projects / GitHub / LineageOS / android_kernel_motorola_exynos9610.git / blame
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (incremental) | history | HEAD
ext4: unregister features interface on module unload
[GitHub/LineageOS/android_kernel_motorola_exynos9610.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;
bfff6873
LC		 57struct ext4_lazy_init *ext4_li_info;
58struct mutex ext4_li_mtx;
857ac889 59struct 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);
0b8e58a1 836
ac27a0ec
DK		 837 return &ei->vfs_inode;
838}
839
7ff9c073
TT		 840static int ext4_drop_inode(struct inode *inode)
841{
842 int drop = generic_drop_inode(inode);
843
844 trace_ext4_drop_inode(inode, drop);
845 return drop;
846}
847
fa0d7e3d
NP		 848static void ext4_i_callback(struct rcu_head *head)
849{
850 struct inode *inode = container_of(head, struct inode, i_rcu);
851 INIT_LIST_HEAD(&inode->i_dentry);
852 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
853}
854
617ba13b 855static void ext4_destroy_inode(struct inode *inode)
ac27a0ec 856{
f7ad6d2e 857 ext4_ioend_wait(inode);
9f7dd93d 858 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
b31e1552
ES		 859 ext4_msg(inode->i_sb, KERN_ERR,
860 "Inode %lu (%p): orphan list check failed!",
861 inode->i_ino, EXT4_I(inode));
9f7dd93d
VA		 862 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
863 EXT4_I(inode), sizeof(struct ext4_inode_info),
864 true);
865 dump_stack();
866 }
fa0d7e3d 867 call_rcu(&inode->i_rcu, ext4_i_callback);
ac27a0ec
DK		 868}
869
51cc5068 870static void init_once(void *foo)
ac27a0ec 871{
617ba13b 872 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
ac27a0ec 873
a35afb83 874 INIT_LIST_HEAD(&ei->i_orphan);
03010a33 875#ifdef CONFIG_EXT4_FS_XATTR
a35afb83 876 init_rwsem(&ei->xattr_sem);
ac27a0ec 877#endif
0e855ac8 878 init_rwsem(&ei->i_data_sem);
a35afb83 879 inode_init_once(&ei->vfs_inode);
ac27a0ec
DK		 880}
881
882static int init_inodecache(void)
883{
617ba13b
MC		 884 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
885 sizeof(struct ext4_inode_info),
ac27a0ec
DK		 886 0, (SLAB_RECLAIM_ACCOUNT|
887 SLAB_MEM_SPREAD),
20c2df83 888 init_once);
617ba13b 889 if (ext4_inode_cachep == NULL)
ac27a0ec
DK		 890 return -ENOMEM;
891 return 0;
892}
893
894static void destroy_inodecache(void)
895{
617ba13b 896 kmem_cache_destroy(ext4_inode_cachep);
ac27a0ec
DK		 897}
898
0930fcc1 899void ext4_clear_inode(struct inode *inode)
ac27a0ec 900{
0930fcc1
AV		 901 invalidate_inode_buffers(inode);
902 end_writeback(inode);
9f754758 903 dquot_drop(inode);
c2ea3fde 904 ext4_discard_preallocations(inode);
8aefcd55
TT		 905 if (EXT4_I(inode)->jinode) {
906 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
907 EXT4_I(inode)->jinode);
908 jbd2_free_inode(EXT4_I(inode)->jinode);
909 EXT4_I(inode)->jinode = NULL;
910 }
ac27a0ec
DK		 911}
912
2b2d6d01
TT		 913static inline void ext4_show_quota_options(struct seq_file *seq,
914 struct super_block *sb)
ac27a0ec
DK		 915{
916#if defined(CONFIG_QUOTA)
617ba13b 917 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 918
5a20bdfc
JK		 919 if (sbi->s_jquota_fmt) {
920 char *fmtname = "";
921
922 switch (sbi->s_jquota_fmt) {
923 case QFMT_VFS_OLD:
924 fmtname = "vfsold";
925 break;
926 case QFMT_VFS_V0:
927 fmtname = "vfsv0";
928 break;
929 case QFMT_VFS_V1:
930 fmtname = "vfsv1";
931 break;
932 }
933 seq_printf(seq, ",jqfmt=%s", fmtname);
934 }
ac27a0ec
DK		 935
936 if (sbi->s_qf_names[USRQUOTA])
937 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
938
939 if (sbi->s_qf_names[GRPQUOTA])
940 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
941
482a7425 942 if (test_opt(sb, USRQUOTA))
ac27a0ec
DK		 943 seq_puts(seq, ",usrquota");
944
482a7425 945 if (test_opt(sb, GRPQUOTA))
ac27a0ec
DK		 946 seq_puts(seq, ",grpquota");
947#endif
948}
949
d9c9bef1
MS		 950/*
951 * Show an option if
952 * - it's set to a non-default value OR
953 * - if the per-sb default is different from the global default
954 */
617ba13b 955static int ext4_show_options(struct seq_file *seq, struct vfsmount *vfs)
ac27a0ec 956{
aa22df2c
AK		 957 int def_errors;
958 unsigned long def_mount_opts;
ac27a0ec 959 struct super_block *sb = vfs->mnt_sb;
d9c9bef1
MS		 960 struct ext4_sb_info *sbi = EXT4_SB(sb);
961 struct ext4_super_block *es = sbi->s_es;
d9c9bef1
MS		 962
963 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
aa22df2c 964 def_errors = le16_to_cpu(es->s_errors);
d9c9bef1
MS		 965
966 if (sbi->s_sb_block != 1)
967 seq_printf(seq, ",sb=%llu", sbi->s_sb_block);
968 if (test_opt(sb, MINIX_DF))
969 seq_puts(seq, ",minixdf");
aa22df2c 970 if (test_opt(sb, GRPID) && !(def_mount_opts & EXT4_DEFM_BSDGROUPS))
d9c9bef1
MS		 971 seq_puts(seq, ",grpid");
972 if (!test_opt(sb, GRPID) && (def_mount_opts & EXT4_DEFM_BSDGROUPS))
973 seq_puts(seq, ",nogrpid");
974 if (sbi->s_resuid != EXT4_DEF_RESUID ||
975 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID) {
976 seq_printf(seq, ",resuid=%u", sbi->s_resuid);
977 }
978 if (sbi->s_resgid != EXT4_DEF_RESGID ||
979 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID) {
980 seq_printf(seq, ",resgid=%u", sbi->s_resgid);
981 }
bb4f397a 982 if (test_opt(sb, ERRORS_RO)) {
d9c9bef1 983 if (def_errors == EXT4_ERRORS_PANIC ||
bb4f397a
AK		 984 def_errors == EXT4_ERRORS_CONTINUE) {
985 seq_puts(seq, ",errors=remount-ro");
d9c9bef1
MS		 986 }
987 }
aa22df2c 988 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
bb4f397a 989 seq_puts(seq, ",errors=continue");
aa22df2c 990 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
d9c9bef1 991 seq_puts(seq, ",errors=panic");
aa22df2c 992 if (test_opt(sb, NO_UID32) && !(def_mount_opts & EXT4_DEFM_UID16))
d9c9bef1 993 seq_puts(seq, ",nouid32");
aa22df2c 994 if (test_opt(sb, DEBUG) && !(def_mount_opts & EXT4_DEFM_DEBUG))
d9c9bef1
MS		 995 seq_puts(seq, ",debug");
996 if (test_opt(sb, OLDALLOC))
997 seq_puts(seq, ",oldalloc");
03010a33 998#ifdef CONFIG_EXT4_FS_XATTR
aa22df2c
AK		 999 if (test_opt(sb, XATTR_USER) &&
1000 !(def_mount_opts & EXT4_DEFM_XATTR_USER))
d9c9bef1
MS		 1001 seq_puts(seq, ",user_xattr");
1002 if (!test_opt(sb, XATTR_USER) &&
1003 (def_mount_opts & EXT4_DEFM_XATTR_USER)) {
1004 seq_puts(seq, ",nouser_xattr");
1005 }
1006#endif
03010a33 1007#ifdef CONFIG_EXT4_FS_POSIX_ACL
aa22df2c 1008 if (test_opt(sb, POSIX_ACL) && !(def_mount_opts & EXT4_DEFM_ACL))
d9c9bef1
MS		 1009 seq_puts(seq, ",acl");
1010 if (!test_opt(sb, POSIX_ACL) && (def_mount_opts & EXT4_DEFM_ACL))
1011 seq_puts(seq, ",noacl");
1012#endif
30773840 1013 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
d9c9bef1
MS		 1014 seq_printf(seq, ",commit=%u",
1015 (unsigned) (sbi->s_commit_interval / HZ));
1016 }
30773840
TT		 1017 if (sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME) {
1018 seq_printf(seq, ",min_batch_time=%u",
1019 (unsigned) sbi->s_min_batch_time);
1020 }
1021 if (sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME) {
1022 seq_printf(seq, ",max_batch_time=%u",
1023 (unsigned) sbi->s_min_batch_time);
1024 }
1025
571640ca
ES		 1026 /*
1027 * We're changing the default of barrier mount option, so
1028 * let's always display its mount state so it's clear what its
1029 * status is.
1030 */
1031 seq_puts(seq, ",barrier=");
1032 seq_puts(seq, test_opt(sb, BARRIER) ? "1" : "0");
cd0b6a39
TT		 1033 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
1034 seq_puts(seq, ",journal_async_commit");
39a4bade
JK		 1035 else if (test_opt(sb, JOURNAL_CHECKSUM))
1036 seq_puts(seq, ",journal_checksum");
25ec56b5
JNC		 1037 if (test_opt(sb, I_VERSION))
1038 seq_puts(seq, ",i_version");
8b67f04a
TT		 1039 if (!test_opt(sb, DELALLOC) &&
1040 !(def_mount_opts & EXT4_DEFM_NODELALLOC))
dd919b98
AK		 1041 seq_puts(seq, ",nodelalloc");
1042
1449032b
TT		 1043 if (test_opt(sb, MBLK_IO_SUBMIT))
1044 seq_puts(seq, ",mblk_io_submit");
cb45bbe4
MS		 1045 if (sbi->s_stripe)
1046 seq_printf(seq, ",stripe=%lu", sbi->s_stripe);
aa22df2c
AK		 1047 /*
1048 * journal mode get enabled in different ways
1049 * So just print the value even if we didn't specify it
1050 */
617ba13b 1051 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
ac27a0ec 1052 seq_puts(seq, ",data=journal");
617ba13b 1053 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
ac27a0ec 1054 seq_puts(seq, ",data=ordered");
617ba13b 1055 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
ac27a0ec
DK		 1056 seq_puts(seq, ",data=writeback");
1057
240799cd
TT		 1058 if (sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1059 seq_printf(seq, ",inode_readahead_blks=%u",
1060 sbi->s_inode_readahead_blks);
1061
5bf5683a
HK		 1062 if (test_opt(sb, DATA_ERR_ABORT))
1063 seq_puts(seq, ",data_err=abort");
1064
afd4672d 1065 if (test_opt(sb, NO_AUTO_DA_ALLOC))
06705bff 1066 seq_puts(seq, ",noauto_da_alloc");
afd4672d 1067
8b67f04a 1068 if (test_opt(sb, DISCARD) && !(def_mount_opts & EXT4_DEFM_DISCARD))
5328e635
ES		 1069 seq_puts(seq, ",discard");
1070
e3bb52ae
ES		 1071 if (test_opt(sb, NOLOAD))
1072 seq_puts(seq, ",norecovery");
1073
744692dc
JZ		 1074 if (test_opt(sb, DIOREAD_NOLOCK))
1075 seq_puts(seq, ",dioread_nolock");
1076
8b67f04a
TT		 1077 if (test_opt(sb, BLOCK_VALIDITY) &&
1078 !(def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY))
1079 seq_puts(seq, ",block_validity");
1080
bfff6873
LC		 1081 if (!test_opt(sb, INIT_INODE_TABLE))
1082 seq_puts(seq, ",noinit_inode_table");
1083 else if (sbi->s_li_wait_mult)
1084 seq_printf(seq, ",init_inode_table=%u",
1085 (unsigned) sbi->s_li_wait_mult);
1086
617ba13b 1087 ext4_show_quota_options(seq, sb);
0b8e58a1 1088
ac27a0ec
DK		 1089 return 0;
1090}
1091
1b961ac0 1092static struct inode *ext4_nfs_get_inode(struct super_block *sb,
0b8e58a1 1093 u64 ino, u32 generation)
ac27a0ec 1094{
ac27a0ec 1095 struct inode *inode;
ac27a0ec 1096
617ba13b 1097 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
ac27a0ec 1098 return ERR_PTR(-ESTALE);
617ba13b 1099 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
ac27a0ec
DK		 1100 return ERR_PTR(-ESTALE);
1101
1102 /* iget isn't really right if the inode is currently unallocated!!
1103 *
617ba13b 1104 * ext4_read_inode will return a bad_inode if the inode had been
ac27a0ec
DK		 1105 * deleted, so we should be safe.
1106 *
1107 * Currently we don't know the generation for parent directory, so
1108 * a generation of 0 means "accept any"
1109 */
1d1fe1ee
DH		 1110 inode = ext4_iget(sb, ino);
1111 if (IS_ERR(inode))
1112 return ERR_CAST(inode);
1113 if (generation && inode->i_generation != generation) {
ac27a0ec
DK		 1114 iput(inode);
1115 return ERR_PTR(-ESTALE);
1116 }
1b961ac0
CH		 1117
1118 return inode;
1119}
1120
1121static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
0b8e58a1 1122 int fh_len, int fh_type)
1b961ac0
CH		 1123{
1124 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1125 ext4_nfs_get_inode);
1126}
1127
1128static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
0b8e58a1 1129 int fh_len, int fh_type)
1b961ac0
CH		 1130{
1131 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1132 ext4_nfs_get_inode);
ac27a0ec
DK		 1133}
1134
c39a7f84
TO		 1135/*
1136 * Try to release metadata pages (indirect blocks, directories) which are
1137 * mapped via the block device. Since these pages could have journal heads
1138 * which would prevent try_to_free_buffers() from freeing them, we must use
1139 * jbd2 layer's try_to_free_buffers() function to release them.
1140 */
0b8e58a1
AD		 1141static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1142 gfp_t wait)
c39a7f84
TO		 1143{
1144 journal_t *journal = EXT4_SB(sb)->s_journal;
1145
1146 WARN_ON(PageChecked(page));
1147 if (!page_has_buffers(page))
1148 return 0;
1149 if (journal)
1150 return jbd2_journal_try_to_free_buffers(journal, page,
1151 wait & ~__GFP_WAIT);
1152 return try_to_free_buffers(page);
1153}
1154
ac27a0ec 1155#ifdef CONFIG_QUOTA
af5bc92d 1156#define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
2b2d6d01 1157#define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
ac27a0ec 1158
617ba13b
MC		 1159static int ext4_write_dquot(struct dquot *dquot);
1160static int ext4_acquire_dquot(struct dquot *dquot);
1161static int ext4_release_dquot(struct dquot *dquot);
1162static int ext4_mark_dquot_dirty(struct dquot *dquot);
1163static int ext4_write_info(struct super_block *sb, int type);
6f28e087 1164static int ext4_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 1165 struct path *path);
ca0e05e4 1166static int ext4_quota_off(struct super_block *sb, int type);
617ba13b
MC		 1167static int ext4_quota_on_mount(struct super_block *sb, int type);
1168static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec 1169 size_t len, loff_t off);
617ba13b 1170static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 1171 const char *data, size_t len, loff_t off);
1172
61e225dc 1173static const struct dquot_operations ext4_quota_operations = {
a9e7f447 1174#ifdef CONFIG_QUOTA
60e58e0f 1175 .get_reserved_space = ext4_get_reserved_space,
a9e7f447 1176#endif
617ba13b
MC		 1177 .write_dquot = ext4_write_dquot,
1178 .acquire_dquot = ext4_acquire_dquot,
1179 .release_dquot = ext4_release_dquot,
1180 .mark_dirty = ext4_mark_dquot_dirty,
a5b5ee32
JK		 1181 .write_info = ext4_write_info,
1182 .alloc_dquot = dquot_alloc,
1183 .destroy_dquot = dquot_destroy,
ac27a0ec
DK		 1184};
1185
0d54b217 1186static const struct quotactl_ops ext4_qctl_operations = {
617ba13b 1187 .quota_on = ext4_quota_on,
ca0e05e4 1188 .quota_off = ext4_quota_off,
287a8095
CH		 1189 .quota_sync = dquot_quota_sync,
1190 .get_info = dquot_get_dqinfo,
1191 .set_info = dquot_set_dqinfo,
1192 .get_dqblk = dquot_get_dqblk,
1193 .set_dqblk = dquot_set_dqblk
ac27a0ec
DK		 1194};
1195#endif
1196
ee9b6d61 1197static const struct super_operations ext4_sops = {
617ba13b
MC		 1198 .alloc_inode = ext4_alloc_inode,
1199 .destroy_inode = ext4_destroy_inode,
617ba13b
MC		 1200 .write_inode = ext4_write_inode,
1201 .dirty_inode = ext4_dirty_inode,
7ff9c073 1202 .drop_inode = ext4_drop_inode,
0930fcc1 1203 .evict_inode = ext4_evict_inode,
617ba13b 1204 .put_super = ext4_put_super,
617ba13b 1205 .sync_fs = ext4_sync_fs,
c4be0c1d
TS		 1206 .freeze_fs = ext4_freeze,
1207 .unfreeze_fs = ext4_unfreeze,
617ba13b
MC		 1208 .statfs = ext4_statfs,
1209 .remount_fs = ext4_remount,
617ba13b 1210 .show_options = ext4_show_options,
ac27a0ec 1211#ifdef CONFIG_QUOTA
617ba13b
MC		 1212 .quota_read = ext4_quota_read,
1213 .quota_write = ext4_quota_write,
ac27a0ec 1214#endif
c39a7f84 1215 .bdev_try_to_free_page = bdev_try_to_free_page,
ac27a0ec
DK		 1216};
1217
9ca92389
TT		 1218static const struct super_operations ext4_nojournal_sops = {
1219 .alloc_inode = ext4_alloc_inode,
1220 .destroy_inode = ext4_destroy_inode,
1221 .write_inode = ext4_write_inode,
1222 .dirty_inode = ext4_dirty_inode,
7ff9c073 1223 .drop_inode = ext4_drop_inode,
0930fcc1 1224 .evict_inode = ext4_evict_inode,
9ca92389
TT		 1225 .write_super = ext4_write_super,
1226 .put_super = ext4_put_super,
1227 .statfs = ext4_statfs,
1228 .remount_fs = ext4_remount,
9ca92389
TT		 1229 .show_options = ext4_show_options,
1230#ifdef CONFIG_QUOTA
1231 .quota_read = ext4_quota_read,
1232 .quota_write = ext4_quota_write,
1233#endif
1234 .bdev_try_to_free_page = bdev_try_to_free_page,
1235};
1236
39655164 1237static const struct export_operations ext4_export_ops = {
1b961ac0
CH		 1238 .fh_to_dentry = ext4_fh_to_dentry,
1239 .fh_to_parent = ext4_fh_to_parent,
617ba13b 1240 .get_parent = ext4_get_parent,
ac27a0ec
DK		 1241};
1242
1243enum {
1244 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1245 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
01436ef2 1246 Opt_nouid32, Opt_debug, Opt_oldalloc, Opt_orlov,
ac27a0ec 1247 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
06705bff 1248 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload, Opt_nobh, Opt_bh,
30773840 1249 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
c3191067 1250 Opt_journal_update, Opt_journal_dev,
818d276c 1251 Opt_journal_checksum, Opt_journal_async_commit,
ac27a0ec 1252 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
296c355c 1253 Opt_data_err_abort, Opt_data_err_ignore,
ac27a0ec 1254 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
5a20bdfc
JK		 1255 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1256 Opt_noquota, Opt_ignore, Opt_barrier, Opt_nobarrier, Opt_err,
1257 Opt_resize, Opt_usrquota, Opt_grpquota, Opt_i_version,
1449032b
TT		 1258 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1259 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
5328e635 1260 Opt_inode_readahead_blks, Opt_journal_ioprio,
744692dc 1261 Opt_dioread_nolock, Opt_dioread_lock,
5328e635 1262 Opt_discard, Opt_nodiscard,
bfff6873 1263 Opt_init_inode_table, Opt_noinit_inode_table,
ac27a0ec
DK		 1264};
1265
a447c093 1266static const match_table_t tokens = {
ac27a0ec
DK		 1267 {Opt_bsd_df, "bsddf"},
1268 {Opt_minix_df, "minixdf"},
1269 {Opt_grpid, "grpid"},
1270 {Opt_grpid, "bsdgroups"},
1271 {Opt_nogrpid, "nogrpid"},
1272 {Opt_nogrpid, "sysvgroups"},
1273 {Opt_resgid, "resgid=%u"},
1274 {Opt_resuid, "resuid=%u"},
1275 {Opt_sb, "sb=%u"},
1276 {Opt_err_cont, "errors=continue"},
1277 {Opt_err_panic, "errors=panic"},
1278 {Opt_err_ro, "errors=remount-ro"},
1279 {Opt_nouid32, "nouid32"},
ac27a0ec
DK		 1280 {Opt_debug, "debug"},
1281 {Opt_oldalloc, "oldalloc"},
1282 {Opt_orlov, "orlov"},
1283 {Opt_user_xattr, "user_xattr"},
1284 {Opt_nouser_xattr, "nouser_xattr"},
1285 {Opt_acl, "acl"},
1286 {Opt_noacl, "noacl"},
ac27a0ec 1287 {Opt_noload, "noload"},
e3bb52ae 1288 {Opt_noload, "norecovery"},
ac27a0ec
DK		 1289 {Opt_nobh, "nobh"},
1290 {Opt_bh, "bh"},
1291 {Opt_commit, "commit=%u"},
30773840
TT		 1292 {Opt_min_batch_time, "min_batch_time=%u"},
1293 {Opt_max_batch_time, "max_batch_time=%u"},
ac27a0ec 1294 {Opt_journal_update, "journal=update"},
ac27a0ec 1295 {Opt_journal_dev, "journal_dev=%u"},
818d276c
GS		 1296 {Opt_journal_checksum, "journal_checksum"},
1297 {Opt_journal_async_commit, "journal_async_commit"},
ac27a0ec
DK		 1298 {Opt_abort, "abort"},
1299 {Opt_data_journal, "data=journal"},
1300 {Opt_data_ordered, "data=ordered"},
1301 {Opt_data_writeback, "data=writeback"},
5bf5683a
HK		 1302 {Opt_data_err_abort, "data_err=abort"},
1303 {Opt_data_err_ignore, "data_err=ignore"},
ac27a0ec
DK		 1304 {Opt_offusrjquota, "usrjquota="},
1305 {Opt_usrjquota, "usrjquota=%s"},
1306 {Opt_offgrpjquota, "grpjquota="},
1307 {Opt_grpjquota, "grpjquota=%s"},
1308 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1309 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
5a20bdfc 1310 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
ac27a0ec
DK		 1311 {Opt_grpquota, "grpquota"},
1312 {Opt_noquota, "noquota"},
1313 {Opt_quota, "quota"},
1314 {Opt_usrquota, "usrquota"},
1315 {Opt_barrier, "barrier=%u"},
06705bff
TT		 1316 {Opt_barrier, "barrier"},
1317 {Opt_nobarrier, "nobarrier"},
25ec56b5 1318 {Opt_i_version, "i_version"},
c9de560d 1319 {Opt_stripe, "stripe=%u"},
ac27a0ec 1320 {Opt_resize, "resize"},
64769240 1321 {Opt_delalloc, "delalloc"},
dd919b98 1322 {Opt_nodelalloc, "nodelalloc"},
1449032b
TT		 1323 {Opt_mblk_io_submit, "mblk_io_submit"},
1324 {Opt_nomblk_io_submit, "nomblk_io_submit"},
6fd058f7
TT		 1325 {Opt_block_validity, "block_validity"},
1326 {Opt_noblock_validity, "noblock_validity"},
240799cd 1327 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
b3881f74 1328 {Opt_journal_ioprio, "journal_ioprio=%u"},
afd4672d 1329 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
06705bff
TT		 1330 {Opt_auto_da_alloc, "auto_da_alloc"},
1331 {Opt_noauto_da_alloc, "noauto_da_alloc"},
744692dc
JZ		 1332 {Opt_dioread_nolock, "dioread_nolock"},
1333 {Opt_dioread_lock, "dioread_lock"},
5328e635
ES		 1334 {Opt_discard, "discard"},
1335 {Opt_nodiscard, "nodiscard"},
bfff6873
LC		 1336 {Opt_init_inode_table, "init_itable=%u"},
1337 {Opt_init_inode_table, "init_itable"},
1338 {Opt_noinit_inode_table, "noinit_itable"},
f3f12faa 1339 {Opt_err, NULL},
ac27a0ec
DK		 1340};
1341
617ba13b 1342static ext4_fsblk_t get_sb_block(void **data)
ac27a0ec 1343{
617ba13b 1344 ext4_fsblk_t sb_block;
ac27a0ec
DK		 1345 char *options = (char *) *data;
1346
1347 if (!options || strncmp(options, "sb=", 3) != 0)
1348 return 1; /* Default location */
0b8e58a1 1349
ac27a0ec 1350 options += 3;
0b8e58a1 1351 /* TODO: use simple_strtoll with >32bit ext4 */
ac27a0ec
DK		 1352 sb_block = simple_strtoul(options, &options, 0);
1353 if (*options && *options != ',') {
4776004f 1354 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
ac27a0ec
DK		 1355 (char *) *data);
1356 return 1;
1357 }
1358 if (*options == ',')
1359 options++;
1360 *data = (void *) options;
0b8e58a1 1361
ac27a0ec
DK		 1362 return sb_block;
1363}
1364
b3881f74 1365#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
437ca0fd
DM		 1366static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1367 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
b3881f74 1368
56c50f11
DM		 1369#ifdef CONFIG_QUOTA
1370static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1371{
1372 struct ext4_sb_info *sbi = EXT4_SB(sb);
1373 char *qname;
1374
1375 if (sb_any_quota_loaded(sb) &&
1376 !sbi->s_qf_names[qtype]) {
1377 ext4_msg(sb, KERN_ERR,
1378 "Cannot change journaled "
1379 "quota options when quota turned on");
1380 return 0;
1381 }
1382 qname = match_strdup(args);
1383 if (!qname) {
1384 ext4_msg(sb, KERN_ERR,
1385 "Not enough memory for storing quotafile name");
1386 return 0;
1387 }
1388 if (sbi->s_qf_names[qtype] &&
1389 strcmp(sbi->s_qf_names[qtype], qname)) {
1390 ext4_msg(sb, KERN_ERR,
1391 "%s quota file already specified", QTYPE2NAME(qtype));
1392 kfree(qname);
1393 return 0;
1394 }
1395 sbi->s_qf_names[qtype] = qname;
1396 if (strchr(sbi->s_qf_names[qtype], '/')) {
1397 ext4_msg(sb, KERN_ERR,
1398 "quotafile must be on filesystem root");
1399 kfree(sbi->s_qf_names[qtype]);
1400 sbi->s_qf_names[qtype] = NULL;
1401 return 0;
1402 }
fd8c37ec 1403 set_opt(sb, QUOTA);
56c50f11
DM		 1404 return 1;
1405}
1406
1407static int clear_qf_name(struct super_block *sb, int qtype)
1408{
1409
1410 struct ext4_sb_info *sbi = EXT4_SB(sb);
1411
1412 if (sb_any_quota_loaded(sb) &&
1413 sbi->s_qf_names[qtype]) {
1414 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1415 " when quota turned on");
1416 return 0;
1417 }
1418 /*
1419 * The space will be released later when all options are confirmed
1420 * to be correct
1421 */
1422 sbi->s_qf_names[qtype] = NULL;
1423 return 1;
1424}
1425#endif
1426
2b2d6d01 1427static int parse_options(char *options, struct super_block *sb,
c3191067 1428 unsigned long *journal_devnum,
b3881f74 1429 unsigned int *journal_ioprio,
2b2d6d01 1430 ext4_fsblk_t *n_blocks_count, int is_remount)
ac27a0ec 1431{
617ba13b 1432 struct ext4_sb_info *sbi = EXT4_SB(sb);
2b2d6d01 1433 char *p;
ac27a0ec
DK		 1434 substring_t args[MAX_OPT_ARGS];
1435 int data_opt = 0;
1436 int option;
1437#ifdef CONFIG_QUOTA
56c50f11 1438 int qfmt;
ac27a0ec
DK		 1439#endif
1440
1441 if (!options)
1442 return 1;
1443
2b2d6d01 1444 while ((p = strsep(&options, ",")) != NULL) {
ac27a0ec
DK		 1445 int token;
1446 if (!*p)
1447 continue;
1448
15121c18
ES		 1449 /*
1450 * Initialize args struct so we know whether arg was
1451 * found; some options take optional arguments.
1452 */
1453 args[0].to = args[0].from = 0;
ac27a0ec
DK		 1454 token = match_token(p, tokens, args);
1455 switch (token) {
1456 case Opt_bsd_df:
437ca0fd 1457 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
fd8c37ec 1458 clear_opt(sb, MINIX_DF);
ac27a0ec
DK		 1459 break;
1460 case Opt_minix_df:
437ca0fd 1461 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
fd8c37ec 1462 set_opt(sb, MINIX_DF);
437ca0fd 1463
ac27a0ec
DK		 1464 break;
1465 case Opt_grpid:
437ca0fd 1466 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
fd8c37ec 1467 set_opt(sb, GRPID);
437ca0fd 1468
ac27a0ec
DK		 1469 break;
1470 case Opt_nogrpid:
437ca0fd 1471 ext4_msg(sb, KERN_WARNING, deprecated_msg, p, "2.6.38");
fd8c37ec 1472 clear_opt(sb, GRPID);
437ca0fd 1473
ac27a0ec
DK		 1474 break;
1475 case Opt_resuid:
1476 if (match_int(&args[0], &option))
1477 return 0;
1478 sbi->s_resuid = option;
1479 break;
1480 case Opt_resgid:
1481 if (match_int(&args[0], &option))
1482 return 0;
1483 sbi->s_resgid = option;
1484 break;
1485 case Opt_sb:
1486 /* handled by get_sb_block() instead of here */
1487 /* *sb_block = match_int(&args[0]); */
1488 break;
1489 case Opt_err_panic:
fd8c37ec
TT		 1490 clear_opt(sb, ERRORS_CONT);
1491 clear_opt(sb, ERRORS_RO);
1492 set_opt(sb, ERRORS_PANIC);
ac27a0ec
DK		 1493 break;
1494 case Opt_err_ro:
fd8c37ec
TT		 1495 clear_opt(sb, ERRORS_CONT);
1496 clear_opt(sb, ERRORS_PANIC);
1497 set_opt(sb, ERRORS_RO);
ac27a0ec
DK		 1498 break;
1499 case Opt_err_cont:
fd8c37ec
TT		 1500 clear_opt(sb, ERRORS_RO);
1501 clear_opt(sb, ERRORS_PANIC);
1502 set_opt(sb, ERRORS_CONT);
ac27a0ec
DK		 1503 break;
1504 case Opt_nouid32:
fd8c37ec 1505 set_opt(sb, NO_UID32);
ac27a0ec 1506 break;
ac27a0ec 1507 case Opt_debug:
fd8c37ec 1508 set_opt(sb, DEBUG);
ac27a0ec
DK		 1509 break;
1510 case Opt_oldalloc:
fd8c37ec 1511 set_opt(sb, OLDALLOC);
ac27a0ec
DK		 1512 break;
1513 case Opt_orlov:
fd8c37ec 1514 clear_opt(sb, OLDALLOC);
ac27a0ec 1515 break;
03010a33 1516#ifdef CONFIG_EXT4_FS_XATTR
ac27a0ec 1517 case Opt_user_xattr:
fd8c37ec 1518 set_opt(sb, XATTR_USER);
ac27a0ec
DK		 1519 break;
1520 case Opt_nouser_xattr:
fd8c37ec 1521 clear_opt(sb, XATTR_USER);
ac27a0ec
DK		 1522 break;
1523#else
1524 case Opt_user_xattr:
1525 case Opt_nouser_xattr:
b31e1552 1526 ext4_msg(sb, KERN_ERR, "(no)user_xattr options not supported");
ac27a0ec
DK		 1527 break;
1528#endif
03010a33 1529#ifdef CONFIG_EXT4_FS_POSIX_ACL
ac27a0ec 1530 case Opt_acl:
fd8c37ec 1531 set_opt(sb, POSIX_ACL);
ac27a0ec
DK		 1532 break;
1533 case Opt_noacl:
fd8c37ec 1534 clear_opt(sb, POSIX_ACL);
ac27a0ec
DK		 1535 break;
1536#else
1537 case Opt_acl:
1538 case Opt_noacl:
b31e1552 1539 ext4_msg(sb, KERN_ERR, "(no)acl options not supported");
ac27a0ec
DK		 1540 break;
1541#endif
ac27a0ec
DK		 1542 case Opt_journal_update:
1543 /* @@@ FIXME */
1544 /* Eventually we will want to be able to create
1545 a journal file here. For now, only allow the
1546 user to specify an existing inode to be the
1547 journal file. */
1548 if (is_remount) {
b31e1552
ES		 1549 ext4_msg(sb, KERN_ERR,
1550 "Cannot specify journal on remount");
ac27a0ec
DK		 1551 return 0;
1552 }
fd8c37ec 1553 set_opt(sb, UPDATE_JOURNAL);
ac27a0ec 1554 break;
ac27a0ec
DK		 1555 case Opt_journal_dev:
1556 if (is_remount) {
b31e1552
ES		 1557 ext4_msg(sb, KERN_ERR,
1558 "Cannot specify journal on remount");
ac27a0ec
DK		 1559 return 0;
1560 }
1561 if (match_int(&args[0], &option))
1562 return 0;
1563 *journal_devnum = option;
1564 break;
818d276c 1565 case Opt_journal_checksum:
fd8c37ec 1566 set_opt(sb, JOURNAL_CHECKSUM);
d4da6c9c 1567 break;
818d276c 1568 case Opt_journal_async_commit:
fd8c37ec
TT		 1569 set_opt(sb, JOURNAL_ASYNC_COMMIT);
1570 set_opt(sb, JOURNAL_CHECKSUM);
818d276c 1571 break;
ac27a0ec 1572 case Opt_noload:
fd8c37ec 1573 set_opt(sb, NOLOAD);
ac27a0ec
DK		 1574 break;
1575 case Opt_commit:
1576 if (match_int(&args[0], &option))
1577 return 0;
1578 if (option < 0)
1579 return 0;
1580 if (option == 0)
cd02ff0b 1581 option = JBD2_DEFAULT_MAX_COMMIT_AGE;
ac27a0ec
DK		 1582 sbi->s_commit_interval = HZ * option;
1583 break;
30773840
TT		 1584 case Opt_max_batch_time:
1585 if (match_int(&args[0], &option))
1586 return 0;
1587 if (option < 0)
1588 return 0;
1589 if (option == 0)
1590 option = EXT4_DEF_MAX_BATCH_TIME;
1591 sbi->s_max_batch_time = option;
1592 break;
1593 case Opt_min_batch_time:
1594 if (match_int(&args[0], &option))
1595 return 0;
1596 if (option < 0)
1597 return 0;
1598 sbi->s_min_batch_time = option;
1599 break;
ac27a0ec 1600 case Opt_data_journal:
617ba13b 1601 data_opt = EXT4_MOUNT_JOURNAL_DATA;
ac27a0ec
DK		 1602 goto datacheck;
1603 case Opt_data_ordered:
617ba13b 1604 data_opt = EXT4_MOUNT_ORDERED_DATA;
ac27a0ec
DK		 1605 goto datacheck;
1606 case Opt_data_writeback:
617ba13b 1607 data_opt = EXT4_MOUNT_WRITEBACK_DATA;
ac27a0ec
DK		 1608 datacheck:
1609 if (is_remount) {
482a7425 1610 if (test_opt(sb, DATA_FLAGS) != data_opt) {
b31e1552
ES		 1611 ext4_msg(sb, KERN_ERR,
1612 "Cannot change data mode on remount");
ac27a0ec
DK		 1613 return 0;
1614 }
1615 } else {
fd8c37ec 1616 clear_opt(sb, DATA_FLAGS);
ac27a0ec
DK		 1617 sbi->s_mount_opt |= data_opt;
1618 }
1619 break;
5bf5683a 1620 case Opt_data_err_abort:
fd8c37ec 1621 set_opt(sb, DATA_ERR_ABORT);
5bf5683a
HK		 1622 break;
1623 case Opt_data_err_ignore:
fd8c37ec 1624 clear_opt(sb, DATA_ERR_ABORT);
5bf5683a 1625 break;
ac27a0ec
DK		 1626#ifdef CONFIG_QUOTA
1627 case Opt_usrjquota:
56c50f11 1628 if (!set_qf_name(sb, USRQUOTA, &args[0]))
ac27a0ec 1629 return 0;
56c50f11
DM		 1630 break;
1631 case Opt_grpjquota:
1632 if (!set_qf_name(sb, GRPQUOTA, &args[0]))
ac27a0ec 1633 return 0;
ac27a0ec
DK		 1634 break;
1635 case Opt_offusrjquota:
56c50f11
DM		 1636 if (!clear_qf_name(sb, USRQUOTA))
1637 return 0;
1638 break;
ac27a0ec 1639 case Opt_offgrpjquota:
56c50f11 1640 if (!clear_qf_name(sb, GRPQUOTA))
ac27a0ec 1641 return 0;
ac27a0ec 1642 break;
56c50f11 1643
ac27a0ec 1644 case Opt_jqfmt_vfsold:
dfc5d03f
JK		 1645 qfmt = QFMT_VFS_OLD;
1646 goto set_qf_format;
ac27a0ec 1647 case Opt_jqfmt_vfsv0:
dfc5d03f 1648 qfmt = QFMT_VFS_V0;
5a20bdfc
JK		 1649 goto set_qf_format;
1650 case Opt_jqfmt_vfsv1:
1651 qfmt = QFMT_VFS_V1;
dfc5d03f 1652set_qf_format:
17bd13b3 1653 if (sb_any_quota_loaded(sb) &&
dfc5d03f 1654 sbi->s_jquota_fmt != qfmt) {
b31e1552 1655 ext4_msg(sb, KERN_ERR, "Cannot change "
dfc5d03f 1656 "journaled quota options when "
b31e1552 1657 "quota turned on");
dfc5d03f
JK		 1658 return 0;
1659 }
1660 sbi->s_jquota_fmt = qfmt;
ac27a0ec
DK		 1661 break;
1662 case Opt_quota:
1663 case Opt_usrquota:
fd8c37ec
TT		 1664 set_opt(sb, QUOTA);
1665 set_opt(sb, USRQUOTA);
ac27a0ec
DK		 1666 break;
1667 case Opt_grpquota:
fd8c37ec
TT		 1668 set_opt(sb, QUOTA);
1669 set_opt(sb, GRPQUOTA);
ac27a0ec
DK		 1670 break;
1671 case Opt_noquota:
17bd13b3 1672 if (sb_any_quota_loaded(sb)) {
b31e1552
ES		 1673 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1674 "options when quota turned on");
ac27a0ec
DK		 1675 return 0;
1676 }
fd8c37ec
TT		 1677 clear_opt(sb, QUOTA);
1678 clear_opt(sb, USRQUOTA);
1679 clear_opt(sb, GRPQUOTA);
ac27a0ec
DK		 1680 break;
1681#else
1682 case Opt_quota:
1683 case Opt_usrquota:
1684 case Opt_grpquota:
b31e1552
ES		 1685 ext4_msg(sb, KERN_ERR,
1686 "quota options not supported");
cd59e7b9 1687 break;
ac27a0ec
DK		 1688 case Opt_usrjquota:
1689 case Opt_grpjquota:
1690 case Opt_offusrjquota:
1691 case Opt_offgrpjquota:
1692 case Opt_jqfmt_vfsold:
1693 case Opt_jqfmt_vfsv0:
5a20bdfc 1694 case Opt_jqfmt_vfsv1:
b31e1552
ES		 1695 ext4_msg(sb, KERN_ERR,
1696 "journaled quota options not supported");
ac27a0ec
DK		 1697 break;
1698 case Opt_noquota:
1699 break;
1700#endif
1701 case Opt_abort:
4ab2f15b 1702 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
ac27a0ec 1703 break;
06705bff 1704 case Opt_nobarrier:
fd8c37ec 1705 clear_opt(sb, BARRIER);
06705bff 1706 break;
ac27a0ec 1707 case Opt_barrier:
15121c18
ES		 1708 if (args[0].from) {
1709 if (match_int(&args[0], &option))
1710 return 0;
1711 } else
1712 option = 1; /* No argument, default to 1 */
ac27a0ec 1713 if (option)
fd8c37ec 1714 set_opt(sb, BARRIER);
ac27a0ec 1715 else
fd8c37ec 1716 clear_opt(sb, BARRIER);
ac27a0ec
DK		 1717 break;
1718 case Opt_ignore:
1719 break;
1720 case Opt_resize:
1721 if (!is_remount) {
b31e1552
ES		 1722 ext4_msg(sb, KERN_ERR,
1723 "resize option only available "
1724 "for remount");
ac27a0ec
DK		 1725 return 0;
1726 }
1727 if (match_int(&args[0], &option) != 0)
1728 return 0;
1729 *n_blocks_count = option;
1730 break;
1731 case Opt_nobh:
206f7ab4
CH		 1732 ext4_msg(sb, KERN_WARNING,
1733 "Ignoring deprecated nobh option");
ac27a0ec
DK		 1734 break;
1735 case Opt_bh:
206f7ab4
CH		 1736 ext4_msg(sb, KERN_WARNING,
1737 "Ignoring deprecated bh option");
ac27a0ec 1738 break;
25ec56b5 1739 case Opt_i_version:
fd8c37ec 1740 set_opt(sb, I_VERSION);
25ec56b5
JNC		 1741 sb->s_flags |= MS_I_VERSION;
1742 break;
dd919b98 1743 case Opt_nodelalloc:
fd8c37ec 1744 clear_opt(sb, DELALLOC);
dd919b98 1745 break;
1449032b 1746 case Opt_mblk_io_submit:
fd8c37ec 1747 set_opt(sb, MBLK_IO_SUBMIT);
1449032b
TT		 1748 break;
1749 case Opt_nomblk_io_submit:
fd8c37ec 1750 clear_opt(sb, MBLK_IO_SUBMIT);
1449032b 1751 break;
c9de560d
AT		 1752 case Opt_stripe:
1753 if (match_int(&args[0], &option))
1754 return 0;
1755 if (option < 0)
1756 return 0;
1757 sbi->s_stripe = option;
1758 break;
64769240 1759 case Opt_delalloc:
fd8c37ec 1760 set_opt(sb, DELALLOC);
64769240 1761 break;
6fd058f7 1762 case Opt_block_validity:
fd8c37ec 1763 set_opt(sb, BLOCK_VALIDITY);
6fd058f7
TT		 1764 break;
1765 case Opt_noblock_validity:
fd8c37ec 1766 clear_opt(sb, BLOCK_VALIDITY);
6fd058f7 1767 break;
240799cd
TT		 1768 case Opt_inode_readahead_blks:
1769 if (match_int(&args[0], &option))
1770 return 0;
1771 if (option < 0 || option > (1 << 30))
1772 return 0;
f7c43950 1773 if (!is_power_of_2(option)) {
b31e1552
ES		 1774 ext4_msg(sb, KERN_ERR,
1775 "EXT4-fs: inode_readahead_blks"
1776 " must be a power of 2");
3197ebdb
TT		 1777 return 0;
1778 }
240799cd
TT		 1779 sbi->s_inode_readahead_blks = option;
1780 break;
b3881f74
TT		 1781 case Opt_journal_ioprio:
1782 if (match_int(&args[0], &option))
1783 return 0;
1784 if (option < 0 || option > 7)
1785 break;
1786 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE,
1787 option);
1788 break;
06705bff 1789 case Opt_noauto_da_alloc:
fd8c37ec 1790 set_opt(sb, NO_AUTO_DA_ALLOC);
06705bff 1791 break;
afd4672d 1792 case Opt_auto_da_alloc:
15121c18
ES		 1793 if (args[0].from) {
1794 if (match_int(&args[0], &option))
1795 return 0;
1796 } else
1797 option = 1; /* No argument, default to 1 */
afd4672d 1798 if (option)
fd8c37ec 1799 clear_opt(sb, NO_AUTO_DA_ALLOC);
afd4672d 1800 else
fd8c37ec 1801 set_opt(sb,NO_AUTO_DA_ALLOC);
afd4672d 1802 break;
5328e635 1803 case Opt_discard:
fd8c37ec 1804 set_opt(sb, DISCARD);
5328e635
ES		 1805 break;
1806 case Opt_nodiscard:
fd8c37ec 1807 clear_opt(sb, DISCARD);
5328e635 1808 break;
744692dc 1809 case Opt_dioread_nolock:
fd8c37ec 1810 set_opt(sb, DIOREAD_NOLOCK);
744692dc
JZ		 1811 break;
1812 case Opt_dioread_lock:
fd8c37ec 1813 clear_opt(sb, DIOREAD_NOLOCK);
744692dc 1814 break;
bfff6873 1815 case Opt_init_inode_table:
fd8c37ec 1816 set_opt(sb, INIT_INODE_TABLE);
bfff6873
LC		 1817 if (args[0].from) {
1818 if (match_int(&args[0], &option))
1819 return 0;
1820 } else
1821 option = EXT4_DEF_LI_WAIT_MULT;
1822 if (option < 0)
1823 return 0;
1824 sbi->s_li_wait_mult = option;
1825 break;
1826 case Opt_noinit_inode_table:
fd8c37ec 1827 clear_opt(sb, INIT_INODE_TABLE);
bfff6873 1828 break;
ac27a0ec 1829 default:
b31e1552
ES		 1830 ext4_msg(sb, KERN_ERR,
1831 "Unrecognized mount option \"%s\" "
1832 "or missing value", p);
ac27a0ec
DK		 1833 return 0;
1834 }
1835 }
1836#ifdef CONFIG_QUOTA
1837 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
482a7425 1838 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
fd8c37ec 1839 clear_opt(sb, USRQUOTA);
ac27a0ec 1840
482a7425 1841 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
fd8c37ec 1842 clear_opt(sb, GRPQUOTA);
ac27a0ec 1843
56c50f11 1844 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
b31e1552
ES		 1845 ext4_msg(sb, KERN_ERR, "old and new quota "
1846 "format mixing");
ac27a0ec
DK		 1847 return 0;
1848 }
1849
1850 if (!sbi->s_jquota_fmt) {
b31e1552
ES		 1851 ext4_msg(sb, KERN_ERR, "journaled quota format "
1852 "not specified");
ac27a0ec
DK		 1853 return 0;
1854 }
1855 } else {
1856 if (sbi->s_jquota_fmt) {
b31e1552 1857 ext4_msg(sb, KERN_ERR, "journaled quota format "
2c8be6b2 1858 "specified with no journaling "
b31e1552 1859 "enabled");
ac27a0ec
DK		 1860 return 0;
1861 }
1862 }
1863#endif
1864 return 1;
1865}
1866
617ba13b 1867static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
ac27a0ec
DK		 1868 int read_only)
1869{
617ba13b 1870 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec
DK		 1871 int res = 0;
1872
617ba13b 1873 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
b31e1552
ES		 1874 ext4_msg(sb, KERN_ERR, "revision level too high, "
1875 "forcing read-only mode");
ac27a0ec
DK		 1876 res = MS_RDONLY;
1877 }
1878 if (read_only)
1879 return res;
617ba13b 1880 if (!(sbi->s_mount_state & EXT4_VALID_FS))
b31e1552
ES		 1881 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1882 "running e2fsck is recommended");
617ba13b 1883 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
b31e1552
ES		 1884 ext4_msg(sb, KERN_WARNING,
1885 "warning: mounting fs with errors, "
1886 "running e2fsck is recommended");
ac27a0ec
DK		 1887 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) >= 0 &&
1888 le16_to_cpu(es->s_mnt_count) >=
1889 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
b31e1552
ES		 1890 ext4_msg(sb, KERN_WARNING,
1891 "warning: maximal mount count reached, "
1892 "running e2fsck is recommended");
ac27a0ec
DK		 1893 else if (le32_to_cpu(es->s_checkinterval) &&
1894 (le32_to_cpu(es->s_lastcheck) +
1895 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
b31e1552
ES		 1896 ext4_msg(sb, KERN_WARNING,
1897 "warning: checktime reached, "
1898 "running e2fsck is recommended");
0b8e58a1 1899 if (!sbi->s_journal)
0390131b 1900 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
ac27a0ec 1901 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
617ba13b 1902 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
e8546d06 1903 le16_add_cpu(&es->s_mnt_count, 1);
ac27a0ec 1904 es->s_mtime = cpu_to_le32(get_seconds());
617ba13b 1905 ext4_update_dynamic_rev(sb);
0390131b
FM		 1906 if (sbi->s_journal)
1907 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
ac27a0ec 1908
e2d67052 1909 ext4_commit_super(sb, 1);
ac27a0ec 1910 if (test_opt(sb, DEBUG))
a9df9a49 1911 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
a2595b8a 1912 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
ac27a0ec
DK		 1913 sb->s_blocksize,
1914 sbi->s_groups_count,
617ba13b
MC		 1915 EXT4_BLOCKS_PER_GROUP(sb),
1916 EXT4_INODES_PER_GROUP(sb),
a2595b8a 1917 sbi->s_mount_opt, sbi->s_mount_opt2);
ac27a0ec 1918
ac27a0ec
DK		 1919 return res;
1920}
1921
772cb7c8
JS		 1922static int ext4_fill_flex_info(struct super_block *sb)
1923{
1924 struct ext4_sb_info *sbi = EXT4_SB(sb);
1925 struct ext4_group_desc *gdp = NULL;
772cb7c8
JS		 1926 ext4_group_t flex_group_count;
1927 ext4_group_t flex_group;
1928 int groups_per_flex = 0;
c5ca7c76 1929 size_t size;
772cb7c8
JS		 1930 int i;
1931
503358ae
TT		 1932 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1933 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1934
1935 if (groups_per_flex < 2) {
772cb7c8
JS		 1936 sbi->s_log_groups_per_flex = 0;
1937 return 1;
1938 }
1939
c62a11fd
FB		 1940 /* We allocate both existing and potentially added groups */
1941 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
d94e99a6
AK		 1942 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1943 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
c5ca7c76
TT		 1944 size = flex_group_count * sizeof(struct flex_groups);
1945 sbi->s_flex_groups = kzalloc(size, GFP_KERNEL);
1946 if (sbi->s_flex_groups == NULL) {
94de56ab
JP		 1947 sbi->s_flex_groups = vzalloc(size);
1948 if (sbi->s_flex_groups == NULL) {
1949 ext4_msg(sb, KERN_ERR,
1950 "not enough memory for %u flex groups",
1951 flex_group_count);
1952 goto failed;
1953 }
772cb7c8 1954 }
772cb7c8 1955
772cb7c8 1956 for (i = 0; i < sbi->s_groups_count; i++) {
88b6edd1 1957 gdp = ext4_get_group_desc(sb, i, NULL);
772cb7c8
JS		 1958
1959 flex_group = ext4_flex_group(sbi, i);
7ad9bb65
TT		 1960 atomic_add(ext4_free_inodes_count(sb, gdp),
1961 &sbi->s_flex_groups[flex_group].free_inodes);
1962 atomic_add(ext4_free_blks_count(sb, gdp),
1963 &sbi->s_flex_groups[flex_group].free_blocks);
1964 atomic_add(ext4_used_dirs_count(sb, gdp),
1965 &sbi->s_flex_groups[flex_group].used_dirs);
772cb7c8
JS		 1966 }
1967
1968 return 1;
1969failed:
1970 return 0;
1971}
1972
717d50e4
AD		 1973__le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1974 struct ext4_group_desc *gdp)
1975{
1976 __u16 crc = 0;
1977
1978 if (sbi->s_es->s_feature_ro_compat &
1979 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) {
1980 int offset = offsetof(struct ext4_group_desc, bg_checksum);
1981 __le32 le_group = cpu_to_le32(block_group);
1982
1983 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1984 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1985 crc = crc16(crc, (__u8 *)gdp, offset);
1986 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1987 /* for checksum of struct ext4_group_desc do the rest...*/
1988 if ((sbi->s_es->s_feature_incompat &
1989 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1990 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1991 crc = crc16(crc, (__u8 *)gdp + offset,
1992 le16_to_cpu(sbi->s_es->s_desc_size) -
1993 offset);
1994 }
1995
1996 return cpu_to_le16(crc);
1997}
1998
1999int ext4_group_desc_csum_verify(struct ext4_sb_info *sbi, __u32 block_group,
2000 struct ext4_group_desc *gdp)
2001{
2002 if ((sbi->s_es->s_feature_ro_compat &
2003 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)) &&
2004 (gdp->bg_checksum != ext4_group_desc_csum(sbi, block_group, gdp)))
2005 return 0;
2006
2007 return 1;
2008}
2009
ac27a0ec 2010/* Called at mount-time, super-block is locked */
bfff6873
LC		 2011static int ext4_check_descriptors(struct super_block *sb,
2012 ext4_group_t *first_not_zeroed)
ac27a0ec 2013{
617ba13b
MC		 2014 struct ext4_sb_info *sbi = EXT4_SB(sb);
2015 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2016 ext4_fsblk_t last_block;
bd81d8ee
LV		 2017 ext4_fsblk_t block_bitmap;
2018 ext4_fsblk_t inode_bitmap;
2019 ext4_fsblk_t inode_table;
ce421581 2020 int flexbg_flag = 0;
bfff6873 2021 ext4_group_t i, grp = sbi->s_groups_count;
ac27a0ec 2022
ce421581
JS		 2023 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2024 flexbg_flag = 1;
2025
af5bc92d 2026 ext4_debug("Checking group descriptors");
ac27a0ec 2027
197cd65a
AM		 2028 for (i = 0; i < sbi->s_groups_count; i++) {
2029 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2030
ce421581 2031 if (i == sbi->s_groups_count - 1 || flexbg_flag)
bd81d8ee 2032 last_block = ext4_blocks_count(sbi->s_es) - 1;
ac27a0ec
DK		 2033 else
2034 last_block = first_block +
617ba13b 2035 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
ac27a0ec 2036
bfff6873
LC		 2037 if ((grp == sbi->s_groups_count) &&
2038 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2039 grp = i;
2040
8fadc143 2041 block_bitmap = ext4_block_bitmap(sb, gdp);
2b2d6d01 2042 if (block_bitmap < first_block || block_bitmap > last_block) {
b31e1552 2043 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2044 "Block bitmap for group %u not in group "
b31e1552 2045 "(block %llu)!", i, block_bitmap);
ac27a0ec
DK		 2046 return 0;
2047 }
8fadc143 2048 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2b2d6d01 2049 if (inode_bitmap < first_block || inode_bitmap > last_block) {
b31e1552 2050 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2051 "Inode bitmap for group %u not in group "
b31e1552 2052 "(block %llu)!", i, inode_bitmap);
ac27a0ec
DK		 2053 return 0;
2054 }
8fadc143 2055 inode_table = ext4_inode_table(sb, gdp);
bd81d8ee 2056 if (inode_table < first_block ||
2b2d6d01 2057 inode_table + sbi->s_itb_per_group - 1 > last_block) {
b31e1552 2058 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
a9df9a49 2059 "Inode table for group %u not in group "
b31e1552 2060 "(block %llu)!", i, inode_table);
ac27a0ec
DK		 2061 return 0;
2062 }
955ce5f5 2063 ext4_lock_group(sb, i);
717d50e4 2064 if (!ext4_group_desc_csum_verify(sbi, i, gdp)) {
b31e1552
ES		 2065 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2066 "Checksum for group %u failed (%u!=%u)",
2067 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2068 gdp)), le16_to_cpu(gdp->bg_checksum));
7ee1ec4c 2069 if (!(sb->s_flags & MS_RDONLY)) {
955ce5f5 2070 ext4_unlock_group(sb, i);
8a266467 2071 return 0;
7ee1ec4c 2072 }
717d50e4 2073 }
955ce5f5 2074 ext4_unlock_group(sb, i);
ce421581
JS		 2075 if (!flexbg_flag)
2076 first_block += EXT4_BLOCKS_PER_GROUP(sb);
ac27a0ec 2077 }
bfff6873
LC		 2078 if (NULL != first_not_zeroed)
2079 *first_not_zeroed = grp;
ac27a0ec 2080
bd81d8ee 2081 ext4_free_blocks_count_set(sbi->s_es, ext4_count_free_blocks(sb));
0b8e58a1 2082 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
ac27a0ec
DK		 2083 return 1;
2084}
2085
617ba13b 2086/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
ac27a0ec
DK		 2087 * the superblock) which were deleted from all directories, but held open by
2088 * a process at the time of a crash. We walk the list and try to delete these
2089 * inodes at recovery time (only with a read-write filesystem).
2090 *
2091 * In order to keep the orphan inode chain consistent during traversal (in
2092 * case of crash during recovery), we link each inode into the superblock
2093 * orphan list_head and handle it the same way as an inode deletion during
2094 * normal operation (which journals the operations for us).
2095 *
2096 * We only do an iget() and an iput() on each inode, which is very safe if we
2097 * accidentally point at an in-use or already deleted inode. The worst that
2098 * can happen in this case is that we get a "bit already cleared" message from
617ba13b 2099 * ext4_free_inode(). The only reason we would point at a wrong inode is if
ac27a0ec
DK		 2100 * e2fsck was run on this filesystem, and it must have already done the orphan
2101 * inode cleanup for us, so we can safely abort without any further action.
2102 */
2b2d6d01
TT		 2103static void ext4_orphan_cleanup(struct super_block *sb,
2104 struct ext4_super_block *es)
ac27a0ec
DK		 2105{
2106 unsigned int s_flags = sb->s_flags;
2107 int nr_orphans = 0, nr_truncates = 0;
2108#ifdef CONFIG_QUOTA
2109 int i;
2110#endif
2111 if (!es->s_last_orphan) {
2112 jbd_debug(4, "no orphan inodes to clean up\n");
2113 return;
2114 }
2115
a8f48a95 2116 if (bdev_read_only(sb->s_bdev)) {
b31e1552
ES		 2117 ext4_msg(sb, KERN_ERR, "write access "
2118 "unavailable, skipping orphan cleanup");
a8f48a95
ES		 2119 return;
2120 }
2121
617ba13b 2122 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
ac27a0ec
DK		 2123 if (es->s_last_orphan)
2124 jbd_debug(1, "Errors on filesystem, "
2125 "clearing orphan list.\n");
2126 es->s_last_orphan = 0;
2127 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2128 return;
2129 }
2130
2131 if (s_flags & MS_RDONLY) {
b31e1552 2132 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
ac27a0ec
DK		 2133 sb->s_flags &= ~MS_RDONLY;
2134 }
2135#ifdef CONFIG_QUOTA
2136 /* Needed for iput() to work correctly and not trash data */
2137 sb->s_flags |= MS_ACTIVE;
2138 /* Turn on quotas so that they are updated correctly */
2139 for (i = 0; i < MAXQUOTAS; i++) {
617ba13b
MC		 2140 if (EXT4_SB(sb)->s_qf_names[i]) {
2141 int ret = ext4_quota_on_mount(sb, i);
ac27a0ec 2142 if (ret < 0)
b31e1552
ES		 2143 ext4_msg(sb, KERN_ERR,
2144 "Cannot turn on journaled "
2145 "quota: error %d", ret);
ac27a0ec
DK		 2146 }
2147 }
2148#endif
2149
2150 while (es->s_last_orphan) {
2151 struct inode *inode;
2152
97bd42b9
JB		 2153 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2154 if (IS_ERR(inode)) {
ac27a0ec
DK		 2155 es->s_last_orphan = 0;
2156 break;
2157 }
2158
617ba13b 2159 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
871a2931 2160 dquot_initialize(inode);
ac27a0ec 2161 if (inode->i_nlink) {
b31e1552
ES		 2162 ext4_msg(sb, KERN_DEBUG,
2163 "%s: truncating inode %lu to %lld bytes",
46e665e9 2164 __func__, inode->i_ino, inode->i_size);
e5f8eab8 2165 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
ac27a0ec 2166 inode->i_ino, inode->i_size);
617ba13b 2167 ext4_truncate(inode);
ac27a0ec
DK		 2168 nr_truncates++;
2169 } else {
b31e1552
ES		 2170 ext4_msg(sb, KERN_DEBUG,
2171 "%s: deleting unreferenced inode %lu",
46e665e9 2172 __func__, inode->i_ino);
ac27a0ec
DK		 2173 jbd_debug(2, "deleting unreferenced inode %lu\n",
2174 inode->i_ino);
2175 nr_orphans++;
2176 }
2177 iput(inode); /* The delete magic happens here! */
2178 }
2179
2b2d6d01 2180#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
ac27a0ec
DK		 2181
2182 if (nr_orphans)
b31e1552
ES		 2183 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2184 PLURAL(nr_orphans));
ac27a0ec 2185 if (nr_truncates)
b31e1552
ES		 2186 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2187 PLURAL(nr_truncates));
ac27a0ec
DK		 2188#ifdef CONFIG_QUOTA
2189 /* Turn quotas off */
2190 for (i = 0; i < MAXQUOTAS; i++) {
2191 if (sb_dqopt(sb)->files[i])
287a8095 2192 dquot_quota_off(sb, i);
ac27a0ec
DK		 2193 }
2194#endif
2195 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2196}
0b8e58a1 2197
cd2291a4
ES		 2198/*
2199 * Maximal extent format file size.
2200 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2201 * extent format containers, within a sector_t, and within i_blocks
2202 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2203 * so that won't be a limiting factor.
2204 *
2205 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2206 */
f287a1a5 2207static loff_t ext4_max_size(int blkbits, int has_huge_files)
cd2291a4
ES		 2208{
2209 loff_t res;
2210 loff_t upper_limit = MAX_LFS_FILESIZE;
2211
2212 /* small i_blocks in vfs inode? */
f287a1a5 2213 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
cd2291a4 2214 /*
90c699a9 2215 * CONFIG_LBDAF is not enabled implies the inode
cd2291a4
ES		 2216 * i_block represent total blocks in 512 bytes
2217 * 32 == size of vfs inode i_blocks * 8
2218 */
2219 upper_limit = (1LL << 32) - 1;
2220
2221 /* total blocks in file system block size */
2222 upper_limit >>= (blkbits - 9);
2223 upper_limit <<= blkbits;
2224 }
2225
2226 /* 32-bit extent-start container, ee_block */
2227 res = 1LL << 32;
2228 res <<= blkbits;
2229 res -= 1;
2230
2231 /* Sanity check against vm- & vfs- imposed limits */
2232 if (res > upper_limit)
2233 res = upper_limit;
2234
2235 return res;
2236}
ac27a0ec 2237
ac27a0ec 2238/*
cd2291a4 2239 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
0fc1b451
AK		 2240 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2241 * We need to be 1 filesystem block less than the 2^48 sector limit.
ac27a0ec 2242 */
f287a1a5 2243static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
ac27a0ec 2244{
617ba13b 2245 loff_t res = EXT4_NDIR_BLOCKS;
0fc1b451
AK		 2246 int meta_blocks;
2247 loff_t upper_limit;
0b8e58a1
AD		 2248 /* This is calculated to be the largest file size for a dense, block
2249 * mapped file such that the file's total number of 512-byte sectors,
2250 * including data and all indirect blocks, does not exceed (2^48 - 1).
2251 *
2252 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2253 * number of 512-byte sectors of the file.
0fc1b451
AK		 2254 */
2255
f287a1a5 2256 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
0fc1b451 2257 /*
90c699a9 2258 * !has_huge_files or CONFIG_LBDAF not enabled implies that
0b8e58a1
AD		 2259 * the inode i_block field represents total file blocks in
2260 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
0fc1b451
AK		 2261 */
2262 upper_limit = (1LL << 32) - 1;
2263
2264 /* total blocks in file system block size */
2265 upper_limit >>= (bits - 9);
2266
2267 } else {
8180a562
AK		 2268 /*
2269 * We use 48 bit ext4_inode i_blocks
2270 * With EXT4_HUGE_FILE_FL set the i_blocks
2271 * represent total number of blocks in
2272 * file system block size
2273 */
0fc1b451
AK		 2274 upper_limit = (1LL << 48) - 1;
2275
0fc1b451
AK		 2276 }
2277
2278 /* indirect blocks */
2279 meta_blocks = 1;
2280 /* double indirect blocks */
2281 meta_blocks += 1 + (1LL << (bits-2));
2282 /* tripple indirect blocks */
2283 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2284
2285 upper_limit -= meta_blocks;
2286 upper_limit <<= bits;
ac27a0ec
DK		 2287
2288 res += 1LL << (bits-2);
2289 res += 1LL << (2*(bits-2));
2290 res += 1LL << (3*(bits-2));
2291 res <<= bits;
2292 if (res > upper_limit)
2293 res = upper_limit;
0fc1b451
AK		 2294
2295 if (res > MAX_LFS_FILESIZE)
2296 res = MAX_LFS_FILESIZE;
2297
ac27a0ec
DK		 2298 return res;
2299}
2300
617ba13b 2301static ext4_fsblk_t descriptor_loc(struct super_block *sb,
0b8e58a1 2302 ext4_fsblk_t logical_sb_block, int nr)
ac27a0ec 2303{
617ba13b 2304 struct ext4_sb_info *sbi = EXT4_SB(sb);
fd2d4291 2305 ext4_group_t bg, first_meta_bg;
ac27a0ec
DK		 2306 int has_super = 0;
2307
2308 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2309
617ba13b 2310 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
ac27a0ec 2311 nr < first_meta_bg)
70bbb3e0 2312 return logical_sb_block + nr + 1;
ac27a0ec 2313 bg = sbi->s_desc_per_block * nr;
617ba13b 2314 if (ext4_bg_has_super(sb, bg))
ac27a0ec 2315 has_super = 1;
0b8e58a1 2316
617ba13b 2317 return (has_super + ext4_group_first_block_no(sb, bg));
ac27a0ec
DK		 2318}
2319
c9de560d
AT		 2320/**
2321 * ext4_get_stripe_size: Get the stripe size.
2322 * @sbi: In memory super block info
2323 *
2324 * If we have specified it via mount option, then
2325 * use the mount option value. If the value specified at mount time is
2326 * greater than the blocks per group use the super block value.
2327 * If the super block value is greater than blocks per group return 0.
2328 * Allocator needs it be less than blocks per group.
2329 *
2330 */
2331static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2332{
2333 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2334 unsigned long stripe_width =
2335 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2336
2337 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2338 return sbi->s_stripe;
2339
2340 if (stripe_width <= sbi->s_blocks_per_group)
2341 return stripe_width;
2342
2343 if (stride <= sbi->s_blocks_per_group)
2344 return stride;
2345
2346 return 0;
2347}
ac27a0ec 2348
3197ebdb
TT		 2349/* sysfs supprt */
2350
2351struct ext4_attr {
2352 struct attribute attr;
2353 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
60e6679e 2354 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
3197ebdb
TT		 2355 const char *, size_t);
2356 int offset;
2357};
2358
2359static int parse_strtoul(const char *buf,
2360 unsigned long max, unsigned long *value)
2361{
2362 char *endp;
2363
e7d2860b
AGR		 2364 *value = simple_strtoul(skip_spaces(buf), &endp, 0);
2365 endp = skip_spaces(endp);
3197ebdb
TT		 2366 if (*endp || *value > max)
2367 return -EINVAL;
2368
2369 return 0;
2370}
2371
2372static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2373 struct ext4_sb_info *sbi,
2374 char *buf)
2375{
2376 return snprintf(buf, PAGE_SIZE, "%llu\n",
2377 (s64) percpu_counter_sum(&sbi->s_dirtyblocks_counter));
2378}
2379
2380static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2381 struct ext4_sb_info *sbi, char *buf)
2382{
2383 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2384
f613dfcb
TT		 2385 if (!sb->s_bdev->bd_part)
2386 return snprintf(buf, PAGE_SIZE, "0\n");
3197ebdb
TT		 2387 return snprintf(buf, PAGE_SIZE, "%lu\n",
2388 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2389 sbi->s_sectors_written_start) >> 1);
2390}
2391
2392static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2393 struct ext4_sb_info *sbi, char *buf)
2394{
2395 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2396
f613dfcb
TT		 2397 if (!sb->s_bdev->bd_part)
2398 return snprintf(buf, PAGE_SIZE, "0\n");
3197ebdb 2399 return snprintf(buf, PAGE_SIZE, "%llu\n",
a6b43e38 2400 (unsigned long long)(sbi->s_kbytes_written +
3197ebdb 2401 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
a6b43e38 2402 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
3197ebdb
TT		 2403}
2404
2405static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2406 struct ext4_sb_info *sbi,
2407 const char *buf, size_t count)
2408{
2409 unsigned long t;
2410
2411 if (parse_strtoul(buf, 0x40000000, &t))
2412 return -EINVAL;
2413
f7c43950 2414 if (!is_power_of_2(t))
3197ebdb
TT		 2415 return -EINVAL;
2416
2417 sbi->s_inode_readahead_blks = t;
2418 return count;
2419}
2420
2421static ssize_t sbi_ui_show(struct ext4_attr *a,
0b8e58a1 2422 struct ext4_sb_info *sbi, char *buf)
3197ebdb
TT		 2423{
2424 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2425
2426 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2427}
2428
2429static ssize_t sbi_ui_store(struct ext4_attr *a,
2430 struct ext4_sb_info *sbi,
2431 const char *buf, size_t count)
2432{
2433 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2434 unsigned long t;
2435
2436 if (parse_strtoul(buf, 0xffffffff, &t))
2437 return -EINVAL;
2438 *ui = t;
2439 return count;
2440}
2441
2442#define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2443static struct ext4_attr ext4_attr_##_name = { \
2444 .attr = {.name = __stringify(_name), .mode = _mode }, \
2445 .show = _show, \
2446 .store = _store, \
2447 .offset = offsetof(struct ext4_sb_info, _elname), \
2448}
2449#define EXT4_ATTR(name, mode, show, store) \
2450static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2451
857ac889 2452#define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
3197ebdb
TT		 2453#define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2454#define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2455#define EXT4_RW_ATTR_SBI_UI(name, elname) \
2456 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2457#define ATTR_LIST(name) &ext4_attr_##name.attr
2458
2459EXT4_RO_ATTR(delayed_allocation_blocks);
2460EXT4_RO_ATTR(session_write_kbytes);
2461EXT4_RO_ATTR(lifetime_write_kbytes);
2462EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2463 inode_readahead_blks_store, s_inode_readahead_blks);
11013911 2464EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
3197ebdb
TT		 2465EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2466EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2467EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2468EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2469EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2470EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
55138e0b 2471EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
3197ebdb
TT		 2472
2473static struct attribute *ext4_attrs[] = {
2474 ATTR_LIST(delayed_allocation_blocks),
2475 ATTR_LIST(session_write_kbytes),
2476 ATTR_LIST(lifetime_write_kbytes),
2477 ATTR_LIST(inode_readahead_blks),
11013911 2478 ATTR_LIST(inode_goal),
3197ebdb
TT		 2479 ATTR_LIST(mb_stats),
2480 ATTR_LIST(mb_max_to_scan),
2481 ATTR_LIST(mb_min_to_scan),
2482 ATTR_LIST(mb_order2_req),
2483 ATTR_LIST(mb_stream_req),
2484 ATTR_LIST(mb_group_prealloc),
55138e0b 2485 ATTR_LIST(max_writeback_mb_bump),
3197ebdb
TT		 2486 NULL,
2487};
2488
857ac889
LC		 2489/* Features this copy of ext4 supports */
2490EXT4_INFO_ATTR(lazy_itable_init);
27ee40df 2491EXT4_INFO_ATTR(batched_discard);
857ac889
LC		 2492
2493static struct attribute *ext4_feat_attrs[] = {
2494 ATTR_LIST(lazy_itable_init),
27ee40df 2495 ATTR_LIST(batched_discard),
857ac889
LC		 2496 NULL,
2497};
2498
3197ebdb
TT		 2499static ssize_t ext4_attr_show(struct kobject *kobj,
2500 struct attribute *attr, char *buf)
2501{
2502 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2503 s_kobj);
2504 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2505
2506 return a->show ? a->show(a, sbi, buf) : 0;
2507}
2508
2509static ssize_t ext4_attr_store(struct kobject *kobj,
2510 struct attribute *attr,
2511 const char *buf, size_t len)
2512{
2513 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2514 s_kobj);
2515 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2516
2517 return a->store ? a->store(a, sbi, buf, len) : 0;
2518}
2519
2520static void ext4_sb_release(struct kobject *kobj)
2521{
2522 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2523 s_kobj);
2524 complete(&sbi->s_kobj_unregister);
2525}
2526
52cf25d0 2527static const struct sysfs_ops ext4_attr_ops = {
3197ebdb
TT		 2528 .show = ext4_attr_show,
2529 .store = ext4_attr_store,
2530};
2531
2532static struct kobj_type ext4_ktype = {
2533 .default_attrs = ext4_attrs,
2534 .sysfs_ops = &ext4_attr_ops,
2535 .release = ext4_sb_release,
2536};
2537
857ac889
LC		 2538static void ext4_feat_release(struct kobject *kobj)
2539{
2540 complete(&ext4_feat->f_kobj_unregister);
2541}
2542
2543static struct kobj_type ext4_feat_ktype = {
2544 .default_attrs = ext4_feat_attrs,
2545 .sysfs_ops = &ext4_attr_ops,
2546 .release = ext4_feat_release,
2547};
2548
a13fb1a4
ES		 2549/*
2550 * Check whether this filesystem can be mounted based on
2551 * the features present and the RDONLY/RDWR mount requested.
2552 * Returns 1 if this filesystem can be mounted as requested,
2553 * 0 if it cannot be.
2554 */
2555static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2556{
2557 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2558 ext4_msg(sb, KERN_ERR,
2559 "Couldn't mount because of "
2560 "unsupported optional features (%x)",
2561 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2562 ~EXT4_FEATURE_INCOMPAT_SUPP));
2563 return 0;
2564 }
2565
2566 if (readonly)
2567 return 1;
2568
2569 /* Check that feature set is OK for a read-write mount */
2570 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2571 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2572 "unsupported optional features (%x)",
2573 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2574 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2575 return 0;
2576 }
2577 /*
2578 * Large file size enabled file system can only be mounted
2579 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2580 */
2581 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2582 if (sizeof(blkcnt_t) < sizeof(u64)) {
2583 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2584 "cannot be mounted RDWR without "
2585 "CONFIG_LBDAF");
2586 return 0;
2587 }
2588 }
2589 return 1;
2590}
2591
66e61a9e
TT		 2592/*
2593 * This function is called once a day if we have errors logged
2594 * on the file system
2595 */
2596static void print_daily_error_info(unsigned long arg)
2597{
2598 struct super_block *sb = (struct super_block *) arg;
2599 struct ext4_sb_info *sbi;
2600 struct ext4_super_block *es;
2601
2602 sbi = EXT4_SB(sb);
2603 es = sbi->s_es;
2604
2605 if (es->s_error_count)
2606 ext4_msg(sb, KERN_NOTICE, "error count: %u",
2607 le32_to_cpu(es->s_error_count));
2608 if (es->s_first_error_time) {
2609 printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d",
2610 sb->s_id, le32_to_cpu(es->s_first_error_time),
2611 (int) sizeof(es->s_first_error_func),
2612 es->s_first_error_func,
2613 le32_to_cpu(es->s_first_error_line));
2614 if (es->s_first_error_ino)
2615 printk(": inode %u",
2616 le32_to_cpu(es->s_first_error_ino));
2617 if (es->s_first_error_block)
2618 printk(": block %llu", (unsigned long long)
2619 le64_to_cpu(es->s_first_error_block));
2620 printk("\n");
2621 }
2622 if (es->s_last_error_time) {
2623 printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d",
2624 sb->s_id, le32_to_cpu(es->s_last_error_time),
2625 (int) sizeof(es->s_last_error_func),
2626 es->s_last_error_func,
2627 le32_to_cpu(es->s_last_error_line));
2628 if (es->s_last_error_ino)
2629 printk(": inode %u",
2630 le32_to_cpu(es->s_last_error_ino));
2631 if (es->s_last_error_block)
2632 printk(": block %llu", (unsigned long long)
2633 le64_to_cpu(es->s_last_error_block));
2634 printk("\n");
2635 }
2636 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2637}
2638
bfff6873
LC		 2639static void ext4_lazyinode_timeout(unsigned long data)
2640{
2641 struct task_struct *p = (struct task_struct *)data;
2642 wake_up_process(p);
2643}
2644
2645/* Find next suitable group and run ext4_init_inode_table */
2646static int ext4_run_li_request(struct ext4_li_request *elr)
2647{
2648 struct ext4_group_desc *gdp = NULL;
2649 ext4_group_t group, ngroups;
2650 struct super_block *sb;
2651 unsigned long timeout = 0;
2652 int ret = 0;
2653
2654 sb = elr->lr_super;
2655 ngroups = EXT4_SB(sb)->s_groups_count;
2656
2657 for (group = elr->lr_next_group; group < ngroups; group++) {
2658 gdp = ext4_get_group_desc(sb, group, NULL);
2659 if (!gdp) {
2660 ret = 1;
2661 break;
2662 }
2663
2664 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2665 break;
2666 }
2667
2668 if (group == ngroups)
2669 ret = 1;
2670
2671 if (!ret) {
2672 timeout = jiffies;
2673 ret = ext4_init_inode_table(sb, group,
2674 elr->lr_timeout ? 0 : 1);
2675 if (elr->lr_timeout == 0) {
2676 timeout = jiffies - timeout;
2677 if (elr->lr_sbi->s_li_wait_mult)
2678 timeout *= elr->lr_sbi->s_li_wait_mult;
2679 else
2680 timeout *= 20;
2681 elr->lr_timeout = timeout;
2682 }
2683 elr->lr_next_sched = jiffies + elr->lr_timeout;
2684 elr->lr_next_group = group + 1;
2685 }
2686
2687 return ret;
2688}
2689
2690/*
2691 * Remove lr_request from the list_request and free the
2692 * request tructure. Should be called with li_list_mtx held
2693 */
2694static void ext4_remove_li_request(struct ext4_li_request *elr)
2695{
2696 struct ext4_sb_info *sbi;
2697
2698 if (!elr)
2699 return;
2700
2701 sbi = elr->lr_sbi;
2702
2703 list_del(&elr->lr_request);
2704 sbi->s_li_request = NULL;
2705 kfree(elr);
2706}
2707
2708static void ext4_unregister_li_request(struct super_block *sb)
2709{
2710 struct ext4_li_request *elr = EXT4_SB(sb)->s_li_request;
2711
2712 if (!ext4_li_info)
2713 return;
2714
2715 mutex_lock(&ext4_li_info->li_list_mtx);
2716 ext4_remove_li_request(elr);
2717 mutex_unlock(&ext4_li_info->li_list_mtx);
2718}
2719
8f1f7453
ES		 2720static struct task_struct *ext4_lazyinit_task;
2721
bfff6873
LC		 2722/*
2723 * This is the function where ext4lazyinit thread lives. It walks
2724 * through the request list searching for next scheduled filesystem.
2725 * When such a fs is found, run the lazy initialization request
2726 * (ext4_rn_li_request) and keep track of the time spend in this
2727 * function. Based on that time we compute next schedule time of
2728 * the request. When walking through the list is complete, compute
2729 * next waking time and put itself into sleep.
2730 */
2731static int ext4_lazyinit_thread(void *arg)
2732{
2733 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2734 struct list_head *pos, *n;
2735 struct ext4_li_request *elr;
2736 unsigned long next_wakeup;
2737 DEFINE_WAIT(wait);
bfff6873
LC		 2738
2739 BUG_ON(NULL == eli);
2740
2741 eli->li_timer.data = (unsigned long)current;
2742 eli->li_timer.function = ext4_lazyinode_timeout;
2743
2744 eli->li_task = current;
2745 wake_up(&eli->li_wait_task);
2746
2747cont_thread:
2748 while (true) {
2749 next_wakeup = MAX_JIFFY_OFFSET;
2750
2751 mutex_lock(&eli->li_list_mtx);
2752 if (list_empty(&eli->li_request_list)) {
2753 mutex_unlock(&eli->li_list_mtx);
2754 goto exit_thread;
2755 }
2756
2757 list_for_each_safe(pos, n, &eli->li_request_list) {
2758 elr = list_entry(pos, struct ext4_li_request,
2759 lr_request);
2760
b2c78cd0
TT		 2761 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2762 if (ext4_run_li_request(elr) != 0) {
2763 /* error, remove the lazy_init job */
2764 ext4_remove_li_request(elr);
2765 continue;
2766 }
bfff6873
LC		 2767 }
2768
2769 if (time_before(elr->lr_next_sched, next_wakeup))
2770 next_wakeup = elr->lr_next_sched;
2771 }
2772 mutex_unlock(&eli->li_list_mtx);
2773
2774 if (freezing(current))
2775 refrigerator();
2776
f4245bd4
LC		 2777 if ((time_after_eq(jiffies, next_wakeup)) ||
2778 (MAX_JIFFY_OFFSET == next_wakeup)) {
bfff6873
LC		 2779 cond_resched();
2780 continue;
2781 }
2782
2783 eli->li_timer.expires = next_wakeup;
2784 add_timer(&eli->li_timer);
2785 prepare_to_wait(&eli->li_wait_daemon, &wait,
2786 TASK_INTERRUPTIBLE);
2787 if (time_before(jiffies, next_wakeup))
2788 schedule();
2789 finish_wait(&eli->li_wait_daemon, &wait);
8f1f7453
ES		 2790 if (kthread_should_stop()) {
2791 ext4_clear_request_list();
2792 goto exit_thread;
2793 }
bfff6873
LC		 2794 }
2795
2796exit_thread:
2797 /*
2798 * It looks like the request list is empty, but we need
2799 * to check it under the li_list_mtx lock, to prevent any
2800 * additions into it, and of course we should lock ext4_li_mtx
2801 * to atomically free the list and ext4_li_info, because at
2802 * this point another ext4 filesystem could be registering
2803 * new one.
2804 */
2805 mutex_lock(&ext4_li_mtx);
2806 mutex_lock(&eli->li_list_mtx);
2807 if (!list_empty(&eli->li_request_list)) {
2808 mutex_unlock(&eli->li_list_mtx);
2809 mutex_unlock(&ext4_li_mtx);
2810 goto cont_thread;
2811 }
2812 mutex_unlock(&eli->li_list_mtx);
2813 del_timer_sync(&ext4_li_info->li_timer);
2814 eli->li_task = NULL;
2815 wake_up(&eli->li_wait_task);
2816
2817 kfree(ext4_li_info);
8f1f7453 2818 ext4_lazyinit_task = NULL;
bfff6873
LC		 2819 ext4_li_info = NULL;
2820 mutex_unlock(&ext4_li_mtx);
2821
2822 return 0;
2823}
2824
2825static void ext4_clear_request_list(void)
2826{
2827 struct list_head *pos, *n;
2828 struct ext4_li_request *elr;
2829
2830 mutex_lock(&ext4_li_info->li_list_mtx);
bfff6873
LC		 2831 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2832 elr = list_entry(pos, struct ext4_li_request,
2833 lr_request);
2834 ext4_remove_li_request(elr);
2835 }
2836 mutex_unlock(&ext4_li_info->li_list_mtx);
2837}
2838
2839static int ext4_run_lazyinit_thread(void)
2840{
8f1f7453
ES		 2841 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2842 ext4_li_info, "ext4lazyinit");
2843 if (IS_ERR(ext4_lazyinit_task)) {
2844 int err = PTR_ERR(ext4_lazyinit_task);
bfff6873
LC		 2845 ext4_clear_request_list();
2846 del_timer_sync(&ext4_li_info->li_timer);
2847 kfree(ext4_li_info);
2848 ext4_li_info = NULL;
2849 printk(KERN_CRIT "EXT4: error %d creating inode table "
2850 "initialization thread\n",
2851 err);
2852 return err;
2853 }
2854 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
2855
2856 wait_event(ext4_li_info->li_wait_task, ext4_li_info->li_task != NULL);
2857 return 0;
2858}
2859
2860/*
2861 * Check whether it make sense to run itable init. thread or not.
2862 * If there is at least one uninitialized inode table, return
2863 * corresponding group number, else the loop goes through all
2864 * groups and return total number of groups.
2865 */
2866static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2867{
2868 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2869 struct ext4_group_desc *gdp = NULL;
2870
2871 for (group = 0; group < ngroups; group++) {
2872 gdp = ext4_get_group_desc(sb, group, NULL);
2873 if (!gdp)
2874 continue;
2875
2876 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2877 break;
2878 }
2879
2880 return group;
2881}
2882
2883static int ext4_li_info_new(void)
2884{
2885 struct ext4_lazy_init *eli = NULL;
2886
2887 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2888 if (!eli)
2889 return -ENOMEM;
2890
2891 eli->li_task = NULL;
2892 INIT_LIST_HEAD(&eli->li_request_list);
2893 mutex_init(&eli->li_list_mtx);
2894
2895 init_waitqueue_head(&eli->li_wait_daemon);
2896 init_waitqueue_head(&eli->li_wait_task);
2897 init_timer(&eli->li_timer);
2898 eli->li_state |= EXT4_LAZYINIT_QUIT;
2899
2900 ext4_li_info = eli;
2901
2902 return 0;
2903}
2904
2905static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
2906 ext4_group_t start)
2907{
2908 struct ext4_sb_info *sbi = EXT4_SB(sb);
2909 struct ext4_li_request *elr;
2910 unsigned long rnd;
2911
2912 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
2913 if (!elr)
2914 return NULL;
2915
2916 elr->lr_super = sb;
2917 elr->lr_sbi = sbi;
2918 elr->lr_next_group = start;
2919
2920 /*
2921 * Randomize first schedule time of the request to
2922 * spread the inode table initialization requests
2923 * better.
2924 */
2925 get_random_bytes(&rnd, sizeof(rnd));
2926 elr->lr_next_sched = jiffies + (unsigned long)rnd %
2927 (EXT4_DEF_LI_MAX_START_DELAY * HZ);
2928
2929 return elr;
2930}
2931
2932static int ext4_register_li_request(struct super_block *sb,
2933 ext4_group_t first_not_zeroed)
2934{
2935 struct ext4_sb_info *sbi = EXT4_SB(sb);
2936 struct ext4_li_request *elr;
2937 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
6c5a6cb9 2938 int ret = 0;
bfff6873
LC		 2939
2940 if (sbi->s_li_request != NULL)
beed5ecb 2941 return 0;
bfff6873
LC		 2942
2943 if (first_not_zeroed == ngroups ||
2944 (sb->s_flags & MS_RDONLY) ||
2945 !test_opt(sb, INIT_INODE_TABLE)) {
2946 sbi->s_li_request = NULL;
beed5ecb 2947 return 0;
bfff6873
LC		 2948 }
2949
2950 if (first_not_zeroed == ngroups) {
2951 sbi->s_li_request = NULL;
beed5ecb 2952 return 0;
bfff6873
LC		 2953 }
2954
2955 elr = ext4_li_request_new(sb, first_not_zeroed);
beed5ecb
NK		 2956 if (!elr)
2957 return -ENOMEM;
bfff6873
LC		 2958
2959 mutex_lock(&ext4_li_mtx);
2960
2961 if (NULL == ext4_li_info) {
2962 ret = ext4_li_info_new();
2963 if (ret)
2964 goto out;
2965 }
2966
2967 mutex_lock(&ext4_li_info->li_list_mtx);
2968 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
2969 mutex_unlock(&ext4_li_info->li_list_mtx);
2970
2971 sbi->s_li_request = elr;
2972
2973 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
2974 ret = ext4_run_lazyinit_thread();
2975 if (ret)
2976 goto out;
2977 }
bfff6873 2978out:
beed5ecb
NK		 2979 mutex_unlock(&ext4_li_mtx);
2980 if (ret)
bfff6873 2981 kfree(elr);
bfff6873
LC		 2982 return ret;
2983}
2984
2985/*
2986 * We do not need to lock anything since this is called on
2987 * module unload.
2988 */
2989static void ext4_destroy_lazyinit_thread(void)
2990{
2991 /*
2992 * If thread exited earlier
2993 * there's nothing to be done.
2994 */
8f1f7453 2995 if (!ext4_li_info || !ext4_lazyinit_task)
bfff6873
LC		 2996 return;
2997
8f1f7453 2998 kthread_stop(ext4_lazyinit_task);
bfff6873
LC		 2999}
3000
2b2d6d01 3001static int ext4_fill_super(struct super_block *sb, void *data, int silent)
7477827f
AK		 3002 __releases(kernel_lock)
3003 __acquires(kernel_lock)
ac27a0ec 3004{
d4c402d9 3005 char *orig_data = kstrdup(data, GFP_KERNEL);
2b2d6d01 3006 struct buffer_head *bh;
617ba13b
MC		 3007 struct ext4_super_block *es = NULL;
3008 struct ext4_sb_info *sbi;
3009 ext4_fsblk_t block;
3010 ext4_fsblk_t sb_block = get_sb_block(&data);
70bbb3e0 3011 ext4_fsblk_t logical_sb_block;
ac27a0ec 3012 unsigned long offset = 0;
ac27a0ec
DK		 3013 unsigned long journal_devnum = 0;
3014 unsigned long def_mount_opts;
3015 struct inode *root;
9f6200bb 3016 char *cp;
0390131b 3017 const char *descr;
dcc7dae3 3018 int ret = -ENOMEM;
ac27a0ec 3019 int blocksize;
4ec11028
TT		 3020 unsigned int db_count;
3021 unsigned int i;
f287a1a5 3022 int needs_recovery, has_huge_files;
bd81d8ee 3023 __u64 blocks_count;
833f4077 3024 int err;
b3881f74 3025 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
bfff6873 3026 ext4_group_t first_not_zeroed;
ac27a0ec
DK		 3027
3028 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3029 if (!sbi)
dcc7dae3 3030 goto out_free_orig;
705895b6
PE		 3031
3032 sbi->s_blockgroup_lock =
3033 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3034 if (!sbi->s_blockgroup_lock) {
3035 kfree(sbi);
dcc7dae3 3036 goto out_free_orig;
705895b6 3037 }
ac27a0ec
DK		 3038 sb->s_fs_info = sbi;
3039 sbi->s_mount_opt = 0;
617ba13b
MC		 3040 sbi->s_resuid = EXT4_DEF_RESUID;
3041 sbi->s_resgid = EXT4_DEF_RESGID;
240799cd 3042 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
d9c9bef1 3043 sbi->s_sb_block = sb_block;
f613dfcb
TT		 3044 if (sb->s_bdev->bd_part)
3045 sbi->s_sectors_written_start =
3046 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
ac27a0ec 3047
9f6200bb
TT		 3048 /* Cleanup superblock name */
3049 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3050 *cp = '!';
3051
dcc7dae3 3052 ret = -EINVAL;
617ba13b 3053 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
ac27a0ec 3054 if (!blocksize) {
b31e1552 3055 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
ac27a0ec
DK		 3056 goto out_fail;
3057 }
3058
3059 /*
617ba13b 3060 * The ext4 superblock will not be buffer aligned for other than 1kB
ac27a0ec
DK		 3061 * block sizes. We need to calculate the offset from buffer start.
3062 */
617ba13b 3063 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
70bbb3e0
AM		 3064 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3065 offset = do_div(logical_sb_block, blocksize);
ac27a0ec 3066 } else {
70bbb3e0 3067 logical_sb_block = sb_block;
ac27a0ec
DK		 3068 }
3069
70bbb3e0 3070 if (!(bh = sb_bread(sb, logical_sb_block))) {
b31e1552 3071 ext4_msg(sb, KERN_ERR, "unable to read superblock");
ac27a0ec
DK		 3072 goto out_fail;
3073 }
3074 /*
3075 * Note: s_es must be initialized as soon as possible because
617ba13b 3076 * some ext4 macro-instructions depend on its value
ac27a0ec 3077 */
617ba13b 3078 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
ac27a0ec
DK		 3079 sbi->s_es = es;
3080 sb->s_magic = le16_to_cpu(es->s_magic);
617ba13b
MC		 3081 if (sb->s_magic != EXT4_SUPER_MAGIC)
3082 goto cantfind_ext4;
afc32f7e 3083 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
ac27a0ec
DK		 3084
3085 /* Set defaults before we parse the mount options */
3086 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
fd8c37ec 3087 set_opt(sb, INIT_INODE_TABLE);
617ba13b 3088 if (def_mount_opts & EXT4_DEFM_DEBUG)
fd8c37ec 3089 set_opt(sb, DEBUG);
437ca0fd
DM		 3090 if (def_mount_opts & EXT4_DEFM_BSDGROUPS) {
3091 ext4_msg(sb, KERN_WARNING, deprecated_msg, "bsdgroups",
3092 "2.6.38");
fd8c37ec 3093 set_opt(sb, GRPID);
437ca0fd 3094 }
617ba13b 3095 if (def_mount_opts & EXT4_DEFM_UID16)
fd8c37ec 3096 set_opt(sb, NO_UID32);
03010a33 3097#ifdef CONFIG_EXT4_FS_XATTR
617ba13b 3098 if (def_mount_opts & EXT4_DEFM_XATTR_USER)
fd8c37ec 3099 set_opt(sb, XATTR_USER);
2e7842b8 3100#endif
03010a33 3101#ifdef CONFIG_EXT4_FS_POSIX_ACL
617ba13b 3102 if (def_mount_opts & EXT4_DEFM_ACL)
fd8c37ec 3103 set_opt(sb, POSIX_ACL);
2e7842b8 3104#endif
617ba13b 3105 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
fd8c37ec 3106 set_opt(sb, JOURNAL_DATA);
617ba13b 3107 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
fd8c37ec 3108 set_opt(sb, ORDERED_DATA);
617ba13b 3109 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
fd8c37ec 3110 set_opt(sb, WRITEBACK_DATA);
617ba13b
MC		 3111
3112 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
fd8c37ec 3113 set_opt(sb, ERRORS_PANIC);
bb4f397a 3114 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
fd8c37ec 3115 set_opt(sb, ERRORS_CONT);
bb4f397a 3116 else
fd8c37ec 3117 set_opt(sb, ERRORS_RO);
8b67f04a 3118 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
fd8c37ec 3119 set_opt(sb, BLOCK_VALIDITY);
8b67f04a 3120 if (def_mount_opts & EXT4_DEFM_DISCARD)
fd8c37ec 3121 set_opt(sb, DISCARD);
ac27a0ec
DK		 3122
3123 sbi->s_resuid = le16_to_cpu(es->s_def_resuid);
3124 sbi->s_resgid = le16_to_cpu(es->s_def_resgid);
30773840
TT		 3125 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3126 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3127 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
ac27a0ec 3128
8b67f04a 3129 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
fd8c37ec 3130 set_opt(sb, BARRIER);
ac27a0ec 3131
dd919b98
AK		 3132 /*
3133 * enable delayed allocation by default
3134 * Use -o nodelalloc to turn it off
3135 */
8b67f04a
TT		 3136 if (!IS_EXT3_SB(sb) &&
3137 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
fd8c37ec 3138 set_opt(sb, DELALLOC);
dd919b98 3139
8b67f04a
TT		 3140 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3141 &journal_devnum, &journal_ioprio, NULL, 0)) {
3142 ext4_msg(sb, KERN_WARNING,
3143 "failed to parse options in superblock: %s",
3144 sbi->s_es->s_mount_opts);
3145 }
b3881f74
TT		 3146 if (!parse_options((char *) data, sb, &journal_devnum,
3147 &journal_ioprio, NULL, 0))
ac27a0ec
DK		 3148 goto failed_mount;
3149
3150 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 3151 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec 3152
617ba13b
MC		 3153 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3154 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3155 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3156 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
b31e1552
ES		 3157 ext4_msg(sb, KERN_WARNING,
3158 "feature flags set on rev 0 fs, "
3159 "running e2fsck is recommended");
469108ff 3160
ac27a0ec
DK		 3161 /*
3162 * Check feature flags regardless of the revision level, since we
3163 * previously didn't change the revision level when setting the flags,
3164 * so there is a chance incompat flags are set on a rev 0 filesystem.
3165 */
a13fb1a4 3166 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
ac27a0ec 3167 goto failed_mount;
a13fb1a4 3168
ac27a0ec
DK		 3169 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3170
617ba13b
MC		 3171 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3172 blocksize > EXT4_MAX_BLOCK_SIZE) {
b31e1552
ES		 3173 ext4_msg(sb, KERN_ERR,
3174 "Unsupported filesystem blocksize %d", blocksize);
ac27a0ec
DK		 3175 goto failed_mount;
3176 }
3177
ac27a0ec 3178 if (sb->s_blocksize != blocksize) {
ce40733c
AK		 3179 /* Validate the filesystem blocksize */
3180 if (!sb_set_blocksize(sb, blocksize)) {
b31e1552 3181 ext4_msg(sb, KERN_ERR, "bad block size %d",
ce40733c 3182 blocksize);
ac27a0ec
DK		 3183 goto failed_mount;
3184 }
3185
2b2d6d01 3186 brelse(bh);
70bbb3e0
AM		 3187 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3188 offset = do_div(logical_sb_block, blocksize);
3189 bh = sb_bread(sb, logical_sb_block);
ac27a0ec 3190 if (!bh) {
b31e1552
ES		 3191 ext4_msg(sb, KERN_ERR,
3192 "Can't read superblock on 2nd try");
ac27a0ec
DK		 3193 goto failed_mount;
3194 }
617ba13b 3195 es = (struct ext4_super_block *)(((char *)bh->b_data) + offset);
ac27a0ec 3196 sbi->s_es = es;
617ba13b 3197 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
b31e1552
ES		 3198 ext4_msg(sb, KERN_ERR,
3199 "Magic mismatch, very weird!");
ac27a0ec
DK		 3200 goto failed_mount;
3201 }
3202 }
3203
a13fb1a4
ES		 3204 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3205 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
f287a1a5
TT		 3206 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3207 has_huge_files);
3208 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
ac27a0ec 3209
617ba13b
MC		 3210 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3211 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3212 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
ac27a0ec
DK		 3213 } else {
3214 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3215 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
617ba13b 3216 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
1330593e 3217 (!is_power_of_2(sbi->s_inode_size)) ||
ac27a0ec 3218 (sbi->s_inode_size > blocksize)) {
b31e1552
ES		 3219 ext4_msg(sb, KERN_ERR,
3220 "unsupported inode size: %d",
2b2d6d01 3221 sbi->s_inode_size);
ac27a0ec
DK		 3222 goto failed_mount;
3223 }
ef7f3835
KS		 3224 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3225 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
ac27a0ec 3226 }
0b8e58a1 3227
0d1ee42f
AR		 3228 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3229 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
8fadc143 3230 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
0d1ee42f 3231 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
d8ea6cf8 3232 !is_power_of_2(sbi->s_desc_size)) {
b31e1552
ES		 3233 ext4_msg(sb, KERN_ERR,
3234 "unsupported descriptor size %lu",
0d1ee42f
AR		 3235 sbi->s_desc_size);
3236 goto failed_mount;
3237 }
3238 } else
3239 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
0b8e58a1 3240
ac27a0ec 3241 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
ac27a0ec 3242 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
b47b6f38 3243 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
617ba13b 3244 goto cantfind_ext4;
0b8e58a1 3245
617ba13b 3246 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
ac27a0ec 3247 if (sbi->s_inodes_per_block == 0)
617ba13b 3248 goto cantfind_ext4;
ac27a0ec
DK		 3249 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3250 sbi->s_inodes_per_block;
0d1ee42f 3251 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
ac27a0ec
DK		 3252 sbi->s_sbh = bh;
3253 sbi->s_mount_state = le16_to_cpu(es->s_state);
e57aa839
FW		 3254 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3255 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
0b8e58a1 3256
2b2d6d01 3257 for (i = 0; i < 4; i++)
ac27a0ec
DK		 3258 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3259 sbi->s_def_hash_version = es->s_def_hash_version;
f99b2589
TT		 3260 i = le32_to_cpu(es->s_flags);
3261 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3262 sbi->s_hash_unsigned = 3;
3263 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3264#ifdef __CHAR_UNSIGNED__
3265 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3266 sbi->s_hash_unsigned = 3;
3267#else
3268 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3269#endif
3270 sb->s_dirt = 1;
3271 }
ac27a0ec
DK		 3272
3273 if (sbi->s_blocks_per_group > blocksize * 8) {
b31e1552
ES		 3274 ext4_msg(sb, KERN_ERR,
3275 "#blocks per group too big: %lu",
2b2d6d01 3276 sbi->s_blocks_per_group);
ac27a0ec
DK		 3277 goto failed_mount;
3278 }
ac27a0ec 3279 if (sbi->s_inodes_per_group > blocksize * 8) {
b31e1552
ES		 3280 ext4_msg(sb, KERN_ERR,
3281 "#inodes per group too big: %lu",
2b2d6d01 3282 sbi->s_inodes_per_group);
ac27a0ec
DK		 3283 goto failed_mount;
3284 }
3285
bf43d84b
ES		 3286 /*
3287 * Test whether we have more sectors than will fit in sector_t,
3288 * and whether the max offset is addressable by the page cache.
3289 */
5a9ae68a 3290 err = generic_check_addressable(sb->s_blocksize_bits,
30ca22c7 3291 ext4_blocks_count(es));
5a9ae68a 3292 if (err) {
b31e1552 3293 ext4_msg(sb, KERN_ERR, "filesystem"
bf43d84b 3294 " too large to mount safely on this system");
ac27a0ec 3295 if (sizeof(sector_t) < 8)
90c699a9 3296 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
5a9ae68a 3297 ret = err;
ac27a0ec
DK		 3298 goto failed_mount;
3299 }
3300
617ba13b
MC		 3301 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3302 goto cantfind_ext4;
e7c95593 3303
0f2ddca6
FTN		 3304 /* check blocks count against device size */
3305 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3306 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
b31e1552
ES		 3307 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3308 "exceeds size of device (%llu blocks)",
0f2ddca6
FTN		 3309 ext4_blocks_count(es), blocks_count);
3310 goto failed_mount;
3311 }
3312
0b8e58a1
AD		 3313 /*
3314 * It makes no sense for the first data block to be beyond the end
3315 * of the filesystem.
3316 */
3317 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
b31e1552
ES		 3318 ext4_msg(sb, KERN_WARNING, "bad geometry: first data"
3319 "block %u is beyond end of filesystem (%llu)",
3320 le32_to_cpu(es->s_first_data_block),
3321 ext4_blocks_count(es));
e7c95593
ES		 3322 goto failed_mount;
3323 }
bd81d8ee
LV		 3324 blocks_count = (ext4_blocks_count(es) -
3325 le32_to_cpu(es->s_first_data_block) +
3326 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3327 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4ec11028 3328 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
b31e1552 3329 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
4ec11028 3330 "(block count %llu, first data block %u, "
b31e1552 3331 "blocks per group %lu)", sbi->s_groups_count,
4ec11028
TT		 3332 ext4_blocks_count(es),
3333 le32_to_cpu(es->s_first_data_block),
3334 EXT4_BLOCKS_PER_GROUP(sb));
3335 goto failed_mount;
3336 }
bd81d8ee 3337 sbi->s_groups_count = blocks_count;
fb0a387d
ES		 3338 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3339 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
617ba13b
MC		 3340 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3341 EXT4_DESC_PER_BLOCK(sb);
2b2d6d01 3342 sbi->s_group_desc = kmalloc(db_count * sizeof(struct buffer_head *),
ac27a0ec
DK		 3343 GFP_KERNEL);
3344 if (sbi->s_group_desc == NULL) {
b31e1552 3345 ext4_msg(sb, KERN_ERR, "not enough memory");
ac27a0ec
DK		 3346 goto failed_mount;
3347 }
3348
3244fcb1 3349#ifdef CONFIG_PROC_FS
9f6200bb
TT		 3350 if (ext4_proc_root)
3351 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
3244fcb1 3352#endif
240799cd 3353
705895b6 3354 bgl_lock_init(sbi->s_blockgroup_lock);
ac27a0ec
DK		 3355
3356 for (i = 0; i < db_count; i++) {
70bbb3e0 3357 block = descriptor_loc(sb, logical_sb_block, i);
ac27a0ec
DK		 3358 sbi->s_group_desc[i] = sb_bread(sb, block);
3359 if (!sbi->s_group_desc[i]) {
b31e1552
ES		 3360 ext4_msg(sb, KERN_ERR,
3361 "can't read group descriptor %d", i);
ac27a0ec
DK		 3362 db_count = i;
3363 goto failed_mount2;
3364 }
3365 }
bfff6873 3366 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
b31e1552 3367 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
ac27a0ec
DK		 3368 goto failed_mount2;
3369 }
772cb7c8
JS		 3370 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3371 if (!ext4_fill_flex_info(sb)) {
b31e1552
ES		 3372 ext4_msg(sb, KERN_ERR,
3373 "unable to initialize "
3374 "flex_bg meta info!");
772cb7c8
JS		 3375 goto failed_mount2;
3376 }
3377
ac27a0ec
DK		 3378 sbi->s_gdb_count = db_count;
3379 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3380 spin_lock_init(&sbi->s_next_gen_lock);
3381
ce7e010a
TT		 3382 err = percpu_counter_init(&sbi->s_freeblocks_counter,
3383 ext4_count_free_blocks(sb));
3384 if (!err) {
3385 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3386 ext4_count_free_inodes(sb));
3387 }
3388 if (!err) {
3389 err = percpu_counter_init(&sbi->s_dirs_counter,
3390 ext4_count_dirs(sb));
3391 }
3392 if (!err) {
3393 err = percpu_counter_init(&sbi->s_dirtyblocks_counter, 0);
3394 }
3395 if (err) {
3396 ext4_msg(sb, KERN_ERR, "insufficient memory");
3397 goto failed_mount3;
3398 }
3399
c9de560d 3400 sbi->s_stripe = ext4_get_stripe_size(sbi);
55138e0b 3401 sbi->s_max_writeback_mb_bump = 128;
c9de560d 3402
ac27a0ec
DK		 3403 /*
3404 * set up enough so that it can read an inode
3405 */
9ca92389
TT		 3406 if (!test_opt(sb, NOLOAD) &&
3407 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3408 sb->s_op = &ext4_sops;
3409 else
3410 sb->s_op = &ext4_nojournal_sops;
617ba13b
MC		 3411 sb->s_export_op = &ext4_export_ops;
3412 sb->s_xattr = ext4_xattr_handlers;
ac27a0ec 3413#ifdef CONFIG_QUOTA
617ba13b
MC		 3414 sb->s_qcop = &ext4_qctl_operations;
3415 sb->dq_op = &ext4_quota_operations;
ac27a0ec
DK		 3416#endif
3417 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3b9d4ed2 3418 mutex_init(&sbi->s_orphan_lock);
32ed5058 3419 mutex_init(&sbi->s_resize_lock);
ac27a0ec
DK		 3420
3421 sb->s_root = NULL;
3422
3423 needs_recovery = (es->s_last_orphan != 0 ||
617ba13b
MC		 3424 EXT4_HAS_INCOMPAT_FEATURE(sb,
3425 EXT4_FEATURE_INCOMPAT_RECOVER));
ac27a0ec
DK		 3426
3427 /*
3428 * The first inode we look at is the journal inode. Don't try
3429 * root first: it may be modified in the journal!
3430 */
3431 if (!test_opt(sb, NOLOAD) &&
617ba13b
MC		 3432 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3433 if (ext4_load_journal(sb, es, journal_devnum))
ac27a0ec 3434 goto failed_mount3;
0390131b
FM		 3435 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3436 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
b31e1552
ES		 3437 ext4_msg(sb, KERN_ERR, "required journal recovery "
3438 "suppressed and not mounted read-only");
744692dc 3439 goto failed_mount_wq;
ac27a0ec 3440 } else {
fd8c37ec
TT		 3441 clear_opt(sb, DATA_FLAGS);
3442 set_opt(sb, WRITEBACK_DATA);
0390131b
FM		 3443 sbi->s_journal = NULL;
3444 needs_recovery = 0;
3445 goto no_journal;
ac27a0ec
DK		 3446 }
3447
eb40a09c
JS		 3448 if (ext4_blocks_count(es) > 0xffffffffULL &&
3449 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3450 JBD2_FEATURE_INCOMPAT_64BIT)) {
b31e1552 3451 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
744692dc 3452 goto failed_mount_wq;
eb40a09c
JS		 3453 }
3454
d4da6c9c
LT		 3455 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3456 jbd2_journal_set_features(sbi->s_journal,
3457 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
818d276c 3458 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
d4da6c9c
LT		 3459 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3460 jbd2_journal_set_features(sbi->s_journal,
3461 JBD2_FEATURE_COMPAT_CHECKSUM, 0, 0);
818d276c
GS		 3462 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3463 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
d4da6c9c
LT		 3464 } else {
3465 jbd2_journal_clear_features(sbi->s_journal,
3466 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3467 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3468 }
818d276c 3469
ac27a0ec
DK		 3470 /* We have now updated the journal if required, so we can
3471 * validate the data journaling mode. */
3472 switch (test_opt(sb, DATA_FLAGS)) {
3473 case 0:
3474 /* No mode set, assume a default based on the journal
63f57933
AM		 3475 * capabilities: ORDERED_DATA if the journal can
3476 * cope, else JOURNAL_DATA
3477 */
dab291af
MC		 3478 if (jbd2_journal_check_available_features
3479 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
fd8c37ec 3480 set_opt(sb, ORDERED_DATA);
ac27a0ec 3481 else
fd8c37ec 3482 set_opt(sb, JOURNAL_DATA);
ac27a0ec
DK		 3483 break;
3484
617ba13b
MC		 3485 case EXT4_MOUNT_ORDERED_DATA:
3486 case EXT4_MOUNT_WRITEBACK_DATA:
dab291af
MC		 3487 if (!jbd2_journal_check_available_features
3488 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
b31e1552
ES		 3489 ext4_msg(sb, KERN_ERR, "Journal does not support "
3490 "requested data journaling mode");
744692dc 3491 goto failed_mount_wq;
ac27a0ec
DK		 3492 }
3493 default:
3494 break;
3495 }
b3881f74 3496 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
ac27a0ec 3497
ce7e010a
TT		 3498 /*
3499 * The journal may have updated the bg summary counts, so we
3500 * need to update the global counters.
3501 */
3502 percpu_counter_set(&sbi->s_freeblocks_counter,
3503 ext4_count_free_blocks(sb));
3504 percpu_counter_set(&sbi->s_freeinodes_counter,
3505 ext4_count_free_inodes(sb));
3506 percpu_counter_set(&sbi->s_dirs_counter,
3507 ext4_count_dirs(sb));
3508 percpu_counter_set(&sbi->s_dirtyblocks_counter, 0);
206f7ab4 3509
ce7e010a 3510no_journal:
4c0425ff
MC		 3511 EXT4_SB(sb)->dio_unwritten_wq = create_workqueue("ext4-dio-unwritten");
3512 if (!EXT4_SB(sb)->dio_unwritten_wq) {
3513 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
3514 goto failed_mount_wq;
3515 }
3516
ac27a0ec 3517 /*
dab291af 3518 * The jbd2_journal_load will have done any necessary log recovery,
ac27a0ec
DK		 3519 * so we can safely mount the rest of the filesystem now.
3520 */
3521
1d1fe1ee
DH		 3522 root = ext4_iget(sb, EXT4_ROOT_INO);
3523 if (IS_ERR(root)) {
b31e1552 3524 ext4_msg(sb, KERN_ERR, "get root inode failed");
1d1fe1ee 3525 ret = PTR_ERR(root);
ac27a0ec
DK		 3526 goto failed_mount4;
3527 }
3528 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
1d1fe1ee 3529 iput(root);
b31e1552 3530 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
ac27a0ec
DK		 3531 goto failed_mount4;
3532 }
1d1fe1ee
DH		 3533 sb->s_root = d_alloc_root(root);
3534 if (!sb->s_root) {
b31e1552 3535 ext4_msg(sb, KERN_ERR, "get root dentry failed");
1d1fe1ee
DH		 3536 iput(root);
3537 ret = -ENOMEM;
3538 goto failed_mount4;
3539 }
ac27a0ec 3540
2b2d6d01 3541 ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY);
ef7f3835
KS		 3542
3543 /* determine the minimum size of new large inodes, if present */
3544 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3545 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3546 EXT4_GOOD_OLD_INODE_SIZE;
3547 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3548 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
3549 if (sbi->s_want_extra_isize <
3550 le16_to_cpu(es->s_want_extra_isize))
3551 sbi->s_want_extra_isize =
3552 le16_to_cpu(es->s_want_extra_isize);
3553 if (sbi->s_want_extra_isize <
3554 le16_to_cpu(es->s_min_extra_isize))
3555 sbi->s_want_extra_isize =
3556 le16_to_cpu(es->s_min_extra_isize);
3557 }
3558 }
3559 /* Check if enough inode space is available */
3560 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3561 sbi->s_inode_size) {
3562 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3563 EXT4_GOOD_OLD_INODE_SIZE;
b31e1552
ES		 3564 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3565 "available");
ef7f3835
KS		 3566 }
3567
90576c0b
TT		 3568 if (test_opt(sb, DELALLOC) &&
3569 (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)) {
b31e1552
ES		 3570 ext4_msg(sb, KERN_WARNING, "Ignoring delalloc option - "
3571 "requested data journaling mode");
fd8c37ec 3572 clear_opt(sb, DELALLOC);
90576c0b 3573 }
744692dc
JZ		 3574 if (test_opt(sb, DIOREAD_NOLOCK)) {
3575 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3576 ext4_msg(sb, KERN_WARNING, "Ignoring dioread_nolock "
3577 "option - requested data journaling mode");
fd8c37ec 3578 clear_opt(sb, DIOREAD_NOLOCK);
744692dc
JZ		 3579 }
3580 if (sb->s_blocksize < PAGE_SIZE) {
3581 ext4_msg(sb, KERN_WARNING, "Ignoring dioread_nolock "
3582 "option - block size is too small");
fd8c37ec 3583 clear_opt(sb, DIOREAD_NOLOCK);
744692dc
JZ		 3584 }
3585 }
c2774d84 3586
6fd058f7
TT		 3587 err = ext4_setup_system_zone(sb);
3588 if (err) {
b31e1552 3589 ext4_msg(sb, KERN_ERR, "failed to initialize system "
fbe845dd 3590 "zone (%d)", err);
6fd058f7
TT		 3591 goto failed_mount4;
3592 }
3593
c2774d84
AK		 3594 ext4_ext_init(sb);
3595 err = ext4_mb_init(sb, needs_recovery);
3596 if (err) {
421f91d2 3597 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
b31e1552 3598 err);
c2774d84
AK		 3599 goto failed_mount4;
3600 }
3601
bfff6873
LC		 3602 err = ext4_register_li_request(sb, first_not_zeroed);
3603 if (err)
3604 goto failed_mount4;
3605
3197ebdb
TT		 3606 sbi->s_kobj.kset = ext4_kset;
3607 init_completion(&sbi->s_kobj_unregister);
3608 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
3609 "%s", sb->s_id);
3610 if (err) {
3611 ext4_mb_release(sb);
3612 ext4_ext_release(sb);
3613 goto failed_mount4;
3614 };
3615
617ba13b
MC		 3616 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3617 ext4_orphan_cleanup(sb, es);
3618 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
0390131b 3619 if (needs_recovery) {
b31e1552 3620 ext4_msg(sb, KERN_INFO, "recovery complete");
0390131b
FM		 3621 ext4_mark_recovery_complete(sb, es);
3622 }
3623 if (EXT4_SB(sb)->s_journal) {
3624 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3625 descr = " journalled data mode";
3626 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3627 descr = " ordered data mode";
3628 else
3629 descr = " writeback data mode";
3630 } else
3631 descr = "out journal";
3632
d4c402d9 3633 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
8b67f04a
TT		 3634 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
3635 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
ac27a0ec 3636
66e61a9e
TT		 3637 init_timer(&sbi->s_err_report);
3638 sbi->s_err_report.function = print_daily_error_info;
3639 sbi->s_err_report.data = (unsigned long) sb;
3640 if (es->s_error_count)
3641 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
ac27a0ec 3642
d4c402d9 3643 kfree(orig_data);
ac27a0ec
DK		 3644 return 0;
3645
617ba13b 3646cantfind_ext4:
ac27a0ec 3647 if (!silent)
b31e1552 3648 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
ac27a0ec
DK		 3649 goto failed_mount;
3650
3651failed_mount4:
b31e1552 3652 ext4_msg(sb, KERN_ERR, "mount failed");
4c0425ff
MC		 3653 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
3654failed_mount_wq:
6fd058f7 3655 ext4_release_system_zone(sb);
0390131b
FM		 3656 if (sbi->s_journal) {
3657 jbd2_journal_destroy(sbi->s_journal);
3658 sbi->s_journal = NULL;
3659 }
ac27a0ec 3660failed_mount3:
c5ca7c76
TT		 3661 if (sbi->s_flex_groups) {
3662 if (is_vmalloc_addr(sbi->s_flex_groups))
3663 vfree(sbi->s_flex_groups);
3664 else
3665 kfree(sbi->s_flex_groups);
3666 }
ce7e010a
TT		 3667 percpu_counter_destroy(&sbi->s_freeblocks_counter);
3668 percpu_counter_destroy(&sbi->s_freeinodes_counter);
3669 percpu_counter_destroy(&sbi->s_dirs_counter);
3670 percpu_counter_destroy(&sbi->s_dirtyblocks_counter);
ac27a0ec
DK		 3671failed_mount2:
3672 for (i = 0; i < db_count; i++)
3673 brelse(sbi->s_group_desc[i]);
3674 kfree(sbi->s_group_desc);
3675failed_mount:
240799cd 3676 if (sbi->s_proc) {
9f6200bb 3677 remove_proc_entry(sb->s_id, ext4_proc_root);
240799cd 3678 }
ac27a0ec
DK		 3679#ifdef CONFIG_QUOTA
3680 for (i = 0; i < MAXQUOTAS; i++)
3681 kfree(sbi->s_qf_names[i]);
3682#endif
617ba13b 3683 ext4_blkdev_remove(sbi);
ac27a0ec
DK		 3684 brelse(bh);
3685out_fail:
3686 sb->s_fs_info = NULL;
f6830165 3687 kfree(sbi->s_blockgroup_lock);
ac27a0ec 3688 kfree(sbi);
dcc7dae3 3689out_free_orig:
d4c402d9 3690 kfree(orig_data);
1d1fe1ee 3691 return ret;
ac27a0ec
DK		 3692}
3693
3694/*
3695 * Setup any per-fs journal parameters now. We'll do this both on
3696 * initial mount, once the journal has been initialised but before we've
3697 * done any recovery; and again on any subsequent remount.
3698 */
617ba13b 3699static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
ac27a0ec 3700{
617ba13b 3701 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 3702
30773840
TT		 3703 journal->j_commit_interval = sbi->s_commit_interval;
3704 journal->j_min_batch_time = sbi->s_min_batch_time;
3705 journal->j_max_batch_time = sbi->s_max_batch_time;
ac27a0ec 3706
a931da6a 3707 write_lock(&journal->j_state_lock);
ac27a0ec 3708 if (test_opt(sb, BARRIER))
dab291af 3709 journal->j_flags |= JBD2_BARRIER;
ac27a0ec 3710 else
dab291af 3711 journal->j_flags &= ~JBD2_BARRIER;
5bf5683a
HK		 3712 if (test_opt(sb, DATA_ERR_ABORT))
3713 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
3714 else
3715 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
a931da6a 3716 write_unlock(&journal->j_state_lock);
ac27a0ec
DK		 3717}
3718
617ba13b 3719static journal_t *ext4_get_journal(struct super_block *sb,
ac27a0ec
DK		 3720 unsigned int journal_inum)
3721{
3722 struct inode *journal_inode;
3723 journal_t *journal;
3724
0390131b
FM		 3725 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3726
ac27a0ec
DK		 3727 /* First, test for the existence of a valid inode on disk. Bad
3728 * things happen if we iget() an unused inode, as the subsequent
3729 * iput() will try to delete it. */
3730
1d1fe1ee
DH		 3731 journal_inode = ext4_iget(sb, journal_inum);
3732 if (IS_ERR(journal_inode)) {
b31e1552 3733 ext4_msg(sb, KERN_ERR, "no journal found");
ac27a0ec
DK		 3734 return NULL;
3735 }
3736 if (!journal_inode->i_nlink) {
3737 make_bad_inode(journal_inode);
3738 iput(journal_inode);
b31e1552 3739 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
ac27a0ec
DK		 3740 return NULL;
3741 }
3742
e5f8eab8 3743 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
ac27a0ec 3744 journal_inode, journal_inode->i_size);
1d1fe1ee 3745 if (!S_ISREG(journal_inode->i_mode)) {
b31e1552 3746 ext4_msg(sb, KERN_ERR, "invalid journal inode");
ac27a0ec
DK		 3747 iput(journal_inode);
3748 return NULL;
3749 }
3750
dab291af 3751 journal = jbd2_journal_init_inode(journal_inode);
ac27a0ec 3752 if (!journal) {
b31e1552 3753 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
ac27a0ec
DK		 3754 iput(journal_inode);
3755 return NULL;
3756 }
3757 journal->j_private = sb;
617ba13b 3758 ext4_init_journal_params(sb, journal);
ac27a0ec
DK		 3759 return journal;
3760}
3761
617ba13b 3762static journal_t *ext4_get_dev_journal(struct super_block *sb,
ac27a0ec
DK		 3763 dev_t j_dev)
3764{
2b2d6d01 3765 struct buffer_head *bh;
ac27a0ec 3766 journal_t *journal;
617ba13b
MC		 3767 ext4_fsblk_t start;
3768 ext4_fsblk_t len;
ac27a0ec 3769 int hblock, blocksize;
617ba13b 3770 ext4_fsblk_t sb_block;
ac27a0ec 3771 unsigned long offset;
2b2d6d01 3772 struct ext4_super_block *es;
ac27a0ec
DK		 3773 struct block_device *bdev;
3774
0390131b
FM		 3775 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3776
b31e1552 3777 bdev = ext4_blkdev_get(j_dev, sb);
ac27a0ec
DK		 3778 if (bdev == NULL)
3779 return NULL;
3780
ac27a0ec 3781 blocksize = sb->s_blocksize;
e1defc4f 3782 hblock = bdev_logical_block_size(bdev);
ac27a0ec 3783 if (blocksize < hblock) {
b31e1552
ES		 3784 ext4_msg(sb, KERN_ERR,
3785 "blocksize too small for journal device");
ac27a0ec
DK		 3786 goto out_bdev;
3787 }
3788
617ba13b
MC		 3789 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
3790 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
ac27a0ec
DK		 3791 set_blocksize(bdev, blocksize);
3792 if (!(bh = __bread(bdev, sb_block, blocksize))) {
b31e1552
ES		 3793 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
3794 "external journal");
ac27a0ec
DK		 3795 goto out_bdev;
3796 }
3797
617ba13b
MC		 3798 es = (struct ext4_super_block *) (((char *)bh->b_data) + offset);
3799 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
ac27a0ec 3800 !(le32_to_cpu(es->s_feature_incompat) &
617ba13b 3801 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
b31e1552
ES		 3802 ext4_msg(sb, KERN_ERR, "external journal has "
3803 "bad superblock");
ac27a0ec
DK		 3804 brelse(bh);
3805 goto out_bdev;
3806 }
3807
617ba13b 3808 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
b31e1552 3809 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
ac27a0ec
DK		 3810 brelse(bh);
3811 goto out_bdev;
3812 }
3813
bd81d8ee 3814 len = ext4_blocks_count(es);
ac27a0ec
DK		 3815 start = sb_block + 1;
3816 brelse(bh); /* we're done with the superblock */
3817
dab291af 3818 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
ac27a0ec
DK		 3819 start, len, blocksize);
3820 if (!journal) {
b31e1552 3821 ext4_msg(sb, KERN_ERR, "failed to create device journal");
ac27a0ec
DK		 3822 goto out_bdev;
3823 }
3824 journal->j_private = sb;
3825 ll_rw_block(READ, 1, &journal->j_sb_buffer);
3826 wait_on_buffer(journal->j_sb_buffer);
3827 if (!buffer_uptodate(journal->j_sb_buffer)) {
b31e1552 3828 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
ac27a0ec
DK		 3829 goto out_journal;
3830 }
3831 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
b31e1552
ES		 3832 ext4_msg(sb, KERN_ERR, "External journal has more than one "
3833 "user (unsupported) - %d",
ac27a0ec
DK		 3834 be32_to_cpu(journal->j_superblock->s_nr_users));
3835 goto out_journal;
3836 }
617ba13b
MC		 3837 EXT4_SB(sb)->journal_bdev = bdev;
3838 ext4_init_journal_params(sb, journal);
ac27a0ec 3839 return journal;
0b8e58a1 3840
ac27a0ec 3841out_journal:
dab291af 3842 jbd2_journal_destroy(journal);
ac27a0ec 3843out_bdev:
617ba13b 3844 ext4_blkdev_put(bdev);
ac27a0ec
DK		 3845 return NULL;
3846}
3847
617ba13b
MC		 3848static int ext4_load_journal(struct super_block *sb,
3849 struct ext4_super_block *es,
ac27a0ec
DK		 3850 unsigned long journal_devnum)
3851{
3852 journal_t *journal;
3853 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
3854 dev_t journal_dev;
3855 int err = 0;
3856 int really_read_only;
3857
0390131b
FM		 3858 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
3859
ac27a0ec
DK		 3860 if (journal_devnum &&
3861 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
b31e1552
ES		 3862 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
3863 "numbers have changed");
ac27a0ec
DK		 3864 journal_dev = new_decode_dev(journal_devnum);
3865 } else
3866 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
3867
3868 really_read_only = bdev_read_only(sb->s_bdev);
3869
3870 /*
3871 * Are we loading a blank journal or performing recovery after a
3872 * crash? For recovery, we need to check in advance whether we
3873 * can get read-write access to the device.
3874 */
617ba13b 3875 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
ac27a0ec 3876 if (sb->s_flags & MS_RDONLY) {
b31e1552
ES		 3877 ext4_msg(sb, KERN_INFO, "INFO: recovery "
3878 "required on readonly filesystem");
ac27a0ec 3879 if (really_read_only) {
b31e1552
ES		 3880 ext4_msg(sb, KERN_ERR, "write access "
3881 "unavailable, cannot proceed");
ac27a0ec
DK		 3882 return -EROFS;
3883 }
b31e1552
ES		 3884 ext4_msg(sb, KERN_INFO, "write access will "
3885 "be enabled during recovery");
ac27a0ec
DK		 3886 }
3887 }
3888
3889 if (journal_inum && journal_dev) {
b31e1552
ES		 3890 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
3891 "and inode journals!");
ac27a0ec
DK		 3892 return -EINVAL;
3893 }
3894
3895 if (journal_inum) {
617ba13b 3896 if (!(journal = ext4_get_journal(sb, journal_inum)))
ac27a0ec
DK		 3897 return -EINVAL;
3898 } else {
617ba13b 3899 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
ac27a0ec
DK		 3900 return -EINVAL;
3901 }
3902
90576c0b 3903 if (!(journal->j_flags & JBD2_BARRIER))
b31e1552 3904 ext4_msg(sb, KERN_INFO, "barriers disabled");
4776004f 3905
ac27a0ec 3906 if (!really_read_only && test_opt(sb, UPDATE_JOURNAL)) {
dab291af 3907 err = jbd2_journal_update_format(journal);
ac27a0ec 3908 if (err) {
b31e1552 3909 ext4_msg(sb, KERN_ERR, "error updating journal");
dab291af 3910 jbd2_journal_destroy(journal);
ac27a0ec
DK		 3911 return err;
3912 }
3913 }
3914
617ba13b 3915 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
dab291af 3916 err = jbd2_journal_wipe(journal, !really_read_only);
1c13d5c0
TT		 3917 if (!err) {
3918 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
3919 if (save)
3920 memcpy(save, ((char *) es) +
3921 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
dab291af 3922 err = jbd2_journal_load(journal);
1c13d5c0
TT		 3923 if (save)
3924 memcpy(((char *) es) + EXT4_S_ERR_START,
3925 save, EXT4_S_ERR_LEN);
3926 kfree(save);
3927 }
ac27a0ec
DK		 3928
3929 if (err) {
b31e1552 3930 ext4_msg(sb, KERN_ERR, "error loading journal");
dab291af 3931 jbd2_journal_destroy(journal);
ac27a0ec
DK		 3932 return err;
3933 }
3934
617ba13b
MC		 3935 EXT4_SB(sb)->s_journal = journal;
3936 ext4_clear_journal_err(sb, es);
ac27a0ec 3937
c41303ce 3938 if (!really_read_only && journal_devnum &&
ac27a0ec
DK		 3939 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
3940 es->s_journal_dev = cpu_to_le32(journal_devnum);
ac27a0ec
DK		 3941
3942 /* Make sure we flush the recovery flag to disk. */
e2d67052 3943 ext4_commit_super(sb, 1);
ac27a0ec
DK		 3944 }
3945
3946 return 0;
3947}
3948
e2d67052 3949static int ext4_commit_super(struct super_block *sb, int sync)
ac27a0ec 3950{
e2d67052 3951 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
617ba13b 3952 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
c4be0c1d 3953 int error = 0;
ac27a0ec
DK		 3954
3955 if (!sbh)
c4be0c1d 3956 return error;
914258bf
TT		 3957 if (buffer_write_io_error(sbh)) {
3958 /*
3959 * Oh, dear. A previous attempt to write the
3960 * superblock failed. This could happen because the
3961 * USB device was yanked out. Or it could happen to
3962 * be a transient write error and maybe the block will
3963 * be remapped. Nothing we can do but to retry the
3964 * write and hope for the best.
3965 */
b31e1552
ES		 3966 ext4_msg(sb, KERN_ERR, "previous I/O error to "
3967 "superblock detected");
914258bf
TT		 3968 clear_buffer_write_io_error(sbh);
3969 set_buffer_uptodate(sbh);
3970 }
71290b36
TT		 3971 /*
3972 * If the file system is mounted read-only, don't update the
3973 * superblock write time. This avoids updating the superblock
3974 * write time when we are mounting the root file system
3975 * read/only but we need to replay the journal; at that point,
3976 * for people who are east of GMT and who make their clock
3977 * tick in localtime for Windows bug-for-bug compatibility,
3978 * the clock is set in the future, and this will cause e2fsck
3979 * to complain and force a full file system check.
3980 */
3981 if (!(sb->s_flags & MS_RDONLY))
3982 es->s_wtime = cpu_to_le32(get_seconds());
f613dfcb
TT		 3983 if (sb->s_bdev->bd_part)
3984 es->s_kbytes_written =
3985 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
afc32f7e
TT		 3986 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
3987 EXT4_SB(sb)->s_sectors_written_start) >> 1));
f613dfcb
TT		 3988 else
3989 es->s_kbytes_written =
3990 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
ce7e010a
TT		 3991 ext4_free_blocks_count_set(es, percpu_counter_sum_positive(
3992 &EXT4_SB(sb)->s_freeblocks_counter));
3993 es->s_free_inodes_count =
3994 cpu_to_le32(percpu_counter_sum_positive(
3995 &EXT4_SB(sb)->s_freeinodes_counter));
7234ab2a 3996 sb->s_dirt = 0;
ac27a0ec
DK		 3997 BUFFER_TRACE(sbh, "marking dirty");
3998 mark_buffer_dirty(sbh);
914258bf 3999 if (sync) {
c4be0c1d
TS		 4000 error = sync_dirty_buffer(sbh);
4001 if (error)
4002 return error;
4003
4004 error = buffer_write_io_error(sbh);
4005 if (error) {
b31e1552
ES		 4006 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4007 "superblock");
914258bf
TT		 4008 clear_buffer_write_io_error(sbh);
4009 set_buffer_uptodate(sbh);
4010 }
4011 }
c4be0c1d 4012 return error;
ac27a0ec
DK		 4013}
4014
ac27a0ec
DK		 4015/*
4016 * Have we just finished recovery? If so, and if we are mounting (or
4017 * remounting) the filesystem readonly, then we will end up with a
4018 * consistent fs on disk. Record that fact.
4019 */
2b2d6d01
TT		 4020static void ext4_mark_recovery_complete(struct super_block *sb,
4021 struct ext4_super_block *es)
ac27a0ec 4022{
617ba13b 4023 journal_t *journal = EXT4_SB(sb)->s_journal;
ac27a0ec 4024
0390131b
FM		 4025 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4026 BUG_ON(journal != NULL);
4027 return;
4028 }
dab291af 4029 jbd2_journal_lock_updates(journal);
7ffe1ea8
HK		 4030 if (jbd2_journal_flush(journal) < 0)
4031 goto out;
4032
617ba13b 4033 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
ac27a0ec 4034 sb->s_flags & MS_RDONLY) {
617ba13b 4035 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
e2d67052 4036 ext4_commit_super(sb, 1);
ac27a0ec 4037 }
7ffe1ea8
HK		 4038
4039out:
dab291af 4040 jbd2_journal_unlock_updates(journal);
ac27a0ec
DK		 4041}
4042
4043/*
4044 * If we are mounting (or read-write remounting) a filesystem whose journal
4045 * has recorded an error from a previous lifetime, move that error to the
4046 * main filesystem now.
4047 */
2b2d6d01
TT		 4048static void ext4_clear_journal_err(struct super_block *sb,
4049 struct ext4_super_block *es)
ac27a0ec
DK		 4050{
4051 journal_t *journal;
4052 int j_errno;
4053 const char *errstr;
4054
0390131b
FM		 4055 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4056
617ba13b 4057 journal = EXT4_SB(sb)->s_journal;
ac27a0ec
DK		 4058
4059 /*
4060 * Now check for any error status which may have been recorded in the
617ba13b 4061 * journal by a prior ext4_error() or ext4_abort()
ac27a0ec
DK		 4062 */
4063
dab291af 4064 j_errno = jbd2_journal_errno(journal);
ac27a0ec
DK		 4065 if (j_errno) {
4066 char nbuf[16];
4067
617ba13b 4068 errstr = ext4_decode_error(sb, j_errno, nbuf);
12062ddd 4069 ext4_warning(sb, "Filesystem error recorded "
ac27a0ec 4070 "from previous mount: %s", errstr);
12062ddd 4071 ext4_warning(sb, "Marking fs in need of filesystem check.");
ac27a0ec 4072
617ba13b
MC		 4073 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4074 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
e2d67052 4075 ext4_commit_super(sb, 1);
ac27a0ec 4076
dab291af 4077 jbd2_journal_clear_err(journal);
ac27a0ec
DK		 4078 }
4079}
4080
4081/*
4082 * Force the running and committing transactions to commit,
4083 * and wait on the commit.
4084 */
617ba13b 4085int ext4_force_commit(struct super_block *sb)
ac27a0ec
DK		 4086{
4087 journal_t *journal;
0390131b 4088 int ret = 0;
ac27a0ec
DK		 4089
4090 if (sb->s_flags & MS_RDONLY)
4091 return 0;
4092
617ba13b 4093 journal = EXT4_SB(sb)->s_journal;
6b0310fb 4094 if (journal) {
437f88cc 4095 vfs_check_frozen(sb, SB_FREEZE_TRANS);
0390131b 4096 ret = ext4_journal_force_commit(journal);
6b0310fb 4097 }
0390131b 4098
ac27a0ec
DK		 4099 return ret;
4100}
4101
2b2d6d01 4102static void ext4_write_super(struct super_block *sb)
ac27a0ec 4103{
ebc1ac16 4104 lock_super(sb);
9ca92389 4105 ext4_commit_super(sb, 1);
ebc1ac16 4106 unlock_super(sb);
ac27a0ec
DK		 4107}
4108
617ba13b 4109static int ext4_sync_fs(struct super_block *sb, int wait)
ac27a0ec 4110{
14ce0cb4 4111 int ret = 0;
9eddacf9 4112 tid_t target;
8d5d02e6 4113 struct ext4_sb_info *sbi = EXT4_SB(sb);
ac27a0ec 4114
9bffad1e 4115 trace_ext4_sync_fs(sb, wait);
8d5d02e6
MC		 4116 flush_workqueue(sbi->dio_unwritten_wq);
4117 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
9ca92389 4118 if (wait)
8d5d02e6 4119 jbd2_log_wait_commit(sbi->s_journal, target);
0390131b 4120 }
14ce0cb4 4121 return ret;
ac27a0ec
DK		 4122}
4123
4124/*
4125 * LVM calls this function before a (read-only) snapshot is created. This
4126 * gives us a chance to flush the journal completely and mark the fs clean.
4127 */
c4be0c1d 4128static int ext4_freeze(struct super_block *sb)
ac27a0ec 4129{
c4be0c1d
TS		 4130 int error = 0;
4131 journal_t *journal;
ac27a0ec 4132
9ca92389
TT		 4133 if (sb->s_flags & MS_RDONLY)
4134 return 0;
ac27a0ec 4135
9ca92389 4136 journal = EXT4_SB(sb)->s_journal;
7ffe1ea8 4137
9ca92389
TT		 4138 /* Now we set up the journal barrier. */
4139 jbd2_journal_lock_updates(journal);
ac27a0ec 4140
9ca92389
TT		 4141 /*
4142 * Don't clear the needs_recovery flag if we failed to flush
4143 * the journal.
4144 */
4145 error = jbd2_journal_flush(journal);
6b0310fb
ES		 4146 if (error < 0)
4147 goto out;
9ca92389
TT		 4148
4149 /* Journal blocked and flushed, clear needs_recovery flag. */
4150 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4151 error = ext4_commit_super(sb, 1);
6b0310fb
ES		 4152out:
4153 /* we rely on s_frozen to stop further updates */
4154 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4155 return error;
ac27a0ec
DK		 4156}
4157
4158/*
4159 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4160 * flag here, even though the filesystem is not technically dirty yet.
4161 */
c4be0c1d 4162static int ext4_unfreeze(struct super_block *sb)
ac27a0ec 4163{
9ca92389
TT		 4164 if (sb->s_flags & MS_RDONLY)
4165 return 0;
4166
4167 lock_super(sb);
4168 /* Reset the needs_recovery flag before the fs is unlocked. */
4169 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4170 ext4_commit_super(sb, 1);
4171 unlock_super(sb);
c4be0c1d 4172 return 0;
ac27a0ec
DK		 4173}
4174
673c6100
TT		 4175/*
4176 * Structure to save mount options for ext4_remount's benefit
4177 */
4178struct ext4_mount_options {
4179 unsigned long s_mount_opt;
a2595b8a 4180 unsigned long s_mount_opt2;
673c6100
TT		 4181 uid_t s_resuid;
4182 gid_t s_resgid;
4183 unsigned long s_commit_interval;
4184 u32 s_min_batch_time, s_max_batch_time;
4185#ifdef CONFIG_QUOTA
4186 int s_jquota_fmt;
4187 char *s_qf_names[MAXQUOTAS];
4188#endif
4189};
4190
2b2d6d01 4191static int ext4_remount(struct super_block *sb, int *flags, char *data)
ac27a0ec 4192{
2b2d6d01 4193 struct ext4_super_block *es;
617ba13b
MC		 4194 struct ext4_sb_info *sbi = EXT4_SB(sb);
4195 ext4_fsblk_t n_blocks_count = 0;
ac27a0ec 4196 unsigned long old_sb_flags;
617ba13b 4197 struct ext4_mount_options old_opts;
c79d967d 4198 int enable_quota = 0;
8a266467 4199 ext4_group_t g;
b3881f74 4200 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
ac27a0ec
DK		 4201 int err;
4202#ifdef CONFIG_QUOTA
4203 int i;
4204#endif
d4c402d9 4205 char *orig_data = kstrdup(data, GFP_KERNEL);
ac27a0ec
DK		 4206
4207 /* Store the original options */
bbd6851a 4208 lock_super(sb);
ac27a0ec
DK		 4209 old_sb_flags = sb->s_flags;
4210 old_opts.s_mount_opt = sbi->s_mount_opt;
a2595b8a 4211 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
ac27a0ec
DK		 4212 old_opts.s_resuid = sbi->s_resuid;
4213 old_opts.s_resgid = sbi->s_resgid;
4214 old_opts.s_commit_interval = sbi->s_commit_interval;
30773840
TT		 4215 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4216 old_opts.s_max_batch_time = sbi->s_max_batch_time;
ac27a0ec
DK		 4217#ifdef CONFIG_QUOTA
4218 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4219 for (i = 0; i < MAXQUOTAS; i++)
4220 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
4221#endif
b3881f74
TT		 4222 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4223 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
ac27a0ec
DK		 4224
4225 /*
4226 * Allow the "check" option to be passed as a remount option.
4227 */
b3881f74
TT		 4228 if (!parse_options(data, sb, NULL, &journal_ioprio,
4229 &n_blocks_count, 1)) {
ac27a0ec
DK		 4230 err = -EINVAL;
4231 goto restore_opts;
4232 }
4233
4ab2f15b 4234 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
c67d859e 4235 ext4_abort(sb, "Abort forced by user");
ac27a0ec
DK		 4236
4237 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
482a7425 4238 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
ac27a0ec
DK		 4239
4240 es = sbi->s_es;
4241
b3881f74 4242 if (sbi->s_journal) {
0390131b 4243 ext4_init_journal_params(sb, sbi->s_journal);
b3881f74
TT		 4244 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4245 }
ac27a0ec
DK		 4246
4247 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY) ||
bd81d8ee 4248 n_blocks_count > ext4_blocks_count(es)) {
4ab2f15b 4249 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
ac27a0ec
DK		 4250 err = -EROFS;
4251 goto restore_opts;
4252 }
4253
4254 if (*flags & MS_RDONLY) {
0f0dd62f
CH		 4255 err = dquot_suspend(sb, -1);
4256 if (err < 0)
c79d967d 4257 goto restore_opts;
c79d967d 4258
ac27a0ec
DK		 4259 /*
4260 * First of all, the unconditional stuff we have to do
4261 * to disable replay of the journal when we next remount
4262 */
4263 sb->s_flags |= MS_RDONLY;
4264
4265 /*
4266 * OK, test if we are remounting a valid rw partition
4267 * readonly, and if so set the rdonly flag and then
4268 * mark the partition as valid again.
4269 */
617ba13b
MC		 4270 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4271 (sbi->s_mount_state & EXT4_VALID_FS))
ac27a0ec
DK		 4272 es->s_state = cpu_to_le16(sbi->s_mount_state);
4273
a63c9eb2 4274 if (sbi->s_journal)
0390131b 4275 ext4_mark_recovery_complete(sb, es);
ac27a0ec 4276 } else {
a13fb1a4
ES		 4277 /* Make sure we can mount this feature set readwrite */
4278 if (!ext4_feature_set_ok(sb, 0)) {
ac27a0ec
DK		 4279 err = -EROFS;
4280 goto restore_opts;
4281 }
8a266467
TT		 4282 /*
4283 * Make sure the group descriptor checksums
0b8e58a1 4284 * are sane. If they aren't, refuse to remount r/w.
8a266467
TT		 4285 */
4286 for (g = 0; g < sbi->s_groups_count; g++) {
4287 struct ext4_group_desc *gdp =
4288 ext4_get_group_desc(sb, g, NULL);
4289
4290 if (!ext4_group_desc_csum_verify(sbi, g, gdp)) {
b31e1552
ES		 4291 ext4_msg(sb, KERN_ERR,
4292 "ext4_remount: Checksum for group %u failed (%u!=%u)",
8a266467
TT		 4293 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4294 le16_to_cpu(gdp->bg_checksum));
4295 err = -EINVAL;
4296 goto restore_opts;
4297 }
4298 }
4299
ead6596b
ES		 4300 /*
4301 * If we have an unprocessed orphan list hanging
4302 * around from a previously readonly bdev mount,
4303 * require a full umount/remount for now.
4304 */
4305 if (es->s_last_orphan) {
b31e1552 4306 ext4_msg(sb, KERN_WARNING, "Couldn't "
ead6596b
ES		 4307 "remount RDWR because of unprocessed "
4308 "orphan inode list. Please "
b31e1552 4309 "umount/remount instead");
ead6596b
ES		 4310 err = -EINVAL;
4311 goto restore_opts;
4312 }
4313
ac27a0ec
DK		 4314 /*
4315 * Mounting a RDONLY partition read-write, so reread
4316 * and store the current valid flag. (It may have
4317 * been changed by e2fsck since we originally mounted
4318 * the partition.)
4319 */
0390131b
FM		 4320 if (sbi->s_journal)
4321 ext4_clear_journal_err(sb, es);
ac27a0ec 4322 sbi->s_mount_state = le16_to_cpu(es->s_state);
617ba13b 4323 if ((err = ext4_group_extend(sb, es, n_blocks_count)))
ac27a0ec 4324 goto restore_opts;
2b2d6d01 4325 if (!ext4_setup_super(sb, es, 0))
ac27a0ec 4326 sb->s_flags &= ~MS_RDONLY;
c79d967d 4327 enable_quota = 1;
ac27a0ec
DK		 4328 }
4329 }
bfff6873
LC		 4330
4331 /*
4332 * Reinitialize lazy itable initialization thread based on
4333 * current settings
4334 */
4335 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4336 ext4_unregister_li_request(sb);
4337 else {
4338 ext4_group_t first_not_zeroed;
4339 first_not_zeroed = ext4_has_uninit_itable(sb);
4340 ext4_register_li_request(sb, first_not_zeroed);
4341 }
4342
6fd058f7 4343 ext4_setup_system_zone(sb);
0390131b 4344 if (sbi->s_journal == NULL)
e2d67052 4345 ext4_commit_super(sb, 1);
0390131b 4346
ac27a0ec
DK		 4347#ifdef CONFIG_QUOTA
4348 /* Release old quota file names */
4349 for (i = 0; i < MAXQUOTAS; i++)
4350 if (old_opts.s_qf_names[i] &&
4351 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4352 kfree(old_opts.s_qf_names[i]);
4353#endif
bbd6851a 4354 unlock_super(sb);
c79d967d 4355 if (enable_quota)
0f0dd62f 4356 dquot_resume(sb, -1);
d4c402d9
CW		 4357
4358 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4359 kfree(orig_data);
ac27a0ec 4360 return 0;
0b8e58a1 4361
ac27a0ec
DK		 4362restore_opts:
4363 sb->s_flags = old_sb_flags;
4364 sbi->s_mount_opt = old_opts.s_mount_opt;
a2595b8a 4365 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
ac27a0ec
DK		 4366 sbi->s_resuid = old_opts.s_resuid;
4367 sbi->s_resgid = old_opts.s_resgid;
4368 sbi->s_commit_interval = old_opts.s_commit_interval;
30773840
TT		 4369 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4370 sbi->s_max_batch_time = old_opts.s_max_batch_time;
ac27a0ec
DK		 4371#ifdef CONFIG_QUOTA
4372 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4373 for (i = 0; i < MAXQUOTAS; i++) {
4374 if (sbi->s_qf_names[i] &&
4375 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4376 kfree(sbi->s_qf_names[i]);
4377 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4378 }
4379#endif
bbd6851a 4380 unlock_super(sb);
d4c402d9 4381 kfree(orig_data);
ac27a0ec
DK		 4382 return err;
4383}
4384
2b2d6d01 4385static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
ac27a0ec
DK		 4386{
4387 struct super_block *sb = dentry->d_sb;
617ba13b
MC		 4388 struct ext4_sb_info *sbi = EXT4_SB(sb);
4389 struct ext4_super_block *es = sbi->s_es;
960cc398 4390 u64 fsid;
ac27a0ec 4391
5e70030d
BP		 4392 if (test_opt(sb, MINIX_DF)) {
4393 sbi->s_overhead_last = 0;
6bc9feff 4394 } else if (sbi->s_blocks_last != ext4_blocks_count(es)) {
8df9675f 4395 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
5e70030d 4396 ext4_fsblk_t overhead = 0;
ac27a0ec
DK		 4397
4398 /*
5e70030d
BP		 4399 * Compute the overhead (FS structures). This is constant
4400 * for a given filesystem unless the number of block groups
4401 * changes so we cache the previous value until it does.
ac27a0ec
DK		 4402 */
4403
4404 /*
4405 * All of the blocks before first_data_block are
4406 * overhead
4407 */
4408 overhead = le32_to_cpu(es->s_first_data_block);
4409
4410 /*
4411 * Add the overhead attributed to the superblock and
4412 * block group descriptors. If the sparse superblocks
4413 * feature is turned on, then not all groups have this.
4414 */
4415 for (i = 0; i < ngroups; i++) {
617ba13b
MC		 4416 overhead += ext4_bg_has_super(sb, i) +
4417 ext4_bg_num_gdb(sb, i);
ac27a0ec
DK		 4418 cond_resched();
4419 }
4420
4421 /*
4422 * Every block group has an inode bitmap, a block
4423 * bitmap, and an inode table.
4424 */
5e70030d
BP		 4425 overhead += ngroups * (2 + sbi->s_itb_per_group);
4426 sbi->s_overhead_last = overhead;
4427 smp_wmb();
6bc9feff 4428 sbi->s_blocks_last = ext4_blocks_count(es);
ac27a0ec
DK		 4429 }
4430
617ba13b 4431 buf->f_type = EXT4_SUPER_MAGIC;
ac27a0ec 4432 buf->f_bsize = sb->s_blocksize;
5e70030d 4433 buf->f_blocks = ext4_blocks_count(es) - sbi->s_overhead_last;
6bc6e63f
AK		 4434 buf->f_bfree = percpu_counter_sum_positive(&sbi->s_freeblocks_counter) -
4435 percpu_counter_sum_positive(&sbi->s_dirtyblocks_counter);
bd81d8ee
LV		 4436 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
4437 if (buf->f_bfree < ext4_r_blocks_count(es))
ac27a0ec
DK		 4438 buf->f_bavail = 0;
4439 buf->f_files = le32_to_cpu(es->s_inodes_count);
52d9f3b4 4440 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
617ba13b 4441 buf->f_namelen = EXT4_NAME_LEN;
960cc398
PE		 4442 fsid = le64_to_cpup((void *)es->s_uuid) ^
4443 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4444 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4445 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
0b8e58a1 4446
ac27a0ec
DK		 4447 return 0;
4448}
4449
0b8e58a1
AD		 4450/* Helper function for writing quotas on sync - we need to start transaction
4451 * before quota file is locked for write. Otherwise the are possible deadlocks:
ac27a0ec 4452 * Process 1 Process 2
617ba13b 4453 * ext4_create() quota_sync()
a269eb18 4454 * jbd2_journal_start() write_dquot()
871a2931 4455 * dquot_initialize() down(dqio_mutex)
dab291af 4456 * down(dqio_mutex) jbd2_journal_start()
ac27a0ec
DK		 4457 *
4458 */
4459
4460#ifdef CONFIG_QUOTA
4461
4462static inline struct inode *dquot_to_inode(struct dquot *dquot)
4463{
4464 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
4465}
4466
617ba13b 4467static int ext4_write_dquot(struct dquot *dquot)
ac27a0ec
DK		 4468{
4469 int ret, err;
4470 handle_t *handle;
4471 struct inode *inode;
4472
4473 inode = dquot_to_inode(dquot);
617ba13b 4474 handle = ext4_journal_start(inode,
0b8e58a1 4475 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 4476 if (IS_ERR(handle))
4477 return PTR_ERR(handle);
4478 ret = dquot_commit(dquot);
617ba13b 4479 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4480 if (!ret)
4481 ret = err;
4482 return ret;
4483}
4484
617ba13b 4485static int ext4_acquire_dquot(struct dquot *dquot)
ac27a0ec
DK		 4486{
4487 int ret, err;
4488 handle_t *handle;
4489
617ba13b 4490 handle = ext4_journal_start(dquot_to_inode(dquot),
0b8e58a1 4491 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
ac27a0ec
DK		 4492 if (IS_ERR(handle))
4493 return PTR_ERR(handle);
4494 ret = dquot_acquire(dquot);
617ba13b 4495 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4496 if (!ret)
4497 ret = err;
4498 return ret;
4499}
4500
617ba13b 4501static int ext4_release_dquot(struct dquot *dquot)
ac27a0ec
DK		 4502{
4503 int ret, err;
4504 handle_t *handle;
4505
617ba13b 4506 handle = ext4_journal_start(dquot_to_inode(dquot),
0b8e58a1 4507 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
9c3013e9
JK		 4508 if (IS_ERR(handle)) {
4509 /* Release dquot anyway to avoid endless cycle in dqput() */
4510 dquot_release(dquot);
ac27a0ec 4511 return PTR_ERR(handle);
9c3013e9 4512 }
ac27a0ec 4513 ret = dquot_release(dquot);
617ba13b 4514 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4515 if (!ret)
4516 ret = err;
4517 return ret;
4518}
4519
617ba13b 4520static int ext4_mark_dquot_dirty(struct dquot *dquot)
ac27a0ec 4521{
2c8be6b2 4522 /* Are we journaling quotas? */
617ba13b
MC		 4523 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
4524 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
ac27a0ec 4525 dquot_mark_dquot_dirty(dquot);
617ba13b 4526 return ext4_write_dquot(dquot);
ac27a0ec
DK		 4527 } else {
4528 return dquot_mark_dquot_dirty(dquot);
4529 }
4530}
4531
617ba13b 4532static int ext4_write_info(struct super_block *sb, int type)
ac27a0ec
DK		 4533{
4534 int ret, err;
4535 handle_t *handle;
4536
4537 /* Data block + inode block */
617ba13b 4538 handle = ext4_journal_start(sb->s_root->d_inode, 2);
ac27a0ec
DK		 4539 if (IS_ERR(handle))
4540 return PTR_ERR(handle);
4541 ret = dquot_commit_info(sb, type);
617ba13b 4542 err = ext4_journal_stop(handle);
ac27a0ec
DK		 4543 if (!ret)
4544 ret = err;
4545 return ret;
4546}
4547
4548/*
4549 * Turn on quotas during mount time - we need to find
4550 * the quota file and such...
4551 */
617ba13b 4552static int ext4_quota_on_mount(struct super_block *sb, int type)
ac27a0ec 4553{
287a8095
CH		 4554 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4555 EXT4_SB(sb)->s_jquota_fmt, type);
ac27a0ec
DK		 4556}
4557
4558/*
4559 * Standard function to be called on quota_on
4560 */
617ba13b 4561static int ext4_quota_on(struct super_block *sb, int type, int format_id,
f00c9e44 4562 struct path *path)
ac27a0ec
DK		 4563{
4564 int err;
ac27a0ec
DK		 4565
4566 if (!test_opt(sb, QUOTA))
4567 return -EINVAL;
0623543b 4568
ac27a0ec 4569 /* Quotafile not on the same filesystem? */
f00c9e44 4570 if (path->mnt->mnt_sb != sb)
ac27a0ec 4571 return -EXDEV;
0623543b
JK		 4572 /* Journaling quota? */
4573 if (EXT4_SB(sb)->s_qf_names[type]) {
2b2d6d01 4574 /* Quotafile not in fs root? */
f00c9e44 4575 if (path->dentry->d_parent != sb->s_root)
b31e1552
ES		 4576 ext4_msg(sb, KERN_WARNING,
4577 "Quota file not on filesystem root. "
4578 "Journaled quota will not work");
2b2d6d01 4579 }
0623543b
JK		 4580
4581 /*
4582 * When we journal data on quota file, we have to flush journal to see
4583 * all updates to the file when we bypass pagecache...
4584 */
0390131b 4585 if (EXT4_SB(sb)->s_journal &&
f00c9e44 4586 ext4_should_journal_data(path->dentry->d_inode)) {
0623543b
JK		 4587 /*
4588 * We don't need to lock updates but journal_flush() could
4589 * otherwise be livelocked...
4590 */
4591 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
7ffe1ea8 4592 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
0623543b 4593 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
f00c9e44 4594 if (err)
7ffe1ea8 4595 return err;
0623543b
JK		 4596 }
4597
f00c9e44 4598 return dquot_quota_on(sb, type, format_id, path);
ac27a0ec
DK		 4599}
4600
ca0e05e4
DM		 4601static int ext4_quota_off(struct super_block *sb, int type)
4602{
87009d86
DM		 4603 /* Force all delayed allocation blocks to be allocated.
4604 * Caller already holds s_umount sem */
4605 if (test_opt(sb, DELALLOC))
ca0e05e4 4606 sync_filesystem(sb);
ca0e05e4
DM		 4607
4608 return dquot_quota_off(sb, type);
4609}
4610
ac27a0ec
DK		 4611/* Read data from quotafile - avoid pagecache and such because we cannot afford
4612 * acquiring the locks... As quota files are never truncated and quota code
4613 * itself serializes the operations (and noone else should touch the files)
4614 * we don't have to be afraid of races */
617ba13b 4615static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
ac27a0ec
DK		 4616 size_t len, loff_t off)
4617{
4618 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 4619 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 4620 int err = 0;
4621 int offset = off & (sb->s_blocksize - 1);
4622 int tocopy;
4623 size_t toread;
4624 struct buffer_head *bh;
4625 loff_t i_size = i_size_read(inode);
4626
4627 if (off > i_size)
4628 return 0;
4629 if (off+len > i_size)
4630 len = i_size-off;
4631 toread = len;
4632 while (toread > 0) {
4633 tocopy = sb->s_blocksize - offset < toread ?
4634 sb->s_blocksize - offset : toread;
617ba13b 4635 bh = ext4_bread(NULL, inode, blk, 0, &err);
ac27a0ec
DK		 4636 if (err)
4637 return err;
4638 if (!bh) /* A hole? */
4639 memset(data, 0, tocopy);
4640 else
4641 memcpy(data, bh->b_data+offset, tocopy);
4642 brelse(bh);
4643 offset = 0;
4644 toread -= tocopy;
4645 data += tocopy;
4646 blk++;
4647 }
4648 return len;
4649}
4650
4651/* Write to quotafile (we know the transaction is already started and has
4652 * enough credits) */
617ba13b 4653static ssize_t ext4_quota_write(struct super_block *sb, int type,
ac27a0ec
DK		 4654 const char *data, size_t len, loff_t off)
4655{
4656 struct inode *inode = sb_dqopt(sb)->files[type];
725d26d3 4657 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
ac27a0ec
DK		 4658 int err = 0;
4659 int offset = off & (sb->s_blocksize - 1);
ac27a0ec
DK		 4660 struct buffer_head *bh;
4661 handle_t *handle = journal_current_handle();
4662
0390131b 4663 if (EXT4_SB(sb)->s_journal && !handle) {
b31e1552
ES		 4664 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4665 " cancelled because transaction is not started",
9c3013e9
JK		 4666 (unsigned long long)off, (unsigned long long)len);
4667 return -EIO;
4668 }
67eeb568
DM		 4669 /*
4670 * Since we account only one data block in transaction credits,
4671 * then it is impossible to cross a block boundary.
4672 */
4673 if (sb->s_blocksize - offset < len) {
4674 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4675 " cancelled because not block aligned",
4676 (unsigned long long)off, (unsigned long long)len);
4677 return -EIO;
4678 }
4679
ac27a0ec 4680 mutex_lock_nested(&inode->i_mutex, I_MUTEX_QUOTA);
67eeb568
DM		 4681 bh = ext4_bread(handle, inode, blk, 1, &err);
4682 if (!bh)
4683 goto out;
62d2b5f2
JK		 4684 err = ext4_journal_get_write_access(handle, bh);
4685 if (err) {
4686 brelse(bh);
4687 goto out;
ac27a0ec 4688 }
67eeb568
DM		 4689 lock_buffer(bh);
4690 memcpy(bh->b_data+offset, data, len);
4691 flush_dcache_page(bh->b_page);
4692 unlock_buffer(bh);
62d2b5f2 4693 err = ext4_handle_dirty_metadata(handle, NULL, bh);
67eeb568 4694 brelse(bh);
ac27a0ec 4695out:
67eeb568 4696 if (err) {
4d04e4fb 4697 mutex_unlock(&inode->i_mutex);
ac27a0ec 4698 return err;
4d04e4fb 4699 }
67eeb568
DM		 4700 if (inode->i_size < off + len) {
4701 i_size_write(inode, off + len);
617ba13b 4702 EXT4_I(inode)->i_disksize = inode->i_size;
ac27a0ec 4703 }
ac27a0ec 4704 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
617ba13b 4705 ext4_mark_inode_dirty(handle, inode);
ac27a0ec 4706 mutex_unlock(&inode->i_mutex);
67eeb568 4707 return len;
ac27a0ec
DK		 4708}
4709
4710#endif
4711
152a0836
AV		 4712static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
4713 const char *dev_name, void *data)
ac27a0ec 4714{
152a0836 4715 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
ac27a0ec
DK		 4716}
4717
37f328eb 4718#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
24b58424
TT		 4719static struct file_system_type ext2_fs_type = {
4720 .owner = THIS_MODULE,
4721 .name = "ext2",
152a0836 4722 .mount = ext4_mount,
24b58424
TT		 4723 .kill_sb = kill_block_super,
4724 .fs_flags = FS_REQUIRES_DEV,
4725};
4726
4727static inline void register_as_ext2(void)
4728{
4729 int err = register_filesystem(&ext2_fs_type);
4730 if (err)
4731 printk(KERN_WARNING
4732 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
4733}
4734
4735static inline void unregister_as_ext2(void)
4736{
4737 unregister_filesystem(&ext2_fs_type);
4738}
51b7e3c9 4739MODULE_ALIAS("ext2");
24b58424
TT		 4740#else
4741static inline void register_as_ext2(void) { }
4742static inline void unregister_as_ext2(void) { }
4743#endif
4744
37f328eb 4745#if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
24b58424
TT		 4746static inline void register_as_ext3(void)
4747{
4748 int err = register_filesystem(&ext3_fs_type);
4749 if (err)
4750 printk(KERN_WARNING
4751 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
4752}
4753
4754static inline void unregister_as_ext3(void)
4755{
4756 unregister_filesystem(&ext3_fs_type);
4757}
51b7e3c9 4758MODULE_ALIAS("ext3");
24b58424
TT		 4759#else
4760static inline void register_as_ext3(void) { }
4761static inline void unregister_as_ext3(void) { }
4762#endif
4763
03010a33
TT		 4764static struct file_system_type ext4_fs_type = {
4765 .owner = THIS_MODULE,
4766 .name = "ext4",
152a0836 4767 .mount = ext4_mount,
03010a33
TT		 4768 .kill_sb = kill_block_super,
4769 .fs_flags = FS_REQUIRES_DEV,
4770};
4771
8f021222 4772static int __init ext4_init_feat_adverts(void)
857ac889
LC		 4773{
4774 struct ext4_features *ef;
4775 int ret = -ENOMEM;
4776
4777 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
4778 if (!ef)
4779 goto out;
4780
4781 ef->f_kobj.kset = ext4_kset;
4782 init_completion(&ef->f_kobj_unregister);
4783 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
4784 "features");
4785 if (ret) {
4786 kfree(ef);
4787 goto out;
4788 }
4789
4790 ext4_feat = ef;
4791 ret = 0;
4792out:
4793 return ret;
4794}
4795
8f021222
LC		 4796static void ext4_exit_feat_adverts(void)
4797{
4798 kobject_put(&ext4_feat->f_kobj);
4799 wait_for_completion(&ext4_feat->f_kobj_unregister);
4800 kfree(ext4_feat);
4801}
4802
5dabfc78 4803static int __init ext4_init_fs(void)
ac27a0ec 4804{
c9de560d
AT		 4805 int err;
4806
12e9b892 4807 ext4_check_flag_values();
5dabfc78 4808 err = ext4_init_pageio();
6fd058f7
TT		 4809 if (err)
4810 return err;
5dabfc78 4811 err = ext4_init_system_zone();
bd2d0210
TT		 4812 if (err)
4813 goto out5;
3197ebdb
TT		 4814 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
4815 if (!ext4_kset)
6fd058f7 4816 goto out4;
9f6200bb 4817 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
857ac889
LC		 4818
4819 err = ext4_init_feat_adverts();
4820
5dabfc78 4821 err = ext4_init_mballoc();
ac27a0ec 4822 if (err)
6fd058f7 4823 goto out3;
c9de560d 4824
5dabfc78 4825 err = ext4_init_xattr();
c9de560d
AT		 4826 if (err)
4827 goto out2;
ac27a0ec
DK		 4828 err = init_inodecache();
4829 if (err)
4830 goto out1;
24b58424
TT		 4831 register_as_ext2();
4832 register_as_ext3();
03010a33 4833 err = register_filesystem(&ext4_fs_type);
ac27a0ec
DK		 4834 if (err)
4835 goto out;
bfff6873
LC		 4836
4837 ext4_li_info = NULL;
4838 mutex_init(&ext4_li_mtx);
ac27a0ec
DK		 4839 return 0;
4840out:
24b58424
TT		 4841 unregister_as_ext2();
4842 unregister_as_ext3();
ac27a0ec
DK		 4843 destroy_inodecache();
4844out1:
5dabfc78 4845 ext4_exit_xattr();
c9de560d 4846out2:
5dabfc78 4847 ext4_exit_mballoc();
6fd058f7 4848out3:
8f021222 4849 ext4_exit_feat_adverts();
6fd058f7
TT		 4850 remove_proc_entry("fs/ext4", NULL);
4851 kset_unregister(ext4_kset);
4852out4:
5dabfc78 4853 ext4_exit_system_zone();
bd2d0210 4854out5:
5dabfc78 4855 ext4_exit_pageio();
ac27a0ec
DK		 4856 return err;
4857}
4858
5dabfc78 4859static void __exit ext4_exit_fs(void)
ac27a0ec 4860{
bfff6873 4861 ext4_destroy_lazyinit_thread();
24b58424
TT		 4862 unregister_as_ext2();
4863 unregister_as_ext3();
03010a33 4864 unregister_filesystem(&ext4_fs_type);
ac27a0ec 4865 destroy_inodecache();
5dabfc78
TT		 4866 ext4_exit_xattr();
4867 ext4_exit_mballoc();
8f021222 4868 ext4_exit_feat_adverts();
9f6200bb 4869 remove_proc_entry("fs/ext4", NULL);
3197ebdb 4870 kset_unregister(ext4_kset);
5dabfc78
TT		 4871 ext4_exit_system_zone();
4872 ext4_exit_pageio();
ac27a0ec
DK		 4873}
4874
4875MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
83982b6f 4876MODULE_DESCRIPTION("Fourth Extended Filesystem");
ac27a0ec 4877MODULE_LICENSE("GPL");
5dabfc78
TT		 4878module_init(ext4_init_fs)
4879module_exit(ext4_exit_fs)