Loading...
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/jbd2.h>
25#include <linux/slab.h>
26#include <linux/init.h>
27#include <linux/blkdev.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/proc_fs.h>
38#include <linux/ctype.h>
39#include <linux/log2.h>
40#include <linux/crc16.h>
41#include <linux/cleancache.h>
42#include <asm/uaccess.h>
43
44#include <linux/kthread.h>
45#include <linux/freezer.h>
46
47#include "ext4.h"
48#include "ext4_extents.h"
49#include "ext4_jbd2.h"
50#include "xattr.h"
51#include "acl.h"
52#include "mballoc.h"
53
54#define CREATE_TRACE_POINTS
55#include <trace/events/ext4.h>
56
57static struct proc_dir_entry *ext4_proc_root;
58static struct kset *ext4_kset;
59static struct ext4_lazy_init *ext4_li_info;
60static struct mutex ext4_li_mtx;
61static struct ext4_features *ext4_feat;
62
63static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
64 unsigned long journal_devnum);
65static int ext4_show_options(struct seq_file *seq, struct dentry *root);
66static int ext4_commit_super(struct super_block *sb, int sync);
67static void ext4_mark_recovery_complete(struct super_block *sb,
68 struct ext4_super_block *es);
69static void ext4_clear_journal_err(struct super_block *sb,
70 struct ext4_super_block *es);
71static int ext4_sync_fs(struct super_block *sb, int wait);
72static const char *ext4_decode_error(struct super_block *sb, int errno,
73 char nbuf[16]);
74static int ext4_remount(struct super_block *sb, int *flags, char *data);
75static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
76static int ext4_unfreeze(struct super_block *sb);
77static void ext4_write_super(struct super_block *sb);
78static int ext4_freeze(struct super_block *sb);
79static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
80 const char *dev_name, void *data);
81static inline int ext2_feature_set_ok(struct super_block *sb);
82static inline int ext3_feature_set_ok(struct super_block *sb);
83static int ext4_feature_set_ok(struct super_block *sb, int readonly);
84static void ext4_destroy_lazyinit_thread(void);
85static void ext4_unregister_li_request(struct super_block *sb);
86static void ext4_clear_request_list(void);
87
88#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
89static struct file_system_type ext2_fs_type = {
90 .owner = THIS_MODULE,
91 .name = "ext2",
92 .mount = ext4_mount,
93 .kill_sb = kill_block_super,
94 .fs_flags = FS_REQUIRES_DEV,
95};
96#define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
97#else
98#define IS_EXT2_SB(sb) (0)
99#endif
100
101
102#if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
103static struct file_system_type ext3_fs_type = {
104 .owner = THIS_MODULE,
105 .name = "ext3",
106 .mount = ext4_mount,
107 .kill_sb = kill_block_super,
108 .fs_flags = FS_REQUIRES_DEV,
109};
110#define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
111#else
112#define IS_EXT3_SB(sb) (0)
113#endif
114
115static int ext4_verify_csum_type(struct super_block *sb,
116 struct ext4_super_block *es)
117{
118 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
119 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
120 return 1;
121
122 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
123}
124
125static __le32 ext4_superblock_csum(struct super_block *sb,
126 struct ext4_super_block *es)
127{
128 struct ext4_sb_info *sbi = EXT4_SB(sb);
129 int offset = offsetof(struct ext4_super_block, s_checksum);
130 __u32 csum;
131
132 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
133
134 return cpu_to_le32(csum);
135}
136
137int ext4_superblock_csum_verify(struct super_block *sb,
138 struct ext4_super_block *es)
139{
140 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
141 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
142 return 1;
143
144 return es->s_checksum == ext4_superblock_csum(sb, es);
145}
146
147void ext4_superblock_csum_set(struct super_block *sb,
148 struct ext4_super_block *es)
149{
150 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
151 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
152 return;
153
154 es->s_checksum = ext4_superblock_csum(sb, es);
155}
156
157void *ext4_kvmalloc(size_t size, gfp_t flags)
158{
159 void *ret;
160
161 ret = kmalloc(size, flags);
162 if (!ret)
163 ret = __vmalloc(size, flags, PAGE_KERNEL);
164 return ret;
165}
166
167void *ext4_kvzalloc(size_t size, gfp_t flags)
168{
169 void *ret;
170
171 ret = kzalloc(size, flags);
172 if (!ret)
173 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
174 return ret;
175}
176
177void ext4_kvfree(void *ptr)
178{
179 if (is_vmalloc_addr(ptr))
180 vfree(ptr);
181 else
182 kfree(ptr);
183
184}
185
186ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
187 struct ext4_group_desc *bg)
188{
189 return le32_to_cpu(bg->bg_block_bitmap_lo) |
190 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
191 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
192}
193
194ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
195 struct ext4_group_desc *bg)
196{
197 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
198 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
199 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
200}
201
202ext4_fsblk_t ext4_inode_table(struct super_block *sb,
203 struct ext4_group_desc *bg)
204{
205 return le32_to_cpu(bg->bg_inode_table_lo) |
206 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
207 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
208}
209
210__u32 ext4_free_group_clusters(struct super_block *sb,
211 struct ext4_group_desc *bg)
212{
213 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
214 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
215 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
216}
217
218__u32 ext4_free_inodes_count(struct super_block *sb,
219 struct ext4_group_desc *bg)
220{
221 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
222 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
223 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
224}
225
226__u32 ext4_used_dirs_count(struct super_block *sb,
227 struct ext4_group_desc *bg)
228{
229 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
230 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
231 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
232}
233
234__u32 ext4_itable_unused_count(struct super_block *sb,
235 struct ext4_group_desc *bg)
236{
237 return le16_to_cpu(bg->bg_itable_unused_lo) |
238 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
239 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
240}
241
242void ext4_block_bitmap_set(struct super_block *sb,
243 struct ext4_group_desc *bg, ext4_fsblk_t blk)
244{
245 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
246 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
247 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
248}
249
250void ext4_inode_bitmap_set(struct super_block *sb,
251 struct ext4_group_desc *bg, ext4_fsblk_t blk)
252{
253 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
254 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
255 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
256}
257
258void ext4_inode_table_set(struct super_block *sb,
259 struct ext4_group_desc *bg, ext4_fsblk_t blk)
260{
261 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
262 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
263 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
264}
265
266void ext4_free_group_clusters_set(struct super_block *sb,
267 struct ext4_group_desc *bg, __u32 count)
268{
269 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
270 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
271 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
272}
273
274void ext4_free_inodes_set(struct super_block *sb,
275 struct ext4_group_desc *bg, __u32 count)
276{
277 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
278 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
279 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
280}
281
282void ext4_used_dirs_set(struct super_block *sb,
283 struct ext4_group_desc *bg, __u32 count)
284{
285 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
286 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
287 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
288}
289
290void ext4_itable_unused_set(struct super_block *sb,
291 struct ext4_group_desc *bg, __u32 count)
292{
293 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
294 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
295 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
296}
297
298
299/* Just increment the non-pointer handle value */
300static handle_t *ext4_get_nojournal(void)
301{
302 handle_t *handle = current->journal_info;
303 unsigned long ref_cnt = (unsigned long)handle;
304
305 BUG_ON(ref_cnt >= EXT4_NOJOURNAL_MAX_REF_COUNT);
306
307 ref_cnt++;
308 handle = (handle_t *)ref_cnt;
309
310 current->journal_info = handle;
311 return handle;
312}
313
314
315/* Decrement the non-pointer handle value */
316static void ext4_put_nojournal(handle_t *handle)
317{
318 unsigned long ref_cnt = (unsigned long)handle;
319
320 BUG_ON(ref_cnt == 0);
321
322 ref_cnt--;
323 handle = (handle_t *)ref_cnt;
324
325 current->journal_info = handle;
326}
327
328/*
329 * Wrappers for jbd2_journal_start/end.
330 *
331 * The only special thing we need to do here is to make sure that all
332 * journal_end calls result in the superblock being marked dirty, so
333 * that sync() will call the filesystem's write_super callback if
334 * appropriate.
335 *
336 * To avoid j_barrier hold in userspace when a user calls freeze(),
337 * ext4 prevents a new handle from being started by s_frozen, which
338 * is in an upper layer.
339 */
340handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
341{
342 journal_t *journal;
343 handle_t *handle;
344
345 trace_ext4_journal_start(sb, nblocks, _RET_IP_);
346 if (sb->s_flags & MS_RDONLY)
347 return ERR_PTR(-EROFS);
348
349 journal = EXT4_SB(sb)->s_journal;
350 handle = ext4_journal_current_handle();
351
352 /*
353 * If a handle has been started, it should be allowed to
354 * finish, otherwise deadlock could happen between freeze
355 * and others(e.g. truncate) due to the restart of the
356 * journal handle if the filesystem is forzen and active
357 * handles are not stopped.
358 */
359 if (!handle)
360 vfs_check_frozen(sb, SB_FREEZE_TRANS);
361
362 if (!journal)
363 return ext4_get_nojournal();
364 /*
365 * Special case here: if the journal has aborted behind our
366 * backs (eg. EIO in the commit thread), then we still need to
367 * take the FS itself readonly cleanly.
368 */
369 if (is_journal_aborted(journal)) {
370 ext4_abort(sb, "Detected aborted journal");
371 return ERR_PTR(-EROFS);
372 }
373 return jbd2_journal_start(journal, nblocks);
374}
375
376/*
377 * The only special thing we need to do here is to make sure that all
378 * jbd2_journal_stop calls result in the superblock being marked dirty, so
379 * that sync() will call the filesystem's write_super callback if
380 * appropriate.
381 */
382int __ext4_journal_stop(const char *where, unsigned int line, handle_t *handle)
383{
384 struct super_block *sb;
385 int err;
386 int rc;
387
388 if (!ext4_handle_valid(handle)) {
389 ext4_put_nojournal(handle);
390 return 0;
391 }
392 sb = handle->h_transaction->t_journal->j_private;
393 err = handle->h_err;
394 rc = jbd2_journal_stop(handle);
395
396 if (!err)
397 err = rc;
398 if (err)
399 __ext4_std_error(sb, where, line, err);
400 return err;
401}
402
403void ext4_journal_abort_handle(const char *caller, unsigned int line,
404 const char *err_fn, struct buffer_head *bh,
405 handle_t *handle, int err)
406{
407 char nbuf[16];
408 const char *errstr = ext4_decode_error(NULL, err, nbuf);
409
410 BUG_ON(!ext4_handle_valid(handle));
411
412 if (bh)
413 BUFFER_TRACE(bh, "abort");
414
415 if (!handle->h_err)
416 handle->h_err = err;
417
418 if (is_handle_aborted(handle))
419 return;
420
421 printk(KERN_ERR "EXT4-fs: %s:%d: aborting transaction: %s in %s\n",
422 caller, line, errstr, err_fn);
423
424 jbd2_journal_abort_handle(handle);
425}
426
427static void __save_error_info(struct super_block *sb, const char *func,
428 unsigned int line)
429{
430 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
431
432 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
433 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
434 es->s_last_error_time = cpu_to_le32(get_seconds());
435 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
436 es->s_last_error_line = cpu_to_le32(line);
437 if (!es->s_first_error_time) {
438 es->s_first_error_time = es->s_last_error_time;
439 strncpy(es->s_first_error_func, func,
440 sizeof(es->s_first_error_func));
441 es->s_first_error_line = cpu_to_le32(line);
442 es->s_first_error_ino = es->s_last_error_ino;
443 es->s_first_error_block = es->s_last_error_block;
444 }
445 /*
446 * Start the daily error reporting function if it hasn't been
447 * started already
448 */
449 if (!es->s_error_count)
450 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
451 es->s_error_count = cpu_to_le32(le32_to_cpu(es->s_error_count) + 1);
452}
453
454static void save_error_info(struct super_block *sb, const char *func,
455 unsigned int line)
456{
457 __save_error_info(sb, func, line);
458 ext4_commit_super(sb, 1);
459}
460
461/*
462 * The del_gendisk() function uninitializes the disk-specific data
463 * structures, including the bdi structure, without telling anyone
464 * else. Once this happens, any attempt to call mark_buffer_dirty()
465 * (for example, by ext4_commit_super), will cause a kernel OOPS.
466 * This is a kludge to prevent these oops until we can put in a proper
467 * hook in del_gendisk() to inform the VFS and file system layers.
468 */
469static int block_device_ejected(struct super_block *sb)
470{
471 struct inode *bd_inode = sb->s_bdev->bd_inode;
472 struct backing_dev_info *bdi = bd_inode->i_mapping->backing_dev_info;
473
474 return bdi->dev == NULL;
475}
476
477static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
478{
479 struct super_block *sb = journal->j_private;
480 struct ext4_sb_info *sbi = EXT4_SB(sb);
481 int error = is_journal_aborted(journal);
482 struct ext4_journal_cb_entry *jce, *tmp;
483
484 spin_lock(&sbi->s_md_lock);
485 list_for_each_entry_safe(jce, tmp, &txn->t_private_list, jce_list) {
486 list_del_init(&jce->jce_list);
487 spin_unlock(&sbi->s_md_lock);
488 jce->jce_func(sb, jce, error);
489 spin_lock(&sbi->s_md_lock);
490 }
491 spin_unlock(&sbi->s_md_lock);
492}
493
494/* Deal with the reporting of failure conditions on a filesystem such as
495 * inconsistencies detected or read IO failures.
496 *
497 * On ext2, we can store the error state of the filesystem in the
498 * superblock. That is not possible on ext4, because we may have other
499 * write ordering constraints on the superblock which prevent us from
500 * writing it out straight away; and given that the journal is about to
501 * be aborted, we can't rely on the current, or future, transactions to
502 * write out the superblock safely.
503 *
504 * We'll just use the jbd2_journal_abort() error code to record an error in
505 * the journal instead. On recovery, the journal will complain about
506 * that error until we've noted it down and cleared it.
507 */
508
509static void ext4_handle_error(struct super_block *sb)
510{
511 if (sb->s_flags & MS_RDONLY)
512 return;
513
514 if (!test_opt(sb, ERRORS_CONT)) {
515 journal_t *journal = EXT4_SB(sb)->s_journal;
516
517 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
518 if (journal)
519 jbd2_journal_abort(journal, -EIO);
520 }
521 if (test_opt(sb, ERRORS_RO)) {
522 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
523 sb->s_flags |= MS_RDONLY;
524 }
525 if (test_opt(sb, ERRORS_PANIC))
526 panic("EXT4-fs (device %s): panic forced after error\n",
527 sb->s_id);
528}
529
530void __ext4_error(struct super_block *sb, const char *function,
531 unsigned int line, const char *fmt, ...)
532{
533 struct va_format vaf;
534 va_list args;
535
536 va_start(args, fmt);
537 vaf.fmt = fmt;
538 vaf.va = &args;
539 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
540 sb->s_id, function, line, current->comm, &vaf);
541 va_end(args);
542 save_error_info(sb, function, line);
543
544 ext4_handle_error(sb);
545}
546
547void ext4_error_inode(struct inode *inode, const char *function,
548 unsigned int line, ext4_fsblk_t block,
549 const char *fmt, ...)
550{
551 va_list args;
552 struct va_format vaf;
553 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
554
555 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
556 es->s_last_error_block = cpu_to_le64(block);
557 save_error_info(inode->i_sb, function, line);
558 va_start(args, fmt);
559 vaf.fmt = fmt;
560 vaf.va = &args;
561 if (block)
562 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
563 "inode #%lu: block %llu: comm %s: %pV\n",
564 inode->i_sb->s_id, function, line, inode->i_ino,
565 block, current->comm, &vaf);
566 else
567 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
568 "inode #%lu: comm %s: %pV\n",
569 inode->i_sb->s_id, function, line, inode->i_ino,
570 current->comm, &vaf);
571 va_end(args);
572
573 ext4_handle_error(inode->i_sb);
574}
575
576void ext4_error_file(struct file *file, const char *function,
577 unsigned int line, ext4_fsblk_t block,
578 const char *fmt, ...)
579{
580 va_list args;
581 struct va_format vaf;
582 struct ext4_super_block *es;
583 struct inode *inode = file->f_dentry->d_inode;
584 char pathname[80], *path;
585
586 es = EXT4_SB(inode->i_sb)->s_es;
587 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
588 save_error_info(inode->i_sb, function, line);
589 path = d_path(&(file->f_path), pathname, sizeof(pathname));
590 if (IS_ERR(path))
591 path = "(unknown)";
592 va_start(args, fmt);
593 vaf.fmt = fmt;
594 vaf.va = &args;
595 if (block)
596 printk(KERN_CRIT
597 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
598 "block %llu: comm %s: path %s: %pV\n",
599 inode->i_sb->s_id, function, line, inode->i_ino,
600 block, current->comm, path, &vaf);
601 else
602 printk(KERN_CRIT
603 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
604 "comm %s: path %s: %pV\n",
605 inode->i_sb->s_id, function, line, inode->i_ino,
606 current->comm, path, &vaf);
607 va_end(args);
608
609 ext4_handle_error(inode->i_sb);
610}
611
612static const char *ext4_decode_error(struct super_block *sb, int errno,
613 char nbuf[16])
614{
615 char *errstr = NULL;
616
617 switch (errno) {
618 case -EIO:
619 errstr = "IO failure";
620 break;
621 case -ENOMEM:
622 errstr = "Out of memory";
623 break;
624 case -EROFS:
625 if (!sb || (EXT4_SB(sb)->s_journal &&
626 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
627 errstr = "Journal has aborted";
628 else
629 errstr = "Readonly filesystem";
630 break;
631 default:
632 /* If the caller passed in an extra buffer for unknown
633 * errors, textualise them now. Else we just return
634 * NULL. */
635 if (nbuf) {
636 /* Check for truncated error codes... */
637 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
638 errstr = nbuf;
639 }
640 break;
641 }
642
643 return errstr;
644}
645
646/* __ext4_std_error decodes expected errors from journaling functions
647 * automatically and invokes the appropriate error response. */
648
649void __ext4_std_error(struct super_block *sb, const char *function,
650 unsigned int line, int errno)
651{
652 char nbuf[16];
653 const char *errstr;
654
655 /* Special case: if the error is EROFS, and we're not already
656 * inside a transaction, then there's really no point in logging
657 * an error. */
658 if (errno == -EROFS && journal_current_handle() == NULL &&
659 (sb->s_flags & MS_RDONLY))
660 return;
661
662 errstr = ext4_decode_error(sb, errno, nbuf);
663 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
664 sb->s_id, function, line, errstr);
665 save_error_info(sb, function, line);
666
667 ext4_handle_error(sb);
668}
669
670/*
671 * ext4_abort is a much stronger failure handler than ext4_error. The
672 * abort function may be used to deal with unrecoverable failures such
673 * as journal IO errors or ENOMEM at a critical moment in log management.
674 *
675 * We unconditionally force the filesystem into an ABORT|READONLY state,
676 * unless the error response on the fs has been set to panic in which
677 * case we take the easy way out and panic immediately.
678 */
679
680void __ext4_abort(struct super_block *sb, const char *function,
681 unsigned int line, const char *fmt, ...)
682{
683 va_list args;
684
685 save_error_info(sb, function, line);
686 va_start(args, fmt);
687 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
688 function, line);
689 vprintk(fmt, args);
690 printk("\n");
691 va_end(args);
692
693 if ((sb->s_flags & MS_RDONLY) == 0) {
694 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
695 sb->s_flags |= MS_RDONLY;
696 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
697 if (EXT4_SB(sb)->s_journal)
698 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
699 save_error_info(sb, function, line);
700 }
701 if (test_opt(sb, ERRORS_PANIC))
702 panic("EXT4-fs panic from previous error\n");
703}
704
705void ext4_msg(struct super_block *sb, const char *prefix, const char *fmt, ...)
706{
707 struct va_format vaf;
708 va_list args;
709
710 va_start(args, fmt);
711 vaf.fmt = fmt;
712 vaf.va = &args;
713 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
714 va_end(args);
715}
716
717void __ext4_warning(struct super_block *sb, const char *function,
718 unsigned int line, const char *fmt, ...)
719{
720 struct va_format vaf;
721 va_list args;
722
723 va_start(args, fmt);
724 vaf.fmt = fmt;
725 vaf.va = &args;
726 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
727 sb->s_id, function, line, &vaf);
728 va_end(args);
729}
730
731void __ext4_grp_locked_error(const char *function, unsigned int line,
732 struct super_block *sb, ext4_group_t grp,
733 unsigned long ino, ext4_fsblk_t block,
734 const char *fmt, ...)
735__releases(bitlock)
736__acquires(bitlock)
737{
738 struct va_format vaf;
739 va_list args;
740 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
741
742 es->s_last_error_ino = cpu_to_le32(ino);
743 es->s_last_error_block = cpu_to_le64(block);
744 __save_error_info(sb, function, line);
745
746 va_start(args, fmt);
747
748 vaf.fmt = fmt;
749 vaf.va = &args;
750 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
751 sb->s_id, function, line, grp);
752 if (ino)
753 printk(KERN_CONT "inode %lu: ", ino);
754 if (block)
755 printk(KERN_CONT "block %llu:", (unsigned long long) block);
756 printk(KERN_CONT "%pV\n", &vaf);
757 va_end(args);
758
759 if (test_opt(sb, ERRORS_CONT)) {
760 ext4_commit_super(sb, 0);
761 return;
762 }
763
764 ext4_unlock_group(sb, grp);
765 ext4_handle_error(sb);
766 /*
767 * We only get here in the ERRORS_RO case; relocking the group
768 * may be dangerous, but nothing bad will happen since the
769 * filesystem will have already been marked read/only and the
770 * journal has been aborted. We return 1 as a hint to callers
771 * who might what to use the return value from
772 * ext4_grp_locked_error() to distinguish between the
773 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
774 * aggressively from the ext4 function in question, with a
775 * more appropriate error code.
776 */
777 ext4_lock_group(sb, grp);
778 return;
779}
780
781void ext4_update_dynamic_rev(struct super_block *sb)
782{
783 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
784
785 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
786 return;
787
788 ext4_warning(sb,
789 "updating to rev %d because of new feature flag, "
790 "running e2fsck is recommended",
791 EXT4_DYNAMIC_REV);
792
793 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
794 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
795 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
796 /* leave es->s_feature_*compat flags alone */
797 /* es->s_uuid will be set by e2fsck if empty */
798
799 /*
800 * The rest of the superblock fields should be zero, and if not it
801 * means they are likely already in use, so leave them alone. We
802 * can leave it up to e2fsck to clean up any inconsistencies there.
803 */
804}
805
806/*
807 * Open the external journal device
808 */
809static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
810{
811 struct block_device *bdev;
812 char b[BDEVNAME_SIZE];
813
814 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
815 if (IS_ERR(bdev))
816 goto fail;
817 return bdev;
818
819fail:
820 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
821 __bdevname(dev, b), PTR_ERR(bdev));
822 return NULL;
823}
824
825/*
826 * Release the journal device
827 */
828static int ext4_blkdev_put(struct block_device *bdev)
829{
830 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
831}
832
833static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
834{
835 struct block_device *bdev;
836 int ret = -ENODEV;
837
838 bdev = sbi->journal_bdev;
839 if (bdev) {
840 ret = ext4_blkdev_put(bdev);
841 sbi->journal_bdev = NULL;
842 }
843 return ret;
844}
845
846static inline struct inode *orphan_list_entry(struct list_head *l)
847{
848 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
849}
850
851static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
852{
853 struct list_head *l;
854
855 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
856 le32_to_cpu(sbi->s_es->s_last_orphan));
857
858 printk(KERN_ERR "sb_info orphan list:\n");
859 list_for_each(l, &sbi->s_orphan) {
860 struct inode *inode = orphan_list_entry(l);
861 printk(KERN_ERR " "
862 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
863 inode->i_sb->s_id, inode->i_ino, inode,
864 inode->i_mode, inode->i_nlink,
865 NEXT_ORPHAN(inode));
866 }
867}
868
869static void ext4_put_super(struct super_block *sb)
870{
871 struct ext4_sb_info *sbi = EXT4_SB(sb);
872 struct ext4_super_block *es = sbi->s_es;
873 int i, err;
874
875 ext4_unregister_li_request(sb);
876 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
877
878 flush_workqueue(sbi->dio_unwritten_wq);
879 destroy_workqueue(sbi->dio_unwritten_wq);
880
881 lock_super(sb);
882 if (sbi->s_journal) {
883 err = jbd2_journal_destroy(sbi->s_journal);
884 sbi->s_journal = NULL;
885 if (err < 0)
886 ext4_abort(sb, "Couldn't clean up the journal");
887 }
888
889 del_timer(&sbi->s_err_report);
890 ext4_release_system_zone(sb);
891 ext4_mb_release(sb);
892 ext4_ext_release(sb);
893 ext4_xattr_put_super(sb);
894
895 if (!(sb->s_flags & MS_RDONLY)) {
896 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
897 es->s_state = cpu_to_le16(sbi->s_mount_state);
898 }
899 if (sb->s_dirt || !(sb->s_flags & MS_RDONLY))
900 ext4_commit_super(sb, 1);
901
902 if (sbi->s_proc) {
903 remove_proc_entry("options", sbi->s_proc);
904 remove_proc_entry(sb->s_id, ext4_proc_root);
905 }
906 kobject_del(&sbi->s_kobj);
907
908 for (i = 0; i < sbi->s_gdb_count; i++)
909 brelse(sbi->s_group_desc[i]);
910 ext4_kvfree(sbi->s_group_desc);
911 ext4_kvfree(sbi->s_flex_groups);
912 percpu_counter_destroy(&sbi->s_freeclusters_counter);
913 percpu_counter_destroy(&sbi->s_freeinodes_counter);
914 percpu_counter_destroy(&sbi->s_dirs_counter);
915 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
916 brelse(sbi->s_sbh);
917#ifdef CONFIG_QUOTA
918 for (i = 0; i < MAXQUOTAS; i++)
919 kfree(sbi->s_qf_names[i]);
920#endif
921
922 /* Debugging code just in case the in-memory inode orphan list
923 * isn't empty. The on-disk one can be non-empty if we've
924 * detected an error and taken the fs readonly, but the
925 * in-memory list had better be clean by this point. */
926 if (!list_empty(&sbi->s_orphan))
927 dump_orphan_list(sb, sbi);
928 J_ASSERT(list_empty(&sbi->s_orphan));
929
930 invalidate_bdev(sb->s_bdev);
931 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
932 /*
933 * Invalidate the journal device's buffers. We don't want them
934 * floating about in memory - the physical journal device may
935 * hotswapped, and it breaks the `ro-after' testing code.
936 */
937 sync_blockdev(sbi->journal_bdev);
938 invalidate_bdev(sbi->journal_bdev);
939 ext4_blkdev_remove(sbi);
940 }
941 if (sbi->s_mmp_tsk)
942 kthread_stop(sbi->s_mmp_tsk);
943 sb->s_fs_info = NULL;
944 /*
945 * Now that we are completely done shutting down the
946 * superblock, we need to actually destroy the kobject.
947 */
948 unlock_super(sb);
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 kfree(sbi);
955}
956
957static struct kmem_cache *ext4_inode_cachep;
958
959/*
960 * Called inside transaction, so use GFP_NOFS
961 */
962static struct inode *ext4_alloc_inode(struct super_block *sb)
963{
964 struct ext4_inode_info *ei;
965
966 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
967 if (!ei)
968 return NULL;
969
970 ei->vfs_inode.i_version = 1;
971 ei->vfs_inode.i_data.writeback_index = 0;
972 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
973 INIT_LIST_HEAD(&ei->i_prealloc_list);
974 spin_lock_init(&ei->i_prealloc_lock);
975 ei->i_reserved_data_blocks = 0;
976 ei->i_reserved_meta_blocks = 0;
977 ei->i_allocated_meta_blocks = 0;
978 ei->i_da_metadata_calc_len = 0;
979 ei->i_da_metadata_calc_last_lblock = 0;
980 spin_lock_init(&(ei->i_block_reservation_lock));
981#ifdef CONFIG_QUOTA
982 ei->i_reserved_quota = 0;
983#endif
984 ei->jinode = NULL;
985 INIT_LIST_HEAD(&ei->i_completed_io_list);
986 spin_lock_init(&ei->i_completed_io_lock);
987 ei->cur_aio_dio = NULL;
988 ei->i_sync_tid = 0;
989 ei->i_datasync_tid = 0;
990 atomic_set(&ei->i_ioend_count, 0);
991 atomic_set(&ei->i_aiodio_unwritten, 0);
992
993 return &ei->vfs_inode;
994}
995
996static int ext4_drop_inode(struct inode *inode)
997{
998 int drop = generic_drop_inode(inode);
999
1000 trace_ext4_drop_inode(inode, drop);
1001 return drop;
1002}
1003
1004static void ext4_i_callback(struct rcu_head *head)
1005{
1006 struct inode *inode = container_of(head, struct inode, i_rcu);
1007 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
1008}
1009
1010static void ext4_destroy_inode(struct inode *inode)
1011{
1012 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
1013 ext4_msg(inode->i_sb, KERN_ERR,
1014 "Inode %lu (%p): orphan list check failed!",
1015 inode->i_ino, EXT4_I(inode));
1016 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
1017 EXT4_I(inode), sizeof(struct ext4_inode_info),
1018 true);
1019 dump_stack();
1020 }
1021 call_rcu(&inode->i_rcu, ext4_i_callback);
1022}
1023
1024static void init_once(void *foo)
1025{
1026 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
1027
1028 INIT_LIST_HEAD(&ei->i_orphan);
1029#ifdef CONFIG_EXT4_FS_XATTR
1030 init_rwsem(&ei->xattr_sem);
1031#endif
1032 init_rwsem(&ei->i_data_sem);
1033 inode_init_once(&ei->vfs_inode);
1034}
1035
1036static int init_inodecache(void)
1037{
1038 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
1039 sizeof(struct ext4_inode_info),
1040 0, (SLAB_RECLAIM_ACCOUNT|
1041 SLAB_MEM_SPREAD),
1042 init_once);
1043 if (ext4_inode_cachep == NULL)
1044 return -ENOMEM;
1045 return 0;
1046}
1047
1048static void destroy_inodecache(void)
1049{
1050 kmem_cache_destroy(ext4_inode_cachep);
1051}
1052
1053void ext4_clear_inode(struct inode *inode)
1054{
1055 invalidate_inode_buffers(inode);
1056 clear_inode(inode);
1057 dquot_drop(inode);
1058 ext4_discard_preallocations(inode);
1059 if (EXT4_I(inode)->jinode) {
1060 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1061 EXT4_I(inode)->jinode);
1062 jbd2_free_inode(EXT4_I(inode)->jinode);
1063 EXT4_I(inode)->jinode = NULL;
1064 }
1065}
1066
1067static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1068 u64 ino, u32 generation)
1069{
1070 struct inode *inode;
1071
1072 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
1073 return ERR_PTR(-ESTALE);
1074 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
1075 return ERR_PTR(-ESTALE);
1076
1077 /* iget isn't really right if the inode is currently unallocated!!
1078 *
1079 * ext4_read_inode will return a bad_inode if the inode had been
1080 * deleted, so we should be safe.
1081 *
1082 * Currently we don't know the generation for parent directory, so
1083 * a generation of 0 means "accept any"
1084 */
1085 inode = ext4_iget(sb, ino);
1086 if (IS_ERR(inode))
1087 return ERR_CAST(inode);
1088 if (generation && inode->i_generation != generation) {
1089 iput(inode);
1090 return ERR_PTR(-ESTALE);
1091 }
1092
1093 return inode;
1094}
1095
1096static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1097 int fh_len, int fh_type)
1098{
1099 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1100 ext4_nfs_get_inode);
1101}
1102
1103static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1104 int fh_len, int fh_type)
1105{
1106 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1107 ext4_nfs_get_inode);
1108}
1109
1110/*
1111 * Try to release metadata pages (indirect blocks, directories) which are
1112 * mapped via the block device. Since these pages could have journal heads
1113 * which would prevent try_to_free_buffers() from freeing them, we must use
1114 * jbd2 layer's try_to_free_buffers() function to release them.
1115 */
1116static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1117 gfp_t wait)
1118{
1119 journal_t *journal = EXT4_SB(sb)->s_journal;
1120
1121 WARN_ON(PageChecked(page));
1122 if (!page_has_buffers(page))
1123 return 0;
1124 if (journal)
1125 return jbd2_journal_try_to_free_buffers(journal, page,
1126 wait & ~__GFP_WAIT);
1127 return try_to_free_buffers(page);
1128}
1129
1130#ifdef CONFIG_QUOTA
1131#define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1132#define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1133
1134static int ext4_write_dquot(struct dquot *dquot);
1135static int ext4_acquire_dquot(struct dquot *dquot);
1136static int ext4_release_dquot(struct dquot *dquot);
1137static int ext4_mark_dquot_dirty(struct dquot *dquot);
1138static int ext4_write_info(struct super_block *sb, int type);
1139static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1140 struct path *path);
1141static int ext4_quota_off(struct super_block *sb, int type);
1142static int ext4_quota_on_mount(struct super_block *sb, int type);
1143static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1144 size_t len, loff_t off);
1145static ssize_t ext4_quota_write(struct super_block *sb, int type,
1146 const char *data, size_t len, loff_t off);
1147
1148static const struct dquot_operations ext4_quota_operations = {
1149 .get_reserved_space = ext4_get_reserved_space,
1150 .write_dquot = ext4_write_dquot,
1151 .acquire_dquot = ext4_acquire_dquot,
1152 .release_dquot = ext4_release_dquot,
1153 .mark_dirty = ext4_mark_dquot_dirty,
1154 .write_info = ext4_write_info,
1155 .alloc_dquot = dquot_alloc,
1156 .destroy_dquot = dquot_destroy,
1157};
1158
1159static const struct quotactl_ops ext4_qctl_operations = {
1160 .quota_on = ext4_quota_on,
1161 .quota_off = ext4_quota_off,
1162 .quota_sync = dquot_quota_sync,
1163 .get_info = dquot_get_dqinfo,
1164 .set_info = dquot_set_dqinfo,
1165 .get_dqblk = dquot_get_dqblk,
1166 .set_dqblk = dquot_set_dqblk
1167};
1168#endif
1169
1170static const struct super_operations ext4_sops = {
1171 .alloc_inode = ext4_alloc_inode,
1172 .destroy_inode = ext4_destroy_inode,
1173 .write_inode = ext4_write_inode,
1174 .dirty_inode = ext4_dirty_inode,
1175 .drop_inode = ext4_drop_inode,
1176 .evict_inode = ext4_evict_inode,
1177 .put_super = ext4_put_super,
1178 .sync_fs = ext4_sync_fs,
1179 .freeze_fs = ext4_freeze,
1180 .unfreeze_fs = ext4_unfreeze,
1181 .statfs = ext4_statfs,
1182 .remount_fs = ext4_remount,
1183 .show_options = ext4_show_options,
1184#ifdef CONFIG_QUOTA
1185 .quota_read = ext4_quota_read,
1186 .quota_write = ext4_quota_write,
1187#endif
1188 .bdev_try_to_free_page = bdev_try_to_free_page,
1189};
1190
1191static const struct super_operations ext4_nojournal_sops = {
1192 .alloc_inode = ext4_alloc_inode,
1193 .destroy_inode = ext4_destroy_inode,
1194 .write_inode = ext4_write_inode,
1195 .dirty_inode = ext4_dirty_inode,
1196 .drop_inode = ext4_drop_inode,
1197 .evict_inode = ext4_evict_inode,
1198 .write_super = ext4_write_super,
1199 .put_super = ext4_put_super,
1200 .statfs = ext4_statfs,
1201 .remount_fs = ext4_remount,
1202 .show_options = ext4_show_options,
1203#ifdef CONFIG_QUOTA
1204 .quota_read = ext4_quota_read,
1205 .quota_write = ext4_quota_write,
1206#endif
1207 .bdev_try_to_free_page = bdev_try_to_free_page,
1208};
1209
1210static const struct export_operations ext4_export_ops = {
1211 .fh_to_dentry = ext4_fh_to_dentry,
1212 .fh_to_parent = ext4_fh_to_parent,
1213 .get_parent = ext4_get_parent,
1214};
1215
1216enum {
1217 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1218 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1219 Opt_nouid32, Opt_debug, Opt_removed,
1220 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1221 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1222 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
1223 Opt_journal_dev, Opt_journal_checksum, Opt_journal_async_commit,
1224 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1225 Opt_data_err_abort, Opt_data_err_ignore,
1226 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1227 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1228 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1229 Opt_usrquota, Opt_grpquota, Opt_i_version,
1230 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1231 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1232 Opt_inode_readahead_blks, Opt_journal_ioprio,
1233 Opt_dioread_nolock, Opt_dioread_lock,
1234 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1235};
1236
1237static const match_table_t tokens = {
1238 {Opt_bsd_df, "bsddf"},
1239 {Opt_minix_df, "minixdf"},
1240 {Opt_grpid, "grpid"},
1241 {Opt_grpid, "bsdgroups"},
1242 {Opt_nogrpid, "nogrpid"},
1243 {Opt_nogrpid, "sysvgroups"},
1244 {Opt_resgid, "resgid=%u"},
1245 {Opt_resuid, "resuid=%u"},
1246 {Opt_sb, "sb=%u"},
1247 {Opt_err_cont, "errors=continue"},
1248 {Opt_err_panic, "errors=panic"},
1249 {Opt_err_ro, "errors=remount-ro"},
1250 {Opt_nouid32, "nouid32"},
1251 {Opt_debug, "debug"},
1252 {Opt_removed, "oldalloc"},
1253 {Opt_removed, "orlov"},
1254 {Opt_user_xattr, "user_xattr"},
1255 {Opt_nouser_xattr, "nouser_xattr"},
1256 {Opt_acl, "acl"},
1257 {Opt_noacl, "noacl"},
1258 {Opt_noload, "norecovery"},
1259 {Opt_noload, "noload"},
1260 {Opt_removed, "nobh"},
1261 {Opt_removed, "bh"},
1262 {Opt_commit, "commit=%u"},
1263 {Opt_min_batch_time, "min_batch_time=%u"},
1264 {Opt_max_batch_time, "max_batch_time=%u"},
1265 {Opt_journal_dev, "journal_dev=%u"},
1266 {Opt_journal_checksum, "journal_checksum"},
1267 {Opt_journal_async_commit, "journal_async_commit"},
1268 {Opt_abort, "abort"},
1269 {Opt_data_journal, "data=journal"},
1270 {Opt_data_ordered, "data=ordered"},
1271 {Opt_data_writeback, "data=writeback"},
1272 {Opt_data_err_abort, "data_err=abort"},
1273 {Opt_data_err_ignore, "data_err=ignore"},
1274 {Opt_offusrjquota, "usrjquota="},
1275 {Opt_usrjquota, "usrjquota=%s"},
1276 {Opt_offgrpjquota, "grpjquota="},
1277 {Opt_grpjquota, "grpjquota=%s"},
1278 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1279 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1280 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1281 {Opt_grpquota, "grpquota"},
1282 {Opt_noquota, "noquota"},
1283 {Opt_quota, "quota"},
1284 {Opt_usrquota, "usrquota"},
1285 {Opt_barrier, "barrier=%u"},
1286 {Opt_barrier, "barrier"},
1287 {Opt_nobarrier, "nobarrier"},
1288 {Opt_i_version, "i_version"},
1289 {Opt_stripe, "stripe=%u"},
1290 {Opt_delalloc, "delalloc"},
1291 {Opt_nodelalloc, "nodelalloc"},
1292 {Opt_mblk_io_submit, "mblk_io_submit"},
1293 {Opt_nomblk_io_submit, "nomblk_io_submit"},
1294 {Opt_block_validity, "block_validity"},
1295 {Opt_noblock_validity, "noblock_validity"},
1296 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1297 {Opt_journal_ioprio, "journal_ioprio=%u"},
1298 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1299 {Opt_auto_da_alloc, "auto_da_alloc"},
1300 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1301 {Opt_dioread_nolock, "dioread_nolock"},
1302 {Opt_dioread_lock, "dioread_lock"},
1303 {Opt_discard, "discard"},
1304 {Opt_nodiscard, "nodiscard"},
1305 {Opt_init_itable, "init_itable=%u"},
1306 {Opt_init_itable, "init_itable"},
1307 {Opt_noinit_itable, "noinit_itable"},
1308 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1309 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1310 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1311 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1312 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
1313 {Opt_err, NULL},
1314};
1315
1316static ext4_fsblk_t get_sb_block(void **data)
1317{
1318 ext4_fsblk_t sb_block;
1319 char *options = (char *) *data;
1320
1321 if (!options || strncmp(options, "sb=", 3) != 0)
1322 return 1; /* Default location */
1323
1324 options += 3;
1325 /* TODO: use simple_strtoll with >32bit ext4 */
1326 sb_block = simple_strtoul(options, &options, 0);
1327 if (*options && *options != ',') {
1328 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1329 (char *) *data);
1330 return 1;
1331 }
1332 if (*options == ',')
1333 options++;
1334 *data = (void *) options;
1335
1336 return sb_block;
1337}
1338
1339#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1340static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1341 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1342
1343#ifdef CONFIG_QUOTA
1344static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1345{
1346 struct ext4_sb_info *sbi = EXT4_SB(sb);
1347 char *qname;
1348
1349 if (sb_any_quota_loaded(sb) &&
1350 !sbi->s_qf_names[qtype]) {
1351 ext4_msg(sb, KERN_ERR,
1352 "Cannot change journaled "
1353 "quota options when quota turned on");
1354 return -1;
1355 }
1356 qname = match_strdup(args);
1357 if (!qname) {
1358 ext4_msg(sb, KERN_ERR,
1359 "Not enough memory for storing quotafile name");
1360 return -1;
1361 }
1362 if (sbi->s_qf_names[qtype] &&
1363 strcmp(sbi->s_qf_names[qtype], qname)) {
1364 ext4_msg(sb, KERN_ERR,
1365 "%s quota file already specified", QTYPE2NAME(qtype));
1366 kfree(qname);
1367 return -1;
1368 }
1369 sbi->s_qf_names[qtype] = qname;
1370 if (strchr(sbi->s_qf_names[qtype], '/')) {
1371 ext4_msg(sb, KERN_ERR,
1372 "quotafile must be on filesystem root");
1373 kfree(sbi->s_qf_names[qtype]);
1374 sbi->s_qf_names[qtype] = NULL;
1375 return -1;
1376 }
1377 set_opt(sb, QUOTA);
1378 return 1;
1379}
1380
1381static int clear_qf_name(struct super_block *sb, int qtype)
1382{
1383
1384 struct ext4_sb_info *sbi = EXT4_SB(sb);
1385
1386 if (sb_any_quota_loaded(sb) &&
1387 sbi->s_qf_names[qtype]) {
1388 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1389 " when quota turned on");
1390 return -1;
1391 }
1392 /*
1393 * The space will be released later when all options are confirmed
1394 * to be correct
1395 */
1396 sbi->s_qf_names[qtype] = NULL;
1397 return 1;
1398}
1399#endif
1400
1401#define MOPT_SET 0x0001
1402#define MOPT_CLEAR 0x0002
1403#define MOPT_NOSUPPORT 0x0004
1404#define MOPT_EXPLICIT 0x0008
1405#define MOPT_CLEAR_ERR 0x0010
1406#define MOPT_GTE0 0x0020
1407#ifdef CONFIG_QUOTA
1408#define MOPT_Q 0
1409#define MOPT_QFMT 0x0040
1410#else
1411#define MOPT_Q MOPT_NOSUPPORT
1412#define MOPT_QFMT MOPT_NOSUPPORT
1413#endif
1414#define MOPT_DATAJ 0x0080
1415
1416static const struct mount_opts {
1417 int token;
1418 int mount_opt;
1419 int flags;
1420} ext4_mount_opts[] = {
1421 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1422 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1423 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1424 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1425 {Opt_mblk_io_submit, EXT4_MOUNT_MBLK_IO_SUBMIT, MOPT_SET},
1426 {Opt_nomblk_io_submit, EXT4_MOUNT_MBLK_IO_SUBMIT, MOPT_CLEAR},
1427 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1428 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1429 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK, MOPT_SET},
1430 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK, MOPT_CLEAR},
1431 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1432 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1433 {Opt_delalloc, EXT4_MOUNT_DELALLOC, MOPT_SET | MOPT_EXPLICIT},
1434 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC, MOPT_CLEAR | MOPT_EXPLICIT},
1435 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM, MOPT_SET},
1436 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1437 EXT4_MOUNT_JOURNAL_CHECKSUM), MOPT_SET},
1438 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_SET},
1439 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1440 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1441 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1442 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT, MOPT_SET},
1443 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT, MOPT_CLEAR},
1444 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1445 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1446 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1447 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1448 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1449 {Opt_commit, 0, MOPT_GTE0},
1450 {Opt_max_batch_time, 0, MOPT_GTE0},
1451 {Opt_min_batch_time, 0, MOPT_GTE0},
1452 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1453 {Opt_init_itable, 0, MOPT_GTE0},
1454 {Opt_stripe, 0, MOPT_GTE0},
1455 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_DATAJ},
1456 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_DATAJ},
1457 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA, MOPT_DATAJ},
1458#ifdef CONFIG_EXT4_FS_XATTR
1459 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1460 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1461#else
1462 {Opt_user_xattr, 0, MOPT_NOSUPPORT},
1463 {Opt_nouser_xattr, 0, MOPT_NOSUPPORT},
1464#endif
1465#ifdef CONFIG_EXT4_FS_POSIX_ACL
1466 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1467 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1468#else
1469 {Opt_acl, 0, MOPT_NOSUPPORT},
1470 {Opt_noacl, 0, MOPT_NOSUPPORT},
1471#endif
1472 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1473 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1474 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1475 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1476 MOPT_SET | MOPT_Q},
1477 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1478 MOPT_SET | MOPT_Q},
1479 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1480 EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1481 {Opt_usrjquota, 0, MOPT_Q},
1482 {Opt_grpjquota, 0, MOPT_Q},
1483 {Opt_offusrjquota, 0, MOPT_Q},
1484 {Opt_offgrpjquota, 0, MOPT_Q},
1485 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1486 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1487 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1488 {Opt_err, 0, 0}
1489};
1490
1491static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1492 substring_t *args, unsigned long *journal_devnum,
1493 unsigned int *journal_ioprio, int is_remount)
1494{
1495 struct ext4_sb_info *sbi = EXT4_SB(sb);
1496 const struct mount_opts *m;
1497 kuid_t uid;
1498 kgid_t gid;
1499 int arg = 0;
1500
1501#ifdef CONFIG_QUOTA
1502 if (token == Opt_usrjquota)
1503 return set_qf_name(sb, USRQUOTA, &args[0]);
1504 else if (token == Opt_grpjquota)
1505 return set_qf_name(sb, GRPQUOTA, &args[0]);
1506 else if (token == Opt_offusrjquota)
1507 return clear_qf_name(sb, USRQUOTA);
1508 else if (token == Opt_offgrpjquota)
1509 return clear_qf_name(sb, GRPQUOTA);
1510#endif
1511 if (args->from && match_int(args, &arg))
1512 return -1;
1513 switch (token) {
1514 case Opt_noacl:
1515 case Opt_nouser_xattr:
1516 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1517 break;
1518 case Opt_sb:
1519 return 1; /* handled by get_sb_block() */
1520 case Opt_removed:
1521 ext4_msg(sb, KERN_WARNING,
1522 "Ignoring removed %s option", opt);
1523 return 1;
1524 case Opt_resuid:
1525 uid = make_kuid(current_user_ns(), arg);
1526 if (!uid_valid(uid)) {
1527 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1528 return -1;
1529 }
1530 sbi->s_resuid = uid;
1531 return 1;
1532 case Opt_resgid:
1533 gid = make_kgid(current_user_ns(), arg);
1534 if (!gid_valid(gid)) {
1535 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1536 return -1;
1537 }
1538 sbi->s_resgid = gid;
1539 return 1;
1540 case Opt_abort:
1541 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1542 return 1;
1543 case Opt_i_version:
1544 sb->s_flags |= MS_I_VERSION;
1545 return 1;
1546 case Opt_journal_dev:
1547 if (is_remount) {
1548 ext4_msg(sb, KERN_ERR,
1549 "Cannot specify journal on remount");
1550 return -1;
1551 }
1552 *journal_devnum = arg;
1553 return 1;
1554 case Opt_journal_ioprio:
1555 if (arg < 0 || arg > 7)
1556 return -1;
1557 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1558 return 1;
1559 }
1560
1561 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1562 if (token != m->token)
1563 continue;
1564 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1565 return -1;
1566 if (m->flags & MOPT_EXPLICIT)
1567 set_opt2(sb, EXPLICIT_DELALLOC);
1568 if (m->flags & MOPT_CLEAR_ERR)
1569 clear_opt(sb, ERRORS_MASK);
1570 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1571 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1572 "options when quota turned on");
1573 return -1;
1574 }
1575
1576 if (m->flags & MOPT_NOSUPPORT) {
1577 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1578 } else if (token == Opt_commit) {
1579 if (arg == 0)
1580 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1581 sbi->s_commit_interval = HZ * arg;
1582 } else if (token == Opt_max_batch_time) {
1583 if (arg == 0)
1584 arg = EXT4_DEF_MAX_BATCH_TIME;
1585 sbi->s_max_batch_time = arg;
1586 } else if (token == Opt_min_batch_time) {
1587 sbi->s_min_batch_time = arg;
1588 } else if (token == Opt_inode_readahead_blks) {
1589 if (arg > (1 << 30))
1590 return -1;
1591 if (arg && !is_power_of_2(arg)) {
1592 ext4_msg(sb, KERN_ERR,
1593 "EXT4-fs: inode_readahead_blks"
1594 " must be a power of 2");
1595 return -1;
1596 }
1597 sbi->s_inode_readahead_blks = arg;
1598 } else if (token == Opt_init_itable) {
1599 set_opt(sb, INIT_INODE_TABLE);
1600 if (!args->from)
1601 arg = EXT4_DEF_LI_WAIT_MULT;
1602 sbi->s_li_wait_mult = arg;
1603 } else if (token == Opt_stripe) {
1604 sbi->s_stripe = arg;
1605 } else if (m->flags & MOPT_DATAJ) {
1606 if (is_remount) {
1607 if (!sbi->s_journal)
1608 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1609 else if (test_opt(sb, DATA_FLAGS) !=
1610 m->mount_opt) {
1611 ext4_msg(sb, KERN_ERR,
1612 "Cannot change data mode on remount");
1613 return -1;
1614 }
1615 } else {
1616 clear_opt(sb, DATA_FLAGS);
1617 sbi->s_mount_opt |= m->mount_opt;
1618 }
1619#ifdef CONFIG_QUOTA
1620 } else if (m->flags & MOPT_QFMT) {
1621 if (sb_any_quota_loaded(sb) &&
1622 sbi->s_jquota_fmt != m->mount_opt) {
1623 ext4_msg(sb, KERN_ERR, "Cannot "
1624 "change journaled quota options "
1625 "when quota turned on");
1626 return -1;
1627 }
1628 sbi->s_jquota_fmt = m->mount_opt;
1629#endif
1630 } else {
1631 if (!args->from)
1632 arg = 1;
1633 if (m->flags & MOPT_CLEAR)
1634 arg = !arg;
1635 else if (unlikely(!(m->flags & MOPT_SET))) {
1636 ext4_msg(sb, KERN_WARNING,
1637 "buggy handling of option %s", opt);
1638 WARN_ON(1);
1639 return -1;
1640 }
1641 if (arg != 0)
1642 sbi->s_mount_opt |= m->mount_opt;
1643 else
1644 sbi->s_mount_opt &= ~m->mount_opt;
1645 }
1646 return 1;
1647 }
1648 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1649 "or missing value", opt);
1650 return -1;
1651}
1652
1653static int parse_options(char *options, struct super_block *sb,
1654 unsigned long *journal_devnum,
1655 unsigned int *journal_ioprio,
1656 int is_remount)
1657{
1658#ifdef CONFIG_QUOTA
1659 struct ext4_sb_info *sbi = EXT4_SB(sb);
1660#endif
1661 char *p;
1662 substring_t args[MAX_OPT_ARGS];
1663 int token;
1664
1665 if (!options)
1666 return 1;
1667
1668 while ((p = strsep(&options, ",")) != NULL) {
1669 if (!*p)
1670 continue;
1671 /*
1672 * Initialize args struct so we know whether arg was
1673 * found; some options take optional arguments.
1674 */
1675 args[0].to = args[0].from = 0;
1676 token = match_token(p, tokens, args);
1677 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1678 journal_ioprio, is_remount) < 0)
1679 return 0;
1680 }
1681#ifdef CONFIG_QUOTA
1682 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1683 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1684 clear_opt(sb, USRQUOTA);
1685
1686 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1687 clear_opt(sb, GRPQUOTA);
1688
1689 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1690 ext4_msg(sb, KERN_ERR, "old and new quota "
1691 "format mixing");
1692 return 0;
1693 }
1694
1695 if (!sbi->s_jquota_fmt) {
1696 ext4_msg(sb, KERN_ERR, "journaled quota format "
1697 "not specified");
1698 return 0;
1699 }
1700 } else {
1701 if (sbi->s_jquota_fmt) {
1702 ext4_msg(sb, KERN_ERR, "journaled quota format "
1703 "specified with no journaling "
1704 "enabled");
1705 return 0;
1706 }
1707 }
1708#endif
1709 return 1;
1710}
1711
1712static inline void ext4_show_quota_options(struct seq_file *seq,
1713 struct super_block *sb)
1714{
1715#if defined(CONFIG_QUOTA)
1716 struct ext4_sb_info *sbi = EXT4_SB(sb);
1717
1718 if (sbi->s_jquota_fmt) {
1719 char *fmtname = "";
1720
1721 switch (sbi->s_jquota_fmt) {
1722 case QFMT_VFS_OLD:
1723 fmtname = "vfsold";
1724 break;
1725 case QFMT_VFS_V0:
1726 fmtname = "vfsv0";
1727 break;
1728 case QFMT_VFS_V1:
1729 fmtname = "vfsv1";
1730 break;
1731 }
1732 seq_printf(seq, ",jqfmt=%s", fmtname);
1733 }
1734
1735 if (sbi->s_qf_names[USRQUOTA])
1736 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1737
1738 if (sbi->s_qf_names[GRPQUOTA])
1739 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
1740
1741 if (test_opt(sb, USRQUOTA))
1742 seq_puts(seq, ",usrquota");
1743
1744 if (test_opt(sb, GRPQUOTA))
1745 seq_puts(seq, ",grpquota");
1746#endif
1747}
1748
1749static const char *token2str(int token)
1750{
1751 static const struct match_token *t;
1752
1753 for (t = tokens; t->token != Opt_err; t++)
1754 if (t->token == token && !strchr(t->pattern, '='))
1755 break;
1756 return t->pattern;
1757}
1758
1759/*
1760 * Show an option if
1761 * - it's set to a non-default value OR
1762 * - if the per-sb default is different from the global default
1763 */
1764static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1765 int nodefs)
1766{
1767 struct ext4_sb_info *sbi = EXT4_SB(sb);
1768 struct ext4_super_block *es = sbi->s_es;
1769 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
1770 const struct mount_opts *m;
1771 char sep = nodefs ? '\n' : ',';
1772
1773#define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1774#define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
1775
1776 if (sbi->s_sb_block != 1)
1777 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1778
1779 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1780 int want_set = m->flags & MOPT_SET;
1781 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1782 (m->flags & MOPT_CLEAR_ERR))
1783 continue;
1784 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
1785 continue; /* skip if same as the default */
1786 if ((want_set &&
1787 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1788 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1789 continue; /* select Opt_noFoo vs Opt_Foo */
1790 SEQ_OPTS_PRINT("%s", token2str(m->token));
1791 }
1792
1793 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
1794 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
1795 SEQ_OPTS_PRINT("resuid=%u",
1796 from_kuid_munged(&init_user_ns, sbi->s_resuid));
1797 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
1798 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
1799 SEQ_OPTS_PRINT("resgid=%u",
1800 from_kgid_munged(&init_user_ns, sbi->s_resgid));
1801 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
1802 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1803 SEQ_OPTS_PUTS("errors=remount-ro");
1804 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
1805 SEQ_OPTS_PUTS("errors=continue");
1806 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
1807 SEQ_OPTS_PUTS("errors=panic");
1808 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
1809 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
1810 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
1811 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
1812 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
1813 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
1814 if (sb->s_flags & MS_I_VERSION)
1815 SEQ_OPTS_PUTS("i_version");
1816 if (nodefs || sbi->s_stripe)
1817 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
1818 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
1819 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1820 SEQ_OPTS_PUTS("data=journal");
1821 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1822 SEQ_OPTS_PUTS("data=ordered");
1823 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1824 SEQ_OPTS_PUTS("data=writeback");
1825 }
1826 if (nodefs ||
1827 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1828 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1829 sbi->s_inode_readahead_blks);
1830
1831 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1832 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
1833 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
1834
1835 ext4_show_quota_options(seq, sb);
1836 return 0;
1837}
1838
1839static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1840{
1841 return _ext4_show_options(seq, root->d_sb, 0);
1842}
1843
1844static int options_seq_show(struct seq_file *seq, void *offset)
1845{
1846 struct super_block *sb = seq->private;
1847 int rc;
1848
1849 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1850 rc = _ext4_show_options(seq, sb, 1);
1851 seq_puts(seq, "\n");
1852 return rc;
1853}
1854
1855static int options_open_fs(struct inode *inode, struct file *file)
1856{
1857 return single_open(file, options_seq_show, PDE(inode)->data);
1858}
1859
1860static const struct file_operations ext4_seq_options_fops = {
1861 .owner = THIS_MODULE,
1862 .open = options_open_fs,
1863 .read = seq_read,
1864 .llseek = seq_lseek,
1865 .release = single_release,
1866};
1867
1868static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1869 int read_only)
1870{
1871 struct ext4_sb_info *sbi = EXT4_SB(sb);
1872 int res = 0;
1873
1874 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1875 ext4_msg(sb, KERN_ERR, "revision level too high, "
1876 "forcing read-only mode");
1877 res = MS_RDONLY;
1878 }
1879 if (read_only)
1880 goto done;
1881 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1882 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1883 "running e2fsck is recommended");
1884 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1885 ext4_msg(sb, KERN_WARNING,
1886 "warning: mounting fs with errors, "
1887 "running e2fsck is recommended");
1888 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
1889 le16_to_cpu(es->s_mnt_count) >=
1890 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1891 ext4_msg(sb, KERN_WARNING,
1892 "warning: maximal mount count reached, "
1893 "running e2fsck is recommended");
1894 else if (le32_to_cpu(es->s_checkinterval) &&
1895 (le32_to_cpu(es->s_lastcheck) +
1896 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1897 ext4_msg(sb, KERN_WARNING,
1898 "warning: checktime reached, "
1899 "running e2fsck is recommended");
1900 if (!sbi->s_journal)
1901 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1902 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1903 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1904 le16_add_cpu(&es->s_mnt_count, 1);
1905 es->s_mtime = cpu_to_le32(get_seconds());
1906 ext4_update_dynamic_rev(sb);
1907 if (sbi->s_journal)
1908 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1909
1910 ext4_commit_super(sb, 1);
1911done:
1912 if (test_opt(sb, DEBUG))
1913 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1914 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
1915 sb->s_blocksize,
1916 sbi->s_groups_count,
1917 EXT4_BLOCKS_PER_GROUP(sb),
1918 EXT4_INODES_PER_GROUP(sb),
1919 sbi->s_mount_opt, sbi->s_mount_opt2);
1920
1921 cleancache_init_fs(sb);
1922 return res;
1923}
1924
1925static int ext4_fill_flex_info(struct super_block *sb)
1926{
1927 struct ext4_sb_info *sbi = EXT4_SB(sb);
1928 struct ext4_group_desc *gdp = NULL;
1929 ext4_group_t flex_group_count;
1930 ext4_group_t flex_group;
1931 unsigned int groups_per_flex = 0;
1932 size_t size;
1933 int i;
1934
1935 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1936 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
1937 sbi->s_log_groups_per_flex = 0;
1938 return 1;
1939 }
1940 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1941
1942 /* We allocate both existing and potentially added groups */
1943 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
1944 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1945 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
1946 size = flex_group_count * sizeof(struct flex_groups);
1947 sbi->s_flex_groups = ext4_kvzalloc(size, GFP_KERNEL);
1948 if (sbi->s_flex_groups == NULL) {
1949 ext4_msg(sb, KERN_ERR, "not enough memory for %u flex groups",
1950 flex_group_count);
1951 goto failed;
1952 }
1953
1954 for (i = 0; i < sbi->s_groups_count; i++) {
1955 gdp = ext4_get_group_desc(sb, i, NULL);
1956
1957 flex_group = ext4_flex_group(sbi, i);
1958 atomic_add(ext4_free_inodes_count(sb, gdp),
1959 &sbi->s_flex_groups[flex_group].free_inodes);
1960 atomic_add(ext4_free_group_clusters(sb, gdp),
1961 &sbi->s_flex_groups[flex_group].free_clusters);
1962 atomic_add(ext4_used_dirs_count(sb, gdp),
1963 &sbi->s_flex_groups[flex_group].used_dirs);
1964 }
1965
1966 return 1;
1967failed:
1968 return 0;
1969}
1970
1971static __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1972 struct ext4_group_desc *gdp)
1973{
1974 int offset;
1975 __u16 crc = 0;
1976 __le32 le_group = cpu_to_le32(block_group);
1977
1978 if ((sbi->s_es->s_feature_ro_compat &
1979 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))) {
1980 /* Use new metadata_csum algorithm */
1981 __u16 old_csum;
1982 __u32 csum32;
1983
1984 old_csum = gdp->bg_checksum;
1985 gdp->bg_checksum = 0;
1986 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
1987 sizeof(le_group));
1988 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp,
1989 sbi->s_desc_size);
1990 gdp->bg_checksum = old_csum;
1991
1992 crc = csum32 & 0xFFFF;
1993 goto out;
1994 }
1995
1996 /* old crc16 code */
1997 offset = offsetof(struct ext4_group_desc, bg_checksum);
1998
1999 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2000 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2001 crc = crc16(crc, (__u8 *)gdp, offset);
2002 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2003 /* for checksum of struct ext4_group_desc do the rest...*/
2004 if ((sbi->s_es->s_feature_incompat &
2005 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
2006 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2007 crc = crc16(crc, (__u8 *)gdp + offset,
2008 le16_to_cpu(sbi->s_es->s_desc_size) -
2009 offset);
2010
2011out:
2012 return cpu_to_le16(crc);
2013}
2014
2015int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2016 struct ext4_group_desc *gdp)
2017{
2018 if (ext4_has_group_desc_csum(sb) &&
2019 (gdp->bg_checksum != ext4_group_desc_csum(EXT4_SB(sb),
2020 block_group, gdp)))
2021 return 0;
2022
2023 return 1;
2024}
2025
2026void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2027 struct ext4_group_desc *gdp)
2028{
2029 if (!ext4_has_group_desc_csum(sb))
2030 return;
2031 gdp->bg_checksum = ext4_group_desc_csum(EXT4_SB(sb), block_group, gdp);
2032}
2033
2034/* Called at mount-time, super-block is locked */
2035static int ext4_check_descriptors(struct super_block *sb,
2036 ext4_group_t *first_not_zeroed)
2037{
2038 struct ext4_sb_info *sbi = EXT4_SB(sb);
2039 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2040 ext4_fsblk_t last_block;
2041 ext4_fsblk_t block_bitmap;
2042 ext4_fsblk_t inode_bitmap;
2043 ext4_fsblk_t inode_table;
2044 int flexbg_flag = 0;
2045 ext4_group_t i, grp = sbi->s_groups_count;
2046
2047 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2048 flexbg_flag = 1;
2049
2050 ext4_debug("Checking group descriptors");
2051
2052 for (i = 0; i < sbi->s_groups_count; i++) {
2053 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2054
2055 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2056 last_block = ext4_blocks_count(sbi->s_es) - 1;
2057 else
2058 last_block = first_block +
2059 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2060
2061 if ((grp == sbi->s_groups_count) &&
2062 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2063 grp = i;
2064
2065 block_bitmap = ext4_block_bitmap(sb, gdp);
2066 if (block_bitmap < first_block || block_bitmap > last_block) {
2067 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2068 "Block bitmap for group %u not in group "
2069 "(block %llu)!", i, block_bitmap);
2070 return 0;
2071 }
2072 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2073 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2074 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2075 "Inode bitmap for group %u not in group "
2076 "(block %llu)!", i, inode_bitmap);
2077 return 0;
2078 }
2079 inode_table = ext4_inode_table(sb, gdp);
2080 if (inode_table < first_block ||
2081 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2082 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2083 "Inode table for group %u not in group "
2084 "(block %llu)!", i, inode_table);
2085 return 0;
2086 }
2087 ext4_lock_group(sb, i);
2088 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2089 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2090 "Checksum for group %u failed (%u!=%u)",
2091 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2092 gdp)), le16_to_cpu(gdp->bg_checksum));
2093 if (!(sb->s_flags & MS_RDONLY)) {
2094 ext4_unlock_group(sb, i);
2095 return 0;
2096 }
2097 }
2098 ext4_unlock_group(sb, i);
2099 if (!flexbg_flag)
2100 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2101 }
2102 if (NULL != first_not_zeroed)
2103 *first_not_zeroed = grp;
2104
2105 ext4_free_blocks_count_set(sbi->s_es,
2106 EXT4_C2B(sbi, ext4_count_free_clusters(sb)));
2107 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
2108 return 1;
2109}
2110
2111/* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2112 * the superblock) which were deleted from all directories, but held open by
2113 * a process at the time of a crash. We walk the list and try to delete these
2114 * inodes at recovery time (only with a read-write filesystem).
2115 *
2116 * In order to keep the orphan inode chain consistent during traversal (in
2117 * case of crash during recovery), we link each inode into the superblock
2118 * orphan list_head and handle it the same way as an inode deletion during
2119 * normal operation (which journals the operations for us).
2120 *
2121 * We only do an iget() and an iput() on each inode, which is very safe if we
2122 * accidentally point at an in-use or already deleted inode. The worst that
2123 * can happen in this case is that we get a "bit already cleared" message from
2124 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2125 * e2fsck was run on this filesystem, and it must have already done the orphan
2126 * inode cleanup for us, so we can safely abort without any further action.
2127 */
2128static void ext4_orphan_cleanup(struct super_block *sb,
2129 struct ext4_super_block *es)
2130{
2131 unsigned int s_flags = sb->s_flags;
2132 int nr_orphans = 0, nr_truncates = 0;
2133#ifdef CONFIG_QUOTA
2134 int i;
2135#endif
2136 if (!es->s_last_orphan) {
2137 jbd_debug(4, "no orphan inodes to clean up\n");
2138 return;
2139 }
2140
2141 if (bdev_read_only(sb->s_bdev)) {
2142 ext4_msg(sb, KERN_ERR, "write access "
2143 "unavailable, skipping orphan cleanup");
2144 return;
2145 }
2146
2147 /* Check if feature set would not allow a r/w mount */
2148 if (!ext4_feature_set_ok(sb, 0)) {
2149 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2150 "unknown ROCOMPAT features");
2151 return;
2152 }
2153
2154 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2155 if (es->s_last_orphan)
2156 jbd_debug(1, "Errors on filesystem, "
2157 "clearing orphan list.\n");
2158 es->s_last_orphan = 0;
2159 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2160 return;
2161 }
2162
2163 if (s_flags & MS_RDONLY) {
2164 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2165 sb->s_flags &= ~MS_RDONLY;
2166 }
2167#ifdef CONFIG_QUOTA
2168 /* Needed for iput() to work correctly and not trash data */
2169 sb->s_flags |= MS_ACTIVE;
2170 /* Turn on quotas so that they are updated correctly */
2171 for (i = 0; i < MAXQUOTAS; i++) {
2172 if (EXT4_SB(sb)->s_qf_names[i]) {
2173 int ret = ext4_quota_on_mount(sb, i);
2174 if (ret < 0)
2175 ext4_msg(sb, KERN_ERR,
2176 "Cannot turn on journaled "
2177 "quota: error %d", ret);
2178 }
2179 }
2180#endif
2181
2182 while (es->s_last_orphan) {
2183 struct inode *inode;
2184
2185 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2186 if (IS_ERR(inode)) {
2187 es->s_last_orphan = 0;
2188 break;
2189 }
2190
2191 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2192 dquot_initialize(inode);
2193 if (inode->i_nlink) {
2194 ext4_msg(sb, KERN_DEBUG,
2195 "%s: truncating inode %lu to %lld bytes",
2196 __func__, inode->i_ino, inode->i_size);
2197 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2198 inode->i_ino, inode->i_size);
2199 ext4_truncate(inode);
2200 nr_truncates++;
2201 } else {
2202 ext4_msg(sb, KERN_DEBUG,
2203 "%s: deleting unreferenced inode %lu",
2204 __func__, inode->i_ino);
2205 jbd_debug(2, "deleting unreferenced inode %lu\n",
2206 inode->i_ino);
2207 nr_orphans++;
2208 }
2209 iput(inode); /* The delete magic happens here! */
2210 }
2211
2212#define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2213
2214 if (nr_orphans)
2215 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2216 PLURAL(nr_orphans));
2217 if (nr_truncates)
2218 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2219 PLURAL(nr_truncates));
2220#ifdef CONFIG_QUOTA
2221 /* Turn quotas off */
2222 for (i = 0; i < MAXQUOTAS; i++) {
2223 if (sb_dqopt(sb)->files[i])
2224 dquot_quota_off(sb, i);
2225 }
2226#endif
2227 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2228}
2229
2230/*
2231 * Maximal extent format file size.
2232 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2233 * extent format containers, within a sector_t, and within i_blocks
2234 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2235 * so that won't be a limiting factor.
2236 *
2237 * However there is other limiting factor. We do store extents in the form
2238 * of starting block and length, hence the resulting length of the extent
2239 * covering maximum file size must fit into on-disk format containers as
2240 * well. Given that length is always by 1 unit bigger than max unit (because
2241 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2242 *
2243 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2244 */
2245static loff_t ext4_max_size(int blkbits, int has_huge_files)
2246{
2247 loff_t res;
2248 loff_t upper_limit = MAX_LFS_FILESIZE;
2249
2250 /* small i_blocks in vfs inode? */
2251 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2252 /*
2253 * CONFIG_LBDAF is not enabled implies the inode
2254 * i_block represent total blocks in 512 bytes
2255 * 32 == size of vfs inode i_blocks * 8
2256 */
2257 upper_limit = (1LL << 32) - 1;
2258
2259 /* total blocks in file system block size */
2260 upper_limit >>= (blkbits - 9);
2261 upper_limit <<= blkbits;
2262 }
2263
2264 /*
2265 * 32-bit extent-start container, ee_block. We lower the maxbytes
2266 * by one fs block, so ee_len can cover the extent of maximum file
2267 * size
2268 */
2269 res = (1LL << 32) - 1;
2270 res <<= blkbits;
2271
2272 /* Sanity check against vm- & vfs- imposed limits */
2273 if (res > upper_limit)
2274 res = upper_limit;
2275
2276 return res;
2277}
2278
2279/*
2280 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2281 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2282 * We need to be 1 filesystem block less than the 2^48 sector limit.
2283 */
2284static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2285{
2286 loff_t res = EXT4_NDIR_BLOCKS;
2287 int meta_blocks;
2288 loff_t upper_limit;
2289 /* This is calculated to be the largest file size for a dense, block
2290 * mapped file such that the file's total number of 512-byte sectors,
2291 * including data and all indirect blocks, does not exceed (2^48 - 1).
2292 *
2293 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2294 * number of 512-byte sectors of the file.
2295 */
2296
2297 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2298 /*
2299 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2300 * the inode i_block field represents total file blocks in
2301 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2302 */
2303 upper_limit = (1LL << 32) - 1;
2304
2305 /* total blocks in file system block size */
2306 upper_limit >>= (bits - 9);
2307
2308 } else {
2309 /*
2310 * We use 48 bit ext4_inode i_blocks
2311 * With EXT4_HUGE_FILE_FL set the i_blocks
2312 * represent total number of blocks in
2313 * file system block size
2314 */
2315 upper_limit = (1LL << 48) - 1;
2316
2317 }
2318
2319 /* indirect blocks */
2320 meta_blocks = 1;
2321 /* double indirect blocks */
2322 meta_blocks += 1 + (1LL << (bits-2));
2323 /* tripple indirect blocks */
2324 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2325
2326 upper_limit -= meta_blocks;
2327 upper_limit <<= bits;
2328
2329 res += 1LL << (bits-2);
2330 res += 1LL << (2*(bits-2));
2331 res += 1LL << (3*(bits-2));
2332 res <<= bits;
2333 if (res > upper_limit)
2334 res = upper_limit;
2335
2336 if (res > MAX_LFS_FILESIZE)
2337 res = MAX_LFS_FILESIZE;
2338
2339 return res;
2340}
2341
2342static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2343 ext4_fsblk_t logical_sb_block, int nr)
2344{
2345 struct ext4_sb_info *sbi = EXT4_SB(sb);
2346 ext4_group_t bg, first_meta_bg;
2347 int has_super = 0;
2348
2349 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2350
2351 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2352 nr < first_meta_bg)
2353 return logical_sb_block + nr + 1;
2354 bg = sbi->s_desc_per_block * nr;
2355 if (ext4_bg_has_super(sb, bg))
2356 has_super = 1;
2357
2358 return (has_super + ext4_group_first_block_no(sb, bg));
2359}
2360
2361/**
2362 * ext4_get_stripe_size: Get the stripe size.
2363 * @sbi: In memory super block info
2364 *
2365 * If we have specified it via mount option, then
2366 * use the mount option value. If the value specified at mount time is
2367 * greater than the blocks per group use the super block value.
2368 * If the super block value is greater than blocks per group return 0.
2369 * Allocator needs it be less than blocks per group.
2370 *
2371 */
2372static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2373{
2374 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2375 unsigned long stripe_width =
2376 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2377 int ret;
2378
2379 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2380 ret = sbi->s_stripe;
2381 else if (stripe_width <= sbi->s_blocks_per_group)
2382 ret = stripe_width;
2383 else if (stride <= sbi->s_blocks_per_group)
2384 ret = stride;
2385 else
2386 ret = 0;
2387
2388 /*
2389 * If the stripe width is 1, this makes no sense and
2390 * we set it to 0 to turn off stripe handling code.
2391 */
2392 if (ret <= 1)
2393 ret = 0;
2394
2395 return ret;
2396}
2397
2398/* sysfs supprt */
2399
2400struct ext4_attr {
2401 struct attribute attr;
2402 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2403 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2404 const char *, size_t);
2405 int offset;
2406};
2407
2408static int parse_strtoul(const char *buf,
2409 unsigned long max, unsigned long *value)
2410{
2411 char *endp;
2412
2413 *value = simple_strtoul(skip_spaces(buf), &endp, 0);
2414 endp = skip_spaces(endp);
2415 if (*endp || *value > max)
2416 return -EINVAL;
2417
2418 return 0;
2419}
2420
2421static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2422 struct ext4_sb_info *sbi,
2423 char *buf)
2424{
2425 return snprintf(buf, PAGE_SIZE, "%llu\n",
2426 (s64) EXT4_C2B(sbi,
2427 percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
2428}
2429
2430static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2431 struct ext4_sb_info *sbi, char *buf)
2432{
2433 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2434
2435 if (!sb->s_bdev->bd_part)
2436 return snprintf(buf, PAGE_SIZE, "0\n");
2437 return snprintf(buf, PAGE_SIZE, "%lu\n",
2438 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2439 sbi->s_sectors_written_start) >> 1);
2440}
2441
2442static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2443 struct ext4_sb_info *sbi, char *buf)
2444{
2445 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2446
2447 if (!sb->s_bdev->bd_part)
2448 return snprintf(buf, PAGE_SIZE, "0\n");
2449 return snprintf(buf, PAGE_SIZE, "%llu\n",
2450 (unsigned long long)(sbi->s_kbytes_written +
2451 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2452 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
2453}
2454
2455static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2456 struct ext4_sb_info *sbi,
2457 const char *buf, size_t count)
2458{
2459 unsigned long t;
2460
2461 if (parse_strtoul(buf, 0x40000000, &t))
2462 return -EINVAL;
2463
2464 if (t && !is_power_of_2(t))
2465 return -EINVAL;
2466
2467 sbi->s_inode_readahead_blks = t;
2468 return count;
2469}
2470
2471static ssize_t sbi_ui_show(struct ext4_attr *a,
2472 struct ext4_sb_info *sbi, char *buf)
2473{
2474 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2475
2476 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2477}
2478
2479static ssize_t sbi_ui_store(struct ext4_attr *a,
2480 struct ext4_sb_info *sbi,
2481 const char *buf, size_t count)
2482{
2483 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2484 unsigned long t;
2485
2486 if (parse_strtoul(buf, 0xffffffff, &t))
2487 return -EINVAL;
2488 *ui = t;
2489 return count;
2490}
2491
2492static ssize_t trigger_test_error(struct ext4_attr *a,
2493 struct ext4_sb_info *sbi,
2494 const char *buf, size_t count)
2495{
2496 int len = count;
2497
2498 if (!capable(CAP_SYS_ADMIN))
2499 return -EPERM;
2500
2501 if (len && buf[len-1] == '\n')
2502 len--;
2503
2504 if (len)
2505 ext4_error(sbi->s_sb, "%.*s", len, buf);
2506 return count;
2507}
2508
2509#define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2510static struct ext4_attr ext4_attr_##_name = { \
2511 .attr = {.name = __stringify(_name), .mode = _mode }, \
2512 .show = _show, \
2513 .store = _store, \
2514 .offset = offsetof(struct ext4_sb_info, _elname), \
2515}
2516#define EXT4_ATTR(name, mode, show, store) \
2517static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2518
2519#define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
2520#define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2521#define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2522#define EXT4_RW_ATTR_SBI_UI(name, elname) \
2523 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2524#define ATTR_LIST(name) &ext4_attr_##name.attr
2525
2526EXT4_RO_ATTR(delayed_allocation_blocks);
2527EXT4_RO_ATTR(session_write_kbytes);
2528EXT4_RO_ATTR(lifetime_write_kbytes);
2529EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2530 inode_readahead_blks_store, s_inode_readahead_blks);
2531EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2532EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2533EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2534EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2535EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2536EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2537EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2538EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
2539EXT4_ATTR(trigger_fs_error, 0200, NULL, trigger_test_error);
2540
2541static struct attribute *ext4_attrs[] = {
2542 ATTR_LIST(delayed_allocation_blocks),
2543 ATTR_LIST(session_write_kbytes),
2544 ATTR_LIST(lifetime_write_kbytes),
2545 ATTR_LIST(inode_readahead_blks),
2546 ATTR_LIST(inode_goal),
2547 ATTR_LIST(mb_stats),
2548 ATTR_LIST(mb_max_to_scan),
2549 ATTR_LIST(mb_min_to_scan),
2550 ATTR_LIST(mb_order2_req),
2551 ATTR_LIST(mb_stream_req),
2552 ATTR_LIST(mb_group_prealloc),
2553 ATTR_LIST(max_writeback_mb_bump),
2554 ATTR_LIST(trigger_fs_error),
2555 NULL,
2556};
2557
2558/* Features this copy of ext4 supports */
2559EXT4_INFO_ATTR(lazy_itable_init);
2560EXT4_INFO_ATTR(batched_discard);
2561
2562static struct attribute *ext4_feat_attrs[] = {
2563 ATTR_LIST(lazy_itable_init),
2564 ATTR_LIST(batched_discard),
2565 NULL,
2566};
2567
2568static ssize_t ext4_attr_show(struct kobject *kobj,
2569 struct attribute *attr, char *buf)
2570{
2571 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2572 s_kobj);
2573 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2574
2575 return a->show ? a->show(a, sbi, buf) : 0;
2576}
2577
2578static ssize_t ext4_attr_store(struct kobject *kobj,
2579 struct attribute *attr,
2580 const char *buf, size_t len)
2581{
2582 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2583 s_kobj);
2584 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2585
2586 return a->store ? a->store(a, sbi, buf, len) : 0;
2587}
2588
2589static void ext4_sb_release(struct kobject *kobj)
2590{
2591 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2592 s_kobj);
2593 complete(&sbi->s_kobj_unregister);
2594}
2595
2596static const struct sysfs_ops ext4_attr_ops = {
2597 .show = ext4_attr_show,
2598 .store = ext4_attr_store,
2599};
2600
2601static struct kobj_type ext4_ktype = {
2602 .default_attrs = ext4_attrs,
2603 .sysfs_ops = &ext4_attr_ops,
2604 .release = ext4_sb_release,
2605};
2606
2607static void ext4_feat_release(struct kobject *kobj)
2608{
2609 complete(&ext4_feat->f_kobj_unregister);
2610}
2611
2612static struct kobj_type ext4_feat_ktype = {
2613 .default_attrs = ext4_feat_attrs,
2614 .sysfs_ops = &ext4_attr_ops,
2615 .release = ext4_feat_release,
2616};
2617
2618/*
2619 * Check whether this filesystem can be mounted based on
2620 * the features present and the RDONLY/RDWR mount requested.
2621 * Returns 1 if this filesystem can be mounted as requested,
2622 * 0 if it cannot be.
2623 */
2624static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2625{
2626 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2627 ext4_msg(sb, KERN_ERR,
2628 "Couldn't mount because of "
2629 "unsupported optional features (%x)",
2630 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2631 ~EXT4_FEATURE_INCOMPAT_SUPP));
2632 return 0;
2633 }
2634
2635 if (readonly)
2636 return 1;
2637
2638 /* Check that feature set is OK for a read-write mount */
2639 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2640 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2641 "unsupported optional features (%x)",
2642 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2643 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2644 return 0;
2645 }
2646 /*
2647 * Large file size enabled file system can only be mounted
2648 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2649 */
2650 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2651 if (sizeof(blkcnt_t) < sizeof(u64)) {
2652 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2653 "cannot be mounted RDWR without "
2654 "CONFIG_LBDAF");
2655 return 0;
2656 }
2657 }
2658 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2659 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2660 ext4_msg(sb, KERN_ERR,
2661 "Can't support bigalloc feature without "
2662 "extents feature\n");
2663 return 0;
2664 }
2665 return 1;
2666}
2667
2668/*
2669 * This function is called once a day if we have errors logged
2670 * on the file system
2671 */
2672static void print_daily_error_info(unsigned long arg)
2673{
2674 struct super_block *sb = (struct super_block *) arg;
2675 struct ext4_sb_info *sbi;
2676 struct ext4_super_block *es;
2677
2678 sbi = EXT4_SB(sb);
2679 es = sbi->s_es;
2680
2681 if (es->s_error_count)
2682 ext4_msg(sb, KERN_NOTICE, "error count: %u",
2683 le32_to_cpu(es->s_error_count));
2684 if (es->s_first_error_time) {
2685 printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d",
2686 sb->s_id, le32_to_cpu(es->s_first_error_time),
2687 (int) sizeof(es->s_first_error_func),
2688 es->s_first_error_func,
2689 le32_to_cpu(es->s_first_error_line));
2690 if (es->s_first_error_ino)
2691 printk(": inode %u",
2692 le32_to_cpu(es->s_first_error_ino));
2693 if (es->s_first_error_block)
2694 printk(": block %llu", (unsigned long long)
2695 le64_to_cpu(es->s_first_error_block));
2696 printk("\n");
2697 }
2698 if (es->s_last_error_time) {
2699 printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d",
2700 sb->s_id, le32_to_cpu(es->s_last_error_time),
2701 (int) sizeof(es->s_last_error_func),
2702 es->s_last_error_func,
2703 le32_to_cpu(es->s_last_error_line));
2704 if (es->s_last_error_ino)
2705 printk(": inode %u",
2706 le32_to_cpu(es->s_last_error_ino));
2707 if (es->s_last_error_block)
2708 printk(": block %llu", (unsigned long long)
2709 le64_to_cpu(es->s_last_error_block));
2710 printk("\n");
2711 }
2712 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2713}
2714
2715/* Find next suitable group and run ext4_init_inode_table */
2716static int ext4_run_li_request(struct ext4_li_request *elr)
2717{
2718 struct ext4_group_desc *gdp = NULL;
2719 ext4_group_t group, ngroups;
2720 struct super_block *sb;
2721 unsigned long timeout = 0;
2722 int ret = 0;
2723
2724 sb = elr->lr_super;
2725 ngroups = EXT4_SB(sb)->s_groups_count;
2726
2727 for (group = elr->lr_next_group; group < ngroups; group++) {
2728 gdp = ext4_get_group_desc(sb, group, NULL);
2729 if (!gdp) {
2730 ret = 1;
2731 break;
2732 }
2733
2734 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2735 break;
2736 }
2737
2738 if (group == ngroups)
2739 ret = 1;
2740
2741 if (!ret) {
2742 timeout = jiffies;
2743 ret = ext4_init_inode_table(sb, group,
2744 elr->lr_timeout ? 0 : 1);
2745 if (elr->lr_timeout == 0) {
2746 timeout = (jiffies - timeout) *
2747 elr->lr_sbi->s_li_wait_mult;
2748 elr->lr_timeout = timeout;
2749 }
2750 elr->lr_next_sched = jiffies + elr->lr_timeout;
2751 elr->lr_next_group = group + 1;
2752 }
2753
2754 return ret;
2755}
2756
2757/*
2758 * Remove lr_request from the list_request and free the
2759 * request structure. Should be called with li_list_mtx held
2760 */
2761static void ext4_remove_li_request(struct ext4_li_request *elr)
2762{
2763 struct ext4_sb_info *sbi;
2764
2765 if (!elr)
2766 return;
2767
2768 sbi = elr->lr_sbi;
2769
2770 list_del(&elr->lr_request);
2771 sbi->s_li_request = NULL;
2772 kfree(elr);
2773}
2774
2775static void ext4_unregister_li_request(struct super_block *sb)
2776{
2777 mutex_lock(&ext4_li_mtx);
2778 if (!ext4_li_info) {
2779 mutex_unlock(&ext4_li_mtx);
2780 return;
2781 }
2782
2783 mutex_lock(&ext4_li_info->li_list_mtx);
2784 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2785 mutex_unlock(&ext4_li_info->li_list_mtx);
2786 mutex_unlock(&ext4_li_mtx);
2787}
2788
2789static struct task_struct *ext4_lazyinit_task;
2790
2791/*
2792 * This is the function where ext4lazyinit thread lives. It walks
2793 * through the request list searching for next scheduled filesystem.
2794 * When such a fs is found, run the lazy initialization request
2795 * (ext4_rn_li_request) and keep track of the time spend in this
2796 * function. Based on that time we compute next schedule time of
2797 * the request. When walking through the list is complete, compute
2798 * next waking time and put itself into sleep.
2799 */
2800static int ext4_lazyinit_thread(void *arg)
2801{
2802 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2803 struct list_head *pos, *n;
2804 struct ext4_li_request *elr;
2805 unsigned long next_wakeup, cur;
2806
2807 BUG_ON(NULL == eli);
2808
2809cont_thread:
2810 while (true) {
2811 next_wakeup = MAX_JIFFY_OFFSET;
2812
2813 mutex_lock(&eli->li_list_mtx);
2814 if (list_empty(&eli->li_request_list)) {
2815 mutex_unlock(&eli->li_list_mtx);
2816 goto exit_thread;
2817 }
2818
2819 list_for_each_safe(pos, n, &eli->li_request_list) {
2820 elr = list_entry(pos, struct ext4_li_request,
2821 lr_request);
2822
2823 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2824 if (ext4_run_li_request(elr) != 0) {
2825 /* error, remove the lazy_init job */
2826 ext4_remove_li_request(elr);
2827 continue;
2828 }
2829 }
2830
2831 if (time_before(elr->lr_next_sched, next_wakeup))
2832 next_wakeup = elr->lr_next_sched;
2833 }
2834 mutex_unlock(&eli->li_list_mtx);
2835
2836 try_to_freeze();
2837
2838 cur = jiffies;
2839 if ((time_after_eq(cur, next_wakeup)) ||
2840 (MAX_JIFFY_OFFSET == next_wakeup)) {
2841 cond_resched();
2842 continue;
2843 }
2844
2845 schedule_timeout_interruptible(next_wakeup - cur);
2846
2847 if (kthread_should_stop()) {
2848 ext4_clear_request_list();
2849 goto exit_thread;
2850 }
2851 }
2852
2853exit_thread:
2854 /*
2855 * It looks like the request list is empty, but we need
2856 * to check it under the li_list_mtx lock, to prevent any
2857 * additions into it, and of course we should lock ext4_li_mtx
2858 * to atomically free the list and ext4_li_info, because at
2859 * this point another ext4 filesystem could be registering
2860 * new one.
2861 */
2862 mutex_lock(&ext4_li_mtx);
2863 mutex_lock(&eli->li_list_mtx);
2864 if (!list_empty(&eli->li_request_list)) {
2865 mutex_unlock(&eli->li_list_mtx);
2866 mutex_unlock(&ext4_li_mtx);
2867 goto cont_thread;
2868 }
2869 mutex_unlock(&eli->li_list_mtx);
2870 kfree(ext4_li_info);
2871 ext4_li_info = NULL;
2872 mutex_unlock(&ext4_li_mtx);
2873
2874 return 0;
2875}
2876
2877static void ext4_clear_request_list(void)
2878{
2879 struct list_head *pos, *n;
2880 struct ext4_li_request *elr;
2881
2882 mutex_lock(&ext4_li_info->li_list_mtx);
2883 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2884 elr = list_entry(pos, struct ext4_li_request,
2885 lr_request);
2886 ext4_remove_li_request(elr);
2887 }
2888 mutex_unlock(&ext4_li_info->li_list_mtx);
2889}
2890
2891static int ext4_run_lazyinit_thread(void)
2892{
2893 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2894 ext4_li_info, "ext4lazyinit");
2895 if (IS_ERR(ext4_lazyinit_task)) {
2896 int err = PTR_ERR(ext4_lazyinit_task);
2897 ext4_clear_request_list();
2898 kfree(ext4_li_info);
2899 ext4_li_info = NULL;
2900 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
2901 "initialization thread\n",
2902 err);
2903 return err;
2904 }
2905 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
2906 return 0;
2907}
2908
2909/*
2910 * Check whether it make sense to run itable init. thread or not.
2911 * If there is at least one uninitialized inode table, return
2912 * corresponding group number, else the loop goes through all
2913 * groups and return total number of groups.
2914 */
2915static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2916{
2917 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2918 struct ext4_group_desc *gdp = NULL;
2919
2920 for (group = 0; group < ngroups; group++) {
2921 gdp = ext4_get_group_desc(sb, group, NULL);
2922 if (!gdp)
2923 continue;
2924
2925 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2926 break;
2927 }
2928
2929 return group;
2930}
2931
2932static int ext4_li_info_new(void)
2933{
2934 struct ext4_lazy_init *eli = NULL;
2935
2936 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2937 if (!eli)
2938 return -ENOMEM;
2939
2940 INIT_LIST_HEAD(&eli->li_request_list);
2941 mutex_init(&eli->li_list_mtx);
2942
2943 eli->li_state |= EXT4_LAZYINIT_QUIT;
2944
2945 ext4_li_info = eli;
2946
2947 return 0;
2948}
2949
2950static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
2951 ext4_group_t start)
2952{
2953 struct ext4_sb_info *sbi = EXT4_SB(sb);
2954 struct ext4_li_request *elr;
2955 unsigned long rnd;
2956
2957 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
2958 if (!elr)
2959 return NULL;
2960
2961 elr->lr_super = sb;
2962 elr->lr_sbi = sbi;
2963 elr->lr_next_group = start;
2964
2965 /*
2966 * Randomize first schedule time of the request to
2967 * spread the inode table initialization requests
2968 * better.
2969 */
2970 get_random_bytes(&rnd, sizeof(rnd));
2971 elr->lr_next_sched = jiffies + (unsigned long)rnd %
2972 (EXT4_DEF_LI_MAX_START_DELAY * HZ);
2973
2974 return elr;
2975}
2976
2977static int ext4_register_li_request(struct super_block *sb,
2978 ext4_group_t first_not_zeroed)
2979{
2980 struct ext4_sb_info *sbi = EXT4_SB(sb);
2981 struct ext4_li_request *elr;
2982 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
2983 int ret = 0;
2984
2985 if (sbi->s_li_request != NULL) {
2986 /*
2987 * Reset timeout so it can be computed again, because
2988 * s_li_wait_mult might have changed.
2989 */
2990 sbi->s_li_request->lr_timeout = 0;
2991 return 0;
2992 }
2993
2994 if (first_not_zeroed == ngroups ||
2995 (sb->s_flags & MS_RDONLY) ||
2996 !test_opt(sb, INIT_INODE_TABLE))
2997 return 0;
2998
2999 elr = ext4_li_request_new(sb, first_not_zeroed);
3000 if (!elr)
3001 return -ENOMEM;
3002
3003 mutex_lock(&ext4_li_mtx);
3004
3005 if (NULL == ext4_li_info) {
3006 ret = ext4_li_info_new();
3007 if (ret)
3008 goto out;
3009 }
3010
3011 mutex_lock(&ext4_li_info->li_list_mtx);
3012 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3013 mutex_unlock(&ext4_li_info->li_list_mtx);
3014
3015 sbi->s_li_request = elr;
3016 /*
3017 * set elr to NULL here since it has been inserted to
3018 * the request_list and the removal and free of it is
3019 * handled by ext4_clear_request_list from now on.
3020 */
3021 elr = NULL;
3022
3023 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3024 ret = ext4_run_lazyinit_thread();
3025 if (ret)
3026 goto out;
3027 }
3028out:
3029 mutex_unlock(&ext4_li_mtx);
3030 if (ret)
3031 kfree(elr);
3032 return ret;
3033}
3034
3035/*
3036 * We do not need to lock anything since this is called on
3037 * module unload.
3038 */
3039static void ext4_destroy_lazyinit_thread(void)
3040{
3041 /*
3042 * If thread exited earlier
3043 * there's nothing to be done.
3044 */
3045 if (!ext4_li_info || !ext4_lazyinit_task)
3046 return;
3047
3048 kthread_stop(ext4_lazyinit_task);
3049}
3050
3051static int set_journal_csum_feature_set(struct super_block *sb)
3052{
3053 int ret = 1;
3054 int compat, incompat;
3055 struct ext4_sb_info *sbi = EXT4_SB(sb);
3056
3057 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3058 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3059 /* journal checksum v2 */
3060 compat = 0;
3061 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V2;
3062 } else {
3063 /* journal checksum v1 */
3064 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3065 incompat = 0;
3066 }
3067
3068 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3069 ret = jbd2_journal_set_features(sbi->s_journal,
3070 compat, 0,
3071 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3072 incompat);
3073 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3074 ret = jbd2_journal_set_features(sbi->s_journal,
3075 compat, 0,
3076 incompat);
3077 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3078 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3079 } else {
3080 jbd2_journal_clear_features(sbi->s_journal,
3081 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3082 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3083 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3084 }
3085
3086 return ret;
3087}
3088
3089/*
3090 * Note: calculating the overhead so we can be compatible with
3091 * historical BSD practice is quite difficult in the face of
3092 * clusters/bigalloc. This is because multiple metadata blocks from
3093 * different block group can end up in the same allocation cluster.
3094 * Calculating the exact overhead in the face of clustered allocation
3095 * requires either O(all block bitmaps) in memory or O(number of block
3096 * groups**2) in time. We will still calculate the superblock for
3097 * older file systems --- and if we come across with a bigalloc file
3098 * system with zero in s_overhead_clusters the estimate will be close to
3099 * correct especially for very large cluster sizes --- but for newer
3100 * file systems, it's better to calculate this figure once at mkfs
3101 * time, and store it in the superblock. If the superblock value is
3102 * present (even for non-bigalloc file systems), we will use it.
3103 */
3104static int count_overhead(struct super_block *sb, ext4_group_t grp,
3105 char *buf)
3106{
3107 struct ext4_sb_info *sbi = EXT4_SB(sb);
3108 struct ext4_group_desc *gdp;
3109 ext4_fsblk_t first_block, last_block, b;
3110 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3111 int s, j, count = 0;
3112
3113 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC))
3114 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3115 sbi->s_itb_per_group + 2);
3116
3117 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3118 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3119 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3120 for (i = 0; i < ngroups; i++) {
3121 gdp = ext4_get_group_desc(sb, i, NULL);
3122 b = ext4_block_bitmap(sb, gdp);
3123 if (b >= first_block && b <= last_block) {
3124 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3125 count++;
3126 }
3127 b = ext4_inode_bitmap(sb, gdp);
3128 if (b >= first_block && b <= last_block) {
3129 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3130 count++;
3131 }
3132 b = ext4_inode_table(sb, gdp);
3133 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3134 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3135 int c = EXT4_B2C(sbi, b - first_block);
3136 ext4_set_bit(c, buf);
3137 count++;
3138 }
3139 if (i != grp)
3140 continue;
3141 s = 0;
3142 if (ext4_bg_has_super(sb, grp)) {
3143 ext4_set_bit(s++, buf);
3144 count++;
3145 }
3146 for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
3147 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3148 count++;
3149 }
3150 }
3151 if (!count)
3152 return 0;
3153 return EXT4_CLUSTERS_PER_GROUP(sb) -
3154 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3155}
3156
3157/*
3158 * Compute the overhead and stash it in sbi->s_overhead
3159 */
3160int ext4_calculate_overhead(struct super_block *sb)
3161{
3162 struct ext4_sb_info *sbi = EXT4_SB(sb);
3163 struct ext4_super_block *es = sbi->s_es;
3164 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3165 ext4_fsblk_t overhead = 0;
3166 char *buf = (char *) get_zeroed_page(GFP_KERNEL);
3167
3168 memset(buf, 0, PAGE_SIZE);
3169 if (!buf)
3170 return -ENOMEM;
3171
3172 /*
3173 * Compute the overhead (FS structures). This is constant
3174 * for a given filesystem unless the number of block groups
3175 * changes so we cache the previous value until it does.
3176 */
3177
3178 /*
3179 * All of the blocks before first_data_block are overhead
3180 */
3181 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3182
3183 /*
3184 * Add the overhead found in each block group
3185 */
3186 for (i = 0; i < ngroups; i++) {
3187 int blks;
3188
3189 blks = count_overhead(sb, i, buf);
3190 overhead += blks;
3191 if (blks)
3192 memset(buf, 0, PAGE_SIZE);
3193 cond_resched();
3194 }
3195 sbi->s_overhead = overhead;
3196 smp_wmb();
3197 free_page((unsigned long) buf);
3198 return 0;
3199}
3200
3201static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3202{
3203 char *orig_data = kstrdup(data, GFP_KERNEL);
3204 struct buffer_head *bh;
3205 struct ext4_super_block *es = NULL;
3206 struct ext4_sb_info *sbi;
3207 ext4_fsblk_t block;
3208 ext4_fsblk_t sb_block = get_sb_block(&data);
3209 ext4_fsblk_t logical_sb_block;
3210 unsigned long offset = 0;
3211 unsigned long journal_devnum = 0;
3212 unsigned long def_mount_opts;
3213 struct inode *root;
3214 char *cp;
3215 const char *descr;
3216 int ret = -ENOMEM;
3217 int blocksize, clustersize;
3218 unsigned int db_count;
3219 unsigned int i;
3220 int needs_recovery, has_huge_files, has_bigalloc;
3221 __u64 blocks_count;
3222 int err;
3223 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3224 ext4_group_t first_not_zeroed;
3225
3226 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3227 if (!sbi)
3228 goto out_free_orig;
3229
3230 sbi->s_blockgroup_lock =
3231 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3232 if (!sbi->s_blockgroup_lock) {
3233 kfree(sbi);
3234 goto out_free_orig;
3235 }
3236 sb->s_fs_info = sbi;
3237 sbi->s_sb = sb;
3238 sbi->s_mount_opt = 0;
3239 sbi->s_resuid = make_kuid(&init_user_ns, EXT4_DEF_RESUID);
3240 sbi->s_resgid = make_kgid(&init_user_ns, EXT4_DEF_RESGID);
3241 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3242 sbi->s_sb_block = sb_block;
3243 if (sb->s_bdev->bd_part)
3244 sbi->s_sectors_written_start =
3245 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3246
3247 /* Cleanup superblock name */
3248 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3249 *cp = '!';
3250
3251 ret = -EINVAL;
3252 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3253 if (!blocksize) {
3254 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3255 goto out_fail;
3256 }
3257
3258 /*
3259 * The ext4 superblock will not be buffer aligned for other than 1kB
3260 * block sizes. We need to calculate the offset from buffer start.
3261 */
3262 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3263 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3264 offset = do_div(logical_sb_block, blocksize);
3265 } else {
3266 logical_sb_block = sb_block;
3267 }
3268
3269 if (!(bh = sb_bread(sb, logical_sb_block))) {
3270 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3271 goto out_fail;
3272 }
3273 /*
3274 * Note: s_es must be initialized as soon as possible because
3275 * some ext4 macro-instructions depend on its value
3276 */
3277 es = (struct ext4_super_block *) (bh->b_data + offset);
3278 sbi->s_es = es;
3279 sb->s_magic = le16_to_cpu(es->s_magic);
3280 if (sb->s_magic != EXT4_SUPER_MAGIC)
3281 goto cantfind_ext4;
3282 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3283
3284 /* Warn if metadata_csum and gdt_csum are both set. */
3285 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3286 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
3287 EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_GDT_CSUM))
3288 ext4_warning(sb, KERN_INFO "metadata_csum and uninit_bg are "
3289 "redundant flags; please run fsck.");
3290
3291 /* Check for a known checksum algorithm */
3292 if (!ext4_verify_csum_type(sb, es)) {
3293 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3294 "unknown checksum algorithm.");
3295 silent = 1;
3296 goto cantfind_ext4;
3297 }
3298
3299 /* Load the checksum driver */
3300 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3301 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3302 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3303 if (IS_ERR(sbi->s_chksum_driver)) {
3304 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3305 ret = PTR_ERR(sbi->s_chksum_driver);
3306 sbi->s_chksum_driver = NULL;
3307 goto failed_mount;
3308 }
3309 }
3310
3311 /* Check superblock checksum */
3312 if (!ext4_superblock_csum_verify(sb, es)) {
3313 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3314 "invalid superblock checksum. Run e2fsck?");
3315 silent = 1;
3316 goto cantfind_ext4;
3317 }
3318
3319 /* Precompute checksum seed for all metadata */
3320 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3321 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
3322 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3323 sizeof(es->s_uuid));
3324
3325 /* Set defaults before we parse the mount options */
3326 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3327 set_opt(sb, INIT_INODE_TABLE);
3328 if (def_mount_opts & EXT4_DEFM_DEBUG)
3329 set_opt(sb, DEBUG);
3330 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3331 set_opt(sb, GRPID);
3332 if (def_mount_opts & EXT4_DEFM_UID16)
3333 set_opt(sb, NO_UID32);
3334 /* xattr user namespace & acls are now defaulted on */
3335#ifdef CONFIG_EXT4_FS_XATTR
3336 set_opt(sb, XATTR_USER);
3337#endif
3338#ifdef CONFIG_EXT4_FS_POSIX_ACL
3339 set_opt(sb, POSIX_ACL);
3340#endif
3341 set_opt(sb, MBLK_IO_SUBMIT);
3342 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3343 set_opt(sb, JOURNAL_DATA);
3344 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3345 set_opt(sb, ORDERED_DATA);
3346 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3347 set_opt(sb, WRITEBACK_DATA);
3348
3349 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3350 set_opt(sb, ERRORS_PANIC);
3351 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3352 set_opt(sb, ERRORS_CONT);
3353 else
3354 set_opt(sb, ERRORS_RO);
3355 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
3356 set_opt(sb, BLOCK_VALIDITY);
3357 if (def_mount_opts & EXT4_DEFM_DISCARD)
3358 set_opt(sb, DISCARD);
3359
3360 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3361 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
3362 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3363 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3364 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3365
3366 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3367 set_opt(sb, BARRIER);
3368
3369 /*
3370 * enable delayed allocation by default
3371 * Use -o nodelalloc to turn it off
3372 */
3373 if (!IS_EXT3_SB(sb) &&
3374 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3375 set_opt(sb, DELALLOC);
3376
3377 /*
3378 * set default s_li_wait_mult for lazyinit, for the case there is
3379 * no mount option specified.
3380 */
3381 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3382
3383 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3384 &journal_devnum, &journal_ioprio, 0)) {
3385 ext4_msg(sb, KERN_WARNING,
3386 "failed to parse options in superblock: %s",
3387 sbi->s_es->s_mount_opts);
3388 }
3389 sbi->s_def_mount_opt = sbi->s_mount_opt;
3390 if (!parse_options((char *) data, sb, &journal_devnum,
3391 &journal_ioprio, 0))
3392 goto failed_mount;
3393
3394 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3395 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3396 "with data=journal disables delayed "
3397 "allocation and O_DIRECT support!\n");
3398 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3399 ext4_msg(sb, KERN_ERR, "can't mount with "
3400 "both data=journal and delalloc");
3401 goto failed_mount;
3402 }
3403 if (test_opt(sb, DIOREAD_NOLOCK)) {
3404 ext4_msg(sb, KERN_ERR, "can't mount with "
3405 "both data=journal and delalloc");
3406 goto failed_mount;
3407 }
3408 if (test_opt(sb, DELALLOC))
3409 clear_opt(sb, DELALLOC);
3410 }
3411
3412 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3413 if (test_opt(sb, DIOREAD_NOLOCK)) {
3414 if (blocksize < PAGE_SIZE) {
3415 ext4_msg(sb, KERN_ERR, "can't mount with "
3416 "dioread_nolock if block size != PAGE_SIZE");
3417 goto failed_mount;
3418 }
3419 }
3420
3421 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3422 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3423
3424 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3425 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3426 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3427 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
3428 ext4_msg(sb, KERN_WARNING,
3429 "feature flags set on rev 0 fs, "
3430 "running e2fsck is recommended");
3431
3432 if (IS_EXT2_SB(sb)) {
3433 if (ext2_feature_set_ok(sb))
3434 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3435 "using the ext4 subsystem");
3436 else {
3437 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3438 "to feature incompatibilities");
3439 goto failed_mount;
3440 }
3441 }
3442
3443 if (IS_EXT3_SB(sb)) {
3444 if (ext3_feature_set_ok(sb))
3445 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3446 "using the ext4 subsystem");
3447 else {
3448 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3449 "to feature incompatibilities");
3450 goto failed_mount;
3451 }
3452 }
3453
3454 /*
3455 * Check feature flags regardless of the revision level, since we
3456 * previously didn't change the revision level when setting the flags,
3457 * so there is a chance incompat flags are set on a rev 0 filesystem.
3458 */
3459 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3460 goto failed_mount;
3461
3462 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3463 blocksize > EXT4_MAX_BLOCK_SIZE) {
3464 ext4_msg(sb, KERN_ERR,
3465 "Unsupported filesystem blocksize %d", blocksize);
3466 goto failed_mount;
3467 }
3468
3469 if (sb->s_blocksize != blocksize) {
3470 /* Validate the filesystem blocksize */
3471 if (!sb_set_blocksize(sb, blocksize)) {
3472 ext4_msg(sb, KERN_ERR, "bad block size %d",
3473 blocksize);
3474 goto failed_mount;
3475 }
3476
3477 brelse(bh);
3478 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3479 offset = do_div(logical_sb_block, blocksize);
3480 bh = sb_bread(sb, logical_sb_block);
3481 if (!bh) {
3482 ext4_msg(sb, KERN_ERR,
3483 "Can't read superblock on 2nd try");
3484 goto failed_mount;
3485 }
3486 es = (struct ext4_super_block *)(bh->b_data + offset);
3487 sbi->s_es = es;
3488 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3489 ext4_msg(sb, KERN_ERR,
3490 "Magic mismatch, very weird!");
3491 goto failed_mount;
3492 }
3493 }
3494
3495 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3496 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
3497 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3498 has_huge_files);
3499 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3500
3501 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3502 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3503 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3504 } else {
3505 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3506 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3507 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3508 (!is_power_of_2(sbi->s_inode_size)) ||
3509 (sbi->s_inode_size > blocksize)) {
3510 ext4_msg(sb, KERN_ERR,
3511 "unsupported inode size: %d",
3512 sbi->s_inode_size);
3513 goto failed_mount;
3514 }
3515 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3516 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3517 }
3518
3519 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3520 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
3521 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3522 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3523 !is_power_of_2(sbi->s_desc_size)) {
3524 ext4_msg(sb, KERN_ERR,
3525 "unsupported descriptor size %lu",
3526 sbi->s_desc_size);
3527 goto failed_mount;
3528 }
3529 } else
3530 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3531
3532 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3533 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3534 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
3535 goto cantfind_ext4;
3536
3537 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3538 if (sbi->s_inodes_per_block == 0)
3539 goto cantfind_ext4;
3540 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3541 sbi->s_inodes_per_block;
3542 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3543 sbi->s_sbh = bh;
3544 sbi->s_mount_state = le16_to_cpu(es->s_state);
3545 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3546 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3547
3548 for (i = 0; i < 4; i++)
3549 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3550 sbi->s_def_hash_version = es->s_def_hash_version;
3551 i = le32_to_cpu(es->s_flags);
3552 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3553 sbi->s_hash_unsigned = 3;
3554 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3555#ifdef __CHAR_UNSIGNED__
3556 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3557 sbi->s_hash_unsigned = 3;
3558#else
3559 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3560#endif
3561 }
3562
3563 /* Handle clustersize */
3564 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3565 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3566 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3567 if (has_bigalloc) {
3568 if (clustersize < blocksize) {
3569 ext4_msg(sb, KERN_ERR,
3570 "cluster size (%d) smaller than "
3571 "block size (%d)", clustersize, blocksize);
3572 goto failed_mount;
3573 }
3574 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3575 le32_to_cpu(es->s_log_block_size);
3576 sbi->s_clusters_per_group =
3577 le32_to_cpu(es->s_clusters_per_group);
3578 if (sbi->s_clusters_per_group > blocksize * 8) {
3579 ext4_msg(sb, KERN_ERR,
3580 "#clusters per group too big: %lu",
3581 sbi->s_clusters_per_group);
3582 goto failed_mount;
3583 }
3584 if (sbi->s_blocks_per_group !=
3585 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3586 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3587 "clusters per group (%lu) inconsistent",
3588 sbi->s_blocks_per_group,
3589 sbi->s_clusters_per_group);
3590 goto failed_mount;
3591 }
3592 } else {
3593 if (clustersize != blocksize) {
3594 ext4_warning(sb, "fragment/cluster size (%d) != "
3595 "block size (%d)", clustersize,
3596 blocksize);
3597 clustersize = blocksize;
3598 }
3599 if (sbi->s_blocks_per_group > blocksize * 8) {
3600 ext4_msg(sb, KERN_ERR,
3601 "#blocks per group too big: %lu",
3602 sbi->s_blocks_per_group);
3603 goto failed_mount;
3604 }
3605 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3606 sbi->s_cluster_bits = 0;
3607 }
3608 sbi->s_cluster_ratio = clustersize / blocksize;
3609
3610 if (sbi->s_inodes_per_group > blocksize * 8) {
3611 ext4_msg(sb, KERN_ERR,
3612 "#inodes per group too big: %lu",
3613 sbi->s_inodes_per_group);
3614 goto failed_mount;
3615 }
3616
3617 /*
3618 * Test whether we have more sectors than will fit in sector_t,
3619 * and whether the max offset is addressable by the page cache.
3620 */
3621 err = generic_check_addressable(sb->s_blocksize_bits,
3622 ext4_blocks_count(es));
3623 if (err) {
3624 ext4_msg(sb, KERN_ERR, "filesystem"
3625 " too large to mount safely on this system");
3626 if (sizeof(sector_t) < 8)
3627 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3628 ret = err;
3629 goto failed_mount;
3630 }
3631
3632 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3633 goto cantfind_ext4;
3634
3635 /* check blocks count against device size */
3636 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3637 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3638 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3639 "exceeds size of device (%llu blocks)",
3640 ext4_blocks_count(es), blocks_count);
3641 goto failed_mount;
3642 }
3643
3644 /*
3645 * It makes no sense for the first data block to be beyond the end
3646 * of the filesystem.
3647 */
3648 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3649 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
3650 "block %u is beyond end of filesystem (%llu)",
3651 le32_to_cpu(es->s_first_data_block),
3652 ext4_blocks_count(es));
3653 goto failed_mount;
3654 }
3655 blocks_count = (ext4_blocks_count(es) -
3656 le32_to_cpu(es->s_first_data_block) +
3657 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3658 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3659 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3660 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3661 "(block count %llu, first data block %u, "
3662 "blocks per group %lu)", sbi->s_groups_count,
3663 ext4_blocks_count(es),
3664 le32_to_cpu(es->s_first_data_block),
3665 EXT4_BLOCKS_PER_GROUP(sb));
3666 goto failed_mount;
3667 }
3668 sbi->s_groups_count = blocks_count;
3669 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3670 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3671 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3672 EXT4_DESC_PER_BLOCK(sb);
3673 sbi->s_group_desc = ext4_kvmalloc(db_count *
3674 sizeof(struct buffer_head *),
3675 GFP_KERNEL);
3676 if (sbi->s_group_desc == NULL) {
3677 ext4_msg(sb, KERN_ERR, "not enough memory");
3678 ret = -ENOMEM;
3679 goto failed_mount;
3680 }
3681
3682 if (ext4_proc_root)
3683 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
3684
3685 if (sbi->s_proc)
3686 proc_create_data("options", S_IRUGO, sbi->s_proc,
3687 &ext4_seq_options_fops, sb);
3688
3689 bgl_lock_init(sbi->s_blockgroup_lock);
3690
3691 for (i = 0; i < db_count; i++) {
3692 block = descriptor_loc(sb, logical_sb_block, i);
3693 sbi->s_group_desc[i] = sb_bread(sb, block);
3694 if (!sbi->s_group_desc[i]) {
3695 ext4_msg(sb, KERN_ERR,
3696 "can't read group descriptor %d", i);
3697 db_count = i;
3698 goto failed_mount2;
3699 }
3700 }
3701 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
3702 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3703 goto failed_mount2;
3704 }
3705 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3706 if (!ext4_fill_flex_info(sb)) {
3707 ext4_msg(sb, KERN_ERR,
3708 "unable to initialize "
3709 "flex_bg meta info!");
3710 goto failed_mount2;
3711 }
3712
3713 sbi->s_gdb_count = db_count;
3714 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3715 spin_lock_init(&sbi->s_next_gen_lock);
3716
3717 init_timer(&sbi->s_err_report);
3718 sbi->s_err_report.function = print_daily_error_info;
3719 sbi->s_err_report.data = (unsigned long) sb;
3720
3721 err = percpu_counter_init(&sbi->s_freeclusters_counter,
3722 ext4_count_free_clusters(sb));
3723 if (!err) {
3724 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3725 ext4_count_free_inodes(sb));
3726 }
3727 if (!err) {
3728 err = percpu_counter_init(&sbi->s_dirs_counter,
3729 ext4_count_dirs(sb));
3730 }
3731 if (!err) {
3732 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0);
3733 }
3734 if (err) {
3735 ext4_msg(sb, KERN_ERR, "insufficient memory");
3736 ret = err;
3737 goto failed_mount3;
3738 }
3739
3740 sbi->s_stripe = ext4_get_stripe_size(sbi);
3741 sbi->s_max_writeback_mb_bump = 128;
3742
3743 /*
3744 * set up enough so that it can read an inode
3745 */
3746 if (!test_opt(sb, NOLOAD) &&
3747 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3748 sb->s_op = &ext4_sops;
3749 else
3750 sb->s_op = &ext4_nojournal_sops;
3751 sb->s_export_op = &ext4_export_ops;
3752 sb->s_xattr = ext4_xattr_handlers;
3753#ifdef CONFIG_QUOTA
3754 sb->s_qcop = &ext4_qctl_operations;
3755 sb->dq_op = &ext4_quota_operations;
3756#endif
3757 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3758
3759 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3760 mutex_init(&sbi->s_orphan_lock);
3761 sbi->s_resize_flags = 0;
3762
3763 sb->s_root = NULL;
3764
3765 needs_recovery = (es->s_last_orphan != 0 ||
3766 EXT4_HAS_INCOMPAT_FEATURE(sb,
3767 EXT4_FEATURE_INCOMPAT_RECOVER));
3768
3769 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3770 !(sb->s_flags & MS_RDONLY))
3771 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3772 goto failed_mount3;
3773
3774 /*
3775 * The first inode we look at is the journal inode. Don't try
3776 * root first: it may be modified in the journal!
3777 */
3778 if (!test_opt(sb, NOLOAD) &&
3779 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3780 if (ext4_load_journal(sb, es, journal_devnum))
3781 goto failed_mount3;
3782 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3783 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3784 ext4_msg(sb, KERN_ERR, "required journal recovery "
3785 "suppressed and not mounted read-only");
3786 goto failed_mount_wq;
3787 } else {
3788 clear_opt(sb, DATA_FLAGS);
3789 sbi->s_journal = NULL;
3790 needs_recovery = 0;
3791 goto no_journal;
3792 }
3793
3794 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT) &&
3795 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3796 JBD2_FEATURE_INCOMPAT_64BIT)) {
3797 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
3798 goto failed_mount_wq;
3799 }
3800
3801 if (!set_journal_csum_feature_set(sb)) {
3802 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
3803 "feature set");
3804 goto failed_mount_wq;
3805 }
3806
3807 /* We have now updated the journal if required, so we can
3808 * validate the data journaling mode. */
3809 switch (test_opt(sb, DATA_FLAGS)) {
3810 case 0:
3811 /* No mode set, assume a default based on the journal
3812 * capabilities: ORDERED_DATA if the journal can
3813 * cope, else JOURNAL_DATA
3814 */
3815 if (jbd2_journal_check_available_features
3816 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
3817 set_opt(sb, ORDERED_DATA);
3818 else
3819 set_opt(sb, JOURNAL_DATA);
3820 break;
3821
3822 case EXT4_MOUNT_ORDERED_DATA:
3823 case EXT4_MOUNT_WRITEBACK_DATA:
3824 if (!jbd2_journal_check_available_features
3825 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
3826 ext4_msg(sb, KERN_ERR, "Journal does not support "
3827 "requested data journaling mode");
3828 goto failed_mount_wq;
3829 }
3830 default:
3831 break;
3832 }
3833 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3834
3835 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
3836
3837 /*
3838 * The journal may have updated the bg summary counts, so we
3839 * need to update the global counters.
3840 */
3841 percpu_counter_set(&sbi->s_freeclusters_counter,
3842 ext4_count_free_clusters(sb));
3843 percpu_counter_set(&sbi->s_freeinodes_counter,
3844 ext4_count_free_inodes(sb));
3845 percpu_counter_set(&sbi->s_dirs_counter,
3846 ext4_count_dirs(sb));
3847 percpu_counter_set(&sbi->s_dirtyclusters_counter, 0);
3848
3849no_journal:
3850 /*
3851 * Get the # of file system overhead blocks from the
3852 * superblock if present.
3853 */
3854 if (es->s_overhead_clusters)
3855 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
3856 else {
3857 ret = ext4_calculate_overhead(sb);
3858 if (ret)
3859 goto failed_mount_wq;
3860 }
3861
3862 /*
3863 * The maximum number of concurrent works can be high and
3864 * concurrency isn't really necessary. Limit it to 1.
3865 */
3866 EXT4_SB(sb)->dio_unwritten_wq =
3867 alloc_workqueue("ext4-dio-unwritten", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
3868 if (!EXT4_SB(sb)->dio_unwritten_wq) {
3869 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
3870 goto failed_mount_wq;
3871 }
3872
3873 /*
3874 * The jbd2_journal_load will have done any necessary log recovery,
3875 * so we can safely mount the rest of the filesystem now.
3876 */
3877
3878 root = ext4_iget(sb, EXT4_ROOT_INO);
3879 if (IS_ERR(root)) {
3880 ext4_msg(sb, KERN_ERR, "get root inode failed");
3881 ret = PTR_ERR(root);
3882 root = NULL;
3883 goto failed_mount4;
3884 }
3885 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
3886 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
3887 iput(root);
3888 goto failed_mount4;
3889 }
3890 sb->s_root = d_make_root(root);
3891 if (!sb->s_root) {
3892 ext4_msg(sb, KERN_ERR, "get root dentry failed");
3893 ret = -ENOMEM;
3894 goto failed_mount4;
3895 }
3896
3897 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
3898 sb->s_flags |= MS_RDONLY;
3899
3900 /* determine the minimum size of new large inodes, if present */
3901 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3902 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3903 EXT4_GOOD_OLD_INODE_SIZE;
3904 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3905 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
3906 if (sbi->s_want_extra_isize <
3907 le16_to_cpu(es->s_want_extra_isize))
3908 sbi->s_want_extra_isize =
3909 le16_to_cpu(es->s_want_extra_isize);
3910 if (sbi->s_want_extra_isize <
3911 le16_to_cpu(es->s_min_extra_isize))
3912 sbi->s_want_extra_isize =
3913 le16_to_cpu(es->s_min_extra_isize);
3914 }
3915 }
3916 /* Check if enough inode space is available */
3917 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3918 sbi->s_inode_size) {
3919 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3920 EXT4_GOOD_OLD_INODE_SIZE;
3921 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3922 "available");
3923 }
3924
3925 err = ext4_setup_system_zone(sb);
3926 if (err) {
3927 ext4_msg(sb, KERN_ERR, "failed to initialize system "
3928 "zone (%d)", err);
3929 goto failed_mount4a;
3930 }
3931
3932 ext4_ext_init(sb);
3933 err = ext4_mb_init(sb);
3934 if (err) {
3935 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
3936 err);
3937 goto failed_mount5;
3938 }
3939
3940 err = ext4_register_li_request(sb, first_not_zeroed);
3941 if (err)
3942 goto failed_mount6;
3943
3944 sbi->s_kobj.kset = ext4_kset;
3945 init_completion(&sbi->s_kobj_unregister);
3946 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
3947 "%s", sb->s_id);
3948 if (err)
3949 goto failed_mount7;
3950
3951 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3952 ext4_orphan_cleanup(sb, es);
3953 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
3954 if (needs_recovery) {
3955 ext4_msg(sb, KERN_INFO, "recovery complete");
3956 ext4_mark_recovery_complete(sb, es);
3957 }
3958 if (EXT4_SB(sb)->s_journal) {
3959 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3960 descr = " journalled data mode";
3961 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3962 descr = " ordered data mode";
3963 else
3964 descr = " writeback data mode";
3965 } else
3966 descr = "out journal";
3967
3968 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
3969 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
3970 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
3971
3972 if (es->s_error_count)
3973 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
3974
3975 kfree(orig_data);
3976 return 0;
3977
3978cantfind_ext4:
3979 if (!silent)
3980 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
3981 goto failed_mount;
3982
3983failed_mount7:
3984 ext4_unregister_li_request(sb);
3985failed_mount6:
3986 ext4_mb_release(sb);
3987failed_mount5:
3988 ext4_ext_release(sb);
3989 ext4_release_system_zone(sb);
3990failed_mount4a:
3991 dput(sb->s_root);
3992 sb->s_root = NULL;
3993failed_mount4:
3994 ext4_msg(sb, KERN_ERR, "mount failed");
3995 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
3996failed_mount_wq:
3997 if (sbi->s_journal) {
3998 jbd2_journal_destroy(sbi->s_journal);
3999 sbi->s_journal = NULL;
4000 }
4001failed_mount3:
4002 del_timer(&sbi->s_err_report);
4003 if (sbi->s_flex_groups)
4004 ext4_kvfree(sbi->s_flex_groups);
4005 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4006 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4007 percpu_counter_destroy(&sbi->s_dirs_counter);
4008 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
4009 if (sbi->s_mmp_tsk)
4010 kthread_stop(sbi->s_mmp_tsk);
4011failed_mount2:
4012 for (i = 0; i < db_count; i++)
4013 brelse(sbi->s_group_desc[i]);
4014 ext4_kvfree(sbi->s_group_desc);
4015failed_mount:
4016 if (sbi->s_chksum_driver)
4017 crypto_free_shash(sbi->s_chksum_driver);
4018 if (sbi->s_proc) {
4019 remove_proc_entry("options", sbi->s_proc);
4020 remove_proc_entry(sb->s_id, ext4_proc_root);
4021 }
4022#ifdef CONFIG_QUOTA
4023 for (i = 0; i < MAXQUOTAS; i++)
4024 kfree(sbi->s_qf_names[i]);
4025#endif
4026 ext4_blkdev_remove(sbi);
4027 brelse(bh);
4028out_fail:
4029 sb->s_fs_info = NULL;
4030 kfree(sbi->s_blockgroup_lock);
4031 kfree(sbi);
4032out_free_orig:
4033 kfree(orig_data);
4034 return ret;
4035}
4036
4037/*
4038 * Setup any per-fs journal parameters now. We'll do this both on
4039 * initial mount, once the journal has been initialised but before we've
4040 * done any recovery; and again on any subsequent remount.
4041 */
4042static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
4043{
4044 struct ext4_sb_info *sbi = EXT4_SB(sb);
4045
4046 journal->j_commit_interval = sbi->s_commit_interval;
4047 journal->j_min_batch_time = sbi->s_min_batch_time;
4048 journal->j_max_batch_time = sbi->s_max_batch_time;
4049
4050 write_lock(&journal->j_state_lock);
4051 if (test_opt(sb, BARRIER))
4052 journal->j_flags |= JBD2_BARRIER;
4053 else
4054 journal->j_flags &= ~JBD2_BARRIER;
4055 if (test_opt(sb, DATA_ERR_ABORT))
4056 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4057 else
4058 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
4059 write_unlock(&journal->j_state_lock);
4060}
4061
4062static journal_t *ext4_get_journal(struct super_block *sb,
4063 unsigned int journal_inum)
4064{
4065 struct inode *journal_inode;
4066 journal_t *journal;
4067
4068 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4069
4070 /* First, test for the existence of a valid inode on disk. Bad
4071 * things happen if we iget() an unused inode, as the subsequent
4072 * iput() will try to delete it. */
4073
4074 journal_inode = ext4_iget(sb, journal_inum);
4075 if (IS_ERR(journal_inode)) {
4076 ext4_msg(sb, KERN_ERR, "no journal found");
4077 return NULL;
4078 }
4079 if (!journal_inode->i_nlink) {
4080 make_bad_inode(journal_inode);
4081 iput(journal_inode);
4082 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
4083 return NULL;
4084 }
4085
4086 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4087 journal_inode, journal_inode->i_size);
4088 if (!S_ISREG(journal_inode->i_mode)) {
4089 ext4_msg(sb, KERN_ERR, "invalid journal inode");
4090 iput(journal_inode);
4091 return NULL;
4092 }
4093
4094 journal = jbd2_journal_init_inode(journal_inode);
4095 if (!journal) {
4096 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
4097 iput(journal_inode);
4098 return NULL;
4099 }
4100 journal->j_private = sb;
4101 ext4_init_journal_params(sb, journal);
4102 return journal;
4103}
4104
4105static journal_t *ext4_get_dev_journal(struct super_block *sb,
4106 dev_t j_dev)
4107{
4108 struct buffer_head *bh;
4109 journal_t *journal;
4110 ext4_fsblk_t start;
4111 ext4_fsblk_t len;
4112 int hblock, blocksize;
4113 ext4_fsblk_t sb_block;
4114 unsigned long offset;
4115 struct ext4_super_block *es;
4116 struct block_device *bdev;
4117
4118 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4119
4120 bdev = ext4_blkdev_get(j_dev, sb);
4121 if (bdev == NULL)
4122 return NULL;
4123
4124 blocksize = sb->s_blocksize;
4125 hblock = bdev_logical_block_size(bdev);
4126 if (blocksize < hblock) {
4127 ext4_msg(sb, KERN_ERR,
4128 "blocksize too small for journal device");
4129 goto out_bdev;
4130 }
4131
4132 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4133 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
4134 set_blocksize(bdev, blocksize);
4135 if (!(bh = __bread(bdev, sb_block, blocksize))) {
4136 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4137 "external journal");
4138 goto out_bdev;
4139 }
4140
4141 es = (struct ext4_super_block *) (bh->b_data + offset);
4142 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
4143 !(le32_to_cpu(es->s_feature_incompat) &
4144 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4145 ext4_msg(sb, KERN_ERR, "external journal has "
4146 "bad superblock");
4147 brelse(bh);
4148 goto out_bdev;
4149 }
4150
4151 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4152 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4153 brelse(bh);
4154 goto out_bdev;
4155 }
4156
4157 len = ext4_blocks_count(es);
4158 start = sb_block + 1;
4159 brelse(bh); /* we're done with the superblock */
4160
4161 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4162 start, len, blocksize);
4163 if (!journal) {
4164 ext4_msg(sb, KERN_ERR, "failed to create device journal");
4165 goto out_bdev;
4166 }
4167 journal->j_private = sb;
4168 ll_rw_block(READ, 1, &journal->j_sb_buffer);
4169 wait_on_buffer(journal->j_sb_buffer);
4170 if (!buffer_uptodate(journal->j_sb_buffer)) {
4171 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4172 goto out_journal;
4173 }
4174 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4175 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4176 "user (unsupported) - %d",
4177 be32_to_cpu(journal->j_superblock->s_nr_users));
4178 goto out_journal;
4179 }
4180 EXT4_SB(sb)->journal_bdev = bdev;
4181 ext4_init_journal_params(sb, journal);
4182 return journal;
4183
4184out_journal:
4185 jbd2_journal_destroy(journal);
4186out_bdev:
4187 ext4_blkdev_put(bdev);
4188 return NULL;
4189}
4190
4191static int ext4_load_journal(struct super_block *sb,
4192 struct ext4_super_block *es,
4193 unsigned long journal_devnum)
4194{
4195 journal_t *journal;
4196 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4197 dev_t journal_dev;
4198 int err = 0;
4199 int really_read_only;
4200
4201 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4202
4203 if (journal_devnum &&
4204 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4205 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4206 "numbers have changed");
4207 journal_dev = new_decode_dev(journal_devnum);
4208 } else
4209 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4210
4211 really_read_only = bdev_read_only(sb->s_bdev);
4212
4213 /*
4214 * Are we loading a blank journal or performing recovery after a
4215 * crash? For recovery, we need to check in advance whether we
4216 * can get read-write access to the device.
4217 */
4218 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
4219 if (sb->s_flags & MS_RDONLY) {
4220 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4221 "required on readonly filesystem");
4222 if (really_read_only) {
4223 ext4_msg(sb, KERN_ERR, "write access "
4224 "unavailable, cannot proceed");
4225 return -EROFS;
4226 }
4227 ext4_msg(sb, KERN_INFO, "write access will "
4228 "be enabled during recovery");
4229 }
4230 }
4231
4232 if (journal_inum && journal_dev) {
4233 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4234 "and inode journals!");
4235 return -EINVAL;
4236 }
4237
4238 if (journal_inum) {
4239 if (!(journal = ext4_get_journal(sb, journal_inum)))
4240 return -EINVAL;
4241 } else {
4242 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4243 return -EINVAL;
4244 }
4245
4246 if (!(journal->j_flags & JBD2_BARRIER))
4247 ext4_msg(sb, KERN_INFO, "barriers disabled");
4248
4249 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
4250 err = jbd2_journal_wipe(journal, !really_read_only);
4251 if (!err) {
4252 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4253 if (save)
4254 memcpy(save, ((char *) es) +
4255 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4256 err = jbd2_journal_load(journal);
4257 if (save)
4258 memcpy(((char *) es) + EXT4_S_ERR_START,
4259 save, EXT4_S_ERR_LEN);
4260 kfree(save);
4261 }
4262
4263 if (err) {
4264 ext4_msg(sb, KERN_ERR, "error loading journal");
4265 jbd2_journal_destroy(journal);
4266 return err;
4267 }
4268
4269 EXT4_SB(sb)->s_journal = journal;
4270 ext4_clear_journal_err(sb, es);
4271
4272 if (!really_read_only && journal_devnum &&
4273 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4274 es->s_journal_dev = cpu_to_le32(journal_devnum);
4275
4276 /* Make sure we flush the recovery flag to disk. */
4277 ext4_commit_super(sb, 1);
4278 }
4279
4280 return 0;
4281}
4282
4283static int ext4_commit_super(struct super_block *sb, int sync)
4284{
4285 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4286 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4287 int error = 0;
4288
4289 if (!sbh || block_device_ejected(sb))
4290 return error;
4291 if (buffer_write_io_error(sbh)) {
4292 /*
4293 * Oh, dear. A previous attempt to write the
4294 * superblock failed. This could happen because the
4295 * USB device was yanked out. Or it could happen to
4296 * be a transient write error and maybe the block will
4297 * be remapped. Nothing we can do but to retry the
4298 * write and hope for the best.
4299 */
4300 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4301 "superblock detected");
4302 clear_buffer_write_io_error(sbh);
4303 set_buffer_uptodate(sbh);
4304 }
4305 /*
4306 * If the file system is mounted read-only, don't update the
4307 * superblock write time. This avoids updating the superblock
4308 * write time when we are mounting the root file system
4309 * read/only but we need to replay the journal; at that point,
4310 * for people who are east of GMT and who make their clock
4311 * tick in localtime for Windows bug-for-bug compatibility,
4312 * the clock is set in the future, and this will cause e2fsck
4313 * to complain and force a full file system check.
4314 */
4315 if (!(sb->s_flags & MS_RDONLY))
4316 es->s_wtime = cpu_to_le32(get_seconds());
4317 if (sb->s_bdev->bd_part)
4318 es->s_kbytes_written =
4319 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4320 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4321 EXT4_SB(sb)->s_sectors_written_start) >> 1));
4322 else
4323 es->s_kbytes_written =
4324 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4325 ext4_free_blocks_count_set(es,
4326 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4327 &EXT4_SB(sb)->s_freeclusters_counter)));
4328 es->s_free_inodes_count =
4329 cpu_to_le32(percpu_counter_sum_positive(
4330 &EXT4_SB(sb)->s_freeinodes_counter));
4331 sb->s_dirt = 0;
4332 BUFFER_TRACE(sbh, "marking dirty");
4333 ext4_superblock_csum_set(sb, es);
4334 mark_buffer_dirty(sbh);
4335 if (sync) {
4336 error = sync_dirty_buffer(sbh);
4337 if (error)
4338 return error;
4339
4340 error = buffer_write_io_error(sbh);
4341 if (error) {
4342 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4343 "superblock");
4344 clear_buffer_write_io_error(sbh);
4345 set_buffer_uptodate(sbh);
4346 }
4347 }
4348 return error;
4349}
4350
4351/*
4352 * Have we just finished recovery? If so, and if we are mounting (or
4353 * remounting) the filesystem readonly, then we will end up with a
4354 * consistent fs on disk. Record that fact.
4355 */
4356static void ext4_mark_recovery_complete(struct super_block *sb,
4357 struct ext4_super_block *es)
4358{
4359 journal_t *journal = EXT4_SB(sb)->s_journal;
4360
4361 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4362 BUG_ON(journal != NULL);
4363 return;
4364 }
4365 jbd2_journal_lock_updates(journal);
4366 if (jbd2_journal_flush(journal) < 0)
4367 goto out;
4368
4369 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
4370 sb->s_flags & MS_RDONLY) {
4371 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4372 ext4_commit_super(sb, 1);
4373 }
4374
4375out:
4376 jbd2_journal_unlock_updates(journal);
4377}
4378
4379/*
4380 * If we are mounting (or read-write remounting) a filesystem whose journal
4381 * has recorded an error from a previous lifetime, move that error to the
4382 * main filesystem now.
4383 */
4384static void ext4_clear_journal_err(struct super_block *sb,
4385 struct ext4_super_block *es)
4386{
4387 journal_t *journal;
4388 int j_errno;
4389 const char *errstr;
4390
4391 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4392
4393 journal = EXT4_SB(sb)->s_journal;
4394
4395 /*
4396 * Now check for any error status which may have been recorded in the
4397 * journal by a prior ext4_error() or ext4_abort()
4398 */
4399
4400 j_errno = jbd2_journal_errno(journal);
4401 if (j_errno) {
4402 char nbuf[16];
4403
4404 errstr = ext4_decode_error(sb, j_errno, nbuf);
4405 ext4_warning(sb, "Filesystem error recorded "
4406 "from previous mount: %s", errstr);
4407 ext4_warning(sb, "Marking fs in need of filesystem check.");
4408
4409 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4410 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4411 ext4_commit_super(sb, 1);
4412
4413 jbd2_journal_clear_err(journal);
4414 jbd2_journal_update_sb_errno(journal);
4415 }
4416}
4417
4418/*
4419 * Force the running and committing transactions to commit,
4420 * and wait on the commit.
4421 */
4422int ext4_force_commit(struct super_block *sb)
4423{
4424 journal_t *journal;
4425 int ret = 0;
4426
4427 if (sb->s_flags & MS_RDONLY)
4428 return 0;
4429
4430 journal = EXT4_SB(sb)->s_journal;
4431 if (journal) {
4432 vfs_check_frozen(sb, SB_FREEZE_TRANS);
4433 ret = ext4_journal_force_commit(journal);
4434 }
4435
4436 return ret;
4437}
4438
4439static void ext4_write_super(struct super_block *sb)
4440{
4441 lock_super(sb);
4442 ext4_commit_super(sb, 1);
4443 unlock_super(sb);
4444}
4445
4446static int ext4_sync_fs(struct super_block *sb, int wait)
4447{
4448 int ret = 0;
4449 tid_t target;
4450 struct ext4_sb_info *sbi = EXT4_SB(sb);
4451
4452 trace_ext4_sync_fs(sb, wait);
4453 flush_workqueue(sbi->dio_unwritten_wq);
4454 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4455 if (wait)
4456 jbd2_log_wait_commit(sbi->s_journal, target);
4457 }
4458 return ret;
4459}
4460
4461/*
4462 * LVM calls this function before a (read-only) snapshot is created. This
4463 * gives us a chance to flush the journal completely and mark the fs clean.
4464 *
4465 * Note that only this function cannot bring a filesystem to be in a clean
4466 * state independently, because ext4 prevents a new handle from being started
4467 * by @sb->s_frozen, which stays in an upper layer. It thus needs help from
4468 * the upper layer.
4469 */
4470static int ext4_freeze(struct super_block *sb)
4471{
4472 int error = 0;
4473 journal_t *journal;
4474
4475 if (sb->s_flags & MS_RDONLY)
4476 return 0;
4477
4478 journal = EXT4_SB(sb)->s_journal;
4479
4480 /* Now we set up the journal barrier. */
4481 jbd2_journal_lock_updates(journal);
4482
4483 /*
4484 * Don't clear the needs_recovery flag if we failed to flush
4485 * the journal.
4486 */
4487 error = jbd2_journal_flush(journal);
4488 if (error < 0)
4489 goto out;
4490
4491 /* Journal blocked and flushed, clear needs_recovery flag. */
4492 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4493 error = ext4_commit_super(sb, 1);
4494out:
4495 /* we rely on s_frozen to stop further updates */
4496 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4497 return error;
4498}
4499
4500/*
4501 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4502 * flag here, even though the filesystem is not technically dirty yet.
4503 */
4504static int ext4_unfreeze(struct super_block *sb)
4505{
4506 if (sb->s_flags & MS_RDONLY)
4507 return 0;
4508
4509 lock_super(sb);
4510 /* Reset the needs_recovery flag before the fs is unlocked. */
4511 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4512 ext4_commit_super(sb, 1);
4513 unlock_super(sb);
4514 return 0;
4515}
4516
4517/*
4518 * Structure to save mount options for ext4_remount's benefit
4519 */
4520struct ext4_mount_options {
4521 unsigned long s_mount_opt;
4522 unsigned long s_mount_opt2;
4523 kuid_t s_resuid;
4524 kgid_t s_resgid;
4525 unsigned long s_commit_interval;
4526 u32 s_min_batch_time, s_max_batch_time;
4527#ifdef CONFIG_QUOTA
4528 int s_jquota_fmt;
4529 char *s_qf_names[MAXQUOTAS];
4530#endif
4531};
4532
4533static int ext4_remount(struct super_block *sb, int *flags, char *data)
4534{
4535 struct ext4_super_block *es;
4536 struct ext4_sb_info *sbi = EXT4_SB(sb);
4537 unsigned long old_sb_flags;
4538 struct ext4_mount_options old_opts;
4539 int enable_quota = 0;
4540 ext4_group_t g;
4541 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4542 int err = 0;
4543#ifdef CONFIG_QUOTA
4544 int i;
4545#endif
4546 char *orig_data = kstrdup(data, GFP_KERNEL);
4547
4548 /* Store the original options */
4549 lock_super(sb);
4550 old_sb_flags = sb->s_flags;
4551 old_opts.s_mount_opt = sbi->s_mount_opt;
4552 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4553 old_opts.s_resuid = sbi->s_resuid;
4554 old_opts.s_resgid = sbi->s_resgid;
4555 old_opts.s_commit_interval = sbi->s_commit_interval;
4556 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4557 old_opts.s_max_batch_time = sbi->s_max_batch_time;
4558#ifdef CONFIG_QUOTA
4559 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4560 for (i = 0; i < MAXQUOTAS; i++)
4561 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
4562#endif
4563 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4564 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4565
4566 /*
4567 * Allow the "check" option to be passed as a remount option.
4568 */
4569 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
4570 err = -EINVAL;
4571 goto restore_opts;
4572 }
4573
4574 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4575 ext4_abort(sb, "Abort forced by user");
4576
4577 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
4578 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
4579
4580 es = sbi->s_es;
4581
4582 if (sbi->s_journal) {
4583 ext4_init_journal_params(sb, sbi->s_journal);
4584 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4585 }
4586
4587 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4588 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
4589 err = -EROFS;
4590 goto restore_opts;
4591 }
4592
4593 if (*flags & MS_RDONLY) {
4594 err = dquot_suspend(sb, -1);
4595 if (err < 0)
4596 goto restore_opts;
4597
4598 /*
4599 * First of all, the unconditional stuff we have to do
4600 * to disable replay of the journal when we next remount
4601 */
4602 sb->s_flags |= MS_RDONLY;
4603
4604 /*
4605 * OK, test if we are remounting a valid rw partition
4606 * readonly, and if so set the rdonly flag and then
4607 * mark the partition as valid again.
4608 */
4609 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4610 (sbi->s_mount_state & EXT4_VALID_FS))
4611 es->s_state = cpu_to_le16(sbi->s_mount_state);
4612
4613 if (sbi->s_journal)
4614 ext4_mark_recovery_complete(sb, es);
4615 } else {
4616 /* Make sure we can mount this feature set readwrite */
4617 if (!ext4_feature_set_ok(sb, 0)) {
4618 err = -EROFS;
4619 goto restore_opts;
4620 }
4621 /*
4622 * Make sure the group descriptor checksums
4623 * are sane. If they aren't, refuse to remount r/w.
4624 */
4625 for (g = 0; g < sbi->s_groups_count; g++) {
4626 struct ext4_group_desc *gdp =
4627 ext4_get_group_desc(sb, g, NULL);
4628
4629 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
4630 ext4_msg(sb, KERN_ERR,
4631 "ext4_remount: Checksum for group %u failed (%u!=%u)",
4632 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4633 le16_to_cpu(gdp->bg_checksum));
4634 err = -EINVAL;
4635 goto restore_opts;
4636 }
4637 }
4638
4639 /*
4640 * If we have an unprocessed orphan list hanging
4641 * around from a previously readonly bdev mount,
4642 * require a full umount/remount for now.
4643 */
4644 if (es->s_last_orphan) {
4645 ext4_msg(sb, KERN_WARNING, "Couldn't "
4646 "remount RDWR because of unprocessed "
4647 "orphan inode list. Please "
4648 "umount/remount instead");
4649 err = -EINVAL;
4650 goto restore_opts;
4651 }
4652
4653 /*
4654 * Mounting a RDONLY partition read-write, so reread
4655 * and store the current valid flag. (It may have
4656 * been changed by e2fsck since we originally mounted
4657 * the partition.)
4658 */
4659 if (sbi->s_journal)
4660 ext4_clear_journal_err(sb, es);
4661 sbi->s_mount_state = le16_to_cpu(es->s_state);
4662 if (!ext4_setup_super(sb, es, 0))
4663 sb->s_flags &= ~MS_RDONLY;
4664 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
4665 EXT4_FEATURE_INCOMPAT_MMP))
4666 if (ext4_multi_mount_protect(sb,
4667 le64_to_cpu(es->s_mmp_block))) {
4668 err = -EROFS;
4669 goto restore_opts;
4670 }
4671 enable_quota = 1;
4672 }
4673 }
4674
4675 /*
4676 * Reinitialize lazy itable initialization thread based on
4677 * current settings
4678 */
4679 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4680 ext4_unregister_li_request(sb);
4681 else {
4682 ext4_group_t first_not_zeroed;
4683 first_not_zeroed = ext4_has_uninit_itable(sb);
4684 ext4_register_li_request(sb, first_not_zeroed);
4685 }
4686
4687 ext4_setup_system_zone(sb);
4688 if (sbi->s_journal == NULL)
4689 ext4_commit_super(sb, 1);
4690
4691#ifdef CONFIG_QUOTA
4692 /* Release old quota file names */
4693 for (i = 0; i < MAXQUOTAS; i++)
4694 if (old_opts.s_qf_names[i] &&
4695 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4696 kfree(old_opts.s_qf_names[i]);
4697#endif
4698 unlock_super(sb);
4699 if (enable_quota)
4700 dquot_resume(sb, -1);
4701
4702 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4703 kfree(orig_data);
4704 return 0;
4705
4706restore_opts:
4707 sb->s_flags = old_sb_flags;
4708 sbi->s_mount_opt = old_opts.s_mount_opt;
4709 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
4710 sbi->s_resuid = old_opts.s_resuid;
4711 sbi->s_resgid = old_opts.s_resgid;
4712 sbi->s_commit_interval = old_opts.s_commit_interval;
4713 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4714 sbi->s_max_batch_time = old_opts.s_max_batch_time;
4715#ifdef CONFIG_QUOTA
4716 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4717 for (i = 0; i < MAXQUOTAS; i++) {
4718 if (sbi->s_qf_names[i] &&
4719 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4720 kfree(sbi->s_qf_names[i]);
4721 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4722 }
4723#endif
4724 unlock_super(sb);
4725 kfree(orig_data);
4726 return err;
4727}
4728
4729static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
4730{
4731 struct super_block *sb = dentry->d_sb;
4732 struct ext4_sb_info *sbi = EXT4_SB(sb);
4733 struct ext4_super_block *es = sbi->s_es;
4734 ext4_fsblk_t overhead = 0;
4735 u64 fsid;
4736 s64 bfree;
4737
4738 if (!test_opt(sb, MINIX_DF))
4739 overhead = sbi->s_overhead;
4740
4741 buf->f_type = EXT4_SUPER_MAGIC;
4742 buf->f_bsize = sb->s_blocksize;
4743 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, sbi->s_overhead);
4744 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
4745 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
4746 /* prevent underflow in case that few free space is available */
4747 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
4748 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
4749 if (buf->f_bfree < ext4_r_blocks_count(es))
4750 buf->f_bavail = 0;
4751 buf->f_files = le32_to_cpu(es->s_inodes_count);
4752 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
4753 buf->f_namelen = EXT4_NAME_LEN;
4754 fsid = le64_to_cpup((void *)es->s_uuid) ^
4755 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4756 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4757 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
4758
4759 return 0;
4760}
4761
4762/* Helper function for writing quotas on sync - we need to start transaction
4763 * before quota file is locked for write. Otherwise the are possible deadlocks:
4764 * Process 1 Process 2
4765 * ext4_create() quota_sync()
4766 * jbd2_journal_start() write_dquot()
4767 * dquot_initialize() down(dqio_mutex)
4768 * down(dqio_mutex) jbd2_journal_start()
4769 *
4770 */
4771
4772#ifdef CONFIG_QUOTA
4773
4774static inline struct inode *dquot_to_inode(struct dquot *dquot)
4775{
4776 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
4777}
4778
4779static int ext4_write_dquot(struct dquot *dquot)
4780{
4781 int ret, err;
4782 handle_t *handle;
4783 struct inode *inode;
4784
4785 inode = dquot_to_inode(dquot);
4786 handle = ext4_journal_start(inode,
4787 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
4788 if (IS_ERR(handle))
4789 return PTR_ERR(handle);
4790 ret = dquot_commit(dquot);
4791 err = ext4_journal_stop(handle);
4792 if (!ret)
4793 ret = err;
4794 return ret;
4795}
4796
4797static int ext4_acquire_dquot(struct dquot *dquot)
4798{
4799 int ret, err;
4800 handle_t *handle;
4801
4802 handle = ext4_journal_start(dquot_to_inode(dquot),
4803 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
4804 if (IS_ERR(handle))
4805 return PTR_ERR(handle);
4806 ret = dquot_acquire(dquot);
4807 err = ext4_journal_stop(handle);
4808 if (!ret)
4809 ret = err;
4810 return ret;
4811}
4812
4813static int ext4_release_dquot(struct dquot *dquot)
4814{
4815 int ret, err;
4816 handle_t *handle;
4817
4818 handle = ext4_journal_start(dquot_to_inode(dquot),
4819 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
4820 if (IS_ERR(handle)) {
4821 /* Release dquot anyway to avoid endless cycle in dqput() */
4822 dquot_release(dquot);
4823 return PTR_ERR(handle);
4824 }
4825 ret = dquot_release(dquot);
4826 err = ext4_journal_stop(handle);
4827 if (!ret)
4828 ret = err;
4829 return ret;
4830}
4831
4832static int ext4_mark_dquot_dirty(struct dquot *dquot)
4833{
4834 /* Are we journaling quotas? */
4835 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
4836 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
4837 dquot_mark_dquot_dirty(dquot);
4838 return ext4_write_dquot(dquot);
4839 } else {
4840 return dquot_mark_dquot_dirty(dquot);
4841 }
4842}
4843
4844static int ext4_write_info(struct super_block *sb, int type)
4845{
4846 int ret, err;
4847 handle_t *handle;
4848
4849 /* Data block + inode block */
4850 handle = ext4_journal_start(sb->s_root->d_inode, 2);
4851 if (IS_ERR(handle))
4852 return PTR_ERR(handle);
4853 ret = dquot_commit_info(sb, type);
4854 err = ext4_journal_stop(handle);
4855 if (!ret)
4856 ret = err;
4857 return ret;
4858}
4859
4860/*
4861 * Turn on quotas during mount time - we need to find
4862 * the quota file and such...
4863 */
4864static int ext4_quota_on_mount(struct super_block *sb, int type)
4865{
4866 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4867 EXT4_SB(sb)->s_jquota_fmt, type);
4868}
4869
4870/*
4871 * Standard function to be called on quota_on
4872 */
4873static int ext4_quota_on(struct super_block *sb, int type, int format_id,
4874 struct path *path)
4875{
4876 int err;
4877
4878 if (!test_opt(sb, QUOTA))
4879 return -EINVAL;
4880
4881 /* Quotafile not on the same filesystem? */
4882 if (path->dentry->d_sb != sb)
4883 return -EXDEV;
4884 /* Journaling quota? */
4885 if (EXT4_SB(sb)->s_qf_names[type]) {
4886 /* Quotafile not in fs root? */
4887 if (path->dentry->d_parent != sb->s_root)
4888 ext4_msg(sb, KERN_WARNING,
4889 "Quota file not on filesystem root. "
4890 "Journaled quota will not work");
4891 }
4892
4893 /*
4894 * When we journal data on quota file, we have to flush journal to see
4895 * all updates to the file when we bypass pagecache...
4896 */
4897 if (EXT4_SB(sb)->s_journal &&
4898 ext4_should_journal_data(path->dentry->d_inode)) {
4899 /*
4900 * We don't need to lock updates but journal_flush() could
4901 * otherwise be livelocked...
4902 */
4903 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
4904 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
4905 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4906 if (err)
4907 return err;
4908 }
4909
4910 return dquot_quota_on(sb, type, format_id, path);
4911}
4912
4913static int ext4_quota_off(struct super_block *sb, int type)
4914{
4915 struct inode *inode = sb_dqopt(sb)->files[type];
4916 handle_t *handle;
4917
4918 /* Force all delayed allocation blocks to be allocated.
4919 * Caller already holds s_umount sem */
4920 if (test_opt(sb, DELALLOC))
4921 sync_filesystem(sb);
4922
4923 if (!inode)
4924 goto out;
4925
4926 /* Update modification times of quota files when userspace can
4927 * start looking at them */
4928 handle = ext4_journal_start(inode, 1);
4929 if (IS_ERR(handle))
4930 goto out;
4931 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
4932 ext4_mark_inode_dirty(handle, inode);
4933 ext4_journal_stop(handle);
4934
4935out:
4936 return dquot_quota_off(sb, type);
4937}
4938
4939/* Read data from quotafile - avoid pagecache and such because we cannot afford
4940 * acquiring the locks... As quota files are never truncated and quota code
4941 * itself serializes the operations (and no one else should touch the files)
4942 * we don't have to be afraid of races */
4943static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
4944 size_t len, loff_t off)
4945{
4946 struct inode *inode = sb_dqopt(sb)->files[type];
4947 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
4948 int err = 0;
4949 int offset = off & (sb->s_blocksize - 1);
4950 int tocopy;
4951 size_t toread;
4952 struct buffer_head *bh;
4953 loff_t i_size = i_size_read(inode);
4954
4955 if (off > i_size)
4956 return 0;
4957 if (off+len > i_size)
4958 len = i_size-off;
4959 toread = len;
4960 while (toread > 0) {
4961 tocopy = sb->s_blocksize - offset < toread ?
4962 sb->s_blocksize - offset : toread;
4963 bh = ext4_bread(NULL, inode, blk, 0, &err);
4964 if (err)
4965 return err;
4966 if (!bh) /* A hole? */
4967 memset(data, 0, tocopy);
4968 else
4969 memcpy(data, bh->b_data+offset, tocopy);
4970 brelse(bh);
4971 offset = 0;
4972 toread -= tocopy;
4973 data += tocopy;
4974 blk++;
4975 }
4976 return len;
4977}
4978
4979/* Write to quotafile (we know the transaction is already started and has
4980 * enough credits) */
4981static ssize_t ext4_quota_write(struct super_block *sb, int type,
4982 const char *data, size_t len, loff_t off)
4983{
4984 struct inode *inode = sb_dqopt(sb)->files[type];
4985 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
4986 int err = 0;
4987 int offset = off & (sb->s_blocksize - 1);
4988 struct buffer_head *bh;
4989 handle_t *handle = journal_current_handle();
4990
4991 if (EXT4_SB(sb)->s_journal && !handle) {
4992 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
4993 " cancelled because transaction is not started",
4994 (unsigned long long)off, (unsigned long long)len);
4995 return -EIO;
4996 }
4997 /*
4998 * Since we account only one data block in transaction credits,
4999 * then it is impossible to cross a block boundary.
5000 */
5001 if (sb->s_blocksize - offset < len) {
5002 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5003 " cancelled because not block aligned",
5004 (unsigned long long)off, (unsigned long long)len);
5005 return -EIO;
5006 }
5007
5008 bh = ext4_bread(handle, inode, blk, 1, &err);
5009 if (!bh)
5010 goto out;
5011 err = ext4_journal_get_write_access(handle, bh);
5012 if (err) {
5013 brelse(bh);
5014 goto out;
5015 }
5016 lock_buffer(bh);
5017 memcpy(bh->b_data+offset, data, len);
5018 flush_dcache_page(bh->b_page);
5019 unlock_buffer(bh);
5020 err = ext4_handle_dirty_metadata(handle, NULL, bh);
5021 brelse(bh);
5022out:
5023 if (err)
5024 return err;
5025 if (inode->i_size < off + len) {
5026 i_size_write(inode, off + len);
5027 EXT4_I(inode)->i_disksize = inode->i_size;
5028 ext4_mark_inode_dirty(handle, inode);
5029 }
5030 return len;
5031}
5032
5033#endif
5034
5035static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5036 const char *dev_name, void *data)
5037{
5038 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
5039}
5040
5041#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5042static inline void register_as_ext2(void)
5043{
5044 int err = register_filesystem(&ext2_fs_type);
5045 if (err)
5046 printk(KERN_WARNING
5047 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5048}
5049
5050static inline void unregister_as_ext2(void)
5051{
5052 unregister_filesystem(&ext2_fs_type);
5053}
5054
5055static inline int ext2_feature_set_ok(struct super_block *sb)
5056{
5057 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
5058 return 0;
5059 if (sb->s_flags & MS_RDONLY)
5060 return 1;
5061 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
5062 return 0;
5063 return 1;
5064}
5065MODULE_ALIAS("ext2");
5066#else
5067static inline void register_as_ext2(void) { }
5068static inline void unregister_as_ext2(void) { }
5069static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
5070#endif
5071
5072#if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5073static inline void register_as_ext3(void)
5074{
5075 int err = register_filesystem(&ext3_fs_type);
5076 if (err)
5077 printk(KERN_WARNING
5078 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5079}
5080
5081static inline void unregister_as_ext3(void)
5082{
5083 unregister_filesystem(&ext3_fs_type);
5084}
5085
5086static inline int ext3_feature_set_ok(struct super_block *sb)
5087{
5088 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
5089 return 0;
5090 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
5091 return 0;
5092 if (sb->s_flags & MS_RDONLY)
5093 return 1;
5094 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
5095 return 0;
5096 return 1;
5097}
5098MODULE_ALIAS("ext3");
5099#else
5100static inline void register_as_ext3(void) { }
5101static inline void unregister_as_ext3(void) { }
5102static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
5103#endif
5104
5105static struct file_system_type ext4_fs_type = {
5106 .owner = THIS_MODULE,
5107 .name = "ext4",
5108 .mount = ext4_mount,
5109 .kill_sb = kill_block_super,
5110 .fs_flags = FS_REQUIRES_DEV,
5111};
5112
5113static int __init ext4_init_feat_adverts(void)
5114{
5115 struct ext4_features *ef;
5116 int ret = -ENOMEM;
5117
5118 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
5119 if (!ef)
5120 goto out;
5121
5122 ef->f_kobj.kset = ext4_kset;
5123 init_completion(&ef->f_kobj_unregister);
5124 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
5125 "features");
5126 if (ret) {
5127 kfree(ef);
5128 goto out;
5129 }
5130
5131 ext4_feat = ef;
5132 ret = 0;
5133out:
5134 return ret;
5135}
5136
5137static void ext4_exit_feat_adverts(void)
5138{
5139 kobject_put(&ext4_feat->f_kobj);
5140 wait_for_completion(&ext4_feat->f_kobj_unregister);
5141 kfree(ext4_feat);
5142}
5143
5144/* Shared across all ext4 file systems */
5145wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5146struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5147
5148static int __init ext4_init_fs(void)
5149{
5150 int i, err;
5151
5152 ext4_li_info = NULL;
5153 mutex_init(&ext4_li_mtx);
5154
5155 ext4_check_flag_values();
5156
5157 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5158 mutex_init(&ext4__aio_mutex[i]);
5159 init_waitqueue_head(&ext4__ioend_wq[i]);
5160 }
5161
5162 err = ext4_init_pageio();
5163 if (err)
5164 return err;
5165 err = ext4_init_system_zone();
5166 if (err)
5167 goto out6;
5168 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
5169 if (!ext4_kset)
5170 goto out5;
5171 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
5172
5173 err = ext4_init_feat_adverts();
5174 if (err)
5175 goto out4;
5176
5177 err = ext4_init_mballoc();
5178 if (err)
5179 goto out3;
5180
5181 err = ext4_init_xattr();
5182 if (err)
5183 goto out2;
5184 err = init_inodecache();
5185 if (err)
5186 goto out1;
5187 register_as_ext3();
5188 register_as_ext2();
5189 err = register_filesystem(&ext4_fs_type);
5190 if (err)
5191 goto out;
5192
5193 return 0;
5194out:
5195 unregister_as_ext2();
5196 unregister_as_ext3();
5197 destroy_inodecache();
5198out1:
5199 ext4_exit_xattr();
5200out2:
5201 ext4_exit_mballoc();
5202out3:
5203 ext4_exit_feat_adverts();
5204out4:
5205 if (ext4_proc_root)
5206 remove_proc_entry("fs/ext4", NULL);
5207 kset_unregister(ext4_kset);
5208out5:
5209 ext4_exit_system_zone();
5210out6:
5211 ext4_exit_pageio();
5212 return err;
5213}
5214
5215static void __exit ext4_exit_fs(void)
5216{
5217 ext4_destroy_lazyinit_thread();
5218 unregister_as_ext2();
5219 unregister_as_ext3();
5220 unregister_filesystem(&ext4_fs_type);
5221 destroy_inodecache();
5222 ext4_exit_xattr();
5223 ext4_exit_mballoc();
5224 ext4_exit_feat_adverts();
5225 remove_proc_entry("fs/ext4", NULL);
5226 kset_unregister(ext4_kset);
5227 ext4_exit_system_zone();
5228 ext4_exit_pageio();
5229}
5230
5231MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5232MODULE_DESCRIPTION("Fourth Extended Filesystem");
5233MODULE_LICENSE("GPL");
5234module_init(ext4_init_fs)
5235module_exit(ext4_exit_fs)
1// SPDX-License-Identifier: GPL-2.0
2/*
3 * linux/fs/ext4/super.c
4 *
5 * Copyright (C) 1992, 1993, 1994, 1995
6 * Remy Card (card@masi.ibp.fr)
7 * Laboratoire MASI - Institut Blaise Pascal
8 * Universite Pierre et Marie Curie (Paris VI)
9 *
10 * from
11 *
12 * linux/fs/minix/inode.c
13 *
14 * Copyright (C) 1991, 1992 Linus Torvalds
15 *
16 * Big-endian to little-endian byte-swapping/bitmaps by
17 * David S. Miller (davem@caip.rutgers.edu), 1995
18 */
19
20#include <linux/module.h>
21#include <linux/string.h>
22#include <linux/fs.h>
23#include <linux/time.h>
24#include <linux/vmalloc.h>
25#include <linux/slab.h>
26#include <linux/init.h>
27#include <linux/blkdev.h>
28#include <linux/backing-dev.h>
29#include <linux/parser.h>
30#include <linux/buffer_head.h>
31#include <linux/exportfs.h>
32#include <linux/vfs.h>
33#include <linux/random.h>
34#include <linux/mount.h>
35#include <linux/namei.h>
36#include <linux/quotaops.h>
37#include <linux/seq_file.h>
38#include <linux/ctype.h>
39#include <linux/log2.h>
40#include <linux/crc16.h>
41#include <linux/dax.h>
42#include <linux/uaccess.h>
43#include <linux/iversion.h>
44#include <linux/unicode.h>
45#include <linux/part_stat.h>
46#include <linux/kthread.h>
47#include <linux/freezer.h>
48#include <linux/fsnotify.h>
49#include <linux/fs_context.h>
50#include <linux/fs_parser.h>
51
52#include "ext4.h"
53#include "ext4_extents.h" /* Needed for trace points definition */
54#include "ext4_jbd2.h"
55#include "xattr.h"
56#include "acl.h"
57#include "mballoc.h"
58#include "fsmap.h"
59
60#define CREATE_TRACE_POINTS
61#include <trace/events/ext4.h>
62
63static struct ext4_lazy_init *ext4_li_info;
64static DEFINE_MUTEX(ext4_li_mtx);
65static struct ratelimit_state ext4_mount_msg_ratelimit;
66
67static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
68 unsigned long journal_devnum);
69static int ext4_show_options(struct seq_file *seq, struct dentry *root);
70static void ext4_update_super(struct super_block *sb);
71static int ext4_commit_super(struct super_block *sb);
72static int ext4_mark_recovery_complete(struct super_block *sb,
73 struct ext4_super_block *es);
74static int ext4_clear_journal_err(struct super_block *sb,
75 struct ext4_super_block *es);
76static int ext4_sync_fs(struct super_block *sb, int wait);
77static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
78static int ext4_unfreeze(struct super_block *sb);
79static int ext4_freeze(struct super_block *sb);
80static inline int ext2_feature_set_ok(struct super_block *sb);
81static inline int ext3_feature_set_ok(struct super_block *sb);
82static void ext4_destroy_lazyinit_thread(void);
83static void ext4_unregister_li_request(struct super_block *sb);
84static void ext4_clear_request_list(void);
85static struct inode *ext4_get_journal_inode(struct super_block *sb,
86 unsigned int journal_inum);
87static int ext4_validate_options(struct fs_context *fc);
88static int ext4_check_opt_consistency(struct fs_context *fc,
89 struct super_block *sb);
90static void ext4_apply_options(struct fs_context *fc, struct super_block *sb);
91static int ext4_parse_param(struct fs_context *fc, struct fs_parameter *param);
92static int ext4_get_tree(struct fs_context *fc);
93static int ext4_reconfigure(struct fs_context *fc);
94static void ext4_fc_free(struct fs_context *fc);
95static int ext4_init_fs_context(struct fs_context *fc);
96static void ext4_kill_sb(struct super_block *sb);
97static const struct fs_parameter_spec ext4_param_specs[];
98
99/*
100 * Lock ordering
101 *
102 * page fault path:
103 * mmap_lock -> sb_start_pagefault -> invalidate_lock (r) -> transaction start
104 * -> page lock -> i_data_sem (rw)
105 *
106 * buffered write path:
107 * sb_start_write -> i_mutex -> mmap_lock
108 * sb_start_write -> i_mutex -> transaction start -> page lock ->
109 * i_data_sem (rw)
110 *
111 * truncate:
112 * sb_start_write -> i_mutex -> invalidate_lock (w) -> i_mmap_rwsem (w) ->
113 * page lock
114 * sb_start_write -> i_mutex -> invalidate_lock (w) -> transaction start ->
115 * i_data_sem (rw)
116 *
117 * direct IO:
118 * sb_start_write -> i_mutex -> mmap_lock
119 * sb_start_write -> i_mutex -> transaction start -> i_data_sem (rw)
120 *
121 * writepages:
122 * transaction start -> page lock(s) -> i_data_sem (rw)
123 */
124
125static const struct fs_context_operations ext4_context_ops = {
126 .parse_param = ext4_parse_param,
127 .get_tree = ext4_get_tree,
128 .reconfigure = ext4_reconfigure,
129 .free = ext4_fc_free,
130};
131
132
133#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
134static struct file_system_type ext2_fs_type = {
135 .owner = THIS_MODULE,
136 .name = "ext2",
137 .init_fs_context = ext4_init_fs_context,
138 .parameters = ext4_param_specs,
139 .kill_sb = ext4_kill_sb,
140 .fs_flags = FS_REQUIRES_DEV,
141};
142MODULE_ALIAS_FS("ext2");
143MODULE_ALIAS("ext2");
144#define IS_EXT2_SB(sb) ((sb)->s_type == &ext2_fs_type)
145#else
146#define IS_EXT2_SB(sb) (0)
147#endif
148
149
150static struct file_system_type ext3_fs_type = {
151 .owner = THIS_MODULE,
152 .name = "ext3",
153 .init_fs_context = ext4_init_fs_context,
154 .parameters = ext4_param_specs,
155 .kill_sb = ext4_kill_sb,
156 .fs_flags = FS_REQUIRES_DEV,
157};
158MODULE_ALIAS_FS("ext3");
159MODULE_ALIAS("ext3");
160#define IS_EXT3_SB(sb) ((sb)->s_type == &ext3_fs_type)
161
162
163static inline void __ext4_read_bh(struct buffer_head *bh, blk_opf_t op_flags,
164 bh_end_io_t *end_io)
165{
166 /*
167 * buffer's verified bit is no longer valid after reading from
168 * disk again due to write out error, clear it to make sure we
169 * recheck the buffer contents.
170 */
171 clear_buffer_verified(bh);
172
173 bh->b_end_io = end_io ? end_io : end_buffer_read_sync;
174 get_bh(bh);
175 submit_bh(REQ_OP_READ | op_flags, bh);
176}
177
178void ext4_read_bh_nowait(struct buffer_head *bh, blk_opf_t op_flags,
179 bh_end_io_t *end_io)
180{
181 BUG_ON(!buffer_locked(bh));
182
183 if (ext4_buffer_uptodate(bh)) {
184 unlock_buffer(bh);
185 return;
186 }
187 __ext4_read_bh(bh, op_flags, end_io);
188}
189
190int ext4_read_bh(struct buffer_head *bh, blk_opf_t op_flags, bh_end_io_t *end_io)
191{
192 BUG_ON(!buffer_locked(bh));
193
194 if (ext4_buffer_uptodate(bh)) {
195 unlock_buffer(bh);
196 return 0;
197 }
198
199 __ext4_read_bh(bh, op_flags, end_io);
200
201 wait_on_buffer(bh);
202 if (buffer_uptodate(bh))
203 return 0;
204 return -EIO;
205}
206
207int ext4_read_bh_lock(struct buffer_head *bh, blk_opf_t op_flags, bool wait)
208{
209 lock_buffer(bh);
210 if (!wait) {
211 ext4_read_bh_nowait(bh, op_flags, NULL);
212 return 0;
213 }
214 return ext4_read_bh(bh, op_flags, NULL);
215}
216
217/*
218 * This works like __bread_gfp() except it uses ERR_PTR for error
219 * returns. Currently with sb_bread it's impossible to distinguish
220 * between ENOMEM and EIO situations (since both result in a NULL
221 * return.
222 */
223static struct buffer_head *__ext4_sb_bread_gfp(struct super_block *sb,
224 sector_t block,
225 blk_opf_t op_flags, gfp_t gfp)
226{
227 struct buffer_head *bh;
228 int ret;
229
230 bh = sb_getblk_gfp(sb, block, gfp);
231 if (bh == NULL)
232 return ERR_PTR(-ENOMEM);
233 if (ext4_buffer_uptodate(bh))
234 return bh;
235
236 ret = ext4_read_bh_lock(bh, REQ_META | op_flags, true);
237 if (ret) {
238 put_bh(bh);
239 return ERR_PTR(ret);
240 }
241 return bh;
242}
243
244struct buffer_head *ext4_sb_bread(struct super_block *sb, sector_t block,
245 blk_opf_t op_flags)
246{
247 gfp_t gfp = mapping_gfp_constraint(sb->s_bdev->bd_inode->i_mapping,
248 ~__GFP_FS) | __GFP_MOVABLE;
249
250 return __ext4_sb_bread_gfp(sb, block, op_flags, gfp);
251}
252
253struct buffer_head *ext4_sb_bread_unmovable(struct super_block *sb,
254 sector_t block)
255{
256 gfp_t gfp = mapping_gfp_constraint(sb->s_bdev->bd_inode->i_mapping,
257 ~__GFP_FS);
258
259 return __ext4_sb_bread_gfp(sb, block, 0, gfp);
260}
261
262void ext4_sb_breadahead_unmovable(struct super_block *sb, sector_t block)
263{
264 struct buffer_head *bh = bdev_getblk(sb->s_bdev, block,
265 sb->s_blocksize, GFP_NOWAIT | __GFP_NOWARN);
266
267 if (likely(bh)) {
268 if (trylock_buffer(bh))
269 ext4_read_bh_nowait(bh, REQ_RAHEAD, NULL);
270 brelse(bh);
271 }
272}
273
274static int ext4_verify_csum_type(struct super_block *sb,
275 struct ext4_super_block *es)
276{
277 if (!ext4_has_feature_metadata_csum(sb))
278 return 1;
279
280 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
281}
282
283__le32 ext4_superblock_csum(struct super_block *sb,
284 struct ext4_super_block *es)
285{
286 struct ext4_sb_info *sbi = EXT4_SB(sb);
287 int offset = offsetof(struct ext4_super_block, s_checksum);
288 __u32 csum;
289
290 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
291
292 return cpu_to_le32(csum);
293}
294
295static int ext4_superblock_csum_verify(struct super_block *sb,
296 struct ext4_super_block *es)
297{
298 if (!ext4_has_metadata_csum(sb))
299 return 1;
300
301 return es->s_checksum == ext4_superblock_csum(sb, es);
302}
303
304void ext4_superblock_csum_set(struct super_block *sb)
305{
306 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
307
308 if (!ext4_has_metadata_csum(sb))
309 return;
310
311 es->s_checksum = ext4_superblock_csum(sb, es);
312}
313
314ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
315 struct ext4_group_desc *bg)
316{
317 return le32_to_cpu(bg->bg_block_bitmap_lo) |
318 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
319 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
320}
321
322ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
323 struct ext4_group_desc *bg)
324{
325 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
326 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
327 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
328}
329
330ext4_fsblk_t ext4_inode_table(struct super_block *sb,
331 struct ext4_group_desc *bg)
332{
333 return le32_to_cpu(bg->bg_inode_table_lo) |
334 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
335 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
336}
337
338__u32 ext4_free_group_clusters(struct super_block *sb,
339 struct ext4_group_desc *bg)
340{
341 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
342 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
343 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
344}
345
346__u32 ext4_free_inodes_count(struct super_block *sb,
347 struct ext4_group_desc *bg)
348{
349 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
350 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
351 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
352}
353
354__u32 ext4_used_dirs_count(struct super_block *sb,
355 struct ext4_group_desc *bg)
356{
357 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
358 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
359 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
360}
361
362__u32 ext4_itable_unused_count(struct super_block *sb,
363 struct ext4_group_desc *bg)
364{
365 return le16_to_cpu(bg->bg_itable_unused_lo) |
366 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
367 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
368}
369
370void ext4_block_bitmap_set(struct super_block *sb,
371 struct ext4_group_desc *bg, ext4_fsblk_t blk)
372{
373 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
374 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
375 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
376}
377
378void ext4_inode_bitmap_set(struct super_block *sb,
379 struct ext4_group_desc *bg, ext4_fsblk_t blk)
380{
381 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
382 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
383 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
384}
385
386void ext4_inode_table_set(struct super_block *sb,
387 struct ext4_group_desc *bg, ext4_fsblk_t blk)
388{
389 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
390 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
391 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
392}
393
394void ext4_free_group_clusters_set(struct super_block *sb,
395 struct ext4_group_desc *bg, __u32 count)
396{
397 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
398 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
399 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
400}
401
402void ext4_free_inodes_set(struct super_block *sb,
403 struct ext4_group_desc *bg, __u32 count)
404{
405 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
406 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
407 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
408}
409
410void ext4_used_dirs_set(struct super_block *sb,
411 struct ext4_group_desc *bg, __u32 count)
412{
413 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
414 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
415 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
416}
417
418void ext4_itable_unused_set(struct super_block *sb,
419 struct ext4_group_desc *bg, __u32 count)
420{
421 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
422 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
423 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
424}
425
426static void __ext4_update_tstamp(__le32 *lo, __u8 *hi, time64_t now)
427{
428 now = clamp_val(now, 0, (1ull << 40) - 1);
429
430 *lo = cpu_to_le32(lower_32_bits(now));
431 *hi = upper_32_bits(now);
432}
433
434static time64_t __ext4_get_tstamp(__le32 *lo, __u8 *hi)
435{
436 return ((time64_t)(*hi) << 32) + le32_to_cpu(*lo);
437}
438#define ext4_update_tstamp(es, tstamp) \
439 __ext4_update_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi, \
440 ktime_get_real_seconds())
441#define ext4_get_tstamp(es, tstamp) \
442 __ext4_get_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi)
443
444#define EXT4_SB_REFRESH_INTERVAL_SEC (3600) /* seconds (1 hour) */
445#define EXT4_SB_REFRESH_INTERVAL_KB (16384) /* kilobytes (16MB) */
446
447/*
448 * The ext4_maybe_update_superblock() function checks and updates the
449 * superblock if needed.
450 *
451 * This function is designed to update the on-disk superblock only under
452 * certain conditions to prevent excessive disk writes and unnecessary
453 * waking of the disk from sleep. The superblock will be updated if:
454 * 1. More than an hour has passed since the last superblock update, and
455 * 2. More than 16MB have been written since the last superblock update.
456 *
457 * @sb: The superblock
458 */
459static void ext4_maybe_update_superblock(struct super_block *sb)
460{
461 struct ext4_sb_info *sbi = EXT4_SB(sb);
462 struct ext4_super_block *es = sbi->s_es;
463 journal_t *journal = sbi->s_journal;
464 time64_t now;
465 __u64 last_update;
466 __u64 lifetime_write_kbytes;
467 __u64 diff_size;
468
469 if (sb_rdonly(sb) || !(sb->s_flags & SB_ACTIVE) ||
470 !journal || (journal->j_flags & JBD2_UNMOUNT))
471 return;
472
473 now = ktime_get_real_seconds();
474 last_update = ext4_get_tstamp(es, s_wtime);
475
476 if (likely(now - last_update < EXT4_SB_REFRESH_INTERVAL_SEC))
477 return;
478
479 lifetime_write_kbytes = sbi->s_kbytes_written +
480 ((part_stat_read(sb->s_bdev, sectors[STAT_WRITE]) -
481 sbi->s_sectors_written_start) >> 1);
482
483 /* Get the number of kilobytes not written to disk to account
484 * for statistics and compare with a multiple of 16 MB. This
485 * is used to determine when the next superblock commit should
486 * occur (i.e. not more often than once per 16MB if there was
487 * less written in an hour).
488 */
489 diff_size = lifetime_write_kbytes - le64_to_cpu(es->s_kbytes_written);
490
491 if (diff_size > EXT4_SB_REFRESH_INTERVAL_KB)
492 schedule_work(&EXT4_SB(sb)->s_sb_upd_work);
493}
494
495/*
496 * The del_gendisk() function uninitializes the disk-specific data
497 * structures, including the bdi structure, without telling anyone
498 * else. Once this happens, any attempt to call mark_buffer_dirty()
499 * (for example, by ext4_commit_super), will cause a kernel OOPS.
500 * This is a kludge to prevent these oops until we can put in a proper
501 * hook in del_gendisk() to inform the VFS and file system layers.
502 */
503static int block_device_ejected(struct super_block *sb)
504{
505 struct inode *bd_inode = sb->s_bdev->bd_inode;
506 struct backing_dev_info *bdi = inode_to_bdi(bd_inode);
507
508 return bdi->dev == NULL;
509}
510
511static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
512{
513 struct super_block *sb = journal->j_private;
514 struct ext4_sb_info *sbi = EXT4_SB(sb);
515 int error = is_journal_aborted(journal);
516 struct ext4_journal_cb_entry *jce;
517
518 BUG_ON(txn->t_state == T_FINISHED);
519
520 ext4_process_freed_data(sb, txn->t_tid);
521 ext4_maybe_update_superblock(sb);
522
523 spin_lock(&sbi->s_md_lock);
524 while (!list_empty(&txn->t_private_list)) {
525 jce = list_entry(txn->t_private_list.next,
526 struct ext4_journal_cb_entry, jce_list);
527 list_del_init(&jce->jce_list);
528 spin_unlock(&sbi->s_md_lock);
529 jce->jce_func(sb, jce, error);
530 spin_lock(&sbi->s_md_lock);
531 }
532 spin_unlock(&sbi->s_md_lock);
533}
534
535/*
536 * This writepage callback for write_cache_pages()
537 * takes care of a few cases after page cleaning.
538 *
539 * write_cache_pages() already checks for dirty pages
540 * and calls clear_page_dirty_for_io(), which we want,
541 * to write protect the pages.
542 *
543 * However, we may have to redirty a page (see below.)
544 */
545static int ext4_journalled_writepage_callback(struct folio *folio,
546 struct writeback_control *wbc,
547 void *data)
548{
549 transaction_t *transaction = (transaction_t *) data;
550 struct buffer_head *bh, *head;
551 struct journal_head *jh;
552
553 bh = head = folio_buffers(folio);
554 do {
555 /*
556 * We have to redirty a page in these cases:
557 * 1) If buffer is dirty, it means the page was dirty because it
558 * contains a buffer that needs checkpointing. So the dirty bit
559 * needs to be preserved so that checkpointing writes the buffer
560 * properly.
561 * 2) If buffer is not part of the committing transaction
562 * (we may have just accidentally come across this buffer because
563 * inode range tracking is not exact) or if the currently running
564 * transaction already contains this buffer as well, dirty bit
565 * needs to be preserved so that the buffer gets writeprotected
566 * properly on running transaction's commit.
567 */
568 jh = bh2jh(bh);
569 if (buffer_dirty(bh) ||
570 (jh && (jh->b_transaction != transaction ||
571 jh->b_next_transaction))) {
572 folio_redirty_for_writepage(wbc, folio);
573 goto out;
574 }
575 } while ((bh = bh->b_this_page) != head);
576
577out:
578 return AOP_WRITEPAGE_ACTIVATE;
579}
580
581static int ext4_journalled_submit_inode_data_buffers(struct jbd2_inode *jinode)
582{
583 struct address_space *mapping = jinode->i_vfs_inode->i_mapping;
584 struct writeback_control wbc = {
585 .sync_mode = WB_SYNC_ALL,
586 .nr_to_write = LONG_MAX,
587 .range_start = jinode->i_dirty_start,
588 .range_end = jinode->i_dirty_end,
589 };
590
591 return write_cache_pages(mapping, &wbc,
592 ext4_journalled_writepage_callback,
593 jinode->i_transaction);
594}
595
596static int ext4_journal_submit_inode_data_buffers(struct jbd2_inode *jinode)
597{
598 int ret;
599
600 if (ext4_should_journal_data(jinode->i_vfs_inode))
601 ret = ext4_journalled_submit_inode_data_buffers(jinode);
602 else
603 ret = ext4_normal_submit_inode_data_buffers(jinode);
604 return ret;
605}
606
607static int ext4_journal_finish_inode_data_buffers(struct jbd2_inode *jinode)
608{
609 int ret = 0;
610
611 if (!ext4_should_journal_data(jinode->i_vfs_inode))
612 ret = jbd2_journal_finish_inode_data_buffers(jinode);
613
614 return ret;
615}
616
617static bool system_going_down(void)
618{
619 return system_state == SYSTEM_HALT || system_state == SYSTEM_POWER_OFF
620 || system_state == SYSTEM_RESTART;
621}
622
623struct ext4_err_translation {
624 int code;
625 int errno;
626};
627
628#define EXT4_ERR_TRANSLATE(err) { .code = EXT4_ERR_##err, .errno = err }
629
630static struct ext4_err_translation err_translation[] = {
631 EXT4_ERR_TRANSLATE(EIO),
632 EXT4_ERR_TRANSLATE(ENOMEM),
633 EXT4_ERR_TRANSLATE(EFSBADCRC),
634 EXT4_ERR_TRANSLATE(EFSCORRUPTED),
635 EXT4_ERR_TRANSLATE(ENOSPC),
636 EXT4_ERR_TRANSLATE(ENOKEY),
637 EXT4_ERR_TRANSLATE(EROFS),
638 EXT4_ERR_TRANSLATE(EFBIG),
639 EXT4_ERR_TRANSLATE(EEXIST),
640 EXT4_ERR_TRANSLATE(ERANGE),
641 EXT4_ERR_TRANSLATE(EOVERFLOW),
642 EXT4_ERR_TRANSLATE(EBUSY),
643 EXT4_ERR_TRANSLATE(ENOTDIR),
644 EXT4_ERR_TRANSLATE(ENOTEMPTY),
645 EXT4_ERR_TRANSLATE(ESHUTDOWN),
646 EXT4_ERR_TRANSLATE(EFAULT),
647};
648
649static int ext4_errno_to_code(int errno)
650{
651 int i;
652
653 for (i = 0; i < ARRAY_SIZE(err_translation); i++)
654 if (err_translation[i].errno == errno)
655 return err_translation[i].code;
656 return EXT4_ERR_UNKNOWN;
657}
658
659static void save_error_info(struct super_block *sb, int error,
660 __u32 ino, __u64 block,
661 const char *func, unsigned int line)
662{
663 struct ext4_sb_info *sbi = EXT4_SB(sb);
664
665 /* We default to EFSCORRUPTED error... */
666 if (error == 0)
667 error = EFSCORRUPTED;
668
669 spin_lock(&sbi->s_error_lock);
670 sbi->s_add_error_count++;
671 sbi->s_last_error_code = error;
672 sbi->s_last_error_line = line;
673 sbi->s_last_error_ino = ino;
674 sbi->s_last_error_block = block;
675 sbi->s_last_error_func = func;
676 sbi->s_last_error_time = ktime_get_real_seconds();
677 if (!sbi->s_first_error_time) {
678 sbi->s_first_error_code = error;
679 sbi->s_first_error_line = line;
680 sbi->s_first_error_ino = ino;
681 sbi->s_first_error_block = block;
682 sbi->s_first_error_func = func;
683 sbi->s_first_error_time = sbi->s_last_error_time;
684 }
685 spin_unlock(&sbi->s_error_lock);
686}
687
688/* Deal with the reporting of failure conditions on a filesystem such as
689 * inconsistencies detected or read IO failures.
690 *
691 * On ext2, we can store the error state of the filesystem in the
692 * superblock. That is not possible on ext4, because we may have other
693 * write ordering constraints on the superblock which prevent us from
694 * writing it out straight away; and given that the journal is about to
695 * be aborted, we can't rely on the current, or future, transactions to
696 * write out the superblock safely.
697 *
698 * We'll just use the jbd2_journal_abort() error code to record an error in
699 * the journal instead. On recovery, the journal will complain about
700 * that error until we've noted it down and cleared it.
701 *
702 * If force_ro is set, we unconditionally force the filesystem into an
703 * ABORT|READONLY state, unless the error response on the fs has been set to
704 * panic in which case we take the easy way out and panic immediately. This is
705 * used to deal with unrecoverable failures such as journal IO errors or ENOMEM
706 * at a critical moment in log management.
707 */
708static void ext4_handle_error(struct super_block *sb, bool force_ro, int error,
709 __u32 ino, __u64 block,
710 const char *func, unsigned int line)
711{
712 journal_t *journal = EXT4_SB(sb)->s_journal;
713 bool continue_fs = !force_ro && test_opt(sb, ERRORS_CONT);
714
715 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
716 if (test_opt(sb, WARN_ON_ERROR))
717 WARN_ON_ONCE(1);
718
719 if (!continue_fs && !sb_rdonly(sb)) {
720 set_bit(EXT4_FLAGS_SHUTDOWN, &EXT4_SB(sb)->s_ext4_flags);
721 if (journal)
722 jbd2_journal_abort(journal, -EIO);
723 }
724
725 if (!bdev_read_only(sb->s_bdev)) {
726 save_error_info(sb, error, ino, block, func, line);
727 /*
728 * In case the fs should keep running, we need to writeout
729 * superblock through the journal. Due to lock ordering
730 * constraints, it may not be safe to do it right here so we
731 * defer superblock flushing to a workqueue.
732 */
733 if (continue_fs && journal)
734 schedule_work(&EXT4_SB(sb)->s_sb_upd_work);
735 else
736 ext4_commit_super(sb);
737 }
738
739 /*
740 * We force ERRORS_RO behavior when system is rebooting. Otherwise we
741 * could panic during 'reboot -f' as the underlying device got already
742 * disabled.
743 */
744 if (test_opt(sb, ERRORS_PANIC) && !system_going_down()) {
745 panic("EXT4-fs (device %s): panic forced after error\n",
746 sb->s_id);
747 }
748
749 if (sb_rdonly(sb) || continue_fs)
750 return;
751
752 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
753 /*
754 * Make sure updated value of ->s_mount_flags will be visible before
755 * ->s_flags update
756 */
757 smp_wmb();
758 sb->s_flags |= SB_RDONLY;
759}
760
761static void update_super_work(struct work_struct *work)
762{
763 struct ext4_sb_info *sbi = container_of(work, struct ext4_sb_info,
764 s_sb_upd_work);
765 journal_t *journal = sbi->s_journal;
766 handle_t *handle;
767
768 /*
769 * If the journal is still running, we have to write out superblock
770 * through the journal to avoid collisions of other journalled sb
771 * updates.
772 *
773 * We use directly jbd2 functions here to avoid recursing back into
774 * ext4 error handling code during handling of previous errors.
775 */
776 if (!sb_rdonly(sbi->s_sb) && journal) {
777 struct buffer_head *sbh = sbi->s_sbh;
778 bool call_notify_err = false;
779
780 handle = jbd2_journal_start(journal, 1);
781 if (IS_ERR(handle))
782 goto write_directly;
783 if (jbd2_journal_get_write_access(handle, sbh)) {
784 jbd2_journal_stop(handle);
785 goto write_directly;
786 }
787
788 if (sbi->s_add_error_count > 0)
789 call_notify_err = true;
790
791 ext4_update_super(sbi->s_sb);
792 if (buffer_write_io_error(sbh) || !buffer_uptodate(sbh)) {
793 ext4_msg(sbi->s_sb, KERN_ERR, "previous I/O error to "
794 "superblock detected");
795 clear_buffer_write_io_error(sbh);
796 set_buffer_uptodate(sbh);
797 }
798
799 if (jbd2_journal_dirty_metadata(handle, sbh)) {
800 jbd2_journal_stop(handle);
801 goto write_directly;
802 }
803 jbd2_journal_stop(handle);
804
805 if (call_notify_err)
806 ext4_notify_error_sysfs(sbi);
807
808 return;
809 }
810write_directly:
811 /*
812 * Write through journal failed. Write sb directly to get error info
813 * out and hope for the best.
814 */
815 ext4_commit_super(sbi->s_sb);
816 ext4_notify_error_sysfs(sbi);
817}
818
819#define ext4_error_ratelimit(sb) \
820 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
821 "EXT4-fs error")
822
823void __ext4_error(struct super_block *sb, const char *function,
824 unsigned int line, bool force_ro, int error, __u64 block,
825 const char *fmt, ...)
826{
827 struct va_format vaf;
828 va_list args;
829
830 if (unlikely(ext4_forced_shutdown(sb)))
831 return;
832
833 trace_ext4_error(sb, function, line);
834 if (ext4_error_ratelimit(sb)) {
835 va_start(args, fmt);
836 vaf.fmt = fmt;
837 vaf.va = &args;
838 printk(KERN_CRIT
839 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
840 sb->s_id, function, line, current->comm, &vaf);
841 va_end(args);
842 }
843 fsnotify_sb_error(sb, NULL, error ? error : EFSCORRUPTED);
844
845 ext4_handle_error(sb, force_ro, error, 0, block, function, line);
846}
847
848void __ext4_error_inode(struct inode *inode, const char *function,
849 unsigned int line, ext4_fsblk_t block, int error,
850 const char *fmt, ...)
851{
852 va_list args;
853 struct va_format vaf;
854
855 if (unlikely(ext4_forced_shutdown(inode->i_sb)))
856 return;
857
858 trace_ext4_error(inode->i_sb, function, line);
859 if (ext4_error_ratelimit(inode->i_sb)) {
860 va_start(args, fmt);
861 vaf.fmt = fmt;
862 vaf.va = &args;
863 if (block)
864 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
865 "inode #%lu: block %llu: comm %s: %pV\n",
866 inode->i_sb->s_id, function, line, inode->i_ino,
867 block, current->comm, &vaf);
868 else
869 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
870 "inode #%lu: comm %s: %pV\n",
871 inode->i_sb->s_id, function, line, inode->i_ino,
872 current->comm, &vaf);
873 va_end(args);
874 }
875 fsnotify_sb_error(inode->i_sb, inode, error ? error : EFSCORRUPTED);
876
877 ext4_handle_error(inode->i_sb, false, error, inode->i_ino, block,
878 function, line);
879}
880
881void __ext4_error_file(struct file *file, const char *function,
882 unsigned int line, ext4_fsblk_t block,
883 const char *fmt, ...)
884{
885 va_list args;
886 struct va_format vaf;
887 struct inode *inode = file_inode(file);
888 char pathname[80], *path;
889
890 if (unlikely(ext4_forced_shutdown(inode->i_sb)))
891 return;
892
893 trace_ext4_error(inode->i_sb, function, line);
894 if (ext4_error_ratelimit(inode->i_sb)) {
895 path = file_path(file, pathname, sizeof(pathname));
896 if (IS_ERR(path))
897 path = "(unknown)";
898 va_start(args, fmt);
899 vaf.fmt = fmt;
900 vaf.va = &args;
901 if (block)
902 printk(KERN_CRIT
903 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
904 "block %llu: comm %s: path %s: %pV\n",
905 inode->i_sb->s_id, function, line, inode->i_ino,
906 block, current->comm, path, &vaf);
907 else
908 printk(KERN_CRIT
909 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
910 "comm %s: path %s: %pV\n",
911 inode->i_sb->s_id, function, line, inode->i_ino,
912 current->comm, path, &vaf);
913 va_end(args);
914 }
915 fsnotify_sb_error(inode->i_sb, inode, EFSCORRUPTED);
916
917 ext4_handle_error(inode->i_sb, false, EFSCORRUPTED, inode->i_ino, block,
918 function, line);
919}
920
921const char *ext4_decode_error(struct super_block *sb, int errno,
922 char nbuf[16])
923{
924 char *errstr = NULL;
925
926 switch (errno) {
927 case -EFSCORRUPTED:
928 errstr = "Corrupt filesystem";
929 break;
930 case -EFSBADCRC:
931 errstr = "Filesystem failed CRC";
932 break;
933 case -EIO:
934 errstr = "IO failure";
935 break;
936 case -ENOMEM:
937 errstr = "Out of memory";
938 break;
939 case -EROFS:
940 if (!sb || (EXT4_SB(sb)->s_journal &&
941 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
942 errstr = "Journal has aborted";
943 else
944 errstr = "Readonly filesystem";
945 break;
946 default:
947 /* If the caller passed in an extra buffer for unknown
948 * errors, textualise them now. Else we just return
949 * NULL. */
950 if (nbuf) {
951 /* Check for truncated error codes... */
952 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
953 errstr = nbuf;
954 }
955 break;
956 }
957
958 return errstr;
959}
960
961/* __ext4_std_error decodes expected errors from journaling functions
962 * automatically and invokes the appropriate error response. */
963
964void __ext4_std_error(struct super_block *sb, const char *function,
965 unsigned int line, int errno)
966{
967 char nbuf[16];
968 const char *errstr;
969
970 if (unlikely(ext4_forced_shutdown(sb)))
971 return;
972
973 /* Special case: if the error is EROFS, and we're not already
974 * inside a transaction, then there's really no point in logging
975 * an error. */
976 if (errno == -EROFS && journal_current_handle() == NULL && sb_rdonly(sb))
977 return;
978
979 if (ext4_error_ratelimit(sb)) {
980 errstr = ext4_decode_error(sb, errno, nbuf);
981 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
982 sb->s_id, function, line, errstr);
983 }
984 fsnotify_sb_error(sb, NULL, errno ? errno : EFSCORRUPTED);
985
986 ext4_handle_error(sb, false, -errno, 0, 0, function, line);
987}
988
989void __ext4_msg(struct super_block *sb,
990 const char *prefix, const char *fmt, ...)
991{
992 struct va_format vaf;
993 va_list args;
994
995 if (sb) {
996 atomic_inc(&EXT4_SB(sb)->s_msg_count);
997 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state),
998 "EXT4-fs"))
999 return;
1000 }
1001
1002 va_start(args, fmt);
1003 vaf.fmt = fmt;
1004 vaf.va = &args;
1005 if (sb)
1006 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
1007 else
1008 printk("%sEXT4-fs: %pV\n", prefix, &vaf);
1009 va_end(args);
1010}
1011
1012static int ext4_warning_ratelimit(struct super_block *sb)
1013{
1014 atomic_inc(&EXT4_SB(sb)->s_warning_count);
1015 return ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state),
1016 "EXT4-fs warning");
1017}
1018
1019void __ext4_warning(struct super_block *sb, const char *function,
1020 unsigned int line, const char *fmt, ...)
1021{
1022 struct va_format vaf;
1023 va_list args;
1024
1025 if (!ext4_warning_ratelimit(sb))
1026 return;
1027
1028 va_start(args, fmt);
1029 vaf.fmt = fmt;
1030 vaf.va = &args;
1031 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
1032 sb->s_id, function, line, &vaf);
1033 va_end(args);
1034}
1035
1036void __ext4_warning_inode(const struct inode *inode, const char *function,
1037 unsigned int line, const char *fmt, ...)
1038{
1039 struct va_format vaf;
1040 va_list args;
1041
1042 if (!ext4_warning_ratelimit(inode->i_sb))
1043 return;
1044
1045 va_start(args, fmt);
1046 vaf.fmt = fmt;
1047 vaf.va = &args;
1048 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
1049 "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
1050 function, line, inode->i_ino, current->comm, &vaf);
1051 va_end(args);
1052}
1053
1054void __ext4_grp_locked_error(const char *function, unsigned int line,
1055 struct super_block *sb, ext4_group_t grp,
1056 unsigned long ino, ext4_fsblk_t block,
1057 const char *fmt, ...)
1058__releases(bitlock)
1059__acquires(bitlock)
1060{
1061 struct va_format vaf;
1062 va_list args;
1063
1064 if (unlikely(ext4_forced_shutdown(sb)))
1065 return;
1066
1067 trace_ext4_error(sb, function, line);
1068 if (ext4_error_ratelimit(sb)) {
1069 va_start(args, fmt);
1070 vaf.fmt = fmt;
1071 vaf.va = &args;
1072 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
1073 sb->s_id, function, line, grp);
1074 if (ino)
1075 printk(KERN_CONT "inode %lu: ", ino);
1076 if (block)
1077 printk(KERN_CONT "block %llu:",
1078 (unsigned long long) block);
1079 printk(KERN_CONT "%pV\n", &vaf);
1080 va_end(args);
1081 }
1082
1083 if (test_opt(sb, ERRORS_CONT)) {
1084 if (test_opt(sb, WARN_ON_ERROR))
1085 WARN_ON_ONCE(1);
1086 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
1087 if (!bdev_read_only(sb->s_bdev)) {
1088 save_error_info(sb, EFSCORRUPTED, ino, block, function,
1089 line);
1090 schedule_work(&EXT4_SB(sb)->s_sb_upd_work);
1091 }
1092 return;
1093 }
1094 ext4_unlock_group(sb, grp);
1095 ext4_handle_error(sb, false, EFSCORRUPTED, ino, block, function, line);
1096 /*
1097 * We only get here in the ERRORS_RO case; relocking the group
1098 * may be dangerous, but nothing bad will happen since the
1099 * filesystem will have already been marked read/only and the
1100 * journal has been aborted. We return 1 as a hint to callers
1101 * who might what to use the return value from
1102 * ext4_grp_locked_error() to distinguish between the
1103 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
1104 * aggressively from the ext4 function in question, with a
1105 * more appropriate error code.
1106 */
1107 ext4_lock_group(sb, grp);
1108 return;
1109}
1110
1111void ext4_mark_group_bitmap_corrupted(struct super_block *sb,
1112 ext4_group_t group,
1113 unsigned int flags)
1114{
1115 struct ext4_sb_info *sbi = EXT4_SB(sb);
1116 struct ext4_group_info *grp = ext4_get_group_info(sb, group);
1117 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, group, NULL);
1118 int ret;
1119
1120 if (!grp || !gdp)
1121 return;
1122 if (flags & EXT4_GROUP_INFO_BBITMAP_CORRUPT) {
1123 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT,
1124 &grp->bb_state);
1125 if (!ret)
1126 percpu_counter_sub(&sbi->s_freeclusters_counter,
1127 grp->bb_free);
1128 }
1129
1130 if (flags & EXT4_GROUP_INFO_IBITMAP_CORRUPT) {
1131 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT,
1132 &grp->bb_state);
1133 if (!ret && gdp) {
1134 int count;
1135
1136 count = ext4_free_inodes_count(sb, gdp);
1137 percpu_counter_sub(&sbi->s_freeinodes_counter,
1138 count);
1139 }
1140 }
1141}
1142
1143void ext4_update_dynamic_rev(struct super_block *sb)
1144{
1145 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
1146
1147 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
1148 return;
1149
1150 ext4_warning(sb,
1151 "updating to rev %d because of new feature flag, "
1152 "running e2fsck is recommended",
1153 EXT4_DYNAMIC_REV);
1154
1155 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
1156 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
1157 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
1158 /* leave es->s_feature_*compat flags alone */
1159 /* es->s_uuid will be set by e2fsck if empty */
1160
1161 /*
1162 * The rest of the superblock fields should be zero, and if not it
1163 * means they are likely already in use, so leave them alone. We
1164 * can leave it up to e2fsck to clean up any inconsistencies there.
1165 */
1166}
1167
1168static inline struct inode *orphan_list_entry(struct list_head *l)
1169{
1170 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
1171}
1172
1173static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
1174{
1175 struct list_head *l;
1176
1177 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
1178 le32_to_cpu(sbi->s_es->s_last_orphan));
1179
1180 printk(KERN_ERR "sb_info orphan list:\n");
1181 list_for_each(l, &sbi->s_orphan) {
1182 struct inode *inode = orphan_list_entry(l);
1183 printk(KERN_ERR " "
1184 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
1185 inode->i_sb->s_id, inode->i_ino, inode,
1186 inode->i_mode, inode->i_nlink,
1187 NEXT_ORPHAN(inode));
1188 }
1189}
1190
1191#ifdef CONFIG_QUOTA
1192static int ext4_quota_off(struct super_block *sb, int type);
1193
1194static inline void ext4_quotas_off(struct super_block *sb, int type)
1195{
1196 BUG_ON(type > EXT4_MAXQUOTAS);
1197
1198 /* Use our quota_off function to clear inode flags etc. */
1199 for (type--; type >= 0; type--)
1200 ext4_quota_off(sb, type);
1201}
1202
1203/*
1204 * This is a helper function which is used in the mount/remount
1205 * codepaths (which holds s_umount) to fetch the quota file name.
1206 */
1207static inline char *get_qf_name(struct super_block *sb,
1208 struct ext4_sb_info *sbi,
1209 int type)
1210{
1211 return rcu_dereference_protected(sbi->s_qf_names[type],
1212 lockdep_is_held(&sb->s_umount));
1213}
1214#else
1215static inline void ext4_quotas_off(struct super_block *sb, int type)
1216{
1217}
1218#endif
1219
1220static int ext4_percpu_param_init(struct ext4_sb_info *sbi)
1221{
1222 ext4_fsblk_t block;
1223 int err;
1224
1225 block = ext4_count_free_clusters(sbi->s_sb);
1226 ext4_free_blocks_count_set(sbi->s_es, EXT4_C2B(sbi, block));
1227 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
1228 GFP_KERNEL);
1229 if (!err) {
1230 unsigned long freei = ext4_count_free_inodes(sbi->s_sb);
1231 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
1232 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
1233 GFP_KERNEL);
1234 }
1235 if (!err)
1236 err = percpu_counter_init(&sbi->s_dirs_counter,
1237 ext4_count_dirs(sbi->s_sb), GFP_KERNEL);
1238 if (!err)
1239 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
1240 GFP_KERNEL);
1241 if (!err)
1242 err = percpu_counter_init(&sbi->s_sra_exceeded_retry_limit, 0,
1243 GFP_KERNEL);
1244 if (!err)
1245 err = percpu_init_rwsem(&sbi->s_writepages_rwsem);
1246
1247 if (err)
1248 ext4_msg(sbi->s_sb, KERN_ERR, "insufficient memory");
1249
1250 return err;
1251}
1252
1253static void ext4_percpu_param_destroy(struct ext4_sb_info *sbi)
1254{
1255 percpu_counter_destroy(&sbi->s_freeclusters_counter);
1256 percpu_counter_destroy(&sbi->s_freeinodes_counter);
1257 percpu_counter_destroy(&sbi->s_dirs_counter);
1258 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
1259 percpu_counter_destroy(&sbi->s_sra_exceeded_retry_limit);
1260 percpu_free_rwsem(&sbi->s_writepages_rwsem);
1261}
1262
1263static void ext4_group_desc_free(struct ext4_sb_info *sbi)
1264{
1265 struct buffer_head **group_desc;
1266 int i;
1267
1268 rcu_read_lock();
1269 group_desc = rcu_dereference(sbi->s_group_desc);
1270 for (i = 0; i < sbi->s_gdb_count; i++)
1271 brelse(group_desc[i]);
1272 kvfree(group_desc);
1273 rcu_read_unlock();
1274}
1275
1276static void ext4_flex_groups_free(struct ext4_sb_info *sbi)
1277{
1278 struct flex_groups **flex_groups;
1279 int i;
1280
1281 rcu_read_lock();
1282 flex_groups = rcu_dereference(sbi->s_flex_groups);
1283 if (flex_groups) {
1284 for (i = 0; i < sbi->s_flex_groups_allocated; i++)
1285 kvfree(flex_groups[i]);
1286 kvfree(flex_groups);
1287 }
1288 rcu_read_unlock();
1289}
1290
1291static void ext4_put_super(struct super_block *sb)
1292{
1293 struct ext4_sb_info *sbi = EXT4_SB(sb);
1294 struct ext4_super_block *es = sbi->s_es;
1295 int aborted = 0;
1296 int err;
1297
1298 /*
1299 * Unregister sysfs before destroying jbd2 journal.
1300 * Since we could still access attr_journal_task attribute via sysfs
1301 * path which could have sbi->s_journal->j_task as NULL
1302 * Unregister sysfs before flush sbi->s_sb_upd_work.
1303 * Since user may read /proc/fs/ext4/xx/mb_groups during umount, If
1304 * read metadata verify failed then will queue error work.
1305 * update_super_work will call start_this_handle may trigger
1306 * BUG_ON.
1307 */
1308 ext4_unregister_sysfs(sb);
1309
1310 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs unmount"))
1311 ext4_msg(sb, KERN_INFO, "unmounting filesystem %pU.",
1312 &sb->s_uuid);
1313
1314 ext4_unregister_li_request(sb);
1315 ext4_quotas_off(sb, EXT4_MAXQUOTAS);
1316
1317 flush_work(&sbi->s_sb_upd_work);
1318 destroy_workqueue(sbi->rsv_conversion_wq);
1319 ext4_release_orphan_info(sb);
1320
1321 if (sbi->s_journal) {
1322 aborted = is_journal_aborted(sbi->s_journal);
1323 err = jbd2_journal_destroy(sbi->s_journal);
1324 sbi->s_journal = NULL;
1325 if ((err < 0) && !aborted) {
1326 ext4_abort(sb, -err, "Couldn't clean up the journal");
1327 }
1328 }
1329
1330 ext4_es_unregister_shrinker(sbi);
1331 timer_shutdown_sync(&sbi->s_err_report);
1332 ext4_release_system_zone(sb);
1333 ext4_mb_release(sb);
1334 ext4_ext_release(sb);
1335
1336 if (!sb_rdonly(sb) && !aborted) {
1337 ext4_clear_feature_journal_needs_recovery(sb);
1338 ext4_clear_feature_orphan_present(sb);
1339 es->s_state = cpu_to_le16(sbi->s_mount_state);
1340 }
1341 if (!sb_rdonly(sb))
1342 ext4_commit_super(sb);
1343
1344 ext4_group_desc_free(sbi);
1345 ext4_flex_groups_free(sbi);
1346 ext4_percpu_param_destroy(sbi);
1347#ifdef CONFIG_QUOTA
1348 for (int i = 0; i < EXT4_MAXQUOTAS; i++)
1349 kfree(get_qf_name(sb, sbi, i));
1350#endif
1351
1352 /* Debugging code just in case the in-memory inode orphan list
1353 * isn't empty. The on-disk one can be non-empty if we've
1354 * detected an error and taken the fs readonly, but the
1355 * in-memory list had better be clean by this point. */
1356 if (!list_empty(&sbi->s_orphan))
1357 dump_orphan_list(sb, sbi);
1358 ASSERT(list_empty(&sbi->s_orphan));
1359
1360 sync_blockdev(sb->s_bdev);
1361 invalidate_bdev(sb->s_bdev);
1362 if (sbi->s_journal_bdev_handle) {
1363 /*
1364 * Invalidate the journal device's buffers. We don't want them
1365 * floating about in memory - the physical journal device may
1366 * hotswapped, and it breaks the `ro-after' testing code.
1367 */
1368 sync_blockdev(sbi->s_journal_bdev_handle->bdev);
1369 invalidate_bdev(sbi->s_journal_bdev_handle->bdev);
1370 }
1371
1372 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
1373 sbi->s_ea_inode_cache = NULL;
1374
1375 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
1376 sbi->s_ea_block_cache = NULL;
1377
1378 ext4_stop_mmpd(sbi);
1379
1380 brelse(sbi->s_sbh);
1381 sb->s_fs_info = NULL;
1382 /*
1383 * Now that we are completely done shutting down the
1384 * superblock, we need to actually destroy the kobject.
1385 */
1386 kobject_put(&sbi->s_kobj);
1387 wait_for_completion(&sbi->s_kobj_unregister);
1388 if (sbi->s_chksum_driver)
1389 crypto_free_shash(sbi->s_chksum_driver);
1390 kfree(sbi->s_blockgroup_lock);
1391 fs_put_dax(sbi->s_daxdev, NULL);
1392 fscrypt_free_dummy_policy(&sbi->s_dummy_enc_policy);
1393#if IS_ENABLED(CONFIG_UNICODE)
1394 utf8_unload(sb->s_encoding);
1395#endif
1396 kfree(sbi);
1397}
1398
1399static struct kmem_cache *ext4_inode_cachep;
1400
1401/*
1402 * Called inside transaction, so use GFP_NOFS
1403 */
1404static struct inode *ext4_alloc_inode(struct super_block *sb)
1405{
1406 struct ext4_inode_info *ei;
1407
1408 ei = alloc_inode_sb(sb, ext4_inode_cachep, GFP_NOFS);
1409 if (!ei)
1410 return NULL;
1411
1412 inode_set_iversion(&ei->vfs_inode, 1);
1413 ei->i_flags = 0;
1414 spin_lock_init(&ei->i_raw_lock);
1415 ei->i_prealloc_node = RB_ROOT;
1416 atomic_set(&ei->i_prealloc_active, 0);
1417 rwlock_init(&ei->i_prealloc_lock);
1418 ext4_es_init_tree(&ei->i_es_tree);
1419 rwlock_init(&ei->i_es_lock);
1420 INIT_LIST_HEAD(&ei->i_es_list);
1421 ei->i_es_all_nr = 0;
1422 ei->i_es_shk_nr = 0;
1423 ei->i_es_shrink_lblk = 0;
1424 ei->i_reserved_data_blocks = 0;
1425 spin_lock_init(&(ei->i_block_reservation_lock));
1426 ext4_init_pending_tree(&ei->i_pending_tree);
1427#ifdef CONFIG_QUOTA
1428 ei->i_reserved_quota = 0;
1429 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
1430#endif
1431 ei->jinode = NULL;
1432 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
1433 spin_lock_init(&ei->i_completed_io_lock);
1434 ei->i_sync_tid = 0;
1435 ei->i_datasync_tid = 0;
1436 atomic_set(&ei->i_unwritten, 0);
1437 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
1438 ext4_fc_init_inode(&ei->vfs_inode);
1439 mutex_init(&ei->i_fc_lock);
1440 return &ei->vfs_inode;
1441}
1442
1443static int ext4_drop_inode(struct inode *inode)
1444{
1445 int drop = generic_drop_inode(inode);
1446
1447 if (!drop)
1448 drop = fscrypt_drop_inode(inode);
1449
1450 trace_ext4_drop_inode(inode, drop);
1451 return drop;
1452}
1453
1454static void ext4_free_in_core_inode(struct inode *inode)
1455{
1456 fscrypt_free_inode(inode);
1457 if (!list_empty(&(EXT4_I(inode)->i_fc_list))) {
1458 pr_warn("%s: inode %ld still in fc list",
1459 __func__, inode->i_ino);
1460 }
1461 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
1462}
1463
1464static void ext4_destroy_inode(struct inode *inode)
1465{
1466 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
1467 ext4_msg(inode->i_sb, KERN_ERR,
1468 "Inode %lu (%p): orphan list check failed!",
1469 inode->i_ino, EXT4_I(inode));
1470 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
1471 EXT4_I(inode), sizeof(struct ext4_inode_info),
1472 true);
1473 dump_stack();
1474 }
1475
1476 if (EXT4_I(inode)->i_reserved_data_blocks)
1477 ext4_msg(inode->i_sb, KERN_ERR,
1478 "Inode %lu (%p): i_reserved_data_blocks (%u) not cleared!",
1479 inode->i_ino, EXT4_I(inode),
1480 EXT4_I(inode)->i_reserved_data_blocks);
1481}
1482
1483static void ext4_shutdown(struct super_block *sb)
1484{
1485 ext4_force_shutdown(sb, EXT4_GOING_FLAGS_NOLOGFLUSH);
1486}
1487
1488static void init_once(void *foo)
1489{
1490 struct ext4_inode_info *ei = foo;
1491
1492 INIT_LIST_HEAD(&ei->i_orphan);
1493 init_rwsem(&ei->xattr_sem);
1494 init_rwsem(&ei->i_data_sem);
1495 inode_init_once(&ei->vfs_inode);
1496 ext4_fc_init_inode(&ei->vfs_inode);
1497}
1498
1499static int __init init_inodecache(void)
1500{
1501 ext4_inode_cachep = kmem_cache_create_usercopy("ext4_inode_cache",
1502 sizeof(struct ext4_inode_info), 0,
1503 (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD|
1504 SLAB_ACCOUNT),
1505 offsetof(struct ext4_inode_info, i_data),
1506 sizeof_field(struct ext4_inode_info, i_data),
1507 init_once);
1508 if (ext4_inode_cachep == NULL)
1509 return -ENOMEM;
1510 return 0;
1511}
1512
1513static void destroy_inodecache(void)
1514{
1515 /*
1516 * Make sure all delayed rcu free inodes are flushed before we
1517 * destroy cache.
1518 */
1519 rcu_barrier();
1520 kmem_cache_destroy(ext4_inode_cachep);
1521}
1522
1523void ext4_clear_inode(struct inode *inode)
1524{
1525 ext4_fc_del(inode);
1526 invalidate_inode_buffers(inode);
1527 clear_inode(inode);
1528 ext4_discard_preallocations(inode);
1529 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
1530 dquot_drop(inode);
1531 if (EXT4_I(inode)->jinode) {
1532 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1533 EXT4_I(inode)->jinode);
1534 jbd2_free_inode(EXT4_I(inode)->jinode);
1535 EXT4_I(inode)->jinode = NULL;
1536 }
1537 fscrypt_put_encryption_info(inode);
1538 fsverity_cleanup_inode(inode);
1539}
1540
1541static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1542 u64 ino, u32 generation)
1543{
1544 struct inode *inode;
1545
1546 /*
1547 * Currently we don't know the generation for parent directory, so
1548 * a generation of 0 means "accept any"
1549 */
1550 inode = ext4_iget(sb, ino, EXT4_IGET_HANDLE);
1551 if (IS_ERR(inode))
1552 return ERR_CAST(inode);
1553 if (generation && inode->i_generation != generation) {
1554 iput(inode);
1555 return ERR_PTR(-ESTALE);
1556 }
1557
1558 return inode;
1559}
1560
1561static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1562 int fh_len, int fh_type)
1563{
1564 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1565 ext4_nfs_get_inode);
1566}
1567
1568static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1569 int fh_len, int fh_type)
1570{
1571 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1572 ext4_nfs_get_inode);
1573}
1574
1575static int ext4_nfs_commit_metadata(struct inode *inode)
1576{
1577 struct writeback_control wbc = {
1578 .sync_mode = WB_SYNC_ALL
1579 };
1580
1581 trace_ext4_nfs_commit_metadata(inode);
1582 return ext4_write_inode(inode, &wbc);
1583}
1584
1585#ifdef CONFIG_QUOTA
1586static const char * const quotatypes[] = INITQFNAMES;
1587#define QTYPE2NAME(t) (quotatypes[t])
1588
1589static int ext4_write_dquot(struct dquot *dquot);
1590static int ext4_acquire_dquot(struct dquot *dquot);
1591static int ext4_release_dquot(struct dquot *dquot);
1592static int ext4_mark_dquot_dirty(struct dquot *dquot);
1593static int ext4_write_info(struct super_block *sb, int type);
1594static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1595 const struct path *path);
1596static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1597 size_t len, loff_t off);
1598static ssize_t ext4_quota_write(struct super_block *sb, int type,
1599 const char *data, size_t len, loff_t off);
1600static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1601 unsigned int flags);
1602
1603static struct dquot **ext4_get_dquots(struct inode *inode)
1604{
1605 return EXT4_I(inode)->i_dquot;
1606}
1607
1608static const struct dquot_operations ext4_quota_operations = {
1609 .get_reserved_space = ext4_get_reserved_space,
1610 .write_dquot = ext4_write_dquot,
1611 .acquire_dquot = ext4_acquire_dquot,
1612 .release_dquot = ext4_release_dquot,
1613 .mark_dirty = ext4_mark_dquot_dirty,
1614 .write_info = ext4_write_info,
1615 .alloc_dquot = dquot_alloc,
1616 .destroy_dquot = dquot_destroy,
1617 .get_projid = ext4_get_projid,
1618 .get_inode_usage = ext4_get_inode_usage,
1619 .get_next_id = dquot_get_next_id,
1620};
1621
1622static const struct quotactl_ops ext4_qctl_operations = {
1623 .quota_on = ext4_quota_on,
1624 .quota_off = ext4_quota_off,
1625 .quota_sync = dquot_quota_sync,
1626 .get_state = dquot_get_state,
1627 .set_info = dquot_set_dqinfo,
1628 .get_dqblk = dquot_get_dqblk,
1629 .set_dqblk = dquot_set_dqblk,
1630 .get_nextdqblk = dquot_get_next_dqblk,
1631};
1632#endif
1633
1634static const struct super_operations ext4_sops = {
1635 .alloc_inode = ext4_alloc_inode,
1636 .free_inode = ext4_free_in_core_inode,
1637 .destroy_inode = ext4_destroy_inode,
1638 .write_inode = ext4_write_inode,
1639 .dirty_inode = ext4_dirty_inode,
1640 .drop_inode = ext4_drop_inode,
1641 .evict_inode = ext4_evict_inode,
1642 .put_super = ext4_put_super,
1643 .sync_fs = ext4_sync_fs,
1644 .freeze_fs = ext4_freeze,
1645 .unfreeze_fs = ext4_unfreeze,
1646 .statfs = ext4_statfs,
1647 .show_options = ext4_show_options,
1648 .shutdown = ext4_shutdown,
1649#ifdef CONFIG_QUOTA
1650 .quota_read = ext4_quota_read,
1651 .quota_write = ext4_quota_write,
1652 .get_dquots = ext4_get_dquots,
1653#endif
1654};
1655
1656static const struct export_operations ext4_export_ops = {
1657 .encode_fh = generic_encode_ino32_fh,
1658 .fh_to_dentry = ext4_fh_to_dentry,
1659 .fh_to_parent = ext4_fh_to_parent,
1660 .get_parent = ext4_get_parent,
1661 .commit_metadata = ext4_nfs_commit_metadata,
1662};
1663
1664enum {
1665 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1666 Opt_resgid, Opt_resuid, Opt_sb,
1667 Opt_nouid32, Opt_debug, Opt_removed,
1668 Opt_user_xattr, Opt_acl,
1669 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1670 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1671 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1672 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1673 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
1674 Opt_inlinecrypt,
1675 Opt_usrjquota, Opt_grpjquota, Opt_quota,
1676 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1677 Opt_usrquota, Opt_grpquota, Opt_prjquota,
1678 Opt_dax, Opt_dax_always, Opt_dax_inode, Opt_dax_never,
1679 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_warn_on_error,
1680 Opt_nowarn_on_error, Opt_mblk_io_submit, Opt_debug_want_extra_isize,
1681 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1682 Opt_inode_readahead_blks, Opt_journal_ioprio,
1683 Opt_dioread_nolock, Opt_dioread_lock,
1684 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1685 Opt_max_dir_size_kb, Opt_nojournal_checksum, Opt_nombcache,
1686 Opt_no_prefetch_block_bitmaps, Opt_mb_optimize_scan,
1687 Opt_errors, Opt_data, Opt_data_err, Opt_jqfmt, Opt_dax_type,
1688#ifdef CONFIG_EXT4_DEBUG
1689 Opt_fc_debug_max_replay, Opt_fc_debug_force
1690#endif
1691};
1692
1693static const struct constant_table ext4_param_errors[] = {
1694 {"continue", EXT4_MOUNT_ERRORS_CONT},
1695 {"panic", EXT4_MOUNT_ERRORS_PANIC},
1696 {"remount-ro", EXT4_MOUNT_ERRORS_RO},
1697 {}
1698};
1699
1700static const struct constant_table ext4_param_data[] = {
1701 {"journal", EXT4_MOUNT_JOURNAL_DATA},
1702 {"ordered", EXT4_MOUNT_ORDERED_DATA},
1703 {"writeback", EXT4_MOUNT_WRITEBACK_DATA},
1704 {}
1705};
1706
1707static const struct constant_table ext4_param_data_err[] = {
1708 {"abort", Opt_data_err_abort},
1709 {"ignore", Opt_data_err_ignore},
1710 {}
1711};
1712
1713static const struct constant_table ext4_param_jqfmt[] = {
1714 {"vfsold", QFMT_VFS_OLD},
1715 {"vfsv0", QFMT_VFS_V0},
1716 {"vfsv1", QFMT_VFS_V1},
1717 {}
1718};
1719
1720static const struct constant_table ext4_param_dax[] = {
1721 {"always", Opt_dax_always},
1722 {"inode", Opt_dax_inode},
1723 {"never", Opt_dax_never},
1724 {}
1725};
1726
1727/* String parameter that allows empty argument */
1728#define fsparam_string_empty(NAME, OPT) \
1729 __fsparam(fs_param_is_string, NAME, OPT, fs_param_can_be_empty, NULL)
1730
1731/*
1732 * Mount option specification
1733 * We don't use fsparam_flag_no because of the way we set the
1734 * options and the way we show them in _ext4_show_options(). To
1735 * keep the changes to a minimum, let's keep the negative options
1736 * separate for now.
1737 */
1738static const struct fs_parameter_spec ext4_param_specs[] = {
1739 fsparam_flag ("bsddf", Opt_bsd_df),
1740 fsparam_flag ("minixdf", Opt_minix_df),
1741 fsparam_flag ("grpid", Opt_grpid),
1742 fsparam_flag ("bsdgroups", Opt_grpid),
1743 fsparam_flag ("nogrpid", Opt_nogrpid),
1744 fsparam_flag ("sysvgroups", Opt_nogrpid),
1745 fsparam_u32 ("resgid", Opt_resgid),
1746 fsparam_u32 ("resuid", Opt_resuid),
1747 fsparam_u32 ("sb", Opt_sb),
1748 fsparam_enum ("errors", Opt_errors, ext4_param_errors),
1749 fsparam_flag ("nouid32", Opt_nouid32),
1750 fsparam_flag ("debug", Opt_debug),
1751 fsparam_flag ("oldalloc", Opt_removed),
1752 fsparam_flag ("orlov", Opt_removed),
1753 fsparam_flag ("user_xattr", Opt_user_xattr),
1754 fsparam_flag ("acl", Opt_acl),
1755 fsparam_flag ("norecovery", Opt_noload),
1756 fsparam_flag ("noload", Opt_noload),
1757 fsparam_flag ("bh", Opt_removed),
1758 fsparam_flag ("nobh", Opt_removed),
1759 fsparam_u32 ("commit", Opt_commit),
1760 fsparam_u32 ("min_batch_time", Opt_min_batch_time),
1761 fsparam_u32 ("max_batch_time", Opt_max_batch_time),
1762 fsparam_u32 ("journal_dev", Opt_journal_dev),
1763 fsparam_bdev ("journal_path", Opt_journal_path),
1764 fsparam_flag ("journal_checksum", Opt_journal_checksum),
1765 fsparam_flag ("nojournal_checksum", Opt_nojournal_checksum),
1766 fsparam_flag ("journal_async_commit",Opt_journal_async_commit),
1767 fsparam_flag ("abort", Opt_abort),
1768 fsparam_enum ("data", Opt_data, ext4_param_data),
1769 fsparam_enum ("data_err", Opt_data_err,
1770 ext4_param_data_err),
1771 fsparam_string_empty
1772 ("usrjquota", Opt_usrjquota),
1773 fsparam_string_empty
1774 ("grpjquota", Opt_grpjquota),
1775 fsparam_enum ("jqfmt", Opt_jqfmt, ext4_param_jqfmt),
1776 fsparam_flag ("grpquota", Opt_grpquota),
1777 fsparam_flag ("quota", Opt_quota),
1778 fsparam_flag ("noquota", Opt_noquota),
1779 fsparam_flag ("usrquota", Opt_usrquota),
1780 fsparam_flag ("prjquota", Opt_prjquota),
1781 fsparam_flag ("barrier", Opt_barrier),
1782 fsparam_u32 ("barrier", Opt_barrier),
1783 fsparam_flag ("nobarrier", Opt_nobarrier),
1784 fsparam_flag ("i_version", Opt_removed),
1785 fsparam_flag ("dax", Opt_dax),
1786 fsparam_enum ("dax", Opt_dax_type, ext4_param_dax),
1787 fsparam_u32 ("stripe", Opt_stripe),
1788 fsparam_flag ("delalloc", Opt_delalloc),
1789 fsparam_flag ("nodelalloc", Opt_nodelalloc),
1790 fsparam_flag ("warn_on_error", Opt_warn_on_error),
1791 fsparam_flag ("nowarn_on_error", Opt_nowarn_on_error),
1792 fsparam_u32 ("debug_want_extra_isize",
1793 Opt_debug_want_extra_isize),
1794 fsparam_flag ("mblk_io_submit", Opt_removed),
1795 fsparam_flag ("nomblk_io_submit", Opt_removed),
1796 fsparam_flag ("block_validity", Opt_block_validity),
1797 fsparam_flag ("noblock_validity", Opt_noblock_validity),
1798 fsparam_u32 ("inode_readahead_blks",
1799 Opt_inode_readahead_blks),
1800 fsparam_u32 ("journal_ioprio", Opt_journal_ioprio),
1801 fsparam_u32 ("auto_da_alloc", Opt_auto_da_alloc),
1802 fsparam_flag ("auto_da_alloc", Opt_auto_da_alloc),
1803 fsparam_flag ("noauto_da_alloc", Opt_noauto_da_alloc),
1804 fsparam_flag ("dioread_nolock", Opt_dioread_nolock),
1805 fsparam_flag ("nodioread_nolock", Opt_dioread_lock),
1806 fsparam_flag ("dioread_lock", Opt_dioread_lock),
1807 fsparam_flag ("discard", Opt_discard),
1808 fsparam_flag ("nodiscard", Opt_nodiscard),
1809 fsparam_u32 ("init_itable", Opt_init_itable),
1810 fsparam_flag ("init_itable", Opt_init_itable),
1811 fsparam_flag ("noinit_itable", Opt_noinit_itable),
1812#ifdef CONFIG_EXT4_DEBUG
1813 fsparam_flag ("fc_debug_force", Opt_fc_debug_force),
1814 fsparam_u32 ("fc_debug_max_replay", Opt_fc_debug_max_replay),
1815#endif
1816 fsparam_u32 ("max_dir_size_kb", Opt_max_dir_size_kb),
1817 fsparam_flag ("test_dummy_encryption",
1818 Opt_test_dummy_encryption),
1819 fsparam_string ("test_dummy_encryption",
1820 Opt_test_dummy_encryption),
1821 fsparam_flag ("inlinecrypt", Opt_inlinecrypt),
1822 fsparam_flag ("nombcache", Opt_nombcache),
1823 fsparam_flag ("no_mbcache", Opt_nombcache), /* for backward compatibility */
1824 fsparam_flag ("prefetch_block_bitmaps",
1825 Opt_removed),
1826 fsparam_flag ("no_prefetch_block_bitmaps",
1827 Opt_no_prefetch_block_bitmaps),
1828 fsparam_s32 ("mb_optimize_scan", Opt_mb_optimize_scan),
1829 fsparam_string ("check", Opt_removed), /* mount option from ext2/3 */
1830 fsparam_flag ("nocheck", Opt_removed), /* mount option from ext2/3 */
1831 fsparam_flag ("reservation", Opt_removed), /* mount option from ext2/3 */
1832 fsparam_flag ("noreservation", Opt_removed), /* mount option from ext2/3 */
1833 fsparam_u32 ("journal", Opt_removed), /* mount option from ext2/3 */
1834 {}
1835};
1836
1837#define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1838
1839#define MOPT_SET 0x0001
1840#define MOPT_CLEAR 0x0002
1841#define MOPT_NOSUPPORT 0x0004
1842#define MOPT_EXPLICIT 0x0008
1843#ifdef CONFIG_QUOTA
1844#define MOPT_Q 0
1845#define MOPT_QFMT 0x0010
1846#else
1847#define MOPT_Q MOPT_NOSUPPORT
1848#define MOPT_QFMT MOPT_NOSUPPORT
1849#endif
1850#define MOPT_NO_EXT2 0x0020
1851#define MOPT_NO_EXT3 0x0040
1852#define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1853#define MOPT_SKIP 0x0080
1854#define MOPT_2 0x0100
1855
1856static const struct mount_opts {
1857 int token;
1858 int mount_opt;
1859 int flags;
1860} ext4_mount_opts[] = {
1861 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1862 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1863 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1864 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1865 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1866 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1867 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1868 MOPT_EXT4_ONLY | MOPT_SET},
1869 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1870 MOPT_EXT4_ONLY | MOPT_CLEAR},
1871 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1872 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1873 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1874 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1875 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1876 MOPT_EXT4_ONLY | MOPT_CLEAR},
1877 {Opt_warn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_SET},
1878 {Opt_nowarn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_CLEAR},
1879 {Opt_commit, 0, MOPT_NO_EXT2},
1880 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1881 MOPT_EXT4_ONLY | MOPT_CLEAR},
1882 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1883 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1884 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1885 EXT4_MOUNT_JOURNAL_CHECKSUM),
1886 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1887 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1888 {Opt_data_err, EXT4_MOUNT_DATA_ERR_ABORT, MOPT_NO_EXT2},
1889 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1890 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1891 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1892 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1893 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1894 {Opt_dax_type, 0, MOPT_EXT4_ONLY},
1895 {Opt_journal_dev, 0, MOPT_NO_EXT2},
1896 {Opt_journal_path, 0, MOPT_NO_EXT2},
1897 {Opt_journal_ioprio, 0, MOPT_NO_EXT2},
1898 {Opt_data, 0, MOPT_NO_EXT2},
1899 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1900#ifdef CONFIG_EXT4_FS_POSIX_ACL
1901 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1902#else
1903 {Opt_acl, 0, MOPT_NOSUPPORT},
1904#endif
1905 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1906 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1907 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1908 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1909 MOPT_SET | MOPT_Q},
1910 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1911 MOPT_SET | MOPT_Q},
1912 {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA,
1913 MOPT_SET | MOPT_Q},
1914 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1915 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA),
1916 MOPT_CLEAR | MOPT_Q},
1917 {Opt_usrjquota, 0, MOPT_Q},
1918 {Opt_grpjquota, 0, MOPT_Q},
1919 {Opt_jqfmt, 0, MOPT_QFMT},
1920 {Opt_nombcache, EXT4_MOUNT_NO_MBCACHE, MOPT_SET},
1921 {Opt_no_prefetch_block_bitmaps, EXT4_MOUNT_NO_PREFETCH_BLOCK_BITMAPS,
1922 MOPT_SET},
1923#ifdef CONFIG_EXT4_DEBUG
1924 {Opt_fc_debug_force, EXT4_MOUNT2_JOURNAL_FAST_COMMIT,
1925 MOPT_SET | MOPT_2 | MOPT_EXT4_ONLY},
1926#endif
1927 {Opt_abort, EXT4_MOUNT2_ABORT, MOPT_SET | MOPT_2},
1928 {Opt_err, 0, 0}
1929};
1930
1931#if IS_ENABLED(CONFIG_UNICODE)
1932static const struct ext4_sb_encodings {
1933 __u16 magic;
1934 char *name;
1935 unsigned int version;
1936} ext4_sb_encoding_map[] = {
1937 {EXT4_ENC_UTF8_12_1, "utf8", UNICODE_AGE(12, 1, 0)},
1938};
1939
1940static const struct ext4_sb_encodings *
1941ext4_sb_read_encoding(const struct ext4_super_block *es)
1942{
1943 __u16 magic = le16_to_cpu(es->s_encoding);
1944 int i;
1945
1946 for (i = 0; i < ARRAY_SIZE(ext4_sb_encoding_map); i++)
1947 if (magic == ext4_sb_encoding_map[i].magic)
1948 return &ext4_sb_encoding_map[i];
1949
1950 return NULL;
1951}
1952#endif
1953
1954#define EXT4_SPEC_JQUOTA (1 << 0)
1955#define EXT4_SPEC_JQFMT (1 << 1)
1956#define EXT4_SPEC_DATAJ (1 << 2)
1957#define EXT4_SPEC_SB_BLOCK (1 << 3)
1958#define EXT4_SPEC_JOURNAL_DEV (1 << 4)
1959#define EXT4_SPEC_JOURNAL_IOPRIO (1 << 5)
1960#define EXT4_SPEC_s_want_extra_isize (1 << 7)
1961#define EXT4_SPEC_s_max_batch_time (1 << 8)
1962#define EXT4_SPEC_s_min_batch_time (1 << 9)
1963#define EXT4_SPEC_s_inode_readahead_blks (1 << 10)
1964#define EXT4_SPEC_s_li_wait_mult (1 << 11)
1965#define EXT4_SPEC_s_max_dir_size_kb (1 << 12)
1966#define EXT4_SPEC_s_stripe (1 << 13)
1967#define EXT4_SPEC_s_resuid (1 << 14)
1968#define EXT4_SPEC_s_resgid (1 << 15)
1969#define EXT4_SPEC_s_commit_interval (1 << 16)
1970#define EXT4_SPEC_s_fc_debug_max_replay (1 << 17)
1971#define EXT4_SPEC_s_sb_block (1 << 18)
1972#define EXT4_SPEC_mb_optimize_scan (1 << 19)
1973
1974struct ext4_fs_context {
1975 char *s_qf_names[EXT4_MAXQUOTAS];
1976 struct fscrypt_dummy_policy dummy_enc_policy;
1977 int s_jquota_fmt; /* Format of quota to use */
1978#ifdef CONFIG_EXT4_DEBUG
1979 int s_fc_debug_max_replay;
1980#endif
1981 unsigned short qname_spec;
1982 unsigned long vals_s_flags; /* Bits to set in s_flags */
1983 unsigned long mask_s_flags; /* Bits changed in s_flags */
1984 unsigned long journal_devnum;
1985 unsigned long s_commit_interval;
1986 unsigned long s_stripe;
1987 unsigned int s_inode_readahead_blks;
1988 unsigned int s_want_extra_isize;
1989 unsigned int s_li_wait_mult;
1990 unsigned int s_max_dir_size_kb;
1991 unsigned int journal_ioprio;
1992 unsigned int vals_s_mount_opt;
1993 unsigned int mask_s_mount_opt;
1994 unsigned int vals_s_mount_opt2;
1995 unsigned int mask_s_mount_opt2;
1996 unsigned int opt_flags; /* MOPT flags */
1997 unsigned int spec;
1998 u32 s_max_batch_time;
1999 u32 s_min_batch_time;
2000 kuid_t s_resuid;
2001 kgid_t s_resgid;
2002 ext4_fsblk_t s_sb_block;
2003};
2004
2005static void ext4_fc_free(struct fs_context *fc)
2006{
2007 struct ext4_fs_context *ctx = fc->fs_private;
2008 int i;
2009
2010 if (!ctx)
2011 return;
2012
2013 for (i = 0; i < EXT4_MAXQUOTAS; i++)
2014 kfree(ctx->s_qf_names[i]);
2015
2016 fscrypt_free_dummy_policy(&ctx->dummy_enc_policy);
2017 kfree(ctx);
2018}
2019
2020int ext4_init_fs_context(struct fs_context *fc)
2021{
2022 struct ext4_fs_context *ctx;
2023
2024 ctx = kzalloc(sizeof(struct ext4_fs_context), GFP_KERNEL);
2025 if (!ctx)
2026 return -ENOMEM;
2027
2028 fc->fs_private = ctx;
2029 fc->ops = &ext4_context_ops;
2030
2031 return 0;
2032}
2033
2034#ifdef CONFIG_QUOTA
2035/*
2036 * Note the name of the specified quota file.
2037 */
2038static int note_qf_name(struct fs_context *fc, int qtype,
2039 struct fs_parameter *param)
2040{
2041 struct ext4_fs_context *ctx = fc->fs_private;
2042 char *qname;
2043
2044 if (param->size < 1) {
2045 ext4_msg(NULL, KERN_ERR, "Missing quota name");
2046 return -EINVAL;
2047 }
2048 if (strchr(param->string, '/')) {
2049 ext4_msg(NULL, KERN_ERR,
2050 "quotafile must be on filesystem root");
2051 return -EINVAL;
2052 }
2053 if (ctx->s_qf_names[qtype]) {
2054 if (strcmp(ctx->s_qf_names[qtype], param->string) != 0) {
2055 ext4_msg(NULL, KERN_ERR,
2056 "%s quota file already specified",
2057 QTYPE2NAME(qtype));
2058 return -EINVAL;
2059 }
2060 return 0;
2061 }
2062
2063 qname = kmemdup_nul(param->string, param->size, GFP_KERNEL);
2064 if (!qname) {
2065 ext4_msg(NULL, KERN_ERR,
2066 "Not enough memory for storing quotafile name");
2067 return -ENOMEM;
2068 }
2069 ctx->s_qf_names[qtype] = qname;
2070 ctx->qname_spec |= 1 << qtype;
2071 ctx->spec |= EXT4_SPEC_JQUOTA;
2072 return 0;
2073}
2074
2075/*
2076 * Clear the name of the specified quota file.
2077 */
2078static int unnote_qf_name(struct fs_context *fc, int qtype)
2079{
2080 struct ext4_fs_context *ctx = fc->fs_private;
2081
2082 if (ctx->s_qf_names[qtype])
2083 kfree(ctx->s_qf_names[qtype]);
2084
2085 ctx->s_qf_names[qtype] = NULL;
2086 ctx->qname_spec |= 1 << qtype;
2087 ctx->spec |= EXT4_SPEC_JQUOTA;
2088 return 0;
2089}
2090#endif
2091
2092static int ext4_parse_test_dummy_encryption(const struct fs_parameter *param,
2093 struct ext4_fs_context *ctx)
2094{
2095 int err;
2096
2097 if (!IS_ENABLED(CONFIG_FS_ENCRYPTION)) {
2098 ext4_msg(NULL, KERN_WARNING,
2099 "test_dummy_encryption option not supported");
2100 return -EINVAL;
2101 }
2102 err = fscrypt_parse_test_dummy_encryption(param,
2103 &ctx->dummy_enc_policy);
2104 if (err == -EINVAL) {
2105 ext4_msg(NULL, KERN_WARNING,
2106 "Value of option \"%s\" is unrecognized", param->key);
2107 } else if (err == -EEXIST) {
2108 ext4_msg(NULL, KERN_WARNING,
2109 "Conflicting test_dummy_encryption options");
2110 return -EINVAL;
2111 }
2112 return err;
2113}
2114
2115#define EXT4_SET_CTX(name) \
2116static inline void ctx_set_##name(struct ext4_fs_context *ctx, \
2117 unsigned long flag) \
2118{ \
2119 ctx->mask_s_##name |= flag; \
2120 ctx->vals_s_##name |= flag; \
2121}
2122
2123#define EXT4_CLEAR_CTX(name) \
2124static inline void ctx_clear_##name(struct ext4_fs_context *ctx, \
2125 unsigned long flag) \
2126{ \
2127 ctx->mask_s_##name |= flag; \
2128 ctx->vals_s_##name &= ~flag; \
2129}
2130
2131#define EXT4_TEST_CTX(name) \
2132static inline unsigned long \
2133ctx_test_##name(struct ext4_fs_context *ctx, unsigned long flag) \
2134{ \
2135 return (ctx->vals_s_##name & flag); \
2136}
2137
2138EXT4_SET_CTX(flags); /* set only */
2139EXT4_SET_CTX(mount_opt);
2140EXT4_CLEAR_CTX(mount_opt);
2141EXT4_TEST_CTX(mount_opt);
2142EXT4_SET_CTX(mount_opt2);
2143EXT4_CLEAR_CTX(mount_opt2);
2144EXT4_TEST_CTX(mount_opt2);
2145
2146static int ext4_parse_param(struct fs_context *fc, struct fs_parameter *param)
2147{
2148 struct ext4_fs_context *ctx = fc->fs_private;
2149 struct fs_parse_result result;
2150 const struct mount_opts *m;
2151 int is_remount;
2152 kuid_t uid;
2153 kgid_t gid;
2154 int token;
2155
2156 token = fs_parse(fc, ext4_param_specs, param, &result);
2157 if (token < 0)
2158 return token;
2159 is_remount = fc->purpose == FS_CONTEXT_FOR_RECONFIGURE;
2160
2161 for (m = ext4_mount_opts; m->token != Opt_err; m++)
2162 if (token == m->token)
2163 break;
2164
2165 ctx->opt_flags |= m->flags;
2166
2167 if (m->flags & MOPT_EXPLICIT) {
2168 if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
2169 ctx_set_mount_opt2(ctx, EXT4_MOUNT2_EXPLICIT_DELALLOC);
2170 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
2171 ctx_set_mount_opt2(ctx,
2172 EXT4_MOUNT2_EXPLICIT_JOURNAL_CHECKSUM);
2173 } else
2174 return -EINVAL;
2175 }
2176
2177 if (m->flags & MOPT_NOSUPPORT) {
2178 ext4_msg(NULL, KERN_ERR, "%s option not supported",
2179 param->key);
2180 return 0;
2181 }
2182
2183 switch (token) {
2184#ifdef CONFIG_QUOTA
2185 case Opt_usrjquota:
2186 if (!*param->string)
2187 return unnote_qf_name(fc, USRQUOTA);
2188 else
2189 return note_qf_name(fc, USRQUOTA, param);
2190 case Opt_grpjquota:
2191 if (!*param->string)
2192 return unnote_qf_name(fc, GRPQUOTA);
2193 else
2194 return note_qf_name(fc, GRPQUOTA, param);
2195#endif
2196 case Opt_sb:
2197 if (fc->purpose == FS_CONTEXT_FOR_RECONFIGURE) {
2198 ext4_msg(NULL, KERN_WARNING,
2199 "Ignoring %s option on remount", param->key);
2200 } else {
2201 ctx->s_sb_block = result.uint_32;
2202 ctx->spec |= EXT4_SPEC_s_sb_block;
2203 }
2204 return 0;
2205 case Opt_removed:
2206 ext4_msg(NULL, KERN_WARNING, "Ignoring removed %s option",
2207 param->key);
2208 return 0;
2209 case Opt_inlinecrypt:
2210#ifdef CONFIG_FS_ENCRYPTION_INLINE_CRYPT
2211 ctx_set_flags(ctx, SB_INLINECRYPT);
2212#else
2213 ext4_msg(NULL, KERN_ERR, "inline encryption not supported");
2214#endif
2215 return 0;
2216 case Opt_errors:
2217 ctx_clear_mount_opt(ctx, EXT4_MOUNT_ERRORS_MASK);
2218 ctx_set_mount_opt(ctx, result.uint_32);
2219 return 0;
2220#ifdef CONFIG_QUOTA
2221 case Opt_jqfmt:
2222 ctx->s_jquota_fmt = result.uint_32;
2223 ctx->spec |= EXT4_SPEC_JQFMT;
2224 return 0;
2225#endif
2226 case Opt_data:
2227 ctx_clear_mount_opt(ctx, EXT4_MOUNT_DATA_FLAGS);
2228 ctx_set_mount_opt(ctx, result.uint_32);
2229 ctx->spec |= EXT4_SPEC_DATAJ;
2230 return 0;
2231 case Opt_commit:
2232 if (result.uint_32 == 0)
2233 result.uint_32 = JBD2_DEFAULT_MAX_COMMIT_AGE;
2234 else if (result.uint_32 > INT_MAX / HZ) {
2235 ext4_msg(NULL, KERN_ERR,
2236 "Invalid commit interval %d, "
2237 "must be smaller than %d",
2238 result.uint_32, INT_MAX / HZ);
2239 return -EINVAL;
2240 }
2241 ctx->s_commit_interval = HZ * result.uint_32;
2242 ctx->spec |= EXT4_SPEC_s_commit_interval;
2243 return 0;
2244 case Opt_debug_want_extra_isize:
2245 if ((result.uint_32 & 1) || (result.uint_32 < 4)) {
2246 ext4_msg(NULL, KERN_ERR,
2247 "Invalid want_extra_isize %d", result.uint_32);
2248 return -EINVAL;
2249 }
2250 ctx->s_want_extra_isize = result.uint_32;
2251 ctx->spec |= EXT4_SPEC_s_want_extra_isize;
2252 return 0;
2253 case Opt_max_batch_time:
2254 ctx->s_max_batch_time = result.uint_32;
2255 ctx->spec |= EXT4_SPEC_s_max_batch_time;
2256 return 0;
2257 case Opt_min_batch_time:
2258 ctx->s_min_batch_time = result.uint_32;
2259 ctx->spec |= EXT4_SPEC_s_min_batch_time;
2260 return 0;
2261 case Opt_inode_readahead_blks:
2262 if (result.uint_32 &&
2263 (result.uint_32 > (1 << 30) ||
2264 !is_power_of_2(result.uint_32))) {
2265 ext4_msg(NULL, KERN_ERR,
2266 "EXT4-fs: inode_readahead_blks must be "
2267 "0 or a power of 2 smaller than 2^31");
2268 return -EINVAL;
2269 }
2270 ctx->s_inode_readahead_blks = result.uint_32;
2271 ctx->spec |= EXT4_SPEC_s_inode_readahead_blks;
2272 return 0;
2273 case Opt_init_itable:
2274 ctx_set_mount_opt(ctx, EXT4_MOUNT_INIT_INODE_TABLE);
2275 ctx->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
2276 if (param->type == fs_value_is_string)
2277 ctx->s_li_wait_mult = result.uint_32;
2278 ctx->spec |= EXT4_SPEC_s_li_wait_mult;
2279 return 0;
2280 case Opt_max_dir_size_kb:
2281 ctx->s_max_dir_size_kb = result.uint_32;
2282 ctx->spec |= EXT4_SPEC_s_max_dir_size_kb;
2283 return 0;
2284#ifdef CONFIG_EXT4_DEBUG
2285 case Opt_fc_debug_max_replay:
2286 ctx->s_fc_debug_max_replay = result.uint_32;
2287 ctx->spec |= EXT4_SPEC_s_fc_debug_max_replay;
2288 return 0;
2289#endif
2290 case Opt_stripe:
2291 ctx->s_stripe = result.uint_32;
2292 ctx->spec |= EXT4_SPEC_s_stripe;
2293 return 0;
2294 case Opt_resuid:
2295 uid = make_kuid(current_user_ns(), result.uint_32);
2296 if (!uid_valid(uid)) {
2297 ext4_msg(NULL, KERN_ERR, "Invalid uid value %d",
2298 result.uint_32);
2299 return -EINVAL;
2300 }
2301 ctx->s_resuid = uid;
2302 ctx->spec |= EXT4_SPEC_s_resuid;
2303 return 0;
2304 case Opt_resgid:
2305 gid = make_kgid(current_user_ns(), result.uint_32);
2306 if (!gid_valid(gid)) {
2307 ext4_msg(NULL, KERN_ERR, "Invalid gid value %d",
2308 result.uint_32);
2309 return -EINVAL;
2310 }
2311 ctx->s_resgid = gid;
2312 ctx->spec |= EXT4_SPEC_s_resgid;
2313 return 0;
2314 case Opt_journal_dev:
2315 if (is_remount) {
2316 ext4_msg(NULL, KERN_ERR,
2317 "Cannot specify journal on remount");
2318 return -EINVAL;
2319 }
2320 ctx->journal_devnum = result.uint_32;
2321 ctx->spec |= EXT4_SPEC_JOURNAL_DEV;
2322 return 0;
2323 case Opt_journal_path:
2324 {
2325 struct inode *journal_inode;
2326 struct path path;
2327 int error;
2328
2329 if (is_remount) {
2330 ext4_msg(NULL, KERN_ERR,
2331 "Cannot specify journal on remount");
2332 return -EINVAL;
2333 }
2334
2335 error = fs_lookup_param(fc, param, 1, LOOKUP_FOLLOW, &path);
2336 if (error) {
2337 ext4_msg(NULL, KERN_ERR, "error: could not find "
2338 "journal device path");
2339 return -EINVAL;
2340 }
2341
2342 journal_inode = d_inode(path.dentry);
2343 ctx->journal_devnum = new_encode_dev(journal_inode->i_rdev);
2344 ctx->spec |= EXT4_SPEC_JOURNAL_DEV;
2345 path_put(&path);
2346 return 0;
2347 }
2348 case Opt_journal_ioprio:
2349 if (result.uint_32 > 7) {
2350 ext4_msg(NULL, KERN_ERR, "Invalid journal IO priority"
2351 " (must be 0-7)");
2352 return -EINVAL;
2353 }
2354 ctx->journal_ioprio =
2355 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, result.uint_32);
2356 ctx->spec |= EXT4_SPEC_JOURNAL_IOPRIO;
2357 return 0;
2358 case Opt_test_dummy_encryption:
2359 return ext4_parse_test_dummy_encryption(param, ctx);
2360 case Opt_dax:
2361 case Opt_dax_type:
2362#ifdef CONFIG_FS_DAX
2363 {
2364 int type = (token == Opt_dax) ?
2365 Opt_dax : result.uint_32;
2366
2367 switch (type) {
2368 case Opt_dax:
2369 case Opt_dax_always:
2370 ctx_set_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS);
2371 ctx_clear_mount_opt2(ctx, EXT4_MOUNT2_DAX_NEVER);
2372 break;
2373 case Opt_dax_never:
2374 ctx_set_mount_opt2(ctx, EXT4_MOUNT2_DAX_NEVER);
2375 ctx_clear_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS);
2376 break;
2377 case Opt_dax_inode:
2378 ctx_clear_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS);
2379 ctx_clear_mount_opt2(ctx, EXT4_MOUNT2_DAX_NEVER);
2380 /* Strictly for printing options */
2381 ctx_set_mount_opt2(ctx, EXT4_MOUNT2_DAX_INODE);
2382 break;
2383 }
2384 return 0;
2385 }
2386#else
2387 ext4_msg(NULL, KERN_INFO, "dax option not supported");
2388 return -EINVAL;
2389#endif
2390 case Opt_data_err:
2391 if (result.uint_32 == Opt_data_err_abort)
2392 ctx_set_mount_opt(ctx, m->mount_opt);
2393 else if (result.uint_32 == Opt_data_err_ignore)
2394 ctx_clear_mount_opt(ctx, m->mount_opt);
2395 return 0;
2396 case Opt_mb_optimize_scan:
2397 if (result.int_32 == 1) {
2398 ctx_set_mount_opt2(ctx, EXT4_MOUNT2_MB_OPTIMIZE_SCAN);
2399 ctx->spec |= EXT4_SPEC_mb_optimize_scan;
2400 } else if (result.int_32 == 0) {
2401 ctx_clear_mount_opt2(ctx, EXT4_MOUNT2_MB_OPTIMIZE_SCAN);
2402 ctx->spec |= EXT4_SPEC_mb_optimize_scan;
2403 } else {
2404 ext4_msg(NULL, KERN_WARNING,
2405 "mb_optimize_scan should be set to 0 or 1.");
2406 return -EINVAL;
2407 }
2408 return 0;
2409 }
2410
2411 /*
2412 * At this point we should only be getting options requiring MOPT_SET,
2413 * or MOPT_CLEAR. Anything else is a bug
2414 */
2415 if (m->token == Opt_err) {
2416 ext4_msg(NULL, KERN_WARNING, "buggy handling of option %s",
2417 param->key);
2418 WARN_ON(1);
2419 return -EINVAL;
2420 }
2421
2422 else {
2423 unsigned int set = 0;
2424
2425 if ((param->type == fs_value_is_flag) ||
2426 result.uint_32 > 0)
2427 set = 1;
2428
2429 if (m->flags & MOPT_CLEAR)
2430 set = !set;
2431 else if (unlikely(!(m->flags & MOPT_SET))) {
2432 ext4_msg(NULL, KERN_WARNING,
2433 "buggy handling of option %s",
2434 param->key);
2435 WARN_ON(1);
2436 return -EINVAL;
2437 }
2438 if (m->flags & MOPT_2) {
2439 if (set != 0)
2440 ctx_set_mount_opt2(ctx, m->mount_opt);
2441 else
2442 ctx_clear_mount_opt2(ctx, m->mount_opt);
2443 } else {
2444 if (set != 0)
2445 ctx_set_mount_opt(ctx, m->mount_opt);
2446 else
2447 ctx_clear_mount_opt(ctx, m->mount_opt);
2448 }
2449 }
2450
2451 return 0;
2452}
2453
2454static int parse_options(struct fs_context *fc, char *options)
2455{
2456 struct fs_parameter param;
2457 int ret;
2458 char *key;
2459
2460 if (!options)
2461 return 0;
2462
2463 while ((key = strsep(&options, ",")) != NULL) {
2464 if (*key) {
2465 size_t v_len = 0;
2466 char *value = strchr(key, '=');
2467
2468 param.type = fs_value_is_flag;
2469 param.string = NULL;
2470
2471 if (value) {
2472 if (value == key)
2473 continue;
2474
2475 *value++ = 0;
2476 v_len = strlen(value);
2477 param.string = kmemdup_nul(value, v_len,
2478 GFP_KERNEL);
2479 if (!param.string)
2480 return -ENOMEM;
2481 param.type = fs_value_is_string;
2482 }
2483
2484 param.key = key;
2485 param.size = v_len;
2486
2487 ret = ext4_parse_param(fc, ¶m);
2488 if (param.string)
2489 kfree(param.string);
2490 if (ret < 0)
2491 return ret;
2492 }
2493 }
2494
2495 ret = ext4_validate_options(fc);
2496 if (ret < 0)
2497 return ret;
2498
2499 return 0;
2500}
2501
2502static int parse_apply_sb_mount_options(struct super_block *sb,
2503 struct ext4_fs_context *m_ctx)
2504{
2505 struct ext4_sb_info *sbi = EXT4_SB(sb);
2506 char *s_mount_opts = NULL;
2507 struct ext4_fs_context *s_ctx = NULL;
2508 struct fs_context *fc = NULL;
2509 int ret = -ENOMEM;
2510
2511 if (!sbi->s_es->s_mount_opts[0])
2512 return 0;
2513
2514 s_mount_opts = kstrndup(sbi->s_es->s_mount_opts,
2515 sizeof(sbi->s_es->s_mount_opts),
2516 GFP_KERNEL);
2517 if (!s_mount_opts)
2518 return ret;
2519
2520 fc = kzalloc(sizeof(struct fs_context), GFP_KERNEL);
2521 if (!fc)
2522 goto out_free;
2523
2524 s_ctx = kzalloc(sizeof(struct ext4_fs_context), GFP_KERNEL);
2525 if (!s_ctx)
2526 goto out_free;
2527
2528 fc->fs_private = s_ctx;
2529 fc->s_fs_info = sbi;
2530
2531 ret = parse_options(fc, s_mount_opts);
2532 if (ret < 0)
2533 goto parse_failed;
2534
2535 ret = ext4_check_opt_consistency(fc, sb);
2536 if (ret < 0) {
2537parse_failed:
2538 ext4_msg(sb, KERN_WARNING,
2539 "failed to parse options in superblock: %s",
2540 s_mount_opts);
2541 ret = 0;
2542 goto out_free;
2543 }
2544
2545 if (s_ctx->spec & EXT4_SPEC_JOURNAL_DEV)
2546 m_ctx->journal_devnum = s_ctx->journal_devnum;
2547 if (s_ctx->spec & EXT4_SPEC_JOURNAL_IOPRIO)
2548 m_ctx->journal_ioprio = s_ctx->journal_ioprio;
2549
2550 ext4_apply_options(fc, sb);
2551 ret = 0;
2552
2553out_free:
2554 if (fc) {
2555 ext4_fc_free(fc);
2556 kfree(fc);
2557 }
2558 kfree(s_mount_opts);
2559 return ret;
2560}
2561
2562static void ext4_apply_quota_options(struct fs_context *fc,
2563 struct super_block *sb)
2564{
2565#ifdef CONFIG_QUOTA
2566 bool quota_feature = ext4_has_feature_quota(sb);
2567 struct ext4_fs_context *ctx = fc->fs_private;
2568 struct ext4_sb_info *sbi = EXT4_SB(sb);
2569 char *qname;
2570 int i;
2571
2572 if (quota_feature)
2573 return;
2574
2575 if (ctx->spec & EXT4_SPEC_JQUOTA) {
2576 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2577 if (!(ctx->qname_spec & (1 << i)))
2578 continue;
2579
2580 qname = ctx->s_qf_names[i]; /* May be NULL */
2581 if (qname)
2582 set_opt(sb, QUOTA);
2583 ctx->s_qf_names[i] = NULL;
2584 qname = rcu_replace_pointer(sbi->s_qf_names[i], qname,
2585 lockdep_is_held(&sb->s_umount));
2586 if (qname)
2587 kfree_rcu_mightsleep(qname);
2588 }
2589 }
2590
2591 if (ctx->spec & EXT4_SPEC_JQFMT)
2592 sbi->s_jquota_fmt = ctx->s_jquota_fmt;
2593#endif
2594}
2595
2596/*
2597 * Check quota settings consistency.
2598 */
2599static int ext4_check_quota_consistency(struct fs_context *fc,
2600 struct super_block *sb)
2601{
2602#ifdef CONFIG_QUOTA
2603 struct ext4_fs_context *ctx = fc->fs_private;
2604 struct ext4_sb_info *sbi = EXT4_SB(sb);
2605 bool quota_feature = ext4_has_feature_quota(sb);
2606 bool quota_loaded = sb_any_quota_loaded(sb);
2607 bool usr_qf_name, grp_qf_name, usrquota, grpquota;
2608 int quota_flags, i;
2609
2610 /*
2611 * We do the test below only for project quotas. 'usrquota' and
2612 * 'grpquota' mount options are allowed even without quota feature
2613 * to support legacy quotas in quota files.
2614 */
2615 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_PRJQUOTA) &&
2616 !ext4_has_feature_project(sb)) {
2617 ext4_msg(NULL, KERN_ERR, "Project quota feature not enabled. "
2618 "Cannot enable project quota enforcement.");
2619 return -EINVAL;
2620 }
2621
2622 quota_flags = EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
2623 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA;
2624 if (quota_loaded &&
2625 ctx->mask_s_mount_opt & quota_flags &&
2626 !ctx_test_mount_opt(ctx, quota_flags))
2627 goto err_quota_change;
2628
2629 if (ctx->spec & EXT4_SPEC_JQUOTA) {
2630
2631 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2632 if (!(ctx->qname_spec & (1 << i)))
2633 continue;
2634
2635 if (quota_loaded &&
2636 !!sbi->s_qf_names[i] != !!ctx->s_qf_names[i])
2637 goto err_jquota_change;
2638
2639 if (sbi->s_qf_names[i] && ctx->s_qf_names[i] &&
2640 strcmp(get_qf_name(sb, sbi, i),
2641 ctx->s_qf_names[i]) != 0)
2642 goto err_jquota_specified;
2643 }
2644
2645 if (quota_feature) {
2646 ext4_msg(NULL, KERN_INFO,
2647 "Journaled quota options ignored when "
2648 "QUOTA feature is enabled");
2649 return 0;
2650 }
2651 }
2652
2653 if (ctx->spec & EXT4_SPEC_JQFMT) {
2654 if (sbi->s_jquota_fmt != ctx->s_jquota_fmt && quota_loaded)
2655 goto err_jquota_change;
2656 if (quota_feature) {
2657 ext4_msg(NULL, KERN_INFO, "Quota format mount options "
2658 "ignored when QUOTA feature is enabled");
2659 return 0;
2660 }
2661 }
2662
2663 /* Make sure we don't mix old and new quota format */
2664 usr_qf_name = (get_qf_name(sb, sbi, USRQUOTA) ||
2665 ctx->s_qf_names[USRQUOTA]);
2666 grp_qf_name = (get_qf_name(sb, sbi, GRPQUOTA) ||
2667 ctx->s_qf_names[GRPQUOTA]);
2668
2669 usrquota = (ctx_test_mount_opt(ctx, EXT4_MOUNT_USRQUOTA) ||
2670 test_opt(sb, USRQUOTA));
2671
2672 grpquota = (ctx_test_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA) ||
2673 test_opt(sb, GRPQUOTA));
2674
2675 if (usr_qf_name) {
2676 ctx_clear_mount_opt(ctx, EXT4_MOUNT_USRQUOTA);
2677 usrquota = false;
2678 }
2679 if (grp_qf_name) {
2680 ctx_clear_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA);
2681 grpquota = false;
2682 }
2683
2684 if (usr_qf_name || grp_qf_name) {
2685 if (usrquota || grpquota) {
2686 ext4_msg(NULL, KERN_ERR, "old and new quota "
2687 "format mixing");
2688 return -EINVAL;
2689 }
2690
2691 if (!(ctx->spec & EXT4_SPEC_JQFMT || sbi->s_jquota_fmt)) {
2692 ext4_msg(NULL, KERN_ERR, "journaled quota format "
2693 "not specified");
2694 return -EINVAL;
2695 }
2696 }
2697
2698 return 0;
2699
2700err_quota_change:
2701 ext4_msg(NULL, KERN_ERR,
2702 "Cannot change quota options when quota turned on");
2703 return -EINVAL;
2704err_jquota_change:
2705 ext4_msg(NULL, KERN_ERR, "Cannot change journaled quota "
2706 "options when quota turned on");
2707 return -EINVAL;
2708err_jquota_specified:
2709 ext4_msg(NULL, KERN_ERR, "%s quota file already specified",
2710 QTYPE2NAME(i));
2711 return -EINVAL;
2712#else
2713 return 0;
2714#endif
2715}
2716
2717static int ext4_check_test_dummy_encryption(const struct fs_context *fc,
2718 struct super_block *sb)
2719{
2720 const struct ext4_fs_context *ctx = fc->fs_private;
2721 const struct ext4_sb_info *sbi = EXT4_SB(sb);
2722
2723 if (!fscrypt_is_dummy_policy_set(&ctx->dummy_enc_policy))
2724 return 0;
2725
2726 if (!ext4_has_feature_encrypt(sb)) {
2727 ext4_msg(NULL, KERN_WARNING,
2728 "test_dummy_encryption requires encrypt feature");
2729 return -EINVAL;
2730 }
2731 /*
2732 * This mount option is just for testing, and it's not worthwhile to
2733 * implement the extra complexity (e.g. RCU protection) that would be
2734 * needed to allow it to be set or changed during remount. We do allow
2735 * it to be specified during remount, but only if there is no change.
2736 */
2737 if (fc->purpose == FS_CONTEXT_FOR_RECONFIGURE) {
2738 if (fscrypt_dummy_policies_equal(&sbi->s_dummy_enc_policy,
2739 &ctx->dummy_enc_policy))
2740 return 0;
2741 ext4_msg(NULL, KERN_WARNING,
2742 "Can't set or change test_dummy_encryption on remount");
2743 return -EINVAL;
2744 }
2745 /* Also make sure s_mount_opts didn't contain a conflicting value. */
2746 if (fscrypt_is_dummy_policy_set(&sbi->s_dummy_enc_policy)) {
2747 if (fscrypt_dummy_policies_equal(&sbi->s_dummy_enc_policy,
2748 &ctx->dummy_enc_policy))
2749 return 0;
2750 ext4_msg(NULL, KERN_WARNING,
2751 "Conflicting test_dummy_encryption options");
2752 return -EINVAL;
2753 }
2754 return 0;
2755}
2756
2757static void ext4_apply_test_dummy_encryption(struct ext4_fs_context *ctx,
2758 struct super_block *sb)
2759{
2760 if (!fscrypt_is_dummy_policy_set(&ctx->dummy_enc_policy) ||
2761 /* if already set, it was already verified to be the same */
2762 fscrypt_is_dummy_policy_set(&EXT4_SB(sb)->s_dummy_enc_policy))
2763 return;
2764 EXT4_SB(sb)->s_dummy_enc_policy = ctx->dummy_enc_policy;
2765 memset(&ctx->dummy_enc_policy, 0, sizeof(ctx->dummy_enc_policy));
2766 ext4_msg(sb, KERN_WARNING, "Test dummy encryption mode enabled");
2767}
2768
2769static int ext4_check_opt_consistency(struct fs_context *fc,
2770 struct super_block *sb)
2771{
2772 struct ext4_fs_context *ctx = fc->fs_private;
2773 struct ext4_sb_info *sbi = fc->s_fs_info;
2774 int is_remount = fc->purpose == FS_CONTEXT_FOR_RECONFIGURE;
2775 int err;
2776
2777 if ((ctx->opt_flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
2778 ext4_msg(NULL, KERN_ERR,
2779 "Mount option(s) incompatible with ext2");
2780 return -EINVAL;
2781 }
2782 if ((ctx->opt_flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
2783 ext4_msg(NULL, KERN_ERR,
2784 "Mount option(s) incompatible with ext3");
2785 return -EINVAL;
2786 }
2787
2788 if (ctx->s_want_extra_isize >
2789 (sbi->s_inode_size - EXT4_GOOD_OLD_INODE_SIZE)) {
2790 ext4_msg(NULL, KERN_ERR,
2791 "Invalid want_extra_isize %d",
2792 ctx->s_want_extra_isize);
2793 return -EINVAL;
2794 }
2795
2796 err = ext4_check_test_dummy_encryption(fc, sb);
2797 if (err)
2798 return err;
2799
2800 if ((ctx->spec & EXT4_SPEC_DATAJ) && is_remount) {
2801 if (!sbi->s_journal) {
2802 ext4_msg(NULL, KERN_WARNING,
2803 "Remounting file system with no journal "
2804 "so ignoring journalled data option");
2805 ctx_clear_mount_opt(ctx, EXT4_MOUNT_DATA_FLAGS);
2806 } else if (ctx_test_mount_opt(ctx, EXT4_MOUNT_DATA_FLAGS) !=
2807 test_opt(sb, DATA_FLAGS)) {
2808 ext4_msg(NULL, KERN_ERR, "Cannot change data mode "
2809 "on remount");
2810 return -EINVAL;
2811 }
2812 }
2813
2814 if (is_remount) {
2815 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS) &&
2816 (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)) {
2817 ext4_msg(NULL, KERN_ERR, "can't mount with "
2818 "both data=journal and dax");
2819 return -EINVAL;
2820 }
2821
2822 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS) &&
2823 (!(sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) ||
2824 (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER))) {
2825fail_dax_change_remount:
2826 ext4_msg(NULL, KERN_ERR, "can't change "
2827 "dax mount option while remounting");
2828 return -EINVAL;
2829 } else if (ctx_test_mount_opt2(ctx, EXT4_MOUNT2_DAX_NEVER) &&
2830 (!(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) ||
2831 (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS))) {
2832 goto fail_dax_change_remount;
2833 } else if (ctx_test_mount_opt2(ctx, EXT4_MOUNT2_DAX_INODE) &&
2834 ((sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) ||
2835 (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) ||
2836 !(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_INODE))) {
2837 goto fail_dax_change_remount;
2838 }
2839 }
2840
2841 return ext4_check_quota_consistency(fc, sb);
2842}
2843
2844static void ext4_apply_options(struct fs_context *fc, struct super_block *sb)
2845{
2846 struct ext4_fs_context *ctx = fc->fs_private;
2847 struct ext4_sb_info *sbi = fc->s_fs_info;
2848
2849 sbi->s_mount_opt &= ~ctx->mask_s_mount_opt;
2850 sbi->s_mount_opt |= ctx->vals_s_mount_opt;
2851 sbi->s_mount_opt2 &= ~ctx->mask_s_mount_opt2;
2852 sbi->s_mount_opt2 |= ctx->vals_s_mount_opt2;
2853 sb->s_flags &= ~ctx->mask_s_flags;
2854 sb->s_flags |= ctx->vals_s_flags;
2855
2856#define APPLY(X) ({ if (ctx->spec & EXT4_SPEC_##X) sbi->X = ctx->X; })
2857 APPLY(s_commit_interval);
2858 APPLY(s_stripe);
2859 APPLY(s_max_batch_time);
2860 APPLY(s_min_batch_time);
2861 APPLY(s_want_extra_isize);
2862 APPLY(s_inode_readahead_blks);
2863 APPLY(s_max_dir_size_kb);
2864 APPLY(s_li_wait_mult);
2865 APPLY(s_resgid);
2866 APPLY(s_resuid);
2867
2868#ifdef CONFIG_EXT4_DEBUG
2869 APPLY(s_fc_debug_max_replay);
2870#endif
2871
2872 ext4_apply_quota_options(fc, sb);
2873 ext4_apply_test_dummy_encryption(ctx, sb);
2874}
2875
2876
2877static int ext4_validate_options(struct fs_context *fc)
2878{
2879#ifdef CONFIG_QUOTA
2880 struct ext4_fs_context *ctx = fc->fs_private;
2881 char *usr_qf_name, *grp_qf_name;
2882
2883 usr_qf_name = ctx->s_qf_names[USRQUOTA];
2884 grp_qf_name = ctx->s_qf_names[GRPQUOTA];
2885
2886 if (usr_qf_name || grp_qf_name) {
2887 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_USRQUOTA) && usr_qf_name)
2888 ctx_clear_mount_opt(ctx, EXT4_MOUNT_USRQUOTA);
2889
2890 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA) && grp_qf_name)
2891 ctx_clear_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA);
2892
2893 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_USRQUOTA) ||
2894 ctx_test_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA)) {
2895 ext4_msg(NULL, KERN_ERR, "old and new quota "
2896 "format mixing");
2897 return -EINVAL;
2898 }
2899 }
2900#endif
2901 return 1;
2902}
2903
2904static inline void ext4_show_quota_options(struct seq_file *seq,
2905 struct super_block *sb)
2906{
2907#if defined(CONFIG_QUOTA)
2908 struct ext4_sb_info *sbi = EXT4_SB(sb);
2909 char *usr_qf_name, *grp_qf_name;
2910
2911 if (sbi->s_jquota_fmt) {
2912 char *fmtname = "";
2913
2914 switch (sbi->s_jquota_fmt) {
2915 case QFMT_VFS_OLD:
2916 fmtname = "vfsold";
2917 break;
2918 case QFMT_VFS_V0:
2919 fmtname = "vfsv0";
2920 break;
2921 case QFMT_VFS_V1:
2922 fmtname = "vfsv1";
2923 break;
2924 }
2925 seq_printf(seq, ",jqfmt=%s", fmtname);
2926 }
2927
2928 rcu_read_lock();
2929 usr_qf_name = rcu_dereference(sbi->s_qf_names[USRQUOTA]);
2930 grp_qf_name = rcu_dereference(sbi->s_qf_names[GRPQUOTA]);
2931 if (usr_qf_name)
2932 seq_show_option(seq, "usrjquota", usr_qf_name);
2933 if (grp_qf_name)
2934 seq_show_option(seq, "grpjquota", grp_qf_name);
2935 rcu_read_unlock();
2936#endif
2937}
2938
2939static const char *token2str(int token)
2940{
2941 const struct fs_parameter_spec *spec;
2942
2943 for (spec = ext4_param_specs; spec->name != NULL; spec++)
2944 if (spec->opt == token && !spec->type)
2945 break;
2946 return spec->name;
2947}
2948
2949/*
2950 * Show an option if
2951 * - it's set to a non-default value OR
2952 * - if the per-sb default is different from the global default
2953 */
2954static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
2955 int nodefs)
2956{
2957 struct ext4_sb_info *sbi = EXT4_SB(sb);
2958 struct ext4_super_block *es = sbi->s_es;
2959 int def_errors;
2960 const struct mount_opts *m;
2961 char sep = nodefs ? '\n' : ',';
2962
2963#define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
2964#define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2965
2966 if (sbi->s_sb_block != 1)
2967 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
2968
2969 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
2970 int want_set = m->flags & MOPT_SET;
2971 int opt_2 = m->flags & MOPT_2;
2972 unsigned int mount_opt, def_mount_opt;
2973
2974 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
2975 m->flags & MOPT_SKIP)
2976 continue;
2977
2978 if (opt_2) {
2979 mount_opt = sbi->s_mount_opt2;
2980 def_mount_opt = sbi->s_def_mount_opt2;
2981 } else {
2982 mount_opt = sbi->s_mount_opt;
2983 def_mount_opt = sbi->s_def_mount_opt;
2984 }
2985 /* skip if same as the default */
2986 if (!nodefs && !(m->mount_opt & (mount_opt ^ def_mount_opt)))
2987 continue;
2988 /* select Opt_noFoo vs Opt_Foo */
2989 if ((want_set &&
2990 (mount_opt & m->mount_opt) != m->mount_opt) ||
2991 (!want_set && (mount_opt & m->mount_opt)))
2992 continue;
2993 SEQ_OPTS_PRINT("%s", token2str(m->token));
2994 }
2995
2996 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
2997 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
2998 SEQ_OPTS_PRINT("resuid=%u",
2999 from_kuid_munged(&init_user_ns, sbi->s_resuid));
3000 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
3001 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
3002 SEQ_OPTS_PRINT("resgid=%u",
3003 from_kgid_munged(&init_user_ns, sbi->s_resgid));
3004 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
3005 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
3006 SEQ_OPTS_PUTS("errors=remount-ro");
3007 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
3008 SEQ_OPTS_PUTS("errors=continue");
3009 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
3010 SEQ_OPTS_PUTS("errors=panic");
3011 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
3012 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
3013 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
3014 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
3015 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
3016 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
3017 if (nodefs || sbi->s_stripe)
3018 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
3019 if (nodefs || EXT4_MOUNT_DATA_FLAGS &
3020 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
3021 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3022 SEQ_OPTS_PUTS("data=journal");
3023 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3024 SEQ_OPTS_PUTS("data=ordered");
3025 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
3026 SEQ_OPTS_PUTS("data=writeback");
3027 }
3028 if (nodefs ||
3029 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
3030 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
3031 sbi->s_inode_readahead_blks);
3032
3033 if (test_opt(sb, INIT_INODE_TABLE) && (nodefs ||
3034 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
3035 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
3036 if (nodefs || sbi->s_max_dir_size_kb)
3037 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
3038 if (test_opt(sb, DATA_ERR_ABORT))
3039 SEQ_OPTS_PUTS("data_err=abort");
3040
3041 fscrypt_show_test_dummy_encryption(seq, sep, sb);
3042
3043 if (sb->s_flags & SB_INLINECRYPT)
3044 SEQ_OPTS_PUTS("inlinecrypt");
3045
3046 if (test_opt(sb, DAX_ALWAYS)) {
3047 if (IS_EXT2_SB(sb))
3048 SEQ_OPTS_PUTS("dax");
3049 else
3050 SEQ_OPTS_PUTS("dax=always");
3051 } else if (test_opt2(sb, DAX_NEVER)) {
3052 SEQ_OPTS_PUTS("dax=never");
3053 } else if (test_opt2(sb, DAX_INODE)) {
3054 SEQ_OPTS_PUTS("dax=inode");
3055 }
3056
3057 if (sbi->s_groups_count >= MB_DEFAULT_LINEAR_SCAN_THRESHOLD &&
3058 !test_opt2(sb, MB_OPTIMIZE_SCAN)) {
3059 SEQ_OPTS_PUTS("mb_optimize_scan=0");
3060 } else if (sbi->s_groups_count < MB_DEFAULT_LINEAR_SCAN_THRESHOLD &&
3061 test_opt2(sb, MB_OPTIMIZE_SCAN)) {
3062 SEQ_OPTS_PUTS("mb_optimize_scan=1");
3063 }
3064
3065 ext4_show_quota_options(seq, sb);
3066 return 0;
3067}
3068
3069static int ext4_show_options(struct seq_file *seq, struct dentry *root)
3070{
3071 return _ext4_show_options(seq, root->d_sb, 0);
3072}
3073
3074int ext4_seq_options_show(struct seq_file *seq, void *offset)
3075{
3076 struct super_block *sb = seq->private;
3077 int rc;
3078
3079 seq_puts(seq, sb_rdonly(sb) ? "ro" : "rw");
3080 rc = _ext4_show_options(seq, sb, 1);
3081 seq_puts(seq, "\n");
3082 return rc;
3083}
3084
3085static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
3086 int read_only)
3087{
3088 struct ext4_sb_info *sbi = EXT4_SB(sb);
3089 int err = 0;
3090
3091 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
3092 ext4_msg(sb, KERN_ERR, "revision level too high, "
3093 "forcing read-only mode");
3094 err = -EROFS;
3095 goto done;
3096 }
3097 if (read_only)
3098 goto done;
3099 if (!(sbi->s_mount_state & EXT4_VALID_FS))
3100 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
3101 "running e2fsck is recommended");
3102 else if (sbi->s_mount_state & EXT4_ERROR_FS)
3103 ext4_msg(sb, KERN_WARNING,
3104 "warning: mounting fs with errors, "
3105 "running e2fsck is recommended");
3106 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
3107 le16_to_cpu(es->s_mnt_count) >=
3108 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
3109 ext4_msg(sb, KERN_WARNING,
3110 "warning: maximal mount count reached, "
3111 "running e2fsck is recommended");
3112 else if (le32_to_cpu(es->s_checkinterval) &&
3113 (ext4_get_tstamp(es, s_lastcheck) +
3114 le32_to_cpu(es->s_checkinterval) <= ktime_get_real_seconds()))
3115 ext4_msg(sb, KERN_WARNING,
3116 "warning: checktime reached, "
3117 "running e2fsck is recommended");
3118 if (!sbi->s_journal)
3119 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
3120 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
3121 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
3122 le16_add_cpu(&es->s_mnt_count, 1);
3123 ext4_update_tstamp(es, s_mtime);
3124 if (sbi->s_journal) {
3125 ext4_set_feature_journal_needs_recovery(sb);
3126 if (ext4_has_feature_orphan_file(sb))
3127 ext4_set_feature_orphan_present(sb);
3128 }
3129
3130 err = ext4_commit_super(sb);
3131done:
3132 if (test_opt(sb, DEBUG))
3133 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
3134 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
3135 sb->s_blocksize,
3136 sbi->s_groups_count,
3137 EXT4_BLOCKS_PER_GROUP(sb),
3138 EXT4_INODES_PER_GROUP(sb),
3139 sbi->s_mount_opt, sbi->s_mount_opt2);
3140 return err;
3141}
3142
3143int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
3144{
3145 struct ext4_sb_info *sbi = EXT4_SB(sb);
3146 struct flex_groups **old_groups, **new_groups;
3147 int size, i, j;
3148
3149 if (!sbi->s_log_groups_per_flex)
3150 return 0;
3151
3152 size = ext4_flex_group(sbi, ngroup - 1) + 1;
3153 if (size <= sbi->s_flex_groups_allocated)
3154 return 0;
3155
3156 new_groups = kvzalloc(roundup_pow_of_two(size *
3157 sizeof(*sbi->s_flex_groups)), GFP_KERNEL);
3158 if (!new_groups) {
3159 ext4_msg(sb, KERN_ERR,
3160 "not enough memory for %d flex group pointers", size);
3161 return -ENOMEM;
3162 }
3163 for (i = sbi->s_flex_groups_allocated; i < size; i++) {
3164 new_groups[i] = kvzalloc(roundup_pow_of_two(
3165 sizeof(struct flex_groups)),
3166 GFP_KERNEL);
3167 if (!new_groups[i]) {
3168 for (j = sbi->s_flex_groups_allocated; j < i; j++)
3169 kvfree(new_groups[j]);
3170 kvfree(new_groups);
3171 ext4_msg(sb, KERN_ERR,
3172 "not enough memory for %d flex groups", size);
3173 return -ENOMEM;
3174 }
3175 }
3176 rcu_read_lock();
3177 old_groups = rcu_dereference(sbi->s_flex_groups);
3178 if (old_groups)
3179 memcpy(new_groups, old_groups,
3180 (sbi->s_flex_groups_allocated *
3181 sizeof(struct flex_groups *)));
3182 rcu_read_unlock();
3183 rcu_assign_pointer(sbi->s_flex_groups, new_groups);
3184 sbi->s_flex_groups_allocated = size;
3185 if (old_groups)
3186 ext4_kvfree_array_rcu(old_groups);
3187 return 0;
3188}
3189
3190static int ext4_fill_flex_info(struct super_block *sb)
3191{
3192 struct ext4_sb_info *sbi = EXT4_SB(sb);
3193 struct ext4_group_desc *gdp = NULL;
3194 struct flex_groups *fg;
3195 ext4_group_t flex_group;
3196 int i, err;
3197
3198 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
3199 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
3200 sbi->s_log_groups_per_flex = 0;
3201 return 1;
3202 }
3203
3204 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
3205 if (err)
3206 goto failed;
3207
3208 for (i = 0; i < sbi->s_groups_count; i++) {
3209 gdp = ext4_get_group_desc(sb, i, NULL);
3210
3211 flex_group = ext4_flex_group(sbi, i);
3212 fg = sbi_array_rcu_deref(sbi, s_flex_groups, flex_group);
3213 atomic_add(ext4_free_inodes_count(sb, gdp), &fg->free_inodes);
3214 atomic64_add(ext4_free_group_clusters(sb, gdp),
3215 &fg->free_clusters);
3216 atomic_add(ext4_used_dirs_count(sb, gdp), &fg->used_dirs);
3217 }
3218
3219 return 1;
3220failed:
3221 return 0;
3222}
3223
3224static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
3225 struct ext4_group_desc *gdp)
3226{
3227 int offset = offsetof(struct ext4_group_desc, bg_checksum);
3228 __u16 crc = 0;
3229 __le32 le_group = cpu_to_le32(block_group);
3230 struct ext4_sb_info *sbi = EXT4_SB(sb);
3231
3232 if (ext4_has_metadata_csum(sbi->s_sb)) {
3233 /* Use new metadata_csum algorithm */
3234 __u32 csum32;
3235 __u16 dummy_csum = 0;
3236
3237 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
3238 sizeof(le_group));
3239 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset);
3240 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum,
3241 sizeof(dummy_csum));
3242 offset += sizeof(dummy_csum);
3243 if (offset < sbi->s_desc_size)
3244 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset,
3245 sbi->s_desc_size - offset);
3246
3247 crc = csum32 & 0xFFFF;
3248 goto out;
3249 }
3250
3251 /* old crc16 code */
3252 if (!ext4_has_feature_gdt_csum(sb))
3253 return 0;
3254
3255 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
3256 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
3257 crc = crc16(crc, (__u8 *)gdp, offset);
3258 offset += sizeof(gdp->bg_checksum); /* skip checksum */
3259 /* for checksum of struct ext4_group_desc do the rest...*/
3260 if (ext4_has_feature_64bit(sb) && offset < sbi->s_desc_size)
3261 crc = crc16(crc, (__u8 *)gdp + offset,
3262 sbi->s_desc_size - offset);
3263
3264out:
3265 return cpu_to_le16(crc);
3266}
3267
3268int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
3269 struct ext4_group_desc *gdp)
3270{
3271 if (ext4_has_group_desc_csum(sb) &&
3272 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
3273 return 0;
3274
3275 return 1;
3276}
3277
3278void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
3279 struct ext4_group_desc *gdp)
3280{
3281 if (!ext4_has_group_desc_csum(sb))
3282 return;
3283 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
3284}
3285
3286/* Called at mount-time, super-block is locked */
3287static int ext4_check_descriptors(struct super_block *sb,
3288 ext4_fsblk_t sb_block,
3289 ext4_group_t *first_not_zeroed)
3290{
3291 struct ext4_sb_info *sbi = EXT4_SB(sb);
3292 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
3293 ext4_fsblk_t last_block;
3294 ext4_fsblk_t last_bg_block = sb_block + ext4_bg_num_gdb(sb, 0);
3295 ext4_fsblk_t block_bitmap;
3296 ext4_fsblk_t inode_bitmap;
3297 ext4_fsblk_t inode_table;
3298 int flexbg_flag = 0;
3299 ext4_group_t i, grp = sbi->s_groups_count;
3300
3301 if (ext4_has_feature_flex_bg(sb))
3302 flexbg_flag = 1;
3303
3304 ext4_debug("Checking group descriptors");
3305
3306 for (i = 0; i < sbi->s_groups_count; i++) {
3307 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
3308
3309 if (i == sbi->s_groups_count - 1 || flexbg_flag)
3310 last_block = ext4_blocks_count(sbi->s_es) - 1;
3311 else
3312 last_block = first_block +
3313 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
3314
3315 if ((grp == sbi->s_groups_count) &&
3316 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3317 grp = i;
3318
3319 block_bitmap = ext4_block_bitmap(sb, gdp);
3320 if (block_bitmap == sb_block) {
3321 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3322 "Block bitmap for group %u overlaps "
3323 "superblock", i);
3324 if (!sb_rdonly(sb))
3325 return 0;
3326 }
3327 if (block_bitmap >= sb_block + 1 &&
3328 block_bitmap <= last_bg_block) {
3329 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3330 "Block bitmap for group %u overlaps "
3331 "block group descriptors", i);
3332 if (!sb_rdonly(sb))
3333 return 0;
3334 }
3335 if (block_bitmap < first_block || block_bitmap > last_block) {
3336 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3337 "Block bitmap for group %u not in group "
3338 "(block %llu)!", i, block_bitmap);
3339 return 0;
3340 }
3341 inode_bitmap = ext4_inode_bitmap(sb, gdp);
3342 if (inode_bitmap == sb_block) {
3343 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3344 "Inode bitmap for group %u overlaps "
3345 "superblock", i);
3346 if (!sb_rdonly(sb))
3347 return 0;
3348 }
3349 if (inode_bitmap >= sb_block + 1 &&
3350 inode_bitmap <= last_bg_block) {
3351 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3352 "Inode bitmap for group %u overlaps "
3353 "block group descriptors", i);
3354 if (!sb_rdonly(sb))
3355 return 0;
3356 }
3357 if (inode_bitmap < first_block || inode_bitmap > last_block) {
3358 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3359 "Inode bitmap for group %u not in group "
3360 "(block %llu)!", i, inode_bitmap);
3361 return 0;
3362 }
3363 inode_table = ext4_inode_table(sb, gdp);
3364 if (inode_table == sb_block) {
3365 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3366 "Inode table for group %u overlaps "
3367 "superblock", i);
3368 if (!sb_rdonly(sb))
3369 return 0;
3370 }
3371 if (inode_table >= sb_block + 1 &&
3372 inode_table <= last_bg_block) {
3373 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3374 "Inode table for group %u overlaps "
3375 "block group descriptors", i);
3376 if (!sb_rdonly(sb))
3377 return 0;
3378 }
3379 if (inode_table < first_block ||
3380 inode_table + sbi->s_itb_per_group - 1 > last_block) {
3381 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3382 "Inode table for group %u not in group "
3383 "(block %llu)!", i, inode_table);
3384 return 0;
3385 }
3386 ext4_lock_group(sb, i);
3387 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
3388 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3389 "Checksum for group %u failed (%u!=%u)",
3390 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
3391 gdp)), le16_to_cpu(gdp->bg_checksum));
3392 if (!sb_rdonly(sb)) {
3393 ext4_unlock_group(sb, i);
3394 return 0;
3395 }
3396 }
3397 ext4_unlock_group(sb, i);
3398 if (!flexbg_flag)
3399 first_block += EXT4_BLOCKS_PER_GROUP(sb);
3400 }
3401 if (NULL != first_not_zeroed)
3402 *first_not_zeroed = grp;
3403 return 1;
3404}
3405
3406/*
3407 * Maximal extent format file size.
3408 * Resulting logical blkno at s_maxbytes must fit in our on-disk
3409 * extent format containers, within a sector_t, and within i_blocks
3410 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
3411 * so that won't be a limiting factor.
3412 *
3413 * However there is other limiting factor. We do store extents in the form
3414 * of starting block and length, hence the resulting length of the extent
3415 * covering maximum file size must fit into on-disk format containers as
3416 * well. Given that length is always by 1 unit bigger than max unit (because
3417 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
3418 *
3419 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
3420 */
3421static loff_t ext4_max_size(int blkbits, int has_huge_files)
3422{
3423 loff_t res;
3424 loff_t upper_limit = MAX_LFS_FILESIZE;
3425
3426 BUILD_BUG_ON(sizeof(blkcnt_t) < sizeof(u64));
3427
3428 if (!has_huge_files) {
3429 upper_limit = (1LL << 32) - 1;
3430
3431 /* total blocks in file system block size */
3432 upper_limit >>= (blkbits - 9);
3433 upper_limit <<= blkbits;
3434 }
3435
3436 /*
3437 * 32-bit extent-start container, ee_block. We lower the maxbytes
3438 * by one fs block, so ee_len can cover the extent of maximum file
3439 * size
3440 */
3441 res = (1LL << 32) - 1;
3442 res <<= blkbits;
3443
3444 /* Sanity check against vm- & vfs- imposed limits */
3445 if (res > upper_limit)
3446 res = upper_limit;
3447
3448 return res;
3449}
3450
3451/*
3452 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
3453 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
3454 * We need to be 1 filesystem block less than the 2^48 sector limit.
3455 */
3456static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
3457{
3458 loff_t upper_limit, res = EXT4_NDIR_BLOCKS;
3459 int meta_blocks;
3460 unsigned int ppb = 1 << (bits - 2);
3461
3462 /*
3463 * This is calculated to be the largest file size for a dense, block
3464 * mapped file such that the file's total number of 512-byte sectors,
3465 * including data and all indirect blocks, does not exceed (2^48 - 1).
3466 *
3467 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
3468 * number of 512-byte sectors of the file.
3469 */
3470 if (!has_huge_files) {
3471 /*
3472 * !has_huge_files or implies that the inode i_block field
3473 * represents total file blocks in 2^32 512-byte sectors ==
3474 * size of vfs inode i_blocks * 8
3475 */
3476 upper_limit = (1LL << 32) - 1;
3477
3478 /* total blocks in file system block size */
3479 upper_limit >>= (bits - 9);
3480
3481 } else {
3482 /*
3483 * We use 48 bit ext4_inode i_blocks
3484 * With EXT4_HUGE_FILE_FL set the i_blocks
3485 * represent total number of blocks in
3486 * file system block size
3487 */
3488 upper_limit = (1LL << 48) - 1;
3489
3490 }
3491
3492 /* Compute how many blocks we can address by block tree */
3493 res += ppb;
3494 res += ppb * ppb;
3495 res += ((loff_t)ppb) * ppb * ppb;
3496 /* Compute how many metadata blocks are needed */
3497 meta_blocks = 1;
3498 meta_blocks += 1 + ppb;
3499 meta_blocks += 1 + ppb + ppb * ppb;
3500 /* Does block tree limit file size? */
3501 if (res + meta_blocks <= upper_limit)
3502 goto check_lfs;
3503
3504 res = upper_limit;
3505 /* How many metadata blocks are needed for addressing upper_limit? */
3506 upper_limit -= EXT4_NDIR_BLOCKS;
3507 /* indirect blocks */
3508 meta_blocks = 1;
3509 upper_limit -= ppb;
3510 /* double indirect blocks */
3511 if (upper_limit < ppb * ppb) {
3512 meta_blocks += 1 + DIV_ROUND_UP_ULL(upper_limit, ppb);
3513 res -= meta_blocks;
3514 goto check_lfs;
3515 }
3516 meta_blocks += 1 + ppb;
3517 upper_limit -= ppb * ppb;
3518 /* tripple indirect blocks for the rest */
3519 meta_blocks += 1 + DIV_ROUND_UP_ULL(upper_limit, ppb) +
3520 DIV_ROUND_UP_ULL(upper_limit, ppb*ppb);
3521 res -= meta_blocks;
3522check_lfs:
3523 res <<= bits;
3524 if (res > MAX_LFS_FILESIZE)
3525 res = MAX_LFS_FILESIZE;
3526
3527 return res;
3528}
3529
3530static ext4_fsblk_t descriptor_loc(struct super_block *sb,
3531 ext4_fsblk_t logical_sb_block, int nr)
3532{
3533 struct ext4_sb_info *sbi = EXT4_SB(sb);
3534 ext4_group_t bg, first_meta_bg;
3535 int has_super = 0;
3536
3537 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
3538
3539 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
3540 return logical_sb_block + nr + 1;
3541 bg = sbi->s_desc_per_block * nr;
3542 if (ext4_bg_has_super(sb, bg))
3543 has_super = 1;
3544
3545 /*
3546 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
3547 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
3548 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
3549 * compensate.
3550 */
3551 if (sb->s_blocksize == 1024 && nr == 0 &&
3552 le32_to_cpu(sbi->s_es->s_first_data_block) == 0)
3553 has_super++;
3554
3555 return (has_super + ext4_group_first_block_no(sb, bg));
3556}
3557
3558/**
3559 * ext4_get_stripe_size: Get the stripe size.
3560 * @sbi: In memory super block info
3561 *
3562 * If we have specified it via mount option, then
3563 * use the mount option value. If the value specified at mount time is
3564 * greater than the blocks per group use the super block value.
3565 * If the super block value is greater than blocks per group return 0.
3566 * Allocator needs it be less than blocks per group.
3567 *
3568 */
3569static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
3570{
3571 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
3572 unsigned long stripe_width =
3573 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3574 int ret;
3575
3576 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3577 ret = sbi->s_stripe;
3578 else if (stripe_width && stripe_width <= sbi->s_blocks_per_group)
3579 ret = stripe_width;
3580 else if (stride && stride <= sbi->s_blocks_per_group)
3581 ret = stride;
3582 else
3583 ret = 0;
3584
3585 /*
3586 * If the stripe width is 1, this makes no sense and
3587 * we set it to 0 to turn off stripe handling code.
3588 */
3589 if (ret <= 1)
3590 ret = 0;
3591
3592 return ret;
3593}
3594
3595/*
3596 * Check whether this filesystem can be mounted based on
3597 * the features present and the RDONLY/RDWR mount requested.
3598 * Returns 1 if this filesystem can be mounted as requested,
3599 * 0 if it cannot be.
3600 */
3601int ext4_feature_set_ok(struct super_block *sb, int readonly)
3602{
3603 if (ext4_has_unknown_ext4_incompat_features(sb)) {
3604 ext4_msg(sb, KERN_ERR,
3605 "Couldn't mount because of "
3606 "unsupported optional features (%x)",
3607 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
3608 ~EXT4_FEATURE_INCOMPAT_SUPP));
3609 return 0;
3610 }
3611
3612#if !IS_ENABLED(CONFIG_UNICODE)
3613 if (ext4_has_feature_casefold(sb)) {
3614 ext4_msg(sb, KERN_ERR,
3615 "Filesystem with casefold feature cannot be "
3616 "mounted without CONFIG_UNICODE");
3617 return 0;
3618 }
3619#endif
3620
3621 if (readonly)
3622 return 1;
3623
3624 if (ext4_has_feature_readonly(sb)) {
3625 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
3626 sb->s_flags |= SB_RDONLY;
3627 return 1;
3628 }
3629
3630 /* Check that feature set is OK for a read-write mount */
3631 if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
3632 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
3633 "unsupported optional features (%x)",
3634 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
3635 ~EXT4_FEATURE_RO_COMPAT_SUPP));
3636 return 0;
3637 }
3638 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
3639 ext4_msg(sb, KERN_ERR,
3640 "Can't support bigalloc feature without "
3641 "extents feature\n");
3642 return 0;
3643 }
3644
3645#if !IS_ENABLED(CONFIG_QUOTA) || !IS_ENABLED(CONFIG_QFMT_V2)
3646 if (!readonly && (ext4_has_feature_quota(sb) ||
3647 ext4_has_feature_project(sb))) {
3648 ext4_msg(sb, KERN_ERR,
3649 "The kernel was not built with CONFIG_QUOTA and CONFIG_QFMT_V2");
3650 return 0;
3651 }
3652#endif /* CONFIG_QUOTA */
3653 return 1;
3654}
3655
3656/*
3657 * This function is called once a day if we have errors logged
3658 * on the file system
3659 */
3660static void print_daily_error_info(struct timer_list *t)
3661{
3662 struct ext4_sb_info *sbi = from_timer(sbi, t, s_err_report);
3663 struct super_block *sb = sbi->s_sb;
3664 struct ext4_super_block *es = sbi->s_es;
3665
3666 if (es->s_error_count)
3667 /* fsck newer than v1.41.13 is needed to clean this condition. */
3668 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
3669 le32_to_cpu(es->s_error_count));
3670 if (es->s_first_error_time) {
3671 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %llu: %.*s:%d",
3672 sb->s_id,
3673 ext4_get_tstamp(es, s_first_error_time),
3674 (int) sizeof(es->s_first_error_func),
3675 es->s_first_error_func,
3676 le32_to_cpu(es->s_first_error_line));
3677 if (es->s_first_error_ino)
3678 printk(KERN_CONT ": inode %u",
3679 le32_to_cpu(es->s_first_error_ino));
3680 if (es->s_first_error_block)
3681 printk(KERN_CONT ": block %llu", (unsigned long long)
3682 le64_to_cpu(es->s_first_error_block));
3683 printk(KERN_CONT "\n");
3684 }
3685 if (es->s_last_error_time) {
3686 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %llu: %.*s:%d",
3687 sb->s_id,
3688 ext4_get_tstamp(es, s_last_error_time),
3689 (int) sizeof(es->s_last_error_func),
3690 es->s_last_error_func,
3691 le32_to_cpu(es->s_last_error_line));
3692 if (es->s_last_error_ino)
3693 printk(KERN_CONT ": inode %u",
3694 le32_to_cpu(es->s_last_error_ino));
3695 if (es->s_last_error_block)
3696 printk(KERN_CONT ": block %llu", (unsigned long long)
3697 le64_to_cpu(es->s_last_error_block));
3698 printk(KERN_CONT "\n");
3699 }
3700 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
3701}
3702
3703/* Find next suitable group and run ext4_init_inode_table */
3704static int ext4_run_li_request(struct ext4_li_request *elr)
3705{
3706 struct ext4_group_desc *gdp = NULL;
3707 struct super_block *sb = elr->lr_super;
3708 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3709 ext4_group_t group = elr->lr_next_group;
3710 unsigned int prefetch_ios = 0;
3711 int ret = 0;
3712 int nr = EXT4_SB(sb)->s_mb_prefetch;
3713 u64 start_time;
3714
3715 if (elr->lr_mode == EXT4_LI_MODE_PREFETCH_BBITMAP) {
3716 elr->lr_next_group = ext4_mb_prefetch(sb, group, nr, &prefetch_ios);
3717 ext4_mb_prefetch_fini(sb, elr->lr_next_group, nr);
3718 trace_ext4_prefetch_bitmaps(sb, group, elr->lr_next_group, nr);
3719 if (group >= elr->lr_next_group) {
3720 ret = 1;
3721 if (elr->lr_first_not_zeroed != ngroups &&
3722 !sb_rdonly(sb) && test_opt(sb, INIT_INODE_TABLE)) {
3723 elr->lr_next_group = elr->lr_first_not_zeroed;
3724 elr->lr_mode = EXT4_LI_MODE_ITABLE;
3725 ret = 0;
3726 }
3727 }
3728 return ret;
3729 }
3730
3731 for (; group < ngroups; group++) {
3732 gdp = ext4_get_group_desc(sb, group, NULL);
3733 if (!gdp) {
3734 ret = 1;
3735 break;
3736 }
3737
3738 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3739 break;
3740 }
3741
3742 if (group >= ngroups)
3743 ret = 1;
3744
3745 if (!ret) {
3746 start_time = ktime_get_real_ns();
3747 ret = ext4_init_inode_table(sb, group,
3748 elr->lr_timeout ? 0 : 1);
3749 trace_ext4_lazy_itable_init(sb, group);
3750 if (elr->lr_timeout == 0) {
3751 elr->lr_timeout = nsecs_to_jiffies((ktime_get_real_ns() - start_time) *
3752 EXT4_SB(elr->lr_super)->s_li_wait_mult);
3753 }
3754 elr->lr_next_sched = jiffies + elr->lr_timeout;
3755 elr->lr_next_group = group + 1;
3756 }
3757 return ret;
3758}
3759
3760/*
3761 * Remove lr_request from the list_request and free the
3762 * request structure. Should be called with li_list_mtx held
3763 */
3764static void ext4_remove_li_request(struct ext4_li_request *elr)
3765{
3766 if (!elr)
3767 return;
3768
3769 list_del(&elr->lr_request);
3770 EXT4_SB(elr->lr_super)->s_li_request = NULL;
3771 kfree(elr);
3772}
3773
3774static void ext4_unregister_li_request(struct super_block *sb)
3775{
3776 mutex_lock(&ext4_li_mtx);
3777 if (!ext4_li_info) {
3778 mutex_unlock(&ext4_li_mtx);
3779 return;
3780 }
3781
3782 mutex_lock(&ext4_li_info->li_list_mtx);
3783 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
3784 mutex_unlock(&ext4_li_info->li_list_mtx);
3785 mutex_unlock(&ext4_li_mtx);
3786}
3787
3788static struct task_struct *ext4_lazyinit_task;
3789
3790/*
3791 * This is the function where ext4lazyinit thread lives. It walks
3792 * through the request list searching for next scheduled filesystem.
3793 * When such a fs is found, run the lazy initialization request
3794 * (ext4_rn_li_request) and keep track of the time spend in this
3795 * function. Based on that time we compute next schedule time of
3796 * the request. When walking through the list is complete, compute
3797 * next waking time and put itself into sleep.
3798 */
3799static int ext4_lazyinit_thread(void *arg)
3800{
3801 struct ext4_lazy_init *eli = arg;
3802 struct list_head *pos, *n;
3803 struct ext4_li_request *elr;
3804 unsigned long next_wakeup, cur;
3805
3806 BUG_ON(NULL == eli);
3807 set_freezable();
3808
3809cont_thread:
3810 while (true) {
3811 next_wakeup = MAX_JIFFY_OFFSET;
3812
3813 mutex_lock(&eli->li_list_mtx);
3814 if (list_empty(&eli->li_request_list)) {
3815 mutex_unlock(&eli->li_list_mtx);
3816 goto exit_thread;
3817 }
3818 list_for_each_safe(pos, n, &eli->li_request_list) {
3819 int err = 0;
3820 int progress = 0;
3821 elr = list_entry(pos, struct ext4_li_request,
3822 lr_request);
3823
3824 if (time_before(jiffies, elr->lr_next_sched)) {
3825 if (time_before(elr->lr_next_sched, next_wakeup))
3826 next_wakeup = elr->lr_next_sched;
3827 continue;
3828 }
3829 if (down_read_trylock(&elr->lr_super->s_umount)) {
3830 if (sb_start_write_trylock(elr->lr_super)) {
3831 progress = 1;
3832 /*
3833 * We hold sb->s_umount, sb can not
3834 * be removed from the list, it is
3835 * now safe to drop li_list_mtx
3836 */
3837 mutex_unlock(&eli->li_list_mtx);
3838 err = ext4_run_li_request(elr);
3839 sb_end_write(elr->lr_super);
3840 mutex_lock(&eli->li_list_mtx);
3841 n = pos->next;
3842 }
3843 up_read((&elr->lr_super->s_umount));
3844 }
3845 /* error, remove the lazy_init job */
3846 if (err) {
3847 ext4_remove_li_request(elr);
3848 continue;
3849 }
3850 if (!progress) {
3851 elr->lr_next_sched = jiffies +
3852 get_random_u32_below(EXT4_DEF_LI_MAX_START_DELAY * HZ);
3853 }
3854 if (time_before(elr->lr_next_sched, next_wakeup))
3855 next_wakeup = elr->lr_next_sched;
3856 }
3857 mutex_unlock(&eli->li_list_mtx);
3858
3859 try_to_freeze();
3860
3861 cur = jiffies;
3862 if ((time_after_eq(cur, next_wakeup)) ||
3863 (MAX_JIFFY_OFFSET == next_wakeup)) {
3864 cond_resched();
3865 continue;
3866 }
3867
3868 schedule_timeout_interruptible(next_wakeup - cur);
3869
3870 if (kthread_should_stop()) {
3871 ext4_clear_request_list();
3872 goto exit_thread;
3873 }
3874 }
3875
3876exit_thread:
3877 /*
3878 * It looks like the request list is empty, but we need
3879 * to check it under the li_list_mtx lock, to prevent any
3880 * additions into it, and of course we should lock ext4_li_mtx
3881 * to atomically free the list and ext4_li_info, because at
3882 * this point another ext4 filesystem could be registering
3883 * new one.
3884 */
3885 mutex_lock(&ext4_li_mtx);
3886 mutex_lock(&eli->li_list_mtx);
3887 if (!list_empty(&eli->li_request_list)) {
3888 mutex_unlock(&eli->li_list_mtx);
3889 mutex_unlock(&ext4_li_mtx);
3890 goto cont_thread;
3891 }
3892 mutex_unlock(&eli->li_list_mtx);
3893 kfree(ext4_li_info);
3894 ext4_li_info = NULL;
3895 mutex_unlock(&ext4_li_mtx);
3896
3897 return 0;
3898}
3899
3900static void ext4_clear_request_list(void)
3901{
3902 struct list_head *pos, *n;
3903 struct ext4_li_request *elr;
3904
3905 mutex_lock(&ext4_li_info->li_list_mtx);
3906 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3907 elr = list_entry(pos, struct ext4_li_request,
3908 lr_request);
3909 ext4_remove_li_request(elr);
3910 }
3911 mutex_unlock(&ext4_li_info->li_list_mtx);
3912}
3913
3914static int ext4_run_lazyinit_thread(void)
3915{
3916 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3917 ext4_li_info, "ext4lazyinit");
3918 if (IS_ERR(ext4_lazyinit_task)) {
3919 int err = PTR_ERR(ext4_lazyinit_task);
3920 ext4_clear_request_list();
3921 kfree(ext4_li_info);
3922 ext4_li_info = NULL;
3923 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
3924 "initialization thread\n",
3925 err);
3926 return err;
3927 }
3928 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
3929 return 0;
3930}
3931
3932/*
3933 * Check whether it make sense to run itable init. thread or not.
3934 * If there is at least one uninitialized inode table, return
3935 * corresponding group number, else the loop goes through all
3936 * groups and return total number of groups.
3937 */
3938static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3939{
3940 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3941 struct ext4_group_desc *gdp = NULL;
3942
3943 if (!ext4_has_group_desc_csum(sb))
3944 return ngroups;
3945
3946 for (group = 0; group < ngroups; group++) {
3947 gdp = ext4_get_group_desc(sb, group, NULL);
3948 if (!gdp)
3949 continue;
3950
3951 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3952 break;
3953 }
3954
3955 return group;
3956}
3957
3958static int ext4_li_info_new(void)
3959{
3960 struct ext4_lazy_init *eli = NULL;
3961
3962 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3963 if (!eli)
3964 return -ENOMEM;
3965
3966 INIT_LIST_HEAD(&eli->li_request_list);
3967 mutex_init(&eli->li_list_mtx);
3968
3969 eli->li_state |= EXT4_LAZYINIT_QUIT;
3970
3971 ext4_li_info = eli;
3972
3973 return 0;
3974}
3975
3976static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3977 ext4_group_t start)
3978{
3979 struct ext4_li_request *elr;
3980
3981 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3982 if (!elr)
3983 return NULL;
3984
3985 elr->lr_super = sb;
3986 elr->lr_first_not_zeroed = start;
3987 if (test_opt(sb, NO_PREFETCH_BLOCK_BITMAPS)) {
3988 elr->lr_mode = EXT4_LI_MODE_ITABLE;
3989 elr->lr_next_group = start;
3990 } else {
3991 elr->lr_mode = EXT4_LI_MODE_PREFETCH_BBITMAP;
3992 }
3993
3994 /*
3995 * Randomize first schedule time of the request to
3996 * spread the inode table initialization requests
3997 * better.
3998 */
3999 elr->lr_next_sched = jiffies + get_random_u32_below(EXT4_DEF_LI_MAX_START_DELAY * HZ);
4000 return elr;
4001}
4002
4003int ext4_register_li_request(struct super_block *sb,
4004 ext4_group_t first_not_zeroed)
4005{
4006 struct ext4_sb_info *sbi = EXT4_SB(sb);
4007 struct ext4_li_request *elr = NULL;
4008 ext4_group_t ngroups = sbi->s_groups_count;
4009 int ret = 0;
4010
4011 mutex_lock(&ext4_li_mtx);
4012 if (sbi->s_li_request != NULL) {
4013 /*
4014 * Reset timeout so it can be computed again, because
4015 * s_li_wait_mult might have changed.
4016 */
4017 sbi->s_li_request->lr_timeout = 0;
4018 goto out;
4019 }
4020
4021 if (sb_rdonly(sb) ||
4022 (test_opt(sb, NO_PREFETCH_BLOCK_BITMAPS) &&
4023 (first_not_zeroed == ngroups || !test_opt(sb, INIT_INODE_TABLE))))
4024 goto out;
4025
4026 elr = ext4_li_request_new(sb, first_not_zeroed);
4027 if (!elr) {
4028 ret = -ENOMEM;
4029 goto out;
4030 }
4031
4032 if (NULL == ext4_li_info) {
4033 ret = ext4_li_info_new();
4034 if (ret)
4035 goto out;
4036 }
4037
4038 mutex_lock(&ext4_li_info->li_list_mtx);
4039 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
4040 mutex_unlock(&ext4_li_info->li_list_mtx);
4041
4042 sbi->s_li_request = elr;
4043 /*
4044 * set elr to NULL here since it has been inserted to
4045 * the request_list and the removal and free of it is
4046 * handled by ext4_clear_request_list from now on.
4047 */
4048 elr = NULL;
4049
4050 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
4051 ret = ext4_run_lazyinit_thread();
4052 if (ret)
4053 goto out;
4054 }
4055out:
4056 mutex_unlock(&ext4_li_mtx);
4057 if (ret)
4058 kfree(elr);
4059 return ret;
4060}
4061
4062/*
4063 * We do not need to lock anything since this is called on
4064 * module unload.
4065 */
4066static void ext4_destroy_lazyinit_thread(void)
4067{
4068 /*
4069 * If thread exited earlier
4070 * there's nothing to be done.
4071 */
4072 if (!ext4_li_info || !ext4_lazyinit_task)
4073 return;
4074
4075 kthread_stop(ext4_lazyinit_task);
4076}
4077
4078static int set_journal_csum_feature_set(struct super_block *sb)
4079{
4080 int ret = 1;
4081 int compat, incompat;
4082 struct ext4_sb_info *sbi = EXT4_SB(sb);
4083
4084 if (ext4_has_metadata_csum(sb)) {
4085 /* journal checksum v3 */
4086 compat = 0;
4087 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
4088 } else {
4089 /* journal checksum v1 */
4090 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
4091 incompat = 0;
4092 }
4093
4094 jbd2_journal_clear_features(sbi->s_journal,
4095 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
4096 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
4097 JBD2_FEATURE_INCOMPAT_CSUM_V2);
4098 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4099 ret = jbd2_journal_set_features(sbi->s_journal,
4100 compat, 0,
4101 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
4102 incompat);
4103 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
4104 ret = jbd2_journal_set_features(sbi->s_journal,
4105 compat, 0,
4106 incompat);
4107 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
4108 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
4109 } else {
4110 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
4111 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
4112 }
4113
4114 return ret;
4115}
4116
4117/*
4118 * Note: calculating the overhead so we can be compatible with
4119 * historical BSD practice is quite difficult in the face of
4120 * clusters/bigalloc. This is because multiple metadata blocks from
4121 * different block group can end up in the same allocation cluster.
4122 * Calculating the exact overhead in the face of clustered allocation
4123 * requires either O(all block bitmaps) in memory or O(number of block
4124 * groups**2) in time. We will still calculate the superblock for
4125 * older file systems --- and if we come across with a bigalloc file
4126 * system with zero in s_overhead_clusters the estimate will be close to
4127 * correct especially for very large cluster sizes --- but for newer
4128 * file systems, it's better to calculate this figure once at mkfs
4129 * time, and store it in the superblock. If the superblock value is
4130 * present (even for non-bigalloc file systems), we will use it.
4131 */
4132static int count_overhead(struct super_block *sb, ext4_group_t grp,
4133 char *buf)
4134{
4135 struct ext4_sb_info *sbi = EXT4_SB(sb);
4136 struct ext4_group_desc *gdp;
4137 ext4_fsblk_t first_block, last_block, b;
4138 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
4139 int s, j, count = 0;
4140 int has_super = ext4_bg_has_super(sb, grp);
4141
4142 if (!ext4_has_feature_bigalloc(sb))
4143 return (has_super + ext4_bg_num_gdb(sb, grp) +
4144 (has_super ? le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) : 0) +
4145 sbi->s_itb_per_group + 2);
4146
4147 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
4148 (grp * EXT4_BLOCKS_PER_GROUP(sb));
4149 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
4150 for (i = 0; i < ngroups; i++) {
4151 gdp = ext4_get_group_desc(sb, i, NULL);
4152 b = ext4_block_bitmap(sb, gdp);
4153 if (b >= first_block && b <= last_block) {
4154 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
4155 count++;
4156 }
4157 b = ext4_inode_bitmap(sb, gdp);
4158 if (b >= first_block && b <= last_block) {
4159 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
4160 count++;
4161 }
4162 b = ext4_inode_table(sb, gdp);
4163 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
4164 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
4165 int c = EXT4_B2C(sbi, b - first_block);
4166 ext4_set_bit(c, buf);
4167 count++;
4168 }
4169 if (i != grp)
4170 continue;
4171 s = 0;
4172 if (ext4_bg_has_super(sb, grp)) {
4173 ext4_set_bit(s++, buf);
4174 count++;
4175 }
4176 j = ext4_bg_num_gdb(sb, grp);
4177 if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) {
4178 ext4_error(sb, "Invalid number of block group "
4179 "descriptor blocks: %d", j);
4180 j = EXT4_BLOCKS_PER_GROUP(sb) - s;
4181 }
4182 count += j;
4183 for (; j > 0; j--)
4184 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
4185 }
4186 if (!count)
4187 return 0;
4188 return EXT4_CLUSTERS_PER_GROUP(sb) -
4189 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
4190}
4191
4192/*
4193 * Compute the overhead and stash it in sbi->s_overhead
4194 */
4195int ext4_calculate_overhead(struct super_block *sb)
4196{
4197 struct ext4_sb_info *sbi = EXT4_SB(sb);
4198 struct ext4_super_block *es = sbi->s_es;
4199 struct inode *j_inode;
4200 unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum);
4201 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
4202 ext4_fsblk_t overhead = 0;
4203 char *buf = (char *) get_zeroed_page(GFP_NOFS);
4204
4205 if (!buf)
4206 return -ENOMEM;
4207
4208 /*
4209 * Compute the overhead (FS structures). This is constant
4210 * for a given filesystem unless the number of block groups
4211 * changes so we cache the previous value until it does.
4212 */
4213
4214 /*
4215 * All of the blocks before first_data_block are overhead
4216 */
4217 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
4218
4219 /*
4220 * Add the overhead found in each block group
4221 */
4222 for (i = 0; i < ngroups; i++) {
4223 int blks;
4224
4225 blks = count_overhead(sb, i, buf);
4226 overhead += blks;
4227 if (blks)
4228 memset(buf, 0, PAGE_SIZE);
4229 cond_resched();
4230 }
4231
4232 /*
4233 * Add the internal journal blocks whether the journal has been
4234 * loaded or not
4235 */
4236 if (sbi->s_journal && !sbi->s_journal_bdev_handle)
4237 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_total_len);
4238 else if (ext4_has_feature_journal(sb) && !sbi->s_journal && j_inum) {
4239 /* j_inum for internal journal is non-zero */
4240 j_inode = ext4_get_journal_inode(sb, j_inum);
4241 if (!IS_ERR(j_inode)) {
4242 j_blocks = j_inode->i_size >> sb->s_blocksize_bits;
4243 overhead += EXT4_NUM_B2C(sbi, j_blocks);
4244 iput(j_inode);
4245 } else {
4246 ext4_msg(sb, KERN_ERR, "can't get journal size");
4247 }
4248 }
4249 sbi->s_overhead = overhead;
4250 smp_wmb();
4251 free_page((unsigned long) buf);
4252 return 0;
4253}
4254
4255static void ext4_set_resv_clusters(struct super_block *sb)
4256{
4257 ext4_fsblk_t resv_clusters;
4258 struct ext4_sb_info *sbi = EXT4_SB(sb);
4259
4260 /*
4261 * There's no need to reserve anything when we aren't using extents.
4262 * The space estimates are exact, there are no unwritten extents,
4263 * hole punching doesn't need new metadata... This is needed especially
4264 * to keep ext2/3 backward compatibility.
4265 */
4266 if (!ext4_has_feature_extents(sb))
4267 return;
4268 /*
4269 * By default we reserve 2% or 4096 clusters, whichever is smaller.
4270 * This should cover the situations where we can not afford to run
4271 * out of space like for example punch hole, or converting
4272 * unwritten extents in delalloc path. In most cases such
4273 * allocation would require 1, or 2 blocks, higher numbers are
4274 * very rare.
4275 */
4276 resv_clusters = (ext4_blocks_count(sbi->s_es) >>
4277 sbi->s_cluster_bits);
4278
4279 do_div(resv_clusters, 50);
4280 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
4281
4282 atomic64_set(&sbi->s_resv_clusters, resv_clusters);
4283}
4284
4285static const char *ext4_quota_mode(struct super_block *sb)
4286{
4287#ifdef CONFIG_QUOTA
4288 if (!ext4_quota_capable(sb))
4289 return "none";
4290
4291 if (EXT4_SB(sb)->s_journal && ext4_is_quota_journalled(sb))
4292 return "journalled";
4293 else
4294 return "writeback";
4295#else
4296 return "disabled";
4297#endif
4298}
4299
4300static void ext4_setup_csum_trigger(struct super_block *sb,
4301 enum ext4_journal_trigger_type type,
4302 void (*trigger)(
4303 struct jbd2_buffer_trigger_type *type,
4304 struct buffer_head *bh,
4305 void *mapped_data,
4306 size_t size))
4307{
4308 struct ext4_sb_info *sbi = EXT4_SB(sb);
4309
4310 sbi->s_journal_triggers[type].sb = sb;
4311 sbi->s_journal_triggers[type].tr_triggers.t_frozen = trigger;
4312}
4313
4314static void ext4_free_sbi(struct ext4_sb_info *sbi)
4315{
4316 if (!sbi)
4317 return;
4318
4319 kfree(sbi->s_blockgroup_lock);
4320 fs_put_dax(sbi->s_daxdev, NULL);
4321 kfree(sbi);
4322}
4323
4324static struct ext4_sb_info *ext4_alloc_sbi(struct super_block *sb)
4325{
4326 struct ext4_sb_info *sbi;
4327
4328 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
4329 if (!sbi)
4330 return NULL;
4331
4332 sbi->s_daxdev = fs_dax_get_by_bdev(sb->s_bdev, &sbi->s_dax_part_off,
4333 NULL, NULL);
4334
4335 sbi->s_blockgroup_lock =
4336 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
4337
4338 if (!sbi->s_blockgroup_lock)
4339 goto err_out;
4340
4341 sb->s_fs_info = sbi;
4342 sbi->s_sb = sb;
4343 return sbi;
4344err_out:
4345 fs_put_dax(sbi->s_daxdev, NULL);
4346 kfree(sbi);
4347 return NULL;
4348}
4349
4350static void ext4_set_def_opts(struct super_block *sb,
4351 struct ext4_super_block *es)
4352{
4353 unsigned long def_mount_opts;
4354
4355 /* Set defaults before we parse the mount options */
4356 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
4357 set_opt(sb, INIT_INODE_TABLE);
4358 if (def_mount_opts & EXT4_DEFM_DEBUG)
4359 set_opt(sb, DEBUG);
4360 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
4361 set_opt(sb, GRPID);
4362 if (def_mount_opts & EXT4_DEFM_UID16)
4363 set_opt(sb, NO_UID32);
4364 /* xattr user namespace & acls are now defaulted on */
4365 set_opt(sb, XATTR_USER);
4366#ifdef CONFIG_EXT4_FS_POSIX_ACL
4367 set_opt(sb, POSIX_ACL);
4368#endif
4369 if (ext4_has_feature_fast_commit(sb))
4370 set_opt2(sb, JOURNAL_FAST_COMMIT);
4371 /* don't forget to enable journal_csum when metadata_csum is enabled. */
4372 if (ext4_has_metadata_csum(sb))
4373 set_opt(sb, JOURNAL_CHECKSUM);
4374
4375 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
4376 set_opt(sb, JOURNAL_DATA);
4377 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
4378 set_opt(sb, ORDERED_DATA);
4379 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
4380 set_opt(sb, WRITEBACK_DATA);
4381
4382 if (le16_to_cpu(es->s_errors) == EXT4_ERRORS_PANIC)
4383 set_opt(sb, ERRORS_PANIC);
4384 else if (le16_to_cpu(es->s_errors) == EXT4_ERRORS_CONTINUE)
4385 set_opt(sb, ERRORS_CONT);
4386 else
4387 set_opt(sb, ERRORS_RO);
4388 /* block_validity enabled by default; disable with noblock_validity */
4389 set_opt(sb, BLOCK_VALIDITY);
4390 if (def_mount_opts & EXT4_DEFM_DISCARD)
4391 set_opt(sb, DISCARD);
4392
4393 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
4394 set_opt(sb, BARRIER);
4395
4396 /*
4397 * enable delayed allocation by default
4398 * Use -o nodelalloc to turn it off
4399 */
4400 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
4401 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
4402 set_opt(sb, DELALLOC);
4403
4404 if (sb->s_blocksize <= PAGE_SIZE)
4405 set_opt(sb, DIOREAD_NOLOCK);
4406}
4407
4408static int ext4_handle_clustersize(struct super_block *sb)
4409{
4410 struct ext4_sb_info *sbi = EXT4_SB(sb);
4411 struct ext4_super_block *es = sbi->s_es;
4412 int clustersize;
4413
4414 /* Handle clustersize */
4415 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
4416 if (ext4_has_feature_bigalloc(sb)) {
4417 if (clustersize < sb->s_blocksize) {
4418 ext4_msg(sb, KERN_ERR,
4419 "cluster size (%d) smaller than "
4420 "block size (%lu)", clustersize, sb->s_blocksize);
4421 return -EINVAL;
4422 }
4423 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
4424 le32_to_cpu(es->s_log_block_size);
4425 sbi->s_clusters_per_group =
4426 le32_to_cpu(es->s_clusters_per_group);
4427 if (sbi->s_clusters_per_group > sb->s_blocksize * 8) {
4428 ext4_msg(sb, KERN_ERR,
4429 "#clusters per group too big: %lu",
4430 sbi->s_clusters_per_group);
4431 return -EINVAL;
4432 }
4433 if (sbi->s_blocks_per_group !=
4434 (sbi->s_clusters_per_group * (clustersize / sb->s_blocksize))) {
4435 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
4436 "clusters per group (%lu) inconsistent",
4437 sbi->s_blocks_per_group,
4438 sbi->s_clusters_per_group);
4439 return -EINVAL;
4440 }
4441 } else {
4442 if (clustersize != sb->s_blocksize) {
4443 ext4_msg(sb, KERN_ERR,
4444 "fragment/cluster size (%d) != "
4445 "block size (%lu)", clustersize, sb->s_blocksize);
4446 return -EINVAL;
4447 }
4448 if (sbi->s_blocks_per_group > sb->s_blocksize * 8) {
4449 ext4_msg(sb, KERN_ERR,
4450 "#blocks per group too big: %lu",
4451 sbi->s_blocks_per_group);
4452 return -EINVAL;
4453 }
4454 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
4455 sbi->s_cluster_bits = 0;
4456 }
4457 sbi->s_cluster_ratio = clustersize / sb->s_blocksize;
4458
4459 /* Do we have standard group size of clustersize * 8 blocks ? */
4460 if (sbi->s_blocks_per_group == clustersize << 3)
4461 set_opt2(sb, STD_GROUP_SIZE);
4462
4463 return 0;
4464}
4465
4466static void ext4_fast_commit_init(struct super_block *sb)
4467{
4468 struct ext4_sb_info *sbi = EXT4_SB(sb);
4469
4470 /* Initialize fast commit stuff */
4471 atomic_set(&sbi->s_fc_subtid, 0);
4472 INIT_LIST_HEAD(&sbi->s_fc_q[FC_Q_MAIN]);
4473 INIT_LIST_HEAD(&sbi->s_fc_q[FC_Q_STAGING]);
4474 INIT_LIST_HEAD(&sbi->s_fc_dentry_q[FC_Q_MAIN]);
4475 INIT_LIST_HEAD(&sbi->s_fc_dentry_q[FC_Q_STAGING]);
4476 sbi->s_fc_bytes = 0;
4477 ext4_clear_mount_flag(sb, EXT4_MF_FC_INELIGIBLE);
4478 sbi->s_fc_ineligible_tid = 0;
4479 spin_lock_init(&sbi->s_fc_lock);
4480 memset(&sbi->s_fc_stats, 0, sizeof(sbi->s_fc_stats));
4481 sbi->s_fc_replay_state.fc_regions = NULL;
4482 sbi->s_fc_replay_state.fc_regions_size = 0;
4483 sbi->s_fc_replay_state.fc_regions_used = 0;
4484 sbi->s_fc_replay_state.fc_regions_valid = 0;
4485 sbi->s_fc_replay_state.fc_modified_inodes = NULL;
4486 sbi->s_fc_replay_state.fc_modified_inodes_size = 0;
4487 sbi->s_fc_replay_state.fc_modified_inodes_used = 0;
4488}
4489
4490static int ext4_inode_info_init(struct super_block *sb,
4491 struct ext4_super_block *es)
4492{
4493 struct ext4_sb_info *sbi = EXT4_SB(sb);
4494
4495 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
4496 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
4497 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
4498 } else {
4499 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
4500 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
4501 if (sbi->s_first_ino < EXT4_GOOD_OLD_FIRST_INO) {
4502 ext4_msg(sb, KERN_ERR, "invalid first ino: %u",
4503 sbi->s_first_ino);
4504 return -EINVAL;
4505 }
4506 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
4507 (!is_power_of_2(sbi->s_inode_size)) ||
4508 (sbi->s_inode_size > sb->s_blocksize)) {
4509 ext4_msg(sb, KERN_ERR,
4510 "unsupported inode size: %d",
4511 sbi->s_inode_size);
4512 ext4_msg(sb, KERN_ERR, "blocksize: %lu", sb->s_blocksize);
4513 return -EINVAL;
4514 }
4515 /*
4516 * i_atime_extra is the last extra field available for
4517 * [acm]times in struct ext4_inode. Checking for that
4518 * field should suffice to ensure we have extra space
4519 * for all three.
4520 */
4521 if (sbi->s_inode_size >= offsetof(struct ext4_inode, i_atime_extra) +
4522 sizeof(((struct ext4_inode *)0)->i_atime_extra)) {
4523 sb->s_time_gran = 1;
4524 sb->s_time_max = EXT4_EXTRA_TIMESTAMP_MAX;
4525 } else {
4526 sb->s_time_gran = NSEC_PER_SEC;
4527 sb->s_time_max = EXT4_NON_EXTRA_TIMESTAMP_MAX;
4528 }
4529 sb->s_time_min = EXT4_TIMESTAMP_MIN;
4530 }
4531
4532 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
4533 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4534 EXT4_GOOD_OLD_INODE_SIZE;
4535 if (ext4_has_feature_extra_isize(sb)) {
4536 unsigned v, max = (sbi->s_inode_size -
4537 EXT4_GOOD_OLD_INODE_SIZE);
4538
4539 v = le16_to_cpu(es->s_want_extra_isize);
4540 if (v > max) {
4541 ext4_msg(sb, KERN_ERR,
4542 "bad s_want_extra_isize: %d", v);
4543 return -EINVAL;
4544 }
4545 if (sbi->s_want_extra_isize < v)
4546 sbi->s_want_extra_isize = v;
4547
4548 v = le16_to_cpu(es->s_min_extra_isize);
4549 if (v > max) {
4550 ext4_msg(sb, KERN_ERR,
4551 "bad s_min_extra_isize: %d", v);
4552 return -EINVAL;
4553 }
4554 if (sbi->s_want_extra_isize < v)
4555 sbi->s_want_extra_isize = v;
4556 }
4557 }
4558
4559 return 0;
4560}
4561
4562#if IS_ENABLED(CONFIG_UNICODE)
4563static int ext4_encoding_init(struct super_block *sb, struct ext4_super_block *es)
4564{
4565 const struct ext4_sb_encodings *encoding_info;
4566 struct unicode_map *encoding;
4567 __u16 encoding_flags = le16_to_cpu(es->s_encoding_flags);
4568
4569 if (!ext4_has_feature_casefold(sb) || sb->s_encoding)
4570 return 0;
4571
4572 encoding_info = ext4_sb_read_encoding(es);
4573 if (!encoding_info) {
4574 ext4_msg(sb, KERN_ERR,
4575 "Encoding requested by superblock is unknown");
4576 return -EINVAL;
4577 }
4578
4579 encoding = utf8_load(encoding_info->version);
4580 if (IS_ERR(encoding)) {
4581 ext4_msg(sb, KERN_ERR,
4582 "can't mount with superblock charset: %s-%u.%u.%u "
4583 "not supported by the kernel. flags: 0x%x.",
4584 encoding_info->name,
4585 unicode_major(encoding_info->version),
4586 unicode_minor(encoding_info->version),
4587 unicode_rev(encoding_info->version),
4588 encoding_flags);
4589 return -EINVAL;
4590 }
4591 ext4_msg(sb, KERN_INFO,"Using encoding defined by superblock: "
4592 "%s-%u.%u.%u with flags 0x%hx", encoding_info->name,
4593 unicode_major(encoding_info->version),
4594 unicode_minor(encoding_info->version),
4595 unicode_rev(encoding_info->version),
4596 encoding_flags);
4597
4598 sb->s_encoding = encoding;
4599 sb->s_encoding_flags = encoding_flags;
4600
4601 return 0;
4602}
4603#else
4604static inline int ext4_encoding_init(struct super_block *sb, struct ext4_super_block *es)
4605{
4606 return 0;
4607}
4608#endif
4609
4610static int ext4_init_metadata_csum(struct super_block *sb, struct ext4_super_block *es)
4611{
4612 struct ext4_sb_info *sbi = EXT4_SB(sb);
4613
4614 /* Warn if metadata_csum and gdt_csum are both set. */
4615 if (ext4_has_feature_metadata_csum(sb) &&
4616 ext4_has_feature_gdt_csum(sb))
4617 ext4_warning(sb, "metadata_csum and uninit_bg are "
4618 "redundant flags; please run fsck.");
4619
4620 /* Check for a known checksum algorithm */
4621 if (!ext4_verify_csum_type(sb, es)) {
4622 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
4623 "unknown checksum algorithm.");
4624 return -EINVAL;
4625 }
4626 ext4_setup_csum_trigger(sb, EXT4_JTR_ORPHAN_FILE,
4627 ext4_orphan_file_block_trigger);
4628
4629 /* Load the checksum driver */
4630 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
4631 if (IS_ERR(sbi->s_chksum_driver)) {
4632 int ret = PTR_ERR(sbi->s_chksum_driver);
4633 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
4634 sbi->s_chksum_driver = NULL;
4635 return ret;
4636 }
4637
4638 /* Check superblock checksum */
4639 if (!ext4_superblock_csum_verify(sb, es)) {
4640 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
4641 "invalid superblock checksum. Run e2fsck?");
4642 return -EFSBADCRC;
4643 }
4644
4645 /* Precompute checksum seed for all metadata */
4646 if (ext4_has_feature_csum_seed(sb))
4647 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
4648 else if (ext4_has_metadata_csum(sb) || ext4_has_feature_ea_inode(sb))
4649 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
4650 sizeof(es->s_uuid));
4651 return 0;
4652}
4653
4654static int ext4_check_feature_compatibility(struct super_block *sb,
4655 struct ext4_super_block *es,
4656 int silent)
4657{
4658 struct ext4_sb_info *sbi = EXT4_SB(sb);
4659
4660 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
4661 (ext4_has_compat_features(sb) ||
4662 ext4_has_ro_compat_features(sb) ||
4663 ext4_has_incompat_features(sb)))
4664 ext4_msg(sb, KERN_WARNING,
4665 "feature flags set on rev 0 fs, "
4666 "running e2fsck is recommended");
4667
4668 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
4669 set_opt2(sb, HURD_COMPAT);
4670 if (ext4_has_feature_64bit(sb)) {
4671 ext4_msg(sb, KERN_ERR,
4672 "The Hurd can't support 64-bit file systems");
4673 return -EINVAL;
4674 }
4675
4676 /*
4677 * ea_inode feature uses l_i_version field which is not
4678 * available in HURD_COMPAT mode.
4679 */
4680 if (ext4_has_feature_ea_inode(sb)) {
4681 ext4_msg(sb, KERN_ERR,
4682 "ea_inode feature is not supported for Hurd");
4683 return -EINVAL;
4684 }
4685 }
4686
4687 if (IS_EXT2_SB(sb)) {
4688 if (ext2_feature_set_ok(sb))
4689 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
4690 "using the ext4 subsystem");
4691 else {
4692 /*
4693 * If we're probing be silent, if this looks like
4694 * it's actually an ext[34] filesystem.
4695 */
4696 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
4697 return -EINVAL;
4698 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
4699 "to feature incompatibilities");
4700 return -EINVAL;
4701 }
4702 }
4703
4704 if (IS_EXT3_SB(sb)) {
4705 if (ext3_feature_set_ok(sb))
4706 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
4707 "using the ext4 subsystem");
4708 else {
4709 /*
4710 * If we're probing be silent, if this looks like
4711 * it's actually an ext4 filesystem.
4712 */
4713 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
4714 return -EINVAL;
4715 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
4716 "to feature incompatibilities");
4717 return -EINVAL;
4718 }
4719 }
4720
4721 /*
4722 * Check feature flags regardless of the revision level, since we
4723 * previously didn't change the revision level when setting the flags,
4724 * so there is a chance incompat flags are set on a rev 0 filesystem.
4725 */
4726 if (!ext4_feature_set_ok(sb, (sb_rdonly(sb))))
4727 return -EINVAL;
4728
4729 if (sbi->s_daxdev) {
4730 if (sb->s_blocksize == PAGE_SIZE)
4731 set_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags);
4732 else
4733 ext4_msg(sb, KERN_ERR, "unsupported blocksize for DAX\n");
4734 }
4735
4736 if (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) {
4737 if (ext4_has_feature_inline_data(sb)) {
4738 ext4_msg(sb, KERN_ERR, "Cannot use DAX on a filesystem"
4739 " that may contain inline data");
4740 return -EINVAL;
4741 }
4742 if (!test_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags)) {
4743 ext4_msg(sb, KERN_ERR,
4744 "DAX unsupported by block device.");
4745 return -EINVAL;
4746 }
4747 }
4748
4749 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
4750 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
4751 es->s_encryption_level);
4752 return -EINVAL;
4753 }
4754
4755 return 0;
4756}
4757
4758static int ext4_check_geometry(struct super_block *sb,
4759 struct ext4_super_block *es)
4760{
4761 struct ext4_sb_info *sbi = EXT4_SB(sb);
4762 __u64 blocks_count;
4763 int err;
4764
4765 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (sb->s_blocksize / 4)) {
4766 ext4_msg(sb, KERN_ERR,
4767 "Number of reserved GDT blocks insanely large: %d",
4768 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks));
4769 return -EINVAL;
4770 }
4771 /*
4772 * Test whether we have more sectors than will fit in sector_t,
4773 * and whether the max offset is addressable by the page cache.
4774 */
4775 err = generic_check_addressable(sb->s_blocksize_bits,
4776 ext4_blocks_count(es));
4777 if (err) {
4778 ext4_msg(sb, KERN_ERR, "filesystem"
4779 " too large to mount safely on this system");
4780 return err;
4781 }
4782
4783 /* check blocks count against device size */
4784 blocks_count = sb_bdev_nr_blocks(sb);
4785 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
4786 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
4787 "exceeds size of device (%llu blocks)",
4788 ext4_blocks_count(es), blocks_count);
4789 return -EINVAL;
4790 }
4791
4792 /*
4793 * It makes no sense for the first data block to be beyond the end
4794 * of the filesystem.
4795 */
4796 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
4797 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
4798 "block %u is beyond end of filesystem (%llu)",
4799 le32_to_cpu(es->s_first_data_block),
4800 ext4_blocks_count(es));
4801 return -EINVAL;
4802 }
4803 if ((es->s_first_data_block == 0) && (es->s_log_block_size == 0) &&
4804 (sbi->s_cluster_ratio == 1)) {
4805 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
4806 "block is 0 with a 1k block and cluster size");
4807 return -EINVAL;
4808 }
4809
4810 blocks_count = (ext4_blocks_count(es) -
4811 le32_to_cpu(es->s_first_data_block) +
4812 EXT4_BLOCKS_PER_GROUP(sb) - 1);
4813 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4814 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
4815 ext4_msg(sb, KERN_WARNING, "groups count too large: %llu "
4816 "(block count %llu, first data block %u, "
4817 "blocks per group %lu)", blocks_count,
4818 ext4_blocks_count(es),
4819 le32_to_cpu(es->s_first_data_block),
4820 EXT4_BLOCKS_PER_GROUP(sb));
4821 return -EINVAL;
4822 }
4823 sbi->s_groups_count = blocks_count;
4824 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
4825 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
4826 if (((u64)sbi->s_groups_count * sbi->s_inodes_per_group) !=
4827 le32_to_cpu(es->s_inodes_count)) {
4828 ext4_msg(sb, KERN_ERR, "inodes count not valid: %u vs %llu",
4829 le32_to_cpu(es->s_inodes_count),
4830 ((u64)sbi->s_groups_count * sbi->s_inodes_per_group));
4831 return -EINVAL;
4832 }
4833
4834 return 0;
4835}
4836
4837static int ext4_group_desc_init(struct super_block *sb,
4838 struct ext4_super_block *es,
4839 ext4_fsblk_t logical_sb_block,
4840 ext4_group_t *first_not_zeroed)
4841{
4842 struct ext4_sb_info *sbi = EXT4_SB(sb);
4843 unsigned int db_count;
4844 ext4_fsblk_t block;
4845 int i;
4846
4847 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
4848 EXT4_DESC_PER_BLOCK(sb);
4849 if (ext4_has_feature_meta_bg(sb)) {
4850 if (le32_to_cpu(es->s_first_meta_bg) > db_count) {
4851 ext4_msg(sb, KERN_WARNING,
4852 "first meta block group too large: %u "
4853 "(group descriptor block count %u)",
4854 le32_to_cpu(es->s_first_meta_bg), db_count);
4855 return -EINVAL;
4856 }
4857 }
4858 rcu_assign_pointer(sbi->s_group_desc,
4859 kvmalloc_array(db_count,
4860 sizeof(struct buffer_head *),
4861 GFP_KERNEL));
4862 if (sbi->s_group_desc == NULL) {
4863 ext4_msg(sb, KERN_ERR, "not enough memory");
4864 return -ENOMEM;
4865 }
4866
4867 bgl_lock_init(sbi->s_blockgroup_lock);
4868
4869 /* Pre-read the descriptors into the buffer cache */
4870 for (i = 0; i < db_count; i++) {
4871 block = descriptor_loc(sb, logical_sb_block, i);
4872 ext4_sb_breadahead_unmovable(sb, block);
4873 }
4874
4875 for (i = 0; i < db_count; i++) {
4876 struct buffer_head *bh;
4877
4878 block = descriptor_loc(sb, logical_sb_block, i);
4879 bh = ext4_sb_bread_unmovable(sb, block);
4880 if (IS_ERR(bh)) {
4881 ext4_msg(sb, KERN_ERR,
4882 "can't read group descriptor %d", i);
4883 sbi->s_gdb_count = i;
4884 return PTR_ERR(bh);
4885 }
4886 rcu_read_lock();
4887 rcu_dereference(sbi->s_group_desc)[i] = bh;
4888 rcu_read_unlock();
4889 }
4890 sbi->s_gdb_count = db_count;
4891 if (!ext4_check_descriptors(sb, logical_sb_block, first_not_zeroed)) {
4892 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
4893 return -EFSCORRUPTED;
4894 }
4895
4896 return 0;
4897}
4898
4899static int ext4_load_and_init_journal(struct super_block *sb,
4900 struct ext4_super_block *es,
4901 struct ext4_fs_context *ctx)
4902{
4903 struct ext4_sb_info *sbi = EXT4_SB(sb);
4904 int err;
4905
4906 err = ext4_load_journal(sb, es, ctx->journal_devnum);
4907 if (err)
4908 return err;
4909
4910 if (ext4_has_feature_64bit(sb) &&
4911 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4912 JBD2_FEATURE_INCOMPAT_64BIT)) {
4913 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
4914 goto out;
4915 }
4916
4917 if (!set_journal_csum_feature_set(sb)) {
4918 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
4919 "feature set");
4920 goto out;
4921 }
4922
4923 if (test_opt2(sb, JOURNAL_FAST_COMMIT) &&
4924 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4925 JBD2_FEATURE_INCOMPAT_FAST_COMMIT)) {
4926 ext4_msg(sb, KERN_ERR,
4927 "Failed to set fast commit journal feature");
4928 goto out;
4929 }
4930
4931 /* We have now updated the journal if required, so we can
4932 * validate the data journaling mode. */
4933 switch (test_opt(sb, DATA_FLAGS)) {
4934 case 0:
4935 /* No mode set, assume a default based on the journal
4936 * capabilities: ORDERED_DATA if the journal can
4937 * cope, else JOURNAL_DATA
4938 */
4939 if (jbd2_journal_check_available_features
4940 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4941 set_opt(sb, ORDERED_DATA);
4942 sbi->s_def_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
4943 } else {
4944 set_opt(sb, JOURNAL_DATA);
4945 sbi->s_def_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
4946 }
4947 break;
4948
4949 case EXT4_MOUNT_ORDERED_DATA:
4950 case EXT4_MOUNT_WRITEBACK_DATA:
4951 if (!jbd2_journal_check_available_features
4952 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4953 ext4_msg(sb, KERN_ERR, "Journal does not support "
4954 "requested data journaling mode");
4955 goto out;
4956 }
4957 break;
4958 default:
4959 break;
4960 }
4961
4962 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
4963 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4964 ext4_msg(sb, KERN_ERR, "can't mount with "
4965 "journal_async_commit in data=ordered mode");
4966 goto out;
4967 }
4968
4969 set_task_ioprio(sbi->s_journal->j_task, ctx->journal_ioprio);
4970
4971 sbi->s_journal->j_submit_inode_data_buffers =
4972 ext4_journal_submit_inode_data_buffers;
4973 sbi->s_journal->j_finish_inode_data_buffers =
4974 ext4_journal_finish_inode_data_buffers;
4975
4976 return 0;
4977
4978out:
4979 /* flush s_sb_upd_work before destroying the journal. */
4980 flush_work(&sbi->s_sb_upd_work);
4981 jbd2_journal_destroy(sbi->s_journal);
4982 sbi->s_journal = NULL;
4983 return -EINVAL;
4984}
4985
4986static int ext4_check_journal_data_mode(struct super_block *sb)
4987{
4988 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4989 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting with "
4990 "data=journal disables delayed allocation, "
4991 "dioread_nolock, O_DIRECT and fast_commit support!\n");
4992 /* can't mount with both data=journal and dioread_nolock. */
4993 clear_opt(sb, DIOREAD_NOLOCK);
4994 clear_opt2(sb, JOURNAL_FAST_COMMIT);
4995 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4996 ext4_msg(sb, KERN_ERR, "can't mount with "
4997 "both data=journal and delalloc");
4998 return -EINVAL;
4999 }
5000 if (test_opt(sb, DAX_ALWAYS)) {
5001 ext4_msg(sb, KERN_ERR, "can't mount with "
5002 "both data=journal and dax");
5003 return -EINVAL;
5004 }
5005 if (ext4_has_feature_encrypt(sb)) {
5006 ext4_msg(sb, KERN_WARNING,
5007 "encrypted files will use data=ordered "
5008 "instead of data journaling mode");
5009 }
5010 if (test_opt(sb, DELALLOC))
5011 clear_opt(sb, DELALLOC);
5012 } else {
5013 sb->s_iflags |= SB_I_CGROUPWB;
5014 }
5015
5016 return 0;
5017}
5018
5019static int ext4_load_super(struct super_block *sb, ext4_fsblk_t *lsb,
5020 int silent)
5021{
5022 struct ext4_sb_info *sbi = EXT4_SB(sb);
5023 struct ext4_super_block *es;
5024 ext4_fsblk_t logical_sb_block;
5025 unsigned long offset = 0;
5026 struct buffer_head *bh;
5027 int ret = -EINVAL;
5028 int blocksize;
5029
5030 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
5031 if (!blocksize) {
5032 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
5033 return -EINVAL;
5034 }
5035
5036 /*
5037 * The ext4 superblock will not be buffer aligned for other than 1kB
5038 * block sizes. We need to calculate the offset from buffer start.
5039 */
5040 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
5041 logical_sb_block = sbi->s_sb_block * EXT4_MIN_BLOCK_SIZE;
5042 offset = do_div(logical_sb_block, blocksize);
5043 } else {
5044 logical_sb_block = sbi->s_sb_block;
5045 }
5046
5047 bh = ext4_sb_bread_unmovable(sb, logical_sb_block);
5048 if (IS_ERR(bh)) {
5049 ext4_msg(sb, KERN_ERR, "unable to read superblock");
5050 return PTR_ERR(bh);
5051 }
5052 /*
5053 * Note: s_es must be initialized as soon as possible because
5054 * some ext4 macro-instructions depend on its value
5055 */
5056 es = (struct ext4_super_block *) (bh->b_data + offset);
5057 sbi->s_es = es;
5058 sb->s_magic = le16_to_cpu(es->s_magic);
5059 if (sb->s_magic != EXT4_SUPER_MAGIC) {
5060 if (!silent)
5061 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
5062 goto out;
5063 }
5064
5065 if (le32_to_cpu(es->s_log_block_size) >
5066 (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
5067 ext4_msg(sb, KERN_ERR,
5068 "Invalid log block size: %u",
5069 le32_to_cpu(es->s_log_block_size));
5070 goto out;
5071 }
5072 if (le32_to_cpu(es->s_log_cluster_size) >
5073 (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
5074 ext4_msg(sb, KERN_ERR,
5075 "Invalid log cluster size: %u",
5076 le32_to_cpu(es->s_log_cluster_size));
5077 goto out;
5078 }
5079
5080 blocksize = EXT4_MIN_BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
5081
5082 /*
5083 * If the default block size is not the same as the real block size,
5084 * we need to reload it.
5085 */
5086 if (sb->s_blocksize == blocksize) {
5087 *lsb = logical_sb_block;
5088 sbi->s_sbh = bh;
5089 return 0;
5090 }
5091
5092 /*
5093 * bh must be released before kill_bdev(), otherwise
5094 * it won't be freed and its page also. kill_bdev()
5095 * is called by sb_set_blocksize().
5096 */
5097 brelse(bh);
5098 /* Validate the filesystem blocksize */
5099 if (!sb_set_blocksize(sb, blocksize)) {
5100 ext4_msg(sb, KERN_ERR, "bad block size %d",
5101 blocksize);
5102 bh = NULL;
5103 goto out;
5104 }
5105
5106 logical_sb_block = sbi->s_sb_block * EXT4_MIN_BLOCK_SIZE;
5107 offset = do_div(logical_sb_block, blocksize);
5108 bh = ext4_sb_bread_unmovable(sb, logical_sb_block);
5109 if (IS_ERR(bh)) {
5110 ext4_msg(sb, KERN_ERR, "Can't read superblock on 2nd try");
5111 ret = PTR_ERR(bh);
5112 bh = NULL;
5113 goto out;
5114 }
5115 es = (struct ext4_super_block *)(bh->b_data + offset);
5116 sbi->s_es = es;
5117 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
5118 ext4_msg(sb, KERN_ERR, "Magic mismatch, very weird!");
5119 goto out;
5120 }
5121 *lsb = logical_sb_block;
5122 sbi->s_sbh = bh;
5123 return 0;
5124out:
5125 brelse(bh);
5126 return ret;
5127}
5128
5129static void ext4_hash_info_init(struct super_block *sb)
5130{
5131 struct ext4_sb_info *sbi = EXT4_SB(sb);
5132 struct ext4_super_block *es = sbi->s_es;
5133 unsigned int i;
5134
5135 for (i = 0; i < 4; i++)
5136 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
5137
5138 sbi->s_def_hash_version = es->s_def_hash_version;
5139 if (ext4_has_feature_dir_index(sb)) {
5140 i = le32_to_cpu(es->s_flags);
5141 if (i & EXT2_FLAGS_UNSIGNED_HASH)
5142 sbi->s_hash_unsigned = 3;
5143 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
5144#ifdef __CHAR_UNSIGNED__
5145 if (!sb_rdonly(sb))
5146 es->s_flags |=
5147 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
5148 sbi->s_hash_unsigned = 3;
5149#else
5150 if (!sb_rdonly(sb))
5151 es->s_flags |=
5152 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
5153#endif
5154 }
5155 }
5156}
5157
5158static int ext4_block_group_meta_init(struct super_block *sb, int silent)
5159{
5160 struct ext4_sb_info *sbi = EXT4_SB(sb);
5161 struct ext4_super_block *es = sbi->s_es;
5162 int has_huge_files;
5163
5164 has_huge_files = ext4_has_feature_huge_file(sb);
5165 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
5166 has_huge_files);
5167 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
5168
5169 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
5170 if (ext4_has_feature_64bit(sb)) {
5171 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
5172 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
5173 !is_power_of_2(sbi->s_desc_size)) {
5174 ext4_msg(sb, KERN_ERR,
5175 "unsupported descriptor size %lu",
5176 sbi->s_desc_size);
5177 return -EINVAL;
5178 }
5179 } else
5180 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
5181
5182 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
5183 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
5184
5185 sbi->s_inodes_per_block = sb->s_blocksize / EXT4_INODE_SIZE(sb);
5186 if (sbi->s_inodes_per_block == 0 || sbi->s_blocks_per_group == 0) {
5187 if (!silent)
5188 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
5189 return -EINVAL;
5190 }
5191 if (sbi->s_inodes_per_group < sbi->s_inodes_per_block ||
5192 sbi->s_inodes_per_group > sb->s_blocksize * 8) {
5193 ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n",
5194 sbi->s_inodes_per_group);
5195 return -EINVAL;
5196 }
5197 sbi->s_itb_per_group = sbi->s_inodes_per_group /
5198 sbi->s_inodes_per_block;
5199 sbi->s_desc_per_block = sb->s_blocksize / EXT4_DESC_SIZE(sb);
5200 sbi->s_mount_state = le16_to_cpu(es->s_state) & ~EXT4_FC_REPLAY;
5201 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
5202 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
5203
5204 return 0;
5205}
5206
5207static int __ext4_fill_super(struct fs_context *fc, struct super_block *sb)
5208{
5209 struct ext4_super_block *es = NULL;
5210 struct ext4_sb_info *sbi = EXT4_SB(sb);
5211 ext4_fsblk_t logical_sb_block;
5212 struct inode *root;
5213 int needs_recovery;
5214 int err;
5215 ext4_group_t first_not_zeroed;
5216 struct ext4_fs_context *ctx = fc->fs_private;
5217 int silent = fc->sb_flags & SB_SILENT;
5218
5219 /* Set defaults for the variables that will be set during parsing */
5220 if (!(ctx->spec & EXT4_SPEC_JOURNAL_IOPRIO))
5221 ctx->journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
5222
5223 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
5224 sbi->s_sectors_written_start =
5225 part_stat_read(sb->s_bdev, sectors[STAT_WRITE]);
5226
5227 err = ext4_load_super(sb, &logical_sb_block, silent);
5228 if (err)
5229 goto out_fail;
5230
5231 es = sbi->s_es;
5232 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
5233
5234 err = ext4_init_metadata_csum(sb, es);
5235 if (err)
5236 goto failed_mount;
5237
5238 ext4_set_def_opts(sb, es);
5239
5240 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
5241 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
5242 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
5243 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
5244 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
5245
5246 /*
5247 * set default s_li_wait_mult for lazyinit, for the case there is
5248 * no mount option specified.
5249 */
5250 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
5251
5252 err = ext4_inode_info_init(sb, es);
5253 if (err)
5254 goto failed_mount;
5255
5256 err = parse_apply_sb_mount_options(sb, ctx);
5257 if (err < 0)
5258 goto failed_mount;
5259
5260 sbi->s_def_mount_opt = sbi->s_mount_opt;
5261 sbi->s_def_mount_opt2 = sbi->s_mount_opt2;
5262
5263 err = ext4_check_opt_consistency(fc, sb);
5264 if (err < 0)
5265 goto failed_mount;
5266
5267 ext4_apply_options(fc, sb);
5268
5269 err = ext4_encoding_init(sb, es);
5270 if (err)
5271 goto failed_mount;
5272
5273 err = ext4_check_journal_data_mode(sb);
5274 if (err)
5275 goto failed_mount;
5276
5277 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
5278 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
5279
5280 /* i_version is always enabled now */
5281 sb->s_flags |= SB_I_VERSION;
5282
5283 err = ext4_check_feature_compatibility(sb, es, silent);
5284 if (err)
5285 goto failed_mount;
5286
5287 err = ext4_block_group_meta_init(sb, silent);
5288 if (err)
5289 goto failed_mount;
5290
5291 ext4_hash_info_init(sb);
5292
5293 err = ext4_handle_clustersize(sb);
5294 if (err)
5295 goto failed_mount;
5296
5297 err = ext4_check_geometry(sb, es);
5298 if (err)
5299 goto failed_mount;
5300
5301 timer_setup(&sbi->s_err_report, print_daily_error_info, 0);
5302 spin_lock_init(&sbi->s_error_lock);
5303 INIT_WORK(&sbi->s_sb_upd_work, update_super_work);
5304
5305 err = ext4_group_desc_init(sb, es, logical_sb_block, &first_not_zeroed);
5306 if (err)
5307 goto failed_mount3;
5308
5309 err = ext4_es_register_shrinker(sbi);
5310 if (err)
5311 goto failed_mount3;
5312
5313 sbi->s_stripe = ext4_get_stripe_size(sbi);
5314 /*
5315 * It's hard to get stripe aligned blocks if stripe is not aligned with
5316 * cluster, just disable stripe and alert user to simpfy code and avoid
5317 * stripe aligned allocation which will rarely successes.
5318 */
5319 if (sbi->s_stripe > 0 && sbi->s_cluster_ratio > 1 &&
5320 sbi->s_stripe % sbi->s_cluster_ratio != 0) {
5321 ext4_msg(sb, KERN_WARNING,
5322 "stripe (%lu) is not aligned with cluster size (%u), "
5323 "stripe is disabled",
5324 sbi->s_stripe, sbi->s_cluster_ratio);
5325 sbi->s_stripe = 0;
5326 }
5327 sbi->s_extent_max_zeroout_kb = 32;
5328
5329 /*
5330 * set up enough so that it can read an inode
5331 */
5332 sb->s_op = &ext4_sops;
5333 sb->s_export_op = &ext4_export_ops;
5334 sb->s_xattr = ext4_xattr_handlers;
5335#ifdef CONFIG_FS_ENCRYPTION
5336 sb->s_cop = &ext4_cryptops;
5337#endif
5338#ifdef CONFIG_FS_VERITY
5339 sb->s_vop = &ext4_verityops;
5340#endif
5341#ifdef CONFIG_QUOTA
5342 sb->dq_op = &ext4_quota_operations;
5343 if (ext4_has_feature_quota(sb))
5344 sb->s_qcop = &dquot_quotactl_sysfile_ops;
5345 else
5346 sb->s_qcop = &ext4_qctl_operations;
5347 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
5348#endif
5349 memcpy(&sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
5350
5351 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
5352 mutex_init(&sbi->s_orphan_lock);
5353
5354 ext4_fast_commit_init(sb);
5355
5356 sb->s_root = NULL;
5357
5358 needs_recovery = (es->s_last_orphan != 0 ||
5359 ext4_has_feature_orphan_present(sb) ||
5360 ext4_has_feature_journal_needs_recovery(sb));
5361
5362 if (ext4_has_feature_mmp(sb) && !sb_rdonly(sb)) {
5363 err = ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block));
5364 if (err)
5365 goto failed_mount3a;
5366 }
5367
5368 err = -EINVAL;
5369 /*
5370 * The first inode we look at is the journal inode. Don't try
5371 * root first: it may be modified in the journal!
5372 */
5373 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
5374 err = ext4_load_and_init_journal(sb, es, ctx);
5375 if (err)
5376 goto failed_mount3a;
5377 } else if (test_opt(sb, NOLOAD) && !sb_rdonly(sb) &&
5378 ext4_has_feature_journal_needs_recovery(sb)) {
5379 ext4_msg(sb, KERN_ERR, "required journal recovery "
5380 "suppressed and not mounted read-only");
5381 goto failed_mount3a;
5382 } else {
5383 /* Nojournal mode, all journal mount options are illegal */
5384 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
5385 ext4_msg(sb, KERN_ERR, "can't mount with "
5386 "journal_async_commit, fs mounted w/o journal");
5387 goto failed_mount3a;
5388 }
5389
5390 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) {
5391 ext4_msg(sb, KERN_ERR, "can't mount with "
5392 "journal_checksum, fs mounted w/o journal");
5393 goto failed_mount3a;
5394 }
5395 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
5396 ext4_msg(sb, KERN_ERR, "can't mount with "
5397 "commit=%lu, fs mounted w/o journal",
5398 sbi->s_commit_interval / HZ);
5399 goto failed_mount3a;
5400 }
5401 if (EXT4_MOUNT_DATA_FLAGS &
5402 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
5403 ext4_msg(sb, KERN_ERR, "can't mount with "
5404 "data=, fs mounted w/o journal");
5405 goto failed_mount3a;
5406 }
5407 sbi->s_def_mount_opt &= ~EXT4_MOUNT_JOURNAL_CHECKSUM;
5408 clear_opt(sb, JOURNAL_CHECKSUM);
5409 clear_opt(sb, DATA_FLAGS);
5410 clear_opt2(sb, JOURNAL_FAST_COMMIT);
5411 sbi->s_journal = NULL;
5412 needs_recovery = 0;
5413 }
5414
5415 if (!test_opt(sb, NO_MBCACHE)) {
5416 sbi->s_ea_block_cache = ext4_xattr_create_cache();
5417 if (!sbi->s_ea_block_cache) {
5418 ext4_msg(sb, KERN_ERR,
5419 "Failed to create ea_block_cache");
5420 err = -EINVAL;
5421 goto failed_mount_wq;
5422 }
5423
5424 if (ext4_has_feature_ea_inode(sb)) {
5425 sbi->s_ea_inode_cache = ext4_xattr_create_cache();
5426 if (!sbi->s_ea_inode_cache) {
5427 ext4_msg(sb, KERN_ERR,
5428 "Failed to create ea_inode_cache");
5429 err = -EINVAL;
5430 goto failed_mount_wq;
5431 }
5432 }
5433 }
5434
5435 /*
5436 * Get the # of file system overhead blocks from the
5437 * superblock if present.
5438 */
5439 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
5440 /* ignore the precalculated value if it is ridiculous */
5441 if (sbi->s_overhead > ext4_blocks_count(es))
5442 sbi->s_overhead = 0;
5443 /*
5444 * If the bigalloc feature is not enabled recalculating the
5445 * overhead doesn't take long, so we might as well just redo
5446 * it to make sure we are using the correct value.
5447 */
5448 if (!ext4_has_feature_bigalloc(sb))
5449 sbi->s_overhead = 0;
5450 if (sbi->s_overhead == 0) {
5451 err = ext4_calculate_overhead(sb);
5452 if (err)
5453 goto failed_mount_wq;
5454 }
5455
5456 /*
5457 * The maximum number of concurrent works can be high and
5458 * concurrency isn't really necessary. Limit it to 1.
5459 */
5460 EXT4_SB(sb)->rsv_conversion_wq =
5461 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
5462 if (!EXT4_SB(sb)->rsv_conversion_wq) {
5463 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
5464 err = -ENOMEM;
5465 goto failed_mount4;
5466 }
5467
5468 /*
5469 * The jbd2_journal_load will have done any necessary log recovery,
5470 * so we can safely mount the rest of the filesystem now.
5471 */
5472
5473 root = ext4_iget(sb, EXT4_ROOT_INO, EXT4_IGET_SPECIAL);
5474 if (IS_ERR(root)) {
5475 ext4_msg(sb, KERN_ERR, "get root inode failed");
5476 err = PTR_ERR(root);
5477 root = NULL;
5478 goto failed_mount4;
5479 }
5480 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
5481 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
5482 iput(root);
5483 err = -EFSCORRUPTED;
5484 goto failed_mount4;
5485 }
5486
5487 sb->s_root = d_make_root(root);
5488 if (!sb->s_root) {
5489 ext4_msg(sb, KERN_ERR, "get root dentry failed");
5490 err = -ENOMEM;
5491 goto failed_mount4;
5492 }
5493
5494 err = ext4_setup_super(sb, es, sb_rdonly(sb));
5495 if (err == -EROFS) {
5496 sb->s_flags |= SB_RDONLY;
5497 } else if (err)
5498 goto failed_mount4a;
5499
5500 ext4_set_resv_clusters(sb);
5501
5502 if (test_opt(sb, BLOCK_VALIDITY)) {
5503 err = ext4_setup_system_zone(sb);
5504 if (err) {
5505 ext4_msg(sb, KERN_ERR, "failed to initialize system "
5506 "zone (%d)", err);
5507 goto failed_mount4a;
5508 }
5509 }
5510 ext4_fc_replay_cleanup(sb);
5511
5512 ext4_ext_init(sb);
5513
5514 /*
5515 * Enable optimize_scan if number of groups is > threshold. This can be
5516 * turned off by passing "mb_optimize_scan=0". This can also be
5517 * turned on forcefully by passing "mb_optimize_scan=1".
5518 */
5519 if (!(ctx->spec & EXT4_SPEC_mb_optimize_scan)) {
5520 if (sbi->s_groups_count >= MB_DEFAULT_LINEAR_SCAN_THRESHOLD)
5521 set_opt2(sb, MB_OPTIMIZE_SCAN);
5522 else
5523 clear_opt2(sb, MB_OPTIMIZE_SCAN);
5524 }
5525
5526 err = ext4_mb_init(sb);
5527 if (err) {
5528 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
5529 err);
5530 goto failed_mount5;
5531 }
5532
5533 /*
5534 * We can only set up the journal commit callback once
5535 * mballoc is initialized
5536 */
5537 if (sbi->s_journal)
5538 sbi->s_journal->j_commit_callback =
5539 ext4_journal_commit_callback;
5540
5541 err = ext4_percpu_param_init(sbi);
5542 if (err)
5543 goto failed_mount6;
5544
5545 if (ext4_has_feature_flex_bg(sb))
5546 if (!ext4_fill_flex_info(sb)) {
5547 ext4_msg(sb, KERN_ERR,
5548 "unable to initialize "
5549 "flex_bg meta info!");
5550 err = -ENOMEM;
5551 goto failed_mount6;
5552 }
5553
5554 err = ext4_register_li_request(sb, first_not_zeroed);
5555 if (err)
5556 goto failed_mount6;
5557
5558 err = ext4_register_sysfs(sb);
5559 if (err)
5560 goto failed_mount7;
5561
5562 err = ext4_init_orphan_info(sb);
5563 if (err)
5564 goto failed_mount8;
5565#ifdef CONFIG_QUOTA
5566 /* Enable quota usage during mount. */
5567 if (ext4_has_feature_quota(sb) && !sb_rdonly(sb)) {
5568 err = ext4_enable_quotas(sb);
5569 if (err)
5570 goto failed_mount9;
5571 }
5572#endif /* CONFIG_QUOTA */
5573
5574 /*
5575 * Save the original bdev mapping's wb_err value which could be
5576 * used to detect the metadata async write error.
5577 */
5578 spin_lock_init(&sbi->s_bdev_wb_lock);
5579 errseq_check_and_advance(&sb->s_bdev->bd_inode->i_mapping->wb_err,
5580 &sbi->s_bdev_wb_err);
5581 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
5582 ext4_orphan_cleanup(sb, es);
5583 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
5584 /*
5585 * Update the checksum after updating free space/inode counters and
5586 * ext4_orphan_cleanup. Otherwise the superblock can have an incorrect
5587 * checksum in the buffer cache until it is written out and
5588 * e2fsprogs programs trying to open a file system immediately
5589 * after it is mounted can fail.
5590 */
5591 ext4_superblock_csum_set(sb);
5592 if (needs_recovery) {
5593 ext4_msg(sb, KERN_INFO, "recovery complete");
5594 err = ext4_mark_recovery_complete(sb, es);
5595 if (err)
5596 goto failed_mount10;
5597 }
5598
5599 if (test_opt(sb, DISCARD) && !bdev_max_discard_sectors(sb->s_bdev))
5600 ext4_msg(sb, KERN_WARNING,
5601 "mounting with \"discard\" option, but the device does not support discard");
5602
5603 if (es->s_error_count)
5604 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
5605
5606 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
5607 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
5608 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
5609 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
5610 atomic_set(&sbi->s_warning_count, 0);
5611 atomic_set(&sbi->s_msg_count, 0);
5612
5613 return 0;
5614
5615failed_mount10:
5616 ext4_quotas_off(sb, EXT4_MAXQUOTAS);
5617failed_mount9: __maybe_unused
5618 ext4_release_orphan_info(sb);
5619failed_mount8:
5620 ext4_unregister_sysfs(sb);
5621 kobject_put(&sbi->s_kobj);
5622failed_mount7:
5623 ext4_unregister_li_request(sb);
5624failed_mount6:
5625 ext4_mb_release(sb);
5626 ext4_flex_groups_free(sbi);
5627 ext4_percpu_param_destroy(sbi);
5628failed_mount5:
5629 ext4_ext_release(sb);
5630 ext4_release_system_zone(sb);
5631failed_mount4a:
5632 dput(sb->s_root);
5633 sb->s_root = NULL;
5634failed_mount4:
5635 ext4_msg(sb, KERN_ERR, "mount failed");
5636 if (EXT4_SB(sb)->rsv_conversion_wq)
5637 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
5638failed_mount_wq:
5639 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
5640 sbi->s_ea_inode_cache = NULL;
5641
5642 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
5643 sbi->s_ea_block_cache = NULL;
5644
5645 if (sbi->s_journal) {
5646 /* flush s_sb_upd_work before journal destroy. */
5647 flush_work(&sbi->s_sb_upd_work);
5648 jbd2_journal_destroy(sbi->s_journal);
5649 sbi->s_journal = NULL;
5650 }
5651failed_mount3a:
5652 ext4_es_unregister_shrinker(sbi);
5653failed_mount3:
5654 /* flush s_sb_upd_work before sbi destroy */
5655 flush_work(&sbi->s_sb_upd_work);
5656 del_timer_sync(&sbi->s_err_report);
5657 ext4_stop_mmpd(sbi);
5658 ext4_group_desc_free(sbi);
5659failed_mount:
5660 if (sbi->s_chksum_driver)
5661 crypto_free_shash(sbi->s_chksum_driver);
5662
5663#if IS_ENABLED(CONFIG_UNICODE)
5664 utf8_unload(sb->s_encoding);
5665#endif
5666
5667#ifdef CONFIG_QUOTA
5668 for (unsigned int i = 0; i < EXT4_MAXQUOTAS; i++)
5669 kfree(get_qf_name(sb, sbi, i));
5670#endif
5671 fscrypt_free_dummy_policy(&sbi->s_dummy_enc_policy);
5672 brelse(sbi->s_sbh);
5673 if (sbi->s_journal_bdev_handle) {
5674 invalidate_bdev(sbi->s_journal_bdev_handle->bdev);
5675 bdev_release(sbi->s_journal_bdev_handle);
5676 }
5677out_fail:
5678 invalidate_bdev(sb->s_bdev);
5679 sb->s_fs_info = NULL;
5680 return err;
5681}
5682
5683static int ext4_fill_super(struct super_block *sb, struct fs_context *fc)
5684{
5685 struct ext4_fs_context *ctx = fc->fs_private;
5686 struct ext4_sb_info *sbi;
5687 const char *descr;
5688 int ret;
5689
5690 sbi = ext4_alloc_sbi(sb);
5691 if (!sbi)
5692 return -ENOMEM;
5693
5694 fc->s_fs_info = sbi;
5695
5696 /* Cleanup superblock name */
5697 strreplace(sb->s_id, '/', '!');
5698
5699 sbi->s_sb_block = 1; /* Default super block location */
5700 if (ctx->spec & EXT4_SPEC_s_sb_block)
5701 sbi->s_sb_block = ctx->s_sb_block;
5702
5703 ret = __ext4_fill_super(fc, sb);
5704 if (ret < 0)
5705 goto free_sbi;
5706
5707 if (sbi->s_journal) {
5708 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
5709 descr = " journalled data mode";
5710 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
5711 descr = " ordered data mode";
5712 else
5713 descr = " writeback data mode";
5714 } else
5715 descr = "out journal";
5716
5717 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
5718 ext4_msg(sb, KERN_INFO, "mounted filesystem %pU %s with%s. "
5719 "Quota mode: %s.", &sb->s_uuid,
5720 sb_rdonly(sb) ? "ro" : "r/w", descr,
5721 ext4_quota_mode(sb));
5722
5723 /* Update the s_overhead_clusters if necessary */
5724 ext4_update_overhead(sb, false);
5725 return 0;
5726
5727free_sbi:
5728 ext4_free_sbi(sbi);
5729 fc->s_fs_info = NULL;
5730 return ret;
5731}
5732
5733static int ext4_get_tree(struct fs_context *fc)
5734{
5735 return get_tree_bdev(fc, ext4_fill_super);
5736}
5737
5738/*
5739 * Setup any per-fs journal parameters now. We'll do this both on
5740 * initial mount, once the journal has been initialised but before we've
5741 * done any recovery; and again on any subsequent remount.
5742 */
5743static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
5744{
5745 struct ext4_sb_info *sbi = EXT4_SB(sb);
5746
5747 journal->j_commit_interval = sbi->s_commit_interval;
5748 journal->j_min_batch_time = sbi->s_min_batch_time;
5749 journal->j_max_batch_time = sbi->s_max_batch_time;
5750 ext4_fc_init(sb, journal);
5751
5752 write_lock(&journal->j_state_lock);
5753 if (test_opt(sb, BARRIER))
5754 journal->j_flags |= JBD2_BARRIER;
5755 else
5756 journal->j_flags &= ~JBD2_BARRIER;
5757 if (test_opt(sb, DATA_ERR_ABORT))
5758 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
5759 else
5760 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
5761 /*
5762 * Always enable journal cycle record option, letting the journal
5763 * records log transactions continuously between each mount.
5764 */
5765 journal->j_flags |= JBD2_CYCLE_RECORD;
5766 write_unlock(&journal->j_state_lock);
5767}
5768
5769static struct inode *ext4_get_journal_inode(struct super_block *sb,
5770 unsigned int journal_inum)
5771{
5772 struct inode *journal_inode;
5773
5774 /*
5775 * Test for the existence of a valid inode on disk. Bad things
5776 * happen if we iget() an unused inode, as the subsequent iput()
5777 * will try to delete it.
5778 */
5779 journal_inode = ext4_iget(sb, journal_inum, EXT4_IGET_SPECIAL);
5780 if (IS_ERR(journal_inode)) {
5781 ext4_msg(sb, KERN_ERR, "no journal found");
5782 return ERR_CAST(journal_inode);
5783 }
5784 if (!journal_inode->i_nlink) {
5785 make_bad_inode(journal_inode);
5786 iput(journal_inode);
5787 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
5788 return ERR_PTR(-EFSCORRUPTED);
5789 }
5790 if (!S_ISREG(journal_inode->i_mode) || IS_ENCRYPTED(journal_inode)) {
5791 ext4_msg(sb, KERN_ERR, "invalid journal inode");
5792 iput(journal_inode);
5793 return ERR_PTR(-EFSCORRUPTED);
5794 }
5795
5796 ext4_debug("Journal inode found at %p: %lld bytes\n",
5797 journal_inode, journal_inode->i_size);
5798 return journal_inode;
5799}
5800
5801static int ext4_journal_bmap(journal_t *journal, sector_t *block)
5802{
5803 struct ext4_map_blocks map;
5804 int ret;
5805
5806 if (journal->j_inode == NULL)
5807 return 0;
5808
5809 map.m_lblk = *block;
5810 map.m_len = 1;
5811 ret = ext4_map_blocks(NULL, journal->j_inode, &map, 0);
5812 if (ret <= 0) {
5813 ext4_msg(journal->j_inode->i_sb, KERN_CRIT,
5814 "journal bmap failed: block %llu ret %d\n",
5815 *block, ret);
5816 jbd2_journal_abort(journal, ret ? ret : -EIO);
5817 return ret;
5818 }
5819 *block = map.m_pblk;
5820 return 0;
5821}
5822
5823static journal_t *ext4_open_inode_journal(struct super_block *sb,
5824 unsigned int journal_inum)
5825{
5826 struct inode *journal_inode;
5827 journal_t *journal;
5828
5829 journal_inode = ext4_get_journal_inode(sb, journal_inum);
5830 if (IS_ERR(journal_inode))
5831 return ERR_CAST(journal_inode);
5832
5833 journal = jbd2_journal_init_inode(journal_inode);
5834 if (IS_ERR(journal)) {
5835 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
5836 iput(journal_inode);
5837 return ERR_CAST(journal);
5838 }
5839 journal->j_private = sb;
5840 journal->j_bmap = ext4_journal_bmap;
5841 ext4_init_journal_params(sb, journal);
5842 return journal;
5843}
5844
5845static struct bdev_handle *ext4_get_journal_blkdev(struct super_block *sb,
5846 dev_t j_dev, ext4_fsblk_t *j_start,
5847 ext4_fsblk_t *j_len)
5848{
5849 struct buffer_head *bh;
5850 struct block_device *bdev;
5851 struct bdev_handle *bdev_handle;
5852 int hblock, blocksize;
5853 ext4_fsblk_t sb_block;
5854 unsigned long offset;
5855 struct ext4_super_block *es;
5856 int errno;
5857
5858 bdev_handle = bdev_open_by_dev(j_dev,
5859 BLK_OPEN_READ | BLK_OPEN_WRITE | BLK_OPEN_RESTRICT_WRITES,
5860 sb, &fs_holder_ops);
5861 if (IS_ERR(bdev_handle)) {
5862 ext4_msg(sb, KERN_ERR,
5863 "failed to open journal device unknown-block(%u,%u) %ld",
5864 MAJOR(j_dev), MINOR(j_dev), PTR_ERR(bdev_handle));
5865 return bdev_handle;
5866 }
5867
5868 bdev = bdev_handle->bdev;
5869 blocksize = sb->s_blocksize;
5870 hblock = bdev_logical_block_size(bdev);
5871 if (blocksize < hblock) {
5872 ext4_msg(sb, KERN_ERR,
5873 "blocksize too small for journal device");
5874 errno = -EINVAL;
5875 goto out_bdev;
5876 }
5877
5878 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
5879 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
5880 set_blocksize(bdev, blocksize);
5881 bh = __bread(bdev, sb_block, blocksize);
5882 if (!bh) {
5883 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
5884 "external journal");
5885 errno = -EINVAL;
5886 goto out_bdev;
5887 }
5888
5889 es = (struct ext4_super_block *) (bh->b_data + offset);
5890 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
5891 !(le32_to_cpu(es->s_feature_incompat) &
5892 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
5893 ext4_msg(sb, KERN_ERR, "external journal has bad superblock");
5894 errno = -EFSCORRUPTED;
5895 goto out_bh;
5896 }
5897
5898 if ((le32_to_cpu(es->s_feature_ro_compat) &
5899 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
5900 es->s_checksum != ext4_superblock_csum(sb, es)) {
5901 ext4_msg(sb, KERN_ERR, "external journal has corrupt superblock");
5902 errno = -EFSCORRUPTED;
5903 goto out_bh;
5904 }
5905
5906 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
5907 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
5908 errno = -EFSCORRUPTED;
5909 goto out_bh;
5910 }
5911
5912 *j_start = sb_block + 1;
5913 *j_len = ext4_blocks_count(es);
5914 brelse(bh);
5915 return bdev_handle;
5916
5917out_bh:
5918 brelse(bh);
5919out_bdev:
5920 bdev_release(bdev_handle);
5921 return ERR_PTR(errno);
5922}
5923
5924static journal_t *ext4_open_dev_journal(struct super_block *sb,
5925 dev_t j_dev)
5926{
5927 journal_t *journal;
5928 ext4_fsblk_t j_start;
5929 ext4_fsblk_t j_len;
5930 struct bdev_handle *bdev_handle;
5931 int errno = 0;
5932
5933 bdev_handle = ext4_get_journal_blkdev(sb, j_dev, &j_start, &j_len);
5934 if (IS_ERR(bdev_handle))
5935 return ERR_CAST(bdev_handle);
5936
5937 journal = jbd2_journal_init_dev(bdev_handle->bdev, sb->s_bdev, j_start,
5938 j_len, sb->s_blocksize);
5939 if (IS_ERR(journal)) {
5940 ext4_msg(sb, KERN_ERR, "failed to create device journal");
5941 errno = PTR_ERR(journal);
5942 goto out_bdev;
5943 }
5944 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
5945 ext4_msg(sb, KERN_ERR, "External journal has more than one "
5946 "user (unsupported) - %d",
5947 be32_to_cpu(journal->j_superblock->s_nr_users));
5948 errno = -EINVAL;
5949 goto out_journal;
5950 }
5951 journal->j_private = sb;
5952 EXT4_SB(sb)->s_journal_bdev_handle = bdev_handle;
5953 ext4_init_journal_params(sb, journal);
5954 return journal;
5955
5956out_journal:
5957 jbd2_journal_destroy(journal);
5958out_bdev:
5959 bdev_release(bdev_handle);
5960 return ERR_PTR(errno);
5961}
5962
5963static int ext4_load_journal(struct super_block *sb,
5964 struct ext4_super_block *es,
5965 unsigned long journal_devnum)
5966{
5967 journal_t *journal;
5968 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
5969 dev_t journal_dev;
5970 int err = 0;
5971 int really_read_only;
5972 int journal_dev_ro;
5973
5974 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5975 return -EFSCORRUPTED;
5976
5977 if (journal_devnum &&
5978 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
5979 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
5980 "numbers have changed");
5981 journal_dev = new_decode_dev(journal_devnum);
5982 } else
5983 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
5984
5985 if (journal_inum && journal_dev) {
5986 ext4_msg(sb, KERN_ERR,
5987 "filesystem has both journal inode and journal device!");
5988 return -EINVAL;
5989 }
5990
5991 if (journal_inum) {
5992 journal = ext4_open_inode_journal(sb, journal_inum);
5993 if (IS_ERR(journal))
5994 return PTR_ERR(journal);
5995 } else {
5996 journal = ext4_open_dev_journal(sb, journal_dev);
5997 if (IS_ERR(journal))
5998 return PTR_ERR(journal);
5999 }
6000
6001 journal_dev_ro = bdev_read_only(journal->j_dev);
6002 really_read_only = bdev_read_only(sb->s_bdev) | journal_dev_ro;
6003
6004 if (journal_dev_ro && !sb_rdonly(sb)) {
6005 ext4_msg(sb, KERN_ERR,
6006 "journal device read-only, try mounting with '-o ro'");
6007 err = -EROFS;
6008 goto err_out;
6009 }
6010
6011 /*
6012 * Are we loading a blank journal or performing recovery after a
6013 * crash? For recovery, we need to check in advance whether we
6014 * can get read-write access to the device.
6015 */
6016 if (ext4_has_feature_journal_needs_recovery(sb)) {
6017 if (sb_rdonly(sb)) {
6018 ext4_msg(sb, KERN_INFO, "INFO: recovery "
6019 "required on readonly filesystem");
6020 if (really_read_only) {
6021 ext4_msg(sb, KERN_ERR, "write access "
6022 "unavailable, cannot proceed "
6023 "(try mounting with noload)");
6024 err = -EROFS;
6025 goto err_out;
6026 }
6027 ext4_msg(sb, KERN_INFO, "write access will "
6028 "be enabled during recovery");
6029 }
6030 }
6031
6032 if (!(journal->j_flags & JBD2_BARRIER))
6033 ext4_msg(sb, KERN_INFO, "barriers disabled");
6034
6035 if (!ext4_has_feature_journal_needs_recovery(sb))
6036 err = jbd2_journal_wipe(journal, !really_read_only);
6037 if (!err) {
6038 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
6039 __le16 orig_state;
6040 bool changed = false;
6041
6042 if (save)
6043 memcpy(save, ((char *) es) +
6044 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
6045 err = jbd2_journal_load(journal);
6046 if (save && memcmp(((char *) es) + EXT4_S_ERR_START,
6047 save, EXT4_S_ERR_LEN)) {
6048 memcpy(((char *) es) + EXT4_S_ERR_START,
6049 save, EXT4_S_ERR_LEN);
6050 changed = true;
6051 }
6052 kfree(save);
6053 orig_state = es->s_state;
6054 es->s_state |= cpu_to_le16(EXT4_SB(sb)->s_mount_state &
6055 EXT4_ERROR_FS);
6056 if (orig_state != es->s_state)
6057 changed = true;
6058 /* Write out restored error information to the superblock */
6059 if (changed && !really_read_only) {
6060 int err2;
6061 err2 = ext4_commit_super(sb);
6062 err = err ? : err2;
6063 }
6064 }
6065
6066 if (err) {
6067 ext4_msg(sb, KERN_ERR, "error loading journal");
6068 goto err_out;
6069 }
6070
6071 EXT4_SB(sb)->s_journal = journal;
6072 err = ext4_clear_journal_err(sb, es);
6073 if (err) {
6074 EXT4_SB(sb)->s_journal = NULL;
6075 jbd2_journal_destroy(journal);
6076 return err;
6077 }
6078
6079 if (!really_read_only && journal_devnum &&
6080 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
6081 es->s_journal_dev = cpu_to_le32(journal_devnum);
6082 ext4_commit_super(sb);
6083 }
6084 if (!really_read_only && journal_inum &&
6085 journal_inum != le32_to_cpu(es->s_journal_inum)) {
6086 es->s_journal_inum = cpu_to_le32(journal_inum);
6087 ext4_commit_super(sb);
6088 }
6089
6090 return 0;
6091
6092err_out:
6093 jbd2_journal_destroy(journal);
6094 return err;
6095}
6096
6097/* Copy state of EXT4_SB(sb) into buffer for on-disk superblock */
6098static void ext4_update_super(struct super_block *sb)
6099{
6100 struct ext4_sb_info *sbi = EXT4_SB(sb);
6101 struct ext4_super_block *es = sbi->s_es;
6102 struct buffer_head *sbh = sbi->s_sbh;
6103
6104 lock_buffer(sbh);
6105 /*
6106 * If the file system is mounted read-only, don't update the
6107 * superblock write time. This avoids updating the superblock
6108 * write time when we are mounting the root file system
6109 * read/only but we need to replay the journal; at that point,
6110 * for people who are east of GMT and who make their clock
6111 * tick in localtime for Windows bug-for-bug compatibility,
6112 * the clock is set in the future, and this will cause e2fsck
6113 * to complain and force a full file system check.
6114 */
6115 if (!sb_rdonly(sb))
6116 ext4_update_tstamp(es, s_wtime);
6117 es->s_kbytes_written =
6118 cpu_to_le64(sbi->s_kbytes_written +
6119 ((part_stat_read(sb->s_bdev, sectors[STAT_WRITE]) -
6120 sbi->s_sectors_written_start) >> 1));
6121 if (percpu_counter_initialized(&sbi->s_freeclusters_counter))
6122 ext4_free_blocks_count_set(es,
6123 EXT4_C2B(sbi, percpu_counter_sum_positive(
6124 &sbi->s_freeclusters_counter)));
6125 if (percpu_counter_initialized(&sbi->s_freeinodes_counter))
6126 es->s_free_inodes_count =
6127 cpu_to_le32(percpu_counter_sum_positive(
6128 &sbi->s_freeinodes_counter));
6129 /* Copy error information to the on-disk superblock */
6130 spin_lock(&sbi->s_error_lock);
6131 if (sbi->s_add_error_count > 0) {
6132 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
6133 if (!es->s_first_error_time && !es->s_first_error_time_hi) {
6134 __ext4_update_tstamp(&es->s_first_error_time,
6135 &es->s_first_error_time_hi,
6136 sbi->s_first_error_time);
6137 strncpy(es->s_first_error_func, sbi->s_first_error_func,
6138 sizeof(es->s_first_error_func));
6139 es->s_first_error_line =
6140 cpu_to_le32(sbi->s_first_error_line);
6141 es->s_first_error_ino =
6142 cpu_to_le32(sbi->s_first_error_ino);
6143 es->s_first_error_block =
6144 cpu_to_le64(sbi->s_first_error_block);
6145 es->s_first_error_errcode =
6146 ext4_errno_to_code(sbi->s_first_error_code);
6147 }
6148 __ext4_update_tstamp(&es->s_last_error_time,
6149 &es->s_last_error_time_hi,
6150 sbi->s_last_error_time);
6151 strncpy(es->s_last_error_func, sbi->s_last_error_func,
6152 sizeof(es->s_last_error_func));
6153 es->s_last_error_line = cpu_to_le32(sbi->s_last_error_line);
6154 es->s_last_error_ino = cpu_to_le32(sbi->s_last_error_ino);
6155 es->s_last_error_block = cpu_to_le64(sbi->s_last_error_block);
6156 es->s_last_error_errcode =
6157 ext4_errno_to_code(sbi->s_last_error_code);
6158 /*
6159 * Start the daily error reporting function if it hasn't been
6160 * started already
6161 */
6162 if (!es->s_error_count)
6163 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ);
6164 le32_add_cpu(&es->s_error_count, sbi->s_add_error_count);
6165 sbi->s_add_error_count = 0;
6166 }
6167 spin_unlock(&sbi->s_error_lock);
6168
6169 ext4_superblock_csum_set(sb);
6170 unlock_buffer(sbh);
6171}
6172
6173static int ext4_commit_super(struct super_block *sb)
6174{
6175 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
6176
6177 if (!sbh)
6178 return -EINVAL;
6179 if (block_device_ejected(sb))
6180 return -ENODEV;
6181
6182 ext4_update_super(sb);
6183
6184 lock_buffer(sbh);
6185 /* Buffer got discarded which means block device got invalidated */
6186 if (!buffer_mapped(sbh)) {
6187 unlock_buffer(sbh);
6188 return -EIO;
6189 }
6190
6191 if (buffer_write_io_error(sbh) || !buffer_uptodate(sbh)) {
6192 /*
6193 * Oh, dear. A previous attempt to write the
6194 * superblock failed. This could happen because the
6195 * USB device was yanked out. Or it could happen to
6196 * be a transient write error and maybe the block will
6197 * be remapped. Nothing we can do but to retry the
6198 * write and hope for the best.
6199 */
6200 ext4_msg(sb, KERN_ERR, "previous I/O error to "
6201 "superblock detected");
6202 clear_buffer_write_io_error(sbh);
6203 set_buffer_uptodate(sbh);
6204 }
6205 get_bh(sbh);
6206 /* Clear potential dirty bit if it was journalled update */
6207 clear_buffer_dirty(sbh);
6208 sbh->b_end_io = end_buffer_write_sync;
6209 submit_bh(REQ_OP_WRITE | REQ_SYNC |
6210 (test_opt(sb, BARRIER) ? REQ_FUA : 0), sbh);
6211 wait_on_buffer(sbh);
6212 if (buffer_write_io_error(sbh)) {
6213 ext4_msg(sb, KERN_ERR, "I/O error while writing "
6214 "superblock");
6215 clear_buffer_write_io_error(sbh);
6216 set_buffer_uptodate(sbh);
6217 return -EIO;
6218 }
6219 return 0;
6220}
6221
6222/*
6223 * Have we just finished recovery? If so, and if we are mounting (or
6224 * remounting) the filesystem readonly, then we will end up with a
6225 * consistent fs on disk. Record that fact.
6226 */
6227static int ext4_mark_recovery_complete(struct super_block *sb,
6228 struct ext4_super_block *es)
6229{
6230 int err;
6231 journal_t *journal = EXT4_SB(sb)->s_journal;
6232
6233 if (!ext4_has_feature_journal(sb)) {
6234 if (journal != NULL) {
6235 ext4_error(sb, "Journal got removed while the fs was "
6236 "mounted!");
6237 return -EFSCORRUPTED;
6238 }
6239 return 0;
6240 }
6241 jbd2_journal_lock_updates(journal);
6242 err = jbd2_journal_flush(journal, 0);
6243 if (err < 0)
6244 goto out;
6245
6246 if (sb_rdonly(sb) && (ext4_has_feature_journal_needs_recovery(sb) ||
6247 ext4_has_feature_orphan_present(sb))) {
6248 if (!ext4_orphan_file_empty(sb)) {
6249 ext4_error(sb, "Orphan file not empty on read-only fs.");
6250 err = -EFSCORRUPTED;
6251 goto out;
6252 }
6253 ext4_clear_feature_journal_needs_recovery(sb);
6254 ext4_clear_feature_orphan_present(sb);
6255 ext4_commit_super(sb);
6256 }
6257out:
6258 jbd2_journal_unlock_updates(journal);
6259 return err;
6260}
6261
6262/*
6263 * If we are mounting (or read-write remounting) a filesystem whose journal
6264 * has recorded an error from a previous lifetime, move that error to the
6265 * main filesystem now.
6266 */
6267static int ext4_clear_journal_err(struct super_block *sb,
6268 struct ext4_super_block *es)
6269{
6270 journal_t *journal;
6271 int j_errno;
6272 const char *errstr;
6273
6274 if (!ext4_has_feature_journal(sb)) {
6275 ext4_error(sb, "Journal got removed while the fs was mounted!");
6276 return -EFSCORRUPTED;
6277 }
6278
6279 journal = EXT4_SB(sb)->s_journal;
6280
6281 /*
6282 * Now check for any error status which may have been recorded in the
6283 * journal by a prior ext4_error() or ext4_abort()
6284 */
6285
6286 j_errno = jbd2_journal_errno(journal);
6287 if (j_errno) {
6288 char nbuf[16];
6289
6290 errstr = ext4_decode_error(sb, j_errno, nbuf);
6291 ext4_warning(sb, "Filesystem error recorded "
6292 "from previous mount: %s", errstr);
6293
6294 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
6295 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
6296 j_errno = ext4_commit_super(sb);
6297 if (j_errno)
6298 return j_errno;
6299 ext4_warning(sb, "Marked fs in need of filesystem check.");
6300
6301 jbd2_journal_clear_err(journal);
6302 jbd2_journal_update_sb_errno(journal);
6303 }
6304 return 0;
6305}
6306
6307/*
6308 * Force the running and committing transactions to commit,
6309 * and wait on the commit.
6310 */
6311int ext4_force_commit(struct super_block *sb)
6312{
6313 return ext4_journal_force_commit(EXT4_SB(sb)->s_journal);
6314}
6315
6316static int ext4_sync_fs(struct super_block *sb, int wait)
6317{
6318 int ret = 0;
6319 tid_t target;
6320 bool needs_barrier = false;
6321 struct ext4_sb_info *sbi = EXT4_SB(sb);
6322
6323 if (unlikely(ext4_forced_shutdown(sb)))
6324 return 0;
6325
6326 trace_ext4_sync_fs(sb, wait);
6327 flush_workqueue(sbi->rsv_conversion_wq);
6328 /*
6329 * Writeback quota in non-journalled quota case - journalled quota has
6330 * no dirty dquots
6331 */
6332 dquot_writeback_dquots(sb, -1);
6333 /*
6334 * Data writeback is possible w/o journal transaction, so barrier must
6335 * being sent at the end of the function. But we can skip it if
6336 * transaction_commit will do it for us.
6337 */
6338 if (sbi->s_journal) {
6339 target = jbd2_get_latest_transaction(sbi->s_journal);
6340 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
6341 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
6342 needs_barrier = true;
6343
6344 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
6345 if (wait)
6346 ret = jbd2_log_wait_commit(sbi->s_journal,
6347 target);
6348 }
6349 } else if (wait && test_opt(sb, BARRIER))
6350 needs_barrier = true;
6351 if (needs_barrier) {
6352 int err;
6353 err = blkdev_issue_flush(sb->s_bdev);
6354 if (!ret)
6355 ret = err;
6356 }
6357
6358 return ret;
6359}
6360
6361/*
6362 * LVM calls this function before a (read-only) snapshot is created. This
6363 * gives us a chance to flush the journal completely and mark the fs clean.
6364 *
6365 * Note that only this function cannot bring a filesystem to be in a clean
6366 * state independently. It relies on upper layer to stop all data & metadata
6367 * modifications.
6368 */
6369static int ext4_freeze(struct super_block *sb)
6370{
6371 int error = 0;
6372 journal_t *journal = EXT4_SB(sb)->s_journal;
6373
6374 if (journal) {
6375 /* Now we set up the journal barrier. */
6376 jbd2_journal_lock_updates(journal);
6377
6378 /*
6379 * Don't clear the needs_recovery flag if we failed to
6380 * flush the journal.
6381 */
6382 error = jbd2_journal_flush(journal, 0);
6383 if (error < 0)
6384 goto out;
6385
6386 /* Journal blocked and flushed, clear needs_recovery flag. */
6387 ext4_clear_feature_journal_needs_recovery(sb);
6388 if (ext4_orphan_file_empty(sb))
6389 ext4_clear_feature_orphan_present(sb);
6390 }
6391
6392 error = ext4_commit_super(sb);
6393out:
6394 if (journal)
6395 /* we rely on upper layer to stop further updates */
6396 jbd2_journal_unlock_updates(journal);
6397 return error;
6398}
6399
6400/*
6401 * Called by LVM after the snapshot is done. We need to reset the RECOVER
6402 * flag here, even though the filesystem is not technically dirty yet.
6403 */
6404static int ext4_unfreeze(struct super_block *sb)
6405{
6406 if (ext4_forced_shutdown(sb))
6407 return 0;
6408
6409 if (EXT4_SB(sb)->s_journal) {
6410 /* Reset the needs_recovery flag before the fs is unlocked. */
6411 ext4_set_feature_journal_needs_recovery(sb);
6412 if (ext4_has_feature_orphan_file(sb))
6413 ext4_set_feature_orphan_present(sb);
6414 }
6415
6416 ext4_commit_super(sb);
6417 return 0;
6418}
6419
6420/*
6421 * Structure to save mount options for ext4_remount's benefit
6422 */
6423struct ext4_mount_options {
6424 unsigned long s_mount_opt;
6425 unsigned long s_mount_opt2;
6426 kuid_t s_resuid;
6427 kgid_t s_resgid;
6428 unsigned long s_commit_interval;
6429 u32 s_min_batch_time, s_max_batch_time;
6430#ifdef CONFIG_QUOTA
6431 int s_jquota_fmt;
6432 char *s_qf_names[EXT4_MAXQUOTAS];
6433#endif
6434};
6435
6436static int __ext4_remount(struct fs_context *fc, struct super_block *sb)
6437{
6438 struct ext4_fs_context *ctx = fc->fs_private;
6439 struct ext4_super_block *es;
6440 struct ext4_sb_info *sbi = EXT4_SB(sb);
6441 unsigned long old_sb_flags;
6442 struct ext4_mount_options old_opts;
6443 ext4_group_t g;
6444 int err = 0;
6445 int alloc_ctx;
6446#ifdef CONFIG_QUOTA
6447 int enable_quota = 0;
6448 int i, j;
6449 char *to_free[EXT4_MAXQUOTAS];
6450#endif
6451
6452
6453 /* Store the original options */
6454 old_sb_flags = sb->s_flags;
6455 old_opts.s_mount_opt = sbi->s_mount_opt;
6456 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
6457 old_opts.s_resuid = sbi->s_resuid;
6458 old_opts.s_resgid = sbi->s_resgid;
6459 old_opts.s_commit_interval = sbi->s_commit_interval;
6460 old_opts.s_min_batch_time = sbi->s_min_batch_time;
6461 old_opts.s_max_batch_time = sbi->s_max_batch_time;
6462#ifdef CONFIG_QUOTA
6463 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
6464 for (i = 0; i < EXT4_MAXQUOTAS; i++)
6465 if (sbi->s_qf_names[i]) {
6466 char *qf_name = get_qf_name(sb, sbi, i);
6467
6468 old_opts.s_qf_names[i] = kstrdup(qf_name, GFP_KERNEL);
6469 if (!old_opts.s_qf_names[i]) {
6470 for (j = 0; j < i; j++)
6471 kfree(old_opts.s_qf_names[j]);
6472 return -ENOMEM;
6473 }
6474 } else
6475 old_opts.s_qf_names[i] = NULL;
6476#endif
6477 if (!(ctx->spec & EXT4_SPEC_JOURNAL_IOPRIO)) {
6478 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
6479 ctx->journal_ioprio =
6480 sbi->s_journal->j_task->io_context->ioprio;
6481 else
6482 ctx->journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
6483
6484 }
6485
6486 /*
6487 * Changing the DIOREAD_NOLOCK or DELALLOC mount options may cause
6488 * two calls to ext4_should_dioread_nolock() to return inconsistent
6489 * values, triggering WARN_ON in ext4_add_complete_io(). we grab
6490 * here s_writepages_rwsem to avoid race between writepages ops and
6491 * remount.
6492 */
6493 alloc_ctx = ext4_writepages_down_write(sb);
6494 ext4_apply_options(fc, sb);
6495 ext4_writepages_up_write(sb, alloc_ctx);
6496
6497 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
6498 test_opt(sb, JOURNAL_CHECKSUM)) {
6499 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
6500 "during remount not supported; ignoring");
6501 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
6502 }
6503
6504 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
6505 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
6506 ext4_msg(sb, KERN_ERR, "can't mount with "
6507 "both data=journal and delalloc");
6508 err = -EINVAL;
6509 goto restore_opts;
6510 }
6511 if (test_opt(sb, DIOREAD_NOLOCK)) {
6512 ext4_msg(sb, KERN_ERR, "can't mount with "
6513 "both data=journal and dioread_nolock");
6514 err = -EINVAL;
6515 goto restore_opts;
6516 }
6517 } else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) {
6518 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
6519 ext4_msg(sb, KERN_ERR, "can't mount with "
6520 "journal_async_commit in data=ordered mode");
6521 err = -EINVAL;
6522 goto restore_opts;
6523 }
6524 }
6525
6526 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_NO_MBCACHE) {
6527 ext4_msg(sb, KERN_ERR, "can't enable nombcache during remount");
6528 err = -EINVAL;
6529 goto restore_opts;
6530 }
6531
6532 if (test_opt2(sb, ABORT))
6533 ext4_abort(sb, ESHUTDOWN, "Abort forced by user");
6534
6535 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
6536 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
6537
6538 es = sbi->s_es;
6539
6540 if (sbi->s_journal) {
6541 ext4_init_journal_params(sb, sbi->s_journal);
6542 set_task_ioprio(sbi->s_journal->j_task, ctx->journal_ioprio);
6543 }
6544
6545 /* Flush outstanding errors before changing fs state */
6546 flush_work(&sbi->s_sb_upd_work);
6547
6548 if ((bool)(fc->sb_flags & SB_RDONLY) != sb_rdonly(sb)) {
6549 if (ext4_forced_shutdown(sb)) {
6550 err = -EROFS;
6551 goto restore_opts;
6552 }
6553
6554 if (fc->sb_flags & SB_RDONLY) {
6555 err = sync_filesystem(sb);
6556 if (err < 0)
6557 goto restore_opts;
6558 err = dquot_suspend(sb, -1);
6559 if (err < 0)
6560 goto restore_opts;
6561
6562 /*
6563 * First of all, the unconditional stuff we have to do
6564 * to disable replay of the journal when we next remount
6565 */
6566 sb->s_flags |= SB_RDONLY;
6567
6568 /*
6569 * OK, test if we are remounting a valid rw partition
6570 * readonly, and if so set the rdonly flag and then
6571 * mark the partition as valid again.
6572 */
6573 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
6574 (sbi->s_mount_state & EXT4_VALID_FS))
6575 es->s_state = cpu_to_le16(sbi->s_mount_state);
6576
6577 if (sbi->s_journal) {
6578 /*
6579 * We let remount-ro finish even if marking fs
6580 * as clean failed...
6581 */
6582 ext4_mark_recovery_complete(sb, es);
6583 }
6584 } else {
6585 /* Make sure we can mount this feature set readwrite */
6586 if (ext4_has_feature_readonly(sb) ||
6587 !ext4_feature_set_ok(sb, 0)) {
6588 err = -EROFS;
6589 goto restore_opts;
6590 }
6591 /*
6592 * Make sure the group descriptor checksums
6593 * are sane. If they aren't, refuse to remount r/w.
6594 */
6595 for (g = 0; g < sbi->s_groups_count; g++) {
6596 struct ext4_group_desc *gdp =
6597 ext4_get_group_desc(sb, g, NULL);
6598
6599 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
6600 ext4_msg(sb, KERN_ERR,
6601 "ext4_remount: Checksum for group %u failed (%u!=%u)",
6602 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
6603 le16_to_cpu(gdp->bg_checksum));
6604 err = -EFSBADCRC;
6605 goto restore_opts;
6606 }
6607 }
6608
6609 /*
6610 * If we have an unprocessed orphan list hanging
6611 * around from a previously readonly bdev mount,
6612 * require a full umount/remount for now.
6613 */
6614 if (es->s_last_orphan || !ext4_orphan_file_empty(sb)) {
6615 ext4_msg(sb, KERN_WARNING, "Couldn't "
6616 "remount RDWR because of unprocessed "
6617 "orphan inode list. Please "
6618 "umount/remount instead");
6619 err = -EINVAL;
6620 goto restore_opts;
6621 }
6622
6623 /*
6624 * Mounting a RDONLY partition read-write, so reread
6625 * and store the current valid flag. (It may have
6626 * been changed by e2fsck since we originally mounted
6627 * the partition.)
6628 */
6629 if (sbi->s_journal) {
6630 err = ext4_clear_journal_err(sb, es);
6631 if (err)
6632 goto restore_opts;
6633 }
6634 sbi->s_mount_state = (le16_to_cpu(es->s_state) &
6635 ~EXT4_FC_REPLAY);
6636
6637 err = ext4_setup_super(sb, es, 0);
6638 if (err)
6639 goto restore_opts;
6640
6641 sb->s_flags &= ~SB_RDONLY;
6642 if (ext4_has_feature_mmp(sb)) {
6643 err = ext4_multi_mount_protect(sb,
6644 le64_to_cpu(es->s_mmp_block));
6645 if (err)
6646 goto restore_opts;
6647 }
6648#ifdef CONFIG_QUOTA
6649 enable_quota = 1;
6650#endif
6651 }
6652 }
6653
6654 /*
6655 * Handle creation of system zone data early because it can fail.
6656 * Releasing of existing data is done when we are sure remount will
6657 * succeed.
6658 */
6659 if (test_opt(sb, BLOCK_VALIDITY) && !sbi->s_system_blks) {
6660 err = ext4_setup_system_zone(sb);
6661 if (err)
6662 goto restore_opts;
6663 }
6664
6665 if (sbi->s_journal == NULL && !(old_sb_flags & SB_RDONLY)) {
6666 err = ext4_commit_super(sb);
6667 if (err)
6668 goto restore_opts;
6669 }
6670
6671#ifdef CONFIG_QUOTA
6672 if (enable_quota) {
6673 if (sb_any_quota_suspended(sb))
6674 dquot_resume(sb, -1);
6675 else if (ext4_has_feature_quota(sb)) {
6676 err = ext4_enable_quotas(sb);
6677 if (err)
6678 goto restore_opts;
6679 }
6680 }
6681 /* Release old quota file names */
6682 for (i = 0; i < EXT4_MAXQUOTAS; i++)
6683 kfree(old_opts.s_qf_names[i]);
6684#endif
6685 if (!test_opt(sb, BLOCK_VALIDITY) && sbi->s_system_blks)
6686 ext4_release_system_zone(sb);
6687
6688 /*
6689 * Reinitialize lazy itable initialization thread based on
6690 * current settings
6691 */
6692 if (sb_rdonly(sb) || !test_opt(sb, INIT_INODE_TABLE))
6693 ext4_unregister_li_request(sb);
6694 else {
6695 ext4_group_t first_not_zeroed;
6696 first_not_zeroed = ext4_has_uninit_itable(sb);
6697 ext4_register_li_request(sb, first_not_zeroed);
6698 }
6699
6700 if (!ext4_has_feature_mmp(sb) || sb_rdonly(sb))
6701 ext4_stop_mmpd(sbi);
6702
6703 return 0;
6704
6705restore_opts:
6706 /*
6707 * If there was a failing r/w to ro transition, we may need to
6708 * re-enable quota
6709 */
6710 if (sb_rdonly(sb) && !(old_sb_flags & SB_RDONLY) &&
6711 sb_any_quota_suspended(sb))
6712 dquot_resume(sb, -1);
6713
6714 alloc_ctx = ext4_writepages_down_write(sb);
6715 sb->s_flags = old_sb_flags;
6716 sbi->s_mount_opt = old_opts.s_mount_opt;
6717 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
6718 sbi->s_resuid = old_opts.s_resuid;
6719 sbi->s_resgid = old_opts.s_resgid;
6720 sbi->s_commit_interval = old_opts.s_commit_interval;
6721 sbi->s_min_batch_time = old_opts.s_min_batch_time;
6722 sbi->s_max_batch_time = old_opts.s_max_batch_time;
6723 ext4_writepages_up_write(sb, alloc_ctx);
6724
6725 if (!test_opt(sb, BLOCK_VALIDITY) && sbi->s_system_blks)
6726 ext4_release_system_zone(sb);
6727#ifdef CONFIG_QUOTA
6728 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
6729 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
6730 to_free[i] = get_qf_name(sb, sbi, i);
6731 rcu_assign_pointer(sbi->s_qf_names[i], old_opts.s_qf_names[i]);
6732 }
6733 synchronize_rcu();
6734 for (i = 0; i < EXT4_MAXQUOTAS; i++)
6735 kfree(to_free[i]);
6736#endif
6737 if (!ext4_has_feature_mmp(sb) || sb_rdonly(sb))
6738 ext4_stop_mmpd(sbi);
6739 return err;
6740}
6741
6742static int ext4_reconfigure(struct fs_context *fc)
6743{
6744 struct super_block *sb = fc->root->d_sb;
6745 int ret;
6746
6747 fc->s_fs_info = EXT4_SB(sb);
6748
6749 ret = ext4_check_opt_consistency(fc, sb);
6750 if (ret < 0)
6751 return ret;
6752
6753 ret = __ext4_remount(fc, sb);
6754 if (ret < 0)
6755 return ret;
6756
6757 ext4_msg(sb, KERN_INFO, "re-mounted %pU %s. Quota mode: %s.",
6758 &sb->s_uuid, sb_rdonly(sb) ? "ro" : "r/w",
6759 ext4_quota_mode(sb));
6760
6761 return 0;
6762}
6763
6764#ifdef CONFIG_QUOTA
6765static int ext4_statfs_project(struct super_block *sb,
6766 kprojid_t projid, struct kstatfs *buf)
6767{
6768 struct kqid qid;
6769 struct dquot *dquot;
6770 u64 limit;
6771 u64 curblock;
6772
6773 qid = make_kqid_projid(projid);
6774 dquot = dqget(sb, qid);
6775 if (IS_ERR(dquot))
6776 return PTR_ERR(dquot);
6777 spin_lock(&dquot->dq_dqb_lock);
6778
6779 limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit,
6780 dquot->dq_dqb.dqb_bhardlimit);
6781 limit >>= sb->s_blocksize_bits;
6782
6783 if (limit && buf->f_blocks > limit) {
6784 curblock = (dquot->dq_dqb.dqb_curspace +
6785 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits;
6786 buf->f_blocks = limit;
6787 buf->f_bfree = buf->f_bavail =
6788 (buf->f_blocks > curblock) ?
6789 (buf->f_blocks - curblock) : 0;
6790 }
6791
6792 limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit,
6793 dquot->dq_dqb.dqb_ihardlimit);
6794 if (limit && buf->f_files > limit) {
6795 buf->f_files = limit;
6796 buf->f_ffree =
6797 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
6798 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
6799 }
6800
6801 spin_unlock(&dquot->dq_dqb_lock);
6802 dqput(dquot);
6803 return 0;
6804}
6805#endif
6806
6807static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
6808{
6809 struct super_block *sb = dentry->d_sb;
6810 struct ext4_sb_info *sbi = EXT4_SB(sb);
6811 struct ext4_super_block *es = sbi->s_es;
6812 ext4_fsblk_t overhead = 0, resv_blocks;
6813 s64 bfree;
6814 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
6815
6816 if (!test_opt(sb, MINIX_DF))
6817 overhead = sbi->s_overhead;
6818
6819 buf->f_type = EXT4_SUPER_MAGIC;
6820 buf->f_bsize = sb->s_blocksize;
6821 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
6822 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
6823 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
6824 /* prevent underflow in case that few free space is available */
6825 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
6826 buf->f_bavail = buf->f_bfree -
6827 (ext4_r_blocks_count(es) + resv_blocks);
6828 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
6829 buf->f_bavail = 0;
6830 buf->f_files = le32_to_cpu(es->s_inodes_count);
6831 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
6832 buf->f_namelen = EXT4_NAME_LEN;
6833 buf->f_fsid = uuid_to_fsid(es->s_uuid);
6834
6835#ifdef CONFIG_QUOTA
6836 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) &&
6837 sb_has_quota_limits_enabled(sb, PRJQUOTA))
6838 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf);
6839#endif
6840 return 0;
6841}
6842
6843
6844#ifdef CONFIG_QUOTA
6845
6846/*
6847 * Helper functions so that transaction is started before we acquire dqio_sem
6848 * to keep correct lock ordering of transaction > dqio_sem
6849 */
6850static inline struct inode *dquot_to_inode(struct dquot *dquot)
6851{
6852 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
6853}
6854
6855static int ext4_write_dquot(struct dquot *dquot)
6856{
6857 int ret, err;
6858 handle_t *handle;
6859 struct inode *inode;
6860
6861 inode = dquot_to_inode(dquot);
6862 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
6863 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
6864 if (IS_ERR(handle))
6865 return PTR_ERR(handle);
6866 ret = dquot_commit(dquot);
6867 err = ext4_journal_stop(handle);
6868 if (!ret)
6869 ret = err;
6870 return ret;
6871}
6872
6873static int ext4_acquire_dquot(struct dquot *dquot)
6874{
6875 int ret, err;
6876 handle_t *handle;
6877
6878 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
6879 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
6880 if (IS_ERR(handle))
6881 return PTR_ERR(handle);
6882 ret = dquot_acquire(dquot);
6883 err = ext4_journal_stop(handle);
6884 if (!ret)
6885 ret = err;
6886 return ret;
6887}
6888
6889static int ext4_release_dquot(struct dquot *dquot)
6890{
6891 int ret, err;
6892 handle_t *handle;
6893
6894 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
6895 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
6896 if (IS_ERR(handle)) {
6897 /* Release dquot anyway to avoid endless cycle in dqput() */
6898 dquot_release(dquot);
6899 return PTR_ERR(handle);
6900 }
6901 ret = dquot_release(dquot);
6902 err = ext4_journal_stop(handle);
6903 if (!ret)
6904 ret = err;
6905 return ret;
6906}
6907
6908static int ext4_mark_dquot_dirty(struct dquot *dquot)
6909{
6910 struct super_block *sb = dquot->dq_sb;
6911
6912 if (ext4_is_quota_journalled(sb)) {
6913 dquot_mark_dquot_dirty(dquot);
6914 return ext4_write_dquot(dquot);
6915 } else {
6916 return dquot_mark_dquot_dirty(dquot);
6917 }
6918}
6919
6920static int ext4_write_info(struct super_block *sb, int type)
6921{
6922 int ret, err;
6923 handle_t *handle;
6924
6925 /* Data block + inode block */
6926 handle = ext4_journal_start_sb(sb, EXT4_HT_QUOTA, 2);
6927 if (IS_ERR(handle))
6928 return PTR_ERR(handle);
6929 ret = dquot_commit_info(sb, type);
6930 err = ext4_journal_stop(handle);
6931 if (!ret)
6932 ret = err;
6933 return ret;
6934}
6935
6936static void lockdep_set_quota_inode(struct inode *inode, int subclass)
6937{
6938 struct ext4_inode_info *ei = EXT4_I(inode);
6939
6940 /* The first argument of lockdep_set_subclass has to be
6941 * *exactly* the same as the argument to init_rwsem() --- in
6942 * this case, in init_once() --- or lockdep gets unhappy
6943 * because the name of the lock is set using the
6944 * stringification of the argument to init_rwsem().
6945 */
6946 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */
6947 lockdep_set_subclass(&ei->i_data_sem, subclass);
6948}
6949
6950/*
6951 * Standard function to be called on quota_on
6952 */
6953static int ext4_quota_on(struct super_block *sb, int type, int format_id,
6954 const struct path *path)
6955{
6956 int err;
6957
6958 if (!test_opt(sb, QUOTA))
6959 return -EINVAL;
6960
6961 /* Quotafile not on the same filesystem? */
6962 if (path->dentry->d_sb != sb)
6963 return -EXDEV;
6964
6965 /* Quota already enabled for this file? */
6966 if (IS_NOQUOTA(d_inode(path->dentry)))
6967 return -EBUSY;
6968
6969 /* Journaling quota? */
6970 if (EXT4_SB(sb)->s_qf_names[type]) {
6971 /* Quotafile not in fs root? */
6972 if (path->dentry->d_parent != sb->s_root)
6973 ext4_msg(sb, KERN_WARNING,
6974 "Quota file not on filesystem root. "
6975 "Journaled quota will not work");
6976 sb_dqopt(sb)->flags |= DQUOT_NOLIST_DIRTY;
6977 } else {
6978 /*
6979 * Clear the flag just in case mount options changed since
6980 * last time.
6981 */
6982 sb_dqopt(sb)->flags &= ~DQUOT_NOLIST_DIRTY;
6983 }
6984
6985 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
6986 err = dquot_quota_on(sb, type, format_id, path);
6987 if (!err) {
6988 struct inode *inode = d_inode(path->dentry);
6989 handle_t *handle;
6990
6991 /*
6992 * Set inode flags to prevent userspace from messing with quota
6993 * files. If this fails, we return success anyway since quotas
6994 * are already enabled and this is not a hard failure.
6995 */
6996 inode_lock(inode);
6997 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
6998 if (IS_ERR(handle))
6999 goto unlock_inode;
7000 EXT4_I(inode)->i_flags |= EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL;
7001 inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
7002 S_NOATIME | S_IMMUTABLE);
7003 err = ext4_mark_inode_dirty(handle, inode);
7004 ext4_journal_stop(handle);
7005 unlock_inode:
7006 inode_unlock(inode);
7007 if (err)
7008 dquot_quota_off(sb, type);
7009 }
7010 if (err)
7011 lockdep_set_quota_inode(path->dentry->d_inode,
7012 I_DATA_SEM_NORMAL);
7013 return err;
7014}
7015
7016static inline bool ext4_check_quota_inum(int type, unsigned long qf_inum)
7017{
7018 switch (type) {
7019 case USRQUOTA:
7020 return qf_inum == EXT4_USR_QUOTA_INO;
7021 case GRPQUOTA:
7022 return qf_inum == EXT4_GRP_QUOTA_INO;
7023 case PRJQUOTA:
7024 return qf_inum >= EXT4_GOOD_OLD_FIRST_INO;
7025 default:
7026 BUG();
7027 }
7028}
7029
7030static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
7031 unsigned int flags)
7032{
7033 int err;
7034 struct inode *qf_inode;
7035 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
7036 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
7037 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
7038 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
7039 };
7040
7041 BUG_ON(!ext4_has_feature_quota(sb));
7042
7043 if (!qf_inums[type])
7044 return -EPERM;
7045
7046 if (!ext4_check_quota_inum(type, qf_inums[type])) {
7047 ext4_error(sb, "Bad quota inum: %lu, type: %d",
7048 qf_inums[type], type);
7049 return -EUCLEAN;
7050 }
7051
7052 qf_inode = ext4_iget(sb, qf_inums[type], EXT4_IGET_SPECIAL);
7053 if (IS_ERR(qf_inode)) {
7054 ext4_error(sb, "Bad quota inode: %lu, type: %d",
7055 qf_inums[type], type);
7056 return PTR_ERR(qf_inode);
7057 }
7058
7059 /* Don't account quota for quota files to avoid recursion */
7060 qf_inode->i_flags |= S_NOQUOTA;
7061 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
7062 err = dquot_load_quota_inode(qf_inode, type, format_id, flags);
7063 if (err)
7064 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
7065 iput(qf_inode);
7066
7067 return err;
7068}
7069
7070/* Enable usage tracking for all quota types. */
7071int ext4_enable_quotas(struct super_block *sb)
7072{
7073 int type, err = 0;
7074 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
7075 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
7076 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
7077 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
7078 };
7079 bool quota_mopt[EXT4_MAXQUOTAS] = {
7080 test_opt(sb, USRQUOTA),
7081 test_opt(sb, GRPQUOTA),
7082 test_opt(sb, PRJQUOTA),
7083 };
7084
7085 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NOLIST_DIRTY;
7086 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
7087 if (qf_inums[type]) {
7088 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
7089 DQUOT_USAGE_ENABLED |
7090 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
7091 if (err) {
7092 ext4_warning(sb,
7093 "Failed to enable quota tracking "
7094 "(type=%d, err=%d, ino=%lu). "
7095 "Please run e2fsck to fix.", type,
7096 err, qf_inums[type]);
7097
7098 ext4_quotas_off(sb, type);
7099 return err;
7100 }
7101 }
7102 }
7103 return 0;
7104}
7105
7106static int ext4_quota_off(struct super_block *sb, int type)
7107{
7108 struct inode *inode = sb_dqopt(sb)->files[type];
7109 handle_t *handle;
7110 int err;
7111
7112 /* Force all delayed allocation blocks to be allocated.
7113 * Caller already holds s_umount sem */
7114 if (test_opt(sb, DELALLOC))
7115 sync_filesystem(sb);
7116
7117 if (!inode || !igrab(inode))
7118 goto out;
7119
7120 err = dquot_quota_off(sb, type);
7121 if (err || ext4_has_feature_quota(sb))
7122 goto out_put;
7123 /*
7124 * When the filesystem was remounted read-only first, we cannot cleanup
7125 * inode flags here. Bad luck but people should be using QUOTA feature
7126 * these days anyway.
7127 */
7128 if (sb_rdonly(sb))
7129 goto out_put;
7130
7131 inode_lock(inode);
7132 /*
7133 * Update modification times of quota files when userspace can
7134 * start looking at them. If we fail, we return success anyway since
7135 * this is not a hard failure and quotas are already disabled.
7136 */
7137 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
7138 if (IS_ERR(handle)) {
7139 err = PTR_ERR(handle);
7140 goto out_unlock;
7141 }
7142 EXT4_I(inode)->i_flags &= ~(EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL);
7143 inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
7144 inode_set_mtime_to_ts(inode, inode_set_ctime_current(inode));
7145 err = ext4_mark_inode_dirty(handle, inode);
7146 ext4_journal_stop(handle);
7147out_unlock:
7148 inode_unlock(inode);
7149out_put:
7150 lockdep_set_quota_inode(inode, I_DATA_SEM_NORMAL);
7151 iput(inode);
7152 return err;
7153out:
7154 return dquot_quota_off(sb, type);
7155}
7156
7157/* Read data from quotafile - avoid pagecache and such because we cannot afford
7158 * acquiring the locks... As quota files are never truncated and quota code
7159 * itself serializes the operations (and no one else should touch the files)
7160 * we don't have to be afraid of races */
7161static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
7162 size_t len, loff_t off)
7163{
7164 struct inode *inode = sb_dqopt(sb)->files[type];
7165 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
7166 int offset = off & (sb->s_blocksize - 1);
7167 int tocopy;
7168 size_t toread;
7169 struct buffer_head *bh;
7170 loff_t i_size = i_size_read(inode);
7171
7172 if (off > i_size)
7173 return 0;
7174 if (off+len > i_size)
7175 len = i_size-off;
7176 toread = len;
7177 while (toread > 0) {
7178 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread);
7179 bh = ext4_bread(NULL, inode, blk, 0);
7180 if (IS_ERR(bh))
7181 return PTR_ERR(bh);
7182 if (!bh) /* A hole? */
7183 memset(data, 0, tocopy);
7184 else
7185 memcpy(data, bh->b_data+offset, tocopy);
7186 brelse(bh);
7187 offset = 0;
7188 toread -= tocopy;
7189 data += tocopy;
7190 blk++;
7191 }
7192 return len;
7193}
7194
7195/* Write to quotafile (we know the transaction is already started and has
7196 * enough credits) */
7197static ssize_t ext4_quota_write(struct super_block *sb, int type,
7198 const char *data, size_t len, loff_t off)
7199{
7200 struct inode *inode = sb_dqopt(sb)->files[type];
7201 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
7202 int err = 0, err2 = 0, offset = off & (sb->s_blocksize - 1);
7203 int retries = 0;
7204 struct buffer_head *bh;
7205 handle_t *handle = journal_current_handle();
7206
7207 if (!handle) {
7208 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
7209 " cancelled because transaction is not started",
7210 (unsigned long long)off, (unsigned long long)len);
7211 return -EIO;
7212 }
7213 /*
7214 * Since we account only one data block in transaction credits,
7215 * then it is impossible to cross a block boundary.
7216 */
7217 if (sb->s_blocksize - offset < len) {
7218 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
7219 " cancelled because not block aligned",
7220 (unsigned long long)off, (unsigned long long)len);
7221 return -EIO;
7222 }
7223
7224 do {
7225 bh = ext4_bread(handle, inode, blk,
7226 EXT4_GET_BLOCKS_CREATE |
7227 EXT4_GET_BLOCKS_METADATA_NOFAIL);
7228 } while (PTR_ERR(bh) == -ENOSPC &&
7229 ext4_should_retry_alloc(inode->i_sb, &retries));
7230 if (IS_ERR(bh))
7231 return PTR_ERR(bh);
7232 if (!bh)
7233 goto out;
7234 BUFFER_TRACE(bh, "get write access");
7235 err = ext4_journal_get_write_access(handle, sb, bh, EXT4_JTR_NONE);
7236 if (err) {
7237 brelse(bh);
7238 return err;
7239 }
7240 lock_buffer(bh);
7241 memcpy(bh->b_data+offset, data, len);
7242 flush_dcache_page(bh->b_page);
7243 unlock_buffer(bh);
7244 err = ext4_handle_dirty_metadata(handle, NULL, bh);
7245 brelse(bh);
7246out:
7247 if (inode->i_size < off + len) {
7248 i_size_write(inode, off + len);
7249 EXT4_I(inode)->i_disksize = inode->i_size;
7250 err2 = ext4_mark_inode_dirty(handle, inode);
7251 if (unlikely(err2 && !err))
7252 err = err2;
7253 }
7254 return err ? err : len;
7255}
7256#endif
7257
7258#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
7259static inline void register_as_ext2(void)
7260{
7261 int err = register_filesystem(&ext2_fs_type);
7262 if (err)
7263 printk(KERN_WARNING
7264 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
7265}
7266
7267static inline void unregister_as_ext2(void)
7268{
7269 unregister_filesystem(&ext2_fs_type);
7270}
7271
7272static inline int ext2_feature_set_ok(struct super_block *sb)
7273{
7274 if (ext4_has_unknown_ext2_incompat_features(sb))
7275 return 0;
7276 if (sb_rdonly(sb))
7277 return 1;
7278 if (ext4_has_unknown_ext2_ro_compat_features(sb))
7279 return 0;
7280 return 1;
7281}
7282#else
7283static inline void register_as_ext2(void) { }
7284static inline void unregister_as_ext2(void) { }
7285static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
7286#endif
7287
7288static inline void register_as_ext3(void)
7289{
7290 int err = register_filesystem(&ext3_fs_type);
7291 if (err)
7292 printk(KERN_WARNING
7293 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
7294}
7295
7296static inline void unregister_as_ext3(void)
7297{
7298 unregister_filesystem(&ext3_fs_type);
7299}
7300
7301static inline int ext3_feature_set_ok(struct super_block *sb)
7302{
7303 if (ext4_has_unknown_ext3_incompat_features(sb))
7304 return 0;
7305 if (!ext4_has_feature_journal(sb))
7306 return 0;
7307 if (sb_rdonly(sb))
7308 return 1;
7309 if (ext4_has_unknown_ext3_ro_compat_features(sb))
7310 return 0;
7311 return 1;
7312}
7313
7314static void ext4_kill_sb(struct super_block *sb)
7315{
7316 struct ext4_sb_info *sbi = EXT4_SB(sb);
7317 struct bdev_handle *handle = sbi ? sbi->s_journal_bdev_handle : NULL;
7318
7319 kill_block_super(sb);
7320
7321 if (handle)
7322 bdev_release(handle);
7323}
7324
7325static struct file_system_type ext4_fs_type = {
7326 .owner = THIS_MODULE,
7327 .name = "ext4",
7328 .init_fs_context = ext4_init_fs_context,
7329 .parameters = ext4_param_specs,
7330 .kill_sb = ext4_kill_sb,
7331 .fs_flags = FS_REQUIRES_DEV | FS_ALLOW_IDMAP,
7332};
7333MODULE_ALIAS_FS("ext4");
7334
7335/* Shared across all ext4 file systems */
7336wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
7337
7338static int __init ext4_init_fs(void)
7339{
7340 int i, err;
7341
7342 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
7343 ext4_li_info = NULL;
7344
7345 /* Build-time check for flags consistency */
7346 ext4_check_flag_values();
7347
7348 for (i = 0; i < EXT4_WQ_HASH_SZ; i++)
7349 init_waitqueue_head(&ext4__ioend_wq[i]);
7350
7351 err = ext4_init_es();
7352 if (err)
7353 return err;
7354
7355 err = ext4_init_pending();
7356 if (err)
7357 goto out7;
7358
7359 err = ext4_init_post_read_processing();
7360 if (err)
7361 goto out6;
7362
7363 err = ext4_init_pageio();
7364 if (err)
7365 goto out5;
7366
7367 err = ext4_init_system_zone();
7368 if (err)
7369 goto out4;
7370
7371 err = ext4_init_sysfs();
7372 if (err)
7373 goto out3;
7374
7375 err = ext4_init_mballoc();
7376 if (err)
7377 goto out2;
7378 err = init_inodecache();
7379 if (err)
7380 goto out1;
7381
7382 err = ext4_fc_init_dentry_cache();
7383 if (err)
7384 goto out05;
7385
7386 register_as_ext3();
7387 register_as_ext2();
7388 err = register_filesystem(&ext4_fs_type);
7389 if (err)
7390 goto out;
7391
7392 return 0;
7393out:
7394 unregister_as_ext2();
7395 unregister_as_ext3();
7396 ext4_fc_destroy_dentry_cache();
7397out05:
7398 destroy_inodecache();
7399out1:
7400 ext4_exit_mballoc();
7401out2:
7402 ext4_exit_sysfs();
7403out3:
7404 ext4_exit_system_zone();
7405out4:
7406 ext4_exit_pageio();
7407out5:
7408 ext4_exit_post_read_processing();
7409out6:
7410 ext4_exit_pending();
7411out7:
7412 ext4_exit_es();
7413
7414 return err;
7415}
7416
7417static void __exit ext4_exit_fs(void)
7418{
7419 ext4_destroy_lazyinit_thread();
7420 unregister_as_ext2();
7421 unregister_as_ext3();
7422 unregister_filesystem(&ext4_fs_type);
7423 ext4_fc_destroy_dentry_cache();
7424 destroy_inodecache();
7425 ext4_exit_mballoc();
7426 ext4_exit_sysfs();
7427 ext4_exit_system_zone();
7428 ext4_exit_pageio();
7429 ext4_exit_post_read_processing();
7430 ext4_exit_es();
7431 ext4_exit_pending();
7432}
7433
7434MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
7435MODULE_DESCRIPTION("Fourth Extended Filesystem");
7436MODULE_LICENSE("GPL");
7437MODULE_SOFTDEP("pre: crc32c");
7438module_init(ext4_init_fs)
7439module_exit(ext4_exit_fs)