]> git.ipfire.org Git - thirdparty/kernel/linux.git/blob - fs/ext4/super.c
projects / thirdparty / kernel / linux.git / blob
? search:


c7d39da7e733b13b2760da3341c8991401043a0b
[thirdparty/kernel/linux.git] / fs / ext4 / super.c
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
63 static struct ext4_lazy_init *ext4_li_info;
64 static DEFINE_MUTEX(ext4_li_mtx);
65 static struct ratelimit_state ext4_mount_msg_ratelimit;
66
67 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
68 unsigned long journal_devnum);
69 static int ext4_show_options(struct seq_file *seq, struct dentry *root);
70 static void ext4_update_super(struct super_block *sb);
71 static int ext4_commit_super(struct super_block *sb);
72 static int ext4_mark_recovery_complete(struct super_block *sb,
73 struct ext4_super_block *es);
74 static int ext4_clear_journal_err(struct super_block *sb,
75 struct ext4_super_block *es);
76 static int ext4_sync_fs(struct super_block *sb, int wait);
77 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
78 static int ext4_unfreeze(struct super_block *sb);
79 static int ext4_freeze(struct super_block *sb);
80 static inline int ext2_feature_set_ok(struct super_block *sb);
81 static inline int ext3_feature_set_ok(struct super_block *sb);
82 static void ext4_unregister_li_request(struct super_block *sb);
83 static void ext4_clear_request_list(void);
84 static struct inode *ext4_get_journal_inode(struct super_block *sb,
85 unsigned int journal_inum);
86 static int ext4_validate_options(struct fs_context *fc);
87 static int ext4_check_opt_consistency(struct fs_context *fc,
88 struct super_block *sb);
89 static void ext4_apply_options(struct fs_context *fc, struct super_block *sb);
90 static int ext4_parse_param(struct fs_context *fc, struct fs_parameter *param);
91 static int ext4_get_tree(struct fs_context *fc);
92 static int ext4_reconfigure(struct fs_context *fc);
93 static void ext4_fc_free(struct fs_context *fc);
94 static int ext4_init_fs_context(struct fs_context *fc);
95 static void ext4_kill_sb(struct super_block *sb);
96 static 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
124 static 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)
133 static 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 };
141 MODULE_ALIAS_FS("ext2");
142 MODULE_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
149 static 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 };
157 MODULE_ALIAS_FS("ext3");
158 MODULE_ALIAS("ext3");
159 #define IS_EXT3_SB(sb) ((sb)->s_type == &ext3_fs_type)
160
161
162 static 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
183 void 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
195 int 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
213 int 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, false);
218 return 0;
219 }
220 return ext4_read_bh(bh, op_flags, NULL, 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 */
229 static 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(-ENOMEM);
239 if (ext4_buffer_uptodate(bh))
240 return bh;
241
242 ret = ext4_read_bh_lock(bh, REQ_META | op_flags, true);
243 if (ret) {
244 put_bh(bh);
245 return ERR_PTR(ret);
246 }
247 return bh;
248 }
249
250 struct 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(sb->s_bdev->bd_mapping,
254 ~__GFP_FS) | __GFP_MOVABLE;
255
256 return __ext4_sb_bread_gfp(sb, block, op_flags, gfp);
257 }
258
259 struct buffer_head *ext4_sb_bread_unmovable(struct super_block *sb,
260 sector_t block)
261 {
262 gfp_t gfp = mapping_gfp_constraint(sb->s_bdev->bd_mapping,
263 ~__GFP_FS);
264
265 return __ext4_sb_bread_gfp(sb, block, 0, gfp);
266 }
267
268 void ext4_sb_breadahead_unmovable(struct super_block *sb, sector_t block)
269 {
270 struct buffer_head *bh = bdev_getblk(sb->s_bdev, block,
271 sb->s_blocksize, GFP_NOWAIT | __GFP_NOWARN);
272
273 if (likely(bh)) {
274 if (trylock_buffer(bh))
275 ext4_read_bh_nowait(bh, REQ_RAHEAD, NULL, false);
276 brelse(bh);
277 }
278 }
279
280 static int ext4_verify_csum_type(struct super_block *sb,
281 struct ext4_super_block *es)
282 {
283 if (!ext4_has_feature_metadata_csum(sb))
284 return 1;
285
286 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
287 }
288
289 __le32 ext4_superblock_csum(struct ext4_super_block *es)
290 {
291 int offset = offsetof(struct ext4_super_block, s_checksum);
292 __u32 csum;
293
294 csum = ext4_chksum(~0, (char *)es, offset);
295
296 return cpu_to_le32(csum);
297 }
298
299 static int ext4_superblock_csum_verify(struct super_block *sb,
300 struct ext4_super_block *es)
301 {
302 if (!ext4_has_feature_metadata_csum(sb))
303 return 1;
304
305 return es->s_checksum == ext4_superblock_csum(es);
306 }
307
308 void ext4_superblock_csum_set(struct super_block *sb)
309 {
310 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
311
312 if (!ext4_has_feature_metadata_csum(sb))
313 return;
314
315 es->s_checksum = ext4_superblock_csum(es);
316 }
317
318 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
319 struct ext4_group_desc *bg)
320 {
321 return le32_to_cpu(bg->bg_block_bitmap_lo) |
322 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
323 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
324 }
325
326 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
327 struct ext4_group_desc *bg)
328 {
329 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
330 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
331 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
332 }
333
334 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
335 struct ext4_group_desc *bg)
336 {
337 return le32_to_cpu(bg->bg_inode_table_lo) |
338 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
339 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
340 }
341
342 __u32 ext4_free_group_clusters(struct super_block *sb,
343 struct ext4_group_desc *bg)
344 {
345 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
346 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
347 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
348 }
349
350 __u32 ext4_free_inodes_count(struct super_block *sb,
351 struct ext4_group_desc *bg)
352 {
353 return le16_to_cpu(READ_ONCE(bg->bg_free_inodes_count_lo)) |
354 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
355 (__u32)le16_to_cpu(READ_ONCE(bg->bg_free_inodes_count_hi)) << 16 : 0);
356 }
357
358 __u32 ext4_used_dirs_count(struct super_block *sb,
359 struct ext4_group_desc *bg)
360 {
361 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
362 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
363 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
364 }
365
366 __u32 ext4_itable_unused_count(struct super_block *sb,
367 struct ext4_group_desc *bg)
368 {
369 return le16_to_cpu(bg->bg_itable_unused_lo) |
370 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
371 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
372 }
373
374 void ext4_block_bitmap_set(struct super_block *sb,
375 struct ext4_group_desc *bg, ext4_fsblk_t blk)
376 {
377 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
378 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
379 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
380 }
381
382 void ext4_inode_bitmap_set(struct super_block *sb,
383 struct ext4_group_desc *bg, ext4_fsblk_t blk)
384 {
385 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
386 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
387 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
388 }
389
390 void ext4_inode_table_set(struct super_block *sb,
391 struct ext4_group_desc *bg, ext4_fsblk_t blk)
392 {
393 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
394 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
395 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
396 }
397
398 void ext4_free_group_clusters_set(struct super_block *sb,
399 struct ext4_group_desc *bg, __u32 count)
400 {
401 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
402 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
403 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
404 }
405
406 void ext4_free_inodes_set(struct super_block *sb,
407 struct ext4_group_desc *bg, __u32 count)
408 {
409 WRITE_ONCE(bg->bg_free_inodes_count_lo, cpu_to_le16((__u16)count));
410 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
411 WRITE_ONCE(bg->bg_free_inodes_count_hi, cpu_to_le16(count >> 16));
412 }
413
414 void ext4_used_dirs_set(struct super_block *sb,
415 struct ext4_group_desc *bg, __u32 count)
416 {
417 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
418 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
419 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
420 }
421
422 void ext4_itable_unused_set(struct super_block *sb,
423 struct ext4_group_desc *bg, __u32 count)
424 {
425 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
426 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
427 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
428 }
429
430 static void __ext4_update_tstamp(__le32 *lo, __u8 *hi, time64_t now)
431 {
432 now = clamp_val(now, 0, (1ull << 40) - 1);
433
434 *lo = cpu_to_le32(lower_32_bits(now));
435 *hi = upper_32_bits(now);
436 }
437
438 static time64_t __ext4_get_tstamp(__le32 *lo, __u8 *hi)
439 {
440 return ((time64_t)(*hi) << 32) + le32_to_cpu(*lo);
441 }
442 #define ext4_update_tstamp(es, tstamp) \
443 __ext4_update_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi, \
444 ktime_get_real_seconds())
445 #define ext4_get_tstamp(es, tstamp) \
446 __ext4_get_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi)
447
448 /*
449 * The ext4_maybe_update_superblock() function checks and updates the
450 * superblock if needed.
451 *
452 * This function is designed to update the on-disk superblock only under
453 * certain conditions to prevent excessive disk writes and unnecessary
454 * waking of the disk from sleep. The superblock will be updated if:
455 * 1. More than sbi->s_sb_update_sec (def: 1 hour) has passed since the last
456 * superblock update
457 * 2. More than sbi->s_sb_update_kb (def: 16MB) kbs have been written since the
458 * last superblock update.
459 *
460 * @sb: The superblock
461 */
462 static void ext4_maybe_update_superblock(struct super_block *sb)
463 {
464 struct ext4_sb_info *sbi = EXT4_SB(sb);
465 struct ext4_super_block *es = sbi->s_es;
466 journal_t *journal = sbi->s_journal;
467 time64_t now;
468 __u64 last_update;
469 __u64 lifetime_write_kbytes;
470 __u64 diff_size;
471
472 if (ext4_emergency_state(sb) || sb_rdonly(sb) ||
473 !(sb->s_flags & SB_ACTIVE) || !journal ||
474 journal->j_flags & JBD2_UNMOUNT)
475 return;
476
477 now = ktime_get_real_seconds();
478 last_update = ext4_get_tstamp(es, s_wtime);
479
480 if (likely(now - last_update < sbi->s_sb_update_sec))
481 return;
482
483 lifetime_write_kbytes = sbi->s_kbytes_written +
484 ((part_stat_read(sb->s_bdev, sectors[STAT_WRITE]) -
485 sbi->s_sectors_written_start) >> 1);
486
487 /* Get the number of kilobytes not written to disk to account
488 * for statistics and compare with a multiple of 16 MB. This
489 * is used to determine when the next superblock commit should
490 * occur (i.e. not more often than once per 16MB if there was
491 * less written in an hour).
492 */
493 diff_size = lifetime_write_kbytes - le64_to_cpu(es->s_kbytes_written);
494
495 if (diff_size > sbi->s_sb_update_kb)
496 schedule_work(&EXT4_SB(sb)->s_sb_upd_work);
497 }
498
499 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
500 {
501 struct super_block *sb = journal->j_private;
502
503 BUG_ON(txn->t_state == T_FINISHED);
504
505 ext4_process_freed_data(sb, txn->t_tid);
506 ext4_maybe_update_superblock(sb);
507 }
508
509 static bool ext4_journalled_writepage_needs_redirty(struct jbd2_inode *jinode,
510 struct folio *folio)
511 {
512 struct buffer_head *bh, *head;
513 struct journal_head *jh;
514
515 bh = head = folio_buffers(folio);
516 do {
517 /*
518 * We have to redirty a page in these cases:
519 * 1) If buffer is dirty, it means the page was dirty because it
520 * contains a buffer that needs checkpointing. So the dirty bit
521 * needs to be preserved so that checkpointing writes the buffer
522 * properly.
523 * 2) If buffer is not part of the committing transaction
524 * (we may have just accidentally come across this buffer because
525 * inode range tracking is not exact) or if the currently running
526 * transaction already contains this buffer as well, dirty bit
527 * needs to be preserved so that the buffer gets writeprotected
528 * properly on running transaction's commit.
529 */
530 jh = bh2jh(bh);
531 if (buffer_dirty(bh) ||
532 (jh && (jh->b_transaction != jinode->i_transaction ||
533 jh->b_next_transaction)))
534 return true;
535 } while ((bh = bh->b_this_page) != head);
536
537 return false;
538 }
539
540 static int ext4_journalled_submit_inode_data_buffers(struct jbd2_inode *jinode)
541 {
542 struct address_space *mapping = jinode->i_vfs_inode->i_mapping;
543 struct writeback_control wbc = {
544 .sync_mode = WB_SYNC_ALL,
545 .nr_to_write = LONG_MAX,
546 .range_start = jinode->i_dirty_start,
547 .range_end = jinode->i_dirty_end,
548 };
549 struct folio *folio = NULL;
550 int error;
551
552 /*
553 * writeback_iter() already checks for dirty pages and calls
554 * folio_clear_dirty_for_io(), which we want to write protect the
555 * folios.
556 *
557 * However, we may have to redirty a folio sometimes.
558 */
559 while ((folio = writeback_iter(mapping, &wbc, folio, &error))) {
560 if (ext4_journalled_writepage_needs_redirty(jinode, folio))
561 folio_redirty_for_writepage(&wbc, folio);
562 folio_unlock(folio);
563 }
564
565 return error;
566 }
567
568 static int ext4_journal_submit_inode_data_buffers(struct jbd2_inode *jinode)
569 {
570 int ret;
571
572 if (ext4_should_journal_data(jinode->i_vfs_inode))
573 ret = ext4_journalled_submit_inode_data_buffers(jinode);
574 else
575 ret = ext4_normal_submit_inode_data_buffers(jinode);
576 return ret;
577 }
578
579 static int ext4_journal_finish_inode_data_buffers(struct jbd2_inode *jinode)
580 {
581 int ret = 0;
582
583 if (!ext4_should_journal_data(jinode->i_vfs_inode))
584 ret = jbd2_journal_finish_inode_data_buffers(jinode);
585
586 return ret;
587 }
588
589 static bool system_going_down(void)
590 {
591 return system_state == SYSTEM_HALT || system_state == SYSTEM_POWER_OFF
592 || system_state == SYSTEM_RESTART;
593 }
594
595 struct ext4_err_translation {
596 int code;
597 int errno;
598 };
599
600 #define EXT4_ERR_TRANSLATE(err) { .code = EXT4_ERR_##err, .errno = err }
601
602 static struct ext4_err_translation err_translation[] = {
603 EXT4_ERR_TRANSLATE(EIO),
604 EXT4_ERR_TRANSLATE(ENOMEM),
605 EXT4_ERR_TRANSLATE(EFSBADCRC),
606 EXT4_ERR_TRANSLATE(EFSCORRUPTED),
607 EXT4_ERR_TRANSLATE(ENOSPC),
608 EXT4_ERR_TRANSLATE(ENOKEY),
609 EXT4_ERR_TRANSLATE(EROFS),
610 EXT4_ERR_TRANSLATE(EFBIG),
611 EXT4_ERR_TRANSLATE(EEXIST),
612 EXT4_ERR_TRANSLATE(ERANGE),
613 EXT4_ERR_TRANSLATE(EOVERFLOW),
614 EXT4_ERR_TRANSLATE(EBUSY),
615 EXT4_ERR_TRANSLATE(ENOTDIR),
616 EXT4_ERR_TRANSLATE(ENOTEMPTY),
617 EXT4_ERR_TRANSLATE(ESHUTDOWN),
618 EXT4_ERR_TRANSLATE(EFAULT),
619 };
620
621 static int ext4_errno_to_code(int errno)
622 {
623 int i;
624
625 for (i = 0; i < ARRAY_SIZE(err_translation); i++)
626 if (err_translation[i].errno == errno)
627 return err_translation[i].code;
628 return EXT4_ERR_UNKNOWN;
629 }
630
631 static void save_error_info(struct super_block *sb, int error,
632 __u32 ino, __u64 block,
633 const char *func, unsigned int line)
634 {
635 struct ext4_sb_info *sbi = EXT4_SB(sb);
636
637 /* We default to EFSCORRUPTED error... */
638 if (error == 0)
639 error = EFSCORRUPTED;
640
641 spin_lock(&sbi->s_error_lock);
642 sbi->s_add_error_count++;
643 sbi->s_last_error_code = error;
644 sbi->s_last_error_line = line;
645 sbi->s_last_error_ino = ino;
646 sbi->s_last_error_block = block;
647 sbi->s_last_error_func = func;
648 sbi->s_last_error_time = ktime_get_real_seconds();
649 if (!sbi->s_first_error_time) {
650 sbi->s_first_error_code = error;
651 sbi->s_first_error_line = line;
652 sbi->s_first_error_ino = ino;
653 sbi->s_first_error_block = block;
654 sbi->s_first_error_func = func;
655 sbi->s_first_error_time = sbi->s_last_error_time;
656 }
657 spin_unlock(&sbi->s_error_lock);
658 }
659
660 /* Deal with the reporting of failure conditions on a filesystem such as
661 * inconsistencies detected or read IO failures.
662 *
663 * On ext2, we can store the error state of the filesystem in the
664 * superblock. That is not possible on ext4, because we may have other
665 * write ordering constraints on the superblock which prevent us from
666 * writing it out straight away; and given that the journal is about to
667 * be aborted, we can't rely on the current, or future, transactions to
668 * write out the superblock safely.
669 *
670 * We'll just use the jbd2_journal_abort() error code to record an error in
671 * the journal instead. On recovery, the journal will complain about
672 * that error until we've noted it down and cleared it.
673 *
674 * If force_ro is set, we unconditionally force the filesystem into an
675 * ABORT|READONLY state, unless the error response on the fs has been set to
676 * panic in which case we take the easy way out and panic immediately. This is
677 * used to deal with unrecoverable failures such as journal IO errors or ENOMEM
678 * at a critical moment in log management.
679 */
680 static void ext4_handle_error(struct super_block *sb, bool force_ro, int error,
681 __u32 ino, __u64 block,
682 const char *func, unsigned int line)
683 {
684 journal_t *journal = EXT4_SB(sb)->s_journal;
685 bool continue_fs = !force_ro && test_opt(sb, ERRORS_CONT);
686
687 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
688 if (test_opt(sb, WARN_ON_ERROR))
689 WARN_ON_ONCE(1);
690
691 if (!continue_fs && !ext4_emergency_ro(sb) && journal)
692 jbd2_journal_abort(journal, -EIO);
693
694 if (!bdev_read_only(sb->s_bdev)) {
695 save_error_info(sb, error, ino, block, func, line);
696 /*
697 * In case the fs should keep running, we need to writeout
698 * superblock through the journal. Due to lock ordering
699 * constraints, it may not be safe to do it right here so we
700 * defer superblock flushing to a workqueue. We just need to be
701 * careful when the journal is already shutting down. If we get
702 * here in that case, just update the sb directly as the last
703 * transaction won't commit anyway.
704 */
705 if (continue_fs && journal &&
706 !ext4_test_mount_flag(sb, EXT4_MF_JOURNAL_DESTROY))
707 schedule_work(&EXT4_SB(sb)->s_sb_upd_work);
708 else
709 ext4_commit_super(sb);
710 }
711
712 /*
713 * We force ERRORS_RO behavior when system is rebooting. Otherwise we
714 * could panic during 'reboot -f' as the underlying device got already
715 * disabled.
716 */
717 if (test_opt(sb, ERRORS_PANIC) && !system_going_down()) {
718 panic("EXT4-fs (device %s): panic forced after error\n",
719 sb->s_id);
720 }
721
722 if (ext4_emergency_ro(sb) || continue_fs)
723 return;
724
725 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
726 /*
727 * We don't set SB_RDONLY because that requires sb->s_umount
728 * semaphore and setting it without proper remount procedure is
729 * confusing code such as freeze_super() leading to deadlocks
730 * and other problems.
731 */
732 set_bit(EXT4_FLAGS_EMERGENCY_RO, &EXT4_SB(sb)->s_ext4_flags);
733 }
734
735 static void update_super_work(struct work_struct *work)
736 {
737 struct ext4_sb_info *sbi = container_of(work, struct ext4_sb_info,
738 s_sb_upd_work);
739 journal_t *journal = sbi->s_journal;
740 handle_t *handle;
741
742 /*
743 * If the journal is still running, we have to write out superblock
744 * through the journal to avoid collisions of other journalled sb
745 * updates.
746 *
747 * We use directly jbd2 functions here to avoid recursing back into
748 * ext4 error handling code during handling of previous errors.
749 */
750 if (!ext4_emergency_state(sbi->s_sb) &&
751 !sb_rdonly(sbi->s_sb) && journal) {
752 struct buffer_head *sbh = sbi->s_sbh;
753 bool call_notify_err = false;
754
755 handle = jbd2_journal_start(journal, 1);
756 if (IS_ERR(handle))
757 goto write_directly;
758 if (jbd2_journal_get_write_access(handle, sbh)) {
759 jbd2_journal_stop(handle);
760 goto write_directly;
761 }
762
763 if (sbi->s_add_error_count > 0)
764 call_notify_err = true;
765
766 ext4_update_super(sbi->s_sb);
767 if (buffer_write_io_error(sbh) || !buffer_uptodate(sbh)) {
768 ext4_msg(sbi->s_sb, KERN_ERR, "previous I/O error to "
769 "superblock detected");
770 clear_buffer_write_io_error(sbh);
771 set_buffer_uptodate(sbh);
772 }
773
774 if (jbd2_journal_dirty_metadata(handle, sbh)) {
775 jbd2_journal_stop(handle);
776 goto write_directly;
777 }
778 jbd2_journal_stop(handle);
779
780 if (call_notify_err)
781 ext4_notify_error_sysfs(sbi);
782
783 return;
784 }
785 write_directly:
786 /*
787 * Write through journal failed. Write sb directly to get error info
788 * out and hope for the best.
789 */
790 ext4_commit_super(sbi->s_sb);
791 ext4_notify_error_sysfs(sbi);
792 }
793
794 #define ext4_error_ratelimit(sb) \
795 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
796 "EXT4-fs error")
797
798 void __ext4_error(struct super_block *sb, const char *function,
799 unsigned int line, bool force_ro, int error, __u64 block,
800 const char *fmt, ...)
801 {
802 struct va_format vaf;
803 va_list args;
804
805 if (unlikely(ext4_emergency_state(sb)))
806 return;
807
808 trace_ext4_error(sb, function, line);
809 if (ext4_error_ratelimit(sb)) {
810 va_start(args, fmt);
811 vaf.fmt = fmt;
812 vaf.va = &args;
813 printk(KERN_CRIT
814 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
815 sb->s_id, function, line, current->comm, &vaf);
816 va_end(args);
817 }
818 fsnotify_sb_error(sb, NULL, error ? error : EFSCORRUPTED);
819
820 ext4_handle_error(sb, force_ro, error, 0, block, function, line);
821 }
822
823 void __ext4_error_inode(struct inode *inode, const char *function,
824 unsigned int line, ext4_fsblk_t block, int error,
825 const char *fmt, ...)
826 {
827 va_list args;
828 struct va_format vaf;
829
830 if (unlikely(ext4_emergency_state(inode->i_sb)))
831 return;
832
833 trace_ext4_error(inode->i_sb, function, line);
834 if (ext4_error_ratelimit(inode->i_sb)) {
835 va_start(args, fmt);
836 vaf.fmt = fmt;
837 vaf.va = &args;
838 if (block)
839 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
840 "inode #%lu: block %llu: comm %s: %pV\n",
841 inode->i_sb->s_id, function, line, inode->i_ino,
842 block, current->comm, &vaf);
843 else
844 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
845 "inode #%lu: comm %s: %pV\n",
846 inode->i_sb->s_id, function, line, inode->i_ino,
847 current->comm, &vaf);
848 va_end(args);
849 }
850 fsnotify_sb_error(inode->i_sb, inode, error ? error : EFSCORRUPTED);
851
852 ext4_handle_error(inode->i_sb, false, error, inode->i_ino, block,
853 function, line);
854 }
855
856 void __ext4_error_file(struct file *file, const char *function,
857 unsigned int line, ext4_fsblk_t block,
858 const char *fmt, ...)
859 {
860 va_list args;
861 struct va_format vaf;
862 struct inode *inode = file_inode(file);
863 char pathname[80], *path;
864
865 if (unlikely(ext4_emergency_state(inode->i_sb)))
866 return;
867
868 trace_ext4_error(inode->i_sb, function, line);
869 if (ext4_error_ratelimit(inode->i_sb)) {
870 path = file_path(file, pathname, sizeof(pathname));
871 if (IS_ERR(path))
872 path = "(unknown)";
873 va_start(args, fmt);
874 vaf.fmt = fmt;
875 vaf.va = &args;
876 if (block)
877 printk(KERN_CRIT
878 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
879 "block %llu: comm %s: path %s: %pV\n",
880 inode->i_sb->s_id, function, line, inode->i_ino,
881 block, current->comm, path, &vaf);
882 else
883 printk(KERN_CRIT
884 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
885 "comm %s: path %s: %pV\n",
886 inode->i_sb->s_id, function, line, inode->i_ino,
887 current->comm, path, &vaf);
888 va_end(args);
889 }
890 fsnotify_sb_error(inode->i_sb, inode, EFSCORRUPTED);
891
892 ext4_handle_error(inode->i_sb, false, EFSCORRUPTED, inode->i_ino, block,
893 function, line);
894 }
895
896 const char *ext4_decode_error(struct super_block *sb, int errno,
897 char nbuf[16])
898 {
899 char *errstr = NULL;
900
901 switch (errno) {
902 case -EFSCORRUPTED:
903 errstr = "Corrupt filesystem";
904 break;
905 case -EFSBADCRC:
906 errstr = "Filesystem failed CRC";
907 break;
908 case -EIO:
909 errstr = "IO failure";
910 break;
911 case -ENOMEM:
912 errstr = "Out of memory";
913 break;
914 case -EROFS:
915 if (!sb || (EXT4_SB(sb)->s_journal &&
916 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
917 errstr = "Journal has aborted";
918 else
919 errstr = "Readonly filesystem";
920 break;
921 default:
922 /* If the caller passed in an extra buffer for unknown
923 * errors, textualise them now. Else we just return
924 * NULL. */
925 if (nbuf) {
926 /* Check for truncated error codes... */
927 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
928 errstr = nbuf;
929 }
930 break;
931 }
932
933 return errstr;
934 }
935
936 /* __ext4_std_error decodes expected errors from journaling functions
937 * automatically and invokes the appropriate error response. */
938
939 void __ext4_std_error(struct super_block *sb, const char *function,
940 unsigned int line, int errno)
941 {
942 char nbuf[16];
943 const char *errstr;
944
945 if (unlikely(ext4_emergency_state(sb)))
946 return;
947
948 /* Special case: if the error is EROFS, and we're not already
949 * inside a transaction, then there's really no point in logging
950 * an error. */
951 if (errno == -EROFS && journal_current_handle() == NULL && sb_rdonly(sb))
952 return;
953
954 if (ext4_error_ratelimit(sb)) {
955 errstr = ext4_decode_error(sb, errno, nbuf);
956 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
957 sb->s_id, function, line, errstr);
958 }
959 fsnotify_sb_error(sb, NULL, errno ? errno : EFSCORRUPTED);
960
961 ext4_handle_error(sb, false, -errno, 0, 0, function, line);
962 }
963
964 void __ext4_msg(struct super_block *sb,
965 const char *prefix, const char *fmt, ...)
966 {
967 struct va_format vaf;
968 va_list args;
969
970 if (sb) {
971 atomic_inc(&EXT4_SB(sb)->s_msg_count);
972 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state),
973 "EXT4-fs"))
974 return;
975 }
976
977 va_start(args, fmt);
978 vaf.fmt = fmt;
979 vaf.va = &args;
980 if (sb)
981 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
982 else
983 printk("%sEXT4-fs: %pV\n", prefix, &vaf);
984 va_end(args);
985 }
986
987 static int ext4_warning_ratelimit(struct super_block *sb)
988 {
989 atomic_inc(&EXT4_SB(sb)->s_warning_count);
990 return ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state),
991 "EXT4-fs warning");
992 }
993
994 void __ext4_warning(struct super_block *sb, const char *function,
995 unsigned int line, const char *fmt, ...)
996 {
997 struct va_format vaf;
998 va_list args;
999
1000 if (!ext4_warning_ratelimit(sb))
1001 return;
1002
1003 va_start(args, fmt);
1004 vaf.fmt = fmt;
1005 vaf.va = &args;
1006 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
1007 sb->s_id, function, line, &vaf);
1008 va_end(args);
1009 }
1010
1011 void __ext4_warning_inode(const struct inode *inode, const char *function,
1012 unsigned int line, const char *fmt, ...)
1013 {
1014 struct va_format vaf;
1015 va_list args;
1016
1017 if (!ext4_warning_ratelimit(inode->i_sb))
1018 return;
1019
1020 va_start(args, fmt);
1021 vaf.fmt = fmt;
1022 vaf.va = &args;
1023 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
1024 "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
1025 function, line, inode->i_ino, current->comm, &vaf);
1026 va_end(args);
1027 }
1028
1029 void __ext4_grp_locked_error(const char *function, unsigned int line,
1030 struct super_block *sb, ext4_group_t grp,
1031 unsigned long ino, ext4_fsblk_t block,
1032 const char *fmt, ...)
1033 __releases(bitlock)
1034 __acquires(bitlock)
1035 {
1036 struct va_format vaf;
1037 va_list args;
1038
1039 if (unlikely(ext4_emergency_state(sb)))
1040 return;
1041
1042 trace_ext4_error(sb, function, line);
1043 if (ext4_error_ratelimit(sb)) {
1044 va_start(args, fmt);
1045 vaf.fmt = fmt;
1046 vaf.va = &args;
1047 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
1048 sb->s_id, function, line, grp);
1049 if (ino)
1050 printk(KERN_CONT "inode %lu: ", ino);
1051 if (block)
1052 printk(KERN_CONT "block %llu:",
1053 (unsigned long long) block);
1054 printk(KERN_CONT "%pV\n", &vaf);
1055 va_end(args);
1056 }
1057
1058 if (test_opt(sb, ERRORS_CONT)) {
1059 if (test_opt(sb, WARN_ON_ERROR))
1060 WARN_ON_ONCE(1);
1061 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
1062 if (!bdev_read_only(sb->s_bdev)) {
1063 save_error_info(sb, EFSCORRUPTED, ino, block, function,
1064 line);
1065 schedule_work(&EXT4_SB(sb)->s_sb_upd_work);
1066 }
1067 return;
1068 }
1069 ext4_unlock_group(sb, grp);
1070 ext4_handle_error(sb, false, EFSCORRUPTED, ino, block, function, line);
1071 /*
1072 * We only get here in the ERRORS_RO case; relocking the group
1073 * may be dangerous, but nothing bad will happen since the
1074 * filesystem will have already been marked read/only and the
1075 * journal has been aborted. We return 1 as a hint to callers
1076 * who might what to use the return value from
1077 * ext4_grp_locked_error() to distinguish between the
1078 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
1079 * aggressively from the ext4 function in question, with a
1080 * more appropriate error code.
1081 */
1082 ext4_lock_group(sb, grp);
1083 return;
1084 }
1085
1086 void ext4_mark_group_bitmap_corrupted(struct super_block *sb,
1087 ext4_group_t group,
1088 unsigned int flags)
1089 {
1090 struct ext4_sb_info *sbi = EXT4_SB(sb);
1091 struct ext4_group_info *grp = ext4_get_group_info(sb, group);
1092 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, group, NULL);
1093 int ret;
1094
1095 if (!grp || !gdp)
1096 return;
1097 if (flags & EXT4_GROUP_INFO_BBITMAP_CORRUPT) {
1098 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT,
1099 &grp->bb_state);
1100 if (!ret)
1101 percpu_counter_sub(&sbi->s_freeclusters_counter,
1102 grp->bb_free);
1103 }
1104
1105 if (flags & EXT4_GROUP_INFO_IBITMAP_CORRUPT) {
1106 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT,
1107 &grp->bb_state);
1108 if (!ret && gdp) {
1109 int count;
1110
1111 count = ext4_free_inodes_count(sb, gdp);
1112 percpu_counter_sub(&sbi->s_freeinodes_counter,
1113 count);
1114 }
1115 }
1116 }
1117
1118 void ext4_update_dynamic_rev(struct super_block *sb)
1119 {
1120 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
1121
1122 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
1123 return;
1124
1125 ext4_warning(sb,
1126 "updating to rev %d because of new feature flag, "
1127 "running e2fsck is recommended",
1128 EXT4_DYNAMIC_REV);
1129
1130 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
1131 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
1132 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
1133 /* leave es->s_feature_*compat flags alone */
1134 /* es->s_uuid will be set by e2fsck if empty */
1135
1136 /*
1137 * The rest of the superblock fields should be zero, and if not it
1138 * means they are likely already in use, so leave them alone. We
1139 * can leave it up to e2fsck to clean up any inconsistencies there.
1140 */
1141 }
1142
1143 static inline struct inode *orphan_list_entry(struct list_head *l)
1144 {
1145 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
1146 }
1147
1148 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
1149 {
1150 struct list_head *l;
1151
1152 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
1153 le32_to_cpu(sbi->s_es->s_last_orphan));
1154
1155 printk(KERN_ERR "sb_info orphan list:\n");
1156 list_for_each(l, &sbi->s_orphan) {
1157 struct inode *inode = orphan_list_entry(l);
1158 printk(KERN_ERR " "
1159 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
1160 inode->i_sb->s_id, inode->i_ino, inode,
1161 inode->i_mode, inode->i_nlink,
1162 NEXT_ORPHAN(inode));
1163 }
1164 }
1165
1166 #ifdef CONFIG_QUOTA
1167 static int ext4_quota_off(struct super_block *sb, int type);
1168
1169 static inline void ext4_quotas_off(struct super_block *sb, int type)
1170 {
1171 BUG_ON(type > EXT4_MAXQUOTAS);
1172
1173 /* Use our quota_off function to clear inode flags etc. */
1174 for (type--; type >= 0; type--)
1175 ext4_quota_off(sb, type);
1176 }
1177
1178 /*
1179 * This is a helper function which is used in the mount/remount
1180 * codepaths (which holds s_umount) to fetch the quota file name.
1181 */
1182 static inline char *get_qf_name(struct super_block *sb,
1183 struct ext4_sb_info *sbi,
1184 int type)
1185 {
1186 return rcu_dereference_protected(sbi->s_qf_names[type],
1187 lockdep_is_held(&sb->s_umount));
1188 }
1189 #else
1190 static inline void ext4_quotas_off(struct super_block *sb, int type)
1191 {
1192 }
1193 #endif
1194
1195 static int ext4_percpu_param_init(struct ext4_sb_info *sbi)
1196 {
1197 ext4_fsblk_t block;
1198 int err;
1199
1200 block = ext4_count_free_clusters(sbi->s_sb);
1201 ext4_free_blocks_count_set(sbi->s_es, EXT4_C2B(sbi, block));
1202 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
1203 GFP_KERNEL);
1204 if (!err) {
1205 unsigned long freei = ext4_count_free_inodes(sbi->s_sb);
1206 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
1207 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
1208 GFP_KERNEL);
1209 }
1210 if (!err)
1211 err = percpu_counter_init(&sbi->s_dirs_counter,
1212 ext4_count_dirs(sbi->s_sb), GFP_KERNEL);
1213 if (!err)
1214 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
1215 GFP_KERNEL);
1216 if (!err)
1217 err = percpu_counter_init(&sbi->s_sra_exceeded_retry_limit, 0,
1218 GFP_KERNEL);
1219 if (!err)
1220 err = percpu_init_rwsem(&sbi->s_writepages_rwsem);
1221
1222 if (err)
1223 ext4_msg(sbi->s_sb, KERN_ERR, "insufficient memory");
1224
1225 return err;
1226 }
1227
1228 static void ext4_percpu_param_destroy(struct ext4_sb_info *sbi)
1229 {
1230 percpu_counter_destroy(&sbi->s_freeclusters_counter);
1231 percpu_counter_destroy(&sbi->s_freeinodes_counter);
1232 percpu_counter_destroy(&sbi->s_dirs_counter);
1233 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
1234 percpu_counter_destroy(&sbi->s_sra_exceeded_retry_limit);
1235 percpu_free_rwsem(&sbi->s_writepages_rwsem);
1236 }
1237
1238 static void ext4_group_desc_free(struct ext4_sb_info *sbi)
1239 {
1240 struct buffer_head **group_desc;
1241 int i;
1242
1243 rcu_read_lock();
1244 group_desc = rcu_dereference(sbi->s_group_desc);
1245 for (i = 0; i < sbi->s_gdb_count; i++)
1246 brelse(group_desc[i]);
1247 kvfree(group_desc);
1248 rcu_read_unlock();
1249 }
1250
1251 static void ext4_flex_groups_free(struct ext4_sb_info *sbi)
1252 {
1253 struct flex_groups **flex_groups;
1254 int i;
1255
1256 rcu_read_lock();
1257 flex_groups = rcu_dereference(sbi->s_flex_groups);
1258 if (flex_groups) {
1259 for (i = 0; i < sbi->s_flex_groups_allocated; i++)
1260 kvfree(flex_groups[i]);
1261 kvfree(flex_groups);
1262 }
1263 rcu_read_unlock();
1264 }
1265
1266 static void ext4_put_super(struct super_block *sb)
1267 {
1268 struct ext4_sb_info *sbi = EXT4_SB(sb);
1269 struct ext4_super_block *es = sbi->s_es;
1270 int aborted = 0;
1271 int err;
1272
1273 /*
1274 * Unregister sysfs before destroying jbd2 journal.
1275 * Since we could still access attr_journal_task attribute via sysfs
1276 * path which could have sbi->s_journal->j_task as NULL
1277 * Unregister sysfs before flush sbi->s_sb_upd_work.
1278 * Since user may read /proc/fs/ext4/xx/mb_groups during umount, If
1279 * read metadata verify failed then will queue error work.
1280 * update_super_work will call start_this_handle may trigger
1281 * BUG_ON.
1282 */
1283 ext4_unregister_sysfs(sb);
1284
1285 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs unmount"))
1286 ext4_msg(sb, KERN_INFO, "unmounting filesystem %pU.",
1287 &sb->s_uuid);
1288
1289 ext4_unregister_li_request(sb);
1290 ext4_quotas_off(sb, EXT4_MAXQUOTAS);
1291
1292 destroy_workqueue(sbi->rsv_conversion_wq);
1293 ext4_release_orphan_info(sb);
1294
1295 if (sbi->s_journal) {
1296 aborted = is_journal_aborted(sbi->s_journal);
1297 err = ext4_journal_destroy(sbi, sbi->s_journal);
1298 if ((err < 0) && !aborted) {
1299 ext4_abort(sb, -err, "Couldn't clean up the journal");
1300 }
1301 } else
1302 flush_work(&sbi->s_sb_upd_work);
1303
1304 ext4_es_unregister_shrinker(sbi);
1305 timer_shutdown_sync(&sbi->s_err_report);
1306 ext4_release_system_zone(sb);
1307 ext4_mb_release(sb);
1308 ext4_ext_release(sb);
1309
1310 if (!ext4_emergency_state(sb) && !sb_rdonly(sb)) {
1311 if (!aborted) {
1312 ext4_clear_feature_journal_needs_recovery(sb);
1313 ext4_clear_feature_orphan_present(sb);
1314 es->s_state = cpu_to_le16(sbi->s_mount_state);
1315 }
1316 ext4_commit_super(sb);
1317 }
1318
1319 ext4_group_desc_free(sbi);
1320 ext4_flex_groups_free(sbi);
1321
1322 WARN_ON_ONCE(!(sbi->s_mount_state & EXT4_ERROR_FS) &&
1323 percpu_counter_sum(&sbi->s_dirtyclusters_counter));
1324 ext4_percpu_param_destroy(sbi);
1325 #ifdef CONFIG_QUOTA
1326 for (int i = 0; i < EXT4_MAXQUOTAS; i++)
1327 kfree(get_qf_name(sb, sbi, i));
1328 #endif
1329
1330 /* Debugging code just in case the in-memory inode orphan list
1331 * isn't empty. The on-disk one can be non-empty if we've
1332 * detected an error and taken the fs readonly, but the
1333 * in-memory list had better be clean by this point. */
1334 if (!list_empty(&sbi->s_orphan))
1335 dump_orphan_list(sb, sbi);
1336 ASSERT(list_empty(&sbi->s_orphan));
1337
1338 sync_blockdev(sb->s_bdev);
1339 invalidate_bdev(sb->s_bdev);
1340 if (sbi->s_journal_bdev_file) {
1341 /*
1342 * Invalidate the journal device's buffers. We don't want them
1343 * floating about in memory - the physical journal device may
1344 * hotswapped, and it breaks the `ro-after' testing code.
1345 */
1346 sync_blockdev(file_bdev(sbi->s_journal_bdev_file));
1347 invalidate_bdev(file_bdev(sbi->s_journal_bdev_file));
1348 }
1349
1350 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
1351 sbi->s_ea_inode_cache = NULL;
1352
1353 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
1354 sbi->s_ea_block_cache = NULL;
1355
1356 ext4_stop_mmpd(sbi);
1357
1358 brelse(sbi->s_sbh);
1359 sb->s_fs_info = NULL;
1360 /*
1361 * Now that we are completely done shutting down the
1362 * superblock, we need to actually destroy the kobject.
1363 */
1364 kobject_put(&sbi->s_kobj);
1365 wait_for_completion(&sbi->s_kobj_unregister);
1366 kfree(sbi->s_blockgroup_lock);
1367 fs_put_dax(sbi->s_daxdev, NULL);
1368 fscrypt_free_dummy_policy(&sbi->s_dummy_enc_policy);
1369 #if IS_ENABLED(CONFIG_UNICODE)
1370 utf8_unload(sb->s_encoding);
1371 #endif
1372 kfree(sbi);
1373 }
1374
1375 static struct kmem_cache *ext4_inode_cachep;
1376
1377 /*
1378 * Called inside transaction, so use GFP_NOFS
1379 */
1380 static struct inode *ext4_alloc_inode(struct super_block *sb)
1381 {
1382 struct ext4_inode_info *ei;
1383
1384 ei = alloc_inode_sb(sb, ext4_inode_cachep, GFP_NOFS);
1385 if (!ei)
1386 return NULL;
1387
1388 inode_set_iversion(&ei->vfs_inode, 1);
1389 ei->i_flags = 0;
1390 spin_lock_init(&ei->i_raw_lock);
1391 ei->i_prealloc_node = RB_ROOT;
1392 atomic_set(&ei->i_prealloc_active, 0);
1393 rwlock_init(&ei->i_prealloc_lock);
1394 ext4_es_init_tree(&ei->i_es_tree);
1395 rwlock_init(&ei->i_es_lock);
1396 INIT_LIST_HEAD(&ei->i_es_list);
1397 ei->i_es_all_nr = 0;
1398 ei->i_es_shk_nr = 0;
1399 ei->i_es_shrink_lblk = 0;
1400 ei->i_reserved_data_blocks = 0;
1401 spin_lock_init(&(ei->i_block_reservation_lock));
1402 ext4_init_pending_tree(&ei->i_pending_tree);
1403 #ifdef CONFIG_QUOTA
1404 ei->i_reserved_quota = 0;
1405 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
1406 #endif
1407 ei->jinode = NULL;
1408 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
1409 spin_lock_init(&ei->i_completed_io_lock);
1410 ei->i_sync_tid = 0;
1411 ei->i_datasync_tid = 0;
1412 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
1413 ext4_fc_init_inode(&ei->vfs_inode);
1414 spin_lock_init(&ei->i_fc_lock);
1415 return &ei->vfs_inode;
1416 }
1417
1418 static int ext4_drop_inode(struct inode *inode)
1419 {
1420 int drop = generic_drop_inode(inode);
1421
1422 if (!drop)
1423 drop = fscrypt_drop_inode(inode);
1424
1425 trace_ext4_drop_inode(inode, drop);
1426 return drop;
1427 }
1428
1429 static void ext4_free_in_core_inode(struct inode *inode)
1430 {
1431 fscrypt_free_inode(inode);
1432 if (!list_empty(&(EXT4_I(inode)->i_fc_list))) {
1433 pr_warn("%s: inode %ld still in fc list",
1434 __func__, inode->i_ino);
1435 }
1436 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
1437 }
1438
1439 static void ext4_destroy_inode(struct inode *inode)
1440 {
1441 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
1442 ext4_msg(inode->i_sb, KERN_ERR,
1443 "Inode %lu (%p): orphan list check failed!",
1444 inode->i_ino, EXT4_I(inode));
1445 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
1446 EXT4_I(inode), sizeof(struct ext4_inode_info),
1447 true);
1448 dump_stack();
1449 }
1450
1451 if (!(EXT4_SB(inode->i_sb)->s_mount_state & EXT4_ERROR_FS) &&
1452 WARN_ON_ONCE(EXT4_I(inode)->i_reserved_data_blocks))
1453 ext4_msg(inode->i_sb, KERN_ERR,
1454 "Inode %lu (%p): i_reserved_data_blocks (%u) not cleared!",
1455 inode->i_ino, EXT4_I(inode),
1456 EXT4_I(inode)->i_reserved_data_blocks);
1457 }
1458
1459 static void ext4_shutdown(struct super_block *sb)
1460 {
1461 ext4_force_shutdown(sb, EXT4_GOING_FLAGS_NOLOGFLUSH);
1462 }
1463
1464 static void init_once(void *foo)
1465 {
1466 struct ext4_inode_info *ei = foo;
1467
1468 INIT_LIST_HEAD(&ei->i_orphan);
1469 init_rwsem(&ei->xattr_sem);
1470 init_rwsem(&ei->i_data_sem);
1471 inode_init_once(&ei->vfs_inode);
1472 ext4_fc_init_inode(&ei->vfs_inode);
1473 }
1474
1475 static int __init init_inodecache(void)
1476 {
1477 ext4_inode_cachep = kmem_cache_create_usercopy("ext4_inode_cache",
1478 sizeof(struct ext4_inode_info), 0,
1479 SLAB_RECLAIM_ACCOUNT | SLAB_ACCOUNT,
1480 offsetof(struct ext4_inode_info, i_data),
1481 sizeof_field(struct ext4_inode_info, i_data),
1482 init_once);
1483 if (ext4_inode_cachep == NULL)
1484 return -ENOMEM;
1485 return 0;
1486 }
1487
1488 static void destroy_inodecache(void)
1489 {
1490 /*
1491 * Make sure all delayed rcu free inodes are flushed before we
1492 * destroy cache.
1493 */
1494 rcu_barrier();
1495 kmem_cache_destroy(ext4_inode_cachep);
1496 }
1497
1498 void ext4_clear_inode(struct inode *inode)
1499 {
1500 ext4_fc_del(inode);
1501 invalidate_inode_buffers(inode);
1502 clear_inode(inode);
1503 ext4_discard_preallocations(inode);
1504 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
1505 dquot_drop(inode);
1506 if (EXT4_I(inode)->jinode) {
1507 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1508 EXT4_I(inode)->jinode);
1509 jbd2_free_inode(EXT4_I(inode)->jinode);
1510 EXT4_I(inode)->jinode = NULL;
1511 }
1512 fscrypt_put_encryption_info(inode);
1513 fsverity_cleanup_inode(inode);
1514 }
1515
1516 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1517 u64 ino, u32 generation)
1518 {
1519 struct inode *inode;
1520
1521 /*
1522 * Currently we don't know the generation for parent directory, so
1523 * a generation of 0 means "accept any"
1524 */
1525 inode = ext4_iget(sb, ino, EXT4_IGET_HANDLE);
1526 if (IS_ERR(inode))
1527 return ERR_CAST(inode);
1528 if (generation && inode->i_generation != generation) {
1529 iput(inode);
1530 return ERR_PTR(-ESTALE);
1531 }
1532
1533 return inode;
1534 }
1535
1536 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1537 int fh_len, int fh_type)
1538 {
1539 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1540 ext4_nfs_get_inode);
1541 }
1542
1543 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1544 int fh_len, int fh_type)
1545 {
1546 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1547 ext4_nfs_get_inode);
1548 }
1549
1550 static int ext4_nfs_commit_metadata(struct inode *inode)
1551 {
1552 struct writeback_control wbc = {
1553 .sync_mode = WB_SYNC_ALL
1554 };
1555
1556 trace_ext4_nfs_commit_metadata(inode);
1557 return ext4_write_inode(inode, &wbc);
1558 }
1559
1560 #ifdef CONFIG_QUOTA
1561 static const char * const quotatypes[] = INITQFNAMES;
1562 #define QTYPE2NAME(t) (quotatypes[t])
1563
1564 static int ext4_write_dquot(struct dquot *dquot);
1565 static int ext4_acquire_dquot(struct dquot *dquot);
1566 static int ext4_release_dquot(struct dquot *dquot);
1567 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1568 static int ext4_write_info(struct super_block *sb, int type);
1569 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1570 const struct path *path);
1571 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1572 size_t len, loff_t off);
1573 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1574 const char *data, size_t len, loff_t off);
1575 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1576 unsigned int flags);
1577
1578 static struct dquot __rcu **ext4_get_dquots(struct inode *inode)
1579 {
1580 return EXT4_I(inode)->i_dquot;
1581 }
1582
1583 static const struct dquot_operations ext4_quota_operations = {
1584 .get_reserved_space = ext4_get_reserved_space,
1585 .write_dquot = ext4_write_dquot,
1586 .acquire_dquot = ext4_acquire_dquot,
1587 .release_dquot = ext4_release_dquot,
1588 .mark_dirty = ext4_mark_dquot_dirty,
1589 .write_info = ext4_write_info,
1590 .alloc_dquot = dquot_alloc,
1591 .destroy_dquot = dquot_destroy,
1592 .get_projid = ext4_get_projid,
1593 .get_inode_usage = ext4_get_inode_usage,
1594 .get_next_id = dquot_get_next_id,
1595 };
1596
1597 static const struct quotactl_ops ext4_qctl_operations = {
1598 .quota_on = ext4_quota_on,
1599 .quota_off = ext4_quota_off,
1600 .quota_sync = dquot_quota_sync,
1601 .get_state = dquot_get_state,
1602 .set_info = dquot_set_dqinfo,
1603 .get_dqblk = dquot_get_dqblk,
1604 .set_dqblk = dquot_set_dqblk,
1605 .get_nextdqblk = dquot_get_next_dqblk,
1606 };
1607 #endif
1608
1609 static const struct super_operations ext4_sops = {
1610 .alloc_inode = ext4_alloc_inode,
1611 .free_inode = ext4_free_in_core_inode,
1612 .destroy_inode = ext4_destroy_inode,
1613 .write_inode = ext4_write_inode,
1614 .dirty_inode = ext4_dirty_inode,
1615 .drop_inode = ext4_drop_inode,
1616 .evict_inode = ext4_evict_inode,
1617 .put_super = ext4_put_super,
1618 .sync_fs = ext4_sync_fs,
1619 .freeze_fs = ext4_freeze,
1620 .unfreeze_fs = ext4_unfreeze,
1621 .statfs = ext4_statfs,
1622 .show_options = ext4_show_options,
1623 .shutdown = ext4_shutdown,
1624 #ifdef CONFIG_QUOTA
1625 .quota_read = ext4_quota_read,
1626 .quota_write = ext4_quota_write,
1627 .get_dquots = ext4_get_dquots,
1628 #endif
1629 };
1630
1631 static const struct export_operations ext4_export_ops = {
1632 .encode_fh = generic_encode_ino32_fh,
1633 .fh_to_dentry = ext4_fh_to_dentry,
1634 .fh_to_parent = ext4_fh_to_parent,
1635 .get_parent = ext4_get_parent,
1636 .commit_metadata = ext4_nfs_commit_metadata,
1637 };
1638
1639 enum {
1640 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1641 Opt_resgid, Opt_resuid, Opt_sb,
1642 Opt_nouid32, Opt_debug, Opt_removed,
1643 Opt_user_xattr, Opt_acl,
1644 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1645 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1646 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1647 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1648 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
1649 Opt_inlinecrypt,
1650 Opt_usrjquota, Opt_grpjquota, Opt_quota,
1651 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1652 Opt_usrquota, Opt_grpquota, Opt_prjquota,
1653 Opt_dax, Opt_dax_always, Opt_dax_inode, Opt_dax_never,
1654 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_warn_on_error,
1655 Opt_nowarn_on_error, Opt_mblk_io_submit, Opt_debug_want_extra_isize,
1656 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1657 Opt_inode_readahead_blks, Opt_journal_ioprio,
1658 Opt_dioread_nolock, Opt_dioread_lock,
1659 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1660 Opt_max_dir_size_kb, Opt_nojournal_checksum, Opt_nombcache,
1661 Opt_no_prefetch_block_bitmaps, Opt_mb_optimize_scan,
1662 Opt_errors, Opt_data, Opt_data_err, Opt_jqfmt, Opt_dax_type,
1663 #ifdef CONFIG_EXT4_DEBUG
1664 Opt_fc_debug_max_replay, Opt_fc_debug_force
1665 #endif
1666 };
1667
1668 static const struct constant_table ext4_param_errors[] = {
1669 {"continue", EXT4_MOUNT_ERRORS_CONT},
1670 {"panic", EXT4_MOUNT_ERRORS_PANIC},
1671 {"remount-ro", EXT4_MOUNT_ERRORS_RO},
1672 {}
1673 };
1674
1675 static const struct constant_table ext4_param_data[] = {
1676 {"journal", EXT4_MOUNT_JOURNAL_DATA},
1677 {"ordered", EXT4_MOUNT_ORDERED_DATA},
1678 {"writeback", EXT4_MOUNT_WRITEBACK_DATA},
1679 {}
1680 };
1681
1682 static const struct constant_table ext4_param_data_err[] = {
1683 {"abort", Opt_data_err_abort},
1684 {"ignore", Opt_data_err_ignore},
1685 {}
1686 };
1687
1688 static const struct constant_table ext4_param_jqfmt[] = {
1689 {"vfsold", QFMT_VFS_OLD},
1690 {"vfsv0", QFMT_VFS_V0},
1691 {"vfsv1", QFMT_VFS_V1},
1692 {}
1693 };
1694
1695 static const struct constant_table ext4_param_dax[] = {
1696 {"always", Opt_dax_always},
1697 {"inode", Opt_dax_inode},
1698 {"never", Opt_dax_never},
1699 {}
1700 };
1701
1702 /*
1703 * Mount option specification
1704 * We don't use fsparam_flag_no because of the way we set the
1705 * options and the way we show them in _ext4_show_options(). To
1706 * keep the changes to a minimum, let's keep the negative options
1707 * separate for now.
1708 */
1709 static const struct fs_parameter_spec ext4_param_specs[] = {
1710 fsparam_flag ("bsddf", Opt_bsd_df),
1711 fsparam_flag ("minixdf", Opt_minix_df),
1712 fsparam_flag ("grpid", Opt_grpid),
1713 fsparam_flag ("bsdgroups", Opt_grpid),
1714 fsparam_flag ("nogrpid", Opt_nogrpid),
1715 fsparam_flag ("sysvgroups", Opt_nogrpid),
1716 fsparam_gid ("resgid", Opt_resgid),
1717 fsparam_uid ("resuid", Opt_resuid),
1718 fsparam_u32 ("sb", Opt_sb),
1719 fsparam_enum ("errors", Opt_errors, ext4_param_errors),
1720 fsparam_flag ("nouid32", Opt_nouid32),
1721 fsparam_flag ("debug", Opt_debug),
1722 fsparam_flag ("oldalloc", Opt_removed),
1723 fsparam_flag ("orlov", Opt_removed),
1724 fsparam_flag ("user_xattr", Opt_user_xattr),
1725 fsparam_flag ("acl", Opt_acl),
1726 fsparam_flag ("norecovery", Opt_noload),
1727 fsparam_flag ("noload", Opt_noload),
1728 fsparam_flag ("bh", Opt_removed),
1729 fsparam_flag ("nobh", Opt_removed),
1730 fsparam_u32 ("commit", Opt_commit),
1731 fsparam_u32 ("min_batch_time", Opt_min_batch_time),
1732 fsparam_u32 ("max_batch_time", Opt_max_batch_time),
1733 fsparam_u32 ("journal_dev", Opt_journal_dev),
1734 fsparam_bdev ("journal_path", Opt_journal_path),
1735 fsparam_flag ("journal_checksum", Opt_journal_checksum),
1736 fsparam_flag ("nojournal_checksum", Opt_nojournal_checksum),
1737 fsparam_flag ("journal_async_commit",Opt_journal_async_commit),
1738 fsparam_flag ("abort", Opt_abort),
1739 fsparam_enum ("data", Opt_data, ext4_param_data),
1740 fsparam_enum ("data_err", Opt_data_err,
1741 ext4_param_data_err),
1742 fsparam_string_empty
1743 ("usrjquota", Opt_usrjquota),
1744 fsparam_string_empty
1745 ("grpjquota", Opt_grpjquota),
1746 fsparam_enum ("jqfmt", Opt_jqfmt, ext4_param_jqfmt),
1747 fsparam_flag ("grpquota", Opt_grpquota),
1748 fsparam_flag ("quota", Opt_quota),
1749 fsparam_flag ("noquota", Opt_noquota),
1750 fsparam_flag ("usrquota", Opt_usrquota),
1751 fsparam_flag ("prjquota", Opt_prjquota),
1752 fsparam_flag ("barrier", Opt_barrier),
1753 fsparam_u32 ("barrier", Opt_barrier),
1754 fsparam_flag ("nobarrier", Opt_nobarrier),
1755 fsparam_flag ("i_version", Opt_removed),
1756 fsparam_flag ("dax", Opt_dax),
1757 fsparam_enum ("dax", Opt_dax_type, ext4_param_dax),
1758 fsparam_u32 ("stripe", Opt_stripe),
1759 fsparam_flag ("delalloc", Opt_delalloc),
1760 fsparam_flag ("nodelalloc", Opt_nodelalloc),
1761 fsparam_flag ("warn_on_error", Opt_warn_on_error),
1762 fsparam_flag ("nowarn_on_error", Opt_nowarn_on_error),
1763 fsparam_u32 ("debug_want_extra_isize",
1764 Opt_debug_want_extra_isize),
1765 fsparam_flag ("mblk_io_submit", Opt_removed),
1766 fsparam_flag ("nomblk_io_submit", Opt_removed),
1767 fsparam_flag ("block_validity", Opt_block_validity),
1768 fsparam_flag ("noblock_validity", Opt_noblock_validity),
1769 fsparam_u32 ("inode_readahead_blks",
1770 Opt_inode_readahead_blks),
1771 fsparam_u32 ("journal_ioprio", Opt_journal_ioprio),
1772 fsparam_u32 ("auto_da_alloc", Opt_auto_da_alloc),
1773 fsparam_flag ("auto_da_alloc", Opt_auto_da_alloc),
1774 fsparam_flag ("noauto_da_alloc", Opt_noauto_da_alloc),
1775 fsparam_flag ("dioread_nolock", Opt_dioread_nolock),
1776 fsparam_flag ("nodioread_nolock", Opt_dioread_lock),
1777 fsparam_flag ("dioread_lock", Opt_dioread_lock),
1778 fsparam_flag ("discard", Opt_discard),
1779 fsparam_flag ("nodiscard", Opt_nodiscard),
1780 fsparam_u32 ("init_itable", Opt_init_itable),
1781 fsparam_flag ("init_itable", Opt_init_itable),
1782 fsparam_flag ("noinit_itable", Opt_noinit_itable),
1783 #ifdef CONFIG_EXT4_DEBUG
1784 fsparam_flag ("fc_debug_force", Opt_fc_debug_force),
1785 fsparam_u32 ("fc_debug_max_replay", Opt_fc_debug_max_replay),
1786 #endif
1787 fsparam_u32 ("max_dir_size_kb", Opt_max_dir_size_kb),
1788 fsparam_flag ("test_dummy_encryption",
1789 Opt_test_dummy_encryption),
1790 fsparam_string ("test_dummy_encryption",
1791 Opt_test_dummy_encryption),
1792 fsparam_flag ("inlinecrypt", Opt_inlinecrypt),
1793 fsparam_flag ("nombcache", Opt_nombcache),
1794 fsparam_flag ("no_mbcache", Opt_nombcache), /* for backward compatibility */
1795 fsparam_flag ("prefetch_block_bitmaps",
1796 Opt_removed),
1797 fsparam_flag ("no_prefetch_block_bitmaps",
1798 Opt_no_prefetch_block_bitmaps),
1799 fsparam_s32 ("mb_optimize_scan", Opt_mb_optimize_scan),
1800 fsparam_string ("check", Opt_removed), /* mount option from ext2/3 */
1801 fsparam_flag ("nocheck", Opt_removed), /* mount option from ext2/3 */
1802 fsparam_flag ("reservation", Opt_removed), /* mount option from ext2/3 */
1803 fsparam_flag ("noreservation", Opt_removed), /* mount option from ext2/3 */
1804 fsparam_u32 ("journal", Opt_removed), /* mount option from ext2/3 */
1805 {}
1806 };
1807
1808
1809 #define MOPT_SET 0x0001
1810 #define MOPT_CLEAR 0x0002
1811 #define MOPT_NOSUPPORT 0x0004
1812 #define MOPT_EXPLICIT 0x0008
1813 #ifdef CONFIG_QUOTA
1814 #define MOPT_Q 0
1815 #define MOPT_QFMT 0x0010
1816 #else
1817 #define MOPT_Q MOPT_NOSUPPORT
1818 #define MOPT_QFMT MOPT_NOSUPPORT
1819 #endif
1820 #define MOPT_NO_EXT2 0x0020
1821 #define MOPT_NO_EXT3 0x0040
1822 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1823 #define MOPT_SKIP 0x0080
1824 #define MOPT_2 0x0100
1825
1826 static const struct mount_opts {
1827 int token;
1828 int mount_opt;
1829 int flags;
1830 } ext4_mount_opts[] = {
1831 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1832 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1833 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1834 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1835 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1836 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1837 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1838 MOPT_EXT4_ONLY | MOPT_SET},
1839 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1840 MOPT_EXT4_ONLY | MOPT_CLEAR},
1841 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1842 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1843 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1844 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1845 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1846 MOPT_EXT4_ONLY | MOPT_CLEAR},
1847 {Opt_warn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_SET},
1848 {Opt_nowarn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_CLEAR},
1849 {Opt_commit, 0, MOPT_NO_EXT2},
1850 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1851 MOPT_EXT4_ONLY | MOPT_CLEAR},
1852 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1853 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1854 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1855 EXT4_MOUNT_JOURNAL_CHECKSUM),
1856 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1857 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1858 {Opt_data_err, EXT4_MOUNT_DATA_ERR_ABORT, MOPT_NO_EXT2},
1859 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1860 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1861 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1862 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1863 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1864 {Opt_dax_type, 0, MOPT_EXT4_ONLY},
1865 {Opt_journal_dev, 0, MOPT_NO_EXT2},
1866 {Opt_journal_path, 0, MOPT_NO_EXT2},
1867 {Opt_journal_ioprio, 0, MOPT_NO_EXT2},
1868 {Opt_data, 0, MOPT_NO_EXT2},
1869 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1870 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1871 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1872 #else
1873 {Opt_acl, 0, MOPT_NOSUPPORT},
1874 #endif
1875 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1876 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1877 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1878 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1879 MOPT_SET | MOPT_Q},
1880 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1881 MOPT_SET | MOPT_Q},
1882 {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA,
1883 MOPT_SET | MOPT_Q},
1884 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1885 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA),
1886 MOPT_CLEAR | MOPT_Q},
1887 {Opt_usrjquota, 0, MOPT_Q},
1888 {Opt_grpjquota, 0, MOPT_Q},
1889 {Opt_jqfmt, 0, MOPT_QFMT},
1890 {Opt_nombcache, EXT4_MOUNT_NO_MBCACHE, MOPT_SET},
1891 {Opt_no_prefetch_block_bitmaps, EXT4_MOUNT_NO_PREFETCH_BLOCK_BITMAPS,
1892 MOPT_SET},
1893 #ifdef CONFIG_EXT4_DEBUG
1894 {Opt_fc_debug_force, EXT4_MOUNT2_JOURNAL_FAST_COMMIT,
1895 MOPT_SET | MOPT_2 | MOPT_EXT4_ONLY},
1896 #endif
1897 {Opt_abort, EXT4_MOUNT2_ABORT, MOPT_SET | MOPT_2},
1898 {Opt_err, 0, 0}
1899 };
1900
1901 #if IS_ENABLED(CONFIG_UNICODE)
1902 static const struct ext4_sb_encodings {
1903 __u16 magic;
1904 char *name;
1905 unsigned int version;
1906 } ext4_sb_encoding_map[] = {
1907 {EXT4_ENC_UTF8_12_1, "utf8", UNICODE_AGE(12, 1, 0)},
1908 };
1909
1910 static const struct ext4_sb_encodings *
1911 ext4_sb_read_encoding(const struct ext4_super_block *es)
1912 {
1913 __u16 magic = le16_to_cpu(es->s_encoding);
1914 int i;
1915
1916 for (i = 0; i < ARRAY_SIZE(ext4_sb_encoding_map); i++)
1917 if (magic == ext4_sb_encoding_map[i].magic)
1918 return &ext4_sb_encoding_map[i];
1919
1920 return NULL;
1921 }
1922 #endif
1923
1924 #define EXT4_SPEC_JQUOTA (1 << 0)
1925 #define EXT4_SPEC_JQFMT (1 << 1)
1926 #define EXT4_SPEC_DATAJ (1 << 2)
1927 #define EXT4_SPEC_SB_BLOCK (1 << 3)
1928 #define EXT4_SPEC_JOURNAL_DEV (1 << 4)
1929 #define EXT4_SPEC_JOURNAL_IOPRIO (1 << 5)
1930 #define EXT4_SPEC_s_want_extra_isize (1 << 7)
1931 #define EXT4_SPEC_s_max_batch_time (1 << 8)
1932 #define EXT4_SPEC_s_min_batch_time (1 << 9)
1933 #define EXT4_SPEC_s_inode_readahead_blks (1 << 10)
1934 #define EXT4_SPEC_s_li_wait_mult (1 << 11)
1935 #define EXT4_SPEC_s_max_dir_size_kb (1 << 12)
1936 #define EXT4_SPEC_s_stripe (1 << 13)
1937 #define EXT4_SPEC_s_resuid (1 << 14)
1938 #define EXT4_SPEC_s_resgid (1 << 15)
1939 #define EXT4_SPEC_s_commit_interval (1 << 16)
1940 #define EXT4_SPEC_s_fc_debug_max_replay (1 << 17)
1941 #define EXT4_SPEC_s_sb_block (1 << 18)
1942 #define EXT4_SPEC_mb_optimize_scan (1 << 19)
1943
1944 struct ext4_fs_context {
1945 char *s_qf_names[EXT4_MAXQUOTAS];
1946 struct fscrypt_dummy_policy dummy_enc_policy;
1947 int s_jquota_fmt; /* Format of quota to use */
1948 #ifdef CONFIG_EXT4_DEBUG
1949 int s_fc_debug_max_replay;
1950 #endif
1951 unsigned short qname_spec;
1952 unsigned long vals_s_flags; /* Bits to set in s_flags */
1953 unsigned long mask_s_flags; /* Bits changed in s_flags */
1954 unsigned long journal_devnum;
1955 unsigned long s_commit_interval;
1956 unsigned long s_stripe;
1957 unsigned int s_inode_readahead_blks;
1958 unsigned int s_want_extra_isize;
1959 unsigned int s_li_wait_mult;
1960 unsigned int s_max_dir_size_kb;
1961 unsigned int journal_ioprio;
1962 unsigned int vals_s_mount_opt;
1963 unsigned int mask_s_mount_opt;
1964 unsigned int vals_s_mount_opt2;
1965 unsigned int mask_s_mount_opt2;
1966 unsigned int opt_flags; /* MOPT flags */
1967 unsigned int spec;
1968 u32 s_max_batch_time;
1969 u32 s_min_batch_time;
1970 kuid_t s_resuid;
1971 kgid_t s_resgid;
1972 ext4_fsblk_t s_sb_block;
1973 };
1974
1975 static void ext4_fc_free(struct fs_context *fc)
1976 {
1977 struct ext4_fs_context *ctx = fc->fs_private;
1978 int i;
1979
1980 if (!ctx)
1981 return;
1982
1983 for (i = 0; i < EXT4_MAXQUOTAS; i++)
1984 kfree(ctx->s_qf_names[i]);
1985
1986 fscrypt_free_dummy_policy(&ctx->dummy_enc_policy);
1987 kfree(ctx);
1988 }
1989
1990 int ext4_init_fs_context(struct fs_context *fc)
1991 {
1992 struct ext4_fs_context *ctx;
1993
1994 ctx = kzalloc(sizeof(struct ext4_fs_context), GFP_KERNEL);
1995 if (!ctx)
1996 return -ENOMEM;
1997
1998 fc->fs_private = ctx;
1999 fc->ops = &ext4_context_ops;
2000
2001 return 0;
2002 }
2003
2004 #ifdef CONFIG_QUOTA
2005 /*
2006 * Note the name of the specified quota file.
2007 */
2008 static int note_qf_name(struct fs_context *fc, int qtype,
2009 struct fs_parameter *param)
2010 {
2011 struct ext4_fs_context *ctx = fc->fs_private;
2012 char *qname;
2013
2014 if (param->size < 1) {
2015 ext4_msg(NULL, KERN_ERR, "Missing quota name");
2016 return -EINVAL;
2017 }
2018 if (strchr(param->string, '/')) {
2019 ext4_msg(NULL, KERN_ERR,
2020 "quotafile must be on filesystem root");
2021 return -EINVAL;
2022 }
2023 if (ctx->s_qf_names[qtype]) {
2024 if (strcmp(ctx->s_qf_names[qtype], param->string) != 0) {
2025 ext4_msg(NULL, KERN_ERR,
2026 "%s quota file already specified",
2027 QTYPE2NAME(qtype));
2028 return -EINVAL;
2029 }
2030 return 0;
2031 }
2032
2033 qname = kmemdup_nul(param->string, param->size, GFP_KERNEL);
2034 if (!qname) {
2035 ext4_msg(NULL, KERN_ERR,
2036 "Not enough memory for storing quotafile name");
2037 return -ENOMEM;
2038 }
2039 ctx->s_qf_names[qtype] = qname;
2040 ctx->qname_spec |= 1 << qtype;
2041 ctx->spec |= EXT4_SPEC_JQUOTA;
2042 return 0;
2043 }
2044
2045 /*
2046 * Clear the name of the specified quota file.
2047 */
2048 static int unnote_qf_name(struct fs_context *fc, int qtype)
2049 {
2050 struct ext4_fs_context *ctx = fc->fs_private;
2051
2052 kfree(ctx->s_qf_names[qtype]);
2053
2054 ctx->s_qf_names[qtype] = NULL;
2055 ctx->qname_spec |= 1 << qtype;
2056 ctx->spec |= EXT4_SPEC_JQUOTA;
2057 return 0;
2058 }
2059 #endif
2060
2061 static int ext4_parse_test_dummy_encryption(const struct fs_parameter *param,
2062 struct ext4_fs_context *ctx)
2063 {
2064 int err;
2065
2066 if (!IS_ENABLED(CONFIG_FS_ENCRYPTION)) {
2067 ext4_msg(NULL, KERN_WARNING,
2068 "test_dummy_encryption option not supported");
2069 return -EINVAL;
2070 }
2071 err = fscrypt_parse_test_dummy_encryption(param,
2072 &ctx->dummy_enc_policy);
2073 if (err == -EINVAL) {
2074 ext4_msg(NULL, KERN_WARNING,
2075 "Value of option \"%s\" is unrecognized", param->key);
2076 } else if (err == -EEXIST) {
2077 ext4_msg(NULL, KERN_WARNING,
2078 "Conflicting test_dummy_encryption options");
2079 return -EINVAL;
2080 }
2081 return err;
2082 }
2083
2084 #define EXT4_SET_CTX(name) \
2085 static inline __maybe_unused \
2086 void ctx_set_##name(struct ext4_fs_context *ctx, unsigned long flag) \
2087 { \
2088 ctx->mask_s_##name |= flag; \
2089 ctx->vals_s_##name |= flag; \
2090 }
2091
2092 #define EXT4_CLEAR_CTX(name) \
2093 static inline __maybe_unused \
2094 void ctx_clear_##name(struct ext4_fs_context *ctx, unsigned long flag) \
2095 { \
2096 ctx->mask_s_##name |= flag; \
2097 ctx->vals_s_##name &= ~flag; \
2098 }
2099
2100 #define EXT4_TEST_CTX(name) \
2101 static inline unsigned long \
2102 ctx_test_##name(struct ext4_fs_context *ctx, unsigned long flag) \
2103 { \
2104 return (ctx->vals_s_##name & flag); \
2105 }
2106
2107 EXT4_SET_CTX(flags); /* set only */
2108 EXT4_SET_CTX(mount_opt);
2109 EXT4_CLEAR_CTX(mount_opt);
2110 EXT4_TEST_CTX(mount_opt);
2111 EXT4_SET_CTX(mount_opt2);
2112 EXT4_CLEAR_CTX(mount_opt2);
2113 EXT4_TEST_CTX(mount_opt2);
2114
2115 static int ext4_parse_param(struct fs_context *fc, struct fs_parameter *param)
2116 {
2117 struct ext4_fs_context *ctx = fc->fs_private;
2118 struct fs_parse_result result;
2119 const struct mount_opts *m;
2120 int is_remount;
2121 int token;
2122
2123 token = fs_parse(fc, ext4_param_specs, param, &result);
2124 if (token < 0)
2125 return token;
2126 is_remount = fc->purpose == FS_CONTEXT_FOR_RECONFIGURE;
2127
2128 for (m = ext4_mount_opts; m->token != Opt_err; m++)
2129 if (token == m->token)
2130 break;
2131
2132 ctx->opt_flags |= m->flags;
2133
2134 if (m->flags & MOPT_EXPLICIT) {
2135 if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
2136 ctx_set_mount_opt2(ctx, EXT4_MOUNT2_EXPLICIT_DELALLOC);
2137 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
2138 ctx_set_mount_opt2(ctx,
2139 EXT4_MOUNT2_EXPLICIT_JOURNAL_CHECKSUM);
2140 } else
2141 return -EINVAL;
2142 }
2143
2144 if (m->flags & MOPT_NOSUPPORT) {
2145 ext4_msg(NULL, KERN_ERR, "%s option not supported",
2146 param->key);
2147 return 0;
2148 }
2149
2150 switch (token) {
2151 #ifdef CONFIG_QUOTA
2152 case Opt_usrjquota:
2153 if (!*param->string)
2154 return unnote_qf_name(fc, USRQUOTA);
2155 else
2156 return note_qf_name(fc, USRQUOTA, param);
2157 case Opt_grpjquota:
2158 if (!*param->string)
2159 return unnote_qf_name(fc, GRPQUOTA);
2160 else
2161 return note_qf_name(fc, GRPQUOTA, param);
2162 #endif
2163 case Opt_sb:
2164 if (fc->purpose == FS_CONTEXT_FOR_RECONFIGURE) {
2165 ext4_msg(NULL, KERN_WARNING,
2166 "Ignoring %s option on remount", param->key);
2167 } else {
2168 ctx->s_sb_block = result.uint_32;
2169 ctx->spec |= EXT4_SPEC_s_sb_block;
2170 }
2171 return 0;
2172 case Opt_removed:
2173 ext4_msg(NULL, KERN_WARNING, "Ignoring removed %s option",
2174 param->key);
2175 return 0;
2176 case Opt_inlinecrypt:
2177 #ifdef CONFIG_FS_ENCRYPTION_INLINE_CRYPT
2178 ctx_set_flags(ctx, SB_INLINECRYPT);
2179 #else
2180 ext4_msg(NULL, KERN_ERR, "inline encryption not supported");
2181 #endif
2182 return 0;
2183 case Opt_errors:
2184 ctx_clear_mount_opt(ctx, EXT4_MOUNT_ERRORS_MASK);
2185 ctx_set_mount_opt(ctx, result.uint_32);
2186 return 0;
2187 #ifdef CONFIG_QUOTA
2188 case Opt_jqfmt:
2189 ctx->s_jquota_fmt = result.uint_32;
2190 ctx->spec |= EXT4_SPEC_JQFMT;
2191 return 0;
2192 #endif
2193 case Opt_data:
2194 ctx_clear_mount_opt(ctx, EXT4_MOUNT_DATA_FLAGS);
2195 ctx_set_mount_opt(ctx, result.uint_32);
2196 ctx->spec |= EXT4_SPEC_DATAJ;
2197 return 0;
2198 case Opt_commit:
2199 if (result.uint_32 == 0)
2200 result.uint_32 = JBD2_DEFAULT_MAX_COMMIT_AGE;
2201 else if (result.uint_32 > INT_MAX / HZ) {
2202 ext4_msg(NULL, KERN_ERR,
2203 "Invalid commit interval %d, "
2204 "must be smaller than %d",
2205 result.uint_32, INT_MAX / HZ);
2206 return -EINVAL;
2207 }
2208 ctx->s_commit_interval = HZ * result.uint_32;
2209 ctx->spec |= EXT4_SPEC_s_commit_interval;
2210 return 0;
2211 case Opt_debug_want_extra_isize:
2212 if ((result.uint_32 & 1) || (result.uint_32 < 4)) {
2213 ext4_msg(NULL, KERN_ERR,
2214 "Invalid want_extra_isize %d", result.uint_32);
2215 return -EINVAL;
2216 }
2217 ctx->s_want_extra_isize = result.uint_32;
2218 ctx->spec |= EXT4_SPEC_s_want_extra_isize;
2219 return 0;
2220 case Opt_max_batch_time:
2221 ctx->s_max_batch_time = result.uint_32;
2222 ctx->spec |= EXT4_SPEC_s_max_batch_time;
2223 return 0;
2224 case Opt_min_batch_time:
2225 ctx->s_min_batch_time = result.uint_32;
2226 ctx->spec |= EXT4_SPEC_s_min_batch_time;
2227 return 0;
2228 case Opt_inode_readahead_blks:
2229 if (result.uint_32 &&
2230 (result.uint_32 > (1 << 30) ||
2231 !is_power_of_2(result.uint_32))) {
2232 ext4_msg(NULL, KERN_ERR,
2233 "EXT4-fs: inode_readahead_blks must be "
2234 "0 or a power of 2 smaller than 2^31");
2235 return -EINVAL;
2236 }
2237 ctx->s_inode_readahead_blks = result.uint_32;
2238 ctx->spec |= EXT4_SPEC_s_inode_readahead_blks;
2239 return 0;
2240 case Opt_init_itable:
2241 ctx_set_mount_opt(ctx, EXT4_MOUNT_INIT_INODE_TABLE);
2242 ctx->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
2243 if (param->type == fs_value_is_string)
2244 ctx->s_li_wait_mult = result.uint_32;
2245 ctx->spec |= EXT4_SPEC_s_li_wait_mult;
2246 return 0;
2247 case Opt_max_dir_size_kb:
2248 ctx->s_max_dir_size_kb = result.uint_32;
2249 ctx->spec |= EXT4_SPEC_s_max_dir_size_kb;
2250 return 0;
2251 #ifdef CONFIG_EXT4_DEBUG
2252 case Opt_fc_debug_max_replay:
2253 ctx->s_fc_debug_max_replay = result.uint_32;
2254 ctx->spec |= EXT4_SPEC_s_fc_debug_max_replay;
2255 return 0;
2256 #endif
2257 case Opt_stripe:
2258 ctx->s_stripe = result.uint_32;
2259 ctx->spec |= EXT4_SPEC_s_stripe;
2260 return 0;
2261 case Opt_resuid:
2262 ctx->s_resuid = result.uid;
2263 ctx->spec |= EXT4_SPEC_s_resuid;
2264 return 0;
2265 case Opt_resgid:
2266 ctx->s_resgid = result.gid;
2267 ctx->spec |= EXT4_SPEC_s_resgid;
2268 return 0;
2269 case Opt_journal_dev:
2270 if (is_remount) {
2271 ext4_msg(NULL, KERN_ERR,
2272 "Cannot specify journal on remount");
2273 return -EINVAL;
2274 }
2275 ctx->journal_devnum = result.uint_32;
2276 ctx->spec |= EXT4_SPEC_JOURNAL_DEV;
2277 return 0;
2278 case Opt_journal_path:
2279 {
2280 struct inode *journal_inode;
2281 struct path path;
2282 int error;
2283
2284 if (is_remount) {
2285 ext4_msg(NULL, KERN_ERR,
2286 "Cannot specify journal on remount");
2287 return -EINVAL;
2288 }
2289
2290 error = fs_lookup_param(fc, param, 1, LOOKUP_FOLLOW, &path);
2291 if (error) {
2292 ext4_msg(NULL, KERN_ERR, "error: could not find "
2293 "journal device path");
2294 return -EINVAL;
2295 }
2296
2297 journal_inode = d_inode(path.dentry);
2298 ctx->journal_devnum = new_encode_dev(journal_inode->i_rdev);
2299 ctx->spec |= EXT4_SPEC_JOURNAL_DEV;
2300 path_put(&path);
2301 return 0;
2302 }
2303 case Opt_journal_ioprio:
2304 if (result.uint_32 > 7) {
2305 ext4_msg(NULL, KERN_ERR, "Invalid journal IO priority"
2306 " (must be 0-7)");
2307 return -EINVAL;
2308 }
2309 ctx->journal_ioprio =
2310 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, result.uint_32);
2311 ctx->spec |= EXT4_SPEC_JOURNAL_IOPRIO;
2312 return 0;
2313 case Opt_test_dummy_encryption:
2314 return ext4_parse_test_dummy_encryption(param, ctx);
2315 case Opt_dax:
2316 case Opt_dax_type:
2317 #ifdef CONFIG_FS_DAX
2318 {
2319 int type = (token == Opt_dax) ?
2320 Opt_dax : result.uint_32;
2321
2322 switch (type) {
2323 case Opt_dax:
2324 case Opt_dax_always:
2325 ctx_set_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS);
2326 ctx_clear_mount_opt2(ctx, EXT4_MOUNT2_DAX_NEVER);
2327 break;
2328 case Opt_dax_never:
2329 ctx_set_mount_opt2(ctx, EXT4_MOUNT2_DAX_NEVER);
2330 ctx_clear_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS);
2331 break;
2332 case Opt_dax_inode:
2333 ctx_clear_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS);
2334 ctx_clear_mount_opt2(ctx, EXT4_MOUNT2_DAX_NEVER);
2335 /* Strictly for printing options */
2336 ctx_set_mount_opt2(ctx, EXT4_MOUNT2_DAX_INODE);
2337 break;
2338 }
2339 return 0;
2340 }
2341 #else
2342 ext4_msg(NULL, KERN_INFO, "dax option not supported");
2343 return -EINVAL;
2344 #endif
2345 case Opt_data_err:
2346 if (result.uint_32 == Opt_data_err_abort)
2347 ctx_set_mount_opt(ctx, m->mount_opt);
2348 else if (result.uint_32 == Opt_data_err_ignore)
2349 ctx_clear_mount_opt(ctx, m->mount_opt);
2350 return 0;
2351 case Opt_mb_optimize_scan:
2352 if (result.int_32 == 1) {
2353 ctx_set_mount_opt2(ctx, EXT4_MOUNT2_MB_OPTIMIZE_SCAN);
2354 ctx->spec |= EXT4_SPEC_mb_optimize_scan;
2355 } else if (result.int_32 == 0) {
2356 ctx_clear_mount_opt2(ctx, EXT4_MOUNT2_MB_OPTIMIZE_SCAN);
2357 ctx->spec |= EXT4_SPEC_mb_optimize_scan;
2358 } else {
2359 ext4_msg(NULL, KERN_WARNING,
2360 "mb_optimize_scan should be set to 0 or 1.");
2361 return -EINVAL;
2362 }
2363 return 0;
2364 }
2365
2366 /*
2367 * At this point we should only be getting options requiring MOPT_SET,
2368 * or MOPT_CLEAR. Anything else is a bug
2369 */
2370 if (m->token == Opt_err) {
2371 ext4_msg(NULL, KERN_WARNING, "buggy handling of option %s",
2372 param->key);
2373 WARN_ON(1);
2374 return -EINVAL;
2375 }
2376
2377 else {
2378 unsigned int set = 0;
2379
2380 if ((param->type == fs_value_is_flag) ||
2381 result.uint_32 > 0)
2382 set = 1;
2383
2384 if (m->flags & MOPT_CLEAR)
2385 set = !set;
2386 else if (unlikely(!(m->flags & MOPT_SET))) {
2387 ext4_msg(NULL, KERN_WARNING,
2388 "buggy handling of option %s",
2389 param->key);
2390 WARN_ON(1);
2391 return -EINVAL;
2392 }
2393 if (m->flags & MOPT_2) {
2394 if (set != 0)
2395 ctx_set_mount_opt2(ctx, m->mount_opt);
2396 else
2397 ctx_clear_mount_opt2(ctx, m->mount_opt);
2398 } else {
2399 if (set != 0)
2400 ctx_set_mount_opt(ctx, m->mount_opt);
2401 else
2402 ctx_clear_mount_opt(ctx, m->mount_opt);
2403 }
2404 }
2405
2406 return 0;
2407 }
2408
2409 static int parse_options(struct fs_context *fc, char *options)
2410 {
2411 struct fs_parameter param;
2412 int ret;
2413 char *key;
2414
2415 if (!options)
2416 return 0;
2417
2418 while ((key = strsep(&options, ",")) != NULL) {
2419 if (*key) {
2420 size_t v_len = 0;
2421 char *value = strchr(key, '=');
2422
2423 param.type = fs_value_is_flag;
2424 param.string = NULL;
2425
2426 if (value) {
2427 if (value == key)
2428 continue;
2429
2430 *value++ = 0;
2431 v_len = strlen(value);
2432 param.string = kmemdup_nul(value, v_len,
2433 GFP_KERNEL);
2434 if (!param.string)
2435 return -ENOMEM;
2436 param.type = fs_value_is_string;
2437 }
2438
2439 param.key = key;
2440 param.size = v_len;
2441
2442 ret = ext4_parse_param(fc, &param);
2443 kfree(param.string);
2444 if (ret < 0)
2445 return ret;
2446 }
2447 }
2448
2449 ret = ext4_validate_options(fc);
2450 if (ret < 0)
2451 return ret;
2452
2453 return 0;
2454 }
2455
2456 static int parse_apply_sb_mount_options(struct super_block *sb,
2457 struct ext4_fs_context *m_ctx)
2458 {
2459 struct ext4_sb_info *sbi = EXT4_SB(sb);
2460 char *s_mount_opts = NULL;
2461 struct ext4_fs_context *s_ctx = NULL;
2462 struct fs_context *fc = NULL;
2463 int ret = -ENOMEM;
2464
2465 if (!sbi->s_es->s_mount_opts[0])
2466 return 0;
2467
2468 s_mount_opts = kstrndup(sbi->s_es->s_mount_opts,
2469 sizeof(sbi->s_es->s_mount_opts),
2470 GFP_KERNEL);
2471 if (!s_mount_opts)
2472 return ret;
2473
2474 fc = kzalloc(sizeof(struct fs_context), GFP_KERNEL);
2475 if (!fc)
2476 goto out_free;
2477
2478 s_ctx = kzalloc(sizeof(struct ext4_fs_context), GFP_KERNEL);
2479 if (!s_ctx)
2480 goto out_free;
2481
2482 fc->fs_private = s_ctx;
2483 fc->s_fs_info = sbi;
2484
2485 ret = parse_options(fc, s_mount_opts);
2486 if (ret < 0)
2487 goto parse_failed;
2488
2489 ret = ext4_check_opt_consistency(fc, sb);
2490 if (ret < 0) {
2491 parse_failed:
2492 ext4_msg(sb, KERN_WARNING,
2493 "failed to parse options in superblock: %s",
2494 s_mount_opts);
2495 ret = 0;
2496 goto out_free;
2497 }
2498
2499 if (s_ctx->spec & EXT4_SPEC_JOURNAL_DEV)
2500 m_ctx->journal_devnum = s_ctx->journal_devnum;
2501 if (s_ctx->spec & EXT4_SPEC_JOURNAL_IOPRIO)
2502 m_ctx->journal_ioprio = s_ctx->journal_ioprio;
2503
2504 ext4_apply_options(fc, sb);
2505 ret = 0;
2506
2507 out_free:
2508 if (fc) {
2509 ext4_fc_free(fc);
2510 kfree(fc);
2511 }
2512 kfree(s_mount_opts);
2513 return ret;
2514 }
2515
2516 static void ext4_apply_quota_options(struct fs_context *fc,
2517 struct super_block *sb)
2518 {
2519 #ifdef CONFIG_QUOTA
2520 bool quota_feature = ext4_has_feature_quota(sb);
2521 struct ext4_fs_context *ctx = fc->fs_private;
2522 struct ext4_sb_info *sbi = EXT4_SB(sb);
2523 char *qname;
2524 int i;
2525
2526 if (quota_feature)
2527 return;
2528
2529 if (ctx->spec & EXT4_SPEC_JQUOTA) {
2530 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2531 if (!(ctx->qname_spec & (1 << i)))
2532 continue;
2533
2534 qname = ctx->s_qf_names[i]; /* May be NULL */
2535 if (qname)
2536 set_opt(sb, QUOTA);
2537 ctx->s_qf_names[i] = NULL;
2538 qname = rcu_replace_pointer(sbi->s_qf_names[i], qname,
2539 lockdep_is_held(&sb->s_umount));
2540 if (qname)
2541 kfree_rcu_mightsleep(qname);
2542 }
2543 }
2544
2545 if (ctx->spec & EXT4_SPEC_JQFMT)
2546 sbi->s_jquota_fmt = ctx->s_jquota_fmt;
2547 #endif
2548 }
2549
2550 /*
2551 * Check quota settings consistency.
2552 */
2553 static int ext4_check_quota_consistency(struct fs_context *fc,
2554 struct super_block *sb)
2555 {
2556 #ifdef CONFIG_QUOTA
2557 struct ext4_fs_context *ctx = fc->fs_private;
2558 struct ext4_sb_info *sbi = EXT4_SB(sb);
2559 bool quota_feature = ext4_has_feature_quota(sb);
2560 bool quota_loaded = sb_any_quota_loaded(sb);
2561 bool usr_qf_name, grp_qf_name, usrquota, grpquota;
2562 int quota_flags, i;
2563
2564 /*
2565 * We do the test below only for project quotas. 'usrquota' and
2566 * 'grpquota' mount options are allowed even without quota feature
2567 * to support legacy quotas in quota files.
2568 */
2569 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_PRJQUOTA) &&
2570 !ext4_has_feature_project(sb)) {
2571 ext4_msg(NULL, KERN_ERR, "Project quota feature not enabled. "
2572 "Cannot enable project quota enforcement.");
2573 return -EINVAL;
2574 }
2575
2576 quota_flags = EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
2577 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA;
2578 if (quota_loaded &&
2579 ctx->mask_s_mount_opt & quota_flags &&
2580 !ctx_test_mount_opt(ctx, quota_flags))
2581 goto err_quota_change;
2582
2583 if (ctx->spec & EXT4_SPEC_JQUOTA) {
2584
2585 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2586 if (!(ctx->qname_spec & (1 << i)))
2587 continue;
2588
2589 if (quota_loaded &&
2590 !!sbi->s_qf_names[i] != !!ctx->s_qf_names[i])
2591 goto err_jquota_change;
2592
2593 if (sbi->s_qf_names[i] && ctx->s_qf_names[i] &&
2594 strcmp(get_qf_name(sb, sbi, i),
2595 ctx->s_qf_names[i]) != 0)
2596 goto err_jquota_specified;
2597 }
2598
2599 if (quota_feature) {
2600 ext4_msg(NULL, KERN_INFO,
2601 "Journaled quota options ignored when "
2602 "QUOTA feature is enabled");
2603 return 0;
2604 }
2605 }
2606
2607 if (ctx->spec & EXT4_SPEC_JQFMT) {
2608 if (sbi->s_jquota_fmt != ctx->s_jquota_fmt && quota_loaded)
2609 goto err_jquota_change;
2610 if (quota_feature) {
2611 ext4_msg(NULL, KERN_INFO, "Quota format mount options "
2612 "ignored when QUOTA feature is enabled");
2613 return 0;
2614 }
2615 }
2616
2617 /* Make sure we don't mix old and new quota format */
2618 usr_qf_name = (get_qf_name(sb, sbi, USRQUOTA) ||
2619 ctx->s_qf_names[USRQUOTA]);
2620 grp_qf_name = (get_qf_name(sb, sbi, GRPQUOTA) ||
2621 ctx->s_qf_names[GRPQUOTA]);
2622
2623 usrquota = (ctx_test_mount_opt(ctx, EXT4_MOUNT_USRQUOTA) ||
2624 test_opt(sb, USRQUOTA));
2625
2626 grpquota = (ctx_test_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA) ||
2627 test_opt(sb, GRPQUOTA));
2628
2629 if (usr_qf_name) {
2630 ctx_clear_mount_opt(ctx, EXT4_MOUNT_USRQUOTA);
2631 usrquota = false;
2632 }
2633 if (grp_qf_name) {
2634 ctx_clear_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA);
2635 grpquota = false;
2636 }
2637
2638 if (usr_qf_name || grp_qf_name) {
2639 if (usrquota || grpquota) {
2640 ext4_msg(NULL, KERN_ERR, "old and new quota "
2641 "format mixing");
2642 return -EINVAL;
2643 }
2644
2645 if (!(ctx->spec & EXT4_SPEC_JQFMT || sbi->s_jquota_fmt)) {
2646 ext4_msg(NULL, KERN_ERR, "journaled quota format "
2647 "not specified");
2648 return -EINVAL;
2649 }
2650 }
2651
2652 return 0;
2653
2654 err_quota_change:
2655 ext4_msg(NULL, KERN_ERR,
2656 "Cannot change quota options when quota turned on");
2657 return -EINVAL;
2658 err_jquota_change:
2659 ext4_msg(NULL, KERN_ERR, "Cannot change journaled quota "
2660 "options when quota turned on");
2661 return -EINVAL;
2662 err_jquota_specified:
2663 ext4_msg(NULL, KERN_ERR, "%s quota file already specified",
2664 QTYPE2NAME(i));
2665 return -EINVAL;
2666 #else
2667 return 0;
2668 #endif
2669 }
2670
2671 static int ext4_check_test_dummy_encryption(const struct fs_context *fc,
2672 struct super_block *sb)
2673 {
2674 const struct ext4_fs_context *ctx = fc->fs_private;
2675 const struct ext4_sb_info *sbi = EXT4_SB(sb);
2676
2677 if (!fscrypt_is_dummy_policy_set(&ctx->dummy_enc_policy))
2678 return 0;
2679
2680 if (!ext4_has_feature_encrypt(sb)) {
2681 ext4_msg(NULL, KERN_WARNING,
2682 "test_dummy_encryption requires encrypt feature");
2683 return -EINVAL;
2684 }
2685 /*
2686 * This mount option is just for testing, and it's not worthwhile to
2687 * implement the extra complexity (e.g. RCU protection) that would be
2688 * needed to allow it to be set or changed during remount. We do allow
2689 * it to be specified during remount, but only if there is no change.
2690 */
2691 if (fc->purpose == FS_CONTEXT_FOR_RECONFIGURE) {
2692 if (fscrypt_dummy_policies_equal(&sbi->s_dummy_enc_policy,
2693 &ctx->dummy_enc_policy))
2694 return 0;
2695 ext4_msg(NULL, KERN_WARNING,
2696 "Can't set or change test_dummy_encryption on remount");
2697 return -EINVAL;
2698 }
2699 /* Also make sure s_mount_opts didn't contain a conflicting value. */
2700 if (fscrypt_is_dummy_policy_set(&sbi->s_dummy_enc_policy)) {
2701 if (fscrypt_dummy_policies_equal(&sbi->s_dummy_enc_policy,
2702 &ctx->dummy_enc_policy))
2703 return 0;
2704 ext4_msg(NULL, KERN_WARNING,
2705 "Conflicting test_dummy_encryption options");
2706 return -EINVAL;
2707 }
2708 return 0;
2709 }
2710
2711 static void ext4_apply_test_dummy_encryption(struct ext4_fs_context *ctx,
2712 struct super_block *sb)
2713 {
2714 if (!fscrypt_is_dummy_policy_set(&ctx->dummy_enc_policy) ||
2715 /* if already set, it was already verified to be the same */
2716 fscrypt_is_dummy_policy_set(&EXT4_SB(sb)->s_dummy_enc_policy))
2717 return;
2718 EXT4_SB(sb)->s_dummy_enc_policy = ctx->dummy_enc_policy;
2719 memset(&ctx->dummy_enc_policy, 0, sizeof(ctx->dummy_enc_policy));
2720 ext4_msg(sb, KERN_WARNING, "Test dummy encryption mode enabled");
2721 }
2722
2723 static int ext4_check_opt_consistency(struct fs_context *fc,
2724 struct super_block *sb)
2725 {
2726 struct ext4_fs_context *ctx = fc->fs_private;
2727 struct ext4_sb_info *sbi = fc->s_fs_info;
2728 int is_remount = fc->purpose == FS_CONTEXT_FOR_RECONFIGURE;
2729 int err;
2730
2731 if ((ctx->opt_flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
2732 ext4_msg(NULL, KERN_ERR,
2733 "Mount option(s) incompatible with ext2");
2734 return -EINVAL;
2735 }
2736 if ((ctx->opt_flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
2737 ext4_msg(NULL, KERN_ERR,
2738 "Mount option(s) incompatible with ext3");
2739 return -EINVAL;
2740 }
2741
2742 if (ctx->s_want_extra_isize >
2743 (sbi->s_inode_size - EXT4_GOOD_OLD_INODE_SIZE)) {
2744 ext4_msg(NULL, KERN_ERR,
2745 "Invalid want_extra_isize %d",
2746 ctx->s_want_extra_isize);
2747 return -EINVAL;
2748 }
2749
2750 err = ext4_check_test_dummy_encryption(fc, sb);
2751 if (err)
2752 return err;
2753
2754 if ((ctx->spec & EXT4_SPEC_DATAJ) && is_remount) {
2755 if (!sbi->s_journal) {
2756 ext4_msg(NULL, KERN_WARNING,
2757 "Remounting file system with no journal "
2758 "so ignoring journalled data option");
2759 ctx_clear_mount_opt(ctx, EXT4_MOUNT_DATA_FLAGS);
2760 } else if (ctx_test_mount_opt(ctx, EXT4_MOUNT_DATA_FLAGS) !=
2761 test_opt(sb, DATA_FLAGS)) {
2762 ext4_msg(NULL, KERN_ERR, "Cannot change data mode "
2763 "on remount");
2764 return -EINVAL;
2765 }
2766 }
2767
2768 if (is_remount) {
2769 if (!sbi->s_journal &&
2770 ctx_test_mount_opt(ctx, EXT4_MOUNT_DATA_ERR_ABORT)) {
2771 ext4_msg(NULL, KERN_WARNING,
2772 "Remounting fs w/o journal so ignoring data_err option");
2773 ctx_clear_mount_opt(ctx, EXT4_MOUNT_DATA_ERR_ABORT);
2774 }
2775
2776 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS) &&
2777 (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)) {
2778 ext4_msg(NULL, KERN_ERR, "can't mount with "
2779 "both data=journal and dax");
2780 return -EINVAL;
2781 }
2782
2783 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_DAX_ALWAYS) &&
2784 (!(sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) ||
2785 (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER))) {
2786 fail_dax_change_remount:
2787 ext4_msg(NULL, KERN_ERR, "can't change "
2788 "dax mount option while remounting");
2789 return -EINVAL;
2790 } else if (ctx_test_mount_opt2(ctx, EXT4_MOUNT2_DAX_NEVER) &&
2791 (!(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) ||
2792 (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS))) {
2793 goto fail_dax_change_remount;
2794 } else if (ctx_test_mount_opt2(ctx, EXT4_MOUNT2_DAX_INODE) &&
2795 ((sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) ||
2796 (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) ||
2797 !(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_INODE))) {
2798 goto fail_dax_change_remount;
2799 }
2800 }
2801
2802 return ext4_check_quota_consistency(fc, sb);
2803 }
2804
2805 static void ext4_apply_options(struct fs_context *fc, struct super_block *sb)
2806 {
2807 struct ext4_fs_context *ctx = fc->fs_private;
2808 struct ext4_sb_info *sbi = fc->s_fs_info;
2809
2810 sbi->s_mount_opt &= ~ctx->mask_s_mount_opt;
2811 sbi->s_mount_opt |= ctx->vals_s_mount_opt;
2812 sbi->s_mount_opt2 &= ~ctx->mask_s_mount_opt2;
2813 sbi->s_mount_opt2 |= ctx->vals_s_mount_opt2;
2814 sb->s_flags &= ~ctx->mask_s_flags;
2815 sb->s_flags |= ctx->vals_s_flags;
2816
2817 #define APPLY(X) ({ if (ctx->spec & EXT4_SPEC_##X) sbi->X = ctx->X; })
2818 APPLY(s_commit_interval);
2819 APPLY(s_stripe);
2820 APPLY(s_max_batch_time);
2821 APPLY(s_min_batch_time);
2822 APPLY(s_want_extra_isize);
2823 APPLY(s_inode_readahead_blks);
2824 APPLY(s_max_dir_size_kb);
2825 APPLY(s_li_wait_mult);
2826 APPLY(s_resgid);
2827 APPLY(s_resuid);
2828
2829 #ifdef CONFIG_EXT4_DEBUG
2830 APPLY(s_fc_debug_max_replay);
2831 #endif
2832
2833 ext4_apply_quota_options(fc, sb);
2834 ext4_apply_test_dummy_encryption(ctx, sb);
2835 }
2836
2837
2838 static int ext4_validate_options(struct fs_context *fc)
2839 {
2840 #ifdef CONFIG_QUOTA
2841 struct ext4_fs_context *ctx = fc->fs_private;
2842 char *usr_qf_name, *grp_qf_name;
2843
2844 usr_qf_name = ctx->s_qf_names[USRQUOTA];
2845 grp_qf_name = ctx->s_qf_names[GRPQUOTA];
2846
2847 if (usr_qf_name || grp_qf_name) {
2848 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_USRQUOTA) && usr_qf_name)
2849 ctx_clear_mount_opt(ctx, EXT4_MOUNT_USRQUOTA);
2850
2851 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA) && grp_qf_name)
2852 ctx_clear_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA);
2853
2854 if (ctx_test_mount_opt(ctx, EXT4_MOUNT_USRQUOTA) ||
2855 ctx_test_mount_opt(ctx, EXT4_MOUNT_GRPQUOTA)) {
2856 ext4_msg(NULL, KERN_ERR, "old and new quota "
2857 "format mixing");
2858 return -EINVAL;
2859 }
2860 }
2861 #endif
2862 return 1;
2863 }
2864
2865 static inline void ext4_show_quota_options(struct seq_file *seq,
2866 struct super_block *sb)
2867 {
2868 #if defined(CONFIG_QUOTA)
2869 struct ext4_sb_info *sbi = EXT4_SB(sb);
2870 char *usr_qf_name, *grp_qf_name;
2871
2872 if (sbi->s_jquota_fmt) {
2873 char *fmtname = "";
2874
2875 switch (sbi->s_jquota_fmt) {
2876 case QFMT_VFS_OLD:
2877 fmtname = "vfsold";
2878 break;
2879 case QFMT_VFS_V0:
2880 fmtname = "vfsv0";
2881 break;
2882 case QFMT_VFS_V1:
2883 fmtname = "vfsv1";
2884 break;
2885 }
2886 seq_printf(seq, ",jqfmt=%s", fmtname);
2887 }
2888
2889 rcu_read_lock();
2890 usr_qf_name = rcu_dereference(sbi->s_qf_names[USRQUOTA]);
2891 grp_qf_name = rcu_dereference(sbi->s_qf_names[GRPQUOTA]);
2892 if (usr_qf_name)
2893 seq_show_option(seq, "usrjquota", usr_qf_name);
2894 if (grp_qf_name)
2895 seq_show_option(seq, "grpjquota", grp_qf_name);
2896 rcu_read_unlock();
2897 #endif
2898 }
2899
2900 static const char *token2str(int token)
2901 {
2902 const struct fs_parameter_spec *spec;
2903
2904 for (spec = ext4_param_specs; spec->name != NULL; spec++)
2905 if (spec->opt == token && !spec->type)
2906 break;
2907 return spec->name;
2908 }
2909
2910 /*
2911 * Show an option if
2912 * - it's set to a non-default value OR
2913 * - if the per-sb default is different from the global default
2914 */
2915 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
2916 int nodefs)
2917 {
2918 struct ext4_sb_info *sbi = EXT4_SB(sb);
2919 struct ext4_super_block *es = sbi->s_es;
2920 int def_errors;
2921 const struct mount_opts *m;
2922 char sep = nodefs ? '\n' : ',';
2923
2924 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
2925 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2926
2927 if (sbi->s_sb_block != 1)
2928 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
2929
2930 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
2931 int want_set = m->flags & MOPT_SET;
2932 int opt_2 = m->flags & MOPT_2;
2933 unsigned int mount_opt, def_mount_opt;
2934
2935 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
2936 m->flags & MOPT_SKIP)
2937 continue;
2938
2939 if (opt_2) {
2940 mount_opt = sbi->s_mount_opt2;
2941 def_mount_opt = sbi->s_def_mount_opt2;
2942 } else {
2943 mount_opt = sbi->s_mount_opt;
2944 def_mount_opt = sbi->s_def_mount_opt;
2945 }
2946 /* skip if same as the default */
2947 if (!nodefs && !(m->mount_opt & (mount_opt ^ def_mount_opt)))
2948 continue;
2949 /* select Opt_noFoo vs Opt_Foo */
2950 if ((want_set &&
2951 (mount_opt & m->mount_opt) != m->mount_opt) ||
2952 (!want_set && (mount_opt & m->mount_opt)))
2953 continue;
2954 SEQ_OPTS_PRINT("%s", token2str(m->token));
2955 }
2956
2957 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
2958 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
2959 SEQ_OPTS_PRINT("resuid=%u",
2960 from_kuid_munged(&init_user_ns, sbi->s_resuid));
2961 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
2962 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
2963 SEQ_OPTS_PRINT("resgid=%u",
2964 from_kgid_munged(&init_user_ns, sbi->s_resgid));
2965 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
2966 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
2967 SEQ_OPTS_PUTS("errors=remount-ro");
2968 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
2969 SEQ_OPTS_PUTS("errors=continue");
2970 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
2971 SEQ_OPTS_PUTS("errors=panic");
2972 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
2973 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
2974 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
2975 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
2976 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
2977 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2978 if (nodefs || sbi->s_stripe)
2979 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
2980 if (nodefs || EXT4_MOUNT_DATA_FLAGS &
2981 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
2982 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2983 SEQ_OPTS_PUTS("data=journal");
2984 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2985 SEQ_OPTS_PUTS("data=ordered");
2986 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
2987 SEQ_OPTS_PUTS("data=writeback");
2988 }
2989 if (nodefs ||
2990 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
2991 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
2992 sbi->s_inode_readahead_blks);
2993
2994 if (test_opt(sb, INIT_INODE_TABLE) && (nodefs ||
2995 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
2996 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
2997 if (nodefs || sbi->s_max_dir_size_kb)
2998 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
2999 if (test_opt(sb, DATA_ERR_ABORT))
3000 SEQ_OPTS_PUTS("data_err=abort");
3001
3002 fscrypt_show_test_dummy_encryption(seq, sep, sb);
3003
3004 if (sb->s_flags & SB_INLINECRYPT)
3005 SEQ_OPTS_PUTS("inlinecrypt");
3006
3007 if (test_opt(sb, DAX_ALWAYS)) {
3008 if (IS_EXT2_SB(sb))
3009 SEQ_OPTS_PUTS("dax");
3010 else
3011 SEQ_OPTS_PUTS("dax=always");
3012 } else if (test_opt2(sb, DAX_NEVER)) {
3013 SEQ_OPTS_PUTS("dax=never");
3014 } else if (test_opt2(sb, DAX_INODE)) {
3015 SEQ_OPTS_PUTS("dax=inode");
3016 }
3017
3018 if (sbi->s_groups_count >= MB_DEFAULT_LINEAR_SCAN_THRESHOLD &&
3019 !test_opt2(sb, MB_OPTIMIZE_SCAN)) {
3020 SEQ_OPTS_PUTS("mb_optimize_scan=0");
3021 } else if (sbi->s_groups_count < MB_DEFAULT_LINEAR_SCAN_THRESHOLD &&
3022 test_opt2(sb, MB_OPTIMIZE_SCAN)) {
3023 SEQ_OPTS_PUTS("mb_optimize_scan=1");
3024 }
3025
3026 if (nodefs && !test_opt(sb, NO_PREFETCH_BLOCK_BITMAPS))
3027 SEQ_OPTS_PUTS("prefetch_block_bitmaps");
3028
3029 if (ext4_emergency_ro(sb))
3030 SEQ_OPTS_PUTS("emergency_ro");
3031
3032 if (ext4_forced_shutdown(sb))
3033 SEQ_OPTS_PUTS("shutdown");
3034
3035 ext4_show_quota_options(seq, sb);
3036 return 0;
3037 }
3038
3039 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
3040 {
3041 return _ext4_show_options(seq, root->d_sb, 0);
3042 }
3043
3044 int ext4_seq_options_show(struct seq_file *seq, void *offset)
3045 {
3046 struct super_block *sb = seq->private;
3047 int rc;
3048
3049 seq_puts(seq, sb_rdonly(sb) ? "ro" : "rw");
3050 rc = _ext4_show_options(seq, sb, 1);
3051 seq_putc(seq, '\n');
3052 return rc;
3053 }
3054
3055 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
3056 int read_only)
3057 {
3058 struct ext4_sb_info *sbi = EXT4_SB(sb);
3059 int err = 0;
3060
3061 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
3062 ext4_msg(sb, KERN_ERR, "revision level too high, "
3063 "forcing read-only mode");
3064 err = -EROFS;
3065 goto done;
3066 }
3067 if (read_only)
3068 goto done;
3069 if (!(sbi->s_mount_state & EXT4_VALID_FS))
3070 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
3071 "running e2fsck is recommended");
3072 else if (sbi->s_mount_state & EXT4_ERROR_FS)
3073 ext4_msg(sb, KERN_WARNING,
3074 "warning: mounting fs with errors, "
3075 "running e2fsck is recommended");
3076 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
3077 le16_to_cpu(es->s_mnt_count) >=
3078 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
3079 ext4_msg(sb, KERN_WARNING,
3080 "warning: maximal mount count reached, "
3081 "running e2fsck is recommended");
3082 else if (le32_to_cpu(es->s_checkinterval) &&
3083 (ext4_get_tstamp(es, s_lastcheck) +
3084 le32_to_cpu(es->s_checkinterval) <= ktime_get_real_seconds()))
3085 ext4_msg(sb, KERN_WARNING,
3086 "warning: checktime reached, "
3087 "running e2fsck is recommended");
3088 if (!sbi->s_journal)
3089 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
3090 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
3091 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
3092 le16_add_cpu(&es->s_mnt_count, 1);
3093 ext4_update_tstamp(es, s_mtime);
3094 if (sbi->s_journal) {
3095 ext4_set_feature_journal_needs_recovery(sb);
3096 if (ext4_has_feature_orphan_file(sb))
3097 ext4_set_feature_orphan_present(sb);
3098 }
3099
3100 err = ext4_commit_super(sb);
3101 done:
3102 if (test_opt(sb, DEBUG))
3103 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
3104 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
3105 sb->s_blocksize,
3106 sbi->s_groups_count,
3107 EXT4_BLOCKS_PER_GROUP(sb),
3108 EXT4_INODES_PER_GROUP(sb),
3109 sbi->s_mount_opt, sbi->s_mount_opt2);
3110 return err;
3111 }
3112
3113 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
3114 {
3115 struct ext4_sb_info *sbi = EXT4_SB(sb);
3116 struct flex_groups **old_groups, **new_groups;
3117 int size, i, j;
3118
3119 if (!sbi->s_log_groups_per_flex)
3120 return 0;
3121
3122 size = ext4_flex_group(sbi, ngroup - 1) + 1;
3123 if (size <= sbi->s_flex_groups_allocated)
3124 return 0;
3125
3126 new_groups = kvzalloc(roundup_pow_of_two(size *
3127 sizeof(*sbi->s_flex_groups)), GFP_KERNEL);
3128 if (!new_groups) {
3129 ext4_msg(sb, KERN_ERR,
3130 "not enough memory for %d flex group pointers", size);
3131 return -ENOMEM;
3132 }
3133 for (i = sbi->s_flex_groups_allocated; i < size; i++) {
3134 new_groups[i] = kvzalloc(roundup_pow_of_two(
3135 sizeof(struct flex_groups)),
3136 GFP_KERNEL);
3137 if (!new_groups[i]) {
3138 for (j = sbi->s_flex_groups_allocated; j < i; j++)
3139 kvfree(new_groups[j]);
3140 kvfree(new_groups);
3141 ext4_msg(sb, KERN_ERR,
3142 "not enough memory for %d flex groups", size);
3143 return -ENOMEM;
3144 }
3145 }
3146 rcu_read_lock();
3147 old_groups = rcu_dereference(sbi->s_flex_groups);
3148 if (old_groups)
3149 memcpy(new_groups, old_groups,
3150 (sbi->s_flex_groups_allocated *
3151 sizeof(struct flex_groups *)));
3152 rcu_read_unlock();
3153 rcu_assign_pointer(sbi->s_flex_groups, new_groups);
3154 sbi->s_flex_groups_allocated = size;
3155 if (old_groups)
3156 ext4_kvfree_array_rcu(old_groups);
3157 return 0;
3158 }
3159
3160 static int ext4_fill_flex_info(struct super_block *sb)
3161 {
3162 struct ext4_sb_info *sbi = EXT4_SB(sb);
3163 struct ext4_group_desc *gdp = NULL;
3164 struct flex_groups *fg;
3165 ext4_group_t flex_group;
3166 int i, err;
3167
3168 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
3169 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
3170 sbi->s_log_groups_per_flex = 0;
3171 return 1;
3172 }
3173
3174 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
3175 if (err)
3176 goto failed;
3177
3178 for (i = 0; i < sbi->s_groups_count; i++) {
3179 gdp = ext4_get_group_desc(sb, i, NULL);
3180
3181 flex_group = ext4_flex_group(sbi, i);
3182 fg = sbi_array_rcu_deref(sbi, s_flex_groups, flex_group);
3183 atomic_add(ext4_free_inodes_count(sb, gdp), &fg->free_inodes);
3184 atomic64_add(ext4_free_group_clusters(sb, gdp),
3185 &fg->free_clusters);
3186 atomic_add(ext4_used_dirs_count(sb, gdp), &fg->used_dirs);
3187 }
3188
3189 return 1;
3190 failed:
3191 return 0;
3192 }
3193
3194 static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
3195 struct ext4_group_desc *gdp)
3196 {
3197 int offset = offsetof(struct ext4_group_desc, bg_checksum);
3198 __u16 crc = 0;
3199 __le32 le_group = cpu_to_le32(block_group);
3200 struct ext4_sb_info *sbi = EXT4_SB(sb);
3201
3202 if (ext4_has_feature_metadata_csum(sbi->s_sb)) {
3203 /* Use new metadata_csum algorithm */
3204 __u32 csum32;
3205 __u16 dummy_csum = 0;
3206
3207 csum32 = ext4_chksum(sbi->s_csum_seed, (__u8 *)&le_group,
3208 sizeof(le_group));
3209 csum32 = ext4_chksum(csum32, (__u8 *)gdp, offset);
3210 csum32 = ext4_chksum(csum32, (__u8 *)&dummy_csum,
3211 sizeof(dummy_csum));
3212 offset += sizeof(dummy_csum);
3213 if (offset < sbi->s_desc_size)
3214 csum32 = ext4_chksum(csum32, (__u8 *)gdp + offset,
3215 sbi->s_desc_size - offset);
3216
3217 crc = csum32 & 0xFFFF;
3218 goto out;
3219 }
3220
3221 /* old crc16 code */
3222 if (!ext4_has_feature_gdt_csum(sb))
3223 return 0;
3224
3225 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
3226 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
3227 crc = crc16(crc, (__u8 *)gdp, offset);
3228 offset += sizeof(gdp->bg_checksum); /* skip checksum */
3229 /* for checksum of struct ext4_group_desc do the rest...*/
3230 if (ext4_has_feature_64bit(sb) && offset < sbi->s_desc_size)
3231 crc = crc16(crc, (__u8 *)gdp + offset,
3232 sbi->s_desc_size - offset);
3233
3234 out:
3235 return cpu_to_le16(crc);
3236 }
3237
3238 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
3239 struct ext4_group_desc *gdp)
3240 {
3241 if (ext4_has_group_desc_csum(sb) &&
3242 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
3243 return 0;
3244
3245 return 1;
3246 }
3247
3248 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
3249 struct ext4_group_desc *gdp)
3250 {
3251 if (!ext4_has_group_desc_csum(sb))
3252 return;
3253 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
3254 }
3255
3256 /* Called at mount-time, super-block is locked */
3257 static int ext4_check_descriptors(struct super_block *sb,
3258 ext4_fsblk_t sb_block,
3259 ext4_group_t *first_not_zeroed)
3260 {
3261 struct ext4_sb_info *sbi = EXT4_SB(sb);
3262 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
3263 ext4_fsblk_t last_block;
3264 ext4_fsblk_t last_bg_block = sb_block + ext4_bg_num_gdb(sb, 0);
3265 ext4_fsblk_t block_bitmap;
3266 ext4_fsblk_t inode_bitmap;
3267 ext4_fsblk_t inode_table;
3268 int flexbg_flag = 0;
3269 ext4_group_t i, grp = sbi->s_groups_count;
3270
3271 if (ext4_has_feature_flex_bg(sb))
3272 flexbg_flag = 1;
3273
3274 ext4_debug("Checking group descriptors");
3275
3276 for (i = 0; i < sbi->s_groups_count; i++) {
3277 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
3278
3279 if (i == sbi->s_groups_count - 1 || flexbg_flag)
3280 last_block = ext4_blocks_count(sbi->s_es) - 1;
3281 else
3282 last_block = first_block +
3283 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
3284
3285 if ((grp == sbi->s_groups_count) &&
3286 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3287 grp = i;
3288
3289 block_bitmap = ext4_block_bitmap(sb, gdp);
3290 if (block_bitmap == sb_block) {
3291 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3292 "Block bitmap for group %u overlaps "
3293 "superblock", i);
3294 if (!sb_rdonly(sb))
3295 return 0;
3296 }
3297 if (block_bitmap >= sb_block + 1 &&
3298 block_bitmap <= last_bg_block) {
3299 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3300 "Block bitmap for group %u overlaps "
3301 "block group descriptors", i);
3302 if (!sb_rdonly(sb))
3303 return 0;
3304 }
3305 if (block_bitmap < first_block || block_bitmap > last_block) {
3306 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3307 "Block bitmap for group %u not in group "
3308 "(block %llu)!", i, block_bitmap);
3309 return 0;
3310 }
3311 inode_bitmap = ext4_inode_bitmap(sb, gdp);
3312 if (inode_bitmap == sb_block) {
3313 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3314 "Inode bitmap for group %u overlaps "
3315 "superblock", i);
3316 if (!sb_rdonly(sb))
3317 return 0;
3318 }
3319 if (inode_bitmap >= sb_block + 1 &&
3320 inode_bitmap <= last_bg_block) {
3321 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3322 "Inode bitmap for group %u overlaps "
3323 "block group descriptors", i);
3324 if (!sb_rdonly(sb))
3325 return 0;
3326 }
3327 if (inode_bitmap < first_block || inode_bitmap > last_block) {
3328 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3329 "Inode bitmap for group %u not in group "
3330 "(block %llu)!", i, inode_bitmap);
3331 return 0;
3332 }
3333 inode_table = ext4_inode_table(sb, gdp);
3334 if (inode_table == sb_block) {
3335 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3336 "Inode table for group %u overlaps "
3337 "superblock", i);
3338 if (!sb_rdonly(sb))
3339 return 0;
3340 }
3341 if (inode_table >= sb_block + 1 &&
3342 inode_table <= last_bg_block) {
3343 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3344 "Inode table for group %u overlaps "
3345 "block group descriptors", i);
3346 if (!sb_rdonly(sb))
3347 return 0;
3348 }
3349 if (inode_table < first_block ||
3350 inode_table + sbi->s_itb_per_group - 1 > last_block) {
3351 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3352 "Inode table for group %u not in group "
3353 "(block %llu)!", i, inode_table);
3354 return 0;
3355 }
3356 ext4_lock_group(sb, i);
3357 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
3358 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
3359 "Checksum for group %u failed (%u!=%u)",
3360 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
3361 gdp)), le16_to_cpu(gdp->bg_checksum));
3362 if (!sb_rdonly(sb)) {
3363 ext4_unlock_group(sb, i);
3364 return 0;
3365 }
3366 }
3367 ext4_unlock_group(sb, i);
3368 if (!flexbg_flag)
3369 first_block += EXT4_BLOCKS_PER_GROUP(sb);
3370 }
3371 if (NULL != first_not_zeroed)
3372 *first_not_zeroed = grp;
3373 return 1;
3374 }
3375
3376 /*
3377 * Maximal extent format file size.
3378 * Resulting logical blkno at s_maxbytes must fit in our on-disk
3379 * extent format containers, within a sector_t, and within i_blocks
3380 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
3381 * so that won't be a limiting factor.
3382 *
3383 * However there is other limiting factor. We do store extents in the form
3384 * of starting block and length, hence the resulting length of the extent
3385 * covering maximum file size must fit into on-disk format containers as
3386 * well. Given that length is always by 1 unit bigger than max unit (because
3387 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
3388 *
3389 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
3390 */
3391 static loff_t ext4_max_size(int blkbits, int has_huge_files)
3392 {
3393 loff_t res;
3394 loff_t upper_limit = MAX_LFS_FILESIZE;
3395
3396 BUILD_BUG_ON(sizeof(blkcnt_t) < sizeof(u64));
3397
3398 if (!has_huge_files) {
3399 upper_limit = (1LL << 32) - 1;
3400
3401 /* total blocks in file system block size */
3402 upper_limit >>= (blkbits - 9);
3403 upper_limit <<= blkbits;
3404 }
3405
3406 /*
3407 * 32-bit extent-start container, ee_block. We lower the maxbytes
3408 * by one fs block, so ee_len can cover the extent of maximum file
3409 * size
3410 */
3411 res = (1LL << 32) - 1;
3412 res <<= blkbits;
3413
3414 /* Sanity check against vm- & vfs- imposed limits */
3415 if (res > upper_limit)
3416 res = upper_limit;
3417
3418 return res;
3419 }
3420
3421 /*
3422 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
3423 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
3424 * We need to be 1 filesystem block less than the 2^48 sector limit.
3425 */
3426 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
3427 {
3428 loff_t upper_limit, res = EXT4_NDIR_BLOCKS;
3429 int meta_blocks;
3430 unsigned int ppb = 1 << (bits - 2);
3431
3432 /*
3433 * This is calculated to be the largest file size for a dense, block
3434 * mapped file such that the file's total number of 512-byte sectors,
3435 * including data and all indirect blocks, does not exceed (2^48 - 1).
3436 *
3437 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
3438 * number of 512-byte sectors of the file.
3439 */
3440 if (!has_huge_files) {
3441 /*
3442 * !has_huge_files or implies that the inode i_block field
3443 * represents total file blocks in 2^32 512-byte sectors ==
3444 * size of vfs inode i_blocks * 8
3445 */
3446 upper_limit = (1LL << 32) - 1;
3447
3448 /* total blocks in file system block size */
3449 upper_limit >>= (bits - 9);
3450
3451 } else {
3452 /*
3453 * We use 48 bit ext4_inode i_blocks
3454 * With EXT4_HUGE_FILE_FL set the i_blocks
3455 * represent total number of blocks in
3456 * file system block size
3457 */
3458 upper_limit = (1LL << 48) - 1;
3459
3460 }
3461
3462 /* Compute how many blocks we can address by block tree */
3463 res += ppb;
3464 res += ppb * ppb;
3465 res += ((loff_t)ppb) * ppb * ppb;
3466 /* Compute how many metadata blocks are needed */
3467 meta_blocks = 1;
3468 meta_blocks += 1 + ppb;
3469 meta_blocks += 1 + ppb + ppb * ppb;
3470 /* Does block tree limit file size? */
3471 if (res + meta_blocks <= upper_limit)
3472 goto check_lfs;
3473
3474 res = upper_limit;
3475 /* How many metadata blocks are needed for addressing upper_limit? */
3476 upper_limit -= EXT4_NDIR_BLOCKS;
3477 /* indirect blocks */
3478 meta_blocks = 1;
3479 upper_limit -= ppb;
3480 /* double indirect blocks */
3481 if (upper_limit < ppb * ppb) {
3482 meta_blocks += 1 + DIV_ROUND_UP_ULL(upper_limit, ppb);
3483 res -= meta_blocks;
3484 goto check_lfs;
3485 }
3486 meta_blocks += 1 + ppb;
3487 upper_limit -= ppb * ppb;
3488 /* tripple indirect blocks for the rest */
3489 meta_blocks += 1 + DIV_ROUND_UP_ULL(upper_limit, ppb) +
3490 DIV_ROUND_UP_ULL(upper_limit, ppb*ppb);
3491 res -= meta_blocks;
3492 check_lfs:
3493 res <<= bits;
3494 if (res > MAX_LFS_FILESIZE)
3495 res = MAX_LFS_FILESIZE;
3496
3497 return res;
3498 }
3499
3500 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
3501 ext4_fsblk_t logical_sb_block, int nr)
3502 {
3503 struct ext4_sb_info *sbi = EXT4_SB(sb);
3504 ext4_group_t bg, first_meta_bg;
3505 int has_super = 0;
3506
3507 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
3508
3509 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
3510 return logical_sb_block + nr + 1;
3511 bg = sbi->s_desc_per_block * nr;
3512 if (ext4_bg_has_super(sb, bg))
3513 has_super = 1;
3514
3515 /*
3516 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
3517 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
3518 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
3519 * compensate.
3520 */
3521 if (sb->s_blocksize == 1024 && nr == 0 &&
3522 le32_to_cpu(sbi->s_es->s_first_data_block) == 0)
3523 has_super++;
3524
3525 return (has_super + ext4_group_first_block_no(sb, bg));
3526 }
3527
3528 /**
3529 * ext4_get_stripe_size: Get the stripe size.
3530 * @sbi: In memory super block info
3531 *
3532 * If we have specified it via mount option, then
3533 * use the mount option value. If the value specified at mount time is
3534 * greater than the blocks per group use the super block value.
3535 * If the super block value is greater than blocks per group return 0.
3536 * Allocator needs it be less than blocks per group.
3537 *
3538 */
3539 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
3540 {
3541 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
3542 unsigned long stripe_width =
3543 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3544 int ret;
3545
3546 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3547 ret = sbi->s_stripe;
3548 else if (stripe_width && stripe_width <= sbi->s_blocks_per_group)
3549 ret = stripe_width;
3550 else if (stride && stride <= sbi->s_blocks_per_group)
3551 ret = stride;
3552 else
3553 ret = 0;
3554
3555 /*
3556 * If the stripe width is 1, this makes no sense and
3557 * we set it to 0 to turn off stripe handling code.
3558 */
3559 if (ret <= 1)
3560 ret = 0;
3561
3562 return ret;
3563 }
3564
3565 /*
3566 * Check whether this filesystem can be mounted based on
3567 * the features present and the RDONLY/RDWR mount requested.
3568 * Returns 1 if this filesystem can be mounted as requested,
3569 * 0 if it cannot be.
3570 */
3571 int ext4_feature_set_ok(struct super_block *sb, int readonly)
3572 {
3573 if (ext4_has_unknown_ext4_incompat_features(sb)) {
3574 ext4_msg(sb, KERN_ERR,
3575 "Couldn't mount because of "
3576 "unsupported optional features (%x)",
3577 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
3578 ~EXT4_FEATURE_INCOMPAT_SUPP));
3579 return 0;
3580 }
3581
3582 if (!IS_ENABLED(CONFIG_UNICODE) && ext4_has_feature_casefold(sb)) {
3583 ext4_msg(sb, KERN_ERR,
3584 "Filesystem with casefold feature cannot be "
3585 "mounted without CONFIG_UNICODE");
3586 return 0;
3587 }
3588
3589 if (readonly)
3590 return 1;
3591
3592 if (ext4_has_feature_readonly(sb)) {
3593 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
3594 sb->s_flags |= SB_RDONLY;
3595 return 1;
3596 }
3597
3598 /* Check that feature set is OK for a read-write mount */
3599 if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
3600 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
3601 "unsupported optional features (%x)",
3602 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
3603 ~EXT4_FEATURE_RO_COMPAT_SUPP));
3604 return 0;
3605 }
3606 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
3607 ext4_msg(sb, KERN_ERR,
3608 "Can't support bigalloc feature without "
3609 "extents feature\n");
3610 return 0;
3611 }
3612
3613 #if !IS_ENABLED(CONFIG_QUOTA) || !IS_ENABLED(CONFIG_QFMT_V2)
3614 if (!readonly && (ext4_has_feature_quota(sb) ||
3615 ext4_has_feature_project(sb))) {
3616 ext4_msg(sb, KERN_ERR,
3617 "The kernel was not built with CONFIG_QUOTA and CONFIG_QFMT_V2");
3618 return 0;
3619 }
3620 #endif /* CONFIG_QUOTA */
3621 return 1;
3622 }
3623
3624 /*
3625 * This function is called once a day if we have errors logged
3626 * on the file system
3627 */
3628 static void print_daily_error_info(struct timer_list *t)
3629 {
3630 struct ext4_sb_info *sbi = timer_container_of(sbi, t, s_err_report);
3631 struct super_block *sb = sbi->s_sb;
3632 struct ext4_super_block *es = sbi->s_es;
3633
3634 if (es->s_error_count)
3635 /* fsck newer than v1.41.13 is needed to clean this condition. */
3636 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
3637 le32_to_cpu(es->s_error_count));
3638 if (es->s_first_error_time) {
3639 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %llu: %.*s:%d",
3640 sb->s_id,
3641 ext4_get_tstamp(es, s_first_error_time),
3642 (int) sizeof(es->s_first_error_func),
3643 es->s_first_error_func,
3644 le32_to_cpu(es->s_first_error_line));
3645 if (es->s_first_error_ino)
3646 printk(KERN_CONT ": inode %u",
3647 le32_to_cpu(es->s_first_error_ino));
3648 if (es->s_first_error_block)
3649 printk(KERN_CONT ": block %llu", (unsigned long long)
3650 le64_to_cpu(es->s_first_error_block));
3651 printk(KERN_CONT "\n");
3652 }
3653 if (es->s_last_error_time) {
3654 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %llu: %.*s:%d",
3655 sb->s_id,
3656 ext4_get_tstamp(es, s_last_error_time),
3657 (int) sizeof(es->s_last_error_func),
3658 es->s_last_error_func,
3659 le32_to_cpu(es->s_last_error_line));
3660 if (es->s_last_error_ino)
3661 printk(KERN_CONT ": inode %u",
3662 le32_to_cpu(es->s_last_error_ino));
3663 if (es->s_last_error_block)
3664 printk(KERN_CONT ": block %llu", (unsigned long long)
3665 le64_to_cpu(es->s_last_error_block));
3666 printk(KERN_CONT "\n");
3667 }
3668 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
3669 }
3670
3671 /* Find next suitable group and run ext4_init_inode_table */
3672 static int ext4_run_li_request(struct ext4_li_request *elr)
3673 {
3674 struct ext4_group_desc *gdp = NULL;
3675 struct super_block *sb = elr->lr_super;
3676 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3677 ext4_group_t group = elr->lr_next_group;
3678 unsigned int prefetch_ios = 0;
3679 int ret = 0;
3680 int nr = EXT4_SB(sb)->s_mb_prefetch;
3681 u64 start_time;
3682
3683 if (elr->lr_mode == EXT4_LI_MODE_PREFETCH_BBITMAP) {
3684 elr->lr_next_group = ext4_mb_prefetch(sb, group, nr, &prefetch_ios);
3685 ext4_mb_prefetch_fini(sb, elr->lr_next_group, nr);
3686 trace_ext4_prefetch_bitmaps(sb, group, elr->lr_next_group, nr);
3687 if (group >= elr->lr_next_group) {
3688 ret = 1;
3689 if (elr->lr_first_not_zeroed != ngroups &&
3690 !ext4_emergency_state(sb) && !sb_rdonly(sb) &&
3691 test_opt(sb, INIT_INODE_TABLE)) {
3692 elr->lr_next_group = elr->lr_first_not_zeroed;
3693 elr->lr_mode = EXT4_LI_MODE_ITABLE;
3694 ret = 0;
3695 }
3696 }
3697 return ret;
3698 }
3699
3700 for (; group < ngroups; group++) {
3701 gdp = ext4_get_group_desc(sb, group, NULL);
3702 if (!gdp) {
3703 ret = 1;
3704 break;
3705 }
3706
3707 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3708 break;
3709 }
3710
3711 if (group >= ngroups)
3712 ret = 1;
3713
3714 if (!ret) {
3715 start_time = ktime_get_ns();
3716 ret = ext4_init_inode_table(sb, group,
3717 elr->lr_timeout ? 0 : 1);
3718 trace_ext4_lazy_itable_init(sb, group);
3719 if (elr->lr_timeout == 0) {
3720 elr->lr_timeout = nsecs_to_jiffies((ktime_get_ns() - start_time) *
3721 EXT4_SB(elr->lr_super)->s_li_wait_mult);
3722 }
3723 elr->lr_next_sched = jiffies + elr->lr_timeout;
3724 elr->lr_next_group = group + 1;
3725 }
3726 return ret;
3727 }
3728
3729 /*
3730 * Remove lr_request from the list_request and free the
3731 * request structure. Should be called with li_list_mtx held
3732 */
3733 static void ext4_remove_li_request(struct ext4_li_request *elr)
3734 {
3735 if (!elr)
3736 return;
3737
3738 list_del(&elr->lr_request);
3739 EXT4_SB(elr->lr_super)->s_li_request = NULL;
3740 kfree(elr);
3741 }
3742
3743 static void ext4_unregister_li_request(struct super_block *sb)
3744 {
3745 mutex_lock(&ext4_li_mtx);
3746 if (!ext4_li_info) {
3747 mutex_unlock(&ext4_li_mtx);
3748 return;
3749 }
3750
3751 mutex_lock(&ext4_li_info->li_list_mtx);
3752 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
3753 mutex_unlock(&ext4_li_info->li_list_mtx);
3754 mutex_unlock(&ext4_li_mtx);
3755 }
3756
3757 static struct task_struct *ext4_lazyinit_task;
3758
3759 /*
3760 * This is the function where ext4lazyinit thread lives. It walks
3761 * through the request list searching for next scheduled filesystem.
3762 * When such a fs is found, run the lazy initialization request
3763 * (ext4_rn_li_request) and keep track of the time spend in this
3764 * function. Based on that time we compute next schedule time of
3765 * the request. When walking through the list is complete, compute
3766 * next waking time and put itself into sleep.
3767 */
3768 static int ext4_lazyinit_thread(void *arg)
3769 {
3770 struct ext4_lazy_init *eli = arg;
3771 struct list_head *pos, *n;
3772 struct ext4_li_request *elr;
3773 unsigned long next_wakeup, cur;
3774
3775 BUG_ON(NULL == eli);
3776 set_freezable();
3777
3778 cont_thread:
3779 while (true) {
3780 bool next_wakeup_initialized = false;
3781
3782 next_wakeup = 0;
3783 mutex_lock(&eli->li_list_mtx);
3784 if (list_empty(&eli->li_request_list)) {
3785 mutex_unlock(&eli->li_list_mtx);
3786 goto exit_thread;
3787 }
3788 list_for_each_safe(pos, n, &eli->li_request_list) {
3789 int err = 0;
3790 int progress = 0;
3791 elr = list_entry(pos, struct ext4_li_request,
3792 lr_request);
3793
3794 if (time_before(jiffies, elr->lr_next_sched)) {
3795 if (!next_wakeup_initialized ||
3796 time_before(elr->lr_next_sched, next_wakeup)) {
3797 next_wakeup = elr->lr_next_sched;
3798 next_wakeup_initialized = true;
3799 }
3800 continue;
3801 }
3802 if (down_read_trylock(&elr->lr_super->s_umount)) {
3803 if (sb_start_write_trylock(elr->lr_super)) {
3804 progress = 1;
3805 /*
3806 * We hold sb->s_umount, sb can not
3807 * be removed from the list, it is
3808 * now safe to drop li_list_mtx
3809 */
3810 mutex_unlock(&eli->li_list_mtx);
3811 err = ext4_run_li_request(elr);
3812 sb_end_write(elr->lr_super);
3813 mutex_lock(&eli->li_list_mtx);
3814 n = pos->next;
3815 }
3816 up_read((&elr->lr_super->s_umount));
3817 }
3818 /* error, remove the lazy_init job */
3819 if (err) {
3820 ext4_remove_li_request(elr);
3821 continue;
3822 }
3823 if (!progress) {
3824 elr->lr_next_sched = jiffies +
3825 get_random_u32_below(EXT4_DEF_LI_MAX_START_DELAY * HZ);
3826 }
3827 if (!next_wakeup_initialized ||
3828 time_before(elr->lr_next_sched, next_wakeup)) {
3829 next_wakeup = elr->lr_next_sched;
3830 next_wakeup_initialized = true;
3831 }
3832 }
3833 mutex_unlock(&eli->li_list_mtx);
3834
3835 try_to_freeze();
3836
3837 cur = jiffies;
3838 if (!next_wakeup_initialized || time_after_eq(cur, next_wakeup)) {
3839 cond_resched();
3840 continue;
3841 }
3842
3843 schedule_timeout_interruptible(next_wakeup - cur);
3844
3845 if (kthread_should_stop()) {
3846 ext4_clear_request_list();
3847 goto exit_thread;
3848 }
3849 }
3850
3851 exit_thread:
3852 /*
3853 * It looks like the request list is empty, but we need
3854 * to check it under the li_list_mtx lock, to prevent any
3855 * additions into it, and of course we should lock ext4_li_mtx
3856 * to atomically free the list and ext4_li_info, because at
3857 * this point another ext4 filesystem could be registering
3858 * new one.
3859 */
3860 mutex_lock(&ext4_li_mtx);
3861 mutex_lock(&eli->li_list_mtx);
3862 if (!list_empty(&eli->li_request_list)) {
3863 mutex_unlock(&eli->li_list_mtx);
3864 mutex_unlock(&ext4_li_mtx);
3865 goto cont_thread;
3866 }
3867 mutex_unlock(&eli->li_list_mtx);
3868 kfree(ext4_li_info);
3869 ext4_li_info = NULL;
3870 mutex_unlock(&ext4_li_mtx);
3871
3872 return 0;
3873 }
3874
3875 static void ext4_clear_request_list(void)
3876 {
3877 struct list_head *pos, *n;
3878 struct ext4_li_request *elr;
3879
3880 mutex_lock(&ext4_li_info->li_list_mtx);
3881 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3882 elr = list_entry(pos, struct ext4_li_request,
3883 lr_request);
3884 ext4_remove_li_request(elr);
3885 }
3886 mutex_unlock(&ext4_li_info->li_list_mtx);
3887 }
3888
3889 static int ext4_run_lazyinit_thread(void)
3890 {
3891 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3892 ext4_li_info, "ext4lazyinit");
3893 if (IS_ERR(ext4_lazyinit_task)) {
3894 int err = PTR_ERR(ext4_lazyinit_task);
3895 ext4_clear_request_list();
3896 kfree(ext4_li_info);
3897 ext4_li_info = NULL;
3898 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
3899 "initialization thread\n",
3900 err);
3901 return err;
3902 }
3903 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
3904 return 0;
3905 }
3906
3907 /*
3908 * Check whether it make sense to run itable init. thread or not.
3909 * If there is at least one uninitialized inode table, return
3910 * corresponding group number, else the loop goes through all
3911 * groups and return total number of groups.
3912 */
3913 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3914 {
3915 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3916 struct ext4_group_desc *gdp = NULL;
3917
3918 if (!ext4_has_group_desc_csum(sb))
3919 return ngroups;
3920
3921 for (group = 0; group < ngroups; group++) {
3922 gdp = ext4_get_group_desc(sb, group, NULL);
3923 if (!gdp)
3924 continue;
3925
3926 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3927 break;
3928 }
3929
3930 return group;
3931 }
3932
3933 static int ext4_li_info_new(void)
3934 {
3935 struct ext4_lazy_init *eli = NULL;
3936
3937 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3938 if (!eli)
3939 return -ENOMEM;
3940
3941 INIT_LIST_HEAD(&eli->li_request_list);
3942 mutex_init(&eli->li_list_mtx);
3943
3944 eli->li_state |= EXT4_LAZYINIT_QUIT;
3945
3946 ext4_li_info = eli;
3947
3948 return 0;
3949 }
3950
3951 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3952 ext4_group_t start)
3953 {
3954 struct ext4_li_request *elr;
3955
3956 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3957 if (!elr)
3958 return NULL;
3959
3960 elr->lr_super = sb;
3961 elr->lr_first_not_zeroed = start;
3962 if (test_opt(sb, NO_PREFETCH_BLOCK_BITMAPS)) {
3963 elr->lr_mode = EXT4_LI_MODE_ITABLE;
3964 elr->lr_next_group = start;
3965 } else {
3966 elr->lr_mode = EXT4_LI_MODE_PREFETCH_BBITMAP;
3967 }
3968
3969 /*
3970 * Randomize first schedule time of the request to
3971 * spread the inode table initialization requests
3972 * better.
3973 */
3974 elr->lr_next_sched = jiffies + get_random_u32_below(EXT4_DEF_LI_MAX_START_DELAY * HZ);
3975 return elr;
3976 }
3977
3978 int ext4_register_li_request(struct super_block *sb,
3979 ext4_group_t first_not_zeroed)
3980 {
3981 struct ext4_sb_info *sbi = EXT4_SB(sb);
3982 struct ext4_li_request *elr = NULL;
3983 ext4_group_t ngroups = sbi->s_groups_count;
3984 int ret = 0;
3985
3986 mutex_lock(&ext4_li_mtx);
3987 if (sbi->s_li_request != NULL) {
3988 /*
3989 * Reset timeout so it can be computed again, because
3990 * s_li_wait_mult might have changed.
3991 */
3992 sbi->s_li_request->lr_timeout = 0;
3993 goto out;
3994 }
3995
3996 if (ext4_emergency_state(sb) || sb_rdonly(sb) ||
3997 (test_opt(sb, NO_PREFETCH_BLOCK_BITMAPS) &&
3998 (first_not_zeroed == ngroups || !test_opt(sb, INIT_INODE_TABLE))))
3999 goto out;
4000
4001 elr = ext4_li_request_new(sb, first_not_zeroed);
4002 if (!elr) {
4003 ret = -ENOMEM;
4004 goto out;
4005 }
4006
4007 if (NULL == ext4_li_info) {
4008 ret = ext4_li_info_new();
4009 if (ret)
4010 goto out;
4011 }
4012
4013 mutex_lock(&ext4_li_info->li_list_mtx);
4014 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
4015 mutex_unlock(&ext4_li_info->li_list_mtx);
4016
4017 sbi->s_li_request = elr;
4018 /*
4019 * set elr to NULL here since it has been inserted to
4020 * the request_list and the removal and free of it is
4021 * handled by ext4_clear_request_list from now on.
4022 */
4023 elr = NULL;
4024
4025 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
4026 ret = ext4_run_lazyinit_thread();
4027 if (ret)
4028 goto out;
4029 }
4030 out:
4031 mutex_unlock(&ext4_li_mtx);
4032 if (ret)
4033 kfree(elr);
4034 return ret;
4035 }
4036
4037 /*
4038 * We do not need to lock anything since this is called on
4039 * module unload.
4040 */
4041 static void ext4_destroy_lazyinit_thread(void)
4042 {
4043 /*
4044 * If thread exited earlier
4045 * there's nothing to be done.
4046 */
4047 if (!ext4_li_info || !ext4_lazyinit_task)
4048 return;
4049
4050 kthread_stop(ext4_lazyinit_task);
4051 }
4052
4053 static int set_journal_csum_feature_set(struct super_block *sb)
4054 {
4055 int ret = 1;
4056 int compat, incompat;
4057 struct ext4_sb_info *sbi = EXT4_SB(sb);
4058
4059 if (ext4_has_feature_metadata_csum(sb)) {
4060 /* journal checksum v3 */
4061 compat = 0;
4062 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
4063 } else {
4064 /* journal checksum v1 */
4065 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
4066 incompat = 0;
4067 }
4068
4069 jbd2_journal_clear_features(sbi->s_journal,
4070 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
4071 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
4072 JBD2_FEATURE_INCOMPAT_CSUM_V2);
4073 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4074 ret = jbd2_journal_set_features(sbi->s_journal,
4075 compat, 0,
4076 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
4077 incompat);
4078 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
4079 ret = jbd2_journal_set_features(sbi->s_journal,
4080 compat, 0,
4081 incompat);
4082 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
4083 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
4084 } else {
4085 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
4086 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
4087 }
4088
4089 return ret;
4090 }
4091
4092 /*
4093 * Note: calculating the overhead so we can be compatible with
4094 * historical BSD practice is quite difficult in the face of
4095 * clusters/bigalloc. This is because multiple metadata blocks from
4096 * different block group can end up in the same allocation cluster.
4097 * Calculating the exact overhead in the face of clustered allocation
4098 * requires either O(all block bitmaps) in memory or O(number of block
4099 * groups**2) in time. We will still calculate the superblock for
4100 * older file systems --- and if we come across with a bigalloc file
4101 * system with zero in s_overhead_clusters the estimate will be close to
4102 * correct especially for very large cluster sizes --- but for newer
4103 * file systems, it's better to calculate this figure once at mkfs
4104 * time, and store it in the superblock. If the superblock value is
4105 * present (even for non-bigalloc file systems), we will use it.
4106 */
4107 static int count_overhead(struct super_block *sb, ext4_group_t grp,
4108 char *buf)
4109 {
4110 struct ext4_sb_info *sbi = EXT4_SB(sb);
4111 struct ext4_group_desc *gdp;
4112 ext4_fsblk_t first_block, last_block, b;
4113 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
4114 int s, j, count = 0;
4115 int has_super = ext4_bg_has_super(sb, grp);
4116
4117 if (!ext4_has_feature_bigalloc(sb))
4118 return (has_super + ext4_bg_num_gdb(sb, grp) +
4119 (has_super ? le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) : 0) +
4120 sbi->s_itb_per_group + 2);
4121
4122 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
4123 (grp * EXT4_BLOCKS_PER_GROUP(sb));
4124 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
4125 for (i = 0; i < ngroups; i++) {
4126 gdp = ext4_get_group_desc(sb, i, NULL);
4127 b = ext4_block_bitmap(sb, gdp);
4128 if (b >= first_block && b <= last_block) {
4129 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
4130 count++;
4131 }
4132 b = ext4_inode_bitmap(sb, gdp);
4133 if (b >= first_block && b <= last_block) {
4134 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
4135 count++;
4136 }
4137 b = ext4_inode_table(sb, gdp);
4138 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
4139 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
4140 int c = EXT4_B2C(sbi, b - first_block);
4141 ext4_set_bit(c, buf);
4142 count++;
4143 }
4144 if (i != grp)
4145 continue;
4146 s = 0;
4147 if (ext4_bg_has_super(sb, grp)) {
4148 ext4_set_bit(s++, buf);
4149 count++;
4150 }
4151 j = ext4_bg_num_gdb(sb, grp);
4152 if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) {
4153 ext4_error(sb, "Invalid number of block group "
4154 "descriptor blocks: %d", j);
4155 j = EXT4_BLOCKS_PER_GROUP(sb) - s;
4156 }
4157 count += j;
4158 for (; j > 0; j--)
4159 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
4160 }
4161 if (!count)
4162 return 0;
4163 return EXT4_CLUSTERS_PER_GROUP(sb) -
4164 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
4165 }
4166
4167 /*
4168 * Compute the overhead and stash it in sbi->s_overhead
4169 */
4170 int ext4_calculate_overhead(struct super_block *sb)
4171 {
4172 struct ext4_sb_info *sbi = EXT4_SB(sb);
4173 struct ext4_super_block *es = sbi->s_es;
4174 struct inode *j_inode;
4175 unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum);
4176 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
4177 ext4_fsblk_t overhead = 0;
4178 char *buf = (char *) get_zeroed_page(GFP_NOFS);
4179
4180 if (!buf)
4181 return -ENOMEM;
4182
4183 /*
4184 * Compute the overhead (FS structures). This is constant
4185 * for a given filesystem unless the number of block groups
4186 * changes so we cache the previous value until it does.
4187 */
4188
4189 /*
4190 * All of the blocks before first_data_block are overhead
4191 */
4192 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
4193
4194 /*
4195 * Add the overhead found in each block group
4196 */
4197 for (i = 0; i < ngroups; i++) {
4198 int blks;
4199
4200 blks = count_overhead(sb, i, buf);
4201 overhead += blks;
4202 if (blks)
4203 memset(buf, 0, PAGE_SIZE);
4204 cond_resched();
4205 }
4206
4207 /*
4208 * Add the internal journal blocks whether the journal has been
4209 * loaded or not
4210 */
4211 if (sbi->s_journal && !sbi->s_journal_bdev_file)
4212 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_total_len);
4213 else if (ext4_has_feature_journal(sb) && !sbi->s_journal && j_inum) {
4214 /* j_inum for internal journal is non-zero */
4215 j_inode = ext4_get_journal_inode(sb, j_inum);
4216 if (!IS_ERR(j_inode)) {
4217 j_blocks = j_inode->i_size >> sb->s_blocksize_bits;
4218 overhead += EXT4_NUM_B2C(sbi, j_blocks);
4219 iput(j_inode);
4220 } else {
4221 ext4_msg(sb, KERN_ERR, "can't get journal size");
4222 }
4223 }
4224 sbi->s_overhead = overhead;
4225 smp_wmb();
4226 free_page((unsigned long) buf);
4227 return 0;
4228 }
4229
4230 static void ext4_set_resv_clusters(struct super_block *sb)
4231 {
4232 ext4_fsblk_t resv_clusters;
4233 struct ext4_sb_info *sbi = EXT4_SB(sb);
4234
4235 /*
4236 * There's no need to reserve anything when we aren't using extents.
4237 * The space estimates are exact, there are no unwritten extents,
4238 * hole punching doesn't need new metadata... This is needed especially
4239 * to keep ext2/3 backward compatibility.
4240 */
4241 if (!ext4_has_feature_extents(sb))
4242 return;
4243 /*
4244 * By default we reserve 2% or 4096 clusters, whichever is smaller.
4245 * This should cover the situations where we can not afford to run
4246 * out of space like for example punch hole, or converting
4247 * unwritten extents in delalloc path. In most cases such
4248 * allocation would require 1, or 2 blocks, higher numbers are
4249 * very rare.
4250 */
4251 resv_clusters = (ext4_blocks_count(sbi->s_es) >>
4252 sbi->s_cluster_bits);
4253
4254 do_div(resv_clusters, 50);
4255 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
4256
4257 atomic64_set(&sbi->s_resv_clusters, resv_clusters);
4258 }
4259
4260 static const char *ext4_quota_mode(struct super_block *sb)
4261 {
4262 #ifdef CONFIG_QUOTA
4263 if (!ext4_quota_capable(sb))
4264 return "none";
4265
4266 if (EXT4_SB(sb)->s_journal && ext4_is_quota_journalled(sb))
4267 return "journalled";
4268 else
4269 return "writeback";
4270 #else
4271 return "disabled";
4272 #endif
4273 }
4274
4275 static void ext4_setup_csum_trigger(struct super_block *sb,
4276 enum ext4_journal_trigger_type type,
4277 void (*trigger)(
4278 struct jbd2_buffer_trigger_type *type,
4279 struct buffer_head *bh,
4280 void *mapped_data,
4281 size_t size))
4282 {
4283 struct ext4_sb_info *sbi = EXT4_SB(sb);
4284
4285 sbi->s_journal_triggers[type].sb = sb;
4286 sbi->s_journal_triggers[type].tr_triggers.t_frozen = trigger;
4287 }
4288
4289 static void ext4_free_sbi(struct ext4_sb_info *sbi)
4290 {
4291 if (!sbi)
4292 return;
4293
4294 kfree(sbi->s_blockgroup_lock);
4295 fs_put_dax(sbi->s_daxdev, NULL);
4296 kfree(sbi);
4297 }
4298
4299 static struct ext4_sb_info *ext4_alloc_sbi(struct super_block *sb)
4300 {
4301 struct ext4_sb_info *sbi;
4302
4303 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
4304 if (!sbi)
4305 return NULL;
4306
4307 sbi->s_daxdev = fs_dax_get_by_bdev(sb->s_bdev, &sbi->s_dax_part_off,
4308 NULL, NULL);
4309
4310 sbi->s_blockgroup_lock =
4311 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
4312
4313 if (!sbi->s_blockgroup_lock)
4314 goto err_out;
4315
4316 sb->s_fs_info = sbi;
4317 sbi->s_sb = sb;
4318 return sbi;
4319 err_out:
4320 fs_put_dax(sbi->s_daxdev, NULL);
4321 kfree(sbi);
4322 return NULL;
4323 }
4324
4325 static void ext4_set_def_opts(struct super_block *sb,
4326 struct ext4_super_block *es)
4327 {
4328 unsigned long def_mount_opts;
4329
4330 /* Set defaults before we parse the mount options */
4331 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
4332 set_opt(sb, INIT_INODE_TABLE);
4333 if (def_mount_opts & EXT4_DEFM_DEBUG)
4334 set_opt(sb, DEBUG);
4335 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
4336 set_opt(sb, GRPID);
4337 if (def_mount_opts & EXT4_DEFM_UID16)
4338 set_opt(sb, NO_UID32);
4339 /* xattr user namespace & acls are now defaulted on */
4340 set_opt(sb, XATTR_USER);
4341 #ifdef CONFIG_EXT4_FS_POSIX_ACL
4342 set_opt(sb, POSIX_ACL);
4343 #endif
4344 if (ext4_has_feature_fast_commit(sb))
4345 set_opt2(sb, JOURNAL_FAST_COMMIT);
4346 /* don't forget to enable journal_csum when metadata_csum is enabled. */
4347 if (ext4_has_feature_metadata_csum(sb))
4348 set_opt(sb, JOURNAL_CHECKSUM);
4349
4350 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
4351 set_opt(sb, JOURNAL_DATA);
4352 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
4353 set_opt(sb, ORDERED_DATA);
4354 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
4355 set_opt(sb, WRITEBACK_DATA);
4356
4357 if (le16_to_cpu(es->s_errors) == EXT4_ERRORS_PANIC)
4358 set_opt(sb, ERRORS_PANIC);
4359 else if (le16_to_cpu(es->s_errors) == EXT4_ERRORS_CONTINUE)
4360 set_opt(sb, ERRORS_CONT);
4361 else
4362 set_opt(sb, ERRORS_RO);
4363 /* block_validity enabled by default; disable with noblock_validity */
4364 set_opt(sb, BLOCK_VALIDITY);
4365 if (def_mount_opts & EXT4_DEFM_DISCARD)
4366 set_opt(sb, DISCARD);
4367
4368 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
4369 set_opt(sb, BARRIER);
4370
4371 /*
4372 * enable delayed allocation by default
4373 * Use -o nodelalloc to turn it off
4374 */
4375 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
4376 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
4377 set_opt(sb, DELALLOC);
4378
4379 if (sb->s_blocksize <= PAGE_SIZE)
4380 set_opt(sb, DIOREAD_NOLOCK);
4381 }
4382
4383 static int ext4_handle_clustersize(struct super_block *sb)
4384 {
4385 struct ext4_sb_info *sbi = EXT4_SB(sb);
4386 struct ext4_super_block *es = sbi->s_es;
4387 int clustersize;
4388
4389 /* Handle clustersize */
4390 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
4391 if (ext4_has_feature_bigalloc(sb)) {
4392 if (clustersize < sb->s_blocksize) {
4393 ext4_msg(sb, KERN_ERR,
4394 "cluster size (%d) smaller than "
4395 "block size (%lu)", clustersize, sb->s_blocksize);
4396 return -EINVAL;
4397 }
4398 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
4399 le32_to_cpu(es->s_log_block_size);
4400 } else {
4401 if (clustersize != sb->s_blocksize) {
4402 ext4_msg(sb, KERN_ERR,
4403 "fragment/cluster size (%d) != "
4404 "block size (%lu)", clustersize, sb->s_blocksize);
4405 return -EINVAL;
4406 }
4407 if (sbi->s_blocks_per_group > sb->s_blocksize * 8) {
4408 ext4_msg(sb, KERN_ERR,
4409 "#blocks per group too big: %lu",
4410 sbi->s_blocks_per_group);
4411 return -EINVAL;
4412 }
4413 sbi->s_cluster_bits = 0;
4414 }
4415 sbi->s_clusters_per_group = le32_to_cpu(es->s_clusters_per_group);
4416 if (sbi->s_clusters_per_group > sb->s_blocksize * 8) {
4417 ext4_msg(sb, KERN_ERR, "#clusters per group too big: %lu",
4418 sbi->s_clusters_per_group);
4419 return -EINVAL;
4420 }
4421 if (sbi->s_blocks_per_group !=
4422 (sbi->s_clusters_per_group * (clustersize / sb->s_blocksize))) {
4423 ext4_msg(sb, KERN_ERR,
4424 "blocks per group (%lu) and clusters per group (%lu) inconsistent",
4425 sbi->s_blocks_per_group, sbi->s_clusters_per_group);
4426 return -EINVAL;
4427 }
4428 sbi->s_cluster_ratio = clustersize / sb->s_blocksize;
4429
4430 /* Do we have standard group size of clustersize * 8 blocks ? */
4431 if (sbi->s_blocks_per_group == clustersize << 3)
4432 set_opt2(sb, STD_GROUP_SIZE);
4433
4434 return 0;
4435 }
4436
4437 /*
4438 * ext4_atomic_write_init: Initializes filesystem min & max atomic write units.
4439 * With non-bigalloc filesystem awu will be based upon filesystem blocksize
4440 * & bdev awu units.
4441 * With bigalloc it will be based upon bigalloc cluster size & bdev awu units.
4442 * @sb: super block
4443 */
4444 static void ext4_atomic_write_init(struct super_block *sb)
4445 {
4446 struct ext4_sb_info *sbi = EXT4_SB(sb);
4447 struct block_device *bdev = sb->s_bdev;
4448 unsigned int clustersize = EXT4_CLUSTER_SIZE(sb);
4449
4450 if (!bdev_can_atomic_write(bdev))
4451 return;
4452
4453 if (!ext4_has_feature_extents(sb))
4454 return;
4455
4456 sbi->s_awu_min = max(sb->s_blocksize,
4457 bdev_atomic_write_unit_min_bytes(bdev));
4458 sbi->s_awu_max = min(clustersize,
4459 bdev_atomic_write_unit_max_bytes(bdev));
4460 if (sbi->s_awu_min && sbi->s_awu_max &&
4461 sbi->s_awu_min <= sbi->s_awu_max) {
4462 ext4_msg(sb, KERN_NOTICE, "Supports (experimental) DIO atomic writes awu_min: %u, awu_max: %u",
4463 sbi->s_awu_min, sbi->s_awu_max);
4464 } else {
4465 sbi->s_awu_min = 0;
4466 sbi->s_awu_max = 0;
4467 }
4468 }
4469
4470 static void ext4_fast_commit_init(struct super_block *sb)
4471 {
4472 struct ext4_sb_info *sbi = EXT4_SB(sb);
4473
4474 /* Initialize fast commit stuff */
4475 atomic_set(&sbi->s_fc_subtid, 0);
4476 INIT_LIST_HEAD(&sbi->s_fc_q[FC_Q_MAIN]);
4477 INIT_LIST_HEAD(&sbi->s_fc_q[FC_Q_STAGING]);
4478 INIT_LIST_HEAD(&sbi->s_fc_dentry_q[FC_Q_MAIN]);
4479 INIT_LIST_HEAD(&sbi->s_fc_dentry_q[FC_Q_STAGING]);
4480 sbi->s_fc_bytes = 0;
4481 ext4_clear_mount_flag(sb, EXT4_MF_FC_INELIGIBLE);
4482 sbi->s_fc_ineligible_tid = 0;
4483 mutex_init(&sbi->s_fc_lock);
4484 memset(&sbi->s_fc_stats, 0, sizeof(sbi->s_fc_stats));
4485 sbi->s_fc_replay_state.fc_regions = NULL;
4486 sbi->s_fc_replay_state.fc_regions_size = 0;
4487 sbi->s_fc_replay_state.fc_regions_used = 0;
4488 sbi->s_fc_replay_state.fc_regions_valid = 0;
4489 sbi->s_fc_replay_state.fc_modified_inodes = NULL;
4490 sbi->s_fc_replay_state.fc_modified_inodes_size = 0;
4491 sbi->s_fc_replay_state.fc_modified_inodes_used = 0;
4492 }
4493
4494 static int ext4_inode_info_init(struct super_block *sb,
4495 struct ext4_super_block *es)
4496 {
4497 struct ext4_sb_info *sbi = EXT4_SB(sb);
4498
4499 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
4500 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
4501 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
4502 } else {
4503 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
4504 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
4505 if (sbi->s_first_ino < EXT4_GOOD_OLD_FIRST_INO) {
4506 ext4_msg(sb, KERN_ERR, "invalid first ino: %u",
4507 sbi->s_first_ino);
4508 return -EINVAL;
4509 }
4510 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
4511 (!is_power_of_2(sbi->s_inode_size)) ||
4512 (sbi->s_inode_size > sb->s_blocksize)) {
4513 ext4_msg(sb, KERN_ERR,
4514 "unsupported inode size: %d",
4515 sbi->s_inode_size);
4516 ext4_msg(sb, KERN_ERR, "blocksize: %lu", sb->s_blocksize);
4517 return -EINVAL;
4518 }
4519 /*
4520 * i_atime_extra is the last extra field available for
4521 * [acm]times in struct ext4_inode. Checking for that
4522 * field should suffice to ensure we have extra space
4523 * for all three.
4524 */
4525 if (sbi->s_inode_size >= offsetof(struct ext4_inode, i_atime_extra) +
4526 sizeof(((struct ext4_inode *)0)->i_atime_extra)) {
4527 sb->s_time_gran = 1;
4528 sb->s_time_max = EXT4_EXTRA_TIMESTAMP_MAX;
4529 } else {
4530 sb->s_time_gran = NSEC_PER_SEC;
4531 sb->s_time_max = EXT4_NON_EXTRA_TIMESTAMP_MAX;
4532 }
4533 sb->s_time_min = EXT4_TIMESTAMP_MIN;
4534 }
4535
4536 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
4537 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4538 EXT4_GOOD_OLD_INODE_SIZE;
4539 if (ext4_has_feature_extra_isize(sb)) {
4540 unsigned v, max = (sbi->s_inode_size -
4541 EXT4_GOOD_OLD_INODE_SIZE);
4542
4543 v = le16_to_cpu(es->s_want_extra_isize);
4544 if (v > max) {
4545 ext4_msg(sb, KERN_ERR,
4546 "bad s_want_extra_isize: %d", v);
4547 return -EINVAL;
4548 }
4549 if (sbi->s_want_extra_isize < v)
4550 sbi->s_want_extra_isize = v;
4551
4552 v = le16_to_cpu(es->s_min_extra_isize);
4553 if (v > max) {
4554 ext4_msg(sb, KERN_ERR,
4555 "bad s_min_extra_isize: %d", v);
4556 return -EINVAL;
4557 }
4558 if (sbi->s_want_extra_isize < v)
4559 sbi->s_want_extra_isize = v;
4560 }
4561 }
4562
4563 return 0;
4564 }
4565
4566 #if IS_ENABLED(CONFIG_UNICODE)
4567 static int ext4_encoding_init(struct super_block *sb, struct ext4_super_block *es)
4568 {
4569 const struct ext4_sb_encodings *encoding_info;
4570 struct unicode_map *encoding;
4571 __u16 encoding_flags = le16_to_cpu(es->s_encoding_flags);
4572
4573 if (!ext4_has_feature_casefold(sb) || sb->s_encoding)
4574 return 0;
4575
4576 encoding_info = ext4_sb_read_encoding(es);
4577 if (!encoding_info) {
4578 ext4_msg(sb, KERN_ERR,
4579 "Encoding requested by superblock is unknown");
4580 return -EINVAL;
4581 }
4582
4583 encoding = utf8_load(encoding_info->version);
4584 if (IS_ERR(encoding)) {
4585 ext4_msg(sb, KERN_ERR,
4586 "can't mount with superblock charset: %s-%u.%u.%u "
4587 "not supported by the kernel. flags: 0x%x.",
4588 encoding_info->name,
4589 unicode_major(encoding_info->version),
4590 unicode_minor(encoding_info->version),
4591 unicode_rev(encoding_info->version),
4592 encoding_flags);
4593 return -EINVAL;
4594 }
4595 ext4_msg(sb, KERN_INFO,"Using encoding defined by superblock: "
4596 "%s-%u.%u.%u with flags 0x%hx", encoding_info->name,
4597 unicode_major(encoding_info->version),
4598 unicode_minor(encoding_info->version),
4599 unicode_rev(encoding_info->version),
4600 encoding_flags);
4601
4602 sb->s_encoding = encoding;
4603 sb->s_encoding_flags = encoding_flags;
4604
4605 return 0;
4606 }
4607 #else
4608 static inline int ext4_encoding_init(struct super_block *sb, struct ext4_super_block *es)
4609 {
4610 return 0;
4611 }
4612 #endif
4613
4614 static int ext4_init_metadata_csum(struct super_block *sb, struct ext4_super_block *es)
4615 {
4616 struct ext4_sb_info *sbi = EXT4_SB(sb);
4617
4618 /* Warn if metadata_csum and gdt_csum are both set. */
4619 if (ext4_has_feature_metadata_csum(sb) &&
4620 ext4_has_feature_gdt_csum(sb))
4621 ext4_warning(sb, "metadata_csum and uninit_bg are "
4622 "redundant flags; please run fsck.");
4623
4624 /* Check for a known checksum algorithm */
4625 if (!ext4_verify_csum_type(sb, es)) {
4626 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
4627 "unknown checksum algorithm.");
4628 return -EINVAL;
4629 }
4630 ext4_setup_csum_trigger(sb, EXT4_JTR_ORPHAN_FILE,
4631 ext4_orphan_file_block_trigger);
4632
4633 /* Check superblock checksum */
4634 if (!ext4_superblock_csum_verify(sb, es)) {
4635 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
4636 "invalid superblock checksum. Run e2fsck?");
4637 return -EFSBADCRC;
4638 }
4639
4640 /* Precompute checksum seed for all metadata */
4641 if (ext4_has_feature_csum_seed(sb))
4642 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
4643 else if (ext4_has_feature_metadata_csum(sb) ||
4644 ext4_has_feature_ea_inode(sb))
4645 sbi->s_csum_seed = ext4_chksum(~0, es->s_uuid,
4646 sizeof(es->s_uuid));
4647 return 0;
4648 }
4649
4650 static int ext4_check_feature_compatibility(struct super_block *sb,
4651 struct ext4_super_block *es,
4652 int silent)
4653 {
4654 struct ext4_sb_info *sbi = EXT4_SB(sb);
4655
4656 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
4657 (ext4_has_compat_features(sb) ||
4658 ext4_has_ro_compat_features(sb) ||
4659 ext4_has_incompat_features(sb)))
4660 ext4_msg(sb, KERN_WARNING,
4661 "feature flags set on rev 0 fs, "
4662 "running e2fsck is recommended");
4663
4664 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
4665 set_opt2(sb, HURD_COMPAT);
4666 if (ext4_has_feature_64bit(sb)) {
4667 ext4_msg(sb, KERN_ERR,
4668 "The Hurd can't support 64-bit file systems");
4669 return -EINVAL;
4670 }
4671
4672 /*
4673 * ea_inode feature uses l_i_version field which is not
4674 * available in HURD_COMPAT mode.
4675 */
4676 if (ext4_has_feature_ea_inode(sb)) {
4677 ext4_msg(sb, KERN_ERR,
4678 "ea_inode feature is not supported for Hurd");
4679 return -EINVAL;
4680 }
4681 }
4682
4683 if (IS_EXT2_SB(sb)) {
4684 if (ext2_feature_set_ok(sb))
4685 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
4686 "using the ext4 subsystem");
4687 else {
4688 /*
4689 * If we're probing be silent, if this looks like
4690 * it's actually an ext[34] filesystem.
4691 */
4692 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
4693 return -EINVAL;
4694 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
4695 "to feature incompatibilities");
4696 return -EINVAL;
4697 }
4698 }
4699
4700 if (IS_EXT3_SB(sb)) {
4701 if (ext3_feature_set_ok(sb))
4702 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
4703 "using the ext4 subsystem");
4704 else {
4705 /*
4706 * If we're probing be silent, if this looks like
4707 * it's actually an ext4 filesystem.
4708 */
4709 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
4710 return -EINVAL;
4711 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
4712 "to feature incompatibilities");
4713 return -EINVAL;
4714 }
4715 }
4716
4717 /*
4718 * Check feature flags regardless of the revision level, since we
4719 * previously didn't change the revision level when setting the flags,
4720 * so there is a chance incompat flags are set on a rev 0 filesystem.
4721 */
4722 if (!ext4_feature_set_ok(sb, (sb_rdonly(sb))))
4723 return -EINVAL;
4724
4725 if (sbi->s_daxdev) {
4726 if (sb->s_blocksize == PAGE_SIZE)
4727 set_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags);
4728 else
4729 ext4_msg(sb, KERN_ERR, "unsupported blocksize for DAX\n");
4730 }
4731
4732 if (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) {
4733 if (ext4_has_feature_inline_data(sb)) {
4734 ext4_msg(sb, KERN_ERR, "Cannot use DAX on a filesystem"
4735 " that may contain inline data");
4736 return -EINVAL;
4737 }
4738 if (!test_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags)) {
4739 ext4_msg(sb, KERN_ERR,
4740 "DAX unsupported by block device.");
4741 return -EINVAL;
4742 }
4743 }
4744
4745 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
4746 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
4747 es->s_encryption_level);
4748 return -EINVAL;
4749 }
4750
4751 return 0;
4752 }
4753
4754 static int ext4_check_geometry(struct super_block *sb,
4755 struct ext4_super_block *es)
4756 {
4757 struct ext4_sb_info *sbi = EXT4_SB(sb);
4758 __u64 blocks_count;
4759 int err;
4760
4761 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (sb->s_blocksize / 4)) {
4762 ext4_msg(sb, KERN_ERR,
4763 "Number of reserved GDT blocks insanely large: %d",
4764 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks));
4765 return -EINVAL;
4766 }
4767 /*
4768 * Test whether we have more sectors than will fit in sector_t,
4769 * and whether the max offset is addressable by the page cache.
4770 */
4771 err = generic_check_addressable(sb->s_blocksize_bits,
4772 ext4_blocks_count(es));
4773 if (err) {
4774 ext4_msg(sb, KERN_ERR, "filesystem"
4775 " too large to mount safely on this system");
4776 return err;
4777 }
4778
4779 /* check blocks count against device size */
4780 blocks_count = sb_bdev_nr_blocks(sb);
4781 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
4782 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
4783 "exceeds size of device (%llu blocks)",
4784 ext4_blocks_count(es), blocks_count);
4785 return -EINVAL;
4786 }
4787
4788 /*
4789 * It makes no sense for the first data block to be beyond the end
4790 * of the filesystem.
4791 */
4792 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
4793 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
4794 "block %u is beyond end of filesystem (%llu)",
4795 le32_to_cpu(es->s_first_data_block),
4796 ext4_blocks_count(es));
4797 return -EINVAL;
4798 }
4799 if ((es->s_first_data_block == 0) && (es->s_log_block_size == 0) &&
4800 (sbi->s_cluster_ratio == 1)) {
4801 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
4802 "block is 0 with a 1k block and cluster size");
4803 return -EINVAL;
4804 }
4805
4806 blocks_count = (ext4_blocks_count(es) -
4807 le32_to_cpu(es->s_first_data_block) +
4808 EXT4_BLOCKS_PER_GROUP(sb) - 1);
4809 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4810 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
4811 ext4_msg(sb, KERN_WARNING, "groups count too large: %llu "
4812 "(block count %llu, first data block %u, "
4813 "blocks per group %lu)", blocks_count,
4814 ext4_blocks_count(es),
4815 le32_to_cpu(es->s_first_data_block),
4816 EXT4_BLOCKS_PER_GROUP(sb));
4817 return -EINVAL;
4818 }
4819 sbi->s_groups_count = blocks_count;
4820 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
4821 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
4822 if (((u64)sbi->s_groups_count * sbi->s_inodes_per_group) !=
4823 le32_to_cpu(es->s_inodes_count)) {
4824 ext4_msg(sb, KERN_ERR, "inodes count not valid: %u vs %llu",
4825 le32_to_cpu(es->s_inodes_count),
4826 ((u64)sbi->s_groups_count * sbi->s_inodes_per_group));
4827 return -EINVAL;
4828 }
4829
4830 return 0;
4831 }
4832
4833 static int ext4_group_desc_init(struct super_block *sb,
4834 struct ext4_super_block *es,
4835 ext4_fsblk_t logical_sb_block,
4836 ext4_group_t *first_not_zeroed)
4837 {
4838 struct ext4_sb_info *sbi = EXT4_SB(sb);
4839 unsigned int db_count;
4840 ext4_fsblk_t block;
4841 int i;
4842
4843 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
4844 EXT4_DESC_PER_BLOCK(sb);
4845 if (ext4_has_feature_meta_bg(sb)) {
4846 if (le32_to_cpu(es->s_first_meta_bg) > db_count) {
4847 ext4_msg(sb, KERN_WARNING,
4848 "first meta block group too large: %u "
4849 "(group descriptor block count %u)",
4850 le32_to_cpu(es->s_first_meta_bg), db_count);
4851 return -EINVAL;
4852 }
4853 }
4854 rcu_assign_pointer(sbi->s_group_desc,
4855 kvmalloc_array(db_count,
4856 sizeof(struct buffer_head *),
4857 GFP_KERNEL));
4858 if (sbi->s_group_desc == NULL) {
4859 ext4_msg(sb, KERN_ERR, "not enough memory");
4860 return -ENOMEM;
4861 }
4862
4863 bgl_lock_init(sbi->s_blockgroup_lock);
4864
4865 /* Pre-read the descriptors into the buffer cache */
4866 for (i = 0; i < db_count; i++) {
4867 block = descriptor_loc(sb, logical_sb_block, i);
4868 ext4_sb_breadahead_unmovable(sb, block);
4869 }
4870
4871 for (i = 0; i < db_count; i++) {
4872 struct buffer_head *bh;
4873
4874 block = descriptor_loc(sb, logical_sb_block, i);
4875 bh = ext4_sb_bread_unmovable(sb, block);
4876 if (IS_ERR(bh)) {
4877 ext4_msg(sb, KERN_ERR,
4878 "can't read group descriptor %d", i);
4879 sbi->s_gdb_count = i;
4880 return PTR_ERR(bh);
4881 }
4882 rcu_read_lock();
4883 rcu_dereference(sbi->s_group_desc)[i] = bh;
4884 rcu_read_unlock();
4885 }
4886 sbi->s_gdb_count = db_count;
4887 if (!ext4_check_descriptors(sb, logical_sb_block, first_not_zeroed)) {
4888 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
4889 return -EFSCORRUPTED;
4890 }
4891
4892 return 0;
4893 }
4894
4895 static int ext4_load_and_init_journal(struct super_block *sb,
4896 struct ext4_super_block *es,
4897 struct ext4_fs_context *ctx)
4898 {
4899 struct ext4_sb_info *sbi = EXT4_SB(sb);
4900 int err;
4901
4902 err = ext4_load_journal(sb, es, ctx->journal_devnum);
4903 if (err)
4904 return err;
4905
4906 if (ext4_has_feature_64bit(sb) &&
4907 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4908 JBD2_FEATURE_INCOMPAT_64BIT)) {
4909 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
4910 goto out;
4911 }
4912
4913 if (!set_journal_csum_feature_set(sb)) {
4914 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
4915 "feature set");
4916 goto out;
4917 }
4918
4919 if (test_opt2(sb, JOURNAL_FAST_COMMIT) &&
4920 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4921 JBD2_FEATURE_INCOMPAT_FAST_COMMIT)) {
4922 ext4_msg(sb, KERN_ERR,
4923 "Failed to set fast commit journal feature");
4924 goto out;
4925 }
4926
4927 /* We have now updated the journal if required, so we can
4928 * validate the data journaling mode. */
4929 switch (test_opt(sb, DATA_FLAGS)) {
4930 case 0:
4931 /* No mode set, assume a default based on the journal
4932 * capabilities: ORDERED_DATA if the journal can
4933 * cope, else JOURNAL_DATA
4934 */
4935 if (jbd2_journal_check_available_features
4936 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4937 set_opt(sb, ORDERED_DATA);
4938 sbi->s_def_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
4939 } else {
4940 set_opt(sb, JOURNAL_DATA);
4941 sbi->s_def_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
4942 }
4943 break;
4944
4945 case EXT4_MOUNT_ORDERED_DATA:
4946 case EXT4_MOUNT_WRITEBACK_DATA:
4947 if (!jbd2_journal_check_available_features
4948 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4949 ext4_msg(sb, KERN_ERR, "Journal does not support "
4950 "requested data journaling mode");
4951 goto out;
4952 }
4953 break;
4954 default:
4955 break;
4956 }
4957
4958 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
4959 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4960 ext4_msg(sb, KERN_ERR, "can't mount with "
4961 "journal_async_commit in data=ordered mode");
4962 goto out;
4963 }
4964
4965 set_task_ioprio(sbi->s_journal->j_task, ctx->journal_ioprio);
4966
4967 sbi->s_journal->j_submit_inode_data_buffers =
4968 ext4_journal_submit_inode_data_buffers;
4969 sbi->s_journal->j_finish_inode_data_buffers =
4970 ext4_journal_finish_inode_data_buffers;
4971
4972 return 0;
4973
4974 out:
4975 ext4_journal_destroy(sbi, sbi->s_journal);
4976 return -EINVAL;
4977 }
4978
4979 static int ext4_check_journal_data_mode(struct super_block *sb)
4980 {
4981 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4982 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting with "
4983 "data=journal disables delayed allocation, "
4984 "dioread_nolock, O_DIRECT and fast_commit support!\n");
4985 /* can't mount with both data=journal and dioread_nolock. */
4986 clear_opt(sb, DIOREAD_NOLOCK);
4987 clear_opt2(sb, JOURNAL_FAST_COMMIT);
4988 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4989 ext4_msg(sb, KERN_ERR, "can't mount with "
4990 "both data=journal and delalloc");
4991 return -EINVAL;
4992 }
4993 if (test_opt(sb, DAX_ALWAYS)) {
4994 ext4_msg(sb, KERN_ERR, "can't mount with "
4995 "both data=journal and dax");
4996 return -EINVAL;
4997 }
4998 if (ext4_has_feature_encrypt(sb)) {
4999 ext4_msg(sb, KERN_WARNING,
5000 "encrypted files will use data=ordered "
5001 "instead of data journaling mode");
5002 }
5003 if (test_opt(sb, DELALLOC))
5004 clear_opt(sb, DELALLOC);
5005 } else {
5006 sb->s_iflags |= SB_I_CGROUPWB;
5007 }
5008
5009 return 0;
5010 }
5011
5012 static const char *ext4_has_journal_option(struct super_block *sb)
5013 {
5014 struct ext4_sb_info *sbi = EXT4_SB(sb);
5015
5016 if (test_opt(sb, JOURNAL_ASYNC_COMMIT))
5017 return "journal_async_commit";
5018 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM))
5019 return "journal_checksum";
5020 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
5021 return "commit=";
5022 if (EXT4_MOUNT_DATA_FLAGS &
5023 (sbi->s_mount_opt ^ sbi->s_def_mount_opt))
5024 return "data=";
5025 if (test_opt(sb, DATA_ERR_ABORT))
5026 return "data_err=abort";
5027 return NULL;
5028 }
5029
5030 static int ext4_load_super(struct super_block *sb, ext4_fsblk_t *lsb,
5031 int silent)
5032 {
5033 struct ext4_sb_info *sbi = EXT4_SB(sb);
5034 struct ext4_super_block *es;
5035 ext4_fsblk_t logical_sb_block;
5036 unsigned long offset = 0;
5037 struct buffer_head *bh;
5038 int ret = -EINVAL;
5039 int blocksize;
5040
5041 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
5042 if (!blocksize) {
5043 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
5044 return -EINVAL;
5045 }
5046
5047 /*
5048 * The ext4 superblock will not be buffer aligned for other than 1kB
5049 * block sizes. We need to calculate the offset from buffer start.
5050 */
5051 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
5052 logical_sb_block = sbi->s_sb_block * EXT4_MIN_BLOCK_SIZE;
5053 offset = do_div(logical_sb_block, blocksize);
5054 } else {
5055 logical_sb_block = sbi->s_sb_block;
5056 }
5057
5058 bh = ext4_sb_bread_unmovable(sb, logical_sb_block);
5059 if (IS_ERR(bh)) {
5060 ext4_msg(sb, KERN_ERR, "unable to read superblock");
5061 return PTR_ERR(bh);
5062 }
5063 /*
5064 * Note: s_es must be initialized as soon as possible because
5065 * some ext4 macro-instructions depend on its value
5066 */
5067 es = (struct ext4_super_block *) (bh->b_data + offset);
5068 sbi->s_es = es;
5069 sb->s_magic = le16_to_cpu(es->s_magic);
5070 if (sb->s_magic != EXT4_SUPER_MAGIC) {
5071 if (!silent)
5072 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
5073 goto out;
5074 }
5075
5076 if (le32_to_cpu(es->s_log_block_size) >
5077 (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
5078 ext4_msg(sb, KERN_ERR,
5079 "Invalid log block size: %u",
5080 le32_to_cpu(es->s_log_block_size));
5081 goto out;
5082 }
5083 if (le32_to_cpu(es->s_log_cluster_size) >
5084 (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
5085 ext4_msg(sb, KERN_ERR,
5086 "Invalid log cluster size: %u",
5087 le32_to_cpu(es->s_log_cluster_size));
5088 goto out;
5089 }
5090
5091 blocksize = EXT4_MIN_BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
5092
5093 /*
5094 * If the default block size is not the same as the real block size,
5095 * we need to reload it.
5096 */
5097 if (sb->s_blocksize == blocksize) {
5098 *lsb = logical_sb_block;
5099 sbi->s_sbh = bh;
5100 return 0;
5101 }
5102
5103 /*
5104 * bh must be released before kill_bdev(), otherwise
5105 * it won't be freed and its page also. kill_bdev()
5106 * is called by sb_set_blocksize().
5107 */
5108 brelse(bh);
5109 /* Validate the filesystem blocksize */
5110 if (!sb_set_blocksize(sb, blocksize)) {
5111 ext4_msg(sb, KERN_ERR, "bad block size %d",
5112 blocksize);
5113 bh = NULL;
5114 goto out;
5115 }
5116
5117 logical_sb_block = sbi->s_sb_block * EXT4_MIN_BLOCK_SIZE;
5118 offset = do_div(logical_sb_block, blocksize);
5119 bh = ext4_sb_bread_unmovable(sb, logical_sb_block);
5120 if (IS_ERR(bh)) {
5121 ext4_msg(sb, KERN_ERR, "Can't read superblock on 2nd try");
5122 ret = PTR_ERR(bh);
5123 bh = NULL;
5124 goto out;
5125 }
5126 es = (struct ext4_super_block *)(bh->b_data + offset);
5127 sbi->s_es = es;
5128 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
5129 ext4_msg(sb, KERN_ERR, "Magic mismatch, very weird!");
5130 goto out;
5131 }
5132 *lsb = logical_sb_block;
5133 sbi->s_sbh = bh;
5134 return 0;
5135 out:
5136 brelse(bh);
5137 return ret;
5138 }
5139
5140 static int ext4_hash_info_init(struct super_block *sb)
5141 {
5142 struct ext4_sb_info *sbi = EXT4_SB(sb);
5143 struct ext4_super_block *es = sbi->s_es;
5144 unsigned int i;
5145
5146 sbi->s_def_hash_version = es->s_def_hash_version;
5147
5148 if (sbi->s_def_hash_version > DX_HASH_LAST) {
5149 ext4_msg(sb, KERN_ERR,
5150 "Invalid default hash set in the superblock");
5151 return -EINVAL;
5152 } else if (sbi->s_def_hash_version == DX_HASH_SIPHASH) {
5153 ext4_msg(sb, KERN_ERR,
5154 "SIPHASH is not a valid default hash value");
5155 return -EINVAL;
5156 }
5157
5158 for (i = 0; i < 4; i++)
5159 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
5160
5161 if (ext4_has_feature_dir_index(sb)) {
5162 i = le32_to_cpu(es->s_flags);
5163 if (i & EXT2_FLAGS_UNSIGNED_HASH)
5164 sbi->s_hash_unsigned = 3;
5165 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
5166 #ifdef __CHAR_UNSIGNED__
5167 if (!sb_rdonly(sb))
5168 es->s_flags |=
5169 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
5170 sbi->s_hash_unsigned = 3;
5171 #else
5172 if (!sb_rdonly(sb))
5173 es->s_flags |=
5174 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
5175 #endif
5176 }
5177 }
5178 return 0;
5179 }
5180
5181 static int ext4_block_group_meta_init(struct super_block *sb, int silent)
5182 {
5183 struct ext4_sb_info *sbi = EXT4_SB(sb);
5184 struct ext4_super_block *es = sbi->s_es;
5185 int has_huge_files;
5186
5187 has_huge_files = ext4_has_feature_huge_file(sb);
5188 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
5189 has_huge_files);
5190 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
5191
5192 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
5193 if (ext4_has_feature_64bit(sb)) {
5194 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
5195 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
5196 !is_power_of_2(sbi->s_desc_size)) {
5197 ext4_msg(sb, KERN_ERR,
5198 "unsupported descriptor size %lu",
5199 sbi->s_desc_size);
5200 return -EINVAL;
5201 }
5202 } else
5203 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
5204
5205 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
5206 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
5207
5208 sbi->s_inodes_per_block = sb->s_blocksize / EXT4_INODE_SIZE(sb);
5209 if (sbi->s_inodes_per_block == 0 || sbi->s_blocks_per_group == 0) {
5210 if (!silent)
5211 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
5212 return -EINVAL;
5213 }
5214 if (sbi->s_inodes_per_group < sbi->s_inodes_per_block ||
5215 sbi->s_inodes_per_group > sb->s_blocksize * 8) {
5216 ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n",
5217 sbi->s_inodes_per_group);
5218 return -EINVAL;
5219 }
5220 sbi->s_itb_per_group = sbi->s_inodes_per_group /
5221 sbi->s_inodes_per_block;
5222 sbi->s_desc_per_block = sb->s_blocksize / EXT4_DESC_SIZE(sb);
5223 sbi->s_mount_state = le16_to_cpu(es->s_state) & ~EXT4_FC_REPLAY;
5224 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
5225 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
5226
5227 return 0;
5228 }
5229
5230 /*
5231 * It's hard to get stripe aligned blocks if stripe is not aligned with
5232 * cluster, just disable stripe and alert user to simplify code and avoid
5233 * stripe aligned allocation which will rarely succeed.
5234 */
5235 static bool ext4_is_stripe_incompatible(struct super_block *sb, unsigned long stripe)
5236 {
5237 struct ext4_sb_info *sbi = EXT4_SB(sb);
5238 return (stripe > 0 && sbi->s_cluster_ratio > 1 &&
5239 stripe % sbi->s_cluster_ratio != 0);
5240 }
5241
5242 static int __ext4_fill_super(struct fs_context *fc, struct super_block *sb)
5243 {
5244 struct ext4_super_block *es = NULL;
5245 struct ext4_sb_info *sbi = EXT4_SB(sb);
5246 ext4_fsblk_t logical_sb_block;
5247 struct inode *root;
5248 int needs_recovery;
5249 int err;
5250 ext4_group_t first_not_zeroed;
5251 struct ext4_fs_context *ctx = fc->fs_private;
5252 int silent = fc->sb_flags & SB_SILENT;
5253
5254 /* Set defaults for the variables that will be set during parsing */
5255 if (!(ctx->spec & EXT4_SPEC_JOURNAL_IOPRIO))
5256 ctx->journal_ioprio = EXT4_DEF_JOURNAL_IOPRIO;
5257
5258 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
5259 sbi->s_sectors_written_start =
5260 part_stat_read(sb->s_bdev, sectors[STAT_WRITE]);
5261
5262 err = ext4_load_super(sb, &logical_sb_block, silent);
5263 if (err)
5264 goto out_fail;
5265
5266 es = sbi->s_es;
5267 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
5268
5269 err = ext4_init_metadata_csum(sb, es);
5270 if (err)
5271 goto failed_mount;
5272
5273 ext4_set_def_opts(sb, es);
5274
5275 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
5276 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
5277 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
5278 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
5279 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
5280 sbi->s_sb_update_kb = EXT4_DEF_SB_UPDATE_INTERVAL_KB;
5281 sbi->s_sb_update_sec = EXT4_DEF_SB_UPDATE_INTERVAL_SEC;
5282
5283 /*
5284 * set default s_li_wait_mult for lazyinit, for the case there is
5285 * no mount option specified.
5286 */
5287 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
5288
5289 err = ext4_inode_info_init(sb, es);
5290 if (err)
5291 goto failed_mount;
5292
5293 err = parse_apply_sb_mount_options(sb, ctx);
5294 if (err < 0)
5295 goto failed_mount;
5296
5297 sbi->s_def_mount_opt = sbi->s_mount_opt;
5298 sbi->s_def_mount_opt2 = sbi->s_mount_opt2;
5299
5300 err = ext4_check_opt_consistency(fc, sb);
5301 if (err < 0)
5302 goto failed_mount;
5303
5304 ext4_apply_options(fc, sb);
5305
5306 err = ext4_encoding_init(sb, es);
5307 if (err)
5308 goto failed_mount;
5309
5310 err = ext4_check_journal_data_mode(sb);
5311 if (err)
5312 goto failed_mount;
5313
5314 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
5315 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
5316
5317 /* i_version is always enabled now */
5318 sb->s_flags |= SB_I_VERSION;
5319
5320 /* HSM events are allowed by default. */
5321 sb->s_iflags |= SB_I_ALLOW_HSM;
5322
5323 err = ext4_check_feature_compatibility(sb, es, silent);
5324 if (err)
5325 goto failed_mount;
5326
5327 err = ext4_block_group_meta_init(sb, silent);
5328 if (err)
5329 goto failed_mount;
5330
5331 err = ext4_hash_info_init(sb);
5332 if (err)
5333 goto failed_mount;
5334
5335 err = ext4_handle_clustersize(sb);
5336 if (err)
5337 goto failed_mount;
5338
5339 err = ext4_check_geometry(sb, es);
5340 if (err)
5341 goto failed_mount;
5342
5343 timer_setup(&sbi->s_err_report, print_daily_error_info, 0);
5344 spin_lock_init(&sbi->s_error_lock);
5345 INIT_WORK(&sbi->s_sb_upd_work, update_super_work);
5346
5347 err = ext4_group_desc_init(sb, es, logical_sb_block, &first_not_zeroed);
5348 if (err)
5349 goto failed_mount3;
5350
5351 err = ext4_es_register_shrinker(sbi);
5352 if (err)
5353 goto failed_mount3;
5354
5355 sbi->s_stripe = ext4_get_stripe_size(sbi);
5356 if (ext4_is_stripe_incompatible(sb, sbi->s_stripe)) {
5357 ext4_msg(sb, KERN_WARNING,
5358 "stripe (%lu) is not aligned with cluster size (%u), "
5359 "stripe is disabled",
5360 sbi->s_stripe, sbi->s_cluster_ratio);
5361 sbi->s_stripe = 0;
5362 }
5363 sbi->s_extent_max_zeroout_kb = 32;
5364
5365 /*
5366 * set up enough so that it can read an inode
5367 */
5368 sb->s_op = &ext4_sops;
5369 sb->s_export_op = &ext4_export_ops;
5370 sb->s_xattr = ext4_xattr_handlers;
5371 #ifdef CONFIG_FS_ENCRYPTION
5372 sb->s_cop = &ext4_cryptops;
5373 #endif
5374 #ifdef CONFIG_FS_VERITY
5375 sb->s_vop = &ext4_verityops;
5376 #endif
5377 #ifdef CONFIG_QUOTA
5378 sb->dq_op = &ext4_quota_operations;
5379 if (ext4_has_feature_quota(sb))
5380 sb->s_qcop = &dquot_quotactl_sysfile_ops;
5381 else
5382 sb->s_qcop = &ext4_qctl_operations;
5383 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
5384 #endif
5385 super_set_uuid(sb, es->s_uuid, sizeof(es->s_uuid));
5386 super_set_sysfs_name_bdev(sb);
5387
5388 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
5389 mutex_init(&sbi->s_orphan_lock);
5390
5391 spin_lock_init(&sbi->s_bdev_wb_lock);
5392
5393 ext4_atomic_write_init(sb);
5394 ext4_fast_commit_init(sb);
5395
5396 sb->s_root = NULL;
5397
5398 needs_recovery = (es->s_last_orphan != 0 ||
5399 ext4_has_feature_orphan_present(sb) ||
5400 ext4_has_feature_journal_needs_recovery(sb));
5401
5402 if (ext4_has_feature_mmp(sb) && !sb_rdonly(sb)) {
5403 err = ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block));
5404 if (err)
5405 goto failed_mount3a;
5406 }
5407
5408 err = -EINVAL;
5409 /*
5410 * The first inode we look at is the journal inode. Don't try
5411 * root first: it may be modified in the journal!
5412 */
5413 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
5414 err = ext4_load_and_init_journal(sb, es, ctx);
5415 if (err)
5416 goto failed_mount3a;
5417 } else if (test_opt(sb, NOLOAD) && !sb_rdonly(sb) &&
5418 ext4_has_feature_journal_needs_recovery(sb)) {
5419 ext4_msg(sb, KERN_ERR, "required journal recovery "
5420 "suppressed and not mounted read-only");
5421 goto failed_mount3a;
5422 } else {
5423 const char *journal_option;
5424
5425 /* Nojournal mode, all journal mount options are illegal */
5426 journal_option = ext4_has_journal_option(sb);
5427 if (journal_option != NULL) {
5428 ext4_msg(sb, KERN_ERR,
5429 "can't mount with %s, fs mounted w/o journal",
5430 journal_option);
5431 goto failed_mount3a;
5432 }
5433
5434 sbi->s_def_mount_opt &= ~EXT4_MOUNT_JOURNAL_CHECKSUM;
5435 clear_opt(sb, JOURNAL_CHECKSUM);
5436 clear_opt(sb, DATA_FLAGS);
5437 clear_opt2(sb, JOURNAL_FAST_COMMIT);
5438 sbi->s_journal = NULL;
5439 needs_recovery = 0;
5440 }
5441
5442 if (!test_opt(sb, NO_MBCACHE)) {
5443 sbi->s_ea_block_cache = ext4_xattr_create_cache();
5444 if (!sbi->s_ea_block_cache) {
5445 ext4_msg(sb, KERN_ERR,
5446 "Failed to create ea_block_cache");
5447 err = -EINVAL;
5448 goto failed_mount_wq;
5449 }
5450
5451 if (ext4_has_feature_ea_inode(sb)) {
5452 sbi->s_ea_inode_cache = ext4_xattr_create_cache();
5453 if (!sbi->s_ea_inode_cache) {
5454 ext4_msg(sb, KERN_ERR,
5455 "Failed to create ea_inode_cache");
5456 err = -EINVAL;
5457 goto failed_mount_wq;
5458 }
5459 }
5460 }
5461
5462 /*
5463 * Get the # of file system overhead blocks from the
5464 * superblock if present.
5465 */
5466 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
5467 /* ignore the precalculated value if it is ridiculous */
5468 if (sbi->s_overhead > ext4_blocks_count(es))
5469 sbi->s_overhead = 0;
5470 /*
5471 * If the bigalloc feature is not enabled recalculating the
5472 * overhead doesn't take long, so we might as well just redo
5473 * it to make sure we are using the correct value.
5474 */
5475 if (!ext4_has_feature_bigalloc(sb))
5476 sbi->s_overhead = 0;
5477 if (sbi->s_overhead == 0) {
5478 err = ext4_calculate_overhead(sb);
5479 if (err)
5480 goto failed_mount_wq;
5481 }
5482
5483 /*
5484 * The maximum number of concurrent works can be high and
5485 * concurrency isn't really necessary. Limit it to 1.
5486 */
5487 EXT4_SB(sb)->rsv_conversion_wq =
5488 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
5489 if (!EXT4_SB(sb)->rsv_conversion_wq) {
5490 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
5491 err = -ENOMEM;
5492 goto failed_mount4;
5493 }
5494
5495 /*
5496 * The jbd2_journal_load will have done any necessary log recovery,
5497 * so we can safely mount the rest of the filesystem now.
5498 */
5499
5500 root = ext4_iget(sb, EXT4_ROOT_INO, EXT4_IGET_SPECIAL);
5501 if (IS_ERR(root)) {
5502 ext4_msg(sb, KERN_ERR, "get root inode failed");
5503 err = PTR_ERR(root);
5504 root = NULL;
5505 goto failed_mount4;
5506 }
5507 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
5508 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
5509 iput(root);
5510 err = -EFSCORRUPTED;
5511 goto failed_mount4;
5512 }
5513
5514 generic_set_sb_d_ops(sb);
5515 sb->s_root = d_make_root(root);
5516 if (!sb->s_root) {
5517 ext4_msg(sb, KERN_ERR, "get root dentry failed");
5518 err = -ENOMEM;
5519 goto failed_mount4;
5520 }
5521
5522 err = ext4_setup_super(sb, es, sb_rdonly(sb));
5523 if (err == -EROFS) {
5524 sb->s_flags |= SB_RDONLY;
5525 } else if (err)
5526 goto failed_mount4a;
5527
5528 ext4_set_resv_clusters(sb);
5529
5530 if (test_opt(sb, BLOCK_VALIDITY)) {
5531 err = ext4_setup_system_zone(sb);
5532 if (err) {
5533 ext4_msg(sb, KERN_ERR, "failed to initialize system "
5534 "zone (%d)", err);
5535 goto failed_mount4a;
5536 }
5537 }
5538 ext4_fc_replay_cleanup(sb);
5539
5540 ext4_ext_init(sb);
5541
5542 /*
5543 * Enable optimize_scan if number of groups is > threshold. This can be
5544 * turned off by passing "mb_optimize_scan=0". This can also be
5545 * turned on forcefully by passing "mb_optimize_scan=1".
5546 */
5547 if (!(ctx->spec & EXT4_SPEC_mb_optimize_scan)) {
5548 if (sbi->s_groups_count >= MB_DEFAULT_LINEAR_SCAN_THRESHOLD)
5549 set_opt2(sb, MB_OPTIMIZE_SCAN);
5550 else
5551 clear_opt2(sb, MB_OPTIMIZE_SCAN);
5552 }
5553
5554 err = ext4_mb_init(sb);
5555 if (err) {
5556 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
5557 err);
5558 goto failed_mount5;
5559 }
5560
5561 /*
5562 * We can only set up the journal commit callback once
5563 * mballoc is initialized
5564 */
5565 if (sbi->s_journal)
5566 sbi->s_journal->j_commit_callback =
5567 ext4_journal_commit_callback;
5568
5569 err = ext4_percpu_param_init(sbi);
5570 if (err)
5571 goto failed_mount6;
5572
5573 if (ext4_has_feature_flex_bg(sb))
5574 if (!ext4_fill_flex_info(sb)) {
5575 ext4_msg(sb, KERN_ERR,
5576 "unable to initialize "
5577 "flex_bg meta info!");
5578 err = -ENOMEM;
5579 goto failed_mount6;
5580 }
5581
5582 err = ext4_register_li_request(sb, first_not_zeroed);
5583 if (err)
5584 goto failed_mount6;
5585
5586 err = ext4_init_orphan_info(sb);
5587 if (err)
5588 goto failed_mount7;
5589 #ifdef CONFIG_QUOTA
5590 /* Enable quota usage during mount. */
5591 if (ext4_has_feature_quota(sb) && !sb_rdonly(sb)) {
5592 err = ext4_enable_quotas(sb);
5593 if (err)
5594 goto failed_mount8;
5595 }
5596 #endif /* CONFIG_QUOTA */
5597
5598 /*
5599 * Save the original bdev mapping's wb_err value which could be
5600 * used to detect the metadata async write error.
5601 */
5602 errseq_check_and_advance(&sb->s_bdev->bd_mapping->wb_err,
5603 &sbi->s_bdev_wb_err);
5604 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
5605 ext4_orphan_cleanup(sb, es);
5606 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
5607 /*
5608 * Update the checksum after updating free space/inode counters and
5609 * ext4_orphan_cleanup. Otherwise the superblock can have an incorrect
5610 * checksum in the buffer cache until it is written out and
5611 * e2fsprogs programs trying to open a file system immediately
5612 * after it is mounted can fail.
5613 */
5614 ext4_superblock_csum_set(sb);
5615 if (needs_recovery) {
5616 ext4_msg(sb, KERN_INFO, "recovery complete");
5617 err = ext4_mark_recovery_complete(sb, es);
5618 if (err)
5619 goto failed_mount9;
5620 }
5621
5622 if (test_opt(sb, DISCARD) && !bdev_max_discard_sectors(sb->s_bdev)) {
5623 ext4_msg(sb, KERN_WARNING,
5624 "mounting with \"discard\" option, but the device does not support discard");
5625 clear_opt(sb, DISCARD);
5626 }
5627
5628 if (es->s_error_count)
5629 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
5630
5631 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
5632 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
5633 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
5634 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
5635 atomic_set(&sbi->s_warning_count, 0);
5636 atomic_set(&sbi->s_msg_count, 0);
5637
5638 /* Register sysfs after all initializations are complete. */
5639 err = ext4_register_sysfs(sb);
5640 if (err)
5641 goto failed_mount9;
5642
5643 return 0;
5644
5645 failed_mount9:
5646 ext4_quotas_off(sb, EXT4_MAXQUOTAS);
5647 failed_mount8: __maybe_unused
5648 ext4_release_orphan_info(sb);
5649 failed_mount7:
5650 ext4_unregister_li_request(sb);
5651 failed_mount6:
5652 ext4_mb_release(sb);
5653 ext4_flex_groups_free(sbi);
5654 ext4_percpu_param_destroy(sbi);
5655 failed_mount5:
5656 ext4_ext_release(sb);
5657 ext4_release_system_zone(sb);
5658 failed_mount4a:
5659 dput(sb->s_root);
5660 sb->s_root = NULL;
5661 failed_mount4:
5662 ext4_msg(sb, KERN_ERR, "mount failed");
5663 if (EXT4_SB(sb)->rsv_conversion_wq)
5664 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
5665 failed_mount_wq:
5666 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
5667 sbi->s_ea_inode_cache = NULL;
5668
5669 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
5670 sbi->s_ea_block_cache = NULL;
5671
5672 if (sbi->s_journal) {
5673 ext4_journal_destroy(sbi, sbi->s_journal);
5674 }
5675 failed_mount3a:
5676 ext4_es_unregister_shrinker(sbi);
5677 failed_mount3:
5678 /* flush s_sb_upd_work before sbi destroy */
5679 flush_work(&sbi->s_sb_upd_work);
5680 ext4_stop_mmpd(sbi);
5681 timer_delete_sync(&sbi->s_err_report);
5682 ext4_group_desc_free(sbi);
5683 failed_mount:
5684 #if IS_ENABLED(CONFIG_UNICODE)
5685 utf8_unload(sb->s_encoding);
5686 #endif
5687
5688 #ifdef CONFIG_QUOTA
5689 for (unsigned int i = 0; i < EXT4_MAXQUOTAS; i++)
5690 kfree(get_qf_name(sb, sbi, i));
5691 #endif
5692 fscrypt_free_dummy_policy(&sbi->s_dummy_enc_policy);
5693 brelse(sbi->s_sbh);
5694 if (sbi->s_journal_bdev_file) {
5695 invalidate_bdev(file_bdev(sbi->s_journal_bdev_file));
5696 bdev_fput(sbi->s_journal_bdev_file);
5697 }
5698 out_fail:
5699 invalidate_bdev(sb->s_bdev);
5700 sb->s_fs_info = NULL;
5701 return err;
5702 }
5703
5704 static int ext4_fill_super(struct super_block *sb, struct fs_context *fc)
5705 {
5706 struct ext4_fs_context *ctx = fc->fs_private;
5707 struct ext4_sb_info *sbi;
5708 const char *descr;
5709 int ret;
5710
5711 sbi = ext4_alloc_sbi(sb);
5712 if (!sbi)
5713 return -ENOMEM;
5714
5715 fc->s_fs_info = sbi;
5716
5717 /* Cleanup superblock name */
5718 strreplace(sb->s_id, '/', '!');
5719
5720 sbi->s_sb_block = 1; /* Default super block location */
5721 if (ctx->spec & EXT4_SPEC_s_sb_block)
5722 sbi->s_sb_block = ctx->s_sb_block;
5723
5724 ret = __ext4_fill_super(fc, sb);
5725 if (ret < 0)
5726 goto free_sbi;
5727
5728 if (sbi->s_journal) {
5729 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
5730 descr = " journalled data mode";
5731 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
5732 descr = " ordered data mode";
5733 else
5734 descr = " writeback data mode";
5735 } else
5736 descr = "out journal";
5737
5738 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
5739 ext4_msg(sb, KERN_INFO, "mounted filesystem %pU %s with%s. "
5740 "Quota mode: %s.", &sb->s_uuid,
5741 sb_rdonly(sb) ? "ro" : "r/w", descr,
5742 ext4_quota_mode(sb));
5743
5744 /* Update the s_overhead_clusters if necessary */
5745 ext4_update_overhead(sb, false);
5746 return 0;
5747
5748 free_sbi:
5749 ext4_free_sbi(sbi);
5750 fc->s_fs_info = NULL;
5751 return ret;
5752 }
5753
5754 static int ext4_get_tree(struct fs_context *fc)
5755 {
5756 return get_tree_bdev(fc, ext4_fill_super);
5757 }
5758
5759 /*
5760 * Setup any per-fs journal parameters now. We'll do this both on
5761 * initial mount, once the journal has been initialised but before we've
5762 * done any recovery; and again on any subsequent remount.
5763 */
5764 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
5765 {
5766 struct ext4_sb_info *sbi = EXT4_SB(sb);
5767
5768 journal->j_commit_interval = sbi->s_commit_interval;
5769 journal->j_min_batch_time = sbi->s_min_batch_time;
5770 journal->j_max_batch_time = sbi->s_max_batch_time;
5771 ext4_fc_init(sb, journal);
5772
5773 write_lock(&journal->j_state_lock);
5774 if (test_opt(sb, BARRIER))
5775 journal->j_flags |= JBD2_BARRIER;
5776 else
5777 journal->j_flags &= ~JBD2_BARRIER;
5778 /*
5779 * Always enable journal cycle record option, letting the journal
5780 * records log transactions continuously between each mount.
5781 */
5782 journal->j_flags |= JBD2_CYCLE_RECORD;
5783 write_unlock(&journal->j_state_lock);
5784 }
5785
5786 static struct inode *ext4_get_journal_inode(struct super_block *sb,
5787 unsigned int journal_inum)
5788 {
5789 struct inode *journal_inode;
5790
5791 /*
5792 * Test for the existence of a valid inode on disk. Bad things
5793 * happen if we iget() an unused inode, as the subsequent iput()
5794 * will try to delete it.
5795 */
5796 journal_inode = ext4_iget(sb, journal_inum, EXT4_IGET_SPECIAL);
5797 if (IS_ERR(journal_inode)) {
5798 ext4_msg(sb, KERN_ERR, "no journal found");
5799 return ERR_CAST(journal_inode);
5800 }
5801 if (!journal_inode->i_nlink) {
5802 make_bad_inode(journal_inode);
5803 iput(journal_inode);
5804 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
5805 return ERR_PTR(-EFSCORRUPTED);
5806 }
5807 if (!S_ISREG(journal_inode->i_mode) || IS_ENCRYPTED(journal_inode)) {
5808 ext4_msg(sb, KERN_ERR, "invalid journal inode");
5809 iput(journal_inode);
5810 return ERR_PTR(-EFSCORRUPTED);
5811 }
5812
5813 ext4_debug("Journal inode found at %p: %lld bytes\n",
5814 journal_inode, journal_inode->i_size);
5815 return journal_inode;
5816 }
5817
5818 static int ext4_journal_bmap(journal_t *journal, sector_t *block)
5819 {
5820 struct ext4_map_blocks map;
5821 int ret;
5822
5823 if (journal->j_inode == NULL)
5824 return 0;
5825
5826 map.m_lblk = *block;
5827 map.m_len = 1;
5828 ret = ext4_map_blocks(NULL, journal->j_inode, &map, 0);
5829 if (ret <= 0) {
5830 ext4_msg(journal->j_inode->i_sb, KERN_CRIT,
5831 "journal bmap failed: block %llu ret %d\n",
5832 *block, ret);
5833 jbd2_journal_abort(journal, ret ? ret : -EIO);
5834 return ret;
5835 }
5836 *block = map.m_pblk;
5837 return 0;
5838 }
5839
5840 static journal_t *ext4_open_inode_journal(struct super_block *sb,
5841 unsigned int journal_inum)
5842 {
5843 struct inode *journal_inode;
5844 journal_t *journal;
5845
5846 journal_inode = ext4_get_journal_inode(sb, journal_inum);
5847 if (IS_ERR(journal_inode))
5848 return ERR_CAST(journal_inode);
5849
5850 journal = jbd2_journal_init_inode(journal_inode);
5851 if (IS_ERR(journal)) {
5852 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
5853 iput(journal_inode);
5854 return ERR_CAST(journal);
5855 }
5856 journal->j_private = sb;
5857 journal->j_bmap = ext4_journal_bmap;
5858 ext4_init_journal_params(sb, journal);
5859 return journal;
5860 }
5861
5862 static struct file *ext4_get_journal_blkdev(struct super_block *sb,
5863 dev_t j_dev, ext4_fsblk_t *j_start,
5864 ext4_fsblk_t *j_len)
5865 {
5866 struct buffer_head *bh;
5867 struct block_device *bdev;
5868 struct file *bdev_file;
5869 int hblock, blocksize;
5870 ext4_fsblk_t sb_block;
5871 unsigned long offset;
5872 struct ext4_super_block *es;
5873 int errno;
5874
5875 bdev_file = bdev_file_open_by_dev(j_dev,
5876 BLK_OPEN_READ | BLK_OPEN_WRITE | BLK_OPEN_RESTRICT_WRITES,
5877 sb, &fs_holder_ops);
5878 if (IS_ERR(bdev_file)) {
5879 ext4_msg(sb, KERN_ERR,
5880 "failed to open journal device unknown-block(%u,%u) %ld",
5881 MAJOR(j_dev), MINOR(j_dev), PTR_ERR(bdev_file));
5882 return bdev_file;
5883 }
5884
5885 bdev = file_bdev(bdev_file);
5886 blocksize = sb->s_blocksize;
5887 hblock = bdev_logical_block_size(bdev);
5888 if (blocksize < hblock) {
5889 ext4_msg(sb, KERN_ERR,
5890 "blocksize too small for journal device");
5891 errno = -EINVAL;
5892 goto out_bdev;
5893 }
5894
5895 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
5896 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
5897 set_blocksize(bdev_file, blocksize);
5898 bh = __bread(bdev, sb_block, blocksize);
5899 if (!bh) {
5900 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
5901 "external journal");
5902 errno = -EINVAL;
5903 goto out_bdev;
5904 }
5905
5906 es = (struct ext4_super_block *) (bh->b_data + offset);
5907 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
5908 !(le32_to_cpu(es->s_feature_incompat) &
5909 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
5910 ext4_msg(sb, KERN_ERR, "external journal has bad superblock");
5911 errno = -EFSCORRUPTED;
5912 goto out_bh;
5913 }
5914
5915 if ((le32_to_cpu(es->s_feature_ro_compat) &
5916 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
5917 es->s_checksum != ext4_superblock_csum(es)) {
5918 ext4_msg(sb, KERN_ERR, "external journal has corrupt superblock");
5919 errno = -EFSCORRUPTED;
5920 goto out_bh;
5921 }
5922
5923 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
5924 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
5925 errno = -EFSCORRUPTED;
5926 goto out_bh;
5927 }
5928
5929 *j_start = sb_block + 1;
5930 *j_len = ext4_blocks_count(es);
5931 brelse(bh);
5932 return bdev_file;
5933
5934 out_bh:
5935 brelse(bh);
5936 out_bdev:
5937 bdev_fput(bdev_file);
5938 return ERR_PTR(errno);
5939 }
5940
5941 static journal_t *ext4_open_dev_journal(struct super_block *sb,
5942 dev_t j_dev)
5943 {
5944 journal_t *journal;
5945 ext4_fsblk_t j_start;
5946 ext4_fsblk_t j_len;
5947 struct file *bdev_file;
5948 int errno = 0;
5949
5950 bdev_file = ext4_get_journal_blkdev(sb, j_dev, &j_start, &j_len);
5951 if (IS_ERR(bdev_file))
5952 return ERR_CAST(bdev_file);
5953
5954 journal = jbd2_journal_init_dev(file_bdev(bdev_file), sb->s_bdev, j_start,
5955 j_len, sb->s_blocksize);
5956 if (IS_ERR(journal)) {
5957 ext4_msg(sb, KERN_ERR, "failed to create device journal");
5958 errno = PTR_ERR(journal);
5959 goto out_bdev;
5960 }
5961 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
5962 ext4_msg(sb, KERN_ERR, "External journal has more than one "
5963 "user (unsupported) - %d",
5964 be32_to_cpu(journal->j_superblock->s_nr_users));
5965 errno = -EINVAL;
5966 goto out_journal;
5967 }
5968 journal->j_private = sb;
5969 EXT4_SB(sb)->s_journal_bdev_file = bdev_file;
5970 ext4_init_journal_params(sb, journal);
5971 return journal;
5972
5973 out_journal:
5974 ext4_journal_destroy(EXT4_SB(sb), journal);
5975 out_bdev:
5976 bdev_fput(bdev_file);
5977 return ERR_PTR(errno);
5978 }
5979
5980 static int ext4_load_journal(struct super_block *sb,
5981 struct ext4_super_block *es,
5982 unsigned long journal_devnum)
5983 {
5984 journal_t *journal;
5985 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
5986 dev_t journal_dev;
5987 int err = 0;
5988 int really_read_only;
5989 int journal_dev_ro;
5990
5991 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5992 return -EFSCORRUPTED;
5993
5994 if (journal_devnum &&
5995 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
5996 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
5997 "numbers have changed");
5998 journal_dev = new_decode_dev(journal_devnum);
5999 } else
6000 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
6001
6002 if (journal_inum && journal_dev) {
6003 ext4_msg(sb, KERN_ERR,
6004 "filesystem has both journal inode and journal device!");
6005 return -EINVAL;
6006 }
6007
6008 if (journal_inum) {
6009 journal = ext4_open_inode_journal(sb, journal_inum);
6010 if (IS_ERR(journal))
6011 return PTR_ERR(journal);
6012 } else {
6013 journal = ext4_open_dev_journal(sb, journal_dev);
6014 if (IS_ERR(journal))
6015 return PTR_ERR(journal);
6016 }
6017
6018 journal_dev_ro = bdev_read_only(journal->j_dev);
6019 really_read_only = bdev_read_only(sb->s_bdev) | journal_dev_ro;
6020
6021 if (journal_dev_ro && !sb_rdonly(sb)) {
6022 ext4_msg(sb, KERN_ERR,
6023 "journal device read-only, try mounting with '-o ro'");
6024 err = -EROFS;
6025 goto err_out;
6026 }
6027
6028 /*
6029 * Are we loading a blank journal or performing recovery after a
6030 * crash? For recovery, we need to check in advance whether we
6031 * can get read-write access to the device.
6032 */
6033 if (ext4_has_feature_journal_needs_recovery(sb)) {
6034 if (sb_rdonly(sb)) {
6035 ext4_msg(sb, KERN_INFO, "INFO: recovery "
6036 "required on readonly filesystem");
6037 if (really_read_only) {
6038 ext4_msg(sb, KERN_ERR, "write access "
6039 "unavailable, cannot proceed "
6040 "(try mounting with noload)");
6041 err = -EROFS;
6042 goto err_out;
6043 }
6044 ext4_msg(sb, KERN_INFO, "write access will "
6045 "be enabled during recovery");
6046 }
6047 }
6048
6049 if (!(journal->j_flags & JBD2_BARRIER))
6050 ext4_msg(sb, KERN_INFO, "barriers disabled");
6051
6052 if (!ext4_has_feature_journal_needs_recovery(sb))
6053 err = jbd2_journal_wipe(journal, !really_read_only);
6054 if (!err) {
6055 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
6056 __le16 orig_state;
6057 bool changed = false;
6058
6059 if (save)
6060 memcpy(save, ((char *) es) +
6061 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
6062 err = jbd2_journal_load(journal);
6063 if (save && memcmp(((char *) es) + EXT4_S_ERR_START,
6064 save, EXT4_S_ERR_LEN)) {
6065 memcpy(((char *) es) + EXT4_S_ERR_START,
6066 save, EXT4_S_ERR_LEN);
6067 changed = true;
6068 }
6069 kfree(save);
6070 orig_state = es->s_state;
6071 es->s_state |= cpu_to_le16(EXT4_SB(sb)->s_mount_state &
6072 EXT4_ERROR_FS);
6073 if (orig_state != es->s_state)
6074 changed = true;
6075 /* Write out restored error information to the superblock */
6076 if (changed && !really_read_only) {
6077 int err2;
6078 err2 = ext4_commit_super(sb);
6079 err = err ? : err2;
6080 }
6081 }
6082
6083 if (err) {
6084 ext4_msg(sb, KERN_ERR, "error loading journal");
6085 goto err_out;
6086 }
6087
6088 EXT4_SB(sb)->s_journal = journal;
6089 err = ext4_clear_journal_err(sb, es);
6090 if (err) {
6091 ext4_journal_destroy(EXT4_SB(sb), journal);
6092 return err;
6093 }
6094
6095 if (!really_read_only && journal_devnum &&
6096 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
6097 es->s_journal_dev = cpu_to_le32(journal_devnum);
6098 ext4_commit_super(sb);
6099 }
6100 if (!really_read_only && journal_inum &&
6101 journal_inum != le32_to_cpu(es->s_journal_inum)) {
6102 es->s_journal_inum = cpu_to_le32(journal_inum);
6103 ext4_commit_super(sb);
6104 }
6105
6106 return 0;
6107
6108 err_out:
6109 ext4_journal_destroy(EXT4_SB(sb), journal);
6110 return err;
6111 }
6112
6113 /* Copy state of EXT4_SB(sb) into buffer for on-disk superblock */
6114 static void ext4_update_super(struct super_block *sb)
6115 {
6116 struct ext4_sb_info *sbi = EXT4_SB(sb);
6117 struct ext4_super_block *es = sbi->s_es;
6118 struct buffer_head *sbh = sbi->s_sbh;
6119
6120 lock_buffer(sbh);
6121 /*
6122 * If the file system is mounted read-only, don't update the
6123 * superblock write time. This avoids updating the superblock
6124 * write time when we are mounting the root file system
6125 * read/only but we need to replay the journal; at that point,
6126 * for people who are east of GMT and who make their clock
6127 * tick in localtime for Windows bug-for-bug compatibility,
6128 * the clock is set in the future, and this will cause e2fsck
6129 * to complain and force a full file system check.
6130 */
6131 if (!sb_rdonly(sb))
6132 ext4_update_tstamp(es, s_wtime);
6133 es->s_kbytes_written =
6134 cpu_to_le64(sbi->s_kbytes_written +
6135 ((part_stat_read(sb->s_bdev, sectors[STAT_WRITE]) -
6136 sbi->s_sectors_written_start) >> 1));
6137 if (percpu_counter_initialized(&sbi->s_freeclusters_counter))
6138 ext4_free_blocks_count_set(es,
6139 EXT4_C2B(sbi, percpu_counter_sum_positive(
6140 &sbi->s_freeclusters_counter)));
6141 if (percpu_counter_initialized(&sbi->s_freeinodes_counter))
6142 es->s_free_inodes_count =
6143 cpu_to_le32(percpu_counter_sum_positive(
6144 &sbi->s_freeinodes_counter));
6145 /* Copy error information to the on-disk superblock */
6146 spin_lock(&sbi->s_error_lock);
6147 if (sbi->s_add_error_count > 0) {
6148 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
6149 if (!es->s_first_error_time && !es->s_first_error_time_hi) {
6150 __ext4_update_tstamp(&es->s_first_error_time,
6151 &es->s_first_error_time_hi,
6152 sbi->s_first_error_time);
6153 strtomem_pad(es->s_first_error_func,
6154 sbi->s_first_error_func, 0);
6155 es->s_first_error_line =
6156 cpu_to_le32(sbi->s_first_error_line);
6157 es->s_first_error_ino =
6158 cpu_to_le32(sbi->s_first_error_ino);
6159 es->s_first_error_block =
6160 cpu_to_le64(sbi->s_first_error_block);
6161 es->s_first_error_errcode =
6162 ext4_errno_to_code(sbi->s_first_error_code);
6163 }
6164 __ext4_update_tstamp(&es->s_last_error_time,
6165 &es->s_last_error_time_hi,
6166 sbi->s_last_error_time);
6167 strtomem_pad(es->s_last_error_func, sbi->s_last_error_func, 0);
6168 es->s_last_error_line = cpu_to_le32(sbi->s_last_error_line);
6169 es->s_last_error_ino = cpu_to_le32(sbi->s_last_error_ino);
6170 es->s_last_error_block = cpu_to_le64(sbi->s_last_error_block);
6171 es->s_last_error_errcode =
6172 ext4_errno_to_code(sbi->s_last_error_code);
6173 /*
6174 * Start the daily error reporting function if it hasn't been
6175 * started already
6176 */
6177 if (!es->s_error_count)
6178 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ);
6179 le32_add_cpu(&es->s_error_count, sbi->s_add_error_count);
6180 sbi->s_add_error_count = 0;
6181 }
6182 spin_unlock(&sbi->s_error_lock);
6183
6184 ext4_superblock_csum_set(sb);
6185 unlock_buffer(sbh);
6186 }
6187
6188 static int ext4_commit_super(struct super_block *sb)
6189 {
6190 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
6191
6192 if (!sbh)
6193 return -EINVAL;
6194
6195 ext4_update_super(sb);
6196
6197 lock_buffer(sbh);
6198 /* Buffer got discarded which means block device got invalidated */
6199 if (!buffer_mapped(sbh)) {
6200 unlock_buffer(sbh);
6201 return -EIO;
6202 }
6203
6204 if (buffer_write_io_error(sbh) || !buffer_uptodate(sbh)) {
6205 /*
6206 * Oh, dear. A previous attempt to write the
6207 * superblock failed. This could happen because the
6208 * USB device was yanked out. Or it could happen to
6209 * be a transient write error and maybe the block will
6210 * be remapped. Nothing we can do but to retry the
6211 * write and hope for the best.
6212 */
6213 ext4_msg(sb, KERN_ERR, "previous I/O error to "
6214 "superblock detected");
6215 clear_buffer_write_io_error(sbh);
6216 set_buffer_uptodate(sbh);
6217 }
6218 get_bh(sbh);
6219 /* Clear potential dirty bit if it was journalled update */
6220 clear_buffer_dirty(sbh);
6221 sbh->b_end_io = end_buffer_write_sync;
6222 submit_bh(REQ_OP_WRITE | REQ_SYNC |
6223 (test_opt(sb, BARRIER) ? REQ_FUA : 0), sbh);
6224 wait_on_buffer(sbh);
6225 if (buffer_write_io_error(sbh)) {
6226 ext4_msg(sb, KERN_ERR, "I/O error while writing "
6227 "superblock");
6228 clear_buffer_write_io_error(sbh);
6229 set_buffer_uptodate(sbh);
6230 return -EIO;
6231 }
6232 return 0;
6233 }
6234
6235 /*
6236 * Have we just finished recovery? If so, and if we are mounting (or
6237 * remounting) the filesystem readonly, then we will end up with a
6238 * consistent fs on disk. Record that fact.
6239 */
6240 static int ext4_mark_recovery_complete(struct super_block *sb,
6241 struct ext4_super_block *es)
6242 {
6243 int err;
6244 journal_t *journal = EXT4_SB(sb)->s_journal;
6245
6246 if (!ext4_has_feature_journal(sb)) {
6247 if (journal != NULL) {
6248 ext4_error(sb, "Journal got removed while the fs was "
6249 "mounted!");
6250 return -EFSCORRUPTED;
6251 }
6252 return 0;
6253 }
6254 jbd2_journal_lock_updates(journal);
6255 err = jbd2_journal_flush(journal, 0);
6256 if (err < 0)
6257 goto out;
6258
6259 if (sb_rdonly(sb) && (ext4_has_feature_journal_needs_recovery(sb) ||
6260 ext4_has_feature_orphan_present(sb))) {
6261 if (!ext4_orphan_file_empty(sb)) {
6262 ext4_error(sb, "Orphan file not empty on read-only fs.");
6263 err = -EFSCORRUPTED;
6264 goto out;
6265 }
6266 ext4_clear_feature_journal_needs_recovery(sb);
6267 ext4_clear_feature_orphan_present(sb);
6268 ext4_commit_super(sb);
6269 }
6270 out:
6271 jbd2_journal_unlock_updates(journal);
6272 return err;
6273 }
6274
6275 /*
6276 * If we are mounting (or read-write remounting) a filesystem whose journal
6277 * has recorded an error from a previous lifetime, move that error to the
6278 * main filesystem now.
6279 */
6280 static int ext4_clear_journal_err(struct super_block *sb,
6281 struct ext4_super_block *es)
6282 {
6283 journal_t *journal;
6284 int j_errno;
6285 const char *errstr;
6286
6287 if (!ext4_has_feature_journal(sb)) {
6288 ext4_error(sb, "Journal got removed while the fs was mounted!");
6289 return -EFSCORRUPTED;
6290 }
6291
6292 journal = EXT4_SB(sb)->s_journal;
6293
6294 /*
6295 * Now check for any error status which may have been recorded in the
6296 * journal by a prior ext4_error() or ext4_abort()
6297 */
6298
6299 j_errno = jbd2_journal_errno(journal);
6300 if (j_errno) {
6301 char nbuf[16];
6302
6303 errstr = ext4_decode_error(sb, j_errno, nbuf);
6304 ext4_warning(sb, "Filesystem error recorded "
6305 "from previous mount: %s", errstr);
6306
6307 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
6308 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
6309 j_errno = ext4_commit_super(sb);
6310 if (j_errno)
6311 return j_errno;
6312 ext4_warning(sb, "Marked fs in need of filesystem check.");
6313
6314 jbd2_journal_clear_err(journal);
6315 jbd2_journal_update_sb_errno(journal);
6316 }
6317 return 0;
6318 }
6319
6320 /*
6321 * Force the running and committing transactions to commit,
6322 * and wait on the commit.
6323 */
6324 int ext4_force_commit(struct super_block *sb)
6325 {
6326 return ext4_journal_force_commit(EXT4_SB(sb)->s_journal);
6327 }
6328
6329 static int ext4_sync_fs(struct super_block *sb, int wait)
6330 {
6331 int ret = 0;
6332 tid_t target;
6333 bool needs_barrier = false;
6334 struct ext4_sb_info *sbi = EXT4_SB(sb);
6335
6336 ret = ext4_emergency_state(sb);
6337 if (unlikely(ret))
6338 return ret;
6339
6340 trace_ext4_sync_fs(sb, wait);
6341 flush_workqueue(sbi->rsv_conversion_wq);
6342 /*
6343 * Writeback quota in non-journalled quota case - journalled quota has
6344 * no dirty dquots
6345 */
6346 dquot_writeback_dquots(sb, -1);
6347 /*
6348 * Data writeback is possible w/o journal transaction, so barrier must
6349 * being sent at the end of the function. But we can skip it if
6350 * transaction_commit will do it for us.
6351 */
6352 if (sbi->s_journal) {
6353 target = jbd2_get_latest_transaction(sbi->s_journal);
6354 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
6355 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
6356 needs_barrier = true;
6357
6358 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
6359 if (wait)
6360 ret = jbd2_log_wait_commit(sbi->s_journal,
6361 target);
6362 }
6363 } else if (wait && test_opt(sb, BARRIER))
6364 needs_barrier = true;
6365 if (needs_barrier) {
6366 int err;
6367 err = blkdev_issue_flush(sb->s_bdev);
6368 if (!ret)
6369 ret = err;
6370 }
6371
6372 return ret;
6373 }
6374
6375 /*
6376 * LVM calls this function before a (read-only) snapshot is created. This
6377 * gives us a chance to flush the journal completely and mark the fs clean.
6378 *
6379 * Note that only this function cannot bring a filesystem to be in a clean
6380 * state independently. It relies on upper layer to stop all data & metadata
6381 * modifications.
6382 */
6383 static int ext4_freeze(struct super_block *sb)
6384 {
6385 int error = 0;
6386 journal_t *journal = EXT4_SB(sb)->s_journal;
6387
6388 if (journal) {
6389 /* Now we set up the journal barrier. */
6390 jbd2_journal_lock_updates(journal);
6391
6392 /*
6393 * Don't clear the needs_recovery flag if we failed to
6394 * flush the journal.
6395 */
6396 error = jbd2_journal_flush(journal, 0);
6397 if (error < 0)
6398 goto out;
6399
6400 /* Journal blocked and flushed, clear needs_recovery flag. */
6401 ext4_clear_feature_journal_needs_recovery(sb);
6402 if (ext4_orphan_file_empty(sb))
6403 ext4_clear_feature_orphan_present(sb);
6404 }
6405
6406 error = ext4_commit_super(sb);
6407 out:
6408 if (journal)
6409 /* we rely on upper layer to stop further updates */
6410 jbd2_journal_unlock_updates(journal);
6411 return error;
6412 }
6413
6414 /*
6415 * Called by LVM after the snapshot is done. We need to reset the RECOVER
6416 * flag here, even though the filesystem is not technically dirty yet.
6417 */
6418 static int ext4_unfreeze(struct super_block *sb)
6419 {
6420 if (ext4_emergency_state(sb))
6421 return 0;
6422
6423 if (EXT4_SB(sb)->s_journal) {
6424 /* Reset the needs_recovery flag before the fs is unlocked. */
6425 ext4_set_feature_journal_needs_recovery(sb);
6426 if (ext4_has_feature_orphan_file(sb))
6427 ext4_set_feature_orphan_present(sb);
6428 }
6429
6430 ext4_commit_super(sb);
6431 return 0;
6432 }
6433
6434 /*
6435 * Structure to save mount options for ext4_remount's benefit
6436 */
6437 struct ext4_mount_options {
6438 unsigned long s_mount_opt;
6439 unsigned long s_mount_opt2;
6440 kuid_t s_resuid;
6441 kgid_t s_resgid;
6442 unsigned long s_commit_interval;
6443 u32 s_min_batch_time, s_max_batch_time;
6444 #ifdef CONFIG_QUOTA
6445 int s_jquota_fmt;
6446 char *s_qf_names[EXT4_MAXQUOTAS];
6447 #endif
6448 };
6449
6450 static int __ext4_remount(struct fs_context *fc, struct super_block *sb)
6451 {
6452 struct ext4_fs_context *ctx = fc->fs_private;
6453 struct ext4_super_block *es;
6454 struct ext4_sb_info *sbi = EXT4_SB(sb);
6455 unsigned long old_sb_flags;
6456 struct ext4_mount_options old_opts;
6457 ext4_group_t g;
6458 int err = 0;
6459 int alloc_ctx;
6460 #ifdef CONFIG_QUOTA
6461 int enable_quota = 0;
6462 int i, j;
6463 char *to_free[EXT4_MAXQUOTAS];
6464 #endif
6465
6466
6467 /* Store the original options */
6468 old_sb_flags = sb->s_flags;
6469 old_opts.s_mount_opt = sbi->s_mount_opt;
6470 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
6471 old_opts.s_resuid = sbi->s_resuid;
6472 old_opts.s_resgid = sbi->s_resgid;
6473 old_opts.s_commit_interval = sbi->s_commit_interval;
6474 old_opts.s_min_batch_time = sbi->s_min_batch_time;
6475 old_opts.s_max_batch_time = sbi->s_max_batch_time;
6476 #ifdef CONFIG_QUOTA
6477 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
6478 for (i = 0; i < EXT4_MAXQUOTAS; i++)
6479 if (sbi->s_qf_names[i]) {
6480 char *qf_name = get_qf_name(sb, sbi, i);
6481
6482 old_opts.s_qf_names[i] = kstrdup(qf_name, GFP_KERNEL);
6483 if (!old_opts.s_qf_names[i]) {
6484 for (j = 0; j < i; j++)
6485 kfree(old_opts.s_qf_names[j]);
6486 return -ENOMEM;
6487 }
6488 } else
6489 old_opts.s_qf_names[i] = NULL;
6490 #endif
6491 if (!(ctx->spec & EXT4_SPEC_JOURNAL_IOPRIO)) {
6492 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
6493 ctx->journal_ioprio =
6494 sbi->s_journal->j_task->io_context->ioprio;
6495 else
6496 ctx->journal_ioprio = EXT4_DEF_JOURNAL_IOPRIO;
6497
6498 }
6499
6500 if ((ctx->spec & EXT4_SPEC_s_stripe) &&
6501 ext4_is_stripe_incompatible(sb, ctx->s_stripe)) {
6502 ext4_msg(sb, KERN_WARNING,
6503 "stripe (%lu) is not aligned with cluster size (%u), "
6504 "stripe is disabled",
6505 ctx->s_stripe, sbi->s_cluster_ratio);
6506 ctx->s_stripe = 0;
6507 }
6508
6509 /*
6510 * Changing the DIOREAD_NOLOCK or DELALLOC mount options may cause
6511 * two calls to ext4_should_dioread_nolock() to return inconsistent
6512 * values, triggering WARN_ON in ext4_add_complete_io(). we grab
6513 * here s_writepages_rwsem to avoid race between writepages ops and
6514 * remount.
6515 */
6516 alloc_ctx = ext4_writepages_down_write(sb);
6517 ext4_apply_options(fc, sb);
6518 ext4_writepages_up_write(sb, alloc_ctx);
6519
6520 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
6521 test_opt(sb, JOURNAL_CHECKSUM)) {
6522 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
6523 "during remount not supported; ignoring");
6524 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
6525 }
6526
6527 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
6528 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
6529 ext4_msg(sb, KERN_ERR, "can't mount with "
6530 "both data=journal and delalloc");
6531 err = -EINVAL;
6532 goto restore_opts;
6533 }
6534 if (test_opt(sb, DIOREAD_NOLOCK)) {
6535 ext4_msg(sb, KERN_ERR, "can't mount with "
6536 "both data=journal and dioread_nolock");
6537 err = -EINVAL;
6538 goto restore_opts;
6539 }
6540 } else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) {
6541 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
6542 ext4_msg(sb, KERN_ERR, "can't mount with "
6543 "journal_async_commit in data=ordered mode");
6544 err = -EINVAL;
6545 goto restore_opts;
6546 }
6547 }
6548
6549 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_NO_MBCACHE) {
6550 ext4_msg(sb, KERN_ERR, "can't enable nombcache during remount");
6551 err = -EINVAL;
6552 goto restore_opts;
6553 }
6554
6555 if ((old_opts.s_mount_opt & EXT4_MOUNT_DELALLOC) &&
6556 !test_opt(sb, DELALLOC)) {
6557 ext4_msg(sb, KERN_ERR, "can't disable delalloc during remount");
6558 err = -EINVAL;
6559 goto restore_opts;
6560 }
6561
6562 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
6563 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
6564
6565 es = sbi->s_es;
6566
6567 if (sbi->s_journal) {
6568 ext4_init_journal_params(sb, sbi->s_journal);
6569 set_task_ioprio(sbi->s_journal->j_task, ctx->journal_ioprio);
6570 }
6571
6572 /* Flush outstanding errors before changing fs state */
6573 flush_work(&sbi->s_sb_upd_work);
6574
6575 if ((bool)(fc->sb_flags & SB_RDONLY) != sb_rdonly(sb)) {
6576 if (ext4_emergency_state(sb)) {
6577 err = -EROFS;
6578 goto restore_opts;
6579 }
6580
6581 if (fc->sb_flags & SB_RDONLY) {
6582 err = sync_filesystem(sb);
6583 if (err < 0)
6584 goto restore_opts;
6585 err = dquot_suspend(sb, -1);
6586 if (err < 0)
6587 goto restore_opts;
6588
6589 /*
6590 * First of all, the unconditional stuff we have to do
6591 * to disable replay of the journal when we next remount
6592 */
6593 sb->s_flags |= SB_RDONLY;
6594
6595 /*
6596 * OK, test if we are remounting a valid rw partition
6597 * readonly, and if so set the rdonly flag and then
6598 * mark the partition as valid again.
6599 */
6600 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
6601 (sbi->s_mount_state & EXT4_VALID_FS))
6602 es->s_state = cpu_to_le16(sbi->s_mount_state);
6603
6604 if (sbi->s_journal) {
6605 /*
6606 * We let remount-ro finish even if marking fs
6607 * as clean failed...
6608 */
6609 ext4_mark_recovery_complete(sb, es);
6610 }
6611 } else {
6612 /* Make sure we can mount this feature set readwrite */
6613 if (ext4_has_feature_readonly(sb) ||
6614 !ext4_feature_set_ok(sb, 0)) {
6615 err = -EROFS;
6616 goto restore_opts;
6617 }
6618 /*
6619 * Make sure the group descriptor checksums
6620 * are sane. If they aren't, refuse to remount r/w.
6621 */
6622 for (g = 0; g < sbi->s_groups_count; g++) {
6623 struct ext4_group_desc *gdp =
6624 ext4_get_group_desc(sb, g, NULL);
6625
6626 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
6627 ext4_msg(sb, KERN_ERR,
6628 "ext4_remount: Checksum for group %u failed (%u!=%u)",
6629 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
6630 le16_to_cpu(gdp->bg_checksum));
6631 err = -EFSBADCRC;
6632 goto restore_opts;
6633 }
6634 }
6635
6636 /*
6637 * If we have an unprocessed orphan list hanging
6638 * around from a previously readonly bdev mount,
6639 * require a full umount/remount for now.
6640 */
6641 if (es->s_last_orphan || !ext4_orphan_file_empty(sb)) {
6642 ext4_msg(sb, KERN_WARNING, "Couldn't "
6643 "remount RDWR because of unprocessed "
6644 "orphan inode list. Please "
6645 "umount/remount instead");
6646 err = -EINVAL;
6647 goto restore_opts;
6648 }
6649
6650 /*
6651 * Mounting a RDONLY partition read-write, so reread
6652 * and store the current valid flag. (It may have
6653 * been changed by e2fsck since we originally mounted
6654 * the partition.)
6655 */
6656 if (sbi->s_journal) {
6657 err = ext4_clear_journal_err(sb, es);
6658 if (err)
6659 goto restore_opts;
6660 }
6661 sbi->s_mount_state = (le16_to_cpu(es->s_state) &
6662 ~EXT4_FC_REPLAY);
6663
6664 err = ext4_setup_super(sb, es, 0);
6665 if (err)
6666 goto restore_opts;
6667
6668 sb->s_flags &= ~SB_RDONLY;
6669 if (ext4_has_feature_mmp(sb)) {
6670 err = ext4_multi_mount_protect(sb,
6671 le64_to_cpu(es->s_mmp_block));
6672 if (err)
6673 goto restore_opts;
6674 }
6675 #ifdef CONFIG_QUOTA
6676 enable_quota = 1;
6677 #endif
6678 }
6679 }
6680
6681 /*
6682 * Handle creation of system zone data early because it can fail.
6683 * Releasing of existing data is done when we are sure remount will
6684 * succeed.
6685 */
6686 if (test_opt(sb, BLOCK_VALIDITY) && !sbi->s_system_blks) {
6687 err = ext4_setup_system_zone(sb);
6688 if (err)
6689 goto restore_opts;
6690 }
6691
6692 if (sbi->s_journal == NULL && !(old_sb_flags & SB_RDONLY)) {
6693 err = ext4_commit_super(sb);
6694 if (err)
6695 goto restore_opts;
6696 }
6697
6698 #ifdef CONFIG_QUOTA
6699 if (enable_quota) {
6700 if (sb_any_quota_suspended(sb))
6701 dquot_resume(sb, -1);
6702 else if (ext4_has_feature_quota(sb)) {
6703 err = ext4_enable_quotas(sb);
6704 if (err)
6705 goto restore_opts;
6706 }
6707 }
6708 /* Release old quota file names */
6709 for (i = 0; i < EXT4_MAXQUOTAS; i++)
6710 kfree(old_opts.s_qf_names[i]);
6711 #endif
6712 if (!test_opt(sb, BLOCK_VALIDITY) && sbi->s_system_blks)
6713 ext4_release_system_zone(sb);
6714
6715 /*
6716 * Reinitialize lazy itable initialization thread based on
6717 * current settings
6718 */
6719 if (sb_rdonly(sb) || !test_opt(sb, INIT_INODE_TABLE))
6720 ext4_unregister_li_request(sb);
6721 else {
6722 ext4_group_t first_not_zeroed;
6723 first_not_zeroed = ext4_has_uninit_itable(sb);
6724 ext4_register_li_request(sb, first_not_zeroed);
6725 }
6726
6727 if (!ext4_has_feature_mmp(sb) || sb_rdonly(sb))
6728 ext4_stop_mmpd(sbi);
6729
6730 /*
6731 * Handle aborting the filesystem as the last thing during remount to
6732 * avoid obsure errors during remount when some option changes fail to
6733 * apply due to shutdown filesystem.
6734 */
6735 if (test_opt2(sb, ABORT))
6736 ext4_abort(sb, ESHUTDOWN, "Abort forced by user");
6737
6738 return 0;
6739
6740 restore_opts:
6741 /*
6742 * If there was a failing r/w to ro transition, we may need to
6743 * re-enable quota
6744 */
6745 if (sb_rdonly(sb) && !(old_sb_flags & SB_RDONLY) &&
6746 sb_any_quota_suspended(sb))
6747 dquot_resume(sb, -1);
6748
6749 alloc_ctx = ext4_writepages_down_write(sb);
6750 sb->s_flags = old_sb_flags;
6751 sbi->s_mount_opt = old_opts.s_mount_opt;
6752 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
6753 sbi->s_resuid = old_opts.s_resuid;
6754 sbi->s_resgid = old_opts.s_resgid;
6755 sbi->s_commit_interval = old_opts.s_commit_interval;
6756 sbi->s_min_batch_time = old_opts.s_min_batch_time;
6757 sbi->s_max_batch_time = old_opts.s_max_batch_time;
6758 ext4_writepages_up_write(sb, alloc_ctx);
6759
6760 if (!test_opt(sb, BLOCK_VALIDITY) && sbi->s_system_blks)
6761 ext4_release_system_zone(sb);
6762 #ifdef CONFIG_QUOTA
6763 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
6764 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
6765 to_free[i] = get_qf_name(sb, sbi, i);
6766 rcu_assign_pointer(sbi->s_qf_names[i], old_opts.s_qf_names[i]);
6767 }
6768 synchronize_rcu();
6769 for (i = 0; i < EXT4_MAXQUOTAS; i++)
6770 kfree(to_free[i]);
6771 #endif
6772 if (!ext4_has_feature_mmp(sb) || sb_rdonly(sb))
6773 ext4_stop_mmpd(sbi);
6774 return err;
6775 }
6776
6777 static int ext4_reconfigure(struct fs_context *fc)
6778 {
6779 struct super_block *sb = fc->root->d_sb;
6780 int ret;
6781 bool old_ro = sb_rdonly(sb);
6782
6783 fc->s_fs_info = EXT4_SB(sb);
6784
6785 ret = ext4_check_opt_consistency(fc, sb);
6786 if (ret < 0)
6787 return ret;
6788
6789 ret = __ext4_remount(fc, sb);
6790 if (ret < 0)
6791 return ret;
6792
6793 ext4_msg(sb, KERN_INFO, "re-mounted %pU%s.",
6794 &sb->s_uuid,
6795 (old_ro != sb_rdonly(sb)) ? (sb_rdonly(sb) ? " ro" : " r/w") : "");
6796
6797 return 0;
6798 }
6799
6800 #ifdef CONFIG_QUOTA
6801 static int ext4_statfs_project(struct super_block *sb,
6802 kprojid_t projid, struct kstatfs *buf)
6803 {
6804 struct kqid qid;
6805 struct dquot *dquot;
6806 u64 limit;
6807 u64 curblock;
6808
6809 qid = make_kqid_projid(projid);
6810 dquot = dqget(sb, qid);
6811 if (IS_ERR(dquot))
6812 return PTR_ERR(dquot);
6813 spin_lock(&dquot->dq_dqb_lock);
6814
6815 limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit,
6816 dquot->dq_dqb.dqb_bhardlimit);
6817 limit >>= sb->s_blocksize_bits;
6818
6819 if (limit) {
6820 uint64_t remaining = 0;
6821
6822 curblock = (dquot->dq_dqb.dqb_curspace +
6823 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits;
6824 if (limit > curblock)
6825 remaining = limit - curblock;
6826
6827 buf->f_blocks = min(buf->f_blocks, limit);
6828 buf->f_bfree = min(buf->f_bfree, remaining);
6829 buf->f_bavail = min(buf->f_bavail, remaining);
6830 }
6831
6832 limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit,
6833 dquot->dq_dqb.dqb_ihardlimit);
6834 if (limit) {
6835 uint64_t remaining = 0;
6836
6837 if (limit > dquot->dq_dqb.dqb_curinodes)
6838 remaining = limit - dquot->dq_dqb.dqb_curinodes;
6839
6840 buf->f_files = min(buf->f_files, limit);
6841 buf->f_ffree = min(buf->f_ffree, remaining);
6842 }
6843
6844 spin_unlock(&dquot->dq_dqb_lock);
6845 dqput(dquot);
6846 return 0;
6847 }
6848 #endif
6849
6850 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
6851 {
6852 struct super_block *sb = dentry->d_sb;
6853 struct ext4_sb_info *sbi = EXT4_SB(sb);
6854 struct ext4_super_block *es = sbi->s_es;
6855 ext4_fsblk_t overhead = 0, resv_blocks;
6856 s64 bfree;
6857 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
6858
6859 if (!test_opt(sb, MINIX_DF))
6860 overhead = sbi->s_overhead;
6861
6862 buf->f_type = EXT4_SUPER_MAGIC;
6863 buf->f_bsize = sb->s_blocksize;
6864 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
6865 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
6866 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
6867 /* prevent underflow in case that few free space is available */
6868 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
6869 buf->f_bavail = buf->f_bfree -
6870 (ext4_r_blocks_count(es) + resv_blocks);
6871 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
6872 buf->f_bavail = 0;
6873 buf->f_files = le32_to_cpu(es->s_inodes_count);
6874 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
6875 buf->f_namelen = EXT4_NAME_LEN;
6876 buf->f_fsid = uuid_to_fsid(es->s_uuid);
6877
6878 #ifdef CONFIG_QUOTA
6879 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) &&
6880 sb_has_quota_limits_enabled(sb, PRJQUOTA))
6881 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf);
6882 #endif
6883 return 0;
6884 }
6885
6886
6887 #ifdef CONFIG_QUOTA
6888
6889 /*
6890 * Helper functions so that transaction is started before we acquire dqio_sem
6891 * to keep correct lock ordering of transaction > dqio_sem
6892 */
6893 static inline struct inode *dquot_to_inode(struct dquot *dquot)
6894 {
6895 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
6896 }
6897
6898 static int ext4_write_dquot(struct dquot *dquot)
6899 {
6900 int ret, err;
6901 handle_t *handle;
6902 struct inode *inode;
6903
6904 inode = dquot_to_inode(dquot);
6905 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
6906 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
6907 if (IS_ERR(handle))
6908 return PTR_ERR(handle);
6909 ret = dquot_commit(dquot);
6910 if (ret < 0)
6911 ext4_error_err(dquot->dq_sb, -ret,
6912 "Failed to commit dquot type %d",
6913 dquot->dq_id.type);
6914 err = ext4_journal_stop(handle);
6915 if (!ret)
6916 ret = err;
6917 return ret;
6918 }
6919
6920 static int ext4_acquire_dquot(struct dquot *dquot)
6921 {
6922 int ret, err;
6923 handle_t *handle;
6924
6925 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
6926 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
6927 if (IS_ERR(handle))
6928 return PTR_ERR(handle);
6929 ret = dquot_acquire(dquot);
6930 if (ret < 0)
6931 ext4_error_err(dquot->dq_sb, -ret,
6932 "Failed to acquire dquot type %d",
6933 dquot->dq_id.type);
6934 err = ext4_journal_stop(handle);
6935 if (!ret)
6936 ret = err;
6937 return ret;
6938 }
6939
6940 static int ext4_release_dquot(struct dquot *dquot)
6941 {
6942 int ret, err;
6943 handle_t *handle;
6944 bool freeze_protected = false;
6945
6946 /*
6947 * Trying to sb_start_intwrite() in a running transaction
6948 * can result in a deadlock. Further, running transactions
6949 * are already protected from freezing.
6950 */
6951 if (!ext4_journal_current_handle()) {
6952 sb_start_intwrite(dquot->dq_sb);
6953 freeze_protected = true;
6954 }
6955
6956 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
6957 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
6958 if (IS_ERR(handle)) {
6959 /* Release dquot anyway to avoid endless cycle in dqput() */
6960 dquot_release(dquot);
6961 if (freeze_protected)
6962 sb_end_intwrite(dquot->dq_sb);
6963 return PTR_ERR(handle);
6964 }
6965 ret = dquot_release(dquot);
6966 if (ret < 0)
6967 ext4_error_err(dquot->dq_sb, -ret,
6968 "Failed to release dquot type %d",
6969 dquot->dq_id.type);
6970 err = ext4_journal_stop(handle);
6971 if (!ret)
6972 ret = err;
6973
6974 if (freeze_protected)
6975 sb_end_intwrite(dquot->dq_sb);
6976
6977 return ret;
6978 }
6979
6980 static int ext4_mark_dquot_dirty(struct dquot *dquot)
6981 {
6982 struct super_block *sb = dquot->dq_sb;
6983
6984 if (ext4_is_quota_journalled(sb)) {
6985 dquot_mark_dquot_dirty(dquot);
6986 return ext4_write_dquot(dquot);
6987 } else {
6988 return dquot_mark_dquot_dirty(dquot);
6989 }
6990 }
6991
6992 static int ext4_write_info(struct super_block *sb, int type)
6993 {
6994 int ret, err;
6995 handle_t *handle;
6996
6997 /* Data block + inode block */
6998 handle = ext4_journal_start_sb(sb, EXT4_HT_QUOTA, 2);
6999 if (IS_ERR(handle))
7000 return PTR_ERR(handle);
7001 ret = dquot_commit_info(sb, type);
7002 err = ext4_journal_stop(handle);
7003 if (!ret)
7004 ret = err;
7005 return ret;
7006 }
7007
7008 static void lockdep_set_quota_inode(struct inode *inode, int subclass)
7009 {
7010 struct ext4_inode_info *ei = EXT4_I(inode);
7011
7012 /* The first argument of lockdep_set_subclass has to be
7013 * *exactly* the same as the argument to init_rwsem() --- in
7014 * this case, in init_once() --- or lockdep gets unhappy
7015 * because the name of the lock is set using the
7016 * stringification of the argument to init_rwsem().
7017 */
7018 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */
7019 lockdep_set_subclass(&ei->i_data_sem, subclass);
7020 }
7021
7022 /*
7023 * Standard function to be called on quota_on
7024 */
7025 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
7026 const struct path *path)
7027 {
7028 int err;
7029
7030 if (!test_opt(sb, QUOTA))
7031 return -EINVAL;
7032
7033 /* Quotafile not on the same filesystem? */
7034 if (path->dentry->d_sb != sb)
7035 return -EXDEV;
7036
7037 /* Quota already enabled for this file? */
7038 if (IS_NOQUOTA(d_inode(path->dentry)))
7039 return -EBUSY;
7040
7041 /* Journaling quota? */
7042 if (EXT4_SB(sb)->s_qf_names[type]) {
7043 /* Quotafile not in fs root? */
7044 if (path->dentry->d_parent != sb->s_root)
7045 ext4_msg(sb, KERN_WARNING,
7046 "Quota file not on filesystem root. "
7047 "Journaled quota will not work");
7048 sb_dqopt(sb)->flags |= DQUOT_NOLIST_DIRTY;
7049 } else {
7050 /*
7051 * Clear the flag just in case mount options changed since
7052 * last time.
7053 */
7054 sb_dqopt(sb)->flags &= ~DQUOT_NOLIST_DIRTY;
7055 }
7056
7057 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
7058 err = dquot_quota_on(sb, type, format_id, path);
7059 if (!err) {
7060 struct inode *inode = d_inode(path->dentry);
7061 handle_t *handle;
7062
7063 /*
7064 * Set inode flags to prevent userspace from messing with quota
7065 * files. If this fails, we return success anyway since quotas
7066 * are already enabled and this is not a hard failure.
7067 */
7068 inode_lock(inode);
7069 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
7070 if (IS_ERR(handle))
7071 goto unlock_inode;
7072 EXT4_I(inode)->i_flags |= EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL;
7073 inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
7074 S_NOATIME | S_IMMUTABLE);
7075 err = ext4_mark_inode_dirty(handle, inode);
7076 ext4_journal_stop(handle);
7077 unlock_inode:
7078 inode_unlock(inode);
7079 if (err)
7080 dquot_quota_off(sb, type);
7081 }
7082 if (err)
7083 lockdep_set_quota_inode(path->dentry->d_inode,
7084 I_DATA_SEM_NORMAL);
7085 return err;
7086 }
7087
7088 static inline bool ext4_check_quota_inum(int type, unsigned long qf_inum)
7089 {
7090 switch (type) {
7091 case USRQUOTA:
7092 return qf_inum == EXT4_USR_QUOTA_INO;
7093 case GRPQUOTA:
7094 return qf_inum == EXT4_GRP_QUOTA_INO;
7095 case PRJQUOTA:
7096 return qf_inum >= EXT4_GOOD_OLD_FIRST_INO;
7097 default:
7098 BUG();
7099 }
7100 }
7101
7102 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
7103 unsigned int flags)
7104 {
7105 int err;
7106 struct inode *qf_inode;
7107 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
7108 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
7109 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
7110 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
7111 };
7112
7113 BUG_ON(!ext4_has_feature_quota(sb));
7114
7115 if (!qf_inums[type])
7116 return -EPERM;
7117
7118 if (!ext4_check_quota_inum(type, qf_inums[type])) {
7119 ext4_error(sb, "Bad quota inum: %lu, type: %d",
7120 qf_inums[type], type);
7121 return -EUCLEAN;
7122 }
7123
7124 qf_inode = ext4_iget(sb, qf_inums[type], EXT4_IGET_SPECIAL);
7125 if (IS_ERR(qf_inode)) {
7126 ext4_error(sb, "Bad quota inode: %lu, type: %d",
7127 qf_inums[type], type);
7128 return PTR_ERR(qf_inode);
7129 }
7130
7131 /* Don't account quota for quota files to avoid recursion */
7132 qf_inode->i_flags |= S_NOQUOTA;
7133 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
7134 err = dquot_load_quota_inode(qf_inode, type, format_id, flags);
7135 if (err)
7136 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
7137 iput(qf_inode);
7138
7139 return err;
7140 }
7141
7142 /* Enable usage tracking for all quota types. */
7143 int ext4_enable_quotas(struct super_block *sb)
7144 {
7145 int type, err = 0;
7146 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
7147 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
7148 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
7149 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
7150 };
7151 bool quota_mopt[EXT4_MAXQUOTAS] = {
7152 test_opt(sb, USRQUOTA),
7153 test_opt(sb, GRPQUOTA),
7154 test_opt(sb, PRJQUOTA),
7155 };
7156
7157 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NOLIST_DIRTY;
7158 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
7159 if (qf_inums[type]) {
7160 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
7161 DQUOT_USAGE_ENABLED |
7162 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
7163 if (err) {
7164 ext4_warning(sb,
7165 "Failed to enable quota tracking "
7166 "(type=%d, err=%d, ino=%lu). "
7167 "Please run e2fsck to fix.", type,
7168 err, qf_inums[type]);
7169
7170 ext4_quotas_off(sb, type);
7171 return err;
7172 }
7173 }
7174 }
7175 return 0;
7176 }
7177
7178 static int ext4_quota_off(struct super_block *sb, int type)
7179 {
7180 struct inode *inode = sb_dqopt(sb)->files[type];
7181 handle_t *handle;
7182 int err;
7183
7184 /* Force all delayed allocation blocks to be allocated.
7185 * Caller already holds s_umount sem */
7186 if (test_opt(sb, DELALLOC))
7187 sync_filesystem(sb);
7188
7189 if (!inode || !igrab(inode))
7190 goto out;
7191
7192 err = dquot_quota_off(sb, type);
7193 if (err || ext4_has_feature_quota(sb))
7194 goto out_put;
7195 /*
7196 * When the filesystem was remounted read-only first, we cannot cleanup
7197 * inode flags here. Bad luck but people should be using QUOTA feature
7198 * these days anyway.
7199 */
7200 if (sb_rdonly(sb))
7201 goto out_put;
7202
7203 inode_lock(inode);
7204 /*
7205 * Update modification times of quota files when userspace can
7206 * start looking at them. If we fail, we return success anyway since
7207 * this is not a hard failure and quotas are already disabled.
7208 */
7209 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
7210 if (IS_ERR(handle)) {
7211 err = PTR_ERR(handle);
7212 goto out_unlock;
7213 }
7214 EXT4_I(inode)->i_flags &= ~(EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL);
7215 inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
7216 inode_set_mtime_to_ts(inode, inode_set_ctime_current(inode));
7217 err = ext4_mark_inode_dirty(handle, inode);
7218 ext4_journal_stop(handle);
7219 out_unlock:
7220 inode_unlock(inode);
7221 out_put:
7222 lockdep_set_quota_inode(inode, I_DATA_SEM_NORMAL);
7223 iput(inode);
7224 return err;
7225 out:
7226 return dquot_quota_off(sb, type);
7227 }
7228
7229 /* Read data from quotafile - avoid pagecache and such because we cannot afford
7230 * acquiring the locks... As quota files are never truncated and quota code
7231 * itself serializes the operations (and no one else should touch the files)
7232 * we don't have to be afraid of races */
7233 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
7234 size_t len, loff_t off)
7235 {
7236 struct inode *inode = sb_dqopt(sb)->files[type];
7237 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
7238 int offset = off & (sb->s_blocksize - 1);
7239 int tocopy;
7240 size_t toread;
7241 struct buffer_head *bh;
7242 loff_t i_size = i_size_read(inode);
7243
7244 if (off > i_size)
7245 return 0;
7246 if (off+len > i_size)
7247 len = i_size-off;
7248 toread = len;
7249 while (toread > 0) {
7250 tocopy = min_t(unsigned long, sb->s_blocksize - offset, toread);
7251 bh = ext4_bread(NULL, inode, blk, 0);
7252 if (IS_ERR(bh))
7253 return PTR_ERR(bh);
7254 if (!bh) /* A hole? */
7255 memset(data, 0, tocopy);
7256 else
7257 memcpy(data, bh->b_data+offset, tocopy);
7258 brelse(bh);
7259 offset = 0;
7260 toread -= tocopy;
7261 data += tocopy;
7262 blk++;
7263 }
7264 return len;
7265 }
7266
7267 /* Write to quotafile (we know the transaction is already started and has
7268 * enough credits) */
7269 static ssize_t ext4_quota_write(struct super_block *sb, int type,
7270 const char *data, size_t len, loff_t off)
7271 {
7272 struct inode *inode = sb_dqopt(sb)->files[type];
7273 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
7274 int err = 0, err2 = 0, offset = off & (sb->s_blocksize - 1);
7275 int retries = 0;
7276 struct buffer_head *bh;
7277 handle_t *handle = journal_current_handle();
7278
7279 if (!handle) {
7280 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
7281 " cancelled because transaction is not started",
7282 (unsigned long long)off, (unsigned long long)len);
7283 return -EIO;
7284 }
7285 /*
7286 * Since we account only one data block in transaction credits,
7287 * then it is impossible to cross a block boundary.
7288 */
7289 if (sb->s_blocksize - offset < len) {
7290 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
7291 " cancelled because not block aligned",
7292 (unsigned long long)off, (unsigned long long)len);
7293 return -EIO;
7294 }
7295
7296 do {
7297 bh = ext4_bread(handle, inode, blk,
7298 EXT4_GET_BLOCKS_CREATE |
7299 EXT4_GET_BLOCKS_METADATA_NOFAIL);
7300 } while (PTR_ERR(bh) == -ENOSPC &&
7301 ext4_should_retry_alloc(inode->i_sb, &retries));
7302 if (IS_ERR(bh))
7303 return PTR_ERR(bh);
7304 if (!bh)
7305 goto out;
7306 BUFFER_TRACE(bh, "get write access");
7307 err = ext4_journal_get_write_access(handle, sb, bh, EXT4_JTR_NONE);
7308 if (err) {
7309 brelse(bh);
7310 return err;
7311 }
7312 lock_buffer(bh);
7313 memcpy(bh->b_data+offset, data, len);
7314 flush_dcache_folio(bh->b_folio);
7315 unlock_buffer(bh);
7316 err = ext4_handle_dirty_metadata(handle, NULL, bh);
7317 brelse(bh);
7318 out:
7319 if (inode->i_size < off + len) {
7320 i_size_write(inode, off + len);
7321 EXT4_I(inode)->i_disksize = inode->i_size;
7322 err2 = ext4_mark_inode_dirty(handle, inode);
7323 if (unlikely(err2 && !err))
7324 err = err2;
7325 }
7326 return err ? err : len;
7327 }
7328 #endif
7329
7330 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
7331 static inline void register_as_ext2(void)
7332 {
7333 int err = register_filesystem(&ext2_fs_type);
7334 if (err)
7335 printk(KERN_WARNING
7336 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
7337 }
7338
7339 static inline void unregister_as_ext2(void)
7340 {
7341 unregister_filesystem(&ext2_fs_type);
7342 }
7343
7344 static inline int ext2_feature_set_ok(struct super_block *sb)
7345 {
7346 if (ext4_has_unknown_ext2_incompat_features(sb))
7347 return 0;
7348 if (sb_rdonly(sb))
7349 return 1;
7350 if (ext4_has_unknown_ext2_ro_compat_features(sb))
7351 return 0;
7352 return 1;
7353 }
7354 #else
7355 static inline void register_as_ext2(void) { }
7356 static inline void unregister_as_ext2(void) { }
7357 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
7358 #endif
7359
7360 static inline void register_as_ext3(void)
7361 {
7362 int err = register_filesystem(&ext3_fs_type);
7363 if (err)
7364 printk(KERN_WARNING
7365 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
7366 }
7367
7368 static inline void unregister_as_ext3(void)
7369 {
7370 unregister_filesystem(&ext3_fs_type);
7371 }
7372
7373 static inline int ext3_feature_set_ok(struct super_block *sb)
7374 {
7375 if (ext4_has_unknown_ext3_incompat_features(sb))
7376 return 0;
7377 if (!ext4_has_feature_journal(sb))
7378 return 0;
7379 if (sb_rdonly(sb))
7380 return 1;
7381 if (ext4_has_unknown_ext3_ro_compat_features(sb))
7382 return 0;
7383 return 1;
7384 }
7385
7386 static void ext4_kill_sb(struct super_block *sb)
7387 {
7388 struct ext4_sb_info *sbi = EXT4_SB(sb);
7389 struct file *bdev_file = sbi ? sbi->s_journal_bdev_file : NULL;
7390
7391 kill_block_super(sb);
7392
7393 if (bdev_file)
7394 bdev_fput(bdev_file);
7395 }
7396
7397 static struct file_system_type ext4_fs_type = {
7398 .owner = THIS_MODULE,
7399 .name = "ext4",
7400 .init_fs_context = ext4_init_fs_context,
7401 .parameters = ext4_param_specs,
7402 .kill_sb = ext4_kill_sb,
7403 .fs_flags = FS_REQUIRES_DEV | FS_ALLOW_IDMAP | FS_MGTIME,
7404 };
7405 MODULE_ALIAS_FS("ext4");
7406
7407 static int __init ext4_init_fs(void)
7408 {
7409 int err;
7410
7411 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
7412 ext4_li_info = NULL;
7413
7414 /* Build-time check for flags consistency */
7415 ext4_check_flag_values();
7416
7417 err = ext4_init_es();
7418 if (err)
7419 return err;
7420
7421 err = ext4_init_pending();
7422 if (err)
7423 goto out7;
7424
7425 err = ext4_init_post_read_processing();
7426 if (err)
7427 goto out6;
7428
7429 err = ext4_init_pageio();
7430 if (err)
7431 goto out5;
7432
7433 err = ext4_init_system_zone();
7434 if (err)
7435 goto out4;
7436
7437 err = ext4_init_sysfs();
7438 if (err)
7439 goto out3;
7440
7441 err = ext4_init_mballoc();
7442 if (err)
7443 goto out2;
7444 err = init_inodecache();
7445 if (err)
7446 goto out1;
7447
7448 err = ext4_fc_init_dentry_cache();
7449 if (err)
7450 goto out05;
7451
7452 register_as_ext3();
7453 register_as_ext2();
7454 err = register_filesystem(&ext4_fs_type);
7455 if (err)
7456 goto out;
7457
7458 return 0;
7459 out:
7460 unregister_as_ext2();
7461 unregister_as_ext3();
7462 ext4_fc_destroy_dentry_cache();
7463 out05:
7464 destroy_inodecache();
7465 out1:
7466 ext4_exit_mballoc();
7467 out2:
7468 ext4_exit_sysfs();
7469 out3:
7470 ext4_exit_system_zone();
7471 out4:
7472 ext4_exit_pageio();
7473 out5:
7474 ext4_exit_post_read_processing();
7475 out6:
7476 ext4_exit_pending();
7477 out7:
7478 ext4_exit_es();
7479
7480 return err;
7481 }
7482
7483 static void __exit ext4_exit_fs(void)
7484 {
7485 ext4_destroy_lazyinit_thread();
7486 unregister_as_ext2();
7487 unregister_as_ext3();
7488 unregister_filesystem(&ext4_fs_type);
7489 ext4_fc_destroy_dentry_cache();
7490 destroy_inodecache();
7491 ext4_exit_mballoc();
7492 ext4_exit_sysfs();
7493 ext4_exit_system_zone();
7494 ext4_exit_pageio();
7495 ext4_exit_post_read_processing();
7496 ext4_exit_es();
7497 ext4_exit_pending();
7498 }
7499
7500 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
7501 MODULE_DESCRIPTION("Fourth Extended Filesystem");
7502 MODULE_LICENSE("GPL");
7503 module_init(ext4_init_fs)
7504 module_exit(ext4_exit_fs)
A mirror of Linus' kernel repository