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_unregister_li_request(struct super_block *sb);
83static void ext4_clear_request_list(void);
84static struct inode *ext4_get_journal_inode(struct super_block *sb,
85 unsigned int journal_inum);
86static int ext4_validate_options(struct fs_context *fc);
87static int ext4_check_opt_consistency(struct fs_context *fc,
88 struct super_block *sb);
89static void ext4_apply_options(struct fs_context *fc, struct super_block *sb);
90static int ext4_parse_param(struct fs_context *fc, struct fs_parameter *param);
91static int ext4_get_tree(struct fs_context *fc);
92static int ext4_reconfigure(struct fs_context *fc);
93static void ext4_fc_free(struct fs_context *fc);
94static int ext4_init_fs_context(struct fs_context *fc);
95static void ext4_kill_sb(struct super_block *sb);
96static const struct fs_parameter_spec ext4_param_specs[];
97
98/*
99 * Lock ordering
100 *
101 * page fault path:
102 * mmap_lock -> sb_start_pagefault -> invalidate_lock (r) -> transaction start
103 * -> page lock -> i_data_sem (rw)
104 *
105 * buffered write path:
106 * sb_start_write -> i_mutex -> mmap_lock
107 * sb_start_write -> i_mutex -> transaction start -> page lock ->
108 * i_data_sem (rw)
109 *
110 * truncate:
111 * sb_start_write -> i_mutex -> invalidate_lock (w) -> i_mmap_rwsem (w) ->
112 * page lock
113 * sb_start_write -> i_mutex -> invalidate_lock (w) -> transaction start ->
114 * i_data_sem (rw)
115 *
116 * direct IO:
117 * sb_start_write -> i_mutex -> mmap_lock
118 * sb_start_write -> i_mutex -> transaction start -> i_data_sem (rw)
119 *
120 * writepages:
121 * transaction start -> page lock(s) -> i_data_sem (rw)
122 */
123
124static const struct fs_context_operations ext4_context_ops = {
125 .parse_param = ext4_parse_param,
126 .get_tree = ext4_get_tree,
127 .reconfigure = ext4_reconfigure,
128 .free = ext4_fc_free,
129};
130
131
132#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
133static struct file_system_type ext2_fs_type = {
134 .owner = THIS_MODULE,
135 .name = "ext2",
136 .init_fs_context = ext4_init_fs_context,
137 .parameters = ext4_param_specs,
138 .kill_sb = ext4_kill_sb,
139 .fs_flags = FS_REQUIRES_DEV,
140};
141MODULE_ALIAS_FS("ext2");
142MODULE_ALIAS("ext2");
143#define IS_EXT2_SB(sb) ((sb)->s_type == &ext2_fs_type)
144#else
145#define IS_EXT2_SB(sb) (0)
146#endif
147
148
149static struct file_system_type ext3_fs_type = {
150 .owner = THIS_MODULE,
151 .name = "ext3",
152 .init_fs_context = ext4_init_fs_context,
153 .parameters = ext4_param_specs,
154 .kill_sb = ext4_kill_sb,
155 .fs_flags = FS_REQUIRES_DEV,
156};
157MODULE_ALIAS_FS("ext3");
158MODULE_ALIAS("ext3");
159#define IS_EXT3_SB(sb) ((sb)->s_type == &ext3_fs_type)
160
161
162static inline void __ext4_read_bh(struct buffer_head *bh, blk_opf_t op_flags,
163 bh_end_io_t *end_io, bool simu_fail)
164{
165 if (simu_fail) {
166 clear_buffer_uptodate(bh);
167 unlock_buffer(bh);
168 return;
169 }
170
171 /*
172 * buffer's verified bit is no longer valid after reading from
173 * disk again due to write out error, clear it to make sure we
174 * recheck the buffer contents.
175 */
176 clear_buffer_verified(bh);
177
178 bh->b_end_io = end_io ? end_io : end_buffer_read_sync;
179 get_bh(bh);
180 submit_bh(REQ_OP_READ | op_flags, bh);
181}
182
183void ext4_read_bh_nowait(struct buffer_head *bh, blk_opf_t op_flags,
184 bh_end_io_t *end_io, bool simu_fail)
185{
186 BUG_ON(!buffer_locked(bh));
187
188 if (ext4_buffer_uptodate(bh)) {
189 unlock_buffer(bh);
190 return;
191 }
192 __ext4_read_bh(bh, op_flags, end_io, simu_fail);
193}
194
195int ext4_read_bh(struct buffer_head *bh, blk_opf_t op_flags,
196 bh_end_io_t *end_io, bool simu_fail)
197{
198 BUG_ON(!buffer_locked(bh));
199
200 if (ext4_buffer_uptodate(bh)) {
201 unlock_buffer(bh);
202 return 0;
203 }
204
205 __ext4_read_bh(bh, op_flags, end_io, simu_fail);
206
207 wait_on_buffer(bh);
208 if (buffer_uptodate(bh))
209 return 0;
210 return -EIO;
211}
212
213int ext4_read_bh_lock(struct buffer_head *bh, blk_opf_t op_flags, bool wait)
214{
215 lock_buffer(bh);
216 if (!wait) {
217 ext4_read_bh_nowait(bh, op_flags, NULL, simu_fail: false);
218 return 0;
219 }
220 return ext4_read_bh(bh, op_flags, NULL, simu_fail: false);
221}
222
223/*
224 * This works like __bread_gfp() except it uses ERR_PTR for error
225 * returns. Currently with sb_bread it's impossible to distinguish
226 * between ENOMEM and EIO situations (since both result in a NULL
227 * return.
228 */
229static struct buffer_head *__ext4_sb_bread_gfp(struct super_block *sb,
230 sector_t block,
231 blk_opf_t op_flags, gfp_t gfp)
232{
233 struct buffer_head *bh;
234 int ret;
235
236 bh = sb_getblk_gfp(sb, block, gfp);
237 if (bh == NULL)
238 return ERR_PTR(error: -ENOMEM);
239 if (ext4_buffer_uptodate(bh))
240 return bh;
241
242 ret = ext4_read_bh_lock(bh, REQ_META | op_flags, wait: true);
243 if (ret) {
244 put_bh(bh);
245 return ERR_PTR(error: ret);
246 }
247 return bh;
248}
249
250struct buffer_head *ext4_sb_bread(struct super_block *sb, sector_t block,
251 blk_opf_t op_flags)
252{
253 gfp_t gfp = mapping_gfp_constraint(mapping: sb->s_bdev->bd_mapping,
254 gfp_mask: ~__GFP_FS) | __GFP_MOVABLE;
255
256 return __ext4_sb_bread_gfp(sb, block, op_flags, gfp);
257}
258
259struct buffer_head *ext4_sb_bread_unmovable(struct super_block *sb,
260 sector_t block)
261{
262 gfp_t gfp = mapping_gfp_constraint(mapping: sb->s_bdev->bd_mapping,
263 gfp_mask: ~__GFP_FS);
264
265 return __ext4_sb_bread_gfp(sb, block, op_flags: 0, gfp);
266}
267
268struct buffer_head *ext4_sb_bread_nofail(struct super_block *sb,
269 sector_t block)
270{
271 gfp_t gfp = mapping_gfp_constraint(mapping: sb->s_bdev->bd_mapping,
272 gfp_mask: ~__GFP_FS) | __GFP_MOVABLE | __GFP_NOFAIL;
273
274 return __ext4_sb_bread_gfp(sb, block, op_flags: 0, gfp);
275}
276
277void ext4_sb_breadahead_unmovable(struct super_block *sb, sector_t block)
278{
279 struct buffer_head *bh = bdev_getblk(bdev: sb->s_bdev, block,
280 size: sb->s_blocksize, GFP_NOWAIT);
281
282 if (likely(bh)) {
283 if (trylock_buffer(bh))
284 ext4_read_bh_nowait(bh, REQ_RAHEAD, NULL, simu_fail: false);
285 brelse(bh);
286 }
287}
288
289static int ext4_verify_csum_type(struct super_block *sb,
290 struct ext4_super_block *es)
291{
292 if (!ext4_has_feature_metadata_csum(sb))
293 return 1;
294
295 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
296}
297
298__le32 ext4_superblock_csum(struct ext4_super_block *es)
299{
300 int offset = offsetof(struct ext4_super_block, s_checksum);
301 __u32 csum;
302
303 csum = ext4_chksum(crc: ~0, address: (char *)es, length: offset);
304
305 return cpu_to_le32(csum);
306}
307
308static int ext4_superblock_csum_verify(struct super_block *sb,
309 struct ext4_super_block *es)
310{
311 if (!ext4_has_feature_metadata_csum(sb))
312 return 1;
313
314 return es->s_checksum == ext4_superblock_csum(es);
315}
316
317void ext4_superblock_csum_set(struct super_block *sb)
318{
319 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
320
321 if (!ext4_has_feature_metadata_csum(sb))
322 return;
323
324 es->s_checksum = ext4_superblock_csum(es);
325}
326
327ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
328 struct ext4_group_desc *bg)
329{
330 return le32_to_cpu(bg->bg_block_bitmap_lo) |
331 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
332 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
333}
334
335ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
336 struct ext4_group_desc *bg)
337{
338 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
339 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
340 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
341}
342
343ext4_fsblk_t ext4_inode_table(struct super_block *sb,
344 struct ext4_group_desc *bg)
345{
346 return le32_to_cpu(bg->bg_inode_table_lo) |
347 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
348 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
349}
350
351__u32 ext4_free_group_clusters(struct super_block *sb,
352 struct ext4_group_desc *bg)
353{
354 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
355 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
356 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
357}
358
359__u32 ext4_free_inodes_count(struct super_block *sb,
360 struct ext4_group_desc *bg)
361{
362 return le16_to_cpu(READ_ONCE(bg->bg_free_inodes_count_lo)) |
363 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
364 (__u32)le16_to_cpu(READ_ONCE(bg->bg_free_inodes_count_hi)) << 16 : 0);
365}
366
367__u32 ext4_used_dirs_count(struct super_block *sb,
368 struct ext4_group_desc *bg)
369{
370 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
371 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
372 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
373}
374
375__u32 ext4_itable_unused_count(struct super_block *sb,
376 struct ext4_group_desc *bg)
377{
378 return le16_to_cpu(bg->bg_itable_unused_lo) |
379 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
380 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
381}
382
383void ext4_block_bitmap_set(struct super_block *sb,
384 struct ext4_group_desc *bg, ext4_fsblk_t blk)
385{
386 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
387 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
388 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
389}
390
391void ext4_inode_bitmap_set(struct super_block *sb,
392 struct ext4_group_desc *bg, ext4_fsblk_t blk)
393{
394 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
395 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
396 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
397}
398
399void ext4_inode_table_set(struct super_block *sb,
400 struct ext4_group_desc *bg, ext4_fsblk_t blk)
401{
402 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
403 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
404 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
405}
406
407void ext4_free_group_clusters_set(struct super_block *sb,
408 struct ext4_group_desc *bg, __u32 count)
409{
410 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
411 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
412 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
413}
414
415void ext4_free_inodes_set(struct super_block *sb,
416 struct ext4_group_desc *bg, __u32 count)
417{
418 WRITE_ONCE(bg->bg_free_inodes_count_lo, cpu_to_le16((__u16)count));
419 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
420 WRITE_ONCE(bg->bg_free_inodes_count_hi, cpu_to_le16(count >> 16));
421}
422
423void ext4_used_dirs_set(struct super_block *sb,
424 struct ext4_group_desc *bg, __u32 count)
425{
426 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
427 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
428 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
429}
430
431void ext4_itable_unused_set(struct super_block *sb,
432 struct ext4_group_desc *bg, __u32 count)
433{
434 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
435 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
436 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
437}
438
439static void __ext4_update_tstamp(__le32 *lo, __u8 *hi, time64_t now)
440{
441 now = clamp_val(now, 0, (1ull << 40) - 1);
442
443 *lo = cpu_to_le32(lower_32_bits(now));
444 *hi = upper_32_bits(now);
445}
446
447static time64_t __ext4_get_tstamp(__le32 *lo, __u8 *hi)
448{
449 return ((time64_t)(*hi) << 32) + le32_to_cpu(*lo);
450}
451#define ext4_update_tstamp(es, tstamp) \
452 __ext4_update_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi, \
453 ktime_get_real_seconds())
454#define ext4_get_tstamp(es, tstamp) \
455 __ext4_get_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi)
456
457/*
458 * The ext4_maybe_update_superblock() function checks and updates the
459 * superblock if needed.
460 *
461 * This function is designed to update the on-disk superblock only under
462 * certain conditions to prevent excessive disk writes and unnecessary
463 * waking of the disk from sleep. The superblock will be updated if:
464 * 1. More than sbi->s_sb_update_sec (def: 1 hour) has passed since the last
465 * superblock update
466 * 2. More than sbi->s_sb_update_kb (def: 16MB) kbs have been written since the
467 * last superblock update.
468 *
469 * @sb: The superblock
470 */
471static void ext4_maybe_update_superblock(struct super_block *sb)
472{
473 struct ext4_sb_info *sbi = EXT4_SB(sb);
474 struct ext4_super_block *es = sbi->s_es;
475 journal_t *journal = sbi->s_journal;
476 time64_t now;
477 __u64 last_update;
478 __u64 lifetime_write_kbytes;
479 __u64 diff_size;
480
481 if (ext4_emergency_state(sb) || sb_rdonly(sb) ||
482 !(sb->s_flags & SB_ACTIVE) || !journal ||
483 journal->j_flags & JBD2_UNMOUNT)
484 return;
485
486 now = ktime_get_real_seconds();
487 last_update = ext4_get_tstamp(es, s_wtime);
488
489 if (likely(now - last_update < sbi->s_sb_update_sec))
490 return;
491
492 lifetime_write_kbytes = sbi->s_kbytes_written +
493 ((part_stat_read(sb->s_bdev, sectors[STAT_WRITE]) -
494 sbi->s_sectors_written_start) >> 1);
495
496 /* Get the number of kilobytes not written to disk to account
497 * for statistics and compare with a multiple of 16 MB. This
498 * is used to determine when the next superblock commit should
499 * occur (i.e. not more often than once per 16MB if there was
500 * less written in an hour).
501 */
502 diff_size = lifetime_write_kbytes - le64_to_cpu(es->s_kbytes_written);
503
504 if (diff_size > sbi->s_sb_update_kb)
505 schedule_work(work: &EXT4_SB(sb)->s_sb_upd_work);
506}
507
508static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
509{
510 struct super_block *sb = journal->j_private;
511
512 BUG_ON(txn->t_state == T_FINISHED);
513
514 ext4_process_freed_data(sb, commit_tid: txn->t_tid);
515 ext4_maybe_update_superblock(sb);
516}
517
518static bool ext4_journalled_writepage_needs_redirty(struct jbd2_inode *jinode,
519 struct folio *folio)
520{
521 struct buffer_head *bh, *head;
522 struct journal_head *jh;
523
524 bh = head = folio_buffers(folio);
525 do {
526 /*
527 * We have to redirty a page in these cases:
528 * 1) If buffer is dirty, it means the page was dirty because it
529 * contains a buffer that needs checkpointing. So the dirty bit
530 * needs to be preserved so that checkpointing writes the buffer
531 * properly.
532 * 2) If buffer is not part of the committing transaction
533 * (we may have just accidentally come across this buffer because
534 * inode range tracking is not exact) or if the currently running
535 * transaction already contains this buffer as well, dirty bit
536 * needs to be preserved so that the buffer gets writeprotected
537 * properly on running transaction's commit.
538 */
539 jh = bh2jh(bh);
540 if (buffer_dirty(bh) ||
541 (jh && (jh->b_transaction != jinode->i_transaction ||
542 jh->b_next_transaction)))
543 return true;
544 } while ((bh = bh->b_this_page) != head);
545
546 return false;
547}
548
549static int ext4_journalled_submit_inode_data_buffers(struct jbd2_inode *jinode)
550{
551 struct address_space *mapping = jinode->i_vfs_inode->i_mapping;
552 struct writeback_control wbc = {
553 .sync_mode = WB_SYNC_ALL,
554 .nr_to_write = LONG_MAX,
555 .range_start = jinode->i_dirty_start,
556 .range_end = jinode->i_dirty_end,
557 };
558 struct folio *folio = NULL;
559 int error;
560
561 /*
562 * writeback_iter() already checks for dirty pages and calls
563 * folio_clear_dirty_for_io(), which we want to write protect the
564 * folios.
565 *
566 * However, we may have to redirty a folio sometimes.
567 */
568 while ((folio = writeback_iter(mapping, wbc: &wbc, folio, error: &error))) {
569 if (ext4_journalled_writepage_needs_redirty(jinode, folio))
570 folio_redirty_for_writepage(&wbc, folio);
571 folio_unlock(folio);
572 }
573
574 return error;
575}
576
577static int ext4_journal_submit_inode_data_buffers(struct jbd2_inode *jinode)
578{
579 int ret;
580
581 if (ext4_should_journal_data(inode: jinode->i_vfs_inode))
582 ret = ext4_journalled_submit_inode_data_buffers(jinode);
583 else
584 ret = ext4_normal_submit_inode_data_buffers(jinode);
585 return ret;
586}
587
588static int ext4_journal_finish_inode_data_buffers(struct jbd2_inode *jinode)
589{
590 int ret = 0;
591
592 if (!ext4_should_journal_data(inode: jinode->i_vfs_inode))
593 ret = jbd2_journal_finish_inode_data_buffers(jinode);
594
595 return ret;
596}
597
598static bool system_going_down(void)
599{
600 return system_state == SYSTEM_HALT || system_state == SYSTEM_POWER_OFF
601 || system_state == SYSTEM_RESTART;
602}
603
604struct ext4_err_translation {
605 int code;
606 int errno;
607};
608
609#define EXT4_ERR_TRANSLATE(err) { .code = EXT4_ERR_##err, .errno = err }
610
611static struct ext4_err_translation err_translation[] = {
612 EXT4_ERR_TRANSLATE(EIO),
613 EXT4_ERR_TRANSLATE(ENOMEM),
614 EXT4_ERR_TRANSLATE(EFSBADCRC),
615 EXT4_ERR_TRANSLATE(EFSCORRUPTED),
616 EXT4_ERR_TRANSLATE(ENOSPC),
617 EXT4_ERR_TRANSLATE(ENOKEY),
618 EXT4_ERR_TRANSLATE(EROFS),
619 EXT4_ERR_TRANSLATE(EFBIG),
620 EXT4_ERR_TRANSLATE(EEXIST),
621 EXT4_ERR_TRANSLATE(ERANGE),
622 EXT4_ERR_TRANSLATE(EOVERFLOW),
623 EXT4_ERR_TRANSLATE(EBUSY),
624 EXT4_ERR_TRANSLATE(ENOTDIR),
625 EXT4_ERR_TRANSLATE(ENOTEMPTY),
626 EXT4_ERR_TRANSLATE(ESHUTDOWN),
627 EXT4_ERR_TRANSLATE(EFAULT),
628};
629
630static int ext4_errno_to_code(int errno)
631{
632 int i;
633
634 for (i = 0; i < ARRAY_SIZE(err_translation); i++)
635 if (err_translation[i].errno == errno)
636 return err_translation[i].code;
637 return EXT4_ERR_UNKNOWN;
638}
639
640static void save_error_info(struct super_block *sb, int error,
641 __u32 ino, __u64 block,
642 const char *func, unsigned int line)
643{
644 struct ext4_sb_info *sbi = EXT4_SB(sb);
645
646 /* We default to EFSCORRUPTED error... */
647 if (error == 0)
648 error = EFSCORRUPTED;
649
650 spin_lock(lock: &sbi->s_error_lock);
651 sbi->s_add_error_count++;
652 sbi->s_last_error_code = error;
653 sbi->s_last_error_line = line;
654 sbi->s_last_error_ino = ino;
655 sbi->s_last_error_block = block;
656 sbi->s_last_error_func = func;
657 sbi->s_last_error_time = ktime_get_real_seconds();
658 if (!sbi->s_first_error_time) {
659 sbi->s_first_error_code = error;
660 sbi->s_first_error_line = line;
661 sbi->s_first_error_ino = ino;
662 sbi->s_first_error_block = block;
663 sbi->s_first_error_func = func;
664 sbi->s_first_error_time = sbi->s_last_error_time;
665 }
666 spin_unlock(lock: &sbi->s_error_lock);
667}
668
669/* Deal with the reporting of failure conditions on a filesystem such as
670 * inconsistencies detected or read IO failures.
671 *
672 * On ext2, we can store the error state of the filesystem in the
673 * superblock. That is not possible on ext4, because we may have other
674 * write ordering constraints on the superblock which prevent us from
675 * writing it out straight away; and given that the journal is about to
676 * be aborted, we can't rely on the current, or future, transactions to
677 * write out the superblock safely.
678 *
679 * We'll just use the jbd2_journal_abort() error code to record an error in
680 * the journal instead. On recovery, the journal will complain about
681 * that error until we've noted it down and cleared it.
682 *
683 * If force_ro is set, we unconditionally force the filesystem into an
684 * ABORT|READONLY state, unless the error response on the fs has been set to
685 * panic in which case we take the easy way out and panic immediately. This is
686 * used to deal with unrecoverable failures such as journal IO errors or ENOMEM
687 * at a critical moment in log management.
688 */
689static void ext4_handle_error(struct super_block *sb, bool force_ro, int error,
690 __u32 ino, __u64 block,
691 const char *func, unsigned int line)
692{
693 journal_t *journal = EXT4_SB(sb)->s_journal;
694 bool continue_fs = !force_ro && test_opt(sb, ERRORS_CONT);
695
696 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
697 if (test_opt(sb, WARN_ON_ERROR))
698 WARN_ON_ONCE(1);
699
700 if (!continue_fs && !ext4_emergency_ro(sb) && journal)
701 jbd2_journal_abort(journal, -EIO);
702
703 if (!bdev_read_only(bdev: sb->s_bdev)) {
704 save_error_info(sb, error, ino, block, func, line);
705 /*
706 * In case the fs should keep running, we need to writeout
707 * superblock through the journal. Due to lock ordering
708 * constraints, it may not be safe to do it right here so we
709 * defer superblock flushing to a workqueue. We just need to be
710 * careful when the journal is already shutting down. If we get
711 * here in that case, just update the sb directly as the last
712 * transaction won't commit anyway.
713 */
714 if (continue_fs && journal &&
715 !ext4_test_mount_flag(sb, bit: EXT4_MF_JOURNAL_DESTROY))
716 schedule_work(work: &EXT4_SB(sb)->s_sb_upd_work);
717 else
718 ext4_commit_super(sb);
719 }
720
721 /*
722 * We force ERRORS_RO behavior when system is rebooting. Otherwise we
723 * could panic during 'reboot -f' as the underlying device got already
724 * disabled.
725 */
726 if (test_opt(sb, ERRORS_PANIC) && !system_going_down()) {
727 panic(fmt: "EXT4-fs (device %s): panic forced after error\n",
728 sb->s_id);
729 }
730
731 if (ext4_emergency_ro(sb) || continue_fs)
732 return;
733
734 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
735 /*
736 * We don't set SB_RDONLY because that requires sb->s_umount
737 * semaphore and setting it without proper remount procedure is
738 * confusing code such as freeze_super() leading to deadlocks
739 * and other problems.
740 */
741 set_bit(nr: EXT4_FLAGS_EMERGENCY_RO, addr: &EXT4_SB(sb)->s_ext4_flags);
742}
743
744static void update_super_work(struct work_struct *work)
745{
746 struct ext4_sb_info *sbi = container_of(work, struct ext4_sb_info,
747 s_sb_upd_work);
748 journal_t *journal = sbi->s_journal;
749 handle_t *handle;
750
751 /*
752 * If the journal is still running, we have to write out superblock
753 * through the journal to avoid collisions of other journalled sb
754 * updates.
755 *
756 * We use directly jbd2 functions here to avoid recursing back into
757 * ext4 error handling code during handling of previous errors.
758 */
759 if (!ext4_emergency_state(sb: sbi->s_sb) &&
760 !sb_rdonly(sb: sbi->s_sb) && journal) {
761 struct buffer_head *sbh = sbi->s_sbh;
762 bool call_notify_err = false;
763
764 handle = jbd2_journal_start(journal, nblocks: 1);
765 if (IS_ERR(ptr: handle))
766 goto write_directly;
767 if (jbd2_journal_get_write_access(handle, sbh)) {
768 jbd2_journal_stop(handle);
769 goto write_directly;
770 }
771
772 if (sbi->s_add_error_count > 0)
773 call_notify_err = true;
774
775 ext4_update_super(sb: sbi->s_sb);
776 if (buffer_write_io_error(bh: sbh) || !buffer_uptodate(bh: sbh)) {
777 ext4_msg(sbi->s_sb, KERN_ERR, "previous I/O error to "
778 "superblock detected");
779 clear_buffer_write_io_error(bh: sbh);
780 set_buffer_uptodate(sbh);
781 }
782
783 if (jbd2_journal_dirty_metadata(handle, sbh)) {
784 jbd2_journal_stop(handle);
785 goto write_directly;
786 }
787 jbd2_journal_stop(handle);
788
789 if (call_notify_err)
790 ext4_notify_error_sysfs(sbi);
791
792 return;
793 }
794write_directly:
795 /*
796 * Write through journal failed. Write sb directly to get error info
797 * out and hope for the best.
798 */
799 ext4_commit_super(sb: sbi->s_sb);
800 ext4_notify_error_sysfs(sbi);
801}
802
803#define ext4_error_ratelimit(sb) \
804 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
805 "EXT4-fs error")
806
807void __ext4_error(struct super_block *sb, const char *function,
808 unsigned int line, bool force_ro, int error, __u64 block,
809 const char *fmt, ...)
810{
811 struct va_format vaf;
812 va_list args;
813
814 if (unlikely(ext4_emergency_state(sb)))
815 return;
816
817 trace_ext4_error(sb, function, line);
818 if (ext4_error_ratelimit(sb)) {
819 va_start(args, fmt);
820 vaf.fmt = fmt;
821 vaf.va = &args;
822 printk(KERN_CRIT
823 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
824 sb->s_id, function, line, current->comm, &vaf);
825 va_end(args);
826 }
827 fsnotify_sb_error(sb, NULL, error: error ? error : EFSCORRUPTED);
828
829 ext4_handle_error(sb, force_ro, error, ino: 0, block, func: function, line);
830}
831
832void __ext4_error_inode(struct inode *inode, const char *function,
833 unsigned int line, ext4_fsblk_t block, int error,
834 const char *fmt, ...)
835{
836 va_list args;
837 struct va_format vaf;
838
839 if (unlikely(ext4_emergency_state(inode->i_sb)))
840 return;
841
842 trace_ext4_error(sb: inode->i_sb, function, line);
843 if (ext4_error_ratelimit(inode->i_sb)) {
844 va_start(args, fmt);
845 vaf.fmt = fmt;
846 vaf.va = &args;
847 if (block)
848 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
849 "inode #%lu: block %llu: comm %s: %pV\n",
850 inode->i_sb->s_id, function, line, inode->i_ino,
851 block, current->comm, &vaf);
852 else
853 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
854 "inode #%lu: comm %s: %pV\n",
855 inode->i_sb->s_id, function, line, inode->i_ino,
856 current->comm, &vaf);
857 va_end(args);
858 }
859 fsnotify_sb_error(sb: inode->i_sb, inode, error: error ? error : EFSCORRUPTED);
860
861 ext4_handle_error(sb: inode->i_sb, force_ro: false, error, ino: inode->i_ino, block,
862 func: function, line);
863}
864
865void __ext4_error_file(struct file *file, const char *function,
866 unsigned int line, ext4_fsblk_t block,
867 const char *fmt, ...)
868{
869 va_list args;
870 struct va_format vaf;
871 struct inode *inode = file_inode(f: file);
872 char pathname[80], *path;
873
874 if (unlikely(ext4_emergency_state(inode->i_sb)))
875 return;
876
877 trace_ext4_error(sb: inode->i_sb, function, line);
878 if (ext4_error_ratelimit(inode->i_sb)) {
879 path = file_path(file, pathname, sizeof(pathname));
880 if (IS_ERR(ptr: path))
881 path = "(unknown)";
882 va_start(args, fmt);
883 vaf.fmt = fmt;
884 vaf.va = &args;
885 if (block)
886 printk(KERN_CRIT
887 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
888 "block %llu: comm %s: path %s: %pV\n",
889 inode->i_sb->s_id, function, line, inode->i_ino,
890 block, current->comm, path, &vaf);
891 else
892 printk(KERN_CRIT
893 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
894 "comm %s: path %s: %pV\n",
895 inode->i_sb->s_id, function, line, inode->i_ino,
896 current->comm, path, &vaf);
897 va_end(args);
898 }
899 fsnotify_sb_error(sb: inode->i_sb, inode, EFSCORRUPTED);
900
901 ext4_handle_error(sb: inode->i_sb, force_ro: false, EFSCORRUPTED, ino: inode->i_ino, block,
902 func: function, line);
903}
904
905const char *ext4_decode_error(struct super_block *sb, int errno,
906 char nbuf[16])
907{
908 char *errstr = NULL;
909
910 switch (errno) {
911 case -EFSCORRUPTED:
912 errstr = "Corrupt filesystem";
913 break;
914 case -EFSBADCRC:
915 errstr = "Filesystem failed CRC";
916 break;
917 case -EIO:
918 errstr = "IO failure";
919 break;
920 case -ENOMEM:
921 errstr = "Out of memory";
922 break;
923 case -EROFS:
924 if (!sb || (EXT4_SB(sb)->s_journal &&
925 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
926 errstr = "Journal has aborted";
927 else
928 errstr = "Readonly filesystem";
929 break;
930 default:
931 /* If the caller passed in an extra buffer for unknown
932 * errors, textualise them now. Else we just return
933 * NULL. */
934 if (nbuf) {
935 /* Check for truncated error codes... */
936 if (snprintf(buf: nbuf, size: 16, fmt: "error %d", -errno) >= 0)
937 errstr = nbuf;
938 }
939 break;
940 }
941
942 return errstr;
943}
944
945/* __ext4_std_error decodes expected errors from journaling functions
946 * automatically and invokes the appropriate error response. */
947
948void __ext4_std_error(struct super_block *sb, const char *function,
949 unsigned int line, int errno)
950{
951 char nbuf[16];
952 const char *errstr;
953
954 if (unlikely(ext4_emergency_state(sb)))
955 return;
956
957 /* Special case: if the error is EROFS, and we're not already
958 * inside a transaction, then there's really no point in logging
959 * an error. */
960 if (errno == -EROFS && journal_current_handle() == NULL && sb_rdonly(sb))
961 return;
962
963 if (ext4_error_ratelimit(sb)) {
964 errstr = ext4_decode_error(sb, errno, nbuf);
965 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
966 sb->s_id, function, line, errstr);
967 }
968 fsnotify_sb_error(sb, NULL, error: errno ? errno : EFSCORRUPTED);
969
970 ext4_handle_error(sb, force_ro: false, error: -errno, ino: 0, block: 0, func: function, line);
971}
972
973void __ext4_msg(struct super_block *sb,
974 const char *prefix, const char *fmt, ...)
975{
976 struct va_format vaf;
977 va_list args;
978
979 if (sb) {
980 atomic_inc(v: &EXT4_SB(sb)->s_msg_count);
981 if (!___ratelimit(rs: &(EXT4_SB(sb)->s_msg_ratelimit_state),
982 func: "EXT4-fs"))
983 return;
984 }
985
986 va_start(args, fmt);
987 vaf.fmt = fmt;
988 vaf.va = &args;
989 if (sb)
990 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
991 else
992 printk("%sEXT4-fs: %pV\n", prefix, &vaf);
993 va_end(args);
994}
995
996static int ext4_warning_ratelimit(struct super_block *sb)
997{
998 atomic_inc(v: &EXT4_SB(sb)->s_warning_count);
999 return ___ratelimit(rs: &(EXT4_SB(sb)->s_warning_ratelimit_state),
1000 func: "EXT4-fs warning");
1001}
1002
1003void __ext4_warning(struct super_block *sb, const char *function,
1004 unsigned int line, const char *fmt, ...)
1005{
1006 struct va_format vaf;
1007 va_list args;
1008
1009 if (!ext4_warning_ratelimit(sb))
1010 return;
1011
1012 va_start(args, fmt);
1013 vaf.fmt = fmt;
1014 vaf.va = &args;
1015 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
1016 sb->s_id, function, line, &vaf);
1017 va_end(args);
1018}
1019
1020void __ext4_warning_inode(const struct inode *inode, const char *function,
1021 unsigned int line, const char *fmt, ...)
1022{
1023 struct va_format vaf;
1024 va_list args;
1025
1026 if (!ext4_warning_ratelimit(sb: inode->i_sb))
1027 return;
1028
1029 va_start(args, fmt);
1030 vaf.fmt = fmt;
1031 vaf.va = &args;
1032 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
1033 "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
1034 function, line, inode->i_ino, current->comm, &vaf);
1035 va_end(args);
1036}
1037
1038void __ext4_grp_locked_error(const char *function, unsigned int line,
1039 struct super_block *sb, ext4_group_t grp,
1040 unsigned long ino, ext4_fsblk_t block,
1041 const char *fmt, ...)
1042__releases(bitlock)
1043__acquires(bitlock)
1044{
1045 struct va_format vaf;
1046 va_list args;
1047
1048 if (unlikely(ext4_emergency_state(sb)))
1049 return;
1050
1051 trace_ext4_error(sb, function, line);
1052 if (ext4_error_ratelimit(sb)) {
1053 va_start(args, fmt);
1054 vaf.fmt = fmt;
1055 vaf.va = &args;
1056 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
1057 sb->s_id, function, line, grp);
1058 if (ino)
1059 printk(KERN_CONT "inode %lu: ", ino);
1060 if (block)
1061 printk(KERN_CONT "block %llu:",
1062 (unsigned long long) block);
1063 printk(KERN_CONT "%pV\n", &vaf);
1064 va_end(args);
1065 }
1066
1067 if (test_opt(sb, ERRORS_CONT)) {
1068 if (test_opt(sb, WARN_ON_ERROR))
1069 WARN_ON_ONCE(1);
1070 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
1071 if (!bdev_read_only(bdev: sb->s_bdev)) {
1072 save_error_info(sb, EFSCORRUPTED, ino, block, func: function,
1073 line);
1074 schedule_work(work: &EXT4_SB(sb)->s_sb_upd_work);
1075 }
1076 return;
1077 }
1078 ext4_unlock_group(sb, group: grp);
1079 ext4_handle_error(sb, force_ro: false, EFSCORRUPTED, ino, block, func: function, line);
1080 /*
1081 * We only get here in the ERRORS_RO case; relocking the group
1082 * may be dangerous, but nothing bad will happen since the
1083 * filesystem will have already been marked read/only and the
1084 * journal has been aborted. We return 1 as a hint to callers
1085 * who might what to use the return value from
1086 * ext4_grp_locked_error() to distinguish between the
1087 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
1088 * aggressively from the ext4 function in question, with a
1089 * more appropriate error code.
1090 */
1091 ext4_lock_group(sb, group: grp);
1092 return;
1093}
1094
1095void ext4_mark_group_bitmap_corrupted(struct super_block *sb,
1096 ext4_group_t group,
1097 unsigned int flags)
1098{
1099 struct ext4_sb_info *sbi = EXT4_SB(sb);
1100 struct ext4_group_info *grp = ext4_get_group_info(sb, group);
1101 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, block_group: group, NULL);
1102 int ret;
1103
1104 if (!grp || !gdp)
1105 return;
1106 if (flags & EXT4_GROUP_INFO_BBITMAP_CORRUPT) {
1107 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT,
1108 addr: &grp->bb_state);
1109 if (!ret)
1110 percpu_counter_sub(fbc: &sbi->s_freeclusters_counter,
1111 amount: grp->bb_free);
1112 }
1113
1114 if (flags & EXT4_GROUP_INFO_IBITMAP_CORRUPT) {
1115 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT,
1116 addr: &grp->bb_state);
1117 if (!ret && gdp) {
1118 int count;
1119
1120 count = ext4_free_inodes_count(sb, bg: gdp);
1121 percpu_counter_sub(fbc: &sbi->s_freeinodes_counter,
1122 amount: count);
1123 }
1124 }
1125}
1126
1127void ext4_update_dynamic_rev(struct super_block *sb)
1128{
1129 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
1130
1131 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
1132 return;
1133
1134 ext4_warning(sb,
1135 "updating to rev %d because of new feature flag, "
1136 "running e2fsck is recommended",
1137 EXT4_DYNAMIC_REV);
1138
1139 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
1140 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
1141 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
1142 /* leave es->s_feature_*compat flags alone */
1143 /* es->s_uuid will be set by e2fsck if empty */
1144
1145 /*
1146 * The rest of the superblock fields should be zero, and if not it
1147 * means they are likely already in use, so leave them alone. We
1148 * can leave it up to e2fsck to clean up any inconsistencies there.
1149 */
1150}
1151
1152static inline struct inode *orphan_list_entry(struct list_head *l)
1153{
1154 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
1155}
1156
1157static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
1158{
1159 struct list_head *l;
1160
1161 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
1162 le32_to_cpu(sbi->s_es->s_last_orphan));
1163
1164 printk(KERN_ERR "sb_info orphan list:\n");
1165 list_for_each(l, &sbi->s_orphan) {
1166 struct inode *inode = orphan_list_entry(l);
1167 printk(KERN_ERR " "
1168 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
1169 inode->i_sb->s_id, inode->i_ino, inode,
1170 inode->i_mode, inode->i_nlink,
1171 NEXT_ORPHAN(inode));
1172 }
1173}
1174
1175#ifdef CONFIG_QUOTA
1176static int ext4_quota_off(struct super_block *sb, int type);
1177
1178static inline void ext4_quotas_off(struct super_block *sb, int type)
1179{
1180 BUG_ON(type > EXT4_MAXQUOTAS);
1181
1182 /* Use our quota_off function to clear inode flags etc. */
1183 for (type--; type >= 0; type--)
1184 ext4_quota_off(sb, type);
1185}
1186
1187/*
1188 * This is a helper function which is used in the mount/remount
1189 * codepaths (which holds s_umount) to fetch the quota file name.
1190 */
1191static inline char *get_qf_name(struct super_block *sb,
1192 struct ext4_sb_info *sbi,
1193 int type)
1194{
1195 return rcu_dereference_protected(sbi->s_qf_names[type],
1196 lockdep_is_held(&sb->s_umount));
1197}
1198#else
1199static inline void ext4_quotas_off(struct super_block *sb, int type)
1200{
1201}
1202#endif
1203
1204static int ext4_percpu_param_init(struct ext4_sb_info *sbi)
1205{
1206 ext4_fsblk_t block;
1207 int err;
1208
1209 block = ext4_count_free_clusters(sbi->s_sb);
1210 ext4_free_blocks_count_set(es: sbi->s_es, EXT4_C2B(sbi, block));
1211 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
1212 GFP_KERNEL);
1213 if (!err) {
1214 unsigned long freei = ext4_count_free_inodes(sbi->s_sb);
1215 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
1216 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
1217 GFP_KERNEL);
1218 }
1219 if (!err)
1220 err = percpu_counter_init(&sbi->s_dirs_counter,
1221 ext4_count_dirs(sbi->s_sb), GFP_KERNEL);
1222 if (!err)
1223 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
1224 GFP_KERNEL);
1225 if (!err)
1226 err = percpu_counter_init(&sbi->s_sra_exceeded_retry_limit, 0,
1227 GFP_KERNEL);
1228 if (!err)
1229 err = percpu_init_rwsem(&sbi->s_writepages_rwsem);
1230
1231 if (err)
1232 ext4_msg(sbi->s_sb, KERN_ERR, "insufficient memory");
1233
1234 return err;
1235}
1236
1237static void ext4_percpu_param_destroy(struct ext4_sb_info *sbi)
1238{
1239 percpu_counter_destroy(fbc: &sbi->s_freeclusters_counter);
1240 percpu_counter_destroy(fbc: &sbi->s_freeinodes_counter);
1241 percpu_counter_destroy(fbc: &sbi->s_dirs_counter);
1242 percpu_counter_destroy(fbc: &sbi->s_dirtyclusters_counter);
1243 percpu_counter_destroy(fbc: &sbi->s_sra_exceeded_retry_limit);
1244 percpu_free_rwsem(&sbi->s_writepages_rwsem);
1245}
1246
1247static void ext4_group_desc_free(struct ext4_sb_info *sbi)
1248{
1249 struct buffer_head **group_desc;
1250 int i;
1251
1252 rcu_read_lock();
1253 group_desc = rcu_dereference(sbi->s_group_desc);
1254 for (i = 0; i < sbi->s_gdb_count; i++)
1255 brelse(bh: group_desc[i]);
1256 kvfree(addr: group_desc);
1257 rcu_read_unlock();
1258}
1259
1260static void ext4_flex_groups_free(struct ext4_sb_info *sbi)
1261{
1262 struct flex_groups **flex_groups;
1263 int i;
1264
1265 rcu_read_lock();
1266 flex_groups = rcu_dereference(sbi->s_flex_groups);
1267 if (flex_groups) {
1268 for (i = 0; i < sbi->s_flex_groups_allocated; i++)
1269 kvfree(addr: flex_groups[i]);
1270 kvfree(addr: flex_groups);
1271 }
1272 rcu_read_unlock();
1273}
1274
1275static void ext4_put_super(struct super_block *sb)
1276{
1277 struct ext4_sb_info *sbi = EXT4_SB(sb);
1278 struct ext4_super_block *es = sbi->s_es;
1279 int aborted = 0;
1280 int err;
1281
1282 /*
1283 * Unregister sysfs before destroying jbd2 journal.
1284 * Since we could still access attr_journal_task attribute via sysfs
1285 * path which could have sbi->s_journal->j_task as NULL
1286 * Unregister sysfs before flush sbi->s_sb_upd_work.
1287 * Since user may read /proc/fs/ext4/xx/mb_groups during umount, If
1288 * read metadata verify failed then will queue error work.
1289 * update_super_work will call start_this_handle may trigger
1290 * BUG_ON.
1291 */
1292 ext4_unregister_sysfs(sb);
1293
1294 if (___ratelimit(rs: &ext4_mount_msg_ratelimit, func: "EXT4-fs unmount"))
1295 ext4_msg(sb, KERN_INFO, "unmounting filesystem %pU.",
1296 &sb->s_uuid);
1297
1298 ext4_unregister_li_request(sb);
1299 ext4_quotas_off(sb, EXT4_MAXQUOTAS);
1300
1301 destroy_workqueue(wq: sbi->rsv_conversion_wq);
1302 ext4_release_orphan_info(sb);
1303
1304 if (sbi->s_journal) {
1305 aborted = is_journal_aborted(journal: sbi->s_journal);
1306 err = ext4_journal_destroy(sbi, journal: sbi->s_journal);
1307 if ((err < 0) && !aborted) {
1308 ext4_abort(sb, -err, "Couldn't clean up the journal");
1309 }
1310 } else
1311 flush_work(work: &sbi->s_sb_upd_work);
1312
1313 ext4_es_unregister_shrinker(sbi);
1314 timer_shutdown_sync(timer: &sbi->s_err_report);
1315 ext4_release_system_zone(sb);
1316 ext4_mb_release(sb);
1317 ext4_ext_release(sb);
1318
1319 if (!ext4_emergency_state(sb) && !sb_rdonly(sb)) {
1320 if (!aborted) {
1321 ext4_clear_feature_journal_needs_recovery(sb);
1322 ext4_clear_feature_orphan_present(sb);
1323 es->s_state = cpu_to_le16(sbi->s_mount_state);
1324 }
1325 ext4_commit_super(sb);
1326 }
1327
1328 ext4_group_desc_free(sbi);
1329 ext4_flex_groups_free(sbi);
1330
1331 WARN_ON_ONCE(!(sbi->s_mount_state & EXT4_ERROR_FS) &&
1332 percpu_counter_sum(&sbi->s_dirtyclusters_counter));
1333 ext4_percpu_param_destroy(sbi);
1334#ifdef CONFIG_QUOTA
1335 for (int i = 0; i < EXT4_MAXQUOTAS; i++)
1336 kfree(objp: get_qf_name(sb, sbi, type: i));
1337#endif
1338
1339 /* Debugging code just in case the in-memory inode orphan list
1340 * isn't empty. The on-disk one can be non-empty if we've
1341 * detected an error and taken the fs readonly, but the
1342 * in-memory list had better be clean by this point. */
1343 if (!list_empty(head: &sbi->s_orphan))
1344 dump_orphan_list(sb, sbi);
1345 ASSERT(list_empty(&sbi->s_orphan));
1346
1347 sync_blockdev(bdev: sb->s_bdev);
1348 invalidate_bdev(bdev: sb->s_bdev);
1349 if (sbi->s_journal_bdev_file) {
1350 /*
1351 * Invalidate the journal device's buffers. We don't want them
1352 * floating about in memory - the physical journal device may
1353 * hotswapped, and it breaks the `ro-after' testing code.
1354 */
1355 sync_blockdev(bdev: file_bdev(bdev_file: sbi->s_journal_bdev_file));
1356 invalidate_bdev(bdev: file_bdev(bdev_file: sbi->s_journal_bdev_file));
1357 }
1358
1359 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
1360 sbi->s_ea_inode_cache = NULL;
1361
1362 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
1363 sbi->s_ea_block_cache = NULL;
1364
1365 ext4_stop_mmpd(sbi);
1366
1367 brelse(bh: sbi->s_sbh);
1368 sb->s_fs_info = NULL;
1369 /*
1370 * Now that we are completely done shutting down the
1371 * superblock, we need to actually destroy the kobject.
1372 */
1373 kobject_put(kobj: &sbi->s_kobj);
1374 wait_for_completion(&sbi->s_kobj_unregister);
1375 kfree(objp: sbi->s_blockgroup_lock);
1376 fs_put_dax(dax_dev: sbi->s_daxdev, NULL);
1377 fscrypt_free_dummy_policy(dummy_policy: &sbi->s_dummy_enc_policy);
1378#if IS_ENABLED(CONFIG_UNICODE)
1379 utf8_unload(sb->s_encoding);
1380#endif
1381 kfree(objp: sbi);
1382}
1383
1384static struct kmem_cache *ext4_inode_cachep;
1385
1386/*
1387 * Called inside transaction, so use GFP_NOFS
1388 */
1389static struct inode *ext4_alloc_inode(struct super_block *sb)
1390{
1391 struct ext4_inode_info *ei;
1392
1393 ei = alloc_inode_sb(sb, ext4_inode_cachep, GFP_NOFS);
1394 if (!ei)
1395 return NULL;
1396
1397 inode_set_iversion(inode: &ei->vfs_inode, val: 1);
1398 ei->i_flags = 0;
1399 spin_lock_init(&ei->i_raw_lock);
1400 ei->i_prealloc_node = RB_ROOT;
1401 atomic_set(v: &ei->i_prealloc_active, i: 0);
1402 rwlock_init(&ei->i_prealloc_lock);
1403 ext4_es_init_tree(tree: &ei->i_es_tree);
1404 rwlock_init(&ei->i_es_lock);
1405 INIT_LIST_HEAD(list: &ei->i_es_list);
1406 ei->i_es_all_nr = 0;
1407 ei->i_es_shk_nr = 0;
1408 ei->i_es_shrink_lblk = 0;
1409 ei->i_reserved_data_blocks = 0;
1410 spin_lock_init(&(ei->i_block_reservation_lock));
1411 ext4_init_pending_tree(tree: &ei->i_pending_tree);
1412#ifdef CONFIG_QUOTA
1413 ei->i_reserved_quota = 0;
1414 memset(s: &ei->i_dquot, c: 0, n: sizeof(ei->i_dquot));
1415#endif
1416 ei->jinode = NULL;
1417 INIT_LIST_HEAD(list: &ei->i_rsv_conversion_list);
1418 spin_lock_init(&ei->i_completed_io_lock);
1419 ei->i_sync_tid = 0;
1420 ei->i_datasync_tid = 0;
1421 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
1422 ext4_fc_init_inode(inode: &ei->vfs_inode);
1423 spin_lock_init(&ei->i_fc_lock);
1424 return &ei->vfs_inode;
1425}
1426
1427static int ext4_drop_inode(struct inode *inode)
1428{
1429 int drop = inode_generic_drop(inode);
1430
1431 if (!drop)
1432 drop = fscrypt_drop_inode(inode);
1433
1434 trace_ext4_drop_inode(inode, drop);
1435 return drop;
1436}
1437
1438static void ext4_free_in_core_inode(struct inode *inode)
1439{
1440 fscrypt_free_inode(inode);
1441 if (!list_empty(head: &(EXT4_I(inode)->i_fc_list))) {
1442 pr_warn("%s: inode %ld still in fc list",
1443 __func__, inode->i_ino);
1444 }
1445 kmem_cache_free(s: ext4_inode_cachep, EXT4_I(inode));
1446}
1447
1448static void ext4_destroy_inode(struct inode *inode)
1449{
1450 if (ext4_inode_orphan_tracked(inode)) {
1451 ext4_msg(inode->i_sb, KERN_ERR,
1452 "Inode %lu (%p): inode tracked as orphan!",
1453 inode->i_ino, EXT4_I(inode));
1454 print_hex_dump(KERN_INFO, prefix_str: "", prefix_type: DUMP_PREFIX_ADDRESS, rowsize: 16, groupsize: 4,
1455 EXT4_I(inode), len: sizeof(struct ext4_inode_info),
1456 ascii: true);
1457 dump_stack();
1458 }
1459
1460 if (!(EXT4_SB(sb: inode->i_sb)->s_mount_state & EXT4_ERROR_FS) &&
1461 WARN_ON_ONCE(EXT4_I(inode)->i_reserved_data_blocks))
1462 ext4_msg(inode->i_sb, KERN_ERR,
1463 "Inode %lu (%p): i_reserved_data_blocks (%u) not cleared!",
1464 inode->i_ino, EXT4_I(inode),
1465 EXT4_I(inode)->i_reserved_data_blocks);
1466}
1467
1468static void ext4_shutdown(struct super_block *sb)
1469{
1470 ext4_force_shutdown(sb, EXT4_GOING_FLAGS_NOLOGFLUSH);
1471}
1472
1473static void init_once(void *foo)
1474{
1475 struct ext4_inode_info *ei = foo;
1476
1477 INIT_LIST_HEAD(list: &ei->i_orphan);
1478 init_rwsem(&ei->xattr_sem);
1479 init_rwsem(&ei->i_data_sem);
1480 inode_init_once(&ei->vfs_inode);
1481 ext4_fc_init_inode(inode: &ei->vfs_inode);
1482#ifdef CONFIG_FS_ENCRYPTION
1483 ei->i_crypt_info = NULL;
1484#endif
1485#ifdef CONFIG_FS_VERITY
1486 ei->i_verity_info = NULL;
1487#endif
1488}
1489
1490static int __init init_inodecache(void)
1491{
1492 ext4_inode_cachep = kmem_cache_create_usercopy(name: "ext4_inode_cache",
1493 size: sizeof(struct ext4_inode_info), align: 0,
1494 SLAB_RECLAIM_ACCOUNT | SLAB_ACCOUNT,
1495 offsetof(struct ext4_inode_info, i_data),
1496 sizeof_field(struct ext4_inode_info, i_data),
1497 ctor: init_once);
1498 if (ext4_inode_cachep == NULL)
1499 return -ENOMEM;
1500 return 0;
1501}
1502
1503static void destroy_inodecache(void)
1504{
1505 /*
1506 * Make sure all delayed rcu free inodes are flushed before we
1507 * destroy cache.
1508 */
1509 rcu_barrier();
1510 kmem_cache_destroy(s: ext4_inode_cachep);
1511}
1512
1513void ext4_clear_inode(struct inode *inode)
1514{
1515 ext4_fc_del(inode);
1516 invalidate_inode_buffers(inode);
1517 clear_inode(inode);
1518 ext4_discard_preallocations(inode);
1519 ext4_es_remove_extent(inode, lblk: 0, EXT_MAX_BLOCKS);
1520 dquot_drop(inode);
1521 if (EXT4_I(inode)->jinode) {
1522 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1523 EXT4_I(inode)->jinode);
1524 jbd2_free_inode(EXT4_I(inode)->jinode);
1525 EXT4_I(inode)->jinode = NULL;
1526 }
1527 fscrypt_put_encryption_info(inode);
1528 fsverity_cleanup_inode(inode);
1529}
1530
1531static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1532 u64 ino, u32 generation)
1533{
1534 struct inode *inode;
1535
1536 /*
1537 * Currently we don't know the generation for parent directory, so
1538 * a generation of 0 means "accept any"
1539 */
1540 inode = ext4_iget(sb, ino, EXT4_IGET_HANDLE);
1541 if (IS_ERR(ptr: inode))
1542 return ERR_CAST(ptr: inode);
1543 if (generation && inode->i_generation != generation) {
1544 iput(inode);
1545 return ERR_PTR(error: -ESTALE);
1546 }
1547
1548 return inode;
1549}
1550
1551static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1552 int fh_len, int fh_type)
1553{
1554 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1555 get_inode: ext4_nfs_get_inode);
1556}
1557
1558static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1559 int fh_len, int fh_type)
1560{
1561 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1562 get_inode: ext4_nfs_get_inode);
1563}
1564
1565static int ext4_nfs_commit_metadata(struct inode *inode)
1566{
1567 struct writeback_control wbc = {
1568 .sync_mode = WB_SYNC_ALL
1569 };
1570
1571 trace_ext4_nfs_commit_metadata(inode);
1572 return ext4_write_inode(inode, &wbc);
1573}
1574
1575#ifdef CONFIG_QUOTA
1576static const char * const quotatypes[] = INITQFNAMES;
1577#define QTYPE2NAME(t) (quotatypes[t])
1578
1579static int ext4_write_dquot(struct dquot *dquot);
1580static int ext4_acquire_dquot(struct dquot *dquot);
1581static int ext4_release_dquot(struct dquot *dquot);
1582static int ext4_mark_dquot_dirty(struct dquot *dquot);
1583static int ext4_write_info(struct super_block *sb, int type);
1584static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1585 const struct path *path);
1586static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1587 size_t len, loff_t off);
1588static ssize_t ext4_quota_write(struct super_block *sb, int type,
1589 const char *data, size_t len, loff_t off);
1590static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1591 unsigned int flags);
1592
1593static struct dquot __rcu **ext4_get_dquots(struct inode *inode)
1594{
1595 return EXT4_I(inode)->i_dquot;
1596}
1597
1598static const struct dquot_operations ext4_quota_operations = {
1599 .get_reserved_space = ext4_get_reserved_space,
1600 .write_dquot = ext4_write_dquot,
1601 .acquire_dquot = ext4_acquire_dquot,
1602 .release_dquot = ext4_release_dquot,
1603 .mark_dirty = ext4_mark_dquot_dirty,
1604 .write_info = ext4_write_info,
1605 .alloc_dquot = dquot_alloc,
1606 .destroy_dquot = dquot_destroy,
1607 .get_projid = ext4_get_projid,
1608 .get_inode_usage = ext4_get_inode_usage,
1609 .get_next_id = dquot_get_next_id,
1610};
1611
1612static const struct quotactl_ops ext4_qctl_operations = {
1613 .quota_on = ext4_quota_on,
1614 .quota_off = ext4_quota_off,
1615 .quota_sync = dquot_quota_sync,
1616 .get_state = dquot_get_state,
1617 .set_info = dquot_set_dqinfo,
1618 .get_dqblk = dquot_get_dqblk,
1619 .set_dqblk = dquot_set_dqblk,
1620 .get_nextdqblk = dquot_get_next_dqblk,
1621};
1622#endif
1623
1624static const struct super_operations ext4_sops = {
1625 .alloc_inode = ext4_alloc_inode,
1626 .free_inode = ext4_free_in_core_inode,
1627 .destroy_inode = ext4_destroy_inode,
1628 .write_inode = ext4_write_inode,
1629 .dirty_inode = ext4_dirty_inode,
1630 .drop_inode = ext4_drop_inode,
1631 .evict_inode = ext4_evict_inode,
1632 .put_super = ext4_put_super,
1633 .sync_fs = ext4_sync_fs,
1634 .freeze_fs = ext4_freeze,
1635 .unfreeze_fs = ext4_unfreeze,
1636 .statfs = ext4_statfs,
1637 .show_options = ext4_show_options,
1638 .shutdown = ext4_shutdown,
1639#ifdef CONFIG_QUOTA
1640 .quota_read = ext4_quota_read,
1641 .quota_write = ext4_quota_write,
1642 .get_dquots = ext4_get_dquots,
1643#endif
1644};
1645
1646static const struct export_operations ext4_export_ops = {
1647 .encode_fh = generic_encode_ino32_fh,
1648 .fh_to_dentry = ext4_fh_to_dentry,
1649 .fh_to_parent = ext4_fh_to_parent,
1650 .get_parent = ext4_get_parent,
1651 .commit_metadata = ext4_nfs_commit_metadata,
1652};
1653
1654enum {
1655 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1656 Opt_resgid, Opt_resuid, Opt_sb,
1657 Opt_nouid32, Opt_debug, Opt_removed,
1658 Opt_user_xattr, Opt_acl,
1659 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1660 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1661 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1662 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1663 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
1664 Opt_inlinecrypt,
1665 Opt_usrjquota, Opt_grpjquota, Opt_quota,
1666 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1667 Opt_usrquota, Opt_grpquota, Opt_prjquota,
1668 Opt_dax, Opt_dax_always, Opt_dax_inode, Opt_dax_never,
1669 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_warn_on_error,
1670 Opt_nowarn_on_error, Opt_mblk_io_submit, Opt_debug_want_extra_isize,
1671 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1672 Opt_inode_readahead_blks, Opt_journal_ioprio,
1673 Opt_dioread_nolock, Opt_dioread_lock,
1674 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1675 Opt_max_dir_size_kb, Opt_nojournal_checksum, Opt_nombcache,
1676 Opt_no_prefetch_block_bitmaps, Opt_mb_optimize_scan,
1677 Opt_errors, Opt_data, Opt_data_err, Opt_jqfmt, Opt_dax_type,
1678#ifdef CONFIG_EXT4_DEBUG
1679 Opt_fc_debug_max_replay, Opt_fc_debug_force
1680#endif
1681};
1682
1683static const struct constant_table ext4_param_errors[] = {
1684 {"continue", EXT4_MOUNT_ERRORS_CONT},
1685 {.name: "panic", EXT4_MOUNT_ERRORS_PANIC},
1686 {.name: "remount-ro", EXT4_MOUNT_ERRORS_RO},
1687 {}
1688};
1689
1690static const struct constant_table ext4_param_data[] = {
1691 {"journal", EXT4_MOUNT_JOURNAL_DATA},
1692 {.name: "ordered", EXT4_MOUNT_ORDERED_DATA},
1693 {.name: "writeback", EXT4_MOUNT_WRITEBACK_DATA},
1694 {}
1695};
1696
1697static const struct constant_table ext4_param_data_err[] = {
1698 {"abort", Opt_data_err_abort},
1699 {.name: "ignore", .value: Opt_data_err_ignore},
1700 {}
1701};
1702
1703static const struct constant_table ext4_param_jqfmt[] = {
1704 {"vfsold", QFMT_VFS_OLD},
1705 {.name: "vfsv0", QFMT_VFS_V0},
1706 {.name: "vfsv1", QFMT_VFS_V1},
1707 {}
1708};
1709
1710static const struct constant_table ext4_param_dax[] = {
1711 {"always", Opt_dax_always},
1712 {.name: "inode", .value: Opt_dax_inode},
1713 {.name: "never", .value: Opt_dax_never},
1714 {}
1715};
1716
1717/*
1718 * Mount option specification
1719 * We don't use fsparam_flag_no because of the way we set the
1720 * options and the way we show them in _ext4_show_options(). To
1721 * keep the changes to a minimum, let's keep the negative options
1722 * separate for now.
1723 */
1724static const struct fs_parameter_spec ext4_param_specs[] = {
1725 fsparam_flag ("bsddf", Opt_bsd_df),
1726 fsparam_flag ("minixdf", Opt_minix_df),
1727 fsparam_flag ("grpid", Opt_grpid),
1728 fsparam_flag ("bsdgroups", Opt_grpid),
1729 fsparam_flag ("nogrpid", Opt_nogrpid),
1730 fsparam_flag ("sysvgroups", Opt_nogrpid),
1731 fsparam_gid ("resgid", Opt_resgid),
1732 fsparam_uid ("resuid", Opt_resuid),
1733 fsparam_u32 ("sb", Opt_sb),
1734 fsparam_enum ("errors", Opt_errors, ext4_param_errors),
1735 fsparam_flag ("nouid32", Opt_nouid32),
1736 fsparam_flag ("debug", Opt_debug),
1737 fsparam_flag ("oldalloc", Opt_removed),
1738 fsparam_flag ("orlov", Opt_removed),
1739 fsparam_flag ("user_xattr", Opt_user_xattr),
1740 fsparam_flag ("acl", Opt_acl),
1741 fsparam_flag ("norecovery", Opt_noload),
1742 fsparam_flag ("noload", Opt_noload),
1743 fsparam_flag ("bh", Opt_removed),
1744 fsparam_flag ("nobh", Opt_removed),
1745 fsparam_u32 ("commit", Opt_commit),
1746 fsparam_u32 ("min_batch_time", Opt_min_batch_time),
1747 fsparam_u32 ("max_batch_time", Opt_max_batch_time),
1748 fsparam_u32 ("journal_dev", Opt_journal_dev),
1749 fsparam_bdev ("journal_path", Opt_journal_path),
1750 fsparam_flag ("journal_checksum", Opt_journal_checksum),
1751 fsparam_flag ("nojournal_checksum", Opt_nojournal_checksum),
1752 fsparam_flag ("journal_async_commit",Opt_journal_async_commit),
1753 fsparam_flag ("abort", Opt_abort),
1754 fsparam_enum ("data", Opt_data, ext4_param_data),
1755 fsparam_enum ("data_err", Opt_data_err,
1756 ext4_param_data_err),
1757 fsparam_string_empty
1758 ("usrjquota", Opt_usrjquota),
1759 fsparam_string_empty
1760 ("grpjquota", Opt_grpjquota),
1761 fsparam_enum ("jqfmt", Opt_jqfmt, ext4_param_jqfmt),
1762 fsparam_flag ("grpquota", Opt_grpquota),
1763 fsparam_flag ("quota", Opt_quota),
1764 fsparam_flag ("noquota", Opt_noquota),
1765 fsparam_flag ("usrquota", Opt_usrquota),
1766 fsparam_flag ("prjquota", Opt_prjquota),
1767 fsparam_flag ("barrier", Opt_barrier),
1768 fsparam_u32 ("barrier", Opt_barrier),
1769 fsparam_flag ("nobarrier", Opt_nobarrier),
1770 fsparam_flag ("i_version", Opt_removed),
1771 fsparam_flag ("dax", Opt_dax),
1772 fsparam_enum ("dax", Opt_dax_type, ext4_param_dax),
1773 fsparam_u32 ("stripe", Opt_stripe),
1774 fsparam_flag ("delalloc", Opt_delalloc),
1775 fsparam_flag ("nodelalloc", Opt_nodelalloc),
1776 fsparam_flag ("warn_on_error", Opt_warn_on_error),
1777 fsparam_flag ("nowarn_on_error", Opt_nowarn_on_error),
1778 fsparam_u32 ("debug_want_extra_isize",
1779 Opt_debug_want_extra_isize),
1780 fsparam_flag ("mblk_io_submit", Opt_removed),
1781 fsparam_flag ("nomblk_io_submit", Opt_removed),
1782 fsparam_flag ("block_validity", Opt_block_validity),
1783 fsparam_flag ("noblock_validity", Opt_noblock_validity),
1784 fsparam_u32 ("inode_readahead_blks",
1785 Opt_inode_readahead_blks),
1786 fsparam_u32 ("journal_ioprio", Opt_journal_ioprio),
1787 fsparam_u32 ("auto_da_alloc", Opt_auto_da_alloc),
1788 fsparam_flag ("auto_da_alloc", Opt_auto_da_alloc),
1789 fsparam_flag ("noauto_da_alloc", Opt_noauto_da_alloc),
1790 fsparam_flag ("dioread_nolock", Opt_dioread_nolock),
1791 fsparam_flag ("nodioread_nolock", Opt_dioread_lock),
1792 fsparam_flag ("dioread_lock", Opt_dioread_lock),
1793 fsparam_flag ("discard", Opt_discard),
1794 fsparam_flag ("nodiscard", Opt_nodiscard),
1795 fsparam_u32 ("init_itable", Opt_init_itable),
1796 fsparam_flag ("init_itable", Opt_init_itable),
1797 fsparam_flag ("noinit_itable", Opt_noinit_itable),
1798#ifdef CONFIG_EXT4_DEBUG
1799 fsparam_flag ("fc_debug_force", Opt_fc_debug_force),
1800 fsparam_u32 ("fc_debug_max_replay", Opt_fc_debug_max_replay),
1801#endif
1802 fsparam_u32 ("max_dir_size_kb", Opt_max_dir_size_kb),
1803 fsparam_flag ("test_dummy_encryption",
1804 Opt_test_dummy_encryption),
1805 fsparam_string ("test_dummy_encryption",
1806 Opt_test_dummy_encryption),
1807 fsparam_flag ("inlinecrypt", Opt_inlinecrypt),
1808 fsparam_flag ("nombcache", Opt_nombcache),
1809 fsparam_flag ("no_mbcache", Opt_nombcache), /* for backward compatibility */
1810 fsparam_flag ("prefetch_block_bitmaps",
1811 Opt_removed),
1812 fsparam_flag ("no_prefetch_block_bitmaps",
1813 Opt_no_prefetch_block_bitmaps),
1814 fsparam_s32 ("mb_optimize_scan", Opt_mb_optimize_scan),
1815 fsparam_string ("check", Opt_removed), /* mount option from ext2/3 */
1816 fsparam_flag ("nocheck", Opt_removed), /* mount option from ext2/3 */
1817 fsparam_flag ("reservation", Opt_removed), /* mount option from ext2/3 */
1818 fsparam_flag ("noreservation", Opt_removed), /* mount option from ext2/3 */
1819 fsparam_u32 ("journal", Opt_removed), /* mount option from ext2/3 */
1820 {}
1821};
1822
1823
1824#define MOPT_SET 0x0001
1825#define MOPT_CLEAR 0x0002
1826#define MOPT_NOSUPPORT 0x0004
1827#define MOPT_EXPLICIT 0x0008
1828#ifdef CONFIG_QUOTA
1829#define MOPT_Q 0
1830#define MOPT_QFMT 0x0010
1831#else
1832#define MOPT_Q MOPT_NOSUPPORT
1833#define MOPT_QFMT MOPT_NOSUPPORT
1834#endif
1835#define MOPT_NO_EXT2 0x0020
1836#define MOPT_NO_EXT3 0x0040
1837#define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1838#define MOPT_SKIP 0x0080
1839#define MOPT_2 0x0100
1840
1841static const struct mount_opts {
1842 int token;
1843 int mount_opt;
1844 int flags;
1845} ext4_mount_opts[] = {
1846 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1847 {.token: Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1848 {.token: Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1849 {.token: Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1850 {.token: Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1851 {.token: Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1852 {.token: Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1853 MOPT_EXT4_ONLY | MOPT_SET},
1854 {.token: Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1855 MOPT_EXT4_ONLY | MOPT_CLEAR},
1856 {.token: Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1857 {.token: Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1858 {.token: Opt_delalloc, EXT4_MOUNT_DELALLOC,
1859 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1860 {.token: Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1861 MOPT_EXT4_ONLY | MOPT_CLEAR},
1862 {.token: Opt_warn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_SET},
1863 {.token: Opt_nowarn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_CLEAR},
1864 {.token: Opt_commit, .mount_opt: 0, MOPT_NO_EXT2},
1865 {.token: Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1866 MOPT_EXT4_ONLY | MOPT_CLEAR},
1867 {.token: Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1868 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1869 {.token: Opt_journal_async_commit, .mount_opt: (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1870 EXT4_MOUNT_JOURNAL_CHECKSUM),
1871 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1872 {.token: Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1873 {.token: Opt_data_err, EXT4_MOUNT_DATA_ERR_ABORT, MOPT_NO_EXT2},
1874 {.token: Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1875 {.token: Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1876 {.token: Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1877 {.token: Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1878 {.token: Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1879 {.token: Opt_dax_type, .mount_opt: 0, MOPT_EXT4_ONLY},
1880 {.token: Opt_journal_dev, .mount_opt: 0, MOPT_NO_EXT2},
1881 {.token: Opt_journal_path, .mount_opt: 0, MOPT_NO_EXT2},
1882 {.token: Opt_journal_ioprio, .mount_opt: 0, MOPT_NO_EXT2},
1883 {.token: Opt_data, .mount_opt: 0, MOPT_NO_EXT2},
1884 {.token: Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1885#ifdef CONFIG_EXT4_FS_POSIX_ACL
1886 {.token: Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1887#else
1888 {Opt_acl, 0, MOPT_NOSUPPORT},
1889#endif
1890 {.token: Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1891 {.token: Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1892 {.token: Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1893 {.token: Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1894 MOPT_SET | MOPT_Q},
1895 {.token: Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1896 MOPT_SET | MOPT_Q},
1897 {.token: Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA,
1898 MOPT_SET | MOPT_Q},
1899 {.token: Opt_noquota, .mount_opt: (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1900 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA),
1901 MOPT_CLEAR | MOPT_Q},
1902 {.token: Opt_usrjquota, .mount_opt: 0, MOPT_Q},
1903 {.token: Opt_grpjquota, .mount_opt: 0, MOPT_Q},
1904 {.token: Opt_jqfmt, .mount_opt: 0, MOPT_QFMT},
1905 {.token: Opt_nombcache, EXT4_MOUNT_NO_MBCACHE, MOPT_SET},
1906 {.token: Opt_no_prefetch_block_bitmaps, EXT4_MOUNT_NO_PREFETCH_BLOCK_BITMAPS,
1907 MOPT_SET},
1908#ifdef CONFIG_EXT4_DEBUG
1909 {Opt_fc_debug_force, EXT4_MOUNT2_JOURNAL_FAST_COMMIT,
1910 MOPT_SET | MOPT_2 | MOPT_EXT4_ONLY},
1911#endif
1912 {.token: Opt_abort, EXT4_MOUNT2_ABORT, MOPT_SET | MOPT_2},
1913 {.token: Opt_err, .mount_opt: 0, .flags: 0}
1914};
1915
1916#if IS_ENABLED(CONFIG_UNICODE)
1917static const struct ext4_sb_encodings {
1918 __u16 magic;
1919 char *name;
1920 unsigned int version;
1921} ext4_sb_encoding_map[] = {
1922 {EXT4_ENC_UTF8_12_1, "utf8", UNICODE_AGE(12, 1, 0)},
1923};
1924
1925static const struct ext4_sb_encodings *
1926ext4_sb_read_encoding(const struct ext4_super_block *es)
1927{
1928 __u16 magic = le16_to_cpu(es->s_encoding);
1929 int i;
1930
1931 for (i = 0; i < ARRAY_SIZE(ext4_sb_encoding_map); i++)
1932 if (magic == ext4_sb_encoding_map[i].magic)
1933 return &ext4_sb_encoding_map[i];
1934
1935 return NULL;
1936}
1937#endif
1938
1939#define EXT4_SPEC_JQUOTA (1 << 0)
1940#define EXT4_SPEC_JQFMT (1 << 1)
1941#define EXT4_SPEC_DATAJ (1 << 2)
1942#define EXT4_SPEC_SB_BLOCK (1 << 3)
1943#define EXT4_SPEC_JOURNAL_DEV (1 << 4)
1944#define EXT4_SPEC_JOURNAL_IOPRIO (1 << 5)
1945#define EXT4_SPEC_s_want_extra_isize (1 << 7)
1946#define EXT4_SPEC_s_max_batch_time (1 << 8)
1947#define EXT4_SPEC_s_min_batch_time (1 << 9)
1948#define EXT4_SPEC_s_inode_readahead_blks (1 << 10)
1949#define EXT4_SPEC_s_li_wait_mult (1 << 11)
1950#define EXT4_SPEC_s_max_dir_size_kb (1 << 12)
1951#define EXT4_SPEC_s_stripe (1 << 13)
1952#define EXT4_SPEC_s_resuid (1 << 14)
1953#define EXT4_SPEC_s_resgid (1 << 15)
1954#define EXT4_SPEC_s_commit_interval (1 << 16)
1955#define EXT4_SPEC_s_fc_debug_max_replay (1 << 17)
1956#define EXT4_SPEC_s_sb_block (1 << 18)
1957#define EXT4_SPEC_mb_optimize_scan (1 << 19)
1958
1959struct ext4_fs_context {
1960 char *s_qf_names[EXT4_MAXQUOTAS];
1961 struct fscrypt_dummy_policy dummy_enc_policy;
1962 int s_jquota_fmt; /* Format of quota to use */
1963#ifdef CONFIG_EXT4_DEBUG
1964 int s_fc_debug_max_replay;
1965#endif
1966 unsigned short qname_spec;
1967 unsigned long vals_s_flags; /* Bits to set in s_flags */
1968 unsigned long mask_s_flags; /* Bits changed in s_flags */
1969 unsigned long journal_devnum;
1970 unsigned long s_commit_interval;
1971 unsigned long s_stripe;
1972 unsigned int s_inode_readahead_blks;
1973 unsigned int s_want_extra_isize;
1974 unsigned int s_li_wait_mult;
1975 unsigned int s_max_dir_size_kb;
1976 unsigned int journal_ioprio;
1977 unsigned int vals_s_mount_opt;
1978 unsigned int mask_s_mount_opt;
1979 unsigned int vals_s_mount_opt2;
1980 unsigned int mask_s_mount_opt2;
1981 unsigned int opt_flags; /* MOPT flags */
1982 unsigned int spec;
1983 u32 s_max_batch_time;
1984 u32 s_min_batch_time;
1985 kuid_t s_resuid;
1986 kgid_t s_resgid;
1987 ext4_fsblk_t s_sb_block;
1988};
1989
1990static void ext4_fc_free(struct fs_context *fc)
1991{
1992 struct ext4_fs_context *ctx = fc->fs_private;
1993 int i;
1994
1995 if (!ctx)
1996 return;
1997
1998 for (i = 0; i < EXT4_MAXQUOTAS; i++)
1999 kfree(objp: ctx->s_qf_names[i]);
2000
2001 fscrypt_free_dummy_policy(dummy_policy: &ctx->dummy_enc_policy);
2002 kfree(objp: ctx);
2003}
2004
2005int ext4_init_fs_context(struct fs_context *fc)
2006{
2007 struct ext4_fs_context *ctx;
2008
2009 ctx = kzalloc(sizeof(struct ext4_fs_context), GFP_KERNEL);
2010 if (!ctx)
2011 return -ENOMEM;
2012
2013 fc->fs_private = ctx;
2014 fc->ops = &ext4_context_ops;
2015
2016 /* i_version is always enabled now */
2017 fc->sb_flags |= SB_I_VERSION;
2018
2019 return 0;
2020}
2021
2022#ifdef CONFIG_QUOTA
2023/*
2024 * Note the name of the specified quota file.
2025 */
2026static int note_qf_name(struct fs_context *fc, int qtype,
2027 struct fs_parameter *param)
2028{
2029 struct ext4_fs_context *ctx = fc->fs_private;
2030 char *qname;
2031
2032 if (param->size < 1) {
2033 ext4_msg(NULL, KERN_ERR, "Missing quota name");
2034 return -EINVAL;
2035 }
2036 if (strchr(param->string, '/')) {
2037 ext4_msg(NULL, KERN_ERR,
2038 "quotafile must be on filesystem root");
2039 return -EINVAL;
2040 }
2041 if (ctx->s_qf_names[qtype]) {
2042 if (strcmp(ctx->s_qf_names[qtype], param->string) != 0) {
2043 ext4_msg(NULL, KERN_ERR,
2044 "%s quota file already specified",
2045 QTYPE2NAME(qtype));
2046 return -EINVAL;
2047 }
2048 return 0;
2049 }
2050
2051 qname = kmemdup_nul(s: param->string, len: param->size, GFP_KERNEL);
2052 if (!qname) {
2053 ext4_msg(NULL, KERN_ERR,
2054 "Not enough memory for storing quotafile name");
2055 return -ENOMEM;
2056 }
2057 ctx->s_qf_names[qtype] = qname;
2058 ctx->qname_spec |= 1 << qtype;
2059 ctx->spec |= EXT4_SPEC_JQUOTA;
2060 return 0;
2061}
2062
2063/*
2064 * Clear the name of the specified quota file.
2065 */
2066static int unnote_qf_name(struct fs_context *fc, int qtype)
2067{
2068 struct ext4_fs_context *ctx = fc->fs_private;
2069
2070 kfree(objp: ctx->s_qf_names[qtype]);
2071
2072 ctx->s_qf_names[qtype] = NULL;
2073 ctx->qname_spec |= 1 << qtype;
2074 ctx->spec |= EXT4_SPEC_JQUOTA;
2075 return 0;
2076}
2077#endif
2078
2079static int ext4_parse_test_dummy_encryption(const struct fs_parameter *param,
2080 struct ext4_fs_context *ctx)
2081{
2082 int err;
2083
2084 if (!IS_ENABLED(CONFIG_FS_ENCRYPTION)) {
2085 ext4_msg(NULL, KERN_WARNING,
2086 "test_dummy_encryption option not supported");
2087 return -EINVAL;
2088 }
2089 err = fscrypt_parse_test_dummy_encryption(param,
2090 dummy_policy: &ctx->dummy_enc_policy);
2091 if (err == -EINVAL) {
2092 ext4_msg(NULL, KERN_WARNING,
2093 "Value of option \"%s\" is unrecognized", param->key);
2094 } else if (err == -EEXIST) {
2095 ext4_msg(NULL, KERN_WARNING,
2096 "Conflicting test_dummy_encryption options");
2097 return -EINVAL;
2098 }
2099 return err;
2100}
2101
2102#define EXT4_SET_CTX(name) \
2103static inline __maybe_unused \
2104void ctx_set_##name(struct ext4_fs_context *ctx, unsigned long flag) \
2105{ \
2106 ctx->mask_s_##name |= flag; \
2107 ctx->vals_s_##name |= flag; \
2108}
2109
2110#define EXT4_CLEAR_CTX(name) \
2111static inline __maybe_unused \
2112void ctx_clear_##name(struct ext4_fs_context *ctx, unsigned long flag) \
2113{ \
2114 ctx->mask_s_##name |= flag; \
2115 ctx->vals_s_##name &= ~flag; \
2116}
2117
2118#define EXT4_TEST_CTX(name) \
2119static inline unsigned long \
2120ctx_test_##name(struct ext4_fs_context *ctx, unsigned long flag) \
2121{ \
2122 return (ctx->vals_s_##name & flag); \
2123}
2124
2125EXT4_SET_CTX(flags); /* set only */
2126EXT4_SET_CTX(mount_opt);
2127EXT4_CLEAR_CTX(mount_opt);
2128EXT4_TEST_CTX(mount_opt);
2129EXT4_SET_CTX(mount_opt2);
2130EXT4_CLEAR_CTX(mount_opt2);
2131EXT4_TEST_CTX(mount_opt2);
2132
2133static int ext4_parse_param(struct fs_context *fc, struct fs_parameter *param)
2134{
2135 struct ext4_fs_context *ctx = fc->fs_private;
2136 struct fs_parse_result result;
2137 const struct mount_opts *m;
2138 int is_remount;
2139 int token;
2140
2141 token = fs_parse(fc, desc: ext4_param_specs, param, result: &result);
2142 if (token < 0)
2143 return token;
2144 is_remount = fc->purpose == FS_CONTEXT_FOR_RECONFIGURE;
2145
2146 for (m = ext4_mount_opts; m->token != Opt_err; m++)
2147 if (token == m->token)
2148 break;
2149
2150 ctx->opt_flags |= m->flags;
2151
2152 if (m->flags & MOPT_EXPLICIT) {
2153 if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
2154 ctx_set_mount_opt2(ctx, EXT4_MOUNT2_EXPLICIT_DELALLOC);
2155 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
2156 ctx_set_mount_opt2(ctx,
2157 EXT4_MOUNT2_EXPLICIT_JOURNAL_CHECKSUM);
2158 } else
2159 return -EINVAL;
2160 }
2161
2162 if (m->flags & MOPT_NOSUPPORT) {
2163 ext4_msg(NULL, KERN_ERR, "%s option not supported",
2164 param->key);
2165 return 0;
2166 }
2167
2168 switch (token) {
2169#ifdef CONFIG_QUOTA
2170 case Opt_usrjquota:
2171 if (!*param->string)
2172 return unnote_qf_name(fc, qtype: USRQUOTA);
2173 else
2174 return note_qf_name(fc, qtype: USRQUOTA, param);
2175 case Opt_grpjquota:
2176 if (!*param->string)
2177 return unnote_qf_name(fc, qtype: GRPQUOTA);
2178 else
2179 return note_qf_name(fc, qtype: GRPQUOTA, param);
2180#endif
2181 case Opt_sb:
2182 if (fc->purpose == FS_CONTEXT_FOR_RECONFIGURE) {
2183 ext4_msg(NULL, KERN_WARNING,
2184 "Ignoring %s option on remount", param->key);
2185 } else {
2186 ctx->s_sb_block = result.uint_32;
2187 ctx->spec |= EXT4_SPEC_s_sb_block;
2188 }
2189 return 0;
2190 case Opt_removed:
2191 ext4_msg(NULL, KERN_WARNING, "Ignoring removed %s option",
2192 param->key);
2193 return 0;
2194 case Opt_inlinecrypt:
2195#ifdef CONFIG_FS_ENCRYPTION_INLINE_CRYPT
2196 ctx_set_flags(ctx, SB_INLINECRYPT);
2197#else
2198 ext4_msg(NULL, KERN_ERR, "inline encryption not supported");
2199#endif
2200 return 0;
2201 case Opt_errors:
2202 ctx_clear_mount_opt(ctx, EXT4_MOUNT_ERRORS_MASK);
2203 ctx_set_mount_opt(ctx, flag: result.uint_32);
2204 return 0;
2205#ifdef CONFIG_QUOTA
2206 case Opt_jqfmt:
2207 ctx->s_jquota_fmt = result.uint_32;
2208 ctx->spec |= EXT4_SPEC_JQFMT;
2209 return 0;
2210#endif
2211 case Opt_data:
2212 ctx_clear_mount_opt(ctx, EXT4_MOUNT_DATA_FLAGS);
2213 ctx_set_mount_opt(ctx, flag: result.uint_32);
2214 ctx->spec |= EXT4_SPEC_DATAJ;
2215 return 0;
2216 case Opt_commit:
2217 if (result.uint_32 == 0)
2218 result.uint_32 = JBD2_DEFAULT_MAX_COMMIT_AGE;
2219 else if (result.uint_32 > INT_MAX / HZ) {
2220 ext4_msg(NULL, KERN_ERR,
2221 "Invalid commit interval %d, "
2222 "must be smaller than %d",
2223 result.uint_32, INT_MAX / HZ);
2224 return -EINVAL;
2225 }
2226 ctx->s_commit_interval = HZ * result.uint_32;
2227 ctx->spec |= EXT4_SPEC_s_commit_interval;
2228 return 0;
2229 case Opt_debug_want_extra_isize:
2230 if ((result.uint_32 & 1) || (result.uint_32 < 4)) {
2231 ext4_msg(NULL, KERN_ERR,
2232 "Invalid want_extra_isize %d", result.uint_32);
2233 return -EINVAL;
2234 }
2235 ctx->s_want_extra_isize = result.uint_32;
2236 ctx->spec |= EXT4_SPEC_s_want_extra_isize;
2237 return 0;
2238 case Opt_max_batch_time:
2239 ctx->s_max_batch_time = result.uint_32;
2240 ctx->spec |= EXT4_SPEC_s_max_batch_time;
2241 return 0;
2242 case Opt_min_batch_time:
2243 ctx->s_min_batch_time = result.uint_32;
2244 ctx->spec |= EXT4_SPEC_s_min_batch_time;
2245 return 0;
2246 case Opt_inode_readahead_blks:
2247 if (result.uint_32 &&
2248 (result.uint_32 > (1 << 30) ||
2249 !is_power_of_2(n: result.uint_32))) {
2250 ext4_msg(NULL, KERN_ERR,
2251 "EXT4-fs: inode_readahead_blks must be "
2252 "0 or a power of 2 smaller than 2^31");
2253 return -EINVAL;
2254 }
2255 ctx->s_inode_readahead_blks = result.uint_32;
2256 ctx->spec |= EXT4_SPEC_s_inode_readahead_blks;
2257 return 0;
2258 case Opt_init_itable:
2259 ctx_set_mount_opt(ctx, EXT4_MOUNT_INIT_INODE_TABLE);
2260 ctx->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
2261 if (param->type == fs_value_is_string)
2262 ctx->s_li_wait_mult = result.uint_32;
2263 ctx->spec |= EXT4_SPEC_s_li_wait_mult;
2264 return 0;
2265 case Opt_max_dir_size_kb:
2266 ctx->s_max_dir_size_kb = result.uint_32;
2267 ctx->spec |= EXT4_SPEC_s_max_dir_size_kb;
2268 return 0;
2269#ifdef CONFIG_EXT4_DEBUG
2270 case Opt_fc_debug_max_replay:
2271 ctx->s_fc_debug_max_replay = result.uint_32;
2272 ctx->spec |= EXT4_SPEC_s_fc_debug_max_replay;
2273 return 0;
2274#endif
2275 case Opt_stripe:
2276 ctx->s_stripe = result.uint_32;
2277 ctx->spec |= EXT4_SPEC_s_stripe;
2278 return 0;
2279 case Opt_resuid:
2280 ctx->s_resuid = result.uid;
2281 ctx->spec |= EXT4_SPEC_s_resuid;
2282 return 0;
2283 case Opt_resgid:
2284 ctx->s_resgid = result.gid;
2285 ctx->spec |= EXT4_SPEC_s_resgid;
2286 return 0;
2287 case Opt_journal_dev:
2288 if (is_remount) {
2289 ext4_msg(NULL, KERN_ERR,
2290 "Cannot specify journal on remount");
2291 return -EINVAL;
2292 }
2293 ctx->journal_devnum = result.uint_32;
2294 ctx->spec |= EXT4_SPEC_JOURNAL_DEV;
2295 return 0;
2296 case Opt_journal_path:
2297 {
2298 struct inode *journal_inode;
2299 struct path path;
2300 int error;
2301
2302 if (is_remount) {
2303 ext4_msg(NULL, KERN_ERR,
2304 "Cannot specify journal on remount");
2305 return -EINVAL;
2306 }
2307
2308 error = fs_lookup_param(fc, param, want_bdev: 1, LOOKUP_FOLLOW, path: &path);
2309 if (error) {
2310 ext4_msg(NULL, KERN_ERR, "error: could not find "
2311 "journal device path");
2312 return -EINVAL;
2313 }
2314
2315 journal_inode = d_inode(dentry: path.dentry);
2316 ctx->journal_devnum = new_encode_dev(dev: journal_inode->i_rdev);
2317 ctx->spec |= EXT4_SPEC_JOURNAL_DEV;
2318 path_put(&path);
2319 return 0;
2320 }
2321 case Opt_journal_ioprio:
2322 if (result.uint_32 > 7) {
2323 ext4_msg(NULL, KERN_ERR, "Invalid journal IO priority"
2324 " (must be 0-7)");
2325 return -EINVAL;
2326 }
2327 ctx->journal_ioprio =
2328 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, result.uint_32);
2329 ctx->spec |= EXT4_SPEC_JOURNAL_IOPRIO;
2330 return 0;
2331 case Opt_test_dummy_encryption:
2332 return ext4_parse_test_dummy_encryption(param, ctx);
2333 case Opt_dax:
2334 case Opt_dax_type:
2335#ifdef CONFIG_FS_DAX
2336 {
2337 int type = (token == Opt_dax) ?
2338 Opt_dax : result.uint_32;
2339
2340 switch (type) {
2341 case Opt_dax:
2342 case Opt_dax_always:
2343 ctx_set_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS);
2344 ctx_clear_mount_opt2(ctx, EXT4_MOUNT2_DAX_NEVER);
2345 break;
2346 case Opt_dax_never:
2347 ctx_set_mount_opt2(ctx, EXT4_MOUNT2_DAX_NEVER);
2348 ctx_clear_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS);
2349 break;
2350 case Opt_dax_inode:
2351 ctx_clear_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS);
2352 ctx_clear_mount_opt2(ctx, EXT4_MOUNT2_DAX_NEVER);
2353 /* Strictly for printing options */
2354 ctx_set_mount_opt2(ctx, EXT4_MOUNT2_DAX_INODE);
2355 break;
2356 }
2357 return 0;
2358 }
2359#else
2360 ext4_msg(NULL, KERN_INFO, "dax option not supported");
2361 return -EINVAL;
2362#endif
2363 case Opt_data_err:
2364 if (result.uint_32 == Opt_data_err_abort)
2365 ctx_set_mount_opt(ctx, flag: m->mount_opt);
2366 else if (result.uint_32 == Opt_data_err_ignore)
2367 ctx_clear_mount_opt(ctx, flag: m->mount_opt);
2368 return 0;
2369 case Opt_mb_optimize_scan:
2370 if (result.int_32 == 1) {
2371 ctx_set_mount_opt2(ctx, EXT4_MOUNT2_MB_OPTIMIZE_SCAN);
2372 ctx->spec |= EXT4_SPEC_mb_optimize_scan;
2373 } else if (result.int_32 == 0) {
2374 ctx_clear_mount_opt2(ctx, EXT4_MOUNT2_MB_OPTIMIZE_SCAN);
2375 ctx->spec |= EXT4_SPEC_mb_optimize_scan;
2376 } else {
2377 ext4_msg(NULL, KERN_WARNING,
2378 "mb_optimize_scan should be set to 0 or 1.");
2379 return -EINVAL;
2380 }
2381 return 0;
2382 }
2383
2384 /*
2385 * At this point we should only be getting options requiring MOPT_SET,
2386 * or MOPT_CLEAR. Anything else is a bug
2387 */
2388 if (m->token == Opt_err) {
2389 ext4_msg(NULL, KERN_WARNING, "buggy handling of option %s",
2390 param->key);
2391 WARN_ON(1);
2392 return -EINVAL;
2393 }
2394
2395 else {
2396 unsigned int set = 0;
2397
2398 if ((param->type == fs_value_is_flag) ||
2399 result.uint_32 > 0)
2400 set = 1;
2401
2402 if (m->flags & MOPT_CLEAR)
2403 set = !set;
2404 else if (unlikely(!(m->flags & MOPT_SET))) {
2405 ext4_msg(NULL, KERN_WARNING,
2406 "buggy handling of option %s",
2407 param->key);
2408 WARN_ON(1);
2409 return -EINVAL;
2410 }
2411 if (m->flags & MOPT_2) {
2412 if (set != 0)
2413 ctx_set_mount_opt2(ctx, flag: m->mount_opt);
2414 else
2415 ctx_clear_mount_opt2(ctx, flag: m->mount_opt);
2416 } else {
2417 if (set != 0)
2418 ctx_set_mount_opt(ctx, flag: m->mount_opt);
2419 else
2420 ctx_clear_mount_opt(ctx, flag: m->mount_opt);
2421 }
2422 }
2423
2424 return 0;
2425}
2426
2427static int parse_options(struct fs_context *fc, char *options)
2428{
2429 struct fs_parameter param;
2430 int ret;
2431 char *key;
2432
2433 if (!options)
2434 return 0;
2435
2436 while ((key = strsep(&options, ",")) != NULL) {
2437 if (*key) {
2438 size_t v_len = 0;
2439 char *value = strchr(key, '=');
2440
2441 param.type = fs_value_is_flag;
2442 param.string = NULL;
2443
2444 if (value) {
2445 if (value == key)
2446 continue;
2447
2448 *value++ = 0;
2449 v_len = strlen(value);
2450 param.string = kmemdup_nul(s: value, len: v_len,
2451 GFP_KERNEL);
2452 if (!param.string)
2453 return -ENOMEM;
2454 param.type = fs_value_is_string;
2455 }
2456
2457 param.key = key;
2458 param.size = v_len;
2459
2460 ret = ext4_parse_param(fc, param: &param);
2461 kfree(objp: param.string);
2462 if (ret < 0)
2463 return ret;
2464 }
2465 }
2466
2467 ret = ext4_validate_options(fc);
2468 if (ret < 0)
2469 return ret;
2470
2471 return 0;
2472}
2473
2474static int parse_apply_sb_mount_options(struct super_block *sb,
2475 struct ext4_fs_context *m_ctx)
2476{
2477 struct ext4_sb_info *sbi = EXT4_SB(sb);
2478 char s_mount_opts[65];
2479 struct ext4_fs_context *s_ctx = NULL;
2480 struct fs_context *fc = NULL;
2481 int ret = -ENOMEM;
2482
2483 if (!sbi->s_es->s_mount_opts[0])
2484 return 0;
2485
2486 strscpy_pad(s_mount_opts, sbi->s_es->s_mount_opts);
2487
2488 fc = kzalloc(sizeof(struct fs_context), GFP_KERNEL);
2489 if (!fc)
2490 return -ENOMEM;
2491
2492 s_ctx = kzalloc(sizeof(struct ext4_fs_context), GFP_KERNEL);
2493 if (!s_ctx)
2494 goto out_free;
2495
2496 fc->fs_private = s_ctx;
2497 fc->s_fs_info = sbi;
2498
2499 ret = parse_options(fc, options: s_mount_opts);
2500 if (ret < 0)
2501 goto parse_failed;
2502
2503 ret = ext4_check_opt_consistency(fc, sb);
2504 if (ret < 0) {
2505parse_failed:
2506 ext4_msg(sb, KERN_WARNING,
2507 "failed to parse options in superblock: %s",
2508 s_mount_opts);
2509 ret = 0;
2510 goto out_free;
2511 }
2512
2513 if (s_ctx->spec & EXT4_SPEC_JOURNAL_DEV)
2514 m_ctx->journal_devnum = s_ctx->journal_devnum;
2515 if (s_ctx->spec & EXT4_SPEC_JOURNAL_IOPRIO)
2516 m_ctx->journal_ioprio = s_ctx->journal_ioprio;
2517
2518 ext4_apply_options(fc, sb);
2519 ret = 0;
2520
2521out_free:
2522 ext4_fc_free(fc);
2523 kfree(objp: fc);
2524 return ret;
2525}
2526
2527static void ext4_apply_quota_options(struct fs_context *fc,
2528 struct super_block *sb)
2529{
2530#ifdef CONFIG_QUOTA
2531 bool quota_feature = ext4_has_feature_quota(sb);
2532 struct ext4_fs_context *ctx = fc->fs_private;
2533 struct ext4_sb_info *sbi = EXT4_SB(sb);
2534 char *qname;
2535 int i;
2536
2537 if (quota_feature)
2538 return;
2539
2540 if (ctx->spec & EXT4_SPEC_JQUOTA) {
2541 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2542 if (!(ctx->qname_spec & (1 << i)))
2543 continue;
2544
2545 qname = ctx->s_qf_names[i]; /* May be NULL */
2546 if (qname)
2547 set_opt(sb, QUOTA);
2548 ctx->s_qf_names[i] = NULL;
2549 qname = rcu_replace_pointer(sbi->s_qf_names[i], qname,
2550 lockdep_is_held(&sb->s_umount));
2551 if (qname)
2552 kfree_rcu_mightsleep(qname);
2553 }
2554 }
2555
2556 if (ctx->spec & EXT4_SPEC_JQFMT)
2557 sbi->s_jquota_fmt = ctx->s_jquota_fmt;
2558#endif
2559}
2560
2561/*
2562 * Check quota settings consistency.
2563 */
2564static int ext4_check_quota_consistency(struct fs_context *fc,
2565 struct super_block *sb)
2566{
2567#ifdef CONFIG_QUOTA
2568 struct ext4_fs_context *ctx = fc->fs_private;
2569 struct ext4_sb_info *sbi = EXT4_SB(sb);
2570 bool quota_feature = ext4_has_feature_quota(sb);
2571 bool quota_loaded = sb_any_quota_loaded(sb);
2572 bool usr_qf_name, grp_qf_name, usrquota, grpquota;
2573 int quota_flags, i;
2574
2575 /*
2576 * We do the test below only for project quotas. 'usrquota' and
2577 * 'grpquota' mount options are allowed even without quota feature
2578 * to support legacy quotas in quota files.
2579 */
2580 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_PRJQUOTA) &&
2581 !ext4_has_feature_project(sb)) {
2582 ext4_msg(NULL, KERN_ERR, "Project quota feature not enabled. "
2583 "Cannot enable project quota enforcement.");
2584 return -EINVAL;
2585 }
2586
2587 quota_flags = EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
2588 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA;
2589 if (quota_loaded &&
2590 ctx->mask_s_mount_opt & quota_flags &&
2591 !ctx_test_mount_opt(ctx, flag: quota_flags))
2592 goto err_quota_change;
2593
2594 if (ctx->spec & EXT4_SPEC_JQUOTA) {
2595
2596 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2597 if (!(ctx->qname_spec & (1 << i)))
2598 continue;
2599
2600 if (quota_loaded &&
2601 !!sbi->s_qf_names[i] != !!ctx->s_qf_names[i])
2602 goto err_jquota_change;
2603
2604 if (sbi->s_qf_names[i] && ctx->s_qf_names[i] &&
2605 strcmp(get_qf_name(sb, sbi, type: i),
2606 ctx->s_qf_names[i]) != 0)
2607 goto err_jquota_specified;
2608 }
2609
2610 if (quota_feature) {
2611 ext4_msg(NULL, KERN_INFO,
2612 "Journaled quota options ignored when "
2613 "QUOTA feature is enabled");
2614 return 0;
2615 }
2616 }
2617
2618 if (ctx->spec & EXT4_SPEC_JQFMT) {
2619 if (sbi->s_jquota_fmt != ctx->s_jquota_fmt && quota_loaded)
2620 goto err_jquota_change;
2621 if (quota_feature) {
2622 ext4_msg(NULL, KERN_INFO, "Quota format mount options "
2623 "ignored when QUOTA feature is enabled");
2624 return 0;
2625 }
2626 }
2627
2628 /* Make sure we don't mix old and new quota format */
2629 usr_qf_name = (get_qf_name(sb, sbi, type: USRQUOTA) ||
2630 ctx->s_qf_names[USRQUOTA]);
2631 grp_qf_name = (get_qf_name(sb, sbi, type: GRPQUOTA) ||
2632 ctx->s_qf_names[GRPQUOTA]);
2633
2634 usrquota = (ctx_test_mount_opt(ctx, EXT4_MOUNT_USRQUOTA) ||
2635 test_opt(sb, USRQUOTA));
2636
2637 grpquota = (ctx_test_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA) ||
2638 test_opt(sb, GRPQUOTA));
2639
2640 if (usr_qf_name) {
2641 ctx_clear_mount_opt(ctx, EXT4_MOUNT_USRQUOTA);
2642 usrquota = false;
2643 }
2644 if (grp_qf_name) {
2645 ctx_clear_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA);
2646 grpquota = false;
2647 }
2648
2649 if (usr_qf_name || grp_qf_name) {
2650 if (usrquota || grpquota) {
2651 ext4_msg(NULL, KERN_ERR, "old and new quota "
2652 "format mixing");
2653 return -EINVAL;
2654 }
2655
2656 if (!(ctx->spec & EXT4_SPEC_JQFMT || sbi->s_jquota_fmt)) {
2657 ext4_msg(NULL, KERN_ERR, "journaled quota format "
2658 "not specified");
2659 return -EINVAL;
2660 }
2661 }
2662
2663 return 0;
2664
2665err_quota_change:
2666 ext4_msg(NULL, KERN_ERR,
2667 "Cannot change quota options when quota turned on");
2668 return -EINVAL;
2669err_jquota_change:
2670 ext4_msg(NULL, KERN_ERR, "Cannot change journaled quota "
2671 "options when quota turned on");
2672 return -EINVAL;
2673err_jquota_specified:
2674 ext4_msg(NULL, KERN_ERR, "%s quota file already specified",
2675 QTYPE2NAME(i));
2676 return -EINVAL;
2677#else
2678 return 0;
2679#endif
2680}
2681
2682static int ext4_check_test_dummy_encryption(const struct fs_context *fc,
2683 struct super_block *sb)
2684{
2685 const struct ext4_fs_context *ctx = fc->fs_private;
2686 const struct ext4_sb_info *sbi = EXT4_SB(sb);
2687
2688 if (!fscrypt_is_dummy_policy_set(dummy_policy: &ctx->dummy_enc_policy))
2689 return 0;
2690
2691 if (!ext4_has_feature_encrypt(sb)) {
2692 ext4_msg(NULL, KERN_WARNING,
2693 "test_dummy_encryption requires encrypt feature");
2694 return -EINVAL;
2695 }
2696 /*
2697 * This mount option is just for testing, and it's not worthwhile to
2698 * implement the extra complexity (e.g. RCU protection) that would be
2699 * needed to allow it to be set or changed during remount. We do allow
2700 * it to be specified during remount, but only if there is no change.
2701 */
2702 if (fc->purpose == FS_CONTEXT_FOR_RECONFIGURE) {
2703 if (fscrypt_dummy_policies_equal(p1: &sbi->s_dummy_enc_policy,
2704 p2: &ctx->dummy_enc_policy))
2705 return 0;
2706 ext4_msg(NULL, KERN_WARNING,
2707 "Can't set or change test_dummy_encryption on remount");
2708 return -EINVAL;
2709 }
2710 /* Also make sure s_mount_opts didn't contain a conflicting value. */
2711 if (fscrypt_is_dummy_policy_set(dummy_policy: &sbi->s_dummy_enc_policy)) {
2712 if (fscrypt_dummy_policies_equal(p1: &sbi->s_dummy_enc_policy,
2713 p2: &ctx->dummy_enc_policy))
2714 return 0;
2715 ext4_msg(NULL, KERN_WARNING,
2716 "Conflicting test_dummy_encryption options");
2717 return -EINVAL;
2718 }
2719 return 0;
2720}
2721
2722static void ext4_apply_test_dummy_encryption(struct ext4_fs_context *ctx,
2723 struct super_block *sb)
2724{
2725 if (!fscrypt_is_dummy_policy_set(dummy_policy: &ctx->dummy_enc_policy) ||
2726 /* if already set, it was already verified to be the same */
2727 fscrypt_is_dummy_policy_set(dummy_policy: &EXT4_SB(sb)->s_dummy_enc_policy))
2728 return;
2729 EXT4_SB(sb)->s_dummy_enc_policy = ctx->dummy_enc_policy;
2730 memset(s: &ctx->dummy_enc_policy, c: 0, n: sizeof(ctx->dummy_enc_policy));
2731 ext4_msg(sb, KERN_WARNING, "Test dummy encryption mode enabled");
2732}
2733
2734static int ext4_check_opt_consistency(struct fs_context *fc,
2735 struct super_block *sb)
2736{
2737 struct ext4_fs_context *ctx = fc->fs_private;
2738 struct ext4_sb_info *sbi = fc->s_fs_info;
2739 int is_remount = fc->purpose == FS_CONTEXT_FOR_RECONFIGURE;
2740 int err;
2741
2742 if ((ctx->opt_flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
2743 ext4_msg(NULL, KERN_ERR,
2744 "Mount option(s) incompatible with ext2");
2745 return -EINVAL;
2746 }
2747 if ((ctx->opt_flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
2748 ext4_msg(NULL, KERN_ERR,
2749 "Mount option(s) incompatible with ext3");
2750 return -EINVAL;
2751 }
2752
2753 if (ctx->s_want_extra_isize >
2754 (sbi->s_inode_size - EXT4_GOOD_OLD_INODE_SIZE)) {
2755 ext4_msg(NULL, KERN_ERR,
2756 "Invalid want_extra_isize %d",
2757 ctx->s_want_extra_isize);
2758 return -EINVAL;
2759 }
2760
2761 err = ext4_check_test_dummy_encryption(fc, sb);
2762 if (err)
2763 return err;
2764
2765 if ((ctx->spec & EXT4_SPEC_DATAJ) && is_remount) {
2766 if (!sbi->s_journal) {
2767 ext4_msg(NULL, KERN_WARNING,
2768 "Remounting file system with no journal "
2769 "so ignoring journalled data option");
2770 ctx_clear_mount_opt(ctx, EXT4_MOUNT_DATA_FLAGS);
2771 } else if (ctx_test_mount_opt(ctx, EXT4_MOUNT_DATA_FLAGS) !=
2772 test_opt(sb, DATA_FLAGS)) {
2773 ext4_msg(NULL, KERN_ERR, "Cannot change data mode "
2774 "on remount");
2775 return -EINVAL;
2776 }
2777 }
2778
2779 if (is_remount) {
2780 if (!sbi->s_journal &&
2781 ctx_test_mount_opt(ctx, EXT4_MOUNT_DATA_ERR_ABORT)) {
2782 ext4_msg(NULL, KERN_WARNING,
2783 "Remounting fs w/o journal so ignoring data_err option");
2784 ctx_clear_mount_opt(ctx, EXT4_MOUNT_DATA_ERR_ABORT);
2785 }
2786
2787 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS) &&
2788 (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)) {
2789 ext4_msg(NULL, KERN_ERR, "can't mount with "
2790 "both data=journal and dax");
2791 return -EINVAL;
2792 }
2793
2794 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS) &&
2795 (!(sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) ||
2796 (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER))) {
2797fail_dax_change_remount:
2798 ext4_msg(NULL, KERN_ERR, "can't change "
2799 "dax mount option while remounting");
2800 return -EINVAL;
2801 } else if (ctx_test_mount_opt2(ctx, EXT4_MOUNT2_DAX_NEVER) &&
2802 (!(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) ||
2803 (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS))) {
2804 goto fail_dax_change_remount;
2805 } else if (ctx_test_mount_opt2(ctx, EXT4_MOUNT2_DAX_INODE) &&
2806 ((sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) ||
2807 (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) ||
2808 !(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_INODE))) {
2809 goto fail_dax_change_remount;
2810 }
2811 }
2812
2813 return ext4_check_quota_consistency(fc, sb);
2814}
2815
2816static void ext4_apply_options(struct fs_context *fc, struct super_block *sb)
2817{
2818 struct ext4_fs_context *ctx = fc->fs_private;
2819 struct ext4_sb_info *sbi = fc->s_fs_info;
2820
2821 sbi->s_mount_opt &= ~ctx->mask_s_mount_opt;
2822 sbi->s_mount_opt |= ctx->vals_s_mount_opt;
2823 sbi->s_mount_opt2 &= ~ctx->mask_s_mount_opt2;
2824 sbi->s_mount_opt2 |= ctx->vals_s_mount_opt2;
2825 sb->s_flags &= ~ctx->mask_s_flags;
2826 sb->s_flags |= ctx->vals_s_flags;
2827
2828#define APPLY(X) ({ if (ctx->spec & EXT4_SPEC_##X) sbi->X = ctx->X; })
2829 APPLY(s_commit_interval);
2830 APPLY(s_stripe);
2831 APPLY(s_max_batch_time);
2832 APPLY(s_min_batch_time);
2833 APPLY(s_want_extra_isize);
2834 APPLY(s_inode_readahead_blks);
2835 APPLY(s_max_dir_size_kb);
2836 APPLY(s_li_wait_mult);
2837 APPLY(s_resgid);
2838 APPLY(s_resuid);
2839
2840#ifdef CONFIG_EXT4_DEBUG
2841 APPLY(s_fc_debug_max_replay);
2842#endif
2843
2844 ext4_apply_quota_options(fc, sb);
2845 ext4_apply_test_dummy_encryption(ctx, sb);
2846}
2847
2848
2849static int ext4_validate_options(struct fs_context *fc)
2850{
2851#ifdef CONFIG_QUOTA
2852 struct ext4_fs_context *ctx = fc->fs_private;
2853 char *usr_qf_name, *grp_qf_name;
2854
2855 usr_qf_name = ctx->s_qf_names[USRQUOTA];
2856 grp_qf_name = ctx->s_qf_names[GRPQUOTA];
2857
2858 if (usr_qf_name || grp_qf_name) {
2859 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_USRQUOTA) && usr_qf_name)
2860 ctx_clear_mount_opt(ctx, EXT4_MOUNT_USRQUOTA);
2861
2862 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA) && grp_qf_name)
2863 ctx_clear_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA);
2864
2865 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_USRQUOTA) ||
2866 ctx_test_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA)) {
2867 ext4_msg(NULL, KERN_ERR, "old and new quota "
2868 "format mixing");
2869 return -EINVAL;
2870 }
2871 }
2872#endif
2873 return 1;
2874}
2875
2876static inline void ext4_show_quota_options(struct seq_file *seq,
2877 struct super_block *sb)
2878{
2879#if defined(CONFIG_QUOTA)
2880 struct ext4_sb_info *sbi = EXT4_SB(sb);
2881 char *usr_qf_name, *grp_qf_name;
2882
2883 if (sbi->s_jquota_fmt) {
2884 char *fmtname = "";
2885
2886 switch (sbi->s_jquota_fmt) {
2887 case QFMT_VFS_OLD:
2888 fmtname = "vfsold";
2889 break;
2890 case QFMT_VFS_V0:
2891 fmtname = "vfsv0";
2892 break;
2893 case QFMT_VFS_V1:
2894 fmtname = "vfsv1";
2895 break;
2896 }
2897 seq_printf(m: seq, fmt: ",jqfmt=%s", fmtname);
2898 }
2899
2900 rcu_read_lock();
2901 usr_qf_name = rcu_dereference(sbi->s_qf_names[USRQUOTA]);
2902 grp_qf_name = rcu_dereference(sbi->s_qf_names[GRPQUOTA]);
2903 if (usr_qf_name)
2904 seq_show_option(m: seq, name: "usrjquota", value: usr_qf_name);
2905 if (grp_qf_name)
2906 seq_show_option(m: seq, name: "grpjquota", value: grp_qf_name);
2907 rcu_read_unlock();
2908#endif
2909}
2910
2911static const char *token2str(int token)
2912{
2913 const struct fs_parameter_spec *spec;
2914
2915 for (spec = ext4_param_specs; spec->name != NULL; spec++)
2916 if (spec->opt == token && !spec->type)
2917 break;
2918 return spec->name;
2919}
2920
2921/*
2922 * Show an option if
2923 * - it's set to a non-default value OR
2924 * - if the per-sb default is different from the global default
2925 */
2926static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
2927 int nodefs)
2928{
2929 struct ext4_sb_info *sbi = EXT4_SB(sb);
2930 struct ext4_super_block *es = sbi->s_es;
2931 int def_errors;
2932 const struct mount_opts *m;
2933 char sep = nodefs ? '\n' : ',';
2934
2935#define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
2936#define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2937
2938 if (sbi->s_sb_block != 1)
2939 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
2940
2941 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
2942 int want_set = m->flags & MOPT_SET;
2943 int opt_2 = m->flags & MOPT_2;
2944 unsigned int mount_opt, def_mount_opt;
2945
2946 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
2947 m->flags & MOPT_SKIP)
2948 continue;
2949
2950 if (opt_2) {
2951 mount_opt = sbi->s_mount_opt2;
2952 def_mount_opt = sbi->s_def_mount_opt2;
2953 } else {
2954 mount_opt = sbi->s_mount_opt;
2955 def_mount_opt = sbi->s_def_mount_opt;
2956 }
2957 /* skip if same as the default */
2958 if (!nodefs && !(m->mount_opt & (mount_opt ^ def_mount_opt)))
2959 continue;
2960 /* select Opt_noFoo vs Opt_Foo */
2961 if ((want_set &&
2962 (mount_opt & m->mount_opt) != m->mount_opt) ||
2963 (!want_set && (mount_opt & m->mount_opt)))
2964 continue;
2965 SEQ_OPTS_PRINT("%s", token2str(m->token));
2966 }
2967
2968 if (nodefs || !uid_eq(left: sbi->s_resuid, right: make_kuid(from: &init_user_ns, EXT4_DEF_RESUID)) ||
2969 ext4_get_resuid(es) != EXT4_DEF_RESUID)
2970 SEQ_OPTS_PRINT("resuid=%u",
2971 from_kuid_munged(&init_user_ns, sbi->s_resuid));
2972 if (nodefs || !gid_eq(left: sbi->s_resgid, right: make_kgid(from: &init_user_ns, EXT4_DEF_RESGID)) ||
2973 ext4_get_resgid(es) != EXT4_DEF_RESGID)
2974 SEQ_OPTS_PRINT("resgid=%u",
2975 from_kgid_munged(&init_user_ns, sbi->s_resgid));
2976 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
2977 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
2978 SEQ_OPTS_PUTS("errors=remount-ro");
2979 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
2980 SEQ_OPTS_PUTS("errors=continue");
2981 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
2982 SEQ_OPTS_PUTS("errors=panic");
2983 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
2984 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
2985 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
2986 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
2987 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
2988 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2989 if (nodefs && sb->s_flags & SB_I_VERSION)
2990 SEQ_OPTS_PUTS("i_version");
2991 if (nodefs || sbi->s_stripe)
2992 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
2993 if (nodefs || EXT4_MOUNT_DATA_FLAGS &
2994 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
2995 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2996 SEQ_OPTS_PUTS("data=journal");
2997 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2998 SEQ_OPTS_PUTS("data=ordered");
2999 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
3000 SEQ_OPTS_PUTS("data=writeback");
3001 }
3002 if (nodefs ||
3003 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
3004 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
3005 sbi->s_inode_readahead_blks);
3006
3007 if (test_opt(sb, INIT_INODE_TABLE) && (nodefs ||
3008 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
3009 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
3010 if (nodefs || sbi->s_max_dir_size_kb)
3011 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
3012 if (test_opt(sb, DATA_ERR_ABORT))
3013 SEQ_OPTS_PUTS("data_err=abort");
3014
3015 fscrypt_show_test_dummy_encryption(seq, sep, sb);
3016
3017 if (sb->s_flags & SB_INLINECRYPT)
3018 SEQ_OPTS_PUTS("inlinecrypt");
3019
3020 if (test_opt(sb, DAX_ALWAYS)) {
3021 if (IS_EXT2_SB(sb))
3022 SEQ_OPTS_PUTS("dax");
3023 else
3024 SEQ_OPTS_PUTS("dax=always");
3025 } else if (test_opt2(sb, DAX_NEVER)) {
3026 SEQ_OPTS_PUTS("dax=never");
3027 } else if (test_opt2(sb, DAX_INODE)) {
3028 SEQ_OPTS_PUTS("dax=inode");
3029 }
3030
3031 if (sbi->s_groups_count >= MB_DEFAULT_LINEAR_SCAN_THRESHOLD &&
3032 !test_opt2(sb, MB_OPTIMIZE_SCAN)) {
3033 SEQ_OPTS_PUTS("mb_optimize_scan=0");
3034 } else if (sbi->s_groups_count < MB_DEFAULT_LINEAR_SCAN_THRESHOLD &&
3035 test_opt2(sb, MB_OPTIMIZE_SCAN)) {
3036 SEQ_OPTS_PUTS("mb_optimize_scan=1");
3037 }
3038
3039 if (nodefs && !test_opt(sb, NO_PREFETCH_BLOCK_BITMAPS))
3040 SEQ_OPTS_PUTS("prefetch_block_bitmaps");
3041
3042 if (ext4_emergency_ro(sb))
3043 SEQ_OPTS_PUTS("emergency_ro");
3044
3045 if (ext4_forced_shutdown(sb))
3046 SEQ_OPTS_PUTS("shutdown");
3047
3048 ext4_show_quota_options(seq, sb);
3049 return 0;
3050}
3051
3052static int ext4_show_options(struct seq_file *seq, struct dentry *root)
3053{
3054 return _ext4_show_options(seq, sb: root->d_sb, nodefs: 0);
3055}
3056
3057int ext4_seq_options_show(struct seq_file *seq, void *offset)
3058{
3059 struct super_block *sb = seq->private;
3060 int rc;
3061
3062 seq_puts(m: seq, s: sb_rdonly(sb) ? "ro" : "rw");
3063 rc = _ext4_show_options(seq, sb, nodefs: 1);
3064 seq_putc(m: seq, c: '\n');
3065 return rc;
3066}
3067
3068static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
3069 int read_only)
3070{
3071 struct ext4_sb_info *sbi = EXT4_SB(sb);
3072 int err = 0;
3073
3074 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
3075 ext4_msg(sb, KERN_ERR, "revision level too high, "
3076 "forcing read-only mode");
3077 err = -EROFS;
3078 goto done;
3079 }
3080 if (read_only)
3081 goto done;
3082 if (!(sbi->s_mount_state & EXT4_VALID_FS))
3083 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
3084 "running e2fsck is recommended");
3085 else if (sbi->s_mount_state & EXT4_ERROR_FS)
3086 ext4_msg(sb, KERN_WARNING,
3087 "warning: mounting fs with errors, "
3088 "running e2fsck is recommended");
3089 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
3090 le16_to_cpu(es->s_mnt_count) >=
3091 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
3092 ext4_msg(sb, KERN_WARNING,
3093 "warning: maximal mount count reached, "
3094 "running e2fsck is recommended");
3095 else if (le32_to_cpu(es->s_checkinterval) &&
3096 (ext4_get_tstamp(es, s_lastcheck) +
3097 le32_to_cpu(es->s_checkinterval) <= ktime_get_real_seconds()))
3098 ext4_msg(sb, KERN_WARNING,
3099 "warning: checktime reached, "
3100 "running e2fsck is recommended");
3101 if (!sbi->s_journal)
3102 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
3103 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
3104 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
3105 le16_add_cpu(var: &es->s_mnt_count, val: 1);
3106 ext4_update_tstamp(es, s_mtime);
3107 if (sbi->s_journal) {
3108 ext4_set_feature_journal_needs_recovery(sb);
3109 if (ext4_has_feature_orphan_file(sb))
3110 ext4_set_feature_orphan_present(sb);
3111 }
3112
3113 err = ext4_commit_super(sb);
3114done:
3115 if (test_opt(sb, DEBUG))
3116 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
3117 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
3118 sb->s_blocksize,
3119 sbi->s_groups_count,
3120 EXT4_BLOCKS_PER_GROUP(sb),
3121 EXT4_INODES_PER_GROUP(sb),
3122 sbi->s_mount_opt, sbi->s_mount_opt2);
3123 return err;
3124}
3125
3126int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
3127{
3128 struct ext4_sb_info *sbi = EXT4_SB(sb);
3129 struct flex_groups **old_groups, **new_groups;
3130 int size, i, j;
3131
3132 if (!sbi->s_log_groups_per_flex)
3133 return 0;
3134
3135 size = ext4_flex_group(sbi, block_group: ngroup - 1) + 1;
3136 if (size <= sbi->s_flex_groups_allocated)
3137 return 0;
3138
3139 new_groups = kvzalloc(roundup_pow_of_two(size *
3140 sizeof(*sbi->s_flex_groups)), GFP_KERNEL);
3141 if (!new_groups) {
3142 ext4_msg(sb, KERN_ERR,
3143 "not enough memory for %d flex group pointers", size);
3144 return -ENOMEM;
3145 }
3146 for (i = sbi->s_flex_groups_allocated; i < size; i++) {
3147 new_groups[i] = kvzalloc(roundup_pow_of_two(
3148 sizeof(struct flex_groups)),
3149 GFP_KERNEL);
3150 if (!new_groups[i]) {
3151 for (j = sbi->s_flex_groups_allocated; j < i; j++)
3152 kvfree(addr: new_groups[j]);
3153 kvfree(addr: new_groups);
3154 ext4_msg(sb, KERN_ERR,
3155 "not enough memory for %d flex groups", size);
3156 return -ENOMEM;
3157 }
3158 }
3159 rcu_read_lock();
3160 old_groups = rcu_dereference(sbi->s_flex_groups);
3161 if (old_groups)
3162 memcpy(to: new_groups, from: old_groups,
3163 len: (sbi->s_flex_groups_allocated *
3164 sizeof(struct flex_groups *)));
3165 rcu_read_unlock();
3166 rcu_assign_pointer(sbi->s_flex_groups, new_groups);
3167 sbi->s_flex_groups_allocated = size;
3168 if (old_groups)
3169 ext4_kvfree_array_rcu(to_free: old_groups);
3170 return 0;
3171}
3172
3173static int ext4_fill_flex_info(struct super_block *sb)
3174{
3175 struct ext4_sb_info *sbi = EXT4_SB(sb);
3176 struct ext4_group_desc *gdp = NULL;
3177 struct flex_groups *fg;
3178 ext4_group_t flex_group;
3179 int i, err;
3180
3181 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
3182 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
3183 sbi->s_log_groups_per_flex = 0;
3184 return 1;
3185 }
3186
3187 err = ext4_alloc_flex_bg_array(sb, ngroup: sbi->s_groups_count);
3188 if (err)
3189 goto failed;
3190
3191 for (i = 0; i < sbi->s_groups_count; i++) {
3192 gdp = ext4_get_group_desc(sb, block_group: i, NULL);
3193
3194 flex_group = ext4_flex_group(sbi, block_group: i);
3195 fg = sbi_array_rcu_deref(sbi, s_flex_groups, flex_group);
3196 atomic_add(i: ext4_free_inodes_count(sb, bg: gdp), v: &fg->free_inodes);
3197 atomic64_add(i: ext4_free_group_clusters(sb, bg: gdp),
3198 v: &fg->free_clusters);
3199 atomic_add(i: ext4_used_dirs_count(sb, bg: gdp), v: &fg->used_dirs);
3200 }
3201
3202 return 1;
3203failed:
3204 return 0;
3205}
3206
3207static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
3208 struct ext4_group_desc *gdp)
3209{
3210 int offset = offsetof(struct ext4_group_desc, bg_checksum);
3211 __u16 crc = 0;
3212 __le32 le_group = cpu_to_le32(block_group);
3213 struct ext4_sb_info *sbi = EXT4_SB(sb);
3214
3215 if (ext4_has_feature_metadata_csum(sb: sbi->s_sb)) {
3216 /* Use new metadata_csum algorithm */
3217 __u32 csum32;
3218 __u16 dummy_csum = 0;
3219
3220 csum32 = ext4_chksum(crc: sbi->s_csum_seed, address: (__u8 *)&le_group,
3221 length: sizeof(le_group));
3222 csum32 = ext4_chksum(crc: csum32, address: (__u8 *)gdp, length: offset);
3223 csum32 = ext4_chksum(crc: csum32, address: (__u8 *)&dummy_csum,
3224 length: sizeof(dummy_csum));
3225 offset += sizeof(dummy_csum);
3226 if (offset < sbi->s_desc_size)
3227 csum32 = ext4_chksum(crc: csum32, address: (__u8 *)gdp + offset,
3228 length: sbi->s_desc_size - offset);
3229
3230 crc = csum32 & 0xFFFF;
3231 goto out;
3232 }
3233
3234 /* old crc16 code */
3235 if (!ext4_has_feature_gdt_csum(sb))
3236 return 0;
3237
3238 crc = crc16(crc: ~0, p: sbi->s_es->s_uuid, len: sizeof(sbi->s_es->s_uuid));
3239 crc = crc16(crc, p: (__u8 *)&le_group, len: sizeof(le_group));
3240 crc = crc16(crc, p: (__u8 *)gdp, len: offset);
3241 offset += sizeof(gdp->bg_checksum); /* skip checksum */
3242 /* for checksum of struct ext4_group_desc do the rest...*/
3243 if (ext4_has_feature_64bit(sb) && offset < sbi->s_desc_size)
3244 crc = crc16(crc, p: (__u8 *)gdp + offset,
3245 len: sbi->s_desc_size - offset);
3246
3247out:
3248 return cpu_to_le16(crc);
3249}
3250
3251int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
3252 struct ext4_group_desc *gdp)
3253{
3254 if (ext4_has_group_desc_csum(sb) &&
3255 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
3256 return 0;
3257
3258 return 1;
3259}
3260
3261void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
3262 struct ext4_group_desc *gdp)
3263{
3264 if (!ext4_has_group_desc_csum(sb))
3265 return;
3266 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
3267}
3268
3269/* Called at mount-time, super-block is locked */
3270static int ext4_check_descriptors(struct super_block *sb,
3271 ext4_fsblk_t sb_block,
3272 ext4_group_t *first_not_zeroed)
3273{
3274 struct ext4_sb_info *sbi = EXT4_SB(sb);
3275 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
3276 ext4_fsblk_t last_block;
3277 ext4_fsblk_t last_bg_block = sb_block + ext4_bg_num_gdb(sb, group: 0);
3278 ext4_fsblk_t block_bitmap;
3279 ext4_fsblk_t inode_bitmap;
3280 ext4_fsblk_t inode_table;
3281 int flexbg_flag = 0;
3282 ext4_group_t i, grp = sbi->s_groups_count;
3283
3284 if (ext4_has_feature_flex_bg(sb))
3285 flexbg_flag = 1;
3286
3287 ext4_debug("Checking group descriptors");
3288
3289 for (i = 0; i < sbi->s_groups_count; i++) {
3290 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, block_group: i, NULL);
3291
3292 if (i == sbi->s_groups_count - 1 || flexbg_flag)
3293 last_block = ext4_blocks_count(es: sbi->s_es) - 1;
3294 else
3295 last_block = first_block +
3296 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
3297
3298 if ((grp == sbi->s_groups_count) &&
3299 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3300 grp = i;
3301
3302 block_bitmap = ext4_block_bitmap(sb, bg: gdp);
3303 if (block_bitmap == sb_block) {
3304 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3305 "Block bitmap for group %u overlaps "
3306 "superblock", i);
3307 if (!sb_rdonly(sb))
3308 return 0;
3309 }
3310 if (block_bitmap >= sb_block + 1 &&
3311 block_bitmap <= last_bg_block) {
3312 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3313 "Block bitmap for group %u overlaps "
3314 "block group descriptors", i);
3315 if (!sb_rdonly(sb))
3316 return 0;
3317 }
3318 if (block_bitmap < first_block || block_bitmap > last_block) {
3319 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3320 "Block bitmap for group %u not in group "
3321 "(block %llu)!", i, block_bitmap);
3322 return 0;
3323 }
3324 inode_bitmap = ext4_inode_bitmap(sb, bg: gdp);
3325 if (inode_bitmap == sb_block) {
3326 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3327 "Inode bitmap for group %u overlaps "
3328 "superblock", i);
3329 if (!sb_rdonly(sb))
3330 return 0;
3331 }
3332 if (inode_bitmap >= sb_block + 1 &&
3333 inode_bitmap <= last_bg_block) {
3334 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3335 "Inode bitmap for group %u overlaps "
3336 "block group descriptors", i);
3337 if (!sb_rdonly(sb))
3338 return 0;
3339 }
3340 if (inode_bitmap < first_block || inode_bitmap > last_block) {
3341 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3342 "Inode bitmap for group %u not in group "
3343 "(block %llu)!", i, inode_bitmap);
3344 return 0;
3345 }
3346 inode_table = ext4_inode_table(sb, bg: gdp);
3347 if (inode_table == sb_block) {
3348 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3349 "Inode table for group %u overlaps "
3350 "superblock", i);
3351 if (!sb_rdonly(sb))
3352 return 0;
3353 }
3354 if (inode_table >= sb_block + 1 &&
3355 inode_table <= last_bg_block) {
3356 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3357 "Inode table for group %u overlaps "
3358 "block group descriptors", i);
3359 if (!sb_rdonly(sb))
3360 return 0;
3361 }
3362 if (inode_table < first_block ||
3363 inode_table + sbi->s_itb_per_group - 1 > last_block) {
3364 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3365 "Inode table for group %u not in group "
3366 "(block %llu)!", i, inode_table);
3367 return 0;
3368 }
3369 ext4_lock_group(sb, group: i);
3370 if (!ext4_group_desc_csum_verify(sb, block_group: i, gdp)) {
3371 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3372 "Checksum for group %u failed (%u!=%u)",
3373 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
3374 gdp)), le16_to_cpu(gdp->bg_checksum));
3375 if (!sb_rdonly(sb)) {
3376 ext4_unlock_group(sb, group: i);
3377 return 0;
3378 }
3379 }
3380 ext4_unlock_group(sb, group: i);
3381 if (!flexbg_flag)
3382 first_block += EXT4_BLOCKS_PER_GROUP(sb);
3383 }
3384 if (NULL != first_not_zeroed)
3385 *first_not_zeroed = grp;
3386 return 1;
3387}
3388
3389/*
3390 * Maximal extent format file size.
3391 * Resulting logical blkno at s_maxbytes must fit in our on-disk
3392 * extent format containers, within a sector_t, and within i_blocks
3393 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
3394 * so that won't be a limiting factor.
3395 *
3396 * However there is other limiting factor. We do store extents in the form
3397 * of starting block and length, hence the resulting length of the extent
3398 * covering maximum file size must fit into on-disk format containers as
3399 * well. Given that length is always by 1 unit bigger than max unit (because
3400 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
3401 *
3402 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
3403 */
3404static loff_t ext4_max_size(int blkbits, int has_huge_files)
3405{
3406 loff_t res;
3407 loff_t upper_limit = MAX_LFS_FILESIZE;
3408
3409 BUILD_BUG_ON(sizeof(blkcnt_t) < sizeof(u64));
3410
3411 if (!has_huge_files) {
3412 upper_limit = (1LL << 32) - 1;
3413
3414 /* total blocks in file system block size */
3415 upper_limit >>= (blkbits - 9);
3416 upper_limit <<= blkbits;
3417 }
3418
3419 /*
3420 * 32-bit extent-start container, ee_block. We lower the maxbytes
3421 * by one fs block, so ee_len can cover the extent of maximum file
3422 * size
3423 */
3424 res = (1LL << 32) - 1;
3425 res <<= blkbits;
3426
3427 /* Sanity check against vm- & vfs- imposed limits */
3428 if (res > upper_limit)
3429 res = upper_limit;
3430
3431 return res;
3432}
3433
3434/*
3435 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
3436 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
3437 * We need to be 1 filesystem block less than the 2^48 sector limit.
3438 */
3439static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
3440{
3441 loff_t upper_limit, res = EXT4_NDIR_BLOCKS;
3442 int meta_blocks;
3443 unsigned int ppb = 1 << (bits - 2);
3444
3445 /*
3446 * This is calculated to be the largest file size for a dense, block
3447 * mapped file such that the file's total number of 512-byte sectors,
3448 * including data and all indirect blocks, does not exceed (2^48 - 1).
3449 *
3450 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
3451 * number of 512-byte sectors of the file.
3452 */
3453 if (!has_huge_files) {
3454 /*
3455 * !has_huge_files or implies that the inode i_block field
3456 * represents total file blocks in 2^32 512-byte sectors ==
3457 * size of vfs inode i_blocks * 8
3458 */
3459 upper_limit = (1LL << 32) - 1;
3460
3461 /* total blocks in file system block size */
3462 upper_limit >>= (bits - 9);
3463
3464 } else {
3465 /*
3466 * We use 48 bit ext4_inode i_blocks
3467 * With EXT4_HUGE_FILE_FL set the i_blocks
3468 * represent total number of blocks in
3469 * file system block size
3470 */
3471 upper_limit = (1LL << 48) - 1;
3472
3473 }
3474
3475 /* Compute how many blocks we can address by block tree */
3476 res += ppb;
3477 res += ppb * ppb;
3478 res += ((loff_t)ppb) * ppb * ppb;
3479 /* Compute how many metadata blocks are needed */
3480 meta_blocks = 1;
3481 meta_blocks += 1 + ppb;
3482 meta_blocks += 1 + ppb + ppb * ppb;
3483 /* Does block tree limit file size? */
3484 if (res + meta_blocks <= upper_limit)
3485 goto check_lfs;
3486
3487 res = upper_limit;
3488 /* How many metadata blocks are needed for addressing upper_limit? */
3489 upper_limit -= EXT4_NDIR_BLOCKS;
3490 /* indirect blocks */
3491 meta_blocks = 1;
3492 upper_limit -= ppb;
3493 /* double indirect blocks */
3494 if (upper_limit < ppb * ppb) {
3495 meta_blocks += 1 + DIV_ROUND_UP_ULL(upper_limit, ppb);
3496 res -= meta_blocks;
3497 goto check_lfs;
3498 }
3499 meta_blocks += 1 + ppb;
3500 upper_limit -= ppb * ppb;
3501 /* tripple indirect blocks for the rest */
3502 meta_blocks += 1 + DIV_ROUND_UP_ULL(upper_limit, ppb) +
3503 DIV_ROUND_UP_ULL(upper_limit, ppb*ppb);
3504 res -= meta_blocks;
3505check_lfs:
3506 res <<= bits;
3507 if (res > MAX_LFS_FILESIZE)
3508 res = MAX_LFS_FILESIZE;
3509
3510 return res;
3511}
3512
3513static ext4_fsblk_t descriptor_loc(struct super_block *sb,
3514 ext4_fsblk_t logical_sb_block, int nr)
3515{
3516 struct ext4_sb_info *sbi = EXT4_SB(sb);
3517 ext4_group_t bg, first_meta_bg;
3518 int has_super = 0;
3519
3520 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
3521
3522 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
3523 return logical_sb_block + nr + 1;
3524 bg = sbi->s_desc_per_block * nr;
3525 if (ext4_bg_has_super(sb, group: bg))
3526 has_super = 1;
3527
3528 /*
3529 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
3530 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
3531 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
3532 * compensate.
3533 */
3534 if (sb->s_blocksize == 1024 && nr == 0 &&
3535 le32_to_cpu(sbi->s_es->s_first_data_block) == 0)
3536 has_super++;
3537
3538 return (has_super + ext4_group_first_block_no(sb, group_no: bg));
3539}
3540
3541/**
3542 * ext4_get_stripe_size: Get the stripe size.
3543 * @sbi: In memory super block info
3544 *
3545 * If we have specified it via mount option, then
3546 * use the mount option value. If the value specified at mount time is
3547 * greater than the blocks per group use the super block value.
3548 * If the super block value is greater than blocks per group return 0.
3549 * Allocator needs it be less than blocks per group.
3550 *
3551 */
3552static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
3553{
3554 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
3555 unsigned long stripe_width =
3556 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3557 int ret;
3558
3559 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3560 ret = sbi->s_stripe;
3561 else if (stripe_width && stripe_width <= sbi->s_blocks_per_group)
3562 ret = stripe_width;
3563 else if (stride && stride <= sbi->s_blocks_per_group)
3564 ret = stride;
3565 else
3566 ret = 0;
3567
3568 /*
3569 * If the stripe width is 1, this makes no sense and
3570 * we set it to 0 to turn off stripe handling code.
3571 */
3572 if (ret <= 1)
3573 ret = 0;
3574
3575 return ret;
3576}
3577
3578/*
3579 * Check whether this filesystem can be mounted based on
3580 * the features present and the RDONLY/RDWR mount requested.
3581 * Returns 1 if this filesystem can be mounted as requested,
3582 * 0 if it cannot be.
3583 */
3584int ext4_feature_set_ok(struct super_block *sb, int readonly)
3585{
3586 if (ext4_has_unknown_ext4_incompat_features(sb)) {
3587 ext4_msg(sb, KERN_ERR,
3588 "Couldn't mount because of "
3589 "unsupported optional features (%x)",
3590 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
3591 ~EXT4_FEATURE_INCOMPAT_SUPP));
3592 return 0;
3593 }
3594
3595 if (!IS_ENABLED(CONFIG_UNICODE) && ext4_has_feature_casefold(sb)) {
3596 ext4_msg(sb, KERN_ERR,
3597 "Filesystem with casefold feature cannot be "
3598 "mounted without CONFIG_UNICODE");
3599 return 0;
3600 }
3601
3602 if (readonly)
3603 return 1;
3604
3605 if (ext4_has_feature_readonly(sb)) {
3606 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
3607 sb->s_flags |= SB_RDONLY;
3608 return 1;
3609 }
3610
3611 /* Check that feature set is OK for a read-write mount */
3612 if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
3613 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
3614 "unsupported optional features (%x)",
3615 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
3616 ~EXT4_FEATURE_RO_COMPAT_SUPP));
3617 return 0;
3618 }
3619 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
3620 ext4_msg(sb, KERN_ERR,
3621 "Can't support bigalloc feature without "
3622 "extents feature\n");
3623 return 0;
3624 }
3625
3626#if !IS_ENABLED(CONFIG_QUOTA) || !IS_ENABLED(CONFIG_QFMT_V2)
3627 if (!readonly && (ext4_has_feature_quota(sb) ||
3628 ext4_has_feature_project(sb))) {
3629 ext4_msg(sb, KERN_ERR,
3630 "The kernel was not built with CONFIG_QUOTA and CONFIG_QFMT_V2");
3631 return 0;
3632 }
3633#endif /* CONFIG_QUOTA */
3634 return 1;
3635}
3636
3637/*
3638 * This function is called once a day if we have errors logged
3639 * on the file system
3640 */
3641static void print_daily_error_info(struct timer_list *t)
3642{
3643 struct ext4_sb_info *sbi = timer_container_of(sbi, t, s_err_report);
3644 struct super_block *sb = sbi->s_sb;
3645 struct ext4_super_block *es = sbi->s_es;
3646
3647 if (es->s_error_count)
3648 /* fsck newer than v1.41.13 is needed to clean this condition. */
3649 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
3650 le32_to_cpu(es->s_error_count));
3651 if (es->s_first_error_time) {
3652 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %llu: %.*s:%d",
3653 sb->s_id,
3654 ext4_get_tstamp(es, s_first_error_time),
3655 (int) sizeof(es->s_first_error_func),
3656 es->s_first_error_func,
3657 le32_to_cpu(es->s_first_error_line));
3658 if (es->s_first_error_ino)
3659 printk(KERN_CONT ": inode %u",
3660 le32_to_cpu(es->s_first_error_ino));
3661 if (es->s_first_error_block)
3662 printk(KERN_CONT ": block %llu", (unsigned long long)
3663 le64_to_cpu(es->s_first_error_block));
3664 printk(KERN_CONT "\n");
3665 }
3666 if (es->s_last_error_time) {
3667 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %llu: %.*s:%d",
3668 sb->s_id,
3669 ext4_get_tstamp(es, s_last_error_time),
3670 (int) sizeof(es->s_last_error_func),
3671 es->s_last_error_func,
3672 le32_to_cpu(es->s_last_error_line));
3673 if (es->s_last_error_ino)
3674 printk(KERN_CONT ": inode %u",
3675 le32_to_cpu(es->s_last_error_ino));
3676 if (es->s_last_error_block)
3677 printk(KERN_CONT ": block %llu", (unsigned long long)
3678 le64_to_cpu(es->s_last_error_block));
3679 printk(KERN_CONT "\n");
3680 }
3681 mod_timer(timer: &sbi->s_err_report, expires: jiffies + 24*60*60*HZ); /* Once a day */
3682}
3683
3684/* Find next suitable group and run ext4_init_inode_table */
3685static int ext4_run_li_request(struct ext4_li_request *elr)
3686{
3687 struct ext4_group_desc *gdp = NULL;
3688 struct super_block *sb = elr->lr_super;
3689 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3690 ext4_group_t group = elr->lr_next_group;
3691 unsigned int prefetch_ios = 0;
3692 int ret = 0;
3693 int nr = EXT4_SB(sb)->s_mb_prefetch;
3694 u64 start_time;
3695
3696 if (elr->lr_mode == EXT4_LI_MODE_PREFETCH_BBITMAP) {
3697 elr->lr_next_group = ext4_mb_prefetch(sb, group, nr, cnt: &prefetch_ios);
3698 ext4_mb_prefetch_fini(sb, group: elr->lr_next_group, nr);
3699 trace_ext4_prefetch_bitmaps(sb, group, next: elr->lr_next_group, prefetch_ios: nr);
3700 if (group >= elr->lr_next_group) {
3701 ret = 1;
3702 if (elr->lr_first_not_zeroed != ngroups &&
3703 !ext4_emergency_state(sb) && !sb_rdonly(sb) &&
3704 test_opt(sb, INIT_INODE_TABLE)) {
3705 elr->lr_next_group = elr->lr_first_not_zeroed;
3706 elr->lr_mode = EXT4_LI_MODE_ITABLE;
3707 ret = 0;
3708 }
3709 }
3710 return ret;
3711 }
3712
3713 for (; group < ngroups; group++) {
3714 gdp = ext4_get_group_desc(sb, block_group: group, NULL);
3715 if (!gdp) {
3716 ret = 1;
3717 break;
3718 }
3719
3720 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3721 break;
3722 }
3723
3724 if (group >= ngroups)
3725 ret = 1;
3726
3727 if (!ret) {
3728 start_time = ktime_get_ns();
3729 ret = ext4_init_inode_table(sb, group,
3730 barrier: elr->lr_timeout ? 0 : 1);
3731 trace_ext4_lazy_itable_init(sb, group);
3732 if (elr->lr_timeout == 0) {
3733 elr->lr_timeout = nsecs_to_jiffies(n: (ktime_get_ns() - start_time) *
3734 EXT4_SB(sb: elr->lr_super)->s_li_wait_mult);
3735 }
3736 elr->lr_next_sched = jiffies + elr->lr_timeout;
3737 elr->lr_next_group = group + 1;
3738 }
3739 return ret;
3740}
3741
3742/*
3743 * Remove lr_request from the list_request and free the
3744 * request structure. Should be called with li_list_mtx held
3745 */
3746static void ext4_remove_li_request(struct ext4_li_request *elr)
3747{
3748 if (!elr)
3749 return;
3750
3751 list_del(entry: &elr->lr_request);
3752 EXT4_SB(sb: elr->lr_super)->s_li_request = NULL;
3753 kfree(objp: elr);
3754}
3755
3756static void ext4_unregister_li_request(struct super_block *sb)
3757{
3758 mutex_lock(lock: &ext4_li_mtx);
3759 if (!ext4_li_info) {
3760 mutex_unlock(lock: &ext4_li_mtx);
3761 return;
3762 }
3763
3764 mutex_lock(lock: &ext4_li_info->li_list_mtx);
3765 ext4_remove_li_request(elr: EXT4_SB(sb)->s_li_request);
3766 mutex_unlock(lock: &ext4_li_info->li_list_mtx);
3767 mutex_unlock(lock: &ext4_li_mtx);
3768}
3769
3770static struct task_struct *ext4_lazyinit_task;
3771
3772/*
3773 * This is the function where ext4lazyinit thread lives. It walks
3774 * through the request list searching for next scheduled filesystem.
3775 * When such a fs is found, run the lazy initialization request
3776 * (ext4_rn_li_request) and keep track of the time spend in this
3777 * function. Based on that time we compute next schedule time of
3778 * the request. When walking through the list is complete, compute
3779 * next waking time and put itself into sleep.
3780 */
3781static int ext4_lazyinit_thread(void *arg)
3782{
3783 struct ext4_lazy_init *eli = arg;
3784 struct list_head *pos, *n;
3785 struct ext4_li_request *elr;
3786 unsigned long next_wakeup, cur;
3787
3788 BUG_ON(NULL == eli);
3789 set_freezable();
3790
3791cont_thread:
3792 while (true) {
3793 bool next_wakeup_initialized = false;
3794
3795 next_wakeup = 0;
3796 mutex_lock(lock: &eli->li_list_mtx);
3797 if (list_empty(head: &eli->li_request_list)) {
3798 mutex_unlock(lock: &eli->li_list_mtx);
3799 goto exit_thread;
3800 }
3801 list_for_each_safe(pos, n, &eli->li_request_list) {
3802 int err = 0;
3803 int progress = 0;
3804 elr = list_entry(pos, struct ext4_li_request,
3805 lr_request);
3806
3807 if (time_before(jiffies, elr->lr_next_sched)) {
3808 if (!next_wakeup_initialized ||
3809 time_before(elr->lr_next_sched, next_wakeup)) {
3810 next_wakeup = elr->lr_next_sched;
3811 next_wakeup_initialized = true;
3812 }
3813 continue;
3814 }
3815 if (down_read_trylock(sem: &elr->lr_super->s_umount)) {
3816 if (sb_start_write_trylock(sb: elr->lr_super)) {
3817 progress = 1;
3818 /*
3819 * We hold sb->s_umount, sb can not
3820 * be removed from the list, it is
3821 * now safe to drop li_list_mtx
3822 */
3823 mutex_unlock(lock: &eli->li_list_mtx);
3824 err = ext4_run_li_request(elr);
3825 sb_end_write(sb: elr->lr_super);
3826 mutex_lock(lock: &eli->li_list_mtx);
3827 n = pos->next;
3828 }
3829 up_read(sem: (&elr->lr_super->s_umount));
3830 }
3831 /* error, remove the lazy_init job */
3832 if (err) {
3833 ext4_remove_li_request(elr);
3834 continue;
3835 }
3836 if (!progress) {
3837 elr->lr_next_sched = jiffies +
3838 get_random_u32_below(EXT4_DEF_LI_MAX_START_DELAY * HZ);
3839 }
3840 if (!next_wakeup_initialized ||
3841 time_before(elr->lr_next_sched, next_wakeup)) {
3842 next_wakeup = elr->lr_next_sched;
3843 next_wakeup_initialized = true;
3844 }
3845 }
3846 mutex_unlock(lock: &eli->li_list_mtx);
3847
3848 try_to_freeze();
3849
3850 cur = jiffies;
3851 if (!next_wakeup_initialized || time_after_eq(cur, next_wakeup)) {
3852 cond_resched();
3853 continue;
3854 }
3855
3856 schedule_timeout_interruptible(timeout: next_wakeup - cur);
3857
3858 if (kthread_should_stop()) {
3859 ext4_clear_request_list();
3860 goto exit_thread;
3861 }
3862 }
3863
3864exit_thread:
3865 /*
3866 * It looks like the request list is empty, but we need
3867 * to check it under the li_list_mtx lock, to prevent any
3868 * additions into it, and of course we should lock ext4_li_mtx
3869 * to atomically free the list and ext4_li_info, because at
3870 * this point another ext4 filesystem could be registering
3871 * new one.
3872 */
3873 mutex_lock(lock: &ext4_li_mtx);
3874 mutex_lock(lock: &eli->li_list_mtx);
3875 if (!list_empty(head: &eli->li_request_list)) {
3876 mutex_unlock(lock: &eli->li_list_mtx);
3877 mutex_unlock(lock: &ext4_li_mtx);
3878 goto cont_thread;
3879 }
3880 mutex_unlock(lock: &eli->li_list_mtx);
3881 kfree(objp: ext4_li_info);
3882 ext4_li_info = NULL;
3883 mutex_unlock(lock: &ext4_li_mtx);
3884
3885 return 0;
3886}
3887
3888static void ext4_clear_request_list(void)
3889{
3890 struct list_head *pos, *n;
3891 struct ext4_li_request *elr;
3892
3893 mutex_lock(lock: &ext4_li_info->li_list_mtx);
3894 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3895 elr = list_entry(pos, struct ext4_li_request,
3896 lr_request);
3897 ext4_remove_li_request(elr);
3898 }
3899 mutex_unlock(lock: &ext4_li_info->li_list_mtx);
3900}
3901
3902static int ext4_run_lazyinit_thread(void)
3903{
3904 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3905 ext4_li_info, "ext4lazyinit");
3906 if (IS_ERR(ptr: ext4_lazyinit_task)) {
3907 int err = PTR_ERR(ptr: ext4_lazyinit_task);
3908 ext4_clear_request_list();
3909 kfree(objp: ext4_li_info);
3910 ext4_li_info = NULL;
3911 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
3912 "initialization thread\n",
3913 err);
3914 return err;
3915 }
3916 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
3917 return 0;
3918}
3919
3920/*
3921 * Check whether it make sense to run itable init. thread or not.
3922 * If there is at least one uninitialized inode table, return
3923 * corresponding group number, else the loop goes through all
3924 * groups and return total number of groups.
3925 */
3926static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3927{
3928 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3929 struct ext4_group_desc *gdp = NULL;
3930
3931 if (!ext4_has_group_desc_csum(sb))
3932 return ngroups;
3933
3934 for (group = 0; group < ngroups; group++) {
3935 gdp = ext4_get_group_desc(sb, block_group: group, NULL);
3936 if (!gdp)
3937 continue;
3938
3939 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3940 break;
3941 }
3942
3943 return group;
3944}
3945
3946static int ext4_li_info_new(void)
3947{
3948 struct ext4_lazy_init *eli = NULL;
3949
3950 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3951 if (!eli)
3952 return -ENOMEM;
3953
3954 INIT_LIST_HEAD(list: &eli->li_request_list);
3955 mutex_init(&eli->li_list_mtx);
3956
3957 eli->li_state |= EXT4_LAZYINIT_QUIT;
3958
3959 ext4_li_info = eli;
3960
3961 return 0;
3962}
3963
3964static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3965 ext4_group_t start)
3966{
3967 struct ext4_li_request *elr;
3968
3969 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3970 if (!elr)
3971 return NULL;
3972
3973 elr->lr_super = sb;
3974 elr->lr_first_not_zeroed = start;
3975 if (test_opt(sb, NO_PREFETCH_BLOCK_BITMAPS)) {
3976 elr->lr_mode = EXT4_LI_MODE_ITABLE;
3977 elr->lr_next_group = start;
3978 } else {
3979 elr->lr_mode = EXT4_LI_MODE_PREFETCH_BBITMAP;
3980 }
3981
3982 /*
3983 * Randomize first schedule time of the request to
3984 * spread the inode table initialization requests
3985 * better.
3986 */
3987 elr->lr_next_sched = jiffies + get_random_u32_below(EXT4_DEF_LI_MAX_START_DELAY * HZ);
3988 return elr;
3989}
3990
3991int ext4_register_li_request(struct super_block *sb,
3992 ext4_group_t first_not_zeroed)
3993{
3994 struct ext4_sb_info *sbi = EXT4_SB(sb);
3995 struct ext4_li_request *elr = NULL;
3996 ext4_group_t ngroups = sbi->s_groups_count;
3997 int ret = 0;
3998
3999 mutex_lock(lock: &ext4_li_mtx);
4000 if (sbi->s_li_request != NULL) {
4001 /*
4002 * Reset timeout so it can be computed again, because
4003 * s_li_wait_mult might have changed.
4004 */
4005 sbi->s_li_request->lr_timeout = 0;
4006 goto out;
4007 }
4008
4009 if (ext4_emergency_state(sb) || sb_rdonly(sb) ||
4010 (test_opt(sb, NO_PREFETCH_BLOCK_BITMAPS) &&
4011 (first_not_zeroed == ngroups || !test_opt(sb, INIT_INODE_TABLE))))
4012 goto out;
4013
4014 elr = ext4_li_request_new(sb, start: first_not_zeroed);
4015 if (!elr) {
4016 ret = -ENOMEM;
4017 goto out;
4018 }
4019
4020 if (NULL == ext4_li_info) {
4021 ret = ext4_li_info_new();
4022 if (ret)
4023 goto out;
4024 }
4025
4026 mutex_lock(lock: &ext4_li_info->li_list_mtx);
4027 list_add(new: &elr->lr_request, head: &ext4_li_info->li_request_list);
4028 mutex_unlock(lock: &ext4_li_info->li_list_mtx);
4029
4030 sbi->s_li_request = elr;
4031 /*
4032 * set elr to NULL here since it has been inserted to
4033 * the request_list and the removal and free of it is
4034 * handled by ext4_clear_request_list from now on.
4035 */
4036 elr = NULL;
4037
4038 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
4039 ret = ext4_run_lazyinit_thread();
4040 if (ret)
4041 goto out;
4042 }
4043out:
4044 mutex_unlock(lock: &ext4_li_mtx);
4045 if (ret)
4046 kfree(objp: elr);
4047 return ret;
4048}
4049
4050/*
4051 * We do not need to lock anything since this is called on
4052 * module unload.
4053 */
4054static void ext4_destroy_lazyinit_thread(void)
4055{
4056 /*
4057 * If thread exited earlier
4058 * there's nothing to be done.
4059 */
4060 if (!ext4_li_info || !ext4_lazyinit_task)
4061 return;
4062
4063 kthread_stop(k: ext4_lazyinit_task);
4064}
4065
4066static int set_journal_csum_feature_set(struct super_block *sb)
4067{
4068 int ret = 1;
4069 int compat, incompat;
4070 struct ext4_sb_info *sbi = EXT4_SB(sb);
4071
4072 if (ext4_has_feature_metadata_csum(sb)) {
4073 /* journal checksum v3 */
4074 compat = 0;
4075 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
4076 } else {
4077 /* journal checksum v1 */
4078 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
4079 incompat = 0;
4080 }
4081
4082 jbd2_journal_clear_features(sbi->s_journal,
4083 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
4084 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
4085 JBD2_FEATURE_INCOMPAT_CSUM_V2);
4086 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4087 ret = jbd2_journal_set_features(sbi->s_journal,
4088 compat, 0,
4089 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
4090 incompat);
4091 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
4092 ret = jbd2_journal_set_features(sbi->s_journal,
4093 compat, 0,
4094 incompat);
4095 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
4096 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
4097 } else {
4098 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
4099 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
4100 }
4101
4102 return ret;
4103}
4104
4105/*
4106 * Note: calculating the overhead so we can be compatible with
4107 * historical BSD practice is quite difficult in the face of
4108 * clusters/bigalloc. This is because multiple metadata blocks from
4109 * different block group can end up in the same allocation cluster.
4110 * Calculating the exact overhead in the face of clustered allocation
4111 * requires either O(all block bitmaps) in memory or O(number of block
4112 * groups**2) in time. We will still calculate the superblock for
4113 * older file systems --- and if we come across with a bigalloc file
4114 * system with zero in s_overhead_clusters the estimate will be close to
4115 * correct especially for very large cluster sizes --- but for newer
4116 * file systems, it's better to calculate this figure once at mkfs
4117 * time, and store it in the superblock. If the superblock value is
4118 * present (even for non-bigalloc file systems), we will use it.
4119 */
4120static int count_overhead(struct super_block *sb, ext4_group_t grp,
4121 char *buf)
4122{
4123 struct ext4_sb_info *sbi = EXT4_SB(sb);
4124 struct ext4_group_desc *gdp;
4125 ext4_fsblk_t first_block, last_block, b;
4126 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
4127 int s, j, count = 0;
4128 int has_super = ext4_bg_has_super(sb, group: grp);
4129
4130 if (!ext4_has_feature_bigalloc(sb))
4131 return (has_super + ext4_bg_num_gdb(sb, group: grp) +
4132 (has_super ? le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) : 0) +
4133 sbi->s_itb_per_group + 2);
4134
4135 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
4136 (grp * EXT4_BLOCKS_PER_GROUP(sb));
4137 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
4138 for (i = 0; i < ngroups; i++) {
4139 gdp = ext4_get_group_desc(sb, block_group: i, NULL);
4140 b = ext4_block_bitmap(sb, bg: gdp);
4141 if (b >= first_block && b <= last_block) {
4142 ext4_set_bit(EXT4_B2C(sbi, b - first_block), addr: buf);
4143 count++;
4144 }
4145 b = ext4_inode_bitmap(sb, bg: gdp);
4146 if (b >= first_block && b <= last_block) {
4147 ext4_set_bit(EXT4_B2C(sbi, b - first_block), addr: buf);
4148 count++;
4149 }
4150 b = ext4_inode_table(sb, bg: gdp);
4151 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
4152 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
4153 int c = EXT4_B2C(sbi, b - first_block);
4154 ext4_set_bit(nr: c, addr: buf);
4155 count++;
4156 }
4157 if (i != grp)
4158 continue;
4159 s = 0;
4160 if (ext4_bg_has_super(sb, group: grp)) {
4161 ext4_set_bit(nr: s++, addr: buf);
4162 count++;
4163 }
4164 j = ext4_bg_num_gdb(sb, group: grp);
4165 if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) {
4166 ext4_error(sb, "Invalid number of block group "
4167 "descriptor blocks: %d", j);
4168 j = EXT4_BLOCKS_PER_GROUP(sb) - s;
4169 }
4170 count += j;
4171 for (; j > 0; j--)
4172 ext4_set_bit(EXT4_B2C(sbi, s++), addr: buf);
4173 }
4174 if (!count)
4175 return 0;
4176 return EXT4_CLUSTERS_PER_GROUP(sb) -
4177 ext4_count_free(bitmap: buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
4178}
4179
4180/*
4181 * Compute the overhead and stash it in sbi->s_overhead
4182 */
4183int ext4_calculate_overhead(struct super_block *sb)
4184{
4185 struct ext4_sb_info *sbi = EXT4_SB(sb);
4186 struct ext4_super_block *es = sbi->s_es;
4187 struct inode *j_inode;
4188 unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum);
4189 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
4190 ext4_fsblk_t overhead = 0;
4191 char *buf = (char *) get_zeroed_page(GFP_NOFS);
4192
4193 if (!buf)
4194 return -ENOMEM;
4195
4196 /*
4197 * Compute the overhead (FS structures). This is constant
4198 * for a given filesystem unless the number of block groups
4199 * changes so we cache the previous value until it does.
4200 */
4201
4202 /*
4203 * All of the blocks before first_data_block are overhead
4204 */
4205 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
4206
4207 /*
4208 * Add the overhead found in each block group
4209 */
4210 for (i = 0; i < ngroups; i++) {
4211 int blks;
4212
4213 blks = count_overhead(sb, grp: i, buf);
4214 overhead += blks;
4215 if (blks)
4216 memset(s: buf, c: 0, PAGE_SIZE);
4217 cond_resched();
4218 }
4219
4220 /*
4221 * Add the internal journal blocks whether the journal has been
4222 * loaded or not
4223 */
4224 if (sbi->s_journal && !sbi->s_journal_bdev_file)
4225 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_total_len);
4226 else if (ext4_has_feature_journal(sb) && !sbi->s_journal && j_inum) {
4227 /* j_inum for internal journal is non-zero */
4228 j_inode = ext4_get_journal_inode(sb, journal_inum: j_inum);
4229 if (!IS_ERR(ptr: j_inode)) {
4230 j_blocks = j_inode->i_size >> sb->s_blocksize_bits;
4231 overhead += EXT4_NUM_B2C(sbi, j_blocks);
4232 iput(j_inode);
4233 } else {
4234 ext4_msg(sb, KERN_ERR, "can't get journal size");
4235 }
4236 }
4237 sbi->s_overhead = overhead;
4238 smp_wmb();
4239 free_page((unsigned long) buf);
4240 return 0;
4241}
4242
4243static void ext4_set_resv_clusters(struct super_block *sb)
4244{
4245 ext4_fsblk_t resv_clusters;
4246 struct ext4_sb_info *sbi = EXT4_SB(sb);
4247
4248 /*
4249 * There's no need to reserve anything when we aren't using extents.
4250 * The space estimates are exact, there are no unwritten extents,
4251 * hole punching doesn't need new metadata... This is needed especially
4252 * to keep ext2/3 backward compatibility.
4253 */
4254 if (!ext4_has_feature_extents(sb))
4255 return;
4256 /*
4257 * By default we reserve 2% or 4096 clusters, whichever is smaller.
4258 * This should cover the situations where we can not afford to run
4259 * out of space like for example punch hole, or converting
4260 * unwritten extents in delalloc path. In most cases such
4261 * allocation would require 1, or 2 blocks, higher numbers are
4262 * very rare.
4263 */
4264 resv_clusters = (ext4_blocks_count(es: sbi->s_es) >>
4265 sbi->s_cluster_bits);
4266
4267 do_div(resv_clusters, 50);
4268 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
4269
4270 atomic64_set(v: &sbi->s_resv_clusters, i: resv_clusters);
4271}
4272
4273static const char *ext4_quota_mode(struct super_block *sb)
4274{
4275#ifdef CONFIG_QUOTA
4276 if (!ext4_quota_capable(sb))
4277 return "none";
4278
4279 if (EXT4_SB(sb)->s_journal && ext4_is_quota_journalled(sb))
4280 return "journalled";
4281 else
4282 return "writeback";
4283#else
4284 return "disabled";
4285#endif
4286}
4287
4288static void ext4_setup_csum_trigger(struct super_block *sb,
4289 enum ext4_journal_trigger_type type,
4290 void (*trigger)(
4291 struct jbd2_buffer_trigger_type *type,
4292 struct buffer_head *bh,
4293 void *mapped_data,
4294 size_t size))
4295{
4296 struct ext4_sb_info *sbi = EXT4_SB(sb);
4297
4298 sbi->s_journal_triggers[type].sb = sb;
4299 sbi->s_journal_triggers[type].tr_triggers.t_frozen = trigger;
4300}
4301
4302static void ext4_free_sbi(struct ext4_sb_info *sbi)
4303{
4304 if (!sbi)
4305 return;
4306
4307 kfree(objp: sbi->s_blockgroup_lock);
4308 fs_put_dax(dax_dev: sbi->s_daxdev, NULL);
4309 kfree(objp: sbi);
4310}
4311
4312static struct ext4_sb_info *ext4_alloc_sbi(struct super_block *sb)
4313{
4314 struct ext4_sb_info *sbi;
4315
4316 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
4317 if (!sbi)
4318 return NULL;
4319
4320 sbi->s_daxdev = fs_dax_get_by_bdev(bdev: sb->s_bdev, start_off: &sbi->s_dax_part_off,
4321 NULL, NULL);
4322
4323 sbi->s_blockgroup_lock =
4324 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
4325
4326 if (!sbi->s_blockgroup_lock)
4327 goto err_out;
4328
4329 sb->s_fs_info = sbi;
4330 sbi->s_sb = sb;
4331 return sbi;
4332err_out:
4333 fs_put_dax(dax_dev: sbi->s_daxdev, NULL);
4334 kfree(objp: sbi);
4335 return NULL;
4336}
4337
4338static void ext4_set_def_opts(struct super_block *sb,
4339 struct ext4_super_block *es)
4340{
4341 unsigned long def_mount_opts;
4342
4343 /* Set defaults before we parse the mount options */
4344 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
4345 set_opt(sb, INIT_INODE_TABLE);
4346 if (def_mount_opts & EXT4_DEFM_DEBUG)
4347 set_opt(sb, DEBUG);
4348 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
4349 set_opt(sb, GRPID);
4350 if (def_mount_opts & EXT4_DEFM_UID16)
4351 set_opt(sb, NO_UID32);
4352 /* xattr user namespace & acls are now defaulted on */
4353 set_opt(sb, XATTR_USER);
4354#ifdef CONFIG_EXT4_FS_POSIX_ACL
4355 set_opt(sb, POSIX_ACL);
4356#endif
4357 if (ext4_has_feature_fast_commit(sb))
4358 set_opt2(sb, JOURNAL_FAST_COMMIT);
4359 /* don't forget to enable journal_csum when metadata_csum is enabled. */
4360 if (ext4_has_feature_metadata_csum(sb))
4361 set_opt(sb, JOURNAL_CHECKSUM);
4362
4363 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
4364 set_opt(sb, JOURNAL_DATA);
4365 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
4366 set_opt(sb, ORDERED_DATA);
4367 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
4368 set_opt(sb, WRITEBACK_DATA);
4369
4370 if (le16_to_cpu(es->s_errors) == EXT4_ERRORS_PANIC)
4371 set_opt(sb, ERRORS_PANIC);
4372 else if (le16_to_cpu(es->s_errors) == EXT4_ERRORS_CONTINUE)
4373 set_opt(sb, ERRORS_CONT);
4374 else
4375 set_opt(sb, ERRORS_RO);
4376 /* block_validity enabled by default; disable with noblock_validity */
4377 set_opt(sb, BLOCK_VALIDITY);
4378 if (def_mount_opts & EXT4_DEFM_DISCARD)
4379 set_opt(sb, DISCARD);
4380
4381 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
4382 set_opt(sb, BARRIER);
4383
4384 /*
4385 * enable delayed allocation by default
4386 * Use -o nodelalloc to turn it off
4387 */
4388 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
4389 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
4390 set_opt(sb, DELALLOC);
4391
4392 if (sb->s_blocksize <= PAGE_SIZE)
4393 set_opt(sb, DIOREAD_NOLOCK);
4394}
4395
4396static int ext4_handle_clustersize(struct super_block *sb)
4397{
4398 struct ext4_sb_info *sbi = EXT4_SB(sb);
4399 struct ext4_super_block *es = sbi->s_es;
4400 int clustersize;
4401
4402 /* Handle clustersize */
4403 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
4404 if (ext4_has_feature_bigalloc(sb)) {
4405 if (clustersize < sb->s_blocksize) {
4406 ext4_msg(sb, KERN_ERR,
4407 "cluster size (%d) smaller than "
4408 "block size (%lu)", clustersize, sb->s_blocksize);
4409 return -EINVAL;
4410 }
4411 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
4412 le32_to_cpu(es->s_log_block_size);
4413 } else {
4414 if (clustersize != sb->s_blocksize) {
4415 ext4_msg(sb, KERN_ERR,
4416 "fragment/cluster size (%d) != "
4417 "block size (%lu)", clustersize, sb->s_blocksize);
4418 return -EINVAL;
4419 }
4420 if (sbi->s_blocks_per_group > sb->s_blocksize * 8) {
4421 ext4_msg(sb, KERN_ERR,
4422 "#blocks per group too big: %lu",
4423 sbi->s_blocks_per_group);
4424 return -EINVAL;
4425 }
4426 sbi->s_cluster_bits = 0;
4427 }
4428 sbi->s_clusters_per_group = le32_to_cpu(es->s_clusters_per_group);
4429 if (sbi->s_clusters_per_group > sb->s_blocksize * 8) {
4430 ext4_msg(sb, KERN_ERR, "#clusters per group too big: %lu",
4431 sbi->s_clusters_per_group);
4432 return -EINVAL;
4433 }
4434 if (sbi->s_blocks_per_group !=
4435 (sbi->s_clusters_per_group * (clustersize / sb->s_blocksize))) {
4436 ext4_msg(sb, KERN_ERR,
4437 "blocks per group (%lu) and clusters per group (%lu) inconsistent",
4438 sbi->s_blocks_per_group, sbi->s_clusters_per_group);
4439 return -EINVAL;
4440 }
4441 sbi->s_cluster_ratio = clustersize / sb->s_blocksize;
4442
4443 /* Do we have standard group size of clustersize * 8 blocks ? */
4444 if (sbi->s_blocks_per_group == clustersize << 3)
4445 set_opt2(sb, STD_GROUP_SIZE);
4446
4447 return 0;
4448}
4449
4450/*
4451 * ext4_atomic_write_init: Initializes filesystem min & max atomic write units.
4452 * With non-bigalloc filesystem awu will be based upon filesystem blocksize
4453 * & bdev awu units.
4454 * With bigalloc it will be based upon bigalloc cluster size & bdev awu units.
4455 * @sb: super block
4456 */
4457static void ext4_atomic_write_init(struct super_block *sb)
4458{
4459 struct ext4_sb_info *sbi = EXT4_SB(sb);
4460 struct block_device *bdev = sb->s_bdev;
4461 unsigned int clustersize = EXT4_CLUSTER_SIZE(sb);
4462
4463 if (!bdev_can_atomic_write(bdev))
4464 return;
4465
4466 if (!ext4_has_feature_extents(sb))
4467 return;
4468
4469 sbi->s_awu_min = max(sb->s_blocksize,
4470 bdev_atomic_write_unit_min_bytes(bdev));
4471 sbi->s_awu_max = min(clustersize,
4472 bdev_atomic_write_unit_max_bytes(bdev));
4473 if (sbi->s_awu_min && sbi->s_awu_max &&
4474 sbi->s_awu_min <= sbi->s_awu_max) {
4475 ext4_msg(sb, KERN_NOTICE, "Supports (experimental) DIO atomic writes awu_min: %u, awu_max: %u",
4476 sbi->s_awu_min, sbi->s_awu_max);
4477 } else {
4478 sbi->s_awu_min = 0;
4479 sbi->s_awu_max = 0;
4480 }
4481}
4482
4483static void ext4_fast_commit_init(struct super_block *sb)
4484{
4485 struct ext4_sb_info *sbi = EXT4_SB(sb);
4486
4487 /* Initialize fast commit stuff */
4488 atomic_set(v: &sbi->s_fc_subtid, i: 0);
4489 INIT_LIST_HEAD(list: &sbi->s_fc_q[FC_Q_MAIN]);
4490 INIT_LIST_HEAD(list: &sbi->s_fc_q[FC_Q_STAGING]);
4491 INIT_LIST_HEAD(list: &sbi->s_fc_dentry_q[FC_Q_MAIN]);
4492 INIT_LIST_HEAD(list: &sbi->s_fc_dentry_q[FC_Q_STAGING]);
4493 sbi->s_fc_bytes = 0;
4494 ext4_clear_mount_flag(sb, bit: EXT4_MF_FC_INELIGIBLE);
4495 sbi->s_fc_ineligible_tid = 0;
4496 mutex_init(&sbi->s_fc_lock);
4497 memset(s: &sbi->s_fc_stats, c: 0, n: sizeof(sbi->s_fc_stats));
4498 sbi->s_fc_replay_state.fc_regions = NULL;
4499 sbi->s_fc_replay_state.fc_regions_size = 0;
4500 sbi->s_fc_replay_state.fc_regions_used = 0;
4501 sbi->s_fc_replay_state.fc_regions_valid = 0;
4502 sbi->s_fc_replay_state.fc_modified_inodes = NULL;
4503 sbi->s_fc_replay_state.fc_modified_inodes_size = 0;
4504 sbi->s_fc_replay_state.fc_modified_inodes_used = 0;
4505}
4506
4507static int ext4_inode_info_init(struct super_block *sb,
4508 struct ext4_super_block *es)
4509{
4510 struct ext4_sb_info *sbi = EXT4_SB(sb);
4511
4512 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
4513 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
4514 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
4515 } else {
4516 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
4517 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
4518 if (sbi->s_first_ino < EXT4_GOOD_OLD_FIRST_INO) {
4519 ext4_msg(sb, KERN_ERR, "invalid first ino: %u",
4520 sbi->s_first_ino);
4521 return -EINVAL;
4522 }
4523 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
4524 (!is_power_of_2(n: sbi->s_inode_size)) ||
4525 (sbi->s_inode_size > sb->s_blocksize)) {
4526 ext4_msg(sb, KERN_ERR,
4527 "unsupported inode size: %d",
4528 sbi->s_inode_size);
4529 ext4_msg(sb, KERN_ERR, "blocksize: %lu", sb->s_blocksize);
4530 return -EINVAL;
4531 }
4532 /*
4533 * i_atime_extra is the last extra field available for
4534 * [acm]times in struct ext4_inode. Checking for that
4535 * field should suffice to ensure we have extra space
4536 * for all three.
4537 */
4538 if (sbi->s_inode_size >= offsetof(struct ext4_inode, i_atime_extra) +
4539 sizeof(((struct ext4_inode *)0)->i_atime_extra)) {
4540 sb->s_time_gran = 1;
4541 sb->s_time_max = EXT4_EXTRA_TIMESTAMP_MAX;
4542 } else {
4543 sb->s_time_gran = NSEC_PER_SEC;
4544 sb->s_time_max = EXT4_NON_EXTRA_TIMESTAMP_MAX;
4545 }
4546 sb->s_time_min = EXT4_TIMESTAMP_MIN;
4547 }
4548
4549 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
4550 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4551 EXT4_GOOD_OLD_INODE_SIZE;
4552 if (ext4_has_feature_extra_isize(sb)) {
4553 unsigned v, max = (sbi->s_inode_size -
4554 EXT4_GOOD_OLD_INODE_SIZE);
4555
4556 v = le16_to_cpu(es->s_want_extra_isize);
4557 if (v > max) {
4558 ext4_msg(sb, KERN_ERR,
4559 "bad s_want_extra_isize: %d", v);
4560 return -EINVAL;
4561 }
4562 if (sbi->s_want_extra_isize < v)
4563 sbi->s_want_extra_isize = v;
4564
4565 v = le16_to_cpu(es->s_min_extra_isize);
4566 if (v > max) {
4567 ext4_msg(sb, KERN_ERR,
4568 "bad s_min_extra_isize: %d", v);
4569 return -EINVAL;
4570 }
4571 if (sbi->s_want_extra_isize < v)
4572 sbi->s_want_extra_isize = v;
4573 }
4574 }
4575
4576 return 0;
4577}
4578
4579#if IS_ENABLED(CONFIG_UNICODE)
4580static int ext4_encoding_init(struct super_block *sb, struct ext4_super_block *es)
4581{
4582 const struct ext4_sb_encodings *encoding_info;
4583 struct unicode_map *encoding;
4584 __u16 encoding_flags = le16_to_cpu(es->s_encoding_flags);
4585
4586 if (!ext4_has_feature_casefold(sb) || sb->s_encoding)
4587 return 0;
4588
4589 encoding_info = ext4_sb_read_encoding(es);
4590 if (!encoding_info) {
4591 ext4_msg(sb, KERN_ERR,
4592 "Encoding requested by superblock is unknown");
4593 return -EINVAL;
4594 }
4595
4596 encoding = utf8_load(encoding_info->version);
4597 if (IS_ERR(encoding)) {
4598 ext4_msg(sb, KERN_ERR,
4599 "can't mount with superblock charset: %s-%u.%u.%u "
4600 "not supported by the kernel. flags: 0x%x.",
4601 encoding_info->name,
4602 unicode_major(encoding_info->version),
4603 unicode_minor(encoding_info->version),
4604 unicode_rev(encoding_info->version),
4605 encoding_flags);
4606 return -EINVAL;
4607 }
4608 ext4_msg(sb, KERN_INFO,"Using encoding defined by superblock: "
4609 "%s-%u.%u.%u with flags 0x%hx", encoding_info->name,
4610 unicode_major(encoding_info->version),
4611 unicode_minor(encoding_info->version),
4612 unicode_rev(encoding_info->version),
4613 encoding_flags);
4614
4615 sb->s_encoding = encoding;
4616 sb->s_encoding_flags = encoding_flags;
4617
4618 return 0;
4619}
4620#else
4621static inline int ext4_encoding_init(struct super_block *sb, struct ext4_super_block *es)
4622{
4623 return 0;
4624}
4625#endif
4626
4627static int ext4_init_metadata_csum(struct super_block *sb, struct ext4_super_block *es)
4628{
4629 struct ext4_sb_info *sbi = EXT4_SB(sb);
4630
4631 /* Warn if metadata_csum and gdt_csum are both set. */
4632 if (ext4_has_feature_metadata_csum(sb) &&
4633 ext4_has_feature_gdt_csum(sb))
4634 ext4_warning(sb, "metadata_csum and uninit_bg are "
4635 "redundant flags; please run fsck.");
4636
4637 /* Check for a known checksum algorithm */
4638 if (!ext4_verify_csum_type(sb, es)) {
4639 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
4640 "unknown checksum algorithm.");
4641 return -EINVAL;
4642 }
4643 ext4_setup_csum_trigger(sb, type: EXT4_JTR_ORPHAN_FILE,
4644 trigger: ext4_orphan_file_block_trigger);
4645
4646 /* Check superblock checksum */
4647 if (!ext4_superblock_csum_verify(sb, es)) {
4648 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
4649 "invalid superblock checksum. Run e2fsck?");
4650 return -EFSBADCRC;
4651 }
4652
4653 /* Precompute checksum seed for all metadata */
4654 if (ext4_has_feature_csum_seed(sb))
4655 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
4656 else if (ext4_has_feature_metadata_csum(sb) ||
4657 ext4_has_feature_ea_inode(sb))
4658 sbi->s_csum_seed = ext4_chksum(crc: ~0, address: es->s_uuid,
4659 length: sizeof(es->s_uuid));
4660 return 0;
4661}
4662
4663static int ext4_check_feature_compatibility(struct super_block *sb,
4664 struct ext4_super_block *es,
4665 int silent)
4666{
4667 struct ext4_sb_info *sbi = EXT4_SB(sb);
4668
4669 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
4670 (ext4_has_compat_features(sb) ||
4671 ext4_has_ro_compat_features(sb) ||
4672 ext4_has_incompat_features(sb)))
4673 ext4_msg(sb, KERN_WARNING,
4674 "feature flags set on rev 0 fs, "
4675 "running e2fsck is recommended");
4676
4677 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
4678 set_opt2(sb, HURD_COMPAT);
4679 if (ext4_has_feature_64bit(sb)) {
4680 ext4_msg(sb, KERN_ERR,
4681 "The Hurd can't support 64-bit file systems");
4682 return -EINVAL;
4683 }
4684
4685 /*
4686 * ea_inode feature uses l_i_version field which is not
4687 * available in HURD_COMPAT mode.
4688 */
4689 if (ext4_has_feature_ea_inode(sb)) {
4690 ext4_msg(sb, KERN_ERR,
4691 "ea_inode feature is not supported for Hurd");
4692 return -EINVAL;
4693 }
4694 }
4695
4696 if (IS_EXT2_SB(sb)) {
4697 if (ext2_feature_set_ok(sb))
4698 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
4699 "using the ext4 subsystem");
4700 else {
4701 /*
4702 * If we're probing be silent, if this looks like
4703 * it's actually an ext[34] filesystem.
4704 */
4705 if (silent && ext4_feature_set_ok(sb, readonly: sb_rdonly(sb)))
4706 return -EINVAL;
4707 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
4708 "to feature incompatibilities");
4709 return -EINVAL;
4710 }
4711 }
4712
4713 if (IS_EXT3_SB(sb)) {
4714 if (ext3_feature_set_ok(sb))
4715 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
4716 "using the ext4 subsystem");
4717 else {
4718 /*
4719 * If we're probing be silent, if this looks like
4720 * it's actually an ext4 filesystem.
4721 */
4722 if (silent && ext4_feature_set_ok(sb, readonly: sb_rdonly(sb)))
4723 return -EINVAL;
4724 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
4725 "to feature incompatibilities");
4726 return -EINVAL;
4727 }
4728 }
4729
4730 /*
4731 * Check feature flags regardless of the revision level, since we
4732 * previously didn't change the revision level when setting the flags,
4733 * so there is a chance incompat flags are set on a rev 0 filesystem.
4734 */
4735 if (!ext4_feature_set_ok(sb, readonly: (sb_rdonly(sb))))
4736 return -EINVAL;
4737
4738 if (sbi->s_daxdev) {
4739 if (sb->s_blocksize == PAGE_SIZE)
4740 set_bit(nr: EXT4_FLAGS_BDEV_IS_DAX, addr: &sbi->s_ext4_flags);
4741 else
4742 ext4_msg(sb, KERN_ERR, "unsupported blocksize for DAX\n");
4743 }
4744
4745 if (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) {
4746 if (ext4_has_feature_inline_data(sb)) {
4747 ext4_msg(sb, KERN_ERR, "Cannot use DAX on a filesystem"
4748 " that may contain inline data");
4749 return -EINVAL;
4750 }
4751 if (!test_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags)) {
4752 ext4_msg(sb, KERN_ERR,
4753 "DAX unsupported by block device.");
4754 return -EINVAL;
4755 }
4756 }
4757
4758 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
4759 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
4760 es->s_encryption_level);
4761 return -EINVAL;
4762 }
4763
4764 return 0;
4765}
4766
4767static int ext4_check_geometry(struct super_block *sb,
4768 struct ext4_super_block *es)
4769{
4770 struct ext4_sb_info *sbi = EXT4_SB(sb);
4771 __u64 blocks_count;
4772 int err;
4773
4774 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (sb->s_blocksize / 4)) {
4775 ext4_msg(sb, KERN_ERR,
4776 "Number of reserved GDT blocks insanely large: %d",
4777 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks));
4778 return -EINVAL;
4779 }
4780 /*
4781 * Test whether we have more sectors than will fit in sector_t,
4782 * and whether the max offset is addressable by the page cache.
4783 */
4784 err = generic_check_addressable(sb->s_blocksize_bits,
4785 ext4_blocks_count(es));
4786 if (err) {
4787 ext4_msg(sb, KERN_ERR, "filesystem"
4788 " too large to mount safely on this system");
4789 return err;
4790 }
4791
4792 /* check blocks count against device size */
4793 blocks_count = sb_bdev_nr_blocks(sb);
4794 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
4795 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
4796 "exceeds size of device (%llu blocks)",
4797 ext4_blocks_count(es), blocks_count);
4798 return -EINVAL;
4799 }
4800
4801 /*
4802 * It makes no sense for the first data block to be beyond the end
4803 * of the filesystem.
4804 */
4805 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
4806 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
4807 "block %u is beyond end of filesystem (%llu)",
4808 le32_to_cpu(es->s_first_data_block),
4809 ext4_blocks_count(es));
4810 return -EINVAL;
4811 }
4812 if ((es->s_first_data_block == 0) && (es->s_log_block_size == 0) &&
4813 (sbi->s_cluster_ratio == 1)) {
4814 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
4815 "block is 0 with a 1k block and cluster size");
4816 return -EINVAL;
4817 }
4818
4819 blocks_count = (ext4_blocks_count(es) -
4820 le32_to_cpu(es->s_first_data_block) +
4821 EXT4_BLOCKS_PER_GROUP(sb) - 1);
4822 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4823 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
4824 ext4_msg(sb, KERN_WARNING, "groups count too large: %llu "
4825 "(block count %llu, first data block %u, "
4826 "blocks per group %lu)", blocks_count,
4827 ext4_blocks_count(es),
4828 le32_to_cpu(es->s_first_data_block),
4829 EXT4_BLOCKS_PER_GROUP(sb));
4830 return -EINVAL;
4831 }
4832 sbi->s_groups_count = blocks_count;
4833 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
4834 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
4835 if (((u64)sbi->s_groups_count * sbi->s_inodes_per_group) !=
4836 le32_to_cpu(es->s_inodes_count)) {
4837 ext4_msg(sb, KERN_ERR, "inodes count not valid: %u vs %llu",
4838 le32_to_cpu(es->s_inodes_count),
4839 ((u64)sbi->s_groups_count * sbi->s_inodes_per_group));
4840 return -EINVAL;
4841 }
4842
4843 return 0;
4844}
4845
4846static int ext4_group_desc_init(struct super_block *sb,
4847 struct ext4_super_block *es,
4848 ext4_fsblk_t logical_sb_block,
4849 ext4_group_t *first_not_zeroed)
4850{
4851 struct ext4_sb_info *sbi = EXT4_SB(sb);
4852 unsigned int db_count;
4853 ext4_fsblk_t block;
4854 int i;
4855
4856 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
4857 EXT4_DESC_PER_BLOCK(sb);
4858 if (ext4_has_feature_meta_bg(sb)) {
4859 if (le32_to_cpu(es->s_first_meta_bg) > db_count) {
4860 ext4_msg(sb, KERN_WARNING,
4861 "first meta block group too large: %u "
4862 "(group descriptor block count %u)",
4863 le32_to_cpu(es->s_first_meta_bg), db_count);
4864 return -EINVAL;
4865 }
4866 }
4867 rcu_assign_pointer(sbi->s_group_desc,
4868 kvmalloc_array(db_count,
4869 sizeof(struct buffer_head *),
4870 GFP_KERNEL));
4871 if (sbi->s_group_desc == NULL) {
4872 ext4_msg(sb, KERN_ERR, "not enough memory");
4873 return -ENOMEM;
4874 }
4875
4876 bgl_lock_init(bgl: sbi->s_blockgroup_lock);
4877
4878 /* Pre-read the descriptors into the buffer cache */
4879 for (i = 0; i < db_count; i++) {
4880 block = descriptor_loc(sb, logical_sb_block, nr: i);
4881 ext4_sb_breadahead_unmovable(sb, block);
4882 }
4883
4884 for (i = 0; i < db_count; i++) {
4885 struct buffer_head *bh;
4886
4887 block = descriptor_loc(sb, logical_sb_block, nr: i);
4888 bh = ext4_sb_bread_unmovable(sb, block);
4889 if (IS_ERR(ptr: bh)) {
4890 ext4_msg(sb, KERN_ERR,
4891 "can't read group descriptor %d", i);
4892 sbi->s_gdb_count = i;
4893 return PTR_ERR(ptr: bh);
4894 }
4895 rcu_read_lock();
4896 rcu_dereference(sbi->s_group_desc)[i] = bh;
4897 rcu_read_unlock();
4898 }
4899 sbi->s_gdb_count = db_count;
4900 if (!ext4_check_descriptors(sb, sb_block: logical_sb_block, first_not_zeroed)) {
4901 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
4902 return -EFSCORRUPTED;
4903 }
4904
4905 return 0;
4906}
4907
4908static int ext4_load_and_init_journal(struct super_block *sb,
4909 struct ext4_super_block *es,
4910 struct ext4_fs_context *ctx)
4911{
4912 struct ext4_sb_info *sbi = EXT4_SB(sb);
4913 int err;
4914
4915 err = ext4_load_journal(sb, es, journal_devnum: ctx->journal_devnum);
4916 if (err)
4917 return err;
4918
4919 if (ext4_has_feature_64bit(sb) &&
4920 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4921 JBD2_FEATURE_INCOMPAT_64BIT)) {
4922 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
4923 goto out;
4924 }
4925
4926 if (!set_journal_csum_feature_set(sb)) {
4927 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
4928 "feature set");
4929 goto out;
4930 }
4931
4932 if (test_opt2(sb, JOURNAL_FAST_COMMIT) &&
4933 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4934 JBD2_FEATURE_INCOMPAT_FAST_COMMIT)) {
4935 ext4_msg(sb, KERN_ERR,
4936 "Failed to set fast commit journal feature");
4937 goto out;
4938 }
4939
4940 /* We have now updated the journal if required, so we can
4941 * validate the data journaling mode. */
4942 switch (test_opt(sb, DATA_FLAGS)) {
4943 case 0:
4944 /* No mode set, assume a default based on the journal
4945 * capabilities: ORDERED_DATA if the journal can
4946 * cope, else JOURNAL_DATA
4947 */
4948 if (jbd2_journal_check_available_features
4949 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4950 set_opt(sb, ORDERED_DATA);
4951 sbi->s_def_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
4952 } else {
4953 set_opt(sb, JOURNAL_DATA);
4954 sbi->s_def_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
4955 }
4956 break;
4957
4958 case EXT4_MOUNT_ORDERED_DATA:
4959 case EXT4_MOUNT_WRITEBACK_DATA:
4960 if (!jbd2_journal_check_available_features
4961 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4962 ext4_msg(sb, KERN_ERR, "Journal does not support "
4963 "requested data journaling mode");
4964 goto out;
4965 }
4966 break;
4967 default:
4968 break;
4969 }
4970
4971 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
4972 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4973 ext4_msg(sb, KERN_ERR, "can't mount with "
4974 "journal_async_commit in data=ordered mode");
4975 goto out;
4976 }
4977
4978 set_task_ioprio(task: sbi->s_journal->j_task, ioprio: ctx->journal_ioprio);
4979
4980 sbi->s_journal->j_submit_inode_data_buffers =
4981 ext4_journal_submit_inode_data_buffers;
4982 sbi->s_journal->j_finish_inode_data_buffers =
4983 ext4_journal_finish_inode_data_buffers;
4984
4985 return 0;
4986
4987out:
4988 ext4_journal_destroy(sbi, journal: sbi->s_journal);
4989 return -EINVAL;
4990}
4991
4992static int ext4_check_journal_data_mode(struct super_block *sb)
4993{
4994 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4995 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting with "
4996 "data=journal disables delayed allocation, "
4997 "dioread_nolock, O_DIRECT and fast_commit support!\n");
4998 /* can't mount with both data=journal and dioread_nolock. */
4999 clear_opt(sb, DIOREAD_NOLOCK);
5000 clear_opt2(sb, JOURNAL_FAST_COMMIT);
5001 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
5002 ext4_msg(sb, KERN_ERR, "can't mount with "
5003 "both data=journal and delalloc");
5004 return -EINVAL;
5005 }
5006 if (test_opt(sb, DAX_ALWAYS)) {
5007 ext4_msg(sb, KERN_ERR, "can't mount with "
5008 "both data=journal and dax");
5009 return -EINVAL;
5010 }
5011 if (ext4_has_feature_encrypt(sb)) {
5012 ext4_msg(sb, KERN_WARNING,
5013 "encrypted files will use data=ordered "
5014 "instead of data journaling mode");
5015 }
5016 if (test_opt(sb, DELALLOC))
5017 clear_opt(sb, DELALLOC);
5018 } else {
5019 sb->s_iflags |= SB_I_CGROUPWB;
5020 }
5021
5022 return 0;
5023}
5024
5025static const char *ext4_has_journal_option(struct super_block *sb)
5026{
5027 struct ext4_sb_info *sbi = EXT4_SB(sb);
5028
5029 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
5030 return "journal_async_commit";
5031 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM))
5032 return "journal_checksum";
5033 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
5034 return "commit=";
5035 if (EXT4_MOUNT_DATA_FLAGS &
5036 (sbi->s_mount_opt ^ sbi->s_def_mount_opt))
5037 return "data=";
5038 if (test_opt(sb, DATA_ERR_ABORT))
5039 return "data_err=abort";
5040 return NULL;
5041}
5042
5043static int ext4_load_super(struct super_block *sb, ext4_fsblk_t *lsb,
5044 int silent)
5045{
5046 struct ext4_sb_info *sbi = EXT4_SB(sb);
5047 struct ext4_super_block *es;
5048 ext4_fsblk_t logical_sb_block;
5049 unsigned long offset = 0;
5050 struct buffer_head *bh;
5051 int ret = -EINVAL;
5052 int blocksize;
5053
5054 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
5055 if (!blocksize) {
5056 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
5057 return -EINVAL;
5058 }
5059
5060 /*
5061 * The ext4 superblock will not be buffer aligned for other than 1kB
5062 * block sizes. We need to calculate the offset from buffer start.
5063 */
5064 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
5065 logical_sb_block = sbi->s_sb_block * EXT4_MIN_BLOCK_SIZE;
5066 offset = do_div(logical_sb_block, blocksize);
5067 } else {
5068 logical_sb_block = sbi->s_sb_block;
5069 }
5070
5071 bh = ext4_sb_bread_unmovable(sb, block: logical_sb_block);
5072 if (IS_ERR(ptr: bh)) {
5073 ext4_msg(sb, KERN_ERR, "unable to read superblock");
5074 return PTR_ERR(ptr: bh);
5075 }
5076 /*
5077 * Note: s_es must be initialized as soon as possible because
5078 * some ext4 macro-instructions depend on its value
5079 */
5080 es = (struct ext4_super_block *) (bh->b_data + offset);
5081 sbi->s_es = es;
5082 sb->s_magic = le16_to_cpu(es->s_magic);
5083 if (sb->s_magic != EXT4_SUPER_MAGIC) {
5084 if (!silent)
5085 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
5086 goto out;
5087 }
5088
5089 if (le32_to_cpu(es->s_log_block_size) >
5090 (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
5091 ext4_msg(sb, KERN_ERR,
5092 "Invalid log block size: %u",
5093 le32_to_cpu(es->s_log_block_size));
5094 goto out;
5095 }
5096 if (le32_to_cpu(es->s_log_cluster_size) >
5097 (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
5098 ext4_msg(sb, KERN_ERR,
5099 "Invalid log cluster size: %u",
5100 le32_to_cpu(es->s_log_cluster_size));
5101 goto out;
5102 }
5103
5104 blocksize = EXT4_MIN_BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
5105
5106 /*
5107 * If the default block size is not the same as the real block size,
5108 * we need to reload it.
5109 */
5110 if (sb->s_blocksize == blocksize) {
5111 *lsb = logical_sb_block;
5112 sbi->s_sbh = bh;
5113 return 0;
5114 }
5115
5116 /*
5117 * bh must be released before kill_bdev(), otherwise
5118 * it won't be freed and its page also. kill_bdev()
5119 * is called by sb_set_blocksize().
5120 */
5121 brelse(bh);
5122 /* Validate the filesystem blocksize */
5123 if (!sb_set_blocksize(sb, blocksize)) {
5124 ext4_msg(sb, KERN_ERR, "bad block size %d",
5125 blocksize);
5126 bh = NULL;
5127 goto out;
5128 }
5129
5130 logical_sb_block = sbi->s_sb_block * EXT4_MIN_BLOCK_SIZE;
5131 offset = do_div(logical_sb_block, blocksize);
5132 bh = ext4_sb_bread_unmovable(sb, block: logical_sb_block);
5133 if (IS_ERR(ptr: bh)) {
5134 ext4_msg(sb, KERN_ERR, "Can't read superblock on 2nd try");
5135 ret = PTR_ERR(ptr: bh);
5136 bh = NULL;
5137 goto out;
5138 }
5139 es = (struct ext4_super_block *)(bh->b_data + offset);
5140 sbi->s_es = es;
5141 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
5142 ext4_msg(sb, KERN_ERR, "Magic mismatch, very weird!");
5143 goto out;
5144 }
5145 *lsb = logical_sb_block;
5146 sbi->s_sbh = bh;
5147 return 0;
5148out:
5149 brelse(bh);
5150 return ret;
5151}
5152
5153static int ext4_hash_info_init(struct super_block *sb)
5154{
5155 struct ext4_sb_info *sbi = EXT4_SB(sb);
5156 struct ext4_super_block *es = sbi->s_es;
5157 unsigned int i;
5158
5159 sbi->s_def_hash_version = es->s_def_hash_version;
5160
5161 if (sbi->s_def_hash_version > DX_HASH_LAST) {
5162 ext4_msg(sb, KERN_ERR,
5163 "Invalid default hash set in the superblock");
5164 return -EINVAL;
5165 } else if (sbi->s_def_hash_version == DX_HASH_SIPHASH) {
5166 ext4_msg(sb, KERN_ERR,
5167 "SIPHASH is not a valid default hash value");
5168 return -EINVAL;
5169 }
5170
5171 for (i = 0; i < 4; i++)
5172 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
5173
5174 if (ext4_has_feature_dir_index(sb)) {
5175 i = le32_to_cpu(es->s_flags);
5176 if (i & EXT2_FLAGS_UNSIGNED_HASH)
5177 sbi->s_hash_unsigned = 3;
5178 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
5179#ifdef __CHAR_UNSIGNED__
5180 if (!sb_rdonly(sb))
5181 es->s_flags |=
5182 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
5183 sbi->s_hash_unsigned = 3;
5184#else
5185 if (!sb_rdonly(sb))
5186 es->s_flags |=
5187 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
5188#endif
5189 }
5190 }
5191 return 0;
5192}
5193
5194static int ext4_block_group_meta_init(struct super_block *sb, int silent)
5195{
5196 struct ext4_sb_info *sbi = EXT4_SB(sb);
5197 struct ext4_super_block *es = sbi->s_es;
5198 int has_huge_files;
5199
5200 has_huge_files = ext4_has_feature_huge_file(sb);
5201 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(bits: sb->s_blocksize_bits,
5202 has_huge_files);
5203 sb->s_maxbytes = ext4_max_size(blkbits: sb->s_blocksize_bits, has_huge_files);
5204
5205 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
5206 if (ext4_has_feature_64bit(sb)) {
5207 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
5208 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
5209 !is_power_of_2(n: sbi->s_desc_size)) {
5210 ext4_msg(sb, KERN_ERR,
5211 "unsupported descriptor size %lu",
5212 sbi->s_desc_size);
5213 return -EINVAL;
5214 }
5215 } else
5216 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
5217
5218 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
5219 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
5220
5221 sbi->s_inodes_per_block = sb->s_blocksize / EXT4_INODE_SIZE(sb);
5222 if (sbi->s_inodes_per_block == 0 || sbi->s_blocks_per_group == 0) {
5223 if (!silent)
5224 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
5225 return -EINVAL;
5226 }
5227 if (sbi->s_inodes_per_group < sbi->s_inodes_per_block ||
5228 sbi->s_inodes_per_group > sb->s_blocksize * 8) {
5229 ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n",
5230 sbi->s_inodes_per_group);
5231 return -EINVAL;
5232 }
5233 sbi->s_itb_per_group = sbi->s_inodes_per_group /
5234 sbi->s_inodes_per_block;
5235 sbi->s_desc_per_block = sb->s_blocksize / EXT4_DESC_SIZE(sb);
5236 sbi->s_mount_state = le16_to_cpu(es->s_state) & ~EXT4_FC_REPLAY;
5237 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
5238 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
5239
5240 return 0;
5241}
5242
5243/*
5244 * It's hard to get stripe aligned blocks if stripe is not aligned with
5245 * cluster, just disable stripe and alert user to simplify code and avoid
5246 * stripe aligned allocation which will rarely succeed.
5247 */
5248static bool ext4_is_stripe_incompatible(struct super_block *sb, unsigned long stripe)
5249{
5250 struct ext4_sb_info *sbi = EXT4_SB(sb);
5251 return (stripe > 0 && sbi->s_cluster_ratio > 1 &&
5252 stripe % sbi->s_cluster_ratio != 0);
5253}
5254
5255static int __ext4_fill_super(struct fs_context *fc, struct super_block *sb)
5256{
5257 struct ext4_super_block *es = NULL;
5258 struct ext4_sb_info *sbi = EXT4_SB(sb);
5259 ext4_fsblk_t logical_sb_block;
5260 struct inode *root;
5261 int needs_recovery;
5262 int err;
5263 ext4_group_t first_not_zeroed;
5264 struct ext4_fs_context *ctx = fc->fs_private;
5265 int silent = fc->sb_flags & SB_SILENT;
5266
5267 /* Set defaults for the variables that will be set during parsing */
5268 if (!(ctx->spec & EXT4_SPEC_JOURNAL_IOPRIO))
5269 ctx->journal_ioprio = EXT4_DEF_JOURNAL_IOPRIO;
5270
5271 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
5272 sbi->s_sectors_written_start =
5273 part_stat_read(sb->s_bdev, sectors[STAT_WRITE]);
5274
5275 err = ext4_load_super(sb, lsb: &logical_sb_block, silent);
5276 if (err)
5277 goto out_fail;
5278
5279 es = sbi->s_es;
5280 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
5281
5282 err = ext4_init_metadata_csum(sb, es);
5283 if (err)
5284 goto failed_mount;
5285
5286 ext4_set_def_opts(sb, es);
5287
5288 sbi->s_resuid = make_kuid(from: &init_user_ns, uid: ext4_get_resuid(es));
5289 sbi->s_resgid = make_kgid(from: &init_user_ns, gid: ext4_get_resuid(es));
5290 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
5291 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
5292 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
5293 sbi->s_sb_update_kb = EXT4_DEF_SB_UPDATE_INTERVAL_KB;
5294 sbi->s_sb_update_sec = EXT4_DEF_SB_UPDATE_INTERVAL_SEC;
5295
5296 /*
5297 * set default s_li_wait_mult for lazyinit, for the case there is
5298 * no mount option specified.
5299 */
5300 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
5301
5302 err = ext4_inode_info_init(sb, es);
5303 if (err)
5304 goto failed_mount;
5305
5306 err = parse_apply_sb_mount_options(sb, m_ctx: ctx);
5307 if (err < 0)
5308 goto failed_mount;
5309
5310 sbi->s_def_mount_opt = sbi->s_mount_opt;
5311 sbi->s_def_mount_opt2 = sbi->s_mount_opt2;
5312
5313 err = ext4_check_opt_consistency(fc, sb);
5314 if (err < 0)
5315 goto failed_mount;
5316
5317 ext4_apply_options(fc, sb);
5318
5319 err = ext4_encoding_init(sb, es);
5320 if (err)
5321 goto failed_mount;
5322
5323 err = ext4_check_journal_data_mode(sb);
5324 if (err)
5325 goto failed_mount;
5326
5327 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
5328 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
5329
5330 /* HSM events are allowed by default. */
5331 sb->s_iflags |= SB_I_ALLOW_HSM;
5332
5333 err = ext4_check_feature_compatibility(sb, es, silent);
5334 if (err)
5335 goto failed_mount;
5336
5337 err = ext4_block_group_meta_init(sb, silent);
5338 if (err)
5339 goto failed_mount;
5340
5341 err = ext4_hash_info_init(sb);
5342 if (err)
5343 goto failed_mount;
5344
5345 err = ext4_handle_clustersize(sb);
5346 if (err)
5347 goto failed_mount;
5348
5349 err = ext4_check_geometry(sb, es);
5350 if (err)
5351 goto failed_mount;
5352
5353 timer_setup(&sbi->s_err_report, print_daily_error_info, 0);
5354 spin_lock_init(&sbi->s_error_lock);
5355 INIT_WORK(&sbi->s_sb_upd_work, update_super_work);
5356
5357 err = ext4_group_desc_init(sb, es, logical_sb_block, first_not_zeroed: &first_not_zeroed);
5358 if (err)
5359 goto failed_mount3;
5360
5361 err = ext4_es_register_shrinker(sbi);
5362 if (err)
5363 goto failed_mount3;
5364
5365 sbi->s_stripe = ext4_get_stripe_size(sbi);
5366 if (ext4_is_stripe_incompatible(sb, stripe: sbi->s_stripe)) {
5367 ext4_msg(sb, KERN_WARNING,
5368 "stripe (%lu) is not aligned with cluster size (%u), "
5369 "stripe is disabled",
5370 sbi->s_stripe, sbi->s_cluster_ratio);
5371 sbi->s_stripe = 0;
5372 }
5373 sbi->s_extent_max_zeroout_kb = 32;
5374
5375 /*
5376 * set up enough so that it can read an inode
5377 */
5378 sb->s_op = &ext4_sops;
5379 sb->s_export_op = &ext4_export_ops;
5380 sb->s_xattr = ext4_xattr_handlers;
5381#ifdef CONFIG_FS_ENCRYPTION
5382 sb->s_cop = &ext4_cryptops;
5383#endif
5384#ifdef CONFIG_FS_VERITY
5385 sb->s_vop = &ext4_verityops;
5386#endif
5387#ifdef CONFIG_QUOTA
5388 sb->dq_op = &ext4_quota_operations;
5389 if (ext4_has_feature_quota(sb))
5390 sb->s_qcop = &dquot_quotactl_sysfile_ops;
5391 else
5392 sb->s_qcop = &ext4_qctl_operations;
5393 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
5394#endif
5395 super_set_uuid(sb, uuid: es->s_uuid, len: sizeof(es->s_uuid));
5396 super_set_sysfs_name_bdev(sb);
5397
5398 INIT_LIST_HEAD(list: &sbi->s_orphan); /* unlinked but open files */
5399 mutex_init(&sbi->s_orphan_lock);
5400
5401 spin_lock_init(&sbi->s_bdev_wb_lock);
5402
5403 ext4_atomic_write_init(sb);
5404 ext4_fast_commit_init(sb);
5405
5406 sb->s_root = NULL;
5407
5408 needs_recovery = (es->s_last_orphan != 0 ||
5409 ext4_has_feature_orphan_present(sb) ||
5410 ext4_has_feature_journal_needs_recovery(sb));
5411
5412 if (ext4_has_feature_mmp(sb) && !sb_rdonly(sb)) {
5413 err = ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block));
5414 if (err)
5415 goto failed_mount3a;
5416 }
5417
5418 err = -EINVAL;
5419 /*
5420 * The first inode we look at is the journal inode. Don't try
5421 * root first: it may be modified in the journal!
5422 */
5423 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
5424 err = ext4_load_and_init_journal(sb, es, ctx);
5425 if (err)
5426 goto failed_mount3a;
5427 if (bdev_read_only(bdev: sb->s_bdev))
5428 needs_recovery = 0;
5429 } else if (test_opt(sb, NOLOAD) && !sb_rdonly(sb) &&
5430 ext4_has_feature_journal_needs_recovery(sb)) {
5431 ext4_msg(sb, KERN_ERR, "required journal recovery "
5432 "suppressed and not mounted read-only");
5433 goto failed_mount3a;
5434 } else {
5435 const char *journal_option;
5436
5437 /* Nojournal mode, all journal mount options are illegal */
5438 journal_option = ext4_has_journal_option(sb);
5439 if (journal_option != NULL) {
5440 ext4_msg(sb, KERN_ERR,
5441 "can't mount with %s, fs mounted w/o journal",
5442 journal_option);
5443 goto failed_mount3a;
5444 }
5445
5446 sbi->s_def_mount_opt &= ~EXT4_MOUNT_JOURNAL_CHECKSUM;
5447 clear_opt(sb, JOURNAL_CHECKSUM);
5448 clear_opt(sb, DATA_FLAGS);
5449 clear_opt2(sb, JOURNAL_FAST_COMMIT);
5450 sbi->s_journal = NULL;
5451 needs_recovery = 0;
5452 }
5453
5454 if (!test_opt(sb, NO_MBCACHE)) {
5455 sbi->s_ea_block_cache = ext4_xattr_create_cache();
5456 if (!sbi->s_ea_block_cache) {
5457 ext4_msg(sb, KERN_ERR,
5458 "Failed to create ea_block_cache");
5459 err = -EINVAL;
5460 goto failed_mount_wq;
5461 }
5462
5463 if (ext4_has_feature_ea_inode(sb)) {
5464 sbi->s_ea_inode_cache = ext4_xattr_create_cache();
5465 if (!sbi->s_ea_inode_cache) {
5466 ext4_msg(sb, KERN_ERR,
5467 "Failed to create ea_inode_cache");
5468 err = -EINVAL;
5469 goto failed_mount_wq;
5470 }
5471 }
5472 }
5473
5474 /*
5475 * Get the # of file system overhead blocks from the
5476 * superblock if present.
5477 */
5478 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
5479 /* ignore the precalculated value if it is ridiculous */
5480 if (sbi->s_overhead > ext4_blocks_count(es))
5481 sbi->s_overhead = 0;
5482 /*
5483 * If the bigalloc feature is not enabled recalculating the
5484 * overhead doesn't take long, so we might as well just redo
5485 * it to make sure we are using the correct value.
5486 */
5487 if (!ext4_has_feature_bigalloc(sb))
5488 sbi->s_overhead = 0;
5489 if (sbi->s_overhead == 0) {
5490 err = ext4_calculate_overhead(sb);
5491 if (err)
5492 goto failed_mount_wq;
5493 }
5494
5495 /*
5496 * The maximum number of concurrent works can be high and
5497 * concurrency isn't really necessary. Limit it to 1.
5498 */
5499 EXT4_SB(sb)->rsv_conversion_wq =
5500 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
5501 if (!EXT4_SB(sb)->rsv_conversion_wq) {
5502 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
5503 err = -ENOMEM;
5504 goto failed_mount4;
5505 }
5506
5507 /*
5508 * The jbd2_journal_load will have done any necessary log recovery,
5509 * so we can safely mount the rest of the filesystem now.
5510 */
5511
5512 root = ext4_iget(sb, EXT4_ROOT_INO, EXT4_IGET_SPECIAL);
5513 if (IS_ERR(ptr: root)) {
5514 ext4_msg(sb, KERN_ERR, "get root inode failed");
5515 err = PTR_ERR(ptr: root);
5516 root = NULL;
5517 goto failed_mount4;
5518 }
5519 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
5520 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
5521 iput(root);
5522 err = -EFSCORRUPTED;
5523 goto failed_mount4;
5524 }
5525
5526 generic_set_sb_d_ops(sb);
5527 sb->s_root = d_make_root(root);
5528 if (!sb->s_root) {
5529 ext4_msg(sb, KERN_ERR, "get root dentry failed");
5530 err = -ENOMEM;
5531 goto failed_mount4;
5532 }
5533
5534 err = ext4_setup_super(sb, es, read_only: sb_rdonly(sb));
5535 if (err == -EROFS) {
5536 sb->s_flags |= SB_RDONLY;
5537 } else if (err)
5538 goto failed_mount4a;
5539
5540 ext4_set_resv_clusters(sb);
5541
5542 if (test_opt(sb, BLOCK_VALIDITY)) {
5543 err = ext4_setup_system_zone(sb);
5544 if (err) {
5545 ext4_msg(sb, KERN_ERR, "failed to initialize system "
5546 "zone (%d)", err);
5547 goto failed_mount4a;
5548 }
5549 }
5550 ext4_fc_replay_cleanup(sb);
5551
5552 ext4_ext_init(sb);
5553
5554 /*
5555 * Enable optimize_scan if number of groups is > threshold. This can be
5556 * turned off by passing "mb_optimize_scan=0". This can also be
5557 * turned on forcefully by passing "mb_optimize_scan=1".
5558 */
5559 if (!(ctx->spec & EXT4_SPEC_mb_optimize_scan)) {
5560 if (sbi->s_groups_count >= MB_DEFAULT_LINEAR_SCAN_THRESHOLD)
5561 set_opt2(sb, MB_OPTIMIZE_SCAN);
5562 else
5563 clear_opt2(sb, MB_OPTIMIZE_SCAN);
5564 }
5565
5566 err = ext4_mb_init(sb);
5567 if (err) {
5568 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
5569 err);
5570 goto failed_mount5;
5571 }
5572
5573 /*
5574 * We can only set up the journal commit callback once
5575 * mballoc is initialized
5576 */
5577 if (sbi->s_journal)
5578 sbi->s_journal->j_commit_callback =
5579 ext4_journal_commit_callback;
5580
5581 err = ext4_percpu_param_init(sbi);
5582 if (err)
5583 goto failed_mount6;
5584
5585 if (ext4_has_feature_flex_bg(sb))
5586 if (!ext4_fill_flex_info(sb)) {
5587 ext4_msg(sb, KERN_ERR,
5588 "unable to initialize "
5589 "flex_bg meta info!");
5590 err = -ENOMEM;
5591 goto failed_mount6;
5592 }
5593
5594 err = ext4_register_li_request(sb, first_not_zeroed);
5595 if (err)
5596 goto failed_mount6;
5597
5598 err = ext4_init_orphan_info(sb);
5599 if (err)
5600 goto failed_mount7;
5601#ifdef CONFIG_QUOTA
5602 /* Enable quota usage during mount. */
5603 if (ext4_has_feature_quota(sb) && !sb_rdonly(sb)) {
5604 err = ext4_enable_quotas(sb);
5605 if (err)
5606 goto failed_mount8;
5607 }
5608#endif /* CONFIG_QUOTA */
5609
5610 /*
5611 * Save the original bdev mapping's wb_err value which could be
5612 * used to detect the metadata async write error.
5613 */
5614 errseq_check_and_advance(eseq: &sb->s_bdev->bd_mapping->wb_err,
5615 since: &sbi->s_bdev_wb_err);
5616 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
5617 ext4_orphan_cleanup(sb, es);
5618 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
5619 /*
5620 * Update the checksum after updating free space/inode counters and
5621 * ext4_orphan_cleanup. Otherwise the superblock can have an incorrect
5622 * checksum in the buffer cache until it is written out and
5623 * e2fsprogs programs trying to open a file system immediately
5624 * after it is mounted can fail.
5625 */
5626 ext4_superblock_csum_set(sb);
5627 if (needs_recovery) {
5628 ext4_msg(sb, KERN_INFO, "recovery complete");
5629 err = ext4_mark_recovery_complete(sb, es);
5630 if (err)
5631 goto failed_mount9;
5632 }
5633
5634 if (test_opt(sb, DISCARD) && !bdev_max_discard_sectors(bdev: sb->s_bdev)) {
5635 ext4_msg(sb, KERN_WARNING,
5636 "mounting with \"discard\" option, but the device does not support discard");
5637 clear_opt(sb, DISCARD);
5638 }
5639
5640 if (es->s_error_count)
5641 mod_timer(timer: &sbi->s_err_report, expires: jiffies + 300*HZ); /* 5 minutes */
5642
5643 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
5644 ratelimit_state_init(rs: &sbi->s_err_ratelimit_state, interval: 5 * HZ, burst: 10);
5645 ratelimit_state_init(rs: &sbi->s_warning_ratelimit_state, interval: 5 * HZ, burst: 10);
5646 ratelimit_state_init(rs: &sbi->s_msg_ratelimit_state, interval: 5 * HZ, burst: 10);
5647 atomic_set(v: &sbi->s_warning_count, i: 0);
5648 atomic_set(v: &sbi->s_msg_count, i: 0);
5649
5650 /* Register sysfs after all initializations are complete. */
5651 err = ext4_register_sysfs(sb);
5652 if (err)
5653 goto failed_mount9;
5654
5655 return 0;
5656
5657failed_mount9:
5658 ext4_quotas_off(sb, EXT4_MAXQUOTAS);
5659failed_mount8: __maybe_unused
5660 ext4_release_orphan_info(sb);
5661failed_mount7:
5662 ext4_unregister_li_request(sb);
5663failed_mount6:
5664 ext4_mb_release(sb);
5665 ext4_flex_groups_free(sbi);
5666 ext4_percpu_param_destroy(sbi);
5667failed_mount5:
5668 ext4_ext_release(sb);
5669 ext4_release_system_zone(sb);
5670failed_mount4a:
5671 dput(sb->s_root);
5672 sb->s_root = NULL;
5673failed_mount4:
5674 ext4_msg(sb, KERN_ERR, "mount failed");
5675 if (EXT4_SB(sb)->rsv_conversion_wq)
5676 destroy_workqueue(wq: EXT4_SB(sb)->rsv_conversion_wq);
5677failed_mount_wq:
5678 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
5679 sbi->s_ea_inode_cache = NULL;
5680
5681 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
5682 sbi->s_ea_block_cache = NULL;
5683
5684 if (sbi->s_journal) {
5685 ext4_journal_destroy(sbi, journal: sbi->s_journal);
5686 }
5687failed_mount3a:
5688 ext4_es_unregister_shrinker(sbi);
5689failed_mount3:
5690 /* flush s_sb_upd_work before sbi destroy */
5691 flush_work(work: &sbi->s_sb_upd_work);
5692 ext4_stop_mmpd(sbi);
5693 timer_delete_sync(timer: &sbi->s_err_report);
5694 ext4_group_desc_free(sbi);
5695failed_mount:
5696#if IS_ENABLED(CONFIG_UNICODE)
5697 utf8_unload(sb->s_encoding);
5698#endif
5699
5700#ifdef CONFIG_QUOTA
5701 for (unsigned int i = 0; i < EXT4_MAXQUOTAS; i++)
5702 kfree(objp: get_qf_name(sb, sbi, type: i));
5703#endif
5704 fscrypt_free_dummy_policy(dummy_policy: &sbi->s_dummy_enc_policy);
5705 brelse(bh: sbi->s_sbh);
5706 if (sbi->s_journal_bdev_file) {
5707 invalidate_bdev(bdev: file_bdev(bdev_file: sbi->s_journal_bdev_file));
5708 bdev_fput(bdev_file: sbi->s_journal_bdev_file);
5709 }
5710out_fail:
5711 invalidate_bdev(bdev: sb->s_bdev);
5712 sb->s_fs_info = NULL;
5713 return err;
5714}
5715
5716static int ext4_fill_super(struct super_block *sb, struct fs_context *fc)
5717{
5718 struct ext4_fs_context *ctx = fc->fs_private;
5719 struct ext4_sb_info *sbi;
5720 const char *descr;
5721 int ret;
5722
5723 sbi = ext4_alloc_sbi(sb);
5724 if (!sbi)
5725 return -ENOMEM;
5726
5727 fc->s_fs_info = sbi;
5728
5729 /* Cleanup superblock name */
5730 strreplace(str: sb->s_id, old: '/', new: '!');
5731
5732 sbi->s_sb_block = 1; /* Default super block location */
5733 if (ctx->spec & EXT4_SPEC_s_sb_block)
5734 sbi->s_sb_block = ctx->s_sb_block;
5735
5736 ret = __ext4_fill_super(fc, sb);
5737 if (ret < 0)
5738 goto free_sbi;
5739
5740 if (sbi->s_journal) {
5741 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
5742 descr = " journalled data mode";
5743 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
5744 descr = " ordered data mode";
5745 else
5746 descr = " writeback data mode";
5747 } else
5748 descr = "out journal";
5749
5750 if (___ratelimit(rs: &ext4_mount_msg_ratelimit, func: "EXT4-fs mount"))
5751 ext4_msg(sb, KERN_INFO, "mounted filesystem %pU %s with%s. "
5752 "Quota mode: %s.", &sb->s_uuid,
5753 sb_rdonly(sb) ? "ro" : "r/w", descr,
5754 ext4_quota_mode(sb));
5755
5756 /* Update the s_overhead_clusters if necessary */
5757 ext4_update_overhead(sb, force: false);
5758 return 0;
5759
5760free_sbi:
5761 ext4_free_sbi(sbi);
5762 fc->s_fs_info = NULL;
5763 return ret;
5764}
5765
5766static int ext4_get_tree(struct fs_context *fc)
5767{
5768 return get_tree_bdev(fc, fill_super: ext4_fill_super);
5769}
5770
5771/*
5772 * Setup any per-fs journal parameters now. We'll do this both on
5773 * initial mount, once the journal has been initialised but before we've
5774 * done any recovery; and again on any subsequent remount.
5775 */
5776static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
5777{
5778 struct ext4_sb_info *sbi = EXT4_SB(sb);
5779
5780 journal->j_commit_interval = sbi->s_commit_interval;
5781 journal->j_min_batch_time = sbi->s_min_batch_time;
5782 journal->j_max_batch_time = sbi->s_max_batch_time;
5783 ext4_fc_init(sb, journal);
5784
5785 write_lock(&journal->j_state_lock);
5786 if (test_opt(sb, BARRIER))
5787 journal->j_flags |= JBD2_BARRIER;
5788 else
5789 journal->j_flags &= ~JBD2_BARRIER;
5790 /*
5791 * Always enable journal cycle record option, letting the journal
5792 * records log transactions continuously between each mount.
5793 */
5794 journal->j_flags |= JBD2_CYCLE_RECORD;
5795 write_unlock(&journal->j_state_lock);
5796}
5797
5798static struct inode *ext4_get_journal_inode(struct super_block *sb,
5799 unsigned int journal_inum)
5800{
5801 struct inode *journal_inode;
5802
5803 /*
5804 * Test for the existence of a valid inode on disk. Bad things
5805 * happen if we iget() an unused inode, as the subsequent iput()
5806 * will try to delete it.
5807 */
5808 journal_inode = ext4_iget(sb, journal_inum, EXT4_IGET_SPECIAL);
5809 if (IS_ERR(ptr: journal_inode)) {
5810 ext4_msg(sb, KERN_ERR, "no journal found");
5811 return ERR_CAST(ptr: journal_inode);
5812 }
5813 if (!journal_inode->i_nlink) {
5814 make_bad_inode(journal_inode);
5815 iput(journal_inode);
5816 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
5817 return ERR_PTR(error: -EFSCORRUPTED);
5818 }
5819 if (!S_ISREG(journal_inode->i_mode) || IS_ENCRYPTED(journal_inode)) {
5820 ext4_msg(sb, KERN_ERR, "invalid journal inode");
5821 iput(journal_inode);
5822 return ERR_PTR(error: -EFSCORRUPTED);
5823 }
5824
5825 ext4_debug("Journal inode found at %p: %lld bytes\n",
5826 journal_inode, journal_inode->i_size);
5827 return journal_inode;
5828}
5829
5830static int ext4_journal_bmap(journal_t *journal, sector_t *block)
5831{
5832 struct ext4_map_blocks map;
5833 int ret;
5834
5835 if (journal->j_inode == NULL)
5836 return 0;
5837
5838 map.m_lblk = *block;
5839 map.m_len = 1;
5840 ret = ext4_map_blocks(NULL, inode: journal->j_inode, map: &map, flags: 0);
5841 if (ret <= 0) {
5842 ext4_msg(journal->j_inode->i_sb, KERN_CRIT,
5843 "journal bmap failed: block %llu ret %d\n",
5844 *block, ret);
5845 jbd2_journal_abort(journal, ret ? ret : -EIO);
5846 return ret;
5847 }
5848 *block = map.m_pblk;
5849 return 0;
5850}
5851
5852static journal_t *ext4_open_inode_journal(struct super_block *sb,
5853 unsigned int journal_inum)
5854{
5855 struct inode *journal_inode;
5856 journal_t *journal;
5857
5858 journal_inode = ext4_get_journal_inode(sb, journal_inum);
5859 if (IS_ERR(ptr: journal_inode))
5860 return ERR_CAST(ptr: journal_inode);
5861
5862 journal = jbd2_journal_init_inode(journal_inode);
5863 if (IS_ERR(ptr: journal)) {
5864 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
5865 iput(journal_inode);
5866 return ERR_CAST(ptr: journal);
5867 }
5868 journal->j_private = sb;
5869 journal->j_bmap = ext4_journal_bmap;
5870 ext4_init_journal_params(sb, journal);
5871 return journal;
5872}
5873
5874static struct file *ext4_get_journal_blkdev(struct super_block *sb,
5875 dev_t j_dev, ext4_fsblk_t *j_start,
5876 ext4_fsblk_t *j_len)
5877{
5878 struct buffer_head *bh;
5879 struct block_device *bdev;
5880 struct file *bdev_file;
5881 int hblock, blocksize;
5882 ext4_fsblk_t sb_block;
5883 unsigned long offset;
5884 struct ext4_super_block *es;
5885 int errno;
5886
5887 bdev_file = bdev_file_open_by_dev(dev: j_dev,
5888 BLK_OPEN_READ | BLK_OPEN_WRITE | BLK_OPEN_RESTRICT_WRITES,
5889 holder: sb, hops: &fs_holder_ops);
5890 if (IS_ERR(ptr: bdev_file)) {
5891 ext4_msg(sb, KERN_ERR,
5892 "failed to open journal device unknown-block(%u,%u) %ld",
5893 MAJOR(j_dev), MINOR(j_dev), PTR_ERR(bdev_file));
5894 return bdev_file;
5895 }
5896
5897 bdev = file_bdev(bdev_file);
5898 blocksize = sb->s_blocksize;
5899 hblock = bdev_logical_block_size(bdev);
5900 if (blocksize < hblock) {
5901 ext4_msg(sb, KERN_ERR,
5902 "blocksize too small for journal device");
5903 errno = -EINVAL;
5904 goto out_bdev;
5905 }
5906
5907 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
5908 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
5909 set_blocksize(file: bdev_file, size: blocksize);
5910 bh = __bread(bdev, block: sb_block, size: blocksize);
5911 if (!bh) {
5912 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
5913 "external journal");
5914 errno = -EINVAL;
5915 goto out_bdev;
5916 }
5917
5918 es = (struct ext4_super_block *) (bh->b_data + offset);
5919 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
5920 !(le32_to_cpu(es->s_feature_incompat) &
5921 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
5922 ext4_msg(sb, KERN_ERR, "external journal has bad superblock");
5923 errno = -EFSCORRUPTED;
5924 goto out_bh;
5925 }
5926
5927 if ((le32_to_cpu(es->s_feature_ro_compat) &
5928 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
5929 es->s_checksum != ext4_superblock_csum(es)) {
5930 ext4_msg(sb, KERN_ERR, "external journal has corrupt superblock");
5931 errno = -EFSCORRUPTED;
5932 goto out_bh;
5933 }
5934
5935 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
5936 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
5937 errno = -EFSCORRUPTED;
5938 goto out_bh;
5939 }
5940
5941 *j_start = sb_block + 1;
5942 *j_len = ext4_blocks_count(es);
5943 brelse(bh);
5944 return bdev_file;
5945
5946out_bh:
5947 brelse(bh);
5948out_bdev:
5949 bdev_fput(bdev_file);
5950 return ERR_PTR(error: errno);
5951}
5952
5953static journal_t *ext4_open_dev_journal(struct super_block *sb,
5954 dev_t j_dev)
5955{
5956 journal_t *journal;
5957 ext4_fsblk_t j_start;
5958 ext4_fsblk_t j_len;
5959 struct file *bdev_file;
5960 int errno = 0;
5961
5962 bdev_file = ext4_get_journal_blkdev(sb, j_dev, j_start: &j_start, j_len: &j_len);
5963 if (IS_ERR(ptr: bdev_file))
5964 return ERR_CAST(ptr: bdev_file);
5965
5966 journal = jbd2_journal_init_dev(bdev: file_bdev(bdev_file), fs_dev: sb->s_bdev, start: j_start,
5967 len: j_len, bsize: sb->s_blocksize);
5968 if (IS_ERR(ptr: journal)) {
5969 ext4_msg(sb, KERN_ERR, "failed to create device journal");
5970 errno = PTR_ERR(ptr: journal);
5971 goto out_bdev;
5972 }
5973 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
5974 ext4_msg(sb, KERN_ERR, "External journal has more than one "
5975 "user (unsupported) - %d",
5976 be32_to_cpu(journal->j_superblock->s_nr_users));
5977 errno = -EINVAL;
5978 goto out_journal;
5979 }
5980 journal->j_private = sb;
5981 EXT4_SB(sb)->s_journal_bdev_file = bdev_file;
5982 ext4_init_journal_params(sb, journal);
5983 return journal;
5984
5985out_journal:
5986 ext4_journal_destroy(sbi: EXT4_SB(sb), journal);
5987out_bdev:
5988 bdev_fput(bdev_file);
5989 return ERR_PTR(error: errno);
5990}
5991
5992static int ext4_load_journal(struct super_block *sb,
5993 struct ext4_super_block *es,
5994 unsigned long journal_devnum)
5995{
5996 journal_t *journal;
5997 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
5998 dev_t journal_dev;
5999 int err = 0;
6000 int really_read_only;
6001 int journal_dev_ro;
6002
6003 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
6004 return -EFSCORRUPTED;
6005
6006 if (journal_devnum &&
6007 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
6008 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
6009 "numbers have changed");
6010 journal_dev = new_decode_dev(dev: journal_devnum);
6011 } else
6012 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
6013
6014 if (journal_inum && journal_dev) {
6015 ext4_msg(sb, KERN_ERR,
6016 "filesystem has both journal inode and journal device!");
6017 return -EINVAL;
6018 }
6019
6020 if (journal_inum) {
6021 journal = ext4_open_inode_journal(sb, journal_inum);
6022 if (IS_ERR(ptr: journal))
6023 return PTR_ERR(ptr: journal);
6024 } else {
6025 journal = ext4_open_dev_journal(sb, j_dev: journal_dev);
6026 if (IS_ERR(ptr: journal))
6027 return PTR_ERR(ptr: journal);
6028 }
6029
6030 journal_dev_ro = bdev_read_only(bdev: journal->j_dev);
6031 really_read_only = bdev_read_only(bdev: sb->s_bdev) | journal_dev_ro;
6032
6033 if (journal_dev_ro && !sb_rdonly(sb)) {
6034 ext4_msg(sb, KERN_ERR,
6035 "journal device read-only, try mounting with '-o ro'");
6036 err = -EROFS;
6037 goto err_out;
6038 }
6039
6040 /*
6041 * Are we loading a blank journal or performing recovery after a
6042 * crash? For recovery, we need to check in advance whether we
6043 * can get read-write access to the device.
6044 */
6045 if (ext4_has_feature_journal_needs_recovery(sb)) {
6046 if (sb_rdonly(sb)) {
6047 ext4_msg(sb, KERN_INFO, "INFO: recovery "
6048 "required on readonly filesystem");
6049 if (really_read_only) {
6050 ext4_msg(sb, KERN_ERR, "write access "
6051 "unavailable, cannot proceed "
6052 "(try mounting with noload)");
6053 err = -EROFS;
6054 goto err_out;
6055 }
6056 ext4_msg(sb, KERN_INFO, "write access will "
6057 "be enabled during recovery");
6058 }
6059 }
6060
6061 if (!(journal->j_flags & JBD2_BARRIER))
6062 ext4_msg(sb, KERN_INFO, "barriers disabled");
6063
6064 if (!ext4_has_feature_journal_needs_recovery(sb))
6065 err = jbd2_journal_wipe(journal, !really_read_only);
6066 if (!err) {
6067 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
6068 __le16 orig_state;
6069 bool changed = false;
6070
6071 if (save)
6072 memcpy(to: save, from: ((char *) es) +
6073 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
6074 err = jbd2_journal_load(journal);
6075 if (save && memcmp(((char *) es) + EXT4_S_ERR_START,
6076 save, EXT4_S_ERR_LEN)) {
6077 memcpy(to: ((char *) es) + EXT4_S_ERR_START,
6078 from: save, EXT4_S_ERR_LEN);
6079 changed = true;
6080 }
6081 kfree(objp: save);
6082 orig_state = es->s_state;
6083 es->s_state |= cpu_to_le16(EXT4_SB(sb)->s_mount_state &
6084 EXT4_ERROR_FS);
6085 if (orig_state != es->s_state)
6086 changed = true;
6087 /* Write out restored error information to the superblock */
6088 if (changed && !really_read_only) {
6089 int err2;
6090 err2 = ext4_commit_super(sb);
6091 err = err ? : err2;
6092 }
6093 }
6094
6095 if (err) {
6096 ext4_msg(sb, KERN_ERR, "error loading journal");
6097 goto err_out;
6098 }
6099
6100 EXT4_SB(sb)->s_journal = journal;
6101 err = ext4_clear_journal_err(sb, es);
6102 if (err) {
6103 ext4_journal_destroy(sbi: EXT4_SB(sb), journal);
6104 return err;
6105 }
6106
6107 if (!really_read_only && journal_devnum &&
6108 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
6109 es->s_journal_dev = cpu_to_le32(journal_devnum);
6110 ext4_commit_super(sb);
6111 }
6112 if (!really_read_only && journal_inum &&
6113 journal_inum != le32_to_cpu(es->s_journal_inum)) {
6114 es->s_journal_inum = cpu_to_le32(journal_inum);
6115 ext4_commit_super(sb);
6116 }
6117
6118 return 0;
6119
6120err_out:
6121 ext4_journal_destroy(sbi: EXT4_SB(sb), journal);
6122 return err;
6123}
6124
6125/* Copy state of EXT4_SB(sb) into buffer for on-disk superblock */
6126static void ext4_update_super(struct super_block *sb)
6127{
6128 struct ext4_sb_info *sbi = EXT4_SB(sb);
6129 struct ext4_super_block *es = sbi->s_es;
6130 struct buffer_head *sbh = sbi->s_sbh;
6131
6132 lock_buffer(bh: sbh);
6133 /*
6134 * If the file system is mounted read-only, don't update the
6135 * superblock write time. This avoids updating the superblock
6136 * write time when we are mounting the root file system
6137 * read/only but we need to replay the journal; at that point,
6138 * for people who are east of GMT and who make their clock
6139 * tick in localtime for Windows bug-for-bug compatibility,
6140 * the clock is set in the future, and this will cause e2fsck
6141 * to complain and force a full file system check.
6142 */
6143 if (!sb_rdonly(sb))
6144 ext4_update_tstamp(es, s_wtime);
6145 es->s_kbytes_written =
6146 cpu_to_le64(sbi->s_kbytes_written +
6147 ((part_stat_read(sb->s_bdev, sectors[STAT_WRITE]) -
6148 sbi->s_sectors_written_start) >> 1));
6149 if (percpu_counter_initialized(fbc: &sbi->s_freeclusters_counter))
6150 ext4_free_blocks_count_set(es,
6151 EXT4_C2B(sbi, percpu_counter_sum_positive(
6152 &sbi->s_freeclusters_counter)));
6153 if (percpu_counter_initialized(fbc: &sbi->s_freeinodes_counter))
6154 es->s_free_inodes_count =
6155 cpu_to_le32(percpu_counter_sum_positive(
6156 &sbi->s_freeinodes_counter));
6157 /* Copy error information to the on-disk superblock */
6158 spin_lock(lock: &sbi->s_error_lock);
6159 if (sbi->s_add_error_count > 0) {
6160 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
6161 if (!es->s_first_error_time && !es->s_first_error_time_hi) {
6162 __ext4_update_tstamp(lo: &es->s_first_error_time,
6163 hi: &es->s_first_error_time_hi,
6164 now: sbi->s_first_error_time);
6165 strtomem_pad(es->s_first_error_func,
6166 sbi->s_first_error_func, 0);
6167 es->s_first_error_line =
6168 cpu_to_le32(sbi->s_first_error_line);
6169 es->s_first_error_ino =
6170 cpu_to_le32(sbi->s_first_error_ino);
6171 es->s_first_error_block =
6172 cpu_to_le64(sbi->s_first_error_block);
6173 es->s_first_error_errcode =
6174 ext4_errno_to_code(errno: sbi->s_first_error_code);
6175 }
6176 __ext4_update_tstamp(lo: &es->s_last_error_time,
6177 hi: &es->s_last_error_time_hi,
6178 now: sbi->s_last_error_time);
6179 strtomem_pad(es->s_last_error_func, sbi->s_last_error_func, 0);
6180 es->s_last_error_line = cpu_to_le32(sbi->s_last_error_line);
6181 es->s_last_error_ino = cpu_to_le32(sbi->s_last_error_ino);
6182 es->s_last_error_block = cpu_to_le64(sbi->s_last_error_block);
6183 es->s_last_error_errcode =
6184 ext4_errno_to_code(errno: sbi->s_last_error_code);
6185 /*
6186 * Start the daily error reporting function if it hasn't been
6187 * started already
6188 */
6189 if (!es->s_error_count)
6190 mod_timer(timer: &sbi->s_err_report, expires: jiffies + 24*60*60*HZ);
6191 le32_add_cpu(var: &es->s_error_count, val: sbi->s_add_error_count);
6192 sbi->s_add_error_count = 0;
6193 }
6194 spin_unlock(lock: &sbi->s_error_lock);
6195
6196 ext4_superblock_csum_set(sb);
6197 unlock_buffer(bh: sbh);
6198}
6199
6200static int ext4_commit_super(struct super_block *sb)
6201{
6202 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
6203
6204 if (!sbh)
6205 return -EINVAL;
6206
6207 ext4_update_super(sb);
6208
6209 lock_buffer(bh: sbh);
6210 /* Buffer got discarded which means block device got invalidated */
6211 if (!buffer_mapped(bh: sbh)) {
6212 unlock_buffer(bh: sbh);
6213 return -EIO;
6214 }
6215
6216 if (buffer_write_io_error(bh: sbh) || !buffer_uptodate(bh: sbh)) {
6217 /*
6218 * Oh, dear. A previous attempt to write the
6219 * superblock failed. This could happen because the
6220 * USB device was yanked out. Or it could happen to
6221 * be a transient write error and maybe the block will
6222 * be remapped. Nothing we can do but to retry the
6223 * write and hope for the best.
6224 */
6225 ext4_msg(sb, KERN_ERR, "previous I/O error to "
6226 "superblock detected");
6227 clear_buffer_write_io_error(bh: sbh);
6228 set_buffer_uptodate(sbh);
6229 }
6230 get_bh(bh: sbh);
6231 /* Clear potential dirty bit if it was journalled update */
6232 clear_buffer_dirty(bh: sbh);
6233 sbh->b_end_io = end_buffer_write_sync;
6234 submit_bh(REQ_OP_WRITE | REQ_SYNC |
6235 (test_opt(sb, BARRIER) ? REQ_FUA : 0), sbh);
6236 wait_on_buffer(bh: sbh);
6237 if (buffer_write_io_error(bh: sbh)) {
6238 ext4_msg(sb, KERN_ERR, "I/O error while writing "
6239 "superblock");
6240 clear_buffer_write_io_error(bh: sbh);
6241 set_buffer_uptodate(sbh);
6242 return -EIO;
6243 }
6244 return 0;
6245}
6246
6247/*
6248 * Have we just finished recovery? If so, and if we are mounting (or
6249 * remounting) the filesystem readonly, then we will end up with a
6250 * consistent fs on disk. Record that fact.
6251 */
6252static int ext4_mark_recovery_complete(struct super_block *sb,
6253 struct ext4_super_block *es)
6254{
6255 int err;
6256 journal_t *journal = EXT4_SB(sb)->s_journal;
6257
6258 if (!ext4_has_feature_journal(sb)) {
6259 if (journal != NULL) {
6260 ext4_error(sb, "Journal got removed while the fs was "
6261 "mounted!");
6262 return -EFSCORRUPTED;
6263 }
6264 return 0;
6265 }
6266 jbd2_journal_lock_updates(journal);
6267 err = jbd2_journal_flush(journal, flags: 0);
6268 if (err < 0)
6269 goto out;
6270
6271 if (sb_rdonly(sb) && (ext4_has_feature_journal_needs_recovery(sb) ||
6272 ext4_has_feature_orphan_present(sb))) {
6273 if (!ext4_orphan_file_empty(sb)) {
6274 ext4_error(sb, "Orphan file not empty on read-only fs.");
6275 err = -EFSCORRUPTED;
6276 goto out;
6277 }
6278 ext4_clear_feature_journal_needs_recovery(sb);
6279 ext4_clear_feature_orphan_present(sb);
6280 ext4_commit_super(sb);
6281 }
6282out:
6283 jbd2_journal_unlock_updates(journal);
6284 return err;
6285}
6286
6287/*
6288 * If we are mounting (or read-write remounting) a filesystem whose journal
6289 * has recorded an error from a previous lifetime, move that error to the
6290 * main filesystem now.
6291 */
6292static int ext4_clear_journal_err(struct super_block *sb,
6293 struct ext4_super_block *es)
6294{
6295 journal_t *journal;
6296 int j_errno;
6297 const char *errstr;
6298
6299 if (!ext4_has_feature_journal(sb)) {
6300 ext4_error(sb, "Journal got removed while the fs was mounted!");
6301 return -EFSCORRUPTED;
6302 }
6303
6304 journal = EXT4_SB(sb)->s_journal;
6305
6306 /*
6307 * Now check for any error status which may have been recorded in the
6308 * journal by a prior ext4_error() or ext4_abort()
6309 */
6310
6311 j_errno = jbd2_journal_errno(journal);
6312 if (j_errno) {
6313 char nbuf[16];
6314
6315 errstr = ext4_decode_error(sb, errno: j_errno, nbuf);
6316 ext4_warning(sb, "Filesystem error recorded "
6317 "from previous mount: %s", errstr);
6318
6319 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
6320 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
6321 j_errno = ext4_commit_super(sb);
6322 if (j_errno)
6323 return j_errno;
6324 ext4_warning(sb, "Marked fs in need of filesystem check.");
6325
6326 jbd2_journal_clear_err(journal);
6327 jbd2_journal_update_sb_errno(journal);
6328 }
6329 return 0;
6330}
6331
6332/*
6333 * Force the running and committing transactions to commit,
6334 * and wait on the commit.
6335 */
6336int ext4_force_commit(struct super_block *sb)
6337{
6338 return ext4_journal_force_commit(journal: EXT4_SB(sb)->s_journal);
6339}
6340
6341static int ext4_sync_fs(struct super_block *sb, int wait)
6342{
6343 int ret = 0;
6344 tid_t target;
6345 bool needs_barrier = false;
6346 struct ext4_sb_info *sbi = EXT4_SB(sb);
6347
6348 ret = ext4_emergency_state(sb);
6349 if (unlikely(ret))
6350 return ret;
6351
6352 trace_ext4_sync_fs(sb, wait);
6353 flush_workqueue(sbi->rsv_conversion_wq);
6354 /*
6355 * Writeback quota in non-journalled quota case - journalled quota has
6356 * no dirty dquots
6357 */
6358 dquot_writeback_dquots(sb, type: -1);
6359 /*
6360 * Data writeback is possible w/o journal transaction, so barrier must
6361 * being sent at the end of the function. But we can skip it if
6362 * transaction_commit will do it for us.
6363 */
6364 if (sbi->s_journal) {
6365 target = jbd2_get_latest_transaction(journal: sbi->s_journal);
6366 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
6367 !jbd2_trans_will_send_data_barrier(journal: sbi->s_journal, tid: target))
6368 needs_barrier = true;
6369
6370 if (jbd2_journal_start_commit(journal: sbi->s_journal, tid: &target)) {
6371 if (wait)
6372 ret = jbd2_log_wait_commit(journal: sbi->s_journal,
6373 tid: target);
6374 }
6375 } else if (wait && test_opt(sb, BARRIER))
6376 needs_barrier = true;
6377 if (needs_barrier) {
6378 int err;
6379 err = blkdev_issue_flush(bdev: sb->s_bdev);
6380 if (!ret)
6381 ret = err;
6382 }
6383
6384 return ret;
6385}
6386
6387/*
6388 * LVM calls this function before a (read-only) snapshot is created. This
6389 * gives us a chance to flush the journal completely and mark the fs clean.
6390 *
6391 * Note that only this function cannot bring a filesystem to be in a clean
6392 * state independently. It relies on upper layer to stop all data & metadata
6393 * modifications.
6394 */
6395static int ext4_freeze(struct super_block *sb)
6396{
6397 int error = 0;
6398 journal_t *journal = EXT4_SB(sb)->s_journal;
6399
6400 if (journal) {
6401 /* Now we set up the journal barrier. */
6402 jbd2_journal_lock_updates(journal);
6403
6404 /*
6405 * Don't clear the needs_recovery flag if we failed to
6406 * flush the journal.
6407 */
6408 error = jbd2_journal_flush(journal, flags: 0);
6409 if (error < 0)
6410 goto out;
6411
6412 /* Journal blocked and flushed, clear needs_recovery flag. */
6413 ext4_clear_feature_journal_needs_recovery(sb);
6414 if (ext4_orphan_file_empty(sb))
6415 ext4_clear_feature_orphan_present(sb);
6416 }
6417
6418 error = ext4_commit_super(sb);
6419out:
6420 if (journal)
6421 /* we rely on upper layer to stop further updates */
6422 jbd2_journal_unlock_updates(journal);
6423 return error;
6424}
6425
6426/*
6427 * Called by LVM after the snapshot is done. We need to reset the RECOVER
6428 * flag here, even though the filesystem is not technically dirty yet.
6429 */
6430static int ext4_unfreeze(struct super_block *sb)
6431{
6432 if (ext4_emergency_state(sb))
6433 return 0;
6434
6435 if (EXT4_SB(sb)->s_journal) {
6436 /* Reset the needs_recovery flag before the fs is unlocked. */
6437 ext4_set_feature_journal_needs_recovery(sb);
6438 if (ext4_has_feature_orphan_file(sb))
6439 ext4_set_feature_orphan_present(sb);
6440 }
6441
6442 ext4_commit_super(sb);
6443 return 0;
6444}
6445
6446/*
6447 * Structure to save mount options for ext4_remount's benefit
6448 */
6449struct ext4_mount_options {
6450 unsigned long s_mount_opt;
6451 unsigned long s_mount_opt2;
6452 kuid_t s_resuid;
6453 kgid_t s_resgid;
6454 unsigned long s_commit_interval;
6455 u32 s_min_batch_time, s_max_batch_time;
6456#ifdef CONFIG_QUOTA
6457 int s_jquota_fmt;
6458 char *s_qf_names[EXT4_MAXQUOTAS];
6459#endif
6460};
6461
6462static int __ext4_remount(struct fs_context *fc, struct super_block *sb)
6463{
6464 struct ext4_fs_context *ctx = fc->fs_private;
6465 struct ext4_super_block *es;
6466 struct ext4_sb_info *sbi = EXT4_SB(sb);
6467 unsigned long old_sb_flags;
6468 struct ext4_mount_options old_opts;
6469 ext4_group_t g;
6470 int err = 0;
6471 int alloc_ctx;
6472#ifdef CONFIG_QUOTA
6473 int enable_quota = 0;
6474 int i, j;
6475 char *to_free[EXT4_MAXQUOTAS];
6476#endif
6477
6478
6479 /* Store the original options */
6480 old_sb_flags = sb->s_flags;
6481 old_opts.s_mount_opt = sbi->s_mount_opt;
6482 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
6483 old_opts.s_resuid = sbi->s_resuid;
6484 old_opts.s_resgid = sbi->s_resgid;
6485 old_opts.s_commit_interval = sbi->s_commit_interval;
6486 old_opts.s_min_batch_time = sbi->s_min_batch_time;
6487 old_opts.s_max_batch_time = sbi->s_max_batch_time;
6488#ifdef CONFIG_QUOTA
6489 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
6490 for (i = 0; i < EXT4_MAXQUOTAS; i++)
6491 if (sbi->s_qf_names[i]) {
6492 char *qf_name = get_qf_name(sb, sbi, type: i);
6493
6494 old_opts.s_qf_names[i] = kstrdup(s: qf_name, GFP_KERNEL);
6495 if (!old_opts.s_qf_names[i]) {
6496 for (j = 0; j < i; j++)
6497 kfree(objp: old_opts.s_qf_names[j]);
6498 return -ENOMEM;
6499 }
6500 } else
6501 old_opts.s_qf_names[i] = NULL;
6502#endif
6503 if (!(ctx->spec & EXT4_SPEC_JOURNAL_IOPRIO)) {
6504 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
6505 ctx->journal_ioprio =
6506 sbi->s_journal->j_task->io_context->ioprio;
6507 else
6508 ctx->journal_ioprio = EXT4_DEF_JOURNAL_IOPRIO;
6509
6510 }
6511
6512 if ((ctx->spec & EXT4_SPEC_s_stripe) &&
6513 ext4_is_stripe_incompatible(sb, stripe: ctx->s_stripe)) {
6514 ext4_msg(sb, KERN_WARNING,
6515 "stripe (%lu) is not aligned with cluster size (%u), "
6516 "stripe is disabled",
6517 ctx->s_stripe, sbi->s_cluster_ratio);
6518 ctx->s_stripe = 0;
6519 }
6520
6521 /*
6522 * Changing the DIOREAD_NOLOCK or DELALLOC mount options may cause
6523 * two calls to ext4_should_dioread_nolock() to return inconsistent
6524 * values, triggering WARN_ON in ext4_add_complete_io(). we grab
6525 * here s_writepages_rwsem to avoid race between writepages ops and
6526 * remount.
6527 */
6528 alloc_ctx = ext4_writepages_down_write(sb);
6529 ext4_apply_options(fc, sb);
6530 ext4_writepages_up_write(sb, ctx: alloc_ctx);
6531
6532 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
6533 test_opt(sb, JOURNAL_CHECKSUM)) {
6534 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
6535 "during remount not supported; ignoring");
6536 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
6537 }
6538
6539 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
6540 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
6541 ext4_msg(sb, KERN_ERR, "can't mount with "
6542 "both data=journal and delalloc");
6543 err = -EINVAL;
6544 goto restore_opts;
6545 }
6546 if (test_opt(sb, DIOREAD_NOLOCK)) {
6547 ext4_msg(sb, KERN_ERR, "can't mount with "
6548 "both data=journal and dioread_nolock");
6549 err = -EINVAL;
6550 goto restore_opts;
6551 }
6552 } else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) {
6553 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
6554 ext4_msg(sb, KERN_ERR, "can't mount with "
6555 "journal_async_commit in data=ordered mode");
6556 err = -EINVAL;
6557 goto restore_opts;
6558 }
6559 }
6560
6561 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_NO_MBCACHE) {
6562 ext4_msg(sb, KERN_ERR, "can't enable nombcache during remount");
6563 err = -EINVAL;
6564 goto restore_opts;
6565 }
6566
6567 if ((old_opts.s_mount_opt & EXT4_MOUNT_DELALLOC) &&
6568 !test_opt(sb, DELALLOC)) {
6569 ext4_msg(sb, KERN_ERR, "can't disable delalloc during remount");
6570 err = -EINVAL;
6571 goto restore_opts;
6572 }
6573
6574 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
6575 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
6576
6577 es = sbi->s_es;
6578
6579 if (sbi->s_journal) {
6580 ext4_init_journal_params(sb, journal: sbi->s_journal);
6581 set_task_ioprio(task: sbi->s_journal->j_task, ioprio: ctx->journal_ioprio);
6582 }
6583
6584 /* Flush outstanding errors before changing fs state */
6585 flush_work(work: &sbi->s_sb_upd_work);
6586
6587 if ((bool)(fc->sb_flags & SB_RDONLY) != sb_rdonly(sb)) {
6588 if (ext4_emergency_state(sb)) {
6589 err = -EROFS;
6590 goto restore_opts;
6591 }
6592
6593 if (fc->sb_flags & SB_RDONLY) {
6594 err = sync_filesystem(sb);
6595 if (err < 0)
6596 goto restore_opts;
6597 err = dquot_suspend(sb, type: -1);
6598 if (err < 0)
6599 goto restore_opts;
6600
6601 /*
6602 * First of all, the unconditional stuff we have to do
6603 * to disable replay of the journal when we next remount
6604 */
6605 sb->s_flags |= SB_RDONLY;
6606
6607 /*
6608 * OK, test if we are remounting a valid rw partition
6609 * readonly, and if so set the rdonly flag and then
6610 * mark the partition as valid again.
6611 */
6612 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
6613 (sbi->s_mount_state & EXT4_VALID_FS))
6614 es->s_state = cpu_to_le16(sbi->s_mount_state);
6615
6616 if (sbi->s_journal) {
6617 /*
6618 * We let remount-ro finish even if marking fs
6619 * as clean failed...
6620 */
6621 ext4_mark_recovery_complete(sb, es);
6622 }
6623 } else {
6624 /* Make sure we can mount this feature set readwrite */
6625 if (ext4_has_feature_readonly(sb) ||
6626 !ext4_feature_set_ok(sb, readonly: 0)) {
6627 err = -EROFS;
6628 goto restore_opts;
6629 }
6630 /*
6631 * Make sure the group descriptor checksums
6632 * are sane. If they aren't, refuse to remount r/w.
6633 */
6634 for (g = 0; g < sbi->s_groups_count; g++) {
6635 struct ext4_group_desc *gdp =
6636 ext4_get_group_desc(sb, block_group: g, NULL);
6637
6638 if (!ext4_group_desc_csum_verify(sb, block_group: g, gdp)) {
6639 ext4_msg(sb, KERN_ERR,
6640 "ext4_remount: Checksum for group %u failed (%u!=%u)",
6641 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
6642 le16_to_cpu(gdp->bg_checksum));
6643 err = -EFSBADCRC;
6644 goto restore_opts;
6645 }
6646 }
6647
6648 /*
6649 * If we have an unprocessed orphan list hanging
6650 * around from a previously readonly bdev mount,
6651 * require a full umount/remount for now.
6652 */
6653 if (es->s_last_orphan || !ext4_orphan_file_empty(sb)) {
6654 ext4_msg(sb, KERN_WARNING, "Couldn't "
6655 "remount RDWR because of unprocessed "
6656 "orphan inode list. Please "
6657 "umount/remount instead");
6658 err = -EINVAL;
6659 goto restore_opts;
6660 }
6661
6662 /*
6663 * Mounting a RDONLY partition read-write, so reread
6664 * and store the current valid flag. (It may have
6665 * been changed by e2fsck since we originally mounted
6666 * the partition.)
6667 */
6668 if (sbi->s_journal) {
6669 err = ext4_clear_journal_err(sb, es);
6670 if (err)
6671 goto restore_opts;
6672 }
6673 sbi->s_mount_state = (le16_to_cpu(es->s_state) &
6674 ~EXT4_FC_REPLAY);
6675
6676 err = ext4_setup_super(sb, es, read_only: 0);
6677 if (err)
6678 goto restore_opts;
6679
6680 sb->s_flags &= ~SB_RDONLY;
6681 if (ext4_has_feature_mmp(sb)) {
6682 err = ext4_multi_mount_protect(sb,
6683 le64_to_cpu(es->s_mmp_block));
6684 if (err)
6685 goto restore_opts;
6686 }
6687#ifdef CONFIG_QUOTA
6688 enable_quota = 1;
6689#endif
6690 }
6691 }
6692
6693 /*
6694 * Handle creation of system zone data early because it can fail.
6695 * Releasing of existing data is done when we are sure remount will
6696 * succeed.
6697 */
6698 if (test_opt(sb, BLOCK_VALIDITY) && !sbi->s_system_blks) {
6699 err = ext4_setup_system_zone(sb);
6700 if (err)
6701 goto restore_opts;
6702 }
6703
6704 if (sbi->s_journal == NULL && !(old_sb_flags & SB_RDONLY)) {
6705 err = ext4_commit_super(sb);
6706 if (err)
6707 goto restore_opts;
6708 }
6709
6710#ifdef CONFIG_QUOTA
6711 if (enable_quota) {
6712 if (sb_any_quota_suspended(sb))
6713 dquot_resume(sb, type: -1);
6714 else if (ext4_has_feature_quota(sb)) {
6715 err = ext4_enable_quotas(sb);
6716 if (err)
6717 goto restore_opts;
6718 }
6719 }
6720 /* Release old quota file names */
6721 for (i = 0; i < EXT4_MAXQUOTAS; i++)
6722 kfree(objp: old_opts.s_qf_names[i]);
6723#endif
6724 if (!test_opt(sb, BLOCK_VALIDITY) && sbi->s_system_blks)
6725 ext4_release_system_zone(sb);
6726
6727 /*
6728 * Reinitialize lazy itable initialization thread based on
6729 * current settings
6730 */
6731 if (sb_rdonly(sb) || !test_opt(sb, INIT_INODE_TABLE))
6732 ext4_unregister_li_request(sb);
6733 else {
6734 ext4_group_t first_not_zeroed;
6735 first_not_zeroed = ext4_has_uninit_itable(sb);
6736 ext4_register_li_request(sb, first_not_zeroed);
6737 }
6738
6739 if (!ext4_has_feature_mmp(sb) || sb_rdonly(sb))
6740 ext4_stop_mmpd(sbi);
6741
6742 /*
6743 * Handle aborting the filesystem as the last thing during remount to
6744 * avoid obsure errors during remount when some option changes fail to
6745 * apply due to shutdown filesystem.
6746 */
6747 if (test_opt2(sb, ABORT))
6748 ext4_abort(sb, ESHUTDOWN, "Abort forced by user");
6749
6750 return 0;
6751
6752restore_opts:
6753 /*
6754 * If there was a failing r/w to ro transition, we may need to
6755 * re-enable quota
6756 */
6757 if (sb_rdonly(sb) && !(old_sb_flags & SB_RDONLY) &&
6758 sb_any_quota_suspended(sb))
6759 dquot_resume(sb, type: -1);
6760
6761 alloc_ctx = ext4_writepages_down_write(sb);
6762 sb->s_flags = old_sb_flags;
6763 sbi->s_mount_opt = old_opts.s_mount_opt;
6764 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
6765 sbi->s_resuid = old_opts.s_resuid;
6766 sbi->s_resgid = old_opts.s_resgid;
6767 sbi->s_commit_interval = old_opts.s_commit_interval;
6768 sbi->s_min_batch_time = old_opts.s_min_batch_time;
6769 sbi->s_max_batch_time = old_opts.s_max_batch_time;
6770 ext4_writepages_up_write(sb, ctx: alloc_ctx);
6771
6772 if (!test_opt(sb, BLOCK_VALIDITY) && sbi->s_system_blks)
6773 ext4_release_system_zone(sb);
6774#ifdef CONFIG_QUOTA
6775 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
6776 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
6777 to_free[i] = get_qf_name(sb, sbi, type: i);
6778 rcu_assign_pointer(sbi->s_qf_names[i], old_opts.s_qf_names[i]);
6779 }
6780 synchronize_rcu();
6781 for (i = 0; i < EXT4_MAXQUOTAS; i++)
6782 kfree(objp: to_free[i]);
6783#endif
6784 if (!ext4_has_feature_mmp(sb) || sb_rdonly(sb))
6785 ext4_stop_mmpd(sbi);
6786 return err;
6787}
6788
6789static int ext4_reconfigure(struct fs_context *fc)
6790{
6791 struct super_block *sb = fc->root->d_sb;
6792 int ret;
6793 bool old_ro = sb_rdonly(sb);
6794
6795 fc->s_fs_info = EXT4_SB(sb);
6796
6797 ret = ext4_check_opt_consistency(fc, sb);
6798 if (ret < 0)
6799 return ret;
6800
6801 ret = __ext4_remount(fc, sb);
6802 if (ret < 0)
6803 return ret;
6804
6805 ext4_msg(sb, KERN_INFO, "re-mounted %pU%s.",
6806 &sb->s_uuid,
6807 (old_ro != sb_rdonly(sb)) ? (sb_rdonly(sb) ? " ro" : " r/w") : "");
6808
6809 return 0;
6810}
6811
6812#ifdef CONFIG_QUOTA
6813static int ext4_statfs_project(struct super_block *sb,
6814 kprojid_t projid, struct kstatfs *buf)
6815{
6816 struct kqid qid;
6817 struct dquot *dquot;
6818 u64 limit;
6819 u64 curblock;
6820
6821 qid = make_kqid_projid(projid);
6822 dquot = dqget(sb, qid);
6823 if (IS_ERR(ptr: dquot))
6824 return PTR_ERR(ptr: dquot);
6825 spin_lock(lock: &dquot->dq_dqb_lock);
6826
6827 limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit,
6828 dquot->dq_dqb.dqb_bhardlimit);
6829 limit >>= sb->s_blocksize_bits;
6830
6831 if (limit) {
6832 uint64_t remaining = 0;
6833
6834 curblock = (dquot->dq_dqb.dqb_curspace +
6835 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits;
6836 if (limit > curblock)
6837 remaining = limit - curblock;
6838
6839 buf->f_blocks = min(buf->f_blocks, limit);
6840 buf->f_bfree = min(buf->f_bfree, remaining);
6841 buf->f_bavail = min(buf->f_bavail, remaining);
6842 }
6843
6844 limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit,
6845 dquot->dq_dqb.dqb_ihardlimit);
6846 if (limit) {
6847 uint64_t remaining = 0;
6848
6849 if (limit > dquot->dq_dqb.dqb_curinodes)
6850 remaining = limit - dquot->dq_dqb.dqb_curinodes;
6851
6852 buf->f_files = min(buf->f_files, limit);
6853 buf->f_ffree = min(buf->f_ffree, remaining);
6854 }
6855
6856 spin_unlock(lock: &dquot->dq_dqb_lock);
6857 dqput(dquot);
6858 return 0;
6859}
6860#endif
6861
6862static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
6863{
6864 struct super_block *sb = dentry->d_sb;
6865 struct ext4_sb_info *sbi = EXT4_SB(sb);
6866 struct ext4_super_block *es = sbi->s_es;
6867 ext4_fsblk_t overhead = 0, resv_blocks;
6868 s64 bfree;
6869 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
6870
6871 if (!test_opt(sb, MINIX_DF))
6872 overhead = sbi->s_overhead;
6873
6874 buf->f_type = EXT4_SUPER_MAGIC;
6875 buf->f_bsize = sb->s_blocksize;
6876 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
6877 bfree = percpu_counter_sum_positive(fbc: &sbi->s_freeclusters_counter) -
6878 percpu_counter_sum_positive(fbc: &sbi->s_dirtyclusters_counter);
6879 /* prevent underflow in case that few free space is available */
6880 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
6881 buf->f_bavail = buf->f_bfree -
6882 (ext4_r_blocks_count(es) + resv_blocks);
6883 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
6884 buf->f_bavail = 0;
6885 buf->f_files = le32_to_cpu(es->s_inodes_count);
6886 buf->f_ffree = percpu_counter_sum_positive(fbc: &sbi->s_freeinodes_counter);
6887 buf->f_namelen = EXT4_NAME_LEN;
6888 buf->f_fsid = uuid_to_fsid(uuid: es->s_uuid);
6889
6890#ifdef CONFIG_QUOTA
6891 if (ext4_test_inode_flag(inode: dentry->d_inode, bit: EXT4_INODE_PROJINHERIT) &&
6892 sb_has_quota_limits_enabled(sb, type: PRJQUOTA))
6893 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf);
6894#endif
6895 return 0;
6896}
6897
6898
6899#ifdef CONFIG_QUOTA
6900
6901/*
6902 * Helper functions so that transaction is started before we acquire dqio_sem
6903 * to keep correct lock ordering of transaction > dqio_sem
6904 */
6905static inline struct inode *dquot_to_inode(struct dquot *dquot)
6906{
6907 return sb_dqopt(sb: dquot->dq_sb)->files[dquot->dq_id.type];
6908}
6909
6910static int ext4_write_dquot(struct dquot *dquot)
6911{
6912 int ret, err;
6913 handle_t *handle;
6914 struct inode *inode;
6915
6916 inode = dquot_to_inode(dquot);
6917 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
6918 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
6919 if (IS_ERR(ptr: handle))
6920 return PTR_ERR(ptr: handle);
6921 ret = dquot_commit(dquot);
6922 if (ret < 0)
6923 ext4_error_err(dquot->dq_sb, -ret,
6924 "Failed to commit dquot type %d",
6925 dquot->dq_id.type);
6926 err = ext4_journal_stop(handle);
6927 if (!ret)
6928 ret = err;
6929 return ret;
6930}
6931
6932static int ext4_acquire_dquot(struct dquot *dquot)
6933{
6934 int ret, err;
6935 handle_t *handle;
6936
6937 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
6938 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
6939 if (IS_ERR(ptr: handle))
6940 return PTR_ERR(ptr: handle);
6941 ret = dquot_acquire(dquot);
6942 if (ret < 0)
6943 ext4_error_err(dquot->dq_sb, -ret,
6944 "Failed to acquire dquot type %d",
6945 dquot->dq_id.type);
6946 err = ext4_journal_stop(handle);
6947 if (!ret)
6948 ret = err;
6949 return ret;
6950}
6951
6952static int ext4_release_dquot(struct dquot *dquot)
6953{
6954 int ret, err;
6955 handle_t *handle;
6956 bool freeze_protected = false;
6957
6958 /*
6959 * Trying to sb_start_intwrite() in a running transaction
6960 * can result in a deadlock. Further, running transactions
6961 * are already protected from freezing.
6962 */
6963 if (!ext4_journal_current_handle()) {
6964 sb_start_intwrite(sb: dquot->dq_sb);
6965 freeze_protected = true;
6966 }
6967
6968 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
6969 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
6970 if (IS_ERR(ptr: handle)) {
6971 /* Release dquot anyway to avoid endless cycle in dqput() */
6972 dquot_release(dquot);
6973 if (freeze_protected)
6974 sb_end_intwrite(sb: dquot->dq_sb);
6975 return PTR_ERR(ptr: handle);
6976 }
6977 ret = dquot_release(dquot);
6978 if (ret < 0)
6979 ext4_error_err(dquot->dq_sb, -ret,
6980 "Failed to release dquot type %d",
6981 dquot->dq_id.type);
6982 err = ext4_journal_stop(handle);
6983 if (!ret)
6984 ret = err;
6985
6986 if (freeze_protected)
6987 sb_end_intwrite(sb: dquot->dq_sb);
6988
6989 return ret;
6990}
6991
6992static int ext4_mark_dquot_dirty(struct dquot *dquot)
6993{
6994 struct super_block *sb = dquot->dq_sb;
6995
6996 if (ext4_is_quota_journalled(sb)) {
6997 dquot_mark_dquot_dirty(dquot);
6998 return ext4_write_dquot(dquot);
6999 } else {
7000 return dquot_mark_dquot_dirty(dquot);
7001 }
7002}
7003
7004static int ext4_write_info(struct super_block *sb, int type)
7005{
7006 int ret, err;
7007 handle_t *handle;
7008
7009 /* Data block + inode block */
7010 handle = ext4_journal_start_sb(sb, EXT4_HT_QUOTA, 2);
7011 if (IS_ERR(ptr: handle))
7012 return PTR_ERR(ptr: handle);
7013 ret = dquot_commit_info(sb, type);
7014 err = ext4_journal_stop(handle);
7015 if (!ret)
7016 ret = err;
7017 return ret;
7018}
7019
7020static void lockdep_set_quota_inode(struct inode *inode, int subclass)
7021{
7022 struct ext4_inode_info *ei = EXT4_I(inode);
7023
7024 /* The first argument of lockdep_set_subclass has to be
7025 * *exactly* the same as the argument to init_rwsem() --- in
7026 * this case, in init_once() --- or lockdep gets unhappy
7027 * because the name of the lock is set using the
7028 * stringification of the argument to init_rwsem().
7029 */
7030 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */
7031 lockdep_set_subclass(&ei->i_data_sem, subclass);
7032}
7033
7034/*
7035 * Standard function to be called on quota_on
7036 */
7037static int ext4_quota_on(struct super_block *sb, int type, int format_id,
7038 const struct path *path)
7039{
7040 int err;
7041
7042 if (!test_opt(sb, QUOTA))
7043 return -EINVAL;
7044
7045 /* Quotafile not on the same filesystem? */
7046 if (path->dentry->d_sb != sb)
7047 return -EXDEV;
7048
7049 /* Quota already enabled for this file? */
7050 if (IS_NOQUOTA(d_inode(path->dentry)))
7051 return -EBUSY;
7052
7053 /* Journaling quota? */
7054 if (EXT4_SB(sb)->s_qf_names[type]) {
7055 /* Quotafile not in fs root? */
7056 if (path->dentry->d_parent != sb->s_root)
7057 ext4_msg(sb, KERN_WARNING,
7058 "Quota file not on filesystem root. "
7059 "Journaled quota will not work");
7060 sb_dqopt(sb)->flags |= DQUOT_NOLIST_DIRTY;
7061 } else {
7062 /*
7063 * Clear the flag just in case mount options changed since
7064 * last time.
7065 */
7066 sb_dqopt(sb)->flags &= ~DQUOT_NOLIST_DIRTY;
7067 }
7068
7069 lockdep_set_quota_inode(inode: path->dentry->d_inode, subclass: I_DATA_SEM_QUOTA);
7070 err = dquot_quota_on(sb, type, format_id, path);
7071 if (!err) {
7072 struct inode *inode = d_inode(dentry: path->dentry);
7073 handle_t *handle;
7074
7075 /*
7076 * Set inode flags to prevent userspace from messing with quota
7077 * files. If this fails, we return success anyway since quotas
7078 * are already enabled and this is not a hard failure.
7079 */
7080 inode_lock(inode);
7081 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
7082 if (IS_ERR(ptr: handle))
7083 goto unlock_inode;
7084 EXT4_I(inode)->i_flags |= EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL;
7085 inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
7086 S_NOATIME | S_IMMUTABLE);
7087 err = ext4_mark_inode_dirty(handle, inode);
7088 ext4_journal_stop(handle);
7089 unlock_inode:
7090 inode_unlock(inode);
7091 if (err)
7092 dquot_quota_off(sb, type);
7093 }
7094 if (err)
7095 lockdep_set_quota_inode(inode: path->dentry->d_inode,
7096 subclass: I_DATA_SEM_NORMAL);
7097 return err;
7098}
7099
7100static inline bool ext4_check_quota_inum(int type, unsigned long qf_inum)
7101{
7102 switch (type) {
7103 case USRQUOTA:
7104 return qf_inum == EXT4_USR_QUOTA_INO;
7105 case GRPQUOTA:
7106 return qf_inum == EXT4_GRP_QUOTA_INO;
7107 case PRJQUOTA:
7108 return qf_inum >= EXT4_GOOD_OLD_FIRST_INO;
7109 default:
7110 BUG();
7111 }
7112}
7113
7114static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
7115 unsigned int flags)
7116{
7117 int err;
7118 struct inode *qf_inode;
7119 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
7120 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
7121 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
7122 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
7123 };
7124
7125 BUG_ON(!ext4_has_feature_quota(sb));
7126
7127 if (!qf_inums[type])
7128 return -EPERM;
7129
7130 if (!ext4_check_quota_inum(type, qf_inum: qf_inums[type])) {
7131 ext4_error(sb, "Bad quota inum: %lu, type: %d",
7132 qf_inums[type], type);
7133 return -EUCLEAN;
7134 }
7135
7136 qf_inode = ext4_iget(sb, qf_inums[type], EXT4_IGET_SPECIAL);
7137 if (IS_ERR(ptr: qf_inode)) {
7138 ext4_error(sb, "Bad quota inode: %lu, type: %d",
7139 qf_inums[type], type);
7140 return PTR_ERR(ptr: qf_inode);
7141 }
7142
7143 /* Don't account quota for quota files to avoid recursion */
7144 qf_inode->i_flags |= S_NOQUOTA;
7145 lockdep_set_quota_inode(inode: qf_inode, subclass: I_DATA_SEM_QUOTA);
7146 err = dquot_load_quota_inode(inode: qf_inode, type, format_id, flags);
7147 if (err)
7148 lockdep_set_quota_inode(inode: qf_inode, subclass: I_DATA_SEM_NORMAL);
7149 iput(qf_inode);
7150
7151 return err;
7152}
7153
7154/* Enable usage tracking for all quota types. */
7155int ext4_enable_quotas(struct super_block *sb)
7156{
7157 int type, err = 0;
7158 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
7159 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
7160 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
7161 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
7162 };
7163 bool quota_mopt[EXT4_MAXQUOTAS] = {
7164 test_opt(sb, USRQUOTA),
7165 test_opt(sb, GRPQUOTA),
7166 test_opt(sb, PRJQUOTA),
7167 };
7168
7169 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NOLIST_DIRTY;
7170 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
7171 if (qf_inums[type]) {
7172 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
7173 DQUOT_USAGE_ENABLED |
7174 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
7175 if (err) {
7176 ext4_warning(sb,
7177 "Failed to enable quota tracking "
7178 "(type=%d, err=%d, ino=%lu). "
7179 "Please run e2fsck to fix.", type,
7180 err, qf_inums[type]);
7181
7182 ext4_quotas_off(sb, type);
7183 return err;
7184 }
7185 }
7186 }
7187 return 0;
7188}
7189
7190static int ext4_quota_off(struct super_block *sb, int type)
7191{
7192 struct inode *inode = sb_dqopt(sb)->files[type];
7193 handle_t *handle;
7194 int err;
7195
7196 /* Force all delayed allocation blocks to be allocated.
7197 * Caller already holds s_umount sem */
7198 if (test_opt(sb, DELALLOC))
7199 sync_filesystem(sb);
7200
7201 if (!inode || !igrab(inode))
7202 goto out;
7203
7204 err = dquot_quota_off(sb, type);
7205 if (err || ext4_has_feature_quota(sb))
7206 goto out_put;
7207 /*
7208 * When the filesystem was remounted read-only first, we cannot cleanup
7209 * inode flags here. Bad luck but people should be using QUOTA feature
7210 * these days anyway.
7211 */
7212 if (sb_rdonly(sb))
7213 goto out_put;
7214
7215 inode_lock(inode);
7216 /*
7217 * Update modification times of quota files when userspace can
7218 * start looking at them. If we fail, we return success anyway since
7219 * this is not a hard failure and quotas are already disabled.
7220 */
7221 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
7222 if (IS_ERR(ptr: handle)) {
7223 err = PTR_ERR(ptr: handle);
7224 goto out_unlock;
7225 }
7226 EXT4_I(inode)->i_flags &= ~(EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL);
7227 inode_set_flags(inode, flags: 0, S_NOATIME | S_IMMUTABLE);
7228 inode_set_mtime_to_ts(inode, ts: inode_set_ctime_current(inode));
7229 err = ext4_mark_inode_dirty(handle, inode);
7230 ext4_journal_stop(handle);
7231out_unlock:
7232 inode_unlock(inode);
7233out_put:
7234 lockdep_set_quota_inode(inode, subclass: I_DATA_SEM_NORMAL);
7235 iput(inode);
7236 return err;
7237out:
7238 return dquot_quota_off(sb, type);
7239}
7240
7241/* Read data from quotafile - avoid pagecache and such because we cannot afford
7242 * acquiring the locks... As quota files are never truncated and quota code
7243 * itself serializes the operations (and no one else should touch the files)
7244 * we don't have to be afraid of races */
7245static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
7246 size_t len, loff_t off)
7247{
7248 struct inode *inode = sb_dqopt(sb)->files[type];
7249 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
7250 int offset = off & (sb->s_blocksize - 1);
7251 int tocopy;
7252 size_t toread;
7253 struct buffer_head *bh;
7254 loff_t i_size = i_size_read(inode);
7255
7256 if (off > i_size)
7257 return 0;
7258 if (off+len > i_size)
7259 len = i_size-off;
7260 toread = len;
7261 while (toread > 0) {
7262 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread);
7263 bh = ext4_bread(NULL, inode, blk, 0);
7264 if (IS_ERR(ptr: bh))
7265 return PTR_ERR(ptr: bh);
7266 if (!bh) /* A hole? */
7267 memset(s: data, c: 0, n: tocopy);
7268 else
7269 memcpy(to: data, from: bh->b_data+offset, len: tocopy);
7270 brelse(bh);
7271 offset = 0;
7272 toread -= tocopy;
7273 data += tocopy;
7274 blk++;
7275 }
7276 return len;
7277}
7278
7279/* Write to quotafile (we know the transaction is already started and has
7280 * enough credits) */
7281static ssize_t ext4_quota_write(struct super_block *sb, int type,
7282 const char *data, size_t len, loff_t off)
7283{
7284 struct inode *inode = sb_dqopt(sb)->files[type];
7285 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
7286 int err = 0, err2 = 0, offset = off & (sb->s_blocksize - 1);
7287 int retries = 0;
7288 struct buffer_head *bh;
7289 handle_t *handle = journal_current_handle();
7290
7291 if (!handle) {
7292 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
7293 " cancelled because transaction is not started",
7294 (unsigned long long)off, (unsigned long long)len);
7295 return -EIO;
7296 }
7297 /*
7298 * Since we account only one data block in transaction credits,
7299 * then it is impossible to cross a block boundary.
7300 */
7301 if (sb->s_blocksize - offset < len) {
7302 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
7303 " cancelled because not block aligned",
7304 (unsigned long long)off, (unsigned long long)len);
7305 return -EIO;
7306 }
7307
7308 do {
7309 bh = ext4_bread(handle, inode, blk,
7310 EXT4_GET_BLOCKS_CREATE |
7311 EXT4_GET_BLOCKS_METADATA_NOFAIL);
7312 } while (PTR_ERR(ptr: bh) == -ENOSPC &&
7313 ext4_should_retry_alloc(sb: inode->i_sb, retries: &retries));
7314 if (IS_ERR(ptr: bh))
7315 return PTR_ERR(ptr: bh);
7316 if (!bh)
7317 goto out;
7318 BUFFER_TRACE(bh, "get write access");
7319 err = ext4_journal_get_write_access(handle, sb, bh, EXT4_JTR_NONE);
7320 if (err) {
7321 brelse(bh);
7322 return err;
7323 }
7324 lock_buffer(bh);
7325 memcpy(to: bh->b_data+offset, from: data, len);
7326 flush_dcache_folio(folio: bh->b_folio);
7327 unlock_buffer(bh);
7328 err = ext4_handle_dirty_metadata(handle, NULL, bh);
7329 brelse(bh);
7330out:
7331 if (inode->i_size < off + len) {
7332 i_size_write(inode, i_size: off + len);
7333 EXT4_I(inode)->i_disksize = inode->i_size;
7334 err2 = ext4_mark_inode_dirty(handle, inode);
7335 if (unlikely(err2 && !err))
7336 err = err2;
7337 }
7338 return err ? err : len;
7339}
7340#endif
7341
7342#if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
7343static inline void register_as_ext2(void)
7344{
7345 int err = register_filesystem(&ext2_fs_type);
7346 if (err)
7347 printk(KERN_WARNING
7348 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
7349}
7350
7351static inline void unregister_as_ext2(void)
7352{
7353 unregister_filesystem(&ext2_fs_type);
7354}
7355
7356static inline int ext2_feature_set_ok(struct super_block *sb)
7357{
7358 if (ext4_has_unknown_ext2_incompat_features(sb))
7359 return 0;
7360 if (sb_rdonly(sb))
7361 return 1;
7362 if (ext4_has_unknown_ext2_ro_compat_features(sb))
7363 return 0;
7364 return 1;
7365}
7366#else
7367static inline void register_as_ext2(void) { }
7368static inline void unregister_as_ext2(void) { }
7369static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
7370#endif
7371
7372static inline void register_as_ext3(void)
7373{
7374 int err = register_filesystem(&ext3_fs_type);
7375 if (err)
7376 printk(KERN_WARNING
7377 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
7378}
7379
7380static inline void unregister_as_ext3(void)
7381{
7382 unregister_filesystem(&ext3_fs_type);
7383}
7384
7385static inline int ext3_feature_set_ok(struct super_block *sb)
7386{
7387 if (ext4_has_unknown_ext3_incompat_features(sb))
7388 return 0;
7389 if (!ext4_has_feature_journal(sb))
7390 return 0;
7391 if (sb_rdonly(sb))
7392 return 1;
7393 if (ext4_has_unknown_ext3_ro_compat_features(sb))
7394 return 0;
7395 return 1;
7396}
7397
7398static void ext4_kill_sb(struct super_block *sb)
7399{
7400 struct ext4_sb_info *sbi = EXT4_SB(sb);
7401 struct file *bdev_file = sbi ? sbi->s_journal_bdev_file : NULL;
7402
7403 kill_block_super(sb);
7404
7405 if (bdev_file)
7406 bdev_fput(bdev_file);
7407}
7408
7409static struct file_system_type ext4_fs_type = {
7410 .owner = THIS_MODULE,
7411 .name = "ext4",
7412 .init_fs_context = ext4_init_fs_context,
7413 .parameters = ext4_param_specs,
7414 .kill_sb = ext4_kill_sb,
7415 .fs_flags = FS_REQUIRES_DEV | FS_ALLOW_IDMAP | FS_MGTIME,
7416};
7417MODULE_ALIAS_FS("ext4");
7418
7419static int __init ext4_init_fs(void)
7420{
7421 int err;
7422
7423 ratelimit_state_init(rs: &ext4_mount_msg_ratelimit, interval: 30 * HZ, burst: 64);
7424 ext4_li_info = NULL;
7425
7426 /* Build-time check for flags consistency */
7427 ext4_check_flag_values();
7428
7429 err = ext4_init_es();
7430 if (err)
7431 return err;
7432
7433 err = ext4_init_pending();
7434 if (err)
7435 goto out7;
7436
7437 err = ext4_init_post_read_processing();
7438 if (err)
7439 goto out6;
7440
7441 err = ext4_init_pageio();
7442 if (err)
7443 goto out5;
7444
7445 err = ext4_init_system_zone();
7446 if (err)
7447 goto out4;
7448
7449 err = ext4_init_sysfs();
7450 if (err)
7451 goto out3;
7452
7453 err = ext4_init_mballoc();
7454 if (err)
7455 goto out2;
7456 err = init_inodecache();
7457 if (err)
7458 goto out1;
7459
7460 err = ext4_fc_init_dentry_cache();
7461 if (err)
7462 goto out05;
7463
7464 register_as_ext3();
7465 register_as_ext2();
7466 err = register_filesystem(&ext4_fs_type);
7467 if (err)
7468 goto out;
7469
7470 return 0;
7471out:
7472 unregister_as_ext2();
7473 unregister_as_ext3();
7474 ext4_fc_destroy_dentry_cache();
7475out05:
7476 destroy_inodecache();
7477out1:
7478 ext4_exit_mballoc();
7479out2:
7480 ext4_exit_sysfs();
7481out3:
7482 ext4_exit_system_zone();
7483out4:
7484 ext4_exit_pageio();
7485out5:
7486 ext4_exit_post_read_processing();
7487out6:
7488 ext4_exit_pending();
7489out7:
7490 ext4_exit_es();
7491
7492 return err;
7493}
7494
7495static void __exit ext4_exit_fs(void)
7496{
7497 ext4_destroy_lazyinit_thread();
7498 unregister_as_ext2();
7499 unregister_as_ext3();
7500 unregister_filesystem(&ext4_fs_type);
7501 ext4_fc_destroy_dentry_cache();
7502 destroy_inodecache();
7503 ext4_exit_mballoc();
7504 ext4_exit_sysfs();
7505 ext4_exit_system_zone();
7506 ext4_exit_pageio();
7507 ext4_exit_post_read_processing();
7508 ext4_exit_es();
7509 ext4_exit_pending();
7510}
7511
7512MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
7513MODULE_DESCRIPTION("Fourth Extended Filesystem");
7514MODULE_LICENSE("GPL");
7515module_init(ext4_init_fs)
7516module_exit(ext4_exit_fs)
7517