projects / linux-2.6-block.git / blame_incremental
? search:
summary | shortlog | log | commit | commitdiff | tree
blob | blame (non-incremental) | history | HEAD
libnvdimm, nfit: move the check on nd_reserved2 to the endpoint
[linux-2.6-block.git] / fs / ext4 / super.c
... / ...
CommitLineData
1/*
2 * linux/fs/ext4/super.c
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>
23#include <linux/vmalloc.h>
24#include <linux/slab.h>
25#include <linux/init.h>
26#include <linux/blkdev.h>
27#include <linux/backing-dev.h>
28#include <linux/parser.h>
29#include <linux/buffer_head.h>
30#include <linux/exportfs.h>
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>
37#include <linux/ctype.h>
38#include <linux/log2.h>
39#include <linux/crc16.h>
40#include <linux/dax.h>
41#include <linux/cleancache.h>
42#include <linux/uaccess.h>
43
44#include <linux/kthread.h>
45#include <linux/freezer.h>
46
47#include "ext4.h"
48#include "ext4_extents.h" /* Needed for trace points definition */
49#include "ext4_jbd2.h"
50#include "xattr.h"
51#include "acl.h"
52#include "mballoc.h"
53#include "fsmap.h"
54
55#define CREATE_TRACE_POINTS
56#include <trace/events/ext4.h>
57
58static struct ext4_lazy_init *ext4_li_info;
59static struct mutex ext4_li_mtx;
60static struct ratelimit_state ext4_mount_msg_ratelimit;
61
62static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
63 unsigned long journal_devnum);
64static int ext4_show_options(struct seq_file *seq, struct dentry *root);
65static int ext4_commit_super(struct super_block *sb, int sync);
66static void ext4_mark_recovery_complete(struct super_block *sb,
67 struct ext4_super_block *es);
68static void ext4_clear_journal_err(struct super_block *sb,
69 struct ext4_super_block *es);
70static int ext4_sync_fs(struct super_block *sb, int wait);
71static int ext4_remount(struct super_block *sb, int *flags, char *data);
72static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
73static int ext4_unfreeze(struct super_block *sb);
74static int ext4_freeze(struct super_block *sb);
75static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
76 const char *dev_name, void *data);
77static inline int ext2_feature_set_ok(struct super_block *sb);
78static inline int ext3_feature_set_ok(struct super_block *sb);
79static int ext4_feature_set_ok(struct super_block *sb, int readonly);
80static void ext4_destroy_lazyinit_thread(void);
81static void ext4_unregister_li_request(struct super_block *sb);
82static void ext4_clear_request_list(void);
83static struct inode *ext4_get_journal_inode(struct super_block *sb,
84 unsigned int journal_inum);
85
86/*
87 * Lock ordering
88 *
89 * Note the difference between i_mmap_sem (EXT4_I(inode)->i_mmap_sem) and
90 * i_mmap_rwsem (inode->i_mmap_rwsem)!
91 *
92 * page fault path:
93 * mmap_sem -> sb_start_pagefault -> i_mmap_sem (r) -> transaction start ->
94 * page lock -> i_data_sem (rw)
95 *
96 * buffered write path:
97 * sb_start_write -> i_mutex -> mmap_sem
98 * sb_start_write -> i_mutex -> transaction start -> page lock ->
99 * i_data_sem (rw)
100 *
101 * truncate:
102 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (w) -> i_mmap_sem (w) ->
103 * i_mmap_rwsem (w) -> page lock
104 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (w) -> i_mmap_sem (w) ->
105 * transaction start -> i_data_sem (rw)
106 *
107 * direct IO:
108 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (r) -> mmap_sem
109 * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (r) ->
110 * transaction start -> i_data_sem (rw)
111 *
112 * writepages:
113 * transaction start -> page lock(s) -> i_data_sem (rw)
114 */
115
116#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
117static struct file_system_type ext2_fs_type = {
118 .owner = THIS_MODULE,
119 .name = "ext2",
120 .mount = ext4_mount,
121 .kill_sb = kill_block_super,
122 .fs_flags = FS_REQUIRES_DEV,
123};
124MODULE_ALIAS_FS("ext2");
125MODULE_ALIAS("ext2");
126#define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
127#else
128#define IS_EXT2_SB(sb) (0)
129#endif
130
131
132static struct file_system_type ext3_fs_type = {
133 .owner = THIS_MODULE,
134 .name = "ext3",
135 .mount = ext4_mount,
136 .kill_sb = kill_block_super,
137 .fs_flags = FS_REQUIRES_DEV,
138};
139MODULE_ALIAS_FS("ext3");
140MODULE_ALIAS("ext3");
141#define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
142
143static int ext4_verify_csum_type(struct super_block *sb,
144 struct ext4_super_block *es)
145{
146 if (!ext4_has_feature_metadata_csum(sb))
147 return 1;
148
149 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
150}
151
152static __le32 ext4_superblock_csum(struct super_block *sb,
153 struct ext4_super_block *es)
154{
155 struct ext4_sb_info *sbi = EXT4_SB(sb);
156 int offset = offsetof(struct ext4_super_block, s_checksum);
157 __u32 csum;
158
159 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
160
161 return cpu_to_le32(csum);
162}
163
164static int ext4_superblock_csum_verify(struct super_block *sb,
165 struct ext4_super_block *es)
166{
167 if (!ext4_has_metadata_csum(sb))
168 return 1;
169
170 return es->s_checksum == ext4_superblock_csum(sb, es);
171}
172
173void ext4_superblock_csum_set(struct super_block *sb)
174{
175 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
176
177 if (!ext4_has_metadata_csum(sb))
178 return;
179
180 es->s_checksum = ext4_superblock_csum(sb, es);
181}
182
183void *ext4_kvmalloc(size_t size, gfp_t flags)
184{
185 void *ret;
186
187 ret = kmalloc(size, flags | __GFP_NOWARN);
188 if (!ret)
189 ret = __vmalloc(size, flags, PAGE_KERNEL);
190 return ret;
191}
192
193void *ext4_kvzalloc(size_t size, gfp_t flags)
194{
195 void *ret;
196
197 ret = kzalloc(size, flags | __GFP_NOWARN);
198 if (!ret)
199 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
200 return ret;
201}
202
203ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
204 struct ext4_group_desc *bg)
205{
206 return le32_to_cpu(bg->bg_block_bitmap_lo) |
207 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
208 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
209}
210
211ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
212 struct ext4_group_desc *bg)
213{
214 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
215 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
216 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
217}
218
219ext4_fsblk_t ext4_inode_table(struct super_block *sb,
220 struct ext4_group_desc *bg)
221{
222 return le32_to_cpu(bg->bg_inode_table_lo) |
223 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
224 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
225}
226
227__u32 ext4_free_group_clusters(struct super_block *sb,
228 struct ext4_group_desc *bg)
229{
230 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
231 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
232 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
233}
234
235__u32 ext4_free_inodes_count(struct super_block *sb,
236 struct ext4_group_desc *bg)
237{
238 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
239 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
240 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
241}
242
243__u32 ext4_used_dirs_count(struct super_block *sb,
244 struct ext4_group_desc *bg)
245{
246 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
247 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
248 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
249}
250
251__u32 ext4_itable_unused_count(struct super_block *sb,
252 struct ext4_group_desc *bg)
253{
254 return le16_to_cpu(bg->bg_itable_unused_lo) |
255 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
256 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
257}
258
259void ext4_block_bitmap_set(struct super_block *sb,
260 struct ext4_group_desc *bg, ext4_fsblk_t blk)
261{
262 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
263 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
264 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
265}
266
267void ext4_inode_bitmap_set(struct super_block *sb,
268 struct ext4_group_desc *bg, ext4_fsblk_t blk)
269{
270 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
271 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
272 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
273}
274
275void ext4_inode_table_set(struct super_block *sb,
276 struct ext4_group_desc *bg, ext4_fsblk_t blk)
277{
278 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
279 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
280 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
281}
282
283void ext4_free_group_clusters_set(struct super_block *sb,
284 struct ext4_group_desc *bg, __u32 count)
285{
286 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
287 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
288 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
289}
290
291void ext4_free_inodes_set(struct super_block *sb,
292 struct ext4_group_desc *bg, __u32 count)
293{
294 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
295 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
296 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
297}
298
299void ext4_used_dirs_set(struct super_block *sb,
300 struct ext4_group_desc *bg, __u32 count)
301{
302 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
303 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
304 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
305}
306
307void ext4_itable_unused_set(struct super_block *sb,
308 struct ext4_group_desc *bg, __u32 count)
309{
310 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
311 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
312 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
313}
314
315
316static void __save_error_info(struct super_block *sb, const char *func,
317 unsigned int line)
318{
319 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
320
321 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
322 if (bdev_read_only(sb->s_bdev))
323 return;
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 }
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);
342 le32_add_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/*
353 * The del_gendisk() function uninitializes the disk-specific data
354 * structures, including the bdi structure, without telling anyone
355 * else. Once this happens, any attempt to call mark_buffer_dirty()
356 * (for example, by ext4_commit_super), will cause a kernel OOPS.
357 * This is a kludge to prevent these oops until we can put in a proper
358 * hook in del_gendisk() to inform the VFS and file system layers.
359 */
360static int block_device_ejected(struct super_block *sb)
361{
362 struct inode *bd_inode = sb->s_bdev->bd_inode;
363 struct backing_dev_info *bdi = inode_to_bdi(bd_inode);
364
365 return bdi->dev == NULL;
366}
367
368static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
369{
370 struct super_block *sb = journal->j_private;
371 struct ext4_sb_info *sbi = EXT4_SB(sb);
372 int error = is_journal_aborted(journal);
373 struct ext4_journal_cb_entry *jce;
374
375 BUG_ON(txn->t_state == T_FINISHED);
376
377 ext4_process_freed_data(sb, txn->t_tid);
378
379 spin_lock(&sbi->s_md_lock);
380 while (!list_empty(&txn->t_private_list)) {
381 jce = list_entry(txn->t_private_list.next,
382 struct ext4_journal_cb_entry, jce_list);
383 list_del_init(&jce->jce_list);
384 spin_unlock(&sbi->s_md_lock);
385 jce->jce_func(sb, jce, error);
386 spin_lock(&sbi->s_md_lock);
387 }
388 spin_unlock(&sbi->s_md_lock);
389}
390
391/* Deal with the reporting of failure conditions on a filesystem such as
392 * inconsistencies detected or read IO failures.
393 *
394 * On ext2, we can store the error state of the filesystem in the
395 * superblock. That is not possible on ext4, because we may have other
396 * write ordering constraints on the superblock which prevent us from
397 * writing it out straight away; and given that the journal is about to
398 * be aborted, we can't rely on the current, or future, transactions to
399 * write out the superblock safely.
400 *
401 * We'll just use the jbd2_journal_abort() error code to record an error in
402 * the journal instead. On recovery, the journal will complain about
403 * that error until we've noted it down and cleared it.
404 */
405
406static void ext4_handle_error(struct super_block *sb)
407{
408 if (sb->s_flags & MS_RDONLY)
409 return;
410
411 if (!test_opt(sb, ERRORS_CONT)) {
412 journal_t *journal = EXT4_SB(sb)->s_journal;
413
414 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
415 if (journal)
416 jbd2_journal_abort(journal, -EIO);
417 }
418 if (test_opt(sb, ERRORS_RO)) {
419 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
420 /*
421 * Make sure updated value of ->s_mount_flags will be visible
422 * before ->s_flags update
423 */
424 smp_wmb();
425 sb->s_flags |= MS_RDONLY;
426 }
427 if (test_opt(sb, ERRORS_PANIC)) {
428 if (EXT4_SB(sb)->s_journal &&
429 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
430 return;
431 panic("EXT4-fs (device %s): panic forced after error\n",
432 sb->s_id);
433 }
434}
435
436#define ext4_error_ratelimit(sb) \
437 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
438 "EXT4-fs error")
439
440void __ext4_error(struct super_block *sb, const char *function,
441 unsigned int line, const char *fmt, ...)
442{
443 struct va_format vaf;
444 va_list args;
445
446 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
447 return;
448
449 if (ext4_error_ratelimit(sb)) {
450 va_start(args, fmt);
451 vaf.fmt = fmt;
452 vaf.va = &args;
453 printk(KERN_CRIT
454 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
455 sb->s_id, function, line, current->comm, &vaf);
456 va_end(args);
457 }
458 save_error_info(sb, function, line);
459 ext4_handle_error(sb);
460}
461
462void __ext4_error_inode(struct inode *inode, const char *function,
463 unsigned int line, ext4_fsblk_t block,
464 const char *fmt, ...)
465{
466 va_list args;
467 struct va_format vaf;
468 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
469
470 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
471 return;
472
473 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
474 es->s_last_error_block = cpu_to_le64(block);
475 if (ext4_error_ratelimit(inode->i_sb)) {
476 va_start(args, fmt);
477 vaf.fmt = fmt;
478 vaf.va = &args;
479 if (block)
480 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
481 "inode #%lu: block %llu: comm %s: %pV\n",
482 inode->i_sb->s_id, function, line, inode->i_ino,
483 block, current->comm, &vaf);
484 else
485 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
486 "inode #%lu: comm %s: %pV\n",
487 inode->i_sb->s_id, function, line, inode->i_ino,
488 current->comm, &vaf);
489 va_end(args);
490 }
491 save_error_info(inode->i_sb, function, line);
492 ext4_handle_error(inode->i_sb);
493}
494
495void __ext4_error_file(struct file *file, const char *function,
496 unsigned int line, ext4_fsblk_t block,
497 const char *fmt, ...)
498{
499 va_list args;
500 struct va_format vaf;
501 struct ext4_super_block *es;
502 struct inode *inode = file_inode(file);
503 char pathname[80], *path;
504
505 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
506 return;
507
508 es = EXT4_SB(inode->i_sb)->s_es;
509 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
510 if (ext4_error_ratelimit(inode->i_sb)) {
511 path = file_path(file, pathname, sizeof(pathname));
512 if (IS_ERR(path))
513 path = "(unknown)";
514 va_start(args, fmt);
515 vaf.fmt = fmt;
516 vaf.va = &args;
517 if (block)
518 printk(KERN_CRIT
519 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
520 "block %llu: comm %s: path %s: %pV\n",
521 inode->i_sb->s_id, function, line, inode->i_ino,
522 block, current->comm, path, &vaf);
523 else
524 printk(KERN_CRIT
525 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
526 "comm %s: path %s: %pV\n",
527 inode->i_sb->s_id, function, line, inode->i_ino,
528 current->comm, path, &vaf);
529 va_end(args);
530 }
531 save_error_info(inode->i_sb, function, line);
532 ext4_handle_error(inode->i_sb);
533}
534
535const char *ext4_decode_error(struct super_block *sb, int errno,
536 char nbuf[16])
537{
538 char *errstr = NULL;
539
540 switch (errno) {
541 case -EFSCORRUPTED:
542 errstr = "Corrupt filesystem";
543 break;
544 case -EFSBADCRC:
545 errstr = "Filesystem failed CRC";
546 break;
547 case -EIO:
548 errstr = "IO failure";
549 break;
550 case -ENOMEM:
551 errstr = "Out of memory";
552 break;
553 case -EROFS:
554 if (!sb || (EXT4_SB(sb)->s_journal &&
555 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
556 errstr = "Journal has aborted";
557 else
558 errstr = "Readonly filesystem";
559 break;
560 default:
561 /* If the caller passed in an extra buffer for unknown
562 * errors, textualise them now. Else we just return
563 * NULL. */
564 if (nbuf) {
565 /* Check for truncated error codes... */
566 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
567 errstr = nbuf;
568 }
569 break;
570 }
571
572 return errstr;
573}
574
575/* __ext4_std_error decodes expected errors from journaling functions
576 * automatically and invokes the appropriate error response. */
577
578void __ext4_std_error(struct super_block *sb, const char *function,
579 unsigned int line, int errno)
580{
581 char nbuf[16];
582 const char *errstr;
583
584 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
585 return;
586
587 /* Special case: if the error is EROFS, and we're not already
588 * inside a transaction, then there's really no point in logging
589 * an error. */
590 if (errno == -EROFS && journal_current_handle() == NULL &&
591 (sb->s_flags & MS_RDONLY))
592 return;
593
594 if (ext4_error_ratelimit(sb)) {
595 errstr = ext4_decode_error(sb, errno, nbuf);
596 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
597 sb->s_id, function, line, errstr);
598 }
599
600 save_error_info(sb, function, line);
601 ext4_handle_error(sb);
602}
603
604/*
605 * ext4_abort is a much stronger failure handler than ext4_error. The
606 * abort function may be used to deal with unrecoverable failures such
607 * as journal IO errors or ENOMEM at a critical moment in log management.
608 *
609 * We unconditionally force the filesystem into an ABORT|READONLY state,
610 * unless the error response on the fs has been set to panic in which
611 * case we take the easy way out and panic immediately.
612 */
613
614void __ext4_abort(struct super_block *sb, const char *function,
615 unsigned int line, const char *fmt, ...)
616{
617 struct va_format vaf;
618 va_list args;
619
620 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
621 return;
622
623 save_error_info(sb, function, line);
624 va_start(args, fmt);
625 vaf.fmt = fmt;
626 vaf.va = &args;
627 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: %pV\n",
628 sb->s_id, function, line, &vaf);
629 va_end(args);
630
631 if ((sb->s_flags & MS_RDONLY) == 0) {
632 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
633 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
634 /*
635 * Make sure updated value of ->s_mount_flags will be visible
636 * before ->s_flags update
637 */
638 smp_wmb();
639 sb->s_flags |= MS_RDONLY;
640 if (EXT4_SB(sb)->s_journal)
641 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
642 save_error_info(sb, function, line);
643 }
644 if (test_opt(sb, ERRORS_PANIC)) {
645 if (EXT4_SB(sb)->s_journal &&
646 !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
647 return;
648 panic("EXT4-fs panic from previous error\n");
649 }
650}
651
652void __ext4_msg(struct super_block *sb,
653 const char *prefix, const char *fmt, ...)
654{
655 struct va_format vaf;
656 va_list args;
657
658 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
659 return;
660
661 va_start(args, fmt);
662 vaf.fmt = fmt;
663 vaf.va = &args;
664 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
665 va_end(args);
666}
667
668#define ext4_warning_ratelimit(sb) \
669 ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state), \
670 "EXT4-fs warning")
671
672void __ext4_warning(struct super_block *sb, const char *function,
673 unsigned int line, const char *fmt, ...)
674{
675 struct va_format vaf;
676 va_list args;
677
678 if (!ext4_warning_ratelimit(sb))
679 return;
680
681 va_start(args, fmt);
682 vaf.fmt = fmt;
683 vaf.va = &args;
684 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
685 sb->s_id, function, line, &vaf);
686 va_end(args);
687}
688
689void __ext4_warning_inode(const struct inode *inode, const char *function,
690 unsigned int line, const char *fmt, ...)
691{
692 struct va_format vaf;
693 va_list args;
694
695 if (!ext4_warning_ratelimit(inode->i_sb))
696 return;
697
698 va_start(args, fmt);
699 vaf.fmt = fmt;
700 vaf.va = &args;
701 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
702 "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
703 function, line, inode->i_ino, current->comm, &vaf);
704 va_end(args);
705}
706
707void __ext4_grp_locked_error(const char *function, unsigned int line,
708 struct super_block *sb, ext4_group_t grp,
709 unsigned long ino, ext4_fsblk_t block,
710 const char *fmt, ...)
711__releases(bitlock)
712__acquires(bitlock)
713{
714 struct va_format vaf;
715 va_list args;
716 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
717
718 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
719 return;
720
721 es->s_last_error_ino = cpu_to_le32(ino);
722 es->s_last_error_block = cpu_to_le64(block);
723 __save_error_info(sb, function, line);
724
725 if (ext4_error_ratelimit(sb)) {
726 va_start(args, fmt);
727 vaf.fmt = fmt;
728 vaf.va = &args;
729 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
730 sb->s_id, function, line, grp);
731 if (ino)
732 printk(KERN_CONT "inode %lu: ", ino);
733 if (block)
734 printk(KERN_CONT "block %llu:",
735 (unsigned long long) block);
736 printk(KERN_CONT "%pV\n", &vaf);
737 va_end(args);
738 }
739
740 if (test_opt(sb, ERRORS_CONT)) {
741 ext4_commit_super(sb, 0);
742 return;
743 }
744
745 ext4_unlock_group(sb, grp);
746 ext4_handle_error(sb);
747 /*
748 * We only get here in the ERRORS_RO case; relocking the group
749 * may be dangerous, but nothing bad will happen since the
750 * filesystem will have already been marked read/only and the
751 * journal has been aborted. We return 1 as a hint to callers
752 * who might what to use the return value from
753 * ext4_grp_locked_error() to distinguish between the
754 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
755 * aggressively from the ext4 function in question, with a
756 * more appropriate error code.
757 */
758 ext4_lock_group(sb, grp);
759 return;
760}
761
762void ext4_update_dynamic_rev(struct super_block *sb)
763{
764 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
765
766 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
767 return;
768
769 ext4_warning(sb,
770 "updating to rev %d because of new feature flag, "
771 "running e2fsck is recommended",
772 EXT4_DYNAMIC_REV);
773
774 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
775 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
776 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
777 /* leave es->s_feature_*compat flags alone */
778 /* es->s_uuid will be set by e2fsck if empty */
779
780 /*
781 * The rest of the superblock fields should be zero, and if not it
782 * means they are likely already in use, so leave them alone. We
783 * can leave it up to e2fsck to clean up any inconsistencies there.
784 */
785}
786
787/*
788 * Open the external journal device
789 */
790static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
791{
792 struct block_device *bdev;
793 char b[BDEVNAME_SIZE];
794
795 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
796 if (IS_ERR(bdev))
797 goto fail;
798 return bdev;
799
800fail:
801 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
802 __bdevname(dev, b), PTR_ERR(bdev));
803 return NULL;
804}
805
806/*
807 * Release the journal device
808 */
809static void ext4_blkdev_put(struct block_device *bdev)
810{
811 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
812}
813
814static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
815{
816 struct block_device *bdev;
817 bdev = sbi->journal_bdev;
818 if (bdev) {
819 ext4_blkdev_put(bdev);
820 sbi->journal_bdev = NULL;
821 }
822}
823
824static inline struct inode *orphan_list_entry(struct list_head *l)
825{
826 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
827}
828
829static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
830{
831 struct list_head *l;
832
833 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
834 le32_to_cpu(sbi->s_es->s_last_orphan));
835
836 printk(KERN_ERR "sb_info orphan list:\n");
837 list_for_each(l, &sbi->s_orphan) {
838 struct inode *inode = orphan_list_entry(l);
839 printk(KERN_ERR " "
840 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
841 inode->i_sb->s_id, inode->i_ino, inode,
842 inode->i_mode, inode->i_nlink,
843 NEXT_ORPHAN(inode));
844 }
845}
846
847#ifdef CONFIG_QUOTA
848static int ext4_quota_off(struct super_block *sb, int type);
849
850static inline void ext4_quota_off_umount(struct super_block *sb)
851{
852 int type;
853
854 /* Use our quota_off function to clear inode flags etc. */
855 for (type = 0; type < EXT4_MAXQUOTAS; type++)
856 ext4_quota_off(sb, type);
857}
858#else
859static inline void ext4_quota_off_umount(struct super_block *sb)
860{
861}
862#endif
863
864static void ext4_put_super(struct super_block *sb)
865{
866 struct ext4_sb_info *sbi = EXT4_SB(sb);
867 struct ext4_super_block *es = sbi->s_es;
868 int aborted = 0;
869 int i, err;
870
871 ext4_unregister_li_request(sb);
872 ext4_quota_off_umount(sb);
873
874 flush_workqueue(sbi->rsv_conversion_wq);
875 destroy_workqueue(sbi->rsv_conversion_wq);
876
877 if (sbi->s_journal) {
878 aborted = is_journal_aborted(sbi->s_journal);
879 err = jbd2_journal_destroy(sbi->s_journal);
880 sbi->s_journal = NULL;
881 if ((err < 0) && !aborted)
882 ext4_abort(sb, "Couldn't clean up the journal");
883 }
884
885 ext4_unregister_sysfs(sb);
886 ext4_es_unregister_shrinker(sbi);
887 del_timer_sync(&sbi->s_err_report);
888 ext4_release_system_zone(sb);
889 ext4_mb_release(sb);
890 ext4_ext_release(sb);
891
892 if (!(sb->s_flags & MS_RDONLY) && !aborted) {
893 ext4_clear_feature_journal_needs_recovery(sb);
894 es->s_state = cpu_to_le16(sbi->s_mount_state);
895 }
896 if (!(sb->s_flags & MS_RDONLY))
897 ext4_commit_super(sb, 1);
898
899 for (i = 0; i < sbi->s_gdb_count; i++)
900 brelse(sbi->s_group_desc[i]);
901 kvfree(sbi->s_group_desc);
902 kvfree(sbi->s_flex_groups);
903 percpu_counter_destroy(&sbi->s_freeclusters_counter);
904 percpu_counter_destroy(&sbi->s_freeinodes_counter);
905 percpu_counter_destroy(&sbi->s_dirs_counter);
906 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
907 percpu_free_rwsem(&sbi->s_journal_flag_rwsem);
908#ifdef CONFIG_QUOTA
909 for (i = 0; i < EXT4_MAXQUOTAS; i++)
910 kfree(sbi->s_qf_names[i]);
911#endif
912
913 /* Debugging code just in case the in-memory inode orphan list
914 * isn't empty. The on-disk one can be non-empty if we've
915 * detected an error and taken the fs readonly, but the
916 * in-memory list had better be clean by this point. */
917 if (!list_empty(&sbi->s_orphan))
918 dump_orphan_list(sb, sbi);
919 J_ASSERT(list_empty(&sbi->s_orphan));
920
921 sync_blockdev(sb->s_bdev);
922 invalidate_bdev(sb->s_bdev);
923 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
924 /*
925 * Invalidate the journal device's buffers. We don't want them
926 * floating about in memory - the physical journal device may
927 * hotswapped, and it breaks the `ro-after' testing code.
928 */
929 sync_blockdev(sbi->journal_bdev);
930 invalidate_bdev(sbi->journal_bdev);
931 ext4_blkdev_remove(sbi);
932 }
933 if (sbi->s_ea_inode_cache) {
934 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
935 sbi->s_ea_inode_cache = NULL;
936 }
937 if (sbi->s_ea_block_cache) {
938 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
939 sbi->s_ea_block_cache = NULL;
940 }
941 if (sbi->s_mmp_tsk)
942 kthread_stop(sbi->s_mmp_tsk);
943 brelse(sbi->s_sbh);
944 sb->s_fs_info = NULL;
945 /*
946 * Now that we are completely done shutting down the
947 * superblock, we need to actually destroy the kobject.
948 */
949 kobject_put(&sbi->s_kobj);
950 wait_for_completion(&sbi->s_kobj_unregister);
951 if (sbi->s_chksum_driver)
952 crypto_free_shash(sbi->s_chksum_driver);
953 kfree(sbi->s_blockgroup_lock);
954 fs_put_dax(sbi->s_daxdev);
955 kfree(sbi);
956}
957
958static struct kmem_cache *ext4_inode_cachep;
959
960/*
961 * Called inside transaction, so use GFP_NOFS
962 */
963static struct inode *ext4_alloc_inode(struct super_block *sb)
964{
965 struct ext4_inode_info *ei;
966
967 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
968 if (!ei)
969 return NULL;
970
971 ei->vfs_inode.i_version = 1;
972 spin_lock_init(&ei->i_raw_lock);
973 INIT_LIST_HEAD(&ei->i_prealloc_list);
974 spin_lock_init(&ei->i_prealloc_lock);
975 ext4_es_init_tree(&ei->i_es_tree);
976 rwlock_init(&ei->i_es_lock);
977 INIT_LIST_HEAD(&ei->i_es_list);
978 ei->i_es_all_nr = 0;
979 ei->i_es_shk_nr = 0;
980 ei->i_es_shrink_lblk = 0;
981 ei->i_reserved_data_blocks = 0;
982 ei->i_da_metadata_calc_len = 0;
983 ei->i_da_metadata_calc_last_lblock = 0;
984 spin_lock_init(&(ei->i_block_reservation_lock));
985#ifdef CONFIG_QUOTA
986 ei->i_reserved_quota = 0;
987 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
988#endif
989 ei->jinode = NULL;
990 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
991 spin_lock_init(&ei->i_completed_io_lock);
992 ei->i_sync_tid = 0;
993 ei->i_datasync_tid = 0;
994 atomic_set(&ei->i_unwritten, 0);
995 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
996 return &ei->vfs_inode;
997}
998
999static int ext4_drop_inode(struct inode *inode)
1000{
1001 int drop = generic_drop_inode(inode);
1002
1003 trace_ext4_drop_inode(inode, drop);
1004 return drop;
1005}
1006
1007static void ext4_i_callback(struct rcu_head *head)
1008{
1009 struct inode *inode = container_of(head, struct inode, i_rcu);
1010 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
1011}
1012
1013static void ext4_destroy_inode(struct inode *inode)
1014{
1015 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
1016 ext4_msg(inode->i_sb, KERN_ERR,
1017 "Inode %lu (%p): orphan list check failed!",
1018 inode->i_ino, EXT4_I(inode));
1019 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
1020 EXT4_I(inode), sizeof(struct ext4_inode_info),
1021 true);
1022 dump_stack();
1023 }
1024 call_rcu(&inode->i_rcu, ext4_i_callback);
1025}
1026
1027static void init_once(void *foo)
1028{
1029 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
1030
1031 INIT_LIST_HEAD(&ei->i_orphan);
1032 init_rwsem(&ei->xattr_sem);
1033 init_rwsem(&ei->i_data_sem);
1034 init_rwsem(&ei->i_mmap_sem);
1035 inode_init_once(&ei->vfs_inode);
1036}
1037
1038static int __init init_inodecache(void)
1039{
1040 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
1041 sizeof(struct ext4_inode_info),
1042 0, (SLAB_RECLAIM_ACCOUNT|
1043 SLAB_MEM_SPREAD|SLAB_ACCOUNT),
1044 init_once);
1045 if (ext4_inode_cachep == NULL)
1046 return -ENOMEM;
1047 return 0;
1048}
1049
1050static void destroy_inodecache(void)
1051{
1052 /*
1053 * Make sure all delayed rcu free inodes are flushed before we
1054 * destroy cache.
1055 */
1056 rcu_barrier();
1057 kmem_cache_destroy(ext4_inode_cachep);
1058}
1059
1060void ext4_clear_inode(struct inode *inode)
1061{
1062 invalidate_inode_buffers(inode);
1063 clear_inode(inode);
1064 dquot_drop(inode);
1065 ext4_discard_preallocations(inode);
1066 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
1067 if (EXT4_I(inode)->jinode) {
1068 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1069 EXT4_I(inode)->jinode);
1070 jbd2_free_inode(EXT4_I(inode)->jinode);
1071 EXT4_I(inode)->jinode = NULL;
1072 }
1073#ifdef CONFIG_EXT4_FS_ENCRYPTION
1074 fscrypt_put_encryption_info(inode, NULL);
1075#endif
1076}
1077
1078static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1079 u64 ino, u32 generation)
1080{
1081 struct inode *inode;
1082
1083 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
1084 return ERR_PTR(-ESTALE);
1085 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
1086 return ERR_PTR(-ESTALE);
1087
1088 /* iget isn't really right if the inode is currently unallocated!!
1089 *
1090 * ext4_read_inode will return a bad_inode if the inode had been
1091 * deleted, so we should be safe.
1092 *
1093 * Currently we don't know the generation for parent directory, so
1094 * a generation of 0 means "accept any"
1095 */
1096 inode = ext4_iget_normal(sb, ino);
1097 if (IS_ERR(inode))
1098 return ERR_CAST(inode);
1099 if (generation && inode->i_generation != generation) {
1100 iput(inode);
1101 return ERR_PTR(-ESTALE);
1102 }
1103
1104 return inode;
1105}
1106
1107static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1108 int fh_len, int fh_type)
1109{
1110 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1111 ext4_nfs_get_inode);
1112}
1113
1114static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1115 int fh_len, int fh_type)
1116{
1117 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1118 ext4_nfs_get_inode);
1119}
1120
1121/*
1122 * Try to release metadata pages (indirect blocks, directories) which are
1123 * mapped via the block device. Since these pages could have journal heads
1124 * which would prevent try_to_free_buffers() from freeing them, we must use
1125 * jbd2 layer's try_to_free_buffers() function to release them.
1126 */
1127static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1128 gfp_t wait)
1129{
1130 journal_t *journal = EXT4_SB(sb)->s_journal;
1131
1132 WARN_ON(PageChecked(page));
1133 if (!page_has_buffers(page))
1134 return 0;
1135 if (journal)
1136 return jbd2_journal_try_to_free_buffers(journal, page,
1137 wait & ~__GFP_DIRECT_RECLAIM);
1138 return try_to_free_buffers(page);
1139}
1140
1141#ifdef CONFIG_EXT4_FS_ENCRYPTION
1142static int ext4_get_context(struct inode *inode, void *ctx, size_t len)
1143{
1144 return ext4_xattr_get(inode, EXT4_XATTR_INDEX_ENCRYPTION,
1145 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx, len);
1146}
1147
1148static int ext4_set_context(struct inode *inode, const void *ctx, size_t len,
1149 void *fs_data)
1150{
1151 handle_t *handle = fs_data;
1152 int res, res2, credits, retries = 0;
1153
1154 /*
1155 * Encrypting the root directory is not allowed because e2fsck expects
1156 * lost+found to exist and be unencrypted, and encrypting the root
1157 * directory would imply encrypting the lost+found directory as well as
1158 * the filename "lost+found" itself.
1159 */
1160 if (inode->i_ino == EXT4_ROOT_INO)
1161 return -EPERM;
1162
1163 res = ext4_convert_inline_data(inode);
1164 if (res)
1165 return res;
1166
1167 /*
1168 * If a journal handle was specified, then the encryption context is
1169 * being set on a new inode via inheritance and is part of a larger
1170 * transaction to create the inode. Otherwise the encryption context is
1171 * being set on an existing inode in its own transaction. Only in the
1172 * latter case should the "retry on ENOSPC" logic be used.
1173 */
1174
1175 if (handle) {
1176 res = ext4_xattr_set_handle(handle, inode,
1177 EXT4_XATTR_INDEX_ENCRYPTION,
1178 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1179 ctx, len, 0);
1180 if (!res) {
1181 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1182 ext4_clear_inode_state(inode,
1183 EXT4_STATE_MAY_INLINE_DATA);
1184 /*
1185 * Update inode->i_flags - e.g. S_DAX may get disabled
1186 */
1187 ext4_set_inode_flags(inode);
1188 }
1189 return res;
1190 }
1191
1192 res = dquot_initialize(inode);
1193 if (res)
1194 return res;
1195retry:
1196 res = ext4_xattr_set_credits(inode, len, false /* is_create */,
1197 &credits);
1198 if (res)
1199 return res;
1200
1201 handle = ext4_journal_start(inode, EXT4_HT_MISC, credits);
1202 if (IS_ERR(handle))
1203 return PTR_ERR(handle);
1204
1205 res = ext4_xattr_set_handle(handle, inode, EXT4_XATTR_INDEX_ENCRYPTION,
1206 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1207 ctx, len, 0);
1208 if (!res) {
1209 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1210 /* Update inode->i_flags - e.g. S_DAX may get disabled */
1211 ext4_set_inode_flags(inode);
1212 res = ext4_mark_inode_dirty(handle, inode);
1213 if (res)
1214 EXT4_ERROR_INODE(inode, "Failed to mark inode dirty");
1215 }
1216 res2 = ext4_journal_stop(handle);
1217
1218 if (res == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries))
1219 goto retry;
1220 if (!res)
1221 res = res2;
1222 return res;
1223}
1224
1225static bool ext4_dummy_context(struct inode *inode)
1226{
1227 return DUMMY_ENCRYPTION_ENABLED(EXT4_SB(inode->i_sb));
1228}
1229
1230static unsigned ext4_max_namelen(struct inode *inode)
1231{
1232 return S_ISLNK(inode->i_mode) ? inode->i_sb->s_blocksize :
1233 EXT4_NAME_LEN;
1234}
1235
1236static const struct fscrypt_operations ext4_cryptops = {
1237 .key_prefix = "ext4:",
1238 .get_context = ext4_get_context,
1239 .set_context = ext4_set_context,
1240 .dummy_context = ext4_dummy_context,
1241 .is_encrypted = ext4_encrypted_inode,
1242 .empty_dir = ext4_empty_dir,
1243 .max_namelen = ext4_max_namelen,
1244};
1245#else
1246static const struct fscrypt_operations ext4_cryptops = {
1247 .is_encrypted = ext4_encrypted_inode,
1248};
1249#endif
1250
1251#ifdef CONFIG_QUOTA
1252static const char * const quotatypes[] = INITQFNAMES;
1253#define QTYPE2NAME(t) (quotatypes[t])
1254
1255static int ext4_write_dquot(struct dquot *dquot);
1256static int ext4_acquire_dquot(struct dquot *dquot);
1257static int ext4_release_dquot(struct dquot *dquot);
1258static int ext4_mark_dquot_dirty(struct dquot *dquot);
1259static int ext4_write_info(struct super_block *sb, int type);
1260static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1261 const struct path *path);
1262static int ext4_quota_on_mount(struct super_block *sb, int type);
1263static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1264 size_t len, loff_t off);
1265static ssize_t ext4_quota_write(struct super_block *sb, int type,
1266 const char *data, size_t len, loff_t off);
1267static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1268 unsigned int flags);
1269static int ext4_enable_quotas(struct super_block *sb);
1270static int ext4_get_next_id(struct super_block *sb, struct kqid *qid);
1271
1272static struct dquot **ext4_get_dquots(struct inode *inode)
1273{
1274 return EXT4_I(inode)->i_dquot;
1275}
1276
1277static const struct dquot_operations ext4_quota_operations = {
1278 .get_reserved_space = ext4_get_reserved_space,
1279 .write_dquot = ext4_write_dquot,
1280 .acquire_dquot = ext4_acquire_dquot,
1281 .release_dquot = ext4_release_dquot,
1282 .mark_dirty = ext4_mark_dquot_dirty,
1283 .write_info = ext4_write_info,
1284 .alloc_dquot = dquot_alloc,
1285 .destroy_dquot = dquot_destroy,
1286 .get_projid = ext4_get_projid,
1287 .get_inode_usage = ext4_get_inode_usage,
1288 .get_next_id = ext4_get_next_id,
1289};
1290
1291static const struct quotactl_ops ext4_qctl_operations = {
1292 .quota_on = ext4_quota_on,
1293 .quota_off = ext4_quota_off,
1294 .quota_sync = dquot_quota_sync,
1295 .get_state = dquot_get_state,
1296 .set_info = dquot_set_dqinfo,
1297 .get_dqblk = dquot_get_dqblk,
1298 .set_dqblk = dquot_set_dqblk,
1299 .get_nextdqblk = dquot_get_next_dqblk,
1300};
1301#endif
1302
1303static const struct super_operations ext4_sops = {
1304 .alloc_inode = ext4_alloc_inode,
1305 .destroy_inode = ext4_destroy_inode,
1306 .write_inode = ext4_write_inode,
1307 .dirty_inode = ext4_dirty_inode,
1308 .drop_inode = ext4_drop_inode,
1309 .evict_inode = ext4_evict_inode,
1310 .put_super = ext4_put_super,
1311 .sync_fs = ext4_sync_fs,
1312 .freeze_fs = ext4_freeze,
1313 .unfreeze_fs = ext4_unfreeze,
1314 .statfs = ext4_statfs,
1315 .remount_fs = ext4_remount,
1316 .show_options = ext4_show_options,
1317#ifdef CONFIG_QUOTA
1318 .quota_read = ext4_quota_read,
1319 .quota_write = ext4_quota_write,
1320 .get_dquots = ext4_get_dquots,
1321#endif
1322 .bdev_try_to_free_page = bdev_try_to_free_page,
1323};
1324
1325static const struct export_operations ext4_export_ops = {
1326 .fh_to_dentry = ext4_fh_to_dentry,
1327 .fh_to_parent = ext4_fh_to_parent,
1328 .get_parent = ext4_get_parent,
1329};
1330
1331enum {
1332 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1333 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1334 Opt_nouid32, Opt_debug, Opt_removed,
1335 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1336 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1337 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1338 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1339 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1340 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
1341 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1342 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1343 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1344 Opt_usrquota, Opt_grpquota, Opt_prjquota, Opt_i_version, Opt_dax,
1345 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1346 Opt_lazytime, Opt_nolazytime, Opt_debug_want_extra_isize,
1347 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1348 Opt_inode_readahead_blks, Opt_journal_ioprio,
1349 Opt_dioread_nolock, Opt_dioread_lock,
1350 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1351 Opt_max_dir_size_kb, Opt_nojournal_checksum, Opt_nombcache,
1352};
1353
1354static const match_table_t tokens = {
1355 {Opt_bsd_df, "bsddf"},
1356 {Opt_minix_df, "minixdf"},
1357 {Opt_grpid, "grpid"},
1358 {Opt_grpid, "bsdgroups"},
1359 {Opt_nogrpid, "nogrpid"},
1360 {Opt_nogrpid, "sysvgroups"},
1361 {Opt_resgid, "resgid=%u"},
1362 {Opt_resuid, "resuid=%u"},
1363 {Opt_sb, "sb=%u"},
1364 {Opt_err_cont, "errors=continue"},
1365 {Opt_err_panic, "errors=panic"},
1366 {Opt_err_ro, "errors=remount-ro"},
1367 {Opt_nouid32, "nouid32"},
1368 {Opt_debug, "debug"},
1369 {Opt_removed, "oldalloc"},
1370 {Opt_removed, "orlov"},
1371 {Opt_user_xattr, "user_xattr"},
1372 {Opt_nouser_xattr, "nouser_xattr"},
1373 {Opt_acl, "acl"},
1374 {Opt_noacl, "noacl"},
1375 {Opt_noload, "norecovery"},
1376 {Opt_noload, "noload"},
1377 {Opt_removed, "nobh"},
1378 {Opt_removed, "bh"},
1379 {Opt_commit, "commit=%u"},
1380 {Opt_min_batch_time, "min_batch_time=%u"},
1381 {Opt_max_batch_time, "max_batch_time=%u"},
1382 {Opt_journal_dev, "journal_dev=%u"},
1383 {Opt_journal_path, "journal_path=%s"},
1384 {Opt_journal_checksum, "journal_checksum"},
1385 {Opt_nojournal_checksum, "nojournal_checksum"},
1386 {Opt_journal_async_commit, "journal_async_commit"},
1387 {Opt_abort, "abort"},
1388 {Opt_data_journal, "data=journal"},
1389 {Opt_data_ordered, "data=ordered"},
1390 {Opt_data_writeback, "data=writeback"},
1391 {Opt_data_err_abort, "data_err=abort"},
1392 {Opt_data_err_ignore, "data_err=ignore"},
1393 {Opt_offusrjquota, "usrjquota="},
1394 {Opt_usrjquota, "usrjquota=%s"},
1395 {Opt_offgrpjquota, "grpjquota="},
1396 {Opt_grpjquota, "grpjquota=%s"},
1397 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1398 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1399 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1400 {Opt_grpquota, "grpquota"},
1401 {Opt_noquota, "noquota"},
1402 {Opt_quota, "quota"},
1403 {Opt_usrquota, "usrquota"},
1404 {Opt_prjquota, "prjquota"},
1405 {Opt_barrier, "barrier=%u"},
1406 {Opt_barrier, "barrier"},
1407 {Opt_nobarrier, "nobarrier"},
1408 {Opt_i_version, "i_version"},
1409 {Opt_dax, "dax"},
1410 {Opt_stripe, "stripe=%u"},
1411 {Opt_delalloc, "delalloc"},
1412 {Opt_lazytime, "lazytime"},
1413 {Opt_nolazytime, "nolazytime"},
1414 {Opt_debug_want_extra_isize, "debug_want_extra_isize=%u"},
1415 {Opt_nodelalloc, "nodelalloc"},
1416 {Opt_removed, "mblk_io_submit"},
1417 {Opt_removed, "nomblk_io_submit"},
1418 {Opt_block_validity, "block_validity"},
1419 {Opt_noblock_validity, "noblock_validity"},
1420 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1421 {Opt_journal_ioprio, "journal_ioprio=%u"},
1422 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1423 {Opt_auto_da_alloc, "auto_da_alloc"},
1424 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1425 {Opt_dioread_nolock, "dioread_nolock"},
1426 {Opt_dioread_lock, "dioread_lock"},
1427 {Opt_discard, "discard"},
1428 {Opt_nodiscard, "nodiscard"},
1429 {Opt_init_itable, "init_itable=%u"},
1430 {Opt_init_itable, "init_itable"},
1431 {Opt_noinit_itable, "noinit_itable"},
1432 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1433 {Opt_test_dummy_encryption, "test_dummy_encryption"},
1434 {Opt_nombcache, "nombcache"},
1435 {Opt_nombcache, "no_mbcache"}, /* for backward compatibility */
1436 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1437 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1438 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1439 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1440 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
1441 {Opt_err, NULL},
1442};
1443
1444static ext4_fsblk_t get_sb_block(void **data)
1445{
1446 ext4_fsblk_t sb_block;
1447 char *options = (char *) *data;
1448
1449 if (!options || strncmp(options, "sb=", 3) != 0)
1450 return 1; /* Default location */
1451
1452 options += 3;
1453 /* TODO: use simple_strtoll with >32bit ext4 */
1454 sb_block = simple_strtoul(options, &options, 0);
1455 if (*options && *options != ',') {
1456 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1457 (char *) *data);
1458 return 1;
1459 }
1460 if (*options == ',')
1461 options++;
1462 *data = (void *) options;
1463
1464 return sb_block;
1465}
1466
1467#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1468static const char deprecated_msg[] =
1469 "Mount option \"%s\" will be removed by %s\n"
1470 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1471
1472#ifdef CONFIG_QUOTA
1473static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1474{
1475 struct ext4_sb_info *sbi = EXT4_SB(sb);
1476 char *qname;
1477 int ret = -1;
1478
1479 if (sb_any_quota_loaded(sb) &&
1480 !sbi->s_qf_names[qtype]) {
1481 ext4_msg(sb, KERN_ERR,
1482 "Cannot change journaled "
1483 "quota options when quota turned on");
1484 return -1;
1485 }
1486 if (ext4_has_feature_quota(sb)) {
1487 ext4_msg(sb, KERN_INFO, "Journaled quota options "
1488 "ignored when QUOTA feature is enabled");
1489 return 1;
1490 }
1491 qname = match_strdup(args);
1492 if (!qname) {
1493 ext4_msg(sb, KERN_ERR,
1494 "Not enough memory for storing quotafile name");
1495 return -1;
1496 }
1497 if (sbi->s_qf_names[qtype]) {
1498 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1499 ret = 1;
1500 else
1501 ext4_msg(sb, KERN_ERR,
1502 "%s quota file already specified",
1503 QTYPE2NAME(qtype));
1504 goto errout;
1505 }
1506 if (strchr(qname, '/')) {
1507 ext4_msg(sb, KERN_ERR,
1508 "quotafile must be on filesystem root");
1509 goto errout;
1510 }
1511 sbi->s_qf_names[qtype] = qname;
1512 set_opt(sb, QUOTA);
1513 return 1;
1514errout:
1515 kfree(qname);
1516 return ret;
1517}
1518
1519static int clear_qf_name(struct super_block *sb, int qtype)
1520{
1521
1522 struct ext4_sb_info *sbi = EXT4_SB(sb);
1523
1524 if (sb_any_quota_loaded(sb) &&
1525 sbi->s_qf_names[qtype]) {
1526 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1527 " when quota turned on");
1528 return -1;
1529 }
1530 kfree(sbi->s_qf_names[qtype]);
1531 sbi->s_qf_names[qtype] = NULL;
1532 return 1;
1533}
1534#endif
1535
1536#define MOPT_SET 0x0001
1537#define MOPT_CLEAR 0x0002
1538#define MOPT_NOSUPPORT 0x0004
1539#define MOPT_EXPLICIT 0x0008
1540#define MOPT_CLEAR_ERR 0x0010
1541#define MOPT_GTE0 0x0020
1542#ifdef CONFIG_QUOTA
1543#define MOPT_Q 0
1544#define MOPT_QFMT 0x0040
1545#else
1546#define MOPT_Q MOPT_NOSUPPORT
1547#define MOPT_QFMT MOPT_NOSUPPORT
1548#endif
1549#define MOPT_DATAJ 0x0080
1550#define MOPT_NO_EXT2 0x0100
1551#define MOPT_NO_EXT3 0x0200
1552#define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1553#define MOPT_STRING 0x0400
1554
1555static const struct mount_opts {
1556 int token;
1557 int mount_opt;
1558 int flags;
1559} ext4_mount_opts[] = {
1560 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1561 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1562 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1563 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1564 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1565 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1566 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1567 MOPT_EXT4_ONLY | MOPT_SET},
1568 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1569 MOPT_EXT4_ONLY | MOPT_CLEAR},
1570 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1571 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1572 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1573 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1574 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1575 MOPT_EXT4_ONLY | MOPT_CLEAR},
1576 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1577 MOPT_EXT4_ONLY | MOPT_CLEAR},
1578 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1579 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1580 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1581 EXT4_MOUNT_JOURNAL_CHECKSUM),
1582 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1583 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1584 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1585 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1586 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1587 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1588 MOPT_NO_EXT2},
1589 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1590 MOPT_NO_EXT2},
1591 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1592 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1593 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1594 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1595 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1596 {Opt_commit, 0, MOPT_GTE0},
1597 {Opt_max_batch_time, 0, MOPT_GTE0},
1598 {Opt_min_batch_time, 0, MOPT_GTE0},
1599 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1600 {Opt_init_itable, 0, MOPT_GTE0},
1601 {Opt_dax, EXT4_MOUNT_DAX, MOPT_SET},
1602 {Opt_stripe, 0, MOPT_GTE0},
1603 {Opt_resuid, 0, MOPT_GTE0},
1604 {Opt_resgid, 0, MOPT_GTE0},
1605 {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1606 {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING},
1607 {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1608 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1609 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1610 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1611 MOPT_NO_EXT2 | MOPT_DATAJ},
1612 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1613 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1614#ifdef CONFIG_EXT4_FS_POSIX_ACL
1615 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1616 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1617#else
1618 {Opt_acl, 0, MOPT_NOSUPPORT},
1619 {Opt_noacl, 0, MOPT_NOSUPPORT},
1620#endif
1621 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1622 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1623 {Opt_debug_want_extra_isize, 0, MOPT_GTE0},
1624 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1625 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1626 MOPT_SET | MOPT_Q},
1627 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1628 MOPT_SET | MOPT_Q},
1629 {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA,
1630 MOPT_SET | MOPT_Q},
1631 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1632 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA),
1633 MOPT_CLEAR | MOPT_Q},
1634 {Opt_usrjquota, 0, MOPT_Q},
1635 {Opt_grpjquota, 0, MOPT_Q},
1636 {Opt_offusrjquota, 0, MOPT_Q},
1637 {Opt_offgrpjquota, 0, MOPT_Q},
1638 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1639 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1640 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1641 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
1642 {Opt_test_dummy_encryption, 0, MOPT_GTE0},
1643 {Opt_nombcache, EXT4_MOUNT_NO_MBCACHE, MOPT_SET},
1644 {Opt_err, 0, 0}
1645};
1646
1647static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1648 substring_t *args, unsigned long *journal_devnum,
1649 unsigned int *journal_ioprio, int is_remount)
1650{
1651 struct ext4_sb_info *sbi = EXT4_SB(sb);
1652 const struct mount_opts *m;
1653 kuid_t uid;
1654 kgid_t gid;
1655 int arg = 0;
1656
1657#ifdef CONFIG_QUOTA
1658 if (token == Opt_usrjquota)
1659 return set_qf_name(sb, USRQUOTA, &args[0]);
1660 else if (token == Opt_grpjquota)
1661 return set_qf_name(sb, GRPQUOTA, &args[0]);
1662 else if (token == Opt_offusrjquota)
1663 return clear_qf_name(sb, USRQUOTA);
1664 else if (token == Opt_offgrpjquota)
1665 return clear_qf_name(sb, GRPQUOTA);
1666#endif
1667 switch (token) {
1668 case Opt_noacl:
1669 case Opt_nouser_xattr:
1670 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1671 break;
1672 case Opt_sb:
1673 return 1; /* handled by get_sb_block() */
1674 case Opt_removed:
1675 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
1676 return 1;
1677 case Opt_abort:
1678 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1679 return 1;
1680 case Opt_i_version:
1681 sb->s_flags |= MS_I_VERSION;
1682 return 1;
1683 case Opt_lazytime:
1684 sb->s_flags |= MS_LAZYTIME;
1685 return 1;
1686 case Opt_nolazytime:
1687 sb->s_flags &= ~MS_LAZYTIME;
1688 return 1;
1689 }
1690
1691 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1692 if (token == m->token)
1693 break;
1694
1695 if (m->token == Opt_err) {
1696 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1697 "or missing value", opt);
1698 return -1;
1699 }
1700
1701 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1702 ext4_msg(sb, KERN_ERR,
1703 "Mount option \"%s\" incompatible with ext2", opt);
1704 return -1;
1705 }
1706 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1707 ext4_msg(sb, KERN_ERR,
1708 "Mount option \"%s\" incompatible with ext3", opt);
1709 return -1;
1710 }
1711
1712 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
1713 return -1;
1714 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1715 return -1;
1716 if (m->flags & MOPT_EXPLICIT) {
1717 if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
1718 set_opt2(sb, EXPLICIT_DELALLOC);
1719 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
1720 set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM);
1721 } else
1722 return -1;
1723 }
1724 if (m->flags & MOPT_CLEAR_ERR)
1725 clear_opt(sb, ERRORS_MASK);
1726 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1727 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1728 "options when quota turned on");
1729 return -1;
1730 }
1731
1732 if (m->flags & MOPT_NOSUPPORT) {
1733 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1734 } else if (token == Opt_commit) {
1735 if (arg == 0)
1736 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1737 sbi->s_commit_interval = HZ * arg;
1738 } else if (token == Opt_debug_want_extra_isize) {
1739 sbi->s_want_extra_isize = arg;
1740 } else if (token == Opt_max_batch_time) {
1741 sbi->s_max_batch_time = arg;
1742 } else if (token == Opt_min_batch_time) {
1743 sbi->s_min_batch_time = arg;
1744 } else if (token == Opt_inode_readahead_blks) {
1745 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1746 ext4_msg(sb, KERN_ERR,
1747 "EXT4-fs: inode_readahead_blks must be "
1748 "0 or a power of 2 smaller than 2^31");
1749 return -1;
1750 }
1751 sbi->s_inode_readahead_blks = arg;
1752 } else if (token == Opt_init_itable) {
1753 set_opt(sb, INIT_INODE_TABLE);
1754 if (!args->from)
1755 arg = EXT4_DEF_LI_WAIT_MULT;
1756 sbi->s_li_wait_mult = arg;
1757 } else if (token == Opt_max_dir_size_kb) {
1758 sbi->s_max_dir_size_kb = arg;
1759 } else if (token == Opt_stripe) {
1760 sbi->s_stripe = arg;
1761 } else if (token == Opt_resuid) {
1762 uid = make_kuid(current_user_ns(), arg);
1763 if (!uid_valid(uid)) {
1764 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1765 return -1;
1766 }
1767 sbi->s_resuid = uid;
1768 } else if (token == Opt_resgid) {
1769 gid = make_kgid(current_user_ns(), arg);
1770 if (!gid_valid(gid)) {
1771 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1772 return -1;
1773 }
1774 sbi->s_resgid = gid;
1775 } else if (token == Opt_journal_dev) {
1776 if (is_remount) {
1777 ext4_msg(sb, KERN_ERR,
1778 "Cannot specify journal on remount");
1779 return -1;
1780 }
1781 *journal_devnum = arg;
1782 } else if (token == Opt_journal_path) {
1783 char *journal_path;
1784 struct inode *journal_inode;
1785 struct path path;
1786 int error;
1787
1788 if (is_remount) {
1789 ext4_msg(sb, KERN_ERR,
1790 "Cannot specify journal on remount");
1791 return -1;
1792 }
1793 journal_path = match_strdup(&args[0]);
1794 if (!journal_path) {
1795 ext4_msg(sb, KERN_ERR, "error: could not dup "
1796 "journal device string");
1797 return -1;
1798 }
1799
1800 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1801 if (error) {
1802 ext4_msg(sb, KERN_ERR, "error: could not find "
1803 "journal device path: error %d", error);
1804 kfree(journal_path);
1805 return -1;
1806 }
1807
1808 journal_inode = d_inode(path.dentry);
1809 if (!S_ISBLK(journal_inode->i_mode)) {
1810 ext4_msg(sb, KERN_ERR, "error: journal path %s "
1811 "is not a block device", journal_path);
1812 path_put(&path);
1813 kfree(journal_path);
1814 return -1;
1815 }
1816
1817 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
1818 path_put(&path);
1819 kfree(journal_path);
1820 } else if (token == Opt_journal_ioprio) {
1821 if (arg > 7) {
1822 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1823 " (must be 0-7)");
1824 return -1;
1825 }
1826 *journal_ioprio =
1827 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1828 } else if (token == Opt_test_dummy_encryption) {
1829#ifdef CONFIG_EXT4_FS_ENCRYPTION
1830 sbi->s_mount_flags |= EXT4_MF_TEST_DUMMY_ENCRYPTION;
1831 ext4_msg(sb, KERN_WARNING,
1832 "Test dummy encryption mode enabled");
1833#else
1834 ext4_msg(sb, KERN_WARNING,
1835 "Test dummy encryption mount option ignored");
1836#endif
1837 } else if (m->flags & MOPT_DATAJ) {
1838 if (is_remount) {
1839 if (!sbi->s_journal)
1840 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1841 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
1842 ext4_msg(sb, KERN_ERR,
1843 "Cannot change data mode on remount");
1844 return -1;
1845 }
1846 } else {
1847 clear_opt(sb, DATA_FLAGS);
1848 sbi->s_mount_opt |= m->mount_opt;
1849 }
1850#ifdef CONFIG_QUOTA
1851 } else if (m->flags & MOPT_QFMT) {
1852 if (sb_any_quota_loaded(sb) &&
1853 sbi->s_jquota_fmt != m->mount_opt) {
1854 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1855 "quota options when quota turned on");
1856 return -1;
1857 }
1858 if (ext4_has_feature_quota(sb)) {
1859 ext4_msg(sb, KERN_INFO,
1860 "Quota format mount options ignored "
1861 "when QUOTA feature is enabled");
1862 return 1;
1863 }
1864 sbi->s_jquota_fmt = m->mount_opt;
1865#endif
1866 } else if (token == Opt_dax) {
1867#ifdef CONFIG_FS_DAX
1868 ext4_msg(sb, KERN_WARNING,
1869 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
1870 sbi->s_mount_opt |= m->mount_opt;
1871#else
1872 ext4_msg(sb, KERN_INFO, "dax option not supported");
1873 return -1;
1874#endif
1875 } else if (token == Opt_data_err_abort) {
1876 sbi->s_mount_opt |= m->mount_opt;
1877 } else if (token == Opt_data_err_ignore) {
1878 sbi->s_mount_opt &= ~m->mount_opt;
1879 } else {
1880 if (!args->from)
1881 arg = 1;
1882 if (m->flags & MOPT_CLEAR)
1883 arg = !arg;
1884 else if (unlikely(!(m->flags & MOPT_SET))) {
1885 ext4_msg(sb, KERN_WARNING,
1886 "buggy handling of option %s", opt);
1887 WARN_ON(1);
1888 return -1;
1889 }
1890 if (arg != 0)
1891 sbi->s_mount_opt |= m->mount_opt;
1892 else
1893 sbi->s_mount_opt &= ~m->mount_opt;
1894 }
1895 return 1;
1896}
1897
1898static int parse_options(char *options, struct super_block *sb,
1899 unsigned long *journal_devnum,
1900 unsigned int *journal_ioprio,
1901 int is_remount)
1902{
1903 struct ext4_sb_info *sbi = EXT4_SB(sb);
1904 char *p;
1905 substring_t args[MAX_OPT_ARGS];
1906 int token;
1907
1908 if (!options)
1909 return 1;
1910
1911 while ((p = strsep(&options, ",")) != NULL) {
1912 if (!*p)
1913 continue;
1914 /*
1915 * Initialize args struct so we know whether arg was
1916 * found; some options take optional arguments.
1917 */
1918 args[0].to = args[0].from = NULL;
1919 token = match_token(p, tokens, args);
1920 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1921 journal_ioprio, is_remount) < 0)
1922 return 0;
1923 }
1924#ifdef CONFIG_QUOTA
1925 /*
1926 * We do the test below only for project quotas. 'usrquota' and
1927 * 'grpquota' mount options are allowed even without quota feature
1928 * to support legacy quotas in quota files.
1929 */
1930 if (test_opt(sb, PRJQUOTA) && !ext4_has_feature_project(sb)) {
1931 ext4_msg(sb, KERN_ERR, "Project quota feature not enabled. "
1932 "Cannot enable project quota enforcement.");
1933 return 0;
1934 }
1935 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1936 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1937 clear_opt(sb, USRQUOTA);
1938
1939 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1940 clear_opt(sb, GRPQUOTA);
1941
1942 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1943 ext4_msg(sb, KERN_ERR, "old and new quota "
1944 "format mixing");
1945 return 0;
1946 }
1947
1948 if (!sbi->s_jquota_fmt) {
1949 ext4_msg(sb, KERN_ERR, "journaled quota format "
1950 "not specified");
1951 return 0;
1952 }
1953 }
1954#endif
1955 if (test_opt(sb, DIOREAD_NOLOCK)) {
1956 int blocksize =
1957 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1958
1959 if (blocksize < PAGE_SIZE) {
1960 ext4_msg(sb, KERN_ERR, "can't mount with "
1961 "dioread_nolock if block size != PAGE_SIZE");
1962 return 0;
1963 }
1964 }
1965 return 1;
1966}
1967
1968static inline void ext4_show_quota_options(struct seq_file *seq,
1969 struct super_block *sb)
1970{
1971#if defined(CONFIG_QUOTA)
1972 struct ext4_sb_info *sbi = EXT4_SB(sb);
1973
1974 if (sbi->s_jquota_fmt) {
1975 char *fmtname = "";
1976
1977 switch (sbi->s_jquota_fmt) {
1978 case QFMT_VFS_OLD:
1979 fmtname = "vfsold";
1980 break;
1981 case QFMT_VFS_V0:
1982 fmtname = "vfsv0";
1983 break;
1984 case QFMT_VFS_V1:
1985 fmtname = "vfsv1";
1986 break;
1987 }
1988 seq_printf(seq, ",jqfmt=%s", fmtname);
1989 }
1990
1991 if (sbi->s_qf_names[USRQUOTA])
1992 seq_show_option(seq, "usrjquota", sbi->s_qf_names[USRQUOTA]);
1993
1994 if (sbi->s_qf_names[GRPQUOTA])
1995 seq_show_option(seq, "grpjquota", sbi->s_qf_names[GRPQUOTA]);
1996#endif
1997}
1998
1999static const char *token2str(int token)
2000{
2001 const struct match_token *t;
2002
2003 for (t = tokens; t->token != Opt_err; t++)
2004 if (t->token == token && !strchr(t->pattern, '='))
2005 break;
2006 return t->pattern;
2007}
2008
2009/*
2010 * Show an option if
2011 * - it's set to a non-default value OR
2012 * - if the per-sb default is different from the global default
2013 */
2014static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
2015 int nodefs)
2016{
2017 struct ext4_sb_info *sbi = EXT4_SB(sb);
2018 struct ext4_super_block *es = sbi->s_es;
2019 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
2020 const struct mount_opts *m;
2021 char sep = nodefs ? '\n' : ',';
2022
2023#define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
2024#define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2025
2026 if (sbi->s_sb_block != 1)
2027 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
2028
2029 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
2030 int want_set = m->flags & MOPT_SET;
2031 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
2032 (m->flags & MOPT_CLEAR_ERR))
2033 continue;
2034 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
2035 continue; /* skip if same as the default */
2036 if ((want_set &&
2037 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
2038 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
2039 continue; /* select Opt_noFoo vs Opt_Foo */
2040 SEQ_OPTS_PRINT("%s", token2str(m->token));
2041 }
2042
2043 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
2044 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
2045 SEQ_OPTS_PRINT("resuid=%u",
2046 from_kuid_munged(&init_user_ns, sbi->s_resuid));
2047 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
2048 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
2049 SEQ_OPTS_PRINT("resgid=%u",
2050 from_kgid_munged(&init_user_ns, sbi->s_resgid));
2051 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
2052 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
2053 SEQ_OPTS_PUTS("errors=remount-ro");
2054 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
2055 SEQ_OPTS_PUTS("errors=continue");
2056 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
2057 SEQ_OPTS_PUTS("errors=panic");
2058 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
2059 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
2060 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
2061 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
2062 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
2063 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2064 if (sb->s_flags & MS_I_VERSION)
2065 SEQ_OPTS_PUTS("i_version");
2066 if (nodefs || sbi->s_stripe)
2067 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
2068 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
2069 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2070 SEQ_OPTS_PUTS("data=journal");
2071 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2072 SEQ_OPTS_PUTS("data=ordered");
2073 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
2074 SEQ_OPTS_PUTS("data=writeback");
2075 }
2076 if (nodefs ||
2077 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
2078 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
2079 sbi->s_inode_readahead_blks);
2080
2081 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
2082 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
2083 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
2084 if (nodefs || sbi->s_max_dir_size_kb)
2085 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
2086 if (test_opt(sb, DATA_ERR_ABORT))
2087 SEQ_OPTS_PUTS("data_err=abort");
2088
2089 ext4_show_quota_options(seq, sb);
2090 return 0;
2091}
2092
2093static int ext4_show_options(struct seq_file *seq, struct dentry *root)
2094{
2095 return _ext4_show_options(seq, root->d_sb, 0);
2096}
2097
2098int ext4_seq_options_show(struct seq_file *seq, void *offset)
2099{
2100 struct super_block *sb = seq->private;
2101 int rc;
2102
2103 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
2104 rc = _ext4_show_options(seq, sb, 1);
2105 seq_puts(seq, "\n");
2106 return rc;
2107}
2108
2109static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
2110 int read_only)
2111{
2112 struct ext4_sb_info *sbi = EXT4_SB(sb);
2113 int res = 0;
2114
2115 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
2116 ext4_msg(sb, KERN_ERR, "revision level too high, "
2117 "forcing read-only mode");
2118 res = MS_RDONLY;
2119 }
2120 if (read_only)
2121 goto done;
2122 if (!(sbi->s_mount_state & EXT4_VALID_FS))
2123 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
2124 "running e2fsck is recommended");
2125 else if (sbi->s_mount_state & EXT4_ERROR_FS)
2126 ext4_msg(sb, KERN_WARNING,
2127 "warning: mounting fs with errors, "
2128 "running e2fsck is recommended");
2129 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
2130 le16_to_cpu(es->s_mnt_count) >=
2131 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
2132 ext4_msg(sb, KERN_WARNING,
2133 "warning: maximal mount count reached, "
2134 "running e2fsck is recommended");
2135 else if (le32_to_cpu(es->s_checkinterval) &&
2136 (le32_to_cpu(es->s_lastcheck) +
2137 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
2138 ext4_msg(sb, KERN_WARNING,
2139 "warning: checktime reached, "
2140 "running e2fsck is recommended");
2141 if (!sbi->s_journal)
2142 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
2143 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
2144 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
2145 le16_add_cpu(&es->s_mnt_count, 1);
2146 es->s_mtime = cpu_to_le32(get_seconds());
2147 ext4_update_dynamic_rev(sb);
2148 if (sbi->s_journal)
2149 ext4_set_feature_journal_needs_recovery(sb);
2150
2151 ext4_commit_super(sb, 1);
2152done:
2153 if (test_opt(sb, DEBUG))
2154 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
2155 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
2156 sb->s_blocksize,
2157 sbi->s_groups_count,
2158 EXT4_BLOCKS_PER_GROUP(sb),
2159 EXT4_INODES_PER_GROUP(sb),
2160 sbi->s_mount_opt, sbi->s_mount_opt2);
2161
2162 cleancache_init_fs(sb);
2163 return res;
2164}
2165
2166int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
2167{
2168 struct ext4_sb_info *sbi = EXT4_SB(sb);
2169 struct flex_groups *new_groups;
2170 int size;
2171
2172 if (!sbi->s_log_groups_per_flex)
2173 return 0;
2174
2175 size = ext4_flex_group(sbi, ngroup - 1) + 1;
2176 if (size <= sbi->s_flex_groups_allocated)
2177 return 0;
2178
2179 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
2180 new_groups = kvzalloc(size, GFP_KERNEL);
2181 if (!new_groups) {
2182 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
2183 size / (int) sizeof(struct flex_groups));
2184 return -ENOMEM;
2185 }
2186
2187 if (sbi->s_flex_groups) {
2188 memcpy(new_groups, sbi->s_flex_groups,
2189 (sbi->s_flex_groups_allocated *
2190 sizeof(struct flex_groups)));
2191 kvfree(sbi->s_flex_groups);
2192 }
2193 sbi->s_flex_groups = new_groups;
2194 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
2195 return 0;
2196}
2197
2198static int ext4_fill_flex_info(struct super_block *sb)
2199{
2200 struct ext4_sb_info *sbi = EXT4_SB(sb);
2201 struct ext4_group_desc *gdp = NULL;
2202 ext4_group_t flex_group;
2203 int i, err;
2204
2205 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
2206 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
2207 sbi->s_log_groups_per_flex = 0;
2208 return 1;
2209 }
2210
2211 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
2212 if (err)
2213 goto failed;
2214
2215 for (i = 0; i < sbi->s_groups_count; i++) {
2216 gdp = ext4_get_group_desc(sb, i, NULL);
2217
2218 flex_group = ext4_flex_group(sbi, i);
2219 atomic_add(ext4_free_inodes_count(sb, gdp),
2220 &sbi->s_flex_groups[flex_group].free_inodes);
2221 atomic64_add(ext4_free_group_clusters(sb, gdp),
2222 &sbi->s_flex_groups[flex_group].free_clusters);
2223 atomic_add(ext4_used_dirs_count(sb, gdp),
2224 &sbi->s_flex_groups[flex_group].used_dirs);
2225 }
2226
2227 return 1;
2228failed:
2229 return 0;
2230}
2231
2232static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
2233 struct ext4_group_desc *gdp)
2234{
2235 int offset = offsetof(struct ext4_group_desc, bg_checksum);
2236 __u16 crc = 0;
2237 __le32 le_group = cpu_to_le32(block_group);
2238 struct ext4_sb_info *sbi = EXT4_SB(sb);
2239
2240 if (ext4_has_metadata_csum(sbi->s_sb)) {
2241 /* Use new metadata_csum algorithm */
2242 __u32 csum32;
2243 __u16 dummy_csum = 0;
2244
2245 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2246 sizeof(le_group));
2247 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset);
2248 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum,
2249 sizeof(dummy_csum));
2250 offset += sizeof(dummy_csum);
2251 if (offset < sbi->s_desc_size)
2252 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset,
2253 sbi->s_desc_size - offset);
2254
2255 crc = csum32 & 0xFFFF;
2256 goto out;
2257 }
2258
2259 /* old crc16 code */
2260 if (!ext4_has_feature_gdt_csum(sb))
2261 return 0;
2262
2263 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2264 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2265 crc = crc16(crc, (__u8 *)gdp, offset);
2266 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2267 /* for checksum of struct ext4_group_desc do the rest...*/
2268 if (ext4_has_feature_64bit(sb) &&
2269 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2270 crc = crc16(crc, (__u8 *)gdp + offset,
2271 le16_to_cpu(sbi->s_es->s_desc_size) -
2272 offset);
2273
2274out:
2275 return cpu_to_le16(crc);
2276}
2277
2278int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2279 struct ext4_group_desc *gdp)
2280{
2281 if (ext4_has_group_desc_csum(sb) &&
2282 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
2283 return 0;
2284
2285 return 1;
2286}
2287
2288void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2289 struct ext4_group_desc *gdp)
2290{
2291 if (!ext4_has_group_desc_csum(sb))
2292 return;
2293 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
2294}
2295
2296/* Called at mount-time, super-block is locked */
2297static int ext4_check_descriptors(struct super_block *sb,
2298 ext4_fsblk_t sb_block,
2299 ext4_group_t *first_not_zeroed)
2300{
2301 struct ext4_sb_info *sbi = EXT4_SB(sb);
2302 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2303 ext4_fsblk_t last_block;
2304 ext4_fsblk_t block_bitmap;
2305 ext4_fsblk_t inode_bitmap;
2306 ext4_fsblk_t inode_table;
2307 int flexbg_flag = 0;
2308 ext4_group_t i, grp = sbi->s_groups_count;
2309
2310 if (ext4_has_feature_flex_bg(sb))
2311 flexbg_flag = 1;
2312
2313 ext4_debug("Checking group descriptors");
2314
2315 for (i = 0; i < sbi->s_groups_count; i++) {
2316 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2317
2318 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2319 last_block = ext4_blocks_count(sbi->s_es) - 1;
2320 else
2321 last_block = first_block +
2322 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2323
2324 if ((grp == sbi->s_groups_count) &&
2325 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2326 grp = i;
2327
2328 block_bitmap = ext4_block_bitmap(sb, gdp);
2329 if (block_bitmap == sb_block) {
2330 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2331 "Block bitmap for group %u overlaps "
2332 "superblock", i);
2333 }
2334 if (block_bitmap < first_block || block_bitmap > last_block) {
2335 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2336 "Block bitmap for group %u not in group "
2337 "(block %llu)!", i, block_bitmap);
2338 return 0;
2339 }
2340 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2341 if (inode_bitmap == sb_block) {
2342 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2343 "Inode bitmap for group %u overlaps "
2344 "superblock", i);
2345 }
2346 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2347 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2348 "Inode bitmap for group %u not in group "
2349 "(block %llu)!", i, inode_bitmap);
2350 return 0;
2351 }
2352 inode_table = ext4_inode_table(sb, gdp);
2353 if (inode_table == sb_block) {
2354 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2355 "Inode table for group %u overlaps "
2356 "superblock", i);
2357 }
2358 if (inode_table < first_block ||
2359 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2360 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2361 "Inode table for group %u not in group "
2362 "(block %llu)!", i, inode_table);
2363 return 0;
2364 }
2365 ext4_lock_group(sb, i);
2366 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2367 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2368 "Checksum for group %u failed (%u!=%u)",
2369 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
2370 gdp)), le16_to_cpu(gdp->bg_checksum));
2371 if (!(sb->s_flags & MS_RDONLY)) {
2372 ext4_unlock_group(sb, i);
2373 return 0;
2374 }
2375 }
2376 ext4_unlock_group(sb, i);
2377 if (!flexbg_flag)
2378 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2379 }
2380 if (NULL != first_not_zeroed)
2381 *first_not_zeroed = grp;
2382 return 1;
2383}
2384
2385/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2386 * the superblock) which were deleted from all directories, but held open by
2387 * a process at the time of a crash. We walk the list and try to delete these
2388 * inodes at recovery time (only with a read-write filesystem).
2389 *
2390 * In order to keep the orphan inode chain consistent during traversal (in
2391 * case of crash during recovery), we link each inode into the superblock
2392 * orphan list_head and handle it the same way as an inode deletion during
2393 * normal operation (which journals the operations for us).
2394 *
2395 * We only do an iget() and an iput() on each inode, which is very safe if we
2396 * accidentally point at an in-use or already deleted inode. The worst that
2397 * can happen in this case is that we get a "bit already cleared" message from
2398 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2399 * e2fsck was run on this filesystem, and it must have already done the orphan
2400 * inode cleanup for us, so we can safely abort without any further action.
2401 */
2402static void ext4_orphan_cleanup(struct super_block *sb,
2403 struct ext4_super_block *es)
2404{
2405 unsigned int s_flags = sb->s_flags;
2406 int ret, nr_orphans = 0, nr_truncates = 0;
2407#ifdef CONFIG_QUOTA
2408 int i;
2409#endif
2410 if (!es->s_last_orphan) {
2411 jbd_debug(4, "no orphan inodes to clean up\n");
2412 return;
2413 }
2414
2415 if (bdev_read_only(sb->s_bdev)) {
2416 ext4_msg(sb, KERN_ERR, "write access "
2417 "unavailable, skipping orphan cleanup");
2418 return;
2419 }
2420
2421 /* Check if feature set would not allow a r/w mount */
2422 if (!ext4_feature_set_ok(sb, 0)) {
2423 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2424 "unknown ROCOMPAT features");
2425 return;
2426 }
2427
2428 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2429 /* don't clear list on RO mount w/ errors */
2430 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
2431 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
2432 "clearing orphan list.\n");
2433 es->s_last_orphan = 0;
2434 }
2435 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2436 return;
2437 }
2438
2439 if (s_flags & MS_RDONLY) {
2440 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2441 sb->s_flags &= ~MS_RDONLY;
2442 }
2443#ifdef CONFIG_QUOTA
2444 /* Needed for iput() to work correctly and not trash data */
2445 sb->s_flags |= MS_ACTIVE;
2446 /* Turn on quotas so that they are updated correctly */
2447 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2448 if (EXT4_SB(sb)->s_qf_names[i]) {
2449 int ret = ext4_quota_on_mount(sb, i);
2450 if (ret < 0)
2451 ext4_msg(sb, KERN_ERR,
2452 "Cannot turn on journaled "
2453 "quota: error %d", ret);
2454 }
2455 }
2456#endif
2457
2458 while (es->s_last_orphan) {
2459 struct inode *inode;
2460
2461 /*
2462 * We may have encountered an error during cleanup; if
2463 * so, skip the rest.
2464 */
2465 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2466 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2467 es->s_last_orphan = 0;
2468 break;
2469 }
2470
2471 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2472 if (IS_ERR(inode)) {
2473 es->s_last_orphan = 0;
2474 break;
2475 }
2476
2477 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2478 dquot_initialize(inode);
2479 if (inode->i_nlink) {
2480 if (test_opt(sb, DEBUG))
2481 ext4_msg(sb, KERN_DEBUG,
2482 "%s: truncating inode %lu to %lld bytes",
2483 __func__, inode->i_ino, inode->i_size);
2484 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2485 inode->i_ino, inode->i_size);
2486 inode_lock(inode);
2487 truncate_inode_pages(inode->i_mapping, inode->i_size);
2488 ret = ext4_truncate(inode);
2489 if (ret)
2490 ext4_std_error(inode->i_sb, ret);
2491 inode_unlock(inode);
2492 nr_truncates++;
2493 } else {
2494 if (test_opt(sb, DEBUG))
2495 ext4_msg(sb, KERN_DEBUG,
2496 "%s: deleting unreferenced inode %lu",
2497 __func__, inode->i_ino);
2498 jbd_debug(2, "deleting unreferenced inode %lu\n",
2499 inode->i_ino);
2500 nr_orphans++;
2501 }
2502 iput(inode); /* The delete magic happens here! */
2503 }
2504
2505#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2506
2507 if (nr_orphans)
2508 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2509 PLURAL(nr_orphans));
2510 if (nr_truncates)
2511 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2512 PLURAL(nr_truncates));
2513#ifdef CONFIG_QUOTA
2514 /* Turn quotas off */
2515 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2516 if (sb_dqopt(sb)->files[i])
2517 dquot_quota_off(sb, i);
2518 }
2519#endif
2520 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2521}
2522
2523/*
2524 * Maximal extent format file size.
2525 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2526 * extent format containers, within a sector_t, and within i_blocks
2527 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2528 * so that won't be a limiting factor.
2529 *
2530 * However there is other limiting factor. We do store extents in the form
2531 * of starting block and length, hence the resulting length of the extent
2532 * covering maximum file size must fit into on-disk format containers as
2533 * well. Given that length is always by 1 unit bigger than max unit (because
2534 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2535 *
2536 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2537 */
2538static loff_t ext4_max_size(int blkbits, int has_huge_files)
2539{
2540 loff_t res;
2541 loff_t upper_limit = MAX_LFS_FILESIZE;
2542
2543 /* small i_blocks in vfs inode? */
2544 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2545 /*
2546 * CONFIG_LBDAF is not enabled implies the inode
2547 * i_block represent total blocks in 512 bytes
2548 * 32 == size of vfs inode i_blocks * 8
2549 */
2550 upper_limit = (1LL << 32) - 1;
2551
2552 /* total blocks in file system block size */
2553 upper_limit >>= (blkbits - 9);
2554 upper_limit <<= blkbits;
2555 }
2556
2557 /*
2558 * 32-bit extent-start container, ee_block. We lower the maxbytes
2559 * by one fs block, so ee_len can cover the extent of maximum file
2560 * size
2561 */
2562 res = (1LL << 32) - 1;
2563 res <<= blkbits;
2564
2565 /* Sanity check against vm- & vfs- imposed limits */
2566 if (res > upper_limit)
2567 res = upper_limit;
2568
2569 return res;
2570}
2571
2572/*
2573 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2574 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2575 * We need to be 1 filesystem block less than the 2^48 sector limit.
2576 */
2577static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2578{
2579 loff_t res = EXT4_NDIR_BLOCKS;
2580 int meta_blocks;
2581 loff_t upper_limit;
2582 /* This is calculated to be the largest file size for a dense, block
2583 * mapped file such that the file's total number of 512-byte sectors,
2584 * including data and all indirect blocks, does not exceed (2^48 - 1).
2585 *
2586 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2587 * number of 512-byte sectors of the file.
2588 */
2589
2590 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2591 /*
2592 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2593 * the inode i_block field represents total file blocks in
2594 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2595 */
2596 upper_limit = (1LL << 32) - 1;
2597
2598 /* total blocks in file system block size */
2599 upper_limit >>= (bits - 9);
2600
2601 } else {
2602 /*
2603 * We use 48 bit ext4_inode i_blocks
2604 * With EXT4_HUGE_FILE_FL set the i_blocks
2605 * represent total number of blocks in
2606 * file system block size
2607 */
2608 upper_limit = (1LL << 48) - 1;
2609
2610 }
2611
2612 /* indirect blocks */
2613 meta_blocks = 1;
2614 /* double indirect blocks */
2615 meta_blocks += 1 + (1LL << (bits-2));
2616 /* tripple indirect blocks */
2617 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2618
2619 upper_limit -= meta_blocks;
2620 upper_limit <<= bits;
2621
2622 res += 1LL << (bits-2);
2623 res += 1LL << (2*(bits-2));
2624 res += 1LL << (3*(bits-2));
2625 res <<= bits;
2626 if (res > upper_limit)
2627 res = upper_limit;
2628
2629 if (res > MAX_LFS_FILESIZE)
2630 res = MAX_LFS_FILESIZE;
2631
2632 return res;
2633}
2634
2635static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2636 ext4_fsblk_t logical_sb_block, int nr)
2637{
2638 struct ext4_sb_info *sbi = EXT4_SB(sb);
2639 ext4_group_t bg, first_meta_bg;
2640 int has_super = 0;
2641
2642 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2643
2644 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
2645 return logical_sb_block + nr + 1;
2646 bg = sbi->s_desc_per_block * nr;
2647 if (ext4_bg_has_super(sb, bg))
2648 has_super = 1;
2649
2650 /*
2651 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2652 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2653 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2654 * compensate.
2655 */
2656 if (sb->s_blocksize == 1024 && nr == 0 &&
2657 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block) == 0)
2658 has_super++;
2659
2660 return (has_super + ext4_group_first_block_no(sb, bg));
2661}
2662
2663/**
2664 * ext4_get_stripe_size: Get the stripe size.
2665 * @sbi: In memory super block info
2666 *
2667 * If we have specified it via mount option, then
2668 * use the mount option value. If the value specified at mount time is
2669 * greater than the blocks per group use the super block value.
2670 * If the super block value is greater than blocks per group return 0.
2671 * Allocator needs it be less than blocks per group.
2672 *
2673 */
2674static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2675{
2676 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2677 unsigned long stripe_width =
2678 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2679 int ret;
2680
2681 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2682 ret = sbi->s_stripe;
2683 else if (stripe_width && stripe_width <= sbi->s_blocks_per_group)
2684 ret = stripe_width;
2685 else if (stride && stride <= sbi->s_blocks_per_group)
2686 ret = stride;
2687 else
2688 ret = 0;
2689
2690 /*
2691 * If the stripe width is 1, this makes no sense and
2692 * we set it to 0 to turn off stripe handling code.
2693 */
2694 if (ret <= 1)
2695 ret = 0;
2696
2697 return ret;
2698}
2699
2700/*
2701 * Check whether this filesystem can be mounted based on
2702 * the features present and the RDONLY/RDWR mount requested.
2703 * Returns 1 if this filesystem can be mounted as requested,
2704 * 0 if it cannot be.
2705 */
2706static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2707{
2708 if (ext4_has_unknown_ext4_incompat_features(sb)) {
2709 ext4_msg(sb, KERN_ERR,
2710 "Couldn't mount because of "
2711 "unsupported optional features (%x)",
2712 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2713 ~EXT4_FEATURE_INCOMPAT_SUPP));
2714 return 0;
2715 }
2716
2717 if (readonly)
2718 return 1;
2719
2720 if (ext4_has_feature_readonly(sb)) {
2721 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
2722 sb->s_flags |= MS_RDONLY;
2723 return 1;
2724 }
2725
2726 /* Check that feature set is OK for a read-write mount */
2727 if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
2728 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2729 "unsupported optional features (%x)",
2730 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2731 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2732 return 0;
2733 }
2734 /*
2735 * Large file size enabled file system can only be mounted
2736 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2737 */
2738 if (ext4_has_feature_huge_file(sb)) {
2739 if (sizeof(blkcnt_t) < sizeof(u64)) {
2740 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2741 "cannot be mounted RDWR without "
2742 "CONFIG_LBDAF");
2743 return 0;
2744 }
2745 }
2746 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
2747 ext4_msg(sb, KERN_ERR,
2748 "Can't support bigalloc feature without "
2749 "extents feature\n");
2750 return 0;
2751 }
2752
2753#ifndef CONFIG_QUOTA
2754 if (ext4_has_feature_quota(sb) && !readonly) {
2755 ext4_msg(sb, KERN_ERR,
2756 "Filesystem with quota feature cannot be mounted RDWR "
2757 "without CONFIG_QUOTA");
2758 return 0;
2759 }
2760 if (ext4_has_feature_project(sb) && !readonly) {
2761 ext4_msg(sb, KERN_ERR,
2762 "Filesystem with project quota feature cannot be mounted RDWR "
2763 "without CONFIG_QUOTA");
2764 return 0;
2765 }
2766#endif /* CONFIG_QUOTA */
2767 return 1;
2768}
2769
2770/*
2771 * This function is called once a day if we have errors logged
2772 * on the file system
2773 */
2774static void print_daily_error_info(unsigned long arg)
2775{
2776 struct super_block *sb = (struct super_block *) arg;
2777 struct ext4_sb_info *sbi;
2778 struct ext4_super_block *es;
2779
2780 sbi = EXT4_SB(sb);
2781 es = sbi->s_es;
2782
2783 if (es->s_error_count)
2784 /* fsck newer than v1.41.13 is needed to clean this condition. */
2785 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
2786 le32_to_cpu(es->s_error_count));
2787 if (es->s_first_error_time) {
2788 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d",
2789 sb->s_id, le32_to_cpu(es->s_first_error_time),
2790 (int) sizeof(es->s_first_error_func),
2791 es->s_first_error_func,
2792 le32_to_cpu(es->s_first_error_line));
2793 if (es->s_first_error_ino)
2794 printk(KERN_CONT ": inode %u",
2795 le32_to_cpu(es->s_first_error_ino));
2796 if (es->s_first_error_block)
2797 printk(KERN_CONT ": block %llu", (unsigned long long)
2798 le64_to_cpu(es->s_first_error_block));
2799 printk(KERN_CONT "\n");
2800 }
2801 if (es->s_last_error_time) {
2802 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d",
2803 sb->s_id, le32_to_cpu(es->s_last_error_time),
2804 (int) sizeof(es->s_last_error_func),
2805 es->s_last_error_func,
2806 le32_to_cpu(es->s_last_error_line));
2807 if (es->s_last_error_ino)
2808 printk(KERN_CONT ": inode %u",
2809 le32_to_cpu(es->s_last_error_ino));
2810 if (es->s_last_error_block)
2811 printk(KERN_CONT ": block %llu", (unsigned long long)
2812 le64_to_cpu(es->s_last_error_block));
2813 printk(KERN_CONT "\n");
2814 }
2815 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2816}
2817
2818/* Find next suitable group and run ext4_init_inode_table */
2819static int ext4_run_li_request(struct ext4_li_request *elr)
2820{
2821 struct ext4_group_desc *gdp = NULL;
2822 ext4_group_t group, ngroups;
2823 struct super_block *sb;
2824 unsigned long timeout = 0;
2825 int ret = 0;
2826
2827 sb = elr->lr_super;
2828 ngroups = EXT4_SB(sb)->s_groups_count;
2829
2830 for (group = elr->lr_next_group; group < ngroups; group++) {
2831 gdp = ext4_get_group_desc(sb, group, NULL);
2832 if (!gdp) {
2833 ret = 1;
2834 break;
2835 }
2836
2837 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2838 break;
2839 }
2840
2841 if (group >= ngroups)
2842 ret = 1;
2843
2844 if (!ret) {
2845 timeout = jiffies;
2846 ret = ext4_init_inode_table(sb, group,
2847 elr->lr_timeout ? 0 : 1);
2848 if (elr->lr_timeout == 0) {
2849 timeout = (jiffies - timeout) *
2850 elr->lr_sbi->s_li_wait_mult;
2851 elr->lr_timeout = timeout;
2852 }
2853 elr->lr_next_sched = jiffies + elr->lr_timeout;
2854 elr->lr_next_group = group + 1;
2855 }
2856 return ret;
2857}
2858
2859/*
2860 * Remove lr_request from the list_request and free the
2861 * request structure. Should be called with li_list_mtx held
2862 */
2863static void ext4_remove_li_request(struct ext4_li_request *elr)
2864{
2865 struct ext4_sb_info *sbi;
2866
2867 if (!elr)
2868 return;
2869
2870 sbi = elr->lr_sbi;
2871
2872 list_del(&elr->lr_request);
2873 sbi->s_li_request = NULL;
2874 kfree(elr);
2875}
2876
2877static void ext4_unregister_li_request(struct super_block *sb)
2878{
2879 mutex_lock(&ext4_li_mtx);
2880 if (!ext4_li_info) {
2881 mutex_unlock(&ext4_li_mtx);
2882 return;
2883 }
2884
2885 mutex_lock(&ext4_li_info->li_list_mtx);
2886 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2887 mutex_unlock(&ext4_li_info->li_list_mtx);
2888 mutex_unlock(&ext4_li_mtx);
2889}
2890
2891static struct task_struct *ext4_lazyinit_task;
2892
2893/*
2894 * This is the function where ext4lazyinit thread lives. It walks
2895 * through the request list searching for next scheduled filesystem.
2896 * When such a fs is found, run the lazy initialization request
2897 * (ext4_rn_li_request) and keep track of the time spend in this
2898 * function. Based on that time we compute next schedule time of
2899 * the request. When walking through the list is complete, compute
2900 * next waking time and put itself into sleep.
2901 */
2902static int ext4_lazyinit_thread(void *arg)
2903{
2904 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2905 struct list_head *pos, *n;
2906 struct ext4_li_request *elr;
2907 unsigned long next_wakeup, cur;
2908
2909 BUG_ON(NULL == eli);
2910
2911cont_thread:
2912 while (true) {
2913 next_wakeup = MAX_JIFFY_OFFSET;
2914
2915 mutex_lock(&eli->li_list_mtx);
2916 if (list_empty(&eli->li_request_list)) {
2917 mutex_unlock(&eli->li_list_mtx);
2918 goto exit_thread;
2919 }
2920 list_for_each_safe(pos, n, &eli->li_request_list) {
2921 int err = 0;
2922 int progress = 0;
2923 elr = list_entry(pos, struct ext4_li_request,
2924 lr_request);
2925
2926 if (time_before(jiffies, elr->lr_next_sched)) {
2927 if (time_before(elr->lr_next_sched, next_wakeup))
2928 next_wakeup = elr->lr_next_sched;
2929 continue;
2930 }
2931 if (down_read_trylock(&elr->lr_super->s_umount)) {
2932 if (sb_start_write_trylock(elr->lr_super)) {
2933 progress = 1;
2934 /*
2935 * We hold sb->s_umount, sb can not
2936 * be removed from the list, it is
2937 * now safe to drop li_list_mtx
2938 */
2939 mutex_unlock(&eli->li_list_mtx);
2940 err = ext4_run_li_request(elr);
2941 sb_end_write(elr->lr_super);
2942 mutex_lock(&eli->li_list_mtx);
2943 n = pos->next;
2944 }
2945 up_read((&elr->lr_super->s_umount));
2946 }
2947 /* error, remove the lazy_init job */
2948 if (err) {
2949 ext4_remove_li_request(elr);
2950 continue;
2951 }
2952 if (!progress) {
2953 elr->lr_next_sched = jiffies +
2954 (prandom_u32()
2955 % (EXT4_DEF_LI_MAX_START_DELAY * HZ));
2956 }
2957 if (time_before(elr->lr_next_sched, next_wakeup))
2958 next_wakeup = elr->lr_next_sched;
2959 }
2960 mutex_unlock(&eli->li_list_mtx);
2961
2962 try_to_freeze();
2963
2964 cur = jiffies;
2965 if ((time_after_eq(cur, next_wakeup)) ||
2966 (MAX_JIFFY_OFFSET == next_wakeup)) {
2967 cond_resched();
2968 continue;
2969 }
2970
2971 schedule_timeout_interruptible(next_wakeup - cur);
2972
2973 if (kthread_should_stop()) {
2974 ext4_clear_request_list();
2975 goto exit_thread;
2976 }
2977 }
2978
2979exit_thread:
2980 /*
2981 * It looks like the request list is empty, but we need
2982 * to check it under the li_list_mtx lock, to prevent any
2983 * additions into it, and of course we should lock ext4_li_mtx
2984 * to atomically free the list and ext4_li_info, because at
2985 * this point another ext4 filesystem could be registering
2986 * new one.
2987 */
2988 mutex_lock(&ext4_li_mtx);
2989 mutex_lock(&eli->li_list_mtx);
2990 if (!list_empty(&eli->li_request_list)) {
2991 mutex_unlock(&eli->li_list_mtx);
2992 mutex_unlock(&ext4_li_mtx);
2993 goto cont_thread;
2994 }
2995 mutex_unlock(&eli->li_list_mtx);
2996 kfree(ext4_li_info);
2997 ext4_li_info = NULL;
2998 mutex_unlock(&ext4_li_mtx);
2999
3000 return 0;
3001}
3002
3003static void ext4_clear_request_list(void)
3004{
3005 struct list_head *pos, *n;
3006 struct ext4_li_request *elr;
3007
3008 mutex_lock(&ext4_li_info->li_list_mtx);
3009 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3010 elr = list_entry(pos, struct ext4_li_request,
3011 lr_request);
3012 ext4_remove_li_request(elr);
3013 }
3014 mutex_unlock(&ext4_li_info->li_list_mtx);
3015}
3016
3017static int ext4_run_lazyinit_thread(void)
3018{
3019 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3020 ext4_li_info, "ext4lazyinit");
3021 if (IS_ERR(ext4_lazyinit_task)) {
3022 int err = PTR_ERR(ext4_lazyinit_task);
3023 ext4_clear_request_list();
3024 kfree(ext4_li_info);
3025 ext4_li_info = NULL;
3026 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
3027 "initialization thread\n",
3028 err);
3029 return err;
3030 }
3031 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
3032 return 0;
3033}
3034
3035/*
3036 * Check whether it make sense to run itable init. thread or not.
3037 * If there is at least one uninitialized inode table, return
3038 * corresponding group number, else the loop goes through all
3039 * groups and return total number of groups.
3040 */
3041static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3042{
3043 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3044 struct ext4_group_desc *gdp = NULL;
3045
3046 for (group = 0; group < ngroups; group++) {
3047 gdp = ext4_get_group_desc(sb, group, NULL);
3048 if (!gdp)
3049 continue;
3050
3051 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3052 break;
3053 }
3054
3055 return group;
3056}
3057
3058static int ext4_li_info_new(void)
3059{
3060 struct ext4_lazy_init *eli = NULL;
3061
3062 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3063 if (!eli)
3064 return -ENOMEM;
3065
3066 INIT_LIST_HEAD(&eli->li_request_list);
3067 mutex_init(&eli->li_list_mtx);
3068
3069 eli->li_state |= EXT4_LAZYINIT_QUIT;
3070
3071 ext4_li_info = eli;
3072
3073 return 0;
3074}
3075
3076static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3077 ext4_group_t start)
3078{
3079 struct ext4_sb_info *sbi = EXT4_SB(sb);
3080 struct ext4_li_request *elr;
3081
3082 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3083 if (!elr)
3084 return NULL;
3085
3086 elr->lr_super = sb;
3087 elr->lr_sbi = sbi;
3088 elr->lr_next_group = start;
3089
3090 /*
3091 * Randomize first schedule time of the request to
3092 * spread the inode table initialization requests
3093 * better.
3094 */
3095 elr->lr_next_sched = jiffies + (prandom_u32() %
3096 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
3097 return elr;
3098}
3099
3100int ext4_register_li_request(struct super_block *sb,
3101 ext4_group_t first_not_zeroed)
3102{
3103 struct ext4_sb_info *sbi = EXT4_SB(sb);
3104 struct ext4_li_request *elr = NULL;
3105 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3106 int ret = 0;
3107
3108 mutex_lock(&ext4_li_mtx);
3109 if (sbi->s_li_request != NULL) {
3110 /*
3111 * Reset timeout so it can be computed again, because
3112 * s_li_wait_mult might have changed.
3113 */
3114 sbi->s_li_request->lr_timeout = 0;
3115 goto out;
3116 }
3117
3118 if (first_not_zeroed == ngroups ||
3119 (sb->s_flags & MS_RDONLY) ||
3120 !test_opt(sb, INIT_INODE_TABLE))
3121 goto out;
3122
3123 elr = ext4_li_request_new(sb, first_not_zeroed);
3124 if (!elr) {
3125 ret = -ENOMEM;
3126 goto out;
3127 }
3128
3129 if (NULL == ext4_li_info) {
3130 ret = ext4_li_info_new();
3131 if (ret)
3132 goto out;
3133 }
3134
3135 mutex_lock(&ext4_li_info->li_list_mtx);
3136 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3137 mutex_unlock(&ext4_li_info->li_list_mtx);
3138
3139 sbi->s_li_request = elr;
3140 /*
3141 * set elr to NULL here since it has been inserted to
3142 * the request_list and the removal and free of it is
3143 * handled by ext4_clear_request_list from now on.
3144 */
3145 elr = NULL;
3146
3147 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3148 ret = ext4_run_lazyinit_thread();
3149 if (ret)
3150 goto out;
3151 }
3152out:
3153 mutex_unlock(&ext4_li_mtx);
3154 if (ret)
3155 kfree(elr);
3156 return ret;
3157}
3158
3159/*
3160 * We do not need to lock anything since this is called on
3161 * module unload.
3162 */
3163static void ext4_destroy_lazyinit_thread(void)
3164{
3165 /*
3166 * If thread exited earlier
3167 * there's nothing to be done.
3168 */
3169 if (!ext4_li_info || !ext4_lazyinit_task)
3170 return;
3171
3172 kthread_stop(ext4_lazyinit_task);
3173}
3174
3175static int set_journal_csum_feature_set(struct super_block *sb)
3176{
3177 int ret = 1;
3178 int compat, incompat;
3179 struct ext4_sb_info *sbi = EXT4_SB(sb);
3180
3181 if (ext4_has_metadata_csum(sb)) {
3182 /* journal checksum v3 */
3183 compat = 0;
3184 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
3185 } else {
3186 /* journal checksum v1 */
3187 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3188 incompat = 0;
3189 }
3190
3191 jbd2_journal_clear_features(sbi->s_journal,
3192 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3193 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3194 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3195 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3196 ret = jbd2_journal_set_features(sbi->s_journal,
3197 compat, 0,
3198 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3199 incompat);
3200 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3201 ret = jbd2_journal_set_features(sbi->s_journal,
3202 compat, 0,
3203 incompat);
3204 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3205 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3206 } else {
3207 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3208 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3209 }
3210
3211 return ret;
3212}
3213
3214/*
3215 * Note: calculating the overhead so we can be compatible with
3216 * historical BSD practice is quite difficult in the face of
3217 * clusters/bigalloc. This is because multiple metadata blocks from
3218 * different block group can end up in the same allocation cluster.
3219 * Calculating the exact overhead in the face of clustered allocation
3220 * requires either O(all block bitmaps) in memory or O(number of block
3221 * groups**2) in time. We will still calculate the superblock for
3222 * older file systems --- and if we come across with a bigalloc file
3223 * system with zero in s_overhead_clusters the estimate will be close to
3224 * correct especially for very large cluster sizes --- but for newer
3225 * file systems, it's better to calculate this figure once at mkfs
3226 * time, and store it in the superblock. If the superblock value is
3227 * present (even for non-bigalloc file systems), we will use it.
3228 */
3229static int count_overhead(struct super_block *sb, ext4_group_t grp,
3230 char *buf)
3231{
3232 struct ext4_sb_info *sbi = EXT4_SB(sb);
3233 struct ext4_group_desc *gdp;
3234 ext4_fsblk_t first_block, last_block, b;
3235 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3236 int s, j, count = 0;
3237
3238 if (!ext4_has_feature_bigalloc(sb))
3239 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3240 sbi->s_itb_per_group + 2);
3241
3242 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3243 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3244 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3245 for (i = 0; i < ngroups; i++) {
3246 gdp = ext4_get_group_desc(sb, i, NULL);
3247 b = ext4_block_bitmap(sb, gdp);
3248 if (b >= first_block && b <= last_block) {
3249 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3250 count++;
3251 }
3252 b = ext4_inode_bitmap(sb, gdp);
3253 if (b >= first_block && b <= last_block) {
3254 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3255 count++;
3256 }
3257 b = ext4_inode_table(sb, gdp);
3258 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3259 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3260 int c = EXT4_B2C(sbi, b - first_block);
3261 ext4_set_bit(c, buf);
3262 count++;
3263 }
3264 if (i != grp)
3265 continue;
3266 s = 0;
3267 if (ext4_bg_has_super(sb, grp)) {
3268 ext4_set_bit(s++, buf);
3269 count++;
3270 }
3271 j = ext4_bg_num_gdb(sb, grp);
3272 if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) {
3273 ext4_error(sb, "Invalid number of block group "
3274 "descriptor blocks: %d", j);
3275 j = EXT4_BLOCKS_PER_GROUP(sb) - s;
3276 }
3277 count += j;
3278 for (; j > 0; j--)
3279 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3280 }
3281 if (!count)
3282 return 0;
3283 return EXT4_CLUSTERS_PER_GROUP(sb) -
3284 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3285}
3286
3287/*
3288 * Compute the overhead and stash it in sbi->s_overhead
3289 */
3290int ext4_calculate_overhead(struct super_block *sb)
3291{
3292 struct ext4_sb_info *sbi = EXT4_SB(sb);
3293 struct ext4_super_block *es = sbi->s_es;
3294 struct inode *j_inode;
3295 unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum);
3296 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3297 ext4_fsblk_t overhead = 0;
3298 char *buf = (char *) get_zeroed_page(GFP_NOFS);
3299
3300 if (!buf)
3301 return -ENOMEM;
3302
3303 /*
3304 * Compute the overhead (FS structures). This is constant
3305 * for a given filesystem unless the number of block groups
3306 * changes so we cache the previous value until it does.
3307 */
3308
3309 /*
3310 * All of the blocks before first_data_block are overhead
3311 */
3312 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3313
3314 /*
3315 * Add the overhead found in each block group
3316 */
3317 for (i = 0; i < ngroups; i++) {
3318 int blks;
3319
3320 blks = count_overhead(sb, i, buf);
3321 overhead += blks;
3322 if (blks)
3323 memset(buf, 0, PAGE_SIZE);
3324 cond_resched();
3325 }
3326
3327 /*
3328 * Add the internal journal blocks whether the journal has been
3329 * loaded or not
3330 */
3331 if (sbi->s_journal && !sbi->journal_bdev)
3332 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3333 else if (ext4_has_feature_journal(sb) && !sbi->s_journal) {
3334 j_inode = ext4_get_journal_inode(sb, j_inum);
3335 if (j_inode) {
3336 j_blocks = j_inode->i_size >> sb->s_blocksize_bits;
3337 overhead += EXT4_NUM_B2C(sbi, j_blocks);
3338 iput(j_inode);
3339 } else {
3340 ext4_msg(sb, KERN_ERR, "can't get journal size");
3341 }
3342 }
3343 sbi->s_overhead = overhead;
3344 smp_wmb();
3345 free_page((unsigned long) buf);
3346 return 0;
3347}
3348
3349static void ext4_set_resv_clusters(struct super_block *sb)
3350{
3351 ext4_fsblk_t resv_clusters;
3352 struct ext4_sb_info *sbi = EXT4_SB(sb);
3353
3354 /*
3355 * There's no need to reserve anything when we aren't using extents.
3356 * The space estimates are exact, there are no unwritten extents,
3357 * hole punching doesn't need new metadata... This is needed especially
3358 * to keep ext2/3 backward compatibility.
3359 */
3360 if (!ext4_has_feature_extents(sb))
3361 return;
3362 /*
3363 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3364 * This should cover the situations where we can not afford to run
3365 * out of space like for example punch hole, or converting
3366 * unwritten extents in delalloc path. In most cases such
3367 * allocation would require 1, or 2 blocks, higher numbers are
3368 * very rare.
3369 */
3370 resv_clusters = (ext4_blocks_count(sbi->s_es) >>
3371 sbi->s_cluster_bits);
3372
3373 do_div(resv_clusters, 50);
3374 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3375
3376 atomic64_set(&sbi->s_resv_clusters, resv_clusters);
3377}
3378
3379static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3380{
3381 struct dax_device *dax_dev = fs_dax_get_by_bdev(sb->s_bdev);
3382 char *orig_data = kstrdup(data, GFP_KERNEL);
3383 struct buffer_head *bh;
3384 struct ext4_super_block *es = NULL;
3385 struct ext4_sb_info *sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3386 ext4_fsblk_t block;
3387 ext4_fsblk_t sb_block = get_sb_block(&data);
3388 ext4_fsblk_t logical_sb_block;
3389 unsigned long offset = 0;
3390 unsigned long journal_devnum = 0;
3391 unsigned long def_mount_opts;
3392 struct inode *root;
3393 const char *descr;
3394 int ret = -ENOMEM;
3395 int blocksize, clustersize;
3396 unsigned int db_count;
3397 unsigned int i;
3398 int needs_recovery, has_huge_files, has_bigalloc;
3399 __u64 blocks_count;
3400 int err = 0;
3401 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3402 ext4_group_t first_not_zeroed;
3403
3404 sbi->s_daxdev = dax_dev;
3405 if ((data && !orig_data) || !sbi)
3406 goto out_free_base;
3407
3408 sbi->s_blockgroup_lock =
3409 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3410 if (!sbi->s_blockgroup_lock)
3411 goto out_free_base;
3412
3413 sb->s_fs_info = sbi;
3414 sbi->s_sb = sb;
3415 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3416 sbi->s_sb_block = sb_block;
3417 if (sb->s_bdev->bd_part)
3418 sbi->s_sectors_written_start =
3419 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3420
3421 /* Cleanup superblock name */
3422 strreplace(sb->s_id, '/', '!');
3423
3424 /* -EINVAL is default */
3425 ret = -EINVAL;
3426 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3427 if (!blocksize) {
3428 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3429 goto out_fail;
3430 }
3431
3432 /*
3433 * The ext4 superblock will not be buffer aligned for other than 1kB
3434 * block sizes. We need to calculate the offset from buffer start.
3435 */
3436 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3437 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3438 offset = do_div(logical_sb_block, blocksize);
3439 } else {
3440 logical_sb_block = sb_block;
3441 }
3442
3443 if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
3444 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3445 goto out_fail;
3446 }
3447 /*
3448 * Note: s_es must be initialized as soon as possible because
3449 * some ext4 macro-instructions depend on its value
3450 */
3451 es = (struct ext4_super_block *) (bh->b_data + offset);
3452 sbi->s_es = es;
3453 sb->s_magic = le16_to_cpu(es->s_magic);
3454 if (sb->s_magic != EXT4_SUPER_MAGIC)
3455 goto cantfind_ext4;
3456 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3457
3458 /* Warn if metadata_csum and gdt_csum are both set. */
3459 if (ext4_has_feature_metadata_csum(sb) &&
3460 ext4_has_feature_gdt_csum(sb))
3461 ext4_warning(sb, "metadata_csum and uninit_bg are "
3462 "redundant flags; please run fsck.");
3463
3464 /* Check for a known checksum algorithm */
3465 if (!ext4_verify_csum_type(sb, es)) {
3466 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3467 "unknown checksum algorithm.");
3468 silent = 1;
3469 goto cantfind_ext4;
3470 }
3471
3472 /* Load the checksum driver */
3473 if (ext4_has_feature_metadata_csum(sb) ||
3474 ext4_has_feature_ea_inode(sb)) {
3475 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3476 if (IS_ERR(sbi->s_chksum_driver)) {
3477 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3478 ret = PTR_ERR(sbi->s_chksum_driver);
3479 sbi->s_chksum_driver = NULL;
3480 goto failed_mount;
3481 }
3482 }
3483
3484 /* Check superblock checksum */
3485 if (!ext4_superblock_csum_verify(sb, es)) {
3486 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3487 "invalid superblock checksum. Run e2fsck?");
3488 silent = 1;
3489 ret = -EFSBADCRC;
3490 goto cantfind_ext4;
3491 }
3492
3493 /* Precompute checksum seed for all metadata */
3494 if (ext4_has_feature_csum_seed(sb))
3495 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
3496 else if (ext4_has_metadata_csum(sb) || ext4_has_feature_ea_inode(sb))
3497 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3498 sizeof(es->s_uuid));
3499
3500 /* Set defaults before we parse the mount options */
3501 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3502 set_opt(sb, INIT_INODE_TABLE);
3503 if (def_mount_opts & EXT4_DEFM_DEBUG)
3504 set_opt(sb, DEBUG);
3505 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3506 set_opt(sb, GRPID);
3507 if (def_mount_opts & EXT4_DEFM_UID16)
3508 set_opt(sb, NO_UID32);
3509 /* xattr user namespace & acls are now defaulted on */
3510 set_opt(sb, XATTR_USER);
3511#ifdef CONFIG_EXT4_FS_POSIX_ACL
3512 set_opt(sb, POSIX_ACL);
3513#endif
3514 /* don't forget to enable journal_csum when metadata_csum is enabled. */
3515 if (ext4_has_metadata_csum(sb))
3516 set_opt(sb, JOURNAL_CHECKSUM);
3517
3518 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3519 set_opt(sb, JOURNAL_DATA);
3520 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3521 set_opt(sb, ORDERED_DATA);
3522 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3523 set_opt(sb, WRITEBACK_DATA);
3524
3525 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3526 set_opt(sb, ERRORS_PANIC);
3527 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3528 set_opt(sb, ERRORS_CONT);
3529 else
3530 set_opt(sb, ERRORS_RO);
3531 /* block_validity enabled by default; disable with noblock_validity */
3532 set_opt(sb, BLOCK_VALIDITY);
3533 if (def_mount_opts & EXT4_DEFM_DISCARD)
3534 set_opt(sb, DISCARD);
3535
3536 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3537 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
3538 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3539 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3540 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3541
3542 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3543 set_opt(sb, BARRIER);
3544
3545 /*
3546 * enable delayed allocation by default
3547 * Use -o nodelalloc to turn it off
3548 */
3549 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
3550 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3551 set_opt(sb, DELALLOC);
3552
3553 /*
3554 * set default s_li_wait_mult for lazyinit, for the case there is
3555 * no mount option specified.
3556 */
3557 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3558
3559 if (sbi->s_es->s_mount_opts[0]) {
3560 char *s_mount_opts = kstrndup(sbi->s_es->s_mount_opts,
3561 sizeof(sbi->s_es->s_mount_opts),
3562 GFP_KERNEL);
3563 if (!s_mount_opts)
3564 goto failed_mount;
3565 if (!parse_options(s_mount_opts, sb, &journal_devnum,
3566 &journal_ioprio, 0)) {
3567 ext4_msg(sb, KERN_WARNING,
3568 "failed to parse options in superblock: %s",
3569 s_mount_opts);
3570 }
3571 kfree(s_mount_opts);
3572 }
3573 sbi->s_def_mount_opt = sbi->s_mount_opt;
3574 if (!parse_options((char *) data, sb, &journal_devnum,
3575 &journal_ioprio, 0))
3576 goto failed_mount;
3577
3578 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3579 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3580 "with data=journal disables delayed "
3581 "allocation and O_DIRECT support!\n");
3582 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3583 ext4_msg(sb, KERN_ERR, "can't mount with "
3584 "both data=journal and delalloc");
3585 goto failed_mount;
3586 }
3587 if (test_opt(sb, DIOREAD_NOLOCK)) {
3588 ext4_msg(sb, KERN_ERR, "can't mount with "
3589 "both data=journal and dioread_nolock");
3590 goto failed_mount;
3591 }
3592 if (test_opt(sb, DAX)) {
3593 ext4_msg(sb, KERN_ERR, "can't mount with "
3594 "both data=journal and dax");
3595 goto failed_mount;
3596 }
3597 if (ext4_has_feature_encrypt(sb)) {
3598 ext4_msg(sb, KERN_WARNING,
3599 "encrypted files will use data=ordered "
3600 "instead of data journaling mode");
3601 }
3602 if (test_opt(sb, DELALLOC))
3603 clear_opt(sb, DELALLOC);
3604 } else {
3605 sb->s_iflags |= SB_I_CGROUPWB;
3606 }
3607
3608 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3609 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3610
3611 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3612 (ext4_has_compat_features(sb) ||
3613 ext4_has_ro_compat_features(sb) ||
3614 ext4_has_incompat_features(sb)))
3615 ext4_msg(sb, KERN_WARNING,
3616 "feature flags set on rev 0 fs, "
3617 "running e2fsck is recommended");
3618
3619 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
3620 set_opt2(sb, HURD_COMPAT);
3621 if (ext4_has_feature_64bit(sb)) {
3622 ext4_msg(sb, KERN_ERR,
3623 "The Hurd can't support 64-bit file systems");
3624 goto failed_mount;
3625 }
3626
3627 /*
3628 * ea_inode feature uses l_i_version field which is not
3629 * available in HURD_COMPAT mode.
3630 */
3631 if (ext4_has_feature_ea_inode(sb)) {
3632 ext4_msg(sb, KERN_ERR,
3633 "ea_inode feature is not supported for Hurd");
3634 goto failed_mount;
3635 }
3636 }
3637
3638 if (IS_EXT2_SB(sb)) {
3639 if (ext2_feature_set_ok(sb))
3640 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3641 "using the ext4 subsystem");
3642 else {
3643 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3644 "to feature incompatibilities");
3645 goto failed_mount;
3646 }
3647 }
3648
3649 if (IS_EXT3_SB(sb)) {
3650 if (ext3_feature_set_ok(sb))
3651 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3652 "using the ext4 subsystem");
3653 else {
3654 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3655 "to feature incompatibilities");
3656 goto failed_mount;
3657 }
3658 }
3659
3660 /*
3661 * Check feature flags regardless of the revision level, since we
3662 * previously didn't change the revision level when setting the flags,
3663 * so there is a chance incompat flags are set on a rev 0 filesystem.
3664 */
3665 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3666 goto failed_mount;
3667
3668 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3669 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3670 blocksize > EXT4_MAX_BLOCK_SIZE) {
3671 ext4_msg(sb, KERN_ERR,
3672 "Unsupported filesystem blocksize %d (%d log_block_size)",
3673 blocksize, le32_to_cpu(es->s_log_block_size));
3674 goto failed_mount;
3675 }
3676 if (le32_to_cpu(es->s_log_block_size) >
3677 (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
3678 ext4_msg(sb, KERN_ERR,
3679 "Invalid log block size: %u",
3680 le32_to_cpu(es->s_log_block_size));
3681 goto failed_mount;
3682 }
3683
3684 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (blocksize / 4)) {
3685 ext4_msg(sb, KERN_ERR,
3686 "Number of reserved GDT blocks insanely large: %d",
3687 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks));
3688 goto failed_mount;
3689 }
3690
3691 if (sbi->s_mount_opt & EXT4_MOUNT_DAX) {
3692 err = bdev_dax_supported(sb, blocksize);
3693 if (err)
3694 goto failed_mount;
3695 }
3696
3697 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
3698 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
3699 es->s_encryption_level);
3700 goto failed_mount;
3701 }
3702
3703 if (sb->s_blocksize != blocksize) {
3704 /* Validate the filesystem blocksize */
3705 if (!sb_set_blocksize(sb, blocksize)) {
3706 ext4_msg(sb, KERN_ERR, "bad block size %d",
3707 blocksize);
3708 goto failed_mount;
3709 }
3710
3711 brelse(bh);
3712 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3713 offset = do_div(logical_sb_block, blocksize);
3714 bh = sb_bread_unmovable(sb, logical_sb_block);
3715 if (!bh) {
3716 ext4_msg(sb, KERN_ERR,
3717 "Can't read superblock on 2nd try");
3718 goto failed_mount;
3719 }
3720 es = (struct ext4_super_block *)(bh->b_data + offset);
3721 sbi->s_es = es;
3722 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3723 ext4_msg(sb, KERN_ERR,
3724 "Magic mismatch, very weird!");
3725 goto failed_mount;
3726 }
3727 }
3728
3729 has_huge_files = ext4_has_feature_huge_file(sb);
3730 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3731 has_huge_files);
3732 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3733
3734 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3735 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3736 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3737 } else {
3738 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3739 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3740 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3741 (!is_power_of_2(sbi->s_inode_size)) ||
3742 (sbi->s_inode_size > blocksize)) {
3743 ext4_msg(sb, KERN_ERR,
3744 "unsupported inode size: %d",
3745 sbi->s_inode_size);
3746 goto failed_mount;
3747 }
3748 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3749 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3750 }
3751
3752 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3753 if (ext4_has_feature_64bit(sb)) {
3754 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3755 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3756 !is_power_of_2(sbi->s_desc_size)) {
3757 ext4_msg(sb, KERN_ERR,
3758 "unsupported descriptor size %lu",
3759 sbi->s_desc_size);
3760 goto failed_mount;
3761 }
3762 } else
3763 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3764
3765 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3766 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3767
3768 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3769 if (sbi->s_inodes_per_block == 0)
3770 goto cantfind_ext4;
3771 if (sbi->s_inodes_per_group < sbi->s_inodes_per_block ||
3772 sbi->s_inodes_per_group > blocksize * 8) {
3773 ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n",
3774 sbi->s_blocks_per_group);
3775 goto failed_mount;
3776 }
3777 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3778 sbi->s_inodes_per_block;
3779 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3780 sbi->s_sbh = bh;
3781 sbi->s_mount_state = le16_to_cpu(es->s_state);
3782 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3783 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3784
3785 for (i = 0; i < 4; i++)
3786 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3787 sbi->s_def_hash_version = es->s_def_hash_version;
3788 if (ext4_has_feature_dir_index(sb)) {
3789 i = le32_to_cpu(es->s_flags);
3790 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3791 sbi->s_hash_unsigned = 3;
3792 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3793#ifdef __CHAR_UNSIGNED__
3794 if (!(sb->s_flags & MS_RDONLY))
3795 es->s_flags |=
3796 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3797 sbi->s_hash_unsigned = 3;
3798#else
3799 if (!(sb->s_flags & MS_RDONLY))
3800 es->s_flags |=
3801 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3802#endif
3803 }
3804 }
3805
3806 /* Handle clustersize */
3807 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3808 has_bigalloc = ext4_has_feature_bigalloc(sb);
3809 if (has_bigalloc) {
3810 if (clustersize < blocksize) {
3811 ext4_msg(sb, KERN_ERR,
3812 "cluster size (%d) smaller than "
3813 "block size (%d)", clustersize, blocksize);
3814 goto failed_mount;
3815 }
3816 if (le32_to_cpu(es->s_log_cluster_size) >
3817 (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
3818 ext4_msg(sb, KERN_ERR,
3819 "Invalid log cluster size: %u",
3820 le32_to_cpu(es->s_log_cluster_size));
3821 goto failed_mount;
3822 }
3823 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3824 le32_to_cpu(es->s_log_block_size);
3825 sbi->s_clusters_per_group =
3826 le32_to_cpu(es->s_clusters_per_group);
3827 if (sbi->s_clusters_per_group > blocksize * 8) {
3828 ext4_msg(sb, KERN_ERR,
3829 "#clusters per group too big: %lu",
3830 sbi->s_clusters_per_group);
3831 goto failed_mount;
3832 }
3833 if (sbi->s_blocks_per_group !=
3834 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3835 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3836 "clusters per group (%lu) inconsistent",
3837 sbi->s_blocks_per_group,
3838 sbi->s_clusters_per_group);
3839 goto failed_mount;
3840 }
3841 } else {
3842 if (clustersize != blocksize) {
3843 ext4_warning(sb, "fragment/cluster size (%d) != "
3844 "block size (%d)", clustersize,
3845 blocksize);
3846 clustersize = blocksize;
3847 }
3848 if (sbi->s_blocks_per_group > blocksize * 8) {
3849 ext4_msg(sb, KERN_ERR,
3850 "#blocks per group too big: %lu",
3851 sbi->s_blocks_per_group);
3852 goto failed_mount;
3853 }
3854 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3855 sbi->s_cluster_bits = 0;
3856 }
3857 sbi->s_cluster_ratio = clustersize / blocksize;
3858
3859 /* Do we have standard group size of clustersize * 8 blocks ? */
3860 if (sbi->s_blocks_per_group == clustersize << 3)
3861 set_opt2(sb, STD_GROUP_SIZE);
3862
3863 /*
3864 * Test whether we have more sectors than will fit in sector_t,
3865 * and whether the max offset is addressable by the page cache.
3866 */
3867 err = generic_check_addressable(sb->s_blocksize_bits,
3868 ext4_blocks_count(es));
3869 if (err) {
3870 ext4_msg(sb, KERN_ERR, "filesystem"
3871 " too large to mount safely on this system");
3872 if (sizeof(sector_t) < 8)
3873 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3874 goto failed_mount;
3875 }
3876
3877 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3878 goto cantfind_ext4;
3879
3880 /* check blocks count against device size */
3881 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3882 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3883 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3884 "exceeds size of device (%llu blocks)",
3885 ext4_blocks_count(es), blocks_count);
3886 goto failed_mount;
3887 }
3888
3889 /*
3890 * It makes no sense for the first data block to be beyond the end
3891 * of the filesystem.
3892 */
3893 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3894 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
3895 "block %u is beyond end of filesystem (%llu)",
3896 le32_to_cpu(es->s_first_data_block),
3897 ext4_blocks_count(es));
3898 goto failed_mount;
3899 }
3900 blocks_count = (ext4_blocks_count(es) -
3901 le32_to_cpu(es->s_first_data_block) +
3902 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3903 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3904 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3905 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3906 "(block count %llu, first data block %u, "
3907 "blocks per group %lu)", sbi->s_groups_count,
3908 ext4_blocks_count(es),
3909 le32_to_cpu(es->s_first_data_block),
3910 EXT4_BLOCKS_PER_GROUP(sb));
3911 goto failed_mount;
3912 }
3913 sbi->s_groups_count = blocks_count;
3914 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3915 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3916 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3917 EXT4_DESC_PER_BLOCK(sb);
3918 if (ext4_has_feature_meta_bg(sb)) {
3919 if (le32_to_cpu(es->s_first_meta_bg) > db_count) {
3920 ext4_msg(sb, KERN_WARNING,
3921 "first meta block group too large: %u "
3922 "(group descriptor block count %u)",
3923 le32_to_cpu(es->s_first_meta_bg), db_count);
3924 goto failed_mount;
3925 }
3926 }
3927 sbi->s_group_desc = kvmalloc(db_count *
3928 sizeof(struct buffer_head *),
3929 GFP_KERNEL);
3930 if (sbi->s_group_desc == NULL) {
3931 ext4_msg(sb, KERN_ERR, "not enough memory");
3932 ret = -ENOMEM;
3933 goto failed_mount;
3934 }
3935
3936 bgl_lock_init(sbi->s_blockgroup_lock);
3937
3938 /* Pre-read the descriptors into the buffer cache */
3939 for (i = 0; i < db_count; i++) {
3940 block = descriptor_loc(sb, logical_sb_block, i);
3941 sb_breadahead(sb, block);
3942 }
3943
3944 for (i = 0; i < db_count; i++) {
3945 block = descriptor_loc(sb, logical_sb_block, i);
3946 sbi->s_group_desc[i] = sb_bread_unmovable(sb, block);
3947 if (!sbi->s_group_desc[i]) {
3948 ext4_msg(sb, KERN_ERR,
3949 "can't read group descriptor %d", i);
3950 db_count = i;
3951 goto failed_mount2;
3952 }
3953 }
3954 if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) {
3955 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3956 ret = -EFSCORRUPTED;
3957 goto failed_mount2;
3958 }
3959
3960 sbi->s_gdb_count = db_count;
3961 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3962 spin_lock_init(&sbi->s_next_gen_lock);
3963
3964 setup_timer(&sbi->s_err_report, print_daily_error_info,
3965 (unsigned long) sb);
3966
3967 /* Register extent status tree shrinker */
3968 if (ext4_es_register_shrinker(sbi))
3969 goto failed_mount3;
3970
3971 sbi->s_stripe = ext4_get_stripe_size(sbi);
3972 sbi->s_extent_max_zeroout_kb = 32;
3973
3974 /*
3975 * set up enough so that it can read an inode
3976 */
3977 sb->s_op = &ext4_sops;
3978 sb->s_export_op = &ext4_export_ops;
3979 sb->s_xattr = ext4_xattr_handlers;
3980 sb->s_cop = &ext4_cryptops;
3981#ifdef CONFIG_QUOTA
3982 sb->dq_op = &ext4_quota_operations;
3983 if (ext4_has_feature_quota(sb))
3984 sb->s_qcop = &dquot_quotactl_sysfile_ops;
3985 else
3986 sb->s_qcop = &ext4_qctl_operations;
3987 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
3988#endif
3989 memcpy(&sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3990
3991 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3992 mutex_init(&sbi->s_orphan_lock);
3993
3994 sb->s_root = NULL;
3995
3996 needs_recovery = (es->s_last_orphan != 0 ||
3997 ext4_has_feature_journal_needs_recovery(sb));
3998
3999 if (ext4_has_feature_mmp(sb) && !(sb->s_flags & MS_RDONLY))
4000 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
4001 goto failed_mount3a;
4002
4003 /*
4004 * The first inode we look at is the journal inode. Don't try
4005 * root first: it may be modified in the journal!
4006 */
4007 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
4008 err = ext4_load_journal(sb, es, journal_devnum);
4009 if (err)
4010 goto failed_mount3a;
4011 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
4012 ext4_has_feature_journal_needs_recovery(sb)) {
4013 ext4_msg(sb, KERN_ERR, "required journal recovery "
4014 "suppressed and not mounted read-only");
4015 goto failed_mount_wq;
4016 } else {
4017 /* Nojournal mode, all journal mount options are illegal */
4018 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) {
4019 ext4_msg(sb, KERN_ERR, "can't mount with "
4020 "journal_checksum, fs mounted w/o journal");
4021 goto failed_mount_wq;
4022 }
4023 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4024 ext4_msg(sb, KERN_ERR, "can't mount with "
4025 "journal_async_commit, fs mounted w/o journal");
4026 goto failed_mount_wq;
4027 }
4028 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
4029 ext4_msg(sb, KERN_ERR, "can't mount with "
4030 "commit=%lu, fs mounted w/o journal",
4031 sbi->s_commit_interval / HZ);
4032 goto failed_mount_wq;
4033 }
4034 if (EXT4_MOUNT_DATA_FLAGS &
4035 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
4036 ext4_msg(sb, KERN_ERR, "can't mount with "
4037 "data=, fs mounted w/o journal");
4038 goto failed_mount_wq;
4039 }
4040 sbi->s_def_mount_opt &= EXT4_MOUNT_JOURNAL_CHECKSUM;
4041 clear_opt(sb, JOURNAL_CHECKSUM);
4042 clear_opt(sb, DATA_FLAGS);
4043 sbi->s_journal = NULL;
4044 needs_recovery = 0;
4045 goto no_journal;
4046 }
4047
4048 if (ext4_has_feature_64bit(sb) &&
4049 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4050 JBD2_FEATURE_INCOMPAT_64BIT)) {
4051 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
4052 goto failed_mount_wq;
4053 }
4054
4055 if (!set_journal_csum_feature_set(sb)) {
4056 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
4057 "feature set");
4058 goto failed_mount_wq;
4059 }
4060
4061 /* We have now updated the journal if required, so we can
4062 * validate the data journaling mode. */
4063 switch (test_opt(sb, DATA_FLAGS)) {
4064 case 0:
4065 /* No mode set, assume a default based on the journal
4066 * capabilities: ORDERED_DATA if the journal can
4067 * cope, else JOURNAL_DATA
4068 */
4069 if (jbd2_journal_check_available_features
4070 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
4071 set_opt(sb, ORDERED_DATA);
4072 else
4073 set_opt(sb, JOURNAL_DATA);
4074 break;
4075
4076 case EXT4_MOUNT_ORDERED_DATA:
4077 case EXT4_MOUNT_WRITEBACK_DATA:
4078 if (!jbd2_journal_check_available_features
4079 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4080 ext4_msg(sb, KERN_ERR, "Journal does not support "
4081 "requested data journaling mode");
4082 goto failed_mount_wq;
4083 }
4084 default:
4085 break;
4086 }
4087
4088 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
4089 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4090 ext4_msg(sb, KERN_ERR, "can't mount with "
4091 "journal_async_commit in data=ordered mode");
4092 goto failed_mount_wq;
4093 }
4094
4095 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4096
4097 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
4098
4099no_journal:
4100 if (!test_opt(sb, NO_MBCACHE)) {
4101 sbi->s_ea_block_cache = ext4_xattr_create_cache();
4102 if (!sbi->s_ea_block_cache) {
4103 ext4_msg(sb, KERN_ERR,
4104 "Failed to create ea_block_cache");
4105 goto failed_mount_wq;
4106 }
4107
4108 if (ext4_has_feature_ea_inode(sb)) {
4109 sbi->s_ea_inode_cache = ext4_xattr_create_cache();
4110 if (!sbi->s_ea_inode_cache) {
4111 ext4_msg(sb, KERN_ERR,
4112 "Failed to create ea_inode_cache");
4113 goto failed_mount_wq;
4114 }
4115 }
4116 }
4117
4118 if ((DUMMY_ENCRYPTION_ENABLED(sbi) || ext4_has_feature_encrypt(sb)) &&
4119 (blocksize != PAGE_SIZE)) {
4120 ext4_msg(sb, KERN_ERR,
4121 "Unsupported blocksize for fs encryption");
4122 goto failed_mount_wq;
4123 }
4124
4125 if (DUMMY_ENCRYPTION_ENABLED(sbi) && !(sb->s_flags & MS_RDONLY) &&
4126 !ext4_has_feature_encrypt(sb)) {
4127 ext4_set_feature_encrypt(sb);
4128 ext4_commit_super(sb, 1);
4129 }
4130
4131 /*
4132 * Get the # of file system overhead blocks from the
4133 * superblock if present.
4134 */
4135 if (es->s_overhead_clusters)
4136 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4137 else {
4138 err = ext4_calculate_overhead(sb);
4139 if (err)
4140 goto failed_mount_wq;
4141 }
4142
4143 /*
4144 * The maximum number of concurrent works can be high and
4145 * concurrency isn't really necessary. Limit it to 1.
4146 */
4147 EXT4_SB(sb)->rsv_conversion_wq =
4148 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4149 if (!EXT4_SB(sb)->rsv_conversion_wq) {
4150 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
4151 ret = -ENOMEM;
4152 goto failed_mount4;
4153 }
4154
4155 /*
4156 * The jbd2_journal_load will have done any necessary log recovery,
4157 * so we can safely mount the rest of the filesystem now.
4158 */
4159
4160 root = ext4_iget(sb, EXT4_ROOT_INO);
4161 if (IS_ERR(root)) {
4162 ext4_msg(sb, KERN_ERR, "get root inode failed");
4163 ret = PTR_ERR(root);
4164 root = NULL;
4165 goto failed_mount4;
4166 }
4167 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
4168 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
4169 iput(root);
4170 goto failed_mount4;
4171 }
4172 sb->s_root = d_make_root(root);
4173 if (!sb->s_root) {
4174 ext4_msg(sb, KERN_ERR, "get root dentry failed");
4175 ret = -ENOMEM;
4176 goto failed_mount4;
4177 }
4178
4179 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
4180 sb->s_flags |= MS_RDONLY;
4181
4182 /* determine the minimum size of new large inodes, if present */
4183 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE &&
4184 sbi->s_want_extra_isize == 0) {
4185 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4186 EXT4_GOOD_OLD_INODE_SIZE;
4187 if (ext4_has_feature_extra_isize(sb)) {
4188 if (sbi->s_want_extra_isize <
4189 le16_to_cpu(es->s_want_extra_isize))
4190 sbi->s_want_extra_isize =
4191 le16_to_cpu(es->s_want_extra_isize);
4192 if (sbi->s_want_extra_isize <
4193 le16_to_cpu(es->s_min_extra_isize))
4194 sbi->s_want_extra_isize =
4195 le16_to_cpu(es->s_min_extra_isize);
4196 }
4197 }
4198 /* Check if enough inode space is available */
4199 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
4200 sbi->s_inode_size) {
4201 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4202 EXT4_GOOD_OLD_INODE_SIZE;
4203 ext4_msg(sb, KERN_INFO, "required extra inode space not"
4204 "available");
4205 }
4206
4207 ext4_set_resv_clusters(sb);
4208
4209 err = ext4_setup_system_zone(sb);
4210 if (err) {
4211 ext4_msg(sb, KERN_ERR, "failed to initialize system "
4212 "zone (%d)", err);
4213 goto failed_mount4a;
4214 }
4215
4216 ext4_ext_init(sb);
4217 err = ext4_mb_init(sb);
4218 if (err) {
4219 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4220 err);
4221 goto failed_mount5;
4222 }
4223
4224 block = ext4_count_free_clusters(sb);
4225 ext4_free_blocks_count_set(sbi->s_es,
4226 EXT4_C2B(sbi, block));
4227 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
4228 GFP_KERNEL);
4229 if (!err) {
4230 unsigned long freei = ext4_count_free_inodes(sb);
4231 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
4232 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
4233 GFP_KERNEL);
4234 }
4235 if (!err)
4236 err = percpu_counter_init(&sbi->s_dirs_counter,
4237 ext4_count_dirs(sb), GFP_KERNEL);
4238 if (!err)
4239 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
4240 GFP_KERNEL);
4241 if (!err)
4242 err = percpu_init_rwsem(&sbi->s_journal_flag_rwsem);
4243
4244 if (err) {
4245 ext4_msg(sb, KERN_ERR, "insufficient memory");
4246 goto failed_mount6;
4247 }
4248
4249 if (ext4_has_feature_flex_bg(sb))
4250 if (!ext4_fill_flex_info(sb)) {
4251 ext4_msg(sb, KERN_ERR,
4252 "unable to initialize "
4253 "flex_bg meta info!");
4254 goto failed_mount6;
4255 }
4256
4257 err = ext4_register_li_request(sb, first_not_zeroed);
4258 if (err)
4259 goto failed_mount6;
4260
4261 err = ext4_register_sysfs(sb);
4262 if (err)
4263 goto failed_mount7;
4264
4265#ifdef CONFIG_QUOTA
4266 /* Enable quota usage during mount. */
4267 if (ext4_has_feature_quota(sb) && !(sb->s_flags & MS_RDONLY)) {
4268 err = ext4_enable_quotas(sb);
4269 if (err)
4270 goto failed_mount8;
4271 }
4272#endif /* CONFIG_QUOTA */
4273
4274 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4275 ext4_orphan_cleanup(sb, es);
4276 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
4277 if (needs_recovery) {
4278 ext4_msg(sb, KERN_INFO, "recovery complete");
4279 ext4_mark_recovery_complete(sb, es);
4280 }
4281 if (EXT4_SB(sb)->s_journal) {
4282 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4283 descr = " journalled data mode";
4284 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4285 descr = " ordered data mode";
4286 else
4287 descr = " writeback data mode";
4288 } else
4289 descr = "out journal";
4290
4291 if (test_opt(sb, DISCARD)) {
4292 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4293 if (!blk_queue_discard(q))
4294 ext4_msg(sb, KERN_WARNING,
4295 "mounting with \"discard\" option, but "
4296 "the device does not support discard");
4297 }
4298
4299 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
4300 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
4301 "Opts: %.*s%s%s", descr,
4302 (int) sizeof(sbi->s_es->s_mount_opts),
4303 sbi->s_es->s_mount_opts,
4304 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
4305
4306 if (es->s_error_count)
4307 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
4308
4309 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4310 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
4311 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
4312 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4313
4314 kfree(orig_data);
4315 return 0;
4316
4317cantfind_ext4:
4318 if (!silent)
4319 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
4320 goto failed_mount;
4321
4322#ifdef CONFIG_QUOTA
4323failed_mount8:
4324 ext4_unregister_sysfs(sb);
4325#endif
4326failed_mount7:
4327 ext4_unregister_li_request(sb);
4328failed_mount6:
4329 ext4_mb_release(sb);
4330 if (sbi->s_flex_groups)
4331 kvfree(sbi->s_flex_groups);
4332 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4333 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4334 percpu_counter_destroy(&sbi->s_dirs_counter);
4335 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
4336failed_mount5:
4337 ext4_ext_release(sb);
4338 ext4_release_system_zone(sb);
4339failed_mount4a:
4340 dput(sb->s_root);
4341 sb->s_root = NULL;
4342failed_mount4:
4343 ext4_msg(sb, KERN_ERR, "mount failed");
4344 if (EXT4_SB(sb)->rsv_conversion_wq)
4345 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4346failed_mount_wq:
4347 if (sbi->s_ea_inode_cache) {
4348 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
4349 sbi->s_ea_inode_cache = NULL;
4350 }
4351 if (sbi->s_ea_block_cache) {
4352 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
4353 sbi->s_ea_block_cache = NULL;
4354 }
4355 if (sbi->s_journal) {
4356 jbd2_journal_destroy(sbi->s_journal);
4357 sbi->s_journal = NULL;
4358 }
4359failed_mount3a:
4360 ext4_es_unregister_shrinker(sbi);
4361failed_mount3:
4362 del_timer_sync(&sbi->s_err_report);
4363 if (sbi->s_mmp_tsk)
4364 kthread_stop(sbi->s_mmp_tsk);
4365failed_mount2:
4366 for (i = 0; i < db_count; i++)
4367 brelse(sbi->s_group_desc[i]);
4368 kvfree(sbi->s_group_desc);
4369failed_mount:
4370 if (sbi->s_chksum_driver)
4371 crypto_free_shash(sbi->s_chksum_driver);
4372#ifdef CONFIG_QUOTA
4373 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4374 kfree(sbi->s_qf_names[i]);
4375#endif
4376 ext4_blkdev_remove(sbi);
4377 brelse(bh);
4378out_fail:
4379 sb->s_fs_info = NULL;
4380 kfree(sbi->s_blockgroup_lock);
4381out_free_base:
4382 kfree(sbi);
4383 kfree(orig_data);
4384 fs_put_dax(dax_dev);
4385 return err ? err : ret;
4386}
4387
4388/*
4389 * Setup any per-fs journal parameters now. We'll do this both on
4390 * initial mount, once the journal has been initialised but before we've
4391 * done any recovery; and again on any subsequent remount.
4392 */
4393static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
4394{
4395 struct ext4_sb_info *sbi = EXT4_SB(sb);
4396
4397 journal->j_commit_interval = sbi->s_commit_interval;
4398 journal->j_min_batch_time = sbi->s_min_batch_time;
4399 journal->j_max_batch_time = sbi->s_max_batch_time;
4400
4401 write_lock(&journal->j_state_lock);
4402 if (test_opt(sb, BARRIER))
4403 journal->j_flags |= JBD2_BARRIER;
4404 else
4405 journal->j_flags &= ~JBD2_BARRIER;
4406 if (test_opt(sb, DATA_ERR_ABORT))
4407 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4408 else
4409 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
4410 write_unlock(&journal->j_state_lock);
4411}
4412
4413static struct inode *ext4_get_journal_inode(struct super_block *sb,
4414 unsigned int journal_inum)
4415{
4416 struct inode *journal_inode;
4417
4418 /*
4419 * Test for the existence of a valid inode on disk. Bad things
4420 * happen if we iget() an unused inode, as the subsequent iput()
4421 * will try to delete it.
4422 */
4423 journal_inode = ext4_iget(sb, journal_inum);
4424 if (IS_ERR(journal_inode)) {
4425 ext4_msg(sb, KERN_ERR, "no journal found");
4426 return NULL;
4427 }
4428 if (!journal_inode->i_nlink) {
4429 make_bad_inode(journal_inode);
4430 iput(journal_inode);
4431 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
4432 return NULL;
4433 }
4434
4435 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4436 journal_inode, journal_inode->i_size);
4437 if (!S_ISREG(journal_inode->i_mode)) {
4438 ext4_msg(sb, KERN_ERR, "invalid journal inode");
4439 iput(journal_inode);
4440 return NULL;
4441 }
4442 return journal_inode;
4443}
4444
4445static journal_t *ext4_get_journal(struct super_block *sb,
4446 unsigned int journal_inum)
4447{
4448 struct inode *journal_inode;
4449 journal_t *journal;
4450
4451 BUG_ON(!ext4_has_feature_journal(sb));
4452
4453 journal_inode = ext4_get_journal_inode(sb, journal_inum);
4454 if (!journal_inode)
4455 return NULL;
4456
4457 journal = jbd2_journal_init_inode(journal_inode);
4458 if (!journal) {
4459 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
4460 iput(journal_inode);
4461 return NULL;
4462 }
4463 journal->j_private = sb;
4464 ext4_init_journal_params(sb, journal);
4465 return journal;
4466}
4467
4468static journal_t *ext4_get_dev_journal(struct super_block *sb,
4469 dev_t j_dev)
4470{
4471 struct buffer_head *bh;
4472 journal_t *journal;
4473 ext4_fsblk_t start;
4474 ext4_fsblk_t len;
4475 int hblock, blocksize;
4476 ext4_fsblk_t sb_block;
4477 unsigned long offset;
4478 struct ext4_super_block *es;
4479 struct block_device *bdev;
4480
4481 BUG_ON(!ext4_has_feature_journal(sb));
4482
4483 bdev = ext4_blkdev_get(j_dev, sb);
4484 if (bdev == NULL)
4485 return NULL;
4486
4487 blocksize = sb->s_blocksize;
4488 hblock = bdev_logical_block_size(bdev);
4489 if (blocksize < hblock) {
4490 ext4_msg(sb, KERN_ERR,
4491 "blocksize too small for journal device");
4492 goto out_bdev;
4493 }
4494
4495 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4496 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
4497 set_blocksize(bdev, blocksize);
4498 if (!(bh = __bread(bdev, sb_block, blocksize))) {
4499 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4500 "external journal");
4501 goto out_bdev;
4502 }
4503
4504 es = (struct ext4_super_block *) (bh->b_data + offset);
4505 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
4506 !(le32_to_cpu(es->s_feature_incompat) &
4507 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4508 ext4_msg(sb, KERN_ERR, "external journal has "
4509 "bad superblock");
4510 brelse(bh);
4511 goto out_bdev;
4512 }
4513
4514 if ((le32_to_cpu(es->s_feature_ro_compat) &
4515 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
4516 es->s_checksum != ext4_superblock_csum(sb, es)) {
4517 ext4_msg(sb, KERN_ERR, "external journal has "
4518 "corrupt superblock");
4519 brelse(bh);
4520 goto out_bdev;
4521 }
4522
4523 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4524 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4525 brelse(bh);
4526 goto out_bdev;
4527 }
4528
4529 len = ext4_blocks_count(es);
4530 start = sb_block + 1;
4531 brelse(bh); /* we're done with the superblock */
4532
4533 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4534 start, len, blocksize);
4535 if (!journal) {
4536 ext4_msg(sb, KERN_ERR, "failed to create device journal");
4537 goto out_bdev;
4538 }
4539 journal->j_private = sb;
4540 ll_rw_block(REQ_OP_READ, REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
4541 wait_on_buffer(journal->j_sb_buffer);
4542 if (!buffer_uptodate(journal->j_sb_buffer)) {
4543 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4544 goto out_journal;
4545 }
4546 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4547 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4548 "user (unsupported) - %d",
4549 be32_to_cpu(journal->j_superblock->s_nr_users));
4550 goto out_journal;
4551 }
4552 EXT4_SB(sb)->journal_bdev = bdev;
4553 ext4_init_journal_params(sb, journal);
4554 return journal;
4555
4556out_journal:
4557 jbd2_journal_destroy(journal);
4558out_bdev:
4559 ext4_blkdev_put(bdev);
4560 return NULL;
4561}
4562
4563static int ext4_load_journal(struct super_block *sb,
4564 struct ext4_super_block *es,
4565 unsigned long journal_devnum)
4566{
4567 journal_t *journal;
4568 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4569 dev_t journal_dev;
4570 int err = 0;
4571 int really_read_only;
4572
4573 BUG_ON(!ext4_has_feature_journal(sb));
4574
4575 if (journal_devnum &&
4576 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4577 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4578 "numbers have changed");
4579 journal_dev = new_decode_dev(journal_devnum);
4580 } else
4581 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4582
4583 really_read_only = bdev_read_only(sb->s_bdev);
4584
4585 /*
4586 * Are we loading a blank journal or performing recovery after a
4587 * crash? For recovery, we need to check in advance whether we
4588 * can get read-write access to the device.
4589 */
4590 if (ext4_has_feature_journal_needs_recovery(sb)) {
4591 if (sb->s_flags & MS_RDONLY) {
4592 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4593 "required on readonly filesystem");
4594 if (really_read_only) {
4595 ext4_msg(sb, KERN_ERR, "write access "
4596 "unavailable, cannot proceed");
4597 return -EROFS;
4598 }
4599 ext4_msg(sb, KERN_INFO, "write access will "
4600 "be enabled during recovery");
4601 }
4602 }
4603
4604 if (journal_inum && journal_dev) {
4605 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4606 "and inode journals!");
4607 return -EINVAL;
4608 }
4609
4610 if (journal_inum) {
4611 if (!(journal = ext4_get_journal(sb, journal_inum)))
4612 return -EINVAL;
4613 } else {
4614 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4615 return -EINVAL;
4616 }
4617
4618 if (!(journal->j_flags & JBD2_BARRIER))
4619 ext4_msg(sb, KERN_INFO, "barriers disabled");
4620
4621 if (!ext4_has_feature_journal_needs_recovery(sb))
4622 err = jbd2_journal_wipe(journal, !really_read_only);
4623 if (!err) {
4624 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4625 if (save)
4626 memcpy(save, ((char *) es) +
4627 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4628 err = jbd2_journal_load(journal);
4629 if (save)
4630 memcpy(((char *) es) + EXT4_S_ERR_START,
4631 save, EXT4_S_ERR_LEN);
4632 kfree(save);
4633 }
4634
4635 if (err) {
4636 ext4_msg(sb, KERN_ERR, "error loading journal");
4637 jbd2_journal_destroy(journal);
4638 return err;
4639 }
4640
4641 EXT4_SB(sb)->s_journal = journal;
4642 ext4_clear_journal_err(sb, es);
4643
4644 if (!really_read_only && journal_devnum &&
4645 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4646 es->s_journal_dev = cpu_to_le32(journal_devnum);
4647
4648 /* Make sure we flush the recovery flag to disk. */
4649 ext4_commit_super(sb, 1);
4650 }
4651
4652 return 0;
4653}
4654
4655static int ext4_commit_super(struct super_block *sb, int sync)
4656{
4657 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4658 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4659 int error = 0;
4660
4661 if (!sbh || block_device_ejected(sb))
4662 return error;
4663 /*
4664 * If the file system is mounted read-only, don't update the
4665 * superblock write time. This avoids updating the superblock
4666 * write time when we are mounting the root file system
4667 * read/only but we need to replay the journal; at that point,
4668 * for people who are east of GMT and who make their clock
4669 * tick in localtime for Windows bug-for-bug compatibility,
4670 * the clock is set in the future, and this will cause e2fsck
4671 * to complain and force a full file system check.
4672 */
4673 if (!(sb->s_flags & MS_RDONLY))
4674 es->s_wtime = cpu_to_le32(get_seconds());
4675 if (sb->s_bdev->bd_part)
4676 es->s_kbytes_written =
4677 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4678 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4679 EXT4_SB(sb)->s_sectors_written_start) >> 1));
4680 else
4681 es->s_kbytes_written =
4682 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4683 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
4684 ext4_free_blocks_count_set(es,
4685 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4686 &EXT4_SB(sb)->s_freeclusters_counter)));
4687 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
4688 es->s_free_inodes_count =
4689 cpu_to_le32(percpu_counter_sum_positive(
4690 &EXT4_SB(sb)->s_freeinodes_counter));
4691 BUFFER_TRACE(sbh, "marking dirty");
4692 ext4_superblock_csum_set(sb);
4693 if (sync)
4694 lock_buffer(sbh);
4695 if (buffer_write_io_error(sbh)) {
4696 /*
4697 * Oh, dear. A previous attempt to write the
4698 * superblock failed. This could happen because the
4699 * USB device was yanked out. Or it could happen to
4700 * be a transient write error and maybe the block will
4701 * be remapped. Nothing we can do but to retry the
4702 * write and hope for the best.
4703 */
4704 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4705 "superblock detected");
4706 clear_buffer_write_io_error(sbh);
4707 set_buffer_uptodate(sbh);
4708 }
4709 mark_buffer_dirty(sbh);
4710 if (sync) {
4711 unlock_buffer(sbh);
4712 error = __sync_dirty_buffer(sbh,
4713 REQ_SYNC | (test_opt(sb, BARRIER) ? REQ_FUA : 0));
4714 if (error)
4715 return error;
4716
4717 error = buffer_write_io_error(sbh);
4718 if (error) {
4719 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4720 "superblock");
4721 clear_buffer_write_io_error(sbh);
4722 set_buffer_uptodate(sbh);
4723 }
4724 }
4725 return error;
4726}
4727
4728/*
4729 * Have we just finished recovery? If so, and if we are mounting (or
4730 * remounting) the filesystem readonly, then we will end up with a
4731 * consistent fs on disk. Record that fact.
4732 */
4733static void ext4_mark_recovery_complete(struct super_block *sb,
4734 struct ext4_super_block *es)
4735{
4736 journal_t *journal = EXT4_SB(sb)->s_journal;
4737
4738 if (!ext4_has_feature_journal(sb)) {
4739 BUG_ON(journal != NULL);
4740 return;
4741 }
4742 jbd2_journal_lock_updates(journal);
4743 if (jbd2_journal_flush(journal) < 0)
4744 goto out;
4745
4746 if (ext4_has_feature_journal_needs_recovery(sb) &&
4747 sb->s_flags & MS_RDONLY) {
4748 ext4_clear_feature_journal_needs_recovery(sb);
4749 ext4_commit_super(sb, 1);
4750 }
4751
4752out:
4753 jbd2_journal_unlock_updates(journal);
4754}
4755
4756/*
4757 * If we are mounting (or read-write remounting) a filesystem whose journal
4758 * has recorded an error from a previous lifetime, move that error to the
4759 * main filesystem now.
4760 */
4761static void ext4_clear_journal_err(struct super_block *sb,
4762 struct ext4_super_block *es)
4763{
4764 journal_t *journal;
4765 int j_errno;
4766 const char *errstr;
4767
4768 BUG_ON(!ext4_has_feature_journal(sb));
4769
4770 journal = EXT4_SB(sb)->s_journal;
4771
4772 /*
4773 * Now check for any error status which may have been recorded in the
4774 * journal by a prior ext4_error() or ext4_abort()
4775 */
4776
4777 j_errno = jbd2_journal_errno(journal);
4778 if (j_errno) {
4779 char nbuf[16];
4780
4781 errstr = ext4_decode_error(sb, j_errno, nbuf);
4782 ext4_warning(sb, "Filesystem error recorded "
4783 "from previous mount: %s", errstr);
4784 ext4_warning(sb, "Marking fs in need of filesystem check.");
4785
4786 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4787 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4788 ext4_commit_super(sb, 1);
4789
4790 jbd2_journal_clear_err(journal);
4791 jbd2_journal_update_sb_errno(journal);
4792 }
4793}
4794
4795/*
4796 * Force the running and committing transactions to commit,
4797 * and wait on the commit.
4798 */
4799int ext4_force_commit(struct super_block *sb)
4800{
4801 journal_t *journal;
4802
4803 if (sb->s_flags & MS_RDONLY)
4804 return 0;
4805
4806 journal = EXT4_SB(sb)->s_journal;
4807 return ext4_journal_force_commit(journal);
4808}
4809
4810static int ext4_sync_fs(struct super_block *sb, int wait)
4811{
4812 int ret = 0;
4813 tid_t target;
4814 bool needs_barrier = false;
4815 struct ext4_sb_info *sbi = EXT4_SB(sb);
4816
4817 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
4818 return 0;
4819
4820 trace_ext4_sync_fs(sb, wait);
4821 flush_workqueue(sbi->rsv_conversion_wq);
4822 /*
4823 * Writeback quota in non-journalled quota case - journalled quota has
4824 * no dirty dquots
4825 */
4826 dquot_writeback_dquots(sb, -1);
4827 /*
4828 * Data writeback is possible w/o journal transaction, so barrier must
4829 * being sent at the end of the function. But we can skip it if
4830 * transaction_commit will do it for us.
4831 */
4832 if (sbi->s_journal) {
4833 target = jbd2_get_latest_transaction(sbi->s_journal);
4834 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
4835 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
4836 needs_barrier = true;
4837
4838 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4839 if (wait)
4840 ret = jbd2_log_wait_commit(sbi->s_journal,
4841 target);
4842 }
4843 } else if (wait && test_opt(sb, BARRIER))
4844 needs_barrier = true;
4845 if (needs_barrier) {
4846 int err;
4847 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4848 if (!ret)
4849 ret = err;
4850 }
4851
4852 return ret;
4853}
4854
4855/*
4856 * LVM calls this function before a (read-only) snapshot is created. This
4857 * gives us a chance to flush the journal completely and mark the fs clean.
4858 *
4859 * Note that only this function cannot bring a filesystem to be in a clean
4860 * state independently. It relies on upper layer to stop all data & metadata
4861 * modifications.
4862 */
4863static int ext4_freeze(struct super_block *sb)
4864{
4865 int error = 0;
4866 journal_t *journal;
4867
4868 if (sb->s_flags & MS_RDONLY)
4869 return 0;
4870
4871 journal = EXT4_SB(sb)->s_journal;
4872
4873 if (journal) {
4874 /* Now we set up the journal barrier. */
4875 jbd2_journal_lock_updates(journal);
4876
4877 /*
4878 * Don't clear the needs_recovery flag if we failed to
4879 * flush the journal.
4880 */
4881 error = jbd2_journal_flush(journal);
4882 if (error < 0)
4883 goto out;
4884
4885 /* Journal blocked and flushed, clear needs_recovery flag. */
4886 ext4_clear_feature_journal_needs_recovery(sb);
4887 }
4888
4889 error = ext4_commit_super(sb, 1);
4890out:
4891 if (journal)
4892 /* we rely on upper layer to stop further updates */
4893 jbd2_journal_unlock_updates(journal);
4894 return error;
4895}
4896
4897/*
4898 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4899 * flag here, even though the filesystem is not technically dirty yet.
4900 */
4901static int ext4_unfreeze(struct super_block *sb)
4902{
4903 if ((sb->s_flags & MS_RDONLY) || ext4_forced_shutdown(EXT4_SB(sb)))
4904 return 0;
4905
4906 if (EXT4_SB(sb)->s_journal) {
4907 /* Reset the needs_recovery flag before the fs is unlocked. */
4908 ext4_set_feature_journal_needs_recovery(sb);
4909 }
4910
4911 ext4_commit_super(sb, 1);
4912 return 0;
4913}
4914
4915/*
4916 * Structure to save mount options for ext4_remount's benefit
4917 */
4918struct ext4_mount_options {
4919 unsigned long s_mount_opt;
4920 unsigned long s_mount_opt2;
4921 kuid_t s_resuid;
4922 kgid_t s_resgid;
4923 unsigned long s_commit_interval;
4924 u32 s_min_batch_time, s_max_batch_time;
4925#ifdef CONFIG_QUOTA
4926 int s_jquota_fmt;
4927 char *s_qf_names[EXT4_MAXQUOTAS];
4928#endif
4929};
4930
4931static int ext4_remount(struct super_block *sb, int *flags, char *data)
4932{
4933 struct ext4_super_block *es;
4934 struct ext4_sb_info *sbi = EXT4_SB(sb);
4935 unsigned long old_sb_flags;
4936 struct ext4_mount_options old_opts;
4937 int enable_quota = 0;
4938 ext4_group_t g;
4939 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4940 int err = 0;
4941#ifdef CONFIG_QUOTA
4942 int i, j;
4943#endif
4944 char *orig_data = kstrdup(data, GFP_KERNEL);
4945
4946 /* Store the original options */
4947 old_sb_flags = sb->s_flags;
4948 old_opts.s_mount_opt = sbi->s_mount_opt;
4949 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4950 old_opts.s_resuid = sbi->s_resuid;
4951 old_opts.s_resgid = sbi->s_resgid;
4952 old_opts.s_commit_interval = sbi->s_commit_interval;
4953 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4954 old_opts.s_max_batch_time = sbi->s_max_batch_time;
4955#ifdef CONFIG_QUOTA
4956 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4957 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4958 if (sbi->s_qf_names[i]) {
4959 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4960 GFP_KERNEL);
4961 if (!old_opts.s_qf_names[i]) {
4962 for (j = 0; j < i; j++)
4963 kfree(old_opts.s_qf_names[j]);
4964 kfree(orig_data);
4965 return -ENOMEM;
4966 }
4967 } else
4968 old_opts.s_qf_names[i] = NULL;
4969#endif
4970 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4971 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4972
4973 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
4974 err = -EINVAL;
4975 goto restore_opts;
4976 }
4977
4978 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
4979 test_opt(sb, JOURNAL_CHECKSUM)) {
4980 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
4981 "during remount not supported; ignoring");
4982 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
4983 }
4984
4985 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4986 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4987 ext4_msg(sb, KERN_ERR, "can't mount with "
4988 "both data=journal and delalloc");
4989 err = -EINVAL;
4990 goto restore_opts;
4991 }
4992 if (test_opt(sb, DIOREAD_NOLOCK)) {
4993 ext4_msg(sb, KERN_ERR, "can't mount with "
4994 "both data=journal and dioread_nolock");
4995 err = -EINVAL;
4996 goto restore_opts;
4997 }
4998 if (test_opt(sb, DAX)) {
4999 ext4_msg(sb, KERN_ERR, "can't mount with "
5000 "both data=journal and dax");
5001 err = -EINVAL;
5002 goto restore_opts;
5003 }
5004 } else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) {
5005 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
5006 ext4_msg(sb, KERN_ERR, "can't mount with "
5007 "journal_async_commit in data=ordered mode");
5008 err = -EINVAL;
5009 goto restore_opts;
5010 }
5011 }
5012
5013 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_NO_MBCACHE) {
5014 ext4_msg(sb, KERN_ERR, "can't enable nombcache during remount");
5015 err = -EINVAL;
5016 goto restore_opts;
5017 }
5018
5019 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_DAX) {
5020 ext4_msg(sb, KERN_WARNING, "warning: refusing change of "
5021 "dax flag with busy inodes while remounting");
5022 sbi->s_mount_opt ^= EXT4_MOUNT_DAX;
5023 }
5024
5025 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
5026 ext4_abort(sb, "Abort forced by user");
5027
5028 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
5029 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
5030
5031 es = sbi->s_es;
5032
5033 if (sbi->s_journal) {
5034 ext4_init_journal_params(sb, sbi->s_journal);
5035 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
5036 }
5037
5038 if (*flags & MS_LAZYTIME)
5039 sb->s_flags |= MS_LAZYTIME;
5040
5041 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
5042 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
5043 err = -EROFS;
5044 goto restore_opts;
5045 }
5046
5047 if (*flags & MS_RDONLY) {
5048 err = sync_filesystem(sb);
5049 if (err < 0)
5050 goto restore_opts;
5051 err = dquot_suspend(sb, -1);
5052 if (err < 0)
5053 goto restore_opts;
5054
5055 /*
5056 * First of all, the unconditional stuff we have to do
5057 * to disable replay of the journal when we next remount
5058 */
5059 sb->s_flags |= MS_RDONLY;
5060
5061 /*
5062 * OK, test if we are remounting a valid rw partition
5063 * readonly, and if so set the rdonly flag and then
5064 * mark the partition as valid again.
5065 */
5066 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
5067 (sbi->s_mount_state & EXT4_VALID_FS))
5068 es->s_state = cpu_to_le16(sbi->s_mount_state);
5069
5070 if (sbi->s_journal)
5071 ext4_mark_recovery_complete(sb, es);
5072 } else {
5073 /* Make sure we can mount this feature set readwrite */
5074 if (ext4_has_feature_readonly(sb) ||
5075 !ext4_feature_set_ok(sb, 0)) {
5076 err = -EROFS;
5077 goto restore_opts;
5078 }
5079 /*
5080 * Make sure the group descriptor checksums
5081 * are sane. If they aren't, refuse to remount r/w.
5082 */
5083 for (g = 0; g < sbi->s_groups_count; g++) {
5084 struct ext4_group_desc *gdp =
5085 ext4_get_group_desc(sb, g, NULL);
5086
5087 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
5088 ext4_msg(sb, KERN_ERR,
5089 "ext4_remount: Checksum for group %u failed (%u!=%u)",
5090 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
5091 le16_to_cpu(gdp->bg_checksum));
5092 err = -EFSBADCRC;
5093 goto restore_opts;
5094 }
5095 }
5096
5097 /*
5098 * If we have an unprocessed orphan list hanging
5099 * around from a previously readonly bdev mount,
5100 * require a full umount/remount for now.
5101 */
5102 if (es->s_last_orphan) {
5103 ext4_msg(sb, KERN_WARNING, "Couldn't "
5104 "remount RDWR because of unprocessed "
5105 "orphan inode list. Please "
5106 "umount/remount instead");
5107 err = -EINVAL;
5108 goto restore_opts;
5109 }
5110
5111 /*
5112 * Mounting a RDONLY partition read-write, so reread
5113 * and store the current valid flag. (It may have
5114 * been changed by e2fsck since we originally mounted
5115 * the partition.)
5116 */
5117 if (sbi->s_journal)
5118 ext4_clear_journal_err(sb, es);
5119 sbi->s_mount_state = le16_to_cpu(es->s_state);
5120 if (!ext4_setup_super(sb, es, 0))
5121 sb->s_flags &= ~MS_RDONLY;
5122 if (ext4_has_feature_mmp(sb))
5123 if (ext4_multi_mount_protect(sb,
5124 le64_to_cpu(es->s_mmp_block))) {
5125 err = -EROFS;
5126 goto restore_opts;
5127 }
5128 enable_quota = 1;
5129 }
5130 }
5131
5132 /*
5133 * Reinitialize lazy itable initialization thread based on
5134 * current settings
5135 */
5136 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
5137 ext4_unregister_li_request(sb);
5138 else {
5139 ext4_group_t first_not_zeroed;
5140 first_not_zeroed = ext4_has_uninit_itable(sb);
5141 ext4_register_li_request(sb, first_not_zeroed);
5142 }
5143
5144 ext4_setup_system_zone(sb);
5145 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
5146 ext4_commit_super(sb, 1);
5147
5148#ifdef CONFIG_QUOTA
5149 /* Release old quota file names */
5150 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5151 kfree(old_opts.s_qf_names[i]);
5152 if (enable_quota) {
5153 if (sb_any_quota_suspended(sb))
5154 dquot_resume(sb, -1);
5155 else if (ext4_has_feature_quota(sb)) {
5156 err = ext4_enable_quotas(sb);
5157 if (err)
5158 goto restore_opts;
5159 }
5160 }
5161#endif
5162
5163 *flags = (*flags & ~MS_LAZYTIME) | (sb->s_flags & MS_LAZYTIME);
5164 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
5165 kfree(orig_data);
5166 return 0;
5167
5168restore_opts:
5169 sb->s_flags = old_sb_flags;
5170 sbi->s_mount_opt = old_opts.s_mount_opt;
5171 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
5172 sbi->s_resuid = old_opts.s_resuid;
5173 sbi->s_resgid = old_opts.s_resgid;
5174 sbi->s_commit_interval = old_opts.s_commit_interval;
5175 sbi->s_min_batch_time = old_opts.s_min_batch_time;
5176 sbi->s_max_batch_time = old_opts.s_max_batch_time;
5177#ifdef CONFIG_QUOTA
5178 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
5179 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
5180 kfree(sbi->s_qf_names[i]);
5181 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
5182 }
5183#endif
5184 kfree(orig_data);
5185 return err;
5186}
5187
5188#ifdef CONFIG_QUOTA
5189static int ext4_statfs_project(struct super_block *sb,
5190 kprojid_t projid, struct kstatfs *buf)
5191{
5192 struct kqid qid;
5193 struct dquot *dquot;
5194 u64 limit;
5195 u64 curblock;
5196
5197 qid = make_kqid_projid(projid);
5198 dquot = dqget(sb, qid);
5199 if (IS_ERR(dquot))
5200 return PTR_ERR(dquot);
5201 spin_lock(&dq_data_lock);
5202
5203 limit = (dquot->dq_dqb.dqb_bsoftlimit ?
5204 dquot->dq_dqb.dqb_bsoftlimit :
5205 dquot->dq_dqb.dqb_bhardlimit) >> sb->s_blocksize_bits;
5206 if (limit && buf->f_blocks > limit) {
5207 curblock = dquot->dq_dqb.dqb_curspace >> sb->s_blocksize_bits;
5208 buf->f_blocks = limit;
5209 buf->f_bfree = buf->f_bavail =
5210 (buf->f_blocks > curblock) ?
5211 (buf->f_blocks - curblock) : 0;
5212 }
5213
5214 limit = dquot->dq_dqb.dqb_isoftlimit ?
5215 dquot->dq_dqb.dqb_isoftlimit :
5216 dquot->dq_dqb.dqb_ihardlimit;
5217 if (limit && buf->f_files > limit) {
5218 buf->f_files = limit;
5219 buf->f_ffree =
5220 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
5221 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
5222 }
5223
5224 spin_unlock(&dq_data_lock);
5225 dqput(dquot);
5226 return 0;
5227}
5228#endif
5229
5230static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
5231{
5232 struct super_block *sb = dentry->d_sb;
5233 struct ext4_sb_info *sbi = EXT4_SB(sb);
5234 struct ext4_super_block *es = sbi->s_es;
5235 ext4_fsblk_t overhead = 0, resv_blocks;
5236 u64 fsid;
5237 s64 bfree;
5238 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
5239
5240 if (!test_opt(sb, MINIX_DF))
5241 overhead = sbi->s_overhead;
5242
5243 buf->f_type = EXT4_SUPER_MAGIC;
5244 buf->f_bsize = sb->s_blocksize;
5245 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
5246 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
5247 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
5248 /* prevent underflow in case that few free space is available */
5249 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
5250 buf->f_bavail = buf->f_bfree -
5251 (ext4_r_blocks_count(es) + resv_blocks);
5252 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
5253 buf->f_bavail = 0;
5254 buf->f_files = le32_to_cpu(es->s_inodes_count);
5255 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
5256 buf->f_namelen = EXT4_NAME_LEN;
5257 fsid = le64_to_cpup((void *)es->s_uuid) ^
5258 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
5259 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
5260 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
5261
5262#ifdef CONFIG_QUOTA
5263 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) &&
5264 sb_has_quota_limits_enabled(sb, PRJQUOTA))
5265 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf);
5266#endif
5267 return 0;
5268}
5269
5270/* Helper function for writing quotas on sync - we need to start transaction
5271 * before quota file is locked for write. Otherwise the are possible deadlocks:
5272 * Process 1 Process 2
5273 * ext4_create() quota_sync()
5274 * jbd2_journal_start() write_dquot()
5275 * dquot_initialize() down(dqio_mutex)
5276 * down(dqio_mutex) jbd2_journal_start()
5277 *
5278 */
5279
5280#ifdef CONFIG_QUOTA
5281
5282static inline struct inode *dquot_to_inode(struct dquot *dquot)
5283{
5284 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
5285}
5286
5287static int ext4_write_dquot(struct dquot *dquot)
5288{
5289 int ret, err;
5290 handle_t *handle;
5291 struct inode *inode;
5292
5293 inode = dquot_to_inode(dquot);
5294 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
5295 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
5296 if (IS_ERR(handle))
5297 return PTR_ERR(handle);
5298 ret = dquot_commit(dquot);
5299 err = ext4_journal_stop(handle);
5300 if (!ret)
5301 ret = err;
5302 return ret;
5303}
5304
5305static int ext4_acquire_dquot(struct dquot *dquot)
5306{
5307 int ret, err;
5308 handle_t *handle;
5309
5310 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5311 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
5312 if (IS_ERR(handle))
5313 return PTR_ERR(handle);
5314 ret = dquot_acquire(dquot);
5315 err = ext4_journal_stop(handle);
5316 if (!ret)
5317 ret = err;
5318 return ret;
5319}
5320
5321static int ext4_release_dquot(struct dquot *dquot)
5322{
5323 int ret, err;
5324 handle_t *handle;
5325
5326 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5327 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
5328 if (IS_ERR(handle)) {
5329 /* Release dquot anyway to avoid endless cycle in dqput() */
5330 dquot_release(dquot);
5331 return PTR_ERR(handle);
5332 }
5333 ret = dquot_release(dquot);
5334 err = ext4_journal_stop(handle);
5335 if (!ret)
5336 ret = err;
5337 return ret;
5338}
5339
5340static int ext4_mark_dquot_dirty(struct dquot *dquot)
5341{
5342 struct super_block *sb = dquot->dq_sb;
5343 struct ext4_sb_info *sbi = EXT4_SB(sb);
5344
5345 /* Are we journaling quotas? */
5346 if (ext4_has_feature_quota(sb) ||
5347 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
5348 dquot_mark_dquot_dirty(dquot);
5349 return ext4_write_dquot(dquot);
5350 } else {
5351 return dquot_mark_dquot_dirty(dquot);
5352 }
5353}
5354
5355static int ext4_write_info(struct super_block *sb, int type)
5356{
5357 int ret, err;
5358 handle_t *handle;
5359
5360 /* Data block + inode block */
5361 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
5362 if (IS_ERR(handle))
5363 return PTR_ERR(handle);
5364 ret = dquot_commit_info(sb, type);
5365 err = ext4_journal_stop(handle);
5366 if (!ret)
5367 ret = err;
5368 return ret;
5369}
5370
5371/*
5372 * Turn on quotas during mount time - we need to find
5373 * the quota file and such...
5374 */
5375static int ext4_quota_on_mount(struct super_block *sb, int type)
5376{
5377 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
5378 EXT4_SB(sb)->s_jquota_fmt, type);
5379}
5380
5381static void lockdep_set_quota_inode(struct inode *inode, int subclass)
5382{
5383 struct ext4_inode_info *ei = EXT4_I(inode);
5384
5385 /* The first argument of lockdep_set_subclass has to be
5386 * *exactly* the same as the argument to init_rwsem() --- in
5387 * this case, in init_once() --- or lockdep gets unhappy
5388 * because the name of the lock is set using the
5389 * stringification of the argument to init_rwsem().
5390 */
5391 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */
5392 lockdep_set_subclass(&ei->i_data_sem, subclass);
5393}
5394
5395/*
5396 * Standard function to be called on quota_on
5397 */
5398static int ext4_quota_on(struct super_block *sb, int type, int format_id,
5399 const struct path *path)
5400{
5401 int err;
5402
5403 if (!test_opt(sb, QUOTA))
5404 return -EINVAL;
5405
5406 /* Quotafile not on the same filesystem? */
5407 if (path->dentry->d_sb != sb)
5408 return -EXDEV;
5409 /* Journaling quota? */
5410 if (EXT4_SB(sb)->s_qf_names[type]) {
5411 /* Quotafile not in fs root? */
5412 if (path->dentry->d_parent != sb->s_root)
5413 ext4_msg(sb, KERN_WARNING,
5414 "Quota file not on filesystem root. "
5415 "Journaled quota will not work");
5416 }
5417
5418 /*
5419 * When we journal data on quota file, we have to flush journal to see
5420 * all updates to the file when we bypass pagecache...
5421 */
5422 if (EXT4_SB(sb)->s_journal &&
5423 ext4_should_journal_data(d_inode(path->dentry))) {
5424 /*
5425 * We don't need to lock updates but journal_flush() could
5426 * otherwise be livelocked...
5427 */
5428 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
5429 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
5430 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
5431 if (err)
5432 return err;
5433 }
5434
5435 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
5436 err = dquot_quota_on(sb, type, format_id, path);
5437 if (err) {
5438 lockdep_set_quota_inode(path->dentry->d_inode,
5439 I_DATA_SEM_NORMAL);
5440 } else {
5441 struct inode *inode = d_inode(path->dentry);
5442 handle_t *handle;
5443
5444 /*
5445 * Set inode flags to prevent userspace from messing with quota
5446 * files. If this fails, we return success anyway since quotas
5447 * are already enabled and this is not a hard failure.
5448 */
5449 inode_lock(inode);
5450 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5451 if (IS_ERR(handle))
5452 goto unlock_inode;
5453 EXT4_I(inode)->i_flags |= EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL;
5454 inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
5455 S_NOATIME | S_IMMUTABLE);
5456 ext4_mark_inode_dirty(handle, inode);
5457 ext4_journal_stop(handle);
5458 unlock_inode:
5459 inode_unlock(inode);
5460 }
5461 return err;
5462}
5463
5464static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5465 unsigned int flags)
5466{
5467 int err;
5468 struct inode *qf_inode;
5469 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5470 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5471 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5472 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
5473 };
5474
5475 BUG_ON(!ext4_has_feature_quota(sb));
5476
5477 if (!qf_inums[type])
5478 return -EPERM;
5479
5480 qf_inode = ext4_iget(sb, qf_inums[type]);
5481 if (IS_ERR(qf_inode)) {
5482 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5483 return PTR_ERR(qf_inode);
5484 }
5485
5486 /* Don't account quota for quota files to avoid recursion */
5487 qf_inode->i_flags |= S_NOQUOTA;
5488 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
5489 err = dquot_enable(qf_inode, type, format_id, flags);
5490 iput(qf_inode);
5491 if (err)
5492 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
5493
5494 return err;
5495}
5496
5497/* Enable usage tracking for all quota types. */
5498static int ext4_enable_quotas(struct super_block *sb)
5499{
5500 int type, err = 0;
5501 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5502 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5503 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
5504 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
5505 };
5506 bool quota_mopt[EXT4_MAXQUOTAS] = {
5507 test_opt(sb, USRQUOTA),
5508 test_opt(sb, GRPQUOTA),
5509 test_opt(sb, PRJQUOTA),
5510 };
5511
5512 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
5513 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
5514 if (qf_inums[type]) {
5515 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5516 DQUOT_USAGE_ENABLED |
5517 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
5518 if (err) {
5519 ext4_warning(sb,
5520 "Failed to enable quota tracking "
5521 "(type=%d, err=%d). Please run "
5522 "e2fsck to fix.", type, err);
5523 return err;
5524 }
5525 }
5526 }
5527 return 0;
5528}
5529
5530static int ext4_quota_off(struct super_block *sb, int type)
5531{
5532 struct inode *inode = sb_dqopt(sb)->files[type];
5533 handle_t *handle;
5534 int err;
5535
5536 /* Force all delayed allocation blocks to be allocated.
5537 * Caller already holds s_umount sem */
5538 if (test_opt(sb, DELALLOC))
5539 sync_filesystem(sb);
5540
5541 if (!inode || !igrab(inode))
5542 goto out;
5543
5544 err = dquot_quota_off(sb, type);
5545 if (err || ext4_has_feature_quota(sb))
5546 goto out_put;
5547
5548 inode_lock(inode);
5549 /*
5550 * Update modification times of quota files when userspace can
5551 * start looking at them. If we fail, we return success anyway since
5552 * this is not a hard failure and quotas are already disabled.
5553 */
5554 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5555 if (IS_ERR(handle))
5556 goto out_unlock;
5557 EXT4_I(inode)->i_flags &= ~(EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL);
5558 inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
5559 inode->i_mtime = inode->i_ctime = current_time(inode);
5560 ext4_mark_inode_dirty(handle, inode);
5561 ext4_journal_stop(handle);
5562out_unlock:
5563 inode_unlock(inode);
5564out_put:
5565 lockdep_set_quota_inode(inode, I_DATA_SEM_NORMAL);
5566 iput(inode);
5567 return err;
5568out:
5569 return dquot_quota_off(sb, type);
5570}
5571
5572/* Read data from quotafile - avoid pagecache and such because we cannot afford
5573 * acquiring the locks... As quota files are never truncated and quota code
5574 * itself serializes the operations (and no one else should touch the files)
5575 * we don't have to be afraid of races */
5576static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
5577 size_t len, loff_t off)
5578{
5579 struct inode *inode = sb_dqopt(sb)->files[type];
5580 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5581 int offset = off & (sb->s_blocksize - 1);
5582 int tocopy;
5583 size_t toread;
5584 struct buffer_head *bh;
5585 loff_t i_size = i_size_read(inode);
5586
5587 if (off > i_size)
5588 return 0;
5589 if (off+len > i_size)
5590 len = i_size-off;
5591 toread = len;
5592 while (toread > 0) {
5593 tocopy = sb->s_blocksize - offset < toread ?
5594 sb->s_blocksize - offset : toread;
5595 bh = ext4_bread(NULL, inode, blk, 0);
5596 if (IS_ERR(bh))
5597 return PTR_ERR(bh);
5598 if (!bh) /* A hole? */
5599 memset(data, 0, tocopy);
5600 else
5601 memcpy(data, bh->b_data+offset, tocopy);
5602 brelse(bh);
5603 offset = 0;
5604 toread -= tocopy;
5605 data += tocopy;
5606 blk++;
5607 }
5608 return len;
5609}
5610
5611/* Write to quotafile (we know the transaction is already started and has
5612 * enough credits) */
5613static ssize_t ext4_quota_write(struct super_block *sb, int type,
5614 const char *data, size_t len, loff_t off)
5615{
5616 struct inode *inode = sb_dqopt(sb)->files[type];
5617 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5618 int err, offset = off & (sb->s_blocksize - 1);
5619 int retries = 0;
5620 struct buffer_head *bh;
5621 handle_t *handle = journal_current_handle();
5622
5623 if (EXT4_SB(sb)->s_journal && !handle) {
5624 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5625 " cancelled because transaction is not started",
5626 (unsigned long long)off, (unsigned long long)len);
5627 return -EIO;
5628 }
5629 /*
5630 * Since we account only one data block in transaction credits,
5631 * then it is impossible to cross a block boundary.
5632 */
5633 if (sb->s_blocksize - offset < len) {
5634 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5635 " cancelled because not block aligned",
5636 (unsigned long long)off, (unsigned long long)len);
5637 return -EIO;
5638 }
5639
5640 do {
5641 bh = ext4_bread(handle, inode, blk,
5642 EXT4_GET_BLOCKS_CREATE |
5643 EXT4_GET_BLOCKS_METADATA_NOFAIL);
5644 } while (IS_ERR(bh) && (PTR_ERR(bh) == -ENOSPC) &&
5645 ext4_should_retry_alloc(inode->i_sb, &retries));
5646 if (IS_ERR(bh))
5647 return PTR_ERR(bh);
5648 if (!bh)
5649 goto out;
5650 BUFFER_TRACE(bh, "get write access");
5651 err = ext4_journal_get_write_access(handle, bh);
5652 if (err) {
5653 brelse(bh);
5654 return err;
5655 }
5656 lock_buffer(bh);
5657 memcpy(bh->b_data+offset, data, len);
5658 flush_dcache_page(bh->b_page);
5659 unlock_buffer(bh);
5660 err = ext4_handle_dirty_metadata(handle, NULL, bh);
5661 brelse(bh);
5662out:
5663 if (inode->i_size < off + len) {
5664 i_size_write(inode, off + len);
5665 EXT4_I(inode)->i_disksize = inode->i_size;
5666 ext4_mark_inode_dirty(handle, inode);
5667 }
5668 return len;
5669}
5670
5671static int ext4_get_next_id(struct super_block *sb, struct kqid *qid)
5672{
5673 const struct quota_format_ops *ops;
5674
5675 if (!sb_has_quota_loaded(sb, qid->type))
5676 return -ESRCH;
5677 ops = sb_dqopt(sb)->ops[qid->type];
5678 if (!ops || !ops->get_next_id)
5679 return -ENOSYS;
5680 return dquot_get_next_id(sb, qid);
5681}
5682#endif
5683
5684static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5685 const char *dev_name, void *data)
5686{
5687 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
5688}
5689
5690#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
5691static inline void register_as_ext2(void)
5692{
5693 int err = register_filesystem(&ext2_fs_type);
5694 if (err)
5695 printk(KERN_WARNING
5696 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5697}
5698
5699static inline void unregister_as_ext2(void)
5700{
5701 unregister_filesystem(&ext2_fs_type);
5702}
5703
5704static inline int ext2_feature_set_ok(struct super_block *sb)
5705{
5706 if (ext4_has_unknown_ext2_incompat_features(sb))
5707 return 0;
5708 if (sb->s_flags & MS_RDONLY)
5709 return 1;
5710 if (ext4_has_unknown_ext2_ro_compat_features(sb))
5711 return 0;
5712 return 1;
5713}
5714#else
5715static inline void register_as_ext2(void) { }
5716static inline void unregister_as_ext2(void) { }
5717static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
5718#endif
5719
5720static inline void register_as_ext3(void)
5721{
5722 int err = register_filesystem(&ext3_fs_type);
5723 if (err)
5724 printk(KERN_WARNING
5725 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5726}
5727
5728static inline void unregister_as_ext3(void)
5729{
5730 unregister_filesystem(&ext3_fs_type);
5731}
5732
5733static inline int ext3_feature_set_ok(struct super_block *sb)
5734{
5735 if (ext4_has_unknown_ext3_incompat_features(sb))
5736 return 0;
5737 if (!ext4_has_feature_journal(sb))
5738 return 0;
5739 if (sb->s_flags & MS_RDONLY)
5740 return 1;
5741 if (ext4_has_unknown_ext3_ro_compat_features(sb))
5742 return 0;
5743 return 1;
5744}
5745
5746static struct file_system_type ext4_fs_type = {
5747 .owner = THIS_MODULE,
5748 .name = "ext4",
5749 .mount = ext4_mount,
5750 .kill_sb = kill_block_super,
5751 .fs_flags = FS_REQUIRES_DEV,
5752};
5753MODULE_ALIAS_FS("ext4");
5754
5755/* Shared across all ext4 file systems */
5756wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5757
5758static int __init ext4_init_fs(void)
5759{
5760 int i, err;
5761
5762 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
5763 ext4_li_info = NULL;
5764 mutex_init(&ext4_li_mtx);
5765
5766 /* Build-time check for flags consistency */
5767 ext4_check_flag_values();
5768
5769 for (i = 0; i < EXT4_WQ_HASH_SZ; i++)
5770 init_waitqueue_head(&ext4__ioend_wq[i]);
5771
5772 err = ext4_init_es();
5773 if (err)
5774 return err;
5775
5776 err = ext4_init_pageio();
5777 if (err)
5778 goto out5;
5779
5780 err = ext4_init_system_zone();
5781 if (err)
5782 goto out4;
5783
5784 err = ext4_init_sysfs();
5785 if (err)
5786 goto out3;
5787
5788 err = ext4_init_mballoc();
5789 if (err)
5790 goto out2;
5791 err = init_inodecache();
5792 if (err)
5793 goto out1;
5794 register_as_ext3();
5795 register_as_ext2();
5796 err = register_filesystem(&ext4_fs_type);
5797 if (err)
5798 goto out;
5799
5800 return 0;
5801out:
5802 unregister_as_ext2();
5803 unregister_as_ext3();
5804 destroy_inodecache();
5805out1:
5806 ext4_exit_mballoc();
5807out2:
5808 ext4_exit_sysfs();
5809out3:
5810 ext4_exit_system_zone();
5811out4:
5812 ext4_exit_pageio();
5813out5:
5814 ext4_exit_es();
5815
5816 return err;
5817}
5818
5819static void __exit ext4_exit_fs(void)
5820{
5821 ext4_destroy_lazyinit_thread();
5822 unregister_as_ext2();
5823 unregister_as_ext3();
5824 unregister_filesystem(&ext4_fs_type);
5825 destroy_inodecache();
5826 ext4_exit_mballoc();
5827 ext4_exit_sysfs();
5828 ext4_exit_system_zone();
5829 ext4_exit_pageio();
5830 ext4_exit_es();
5831}
5832
5833MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5834MODULE_DESCRIPTION("Fourth Extended Filesystem");
5835MODULE_LICENSE("GPL");
5836module_init(ext4_init_fs)
5837module_exit(ext4_exit_fs)
Linux 6.x block layer and io_uring tree(s)