]> git.proxmox.com Git - mirror_ubuntu-bionic-kernel.git/blame - fs/ext4/super.c
projects / mirror_ubuntu-bionic-kernel.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
ext4: use the nblocks arg to ext4_truncate_restart_trans()
[mirror_ubuntu-bionic-kernel.git] / fs / ext4 / super.c
CommitLineData
ac27a0ec 1/*
617ba13b 2 * linux/fs/ext4/super.c
ac27a0ec
DK		 3 *
4 * Copyright (C) 1992, 1993, 1994, 1995
5 * Remy Card (card@masi.ibp.fr)
6 * Laboratoire MASI - Institut Blaise Pascal
7 * Universite Pierre et Marie Curie (Paris VI)
8 *
9 * from
10 *
11 * linux/fs/minix/inode.c
12 *
13 * Copyright (C) 1991, 1992 Linus Torvalds
14 *
15 * Big-endian to little-endian byte-swapping/bitmaps by
16 * David S. Miller (davem@caip.rutgers.edu), 1995
17 */
18
19#include <linux/module.h>
20#include <linux/string.h>
21#include <linux/fs.h>
22#include <linux/time.h>
c5ca7c76 23#include <linux/vmalloc.h>
dab291af 24#include <linux/jbd2.h>
ac27a0ec
DK		 25#include <linux/slab.h>
26#include <linux/init.h>
27#include <linux/blkdev.h>
28#include <linux/parser.h>
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>
ac27a0ec
DK		 41#include <asm/uaccess.h>
42
bfff6873
LC		 43#include <linux/kthread.h>
44#include <linux/freezer.h>
45
3dcf5451
CH		 46#include "ext4.h"
47#include "ext4_jbd2.h"
ac27a0ec
DK		 48#include "xattr.h"
49#include "acl.h"
3661d286 50#include "mballoc.h"
ac27a0ec 51
9bffad1e
TT		 52#define CREATE_TRACE_POINTS
53#include <trace/events/ext4.h>
54
1f109d5a 55static struct proc_dir_entry *ext4_proc_root;
3197ebdb 56static struct kset *ext4_kset;
0b75a840
LC		 57static struct ext4_lazy_init *ext4_li_info;
58static struct mutex ext4_li_mtx;
59static struct ext4_features *ext4_feat;
9f6200bb 60
617ba13b 61static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
ac27a0ec 62 unsigned long journal_devnum);
e2d67052 63static int ext4_commit_super(struct super_block *sb, int sync);
2b2d6d01
TT		 64static void ext4_mark_recovery_complete(struct super_block *sb,
65 struct ext4_super_block *es);
66static void ext4_clear_journal_err(struct super_block *sb,
67 struct ext4_super_block *es);
617ba13b 68static int ext4_sync_fs(struct super_block *sb, int wait);
2b2d6d01 69static const char *ext4_decode_error(struct super_block *sb, int errno,
ac27a0ec 70 char nbuf[16]);
2b2d6d01
TT		 71static int ext4_remount(struct super_block *sb, int *flags, char *data);
72static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
c4be0c1d 73static int ext4_unfreeze(struct super_block *sb);
2b2d6d01 74static void ext4_write_super(struct super_block *sb);
c4be0c1d 75static int ext4_freeze(struct super_block *sb);
152a0836
AV		 76static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
77 const char *dev_name, void *data);
bfff6873
LC		 78static void ext4_destroy_lazyinit_thread(void);
79static void ext4_unregister_li_request(struct super_block *sb);
8f1f7453 80static void ext4_clear_request_list(void);
ac27a0ec 81
ba69f9ab
JK		 82#if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
83static struct file_system_type ext3_fs_type = {
84 .owner = THIS_MODULE,
85 .name = "ext3",
152a0836 86 .mount = ext4_mount,
ba69f9ab
JK		 87 .kill_sb = kill_block_super,
88 .fs_flags = FS_REQUIRES_DEV,
89};
90#define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
91#else
92#define IS_EXT3_SB(sb) (0)
93#endif
bd81d8ee 94
8fadc143
AR		 95ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
96 struct ext4_group_desc *bg)
bd81d8ee 97{
3a14589c 98 return le32_to_cpu(bg->bg_block_bitmap_lo) |
8fadc143 99 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 100 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 101}
102
8fadc143
AR		 103ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
104 struct ext4_group_desc *bg)
bd81d8ee 105{
5272f837 106 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
8fadc143 107 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 108 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
bd81d8ee
LV		 109}
110
8fadc143
AR		 111ext4_fsblk_t ext4_inode_table(struct super_block *sb,
112 struct ext4_group_desc *bg)
bd81d8ee 113{
5272f837 114 return le32_to_cpu(bg->bg_inode_table_lo) |
8fadc143 115 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 116 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
bd81d8ee
LV		 117}
118
560671a0
AK		 119__u32 ext4_free_blks_count(struct super_block *sb,
120 struct ext4_group_desc *bg)
121{
122 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
123 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 124 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
560671a0
AK		 125}
126
127__u32 ext4_free_inodes_count(struct super_block *sb,
128 struct ext4_group_desc *bg)
129{
130 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
131 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 132 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
560671a0
AK		 133}
134
135__u32 ext4_used_dirs_count(struct super_block *sb,
136 struct ext4_group_desc *bg)
137{
138 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
139 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 140 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
560671a0
AK		 141}
142
143__u32 ext4_itable_unused_count(struct super_block *sb,
144 struct ext4_group_desc *bg)
145{
146 return le16_to_cpu(bg->bg_itable_unused_lo) |
147 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
0b8e58a1 148 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
560671a0
AK		 149}
150
8fadc143
AR		 151void ext4_block_bitmap_set(struct super_block *sb,
152 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 153{
3a14589c 154 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 155 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
156 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 157}
158
8fadc143
AR		 159void ext4_inode_bitmap_set(struct super_block *sb,
160 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 161{
5272f837 162 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 163 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
164 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 165}
166
8fadc143
AR		 167void ext4_inode_table_set(struct super_block *sb,
168 struct ext4_group_desc *bg, ext4_fsblk_t blk)
bd81d8ee 169{
5272f837 170 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
8fadc143
AR		 171 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
172 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
bd81d8ee
LV		 173}
174
560671a0
AK		 175void ext4_free_blks_set(struct super_block *sb,
176 struct ext4_group_desc *bg, __u32 count)
177{
178 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
179 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
180 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
181}
182
183void ext4_free_inodes_set(struct super_block *sb,
184 struct ext4_group_desc *bg, __u32 count)
185{
186 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
187 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
188 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
189}
190
191void ext4_used_dirs_set(struct super_block *sb,
192 struct ext4_group_desc *bg, __u32 count)
193{
194 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
195 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
196 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
197}
198
199void ext4_itable_unused_set(struct super_block *sb,
200 struct ext4_group_desc *bg, __u32 count)
201{
202 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
203 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
204 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
205}
206
d3d1faf6
CW		 207
208/* Just increment the non-pointer handle value */
209static handle_t *ext4_get_nojournal(void)
210{
211 handle_t *handle = current->journal_info;
212 unsigned long ref_cnt = (unsigned long)handle;
213
214 BUG_ON(ref_cnt >= EXT4_NOJOURNAL_MAX_REF_COUNT);
215
216 ref_cnt++;
217 handle = (handle_t *)ref_cnt;
218
219 current->journal_info = handle;
220 return handle;
221}
222
223
224/* Decrement the non-pointer handle value */
225static void ext4_put_nojournal(handle_t *handle)
226{
227 unsigned long ref_cnt = (unsigned long)handle;
228
229 BUG_ON(ref_cnt == 0);
230
231 ref_cnt--;
232 handle = (handle_t *)ref_cnt;
233
234 current->journal_info = handle;
235}
236
ac27a0ec 237/*
dab291af 238 * Wrappers for jbd2_journal_start/end.
ac27a0ec
DK		 239 *
240 * The only special thing we need to do here is to make sure that all
241 * journal_end calls result in the superblock being marked dirty, so
242 * that sync() will call the filesystem's write_super callback if
243 * appropriate.
244 */
617ba13b 245handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
ac27a0ec
DK		 246{
247 journal_t *journal;
248
249 if (sb->s_flags & MS_RDONLY)
250 return ERR_PTR(-EROFS);
251
437f88cc 252 vfs_check_frozen(sb, SB_FREEZE_TRANS);
ac27a0ec
DK		 253 /* Special case here: if the journal has aborted behind our
254 * backs (eg. EIO in the commit thread), then we still need to
255 * take the FS itself readonly cleanly. */
617ba13b 256 journal = EXT4_SB(sb)->s_journal;
0390131b
FM		 257 if (journal) {
258 if (is_journal_aborted(journal)) {
c67d859e 259 ext4_abort(sb, "Detected aborted journal");
0390131b
FM		 260 return ERR_PTR(-EROFS);
261 }
262 return jbd2_journal_start(journal, nblocks);
ac27a0ec 263 }
d3d1faf6 264 return ext4_get_nojournal();
ac27a0ec
DK		 265}
266
267/*
268 * The only special thing we need to do here is to make sure that all
dab291af 269 * jbd2_journal_stop calls result in the superblock being marked dirty, so
ac27a0ec
DK		 270 * that sync() will call the filesystem's write_super callback if
271 * appropriate.
272 */
c398eda0 273int __ext4_journal_stop(const char *where, unsigned int line, handle_t *handle)
ac27a0ec
DK		 274{
275 struct super_block *sb;
276 int err;
277 int rc;
278
0390131b 279 if (!ext4_handle_valid(handle)) {
d3d1faf6 280 ext4_put_nojournal(handle);
0390131b
FM		 281 return 0;
282 }
ac27a0ec
DK		 283 sb = handle->h_transaction->t_journal->j_private;
284 err = handle->h_err;
dab291af 285 rc = jbd2_journal_stop(handle);
ac27a0ec
DK		 286
287 if (!err)
288 err = rc;
289 if (err)
c398eda0 290 __ext4_std_error(sb, where, line, err);
ac27a0ec
DK		 291 return err;
292}
293
90c7201b
TT		 294void ext4_journal_abort_handle(const char *caller, unsigned int line,
295 const char *err_fn, struct buffer_head *bh,
296 handle_t *handle, int err)
ac27a0ec
DK		 297{
298 char nbuf[16];
617ba13b 299 const char *errstr = ext4_decode_error(NULL, err, nbuf);
ac27a0ec 300
0390131b
FM		 301 BUG_ON(!ext4_handle_valid(handle));
302
ac27a0ec
DK		 303 if (bh)
304 BUFFER_TRACE(bh, "abort");
305
306 if (!handle->h_err)
307 handle->h_err = err;
308
309 if (is_handle_aborted(handle))
310 return;
311
90c7201b
TT		 312 printk(KERN_ERR "%s:%d: aborting transaction: %s in %s\n",
313 caller, line, errstr, err_fn);
ac27a0ec 314
dab291af 315 jbd2_journal_abort_handle(handle);
ac27a0ec
DK		 316}
317
1c13d5c0
TT		 318static void __save_error_info(struct super_block *sb, const char *func,
319 unsigned int line)
320{
321 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
322
323 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
324 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
325 es->s_last_error_time = cpu_to_le32(get_seconds());
326 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
327 es->s_last_error_line = cpu_to_le32(line);
328 if (!es->s_first_error_time) {
329 es->s_first_error_time = es->s_last_error_time;
330 strncpy(es->s_first_error_func, func,
331 sizeof(es->s_first_error_func));
332 es->s_first_error_line = cpu_to_le32(line);
333 es->s_first_error_ino = es->s_last_error_ino;
334 es->s_first_error_block = es->s_last_error_block;
335 }
66e61a9e
TT		 336 /*
337 * Start the daily error reporting function if it hasn't been
338 * started already
339 */
340 if (!es->s_error_count)
341 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
1c13d5c0
TT		 342 es->s_error_count = cpu_to_le32(le32_to_cpu(es->s_error_count) + 1);
343}
344
345static void save_error_info(struct super_block *sb, const char *func,
346 unsigned int line)
347{
348 __save_error_info(sb, func, line);
349 ext4_commit_super(sb, 1);
350}
351
352
ac27a0ec
DK		 353/* Deal with the reporting of failure conditions on a filesystem such as
354 * inconsistencies detected or read IO failures.
355 *
356 * On ext2, we can store the error state of the filesystem in the
617ba13b 357 * superblock. That is not possible on ext4, because we may have other
ac27a0ec
DK		 358 * write ordering constraints on the superblock which prevent us from
359 * writing it out straight away; and given that the journal is about to
360 * be aborted, we can't rely on the current, or future, transactions to
361 * write out the superblock safely.
362 *
dab291af 363 * We'll just use the jbd2_journal_abort() error code to record an error in
d6b198bc 364 * the journal instead. On recovery, the journal will complain about
ac27a0ec
DK		 365 * that error until we've noted it down and cleared it.
366 */
367
617ba13b 368static void ext4_handle_error(struct super_block *sb)
ac27a0ec 369{
ac27a0ec
DK		 370 if (sb->s_flags & MS_RDONLY)
371 return;
372
2b2d6d01 373 if (!test_opt(sb, ERRORS_CONT)) {
617ba13b 374 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 375
4ab2f15b 376 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
ac27a0ec 377 if (journal)
dab291af 378 jbd2_journal_abort(journal, -EIO);
ac27a0ec 379 }
2b2d6d01 380 if (test_opt(sb, ERRORS_RO)) {
b31e1552 381 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
ac27a0ec
DK		 382 sb->s_flags |= MS_RDONLY;
383 }
ac27a0ec 384 if (test_opt(sb, ERRORS_PANIC))
617ba13b 385 panic("EXT4-fs (device %s): panic forced after error\n",
ac27a0ec
DK		 386 sb->s_id);
387}
388
12062ddd 389void __ext4_error(struct super_block *sb, const char *function,
c398eda0 390 unsigned int line, const char *fmt, ...)
ac27a0ec 391{
0ff2ea7d 392 struct va_format vaf;
ac27a0ec
DK		 393 va_list args;
394
395 va_start(args, fmt);
0ff2ea7d
JP		 396 vaf.fmt = fmt;
397 vaf.va = &args;
398 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
399 sb->s_id, function, line, current->comm, &vaf);
ac27a0ec
DK		 400 va_end(args);
401
617ba13b 402 ext4_handle_error(sb);
ac27a0ec
DK		 403}
404
c398eda0
TT		 405void ext4_error_inode(struct inode *inode, const char *function,
406 unsigned int line, ext4_fsblk_t block,
273df556
FM		 407 const char *fmt, ...)
408{
409 va_list args;
f7c21177 410 struct va_format vaf;
1c13d5c0 411 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
273df556 412
1c13d5c0
TT		 413 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
414 es->s_last_error_block = cpu_to_le64(block);
415 save_error_info(inode->i_sb, function, line);
273df556 416 va_start(args, fmt);
f7c21177
TT		 417 vaf.fmt = fmt;
418 vaf.va = &args;
c398eda0
TT		 419 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: inode #%lu: ",
420 inode->i_sb->s_id, function, line, inode->i_ino);
421 if (block)
f7c21177
TT		 422 printk(KERN_CONT "block %llu: ", block);
423 printk(KERN_CONT "comm %s: %pV\n", current->comm, &vaf);
273df556
FM		 424 va_end(args);
425
426 ext4_handle_error(inode->i_sb);
427}
428
c398eda0 429void ext4_error_file(struct file *file, const char *function,
f7c21177
TT		 430 unsigned int line, ext4_fsblk_t block,
431 const char *fmt, ...)
273df556
FM		 432{
433 va_list args;
f7c21177 434 struct va_format vaf;
1c13d5c0 435 struct ext4_super_block *es;
273df556
FM		 436 struct inode *inode = file->f_dentry->d_inode;
437 char pathname[80], *path;
438
1c13d5c0
TT		 439 es = EXT4_SB(inode->i_sb)->s_es;
440 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
441 save_error_info(inode->i_sb, function, line);
273df556 442 path = d_path(&(file->f_path), pathname, sizeof(pathname));
f9a62d09 443 if (IS_ERR(path))
273df556
FM		 444 path = "(unknown)";
445 printk(KERN_CRIT
f7c21177
TT		 446 "EXT4-fs error (device %s): %s:%d: inode #%lu: ",
447 inode->i_sb->s_id, function, line, inode->i_ino);
448 if (block)
449 printk(KERN_CONT "block %llu: ", block);
450 va_start(args, fmt);
451 vaf.fmt = fmt;
452 vaf.va = &args;
453 printk(KERN_CONT "comm %s: path %s: %pV\n", current->comm, path, &vaf);
273df556
FM		 454 va_end(args);
455
456 ext4_handle_error(inode->i_sb);
457}
458
2b2d6d01 459static const char *ext4_decode_error(struct super_block *sb, int errno,
ac27a0ec
DK		 460 char nbuf[16])
461{
462 char *errstr = NULL;
463
464 switch (errno) {
465 case -EIO:
466 errstr = "IO failure";
467 break;
468 case -ENOMEM:
469 errstr = "Out of memory";
470 break;
471 case -EROFS:
78f1ddbb
TT		 472 if (!sb || (EXT4_SB(sb)->s_journal &&
473 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
ac27a0ec
DK		 474 errstr = "Journal has aborted";
475 else
476 errstr = "Readonly filesystem";
477 break;
478 default:
479 /* If the caller passed in an extra buffer for unknown
480 * errors, textualise them now. Else we just return
481 * NULL. */
482 if (nbuf) {
483 /* Check for truncated error codes... */
484 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
485 errstr = nbuf;
486 }
487 break;
488 }
489
490 return errstr;
491}
492
617ba13b 493/* __ext4_std_error decodes expected errors from journaling functions
ac27a0ec
DK		 494 * automatically and invokes the appropriate error response. */
495
c398eda0
TT		 496void __ext4_std_error(struct super_block *sb, const char *function,
497 unsigned int line, int errno)
ac27a0ec
DK		 498{
499 char nbuf[16];
500 const char *errstr;
501
502 /* Special case: if the error is EROFS, and we're not already
503 * inside a transaction, then there's really no point in logging
504 * an error. */
505 if (errno == -EROFS && journal_current_handle() == NULL &&
506 (sb->s_flags & MS_RDONLY))
507 return;
508
617ba13b 509 errstr = ext4_decode_error(sb, errno, nbuf);
c398eda0
TT		 510 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
511 sb->s_id, function, line, errstr);
1c13d5c0 512 save_error_info(sb, function, line);
ac27a0ec 513
617ba13b 514 ext4_handle_error(sb);
ac27a0ec
DK		 515}
516
517/*
617ba13b 518 * ext4_abort is a much stronger failure handler than ext4_error. The
ac27a0ec
DK		 519 * abort function may be used to deal with unrecoverable failures such
520 * as journal IO errors or ENOMEM at a critical moment in log management.
521 *
522 * We unconditionally force the filesystem into an ABORT|READONLY state,
523 * unless the error response on the fs has been set to panic in which
524 * case we take the easy way out and panic immediately.
525 */
526
c67d859e 527void __ext4_abort(struct super_block *sb, const char *function,
c398eda0 528 unsigned int line, const char *fmt, ...)
ac27a0ec
DK		 529{
530 va_list args;
531
1c13d5c0 532 save_error_info(sb, function, line);
ac27a0ec 533 va_start(args, fmt);
c398eda0
TT		 534 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
535 function, line);
ac27a0ec
DK		 536 vprintk(fmt, args);
537 printk("\n");
538 va_end(args);
539
1c13d5c0
TT		 540 if ((sb->s_flags & MS_RDONLY) == 0) {
541 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
542 sb->s_flags |= MS_RDONLY;
543 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
544 if (EXT4_SB(sb)->s_journal)
545 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
546 save_error_info(sb, function, line);
547 }
ac27a0ec 548 if (test_opt(sb, ERRORS_PANIC))
617ba13b 549 panic("EXT4-fs panic from previous error\n");
ac27a0ec
DK		 550}
551
0ff2ea7d 552void ext4_msg(struct super_block *sb, const char *prefix, const char *fmt, ...)
b31e1552 553{
0ff2ea7d 554 struct va_format vaf;
b31e1552
ES		 555 va_list args;
556
557 va_start(args, fmt);
0ff2ea7d
JP		 558 vaf.fmt = fmt;
559 vaf.va = &args;
560 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
b31e1552
ES		 561 va_end(args);
562}
563
12062ddd 564void __ext4_warning(struct super_block *sb, const char *function,
c398eda0 565 unsigned int line, const char *fmt, ...)
ac27a0ec 566{
0ff2ea7d 567 struct va_format vaf;
ac27a0ec
DK		 568 va_list args;
569
570 va_start(args, fmt);
0ff2ea7d
JP		 571 vaf.fmt = fmt;
572 vaf.va = &args;
573 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
574 sb->s_id, function, line, &vaf);
ac27a0ec
DK		 575 va_end(args);
576}
577
e29136f8
TT		 578void __ext4_grp_locked_error(const char *function, unsigned int line,
579 struct super_block *sb, ext4_group_t grp,
580 unsigned long ino, ext4_fsblk_t block,
581 const char *fmt, ...)
5d1b1b3f
AK		 582__releases(bitlock)
583__acquires(bitlock)
584{
0ff2ea7d 585 struct va_format vaf;
5d1b1b3f
AK		 586 va_list args;
587 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
588
1c13d5c0
TT		 589 es->s_last_error_ino = cpu_to_le32(ino);
590 es->s_last_error_block = cpu_to_le64(block);
591 __save_error_info(sb, function, line);
0ff2ea7d 592
5d1b1b3f 593 va_start(args, fmt);
0ff2ea7d
JP		 594
595 vaf.fmt = fmt;
596 vaf.va = &args;
e29136f8
TT		 597 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u",
598 sb->s_id, function, line, grp);
599 if (ino)
0ff2ea7d 600 printk(KERN_CONT "inode %lu: ", ino);
e29136f8 601 if (block)
0ff2ea7d
JP		 602 printk(KERN_CONT "block %llu:", (unsigned long long) block);
603 printk(KERN_CONT "%pV\n", &vaf);
5d1b1b3f
AK		 604 va_end(args);
605
606 if (test_opt(sb, ERRORS_CONT)) {
e2d67052 607 ext4_commit_super(sb, 0);
5d1b1b3f
AK		 608 return;
609 }
1c13d5c0 610
5d1b1b3f
AK		 611 ext4_unlock_group(sb, grp);
612 ext4_handle_error(sb);
613 /*
614 * We only get here in the ERRORS_RO case; relocking the group
615 * may be dangerous, but nothing bad will happen since the
616 * filesystem will have already been marked read/only and the
617 * journal has been aborted. We return 1 as a hint to callers
618 * who might what to use the return value from
619 * ext4_grp_locked_error() to distinguish beween the
620 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
621 * aggressively from the ext4 function in question, with a
622 * more appropriate error code.
623 */
624 ext4_lock_group(sb, grp);
625 return;
626}
627
617ba13b 628void ext4_update_dynamic_rev(struct super_block *sb)
ac27a0ec 629{
617ba13b 630 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
ac27a0ec 631
617ba13b 632 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
ac27a0ec
DK		 633 return;
634
12062ddd 635 ext4_warning(sb,
ac27a0ec
DK		 636 "updating to rev %d because of new feature flag, "
637 "running e2fsck is recommended",
617ba13b 638 EXT4_DYNAMIC_REV);
ac27a0ec 639
617ba13b
MC		 640 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
641 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
642 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
ac27a0ec
DK		 643 /* leave es->s_feature_*compat flags alone */
644 /* es->s_uuid will be set by e2fsck if empty */
645
646 /*
647 * The rest of the superblock fields should be zero, and if not it
648 * means they are likely already in use, so leave them alone. We
649 * can leave it up to e2fsck to clean up any inconsistencies there.
650 */
651}
652
653/*
654 * Open the external journal device
655 */
b31e1552 656static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
ac27a0ec
DK		 657{
658 struct block_device *bdev;
659 char b[BDEVNAME_SIZE];
660
d4d77629 661 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
ac27a0ec
DK		 662 if (IS_ERR(bdev))
663 goto fail;
664 return bdev;
665
666fail:
b31e1552 667 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
ac27a0ec
DK		 668 __bdevname(dev, b), PTR_ERR(bdev));
669 return NULL;
670}
671
672/*
673 * Release the journal device
674 */
617ba13b 675static int ext4_blkdev_put(struct block_device *bdev)
ac27a0ec 676{
e525fd89 677 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
ac27a0ec
DK		 678}
679
617ba13b 680static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
ac27a0ec
DK		 681{
682 struct block_device *bdev;
683 int ret = -ENODEV;
684
685 bdev = sbi->journal_bdev;
686 if (bdev) {
617ba13b 687 ret = ext4_blkdev_put(bdev);
ac27a0ec
DK		 688 sbi->journal_bdev = NULL;
689 }
690 return ret;
691}
692
693static inline struct inode *orphan_list_entry(struct list_head *l)
694{
617ba13b 695 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
ac27a0ec
DK		 696}
697
617ba13b 698static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
ac27a0ec
DK		 699{
700 struct list_head *l;
701
b31e1552
ES		 702 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
703 le32_to_cpu(sbi->s_es->s_last_orphan));
ac27a0ec
DK		 704
705 printk(KERN_ERR "sb_info orphan list:\n");
706 list_for_each(l, &sbi->s_orphan) {
707 struct inode *inode = orphan_list_entry(l);
708 printk(KERN_ERR " "
709 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
710 inode->i_sb->s_id, inode->i_ino, inode,
711 inode->i_mode, inode->i_nlink,
712 NEXT_ORPHAN(inode));
713 }
714}
715
2b2d6d01 716static void ext4_put_super(struct super_block *sb)
ac27a0ec 717{
617ba13b
MC		 718 struct ext4_sb_info *sbi = EXT4_SB(sb);
719 struct ext4_super_block *es = sbi->s_es;
ef2cabf7 720 int i, err;
ac27a0ec 721
857ac889 722 ext4_unregister_li_request(sb);
e0ccfd95
CH		 723 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
724
4c0425ff
MC		 725 flush_workqueue(sbi->dio_unwritten_wq);
726 destroy_workqueue(sbi->dio_unwritten_wq);
727
a9e220f8 728 lock_super(sb);
8c85e125 729 if (sb->s_dirt)
ebc1ac16 730 ext4_commit_super(sb, 1);
8c85e125 731
0390131b
FM		 732 if (sbi->s_journal) {
733 err = jbd2_journal_destroy(sbi->s_journal);
734 sbi->s_journal = NULL;
735 if (err < 0)
c67d859e 736 ext4_abort(sb, "Couldn't clean up the journal");
0390131b 737 }
d4edac31 738
a1c6c569 739 del_timer(&sbi->s_err_report);
d4edac31
JB		 740 ext4_release_system_zone(sb);
741 ext4_mb_release(sb);
742 ext4_ext_release(sb);
743 ext4_xattr_put_super(sb);
744
ac27a0ec 745 if (!(sb->s_flags & MS_RDONLY)) {
617ba13b 746 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 747 es->s_state = cpu_to_le16(sbi->s_mount_state);
e2d67052 748 ext4_commit_super(sb, 1);
ac27a0ec 749 }
240799cd 750 if (sbi->s_proc) {
9f6200bb 751 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 752 }
3197ebdb 753 kobject_del(&sbi->s_kobj);
ac27a0ec
DK		 754
755 for (i = 0; i < sbi->s_gdb_count; i++)
756 brelse(sbi->s_group_desc[i]);
757 kfree(sbi->s_group_desc);
c5ca7c76
TT		 758 if (is_vmalloc_addr(sbi->s_flex_groups))
759 vfree(sbi->s_flex_groups);
760 else
761 kfree(sbi->s_flex_groups);
ac27a0ec
DK		 762 percpu_counter_destroy(&sbi->s_freeblocks_counter);
763 percpu_counter_destroy(&sbi->s_freeinodes_counter);
764 percpu_counter_destroy(&sbi->s_dirs_counter);
6bc6e63f 765 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
ac27a0ec
DK		 766 brelse(sbi->s_sbh);
767#ifdef CONFIG_QUOTA
768 for (i = 0; i < MAXQUOTAS; i++)
769 kfree(sbi->s_qf_names[i]);
770#endif
771
772 /* Debugging code just in case the in-memory inode orphan list
773 * isn't empty. The on-disk one can be non-empty if we've
774 * detected an error and taken the fs readonly, but the
775 * in-memory list had better be clean by this point. */
776 if (!list_empty(&sbi->s_orphan))
777 dump_orphan_list(sb, sbi);
778 J_ASSERT(list_empty(&sbi->s_orphan));
779
f98393a6 780 invalidate_bdev(sb->s_bdev);
ac27a0ec
DK		 781 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
782 /*
783 * Invalidate the journal device's buffers. We don't want them
784 * floating about in memory - the physical journal device may
785 * hotswapped, and it breaks the `ro-after' testing code.
786 */
787 sync_blockdev(sbi->journal_bdev);
f98393a6 788 invalidate_bdev(sbi->journal_bdev);
617ba13b 789 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 790 }
791 sb->s_fs_info = NULL;
3197ebdb
TT		 792 /*
793 * Now that we are completely done shutting down the
794 * superblock, we need to actually destroy the kobject.
795 */
3197ebdb
TT		 796 unlock_super(sb);
797 kobject_put(&sbi->s_kobj);
798 wait_for_completion(&sbi->s_kobj_unregister);
705895b6 799 kfree(sbi->s_blockgroup_lock);
ac27a0ec 800 kfree(sbi);
ac27a0ec
DK		 801}
802
e18b890b 803static struct kmem_cache *ext4_inode_cachep;
ac27a0ec
DK		 804
805/*
806 * Called inside transaction, so use GFP_NOFS
807 */
617ba13b 808static struct inode *ext4_alloc_inode(struct super_block *sb)
ac27a0ec 809{
617ba13b 810 struct ext4_inode_info *ei;
ac27a0ec 811
e6b4f8da 812 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
ac27a0ec
DK		 813 if (!ei)
814 return NULL;
0b8e58a1 815
ac27a0ec 816 ei->vfs_inode.i_version = 1;
91246c00 817 ei->vfs_inode.i_data.writeback_index = 0;
a86c6181 818 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
c9de560d
AT		 819 INIT_LIST_HEAD(&ei->i_prealloc_list);
820 spin_lock_init(&ei->i_prealloc_lock);
d2a17637
MC		 821 ei->i_reserved_data_blocks = 0;
822 ei->i_reserved_meta_blocks = 0;
823 ei->i_allocated_meta_blocks = 0;
9d0be502 824 ei->i_da_metadata_calc_len = 0;
d2a17637 825 spin_lock_init(&(ei->i_block_reservation_lock));
a9e7f447
DM		 826#ifdef CONFIG_QUOTA
827 ei->i_reserved_quota = 0;
828#endif
8aefcd55 829 ei->jinode = NULL;
c7064ef1 830 INIT_LIST_HEAD(&ei->i_completed_io_list);
744692dc 831 spin_lock_init(&ei->i_completed_io_lock);
8d5d02e6 832 ei->cur_aio_dio = NULL;
b436b9be
JK		 833 ei->i_sync_tid = 0;
834 ei->i_datasync_tid = 0;
f7ad6d2e 835 atomic_set(&ei->i_ioend_count, 0);
e9e3bcec 836 atomic_set(&ei->i_aiodio_unwritten, 0);
0b8e58a1 837
ac27a0ec
DK		 838 return &ei->vfs_inode;
839}
840
7ff9c073
TT		 841static int ext4_drop_inode(struct inode *inode)
842{
843 int drop = generic_drop_inode(inode);
844
845 trace_ext4_drop_inode(inode, drop);
846 return drop;
847}
848
fa0d7e3d
NP		 849static void ext4_i_callback(struct rcu_head *head)
850{
851 struct inode *inode = container_of(head, struct inode, i_rcu);
852 INIT_LIST_HEAD(&inode->i_dentry);
853 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
854}
855
617ba13b 856static void ext4_destroy_inode(struct inode *inode)
ac27a0ec 857{
f7ad6d2e 858 ext4_ioend_wait(inode);
9f7dd93d 859 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
b31e1552
ES		 860 ext4_msg(inode->i_sb, KERN_ERR,
861 "Inode %lu (%p): orphan list check failed!",
862 inode->i_ino, EXT4_I(inode));
9f7dd93d
VA		 863 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
864 EXT4_I(inode), sizeof(struct ext4_inode_info),
865 true);
866 dump_stack();
867 }
fa0d7e3d 868 call_rcu(&inode->i_rcu, ext4_i_callback);
ac27a0ec
DK		 869}
870
51cc5068 871static void init_once(void *foo)
ac27a0ec 872{
617ba13b 873 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
ac27a0ec 874
a35afb83 875 INIT_LIST_HEAD(&ei->i_orphan);
03010a33 876#ifdef CONFIG_EXT4_FS_XATTR
a35afb83 877 init_rwsem(&ei->xattr_sem);
ac27a0ec 878#endif
0e855ac8 879 init_rwsem(&ei->i_data_sem);
a35afb83 880 inode_init_once(&ei->vfs_inode);
ac27a0ec
DK		 881}
882
883static int init_inodecache(void)
884{
617ba13b
MC		 885 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
886 sizeof(struct ext4_inode_info),
ac27a0ec
DK		 887 0, (SLAB_RECLAIM_ACCOUNT|
888 SLAB_MEM_SPREAD),
20c2df83 889 init_once);
617ba13b 890 if (ext4_inode_cachep == NULL)
ac27a0ec
DK		 891 return -ENOMEM;
892 return 0;
893}
894
895static void destroy_inodecache(void)
896{
617ba13b 897 kmem_cache_destroy(ext4_inode_cachep);
ac27a0ec
DK		 898}
899
0930fcc1 900void ext4_clear_inode(struct inode *inode)
ac27a0ec 901{
0930fcc1
AV		 902 invalidate_inode_buffers(inode);
903 end_writeback(inode);
9f754758 904 dquot_drop(inode);
c2ea3fde 905 ext4_discard_preallocations(inode);
8aefcd55
TT		 906 if (EXT4_I(inode)->jinode) {
907 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
908 EXT4_I(inode)->jinode);
909 jbd2_free_inode(EXT4_I(inode)->jinode);
910 EXT4_I(inode)->jinode = NULL;
911 }
ac27a0ec
DK		 912}
913
2b2d6d01
TT		 914static inline void ext4_show_quota_options(struct seq_file *seq,
915 struct super_block *sb)
ac27a0ec
DK		 916{
917#if defined(CONFIG_QUOTA)
617ba13b 918 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 919
5a20bdfc
JK		 920 if (sbi->s_jquota_fmt) {
921 char *fmtname = "";
922
923 switch (sbi->s_jquota_fmt) {
924 case QFMT_VFS_OLD:
925 fmtname = "vfsold";
926 break;
927 case QFMT_VFS_V0:
928 fmtname = "vfsv0";
929 break;
930 case QFMT_VFS_V1:
931 fmtname = "vfsv1";
932 break;
933 }
934 seq_printf(seq, ",jqfmt=%s", fmtname);
935 }
ac27a0ec
DK		 936
937 if (sbi->s_qf_names[USRQUOTA])
938 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
939
940 if (sbi->s_qf_names[GRPQUOTA])
941 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
942
482a7425 943 if (test_opt(sb, USRQUOTA))
ac27a0ec
DK		 944 seq_puts(seq, ",usrquota");
945
482a7425 946 if (test_opt(sb, GRPQUOTA))
ac27a0ec
DK		 947 seq_puts(seq, ",grpquota");
948#endif
949}
950
d9c9bef1
MS		 951/*
952 * Show an option if
953 * - it's set to a non-default value OR
954 * - if the per-sb default is different from the global default
955 */
617ba13b 956static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
ac27a0ec 957{
aa22df2c
AK		 958 int def_errors;
959 unsigned long def_mount_opts;
ac27a0ec 960 struct super_block *sb = vfs->mnt_sb;
d9c9bef1
MS		 961 struct ext4_sb_info *sbi = EXT4_SB(sb);
962 struct ext4_super_block *es = sbi->s_es;
d9c9bef1
MS		 963
964 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
aa22df2c 965 def_errors = le16_to_cpu(es->s_errors);
d9c9bef1
MS		 966
967 if (sbi->s_sb_block != 1)
968 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
969 if (test_opt(sb, MINIX_DF))
970 seq_puts(seq, ",minixdf");
aa22df2c 971 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
d9c9bef1
MS		 972 seq_puts(seq, ",grpid");
973 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
974 seq_puts(seq, ",nogrpid");
975 if (sbi->s_resuid != EXT4_DEF_RESUID ||
976 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
977 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
978 }
979 if (sbi->s_resgid != EXT4_DEF_RESGID ||
980 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
981 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
982 }
bb4f397a 983 if (test_opt(sb, ERRORS_RO)) {
d9c9bef1 984 if (def_errors == EXT4_ERRORS_PANIC ||
bb4f397a
AK		 985 def_errors == EXT4_ERRORS_CONTINUE) {
986 seq_puts(seq, ",errors=remount-ro");
d9c9bef1
MS		 987 }
988 }
aa22df2c 989 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
bb4f397a 990 seq_puts(seq, ",errors=continue");
aa22df2c 991 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
d9c9bef1 992 seq_puts(seq, ",errors=panic");
aa22df2c 993 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
d9c9bef1 994 seq_puts(seq, ",nouid32");
aa22df2c 995 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
d9c9bef1
MS		 996 seq_puts(seq, ",debug");
997 if (test_opt(sb, OLDALLOC))
998 seq_puts(seq, ",oldalloc");
03010a33 999#ifdef CONFIG_EXT4_FS_XATTR
ea663336 1000 if (test_opt(sb, XATTR_USER))
d9c9bef1 1001 seq_puts(seq, ",user_xattr");
ea663336 1002 if (!test_opt(sb, XATTR_USER))
d9c9bef1 1003 seq_puts(seq, ",nouser_xattr");
d9c9bef1 1004#endif
03010a33 1005#ifdef CONFIG_EXT4_FS_POSIX_ACL
aa22df2c 1006 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
d9c9bef1
MS		 1007 seq_puts(seq, ",acl");
1008 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
1009 seq_puts(seq, ",noacl");
1010#endif
30773840 1011 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
d9c9bef1
MS		 1012 seq_printf(seq, ",commit=%u",
1013 (unsigned) (sbi->s_commit_interval / HZ));
1014 }
30773840
TT		 1015 if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
1016 seq_printf(seq, ",min_batch_time=%u",
1017 (unsigned) sbi->s_min_batch_time);
1018 }
1019 if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
1020 seq_printf(seq, ",max_batch_time=%u",
1021 (unsigned) sbi->s_min_batch_time);
1022 }
1023
571640ca
ES		 1024 /*
1025 * We're changing the default of barrier mount option, so
1026 * let's always display its mount state so it's clear what its
1027 * status is.
1028 */
1029 seq_puts(seq, ",barrier=");
1030 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
cd0b6a39
TT		 1031 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
1032 seq_puts(seq, ",journal_async_commit");
39a4bade
JK		 1033 else if (test_opt(sb, JOURNAL_CHECKSUM))
1034 seq_puts(seq, ",journal_checksum");
25ec56b5
JNC		 1035 if (test_opt(sb, I_VERSION))
1036 seq_puts(seq, ",i_version");
8b67f04a
TT		 1037 if (!test_opt(sb, DELALLOC) &&
1038 !(def_mount_opts & EXT4_DEFM_NODELALLOC))
dd919b98
AK		 1039 seq_puts(seq, ",nodelalloc");
1040
6fd7a467
TT		 1041 if (!test_opt(sb, MBLK_IO_SUBMIT))
1042 seq_puts(seq, ",nomblk_io_submit");
cb45bbe4
MS		 1043 if (sbi->s_stripe)
1044 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
aa22df2c
AK		 1045 /*
1046 * journal mode get enabled in different ways
1047 * So just print the value even if we didn't specify it
1048 */
617ba13b 1049 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
ac27a0ec 1050 seq_puts(seq, ",data=journal");
617ba13b 1051 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
ac27a0ec 1052 seq_puts(seq, ",data=ordered");
617ba13b 1053 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
ac27a0ec
DK		 1054 seq_puts(seq, ",data=writeback");
1055
240799cd
TT		 1056 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1057 seq_printf(seq, ",inode_readahead_blks=%u",
1058 sbi->s_inode_readahead_blks);
1059
5bf5683a
HK		 1060 if (test_opt(sb, DATA_ERR_ABORT))
1061 seq_puts(seq, ",data_err=abort");
1062
afd4672d 1063 if (test_opt(sb, NO_AUTO_DA_ALLOC))
06705bff 1064 seq_puts(seq, ",noauto_da_alloc");
afd4672d 1065
8b67f04a 1066 if (test_opt(sb, DISCARD) && !(def_mount_opts & EXT4_DEFM_DISCARD))
5328e635
ES		 1067 seq_puts(seq, ",discard");
1068
e3bb52ae
ES		 1069 if (test_opt(sb, NOLOAD))
1070 seq_puts(seq, ",norecovery");
1071
744692dc
JZ		 1072 if (test_opt(sb, DIOREAD_NOLOCK))
1073 seq_puts(seq, ",dioread_nolock");
1074
8b67f04a
TT		 1075 if (test_opt(sb, BLOCK_VALIDITY) &&
1076 !(def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY))
1077 seq_puts(seq, ",block_validity");
1078
bfff6873
LC		 1079 if (!test_opt(sb, INIT_INODE_TABLE))
1080 seq_puts(seq, ",noinit_inode_table");
1081 else if (sbi->s_li_wait_mult)
1082 seq_printf(seq, ",init_inode_table=%u",
1083 (unsigned) sbi->s_li_wait_mult);
1084
617ba13b 1085 ext4_show_quota_options(seq, sb);
0b8e58a1 1086
ac27a0ec
DK		 1087 return 0;
1088}
1089
1b961ac0 1090static struct inode *ext4_nfs_get_inode(struct super_block *sb,
0b8e58a1 1091 u64 ino, u32 generation)
ac27a0ec 1092{
ac27a0ec 1093 struct inode *inode;
ac27a0ec 1094
617ba13b 1095 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
ac27a0ec 1096 return ERR_PTR(-ESTALE);
617ba13b 1097 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
ac27a0ec
DK		 1098 return ERR_PTR(-ESTALE);
1099
1100 /* iget isn't really right if the inode is currently unallocated!!
1101 *
617ba13b 1102 * ext4_read_inode will return a bad_inode if the inode had been
ac27a0ec
DK		 1103 * deleted, so we should be safe.
1104 *
1105 * Currently we don't know the generation for parent directory, so
1106 * a generation of 0 means "accept any"
1107 */
1d1fe1ee
DH		 1108 inode = ext4_iget(sb, ino);
1109 if (IS_ERR(inode))
1110 return ERR_CAST(inode);
1111 if (generation && inode->i_generation != generation) {
ac27a0ec
DK		 1112 iput(inode);
1113 return ERR_PTR(-ESTALE);
1114 }
1b961ac0
CH		 1115
1116 return inode;
1117}
1118
1119static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
0b8e58a1 1120 int fh_len, int fh_type)
1b961ac0
CH		 1121{
1122 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1123 ext4_nfs_get_inode);
1124}
1125
1126static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
0b8e58a1 1127 int fh_len, int fh_type)
1b961ac0
CH		 1128{
1129 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1130 ext4_nfs_get_inode);
ac27a0ec
DK		 1131}
1132
c39a7f84
TO		 1133/*
1134 * Try to release metadata pages (indirect blocks, directories) which are
1135 * mapped via the block device. Since these pages could have journal heads
1136 * which would prevent try_to_free_buffers() from freeing them, we must use
1137 * jbd2 layer's try_to_free_buffers() function to release them.
1138 */
0b8e58a1
AD		 1139static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1140 gfp_t wait)
c39a7f84
TO		 1141{
1142 journal_t *journal = EXT4_SB(sb)->s_journal;
1143
1144 WARN_ON(PageChecked(page));
1145 if (!page_has_buffers(page))
1146 return 0;
1147 if (journal)
1148 return jbd2_journal_try_to_free_buffers(journal, page,
1149 wait & ~__GFP_WAIT);
1150 return try_to_free_buffers(page);
1151}
1152
ac27a0ec 1153#ifdef CONFIG_QUOTA
af5bc92d 1154#define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
2b2d6d01 1155#define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
ac27a0ec 1156
617ba13b
MC		 1157static int ext4_write_dquot(struct dquot *dquot);
1158static int ext4_acquire_dquot(struct dquot *dquot);
1159static int ext4_release_dquot(struct dquot *dquot);
1160static int ext4_mark_dquot_dirty(struct dquot *dquot);
1161static int ext4_write_info(struct super_block *sb, int type);
6f28e087 1162static int ext4_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 1163 struct path *path);
ca0e05e4 1164static int ext4_quota_off(struct super_block *sb, int type);
617ba13b
MC		 1165static int ext4_quota_on_mount(struct super_block *sb, int type);
1166static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 1167 size_t len, loff_t off);
617ba13b 1168static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 1169 const char *data, size_t len, loff_t off);
1170
61e225dc 1171static const struct dquot_operations ext4_quota_operations = {
a9e7f447 1172#ifdef CONFIG_QUOTA
60e58e0f 1173 .get_reserved_space = ext4_get_reserved_space,
a9e7f447 1174#endif
617ba13b
MC		 1175 .write_dquot = ext4_write_dquot,
1176 .acquire_dquot = ext4_acquire_dquot,
1177 .release_dquot = ext4_release_dquot,
1178 .mark_dirty = ext4_mark_dquot_dirty,
a5b5ee32
JK		 1179 .write_info = ext4_write_info,
1180 .alloc_dquot = dquot_alloc,
1181 .destroy_dquot = dquot_destroy,
ac27a0ec
DK		 1182};
1183
0d54b217 1184static const struct quotactl_ops ext4_qctl_operations = {
617ba13b 1185 .quota_on = ext4_quota_on,
ca0e05e4 1186 .quota_off = ext4_quota_off,
287a8095
CH		 1187 .quota_sync = dquot_quota_sync,
1188 .get_info = dquot_get_dqinfo,
1189 .set_info = dquot_set_dqinfo,
1190 .get_dqblk = dquot_get_dqblk,
1191 .set_dqblk = dquot_set_dqblk
ac27a0ec
DK		 1192};
1193#endif
1194
ee9b6d61 1195static const struct super_operations ext4_sops = {
617ba13b
MC		 1196 .alloc_inode = ext4_alloc_inode,
1197 .destroy_inode = ext4_destroy_inode,
617ba13b
MC		 1198 .write_inode = ext4_write_inode,
1199 .dirty_inode = ext4_dirty_inode,
7ff9c073 1200 .drop_inode = ext4_drop_inode,
0930fcc1 1201 .evict_inode = ext4_evict_inode,
617ba13b 1202 .put_super = ext4_put_super,
617ba13b 1203 .sync_fs = ext4_sync_fs,
c4be0c1d
TS		 1204 .freeze_fs = ext4_freeze,
1205 .unfreeze_fs = ext4_unfreeze,
617ba13b
MC		 1206 .statfs = ext4_statfs,
1207 .remount_fs = ext4_remount,
617ba13b 1208 .show_options = ext4_show_options,
ac27a0ec 1209#ifdef CONFIG_QUOTA
617ba13b
MC		 1210 .quota_read = ext4_quota_read,
1211 .quota_write = ext4_quota_write,
ac27a0ec 1212#endif
c39a7f84 1213 .bdev_try_to_free_page = bdev_try_to_free_page,
ac27a0ec
DK		 1214};
1215
9ca92389
TT		 1216static const struct super_operations ext4_nojournal_sops = {
1217 .alloc_inode = ext4_alloc_inode,
1218 .destroy_inode = ext4_destroy_inode,
1219 .write_inode = ext4_write_inode,
1220 .dirty_inode = ext4_dirty_inode,
7ff9c073 1221 .drop_inode = ext4_drop_inode,
0930fcc1 1222 .evict_inode = ext4_evict_inode,
9ca92389
TT		 1223 .write_super = ext4_write_super,
1224 .put_super = ext4_put_super,
1225 .statfs = ext4_statfs,
1226 .remount_fs = ext4_remount,
9ca92389
TT		 1227 .show_options = ext4_show_options,
1228#ifdef CONFIG_QUOTA
1229 .quota_read = ext4_quota_read,
1230 .quota_write = ext4_quota_write,
1231#endif
1232 .bdev_try_to_free_page = bdev_try_to_free_page,
1233};
1234
39655164 1235static const struct export_operations ext4_export_ops = {
1b961ac0
CH		 1236 .fh_to_dentry = ext4_fh_to_dentry,
1237 .fh_to_parent = ext4_fh_to_parent,
617ba13b 1238 .get_parent = ext4_get_parent,
ac27a0ec
DK		 1239};
1240
1241enum {
1242 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1243 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
01436ef2 1244 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
ac27a0ec 1245 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
06705bff 1246 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh,
30773840 1247 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
c3191067 1248 Opt_journal_update, Opt_journal_dev,
818d276c 1249 Opt_journal_checksum, Opt_journal_async_commit,
ac27a0ec 1250 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
296c355c 1251 Opt_data_err_abort, Opt_data_err_ignore,
ac27a0ec 1252 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
5a20bdfc
JK		 1253 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1254 Opt_noquota, Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err,
1255 Opt_resize, Opt_usrquota, Opt_grpquota, Opt_i_version,
1449032b
TT		 1256 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1257 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
5328e635 1258 Opt_inode_readahead_blks, Opt_journal_ioprio,
744692dc 1259 Opt_dioread_nolock, Opt_dioread_lock,
5328e635 1260 Opt_discard, Opt_nodiscard,
bfff6873 1261 Opt_init_inode_table, Opt_noinit_inode_table,
ac27a0ec
DK		 1262};
1263
a447c093 1264static const match_table_t tokens = {
ac27a0ec
DK		 1265 {Opt_bsd_df, "bsddf"},
1266 {Opt_minix_df, "minixdf"},
1267 {Opt_grpid, "grpid"},
1268 {Opt_grpid, "bsdgroups"},
1269 {Opt_nogrpid, "nogrpid"},
1270 {Opt_nogrpid, "sysvgroups"},
1271 {Opt_resgid, "resgid=%u"},
1272 {Opt_resuid, "resuid=%u"},
1273 {Opt_sb, "sb=%u"},
1274 {Opt_err_cont, "errors=continue"},
1275 {Opt_err_panic, "errors=panic"},
1276 {Opt_err_ro, "errors=remount-ro"},
1277 {Opt_nouid32, "nouid32"},
ac27a0ec
DK		 1278 {Opt_debug, "debug"},
1279 {Opt_oldalloc, "oldalloc"},
1280 {Opt_orlov, "orlov"},
1281 {Opt_user_xattr, "user_xattr"},
1282 {Opt_nouser_xattr, "nouser_xattr"},
1283 {Opt_acl, "acl"},
1284 {Opt_noacl, "noacl"},
ac27a0ec 1285 {Opt_noload, "noload"},
e3bb52ae 1286 {Opt_noload, "norecovery"},
ac27a0ec
DK		 1287 {Opt_nobh, "nobh"},
1288 {Opt_bh, "bh"},
1289 {Opt_commit, "commit=%u"},
30773840
TT		 1290 {Opt_min_batch_time, "min_batch_time=%u"},
1291 {Opt_max_batch_time, "max_batch_time=%u"},
ac27a0ec 1292 {Opt_journal_update, "journal=update"},
ac27a0ec 1293 {Opt_journal_dev, "journal_dev=%u"},
818d276c
GS		 1294 {Opt_journal_checksum, "journal_checksum"},
1295 {Opt_journal_async_commit, "journal_async_commit"},
ac27a0ec
DK		 1296 {Opt_abort, "abort"},
1297 {Opt_data_journal, "data=journal"},
1298 {Opt_data_ordered, "data=ordered"},
1299 {Opt_data_writeback, "data=writeback"},
5bf5683a
HK		 1300 {Opt_data_err_abort, "data_err=abort"},
1301 {Opt_data_err_ignore, "data_err=ignore"},
ac27a0ec
DK		 1302 {Opt_offusrjquota, "usrjquota="},
1303 {Opt_usrjquota, "usrjquota=%s"},
1304 {Opt_offgrpjquota, "grpjquota="},
1305 {Opt_grpjquota, "grpjquota=%s"},
1306 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1307 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
5a20bdfc 1308 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
ac27a0ec
DK		 1309 {Opt_grpquota, "grpquota"},
1310 {Opt_noquota, "noquota"},
1311 {Opt_quota, "quota"},
1312 {Opt_usrquota, "usrquota"},
1313 {Opt_barrier, "barrier=%u"},
06705bff
TT		 1314 {Opt_barrier, "barrier"},
1315 {Opt_nobarrier, "nobarrier"},
25ec56b5 1316 {Opt_i_version, "i_version"},
c9de560d 1317 {Opt_stripe, "stripe=%u"},
ac27a0ec 1318 {Opt_resize, "resize"},
64769240 1319 {Opt_delalloc, "delalloc"},
dd919b98 1320 {Opt_nodelalloc, "nodelalloc"},
1449032b
TT		 1321 {Opt_mblk_io_submit, "mblk_io_submit"},
1322 {Opt_nomblk_io_submit, "nomblk_io_submit"},
6fd058f7
TT		 1323 {Opt_block_validity, "block_validity"},
1324 {Opt_noblock_validity, "noblock_validity"},
240799cd 1325 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
b3881f74 1326 {Opt_journal_ioprio, "journal_ioprio=%u"},
afd4672d 1327 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
06705bff
TT		 1328 {Opt_auto_da_alloc, "auto_da_alloc"},
1329 {Opt_noauto_da_alloc, "noauto_da_alloc"},
744692dc
JZ		 1330 {Opt_dioread_nolock, "dioread_nolock"},
1331 {Opt_dioread_lock, "dioread_lock"},
5328e635
ES		 1332 {Opt_discard, "discard"},
1333 {Opt_nodiscard, "nodiscard"},
bfff6873
LC		 1334 {Opt_init_inode_table, "init_itable=%u"},
1335 {Opt_init_inode_table, "init_itable"},
1336 {Opt_noinit_inode_table, "noinit_itable"},
f3f12faa 1337 {Opt_err, NULL},
ac27a0ec
DK		 1338};
1339
617ba13b 1340static ext4_fsblk_t get_sb_block(void **data)
ac27a0ec 1341{
617ba13b 1342 ext4_fsblk_t sb_block;
ac27a0ec
DK		 1343 char *options = (char *) *data;
1344
1345 if (!options || strncmp(options, "sb=", 3) != 0)
1346 return 1; /* Default location */
0b8e58a1 1347
ac27a0ec 1348 options += 3;
0b8e58a1 1349 /* TODO: use simple_strtoll with >32bit ext4 */
ac27a0ec
DK		 1350 sb_block = simple_strtoul(options, &options, 0);
1351 if (*options && *options != ',') {
4776004f 1352 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
ac27a0ec
DK		 1353 (char *) *data);
1354 return 1;
1355 }
1356 if (*options == ',')
1357 options++;
1358 *data = (void *) options;
0b8e58a1 1359
ac27a0ec
DK		 1360 return sb_block;
1361}
1362
b3881f74 1363#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
437ca0fd
DM		 1364static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1365 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
b3881f74 1366
56c50f11
DM		 1367#ifdef CONFIG_QUOTA
1368static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1369{
1370 struct ext4_sb_info *sbi = EXT4_SB(sb);
1371 char *qname;
1372
1373 if (sb_any_quota_loaded(sb) &&
1374 !sbi->s_qf_names[qtype]) {
1375 ext4_msg(sb, KERN_ERR,
1376 "Cannot change journaled "
1377 "quota options when quota turned on");
1378 return 0;
1379 }
1380 qname = match_strdup(args);
1381 if (!qname) {
1382 ext4_msg(sb, KERN_ERR,
1383 "Not enough memory for storing quotafile name");
1384 return 0;
1385 }
1386 if (sbi->s_qf_names[qtype] &&
1387 strcmp(sbi->s_qf_names[qtype], qname)) {
1388 ext4_msg(sb, KERN_ERR,
1389 "%s quota file already specified", QTYPE2NAME(qtype));
1390 kfree(qname);
1391 return 0;
1392 }
1393 sbi->s_qf_names[qtype] = qname;
1394 if (strchr(sbi->s_qf_names[qtype], '/')) {
1395 ext4_msg(sb, KERN_ERR,
1396 "quotafile must be on filesystem root");
1397 kfree(sbi->s_qf_names[qtype]);
1398 sbi->s_qf_names[qtype] = NULL;
1399 return 0;
1400 }
fd8c37ec 1401 set_opt(sb, QUOTA);
56c50f11
DM		 1402 return 1;
1403}
1404
1405static int clear_qf_name(struct super_block *sb, int qtype)
1406{
1407
1408 struct ext4_sb_info *sbi = EXT4_SB(sb);
1409
1410 if (sb_any_quota_loaded(sb) &&
1411 sbi->s_qf_names[qtype]) {
1412 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1413 " when quota turned on");
1414 return 0;
1415 }
1416 /*
1417 * The space will be released later when all options are confirmed
1418 * to be correct
1419 */
1420 sbi->s_qf_names[qtype] = NULL;
1421 return 1;
1422}
1423#endif
1424
2b2d6d01 1425static int parse_options(char *options, struct super_block *sb,
c3191067 1426 unsigned long *journal_devnum,
b3881f74 1427 unsigned int *journal_ioprio,
2b2d6d01 1428 ext4_fsblk_t *n_blocks_count, int is_remount)
ac27a0ec 1429{
617ba13b 1430 struct ext4_sb_info *sbi = EXT4_SB(sb);
2b2d6d01 1431 char *p;
ac27a0ec
DK		 1432 substring_t args[MAX_OPT_ARGS];
1433 int data_opt = 0;
1434 int option;
1435#ifdef CONFIG_QUOTA
56c50f11 1436 int qfmt;
ac27a0ec
DK		 1437#endif
1438
1439 if (!options)
1440 return 1;
1441
2b2d6d01 1442 while ((p = strsep(&options, ",")) != NULL) {
ac27a0ec
DK		 1443 int token;
1444 if (!*p)
1445 continue;
1446
15121c18
ES		 1447 /*
1448 * Initialize args struct so we know whether arg was
1449 * found; some options take optional arguments.
1450 */
7dc57615 1451 args[0].to = args[0].from = NULL;
ac27a0ec
DK		 1452 token = match_token(p, tokens, args);
1453 switch (token) {
1454 case Opt_bsd_df:
437ca0fd 1455 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
fd8c37ec 1456 clear_opt(sb, MINIX_DF);
ac27a0ec
DK		 1457 break;
1458 case Opt_minix_df:
437ca0fd 1459 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
fd8c37ec 1460 set_opt(sb, MINIX_DF);
437ca0fd 1461
ac27a0ec
DK		 1462 break;
1463 case Opt_grpid:
437ca0fd 1464 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
fd8c37ec 1465 set_opt(sb, GRPID);
437ca0fd 1466
ac27a0ec
DK		 1467 break;
1468 case Opt_nogrpid:
437ca0fd 1469 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
fd8c37ec 1470 clear_opt(sb, GRPID);
437ca0fd 1471
ac27a0ec
DK		 1472 break;
1473 case Opt_resuid:
1474 if (match_int(&args[0], &option))
1475 return 0;
1476 sbi->s_resuid = option;
1477 break;
1478 case Opt_resgid:
1479 if (match_int(&args[0], &option))
1480 return 0;
1481 sbi->s_resgid = option;
1482 break;
1483 case Opt_sb:
1484 /* handled by get_sb_block() instead of here */
1485 /* *sb_block = match_int(&args[0]); */
1486 break;
1487 case Opt_err_panic:
fd8c37ec
TT		 1488 clear_opt(sb, ERRORS_CONT);
1489 clear_opt(sb, ERRORS_RO);
1490 set_opt(sb, ERRORS_PANIC);
ac27a0ec
DK		 1491 break;
1492 case Opt_err_ro:
fd8c37ec
TT		 1493 clear_opt(sb, ERRORS_CONT);
1494 clear_opt(sb, ERRORS_PANIC);
1495 set_opt(sb, ERRORS_RO);
ac27a0ec
DK		 1496 break;
1497 case Opt_err_cont:
fd8c37ec
TT		 1498 clear_opt(sb, ERRORS_RO);
1499 clear_opt(sb, ERRORS_PANIC);
1500 set_opt(sb, ERRORS_CONT);
ac27a0ec
DK		 1501 break;
1502 case Opt_nouid32:
fd8c37ec 1503 set_opt(sb, NO_UID32);
ac27a0ec 1504 break;
ac27a0ec 1505 case Opt_debug:
fd8c37ec 1506 set_opt(sb, DEBUG);
ac27a0ec
DK		 1507 break;
1508 case Opt_oldalloc:
fd8c37ec 1509 set_opt(sb, OLDALLOC);
ac27a0ec
DK		 1510 break;
1511 case Opt_orlov:
fd8c37ec 1512 clear_opt(sb, OLDALLOC);
ac27a0ec 1513 break;
03010a33 1514#ifdef CONFIG_EXT4_FS_XATTR
ac27a0ec 1515 case Opt_user_xattr:
fd8c37ec 1516 set_opt(sb, XATTR_USER);
ac27a0ec
DK		 1517 break;
1518 case Opt_nouser_xattr:
fd8c37ec 1519 clear_opt(sb, XATTR_USER);
ac27a0ec
DK		 1520 break;
1521#else
1522 case Opt_user_xattr:
1523 case Opt_nouser_xattr:
b31e1552 1524 ext4_msg(sb, KERN_ERR, "(no)user_xattr options not supported");
ac27a0ec
DK		 1525 break;
1526#endif
03010a33 1527#ifdef CONFIG_EXT4_FS_POSIX_ACL
ac27a0ec 1528 case Opt_acl:
fd8c37ec 1529 set_opt(sb, POSIX_ACL);
ac27a0ec
DK		 1530 break;
1531 case Opt_noacl:
fd8c37ec 1532 clear_opt(sb, POSIX_ACL);
ac27a0ec
DK		 1533 break;
1534#else
1535 case Opt_acl:
1536 case Opt_noacl:
b31e1552 1537 ext4_msg(sb, KERN_ERR, "(no)acl options not supported");
ac27a0ec
DK		 1538 break;
1539#endif
ac27a0ec
DK		 1540 case Opt_journal_update:
1541 /* @@@ FIXME */
1542 /* Eventually we will want to be able to create
1543 a journal file here. For now, only allow the
1544 user to specify an existing inode to be the
1545 journal file. */
1546 if (is_remount) {
b31e1552
ES		 1547 ext4_msg(sb, KERN_ERR,
1548 "Cannot specify journal on remount");
ac27a0ec
DK		 1549 return 0;
1550 }
fd8c37ec 1551 set_opt(sb, UPDATE_JOURNAL);
ac27a0ec 1552 break;
ac27a0ec
DK		 1553 case Opt_journal_dev:
1554 if (is_remount) {
b31e1552
ES		 1555 ext4_msg(sb, KERN_ERR,
1556 "Cannot specify journal on remount");
ac27a0ec
DK		 1557 return 0;
1558 }
1559 if (match_int(&args[0], &option))
1560 return 0;
1561 *journal_devnum = option;
1562 break;
818d276c 1563 case Opt_journal_checksum:
fd8c37ec 1564 set_opt(sb, JOURNAL_CHECKSUM);
d4da6c9c 1565 break;
818d276c 1566 case Opt_journal_async_commit:
fd8c37ec
TT		 1567 set_opt(sb, JOURNAL_ASYNC_COMMIT);
1568 set_opt(sb, JOURNAL_CHECKSUM);
818d276c 1569 break;
ac27a0ec 1570 case Opt_noload:
fd8c37ec 1571 set_opt(sb, NOLOAD);
ac27a0ec
DK		 1572 break;
1573 case Opt_commit:
1574 if (match_int(&args[0], &option))
1575 return 0;
1576 if (option < 0)
1577 return 0;
1578 if (option == 0)
cd02ff0b 1579 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
ac27a0ec
DK		 1580 sbi->s_commit_interval = HZ * option;
1581 break;
30773840
TT		 1582 case Opt_max_batch_time:
1583 if (match_int(&args[0], &option))
1584 return 0;
1585 if (option < 0)
1586 return 0;
1587 if (option == 0)
1588 option = EXT4_DEF_MAX_BATCH_TIME;
1589 sbi->s_max_batch_time = option;
1590 break;
1591 case Opt_min_batch_time:
1592 if (match_int(&args[0], &option))
1593 return 0;
1594 if (option < 0)
1595 return 0;
1596 sbi->s_min_batch_time = option;
1597 break;
ac27a0ec 1598 case Opt_data_journal:
617ba13b 1599 data_opt = EXT4_MOUNT_JOURNAL_DATA;
ac27a0ec
DK		 1600 goto datacheck;
1601 case Opt_data_ordered:
617ba13b 1602 data_opt = EXT4_MOUNT_ORDERED_DATA;
ac27a0ec
DK		 1603 goto datacheck;
1604 case Opt_data_writeback:
617ba13b 1605 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
ac27a0ec
DK		 1606 datacheck:
1607 if (is_remount) {
482a7425 1608 if (test_opt(sb, DATA_FLAGS) != data_opt) {
b31e1552
ES		 1609 ext4_msg(sb, KERN_ERR,
1610 "Cannot change data mode on remount");
ac27a0ec
DK		 1611 return 0;
1612 }
1613 } else {
fd8c37ec 1614 clear_opt(sb, DATA_FLAGS);
ac27a0ec
DK		 1615 sbi->s_mount_opt |= data_opt;
1616 }
1617 break;
5bf5683a 1618 case Opt_data_err_abort:
fd8c37ec 1619 set_opt(sb, DATA_ERR_ABORT);
5bf5683a
HK		 1620 break;
1621 case Opt_data_err_ignore:
fd8c37ec 1622 clear_opt(sb, DATA_ERR_ABORT);
5bf5683a 1623 break;
ac27a0ec
DK		 1624#ifdef CONFIG_QUOTA
1625 case Opt_usrjquota:
56c50f11 1626 if (!set_qf_name(sb, USRQUOTA, &args[0]))
ac27a0ec 1627 return 0;
56c50f11
DM		 1628 break;
1629 case Opt_grpjquota:
1630 if (!set_qf_name(sb, GRPQUOTA, &args[0]))
ac27a0ec 1631 return 0;
ac27a0ec
DK		 1632 break;
1633 case Opt_offusrjquota:
56c50f11
DM		 1634 if (!clear_qf_name(sb, USRQUOTA))
1635 return 0;
1636 break;
ac27a0ec 1637 case Opt_offgrpjquota:
56c50f11 1638 if (!clear_qf_name(sb, GRPQUOTA))
ac27a0ec 1639 return 0;
ac27a0ec 1640 break;
56c50f11 1641
ac27a0ec 1642 case Opt_jqfmt_vfsold:
dfc5d03f
JK		 1643 qfmt = QFMT_VFS_OLD;
1644 goto set_qf_format;
ac27a0ec 1645 case Opt_jqfmt_vfsv0:
dfc5d03f 1646 qfmt = QFMT_VFS_V0;
5a20bdfc
JK		 1647 goto set_qf_format;
1648 case Opt_jqfmt_vfsv1:
1649 qfmt = QFMT_VFS_V1;
dfc5d03f 1650set_qf_format:
17bd13b3 1651 if (sb_any_quota_loaded(sb) &&
dfc5d03f 1652 sbi->s_jquota_fmt != qfmt) {
b31e1552 1653 ext4_msg(sb, KERN_ERR, "Cannot change "
dfc5d03f 1654 "journaled quota options when "
b31e1552 1655 "quota turned on");
dfc5d03f
JK		 1656 return 0;
1657 }
1658 sbi->s_jquota_fmt = qfmt;
ac27a0ec
DK		 1659 break;
1660 case Opt_quota:
1661 case Opt_usrquota:
fd8c37ec
TT		 1662 set_opt(sb, QUOTA);
1663 set_opt(sb, USRQUOTA);
ac27a0ec
DK		 1664 break;
1665 case Opt_grpquota:
fd8c37ec
TT		 1666 set_opt(sb, QUOTA);
1667 set_opt(sb, GRPQUOTA);
ac27a0ec
DK		 1668 break;
1669 case Opt_noquota:
17bd13b3 1670 if (sb_any_quota_loaded(sb)) {
b31e1552
ES		 1671 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1672 "options when quota turned on");
ac27a0ec
DK		 1673 return 0;
1674 }
fd8c37ec
TT		 1675 clear_opt(sb, QUOTA);
1676 clear_opt(sb, USRQUOTA);
1677 clear_opt(sb, GRPQUOTA);
ac27a0ec
DK		 1678 break;
1679#else
1680 case Opt_quota:
1681 case Opt_usrquota:
1682 case Opt_grpquota:
b31e1552
ES		 1683 ext4_msg(sb, KERN_ERR,
1684 "quota options not supported");
cd59e7b9 1685 break;
ac27a0ec
DK		 1686 case Opt_usrjquota:
1687 case Opt_grpjquota:
1688 case Opt_offusrjquota:
1689 case Opt_offgrpjquota:
1690 case Opt_jqfmt_vfsold:
1691 case Opt_jqfmt_vfsv0:
5a20bdfc 1692 case Opt_jqfmt_vfsv1:
b31e1552
ES		 1693 ext4_msg(sb, KERN_ERR,
1694 "journaled quota options not supported");
ac27a0ec
DK		 1695 break;
1696 case Opt_noquota:
1697 break;
1698#endif
1699 case Opt_abort:
4ab2f15b 1700 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
ac27a0ec 1701 break;
06705bff 1702 case Opt_nobarrier:
fd8c37ec 1703 clear_opt(sb, BARRIER);
06705bff 1704 break;
ac27a0ec 1705 case Opt_barrier:
15121c18
ES		 1706 if (args[0].from) {
1707 if (match_int(&args[0], &option))
1708 return 0;
1709 } else
1710 option = 1; /* No argument, default to 1 */
ac27a0ec 1711 if (option)
fd8c37ec 1712 set_opt(sb, BARRIER);
ac27a0ec 1713 else
fd8c37ec 1714 clear_opt(sb, BARRIER);
ac27a0ec
DK		 1715 break;
1716 case Opt_ignore:
1717 break;
1718 case Opt_resize:
1719 if (!is_remount) {
b31e1552
ES		 1720 ext4_msg(sb, KERN_ERR,
1721 "resize option only available "
1722 "for remount");
ac27a0ec
DK		 1723 return 0;
1724 }
1725 if (match_int(&args[0], &option) != 0)
1726 return 0;
1727 *n_blocks_count = option;
1728 break;
1729 case Opt_nobh:
206f7ab4
CH		 1730 ext4_msg(sb, KERN_WARNING,
1731 "Ignoring deprecated nobh option");
ac27a0ec
DK		 1732 break;
1733 case Opt_bh:
206f7ab4
CH		 1734 ext4_msg(sb, KERN_WARNING,
1735 "Ignoring deprecated bh option");
ac27a0ec 1736 break;
25ec56b5 1737 case Opt_i_version:
fd8c37ec 1738 set_opt(sb, I_VERSION);
25ec56b5
JNC		 1739 sb->s_flags |= MS_I_VERSION;
1740 break;
dd919b98 1741 case Opt_nodelalloc:
fd8c37ec 1742 clear_opt(sb, DELALLOC);
dd919b98 1743 break;
1449032b 1744 case Opt_mblk_io_submit:
fd8c37ec 1745 set_opt(sb, MBLK_IO_SUBMIT);
1449032b
TT		 1746 break;
1747 case Opt_nomblk_io_submit:
fd8c37ec 1748 clear_opt(sb, MBLK_IO_SUBMIT);
1449032b 1749 break;
c9de560d
AT		 1750 case Opt_stripe:
1751 if (match_int(&args[0], &option))
1752 return 0;
1753 if (option < 0)
1754 return 0;
1755 sbi->s_stripe = option;
1756 break;
64769240 1757 case Opt_delalloc:
fd8c37ec 1758 set_opt(sb, DELALLOC);
64769240 1759 break;
6fd058f7 1760 case Opt_block_validity:
fd8c37ec 1761 set_opt(sb, BLOCK_VALIDITY);
6fd058f7
TT		 1762 break;
1763 case Opt_noblock_validity:
fd8c37ec 1764 clear_opt(sb, BLOCK_VALIDITY);
6fd058f7 1765 break;
240799cd
TT		 1766 case Opt_inode_readahead_blks:
1767 if (match_int(&args[0], &option))
1768 return 0;
1769 if (option < 0 || option > (1 << 30))
1770 return 0;
5dbd571d 1771 if (option && !is_power_of_2(option)) {
b31e1552
ES		 1772 ext4_msg(sb, KERN_ERR,
1773 "EXT4-fs: inode_readahead_blks"
1774 " must be a power of 2");
3197ebdb
TT		 1775 return 0;
1776 }
240799cd
TT		 1777 sbi->s_inode_readahead_blks = option;
1778 break;
b3881f74
TT		 1779 case Opt_journal_ioprio:
1780 if (match_int(&args[0], &option))
1781 return 0;
1782 if (option < 0 || option > 7)
1783 break;
1784 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1785 option);
1786 break;
06705bff 1787 case Opt_noauto_da_alloc:
fd8c37ec 1788 set_opt(sb, NO_AUTO_DA_ALLOC);
06705bff 1789 break;
afd4672d 1790 case Opt_auto_da_alloc:
15121c18
ES		 1791 if (args[0].from) {
1792 if (match_int(&args[0], &option))
1793 return 0;
1794 } else
1795 option = 1; /* No argument, default to 1 */
afd4672d 1796 if (option)
fd8c37ec 1797 clear_opt(sb, NO_AUTO_DA_ALLOC);
afd4672d 1798 else
fd8c37ec 1799 set_opt(sb,NO_AUTO_DA_ALLOC);
afd4672d 1800 break;
5328e635 1801 case Opt_discard:
fd8c37ec 1802 set_opt(sb, DISCARD);
5328e635
ES		 1803 break;
1804 case Opt_nodiscard:
fd8c37ec 1805 clear_opt(sb, DISCARD);
5328e635 1806 break;
744692dc 1807 case Opt_dioread_nolock:
fd8c37ec 1808 set_opt(sb, DIOREAD_NOLOCK);
744692dc
JZ		 1809 break;
1810 case Opt_dioread_lock:
fd8c37ec 1811 clear_opt(sb, DIOREAD_NOLOCK);
744692dc 1812 break;
bfff6873 1813 case Opt_init_inode_table:
fd8c37ec 1814 set_opt(sb, INIT_INODE_TABLE);
bfff6873
LC		 1815 if (args[0].from) {
1816 if (match_int(&args[0], &option))
1817 return 0;
1818 } else
1819 option = EXT4_DEF_LI_WAIT_MULT;
1820 if (option < 0)
1821 return 0;
1822 sbi->s_li_wait_mult = option;
1823 break;
1824 case Opt_noinit_inode_table:
fd8c37ec 1825 clear_opt(sb, INIT_INODE_TABLE);
bfff6873 1826 break;
ac27a0ec 1827 default:
b31e1552
ES		 1828 ext4_msg(sb, KERN_ERR,
1829 "Unrecognized mount option \"%s\" "
1830 "or missing value", p);
ac27a0ec
DK		 1831 return 0;
1832 }
1833 }
1834#ifdef CONFIG_QUOTA
1835 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
482a7425 1836 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
fd8c37ec 1837 clear_opt(sb, USRQUOTA);
ac27a0ec 1838
482a7425 1839 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
fd8c37ec 1840 clear_opt(sb, GRPQUOTA);
ac27a0ec 1841
56c50f11 1842 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
b31e1552
ES		 1843 ext4_msg(sb, KERN_ERR, "old and new quota "
1844 "format mixing");
ac27a0ec
DK		 1845 return 0;
1846 }
1847
1848 if (!sbi->s_jquota_fmt) {
b31e1552
ES		 1849 ext4_msg(sb, KERN_ERR, "journaled quota format "
1850 "not specified");
ac27a0ec
DK		 1851 return 0;
1852 }
1853 } else {
1854 if (sbi->s_jquota_fmt) {
b31e1552 1855 ext4_msg(sb, KERN_ERR, "journaled quota format "
2c8be6b2 1856 "specified with no journaling "
b31e1552 1857 "enabled");
ac27a0ec
DK		 1858 return 0;
1859 }
1860 }
1861#endif
1862 return 1;
1863}
1864
617ba13b 1865static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 1866 int read_only)
1867{
617ba13b 1868 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 1869 int res = 0;
1870
617ba13b 1871 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
b31e1552
ES		 1872 ext4_msg(sb, KERN_ERR, "revision level too high, "
1873 "forcing read-only mode");
ac27a0ec
DK		 1874 res = MS_RDONLY;
1875 }
1876 if (read_only)
1877 return res;
617ba13b 1878 if (!(sbi->s_mount_state & EXT4_VALID_FS))
b31e1552
ES		 1879 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1880 "running e2fsck is recommended");
617ba13b 1881 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
b31e1552
ES		 1882 ext4_msg(sb, KERN_WARNING,
1883 "warning: mounting fs with errors, "
1884 "running e2fsck is recommended");
ac27a0ec
DK		 1885 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1886 le16_to_cpu(es->s_mnt_count) >=
1887 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
b31e1552
ES		 1888 ext4_msg(sb, KERN_WARNING,
1889 "warning: maximal mount count reached, "
1890 "running e2fsck is recommended");
ac27a0ec
DK		 1891 else if (le32_to_cpu(es->s_checkinterval) &&
1892 (le32_to_cpu(es->s_lastcheck) +
1893 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
b31e1552
ES		 1894 ext4_msg(sb, KERN_WARNING,
1895 "warning: checktime reached, "
1896 "running e2fsck is recommended");
0b8e58a1 1897 if (!sbi->s_journal)
0390131b 1898 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec 1899 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 1900 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 1901 le16_add_cpu(&es->s_mnt_count, 1);
ac27a0ec 1902 es->s_mtime = cpu_to_le32(get_seconds());
617ba13b 1903 ext4_update_dynamic_rev(sb);
0390131b
FM		 1904 if (sbi->s_journal)
1905 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 1906
e2d67052 1907 ext4_commit_super(sb, 1);
ac27a0ec 1908 if (test_opt(sb, DEBUG))
a9df9a49 1909 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
a2595b8a 1910 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
ac27a0ec
DK		 1911 sb->s_blocksize,
1912 sbi->s_groups_count,
617ba13b
MC		 1913 EXT4_BLOCKS_PER_GROUP(sb),
1914 EXT4_INODES_PER_GROUP(sb),
a2595b8a 1915 sbi->s_mount_opt, sbi->s_mount_opt2);
ac27a0ec 1916
ac27a0ec
DK		 1917 return res;
1918}
1919
772cb7c8
JS		 1920static int ext4_fill_flex_info(struct super_block *sb)
1921{
1922 struct ext4_sb_info *sbi = EXT4_SB(sb);
1923 struct ext4_group_desc *gdp = NULL;
772cb7c8
JS		 1924 ext4_group_t flex_group_count;
1925 ext4_group_t flex_group;
1926 int groups_per_flex = 0;
c5ca7c76 1927 size_t size;
772cb7c8
JS		 1928 int i;
1929
503358ae
TT		 1930 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1931 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1932
1933 if (groups_per_flex < 2) {
772cb7c8
JS		 1934 sbi->s_log_groups_per_flex = 0;
1935 return 1;
1936 }
1937
c62a11fd
FB		 1938 /* We allocate both existing and potentially added groups */
1939 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
d94e99a6
AK		 1940 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1941 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
c5ca7c76
TT		 1942 size = flex_group_count * sizeof(struct flex_groups);
1943 sbi->s_flex_groups = kzalloc(size, GFP_KERNEL);
1944 if (sbi->s_flex_groups == NULL) {
94de56ab
JP		 1945 sbi->s_flex_groups = vzalloc(size);
1946 if (sbi->s_flex_groups == NULL) {
1947 ext4_msg(sb, KERN_ERR,
1948 "not enough memory for %u flex groups",
1949 flex_group_count);
1950 goto failed;
1951 }
772cb7c8 1952 }
772cb7c8 1953
772cb7c8 1954 for (i = 0; i < sbi->s_groups_count; i++) {
88b6edd1 1955 gdp = ext4_get_group_desc(sb, i, NULL);
772cb7c8
JS		 1956
1957 flex_group = ext4_flex_group(sbi, i);
7ad9bb65
TT		 1958 atomic_add(ext4_free_inodes_count(sb, gdp),
1959 &sbi->s_flex_groups[flex_group].free_inodes);
1960 atomic_add(ext4_free_blks_count(sb, gdp),
1961 &sbi->s_flex_groups[flex_group].free_blocks);
1962 atomic_add(ext4_used_dirs_count(sb, gdp),
1963 &sbi->s_flex_groups[flex_group].used_dirs);
772cb7c8
JS		 1964 }
1965
1966 return 1;
1967failed:
1968 return 0;
1969}
1970
717d50e4
AD		 1971__le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1972 struct ext4_group_desc *gdp)
1973{
1974 __u16 crc = 0;
1975
1976 if (sbi->s_es->s_feature_ro_compat &
1977 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1978 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1979 __le32 le_group = cpu_to_le32(block_group);
1980
1981 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1982 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1983 crc = crc16(crc, (__u8 *)gdp, offset);
1984 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1985 /* for checksum of struct ext4_group_desc do the rest...*/
1986 if ((sbi->s_es->s_feature_incompat &
1987 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1988 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1989 crc = crc16(crc, (__u8 *)gdp + offset,
1990 le16_to_cpu(sbi->s_es->s_desc_size) -
1991 offset);
1992 }
1993
1994 return cpu_to_le16(crc);
1995}
1996
1997int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
1998 struct ext4_group_desc *gdp)
1999{
2000 if ((sbi->s_es->s_feature_ro_compat &
2001 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
2002 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
2003 return 0;
2004
2005 return 1;
2006}
2007
ac27a0ec 2008/* Called at mount-time, super-block is locked */
bfff6873
LC		 2009static int ext4_check_descriptors(struct super_block *sb,
2010 ext4_group_t *first_not_zeroed)
ac27a0ec 2011{
617ba13b
MC		 2012 struct ext4_sb_info *sbi = EXT4_SB(sb);
2013 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2014 ext4_fsblk_t last_block;
bd81d8ee
LV		 2015 ext4_fsblk_t block_bitmap;
2016 ext4_fsblk_t inode_bitmap;
2017 ext4_fsblk_t inode_table;
ce421581 2018 int flexbg_flag = 0;
bfff6873 2019 ext4_group_t i, grp = sbi->s_groups_count;
ac27a0ec 2020
ce421581
JS		 2021 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2022 flexbg_flag = 1;
2023
af5bc92d 2024 ext4_debug("Checking group descriptors");
ac27a0ec 2025
197cd65a
AM		 2026 for (i = 0; i < sbi->s_groups_count; i++) {
2027 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2028
ce421581 2029 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 2030 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 2031 else
2032 last_block = first_block +
617ba13b 2033 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 2034
bfff6873
LC		 2035 if ((grp == sbi->s_groups_count) &&
2036 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2037 grp = i;
2038
8fadc143 2039 block_bitmap = ext4_block_bitmap(sb, gdp);
2b2d6d01 2040 if (block_bitmap < first_block || block_bitmap > last_block) {
b31e1552 2041 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2042 "Block bitmap for group %u not in group "
b31e1552 2043 "(block %llu)!", i, block_bitmap);
ac27a0ec
DK		 2044 return 0;
2045 }
8fadc143 2046 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2b2d6d01 2047 if (inode_bitmap < first_block || inode_bitmap > last_block) {
b31e1552 2048 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2049 "Inode bitmap for group %u not in group "
b31e1552 2050 "(block %llu)!", i, inode_bitmap);
ac27a0ec
DK		 2051 return 0;
2052 }
8fadc143 2053 inode_table = ext4_inode_table(sb, gdp);
bd81d8ee 2054 if (inode_table < first_block ||
2b2d6d01 2055 inode_table + sbi->s_itb_per_group - 1 > last_block) {
b31e1552 2056 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2057 "Inode table for group %u not in group "
b31e1552 2058 "(block %llu)!", i, inode_table);
ac27a0ec
DK		 2059 return 0;
2060 }
955ce5f5 2061 ext4_lock_group(sb, i);
717d50e4 2062 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
b31e1552
ES		 2063 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2064 "Checksum for group %u failed (%u!=%u)",
2065 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2066 gdp)), le16_to_cpu(gdp->bg_checksum));
7ee1ec4c 2067 if (!(sb->s_flags & MS_RDONLY)) {
955ce5f5 2068 ext4_unlock_group(sb, i);
8a266467 2069 return 0;
7ee1ec4c 2070 }
717d50e4 2071 }
955ce5f5 2072 ext4_unlock_group(sb, i);
ce421581
JS		 2073 if (!flexbg_flag)
2074 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec 2075 }
bfff6873
LC		 2076 if (NULL != first_not_zeroed)
2077 *first_not_zeroed = grp;
ac27a0ec 2078
bd81d8ee 2079 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
0b8e58a1 2080 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
ac27a0ec
DK		 2081 return 1;
2082}
2083
617ba13b 2084/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 2085 * the superblock) which were deleted from all directories, but held open by
2086 * a process at the time of a crash. We walk the list and try to delete these
2087 * inodes at recovery time (only with a read-write filesystem).
2088 *
2089 * In order to keep the orphan inode chain consistent during traversal (in
2090 * case of crash during recovery), we link each inode into the superblock
2091 * orphan list_head and handle it the same way as an inode deletion during
2092 * normal operation (which journals the operations for us).
2093 *
2094 * We only do an iget() and an iput() on each inode, which is very safe if we
2095 * accidentally point at an in-use or already deleted inode. The worst that
2096 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 2097 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 2098 * e2fsck was run on this filesystem, and it must have already done the orphan
2099 * inode cleanup for us, so we can safely abort without any further action.
2100 */
2b2d6d01
TT		 2101static void ext4_orphan_cleanup(struct super_block *sb,
2102 struct ext4_super_block *es)
ac27a0ec
DK		 2103{
2104 unsigned int s_flags = sb->s_flags;
2105 int nr_orphans = 0, nr_truncates = 0;
2106#ifdef CONFIG_QUOTA
2107 int i;
2108#endif
2109 if (!es->s_last_orphan) {
2110 jbd_debug(4, "no orphan inodes to clean up\n");
2111 return;
2112 }
2113
a8f48a95 2114 if (bdev_read_only(sb->s_bdev)) {
b31e1552
ES		 2115 ext4_msg(sb, KERN_ERR, "write access "
2116 "unavailable, skipping orphan cleanup");
a8f48a95
ES		 2117 return;
2118 }
2119
617ba13b 2120 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
ac27a0ec
DK		 2121 if (es->s_last_orphan)
2122 jbd_debug(1, "Errors on filesystem, "
2123 "clearing orphan list.\n");
2124 es->s_last_orphan = 0;
2125 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2126 return;
2127 }
2128
2129 if (s_flags & MS_RDONLY) {
b31e1552 2130 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
ac27a0ec
DK		 2131 sb->s_flags &= ~MS_RDONLY;
2132 }
2133#ifdef CONFIG_QUOTA
2134 /* Needed for iput() to work correctly and not trash data */
2135 sb->s_flags |= MS_ACTIVE;
2136 /* Turn on quotas so that they are updated correctly */
2137 for (i = 0; i < MAXQUOTAS; i++) {
617ba13b
MC		 2138 if (EXT4_SB(sb)->s_qf_names[i]) {
2139 int ret = ext4_quota_on_mount(sb, i);
ac27a0ec 2140 if (ret < 0)
b31e1552
ES		 2141 ext4_msg(sb, KERN_ERR,
2142 "Cannot turn on journaled "
2143 "quota: error %d", ret);
ac27a0ec
DK		 2144 }
2145 }
2146#endif
2147
2148 while (es->s_last_orphan) {
2149 struct inode *inode;
2150
97bd42b9
JB		 2151 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2152 if (IS_ERR(inode)) {
ac27a0ec
DK		 2153 es->s_last_orphan = 0;
2154 break;
2155 }
2156
617ba13b 2157 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
871a2931 2158 dquot_initialize(inode);
ac27a0ec 2159 if (inode->i_nlink) {
b31e1552
ES		 2160 ext4_msg(sb, KERN_DEBUG,
2161 "%s: truncating inode %lu to %lld bytes",
46e665e9 2162 __func__, inode->i_ino, inode->i_size);
e5f8eab8 2163 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
ac27a0ec 2164 inode->i_ino, inode->i_size);
617ba13b 2165 ext4_truncate(inode);
ac27a0ec
DK		 2166 nr_truncates++;
2167 } else {
b31e1552
ES		 2168 ext4_msg(sb, KERN_DEBUG,
2169 "%s: deleting unreferenced inode %lu",
46e665e9 2170 __func__, inode->i_ino);
ac27a0ec
DK		 2171 jbd_debug(2, "deleting unreferenced inode %lu\n",
2172 inode->i_ino);
2173 nr_orphans++;
2174 }
2175 iput(inode); /* The delete magic happens here! */
2176 }
2177
2b2d6d01 2178#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
ac27a0ec
DK		 2179
2180 if (nr_orphans)
b31e1552
ES		 2181 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2182 PLURAL(nr_orphans));
ac27a0ec 2183 if (nr_truncates)
b31e1552
ES		 2184 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2185 PLURAL(nr_truncates));
ac27a0ec
DK		 2186#ifdef CONFIG_QUOTA
2187 /* Turn quotas off */
2188 for (i = 0; i < MAXQUOTAS; i++) {
2189 if (sb_dqopt(sb)->files[i])
287a8095 2190 dquot_quota_off(sb, i);
ac27a0ec
DK		 2191 }
2192#endif
2193 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2194}
0b8e58a1 2195
cd2291a4
ES		 2196/*
2197 * Maximal extent format file size.
2198 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2199 * extent format containers, within a sector_t, and within i_blocks
2200 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2201 * so that won't be a limiting factor.
2202 *
2203 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2204 */
f287a1a5 2205static loff_t ext4_max_size(int blkbits, int has_huge_files)
cd2291a4
ES		 2206{
2207 loff_t res;
2208 loff_t upper_limit = MAX_LFS_FILESIZE;
2209
2210 /* small i_blocks in vfs inode? */
f287a1a5 2211 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
cd2291a4 2212 /*
90c699a9 2213 * CONFIG_LBDAF is not enabled implies the inode
cd2291a4
ES		 2214 * i_block represent total blocks in 512 bytes
2215 * 32 == size of vfs inode i_blocks * 8
2216 */
2217 upper_limit = (1LL << 32) - 1;
2218
2219 /* total blocks in file system block size */
2220 upper_limit >>= (blkbits - 9);
2221 upper_limit <<= blkbits;
2222 }
2223
2224 /* 32-bit extent-start container, ee_block */
2225 res = 1LL << 32;
2226 res <<= blkbits;
2227 res -= 1;
2228
2229 /* Sanity check against vm- & vfs- imposed limits */
2230 if (res > upper_limit)
2231 res = upper_limit;
2232
2233 return res;
2234}
ac27a0ec 2235
ac27a0ec 2236/*
cd2291a4 2237 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 2238 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2239 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 2240 */
f287a1a5 2241static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
ac27a0ec 2242{
617ba13b 2243 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 2244 int meta_blocks;
2245 loff_t upper_limit;
0b8e58a1
AD		 2246 /* This is calculated to be the largest file size for a dense, block
2247 * mapped file such that the file's total number of 512-byte sectors,
2248 * including data and all indirect blocks, does not exceed (2^48 - 1).
2249 *
2250 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2251 * number of 512-byte sectors of the file.
0fc1b451
AK		 2252 */
2253
f287a1a5 2254 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
0fc1b451 2255 /*
90c699a9 2256 * !has_huge_files or CONFIG_LBDAF not enabled implies that
0b8e58a1
AD		 2257 * the inode i_block field represents total file blocks in
2258 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
0fc1b451
AK		 2259 */
2260 upper_limit = (1LL << 32) - 1;
2261
2262 /* total blocks in file system block size */
2263 upper_limit >>= (bits - 9);
2264
2265 } else {
8180a562
AK		 2266 /*
2267 * We use 48 bit ext4_inode i_blocks
2268 * With EXT4_HUGE_FILE_FL set the i_blocks
2269 * represent total number of blocks in
2270 * file system block size
2271 */
0fc1b451
AK		 2272 upper_limit = (1LL << 48) - 1;
2273
0fc1b451
AK		 2274 }
2275
2276 /* indirect blocks */
2277 meta_blocks = 1;
2278 /* double indirect blocks */
2279 meta_blocks += 1 + (1LL << (bits-2));
2280 /* tripple indirect blocks */
2281 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2282
2283 upper_limit -= meta_blocks;
2284 upper_limit <<= bits;
ac27a0ec
DK		 2285
2286 res += 1LL << (bits-2);
2287 res += 1LL << (2*(bits-2));
2288 res += 1LL << (3*(bits-2));
2289 res <<= bits;
2290 if (res > upper_limit)
2291 res = upper_limit;
0fc1b451
AK		 2292
2293 if (res > MAX_LFS_FILESIZE)
2294 res = MAX_LFS_FILESIZE;
2295
ac27a0ec
DK		 2296 return res;
2297}
2298
617ba13b 2299static ext4_fsblk_t descriptor_loc(struct super_block *sb,
0b8e58a1 2300 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 2301{
617ba13b 2302 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 2303 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 2304 int has_super = 0;
2305
2306 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2307
617ba13b 2308 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
ac27a0ec 2309 nr < first_meta_bg)
70bbb3e0 2310 return logical_sb_block + nr + 1;
ac27a0ec 2311 bg = sbi->s_desc_per_block * nr;
617ba13b 2312 if (ext4_bg_has_super(sb, bg))
ac27a0ec 2313 has_super = 1;
0b8e58a1 2314
617ba13b 2315 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 2316}
2317
c9de560d
AT		 2318/**
2319 * ext4_get_stripe_size: Get the stripe size.
2320 * @sbi: In memory super block info
2321 *
2322 * If we have specified it via mount option, then
2323 * use the mount option value. If the value specified at mount time is
2324 * greater than the blocks per group use the super block value.
2325 * If the super block value is greater than blocks per group return 0.
2326 * Allocator needs it be less than blocks per group.
2327 *
2328 */
2329static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2330{
2331 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2332 unsigned long stripe_width =
2333 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2334
2335 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2336 return sbi->s_stripe;
2337
2338 if (stripe_width <= sbi->s_blocks_per_group)
2339 return stripe_width;
2340
2341 if (stride <= sbi->s_blocks_per_group)
2342 return stride;
2343
2344 return 0;
2345}
ac27a0ec 2346
3197ebdb
TT		 2347/* sysfs supprt */
2348
2349struct ext4_attr {
2350 struct attribute attr;
2351 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
60e6679e 2352 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
3197ebdb
TT		 2353 const char *, size_t);
2354 int offset;
2355};
2356
2357static int parse_strtoul(const char *buf,
2358 unsigned long max, unsigned long *value)
2359{
2360 char *endp;
2361
e7d2860b
AGR		 2362 *value = simple_strtoul(skip_spaces(buf), &endp, 0);
2363 endp = skip_spaces(endp);
3197ebdb
TT		 2364 if (*endp || *value > max)
2365 return -EINVAL;
2366
2367 return 0;
2368}
2369
2370static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2371 struct ext4_sb_info *sbi,
2372 char *buf)
2373{
2374 return snprintf(buf, PAGE_SIZE, "%llu\n",
2375 (s64) percpu_counter_sum(&sbi->s_dirtyblocks_counter));
2376}
2377
2378static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2379 struct ext4_sb_info *sbi, char *buf)
2380{
2381 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2382
f613dfcb
TT		 2383 if (!sb->s_bdev->bd_part)
2384 return snprintf(buf, PAGE_SIZE, "0\n");
3197ebdb
TT		 2385 return snprintf(buf, PAGE_SIZE, "%lu\n",
2386 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2387 sbi->s_sectors_written_start) >> 1);
2388}
2389
2390static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2391 struct ext4_sb_info *sbi, char *buf)
2392{
2393 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2394
f613dfcb
TT		 2395 if (!sb->s_bdev->bd_part)
2396 return snprintf(buf, PAGE_SIZE, "0\n");
3197ebdb 2397 return snprintf(buf, PAGE_SIZE, "%llu\n",
a6b43e38 2398 (unsigned long long)(sbi->s_kbytes_written +
3197ebdb 2399 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
a6b43e38 2400 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
3197ebdb
TT		 2401}
2402
2403static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2404 struct ext4_sb_info *sbi,
2405 const char *buf, size_t count)
2406{
2407 unsigned long t;
2408
2409 if (parse_strtoul(buf, 0x40000000, &t))
2410 return -EINVAL;
2411
5dbd571d 2412 if (t && !is_power_of_2(t))
3197ebdb
TT		 2413 return -EINVAL;
2414
2415 sbi->s_inode_readahead_blks = t;
2416 return count;
2417}
2418
2419static ssize_t sbi_ui_show(struct ext4_attr *a,
0b8e58a1 2420 struct ext4_sb_info *sbi, char *buf)
3197ebdb
TT		 2421{
2422 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2423
2424 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2425}
2426
2427static ssize_t sbi_ui_store(struct ext4_attr *a,
2428 struct ext4_sb_info *sbi,
2429 const char *buf, size_t count)
2430{
2431 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2432 unsigned long t;
2433
2434 if (parse_strtoul(buf, 0xffffffff, &t))
2435 return -EINVAL;
2436 *ui = t;
2437 return count;
2438}
2439
2440#define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2441static struct ext4_attr ext4_attr_##_name = { \
2442 .attr = {.name = __stringify(_name), .mode = _mode }, \
2443 .show = _show, \
2444 .store = _store, \
2445 .offset = offsetof(struct ext4_sb_info, _elname), \
2446}
2447#define EXT4_ATTR(name, mode, show, store) \
2448static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2449
857ac889 2450#define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
3197ebdb
TT		 2451#define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2452#define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2453#define EXT4_RW_ATTR_SBI_UI(name, elname) \
2454 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2455#define ATTR_LIST(name) &ext4_attr_##name.attr
2456
2457EXT4_RO_ATTR(delayed_allocation_blocks);
2458EXT4_RO_ATTR(session_write_kbytes);
2459EXT4_RO_ATTR(lifetime_write_kbytes);
2460EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2461 inode_readahead_blks_store, s_inode_readahead_blks);
11013911 2462EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
3197ebdb
TT		 2463EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2464EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2465EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2466EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2467EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2468EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
55138e0b 2469EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
3197ebdb
TT		 2470
2471static struct attribute *ext4_attrs[] = {
2472 ATTR_LIST(delayed_allocation_blocks),
2473 ATTR_LIST(session_write_kbytes),
2474 ATTR_LIST(lifetime_write_kbytes),
2475 ATTR_LIST(inode_readahead_blks),
11013911 2476 ATTR_LIST(inode_goal),
3197ebdb
TT		 2477 ATTR_LIST(mb_stats),
2478 ATTR_LIST(mb_max_to_scan),
2479 ATTR_LIST(mb_min_to_scan),
2480 ATTR_LIST(mb_order2_req),
2481 ATTR_LIST(mb_stream_req),
2482 ATTR_LIST(mb_group_prealloc),
55138e0b 2483 ATTR_LIST(max_writeback_mb_bump),
3197ebdb
TT		 2484 NULL,
2485};
2486
857ac889
LC		 2487/* Features this copy of ext4 supports */
2488EXT4_INFO_ATTR(lazy_itable_init);
27ee40df 2489EXT4_INFO_ATTR(batched_discard);
857ac889
LC		 2490
2491static struct attribute *ext4_feat_attrs[] = {
2492 ATTR_LIST(lazy_itable_init),
27ee40df 2493 ATTR_LIST(batched_discard),
857ac889
LC		 2494 NULL,
2495};
2496
3197ebdb
TT		 2497static ssize_t ext4_attr_show(struct kobject *kobj,
2498 struct attribute *attr, char *buf)
2499{
2500 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2501 s_kobj);
2502 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2503
2504 return a->show ? a->show(a, sbi, buf) : 0;
2505}
2506
2507static ssize_t ext4_attr_store(struct kobject *kobj,
2508 struct attribute *attr,
2509 const char *buf, size_t len)
2510{
2511 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2512 s_kobj);
2513 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2514
2515 return a->store ? a->store(a, sbi, buf, len) : 0;
2516}
2517
2518static void ext4_sb_release(struct kobject *kobj)
2519{
2520 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2521 s_kobj);
2522 complete(&sbi->s_kobj_unregister);
2523}
2524
52cf25d0 2525static const struct sysfs_ops ext4_attr_ops = {
3197ebdb
TT		 2526 .show = ext4_attr_show,
2527 .store = ext4_attr_store,
2528};
2529
2530static struct kobj_type ext4_ktype = {
2531 .default_attrs = ext4_attrs,
2532 .sysfs_ops = &ext4_attr_ops,
2533 .release = ext4_sb_release,
2534};
2535
857ac889
LC		 2536static void ext4_feat_release(struct kobject *kobj)
2537{
2538 complete(&ext4_feat->f_kobj_unregister);
2539}
2540
2541static struct kobj_type ext4_feat_ktype = {
2542 .default_attrs = ext4_feat_attrs,
2543 .sysfs_ops = &ext4_attr_ops,
2544 .release = ext4_feat_release,
2545};
2546
a13fb1a4
ES		 2547/*
2548 * Check whether this filesystem can be mounted based on
2549 * the features present and the RDONLY/RDWR mount requested.
2550 * Returns 1 if this filesystem can be mounted as requested,
2551 * 0 if it cannot be.
2552 */
2553static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2554{
2555 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2556 ext4_msg(sb, KERN_ERR,
2557 "Couldn't mount because of "
2558 "unsupported optional features (%x)",
2559 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2560 ~EXT4_FEATURE_INCOMPAT_SUPP));
2561 return 0;
2562 }
2563
2564 if (readonly)
2565 return 1;
2566
2567 /* Check that feature set is OK for a read-write mount */
2568 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2569 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2570 "unsupported optional features (%x)",
2571 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2572 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2573 return 0;
2574 }
2575 /*
2576 * Large file size enabled file system can only be mounted
2577 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2578 */
2579 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2580 if (sizeof(blkcnt_t) < sizeof(u64)) {
2581 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2582 "cannot be mounted RDWR without "
2583 "CONFIG_LBDAF");
2584 return 0;
2585 }
2586 }
2587 return 1;
2588}
2589
66e61a9e
TT		 2590/*
2591 * This function is called once a day if we have errors logged
2592 * on the file system
2593 */
2594static void print_daily_error_info(unsigned long arg)
2595{
2596 struct super_block *sb = (struct super_block *) arg;
2597 struct ext4_sb_info *sbi;
2598 struct ext4_super_block *es;
2599
2600 sbi = EXT4_SB(sb);
2601 es = sbi->s_es;
2602
2603 if (es->s_error_count)
2604 ext4_msg(sb, KERN_NOTICE, "error count: %u",
2605 le32_to_cpu(es->s_error_count));
2606 if (es->s_first_error_time) {
2607 printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d",
2608 sb->s_id, le32_to_cpu(es->s_first_error_time),
2609 (int) sizeof(es->s_first_error_func),
2610 es->s_first_error_func,
2611 le32_to_cpu(es->s_first_error_line));
2612 if (es->s_first_error_ino)
2613 printk(": inode %u",
2614 le32_to_cpu(es->s_first_error_ino));
2615 if (es->s_first_error_block)
2616 printk(": block %llu", (unsigned long long)
2617 le64_to_cpu(es->s_first_error_block));
2618 printk("\n");
2619 }
2620 if (es->s_last_error_time) {
2621 printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d",
2622 sb->s_id, le32_to_cpu(es->s_last_error_time),
2623 (int) sizeof(es->s_last_error_func),
2624 es->s_last_error_func,
2625 le32_to_cpu(es->s_last_error_line));
2626 if (es->s_last_error_ino)
2627 printk(": inode %u",
2628 le32_to_cpu(es->s_last_error_ino));
2629 if (es->s_last_error_block)
2630 printk(": block %llu", (unsigned long long)
2631 le64_to_cpu(es->s_last_error_block));
2632 printk("\n");
2633 }
2634 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2635}
2636
bfff6873
LC		 2637static void ext4_lazyinode_timeout(unsigned long data)
2638{
2639 struct task_struct *p = (struct task_struct *)data;
2640 wake_up_process(p);
2641}
2642
2643/* Find next suitable group and run ext4_init_inode_table */
2644static int ext4_run_li_request(struct ext4_li_request *elr)
2645{
2646 struct ext4_group_desc *gdp = NULL;
2647 ext4_group_t group, ngroups;
2648 struct super_block *sb;
2649 unsigned long timeout = 0;
2650 int ret = 0;
2651
2652 sb = elr->lr_super;
2653 ngroups = EXT4_SB(sb)->s_groups_count;
2654
2655 for (group = elr->lr_next_group; group < ngroups; group++) {
2656 gdp = ext4_get_group_desc(sb, group, NULL);
2657 if (!gdp) {
2658 ret = 1;
2659 break;
2660 }
2661
2662 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2663 break;
2664 }
2665
2666 if (group == ngroups)
2667 ret = 1;
2668
2669 if (!ret) {
2670 timeout = jiffies;
2671 ret = ext4_init_inode_table(sb, group,
2672 elr->lr_timeout ? 0 : 1);
2673 if (elr->lr_timeout == 0) {
2674 timeout = jiffies - timeout;
2675 if (elr->lr_sbi->s_li_wait_mult)
2676 timeout *= elr->lr_sbi->s_li_wait_mult;
2677 else
2678 timeout *= 20;
2679 elr->lr_timeout = timeout;
2680 }
2681 elr->lr_next_sched = jiffies + elr->lr_timeout;
2682 elr->lr_next_group = group + 1;
2683 }
2684
2685 return ret;
2686}
2687
2688/*
2689 * Remove lr_request from the list_request and free the
2690 * request tructure. Should be called with li_list_mtx held
2691 */
2692static void ext4_remove_li_request(struct ext4_li_request *elr)
2693{
2694 struct ext4_sb_info *sbi;
2695
2696 if (!elr)
2697 return;
2698
2699 sbi = elr->lr_sbi;
2700
2701 list_del(&elr->lr_request);
2702 sbi->s_li_request = NULL;
2703 kfree(elr);
2704}
2705
2706static void ext4_unregister_li_request(struct super_block *sb)
2707{
2708 struct ext4_li_request *elr = EXT4_SB(sb)->s_li_request;
2709
2710 if (!ext4_li_info)
2711 return;
2712
2713 mutex_lock(&ext4_li_info->li_list_mtx);
2714 ext4_remove_li_request(elr);
2715 mutex_unlock(&ext4_li_info->li_list_mtx);
2716}
2717
8f1f7453
ES		 2718static struct task_struct *ext4_lazyinit_task;
2719
bfff6873
LC		 2720/*
2721 * This is the function where ext4lazyinit thread lives. It walks
2722 * through the request list searching for next scheduled filesystem.
2723 * When such a fs is found, run the lazy initialization request
2724 * (ext4_rn_li_request) and keep track of the time spend in this
2725 * function. Based on that time we compute next schedule time of
2726 * the request. When walking through the list is complete, compute
2727 * next waking time and put itself into sleep.
2728 */
2729static int ext4_lazyinit_thread(void *arg)
2730{
2731 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2732 struct list_head *pos, *n;
2733 struct ext4_li_request *elr;
2734 unsigned long next_wakeup;
2735 DEFINE_WAIT(wait);
bfff6873
LC		 2736
2737 BUG_ON(NULL == eli);
2738
2739 eli->li_timer.data = (unsigned long)current;
2740 eli->li_timer.function = ext4_lazyinode_timeout;
2741
2742 eli->li_task = current;
2743 wake_up(&eli->li_wait_task);
2744
2745cont_thread:
2746 while (true) {
2747 next_wakeup = MAX_JIFFY_OFFSET;
2748
2749 mutex_lock(&eli->li_list_mtx);
2750 if (list_empty(&eli->li_request_list)) {
2751 mutex_unlock(&eli->li_list_mtx);
2752 goto exit_thread;
2753 }
2754
2755 list_for_each_safe(pos, n, &eli->li_request_list) {
2756 elr = list_entry(pos, struct ext4_li_request,
2757 lr_request);
2758
b2c78cd0
TT		 2759 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2760 if (ext4_run_li_request(elr) != 0) {
2761 /* error, remove the lazy_init job */
2762 ext4_remove_li_request(elr);
2763 continue;
2764 }
bfff6873
LC		 2765 }
2766
2767 if (time_before(elr->lr_next_sched, next_wakeup))
2768 next_wakeup = elr->lr_next_sched;
2769 }
2770 mutex_unlock(&eli->li_list_mtx);
2771
2772 if (freezing(current))
2773 refrigerator();
2774
f4245bd4
LC		 2775 if ((time_after_eq(jiffies, next_wakeup)) ||
2776 (MAX_JIFFY_OFFSET == next_wakeup)) {
bfff6873
LC		 2777 cond_resched();
2778 continue;
2779 }
2780
2781 eli->li_timer.expires = next_wakeup;
2782 add_timer(&eli->li_timer);
2783 prepare_to_wait(&eli->li_wait_daemon, &wait,
2784 TASK_INTERRUPTIBLE);
2785 if (time_before(jiffies, next_wakeup))
2786 schedule();
2787 finish_wait(&eli->li_wait_daemon, &wait);
8f1f7453
ES		 2788 if (kthread_should_stop()) {
2789 ext4_clear_request_list();
2790 goto exit_thread;
2791 }
bfff6873
LC		 2792 }
2793
2794exit_thread:
2795 /*
2796 * It looks like the request list is empty, but we need
2797 * to check it under the li_list_mtx lock, to prevent any
2798 * additions into it, and of course we should lock ext4_li_mtx
2799 * to atomically free the list and ext4_li_info, because at
2800 * this point another ext4 filesystem could be registering
2801 * new one.
2802 */
2803 mutex_lock(&ext4_li_mtx);
2804 mutex_lock(&eli->li_list_mtx);
2805 if (!list_empty(&eli->li_request_list)) {
2806 mutex_unlock(&eli->li_list_mtx);
2807 mutex_unlock(&ext4_li_mtx);
2808 goto cont_thread;
2809 }
2810 mutex_unlock(&eli->li_list_mtx);
2811 del_timer_sync(&ext4_li_info->li_timer);
2812 eli->li_task = NULL;
2813 wake_up(&eli->li_wait_task);
2814
2815 kfree(ext4_li_info);
8f1f7453 2816 ext4_lazyinit_task = NULL;
bfff6873
LC		 2817 ext4_li_info = NULL;
2818 mutex_unlock(&ext4_li_mtx);
2819
2820 return 0;
2821}
2822
2823static void ext4_clear_request_list(void)
2824{
2825 struct list_head *pos, *n;
2826 struct ext4_li_request *elr;
2827
2828 mutex_lock(&ext4_li_info->li_list_mtx);
bfff6873
LC		 2829 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2830 elr = list_entry(pos, struct ext4_li_request,
2831 lr_request);
2832 ext4_remove_li_request(elr);
2833 }
2834 mutex_unlock(&ext4_li_info->li_list_mtx);
2835}
2836
2837static int ext4_run_lazyinit_thread(void)
2838{
8f1f7453
ES		 2839 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2840 ext4_li_info, "ext4lazyinit");
2841 if (IS_ERR(ext4_lazyinit_task)) {
2842 int err = PTR_ERR(ext4_lazyinit_task);
bfff6873
LC		 2843 ext4_clear_request_list();
2844 del_timer_sync(&ext4_li_info->li_timer);
2845 kfree(ext4_li_info);
2846 ext4_li_info = NULL;
2847 printk(KERN_CRIT "EXT4: error %d creating inode table "
2848 "initialization thread\n",
2849 err);
2850 return err;
2851 }
2852 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
2853
2854 wait_event(ext4_li_info->li_wait_task, ext4_li_info->li_task != NULL);
2855 return 0;
2856}
2857
2858/*
2859 * Check whether it make sense to run itable init. thread or not.
2860 * If there is at least one uninitialized inode table, return
2861 * corresponding group number, else the loop goes through all
2862 * groups and return total number of groups.
2863 */
2864static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2865{
2866 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2867 struct ext4_group_desc *gdp = NULL;
2868
2869 for (group = 0; group < ngroups; group++) {
2870 gdp = ext4_get_group_desc(sb, group, NULL);
2871 if (!gdp)
2872 continue;
2873
2874 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2875 break;
2876 }
2877
2878 return group;
2879}
2880
2881static int ext4_li_info_new(void)
2882{
2883 struct ext4_lazy_init *eli = NULL;
2884
2885 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2886 if (!eli)
2887 return -ENOMEM;
2888
2889 eli->li_task = NULL;
2890 INIT_LIST_HEAD(&eli->li_request_list);
2891 mutex_init(&eli->li_list_mtx);
2892
2893 init_waitqueue_head(&eli->li_wait_daemon);
2894 init_waitqueue_head(&eli->li_wait_task);
2895 init_timer(&eli->li_timer);
2896 eli->li_state |= EXT4_LAZYINIT_QUIT;
2897
2898 ext4_li_info = eli;
2899
2900 return 0;
2901}
2902
2903static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
2904 ext4_group_t start)
2905{
2906 struct ext4_sb_info *sbi = EXT4_SB(sb);
2907 struct ext4_li_request *elr;
2908 unsigned long rnd;
2909
2910 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
2911 if (!elr)
2912 return NULL;
2913
2914 elr->lr_super = sb;
2915 elr->lr_sbi = sbi;
2916 elr->lr_next_group = start;
2917
2918 /*
2919 * Randomize first schedule time of the request to
2920 * spread the inode table initialization requests
2921 * better.
2922 */
2923 get_random_bytes(&rnd, sizeof(rnd));
2924 elr->lr_next_sched = jiffies + (unsigned long)rnd %
2925 (EXT4_DEF_LI_MAX_START_DELAY * HZ);
2926
2927 return elr;
2928}
2929
2930static int ext4_register_li_request(struct super_block *sb,
2931 ext4_group_t first_not_zeroed)
2932{
2933 struct ext4_sb_info *sbi = EXT4_SB(sb);
2934 struct ext4_li_request *elr;
2935 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
6c5a6cb9 2936 int ret = 0;
bfff6873
LC		 2937
2938 if (sbi->s_li_request != NULL)
beed5ecb 2939 return 0;
bfff6873
LC		 2940
2941 if (first_not_zeroed == ngroups ||
2942 (sb->s_flags & MS_RDONLY) ||
2943 !test_opt(sb, INIT_INODE_TABLE)) {
2944 sbi->s_li_request = NULL;
beed5ecb 2945 return 0;
bfff6873
LC		 2946 }
2947
2948 if (first_not_zeroed == ngroups) {
2949 sbi->s_li_request = NULL;
beed5ecb 2950 return 0;
bfff6873
LC		 2951 }
2952
2953 elr = ext4_li_request_new(sb, first_not_zeroed);
beed5ecb
NK		 2954 if (!elr)
2955 return -ENOMEM;
bfff6873
LC		 2956
2957 mutex_lock(&ext4_li_mtx);
2958
2959 if (NULL == ext4_li_info) {
2960 ret = ext4_li_info_new();
2961 if (ret)
2962 goto out;
2963 }
2964
2965 mutex_lock(&ext4_li_info->li_list_mtx);
2966 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
2967 mutex_unlock(&ext4_li_info->li_list_mtx);
2968
2969 sbi->s_li_request = elr;
2970
2971 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
2972 ret = ext4_run_lazyinit_thread();
2973 if (ret)
2974 goto out;
2975 }
bfff6873 2976out:
beed5ecb
NK		 2977 mutex_unlock(&ext4_li_mtx);
2978 if (ret)
bfff6873 2979 kfree(elr);
bfff6873
LC		 2980 return ret;
2981}
2982
2983/*
2984 * We do not need to lock anything since this is called on
2985 * module unload.
2986 */
2987static void ext4_destroy_lazyinit_thread(void)
2988{
2989 /*
2990 * If thread exited earlier
2991 * there's nothing to be done.
2992 */
8f1f7453 2993 if (!ext4_li_info || !ext4_lazyinit_task)
bfff6873
LC		 2994 return;
2995
8f1f7453 2996 kthread_stop(ext4_lazyinit_task);
bfff6873
LC		 2997}
2998
2b2d6d01 2999static int ext4_fill_super(struct super_block *sb, void *data, int silent)
7477827f
AK		 3000 __releases(kernel_lock)
3001 __acquires(kernel_lock)
ac27a0ec 3002{
d4c402d9 3003 char *orig_data = kstrdup(data, GFP_KERNEL);
2b2d6d01 3004 struct buffer_head *bh;
617ba13b
MC		 3005 struct ext4_super_block *es = NULL;
3006 struct ext4_sb_info *sbi;
3007 ext4_fsblk_t block;
3008 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 3009 ext4_fsblk_t logical_sb_block;
ac27a0ec 3010 unsigned long offset = 0;
ac27a0ec
DK		 3011 unsigned long journal_devnum = 0;
3012 unsigned long def_mount_opts;
3013 struct inode *root;
9f6200bb 3014 char *cp;
0390131b 3015 const char *descr;
dcc7dae3 3016 int ret = -ENOMEM;
ac27a0ec 3017 int blocksize;
4ec11028
TT		 3018 unsigned int db_count;
3019 unsigned int i;
f287a1a5 3020 int needs_recovery, has_huge_files;
bd81d8ee 3021 __u64 blocks_count;
833f4077 3022 int err;
b3881f74 3023 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
bfff6873 3024 ext4_group_t first_not_zeroed;
ac27a0ec
DK		 3025
3026 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3027 if (!sbi)
dcc7dae3 3028 goto out_free_orig;
705895b6
PE		 3029
3030 sbi->s_blockgroup_lock =
3031 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3032 if (!sbi->s_blockgroup_lock) {
3033 kfree(sbi);
dcc7dae3 3034 goto out_free_orig;
705895b6 3035 }
ac27a0ec
DK		 3036 sb->s_fs_info = sbi;
3037 sbi->s_mount_opt = 0;
617ba13b
MC		 3038 sbi->s_resuid = EXT4_DEF_RESUID;
3039 sbi->s_resgid = EXT4_DEF_RESGID;
240799cd 3040 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
d9c9bef1 3041 sbi->s_sb_block = sb_block;
f613dfcb
TT		 3042 if (sb->s_bdev->bd_part)
3043 sbi->s_sectors_written_start =
3044 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
ac27a0ec 3045
9f6200bb
TT		 3046 /* Cleanup superblock name */
3047 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3048 *cp = '!';
3049
dcc7dae3 3050 ret = -EINVAL;
617ba13b 3051 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 3052 if (!blocksize) {
b31e1552 3053 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
ac27a0ec
DK		 3054 goto out_fail;
3055 }
3056
3057 /*
617ba13b 3058 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 3059 * block sizes. We need to calculate the offset from buffer start.
3060 */
617ba13b 3061 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 3062 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3063 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 3064 } else {
70bbb3e0 3065 logical_sb_block = sb_block;
ac27a0ec
DK		 3066 }
3067
70bbb3e0 3068 if (!(bh = sb_bread(sb, logical_sb_block))) {
b31e1552 3069 ext4_msg(sb, KERN_ERR, "unable to read superblock");
ac27a0ec
DK		 3070 goto out_fail;
3071 }
3072 /*
3073 * Note: s_es must be initialized as soon as possible because
617ba13b 3074 * some ext4 macro-instructions depend on its value
ac27a0ec 3075 */
617ba13b 3076 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
ac27a0ec
DK		 3077 sbi->s_es = es;
3078 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 3079 if (sb->s_magic != EXT4_SUPER_MAGIC)
3080 goto cantfind_ext4;
afc32f7e 3081 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
ac27a0ec
DK		 3082
3083 /* Set defaults before we parse the mount options */
3084 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
fd8c37ec 3085 set_opt(sb, INIT_INODE_TABLE);
617ba13b 3086 if (def_mount_opts & EXT4_DEFM_DEBUG)
fd8c37ec 3087 set_opt(sb, DEBUG);
437ca0fd
DM		 3088 if (def_mount_opts & EXT4_DEFM_BSDGROUPS) {
3089 ext4_msg(sb, KERN_WARNING, deprecated_msg, "bsdgroups",
3090 "2.6.38");
fd8c37ec 3091 set_opt(sb, GRPID);
437ca0fd 3092 }
617ba13b 3093 if (def_mount_opts & EXT4_DEFM_UID16)
fd8c37ec 3094 set_opt(sb, NO_UID32);
ea663336 3095 /* xattr user namespace & acls are now defaulted on */
03010a33 3096#ifdef CONFIG_EXT4_FS_XATTR
ea663336 3097 set_opt(sb, XATTR_USER);
2e7842b8 3098#endif
03010a33 3099#ifdef CONFIG_EXT4_FS_POSIX_ACL
ea663336 3100 set_opt(sb, POSIX_ACL);
2e7842b8 3101#endif
6fd7a467 3102 set_opt(sb, MBLK_IO_SUBMIT);
617ba13b 3103 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
fd8c37ec 3104 set_opt(sb, JOURNAL_DATA);
617ba13b 3105 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
fd8c37ec 3106 set_opt(sb, ORDERED_DATA);
617ba13b 3107 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
fd8c37ec 3108 set_opt(sb, WRITEBACK_DATA);
617ba13b
MC		 3109
3110 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
fd8c37ec 3111 set_opt(sb, ERRORS_PANIC);
bb4f397a 3112 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
fd8c37ec 3113 set_opt(sb, ERRORS_CONT);
bb4f397a 3114 else
fd8c37ec 3115 set_opt(sb, ERRORS_RO);
8b67f04a 3116 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
fd8c37ec 3117 set_opt(sb, BLOCK_VALIDITY);
8b67f04a 3118 if (def_mount_opts & EXT4_DEFM_DISCARD)
fd8c37ec 3119 set_opt(sb, DISCARD);
ac27a0ec
DK		 3120
3121 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
3122 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
30773840
TT		 3123 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3124 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3125 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
ac27a0ec 3126
8b67f04a 3127 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
fd8c37ec 3128 set_opt(sb, BARRIER);
ac27a0ec 3129
dd919b98
AK		 3130 /*
3131 * enable delayed allocation by default
3132 * Use -o nodelalloc to turn it off
3133 */
8b67f04a
TT		 3134 if (!IS_EXT3_SB(sb) &&
3135 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
fd8c37ec 3136 set_opt(sb, DELALLOC);
dd919b98 3137
8b67f04a
TT		 3138 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3139 &journal_devnum, &journal_ioprio, NULL, 0)) {
3140 ext4_msg(sb, KERN_WARNING,
3141 "failed to parse options in superblock: %s",
3142 sbi->s_es->s_mount_opts);
3143 }
b3881f74
TT		 3144 if (!parse_options((char *) data, sb, &journal_devnum,
3145 &journal_ioprio, NULL, 0))
ac27a0ec
DK		 3146 goto failed_mount;
3147
3148 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 3149 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec 3150
617ba13b
MC		 3151 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3152 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3153 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3154 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
b31e1552
ES		 3155 ext4_msg(sb, KERN_WARNING,
3156 "feature flags set on rev 0 fs, "
3157 "running e2fsck is recommended");
469108ff 3158
ac27a0ec
DK		 3159 /*
3160 * Check feature flags regardless of the revision level, since we
3161 * previously didn't change the revision level when setting the flags,
3162 * so there is a chance incompat flags are set on a rev 0 filesystem.
3163 */
a13fb1a4 3164 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
ac27a0ec 3165 goto failed_mount;
a13fb1a4 3166
ac27a0ec
DK		 3167 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3168
617ba13b
MC		 3169 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3170 blocksize > EXT4_MAX_BLOCK_SIZE) {
b31e1552
ES		 3171 ext4_msg(sb, KERN_ERR,
3172 "Unsupported filesystem blocksize %d", blocksize);
ac27a0ec
DK		 3173 goto failed_mount;
3174 }
3175
ac27a0ec 3176 if (sb->s_blocksize != blocksize) {
ce40733c
AK		 3177 /* Validate the filesystem blocksize */
3178 if (!sb_set_blocksize(sb, blocksize)) {
b31e1552 3179 ext4_msg(sb, KERN_ERR, "bad block size %d",
ce40733c 3180 blocksize);
ac27a0ec
DK		 3181 goto failed_mount;
3182 }
3183
2b2d6d01 3184 brelse(bh);
70bbb3e0
AM		 3185 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3186 offset = do_div(logical_sb_block, blocksize);
3187 bh = sb_bread(sb, logical_sb_block);
ac27a0ec 3188 if (!bh) {
b31e1552
ES		 3189 ext4_msg(sb, KERN_ERR,
3190 "Can't read superblock on 2nd try");
ac27a0ec
DK		 3191 goto failed_mount;
3192 }
617ba13b 3193 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
ac27a0ec 3194 sbi->s_es = es;
617ba13b 3195 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
b31e1552
ES		 3196 ext4_msg(sb, KERN_ERR,
3197 "Magic mismatch, very weird!");
ac27a0ec
DK		 3198 goto failed_mount;
3199 }
3200 }
3201
a13fb1a4
ES		 3202 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3203 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
f287a1a5
TT		 3204 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3205 has_huge_files);
3206 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
ac27a0ec 3207
617ba13b
MC		 3208 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3209 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3210 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
ac27a0ec
DK		 3211 } else {
3212 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3213 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
617ba13b 3214 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1330593e 3215 (!is_power_of_2(sbi->s_inode_size)) ||
ac27a0ec 3216 (sbi->s_inode_size > blocksize)) {
b31e1552
ES		 3217 ext4_msg(sb, KERN_ERR,
3218 "unsupported inode size: %d",
2b2d6d01 3219 sbi->s_inode_size);
ac27a0ec
DK		 3220 goto failed_mount;
3221 }
ef7f3835
KS		 3222 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3223 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
ac27a0ec 3224 }
0b8e58a1 3225
0d1ee42f
AR		 3226 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3227 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
8fadc143 3228 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 3229 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 3230 !is_power_of_2(sbi->s_desc_size)) {
b31e1552
ES		 3231 ext4_msg(sb, KERN_ERR,
3232 "unsupported descriptor size %lu",
0d1ee42f
AR		 3233 sbi->s_desc_size);
3234 goto failed_mount;
3235 }
3236 } else
3237 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
0b8e58a1 3238
ac27a0ec 3239 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 3240 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
b47b6f38 3241 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
617ba13b 3242 goto cantfind_ext4;
0b8e58a1 3243
617ba13b 3244 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 3245 if (sbi->s_inodes_per_block == 0)
617ba13b 3246 goto cantfind_ext4;
ac27a0ec
DK		 3247 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3248 sbi->s_inodes_per_block;
0d1ee42f 3249 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 3250 sbi->s_sbh = bh;
3251 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 3252 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3253 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
0b8e58a1 3254
2b2d6d01 3255 for (i = 0; i < 4; i++)
ac27a0ec
DK		 3256 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3257 sbi->s_def_hash_version = es->s_def_hash_version;
f99b2589
TT		 3258 i = le32_to_cpu(es->s_flags);
3259 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3260 sbi->s_hash_unsigned = 3;
3261 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3262#ifdef __CHAR_UNSIGNED__
3263 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3264 sbi->s_hash_unsigned = 3;
3265#else
3266 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3267#endif
3268 sb->s_dirt = 1;
3269 }
ac27a0ec
DK		 3270
3271 if (sbi->s_blocks_per_group > blocksize * 8) {
b31e1552
ES		 3272 ext4_msg(sb, KERN_ERR,
3273 "#blocks per group too big: %lu",
2b2d6d01 3274 sbi->s_blocks_per_group);
ac27a0ec
DK		 3275 goto failed_mount;
3276 }
ac27a0ec 3277 if (sbi->s_inodes_per_group > blocksize * 8) {
b31e1552
ES		 3278 ext4_msg(sb, KERN_ERR,
3279 "#inodes per group too big: %lu",
2b2d6d01 3280 sbi->s_inodes_per_group);
ac27a0ec
DK		 3281 goto failed_mount;
3282 }
3283
bf43d84b
ES		 3284 /*
3285 * Test whether we have more sectors than will fit in sector_t,
3286 * and whether the max offset is addressable by the page cache.
3287 */
5a9ae68a 3288 err = generic_check_addressable(sb->s_blocksize_bits,
30ca22c7 3289 ext4_blocks_count(es));
5a9ae68a 3290 if (err) {
b31e1552 3291 ext4_msg(sb, KERN_ERR, "filesystem"
bf43d84b 3292 " too large to mount safely on this system");
ac27a0ec 3293 if (sizeof(sector_t) < 8)
90c699a9 3294 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
5a9ae68a 3295 ret = err;
ac27a0ec
DK		 3296 goto failed_mount;
3297 }
3298
617ba13b
MC		 3299 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3300 goto cantfind_ext4;
e7c95593 3301
0f2ddca6
FTN		 3302 /* check blocks count against device size */
3303 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3304 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
b31e1552
ES		 3305 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3306 "exceeds size of device (%llu blocks)",
0f2ddca6
FTN		 3307 ext4_blocks_count(es), blocks_count);
3308 goto failed_mount;
3309 }
3310
0b8e58a1
AD		 3311 /*
3312 * It makes no sense for the first data block to be beyond the end
3313 * of the filesystem.
3314 */
3315 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
b31e1552
ES		 3316 ext4_msg(sb, KERN_WARNING, "bad geometry: first data"
3317 "block %u is beyond end of filesystem (%llu)",
3318 le32_to_cpu(es->s_first_data_block),
3319 ext4_blocks_count(es));
e7c95593
ES		 3320 goto failed_mount;
3321 }
bd81d8ee
LV		 3322 blocks_count = (ext4_blocks_count(es) -
3323 le32_to_cpu(es->s_first_data_block) +
3324 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3325 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4ec11028 3326 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
b31e1552 3327 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
4ec11028 3328 "(block count %llu, first data block %u, "
b31e1552 3329 "blocks per group %lu)", sbi->s_groups_count,
4ec11028
TT		 3330 ext4_blocks_count(es),
3331 le32_to_cpu(es->s_first_data_block),
3332 EXT4_BLOCKS_PER_GROUP(sb));
3333 goto failed_mount;
3334 }
bd81d8ee 3335 sbi->s_groups_count = blocks_count;
fb0a387d
ES		 3336 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3337 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
617ba13b
MC		 3338 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3339 EXT4_DESC_PER_BLOCK(sb);
2b2d6d01 3340 sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
ac27a0ec
DK		 3341 GFP_KERNEL);
3342 if (sbi->s_group_desc == NULL) {
b31e1552 3343 ext4_msg(sb, KERN_ERR, "not enough memory");
ac27a0ec
DK		 3344 goto failed_mount;
3345 }
3346
3244fcb1 3347#ifdef CONFIG_PROC_FS
9f6200bb
TT		 3348 if (ext4_proc_root)
3349 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
3244fcb1 3350#endif
240799cd 3351
705895b6 3352 bgl_lock_init(sbi->s_blockgroup_lock);
ac27a0ec
DK		 3353
3354 for (i = 0; i < db_count; i++) {
70bbb3e0 3355 block = descriptor_loc(sb, logical_sb_block, i);
ac27a0ec
DK		 3356 sbi->s_group_desc[i] = sb_bread(sb, block);
3357 if (!sbi->s_group_desc[i]) {
b31e1552
ES		 3358 ext4_msg(sb, KERN_ERR,
3359 "can't read group descriptor %d", i);
ac27a0ec
DK		 3360 db_count = i;
3361 goto failed_mount2;
3362 }
3363 }
bfff6873 3364 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
b31e1552 3365 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
ac27a0ec
DK		 3366 goto failed_mount2;
3367 }
772cb7c8
JS		 3368 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3369 if (!ext4_fill_flex_info(sb)) {
b31e1552
ES		 3370 ext4_msg(sb, KERN_ERR,
3371 "unable to initialize "
3372 "flex_bg meta info!");
772cb7c8
JS		 3373 goto failed_mount2;
3374 }
3375
ac27a0ec
DK		 3376 sbi->s_gdb_count = db_count;
3377 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3378 spin_lock_init(&sbi->s_next_gen_lock);
3379
ce7e010a
TT		 3380 err = percpu_counter_init(&sbi->s_freeblocks_counter,
3381 ext4_count_free_blocks(sb));
3382 if (!err) {
3383 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3384 ext4_count_free_inodes(sb));
3385 }
3386 if (!err) {
3387 err = percpu_counter_init(&sbi->s_dirs_counter,
3388 ext4_count_dirs(sb));
3389 }
3390 if (!err) {
3391 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
3392 }
3393 if (err) {
3394 ext4_msg(sb, KERN_ERR, "insufficient memory");
3395 goto failed_mount3;
3396 }
3397
c9de560d 3398 sbi->s_stripe = ext4_get_stripe_size(sbi);
55138e0b 3399 sbi->s_max_writeback_mb_bump = 128;
c9de560d 3400
ac27a0ec
DK		 3401 /*
3402 * set up enough so that it can read an inode
3403 */
9ca92389
TT		 3404 if (!test_opt(sb, NOLOAD) &&
3405 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3406 sb->s_op = &ext4_sops;
3407 else
3408 sb->s_op = &ext4_nojournal_sops;
617ba13b
MC		 3409 sb->s_export_op = &ext4_export_ops;
3410 sb->s_xattr = ext4_xattr_handlers;
ac27a0ec 3411#ifdef CONFIG_QUOTA
617ba13b
MC		 3412 sb->s_qcop = &ext4_qctl_operations;
3413 sb->dq_op = &ext4_quota_operations;
ac27a0ec
DK		 3414#endif
3415 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3b9d4ed2 3416 mutex_init(&sbi->s_orphan_lock);
32ed5058 3417 mutex_init(&sbi->s_resize_lock);
ac27a0ec
DK		 3418
3419 sb->s_root = NULL;
3420
3421 needs_recovery = (es->s_last_orphan != 0 ||
617ba13b
MC		 3422 EXT4_HAS_INCOMPAT_FEATURE(sb,
3423 EXT4_FEATURE_INCOMPAT_RECOVER));
ac27a0ec
DK		 3424
3425 /*
3426 * The first inode we look at is the journal inode. Don't try
3427 * root first: it may be modified in the journal!
3428 */
3429 if (!test_opt(sb, NOLOAD) &&
617ba13b
MC		 3430 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3431 if (ext4_load_journal(sb, es, journal_devnum))
ac27a0ec 3432 goto failed_mount3;
0390131b
FM		 3433 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3434 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
b31e1552
ES		 3435 ext4_msg(sb, KERN_ERR, "required journal recovery "
3436 "suppressed and not mounted read-only");
744692dc 3437 goto failed_mount_wq;
ac27a0ec 3438 } else {
fd8c37ec
TT		 3439 clear_opt(sb, DATA_FLAGS);
3440 set_opt(sb, WRITEBACK_DATA);
0390131b
FM		 3441 sbi->s_journal = NULL;
3442 needs_recovery = 0;
3443 goto no_journal;
ac27a0ec
DK		 3444 }
3445
eb40a09c
JS		 3446 if (ext4_blocks_count(es) > 0xffffffffULL &&
3447 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3448 JBD2_FEATURE_INCOMPAT_64BIT)) {
b31e1552 3449 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
744692dc 3450 goto failed_mount_wq;
eb40a09c
JS		 3451 }
3452
d4da6c9c
LT		 3453 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3454 jbd2_journal_set_features(sbi->s_journal,
3455 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
818d276c 3456 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
d4da6c9c
LT		 3457 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3458 jbd2_journal_set_features(sbi->s_journal,
3459 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
818d276c
GS		 3460 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3461 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
d4da6c9c
LT		 3462 } else {
3463 jbd2_journal_clear_features(sbi->s_journal,
3464 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3465 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3466 }
818d276c 3467
ac27a0ec
DK		 3468 /* We have now updated the journal if required, so we can
3469 * validate the data journaling mode. */
3470 switch (test_opt(sb, DATA_FLAGS)) {
3471 case 0:
3472 /* No mode set, assume a default based on the journal
63f57933
AM		 3473 * capabilities: ORDERED_DATA if the journal can
3474 * cope, else JOURNAL_DATA
3475 */
dab291af
MC		 3476 if (jbd2_journal_check_available_features
3477 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
fd8c37ec 3478 set_opt(sb, ORDERED_DATA);
ac27a0ec 3479 else
fd8c37ec 3480 set_opt(sb, JOURNAL_DATA);
ac27a0ec
DK		 3481 break;
3482
617ba13b
MC		 3483 case EXT4_MOUNT_ORDERED_DATA:
3484 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 3485 if (!jbd2_journal_check_available_features
3486 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
b31e1552
ES		 3487 ext4_msg(sb, KERN_ERR, "Journal does not support "
3488 "requested data journaling mode");
744692dc 3489 goto failed_mount_wq;
ac27a0ec
DK		 3490 }
3491 default:
3492 break;
3493 }
b3881f74 3494 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
ac27a0ec 3495
ce7e010a
TT		 3496 /*
3497 * The journal may have updated the bg summary counts, so we
3498 * need to update the global counters.
3499 */
3500 percpu_counter_set(&sbi->s_freeblocks_counter,
3501 ext4_count_free_blocks(sb));
3502 percpu_counter_set(&sbi->s_freeinodes_counter,
3503 ext4_count_free_inodes(sb));
3504 percpu_counter_set(&sbi->s_dirs_counter,
3505 ext4_count_dirs(sb));
3506 percpu_counter_set(&sbi->s_dirtyblocks_counter, 0);
206f7ab4 3507
ce7e010a 3508no_journal:
4c0425ff
MC		 3509 EXT4_SB(sb)->dio_unwritten_wq = create_workqueue("ext4-dio-unwritten");
3510 if (!EXT4_SB(sb)->dio_unwritten_wq) {
3511 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
3512 goto failed_mount_wq;
3513 }
3514
ac27a0ec 3515 /*
dab291af 3516 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 3517 * so we can safely mount the rest of the filesystem now.
3518 */
3519
1d1fe1ee
DH		 3520 root = ext4_iget(sb, EXT4_ROOT_INO);
3521 if (IS_ERR(root)) {
b31e1552 3522 ext4_msg(sb, KERN_ERR, "get root inode failed");
1d1fe1ee 3523 ret = PTR_ERR(root);
32a9bb57 3524 root = NULL;
ac27a0ec
DK		 3525 goto failed_mount4;
3526 }
3527 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
b31e1552 3528 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
ac27a0ec
DK		 3529 goto failed_mount4;
3530 }
1d1fe1ee
DH		 3531 sb->s_root = d_alloc_root(root);
3532 if (!sb->s_root) {
b31e1552 3533 ext4_msg(sb, KERN_ERR, "get root dentry failed");
1d1fe1ee
DH		 3534 ret = -ENOMEM;
3535 goto failed_mount4;
3536 }
ac27a0ec 3537
2b2d6d01 3538 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
ef7f3835
KS		 3539
3540 /* determine the minimum size of new large inodes, if present */
3541 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3542 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3543 EXT4_GOOD_OLD_INODE_SIZE;
3544 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3545 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
3546 if (sbi->s_want_extra_isize <
3547 le16_to_cpu(es->s_want_extra_isize))
3548 sbi->s_want_extra_isize =
3549 le16_to_cpu(es->s_want_extra_isize);
3550 if (sbi->s_want_extra_isize <
3551 le16_to_cpu(es->s_min_extra_isize))
3552 sbi->s_want_extra_isize =
3553 le16_to_cpu(es->s_min_extra_isize);
3554 }
3555 }
3556 /* Check if enough inode space is available */
3557 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3558 sbi->s_inode_size) {
3559 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3560 EXT4_GOOD_OLD_INODE_SIZE;
b31e1552
ES		 3561 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3562 "available");
ef7f3835
KS		 3563 }
3564
90576c0b
TT		 3565 if (test_opt(sb, DELALLOC) &&
3566 (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)) {
b31e1552
ES		 3567 ext4_msg(sb, KERN_WARNING, "Ignoring delalloc option - "
3568 "requested data journaling mode");
fd8c37ec 3569 clear_opt(sb, DELALLOC);
90576c0b 3570 }
744692dc
JZ		 3571 if (test_opt(sb, DIOREAD_NOLOCK)) {
3572 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3573 ext4_msg(sb, KERN_WARNING, "Ignoring dioread_nolock "
3574 "option - requested data journaling mode");
fd8c37ec 3575 clear_opt(sb, DIOREAD_NOLOCK);
744692dc
JZ		 3576 }
3577 if (sb->s_blocksize < PAGE_SIZE) {
3578 ext4_msg(sb, KERN_WARNING, "Ignoring dioread_nolock "
3579 "option - block size is too small");
fd8c37ec 3580 clear_opt(sb, DIOREAD_NOLOCK);
744692dc
JZ		 3581 }
3582 }
c2774d84 3583
6fd058f7
TT		 3584 err = ext4_setup_system_zone(sb);
3585 if (err) {
b31e1552 3586 ext4_msg(sb, KERN_ERR, "failed to initialize system "
fbe845dd 3587 "zone (%d)", err);
6fd058f7
TT		 3588 goto failed_mount4;
3589 }
3590
c2774d84
AK		 3591 ext4_ext_init(sb);
3592 err = ext4_mb_init(sb, needs_recovery);
3593 if (err) {
421f91d2 3594 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
b31e1552 3595 err);
c2774d84
AK		 3596 goto failed_mount4;
3597 }
3598
bfff6873
LC		 3599 err = ext4_register_li_request(sb, first_not_zeroed);
3600 if (err)
3601 goto failed_mount4;
3602
3197ebdb
TT		 3603 sbi->s_kobj.kset = ext4_kset;
3604 init_completion(&sbi->s_kobj_unregister);
3605 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
3606 "%s", sb->s_id);
3607 if (err) {
3608 ext4_mb_release(sb);
3609 ext4_ext_release(sb);
3610 goto failed_mount4;
3611 };
3612
617ba13b
MC		 3613 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3614 ext4_orphan_cleanup(sb, es);
3615 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
0390131b 3616 if (needs_recovery) {
b31e1552 3617 ext4_msg(sb, KERN_INFO, "recovery complete");
0390131b
FM		 3618 ext4_mark_recovery_complete(sb, es);
3619 }
3620 if (EXT4_SB(sb)->s_journal) {
3621 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3622 descr = " journalled data mode";
3623 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3624 descr = " ordered data mode";
3625 else
3626 descr = " writeback data mode";
3627 } else
3628 descr = "out journal";
3629
d4c402d9 3630 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
8b67f04a
TT		 3631 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
3632 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
ac27a0ec 3633
66e61a9e
TT		 3634 init_timer(&sbi->s_err_report);
3635 sbi->s_err_report.function = print_daily_error_info;
3636 sbi->s_err_report.data = (unsigned long) sb;
3637 if (es->s_error_count)
3638 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
ac27a0ec 3639
d4c402d9 3640 kfree(orig_data);
ac27a0ec
DK		 3641 return 0;
3642
617ba13b 3643cantfind_ext4:
ac27a0ec 3644 if (!silent)
b31e1552 3645 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
ac27a0ec
DK		 3646 goto failed_mount;
3647
3648failed_mount4:
32a9bb57
MK		 3649 iput(root);
3650 sb->s_root = NULL;
b31e1552 3651 ext4_msg(sb, KERN_ERR, "mount failed");
4c0425ff
MC		 3652 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
3653failed_mount_wq:
6fd058f7 3654 ext4_release_system_zone(sb);
0390131b
FM		 3655 if (sbi->s_journal) {
3656 jbd2_journal_destroy(sbi->s_journal);
3657 sbi->s_journal = NULL;
3658 }
ac27a0ec 3659failed_mount3:
c5ca7c76
TT		 3660 if (sbi->s_flex_groups) {
3661 if (is_vmalloc_addr(sbi->s_flex_groups))
3662 vfree(sbi->s_flex_groups);
3663 else
3664 kfree(sbi->s_flex_groups);
3665 }
ce7e010a
TT		 3666 percpu_counter_destroy(&sbi->s_freeblocks_counter);
3667 percpu_counter_destroy(&sbi->s_freeinodes_counter);
3668 percpu_counter_destroy(&sbi->s_dirs_counter);
3669 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
ac27a0ec
DK		 3670failed_mount2:
3671 for (i = 0; i < db_count; i++)
3672 brelse(sbi->s_group_desc[i]);
3673 kfree(sbi->s_group_desc);
3674failed_mount:
240799cd 3675 if (sbi->s_proc) {
9f6200bb 3676 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 3677 }
ac27a0ec
DK		 3678#ifdef CONFIG_QUOTA
3679 for (i = 0; i < MAXQUOTAS; i++)
3680 kfree(sbi->s_qf_names[i]);
3681#endif
617ba13b 3682 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 3683 brelse(bh);
3684out_fail:
3685 sb->s_fs_info = NULL;
f6830165 3686 kfree(sbi->s_blockgroup_lock);
ac27a0ec 3687 kfree(sbi);
dcc7dae3 3688out_free_orig:
d4c402d9 3689 kfree(orig_data);
1d1fe1ee 3690 return ret;
ac27a0ec
DK		 3691}
3692
3693/*
3694 * Setup any per-fs journal parameters now. We'll do this both on
3695 * initial mount, once the journal has been initialised but before we've
3696 * done any recovery; and again on any subsequent remount.
3697 */
617ba13b 3698static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 3699{
617ba13b 3700 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 3701
30773840
TT		 3702 journal->j_commit_interval = sbi->s_commit_interval;
3703 journal->j_min_batch_time = sbi->s_min_batch_time;
3704 journal->j_max_batch_time = sbi->s_max_batch_time;
ac27a0ec 3705
a931da6a 3706 write_lock(&journal->j_state_lock);
ac27a0ec 3707 if (test_opt(sb, BARRIER))
dab291af 3708 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 3709 else
dab291af 3710 journal->j_flags &= ~JBD2_BARRIER;
5bf5683a
HK		 3711 if (test_opt(sb, DATA_ERR_ABORT))
3712 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
3713 else
3714 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
a931da6a 3715 write_unlock(&journal->j_state_lock);
ac27a0ec
DK		 3716}
3717
617ba13b 3718static journal_t *ext4_get_journal(struct super_block *sb,
ac27a0ec
DK		 3719 unsigned int journal_inum)
3720{
3721 struct inode *journal_inode;
3722 journal_t *journal;
3723
0390131b
FM		 3724 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3725
ac27a0ec
DK		 3726 /* First, test for the existence of a valid inode on disk. Bad
3727 * things happen if we iget() an unused inode, as the subsequent
3728 * iput() will try to delete it. */
3729
1d1fe1ee
DH		 3730 journal_inode = ext4_iget(sb, journal_inum);
3731 if (IS_ERR(journal_inode)) {
b31e1552 3732 ext4_msg(sb, KERN_ERR, "no journal found");
ac27a0ec
DK		 3733 return NULL;
3734 }
3735 if (!journal_inode->i_nlink) {
3736 make_bad_inode(journal_inode);
3737 iput(journal_inode);
b31e1552 3738 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
ac27a0ec
DK		 3739 return NULL;
3740 }
3741
e5f8eab8 3742 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
ac27a0ec 3743 journal_inode, journal_inode->i_size);
1d1fe1ee 3744 if (!S_ISREG(journal_inode->i_mode)) {
b31e1552 3745 ext4_msg(sb, KERN_ERR, "invalid journal inode");
ac27a0ec
DK		 3746 iput(journal_inode);
3747 return NULL;
3748 }
3749
dab291af 3750 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 3751 if (!journal) {
b31e1552 3752 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
ac27a0ec
DK		 3753 iput(journal_inode);
3754 return NULL;
3755 }
3756 journal->j_private = sb;
617ba13b 3757 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 3758 return journal;
3759}
3760
617ba13b 3761static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 3762 dev_t j_dev)
3763{
2b2d6d01 3764 struct buffer_head *bh;
ac27a0ec 3765 journal_t *journal;
617ba13b
MC		 3766 ext4_fsblk_t start;
3767 ext4_fsblk_t len;
ac27a0ec 3768 int hblock, blocksize;
617ba13b 3769 ext4_fsblk_t sb_block;
ac27a0ec 3770 unsigned long offset;
2b2d6d01 3771 struct ext4_super_block *es;
ac27a0ec
DK		 3772 struct block_device *bdev;
3773
0390131b
FM		 3774 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3775
b31e1552 3776 bdev = ext4_blkdev_get(j_dev, sb);
ac27a0ec
DK		 3777 if (bdev == NULL)
3778 return NULL;
3779
ac27a0ec 3780 blocksize = sb->s_blocksize;
e1defc4f 3781 hblock = bdev_logical_block_size(bdev);
ac27a0ec 3782 if (blocksize < hblock) {
b31e1552
ES		 3783 ext4_msg(sb, KERN_ERR,
3784 "blocksize too small for journal device");
ac27a0ec
DK		 3785 goto out_bdev;
3786 }
3787
617ba13b
MC		 3788 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3789 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 3790 set_blocksize(bdev, blocksize);
3791 if (!(bh = __bread(bdev, sb_block, blocksize))) {
b31e1552
ES		 3792 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
3793 "external journal");
ac27a0ec
DK		 3794 goto out_bdev;
3795 }
3796
617ba13b
MC		 3797 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3798 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 3799 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b 3800 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
b31e1552
ES		 3801 ext4_msg(sb, KERN_ERR, "external journal has "
3802 "bad superblock");
ac27a0ec
DK		 3803 brelse(bh);
3804 goto out_bdev;
3805 }
3806
617ba13b 3807 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
b31e1552 3808 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
ac27a0ec
DK		 3809 brelse(bh);
3810 goto out_bdev;
3811 }
3812
bd81d8ee 3813 len = ext4_blocks_count(es);
ac27a0ec
DK		 3814 start = sb_block + 1;
3815 brelse(bh); /* we're done with the superblock */
3816
dab291af 3817 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 3818 start, len, blocksize);
3819 if (!journal) {
b31e1552 3820 ext4_msg(sb, KERN_ERR, "failed to create device journal");
ac27a0ec
DK		 3821 goto out_bdev;
3822 }
3823 journal->j_private = sb;
3824 ll_rw_block(READ, 1, &journal->j_sb_buffer);
3825 wait_on_buffer(journal->j_sb_buffer);
3826 if (!buffer_uptodate(journal->j_sb_buffer)) {
b31e1552 3827 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
ac27a0ec
DK		 3828 goto out_journal;
3829 }
3830 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
b31e1552
ES		 3831 ext4_msg(sb, KERN_ERR, "External journal has more than one "
3832 "user (unsupported) - %d",
ac27a0ec
DK		 3833 be32_to_cpu(journal->j_superblock->s_nr_users));
3834 goto out_journal;
3835 }
617ba13b
MC		 3836 EXT4_SB(sb)->journal_bdev = bdev;
3837 ext4_init_journal_params(sb, journal);
ac27a0ec 3838 return journal;
0b8e58a1 3839
ac27a0ec 3840out_journal:
dab291af 3841 jbd2_journal_destroy(journal);
ac27a0ec 3842out_bdev:
617ba13b 3843 ext4_blkdev_put(bdev);
ac27a0ec
DK		 3844 return NULL;
3845}
3846
617ba13b
MC		 3847static int ext4_load_journal(struct super_block *sb,
3848 struct ext4_super_block *es,
ac27a0ec
DK		 3849 unsigned long journal_devnum)
3850{
3851 journal_t *journal;
3852 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
3853 dev_t journal_dev;
3854 int err = 0;
3855 int really_read_only;
3856
0390131b
FM		 3857 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3858
ac27a0ec
DK		 3859 if (journal_devnum &&
3860 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
b31e1552
ES		 3861 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
3862 "numbers have changed");
ac27a0ec
DK		 3863 journal_dev = new_decode_dev(journal_devnum);
3864 } else
3865 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
3866
3867 really_read_only = bdev_read_only(sb->s_bdev);
3868
3869 /*
3870 * Are we loading a blank journal or performing recovery after a
3871 * crash? For recovery, we need to check in advance whether we
3872 * can get read-write access to the device.
3873 */
617ba13b 3874 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
ac27a0ec 3875 if (sb->s_flags & MS_RDONLY) {
b31e1552
ES		 3876 ext4_msg(sb, KERN_INFO, "INFO: recovery "
3877 "required on readonly filesystem");
ac27a0ec 3878 if (really_read_only) {
b31e1552
ES		 3879 ext4_msg(sb, KERN_ERR, "write access "
3880 "unavailable, cannot proceed");
ac27a0ec
DK		 3881 return -EROFS;
3882 }
b31e1552
ES		 3883 ext4_msg(sb, KERN_INFO, "write access will "
3884 "be enabled during recovery");
ac27a0ec
DK		 3885 }
3886 }
3887
3888 if (journal_inum && journal_dev) {
b31e1552
ES		 3889 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
3890 "and inode journals!");
ac27a0ec
DK		 3891 return -EINVAL;
3892 }
3893
3894 if (journal_inum) {
617ba13b 3895 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 3896 return -EINVAL;
3897 } else {
617ba13b 3898 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 3899 return -EINVAL;
3900 }
3901
90576c0b 3902 if (!(journal->j_flags & JBD2_BARRIER))
b31e1552 3903 ext4_msg(sb, KERN_INFO, "barriers disabled");
4776004f 3904
ac27a0ec 3905 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
dab291af 3906 err = jbd2_journal_update_format(journal);
ac27a0ec 3907 if (err) {
b31e1552 3908 ext4_msg(sb, KERN_ERR, "error updating journal");
dab291af 3909 jbd2_journal_destroy(journal);
ac27a0ec
DK		 3910 return err;
3911 }
3912 }
3913
617ba13b 3914 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
dab291af 3915 err = jbd2_journal_wipe(journal, !really_read_only);
1c13d5c0
TT		 3916 if (!err) {
3917 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
3918 if (save)
3919 memcpy(save, ((char *) es) +
3920 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
dab291af 3921 err = jbd2_journal_load(journal);
1c13d5c0
TT		 3922 if (save)
3923 memcpy(((char *) es) + EXT4_S_ERR_START,
3924 save, EXT4_S_ERR_LEN);
3925 kfree(save);
3926 }
ac27a0ec
DK		 3927
3928 if (err) {
b31e1552 3929 ext4_msg(sb, KERN_ERR, "error loading journal");
dab291af 3930 jbd2_journal_destroy(journal);
ac27a0ec
DK		 3931 return err;
3932 }
3933
617ba13b
MC		 3934 EXT4_SB(sb)->s_journal = journal;
3935 ext4_clear_journal_err(sb, es);
ac27a0ec 3936
c41303ce 3937 if (!really_read_only && journal_devnum &&
ac27a0ec
DK		 3938 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3939 es->s_journal_dev = cpu_to_le32(journal_devnum);
ac27a0ec
DK		 3940
3941 /* Make sure we flush the recovery flag to disk. */
e2d67052 3942 ext4_commit_super(sb, 1);
ac27a0ec
DK		 3943 }
3944
3945 return 0;
3946}
3947
e2d67052 3948static int ext4_commit_super(struct super_block *sb, int sync)
ac27a0ec 3949{
e2d67052 3950 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
617ba13b 3951 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
c4be0c1d 3952 int error = 0;
ac27a0ec
DK		 3953
3954 if (!sbh)
c4be0c1d 3955 return error;
914258bf
TT		 3956 if (buffer_write_io_error(sbh)) {
3957 /*
3958 * Oh, dear. A previous attempt to write the
3959 * superblock failed. This could happen because the
3960 * USB device was yanked out. Or it could happen to
3961 * be a transient write error and maybe the block will
3962 * be remapped. Nothing we can do but to retry the
3963 * write and hope for the best.
3964 */
b31e1552
ES		 3965 ext4_msg(sb, KERN_ERR, "previous I/O error to "
3966 "superblock detected");
914258bf
TT		 3967 clear_buffer_write_io_error(sbh);
3968 set_buffer_uptodate(sbh);
3969 }
71290b36
TT		 3970 /*
3971 * If the file system is mounted read-only, don't update the
3972 * superblock write time. This avoids updating the superblock
3973 * write time when we are mounting the root file system
3974 * read/only but we need to replay the journal; at that point,
3975 * for people who are east of GMT and who make their clock
3976 * tick in localtime for Windows bug-for-bug compatibility,
3977 * the clock is set in the future, and this will cause e2fsck
3978 * to complain and force a full file system check.
3979 */
3980 if (!(sb->s_flags & MS_RDONLY))
3981 es->s_wtime = cpu_to_le32(get_seconds());
f613dfcb
TT		 3982 if (sb->s_bdev->bd_part)
3983 es->s_kbytes_written =
3984 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
afc32f7e
TT		 3985 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
3986 EXT4_SB(sb)->s_sectors_written_start) >> 1));
f613dfcb
TT		 3987 else
3988 es->s_kbytes_written =
3989 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
ce7e010a
TT		 3990 ext4_free_blocks_count_set(es, percpu_counter_sum_positive(
3991 &EXT4_SB(sb)->s_freeblocks_counter));
3992 es->s_free_inodes_count =
3993 cpu_to_le32(percpu_counter_sum_positive(
3994 &EXT4_SB(sb)->s_freeinodes_counter));
7234ab2a 3995 sb->s_dirt = 0;
ac27a0ec
DK		 3996 BUFFER_TRACE(sbh, "marking dirty");
3997 mark_buffer_dirty(sbh);
914258bf 3998 if (sync) {
c4be0c1d
TS		 3999 error = sync_dirty_buffer(sbh);
4000 if (error)
4001 return error;
4002
4003 error = buffer_write_io_error(sbh);
4004 if (error) {
b31e1552
ES		 4005 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4006 "superblock");
914258bf
TT		 4007 clear_buffer_write_io_error(sbh);
4008 set_buffer_uptodate(sbh);
4009 }
4010 }
c4be0c1d 4011 return error;
ac27a0ec
DK		 4012}
4013
ac27a0ec
DK		 4014/*
4015 * Have we just finished recovery? If so, and if we are mounting (or
4016 * remounting) the filesystem readonly, then we will end up with a
4017 * consistent fs on disk. Record that fact.
4018 */
2b2d6d01
TT		 4019static void ext4_mark_recovery_complete(struct super_block *sb,
4020 struct ext4_super_block *es)
ac27a0ec 4021{
617ba13b 4022 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 4023
0390131b
FM		 4024 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4025 BUG_ON(journal != NULL);
4026 return;
4027 }
dab291af 4028 jbd2_journal_lock_updates(journal);
7ffe1ea8
HK		 4029 if (jbd2_journal_flush(journal) < 0)
4030 goto out;
4031
617ba13b 4032 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
ac27a0ec 4033 sb->s_flags & MS_RDONLY) {
617ba13b 4034 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
e2d67052 4035 ext4_commit_super(sb, 1);
ac27a0ec 4036 }
7ffe1ea8
HK		 4037
4038out:
dab291af 4039 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 4040}
4041
4042/*
4043 * If we are mounting (or read-write remounting) a filesystem whose journal
4044 * has recorded an error from a previous lifetime, move that error to the
4045 * main filesystem now.
4046 */
2b2d6d01
TT		 4047static void ext4_clear_journal_err(struct super_block *sb,
4048 struct ext4_super_block *es)
ac27a0ec
DK		 4049{
4050 journal_t *journal;
4051 int j_errno;
4052 const char *errstr;
4053
0390131b
FM		 4054 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4055
617ba13b 4056 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 4057
4058 /*
4059 * Now check for any error status which may have been recorded in the
617ba13b 4060 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 4061 */
4062
dab291af 4063 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 4064 if (j_errno) {
4065 char nbuf[16];
4066
617ba13b 4067 errstr = ext4_decode_error(sb, j_errno, nbuf);
12062ddd 4068 ext4_warning(sb, "Filesystem error recorded "
ac27a0ec 4069 "from previous mount: %s", errstr);
12062ddd 4070 ext4_warning(sb, "Marking fs in need of filesystem check.");
ac27a0ec 4071
617ba13b
MC		 4072 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4073 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
e2d67052 4074 ext4_commit_super(sb, 1);
ac27a0ec 4075
dab291af 4076 jbd2_journal_clear_err(journal);
ac27a0ec
DK		 4077 }
4078}
4079
4080/*
4081 * Force the running and committing transactions to commit,
4082 * and wait on the commit.
4083 */
617ba13b 4084int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 4085{
4086 journal_t *journal;
0390131b 4087 int ret = 0;
ac27a0ec
DK		 4088
4089 if (sb->s_flags & MS_RDONLY)
4090 return 0;
4091
617ba13b 4092 journal = EXT4_SB(sb)->s_journal;
6b0310fb 4093 if (journal) {
437f88cc 4094 vfs_check_frozen(sb, SB_FREEZE_TRANS);
0390131b 4095 ret = ext4_journal_force_commit(journal);
6b0310fb 4096 }
0390131b 4097
ac27a0ec
DK		 4098 return ret;
4099}
4100
2b2d6d01 4101static void ext4_write_super(struct super_block *sb)
ac27a0ec 4102{
ebc1ac16 4103 lock_super(sb);
9ca92389 4104 ext4_commit_super(sb, 1);
ebc1ac16 4105 unlock_super(sb);
ac27a0ec
DK		 4106}
4107
617ba13b 4108static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec 4109{
14ce0cb4 4110 int ret = 0;
9eddacf9 4111 tid_t target;
8d5d02e6 4112 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4113
9bffad1e 4114 trace_ext4_sync_fs(sb, wait);
8d5d02e6
MC		 4115 flush_workqueue(sbi->dio_unwritten_wq);
4116 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
9ca92389 4117 if (wait)
8d5d02e6 4118 jbd2_log_wait_commit(sbi->s_journal, target);
0390131b 4119 }
14ce0cb4 4120 return ret;
ac27a0ec
DK		 4121}
4122
4123/*
4124 * LVM calls this function before a (read-only) snapshot is created. This
4125 * gives us a chance to flush the journal completely and mark the fs clean.
4126 */
c4be0c1d 4127static int ext4_freeze(struct super_block *sb)
ac27a0ec 4128{
c4be0c1d
TS		 4129 int error = 0;
4130 journal_t *journal;
ac27a0ec 4131
9ca92389
TT		 4132 if (sb->s_flags & MS_RDONLY)
4133 return 0;
ac27a0ec 4134
9ca92389 4135 journal = EXT4_SB(sb)->s_journal;
7ffe1ea8 4136
9ca92389
TT		 4137 /* Now we set up the journal barrier. */
4138 jbd2_journal_lock_updates(journal);
ac27a0ec 4139
9ca92389
TT		 4140 /*
4141 * Don't clear the needs_recovery flag if we failed to flush
4142 * the journal.
4143 */
4144 error = jbd2_journal_flush(journal);
6b0310fb
ES		 4145 if (error < 0)
4146 goto out;
9ca92389
TT		 4147
4148 /* Journal blocked and flushed, clear needs_recovery flag. */
4149 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4150 error = ext4_commit_super(sb, 1);
6b0310fb
ES		 4151out:
4152 /* we rely on s_frozen to stop further updates */
4153 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4154 return error;
ac27a0ec
DK		 4155}
4156
4157/*
4158 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4159 * flag here, even though the filesystem is not technically dirty yet.
4160 */
c4be0c1d 4161static int ext4_unfreeze(struct super_block *sb)
ac27a0ec 4162{
9ca92389
TT		 4163 if (sb->s_flags & MS_RDONLY)
4164 return 0;
4165
4166 lock_super(sb);
4167 /* Reset the needs_recovery flag before the fs is unlocked. */
4168 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4169 ext4_commit_super(sb, 1);
4170 unlock_super(sb);
c4be0c1d 4171 return 0;
ac27a0ec
DK		 4172}
4173
673c6100
TT		 4174/*
4175 * Structure to save mount options for ext4_remount's benefit
4176 */
4177struct ext4_mount_options {
4178 unsigned long s_mount_opt;
a2595b8a 4179 unsigned long s_mount_opt2;
673c6100
TT		 4180 uid_t s_resuid;
4181 gid_t s_resgid;
4182 unsigned long s_commit_interval;
4183 u32 s_min_batch_time, s_max_batch_time;
4184#ifdef CONFIG_QUOTA
4185 int s_jquota_fmt;
4186 char *s_qf_names[MAXQUOTAS];
4187#endif
4188};
4189
2b2d6d01 4190static int ext4_remount(struct super_block *sb, int *flags, char *data)
ac27a0ec 4191{
2b2d6d01 4192 struct ext4_super_block *es;
617ba13b
MC		 4193 struct ext4_sb_info *sbi = EXT4_SB(sb);
4194 ext4_fsblk_t n_blocks_count = 0;
ac27a0ec 4195 unsigned long old_sb_flags;
617ba13b 4196 struct ext4_mount_options old_opts;
c79d967d 4197 int enable_quota = 0;
8a266467 4198 ext4_group_t g;
b3881f74 4199 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
ac27a0ec
DK		 4200 int err;
4201#ifdef CONFIG_QUOTA
4202 int i;
4203#endif
d4c402d9 4204 char *orig_data = kstrdup(data, GFP_KERNEL);
ac27a0ec
DK		 4205
4206 /* Store the original options */
bbd6851a 4207 lock_super(sb);
ac27a0ec
DK		 4208 old_sb_flags = sb->s_flags;
4209 old_opts.s_mount_opt = sbi->s_mount_opt;
a2595b8a 4210 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
ac27a0ec
DK		 4211 old_opts.s_resuid = sbi->s_resuid;
4212 old_opts.s_resgid = sbi->s_resgid;
4213 old_opts.s_commit_interval = sbi->s_commit_interval;
30773840
TT		 4214 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4215 old_opts.s_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 4216#ifdef CONFIG_QUOTA
4217 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4218 for (i = 0; i < MAXQUOTAS; i++)
4219 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
4220#endif
b3881f74
TT		 4221 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4222 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
ac27a0ec
DK		 4223
4224 /*
4225 * Allow the "check" option to be passed as a remount option.
4226 */
b3881f74
TT		 4227 if (!parse_options(data, sb, NULL, &journal_ioprio,
4228 &n_blocks_count, 1)) {
ac27a0ec
DK		 4229 err = -EINVAL;
4230 goto restore_opts;
4231 }
4232
4ab2f15b 4233 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
c67d859e 4234 ext4_abort(sb, "Abort forced by user");
ac27a0ec
DK		 4235
4236 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 4237 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec
DK		 4238
4239 es = sbi->s_es;
4240
b3881f74 4241 if (sbi->s_journal) {
0390131b 4242 ext4_init_journal_params(sb, sbi->s_journal);
b3881f74
TT		 4243 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4244 }
ac27a0ec
DK		 4245
4246 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
bd81d8ee 4247 n_blocks_count > ext4_blocks_count(es)) {
4ab2f15b 4248 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
ac27a0ec
DK		 4249 err = -EROFS;
4250 goto restore_opts;
4251 }
4252
4253 if (*flags & MS_RDONLY) {
0f0dd62f
CH		 4254 err = dquot_suspend(sb, -1);
4255 if (err < 0)
c79d967d 4256 goto restore_opts;
c79d967d 4257
ac27a0ec
DK		 4258 /*
4259 * First of all, the unconditional stuff we have to do
4260 * to disable replay of the journal when we next remount
4261 */
4262 sb->s_flags |= MS_RDONLY;
4263
4264 /*
4265 * OK, test if we are remounting a valid rw partition
4266 * readonly, and if so set the rdonly flag and then
4267 * mark the partition as valid again.
4268 */
617ba13b
MC		 4269 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4270 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 4271 es->s_state = cpu_to_le16(sbi->s_mount_state);
4272
a63c9eb2 4273 if (sbi->s_journal)
0390131b 4274 ext4_mark_recovery_complete(sb, es);
ac27a0ec 4275 } else {
a13fb1a4
ES		 4276 /* Make sure we can mount this feature set readwrite */
4277 if (!ext4_feature_set_ok(sb, 0)) {
ac27a0ec
DK		 4278 err = -EROFS;
4279 goto restore_opts;
4280 }
8a266467
TT		 4281 /*
4282 * Make sure the group descriptor checksums
0b8e58a1 4283 * are sane. If they aren't, refuse to remount r/w.
8a266467
TT		 4284 */
4285 for (g = 0; g < sbi->s_groups_count; g++) {
4286 struct ext4_group_desc *gdp =
4287 ext4_get_group_desc(sb, g, NULL);
4288
4289 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
b31e1552
ES		 4290 ext4_msg(sb, KERN_ERR,
4291 "ext4_remount: Checksum for group %u failed (%u!=%u)",
8a266467
TT		 4292 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4293 le16_to_cpu(gdp->bg_checksum));
4294 err = -EINVAL;
4295 goto restore_opts;
4296 }
4297 }
4298
ead6596b
ES		 4299 /*
4300 * If we have an unprocessed orphan list hanging
4301 * around from a previously readonly bdev mount,
4302 * require a full umount/remount for now.
4303 */
4304 if (es->s_last_orphan) {
b31e1552 4305 ext4_msg(sb, KERN_WARNING, "Couldn't "
ead6596b
ES		 4306 "remount RDWR because of unprocessed "
4307 "orphan inode list. Please "
b31e1552 4308 "umount/remount instead");
ead6596b
ES		 4309 err = -EINVAL;
4310 goto restore_opts;
4311 }
4312
ac27a0ec
DK		 4313 /*
4314 * Mounting a RDONLY partition read-write, so reread
4315 * and store the current valid flag. (It may have
4316 * been changed by e2fsck since we originally mounted
4317 * the partition.)
4318 */
0390131b
FM		 4319 if (sbi->s_journal)
4320 ext4_clear_journal_err(sb, es);
ac27a0ec 4321 sbi->s_mount_state = le16_to_cpu(es->s_state);
617ba13b 4322 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
ac27a0ec 4323 goto restore_opts;
2b2d6d01 4324 if (!ext4_setup_super(sb, es, 0))
ac27a0ec 4325 sb->s_flags &= ~MS_RDONLY;
c79d967d 4326 enable_quota = 1;
ac27a0ec
DK		 4327 }
4328 }
bfff6873
LC		 4329
4330 /*
4331 * Reinitialize lazy itable initialization thread based on
4332 * current settings
4333 */
4334 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4335 ext4_unregister_li_request(sb);
4336 else {
4337 ext4_group_t first_not_zeroed;
4338 first_not_zeroed = ext4_has_uninit_itable(sb);
4339 ext4_register_li_request(sb, first_not_zeroed);
4340 }
4341
6fd058f7 4342 ext4_setup_system_zone(sb);
0390131b 4343 if (sbi->s_journal == NULL)
e2d67052 4344 ext4_commit_super(sb, 1);
0390131b 4345
ac27a0ec
DK		 4346#ifdef CONFIG_QUOTA
4347 /* Release old quota file names */
4348 for (i = 0; i < MAXQUOTAS; i++)
4349 if (old_opts.s_qf_names[i] &&
4350 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4351 kfree(old_opts.s_qf_names[i]);
4352#endif
bbd6851a 4353 unlock_super(sb);
c79d967d 4354 if (enable_quota)
0f0dd62f 4355 dquot_resume(sb, -1);
d4c402d9
CW		 4356
4357 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4358 kfree(orig_data);
ac27a0ec 4359 return 0;
0b8e58a1 4360
ac27a0ec
DK		 4361restore_opts:
4362 sb->s_flags = old_sb_flags;
4363 sbi->s_mount_opt = old_opts.s_mount_opt;
a2595b8a 4364 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
ac27a0ec
DK		 4365 sbi->s_resuid = old_opts.s_resuid;
4366 sbi->s_resgid = old_opts.s_resgid;
4367 sbi->s_commit_interval = old_opts.s_commit_interval;
30773840
TT		 4368 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4369 sbi->s_max_batch_time = old_opts.s_max_batch_time;
ac27a0ec
DK		 4370#ifdef CONFIG_QUOTA
4371 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4372 for (i = 0; i < MAXQUOTAS; i++) {
4373 if (sbi->s_qf_names[i] &&
4374 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4375 kfree(sbi->s_qf_names[i]);
4376 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4377 }
4378#endif
bbd6851a 4379 unlock_super(sb);
d4c402d9 4380 kfree(orig_data);
ac27a0ec
DK		 4381 return err;
4382}
4383
2b2d6d01 4384static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
ac27a0ec
DK		 4385{
4386 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 4387 struct ext4_sb_info *sbi = EXT4_SB(sb);
4388 struct ext4_super_block *es = sbi->s_es;
960cc398 4389 u64 fsid;
ac27a0ec 4390
5e70030d
BP		 4391 if (test_opt(sb, MINIX_DF)) {
4392 sbi->s_overhead_last = 0;
6bc9feff 4393 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
8df9675f 4394 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
5e70030d 4395 ext4_fsblk_t overhead = 0;
ac27a0ec
DK		 4396
4397 /*
5e70030d
BP		 4398 * Compute the overhead (FS structures). This is constant
4399 * for a given filesystem unless the number of block groups
4400 * changes so we cache the previous value until it does.
ac27a0ec
DK		 4401 */
4402
4403 /*
4404 * All of the blocks before first_data_block are
4405 * overhead
4406 */
4407 overhead = le32_to_cpu(es->s_first_data_block);
4408
4409 /*
4410 * Add the overhead attributed to the superblock and
4411 * block group descriptors. If the sparse superblocks
4412 * feature is turned on, then not all groups have this.
4413 */
4414 for (i = 0; i < ngroups; i++) {
617ba13b
MC		 4415 overhead += ext4_bg_has_super(sb, i) +
4416 ext4_bg_num_gdb(sb, i);
ac27a0ec
DK		 4417 cond_resched();
4418 }
4419
4420 /*
4421 * Every block group has an inode bitmap, a block
4422 * bitmap, and an inode table.
4423 */
5e70030d
BP		 4424 overhead += ngroups * (2 + sbi->s_itb_per_group);
4425 sbi->s_overhead_last = overhead;
4426 smp_wmb();
6bc9feff 4427 sbi->s_blocks_last = ext4_blocks_count(es);
ac27a0ec
DK		 4428 }
4429
617ba13b 4430 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 4431 buf->f_bsize = sb->s_blocksize;
5e70030d 4432 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
6bc6e63f
AK		 4433 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
4434 percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
bd81d8ee
LV		 4435 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
4436 if (buf->f_bfree < ext4_r_blocks_count(es))
ac27a0ec
DK		 4437 buf->f_bavail = 0;
4438 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 4439 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
617ba13b 4440 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 4441 fsid = le64_to_cpup((void *)es->s_uuid) ^
4442 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4443 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4444 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
0b8e58a1 4445
ac27a0ec
DK		 4446 return 0;
4447}
4448
0b8e58a1
AD		 4449/* Helper function for writing quotas on sync - we need to start transaction
4450 * before quota file is locked for write. Otherwise the are possible deadlocks:
ac27a0ec 4451 * Process 1 Process 2
617ba13b 4452 * ext4_create() quota_sync()
a269eb18 4453 * jbd2_journal_start() write_dquot()
871a2931 4454 * dquot_initialize() down(dqio_mutex)
dab291af 4455 * down(dqio_mutex) jbd2_journal_start()
ac27a0ec
DK		 4456 *
4457 */
4458
4459#ifdef CONFIG_QUOTA
4460
4461static inline struct inode *dquot_to_inode(struct dquot *dquot)
4462{
4463 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
4464}
4465
617ba13b 4466static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 4467{
4468 int ret, err;
4469 handle_t *handle;
4470 struct inode *inode;
4471
4472 inode = dquot_to_inode(dquot);
617ba13b 4473 handle = ext4_journal_start(inode,
0b8e58a1 4474 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 4475 if (IS_ERR(handle))
4476 return PTR_ERR(handle);
4477 ret = dquot_commit(dquot);
617ba13b 4478 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4479 if (!ret)
4480 ret = err;
4481 return ret;
4482}
4483
617ba13b 4484static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 4485{
4486 int ret, err;
4487 handle_t *handle;
4488
617ba13b 4489 handle = ext4_journal_start(dquot_to_inode(dquot),
0b8e58a1 4490 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 4491 if (IS_ERR(handle))
4492 return PTR_ERR(handle);
4493 ret = dquot_acquire(dquot);
617ba13b 4494 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4495 if (!ret)
4496 ret = err;
4497 return ret;
4498}
4499
617ba13b 4500static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 4501{
4502 int ret, err;
4503 handle_t *handle;
4504
617ba13b 4505 handle = ext4_journal_start(dquot_to_inode(dquot),
0b8e58a1 4506 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 4507 if (IS_ERR(handle)) {
4508 /* Release dquot anyway to avoid endless cycle in dqput() */
4509 dquot_release(dquot);
ac27a0ec 4510 return PTR_ERR(handle);
9c3013e9 4511 }
ac27a0ec 4512 ret = dquot_release(dquot);
617ba13b 4513 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4514 if (!ret)
4515 ret = err;
4516 return ret;
4517}
4518
617ba13b 4519static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec 4520{
2c8be6b2 4521 /* Are we journaling quotas? */
617ba13b
MC		 4522 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
4523 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
ac27a0ec 4524 dquot_mark_dquot_dirty(dquot);
617ba13b 4525 return ext4_write_dquot(dquot);
ac27a0ec
DK		 4526 } else {
4527 return dquot_mark_dquot_dirty(dquot);
4528 }
4529}
4530
617ba13b 4531static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 4532{
4533 int ret, err;
4534 handle_t *handle;
4535
4536 /* Data block + inode block */
617ba13b 4537 handle = ext4_journal_start(sb->s_root->d_inode, 2);
ac27a0ec
DK		 4538 if (IS_ERR(handle))
4539 return PTR_ERR(handle);
4540 ret = dquot_commit_info(sb, type);
617ba13b 4541 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4542 if (!ret)
4543 ret = err;
4544 return ret;
4545}
4546
4547/*
4548 * Turn on quotas during mount time - we need to find
4549 * the quota file and such...
4550 */
617ba13b 4551static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 4552{
287a8095
CH		 4553 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4554 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 4555}
4556
4557/*
4558 * Standard function to be called on quota_on
4559 */
617ba13b 4560static int ext4_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 4561 struct path *path)
ac27a0ec
DK		 4562{
4563 int err;
ac27a0ec
DK		 4564
4565 if (!test_opt(sb, QUOTA))
4566 return -EINVAL;
0623543b 4567
ac27a0ec 4568 /* Quotafile not on the same filesystem? */
f00c9e44 4569 if (path->mnt->mnt_sb != sb)
ac27a0ec 4570 return -EXDEV;
0623543b
JK		 4571 /* Journaling quota? */
4572 if (EXT4_SB(sb)->s_qf_names[type]) {
2b2d6d01 4573 /* Quotafile not in fs root? */
f00c9e44 4574 if (path->dentry->d_parent != sb->s_root)
b31e1552
ES		 4575 ext4_msg(sb, KERN_WARNING,
4576 "Quota file not on filesystem root. "
4577 "Journaled quota will not work");
2b2d6d01 4578 }
0623543b
JK		 4579
4580 /*
4581 * When we journal data on quota file, we have to flush journal to see
4582 * all updates to the file when we bypass pagecache...
4583 */
0390131b 4584 if (EXT4_SB(sb)->s_journal &&
f00c9e44 4585 ext4_should_journal_data(path->dentry->d_inode)) {
0623543b
JK		 4586 /*
4587 * We don't need to lock updates but journal_flush() could
4588 * otherwise be livelocked...
4589 */
4590 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 4591 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
0623543b 4592 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
f00c9e44 4593 if (err)
7ffe1ea8 4594 return err;
0623543b
JK		 4595 }
4596
f00c9e44 4597 return dquot_quota_on(sb, type, format_id, path);
ac27a0ec
DK		 4598}
4599
ca0e05e4
DM		 4600static int ext4_quota_off(struct super_block *sb, int type)
4601{
87009d86
DM		 4602 /* Force all delayed allocation blocks to be allocated.
4603 * Caller already holds s_umount sem */
4604 if (test_opt(sb, DELALLOC))
ca0e05e4 4605 sync_filesystem(sb);
ca0e05e4
DM		 4606
4607 return dquot_quota_off(sb, type);
4608}
4609
ac27a0ec
DK		 4610/* Read data from quotafile - avoid pagecache and such because we cannot afford
4611 * acquiring the locks... As quota files are never truncated and quota code
4612 * itself serializes the operations (and noone else should touch the files)
4613 * we don't have to be afraid of races */
617ba13b 4614static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 4615 size_t len, loff_t off)
4616{
4617 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 4618 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 4619 int err = 0;
4620 int offset = off & (sb->s_blocksize - 1);
4621 int tocopy;
4622 size_t toread;
4623 struct buffer_head *bh;
4624 loff_t i_size = i_size_read(inode);
4625
4626 if (off > i_size)
4627 return 0;
4628 if (off+len > i_size)
4629 len = i_size-off;
4630 toread = len;
4631 while (toread > 0) {
4632 tocopy = sb->s_blocksize - offset < toread ?
4633 sb->s_blocksize - offset : toread;
617ba13b 4634 bh = ext4_bread(NULL, inode, blk, 0, &err);
ac27a0ec
DK		 4635 if (err)
4636 return err;
4637 if (!bh) /* A hole? */
4638 memset(data, 0, tocopy);
4639 else
4640 memcpy(data, bh->b_data+offset, tocopy);
4641 brelse(bh);
4642 offset = 0;
4643 toread -= tocopy;
4644 data += tocopy;
4645 blk++;
4646 }
4647 return len;
4648}
4649
4650/* Write to quotafile (we know the transaction is already started and has
4651 * enough credits) */
617ba13b 4652static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 4653 const char *data, size_t len, loff_t off)
4654{
4655 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 4656 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 4657 int err = 0;
4658 int offset = off & (sb->s_blocksize - 1);
ac27a0ec
DK		 4659 struct buffer_head *bh;
4660 handle_t *handle = journal_current_handle();
4661
0390131b 4662 if (EXT4_SB(sb)->s_journal && !handle) {
b31e1552
ES		 4663 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4664 " cancelled because transaction is not started",
9c3013e9
JK		 4665 (unsigned long long)off, (unsigned long long)len);
4666 return -EIO;
4667 }
67eeb568
DM		 4668 /*
4669 * Since we account only one data block in transaction credits,
4670 * then it is impossible to cross a block boundary.
4671 */
4672 if (sb->s_blocksize - offset < len) {
4673 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4674 " cancelled because not block aligned",
4675 (unsigned long long)off, (unsigned long long)len);
4676 return -EIO;
4677 }
4678
ac27a0ec 4679 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
67eeb568
DM		 4680 bh = ext4_bread(handle, inode, blk, 1, &err);
4681 if (!bh)
4682 goto out;
62d2b5f2
JK		 4683 err = ext4_journal_get_write_access(handle, bh);
4684 if (err) {
4685 brelse(bh);
4686 goto out;
ac27a0ec 4687 }
67eeb568
DM		 4688 lock_buffer(bh);
4689 memcpy(bh->b_data+offset, data, len);
4690 flush_dcache_page(bh->b_page);
4691 unlock_buffer(bh);
62d2b5f2 4692 err = ext4_handle_dirty_metadata(handle, NULL, bh);
67eeb568 4693 brelse(bh);
ac27a0ec 4694out:
67eeb568 4695 if (err) {
4d04e4fb 4696 mutex_unlock(&inode->i_mutex);
ac27a0ec 4697 return err;
4d04e4fb 4698 }
67eeb568
DM		 4699 if (inode->i_size < off + len) {
4700 i_size_write(inode, off + len);
617ba13b 4701 EXT4_I(inode)->i_disksize = inode->i_size;
ac27a0ec 4702 }
ac27a0ec 4703 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
617ba13b 4704 ext4_mark_inode_dirty(handle, inode);
ac27a0ec 4705 mutex_unlock(&inode->i_mutex);
67eeb568 4706 return len;
ac27a0ec
DK		 4707}
4708
4709#endif
4710
152a0836
AV		 4711static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
4712 const char *dev_name, void *data)
ac27a0ec 4713{
152a0836 4714 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
ac27a0ec
DK		 4715}
4716
37f328eb 4717#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
24b58424
TT		 4718static struct file_system_type ext2_fs_type = {
4719 .owner = THIS_MODULE,
4720 .name = "ext2",
152a0836 4721 .mount = ext4_mount,
24b58424
TT		 4722 .kill_sb = kill_block_super,
4723 .fs_flags = FS_REQUIRES_DEV,
4724};
4725
4726static inline void register_as_ext2(void)
4727{
4728 int err = register_filesystem(&ext2_fs_type);
4729 if (err)
4730 printk(KERN_WARNING
4731 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
4732}
4733
4734static inline void unregister_as_ext2(void)
4735{
4736 unregister_filesystem(&ext2_fs_type);
4737}
51b7e3c9 4738MODULE_ALIAS("ext2");
24b58424
TT		 4739#else
4740static inline void register_as_ext2(void) { }
4741static inline void unregister_as_ext2(void) { }
4742#endif
4743
37f328eb 4744#if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
24b58424
TT		 4745static inline void register_as_ext3(void)
4746{
4747 int err = register_filesystem(&ext3_fs_type);
4748 if (err)
4749 printk(KERN_WARNING
4750 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
4751}
4752
4753static inline void unregister_as_ext3(void)
4754{
4755 unregister_filesystem(&ext3_fs_type);
4756}
51b7e3c9 4757MODULE_ALIAS("ext3");
24b58424
TT		 4758#else
4759static inline void register_as_ext3(void) { }
4760static inline void unregister_as_ext3(void) { }
4761#endif
4762
03010a33
TT		 4763static struct file_system_type ext4_fs_type = {
4764 .owner = THIS_MODULE,
4765 .name = "ext4",
152a0836 4766 .mount = ext4_mount,
03010a33
TT		 4767 .kill_sb = kill_block_super,
4768 .fs_flags = FS_REQUIRES_DEV,
4769};
4770
8f021222 4771static int __init ext4_init_feat_adverts(void)
857ac889
LC		 4772{
4773 struct ext4_features *ef;
4774 int ret = -ENOMEM;
4775
4776 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
4777 if (!ef)
4778 goto out;
4779
4780 ef->f_kobj.kset = ext4_kset;
4781 init_completion(&ef->f_kobj_unregister);
4782 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
4783 "features");
4784 if (ret) {
4785 kfree(ef);
4786 goto out;
4787 }
4788
4789 ext4_feat = ef;
4790 ret = 0;
4791out:
4792 return ret;
4793}
4794
8f021222
LC		 4795static void ext4_exit_feat_adverts(void)
4796{
4797 kobject_put(&ext4_feat->f_kobj);
4798 wait_for_completion(&ext4_feat->f_kobj_unregister);
4799 kfree(ext4_feat);
4800}
4801
e9e3bcec
ES		 4802/* Shared across all ext4 file systems */
4803wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
4804struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
4805
5dabfc78 4806static int __init ext4_init_fs(void)
ac27a0ec 4807{
e9e3bcec 4808 int i, err;
c9de560d 4809
12e9b892 4810 ext4_check_flag_values();
e9e3bcec
ES		 4811
4812 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
4813 mutex_init(&ext4__aio_mutex[i]);
4814 init_waitqueue_head(&ext4__ioend_wq[i]);
4815 }
4816
5dabfc78 4817 err = ext4_init_pageio();
6fd058f7
TT		 4818 if (err)
4819 return err;
5dabfc78 4820 err = ext4_init_system_zone();
bd2d0210 4821 if (err)
dd68314c 4822 goto out7;
3197ebdb
TT		 4823 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
4824 if (!ext4_kset)
dd68314c 4825 goto out6;
9f6200bb 4826 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
dd68314c
TT		 4827 if (!ext4_proc_root)
4828 goto out5;
857ac889
LC		 4829
4830 err = ext4_init_feat_adverts();
dd68314c
TT		 4831 if (err)
4832 goto out4;
857ac889 4833
5dabfc78 4834 err = ext4_init_mballoc();
ac27a0ec 4835 if (err)
6fd058f7 4836 goto out3;
c9de560d 4837
5dabfc78 4838 err = ext4_init_xattr();
c9de560d
AT		 4839 if (err)
4840 goto out2;
ac27a0ec
DK		 4841 err = init_inodecache();
4842 if (err)
4843 goto out1;
24b58424
TT		 4844 register_as_ext2();
4845 register_as_ext3();
03010a33 4846 err = register_filesystem(&ext4_fs_type);
ac27a0ec
DK		 4847 if (err)
4848 goto out;
bfff6873
LC		 4849
4850 ext4_li_info = NULL;
4851 mutex_init(&ext4_li_mtx);
ac27a0ec
DK		 4852 return 0;
4853out:
24b58424
TT		 4854 unregister_as_ext2();
4855 unregister_as_ext3();
ac27a0ec
DK		 4856 destroy_inodecache();
4857out1:
5dabfc78 4858 ext4_exit_xattr();
c9de560d 4859out2:
5dabfc78 4860 ext4_exit_mballoc();
6fd058f7 4861out3:
8f021222 4862 ext4_exit_feat_adverts();
dd68314c 4863out4:
6fd058f7 4864 remove_proc_entry("fs/ext4", NULL);
dd68314c 4865out5:
6fd058f7 4866 kset_unregister(ext4_kset);
dd68314c 4867out6:
5dabfc78 4868 ext4_exit_system_zone();
dd68314c 4869out7:
5dabfc78 4870 ext4_exit_pageio();
ac27a0ec
DK		 4871 return err;
4872}
4873
5dabfc78 4874static void __exit ext4_exit_fs(void)
ac27a0ec 4875{
bfff6873 4876 ext4_destroy_lazyinit_thread();
24b58424
TT		 4877 unregister_as_ext2();
4878 unregister_as_ext3();
03010a33 4879 unregister_filesystem(&ext4_fs_type);
ac27a0ec 4880 destroy_inodecache();
5dabfc78
TT		 4881 ext4_exit_xattr();
4882 ext4_exit_mballoc();
8f021222 4883 ext4_exit_feat_adverts();
9f6200bb 4884 remove_proc_entry("fs/ext4", NULL);
3197ebdb 4885 kset_unregister(ext4_kset);
5dabfc78
TT		 4886 ext4_exit_system_zone();
4887 ext4_exit_pageio();
ac27a0ec
DK		 4888}
4889
4890MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
83982b6f 4891MODULE_DESCRIPTION("Fourth Extended Filesystem");
ac27a0ec 4892MODULE_LICENSE("GPL");
5dabfc78
TT		 4893module_init(ext4_init_fs)
4894module_exit(ext4_exit_fs)
ubuntu-bionic-kernel mirror