65bb
TT		 223void ext4_free_group_clusters_set(struct super_block *sb,
224 struct ext4_group_desc *bg, __u32 count)
560671a0
AK		 225{
226 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
227 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
228 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
229}
230
231void ext4_free_inodes_set(struct super_block *sb,
232 struct ext4_group_desc *bg, __u32 count)
233{
234 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
235 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
236 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
237}
238
239void ext4_used_dirs_set(struct super_block *sb,
240 struct ext4_group_desc *bg, __u32 count)
241{
242 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
243 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
244 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
245}
246
247void ext4_itable_unused_set(struct super_block *sb,
248 struct ext4_group_desc *bg, __u32 count)
249{
250 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
251 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
252 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
253}
254
d3d1faf6
CW		 255
256/* Just increment the non-pointer handle value */
257static handle_t *ext4_get_nojournal(void)
258{
259 handle_t *handle = current->journal_info;
260 unsigned long ref_cnt = (unsigned long)handle;
261
262 BUG_ON(ref_cnt >= EXT4_NOJOURNAL_MAX_REF_COUNT);
263
264 ref_cnt++;
265 handle = (handle_t *)ref_cnt;
266
267 current->journal_info = handle;
268 return handle;
269}
270
271
272/* Decrement the non-pointer handle value */
273static void ext4_put_nojournal(handle_t *handle)
274{
275 unsigned long ref_cnt = (unsigned long)handle;
276
277 BUG_ON(ref_cnt == 0);
278
279 ref_cnt--;
280 handle = (handle_t *)ref_cnt;
281
282 current->journal_info = handle;
283}
284
ac27a0ec 285/*
dab291af 286 * Wrappers for jbd2_journal_start/end.
ac27a0ec
DK		 287 *
288 * The only special thing we need to do here is to make sure that all
289 * journal_end calls result in the superblock being marked dirty, so
290 * that sync() will call the filesystem's write_super callback if
291 * appropriate.
be4f27d3
YY		 292 *
293 * To avoid j_barrier hold in userspace when a user calls freeze(),
294 * ext4 prevents a new handle from being started by s_frozen, which
295 * is in an upper layer.
ac27a0ec 296 */
617ba13b 297handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
ac27a0ec
DK		 298{
299 journal_t *journal;
be4f27d3 300 handle_t *handle;
ac27a0ec 301
12706394 302 trace_ext4_journal_start(sb, nblocks, _RET_IP_);
ac27a0ec
DK		 303 if (sb->s_flags & MS_RDONLY)
304 return ERR_PTR(-EROFS);
305
617ba13b 306 journal = EXT4_SB(sb)->s_journal;
be4f27d3
YY		 307 handle = ext4_journal_current_handle();
308
309 /*
310 * If a handle has been started, it should be allowed to
311 * finish, otherwise deadlock could happen between freeze
312 * and others(e.g. truncate) due to the restart of the
313 * journal handle if the filesystem is forzen and active
314 * handles are not stopped.
315 */
316 if (!handle)
317 vfs_check_frozen(sb, SB_FREEZE_TRANS);
318
319 if (!journal)
320 return ext4_get_nojournal();
321 /*
322 * Special case here: if the journal has aborted behind our
323 * backs (eg. EIO in the commit thread), then we still need to
324 * take the FS itself readonly cleanly.
325 */
326 if (is_journal_aborted(journal)) {
327 ext4_abort(sb, "Detected aborted journal");
328 return ERR_PTR(-EROFS);
ac27a0ec 329 }
be4f27d3 330 return jbd2_journal_start(journal, nblocks);
ac27a0ec
DK		 331}
332
333/*
334 * The only special thing we need to do here is to make sure that all
dab291af 335 * jbd2_journal_stop calls result in the superblock being marked dirty, so
ac27a0ec
DK		 336 * that sync() will call the filesystem's write_super callback if
337 * appropriate.
338 */
c398eda0 339int __ext4_journal_stop(const char *where, unsigned int line, handle_t *handle)
ac27a0ec
DK		 340{
341 struct super_block *sb;
342 int err;
343 int rc;
344
0390131b 345 if (!ext4_handle_valid(handle)) {
d3d1faf6 346 ext4_put_nojournal(handle);
0390131b
FM		 347 return 0;
348 }
ac27a0ec
DK		 349 sb = handle->h_transaction->t_journal->j_private;
350 err = handle->h_err;
dab291af 351 rc = jbd2_journal_stop(handle);
ac27a0ec
DK		 352
353 if (!err)
354 err = rc;
355 if (err)
c398eda0 356 __ext4_std_error(sb, where, line, err);
ac27a0ec
DK		 357 return err;
358}
359
90c7201b
TT		 360void ext4_journal_abort_handle(const char *caller, unsigned int line,
361 const char *err_fn, struct buffer_head *bh,
362 handle_t *handle, int err)
ac27a0ec
DK		 363{
364 char nbuf[16];
617ba13b 365 const char *errstr = ext4_decode_error(NULL, err, nbuf);
ac27a0ec 366
0390131b
FM		 367 BUG_ON(!ext4_handle_valid(handle));
368
ac27a0ec
DK		 369 if (bh)
370 BUFFER_TRACE(bh, "abort");
371
372 if (!handle->h_err)
373 handle->h_err = err;
374
375 if (is_handle_aborted(handle))
376 return;
377
90c7201b
TT		 378 printk(KERN_ERR "%s:%d: aborting transaction: %s in %s\n",
379 caller, line, errstr, err_fn);
ac27a0ec 380
dab291af 381 jbd2_journal_abort_handle(handle);
ac27a0ec
DK		 382}
383
1c13d5c0
TT		 384static void __save_error_info(struct super_block *sb, const char *func,
385 unsigned int line)
386{
387 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
388
389 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
390 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
391 es->s_last_error_time = cpu_to_le32(get_seconds());
392 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
393 es->s_last_error_line = cpu_to_le32(line);
394 if (!es->s_first_error_time) {
395 es->s_first_error_time = es->s_last_error_time;
396 strncpy(es->s_first_error_func, func,
397 sizeof(es->s_first_error_func));
398 es->s_first_error_line = cpu_to_le32(line);
399 es->s_first_error_ino = es->s_last_error_ino;
400 es->s_first_error_block = es->s_last_error_block;
401 }
66e61a9e
TT		 402 /*
403 * Start the daily error reporting function if it hasn't been
404 * started already
405 */
406 if (!es->s_error_count)
407 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
1c13d5c0
TT		 408 es->s_error_count = cpu_to_le32(le32_to_cpu(es->s_error_count) + 1);
409}
410
411static void save_error_info(struct super_block *sb, const char *func,
412 unsigned int line)
413{
414 __save_error_info(sb, func, line);
415 ext4_commit_super(sb, 1);
416}
417
7c2e7087
TT		 418/*
419 * The del_gendisk() function uninitializes the disk-specific data
420 * structures, including the bdi structure, without telling anyone
421 * else. Once this happens, any attempt to call mark_buffer_dirty()
422 * (for example, by ext4_commit_super), will cause a kernel OOPS.
423 * This is a kludge to prevent these oops until we can put in a proper
424 * hook in del_gendisk() to inform the VFS and file system layers.
425 */
426static int block_device_ejected(struct super_block *sb)
427{
428 struct inode *bd_inode = sb->s_bdev->bd_inode;
429 struct backing_dev_info *bdi = bd_inode->i_mapping->backing_dev_info;
430
431 return bdi->dev == NULL;
432}
433
1c13d5c0 434
ac27a0ec
DK		 435/* Deal with the reporting of failure conditions on a filesystem such as
436 * inconsistencies detected or read IO failures.
437 *
438 * On ext2, we can store the error state of the filesystem in the
617ba13b 439 * superblock. That is not possible on ext4, because we may have other
ac27a0ec
DK		 440 * write ordering constraints on the superblock which prevent us from
441 * writing it out straight away; and given that the journal is about to
442 * be aborted, we can't rely on the current, or future, transactions to
443 * write out the superblock safely.
444 *
dab291af 445 * We'll just use the jbd2_journal_abort() error code to record an error in
d6b198bc 446 * the journal instead. On recovery, the journal will complain about
ac27a0ec
DK		 447 * that error until we've noted it down and cleared it.
448 */
449
617ba13b 450static void ext4_handle_error(struct super_block *sb)
ac27a0ec 451{
ac27a0ec
DK		 452 if (sb->s_flags & MS_RDONLY)
453 return;
454
2b2d6d01 455 if (!test_opt(sb, ERRORS_CONT)) {
617ba13b 456 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 457
4ab2f15b 458 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
ac27a0ec 459 if (journal)
dab291af 460 jbd2_journal_abort(journal, -EIO);
ac27a0ec 461 }
2b2d6d01 462 if (test_opt(sb, ERRORS_RO)) {
b31e1552 463 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
ac27a0ec
DK		 464 sb->s_flags |= MS_RDONLY;
465 }
ac27a0ec 466 if (test_opt(sb, ERRORS_PANIC))
617ba13b 467 panic("EXT4-fs (device %s): panic forced after error\n",
ac27a0ec
DK		 468 sb->s_id);
469}
470
12062ddd 471void __ext4_error(struct super_block *sb, const char *function,
c398eda0 472 unsigned int line, const char *fmt, ...)
ac27a0ec 473{
0ff2ea7d 474 struct va_format vaf;
ac27a0ec
DK		 475 va_list args;
476
477 va_start(args, fmt);
0ff2ea7d
JP		 478 vaf.fmt = fmt;
479 vaf.va = &args;
480 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
481 sb->s_id, function, line, current->comm, &vaf);
ac27a0ec
DK		 482 va_end(args);
483
617ba13b 484 ext4_handle_error(sb);
ac27a0ec
DK		 485}
486
c398eda0
TT		 487void ext4_error_inode(struct inode *inode, const char *function,
488 unsigned int line, ext4_fsblk_t block,
273df556
FM		 489 const char *fmt, ...)
490{
491 va_list args;
f7c21177 492 struct va_format vaf;
1c13d5c0 493 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
273df556 494
1c13d5c0
TT		 495 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
496 es->s_last_error_block = cpu_to_le64(block);
497 save_error_info(inode->i_sb, function, line);
273df556 498 va_start(args, fmt);
f7c21177
TT		 499 vaf.fmt = fmt;
500 vaf.va = &args;
c398eda0
TT		 501 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: inode #%lu: ",
502 inode->i_sb->s_id, function, line, inode->i_ino);
503 if (block)
f7c21177
TT		 504 printk(KERN_CONT "block %llu: ", block);
505 printk(KERN_CONT "comm %s: %pV\n", current->comm, &vaf);
273df556
FM		 506 va_end(args);
507
508 ext4_handle_error(inode->i_sb);
509}
510
c398eda0 511void ext4_error_file(struct file *file, const char *function,
f7c21177
TT		 512 unsigned int line, ext4_fsblk_t block,
513 const char *fmt, ...)
273df556
FM		 514{
515 va_list args;
f7c21177 516 struct va_format vaf;
1c13d5c0 517 struct ext4_super_block *es;
273df556
FM		 518 struct inode *inode = file->f_dentry->d_inode;
519 char pathname[80], *path;
520
1c13d5c0
TT		 521 es = EXT4_SB(inode->i_sb)->s_es;
522 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
523 save_error_info(inode->i_sb, function, line);
273df556 524 path = d_path(&(file->f_path), pathname, sizeof(pathname));
f9a62d09 525 if (IS_ERR(path))
273df556
FM		 526 path = "(unknown)";
527 printk(KERN_CRIT
f7c21177
TT		 528 "EXT4-fs error (device %s): %s:%d: inode #%lu: ",
529 inode->i_sb->s_id, function, line, inode->i_ino);
530 if (block)
531 printk(KERN_CONT "block %llu: ", block);
532 va_start(args, fmt);
533 vaf.fmt = fmt;
534 vaf.va = &args;
535 printk(KERN_CONT "comm %s: path %s: %pV\n", current->comm, path, &vaf);
273df556
FM		 536 va_end(args);
537
538 ext4_handle_error(inode->i_sb);
539}
540
2b2d6d01 541static const char *ext4_decode_error(struct super_block *sb, int errno,
ac27a0ec
DK		 542 char nbuf[16])
543{
544 char *errstr = NULL;
545
546 switch (errno) {
547 case -EIO:
548 errstr = "IO failure";
549 break;
550 case -ENOMEM:
551 errstr = "Out of memory";
552 break;
553 case -EROFS:
78f1ddbb
TT		 554 if (!sb || (EXT4_SB(sb)->s_journal &&
555 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
ac27a0ec
DK		 556 errstr = "Journal has aborted";
557 else
558 errstr = "Readonly filesystem";
559 break;
560 default:
561 /* If the caller passed in an extra buffer for unknown
562 * errors, textualise them now. Else we just return
563 * NULL. */
564 if (nbuf) {
565 /* Check for truncated error codes... */
566 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
567 errstr = nbuf;
568 }
569 break;
570 }
571
572 return errstr;
573}
574
617ba13b 575/* __ext4_std_error decodes expected errors from journaling functions
ac27a0ec
DK		 576 * automatically and invokes the appropriate error response. */
577
c398eda0
TT		 578void __ext4_std_error(struct super_block *sb, const char *function,
579 unsigned int line, int errno)
ac27a0ec
DK		 580{
581 char nbuf[16];
582 const char *errstr;
583
584 /* Special case: if the error is EROFS, and we're not already
585 * inside a transaction, then there's really no point in logging
586 * an error. */
587 if (errno == -EROFS && journal_current_handle() == NULL &&
588 (sb->s_flags & MS_RDONLY))
589 return;
590
617ba13b 591 errstr = ext4_decode_error(sb, errno, nbuf);
c398eda0
TT		 592 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
593 sb->s_id, function, line, errstr);
1c13d5c0 594 save_error_info(sb, function, line);
ac27a0ec 595
617ba13b 596 ext4_handle_error(sb);
ac27a0ec
DK		 597}
598
599/*
617ba13b 600 * ext4_abort is a much stronger failure handler than ext4_error. The
ac27a0ec
DK		 601 * abort function may be used to deal with unrecoverable failures such
602 * as journal IO errors or ENOMEM at a critical moment in log management.
603 *
604 * We unconditionally force the filesystem into an ABORT|READONLY state,
605 * unless the error response on the fs has been set to panic in which
606 * case we take the easy way out and panic immediately.
607 */
608
c67d859e 609void __ext4_abort(struct super_block *sb, const char *function,
c398eda0 610 unsigned int line, const char *fmt, ...)
ac27a0ec
DK		 611{
612 va_list args;
613
1c13d5c0 614 save_error_info(sb, function, line);
ac27a0ec 615 va_start(args, fmt);
c398eda0
TT		 616 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
617 function, line);
ac27a0ec
DK		 618 vprintk(fmt, args);
619 printk("\n");
620 va_end(args);
621
1c13d5c0
TT		 622 if ((sb->s_flags & MS_RDONLY) == 0) {
623 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
624 sb->s_flags |= MS_RDONLY;
625 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
626 if (EXT4_SB(sb)->s_journal)
627 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
628 save_error_info(sb, function, line);
629 }
ac27a0ec 630 if (test_opt(sb, ERRORS_PANIC))
617ba13b 631 panic("EXT4-fs panic from previous error\n");
ac27a0ec
DK		 632}
633
0ff2ea7d 634void ext4_msg(struct super_block *sb, const char *prefix, const char *fmt, ...)
b31e1552 635{
0ff2ea7d 636 struct va_format vaf;
b31e1552
ES		 637 va_list args;
638
639 va_start(args, fmt);
0ff2ea7d
JP		 640 vaf.fmt = fmt;
641 vaf.va = &args;
642 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
b31e1552
ES		 643 va_end(args);
644}
645
12062ddd 646void __ext4_warning(struct super_block *sb, const char *function,
c398eda0 647 unsigned int line, const char *fmt, ...)
ac27a0ec 648{
0ff2ea7d 649 struct va_format vaf;
ac27a0ec
DK		 650 va_list args;
651
652 va_start(args, fmt);
0ff2ea7d
JP		 653 vaf.fmt = fmt;
654 vaf.va = &args;
655 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
656 sb->s_id, function, line, &vaf);
ac27a0ec
DK		 657 va_end(args);
658}
659
e29136f8
TT		 660void __ext4_grp_locked_error(const char *function, unsigned int line,
661 struct super_block *sb, ext4_group_t grp,
662 unsigned long ino, ext4_fsblk_t block,
663 const char *fmt, ...)
5d1b1b3f
AK		 664__releases(bitlock)
665__acquires(bitlock)
666{
0ff2ea7d 667 struct va_format vaf;
5d1b1b3f
AK		 668 va_list args;
669 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
670
1c13d5c0
TT		 671 es->s_last_error_ino = cpu_to_le32(ino);
672 es->s_last_error_block = cpu_to_le64(block);
673 __save_error_info(sb, function, line);
0ff2ea7d 674
5d1b1b3f 675 va_start(args, fmt);
0ff2ea7d
JP		 676
677 vaf.fmt = fmt;
678 vaf.va = &args;
21149d61 679 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
e29136f8
TT		 680 sb->s_id, function, line, grp);
681 if (ino)
0ff2ea7d 682 printk(KERN_CONT "inode %lu: ", ino);
e29136f8 683 if (block)
0ff2ea7d
JP		 684 printk(KERN_CONT "block %llu:", (unsigned long long) block);
685 printk(KERN_CONT "%pV\n", &vaf);
5d1b1b3f
AK		 686 va_end(args);
687
688 if (test_opt(sb, ERRORS_CONT)) {
e2d67052 689 ext4_commit_super(sb, 0);
5d1b1b3f
AK		 690 return;
691 }
1c13d5c0 692
5d1b1b3f
AK		 693 ext4_unlock_group(sb, grp);
694 ext4_handle_error(sb);
695 /*
696 * We only get here in the ERRORS_RO case; relocking the group
697 * may be dangerous, but nothing bad will happen since the
698 * filesystem will have already been marked read/only and the
699 * journal has been aborted. We return 1 as a hint to callers
700 * who might what to use the return value from
25985edc 701 * ext4_grp_locked_error() to distinguish between the
5d1b1b3f
AK		 702 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
703 * aggressively from the ext4 function in question, with a
704 * more appropriate error code.
705 */
706 ext4_lock_group(sb, grp);
707 return;
708}
709
617ba13b 710void ext4_update_dynamic_rev(struct super_block *sb)
ac27a0ec 711{
617ba13b 712 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 713
617ba13b 714 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
ac27a0ec
DK		 715 return;
716
12062ddd 717 ext4_warning(sb,
ac27a0ec
DK		 718 "updating to rev %d because of new feature flag, "
719 "running e2fsck is recommended",
617ba13b 720 EXT4_DYNAMIC_REV);
ac27a0ec 721
617ba13b
MC		 722 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
723 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
724 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
ac27a0ec
DK		 725 /* leave es->s_feature_*compat flags alone */
726 /* es->s_uuid will be set by e2fsck if empty */
727
728 /*
729 * The rest of the superblock fields should be zero, and if not it
730 * means they are likely already in use, so leave them alone. We
731 * can leave it up to e2fsck to clean up any inconsistencies there.
732 */
733}
734
735/*
736 * Open the external journal device
737 */
b31e1552 738static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
ac27a0ec
DK		 739{
740 struct block_device *bdev;
741 char b[BDEVNAME_SIZE];
742
d4d77629 743 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
ac27a0ec
DK		 744 if (IS_ERR(bdev))
745 goto fail;
746 return bdev;
747
748fail:
b31e1552 749 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
ac27a0ec
DK		 750 __bdevname(dev, b), PTR_ERR(bdev));
751 return NULL;
752}
753
754/*
755 * Release the journal device
756 */
617ba13b 757static int ext4_blkdev_put(struct block_device *bdev)
ac27a0ec 758{
e525fd89 759 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
ac27a0ec
DK		 760}
761
617ba13b 762static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
ac27a0ec
DK		 763{
764 struct block_device *bdev;
765 int ret = -ENODEV;
766
767 bdev = sbi->journal_bdev;
768 if (bdev) {
617ba13b 769 ret = ext4_blkdev_put(bdev);
ac27a0ec
DK		 770 sbi->journal_bdev = NULL;
771 }
772 return ret;
773}
774
775static inline struct inode *orphan_list_entry(struct list_head *l)
776{
617ba13b 777 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
ac27a0ec
DK		 778}
779
617ba13b 780static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
ac27a0ec
DK		 781{
782 struct list_head *l;
783
b31e1552
ES		 784 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
785 le32_to_cpu(sbi->s_es->s_last_orphan));
ac27a0ec
DK		 786
787 printk(KERN_ERR "sb_info orphan list:\n");
788 list_for_each(l, &sbi->s_orphan) {
789 struct inode *inode = orphan_list_entry(l);
790 printk(KERN_ERR " "
791 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
792 inode->i_sb->s_id, inode->i_ino, inode,
793 inode->i_mode, inode->i_nlink,
794 NEXT_ORPHAN(inode));
795 }
796}
797
2b2d6d01 798static void ext4_put_super(struct super_block *sb)
ac27a0ec 799{
617ba13b
MC		 800 struct ext4_sb_info *sbi = EXT4_SB(sb);
801 struct ext4_super_block *es = sbi->s_es;
ef2cabf7 802 int i, err;
ac27a0ec 803
857ac889 804 ext4_unregister_li_request(sb);
e0ccfd95
CH		 805 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
806
4c0425ff
MC		 807 flush_workqueue(sbi->dio_unwritten_wq);
808 destroy_workqueue(sbi->dio_unwritten_wq);
809
a9e220f8 810 lock_super(sb);
8c85e125 811 if (sb->s_dirt)
ebc1ac16 812 ext4_commit_super(sb, 1);
8c85e125 813
0390131b
FM		 814 if (sbi->s_journal) {
815 err = jbd2_journal_destroy(sbi->s_journal);
816 sbi->s_journal = NULL;
817 if (err < 0)
c67d859e 818 ext4_abort(sb, "Couldn't clean up the journal");
0390131b 819 }
d4edac31 820
a1c6c569 821 del_timer(&sbi->s_err_report);
d4edac31
JB		 822 ext4_release_system_zone(sb);
823 ext4_mb_release(sb);
824 ext4_ext_release(sb);
825 ext4_xattr_put_super(sb);
826
ac27a0ec 827 if (!(sb->s_flags & MS_RDONLY)) {
617ba13b 828 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 829 es->s_state = cpu_to_le16(sbi->s_mount_state);
e2d67052 830 ext4_commit_super(sb, 1);
ac27a0ec 831 }
240799cd 832 if (sbi->s_proc) {
9f6200bb 833 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 834 }
3197ebdb 835 kobject_del(&sbi->s_kobj);
ac27a0ec
DK		 836
837 for (i = 0; i < sbi->s_gdb_count; i++)
838 brelse(sbi->s_group_desc[i]);
f18a5f21 839 ext4_kvfree(sbi->s_group_desc);
9933fc0a 840 ext4_kvfree(sbi->s_flex_groups);
57042651 841 percpu_counter_destroy(&sbi->s_freeclusters_counter);
ac27a0ec
DK		 842 percpu_counter_destroy(&sbi->s_freeinodes_counter);
843 percpu_counter_destroy(&sbi->s_dirs_counter);
57042651 844 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
ac27a0ec
DK		 845 brelse(sbi->s_sbh);
846#ifdef CONFIG_QUOTA
847 for (i = 0; i < MAXQUOTAS; i++)
848 kfree(sbi->s_qf_names[i]);
849#endif
850
851 /* Debugging code just in case the in-memory inode orphan list
852 * isn't empty. The on-disk one can be non-empty if we've
853 * detected an error and taken the fs readonly, but the
854 * in-memory list had better be clean by this point. */
855 if (!list_empty(&sbi->s_orphan))
856 dump_orphan_list(sb, sbi);
857 J_ASSERT(list_empty(&sbi->s_orphan));
858
f98393a6 859 invalidate_bdev(sb->s_bdev);
ac27a0ec
DK		 860 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
861 /*
862 * Invalidate the journal device's buffers. We don't want them
863 * floating about in memory - the physical journal device may
864 * hotswapped, and it breaks the `ro-after' testing code.
865 */
866 sync_blockdev(sbi->journal_bdev);
f98393a6 867 invalidate_bdev(sbi->journal_bdev);
617ba13b 868 ext4_blkdev_remove(sbi);
ac27a0ec 869 }
c5e06d10
JL		 870 if (sbi->s_mmp_tsk)
871 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec 872 sb->s_fs_info = NULL;
3197ebdb
TT		 873 /*
874 * Now that we are completely done shutting down the
875 * superblock, we need to actually destroy the kobject.
876 */
3197ebdb
TT		 877 unlock_super(sb);
878 kobject_put(&sbi->s_kobj);
879 wait_for_completion(&sbi->s_kobj_unregister);
705895b6 880 kfree(sbi->s_blockgroup_lock);
ac27a0ec 881 kfree(sbi);
ac27a0ec
DK		 882}
883
e18b890b 884static struct kmem_cache *ext4_inode_cachep;
ac27a0ec
DK		 885
886/*
887 * Called inside transaction, so use GFP_NOFS
888 */
617ba13b 889static struct inode *ext4_alloc_inode(struct super_block *sb)
ac27a0ec 890{
617ba13b 891 struct ext4_inode_info *ei;
ac27a0ec 892
e6b4f8da 893 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
ac27a0ec
DK		 894 if (!ei)
895 return NULL;
0b8e58a1 896
ac27a0ec 897 ei->vfs_inode.i_version = 1;
91246c00 898 ei->vfs_inode.i_data.writeback_index = 0;
a86c6181 899 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
c9de560d
AT		 900 INIT_LIST_HEAD(&ei->i_prealloc_list);
901 spin_lock_init(&ei->i_prealloc_lock);
d2a17637
MC		 902 ei->i_reserved_data_blocks = 0;
903 ei->i_reserved_meta_blocks = 0;
904 ei->i_allocated_meta_blocks = 0;
9d0be502 905 ei->i_da_metadata_calc_len = 0;
d2a17637 906 spin_lock_init(&(ei->i_block_reservation_lock));
a9e7f447
DM		 907#ifdef CONFIG_QUOTA
908 ei->i_reserved_quota = 0;
909#endif
8aefcd55 910 ei->jinode = NULL;
c7064ef1 911 INIT_LIST_HEAD(&ei->i_completed_io_list);
744692dc 912 spin_lock_init(&ei->i_completed_io_lock);
8d5d02e6 913 ei->cur_aio_dio = NULL;
b436b9be
JK		 914 ei->i_sync_tid = 0;
915 ei->i_datasync_tid = 0;
f7ad6d2e 916 atomic_set(&ei->i_ioend_count, 0);
e9e3bcec 917 atomic_set(&ei->i_aiodio_unwritten, 0);
0b8e58a1 918
ac27a0ec
DK		 919 return &ei->vfs_inode;
920}
921
7ff9c073
TT		 922static int ext4_drop_inode(struct inode *inode)
923{
924 int drop = generic_drop_inode(inode);
925
926 trace_ext4_drop_inode(inode, drop);
927 return drop;
928}
929
fa0d7e3d
NP		 930static void ext4_i_callback(struct rcu_head *head)
931{
932 struct inode *inode = container_of(head, struct inode, i_rcu);
933 INIT_LIST_HEAD(&inode->i_dentry);
934 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
935}
936
617ba13b 937static void ext4_destroy_inode(struct inode *inode)
ac27a0ec 938{
9f7dd93d 939 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
b31e1552
ES		 940 ext4_msg(inode->i_sb, KERN_ERR,
941 "Inode %lu (%p): orphan list check failed!",
942 inode->i_ino, EXT4_I(inode));
9f7dd93d
VA		 943 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
944 EXT4_I(inode), sizeof(struct ext4_inode_info),
945 true);
946 dump_stack();
947 }
fa0d7e3d 948 call_rcu(&inode->i_rcu, ext4_i_callback);
ac27a0ec
DK		 949}
950
51cc5068 951static void init_once(void *foo)
ac27a0ec 952{
617ba13b 953 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
ac27a0ec 954
a35afb83 955 INIT_LIST_HEAD(&ei->i_orphan);
03010a33 956#ifdef CONFIG_EXT4_FS_XATTR
a35afb83 957 init_rwsem(&ei->xattr_sem);
ac27a0ec 958#endif
0e855ac8 959 init_rwsem(&ei->i_data_sem);
a35afb83 960 inode_init_once(&ei->vfs_inode);
ac27a0ec
DK		 961}
962
963static int init_inodecache(void)
964{
617ba13b
MC		 965 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
966 sizeof(struct ext4_inode_info),
ac27a0ec
DK		 967 0, (SLAB_RECLAIM_ACCOUNT|
968 SLAB_MEM_SPREAD),
20c2df83 969 init_once);
617ba13b 970 if (ext4_inode_cachep == NULL)
ac27a0ec
DK		 971 return -ENOMEM;
972 return 0;
973}
974
975static void destroy_inodecache(void)
976{
617ba13b 977 kmem_cache_destroy(ext4_inode_cachep);
ac27a0ec
DK		 978}
979
0930fcc1 980void ext4_clear_inode(struct inode *inode)
ac27a0ec 981{
0930fcc1
AV		 982 invalidate_inode_buffers(inode);
983 end_writeback(inode);
9f754758 984 dquot_drop(inode);
c2ea3fde 985 ext4_discard_preallocations(inode);
8aefcd55
TT		 986 if (EXT4_I(inode)->jinode) {
987 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
988 EXT4_I(inode)->jinode);
989 jbd2_free_inode(EXT4_I(inode)->jinode);
990 EXT4_I(inode)->jinode = NULL;
991 }
ac27a0ec
DK		 992}
993
2b2d6d01
TT		 994static inline void ext4_show_quota_options(struct seq_file *seq,
995 struct super_block *sb)
ac27a0ec
DK		 996{
997#if defined(CONFIG_QUOTA)
617ba13b 998 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 999
5a20bdfc
JK		 1000 if (sbi->s_jquota_fmt) {
1001 char *fmtname = "";
1002
1003 switch (sbi->s_jquota_fmt) {
1004 case QFMT_VFS_OLD:
1005 fmtname = "vfsold";
1006 break;
1007 case QFMT_VFS_V0:
1008 fmtname = "vfsv0";
1009 break;
1010 case QFMT_VFS_V1:
1011 fmtname = "vfsv1";
1012 break;
1013 }
1014 seq_printf(seq, ",jqfmt=%s", fmtname);
1015 }
ac27a0ec
DK		 1016
1017 if (sbi->s_qf_names[USRQUOTA])
1018 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1019
1020 if (sbi->s_qf_names[GRPQUOTA])
1021 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
1022
482a7425 1023 if (test_opt(sb, USRQUOTA))
ac27a0ec
DK		 1024 seq_puts(seq, ",usrquota");
1025
482a7425 1026 if (test_opt(sb, GRPQUOTA))
ac27a0ec
DK		 1027 seq_puts(seq, ",grpquota");
1028#endif
1029}
1030
d9c9bef1
MS		 1031/*
1032 * Show an option if
1033 * - it's set to a non-default value OR
1034 * - if the per-sb default is different from the global default
1035 */
617ba13b 1036static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
ac27a0ec 1037{
aa22df2c
AK		 1038 int def_errors;
1039 unsigned long def_mount_opts;
ac27a0ec 1040 struct super_block *sb = vfs->mnt_sb;
d9c9bef1
MS		 1041 struct ext4_sb_info *sbi = EXT4_SB(sb);
1042 struct ext4_super_block *es = sbi->s_es;
d9c9bef1
MS		 1043
1044 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
aa22df2c 1045 def_errors = le16_to_cpu(es->s_errors);
d9c9bef1
MS		 1046
1047 if (sbi->s_sb_block != 1)
1048 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
1049 if (test_opt(sb, MINIX_DF))
1050 seq_puts(seq, ",minixdf");
aa22df2c 1051 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
d9c9bef1
MS		 1052 seq_puts(seq, ",grpid");
1053 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
1054 seq_puts(seq, ",nogrpid");
1055 if (sbi->s_resuid != EXT4_DEF_RESUID ||
1056 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
1057 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
1058 }
1059 if (sbi->s_resgid != EXT4_DEF_RESGID ||
1060 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
1061 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
1062 }
bb4f397a 1063 if (test_opt(sb, ERRORS_RO)) {
d9c9bef1 1064 if (def_errors == EXT4_ERRORS_PANIC ||
bb4f397a
AK		 1065 def_errors == EXT4_ERRORS_CONTINUE) {
1066 seq_puts(seq, ",errors=remount-ro");
d9c9bef1
MS		 1067 }
1068 }
aa22df2c 1069 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
bb4f397a 1070 seq_puts(seq, ",errors=continue");
aa22df2c 1071 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
d9c9bef1 1072 seq_puts(seq, ",errors=panic");
aa22df2c 1073 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
d9c9bef1 1074 seq_puts(seq, ",nouid32");
aa22df2c 1075 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
d9c9bef1 1076 seq_puts(seq, ",debug");
03010a33 1077#ifdef CONFIG_EXT4_FS_XATTR
ea663336 1078 if (test_opt(sb, XATTR_USER))
d9c9bef1 1079 seq_puts(seq, ",user_xattr");
ea663336 1080 if (!test_opt(sb, XATTR_USER))
d9c9bef1 1081 seq_puts(seq, ",nouser_xattr");
d9c9bef1 1082#endif
03010a33 1083#ifdef CONFIG_EXT4_FS_POSIX_ACL
aa22df2c 1084 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
d9c9bef1
MS		 1085 seq_puts(seq, ",acl");
1086 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
1087 seq_puts(seq, ",noacl");
1088#endif
30773840 1089 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
d9c9bef1
MS		 1090 seq_printf(seq, ",commit=%u",
1091 (unsigned) (sbi->s_commit_interval / HZ));
1092 }
30773840
TT		 1093 if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
1094 seq_printf(seq, ",min_batch_time=%u",
1095 (unsigned) sbi->s_min_batch_time);
1096 }
1097 if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
1098 seq_printf(seq, ",max_batch_time=%u",
1099 (unsigned) sbi->s_min_batch_time);
1100 }
1101
571640ca
ES		 1102 /*
1103 * We're changing the default of barrier mount option, so
1104 * let's always display its mount state so it's clear what its
1105 * status is.
1106 */
1107 seq_puts(seq, ",barrier=");
1108 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
cd0b6a39
TT		 1109 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
1110 seq_puts(seq, ",journal_async_commit");
39a4bade
JK		 1111 else if (test_opt(sb, JOURNAL_CHECKSUM))
1112 seq_puts(seq, ",journal_checksum");
25ec56b5
JNC		 1113 if (test_opt(sb, I_VERSION))
1114 seq_puts(seq, ",i_version");
8b67f04a
TT		 1115 if (!test_opt(sb, DELALLOC) &&
1116 !(def_mount_opts & EXT4_DEFM_NODELALLOC))
dd919b98
AK		 1117 seq_puts(seq, ",nodelalloc");
1118
6fd7a467
TT		 1119 if (!test_opt(sb, MBLK_IO_SUBMIT))
1120 seq_puts(seq, ",nomblk_io_submit");
cb45bbe4
MS		 1121 if (sbi->s_stripe)
1122 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
aa22df2c
AK		 1123 /*
1124 * journal mode get enabled in different ways
1125 * So just print the value even if we didn't specify it
1126 */
617ba13b 1127 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
ac27a0ec 1128 seq_puts(seq, ",data=journal");
617ba13b 1129 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
ac27a0ec 1130 seq_puts(seq, ",data=ordered");
617ba13b 1131 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
ac27a0ec
DK		 1132 seq_puts(seq, ",data=writeback");
1133
240799cd
TT		 1134 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1135 seq_printf(seq, ",inode_readahead_blks=%u",
1136 sbi->s_inode_readahead_blks);
1137
5bf5683a
HK		 1138 if (test_opt(sb, DATA_ERR_ABORT))
1139 seq_puts(seq, ",data_err=abort");
1140
afd4672d 1141 if (test_opt(sb, NO_AUTO_DA_ALLOC))
06705bff 1142 seq_puts(seq, ",noauto_da_alloc");
afd4672d 1143
8b67f04a 1144 if (test_opt(sb, DISCARD) && !(def_mount_opts & EXT4_DEFM_DISCARD))
5328e635
ES		 1145 seq_puts(seq, ",discard");
1146
e3bb52ae
ES		 1147 if (test_opt(sb, NOLOAD))
1148 seq_puts(seq, ",norecovery");
1149
744692dc
JZ		 1150 if (test_opt(sb, DIOREAD_NOLOCK))
1151 seq_puts(seq, ",dioread_nolock");
1152
8b67f04a
TT		 1153 if (test_opt(sb, BLOCK_VALIDITY) &&
1154 !(def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY))
1155 seq_puts(seq, ",block_validity");
1156
bfff6873
LC		 1157 if (!test_opt(sb, INIT_INODE_TABLE))
1158 seq_puts(seq, ",noinit_inode_table");
51ce6511 1159 else if (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)
bfff6873
LC		 1160 seq_printf(seq, ",init_inode_table=%u",
1161 (unsigned) sbi->s_li_wait_mult);
1162
617ba13b 1163 ext4_show_quota_options(seq, sb);
0b8e58a1 1164
ac27a0ec
DK		 1165 return 0;
1166}
1167
1b961ac0 1168static struct inode *ext4_nfs_get_inode(struct super_block *sb,
0b8e58a1 1169 u64 ino, u32 generation)
ac27a0ec 1170{
ac27a0ec 1171 struct inode *inode;
ac27a0ec 1172
617ba13b 1173 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
ac27a0ec 1174 return ERR_PTR(-ESTALE);
617ba13b 1175 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
ac27a0ec
DK		 1176 return ERR_PTR(-ESTALE);
1177
1178 /* iget isn't really right if the inode is currently unallocated!!
1179 *
617ba13b 1180 * ext4_read_inode will return a bad_inode if the inode had been
ac27a0ec
DK		 1181 * deleted, so we should be safe.
1182 *
1183 * Currently we don't know the generation for parent directory, so
1184 * a generation of 0 means "accept any"
1185 */
1d1fe1ee
DH		 1186 inode = ext4_iget(sb, ino);
1187 if (IS_ERR(inode))
1188 return ERR_CAST(inode);
1189 if (generation && inode->i_generation != generation) {
ac27a0ec
DK		 1190 iput(inode);
1191 return ERR_PTR(-ESTALE);
1192 }
1b961ac0
CH		 1193
1194 return inode;
1195}
1196
1197static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
0b8e58a1 1198 int fh_len, int fh_type)
1b961ac0
CH		 1199{
1200 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1201 ext4_nfs_get_inode);
1202}
1203
1204static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
0b8e58a1 1205 int fh_len, int fh_type)
1b961ac0
CH		 1206{
1207 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1208 ext4_nfs_get_inode);
ac27a0ec
DK		 1209}
1210
c39a7f84
TO		 1211/*
1212 * Try to release metadata pages (indirect blocks, directories) which are
1213 * mapped via the block device. Since these pages could have journal heads
1214 * which would prevent try_to_free_buffers() from freeing them, we must use
1215 * jbd2 layer's try_to_free_buffers() function to release them.
1216 */
0b8e58a1
AD		 1217static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1218 gfp_t wait)
c39a7f84
TO		 1219{
1220 journal_t *journal = EXT4_SB(sb)->s_journal;
1221
1222 WARN_ON(PageChecked(page));
1223 if (!page_has_buffers(page))
1224 return 0;
1225 if (journal)
1226 return jbd2_journal_try_to_free_buffers(journal, page,
1227 wait & ~__GFP_WAIT);
1228 return try_to_free_buffers(page);
1229}
1230
ac27a0ec 1231#ifdef CONFIG_QUOTA
af5bc92d 1232#define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
2b2d6d01 1233#define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
ac27a0ec 1234
617ba13b
MC		 1235static int ext4_write_dquot(struct dquot *dquot);
1236static int ext4_acquire_dquot(struct dquot *dquot);
1237static int ext4_release_dquot(struct dquot *dquot);
1238static int ext4_mark_dquot_dirty(struct dquot *dquot);
1239static int ext4_write_info(struct super_block *sb, int type);
6f28e087 1240static int ext4_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 1241 struct path *path);
ca0e05e4 1242static int ext4_quota_off(struct super_block *sb, int type);
617ba13b
MC		 1243static int ext4_quota_on_mount(struct super_block *sb, int type);
1244static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 1245 size_t len, loff_t off);
617ba13b 1246static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 1247 const char *data, size_t len, loff_t off);
1248
61e225dc 1249static const struct dquot_operations ext4_quota_operations = {
60e58e0f 1250 .get_reserved_space = ext4_get_reserved_space,
617ba13b
MC		 1251 .write_dquot = ext4_write_dquot,
1252 .acquire_dquot = ext4_acquire_dquot,
1253 .release_dquot = ext4_release_dquot,
1254 .mark_dirty = ext4_mark_dquot_dirty,
a5b5ee32
JK		 1255 .write_info = ext4_write_info,
1256 .alloc_dquot = dquot_alloc,
1257 .destroy_dquot = dquot_destroy,
ac27a0ec
DK		 1258};
1259
0d54b217 1260static const struct quotactl_ops ext4_qctl_operations = {
617ba13b 1261 .quota_on = ext4_quota_on,
ca0e05e4 1262 .quota_off = ext4_quota_off,
287a8095
CH		 1263 .quota_sync = dquot_quota_sync,
1264 .get_info = dquot_get_dqinfo,
1265 .set_info = dquot_set_dqinfo,
1266 .get_dqblk = dquot_get_dqblk,
1267 .set_dqblk = dquot_set_dqblk
ac27a0ec
DK		 1268};
1269#endif
1270
ee9b6d61 1271static const struct super_operations ext4_sops = {
617ba13b
MC		 1272 .alloc_inode = ext4_alloc_inode,
1273 .destroy_inode = ext4_destroy_inode,
617ba13b
MC		 1274 .write_inode = ext4_write_inode,
1275 .dirty_inode = ext4_dirty_inode,
7ff9c073 1276 .drop_inode = ext4_drop_inode,
0930fcc1 1277 .evict_inode = ext4_evict_inode,
617ba13b 1278 .put_super = ext4_put_super,
617ba13b 1279 .sync_fs = ext4_sync_fs,
c4be0c1d
TS		 1280 .freeze_fs = ext4_freeze,
1281 .unfreeze_fs = ext4_unfreeze,
617ba13b
MC		 1282 .statfs = ext4_statfs,
1283 .remount_fs = ext4_remount,
617ba13b 1284 .show_options = ext4_show_options,
ac27a0ec 1285#ifdef CONFIG_QUOTA
617ba13b
MC		 1286 .quota_read = ext4_quota_read,
1287 .quota_write = ext4_quota_write,
ac27a0ec 1288#endif
c39a7f84 1289 .bdev_try_to_free_page = bdev_try_to_free_page,
ac27a0ec
DK		 1290};
1291
9ca92389
TT		 1292static const struct super_operations ext4_nojournal_sops = {
1293 .alloc_inode = ext4_alloc_inode,
1294 .destroy_inode = ext4_destroy_inode,
1295 .write_inode = ext4_write_inode,
1296 .dirty_inode = ext4_dirty_inode,
7ff9c073 1297 .drop_inode = ext4_drop_inode,
0930fcc1 1298 .evict_inode = ext4_evict_inode,
9ca92389
TT		 1299 .write_super = ext4_write_super,
1300 .put_super = ext4_put_super,
1301 .statfs = ext4_statfs,
1302 .remount_fs = ext4_remount,
9ca92389
TT		 1303 .show_options = ext4_show_options,
1304#ifdef CONFIG_QUOTA
1305 .quota_read = ext4_quota_read,
1306 .quota_write = ext4_quota_write,
1307#endif
1308 .bdev_try_to_free_page = bdev_try_to_free_page,
1309};
1310
39655164 1311static const struct export_operations ext4_export_ops = {
1b961ac0
CH		 1312 .fh_to_dentry = ext4_fh_to_dentry,
1313 .fh_to_parent = ext4_fh_to_parent,
617ba13b 1314 .get_parent = ext4_get_parent,
ac27a0ec
DK		 1315};
1316
1317enum {
1318 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1319 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
01436ef2 1320 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
ac27a0ec 1321 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
06705bff 1322 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh,
30773840 1323 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
c3191067 1324 Opt_journal_update, Opt_journal_dev,
818d276c 1325 Opt_journal_checksum, Opt_journal_async_commit,
ac27a0ec 1326 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
296c355c 1327 Opt_data_err_abort, Opt_data_err_ignore,
ac27a0ec 1328 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
5a20bdfc
JK		 1329 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1330 Opt_noquota, Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err,
1331 Opt_resize, Opt_usrquota, Opt_grpquota, Opt_i_version,
1449032b
TT		 1332 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1333 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
5328e635 1334 Opt_inode_readahead_blks, Opt_journal_ioprio,
744692dc 1335 Opt_dioread_nolock, Opt_dioread_lock,
5328e635 1336 Opt_discard, Opt_nodiscard,
bfff6873 1337 Opt_init_inode_table, Opt_noinit_inode_table,
ac27a0ec
DK		 1338};
1339
a447c093 1340static const match_table_t tokens = {
ac27a0ec
DK		 1341 {Opt_bsd_df, "bsddf"},
1342 {Opt_minix_df, "minixdf"},
1343 {Opt_grpid, "grpid"},
1344 {Opt_grpid, "bsdgroups"},
1345 {Opt_nogrpid, "nogrpid"},
1346 {Opt_nogrpid, "sysvgroups"},
1347 {Opt_resgid, "resgid=%u"},
1348 {Opt_resuid, "resuid=%u"},
1349 {Opt_sb, "sb=%u"},
1350 {Opt_err_cont, "errors=continue"},
1351 {Opt_err_panic, "errors=panic"},
1352 {Opt_err_ro, "errors=remount-ro"},
1353 {Opt_nouid32, "nouid32"},
ac27a0ec
DK		 1354 {Opt_debug, "debug"},
1355 {Opt_oldalloc, "oldalloc"},
1356 {Opt_orlov, "orlov"},
1357 {Opt_user_xattr, "user_xattr"},
1358 {Opt_nouser_xattr, "nouser_xattr"},
1359 {Opt_acl, "acl"},
1360 {Opt_noacl, "noacl"},
ac27a0ec 1361 {Opt_noload, "noload"},
e3bb52ae 1362 {Opt_noload, "norecovery"},
ac27a0ec
DK		 1363 {Opt_nobh, "nobh"},
1364 {Opt_bh, "bh"},
1365 {Opt_commit, "commit=%u"},
30773840
TT		 1366 {Opt_min_batch_time, "min_batch_time=%u"},
1367 {Opt_max_batch_time, "max_batch_time=%u"},
ac27a0ec 1368 {Opt_journal_update, "journal=update"},
ac27a0ec 1369 {Opt_journal_dev, "journal_dev=%u"},
818d276c
GS		 1370 {Opt_journal_checksum, "journal_checksum"},
1371 {Opt_journal_async_commit, "journal_async_commit"},
ac27a0ec
DK		 1372 {Opt_abort, "abort"},
1373 {Opt_data_journal, "data=journal"},
1374 {Opt_data_ordered, "data=ordered"},
1375 {Opt_data_writeback, "data=writeback"},
5bf5683a
HK		 1376 {Opt_data_err_abort, "data_err=abort"},
1377 {Opt_data_err_ignore, "data_err=ignore"},
ac27a0ec
DK		 1378 {Opt_offusrjquota, "usrjquota="},
1379 {Opt_usrjquota, "usrjquota=%s"},
1380 {Opt_offgrpjquota, "grpjquota="},
1381 {Opt_grpjquota, "grpjquota=%s"},
1382 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1383 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
5a20bdfc 1384 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
ac27a0ec
DK		 1385 {Opt_grpquota, "grpquota"},
1386 {Opt_noquota, "noquota"},
1387 {Opt_quota, "quota"},
1388 {Opt_usrquota, "usrquota"},
1389 {Opt_barrier, "barrier=%u"},
06705bff
TT		 1390 {Opt_barrier, "barrier"},
1391 {Opt_nobarrier, "nobarrier"},
25ec56b5 1392 {Opt_i_version, "i_version"},
c9de560d 1393 {Opt_stripe, "stripe=%u"},
ac27a0ec 1394 {Opt_resize, "resize"},
64769240 1395 {Opt_delalloc, "delalloc"},
dd919b98 1396 {Opt_nodelalloc, "nodelalloc"},
1449032b
TT		 1397 {Opt_mblk_io_submit, "mblk_io_submit"},
1398 {Opt_nomblk_io_submit, "nomblk_io_submit"},
6fd058f7
TT		 1399 {Opt_block_validity, "block_validity"},
1400 {Opt_noblock_validity, "noblock_validity"},
240799cd 1401 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
b3881f74 1402 {Opt_journal_ioprio, "journal_ioprio=%u"},
afd4672d 1403 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
06705bff
TT		 1404 {Opt_auto_da_alloc, "auto_da_alloc"},
1405 {Opt_noauto_da_alloc, "noauto_da_alloc"},
744692dc
JZ		 1406 {Opt_dioread_nolock, "dioread_nolock"},
1407 {Opt_dioread_lock, "dioread_lock"},
5328e635
ES		 1408 {Opt_discard, "discard"},
1409 {Opt_nodiscard, "nodiscard"},
bfff6873
LC		 1410 {Opt_init_inode_table, "init_itable=%u"},
1411 {Opt_init_inode_table, "init_itable"},
1412 {Opt_noinit_inode_table, "noinit_itable"},
f3f12faa 1413 {Opt_err, NULL},
ac27a0ec
DK		 1414};
1415
617ba13b 1416static ext4_fsblk_t get_sb_block(void **data)
ac27a0ec 1417{
617ba13b 1418 ext4_fsblk_t sb_block;
ac27a0ec
DK		 1419 char *options = (char *) *data;
1420
1421 if (!options || strncmp(options, "sb=", 3) != 0)
1422 return 1; /* Default location */
0b8e58a1 1423
ac27a0ec 1424 options += 3;
0b8e58a1 1425 /* TODO: use simple_strtoll with >32bit ext4 */
ac27a0ec
DK		 1426 sb_block = simple_strtoul(options, &options, 0);
1427 if (*options && *options != ',') {
4776004f 1428 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
ac27a0ec
DK		 1429 (char *) *data);
1430 return 1;
1431 }
1432 if (*options == ',')
1433 options++;
1434 *data = (void *) options;
0b8e58a1 1435
ac27a0ec
DK		 1436 return sb_block;
1437}
1438
b3881f74 1439#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
437ca0fd
DM		 1440static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1441 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
b3881f74 1442
56c50f11
DM		 1443#ifdef CONFIG_QUOTA
1444static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1445{
1446 struct ext4_sb_info *sbi = EXT4_SB(sb);
1447 char *qname;
1448
1449 if (sb_any_quota_loaded(sb) &&
1450 !sbi->s_qf_names[qtype]) {
1451 ext4_msg(sb, KERN_ERR,
1452 "Cannot change journaled "
1453 "quota options when quota turned on");
1454 return 0;
1455 }
1456 qname = match_strdup(args);
1457 if (!qname) {
1458 ext4_msg(sb, KERN_ERR,
1459 "Not enough memory for storing quotafile name");
1460 return 0;
1461 }
1462 if (sbi->s_qf_names[qtype] &&
1463 strcmp(sbi->s_qf_names[qtype], qname)) {
1464 ext4_msg(sb, KERN_ERR,
1465 "%s quota file already specified", QTYPE2NAME(qtype));
1466 kfree(qname);
1467 return 0;
1468 }
1469 sbi->s_qf_names[qtype] = qname;
1470 if (strchr(sbi->s_qf_names[qtype], '/')) {
1471 ext4_msg(sb, KERN_ERR,
1472 "quotafile must be on filesystem root");
1473 kfree(sbi->s_qf_names[qtype]);
1474 sbi->s_qf_names[qtype] = NULL;
1475 return 0;
1476 }
fd8c37ec 1477 set_opt(sb, QUOTA);
56c50f11
DM		 1478 return 1;
1479}
1480
1481static int clear_qf_name(struct super_block *sb, int qtype)
1482{
1483
1484 struct ext4_sb_info *sbi = EXT4_SB(sb);
1485
1486 if (sb_any_quota_loaded(sb) &&
1487 sbi->s_qf_names[qtype]) {
1488 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1489 " when quota turned on");
1490 return 0;
1491 }
1492 /*
1493 * The space will be released later when all options are confirmed
1494 * to be correct
1495 */
1496 sbi->s_qf_names[qtype] = NULL;
1497 return 1;
1498}
1499#endif
1500
2b2d6d01 1501static int parse_options(char *options, struct super_block *sb,
c3191067 1502 unsigned long *journal_devnum,
b3881f74 1503 unsigned int *journal_ioprio,
2b2d6d01 1504 ext4_fsblk_t *n_blocks_count, int is_remount)
ac27a0ec 1505{
617ba13b 1506 struct ext4_sb_info *sbi = EXT4_SB(sb);
2b2d6d01 1507 char *p;
ac27a0ec
DK		 1508 substring_t args[MAX_OPT_ARGS];
1509 int data_opt = 0;
1510 int option;
1511#ifdef CONFIG_QUOTA
56c50f11 1512 int qfmt;
ac27a0ec
DK		 1513#endif
1514
1515 if (!options)
1516 return 1;
1517
2b2d6d01 1518 while ((p = strsep(&options, ",")) != NULL) {
ac27a0ec
DK		 1519 int token;
1520 if (!*p)
1521 continue;
1522
15121c18
ES		 1523 /*
1524 * Initialize args struct so we know whether arg was
1525 * found; some options take optional arguments.
1526 */
7dc57615 1527 args[0].to = args[0].from = NULL;
ac27a0ec
DK		 1528 token = match_token(p, tokens, args);
1529 switch (token) {
1530 case Opt_bsd_df:
437ca0fd 1531 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
fd8c37ec 1532 clear_opt(sb, MINIX_DF);
ac27a0ec
DK		 1533 break;
1534 case Opt_minix_df:
437ca0fd 1535 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
fd8c37ec 1536 set_opt(sb, MINIX_DF);
437ca0fd 1537
ac27a0ec
DK		 1538 break;
1539 case Opt_grpid:
437ca0fd 1540 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
fd8c37ec 1541 set_opt(sb, GRPID);
437ca0fd 1542
ac27a0ec
DK		 1543 break;
1544 case Opt_nogrpid:
437ca0fd 1545 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
fd8c37ec 1546 clear_opt(sb, GRPID);
437ca0fd 1547
ac27a0ec
DK		 1548 break;
1549 case Opt_resuid:
1550 if (match_int(&args[0], &option))
1551 return 0;
1552 sbi->s_resuid = option;
1553 break;
1554 case Opt_resgid:
1555 if (match_int(&args[0], &option))
1556 return 0;
1557 sbi->s_resgid = option;
1558 break;
1559 case Opt_sb:
1560 /* handled by get_sb_block() instead of here */
1561 /* *sb_block = match_int(&args[0]); */
1562 break;
1563 case Opt_err_panic:
fd8c37ec
TT		 1564 clear_opt(sb, ERRORS_CONT);
1565 clear_opt(sb, ERRORS_RO);
1566 set_opt(sb, ERRORS_PANIC);
ac27a0ec
DK		 1567 break;
1568 case Opt_err_ro:
fd8c37ec
TT		 1569 clear_opt(sb, ERRORS_CONT);
1570 clear_opt(sb, ERRORS_PANIC);
1571 set_opt(sb, ERRORS_RO);
ac27a0ec
DK		 1572 break;
1573 case Opt_err_cont:
fd8c37ec
TT		 1574 clear_opt(sb, ERRORS_RO);
1575 clear_opt(sb, ERRORS_PANIC);
1576 set_opt(sb, ERRORS_CONT);
ac27a0ec
DK		 1577 break;
1578 case Opt_nouid32:
fd8c37ec 1579 set_opt(sb, NO_UID32);
ac27a0ec 1580 break;
ac27a0ec 1581 case Opt_debug:
fd8c37ec 1582 set_opt(sb, DEBUG);
ac27a0ec
DK		 1583 break;
1584 case Opt_oldalloc:
4113c4ca
LC		 1585 ext4_msg(sb, KERN_WARNING,
1586 "Ignoring deprecated oldalloc option");
ac27a0ec
DK		 1587 break;
1588 case Opt_orlov:
4113c4ca
LC		 1589 ext4_msg(sb, KERN_WARNING,
1590 "Ignoring deprecated orlov option");
ac27a0ec 1591 break;
03010a33 1592#ifdef CONFIG_EXT4_FS_XATTR
ac27a0ec 1593 case Opt_user_xattr:
fd8c37ec 1594 set_opt(sb, XATTR_USER);
ac27a0ec
DK		 1595 break;
1596 case Opt_nouser_xattr:
fd8c37ec 1597 clear_opt(sb, XATTR_USER);
ac27a0ec
DK		 1598 break;
1599#else
1600 case Opt_user_xattr:
1601 case Opt_nouser_xattr:
b31e1552 1602 ext4_msg(sb, KERN_ERR, "(no)user_xattr options not supported");
ac27a0ec
DK		 1603 break;
1604#endif
03010a33 1605#ifdef CONFIG_EXT4_FS_POSIX_ACL
ac27a0ec 1606 case Opt_acl:
fd8c37ec 1607 set_opt(sb, POSIX_ACL);
ac27a0ec
DK		 1608 break;
1609 case Opt_noacl:
fd8c37ec 1610 clear_opt(sb, POSIX_ACL);
ac27a0ec
DK		 1611 break;
1612#else
1613 case Opt_acl:
1614 case Opt_noacl:
b31e1552 1615 ext4_msg(sb, KERN_ERR, "(no)acl options not supported");
ac27a0ec
DK		 1616 break;
1617#endif
ac27a0ec
DK		 1618 case Opt_journal_update:
1619 /* @@@ FIXME */
1620 /* Eventually we will want to be able to create
1621 a journal file here. For now, only allow the
1622 user to specify an existing inode to be the
1623 journal file. */
1624 if (is_remount) {
b31e1552
ES		 1625 ext4_msg(sb, KERN_ERR,
1626 "Cannot specify journal on remount");
ac27a0ec
DK		 1627 return 0;
1628 }
fd8c37ec 1629 set_opt(sb, UPDATE_JOURNAL);
ac27a0ec 1630 break;
ac27a0ec
DK		 1631 case Opt_journal_dev:
1632 if (is_remount) {
b31e1552
ES		 1633 ext4_msg(sb, KERN_ERR,
1634 "Cannot specify journal on remount");
ac27a0ec
DK		 1635 return 0;
1636 }
1637 if (match_int(&args[0], &option))
1638 return 0;
1639 *journal_devnum = option;
1640 break;
818d276c 1641 case Opt_journal_checksum:
fd8c37ec 1642 set_opt(sb, JOURNAL_CHECKSUM);
d4da6c9c 1643 break;
818d276c 1644 case Opt_journal_async_commit:
fd8c37ec
TT		 1645 set_opt(sb, JOURNAL_ASYNC_COMMIT);
1646 set_opt(sb, JOURNAL_CHECKSUM);
818d276c 1647 break;
ac27a0ec 1648 case Opt_noload:
fd8c37ec 1649 set_opt(sb, NOLOAD);
ac27a0ec
DK		 1650 break;
1651 case Opt_commit:
1652 if (match_int(&args[0], &option))
1653 return 0;
1654 if (option < 0)
1655 return 0;
1656 if (option == 0)
cd02ff0b 1657 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
ac27a0ec
DK		 1658 sbi->s_commit_interval = HZ * option;
1659 break;
30773840
TT		 1660 case Opt_max_batch_time:
1661 if (match_int(&args[0], &option))
1662 return 0;
1663 if (option < 0)
1664 return 0;
1665 if (option == 0)
1666 option = EXT4_DEF_MAX_BATCH_TIME;
1667 sbi->s_max_batch_time = option;
1668 break;
1669 case Opt_min_batch_time:
1670 if (match_int(&args[0], &option))
1671 return 0;
1672 if (option < 0)
1673 return 0;
1674 sbi->s_min_batch_time = option;
1675 break;
ac27a0ec 1676 case Opt_data_journal:
617ba13b 1677 data_opt = EXT4_MOUNT_JOURNAL_DATA;
ac27a0ec
DK		 1678 goto datacheck;
1679 case Opt_data_ordered:
617ba13b 1680 data_opt = EXT4_MOUNT_ORDERED_DATA;
ac27a0ec
DK		 1681 goto datacheck;
1682 case Opt_data_writeback:
617ba13b 1683 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
ac27a0ec
DK		 1684 datacheck:
1685 if (is_remount) {
482a7425 1686 if (test_opt(sb, DATA_FLAGS) != data_opt) {
b31e1552
ES		 1687 ext4_msg(sb, KERN_ERR,
1688 "Cannot change data mode on remount");
ac27a0ec
DK		 1689 return 0;
1690 }
1691 } else {
fd8c37ec 1692 clear_opt(sb, DATA_FLAGS);
ac27a0ec
DK		 1693 sbi->s_mount_opt |= data_opt;
1694 }
1695 break;
5bf5683a 1696 case Opt_data_err_abort:
fd8c37ec 1697 set_opt(sb, DATA_ERR_ABORT);
5bf5683a
HK		 1698 break;
1699 case Opt_data_err_ignore:
fd8c37ec 1700 clear_opt(sb, DATA_ERR_ABORT);
5bf5683a 1701 break;
ac27a0ec
DK		 1702#ifdef CONFIG_QUOTA
1703 case Opt_usrjquota:
56c50f11 1704 if (!set_qf_name(sb, USRQUOTA, &args[0]))
ac27a0ec 1705 return 0;
56c50f11
DM		 1706 break;
1707 case Opt_grpjquota:
1708 if (!set_qf_name(sb, GRPQUOTA, &args[0]))
ac27a0ec 1709 return 0;
ac27a0ec
DK		 1710 break;
1711 case Opt_offusrjquota:
56c50f11
DM		 1712 if (!clear_qf_name(sb, USRQUOTA))
1713 return 0;
1714 break;
ac27a0ec 1715 case Opt_offgrpjquota:
56c50f11 1716 if (!clear_qf_name(sb, GRPQUOTA))
ac27a0ec 1717 return 0;
ac27a0ec 1718 break;
56c50f11 1719
ac27a0ec 1720 case Opt_jqfmt_vfsold:
dfc5d03f
JK		 1721 qfmt = QFMT_VFS_OLD;
1722 goto set_qf_format;
ac27a0ec 1723 case Opt_jqfmt_vfsv0:
dfc5d03f 1724 qfmt = QFMT_VFS_V0;
5a20bdfc
JK		 1725 goto set_qf_format;
1726 case Opt_jqfmt_vfsv1:
1727 qfmt = QFMT_VFS_V1;
dfc5d03f 1728set_qf_format:
17bd13b3 1729 if (sb_any_quota_loaded(sb) &&
dfc5d03f 1730 sbi->s_jquota_fmt != qfmt) {
b31e1552 1731 ext4_msg(sb, KERN_ERR, "Cannot change "
dfc5d03f 1732 "journaled quota options when "
b31e1552 1733 "quota turned on");
dfc5d03f
JK		 1734 return 0;
1735 }
1736 sbi->s_jquota_fmt = qfmt;
ac27a0ec
DK		 1737 break;
1738 case Opt_quota:
1739 case Opt_usrquota:
fd8c37ec
TT		 1740 set_opt(sb, QUOTA);
1741 set_opt(sb, USRQUOTA);
ac27a0ec
DK		 1742 break;
1743 case Opt_grpquota:
fd8c37ec
TT		 1744 set_opt(sb, QUOTA);
1745 set_opt(sb, GRPQUOTA);
ac27a0ec
DK		 1746 break;
1747 case Opt_noquota:
17bd13b3 1748 if (sb_any_quota_loaded(sb)) {
b31e1552
ES		 1749 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1750 "options when quota turned on");
ac27a0ec
DK		 1751 return 0;
1752 }
fd8c37ec
TT		 1753 clear_opt(sb, QUOTA);
1754 clear_opt(sb, USRQUOTA);
1755 clear_opt(sb, GRPQUOTA);
ac27a0ec
DK		 1756 break;
1757#else
1758 case Opt_quota:
1759 case Opt_usrquota:
1760 case Opt_grpquota:
b31e1552
ES		 1761 ext4_msg(sb, KERN_ERR,
1762 "quota options not supported");
cd59e7b9 1763 break;
ac27a0ec
DK		 1764 case Opt_usrjquota:
1765 case Opt_grpjquota:
1766 case Opt_offusrjquota:
1767 case Opt_offgrpjquota:
1768 case Opt_jqfmt_vfsold:
1769 case Opt_jqfmt_vfsv0:
5a20bdfc 1770 case Opt_jqfmt_vfsv1:
b31e1552
ES		 1771 ext4_msg(sb, KERN_ERR,
1772 "journaled quota options not supported");
ac27a0ec
DK		 1773 break;
1774 case Opt_noquota:
1775 break;
1776#endif
1777 case Opt_abort:
4ab2f15b 1778 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
ac27a0ec 1779 break;
06705bff 1780 case Opt_nobarrier:
fd8c37ec 1781 clear_opt(sb, BARRIER);
06705bff 1782 break;
ac27a0ec 1783 case Opt_barrier:
15121c18
ES		 1784 if (args[0].from) {
1785 if (match_int(&args[0], &option))
1786 return 0;
1787 } else
1788 option = 1; /* No argument, default to 1 */
ac27a0ec 1789 if (option)
fd8c37ec 1790 set_opt(sb, BARRIER);
ac27a0ec 1791 else
fd8c37ec 1792 clear_opt(sb, BARRIER);
ac27a0ec
DK		 1793 break;
1794 case Opt_ignore:
1795 break;
1796 case Opt_resize:
1797 if (!is_remount) {
b31e1552
ES		 1798 ext4_msg(sb, KERN_ERR,
1799 "resize option only available "
1800 "for remount");
ac27a0ec
DK		 1801 return 0;
1802 }
1803 if (match_int(&args[0], &option) != 0)
1804 return 0;
1805 *n_blocks_count = option;
1806 break;
1807 case Opt_nobh:
206f7ab4
CH		 1808 ext4_msg(sb, KERN_WARNING,
1809 "Ignoring deprecated nobh option");
ac27a0ec
DK		 1810 break;
1811 case Opt_bh:
206f7ab4
CH		 1812 ext4_msg(sb, KERN_WARNING,
1813 "Ignoring deprecated bh option");
ac27a0ec 1814 break;
25ec56b5 1815 case Opt_i_version:
fd8c37ec 1816 set_opt(sb, I_VERSION);
25ec56b5
JNC		 1817 sb->s_flags |= MS_I_VERSION;
1818 break;
dd919b98 1819 case Opt_nodelalloc:
fd8c37ec 1820 clear_opt(sb, DELALLOC);
56889787 1821 clear_opt2(sb, EXPLICIT_DELALLOC);
dd919b98 1822 break;
1449032b 1823 case Opt_mblk_io_submit:
fd8c37ec 1824 set_opt(sb, MBLK_IO_SUBMIT);
1449032b
TT		 1825 break;
1826 case Opt_nomblk_io_submit:
fd8c37ec 1827 clear_opt(sb, MBLK_IO_SUBMIT);
1449032b 1828 break;
c9de560d
AT		 1829 case Opt_stripe:
1830 if (match_int(&args[0], &option))
1831 return 0;
1832 if (option < 0)
1833 return 0;
1834 sbi->s_stripe = option;
1835 break;
64769240 1836 case Opt_delalloc:
fd8c37ec 1837 set_opt(sb, DELALLOC);
56889787 1838 set_opt2(sb, EXPLICIT_DELALLOC);
64769240 1839 break;
6fd058f7 1840 case Opt_block_validity:
fd8c37ec 1841 set_opt(sb, BLOCK_VALIDITY);
6fd058f7
TT		 1842 break;
1843 case Opt_noblock_validity:
fd8c37ec 1844 clear_opt(sb, BLOCK_VALIDITY);
6fd058f7 1845 break;
240799cd
TT		 1846 case Opt_inode_readahead_blks:
1847 if (match_int(&args[0], &option))
1848 return 0;
1849 if (option < 0 || option > (1 << 30))
1850 return 0;
5dbd571d 1851 if (option && !is_power_of_2(option)) {
b31e1552
ES		 1852 ext4_msg(sb, KERN_ERR,
1853 "EXT4-fs: inode_readahead_blks"
1854 " must be a power of 2");
3197ebdb
TT		 1855 return 0;
1856 }
240799cd
TT		 1857 sbi->s_inode_readahead_blks = option;
1858 break;
b3881f74
TT		 1859 case Opt_journal_ioprio:
1860 if (match_int(&args[0], &option))
1861 return 0;
1862 if (option < 0 || option > 7)
1863 break;
1864 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1865 option);
1866 break;
06705bff 1867 case Opt_noauto_da_alloc:
fd8c37ec 1868 set_opt(sb, NO_AUTO_DA_ALLOC);
06705bff 1869 break;
afd4672d 1870 case Opt_auto_da_alloc:
15121c18
ES		 1871 if (args[0].from) {
1872 if (match_int(&args[0], &option))
1873 return 0;
1874 } else
1875 option = 1; /* No argument, default to 1 */
afd4672d 1876 if (option)
fd8c37ec 1877 clear_opt(sb, NO_AUTO_DA_ALLOC);
afd4672d 1878 else
fd8c37ec 1879 set_opt(sb,NO_AUTO_DA_ALLOC);
afd4672d 1880 break;
5328e635 1881 case Opt_discard:
fd8c37ec 1882 set_opt(sb, DISCARD);
5328e635
ES		 1883 break;
1884 case Opt_nodiscard:
fd8c37ec 1885 clear_opt(sb, DISCARD);
5328e635 1886 break;
744692dc 1887 case Opt_dioread_nolock:
fd8c37ec 1888 set_opt(sb, DIOREAD_NOLOCK);
744692dc
JZ		 1889 break;
1890 case Opt_dioread_lock:
fd8c37ec 1891 clear_opt(sb, DIOREAD_NOLOCK);
744692dc 1892 break;
bfff6873 1893 case Opt_init_inode_table:
fd8c37ec 1894 set_opt(sb, INIT_INODE_TABLE);
bfff6873
LC		 1895 if (args[0].from) {
1896 if (match_int(&args[0], &option))
1897 return 0;
1898 } else
1899 option = EXT4_DEF_LI_WAIT_MULT;
1900 if (option < 0)
1901 return 0;
1902 sbi->s_li_wait_mult = option;
1903 break;
1904 case Opt_noinit_inode_table:
fd8c37ec 1905 clear_opt(sb, INIT_INODE_TABLE);
bfff6873 1906 break;
ac27a0ec 1907 default:
b31e1552
ES		 1908 ext4_msg(sb, KERN_ERR,
1909 "Unrecognized mount option \"%s\" "
1910 "or missing value", p);
ac27a0ec
DK		 1911 return 0;
1912 }
1913 }
1914#ifdef CONFIG_QUOTA
1915 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
482a7425 1916 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
fd8c37ec 1917 clear_opt(sb, USRQUOTA);
ac27a0ec 1918
482a7425 1919 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
fd8c37ec 1920 clear_opt(sb, GRPQUOTA);
ac27a0ec 1921
56c50f11 1922 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
b31e1552
ES		 1923 ext4_msg(sb, KERN_ERR, "old and new quota "
1924 "format mixing");
ac27a0ec
DK		 1925 return 0;
1926 }
1927
1928 if (!sbi->s_jquota_fmt) {
b31e1552
ES		 1929 ext4_msg(sb, KERN_ERR, "journaled quota format "
1930 "not specified");
ac27a0ec
DK		 1931 return 0;
1932 }
1933 } else {
1934 if (sbi->s_jquota_fmt) {
b31e1552 1935 ext4_msg(sb, KERN_ERR, "journaled quota format "
2c8be6b2 1936 "specified with no journaling "
b31e1552 1937 "enabled");
ac27a0ec
DK		 1938 return 0;
1939 }
1940 }
1941#endif
1942 return 1;
1943}
1944
617ba13b 1945static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 1946 int read_only)
1947{
617ba13b 1948 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 1949 int res = 0;
1950
617ba13b 1951 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
b31e1552
ES		 1952 ext4_msg(sb, KERN_ERR, "revision level too high, "
1953 "forcing read-only mode");
ac27a0ec
DK		 1954 res = MS_RDONLY;
1955 }
1956 if (read_only)
281b5995 1957 goto done;
617ba13b 1958 if (!(sbi->s_mount_state & EXT4_VALID_FS))
b31e1552
ES		 1959 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1960 "running e2fsck is recommended");
617ba13b 1961 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
b31e1552
ES		 1962 ext4_msg(sb, KERN_WARNING,
1963 "warning: mounting fs with errors, "
1964 "running e2fsck is recommended");
ed3ce80a 1965 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
ac27a0ec
DK		 1966 le16_to_cpu(es->s_mnt_count) >=
1967 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
b31e1552
ES		 1968 ext4_msg(sb, KERN_WARNING,
1969 "warning: maximal mount count reached, "
1970 "running e2fsck is recommended");
ac27a0ec
DK		 1971 else if (le32_to_cpu(es->s_checkinterval) &&
1972 (le32_to_cpu(es->s_lastcheck) +
1973 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
b31e1552
ES		 1974 ext4_msg(sb, KERN_WARNING,
1975 "warning: checktime reached, "
1976 "running e2fsck is recommended");
0b8e58a1 1977 if (!sbi->s_journal)
0390131b 1978 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec 1979 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 1980 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 1981 le16_add_cpu(&es->s_mnt_count, 1);
ac27a0ec 1982 es->s_mtime = cpu_to_le32(get_seconds());
617ba13b 1983 ext4_update_dynamic_rev(sb);
0390131b
FM		 1984 if (sbi->s_journal)
1985 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 1986
e2d67052 1987 ext4_commit_super(sb, 1);
281b5995 1988done:
ac27a0ec 1989 if (test_opt(sb, DEBUG))
a9df9a49 1990 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
a2595b8a 1991 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
ac27a0ec
DK		 1992 sb->s_blocksize,
1993 sbi->s_groups_count,
617ba13b
MC		 1994 EXT4_BLOCKS_PER_GROUP(sb),
1995 EXT4_INODES_PER_GROUP(sb),
a2595b8a 1996 sbi->s_mount_opt, sbi->s_mount_opt2);
ac27a0ec 1997
7abc52c2 1998 cleancache_init_fs(sb);
ac27a0ec
DK		 1999 return res;
2000}
2001
772cb7c8
JS		 2002static int ext4_fill_flex_info(struct super_block *sb)
2003{
2004 struct ext4_sb_info *sbi = EXT4_SB(sb);
2005 struct ext4_group_desc *gdp = NULL;
772cb7c8
JS		 2006 ext4_group_t flex_group_count;
2007 ext4_group_t flex_group;
2008 int groups_per_flex = 0;
c5ca7c76 2009 size_t size;
772cb7c8
JS		 2010 int i;
2011
503358ae
TT		 2012 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
2013 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
2014
2015 if (groups_per_flex < 2) {
772cb7c8
JS		 2016 sbi->s_log_groups_per_flex = 0;
2017 return 1;
2018 }
2019
c62a11fd
FB		 2020 /* We allocate both existing and potentially added groups */
2021 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
d94e99a6
AK		 2022 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
2023 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
c5ca7c76 2024 size = flex_group_count * sizeof(struct flex_groups);
9933fc0a 2025 sbi->s_flex_groups = ext4_kvzalloc(size, GFP_KERNEL);
c5ca7c76 2026 if (sbi->s_flex_groups == NULL) {
9933fc0a
TT		 2027 ext4_msg(sb, KERN_ERR, "not enough memory for %u flex groups",
2028 flex_group_count);
2029 goto failed;
772cb7c8 2030 }
772cb7c8 2031
772cb7c8 2032 for (i = 0; i < sbi->s_groups_count; i++) {
88b6edd1 2033 gdp = ext4_get_group_desc(sb, i, NULL);
772cb7c8
JS		 2034
2035 flex_group = ext4_flex_group(sbi, i);
7ad9bb65
TT		 2036 atomic_add(ext4_free_inodes_count(sb, gdp),
2037 &sbi->s_flex_groups[flex_group].free_inodes);
021b65bb 2038 atomic_add(ext4_free_group_clusters(sb, gdp),
24aaa8ef 2039 &sbi->s_flex_groups[flex_group].free_clusters);
7ad9bb65
TT		 2040 atomic_add(ext4_used_dirs_count(sb, gdp),
2041 &sbi->s_flex_groups[flex_group].used_dirs);
772cb7c8
JS		 2042 }
2043
2044 return 1;
2045failed:
2046 return 0;
2047}
2048
717d50e4
AD		 2049__le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
2050 struct ext4_group_desc *gdp)
2051{
2052 __u16 crc = 0;
2053
2054 if (sbi->s_es->s_feature_ro_compat &
2055 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
2056 int offset = offsetof(struct ext4_group_desc, bg_checksum);
2057 __le32 le_group = cpu_to_le32(block_group);
2058
2059 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2060 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2061 crc = crc16(crc, (__u8 *)gdp, offset);
2062 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2063 /* for checksum of struct ext4_group_desc do the rest...*/
2064 if ((sbi->s_es->s_feature_incompat &
2065 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
2066 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2067 crc = crc16(crc, (__u8 *)gdp + offset,
2068 le16_to_cpu(sbi->s_es->s_desc_size) -
2069 offset);
2070 }
2071
2072 return cpu_to_le16(crc);
2073}
2074
2075int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
2076 struct ext4_group_desc *gdp)
2077{
2078 if ((sbi->s_es->s_feature_ro_compat &
2079 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
2080 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
2081 return 0;
2082
2083 return 1;
2084}
2085
ac27a0ec 2086/* Called at mount-time, super-block is locked */
bfff6873
LC		 2087static int ext4_check_descriptors(struct super_block *sb,
2088 ext4_group_t *first_not_zeroed)
ac27a0ec 2089{
617ba13b
MC		 2090 struct ext4_sb_info *sbi = EXT4_SB(sb);
2091 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2092 ext4_fsblk_t last_block;
bd81d8ee
LV		 2093 ext4_fsblk_t block_bitmap;
2094 ext4_fsblk_t inode_bitmap;
2095 ext4_fsblk_t inode_table;
ce421581 2096 int flexbg_flag = 0;
bfff6873 2097 ext4_group_t i, grp = sbi->s_groups_count;
ac27a0ec 2098
ce421581
JS		 2099 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2100 flexbg_flag = 1;
2101
af5bc92d 2102 ext4_debug("Checking group descriptors");
ac27a0ec 2103
197cd65a
AM		 2104 for (i = 0; i < sbi->s_groups_count; i++) {
2105 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2106
ce421581 2107 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 2108 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 2109 else
2110 last_block = first_block +
617ba13b 2111 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 2112
bfff6873
LC		 2113 if ((grp == sbi->s_groups_count) &&
2114 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2115 grp = i;
2116
8fadc143 2117 block_bitmap = ext4_block_bitmap(sb, gdp);
2b2d6d01 2118 if (block_bitmap < first_block || block_bitmap > last_block) {
b31e1552 2119 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2120 "Block bitmap for group %u not in group "
b31e1552 2121 "(block %llu)!", i, block_bitmap);
ac27a0ec
DK		 2122 return 0;
2123 }
8fadc143 2124 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2b2d6d01 2125 if (inode_bitmap < first_block || inode_bitmap > last_block) {
b31e1552 2126 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2127 "Inode bitmap for group %u not in group "
b31e1552 2128 "(block %llu)!", i, inode_bitmap);
ac27a0ec
DK		 2129 return 0;
2130 }
8fadc143 2131 inode_table = ext4_inode_table(sb, gdp);
bd81d8ee 2132 if (inode_table < first_block ||
2b2d6d01 2133 inode_table + sbi->s_itb_per_group - 1 > last_block) {
b31e1552 2134 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2135 "Inode table for group %u not in group "
b31e1552 2136 "(block %llu)!", i, inode_table);
ac27a0ec
DK		 2137 return 0;
2138 }
955ce5f5 2139 ext4_lock_group(sb, i);
717d50e4 2140 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
b31e1552
ES		 2141 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2142 "Checksum for group %u failed (%u!=%u)",
2143 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2144 gdp)), le16_to_cpu(gdp->bg_checksum));
7ee1ec4c 2145 if (!(sb->s_flags & MS_RDONLY)) {
955ce5f5 2146 ext4_unlock_group(sb, i);
8a266467 2147 return 0;
7ee1ec4c 2148 }
717d50e4 2149 }
955ce5f5 2150 ext4_unlock_group(sb, i);
ce421581
JS		 2151 if (!flexbg_flag)
2152 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec 2153 }
bfff6873
LC		 2154 if (NULL != first_not_zeroed)
2155 *first_not_zeroed = grp;
ac27a0ec 2156
5dee5437
TT		 2157 ext4_free_blocks_count_set(sbi->s_es,
2158 EXT4_C2B(sbi, ext4_count_free_clusters(sb)));
0b8e58a1 2159 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
ac27a0ec
DK		 2160 return 1;
2161}
2162
617ba13b 2163/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 2164 * the superblock) which were deleted from all directories, but held open by
2165 * a process at the time of a crash. We walk the list and try to delete these
2166 * inodes at recovery time (only with a read-write filesystem).
2167 *
2168 * In order to keep the orphan inode chain consistent during traversal (in
2169 * case of crash during recovery), we link each inode into the superblock
2170 * orphan list_head and handle it the same way as an inode deletion during
2171 * normal operation (which journals the operations for us).
2172 *
2173 * We only do an iget() and an iput() on each inode, which is very safe if we
2174 * accidentally point at an in-use or already deleted inode. The worst that
2175 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 2176 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 2177 * e2fsck was run on this filesystem, and it must have already done the orphan
2178 * inode cleanup for us, so we can safely abort without any further action.
2179 */
2b2d6d01
TT		 2180static void ext4_orphan_cleanup(struct super_block *sb,
2181 struct ext4_super_block *es)
ac27a0ec
DK		 2182{
2183 unsigned int s_flags = sb->s_flags;
2184 int nr_orphans = 0, nr_truncates = 0;
2185#ifdef CONFIG_QUOTA
2186 int i;
2187#endif
2188 if (!es->s_last_orphan) {
2189 jbd_debug(4, "no orphan inodes to clean up\n");
2190 return;
2191 }
2192
a8f48a95 2193 if (bdev_read_only(sb->s_bdev)) {
b31e1552
ES		 2194 ext4_msg(sb, KERN_ERR, "write access "
2195 "unavailable, skipping orphan cleanup");
a8f48a95
ES		 2196 return;
2197 }
2198
d39195c3
AG		 2199 /* Check if feature set would not allow a r/w mount */
2200 if (!ext4_feature_set_ok(sb, 0)) {
2201 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2202 "unknown ROCOMPAT features");
2203 return;
2204 }
2205
617ba13b 2206 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
ac27a0ec
DK		 2207 if (es->s_last_orphan)
2208 jbd_debug(1, "Errors on filesystem, "
2209 "clearing orphan list.\n");
2210 es->s_last_orphan = 0;
2211 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2212 return;
2213 }
2214
2215 if (s_flags & MS_RDONLY) {
b31e1552 2216 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
ac27a0ec
DK		 2217 sb->s_flags &= ~MS_RDONLY;
2218 }
2219#ifdef CONFIG_QUOTA
2220 /* Needed for iput() to work correctly and not trash data */
2221 sb->s_flags |= MS_ACTIVE;
2222 /* Turn on quotas so that they are updated correctly */
2223 for (i = 0; i < MAXQUOTAS; i++) {
617ba13b
MC		 2224 if (EXT4_SB(sb)->s_qf_names[i]) {
2225 int ret = ext4_quota_on_mount(sb, i);
ac27a0ec 2226 if (ret < 0)
b31e1552
ES		 2227 ext4_msg(sb, KERN_ERR,
2228 "Cannot turn on journaled "
2229 "quota: error %d", ret);
ac27a0ec
DK		 2230 }
2231 }
2232#endif
2233
2234 while (es->s_last_orphan) {
2235 struct inode *inode;
2236
97bd42b9
JB		 2237 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2238 if (IS_ERR(inode)) {
ac27a0ec
DK		 2239 es->s_last_orphan = 0;
2240 break;
2241 }
2242
617ba13b 2243 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
871a2931 2244 dquot_initialize(inode);
ac27a0ec 2245 if (inode->i_nlink) {
b31e1552
ES		 2246 ext4_msg(sb, KERN_DEBUG,
2247 "%s: truncating inode %lu to %lld bytes",
46e665e9 2248 __func__, inode->i_ino, inode->i_size);
e5f8eab8 2249 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
ac27a0ec 2250 inode->i_ino, inode->i_size);
617ba13b 2251 ext4_truncate(inode);
ac27a0ec
DK		 2252 nr_truncates++;
2253 } else {
b31e1552
ES		 2254 ext4_msg(sb, KERN_DEBUG,
2255 "%s: deleting unreferenced inode %lu",
46e665e9 2256 __func__, inode->i_ino);
ac27a0ec
DK		 2257 jbd_debug(2, "deleting unreferenced inode %lu\n",
2258 inode->i_ino);
2259 nr_orphans++;
2260 }
2261 iput(inode); /* The delete magic happens here! */
2262 }
2263
2b2d6d01 2264#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
ac27a0ec
DK		 2265
2266 if (nr_orphans)
b31e1552
ES		 2267 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2268 PLURAL(nr_orphans));
ac27a0ec 2269 if (nr_truncates)
b31e1552
ES		 2270 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2271 PLURAL(nr_truncates));
ac27a0ec
DK		 2272#ifdef CONFIG_QUOTA
2273 /* Turn quotas off */
2274 for (i = 0; i < MAXQUOTAS; i++) {
2275 if (sb_dqopt(sb)->files[i])
287a8095 2276 dquot_quota_off(sb, i);
ac27a0ec
DK		 2277 }
2278#endif
2279 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2280}
0b8e58a1 2281
cd2291a4
ES		 2282/*
2283 * Maximal extent format file size.
2284 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2285 * extent format containers, within a sector_t, and within i_blocks
2286 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2287 * so that won't be a limiting factor.
2288 *
f17722f9
LC		 2289 * However there is other limiting factor. We do store extents in the form
2290 * of starting block and length, hence the resulting length of the extent
2291 * covering maximum file size must fit into on-disk format containers as
2292 * well. Given that length is always by 1 unit bigger than max unit (because
2293 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2294 *
cd2291a4
ES		 2295 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2296 */
f287a1a5 2297static loff_t ext4_max_size(int blkbits, int has_huge_files)
cd2291a4
ES		 2298{
2299 loff_t res;
2300 loff_t upper_limit = MAX_LFS_FILESIZE;
2301
2302 /* small i_blocks in vfs inode? */
f287a1a5 2303 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
cd2291a4 2304 /*
90c699a9 2305 * CONFIG_LBDAF is not enabled implies the inode
cd2291a4
ES		 2306 * i_block represent total blocks in 512 bytes
2307 * 32 == size of vfs inode i_blocks * 8
2308 */
2309 upper_limit = (1LL << 32) - 1;
2310
2311 /* total blocks in file system block size */
2312 upper_limit >>= (blkbits - 9);
2313 upper_limit <<= blkbits;
2314 }
2315
f17722f9
LC		 2316 /*
2317 * 32-bit extent-start container, ee_block. We lower the maxbytes
2318 * by one fs block, so ee_len can cover the extent of maximum file
2319 * size
2320 */
2321 res = (1LL << 32) - 1;
cd2291a4 2322 res <<= blkbits;
cd2291a4
ES		 2323
2324 /* Sanity check against vm- & vfs- imposed limits */
2325 if (res > upper_limit)
2326 res = upper_limit;
2327
2328 return res;
2329}
ac27a0ec 2330
ac27a0ec 2331/*
cd2291a4 2332 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 2333 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2334 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 2335 */
f287a1a5 2336static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
ac27a0ec 2337{
617ba13b 2338 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 2339 int meta_blocks;
2340 loff_t upper_limit;
0b8e58a1
AD		 2341 /* This is calculated to be the largest file size for a dense, block
2342 * mapped file such that the file's total number of 512-byte sectors,
2343 * including data and all indirect blocks, does not exceed (2^48 - 1).
2344 *
2345 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2346 * number of 512-byte sectors of the file.
0fc1b451
AK		 2347 */
2348
f287a1a5 2349 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
0fc1b451 2350 /*
90c699a9 2351 * !has_huge_files or CONFIG_LBDAF not enabled implies that
0b8e58a1
AD		 2352 * the inode i_block field represents total file blocks in
2353 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
0fc1b451
AK		 2354 */
2355 upper_limit = (1LL << 32) - 1;
2356
2357 /* total blocks in file system block size */
2358 upper_limit >>= (bits - 9);
2359
2360 } else {
8180a562
AK		 2361 /*
2362 * We use 48 bit ext4_inode i_blocks
2363 * With EXT4_HUGE_FILE_FL set the i_blocks
2364 * represent total number of blocks in
2365 * file system block size
2366 */
0fc1b451
AK		 2367 upper_limit = (1LL << 48) - 1;
2368
0fc1b451
AK		 2369 }
2370
2371 /* indirect blocks */
2372 meta_blocks = 1;
2373 /* double indirect blocks */
2374 meta_blocks += 1 + (1LL << (bits-2));
2375 /* tripple indirect blocks */
2376 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2377
2378 upper_limit -= meta_blocks;
2379 upper_limit <<= bits;
ac27a0ec
DK		 2380
2381 res += 1LL << (bits-2);
2382 res += 1LL << (2*(bits-2));
2383 res += 1LL << (3*(bits-2));
2384 res <<= bits;
2385 if (res > upper_limit)
2386 res = upper_limit;
0fc1b451
AK		 2387
2388 if (res > MAX_LFS_FILESIZE)
2389 res = MAX_LFS_FILESIZE;
2390
ac27a0ec
DK		 2391 return res;
2392}
2393
617ba13b 2394static ext4_fsblk_t descriptor_loc(struct super_block *sb,
0b8e58a1 2395 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 2396{
617ba13b 2397 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 2398 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 2399 int has_super = 0;
2400
2401 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2402
617ba13b 2403 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
ac27a0ec 2404 nr < first_meta_bg)
70bbb3e0 2405 return logical_sb_block + nr + 1;
ac27a0ec 2406 bg = sbi->s_desc_per_block * nr;
617ba13b 2407 if (ext4_bg_has_super(sb, bg))
ac27a0ec 2408 has_super = 1;
0b8e58a1 2409
617ba13b 2410 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 2411}
2412
c9de560d
AT		 2413/**
2414 * ext4_get_stripe_size: Get the stripe size.
2415 * @sbi: In memory super block info
2416 *
2417 * If we have specified it via mount option, then
2418 * use the mount option value. If the value specified at mount time is
2419 * greater than the blocks per group use the super block value.
2420 * If the super block value is greater than blocks per group return 0.
2421 * Allocator needs it be less than blocks per group.
2422 *
2423 */
2424static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2425{
2426 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2427 unsigned long stripe_width =
2428 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3eb08658 2429 int ret;
c9de560d
AT		 2430
2431 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3eb08658
DE		 2432 ret = sbi->s_stripe;
2433 else if (stripe_width <= sbi->s_blocks_per_group)
2434 ret = stripe_width;
2435 else if (stride <= sbi->s_blocks_per_group)
2436 ret = stride;
2437 else
2438 ret = 0;
c9de560d 2439
3eb08658
DE		 2440 /*
2441 * If the stripe width is 1, this makes no sense and
2442 * we set it to 0 to turn off stripe handling code.
2443 */
2444 if (ret <= 1)
2445 ret = 0;
c9de560d 2446
3eb08658 2447 return ret;
c9de560d 2448}
ac27a0ec 2449
3197ebdb
TT		 2450/* sysfs supprt */
2451
2452struct ext4_attr {
2453 struct attribute attr;
2454 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
60e6679e 2455 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
3197ebdb
TT		 2456 const char *, size_t);
2457 int offset;
2458};
2459
2460static int parse_strtoul(const char *buf,
2461 unsigned long max, unsigned long *value)
2462{
2463 char *endp;
2464
e7d2860b
AGR		 2465 *value = simple_strtoul(skip_spaces(buf), &endp, 0);
2466 endp = skip_spaces(endp);
3197ebdb
TT		 2467 if (*endp || *value > max)
2468 return -EINVAL;
2469
2470 return 0;
2471}
2472
2473static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2474 struct ext4_sb_info *sbi,
2475 char *buf)
2476{
2477 return snprintf(buf, PAGE_SIZE, "%llu\n",
7b415bf6
AK		 2478 (s64) EXT4_C2B(sbi,
2479 percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
3197ebdb
TT		 2480}
2481
2482static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2483 struct ext4_sb_info *sbi, char *buf)
2484{
2485 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2486
f613dfcb
TT		 2487 if (!sb->s_bdev->bd_part)
2488 return snprintf(buf, PAGE_SIZE, "0\n");
3197ebdb
TT		 2489 return snprintf(buf, PAGE_SIZE, "%lu\n",
2490 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2491 sbi->s_sectors_written_start) >> 1);
2492}
2493
2494static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2495 struct ext4_sb_info *sbi, char *buf)
2496{
2497 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2498
f613dfcb
TT		 2499 if (!sb->s_bdev->bd_part)
2500 return snprintf(buf, PAGE_SIZE, "0\n");
3197ebdb 2501 return snprintf(buf, PAGE_SIZE, "%llu\n",
a6b43e38 2502 (unsigned long long)(sbi->s_kbytes_written +
3197ebdb 2503 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
a6b43e38 2504 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
3197ebdb
TT		 2505}
2506
77f4135f
VH		 2507static ssize_t extent_cache_hits_show(struct ext4_attr *a,
2508 struct ext4_sb_info *sbi, char *buf)
2509{
2510 return snprintf(buf, PAGE_SIZE, "%lu\n", sbi->extent_cache_hits);
2511}
2512
2513static ssize_t extent_cache_misses_show(struct ext4_attr *a,
2514 struct ext4_sb_info *sbi, char *buf)
2515{
2516 return snprintf(buf, PAGE_SIZE, "%lu\n", sbi->extent_cache_misses);
2517}
2518
3197ebdb
TT		 2519static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2520 struct ext4_sb_info *sbi,
2521 const char *buf, size_t count)
2522{
2523 unsigned long t;
2524
2525 if (parse_strtoul(buf, 0x40000000, &t))
2526 return -EINVAL;
2527
5dbd571d 2528 if (t && !is_power_of_2(t))
3197ebdb
TT		 2529 return -EINVAL;
2530
2531 sbi->s_inode_readahead_blks = t;
2532 return count;
2533}
2534
2535static ssize_t sbi_ui_show(struct ext4_attr *a,
0b8e58a1 2536 struct ext4_sb_info *sbi, char *buf)
3197ebdb
TT		 2537{
2538 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2539
2540 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2541}
2542
2543static ssize_t sbi_ui_store(struct ext4_attr *a,
2544 struct ext4_sb_info *sbi,
2545 const char *buf, size_t count)
2546{
2547 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2548 unsigned long t;
2549
2550 if (parse_strtoul(buf, 0xffffffff, &t))
2551 return -EINVAL;
2552 *ui = t;
2553 return count;
2554}
2555
2556#define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2557static struct ext4_attr ext4_attr_##_name = { \
2558 .attr = {.name = __stringify(_name), .mode = _mode }, \
2559 .show = _show, \
2560 .store = _store, \
2561 .offset = offsetof(struct ext4_sb_info, _elname), \
2562}
2563#define EXT4_ATTR(name, mode, show, store) \
2564static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2565
857ac889 2566#define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
3197ebdb
TT		 2567#define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2568#define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2569#define EXT4_RW_ATTR_SBI_UI(name, elname) \
2570 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2571#define ATTR_LIST(name) &ext4_attr_##name.attr
2572
2573EXT4_RO_ATTR(delayed_allocation_blocks);
2574EXT4_RO_ATTR(session_write_kbytes);
2575EXT4_RO_ATTR(lifetime_write_kbytes);
77f4135f
VH		 2576EXT4_RO_ATTR(extent_cache_hits);
2577EXT4_RO_ATTR(extent_cache_misses);
3197ebdb
TT		 2578EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2579 inode_readahead_blks_store, s_inode_readahead_blks);
11013911 2580EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
3197ebdb
TT		 2581EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2582EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2583EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2584EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2585EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2586EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
55138e0b 2587EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
3197ebdb
TT		 2588
2589static struct attribute *ext4_attrs[] = {
2590 ATTR_LIST(delayed_allocation_blocks),
2591 ATTR_LIST(session_write_kbytes),
2592 ATTR_LIST(lifetime_write_kbytes),
77f4135f
VH		 2593 ATTR_LIST(extent_cache_hits),
2594 ATTR_LIST(extent_cache_misses),
3197ebdb 2595 ATTR_LIST(inode_readahead_blks),
11013911 2596 ATTR_LIST(inode_goal),
3197ebdb
TT		 2597 ATTR_LIST(mb_stats),
2598 ATTR_LIST(mb_max_to_scan),
2599 ATTR_LIST(mb_min_to_scan),
2600 ATTR_LIST(mb_order2_req),
2601 ATTR_LIST(mb_stream_req),
2602 ATTR_LIST(mb_group_prealloc),
55138e0b 2603 ATTR_LIST(max_writeback_mb_bump),
3197ebdb
TT		 2604 NULL,
2605};
2606
857ac889
LC		 2607/* Features this copy of ext4 supports */
2608EXT4_INFO_ATTR(lazy_itable_init);
27ee40df 2609EXT4_INFO_ATTR(batched_discard);
857ac889
LC		 2610
2611static struct attribute *ext4_feat_attrs[] = {
2612 ATTR_LIST(lazy_itable_init),
27ee40df 2613 ATTR_LIST(batched_discard),
857ac889
LC		 2614 NULL,
2615};
2616
3197ebdb
TT		 2617static ssize_t ext4_attr_show(struct kobject *kobj,
2618 struct attribute *attr, char *buf)
2619{
2620 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2621 s_kobj);
2622 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2623
2624 return a->show ? a->show(a, sbi, buf) : 0;
2625}
2626
2627static ssize_t ext4_attr_store(struct kobject *kobj,
2628 struct attribute *attr,
2629 const char *buf, size_t len)
2630{
2631 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2632 s_kobj);
2633 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2634
2635 return a->store ? a->store(a, sbi, buf, len) : 0;
2636}
2637
2638static void ext4_sb_release(struct kobject *kobj)
2639{
2640 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2641 s_kobj);
2642 complete(&sbi->s_kobj_unregister);
2643}
2644
52cf25d0 2645static const struct sysfs_ops ext4_attr_ops = {
3197ebdb
TT		 2646 .show = ext4_attr_show,
2647 .store = ext4_attr_store,
2648};
2649
2650static struct kobj_type ext4_ktype = {
2651 .default_attrs = ext4_attrs,
2652 .sysfs_ops = &ext4_attr_ops,
2653 .release = ext4_sb_release,
2654};
2655
857ac889
LC		 2656static void ext4_feat_release(struct kobject *kobj)
2657{
2658 complete(&ext4_feat->f_kobj_unregister);
2659}
2660
2661static struct kobj_type ext4_feat_ktype = {
2662 .default_attrs = ext4_feat_attrs,
2663 .sysfs_ops = &ext4_attr_ops,
2664 .release = ext4_feat_release,
2665};
2666
a13fb1a4
ES		 2667/*
2668 * Check whether this filesystem can be mounted based on
2669 * the features present and the RDONLY/RDWR mount requested.
2670 * Returns 1 if this filesystem can be mounted as requested,
2671 * 0 if it cannot be.
2672 */
2673static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2674{
2675 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2676 ext4_msg(sb, KERN_ERR,
2677 "Couldn't mount because of "
2678 "unsupported optional features (%x)",
2679 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2680 ~EXT4_FEATURE_INCOMPAT_SUPP));
2681 return 0;
2682 }
2683
2684 if (readonly)
2685 return 1;
2686
2687 /* Check that feature set is OK for a read-write mount */
2688 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2689 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2690 "unsupported optional features (%x)",
2691 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2692 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2693 return 0;
2694 }
2695 /*
2696 * Large file size enabled file system can only be mounted
2697 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2698 */
2699 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2700 if (sizeof(blkcnt_t) < sizeof(u64)) {
2701 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2702 "cannot be mounted RDWR without "
2703 "CONFIG_LBDAF");
2704 return 0;
2705 }
2706 }
bab08ab9
TT		 2707 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2708 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2709 ext4_msg(sb, KERN_ERR,
2710 "Can't support bigalloc feature without "
2711 "extents feature\n");
2712 return 0;
2713 }
a13fb1a4
ES		 2714 return 1;
2715}
2716
66e61a9e
TT		 2717/*
2718 * This function is called once a day if we have errors logged
2719 * on the file system
2720 */
2721static void print_daily_error_info(unsigned long arg)
2722{
2723 struct super_block *sb = (struct super_block *) arg;
2724 struct ext4_sb_info *sbi;
2725 struct ext4_super_block *es;
2726
2727 sbi = EXT4_SB(sb);
2728 es = sbi->s_es;
2729
2730 if (es->s_error_count)
2731 ext4_msg(sb, KERN_NOTICE, "error count: %u",
2732 le32_to_cpu(es->s_error_count));
2733 if (es->s_first_error_time) {
2734 printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d",
2735 sb->s_id, le32_to_cpu(es->s_first_error_time),
2736 (int) sizeof(es->s_first_error_func),
2737 es->s_first_error_func,
2738 le32_to_cpu(es->s_first_error_line));
2739 if (es->s_first_error_ino)
2740 printk(": inode %u",
2741 le32_to_cpu(es->s_first_error_ino));
2742 if (es->s_first_error_block)
2743 printk(": block %llu", (unsigned long long)
2744 le64_to_cpu(es->s_first_error_block));
2745 printk("\n");
2746 }
2747 if (es->s_last_error_time) {
2748 printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d",
2749 sb->s_id, le32_to_cpu(es->s_last_error_time),
2750 (int) sizeof(es->s_last_error_func),
2751 es->s_last_error_func,
2752 le32_to_cpu(es->s_last_error_line));
2753 if (es->s_last_error_ino)
2754 printk(": inode %u",
2755 le32_to_cpu(es->s_last_error_ino));
2756 if (es->s_last_error_block)
2757 printk(": block %llu", (unsigned long long)
2758 le64_to_cpu(es->s_last_error_block));
2759 printk("\n");
2760 }
2761 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2762}
2763
bfff6873
LC		 2764/* Find next suitable group and run ext4_init_inode_table */
2765static int ext4_run_li_request(struct ext4_li_request *elr)
2766{
2767 struct ext4_group_desc *gdp = NULL;
2768 ext4_group_t group, ngroups;
2769 struct super_block *sb;
2770 unsigned long timeout = 0;
2771 int ret = 0;
2772
2773 sb = elr->lr_super;
2774 ngroups = EXT4_SB(sb)->s_groups_count;
2775
2776 for (group = elr->lr_next_group; group < ngroups; group++) {
2777 gdp = ext4_get_group_desc(sb, group, NULL);
2778 if (!gdp) {
2779 ret = 1;
2780 break;
2781 }
2782
2783 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2784 break;
2785 }
2786
2787 if (group == ngroups)
2788 ret = 1;
2789
2790 if (!ret) {
2791 timeout = jiffies;
2792 ret = ext4_init_inode_table(sb, group,
2793 elr->lr_timeout ? 0 : 1);
2794 if (elr->lr_timeout == 0) {
51ce6511
LC		 2795 timeout = (jiffies - timeout) *
2796 elr->lr_sbi->s_li_wait_mult;
bfff6873
LC		 2797 elr->lr_timeout = timeout;
2798 }
2799 elr->lr_next_sched = jiffies + elr->lr_timeout;
2800 elr->lr_next_group = group + 1;
2801 }
2802
2803 return ret;
2804}
2805
2806/*
2807 * Remove lr_request from the list_request and free the
4ed5c033 2808 * request structure. Should be called with li_list_mtx held
bfff6873
LC		 2809 */
2810static void ext4_remove_li_request(struct ext4_li_request *elr)
2811{
2812 struct ext4_sb_info *sbi;
2813
2814 if (!elr)
2815 return;
2816
2817 sbi = elr->lr_sbi;
2818
2819 list_del(&elr->lr_request);
2820 sbi->s_li_request = NULL;
2821 kfree(elr);
2822}
2823
2824static void ext4_unregister_li_request(struct super_block *sb)
2825{
1bb933fb
LC		 2826 mutex_lock(&ext4_li_mtx);
2827 if (!ext4_li_info) {
2828 mutex_unlock(&ext4_li_mtx);
bfff6873 2829 return;
1bb933fb 2830 }
bfff6873
LC		 2831
2832 mutex_lock(&ext4_li_info->li_list_mtx);
1bb933fb 2833 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
bfff6873 2834 mutex_unlock(&ext4_li_info->li_list_mtx);
1bb933fb 2835 mutex_unlock(&ext4_li_mtx);
bfff6873
LC		 2836}
2837
8f1f7453
ES		 2838static struct task_struct *ext4_lazyinit_task;
2839
bfff6873
LC		 2840/*
2841 * This is the function where ext4lazyinit thread lives. It walks
2842 * through the request list searching for next scheduled filesystem.
2843 * When such a fs is found, run the lazy initialization request
2844 * (ext4_rn_li_request) and keep track of the time spend in this
2845 * function. Based on that time we compute next schedule time of
2846 * the request. When walking through the list is complete, compute
2847 * next waking time and put itself into sleep.
2848 */
2849static int ext4_lazyinit_thread(void *arg)
2850{
2851 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2852 struct list_head *pos, *n;
2853 struct ext4_li_request *elr;
4ed5c033 2854 unsigned long next_wakeup, cur;
bfff6873
LC		 2855
2856 BUG_ON(NULL == eli);
2857
bfff6873
LC		 2858cont_thread:
2859 while (true) {
2860 next_wakeup = MAX_JIFFY_OFFSET;
2861
2862 mutex_lock(&eli->li_list_mtx);
2863 if (list_empty(&eli->li_request_list)) {
2864 mutex_unlock(&eli->li_list_mtx);
2865 goto exit_thread;
2866 }
2867
2868 list_for_each_safe(pos, n, &eli->li_request_list) {
2869 elr = list_entry(pos, struct ext4_li_request,
2870 lr_request);
2871
b2c78cd0
TT		 2872 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2873 if (ext4_run_li_request(elr) != 0) {
2874 /* error, remove the lazy_init job */
2875 ext4_remove_li_request(elr);
2876 continue;
2877 }
bfff6873
LC		 2878 }
2879
2880 if (time_before(elr->lr_next_sched, next_wakeup))
2881 next_wakeup = elr->lr_next_sched;
2882 }
2883 mutex_unlock(&eli->li_list_mtx);
2884
a0acae0e 2885 try_to_freeze();
bfff6873 2886
4ed5c033
LC		 2887 cur = jiffies;
2888 if ((time_after_eq(cur, next_wakeup)) ||
f4245bd4 2889 (MAX_JIFFY_OFFSET == next_wakeup)) {
bfff6873
LC		 2890 cond_resched();
2891 continue;
2892 }
2893
4ed5c033
LC		 2894 schedule_timeout_interruptible(next_wakeup - cur);
2895
8f1f7453
ES		 2896 if (kthread_should_stop()) {
2897 ext4_clear_request_list();
2898 goto exit_thread;
2899 }
bfff6873
LC		 2900 }
2901
2902exit_thread:
2903 /*
2904 * It looks like the request list is empty, but we need
2905 * to check it under the li_list_mtx lock, to prevent any
2906 * additions into it, and of course we should lock ext4_li_mtx
2907 * to atomically free the list and ext4_li_info, because at
2908 * this point another ext4 filesystem could be registering
2909 * new one.
2910 */
2911 mutex_lock(&ext4_li_mtx);
2912 mutex_lock(&eli->li_list_mtx);
2913 if (!list_empty(&eli->li_request_list)) {
2914 mutex_unlock(&eli->li_list_mtx);
2915 mutex_unlock(&ext4_li_mtx);
2916 goto cont_thread;
2917 }
2918 mutex_unlock(&eli->li_list_mtx);
bfff6873
LC		 2919 kfree(ext4_li_info);
2920 ext4_li_info = NULL;
2921 mutex_unlock(&ext4_li_mtx);
2922
2923 return 0;
2924}
2925
2926static void ext4_clear_request_list(void)
2927{
2928 struct list_head *pos, *n;
2929 struct ext4_li_request *elr;
2930
2931 mutex_lock(&ext4_li_info->li_list_mtx);
bfff6873
LC		 2932 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2933 elr = list_entry(pos, struct ext4_li_request,
2934 lr_request);
2935 ext4_remove_li_request(elr);
2936 }
2937 mutex_unlock(&ext4_li_info->li_list_mtx);
2938}
2939
2940static int ext4_run_lazyinit_thread(void)
2941{
8f1f7453
ES		 2942 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2943 ext4_li_info, "ext4lazyinit");
2944 if (IS_ERR(ext4_lazyinit_task)) {
2945 int err = PTR_ERR(ext4_lazyinit_task);
bfff6873 2946 ext4_clear_request_list();
bfff6873
LC		 2947 kfree(ext4_li_info);
2948 ext4_li_info = NULL;
2949 printk(KERN_CRIT "EXT4: error %d creating inode table "
2950 "initialization thread\n",
2951 err);
2952 return err;
2953 }
2954 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
bfff6873
LC		 2955 return 0;
2956}
2957
2958/*
2959 * Check whether it make sense to run itable init. thread or not.
2960 * If there is at least one uninitialized inode table, return
2961 * corresponding group number, else the loop goes through all
2962 * groups and return total number of groups.
2963 */
2964static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2965{
2966 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2967 struct ext4_group_desc *gdp = NULL;
2968
2969 for (group = 0; group < ngroups; group++) {
2970 gdp = ext4_get_group_desc(sb, group, NULL);
2971 if (!gdp)
2972 continue;
2973
2974 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2975 break;
2976 }
2977
2978 return group;
2979}
2980
2981static int ext4_li_info_new(void)
2982{
2983 struct ext4_lazy_init *eli = NULL;
2984
2985 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2986 if (!eli)
2987 return -ENOMEM;
2988
bfff6873
LC		 2989 INIT_LIST_HEAD(&eli->li_request_list);
2990 mutex_init(&eli->li_list_mtx);
2991
bfff6873
LC		 2992 eli->li_state |= EXT4_LAZYINIT_QUIT;
2993
2994 ext4_li_info = eli;
2995
2996 return 0;
2997}
2998
2999static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3000 ext4_group_t start)
3001{
3002 struct ext4_sb_info *sbi = EXT4_SB(sb);
3003 struct ext4_li_request *elr;
3004 unsigned long rnd;
3005
3006 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3007 if (!elr)
3008 return NULL;
3009
3010 elr->lr_super = sb;
3011 elr->lr_sbi = sbi;
3012 elr->lr_next_group = start;
3013
3014 /*
3015 * Randomize first schedule time of the request to
3016 * spread the inode table initialization requests
3017 * better.
3018 */
3019 get_random_bytes(&rnd, sizeof(rnd));
3020 elr->lr_next_sched = jiffies + (unsigned long)rnd %
3021 (EXT4_DEF_LI_MAX_START_DELAY * HZ);
3022
3023 return elr;
3024}
3025
3026static int ext4_register_li_request(struct super_block *sb,
3027 ext4_group_t first_not_zeroed)
3028{
3029 struct ext4_sb_info *sbi = EXT4_SB(sb);
3030 struct ext4_li_request *elr;
3031 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
6c5a6cb9 3032 int ret = 0;
bfff6873 3033
51ce6511
LC		 3034 if (sbi->s_li_request != NULL) {
3035 /*
3036 * Reset timeout so it can be computed again, because
3037 * s_li_wait_mult might have changed.
3038 */
3039 sbi->s_li_request->lr_timeout = 0;
beed5ecb 3040 return 0;
51ce6511 3041 }
bfff6873
LC		 3042
3043 if (first_not_zeroed == ngroups ||
3044 (sb->s_flags & MS_RDONLY) ||
55ff3840 3045 !test_opt(sb, INIT_INODE_TABLE))
beed5ecb 3046 return 0;
bfff6873
LC		 3047
3048 elr = ext4_li_request_new(sb, first_not_zeroed);
beed5ecb
NK		 3049 if (!elr)
3050 return -ENOMEM;
bfff6873
LC		 3051
3052 mutex_lock(&ext4_li_mtx);
3053
3054 if (NULL == ext4_li_info) {
3055 ret = ext4_li_info_new();
3056 if (ret)
3057 goto out;
3058 }
3059
3060 mutex_lock(&ext4_li_info->li_list_mtx);
3061 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3062 mutex_unlock(&ext4_li_info->li_list_mtx);
3063
3064 sbi->s_li_request = elr;
46e4690b
TM		 3065 /*
3066 * set elr to NULL here since it has been inserted to
3067 * the request_list and the removal and free of it is
3068 * handled by ext4_clear_request_list from now on.
3069 */
3070 elr = NULL;
bfff6873
LC		 3071
3072 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3073 ret = ext4_run_lazyinit_thread();
3074 if (ret)
3075 goto out;
3076 }
bfff6873 3077out:
beed5ecb
NK		 3078 mutex_unlock(&ext4_li_mtx);
3079 if (ret)
bfff6873 3080 kfree(elr);
bfff6873
LC		 3081 return ret;
3082}
3083
3084/*
3085 * We do not need to lock anything since this is called on
3086 * module unload.
3087 */
3088static void ext4_destroy_lazyinit_thread(void)
3089{
3090 /*
3091 * If thread exited earlier
3092 * there's nothing to be done.
3093 */
8f1f7453 3094 if (!ext4_li_info || !ext4_lazyinit_task)
bfff6873
LC		 3095 return;
3096
8f1f7453 3097 kthread_stop(ext4_lazyinit_task);
bfff6873
LC		 3098}
3099
2b2d6d01 3100static int ext4_fill_super(struct super_block *sb, void *data, int silent)
7477827f
AK		 3101 __releases(kernel_lock)
3102 __acquires(kernel_lock)
ac27a0ec 3103{
d4c402d9 3104 char *orig_data = kstrdup(data, GFP_KERNEL);
2b2d6d01 3105 struct buffer_head *bh;
617ba13b
MC		 3106 struct ext4_super_block *es = NULL;
3107 struct ext4_sb_info *sbi;
3108 ext4_fsblk_t block;
3109 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 3110 ext4_fsblk_t logical_sb_block;
ac27a0ec 3111 unsigned long offset = 0;
ac27a0ec
DK		 3112 unsigned long journal_devnum = 0;
3113 unsigned long def_mount_opts;
3114 struct inode *root;
9f6200bb 3115 char *cp;
0390131b 3116 const char *descr;
dcc7dae3 3117 int ret = -ENOMEM;
281b5995 3118 int blocksize, clustersize;
4ec11028
TT		 3119 unsigned int db_count;
3120 unsigned int i;
281b5995 3121 int needs_recovery, has_huge_files, has_bigalloc;
bd81d8ee 3122 __u64 blocks_count;
833f4077 3123 int err;
b3881f74 3124 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
bfff6873 3125 ext4_group_t first_not_zeroed;
ac27a0ec
DK		 3126
3127 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3128 if (!sbi)
dcc7dae3 3129 goto out_free_orig;
705895b6
PE		 3130
3131 sbi->s_blockgroup_lock =
3132 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3133 if (!sbi->s_blockgroup_lock) {
3134 kfree(sbi);
dcc7dae3 3135 goto out_free_orig;
705895b6 3136 }
ac27a0ec
DK		 3137 sb->s_fs_info = sbi;
3138 sbi->s_mount_opt = 0;
617ba13b
MC		 3139 sbi->s_resuid = EXT4_DEF_RESUID;
3140 sbi->s_resgid = EXT4_DEF_RESGID;
240799cd 3141 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
d9c9bef1 3142 sbi->s_sb_block = sb_block;
f613dfcb
TT		 3143 if (sb->s_bdev->bd_part)
3144 sbi->s_sectors_written_start =
3145 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
ac27a0ec 3146
9f6200bb
TT		 3147 /* Cleanup superblock name */
3148 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3149 *cp = '!';
3150
dcc7dae3 3151 ret = -EINVAL;
617ba13b 3152 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 3153 if (!blocksize) {
b31e1552 3154 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
ac27a0ec
DK		 3155 goto out_fail;
3156 }
3157
3158 /*
617ba13b 3159 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 3160 * block sizes. We need to calculate the offset from buffer start.
3161 */
617ba13b 3162 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 3163 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3164 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 3165 } else {
70bbb3e0 3166 logical_sb_block = sb_block;
ac27a0ec
DK		 3167 }
3168
70bbb3e0 3169 if (!(bh = sb_bread(sb, logical_sb_block))) {
b31e1552 3170 ext4_msg(sb, KERN_ERR, "unable to read superblock");
ac27a0ec
DK		 3171 goto out_fail;
3172 }
3173 /*
3174 * Note: s_es must be initialized as soon as possible because
617ba13b 3175 * some ext4 macro-instructions depend on its value
ac27a0ec 3176 */
617ba13b 3177 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
ac27a0ec
DK		 3178 sbi->s_es = es;
3179 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 3180 if (sb->s_magic != EXT4_SUPER_MAGIC)
3181 goto cantfind_ext4;
afc32f7e 3182 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
ac27a0ec
DK		 3183
3184 /* Set defaults before we parse the mount options */
3185 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
fd8c37ec 3186 set_opt(sb, INIT_INODE_TABLE);
617ba13b 3187 if (def_mount_opts & EXT4_DEFM_DEBUG)
fd8c37ec 3188 set_opt(sb, DEBUG);
437ca0fd
DM		 3189 if (def_mount_opts & EXT4_DEFM_BSDGROUPS) {
3190 ext4_msg(sb, KERN_WARNING, deprecated_msg, "bsdgroups",
3191 "2.6.38");
fd8c37ec 3192 set_opt(sb, GRPID);
437ca0fd 3193 }
617ba13b 3194 if (def_mount_opts & EXT4_DEFM_UID16)
fd8c37ec 3195 set_opt(sb, NO_UID32);
ea663336 3196 /* xattr user namespace & acls are now defaulted on */
03010a33 3197#ifdef CONFIG_EXT4_FS_XATTR
ea663336 3198 set_opt(sb, XATTR_USER);
2e7842b8 3199#endif
03010a33 3200#ifdef CONFIG_EXT4_FS_POSIX_ACL
ea663336 3201 set_opt(sb, POSIX_ACL);
2e7842b8 3202#endif
6fd7a467 3203 set_opt(sb, MBLK_IO_SUBMIT);
617ba13b 3204 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
fd8c37ec 3205 set_opt(sb, JOURNAL_DATA);
617ba13b 3206 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
fd8c37ec 3207 set_opt(sb, ORDERED_DATA);
617ba13b 3208 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
fd8c37ec 3209 set_opt(sb, WRITEBACK_DATA);
617ba13b
MC		 3210
3211 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
fd8c37ec 3212 set_opt(sb, ERRORS_PANIC);
bb4f397a 3213 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
fd8c37ec 3214 set_opt(sb, ERRORS_CONT);
bb4f397a 3215 else
fd8c37ec 3216 set_opt(sb, ERRORS_RO);
8b67f04a 3217 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
fd8c37ec 3218 set_opt(sb, BLOCK_VALIDITY);
8b67f04a 3219 if (def_mount_opts & EXT4_DEFM_DISCARD)
fd8c37ec 3220 set_opt(sb, DISCARD);
ac27a0ec
DK		 3221
3222 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
3223 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
30773840
TT		 3224 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3225 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3226 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
ac27a0ec 3227
8b67f04a 3228 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
fd8c37ec 3229 set_opt(sb, BARRIER);
ac27a0ec 3230
dd919b98
AK		 3231 /*
3232 * enable delayed allocation by default
3233 * Use -o nodelalloc to turn it off
3234 */
8b67f04a
TT		 3235 if (!IS_EXT3_SB(sb) &&
3236 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
fd8c37ec 3237 set_opt(sb, DELALLOC);
dd919b98 3238
51ce6511
LC		 3239 /*
3240 * set default s_li_wait_mult for lazyinit, for the case there is
3241 * no mount option specified.
3242 */
3243 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3244
8b67f04a
TT		 3245 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3246 &journal_devnum, &journal_ioprio, NULL, 0)) {
3247 ext4_msg(sb, KERN_WARNING,
3248 "failed to parse options in superblock: %s",
3249 sbi->s_es->s_mount_opts);
3250 }
b3881f74
TT		 3251 if (!parse_options((char *) data, sb, &journal_devnum,
3252 &journal_ioprio, NULL, 0))
ac27a0ec
DK		 3253 goto failed_mount;
3254
56889787
TT		 3255 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3256 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3257 "with data=journal disables delayed "
3258 "allocation and O_DIRECT support!\n");
3259 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3260 ext4_msg(sb, KERN_ERR, "can't mount with "
3261 "both data=journal and delalloc");
3262 goto failed_mount;
3263 }
3264 if (test_opt(sb, DIOREAD_NOLOCK)) {
3265 ext4_msg(sb, KERN_ERR, "can't mount with "
3266 "both data=journal and delalloc");
3267 goto failed_mount;
3268 }
3269 if (test_opt(sb, DELALLOC))
3270 clear_opt(sb, DELALLOC);
3271 }
3272
3273 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3274 if (test_opt(sb, DIOREAD_NOLOCK)) {
3275 if (blocksize < PAGE_SIZE) {
3276 ext4_msg(sb, KERN_ERR, "can't mount with "
3277 "dioread_nolock if block size != PAGE_SIZE");
3278 goto failed_mount;
3279 }
3280 }
3281
ac27a0ec 3282 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 3283 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec 3284
617ba13b
MC		 3285 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3286 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3287 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3288 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
b31e1552
ES		 3289 ext4_msg(sb, KERN_WARNING,
3290 "feature flags set on rev 0 fs, "
3291 "running e2fsck is recommended");
469108ff 3292
2035e776
TT		 3293 if (IS_EXT2_SB(sb)) {
3294 if (ext2_feature_set_ok(sb))
3295 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3296 "using the ext4 subsystem");
3297 else {
3298 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3299 "to feature incompatibilities");
3300 goto failed_mount;
3301 }
3302 }
3303
3304 if (IS_EXT3_SB(sb)) {
3305 if (ext3_feature_set_ok(sb))
3306 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3307 "using the ext4 subsystem");
3308 else {
3309 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3310 "to feature incompatibilities");
3311 goto failed_mount;
3312 }
3313 }
3314
ac27a0ec
DK		 3315 /*
3316 * Check feature flags regardless of the revision level, since we
3317 * previously didn't change the revision level when setting the flags,
3318 * so there is a chance incompat flags are set on a rev 0 filesystem.
3319 */
a13fb1a4 3320 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
ac27a0ec 3321 goto failed_mount;
a13fb1a4 3322
617ba13b
MC		 3323 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3324 blocksize > EXT4_MAX_BLOCK_SIZE) {
b31e1552
ES		 3325 ext4_msg(sb, KERN_ERR,
3326 "Unsupported filesystem blocksize %d", blocksize);
ac27a0ec
DK		 3327 goto failed_mount;
3328 }
3329
ac27a0ec 3330 if (sb->s_blocksize != blocksize) {
ce40733c
AK		 3331 /* Validate the filesystem blocksize */
3332 if (!sb_set_blocksize(sb, blocksize)) {
b31e1552 3333 ext4_msg(sb, KERN_ERR, "bad block size %d",
ce40733c 3334 blocksize);
ac27a0ec
DK		 3335 goto failed_mount;
3336 }
3337
2b2d6d01 3338 brelse(bh);
70bbb3e0
AM		 3339 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3340 offset = do_div(logical_sb_block, blocksize);
3341 bh = sb_bread(sb, logical_sb_block);
ac27a0ec 3342 if (!bh) {
b31e1552
ES		 3343 ext4_msg(sb, KERN_ERR,
3344 "Can't read superblock on 2nd try");
ac27a0ec
DK		 3345 goto failed_mount;
3346 }
617ba13b 3347 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
ac27a0ec 3348 sbi->s_es = es;
617ba13b 3349 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
b31e1552
ES		 3350 ext4_msg(sb, KERN_ERR,
3351 "Magic mismatch, very weird!");
ac27a0ec
DK		 3352 goto failed_mount;
3353 }
3354 }
3355
a13fb1a4
ES		 3356 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3357 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
f287a1a5
TT		 3358 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3359 has_huge_files);
3360 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
ac27a0ec 3361
617ba13b
MC		 3362 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3363 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3364 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
ac27a0ec
DK		 3365 } else {
3366 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3367 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
617ba13b 3368 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1330593e 3369 (!is_power_of_2(sbi->s_inode_size)) ||
ac27a0ec 3370 (sbi->s_inode_size > blocksize)) {
b31e1552
ES		 3371 ext4_msg(sb, KERN_ERR,
3372 "unsupported inode size: %d",
2b2d6d01 3373 sbi->s_inode_size);
ac27a0ec
DK		 3374 goto failed_mount;
3375 }
ef7f3835
KS		 3376 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3377 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
ac27a0ec 3378 }
0b8e58a1 3379
0d1ee42f
AR		 3380 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3381 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
8fadc143 3382 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 3383 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 3384 !is_power_of_2(sbi->s_desc_size)) {
b31e1552
ES		 3385 ext4_msg(sb, KERN_ERR,
3386 "unsupported descriptor size %lu",
0d1ee42f
AR		 3387 sbi->s_desc_size);
3388 goto failed_mount;
3389 }
3390 } else
3391 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
0b8e58a1 3392
ac27a0ec 3393 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 3394 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
b47b6f38 3395 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
617ba13b 3396 goto cantfind_ext4;
0b8e58a1 3397
617ba13b 3398 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 3399 if (sbi->s_inodes_per_block == 0)
617ba13b 3400 goto cantfind_ext4;
ac27a0ec
DK		 3401 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3402 sbi->s_inodes_per_block;
0d1ee42f 3403 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 3404 sbi->s_sbh = bh;
3405 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 3406 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3407 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
0b8e58a1 3408
2b2d6d01 3409 for (i = 0; i < 4; i++)
ac27a0ec
DK		 3410 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3411 sbi->s_def_hash_version = es->s_def_hash_version;
f99b2589
TT		 3412 i = le32_to_cpu(es->s_flags);
3413 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3414 sbi->s_hash_unsigned = 3;
3415 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3416#ifdef __CHAR_UNSIGNED__
3417 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3418 sbi->s_hash_unsigned = 3;
3419#else
3420 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3421#endif
3422 sb->s_dirt = 1;
3423 }
ac27a0ec 3424
281b5995
TT		 3425 /* Handle clustersize */
3426 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3427 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3428 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3429 if (has_bigalloc) {
3430 if (clustersize < blocksize) {
3431 ext4_msg(sb, KERN_ERR,
3432 "cluster size (%d) smaller than "
3433 "block size (%d)", clustersize, blocksize);
3434 goto failed_mount;
3435 }
3436 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3437 le32_to_cpu(es->s_log_block_size);
3438 sbi->s_clusters_per_group =
3439 le32_to_cpu(es->s_clusters_per_group);
3440 if (sbi->s_clusters_per_group > blocksize * 8) {
3441 ext4_msg(sb, KERN_ERR,
3442 "#clusters per group too big: %lu",
3443 sbi->s_clusters_per_group);
3444 goto failed_mount;
3445 }
3446 if (sbi->s_blocks_per_group !=
3447 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3448 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3449 "clusters per group (%lu) inconsistent",
3450 sbi->s_blocks_per_group,
3451 sbi->s_clusters_per_group);
3452 goto failed_mount;
3453 }
3454 } else {
3455 if (clustersize != blocksize) {
3456 ext4_warning(sb, "fragment/cluster size (%d) != "
3457 "block size (%d)", clustersize,
3458 blocksize);
3459 clustersize = blocksize;
3460 }
3461 if (sbi->s_blocks_per_group > blocksize * 8) {
3462 ext4_msg(sb, KERN_ERR,
3463 "#blocks per group too big: %lu",
3464 sbi->s_blocks_per_group);
3465 goto failed_mount;
3466 }
3467 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3468 sbi->s_cluster_bits = 0;
ac27a0ec 3469 }
281b5995
TT		 3470 sbi->s_cluster_ratio = clustersize / blocksize;
3471
ac27a0ec 3472 if (sbi->s_inodes_per_group > blocksize * 8) {
b31e1552
ES		 3473 ext4_msg(sb, KERN_ERR,
3474 "#inodes per group too big: %lu",
2b2d6d01 3475 sbi->s_inodes_per_group);
ac27a0ec
DK		 3476 goto failed_mount;
3477 }
3478
bf43d84b
ES		 3479 /*
3480 * Test whether we have more sectors than will fit in sector_t,
3481 * and whether the max offset is addressable by the page cache.
3482 */
5a9ae68a 3483 err = generic_check_addressable(sb->s_blocksize_bits,
30ca22c7 3484 ext4_blocks_count(es));
5a9ae68a 3485 if (err) {
b31e1552 3486 ext4_msg(sb, KERN_ERR, "filesystem"
bf43d84b 3487 " too large to mount safely on this system");
ac27a0ec 3488 if (sizeof(sector_t) < 8)
90c699a9 3489 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
5a9ae68a 3490 ret = err;
ac27a0ec
DK		 3491 goto failed_mount;
3492 }
3493
617ba13b
MC		 3494 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3495 goto cantfind_ext4;
e7c95593 3496
0f2ddca6
FTN		 3497 /* check blocks count against device size */
3498 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3499 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
b31e1552
ES		 3500 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3501 "exceeds size of device (%llu blocks)",
0f2ddca6
FTN		 3502 ext4_blocks_count(es), blocks_count);
3503 goto failed_mount;
3504 }
3505
0b8e58a1
AD		 3506 /*
3507 * It makes no sense for the first data block to be beyond the end
3508 * of the filesystem.
3509 */
3510 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
b31e1552
ES		 3511 ext4_msg(sb, KERN_WARNING, "bad geometry: first data"
3512 "block %u is beyond end of filesystem (%llu)",
3513 le32_to_cpu(es->s_first_data_block),
3514 ext4_blocks_count(es));
e7c95593
ES		 3515 goto failed_mount;
3516 }
bd81d8ee
LV		 3517 blocks_count = (ext4_blocks_count(es) -
3518 le32_to_cpu(es->s_first_data_block) +
3519 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3520 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4ec11028 3521 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
b31e1552 3522 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
4ec11028 3523 "(block count %llu, first data block %u, "
b31e1552 3524 "blocks per group %lu)", sbi->s_groups_count,
4ec11028
TT		 3525 ext4_blocks_count(es),
3526 le32_to_cpu(es->s_first_data_block),
3527 EXT4_BLOCKS_PER_GROUP(sb));
3528 goto failed_mount;
3529 }
bd81d8ee 3530 sbi->s_groups_count = blocks_count;
fb0a387d
ES		 3531 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3532 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
617ba13b
MC		 3533 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3534 EXT4_DESC_PER_BLOCK(sb);
f18a5f21
TT		 3535 sbi->s_group_desc = ext4_kvmalloc(db_count *
3536 sizeof(struct buffer_head *),
3537 GFP_KERNEL);
ac27a0ec 3538 if (sbi->s_group_desc == NULL) {
b31e1552 3539 ext4_msg(sb, KERN_ERR, "not enough memory");
ac27a0ec
DK		 3540 goto failed_mount;
3541 }
3542
9f6200bb
TT		 3543 if (ext4_proc_root)
3544 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
240799cd 3545
705895b6 3546 bgl_lock_init(sbi->s_blockgroup_lock);
ac27a0ec
DK		 3547
3548 for (i = 0; i < db_count; i++) {
70bbb3e0 3549 block = descriptor_loc(sb, logical_sb_block, i);
ac27a0ec
DK		 3550 sbi->s_group_desc[i] = sb_bread(sb, block);
3551 if (!sbi->s_group_desc[i]) {
b31e1552
ES		 3552 ext4_msg(sb, KERN_ERR,
3553 "can't read group descriptor %d", i);
ac27a0ec
DK		 3554 db_count = i;
3555 goto failed_mount2;
3556 }
3557 }
bfff6873 3558 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
b31e1552 3559 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
ac27a0ec
DK		 3560 goto failed_mount2;
3561 }
772cb7c8
JS		 3562 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3563 if (!ext4_fill_flex_info(sb)) {
b31e1552
ES		 3564 ext4_msg(sb, KERN_ERR,
3565 "unable to initialize "
3566 "flex_bg meta info!");
772cb7c8
JS		 3567 goto failed_mount2;
3568 }
3569
ac27a0ec
DK		 3570 sbi->s_gdb_count = db_count;
3571 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3572 spin_lock_init(&sbi->s_next_gen_lock);
3573
04496411
TM		 3574 init_timer(&sbi->s_err_report);
3575 sbi->s_err_report.function = print_daily_error_info;
3576 sbi->s_err_report.data = (unsigned long) sb;
3577
57042651 3578 err = percpu_counter_init(&sbi->s_freeclusters_counter,
5dee5437 3579 ext4_count_free_clusters(sb));
ce7e010a
TT		 3580 if (!err) {
3581 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3582 ext4_count_free_inodes(sb));
3583 }
3584 if (!err) {
3585 err = percpu_counter_init(&sbi->s_dirs_counter,
3586 ext4_count_dirs(sb));
3587 }
3588 if (!err) {
57042651 3589 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0);
ce7e010a
TT		 3590 }
3591 if (err) {
3592 ext4_msg(sb, KERN_ERR, "insufficient memory");
3593 goto failed_mount3;
3594 }
3595
c9de560d 3596 sbi->s_stripe = ext4_get_stripe_size(sbi);
55138e0b 3597 sbi->s_max_writeback_mb_bump = 128;
c9de560d 3598
ac27a0ec
DK		 3599 /*
3600 * set up enough so that it can read an inode
3601 */
9ca92389
TT		 3602 if (!test_opt(sb, NOLOAD) &&
3603 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3604 sb->s_op = &ext4_sops;
3605 else
3606 sb->s_op = &ext4_nojournal_sops;
617ba13b
MC		 3607 sb->s_export_op = &ext4_export_ops;
3608 sb->s_xattr = ext4_xattr_handlers;
ac27a0ec 3609#ifdef CONFIG_QUOTA
617ba13b
MC		 3610 sb->s_qcop = &ext4_qctl_operations;
3611 sb->dq_op = &ext4_quota_operations;
ac27a0ec 3612#endif
f2fa2ffc
AK		 3613 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3614
ac27a0ec 3615 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3b9d4ed2 3616 mutex_init(&sbi->s_orphan_lock);
8f82f840 3617 sbi->s_resize_flags = 0;
ac27a0ec
DK		 3618
3619 sb->s_root = NULL;
3620
3621 needs_recovery = (es->s_last_orphan != 0 ||
617ba13b
MC		 3622 EXT4_HAS_INCOMPAT_FEATURE(sb,
3623 EXT4_FEATURE_INCOMPAT_RECOVER));
ac27a0ec 3624
c5e06d10
JL		 3625 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3626 !(sb->s_flags & MS_RDONLY))
3627 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3628 goto failed_mount3;
3629
ac27a0ec
DK		 3630 /*
3631 * The first inode we look at is the journal inode. Don't try
3632 * root first: it may be modified in the journal!
3633 */
3634 if (!test_opt(sb, NOLOAD) &&
617ba13b
MC		 3635 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3636 if (ext4_load_journal(sb, es, journal_devnum))
ac27a0ec 3637 goto failed_mount3;
0390131b
FM		 3638 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3639 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
b31e1552
ES		 3640 ext4_msg(sb, KERN_ERR, "required journal recovery "
3641 "suppressed and not mounted read-only");
744692dc 3642 goto failed_mount_wq;
ac27a0ec 3643 } else {
fd8c37ec 3644 clear_opt(sb, DATA_FLAGS);
0390131b
FM		 3645 sbi->s_journal = NULL;
3646 needs_recovery = 0;
3647 goto no_journal;
ac27a0ec
DK		 3648 }
3649
eb40a09c
JS		 3650 if (ext4_blocks_count(es) > 0xffffffffULL &&
3651 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3652 JBD2_FEATURE_INCOMPAT_64BIT)) {
b31e1552 3653 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
744692dc 3654 goto failed_mount_wq;
eb40a09c
JS		 3655 }
3656
d4da6c9c
LT		 3657 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3658 jbd2_journal_set_features(sbi->s_journal,
3659 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
818d276c 3660 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
d4da6c9c
LT		 3661 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3662 jbd2_journal_set_features(sbi->s_journal,
3663 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
818d276c
GS		 3664 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3665 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
d4da6c9c
LT		 3666 } else {
3667 jbd2_journal_clear_features(sbi->s_journal,
3668 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3669 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3670 }
818d276c 3671
ac27a0ec
DK		 3672 /* We have now updated the journal if required, so we can
3673 * validate the data journaling mode. */
3674 switch (test_opt(sb, DATA_FLAGS)) {
3675 case 0:
3676 /* No mode set, assume a default based on the journal
63f57933
AM		 3677 * capabilities: ORDERED_DATA if the journal can
3678 * cope, else JOURNAL_DATA
3679 */
dab291af
MC		 3680 if (jbd2_journal_check_available_features
3681 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
fd8c37ec 3682 set_opt(sb, ORDERED_DATA);
ac27a0ec 3683 else
fd8c37ec 3684 set_opt(sb, JOURNAL_DATA);
ac27a0ec
DK		 3685 break;
3686
617ba13b
MC		 3687 case EXT4_MOUNT_ORDERED_DATA:
3688 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 3689 if (!jbd2_journal_check_available_features
3690 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
b31e1552
ES		 3691 ext4_msg(sb, KERN_ERR, "Journal does not support "
3692 "requested data journaling mode");
744692dc 3693 goto failed_mount_wq;
ac27a0ec
DK		 3694 }
3695 default:
3696 break;
3697 }
b3881f74 3698 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
ac27a0ec 3699
ce7e010a
TT		 3700 /*
3701 * The journal may have updated the bg summary counts, so we
3702 * need to update the global counters.
3703 */
57042651 3704 percpu_counter_set(&sbi->s_freeclusters_counter,
5dee5437 3705 ext4_count_free_clusters(sb));
ce7e010a
TT		 3706 percpu_counter_set(&sbi->s_freeinodes_counter,
3707 ext4_count_free_inodes(sb));
3708 percpu_counter_set(&sbi->s_dirs_counter,
3709 ext4_count_dirs(sb));
57042651 3710 percpu_counter_set(&sbi->s_dirtyclusters_counter, 0);
206f7ab4 3711
ce7e010a 3712no_journal:
fd89d5f2
TH		 3713 /*
3714 * The maximum number of concurrent works can be high and
3715 * concurrency isn't really necessary. Limit it to 1.
3716 */
3717 EXT4_SB(sb)->dio_unwritten_wq =
ae005cbe 3718 alloc_workqueue("ext4-dio-unwritten", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4c0425ff
MC		 3719 if (!EXT4_SB(sb)->dio_unwritten_wq) {
3720 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
3721 goto failed_mount_wq;
3722 }
3723
ac27a0ec 3724 /*
dab291af 3725 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 3726 * so we can safely mount the rest of the filesystem now.
3727 */
3728
1d1fe1ee
DH		 3729 root = ext4_iget(sb, EXT4_ROOT_INO);
3730 if (IS_ERR(root)) {
b31e1552 3731 ext4_msg(sb, KERN_ERR, "get root inode failed");
1d1fe1ee 3732 ret = PTR_ERR(root);
32a9bb57 3733 root = NULL;
ac27a0ec
DK		 3734 goto failed_mount4;
3735 }
3736 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
b31e1552 3737 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
ac27a0ec
DK		 3738 goto failed_mount4;
3739 }
1d1fe1ee
DH		 3740 sb->s_root = d_alloc_root(root);
3741 if (!sb->s_root) {
b31e1552 3742 ext4_msg(sb, KERN_ERR, "get root dentry failed");
1d1fe1ee
DH		 3743 ret = -ENOMEM;
3744 goto failed_mount4;
3745 }
ac27a0ec 3746
2b2d6d01 3747 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
ef7f3835
KS		 3748
3749 /* determine the minimum size of new large inodes, if present */
3750 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3751 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3752 EXT4_GOOD_OLD_INODE_SIZE;
3753 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3754 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
3755 if (sbi->s_want_extra_isize <
3756 le16_to_cpu(es->s_want_extra_isize))
3757 sbi->s_want_extra_isize =
3758 le16_to_cpu(es->s_want_extra_isize);
3759 if (sbi->s_want_extra_isize <
3760 le16_to_cpu(es->s_min_extra_isize))
3761 sbi->s_want_extra_isize =
3762 le16_to_cpu(es->s_min_extra_isize);
3763 }
3764 }
3765 /* Check if enough inode space is available */
3766 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3767 sbi->s_inode_size) {
3768 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3769 EXT4_GOOD_OLD_INODE_SIZE;
b31e1552
ES		 3770 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3771 "available");
ef7f3835
KS		 3772 }
3773
6fd058f7
TT		 3774 err = ext4_setup_system_zone(sb);
3775 if (err) {
b31e1552 3776 ext4_msg(sb, KERN_ERR, "failed to initialize system "
fbe845dd 3777 "zone (%d)", err);
6fd058f7
TT		 3778 goto failed_mount4;
3779 }
3780
c2774d84
AK		 3781 ext4_ext_init(sb);
3782 err = ext4_mb_init(sb, needs_recovery);
3783 if (err) {
421f91d2 3784 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
b31e1552 3785 err);
dcf2d804 3786 goto failed_mount5;
c2774d84
AK		 3787 }
3788
bfff6873
LC		 3789 err = ext4_register_li_request(sb, first_not_zeroed);
3790 if (err)
dcf2d804 3791 goto failed_mount6;
bfff6873 3792
3197ebdb
TT		 3793 sbi->s_kobj.kset = ext4_kset;
3794 init_completion(&sbi->s_kobj_unregister);
3795 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
3796 "%s", sb->s_id);
dcf2d804
TM		 3797 if (err)
3798 goto failed_mount7;
3197ebdb 3799
617ba13b
MC		 3800 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3801 ext4_orphan_cleanup(sb, es);
3802 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
0390131b 3803 if (needs_recovery) {
b31e1552 3804 ext4_msg(sb, KERN_INFO, "recovery complete");
0390131b
FM		 3805 ext4_mark_recovery_complete(sb, es);
3806 }
3807 if (EXT4_SB(sb)->s_journal) {
3808 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3809 descr = " journalled data mode";
3810 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3811 descr = " ordered data mode";
3812 else
3813 descr = " writeback data mode";
3814 } else
3815 descr = "out journal";
3816
d4c402d9 3817 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
8b67f04a
TT		 3818 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
3819 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
ac27a0ec 3820
66e61a9e
TT		 3821 if (es->s_error_count)
3822 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
ac27a0ec 3823
d4c402d9 3824 kfree(orig_data);
ac27a0ec
DK		 3825 return 0;
3826
617ba13b 3827cantfind_ext4:
ac27a0ec 3828 if (!silent)
b31e1552 3829 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
ac27a0ec
DK		 3830 goto failed_mount;
3831
dcf2d804
TM		 3832failed_mount7:
3833 ext4_unregister_li_request(sb);
3834failed_mount6:
3835 ext4_ext_release(sb);
3836failed_mount5:
3837 ext4_mb_release(sb);
3838 ext4_release_system_zone(sb);
ac27a0ec 3839failed_mount4:
32a9bb57
MK		 3840 iput(root);
3841 sb->s_root = NULL;
b31e1552 3842 ext4_msg(sb, KERN_ERR, "mount failed");
4c0425ff
MC		 3843 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
3844failed_mount_wq:
0390131b
FM		 3845 if (sbi->s_journal) {
3846 jbd2_journal_destroy(sbi->s_journal);
3847 sbi->s_journal = NULL;
3848 }
ac27a0ec 3849failed_mount3:
04496411 3850 del_timer(&sbi->s_err_report);
9933fc0a
TT		 3851 if (sbi->s_flex_groups)
3852 ext4_kvfree(sbi->s_flex_groups);
57042651 3853 percpu_counter_destroy(&sbi->s_freeclusters_counter);
ce7e010a
TT		 3854 percpu_counter_destroy(&sbi->s_freeinodes_counter);
3855 percpu_counter_destroy(&sbi->s_dirs_counter);
57042651 3856 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
c5e06d10
JL		 3857 if (sbi->s_mmp_tsk)
3858 kthread_stop(sbi->s_mmp_tsk);
ac27a0ec
DK		 3859failed_mount2:
3860 for (i = 0; i < db_count; i++)
3861 brelse(sbi->s_group_desc[i]);
f18a5f21 3862 ext4_kvfree(sbi->s_group_desc);
ac27a0ec 3863failed_mount:
240799cd 3864 if (sbi->s_proc) {
9f6200bb 3865 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 3866 }
ac27a0ec
DK		 3867#ifdef CONFIG_QUOTA
3868 for (i = 0; i < MAXQUOTAS; i++)
3869 kfree(sbi->s_qf_names[i]);
3870#endif
617ba13b 3871 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 3872 brelse(bh);
3873out_fail:
3874 sb->s_fs_info = NULL;
f6830165 3875 kfree(sbi->s_blockgroup_lock);
ac27a0ec 3876 kfree(sbi);
dcc7dae3 3877out_free_orig:
d4c402d9 3878 kfree(orig_data);
1d1fe1ee 3879 return ret;
ac27a0ec
DK		 3880}
3881
3882/*
3883 * Setup any per-fs journal parameters now. We'll do this both on
3884 * initial mount, once the journal has been initialised but before we've
3885 * done any recovery; and again on any subsequent remount.
3886 */
617ba13b 3887static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 3888{
617ba13b 3889 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 3890
30773840
TT		 3891 journal->j_commit_interval = sbi->s_commit_interval;
3892 journal->j_min_batch_time = sbi->s_min_batch_time;
3893 journal->j_max_batch_time = sbi->s_max_batch_time;
ac27a0ec 3894
a931da6a 3895 write_lock(&journal->j_state_lock);
ac27a0ec 3896 if (test_opt(sb, BARRIER))
dab291af 3897 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 3898 else
dab291af 3899 journal->j_flags &= ~JBD2_BARRIER;
5bf5683a
HK		 3900 if (test_opt(sb, DATA_ERR_ABORT))
3901 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
3902 else
3903 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
a931da6a 3904 write_unlock(&journal->j_state_lock);
ac27a0ec
DK		 3905}
3906
617ba13b 3907static journal_t *ext4_get_journal(struct super_block *sb,
ac27a0ec
DK		 3908 unsigned int journal_inum)
3909{
3910 struct inode *journal_inode;
3911 journal_t *journal;
3912
0390131b
FM		 3913 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3914
ac27a0ec
DK		 3915 /* First, test for the existence of a valid inode on disk. Bad
3916 * things happen if we iget() an unused inode, as the subsequent
3917 * iput() will try to delete it. */
3918
1d1fe1ee
DH		 3919 journal_inode = ext4_iget(sb, journal_inum);
3920 if (IS_ERR(journal_inode)) {
b31e1552 3921 ext4_msg(sb, KERN_ERR, "no journal found");
ac27a0ec
DK		 3922 return NULL;
3923 }
3924 if (!journal_inode->i_nlink) {
3925 make_bad_inode(journal_inode);
3926 iput(journal_inode);
b31e1552 3927 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
ac27a0ec
DK		 3928 return NULL;
3929 }
3930
e5f8eab8 3931 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
ac27a0ec 3932 journal_inode, journal_inode->i_size);
1d1fe1ee 3933 if (!S_ISREG(journal_inode->i_mode)) {
b31e1552 3934 ext4_msg(sb, KERN_ERR, "invalid journal inode");
ac27a0ec
DK		 3935 iput(journal_inode);
3936 return NULL;
3937 }
3938
dab291af 3939 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 3940 if (!journal) {
b31e1552 3941 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
ac27a0ec
DK		 3942 iput(journal_inode);
3943 return NULL;
3944 }
3945 journal->j_private = sb;
617ba13b 3946 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 3947 return journal;
3948}
3949
617ba13b 3950static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 3951 dev_t j_dev)
3952{
2b2d6d01 3953 struct buffer_head *bh;
ac27a0ec 3954 journal_t *journal;
617ba13b
MC		 3955 ext4_fsblk_t start;
3956 ext4_fsblk_t len;
ac27a0ec 3957 int hblock, blocksize;
617ba13b 3958 ext4_fsblk_t sb_block;
ac27a0ec 3959 unsigned long offset;
2b2d6d01 3960 struct ext4_super_block *es;
ac27a0ec
DK		 3961 struct block_device *bdev;
3962
0390131b
FM		 3963 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3964
b31e1552 3965 bdev = ext4_blkdev_get(j_dev, sb);
ac27a0ec
DK		 3966 if (bdev == NULL)
3967 return NULL;
3968
ac27a0ec 3969 blocksize = sb->s_blocksize;
e1defc4f 3970 hblock = bdev_logical_block_size(bdev);
ac27a0ec 3971 if (blocksize < hblock) {
b31e1552
ES		 3972 ext4_msg(sb, KERN_ERR,
3973 "blocksize too small for journal device");
ac27a0ec
DK		 3974 goto out_bdev;
3975 }
3976
617ba13b
MC		 3977 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3978 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 3979 set_blocksize(bdev, blocksize);
3980 if (!(bh = __bread(bdev, sb_block, blocksize))) {
b31e1552
ES		 3981 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
3982 "external journal");
ac27a0ec
DK		 3983 goto out_bdev;
3984 }
3985
617ba13b
MC		 3986 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3987 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 3988 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b 3989 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
b31e1552
ES		 3990 ext4_msg(sb, KERN_ERR, "external journal has "
3991 "bad superblock");
ac27a0ec
DK		 3992 brelse(bh);
3993 goto out_bdev;
3994 }
3995
617ba13b 3996 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
b31e1552 3997 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
ac27a0ec
DK		 3998 brelse(bh);
3999 goto out_bdev;
4000 }
4001
bd81d8ee 4002 len = ext4_blocks_count(es);
ac27a0ec
DK		 4003 start = sb_block + 1;
4004 brelse(bh); /* we're done with the superblock */
4005
dab291af 4006 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 4007 start, len, blocksize);
4008 if (!journal) {
b31e1552 4009 ext4_msg(sb, KERN_ERR, "failed to create device journal");
ac27a0ec
DK		 4010 goto out_bdev;
4011 }
4012 journal->j_private = sb;
4013 ll_rw_block(READ, 1, &journal->j_sb_buffer);
4014 wait_on_buffer(journal->j_sb_buffer);
4015 if (!buffer_uptodate(journal->j_sb_buffer)) {
b31e1552 4016 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
ac27a0ec
DK		 4017 goto out_journal;
4018 }
4019 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
b31e1552
ES		 4020 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4021 "user (unsupported) - %d",
ac27a0ec
DK		 4022 be32_to_cpu(journal->j_superblock->s_nr_users));
4023 goto out_journal;
4024 }
617ba13b
MC		 4025 EXT4_SB(sb)->journal_bdev = bdev;
4026 ext4_init_journal_params(sb, journal);
ac27a0ec 4027 return journal;
0b8e58a1 4028
ac27a0ec 4029out_journal:
dab291af 4030 jbd2_journal_destroy(journal);
ac27a0ec 4031out_bdev:
617ba13b 4032 ext4_blkdev_put(bdev);
ac27a0ec
DK		 4033 return NULL;
4034}
4035
617ba13b
MC		 4036static int ext4_load_journal(struct super_block *sb,
4037 struct ext4_super_block *es,
ac27a0ec
DK		 4038 unsigned long journal_devnum)
4039{
4040 journal_t *journal;
4041 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4042 dev_t journal_dev;
4043 int err = 0;
4044 int really_read_only;
4045
0390131b
FM		 4046 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4047
ac27a0ec
DK		 4048 if (journal_devnum &&
4049 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
b31e1552
ES		 4050 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4051 "numbers have changed");
ac27a0ec
DK		 4052 journal_dev = new_decode_dev(journal_devnum);
4053 } else
4054 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4055
4056 really_read_only = bdev_read_only(sb->s_bdev);
4057
4058 /*
4059 * Are we loading a blank journal or performing recovery after a
4060 * crash? For recovery, we need to check in advance whether we
4061 * can get read-write access to the device.
4062 */
617ba13b 4063 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
ac27a0ec 4064 if (sb->s_flags & MS_RDONLY) {
b31e1552
ES		 4065 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4066 "required on readonly filesystem");
ac27a0ec 4067 if (really_read_only) {
b31e1552
ES		 4068 ext4_msg(sb, KERN_ERR, "write access "
4069 "unavailable, cannot proceed");
ac27a0ec
DK		 4070 return -EROFS;
4071 }
b31e1552
ES		 4072 ext4_msg(sb, KERN_INFO, "write access will "
4073 "be enabled during recovery");
ac27a0ec
DK		 4074 }
4075 }
4076
4077 if (journal_inum && journal_dev) {
b31e1552
ES		 4078 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4079 "and inode journals!");
ac27a0ec
DK		 4080 return -EINVAL;
4081 }
4082
4083 if (journal_inum) {
617ba13b 4084 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 4085 return -EINVAL;
4086 } else {
617ba13b 4087 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 4088 return -EINVAL;
4089 }
4090
90576c0b 4091 if (!(journal->j_flags & JBD2_BARRIER))
b31e1552 4092 ext4_msg(sb, KERN_INFO, "barriers disabled");
4776004f 4093
ac27a0ec 4094 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
dab291af 4095 err = jbd2_journal_update_format(journal);
ac27a0ec 4096 if (err) {
b31e1552 4097 ext4_msg(sb, KERN_ERR, "error updating journal");
dab291af 4098 jbd2_journal_destroy(journal);
ac27a0ec
DK		 4099 return err;
4100 }
4101 }
4102
617ba13b 4103 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
dab291af 4104 err = jbd2_journal_wipe(journal, !really_read_only);
1c13d5c0
TT		 4105 if (!err) {
4106 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4107 if (save)
4108 memcpy(save, ((char *) es) +
4109 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
dab291af 4110 err = jbd2_journal_load(journal);
1c13d5c0
TT		 4111 if (save)
4112 memcpy(((char *) es) + EXT4_S_ERR_START,
4113 save, EXT4_S_ERR_LEN);
4114 kfree(save);
4115 }
ac27a0ec
DK		 4116
4117 if (err) {
b31e1552 4118 ext4_msg(sb, KERN_ERR, "error loading journal");
dab291af 4119 jbd2_journal_destroy(journal);
ac27a0ec
DK		 4120 return err;
4121 }
4122
617ba13b
MC		 4123 EXT4_SB(sb)->s_journal = journal;
4124 ext4_clear_journal_err(sb, es);
ac27a0ec 4125
c41303ce 4126 if (!really_read_only && journal_devnum &&
ac27a0ec
DK		 4127 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4128 es->s_journal_dev = cpu_to_le32(journal_devnum);
ac27a0ec
DK		 4129
4130 /* Make sure we flush the recovery flag to disk. */
e2d67052 4131 ext4_commit_super(sb, 1);
ac27a0ec
DK		 4132 }
4133
4134 return 0;
4135}
4136
e2d67052 4137static int ext4_commit_super(struct super_block *sb, int sync)
ac27a0ec 4138{
e2d67052 4139 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
617ba13b 4140 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
c4be0c1d 4141 int error = 0;
ac27a0ec 4142
7c2e7087 4143 if (!sbh || block_device_ejected(sb))
c4be0c1d 4144 return error;
914258bf
TT		 4145 if (buffer_write_io_error(sbh)) {
4146 /*
4147 * Oh, dear. A previous attempt to write the
4148 * superblock failed. This could happen because the
4149 * USB device was yanked out. Or it could happen to
4150 * be a transient write error and maybe the block will
4151 * be remapped. Nothing we can do but to retry the
4152 * write and hope for the best.
4153 */
b31e1552
ES		 4154 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4155 "superblock detected");
914258bf
TT		 4156 clear_buffer_write_io_error(sbh);
4157 set_buffer_uptodate(sbh);
4158 }
71290b36
TT		 4159 /*
4160 * If the file system is mounted read-only, don't update the
4161 * superblock write time. This avoids updating the superblock
4162 * write time when we are mounting the root file system
4163 * read/only but we need to replay the journal; at that point,
4164 * for people who are east of GMT and who make their clock
4165 * tick in localtime for Windows bug-for-bug compatibility,
4166 * the clock is set in the future, and this will cause e2fsck
4167 * to complain and force a full file system check.
4168 */
4169 if (!(sb->s_flags & MS_RDONLY))
4170 es->s_wtime = cpu_to_le32(get_seconds());
f613dfcb
TT		 4171 if (sb->s_bdev->bd_part)
4172 es->s_kbytes_written =
4173 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
afc32f7e
TT		 4174 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4175 EXT4_SB(sb)->s_sectors_written_start) >> 1));
f613dfcb
TT		 4176 else
4177 es->s_kbytes_written =
4178 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
57042651
TT		 4179 ext4_free_blocks_count_set(es,
4180 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4181 &EXT4_SB(sb)->s_freeclusters_counter)));
ce7e010a
TT		 4182 es->s_free_inodes_count =
4183 cpu_to_le32(percpu_counter_sum_positive(
4184 &EXT4_SB(sb)->s_freeinodes_counter));
7234ab2a 4185 sb->s_dirt = 0;
ac27a0ec
DK		 4186 BUFFER_TRACE(sbh, "marking dirty");
4187 mark_buffer_dirty(sbh);
914258bf 4188 if (sync) {
c4be0c1d
TS		 4189 error = sync_dirty_buffer(sbh);
4190 if (error)
4191 return error;
4192
4193 error = buffer_write_io_error(sbh);
4194 if (error) {
b31e1552
ES		 4195 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4196 "superblock");
914258bf
TT		 4197 clear_buffer_write_io_error(sbh);
4198 set_buffer_uptodate(sbh);
4199 }
4200 }
c4be0c1d 4201 return error;
ac27a0ec
DK		 4202}
4203
ac27a0ec
DK		 4204/*
4205 * Have we just finished recovery? If so, and if we are mounting (or
4206 * remounting) the filesystem readonly, then we will end up with a
4207 * consistent fs on disk. Record that fact.
4208 */
2b2d6d01
TT		 4209static void ext4_mark_recovery_complete(struct super_block *sb,
4210 struct ext4_super_block *es)
ac27a0ec 4211{
617ba13b 4212 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 4213
0390131b
FM		 4214 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4215 BUG_ON(journal != NULL);
4216 return;
4217 }
dab291af 4218 jbd2_journal_lock_updates(journal);
7ffe1ea8
HK		 4219 if (jbd2_journal_flush(journal) < 0)
4220 goto out;
4221
617ba13b 4222 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
ac27a0ec 4223 sb->s_flags & MS_RDONLY) {
617ba13b 4224 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
e2d67052 4225 ext4_commit_super(sb, 1);
ac27a0ec 4226 }
7ffe1ea8
HK		 4227
4228out:
dab291af 4229 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 4230}
4231
4232/*
4233 * If we are mounting (or read-write remounting) a filesystem whose journal
4234 * has recorded an error from a previous lifetime, move that error to the
4235 * main filesystem now.
4236 */
2b2d6d01
TT		 4237static void ext4_clear_journal_err(struct super_block *sb,
4238 struct ext4_super_block *es)
ac27a0ec
DK		 4239{
4240 journal_t *journal;
4241 int j_errno;
4242 const char *errstr;
4243
0390131b
FM		 4244 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4245
617ba13b 4246 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 4247
4248 /*
4249 * Now check for any error status which may have been recorded in the
617ba13b 4250 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 4251 */
4252
dab291af 4253 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 4254 if (j_errno) {
4255 char nbuf[16];
4256
617ba13b 4257 errstr = ext4_decode_error(sb, j_errno, nbuf);
12062ddd 4258 ext4_warning(sb, "Filesystem error recorded "
ac27a0ec 4259 "from previous mount: %s", errstr);
12062ddd 4260 ext4_warning(sb, "Marking fs in need of filesystem check.");
ac27a0ec 4261
617ba13b
MC		 4262 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4263 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
e2d67052 4264 ext4_commit_super(sb, 1);
ac27a0ec 4265
dab291af 4266 jbd2_journal_clear_err(journal);
ac27a0ec
DK		 4267 }
4268}
4269
4270/*
4271 * Force the running and committing transactions to commit,
4272 * and wait on the commit.
4273 */
617ba13b 4274int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 4275{
4276 journal_t *journal;
0390131b 4277 int ret = 0;
ac27a0ec
DK		 4278
4279 if (sb->s_flags & MS_RDONLY)
4280 return 0;
4281
617ba13b 4282 journal = EXT4_SB(sb)->s_journal;
6b0310fb 4283 if (journal) {
437f88cc 4284 vfs_check_frozen(sb, SB_FREEZE_TRANS);
0390131b 4285 ret = ext4_journal_force_commit(journal);
6b0310fb 4286 }
0390131b 4287
ac27a0ec
DK		 4288 return ret;
4289}
4290
2b2d6d01 4291static void ext4_write_super(struct super_block *sb)
ac27a0ec 4292{
ebc1ac16 4293 lock_super(sb);
9ca92389 4294 ext4_commit_super(sb, 1);
ebc1ac16 4295 unlock_super(sb);
ac27a0ec
DK		 4296}
4297
617ba13b 4298static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec 4299{
14ce0cb4 4300 int ret = 0;
9eddacf9 4301 tid_t target;
8d5d02e6 4302 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4303
9bffad1e 4304 trace_ext4_sync_fs(sb, wait);
8d5d02e6
MC		 4305 flush_workqueue(sbi->dio_unwritten_wq);
4306 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
9ca92389 4307 if (wait)
8d5d02e6 4308 jbd2_log_wait_commit(sbi->s_journal, target);
0390131b 4309 }
14ce0cb4 4310 return ret;
ac27a0ec
DK		 4311}
4312
4313/*
4314 * LVM calls this function before a (read-only) snapshot is created. This
4315 * gives us a chance to flush the journal completely and mark the fs clean.
be4f27d3
YY		 4316 *
4317 * Note that only this function cannot bring a filesystem to be in a clean
4318 * state independently, because ext4 prevents a new handle from being started
4319 * by @sb->s_frozen, which stays in an upper layer. It thus needs help from
4320 * the upper layer.
ac27a0ec 4321 */
c4be0c1d 4322static int ext4_freeze(struct super_block *sb)
ac27a0ec 4323{
c4be0c1d
TS		 4324 int error = 0;
4325 journal_t *journal;
ac27a0ec 4326
9ca92389
TT		 4327 if (sb->s_flags & MS_RDONLY)
4328 return 0;
ac27a0ec 4329
9ca92389 4330 journal = EXT4_SB(sb)->s_journal;
7ffe1ea8 4331
9ca92389
TT		 4332 /* Now we set up the journal barrier. */
4333 jbd2_journal_lock_updates(journal);
ac27a0ec 4334
9ca92389
TT		 4335 /*
4336 * Don't clear the needs_recovery flag if we failed to flush
4337 * the journal.
4338 */
4339 error = jbd2_journal_flush(journal);
6b0310fb
ES		 4340 if (error < 0)
4341 goto out;
9ca92389
TT		 4342
4343 /* Journal blocked and flushed, clear needs_recovery flag. */
4344 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4345 error = ext4_commit_super(sb, 1);
6b0310fb
ES		 4346out:
4347 /* we rely on s_frozen to stop further updates */
4348 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4349 return error;
ac27a0ec
DK		 4350}
4351
4352/*
4353 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4354 * flag here, even though the filesystem is not technically dirty yet.
4355 */
c4be0c1d 4356static int ext4_unfreeze(struct super_block *sb)
ac27a0ec 4357{
9ca92389
TT		 4358 if (sb->s_flags & MS_RDONLY)
4359 return 0;
4360
4361 lock_super(sb);
4362 /* Reset the needs_recovery flag before the fs is unlocked. */
4363 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4364 ext4_commit_super(sb, 1);
4365 unlock_super(sb);
c4be0c1d 4366 return 0;
ac27a0ec
DK		 4367}
4368
673c6100
TT		 4369/*
4370 * Structure to save mount options for ext4_remount's benefit
4371 */
4372struct ext4_mount_options {
4373 unsigned long s_mount_opt;
a2595b8a 4374 unsigned long s_mount_opt2;
673c6100
TT		 4375 uid_t s_resuid;
4376 gid_t s_resgid;
4377 unsigned long s_commit_interval;
4378 u32 s_min_batch_time, s_max_batch_time;
4379#ifdef CONFIG_QUOTA
4380 int s_jquota_fmt;
4381 char *s_qf_names[MAXQUOTAS];
4382#endif
4383};
4384
2b2d6d01 4385static int ext4_remount(struct super_block *sb, int *flags, char *data)
ac27a0ec 4386{
2b2d6d01 4387 struct ext4_super_block *es;
617ba13b
MC		 4388 struct ext4_sb_info *sbi = EXT4_SB(sb);
4389 ext4_fsblk_t n_blocks_count = 0;
ac27a0ec 4390 unsigned long old_sb_flags;
617ba13b 4391 struct ext4_mount_options old_opts;
c79d967d 4392 int enable_quota = 0;
8a266467 4393 ext4_group_t g;
b3881f74 4394 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
c5e06d10 4395 int err = 0;
ac27a0ec
DK		 4396#ifdef CONFIG_QUOTA
4397 int i;
4398#endif
d4c402d9 4399 char *orig_data = kstrdup(data, GFP_KERNEL);
ac27a0ec
DK		 4400
4401 /* Store the original options */
bbd6851a 4402 lock_super(sb);
ac27a0ec
DK		 4403 old_sb_flags = sb->s_flags;
4404 old_opts.s_mount_opt = sbi->s_mount_opt;
a2595b8a 4405 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
ac27a0ec
DK		 4406 old_opts.s_resuid = sbi->s_resuid;
4407 old_opts.s_resgid = sbi->s_resgid;
4408 old_opts.s_commit_interval = sbi->s_commit_interval;
30773840
TT		 4409 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4410 old_opts.s_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 4411#ifdef CONFIG_QUOTA
4412 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4413 for (i = 0; i < MAXQUOTAS; i++)
4414 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
4415#endif
b3881f74
TT		 4416 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4417 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
ac27a0ec
DK		 4418
4419 /*
4420 * Allow the "check" option to be passed as a remount option.
4421 */
b3881f74
TT		 4422 if (!parse_options(data, sb, NULL, &journal_ioprio,
4423 &n_blocks_count, 1)) {
ac27a0ec
DK		 4424 err = -EINVAL;
4425 goto restore_opts;
4426 }
4427
4ab2f15b 4428 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
c67d859e 4429 ext4_abort(sb, "Abort forced by user");
ac27a0ec
DK		 4430
4431 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 4432 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec
DK		 4433
4434 es = sbi->s_es;
4435
b3881f74 4436 if (sbi->s_journal) {
0390131b 4437 ext4_init_journal_params(sb, sbi->s_journal);
b3881f74
TT		 4438 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4439 }
ac27a0ec
DK		 4440
4441 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
bd81d8ee 4442 n_blocks_count > ext4_blocks_count(es)) {
4ab2f15b 4443 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
ac27a0ec
DK		 4444 err = -EROFS;
4445 goto restore_opts;
4446 }
4447
4448 if (*flags & MS_RDONLY) {
0f0dd62f
CH		 4449 err = dquot_suspend(sb, -1);
4450 if (err < 0)
c79d967d 4451 goto restore_opts;
c79d967d 4452
ac27a0ec
DK		 4453 /*
4454 * First of all, the unconditional stuff we have to do
4455 * to disable replay of the journal when we next remount
4456 */
4457 sb->s_flags |= MS_RDONLY;
4458
4459 /*
4460 * OK, test if we are remounting a valid rw partition
4461 * readonly, and if so set the rdonly flag and then
4462 * mark the partition as valid again.
4463 */
617ba13b
MC		 4464 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4465 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 4466 es->s_state = cpu_to_le16(sbi->s_mount_state);
4467
a63c9eb2 4468 if (sbi->s_journal)
0390131b 4469 ext4_mark_recovery_complete(sb, es);
ac27a0ec 4470 } else {
a13fb1a4
ES		 4471 /* Make sure we can mount this feature set readwrite */
4472 if (!ext4_feature_set_ok(sb, 0)) {
ac27a0ec
DK		 4473 err = -EROFS;
4474 goto restore_opts;
4475 }
8a266467
TT		 4476 /*
4477 * Make sure the group descriptor checksums
0b8e58a1 4478 * are sane. If they aren't, refuse to remount r/w.
8a266467
TT		 4479 */
4480 for (g = 0; g < sbi->s_groups_count; g++) {
4481 struct ext4_group_desc *gdp =
4482 ext4_get_group_desc(sb, g, NULL);
4483
4484 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
b31e1552
ES		 4485 ext4_msg(sb, KERN_ERR,
4486 "ext4_remount: Checksum for group %u failed (%u!=%u)",
8a266467
TT		 4487 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4488 le16_to_cpu(gdp->bg_checksum));
4489 err = -EINVAL;
4490 goto restore_opts;
4491 }
4492 }
4493
ead6596b
ES		 4494 /*
4495 * If we have an unprocessed orphan list hanging
4496 * around from a previously readonly bdev mount,
4497 * require a full umount/remount for now.
4498 */
4499 if (es->s_last_orphan) {
b31e1552 4500 ext4_msg(sb, KERN_WARNING, "Couldn't "
ead6596b
ES		 4501 "remount RDWR because of unprocessed "
4502 "orphan inode list. Please "
b31e1552 4503 "umount/remount instead");
ead6596b
ES		 4504 err = -EINVAL;
4505 goto restore_opts;
4506 }
4507
ac27a0ec
DK		 4508 /*
4509 * Mounting a RDONLY partition read-write, so reread
4510 * and store the current valid flag. (It may have
4511 * been changed by e2fsck since we originally mounted
4512 * the partition.)
4513 */
0390131b
FM		 4514 if (sbi->s_journal)
4515 ext4_clear_journal_err(sb, es);
ac27a0ec 4516 sbi->s_mount_state = le16_to_cpu(es->s_state);
617ba13b 4517 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
ac27a0ec 4518 goto restore_opts;
2b2d6d01 4519 if (!ext4_setup_super(sb, es, 0))
ac27a0ec 4520 sb->s_flags &= ~MS_RDONLY;
c5e06d10
JL		 4521 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
4522 EXT4_FEATURE_INCOMPAT_MMP))
4523 if (ext4_multi_mount_protect(sb,
4524 le64_to_cpu(es->s_mmp_block))) {
4525 err = -EROFS;
4526 goto restore_opts;
4527 }
c79d967d 4528 enable_quota = 1;
ac27a0ec
DK		 4529 }
4530 }
bfff6873
LC		 4531
4532 /*
4533 * Reinitialize lazy itable initialization thread based on
4534 * current settings
4535 */
4536 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4537 ext4_unregister_li_request(sb);
4538 else {
4539 ext4_group_t first_not_zeroed;
4540 first_not_zeroed = ext4_has_uninit_itable(sb);
4541 ext4_register_li_request(sb, first_not_zeroed);
4542 }
4543
6fd058f7 4544 ext4_setup_system_zone(sb);
0390131b 4545 if (sbi->s_journal == NULL)
e2d67052 4546 ext4_commit_super(sb, 1);
0390131b 4547
ac27a0ec
DK		 4548#ifdef CONFIG_QUOTA
4549 /* Release old quota file names */
4550 for (i = 0; i < MAXQUOTAS; i++)
4551 if (old_opts.s_qf_names[i] &&
4552 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4553 kfree(old_opts.s_qf_names[i]);
4554#endif
bbd6851a 4555 unlock_super(sb);
c79d967d 4556 if (enable_quota)
0f0dd62f 4557 dquot_resume(sb, -1);
d4c402d9
CW		 4558
4559 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4560 kfree(orig_data);
ac27a0ec 4561 return 0;
0b8e58a1 4562
ac27a0ec
DK		 4563restore_opts:
4564 sb->s_flags = old_sb_flags;
4565 sbi->s_mount_opt = old_opts.s_mount_opt;
a2595b8a 4566 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
ac27a0ec
DK		 4567 sbi->s_resuid = old_opts.s_resuid;
4568 sbi->s_resgid = old_opts.s_resgid;
4569 sbi->s_commit_interval = old_opts.s_commit_interval;
30773840
TT		 4570 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4571 sbi->s_max_batch_time = old_opts.s_max_batch_time;
ac27a0ec
DK		 4572#ifdef CONFIG_QUOTA
4573 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4574 for (i = 0; i < MAXQUOTAS; i++) {
4575 if (sbi->s_qf_names[i] &&
4576 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4577 kfree(sbi->s_qf_names[i]);
4578 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4579 }
4580#endif
bbd6851a 4581 unlock_super(sb);
d4c402d9 4582 kfree(orig_data);
ac27a0ec
DK		 4583 return err;
4584}
4585
f975d6bc
TT		 4586/*
4587 * Note: calculating the overhead so we can be compatible with
4588 * historical BSD practice is quite difficult in the face of
4589 * clusters/bigalloc. This is because multiple metadata blocks from
4590 * different block group can end up in the same allocation cluster.
4591 * Calculating the exact overhead in the face of clustered allocation
4592 * requires either O(all block bitmaps) in memory or O(number of block
4593 * groups**2) in time. We will still calculate the superblock for
4594 * older file systems --- and if we come across with a bigalloc file
4595 * system with zero in s_overhead_clusters the estimate will be close to
4596 * correct especially for very large cluster sizes --- but for newer
4597 * file systems, it's better to calculate this figure once at mkfs
4598 * time, and store it in the superblock. If the superblock value is
4599 * present (even for non-bigalloc file systems), we will use it.
4600 */
2b2d6d01 4601static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
ac27a0ec
DK		 4602{
4603 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 4604 struct ext4_sb_info *sbi = EXT4_SB(sb);
4605 struct ext4_super_block *es = sbi->s_es;
f975d6bc 4606 struct ext4_group_desc *gdp;
960cc398 4607 u64 fsid;
d02a9391 4608 s64 bfree;
ac27a0ec 4609
5e70030d
BP		 4610 if (test_opt(sb, MINIX_DF)) {
4611 sbi->s_overhead_last = 0;
f975d6bc
TT		 4612 } else if (es->s_overhead_clusters) {
4613 sbi->s_overhead_last = le32_to_cpu(es->s_overhead_clusters);
6bc9feff 4614 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
8df9675f 4615 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
5e70030d 4616 ext4_fsblk_t overhead = 0;
ac27a0ec
DK		 4617
4618 /*
5e70030d
BP		 4619 * Compute the overhead (FS structures). This is constant
4620 * for a given filesystem unless the number of block groups
4621 * changes so we cache the previous value until it does.
ac27a0ec
DK		 4622 */
4623
4624 /*
4625 * All of the blocks before first_data_block are
4626 * overhead
4627 */
f975d6bc 4628 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
ac27a0ec
DK		 4629
4630 /*
f975d6bc 4631 * Add the overhead found in each block group
ac27a0ec
DK		 4632 */
4633 for (i = 0; i < ngroups; i++) {
f975d6bc
TT		 4634 gdp = ext4_get_group_desc(sb, i, NULL);
4635 overhead += ext4_num_overhead_clusters(sb, i, gdp);
ac27a0ec
DK		 4636 cond_resched();
4637 }
5e70030d
BP		 4638 sbi->s_overhead_last = overhead;
4639 smp_wmb();
6bc9feff 4640 sbi->s_blocks_last = ext4_blocks_count(es);
ac27a0ec
DK		 4641 }
4642
617ba13b 4643 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 4644 buf->f_bsize = sb->s_blocksize;
f975d6bc
TT		 4645 buf->f_blocks = (ext4_blocks_count(es) -
4646 EXT4_C2B(sbi, sbi->s_overhead_last));
57042651
TT		 4647 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
4648 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
d02a9391 4649 /* prevent underflow in case that few free space is available */
57042651 4650 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
bd81d8ee
LV		 4651 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
4652 if (buf->f_bfree < ext4_r_blocks_count(es))
ac27a0ec
DK		 4653 buf->f_bavail = 0;
4654 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 4655 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
617ba13b 4656 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 4657 fsid = le64_to_cpup((void *)es->s_uuid) ^
4658 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4659 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4660 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
0b8e58a1 4661
ac27a0ec
DK		 4662 return 0;
4663}
4664
0b8e58a1
AD		 4665/* Helper function for writing quotas on sync - we need to start transaction
4666 * before quota file is locked for write. Otherwise the are possible deadlocks:
ac27a0ec 4667 * Process 1 Process 2
617ba13b 4668 * ext4_create() quota_sync()
a269eb18 4669 * jbd2_journal_start() write_dquot()
871a2931 4670 * dquot_initialize() down(dqio_mutex)
dab291af 4671 * down(dqio_mutex) jbd2_journal_start()
ac27a0ec
DK		 4672 *
4673 */
4674
4675#ifdef CONFIG_QUOTA
4676
4677static inline struct inode *dquot_to_inode(struct dquot *dquot)
4678{
4679 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
4680}
4681
617ba13b 4682static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 4683{
4684 int ret, err;
4685 handle_t *handle;
4686 struct inode *inode;
4687
4688 inode = dquot_to_inode(dquot);
617ba13b 4689 handle = ext4_journal_start(inode,
0b8e58a1 4690 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 4691 if (IS_ERR(handle))
4692 return PTR_ERR(handle);
4693 ret = dquot_commit(dquot);
617ba13b 4694 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4695 if (!ret)
4696 ret = err;
4697 return ret;
4698}
4699
617ba13b 4700static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 4701{
4702 int ret, err;
4703 handle_t *handle;
4704
617ba13b 4705 handle = ext4_journal_start(dquot_to_inode(dquot),
0b8e58a1 4706 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 4707 if (IS_ERR(handle))
4708 return PTR_ERR(handle);
4709 ret = dquot_acquire(dquot);
617ba13b 4710 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4711 if (!ret)
4712 ret = err;
4713 return ret;
4714}
4715
617ba13b 4716static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 4717{
4718 int ret, err;
4719 handle_t *handle;
4720
617ba13b 4721 handle = ext4_journal_start(dquot_to_inode(dquot),
0b8e58a1 4722 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 4723 if (IS_ERR(handle)) {
4724 /* Release dquot anyway to avoid endless cycle in dqput() */
4725 dquot_release(dquot);
ac27a0ec 4726 return PTR_ERR(handle);
9c3013e9 4727 }
ac27a0ec 4728 ret = dquot_release(dquot);
617ba13b 4729 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4730 if (!ret)
4731 ret = err;
4732 return ret;
4733}
4734
617ba13b 4735static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec 4736{
2c8be6b2 4737 /* Are we journaling quotas? */
617ba13b
MC		 4738 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
4739 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
ac27a0ec 4740 dquot_mark_dquot_dirty(dquot);
617ba13b 4741 return ext4_write_dquot(dquot);
ac27a0ec
DK		 4742 } else {
4743 return dquot_mark_dquot_dirty(dquot);
4744 }
4745}
4746
617ba13b 4747static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 4748{
4749 int ret, err;
4750 handle_t *handle;
4751
4752 /* Data block + inode block */
617ba13b 4753 handle = ext4_journal_start(sb->s_root->d_inode, 2);
ac27a0ec
DK		 4754 if (IS_ERR(handle))
4755 return PTR_ERR(handle);
4756 ret = dquot_commit_info(sb, type);
617ba13b 4757 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4758 if (!ret)
4759 ret = err;
4760 return ret;
4761}
4762
4763/*
4764 * Turn on quotas during mount time - we need to find
4765 * the quota file and such...
4766 */
617ba13b 4767static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 4768{
287a8095
CH		 4769 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4770 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 4771}
4772
4773/*
4774 * Standard function to be called on quota_on
4775 */
617ba13b 4776static int ext4_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 4777 struct path *path)
ac27a0ec
DK		 4778{
4779 int err;
ac27a0ec
DK		 4780
4781 if (!test_opt(sb, QUOTA))
4782 return -EINVAL;
0623543b 4783
ac27a0ec 4784 /* Quotafile not on the same filesystem? */
f00c9e44 4785 if (path->mnt->mnt_sb != sb)
ac27a0ec 4786 return -EXDEV;
0623543b
JK		 4787 /* Journaling quota? */
4788 if (EXT4_SB(sb)->s_qf_names[type]) {
2b2d6d01 4789 /* Quotafile not in fs root? */
f00c9e44 4790 if (path->dentry->d_parent != sb->s_root)
b31e1552
ES		 4791 ext4_msg(sb, KERN_WARNING,
4792 "Quota file not on filesystem root. "
4793 "Journaled quota will not work");
2b2d6d01 4794 }
0623543b
JK		 4795
4796 /*
4797 * When we journal data on quota file, we have to flush journal to see
4798 * all updates to the file when we bypass pagecache...
4799 */
0390131b 4800 if (EXT4_SB(sb)->s_journal &&
f00c9e44 4801 ext4_should_journal_data(path->dentry->d_inode)) {
0623543b
JK		 4802 /*
4803 * We don't need to lock updates but journal_flush() could
4804 * otherwise be livelocked...
4805 */
4806 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 4807 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
0623543b 4808 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
f00c9e44 4809 if (err)
7ffe1ea8 4810 return err;
0623543b
JK		 4811 }
4812
f00c9e44 4813 return dquot_quota_on(sb, type, format_id, path);
ac27a0ec
DK		 4814}
4815
ca0e05e4
DM		 4816static int ext4_quota_off(struct super_block *sb, int type)
4817{
21f97697
JK		 4818 struct inode *inode = sb_dqopt(sb)->files[type];
4819 handle_t *handle;
4820
87009d86
DM		 4821 /* Force all delayed allocation blocks to be allocated.
4822 * Caller already holds s_umount sem */
4823 if (test_opt(sb, DELALLOC))
ca0e05e4 4824 sync_filesystem(sb);
ca0e05e4 4825
0b268590
AG		 4826 if (!inode)
4827 goto out;
4828
21f97697
JK		 4829 /* Update modification times of quota files when userspace can
4830 * start looking at them */
4831 handle = ext4_journal_start(inode, 1);
4832 if (IS_ERR(handle))
4833 goto out;
4834 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
4835 ext4_mark_inode_dirty(handle, inode);
4836 ext4_journal_stop(handle);
4837
4838out:
ca0e05e4
DM		 4839 return dquot_quota_off(sb, type);
4840}
4841
ac27a0ec
DK		 4842/* Read data from quotafile - avoid pagecache and such because we cannot afford
4843 * acquiring the locks... As quota files are never truncated and quota code
25985edc 4844 * itself serializes the operations (and no one else should touch the files)
ac27a0ec 4845 * we don't have to be afraid of races */
617ba13b 4846static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 4847 size_t len, loff_t off)
4848{
4849 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 4850 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 4851 int err = 0;
4852 int offset = off & (sb->s_blocksize - 1);
4853 int tocopy;
4854 size_t toread;
4855 struct buffer_head *bh;
4856 loff_t i_size = i_size_read(inode);
4857
4858 if (off > i_size)
4859 return 0;
4860 if (off+len > i_size)
4861 len = i_size-off;
4862 toread = len;
4863 while (toread > 0) {
4864 tocopy = sb->s_blocksize - offset < toread ?
4865 sb->s_blocksize - offset : toread;
617ba13b 4866 bh = ext4_bread(NULL, inode, blk, 0, &err);
ac27a0ec
DK		 4867 if (err)
4868 return err;
4869 if (!bh) /* A hole? */
4870 memset(data, 0, tocopy);
4871 else
4872 memcpy(data, bh->b_data+offset, tocopy);
4873 brelse(bh);
4874 offset = 0;
4875 toread -= tocopy;
4876 data += tocopy;
4877 blk++;
4878 }
4879 return len;
4880}
4881
4882/* Write to quotafile (we know the transaction is already started and has
4883 * enough credits) */
617ba13b 4884static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 4885 const char *data, size_t len, loff_t off)
4886{
4887 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 4888 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 4889 int err = 0;
4890 int offset = off & (sb->s_blocksize - 1);
ac27a0ec
DK		 4891 struct buffer_head *bh;
4892 handle_t *handle = journal_current_handle();
4893
0390131b 4894 if (EXT4_SB(sb)->s_journal && !handle) {
b31e1552
ES		 4895 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4896 " cancelled because transaction is not started",
9c3013e9
JK		 4897 (unsigned long long)off, (unsigned long long)len);
4898 return -EIO;
4899 }
67eeb568
DM		 4900 /*
4901 * Since we account only one data block in transaction credits,
4902 * then it is impossible to cross a block boundary.
4903 */
4904 if (sb->s_blocksize - offset < len) {
4905 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4906 " cancelled because not block aligned",
4907 (unsigned long long)off, (unsigned long long)len);
4908 return -EIO;
4909 }
4910
ac27a0ec 4911 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
67eeb568
DM		 4912 bh = ext4_bread(handle, inode, blk, 1, &err);
4913 if (!bh)
4914 goto out;
62d2b5f2
JK		 4915 err = ext4_journal_get_write_access(handle, bh);
4916 if (err) {
4917 brelse(bh);
4918 goto out;
ac27a0ec 4919 }
67eeb568
DM		 4920 lock_buffer(bh);
4921 memcpy(bh->b_data+offset, data, len);
4922 flush_dcache_page(bh->b_page);
4923 unlock_buffer(bh);
62d2b5f2 4924 err = ext4_handle_dirty_metadata(handle, NULL, bh);
67eeb568 4925 brelse(bh);
ac27a0ec 4926out:
67eeb568 4927 if (err) {
4d04e4fb 4928 mutex_unlock(&inode->i_mutex);
ac27a0ec 4929 return err;
4d04e4fb 4930 }
67eeb568
DM		 4931 if (inode->i_size < off + len) {
4932 i_size_write(inode, off + len);
617ba13b 4933 EXT4_I(inode)->i_disksize = inode->i_size;
21f97697 4934 ext4_mark_inode_dirty(handle, inode);
ac27a0ec 4935 }
ac27a0ec 4936 mutex_unlock(&inode->i_mutex);
67eeb568 4937 return len;
ac27a0ec
DK		 4938}
4939
4940#endif
4941
152a0836
AV		 4942static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
4943 const char *dev_name, void *data)
ac27a0ec 4944{
152a0836 4945 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
ac27a0ec
DK		 4946}
4947
37f328eb 4948#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
24b58424
TT		 4949static inline void register_as_ext2(void)
4950{
4951 int err = register_filesystem(&ext2_fs_type);
4952 if (err)
4953 printk(KERN_WARNING
4954 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
4955}
4956
4957static inline void unregister_as_ext2(void)
4958{
4959 unregister_filesystem(&ext2_fs_type);
4960}
2035e776
TT		 4961
4962static inline int ext2_feature_set_ok(struct super_block *sb)
4963{
4964 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
4965 return 0;
4966 if (sb->s_flags & MS_RDONLY)
4967 return 1;
4968 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
4969 return 0;
4970 return 1;
4971}
51b7e3c9 4972MODULE_ALIAS("ext2");
24b58424
TT		 4973#else
4974static inline void register_as_ext2(void) { }
4975static inline void unregister_as_ext2(void) { }
2035e776 4976static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 4977#endif
4978
37f328eb 4979#if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
24b58424
TT		 4980static inline void register_as_ext3(void)
4981{
4982 int err = register_filesystem(&ext3_fs_type);
4983 if (err)
4984 printk(KERN_WARNING
4985 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
4986}
4987
4988static inline void unregister_as_ext3(void)
4989{
4990 unregister_filesystem(&ext3_fs_type);
4991}
2035e776
TT		 4992
4993static inline int ext3_feature_set_ok(struct super_block *sb)
4994{
4995 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
4996 return 0;
4997 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
4998 return 0;
4999 if (sb->s_flags & MS_RDONLY)
5000 return 1;
5001 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
5002 return 0;
5003 return 1;
5004}
51b7e3c9 5005MODULE_ALIAS("ext3");
24b58424
TT		 5006#else
5007static inline void register_as_ext3(void) { }
5008static inline void unregister_as_ext3(void) { }
2035e776 5009static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
24b58424
TT		 5010#endif
5011
03010a33
TT		 5012static struct file_system_type ext4_fs_type = {
5013 .owner = THIS_MODULE,
5014 .name = "ext4",
152a0836 5015 .mount = ext4_mount,
03010a33
TT		 5016 .kill_sb = kill_block_super,
5017 .fs_flags = FS_REQUIRES_DEV,
5018};
5019
8f021222 5020static int __init ext4_init_feat_adverts(void)
857ac889
LC		 5021{
5022 struct ext4_features *ef;
5023 int ret = -ENOMEM;
5024
5025 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
5026 if (!ef)
5027 goto out;
5028
5029 ef->f_kobj.kset = ext4_kset;
5030 init_completion(&ef->f_kobj_unregister);
5031 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
5032 "features");
5033 if (ret) {
5034 kfree(ef);
5035 goto out;
5036 }
5037
5038 ext4_feat = ef;
5039 ret = 0;
5040out:
5041 return ret;
5042}
5043
8f021222
LC		 5044static void ext4_exit_feat_adverts(void)
5045{
5046 kobject_put(&ext4_feat->f_kobj);
5047 wait_for_completion(&ext4_feat->f_kobj_unregister);
5048 kfree(ext4_feat);
5049}
5050
e9e3bcec
ES		 5051/* Shared across all ext4 file systems */
5052wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5053struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5054
5dabfc78 5055static int __init ext4_init_fs(void)
ac27a0ec 5056{
e9e3bcec 5057 int i, err;
c9de560d 5058
12e9b892 5059 ext4_check_flag_values();
e9e3bcec
ES		 5060
5061 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5062 mutex_init(&ext4__aio_mutex[i]);
5063 init_waitqueue_head(&ext4__ioend_wq[i]);
5064 }
5065
5dabfc78 5066 err = ext4_init_pageio();
6fd058f7
TT		 5067 if (err)
5068 return err;
5dabfc78 5069 err = ext4_init_system_zone();
bd2d0210 5070 if (err)
d44651d0 5071 goto out6;
3197ebdb
TT		 5072 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
5073 if (!ext4_kset)
dd68314c 5074 goto out5;
d44651d0 5075 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
857ac889
LC		 5076
5077 err = ext4_init_feat_adverts();
dd68314c
TT		 5078 if (err)
5079 goto out4;
857ac889 5080
5dabfc78 5081 err = ext4_init_mballoc();
ac27a0ec 5082 if (err)
6fd058f7 5083 goto out3;
c9de560d 5084
5dabfc78 5085 err = ext4_init_xattr();
c9de560d
AT		 5086 if (err)
5087 goto out2;
ac27a0ec
DK		 5088 err = init_inodecache();
5089 if (err)
5090 goto out1;
24b58424 5091 register_as_ext3();
2035e776 5092 register_as_ext2();
03010a33 5093 err = register_filesystem(&ext4_fs_type);
ac27a0ec
DK		 5094 if (err)
5095 goto out;
bfff6873
LC		 5096
5097 ext4_li_info = NULL;
5098 mutex_init(&ext4_li_mtx);
ac27a0ec
DK		 5099 return 0;
5100out:
24b58424
TT		 5101 unregister_as_ext2();
5102 unregister_as_ext3();
ac27a0ec
DK		 5103 destroy_inodecache();
5104out1:
5dabfc78 5105 ext4_exit_xattr();
c9de560d 5106out2:
5dabfc78 5107 ext4_exit_mballoc();
6fd058f7 5108out3:
8f021222 5109 ext4_exit_feat_adverts();
dd68314c 5110out4:
d44651d0
FJ		 5111 if (ext4_proc_root)
5112 remove_proc_entry("fs/ext4", NULL);
6fd058f7 5113 kset_unregister(ext4_kset);
d44651d0 5114out5:
5dabfc78 5115 ext4_exit_system_zone();
d44651d0 5116out6:
5dabfc78 5117 ext4_exit_pageio();
ac27a0ec
DK		 5118 return err;
5119}
5120
5dabfc78 5121static void __exit ext4_exit_fs(void)
ac27a0ec 5122{
bfff6873 5123 ext4_destroy_lazyinit_thread();
24b58424
TT		 5124 unregister_as_ext2();
5125 unregister_as_ext3();
03010a33 5126 unregister_filesystem(&ext4_fs_type);
ac27a0ec 5127 destroy_inodecache();
5dabfc78
TT		 5128 ext4_exit_xattr();
5129 ext4_exit_mballoc();
8f021222 5130 ext4_exit_feat_adverts();
9f6200bb 5131 remove_proc_entry("fs/ext4", NULL);
3197ebdb 5132 kset_unregister(ext4_kset);
5dabfc78
TT		 5133 ext4_exit_system_zone();
5134 ext4_exit_pageio();
ac27a0ec
DK		 5135}
5136
5137MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
83982b6f 5138MODULE_DESCRIPTION("Fourth Extended Filesystem");
ac27a0ec 5139MODULE_LICENSE("GPL");
5dabfc78
TT		 5140module_init(ext4_init_fs)
5141module_exit(ext4_exit_fs)
Linux 6.x block layer and io_uring tree(s)