?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Merge tag 'regmap-fix-v5.11-rc2' of git://git.kernel.org/pub/scm/linux/kernel/git...
[linux/fpc-iii.git] / fs / ext4 / super.c
blob21121787c874e186ed547177adcffdebbd3891a6
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/cleancache.h>
43 #include <linux/uaccess.h>
44 #include <linux/iversion.h>
45 #include <linux/unicode.h>
46 #include <linux/part_stat.h>
47 #include <linux/kthread.h>
48 #include <linux/freezer.h>
49
50 #include "ext4.h"
51 #include "ext4_extents.h" /* Needed for trace points definition */
52 #include "ext4_jbd2.h"
53 #include "xattr.h"
54 #include "acl.h"
55 #include "mballoc.h"
56 #include "fsmap.h"
57
58 #define CREATE_TRACE_POINTS
59 #include <trace/events/ext4.h>
60
61 static struct ext4_lazy_init *ext4_li_info;
62 static struct mutex ext4_li_mtx;
63 static struct ratelimit_state ext4_mount_msg_ratelimit;
64
65 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
66 unsigned long journal_devnum);
67 static int ext4_show_options(struct seq_file *seq, struct dentry *root);
68 static int ext4_commit_super(struct super_block *sb, int sync);
69 static int ext4_mark_recovery_complete(struct super_block *sb,
70 struct ext4_super_block *es);
71 static int ext4_clear_journal_err(struct super_block *sb,
72 struct ext4_super_block *es);
73 static int ext4_sync_fs(struct super_block *sb, int wait);
74 static int ext4_remount(struct super_block *sb, int *flags, char *data);
75 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
76 static int ext4_unfreeze(struct super_block *sb);
77 static int ext4_freeze(struct super_block *sb);
78 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
79 const char *dev_name, void *data);
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 int ext4_feature_set_ok(struct super_block *sb, int readonly);
83 static void ext4_destroy_lazyinit_thread(void);
84 static void ext4_unregister_li_request(struct super_block *sb);
85 static void ext4_clear_request_list(void);
86 static struct inode *ext4_get_journal_inode(struct super_block *sb,
87 unsigned int journal_inum);
88
89 /*
90 * Lock ordering
91 *
92 * Note the difference between i_mmap_sem (EXT4_I(inode)->i_mmap_sem) and
93 * i_mmap_rwsem (inode->i_mmap_rwsem)!
94 *
95 * page fault path:
96 * mmap_lock -> sb_start_pagefault -> i_mmap_sem (r) -> transaction start ->
97 * page lock -> i_data_sem (rw)
98 *
99 * buffered write path:
100 * sb_start_write -> i_mutex -> mmap_lock
101 * sb_start_write -> i_mutex -> transaction start -> page lock ->
102 * i_data_sem (rw)
103 *
104 * truncate:
105 * sb_start_write -> i_mutex -> i_mmap_sem (w) -> i_mmap_rwsem (w) -> page lock
106 * sb_start_write -> i_mutex -> i_mmap_sem (w) -> transaction start ->
107 * i_data_sem (rw)
108 *
109 * direct IO:
110 * sb_start_write -> i_mutex -> mmap_lock
111 * sb_start_write -> i_mutex -> transaction start -> i_data_sem (rw)
112 *
113 * writepages:
114 * transaction start -> page lock(s) -> i_data_sem (rw)
115 */
116
117 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
118 static struct file_system_type ext2_fs_type = {
119 .owner = THIS_MODULE,
120 .name = "ext2",
121 .mount = ext4_mount,
122 .kill_sb = kill_block_super,
123 .fs_flags = FS_REQUIRES_DEV,
124 };
125 MODULE_ALIAS_FS("ext2");
126 MODULE_ALIAS("ext2");
127 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
128 #else
129 #define IS_EXT2_SB(sb) (0)
130 #endif
131
132
133 static struct file_system_type ext3_fs_type = {
134 .owner = THIS_MODULE,
135 .name = "ext3",
136 .mount = ext4_mount,
137 .kill_sb = kill_block_super,
138 .fs_flags = FS_REQUIRES_DEV,
139 };
140 MODULE_ALIAS_FS("ext3");
141 MODULE_ALIAS("ext3");
142 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
143
144
145 static inline void __ext4_read_bh(struct buffer_head *bh, int op_flags,
146 bh_end_io_t *end_io)
147 {
148 /*
149 * buffer's verified bit is no longer valid after reading from
150 * disk again due to write out error, clear it to make sure we
151 * recheck the buffer contents.
152 */
153 clear_buffer_verified(bh);
154
155 bh->b_end_io = end_io ? end_io : end_buffer_read_sync;
156 get_bh(bh);
157 submit_bh(REQ_OP_READ, op_flags, bh);
158 }
159
160 void ext4_read_bh_nowait(struct buffer_head *bh, int op_flags,
161 bh_end_io_t *end_io)
162 {
163 BUG_ON(!buffer_locked(bh));
164
165 if (ext4_buffer_uptodate(bh)) {
166 unlock_buffer(bh);
167 return;
168 }
169 __ext4_read_bh(bh, op_flags, end_io);
170 }
171
172 int ext4_read_bh(struct buffer_head *bh, int op_flags, bh_end_io_t *end_io)
173 {
174 BUG_ON(!buffer_locked(bh));
175
176 if (ext4_buffer_uptodate(bh)) {
177 unlock_buffer(bh);
178 return 0;
179 }
180
181 __ext4_read_bh(bh, op_flags, end_io);
182
183 wait_on_buffer(bh);
184 if (buffer_uptodate(bh))
185 return 0;
186 return -EIO;
187 }
188
189 int ext4_read_bh_lock(struct buffer_head *bh, int op_flags, bool wait)
190 {
191 if (trylock_buffer(bh)) {
192 if (wait)
193 return ext4_read_bh(bh, op_flags, NULL);
194 ext4_read_bh_nowait(bh, op_flags, NULL);
195 return 0;
196 }
197 if (wait) {
198 wait_on_buffer(bh);
199 if (buffer_uptodate(bh))
200 return 0;
201 return -EIO;
202 }
203 return 0;
204 }
205
206 /*
207 * This works like __bread_gfp() except it uses ERR_PTR for error
208 * returns. Currently with sb_bread it's impossible to distinguish
209 * between ENOMEM and EIO situations (since both result in a NULL
210 * return.
211 */
212 static struct buffer_head *__ext4_sb_bread_gfp(struct super_block *sb,
213 sector_t block, int op_flags,
214 gfp_t gfp)
215 {
216 struct buffer_head *bh;
217 int ret;
218
219 bh = sb_getblk_gfp(sb, block, gfp);
220 if (bh == NULL)
221 return ERR_PTR(-ENOMEM);
222 if (ext4_buffer_uptodate(bh))
223 return bh;
224
225 ret = ext4_read_bh_lock(bh, REQ_META | op_flags, true);
226 if (ret) {
227 put_bh(bh);
228 return ERR_PTR(ret);
229 }
230 return bh;
231 }
232
233 struct buffer_head *ext4_sb_bread(struct super_block *sb, sector_t block,
234 int op_flags)
235 {
236 return __ext4_sb_bread_gfp(sb, block, op_flags, __GFP_MOVABLE);
237 }
238
239 struct buffer_head *ext4_sb_bread_unmovable(struct super_block *sb,
240 sector_t block)
241 {
242 return __ext4_sb_bread_gfp(sb, block, 0, 0);
243 }
244
245 void ext4_sb_breadahead_unmovable(struct super_block *sb, sector_t block)
246 {
247 struct buffer_head *bh = sb_getblk_gfp(sb, block, 0);
248
249 if (likely(bh)) {
250 ext4_read_bh_lock(bh, REQ_RAHEAD, false);
251 brelse(bh);
252 }
253 }
254
255 static int ext4_verify_csum_type(struct super_block *sb,
256 struct ext4_super_block *es)
257 {
258 if (!ext4_has_feature_metadata_csum(sb))
259 return 1;
260
261 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
262 }
263
264 static __le32 ext4_superblock_csum(struct super_block *sb,
265 struct ext4_super_block *es)
266 {
267 struct ext4_sb_info *sbi = EXT4_SB(sb);
268 int offset = offsetof(struct ext4_super_block, s_checksum);
269 __u32 csum;
270
271 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
272
273 return cpu_to_le32(csum);
274 }
275
276 static int ext4_superblock_csum_verify(struct super_block *sb,
277 struct ext4_super_block *es)
278 {
279 if (!ext4_has_metadata_csum(sb))
280 return 1;
281
282 return es->s_checksum == ext4_superblock_csum(sb, es);
283 }
284
285 void ext4_superblock_csum_set(struct super_block *sb)
286 {
287 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
288
289 if (!ext4_has_metadata_csum(sb))
290 return;
291
292 es->s_checksum = ext4_superblock_csum(sb, es);
293 }
294
295 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
296 struct ext4_group_desc *bg)
297 {
298 return le32_to_cpu(bg->bg_block_bitmap_lo) |
299 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
300 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
301 }
302
303 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
304 struct ext4_group_desc *bg)
305 {
306 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
307 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
308 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
309 }
310
311 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
312 struct ext4_group_desc *bg)
313 {
314 return le32_to_cpu(bg->bg_inode_table_lo) |
315 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
316 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
317 }
318
319 __u32 ext4_free_group_clusters(struct super_block *sb,
320 struct ext4_group_desc *bg)
321 {
322 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
323 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
324 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
325 }
326
327 __u32 ext4_free_inodes_count(struct super_block *sb,
328 struct ext4_group_desc *bg)
329 {
330 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
331 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
332 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
333 }
334
335 __u32 ext4_used_dirs_count(struct super_block *sb,
336 struct ext4_group_desc *bg)
337 {
338 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
339 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
340 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
341 }
342
343 __u32 ext4_itable_unused_count(struct super_block *sb,
344 struct ext4_group_desc *bg)
345 {
346 return le16_to_cpu(bg->bg_itable_unused_lo) |
347 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
348 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
349 }
350
351 void ext4_block_bitmap_set(struct super_block *sb,
352 struct ext4_group_desc *bg, ext4_fsblk_t blk)
353 {
354 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
355 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
356 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
357 }
358
359 void ext4_inode_bitmap_set(struct super_block *sb,
360 struct ext4_group_desc *bg, ext4_fsblk_t blk)
361 {
362 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
363 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
364 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
365 }
366
367 void ext4_inode_table_set(struct super_block *sb,
368 struct ext4_group_desc *bg, ext4_fsblk_t blk)
369 {
370 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
371 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
372 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
373 }
374
375 void ext4_free_group_clusters_set(struct super_block *sb,
376 struct ext4_group_desc *bg, __u32 count)
377 {
378 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
379 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
380 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
381 }
382
383 void ext4_free_inodes_set(struct super_block *sb,
384 struct ext4_group_desc *bg, __u32 count)
385 {
386 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
387 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
388 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
389 }
390
391 void ext4_used_dirs_set(struct super_block *sb,
392 struct ext4_group_desc *bg, __u32 count)
393 {
394 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
395 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
396 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
397 }
398
399 void ext4_itable_unused_set(struct super_block *sb,
400 struct ext4_group_desc *bg, __u32 count)
401 {
402 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
403 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
404 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
405 }
406
407 static void __ext4_update_tstamp(__le32 *lo, __u8 *hi, time64_t now)
408 {
409 now = clamp_val(now, 0, (1ull << 40) - 1);
410
411 *lo = cpu_to_le32(lower_32_bits(now));
412 *hi = upper_32_bits(now);
413 }
414
415 static time64_t __ext4_get_tstamp(__le32 *lo, __u8 *hi)
416 {
417 return ((time64_t)(*hi) << 32) + le32_to_cpu(*lo);
418 }
419 #define ext4_update_tstamp(es, tstamp) \
420 __ext4_update_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi, \
421 ktime_get_real_seconds())
422 #define ext4_get_tstamp(es, tstamp) \
423 __ext4_get_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi)
424
425 /*
426 * The del_gendisk() function uninitializes the disk-specific data
427 * structures, including the bdi structure, without telling anyone
428 * else. Once this happens, any attempt to call mark_buffer_dirty()
429 * (for example, by ext4_commit_super), will cause a kernel OOPS.
430 * This is a kludge to prevent these oops until we can put in a proper
431 * hook in del_gendisk() to inform the VFS and file system layers.
432 */
433 static int block_device_ejected(struct super_block *sb)
434 {
435 struct inode *bd_inode = sb->s_bdev->bd_inode;
436 struct backing_dev_info *bdi = inode_to_bdi(bd_inode);
437
438 return bdi->dev == NULL;
439 }
440
441 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
442 {
443 struct super_block *sb = journal->j_private;
444 struct ext4_sb_info *sbi = EXT4_SB(sb);
445 int error = is_journal_aborted(journal);
446 struct ext4_journal_cb_entry *jce;
447
448 BUG_ON(txn->t_state == T_FINISHED);
449
450 ext4_process_freed_data(sb, txn->t_tid);
451
452 spin_lock(&sbi->s_md_lock);
453 while (!list_empty(&txn->t_private_list)) {
454 jce = list_entry(txn->t_private_list.next,
455 struct ext4_journal_cb_entry, jce_list);
456 list_del_init(&jce->jce_list);
457 spin_unlock(&sbi->s_md_lock);
458 jce->jce_func(sb, jce, error);
459 spin_lock(&sbi->s_md_lock);
460 }
461 spin_unlock(&sbi->s_md_lock);
462 }
463
464 /*
465 * This writepage callback for write_cache_pages()
466 * takes care of a few cases after page cleaning.
467 *
468 * write_cache_pages() already checks for dirty pages
469 * and calls clear_page_dirty_for_io(), which we want,
470 * to write protect the pages.
471 *
472 * However, we may have to redirty a page (see below.)
473 */
474 static int ext4_journalled_writepage_callback(struct page *page,
475 struct writeback_control *wbc,
476 void *data)
477 {
478 transaction_t *transaction = (transaction_t *) data;
479 struct buffer_head *bh, *head;
480 struct journal_head *jh;
481
482 bh = head = page_buffers(page);
483 do {
484 /*
485 * We have to redirty a page in these cases:
486 * 1) If buffer is dirty, it means the page was dirty because it
487 * contains a buffer that needs checkpointing. So the dirty bit
488 * needs to be preserved so that checkpointing writes the buffer
489 * properly.
490 * 2) If buffer is not part of the committing transaction
491 * (we may have just accidentally come across this buffer because
492 * inode range tracking is not exact) or if the currently running
493 * transaction already contains this buffer as well, dirty bit
494 * needs to be preserved so that the buffer gets writeprotected
495 * properly on running transaction's commit.
496 */
497 jh = bh2jh(bh);
498 if (buffer_dirty(bh) ||
499 (jh && (jh->b_transaction != transaction ||
500 jh->b_next_transaction))) {
501 redirty_page_for_writepage(wbc, page);
502 goto out;
503 }
504 } while ((bh = bh->b_this_page) != head);
505
506 out:
507 return AOP_WRITEPAGE_ACTIVATE;
508 }
509
510 static int ext4_journalled_submit_inode_data_buffers(struct jbd2_inode *jinode)
511 {
512 struct address_space *mapping = jinode->i_vfs_inode->i_mapping;
513 struct writeback_control wbc = {
514 .sync_mode = WB_SYNC_ALL,
515 .nr_to_write = LONG_MAX,
516 .range_start = jinode->i_dirty_start,
517 .range_end = jinode->i_dirty_end,
518 };
519
520 return write_cache_pages(mapping, &wbc,
521 ext4_journalled_writepage_callback,
522 jinode->i_transaction);
523 }
524
525 static int ext4_journal_submit_inode_data_buffers(struct jbd2_inode *jinode)
526 {
527 int ret;
528
529 if (ext4_should_journal_data(jinode->i_vfs_inode))
530 ret = ext4_journalled_submit_inode_data_buffers(jinode);
531 else
532 ret = jbd2_journal_submit_inode_data_buffers(jinode);
533
534 return ret;
535 }
536
537 static int ext4_journal_finish_inode_data_buffers(struct jbd2_inode *jinode)
538 {
539 int ret = 0;
540
541 if (!ext4_should_journal_data(jinode->i_vfs_inode))
542 ret = jbd2_journal_finish_inode_data_buffers(jinode);
543
544 return ret;
545 }
546
547 static bool system_going_down(void)
548 {
549 return system_state == SYSTEM_HALT || system_state == SYSTEM_POWER_OFF
550 || system_state == SYSTEM_RESTART;
551 }
552
553 struct ext4_err_translation {
554 int code;
555 int errno;
556 };
557
558 #define EXT4_ERR_TRANSLATE(err) { .code = EXT4_ERR_##err, .errno = err }
559
560 static struct ext4_err_translation err_translation[] = {
561 EXT4_ERR_TRANSLATE(EIO),
562 EXT4_ERR_TRANSLATE(ENOMEM),
563 EXT4_ERR_TRANSLATE(EFSBADCRC),
564 EXT4_ERR_TRANSLATE(EFSCORRUPTED),
565 EXT4_ERR_TRANSLATE(ENOSPC),
566 EXT4_ERR_TRANSLATE(ENOKEY),
567 EXT4_ERR_TRANSLATE(EROFS),
568 EXT4_ERR_TRANSLATE(EFBIG),
569 EXT4_ERR_TRANSLATE(EEXIST),
570 EXT4_ERR_TRANSLATE(ERANGE),
571 EXT4_ERR_TRANSLATE(EOVERFLOW),
572 EXT4_ERR_TRANSLATE(EBUSY),
573 EXT4_ERR_TRANSLATE(ENOTDIR),
574 EXT4_ERR_TRANSLATE(ENOTEMPTY),
575 EXT4_ERR_TRANSLATE(ESHUTDOWN),
576 EXT4_ERR_TRANSLATE(EFAULT),
577 };
578
579 static int ext4_errno_to_code(int errno)
580 {
581 int i;
582
583 for (i = 0; i < ARRAY_SIZE(err_translation); i++)
584 if (err_translation[i].errno == errno)
585 return err_translation[i].code;
586 return EXT4_ERR_UNKNOWN;
587 }
588
589 static void __save_error_info(struct super_block *sb, int error,
590 __u32 ino, __u64 block,
591 const char *func, unsigned int line)
592 {
593 struct ext4_sb_info *sbi = EXT4_SB(sb);
594
595 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
596 if (bdev_read_only(sb->s_bdev))
597 return;
598 /* We default to EFSCORRUPTED error... */
599 if (error == 0)
600 error = EFSCORRUPTED;
601
602 spin_lock(&sbi->s_error_lock);
603 sbi->s_add_error_count++;
604 sbi->s_last_error_code = error;
605 sbi->s_last_error_line = line;
606 sbi->s_last_error_ino = ino;
607 sbi->s_last_error_block = block;
608 sbi->s_last_error_func = func;
609 sbi->s_last_error_time = ktime_get_real_seconds();
610 if (!sbi->s_first_error_time) {
611 sbi->s_first_error_code = error;
612 sbi->s_first_error_line = line;
613 sbi->s_first_error_ino = ino;
614 sbi->s_first_error_block = block;
615 sbi->s_first_error_func = func;
616 sbi->s_first_error_time = sbi->s_last_error_time;
617 }
618 spin_unlock(&sbi->s_error_lock);
619 }
620
621 static void save_error_info(struct super_block *sb, int error,
622 __u32 ino, __u64 block,
623 const char *func, unsigned int line)
624 {
625 __save_error_info(sb, error, ino, block, func, line);
626 if (!bdev_read_only(sb->s_bdev))
627 ext4_commit_super(sb, 1);
628 }
629
630 /* Deal with the reporting of failure conditions on a filesystem such as
631 * inconsistencies detected or read IO failures.
632 *
633 * On ext2, we can store the error state of the filesystem in the
634 * superblock. That is not possible on ext4, because we may have other
635 * write ordering constraints on the superblock which prevent us from
636 * writing it out straight away; and given that the journal is about to
637 * be aborted, we can't rely on the current, or future, transactions to
638 * write out the superblock safely.
639 *
640 * We'll just use the jbd2_journal_abort() error code to record an error in
641 * the journal instead. On recovery, the journal will complain about
642 * that error until we've noted it down and cleared it.
643 *
644 * If force_ro is set, we unconditionally force the filesystem into an
645 * ABORT|READONLY state, unless the error response on the fs has been set to
646 * panic in which case we take the easy way out and panic immediately. This is
647 * used to deal with unrecoverable failures such as journal IO errors or ENOMEM
648 * at a critical moment in log management.
649 */
650 static void ext4_handle_error(struct super_block *sb, bool force_ro)
651 {
652 journal_t *journal = EXT4_SB(sb)->s_journal;
653
654 if (test_opt(sb, WARN_ON_ERROR))
655 WARN_ON_ONCE(1);
656
657 if (sb_rdonly(sb) || (!force_ro && test_opt(sb, ERRORS_CONT)))
658 return;
659
660 ext4_set_mount_flag(sb, EXT4_MF_FS_ABORTED);
661 if (journal)
662 jbd2_journal_abort(journal, -EIO);
663 /*
664 * We force ERRORS_RO behavior when system is rebooting. Otherwise we
665 * could panic during 'reboot -f' as the underlying device got already
666 * disabled.
667 */
668 if (test_opt(sb, ERRORS_PANIC) && !system_going_down()) {
669 panic("EXT4-fs (device %s): panic forced after error\n",
670 sb->s_id);
671 }
672 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
673 /*
674 * Make sure updated value of ->s_mount_flags will be visible before
675 * ->s_flags update
676 */
677 smp_wmb();
678 sb->s_flags |= SB_RDONLY;
679 }
680
681 static void flush_stashed_error_work(struct work_struct *work)
682 {
683 struct ext4_sb_info *sbi = container_of(work, struct ext4_sb_info,
684 s_error_work);
685
686 ext4_commit_super(sbi->s_sb, 1);
687 }
688
689 #define ext4_error_ratelimit(sb) \
690 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
691 "EXT4-fs error")
692
693 void __ext4_error(struct super_block *sb, const char *function,
694 unsigned int line, bool force_ro, int error, __u64 block,
695 const char *fmt, ...)
696 {
697 struct va_format vaf;
698 va_list args;
699
700 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
701 return;
702
703 trace_ext4_error(sb, function, line);
704 if (ext4_error_ratelimit(sb)) {
705 va_start(args, fmt);
706 vaf.fmt = fmt;
707 vaf.va = &args;
708 printk(KERN_CRIT
709 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
710 sb->s_id, function, line, current->comm, &vaf);
711 va_end(args);
712 }
713 save_error_info(sb, error, 0, block, function, line);
714 ext4_handle_error(sb, force_ro);
715 }
716
717 void __ext4_error_inode(struct inode *inode, const char *function,
718 unsigned int line, ext4_fsblk_t block, int error,
719 const char *fmt, ...)
720 {
721 va_list args;
722 struct va_format vaf;
723
724 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
725 return;
726
727 trace_ext4_error(inode->i_sb, function, line);
728 if (ext4_error_ratelimit(inode->i_sb)) {
729 va_start(args, fmt);
730 vaf.fmt = fmt;
731 vaf.va = &args;
732 if (block)
733 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
734 "inode #%lu: block %llu: comm %s: %pV\n",
735 inode->i_sb->s_id, function, line, inode->i_ino,
736 block, current->comm, &vaf);
737 else
738 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
739 "inode #%lu: comm %s: %pV\n",
740 inode->i_sb->s_id, function, line, inode->i_ino,
741 current->comm, &vaf);
742 va_end(args);
743 }
744 save_error_info(inode->i_sb, error, inode->i_ino, block,
745 function, line);
746 ext4_handle_error(inode->i_sb, false);
747 }
748
749 void __ext4_error_file(struct file *file, const char *function,
750 unsigned int line, ext4_fsblk_t block,
751 const char *fmt, ...)
752 {
753 va_list args;
754 struct va_format vaf;
755 struct inode *inode = file_inode(file);
756 char pathname[80], *path;
757
758 if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
759 return;
760
761 trace_ext4_error(inode->i_sb, function, line);
762 if (ext4_error_ratelimit(inode->i_sb)) {
763 path = file_path(file, pathname, sizeof(pathname));
764 if (IS_ERR(path))
765 path = "(unknown)";
766 va_start(args, fmt);
767 vaf.fmt = fmt;
768 vaf.va = &args;
769 if (block)
770 printk(KERN_CRIT
771 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
772 "block %llu: comm %s: path %s: %pV\n",
773 inode->i_sb->s_id, function, line, inode->i_ino,
774 block, current->comm, path, &vaf);
775 else
776 printk(KERN_CRIT
777 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
778 "comm %s: path %s: %pV\n",
779 inode->i_sb->s_id, function, line, inode->i_ino,
780 current->comm, path, &vaf);
781 va_end(args);
782 }
783 save_error_info(inode->i_sb, EFSCORRUPTED, inode->i_ino, block,
784 function, line);
785 ext4_handle_error(inode->i_sb, false);
786 }
787
788 const char *ext4_decode_error(struct super_block *sb, int errno,
789 char nbuf[16])
790 {
791 char *errstr = NULL;
792
793 switch (errno) {
794 case -EFSCORRUPTED:
795 errstr = "Corrupt filesystem";
796 break;
797 case -EFSBADCRC:
798 errstr = "Filesystem failed CRC";
799 break;
800 case -EIO:
801 errstr = "IO failure";
802 break;
803 case -ENOMEM:
804 errstr = "Out of memory";
805 break;
806 case -EROFS:
807 if (!sb || (EXT4_SB(sb)->s_journal &&
808 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
809 errstr = "Journal has aborted";
810 else
811 errstr = "Readonly filesystem";
812 break;
813 default:
814 /* If the caller passed in an extra buffer for unknown
815 * errors, textualise them now. Else we just return
816 * NULL. */
817 if (nbuf) {
818 /* Check for truncated error codes... */
819 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
820 errstr = nbuf;
821 }
822 break;
823 }
824
825 return errstr;
826 }
827
828 /* __ext4_std_error decodes expected errors from journaling functions
829 * automatically and invokes the appropriate error response. */
830
831 void __ext4_std_error(struct super_block *sb, const char *function,
832 unsigned int line, int errno)
833 {
834 char nbuf[16];
835 const char *errstr;
836
837 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
838 return;
839
840 /* Special case: if the error is EROFS, and we're not already
841 * inside a transaction, then there's really no point in logging
842 * an error. */
843 if (errno == -EROFS && journal_current_handle() == NULL && sb_rdonly(sb))
844 return;
845
846 if (ext4_error_ratelimit(sb)) {
847 errstr = ext4_decode_error(sb, errno, nbuf);
848 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
849 sb->s_id, function, line, errstr);
850 }
851
852 save_error_info(sb, -errno, 0, 0, function, line);
853 ext4_handle_error(sb, false);
854 }
855
856 void __ext4_msg(struct super_block *sb,
857 const char *prefix, const char *fmt, ...)
858 {
859 struct va_format vaf;
860 va_list args;
861
862 atomic_inc(&EXT4_SB(sb)->s_msg_count);
863 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
864 return;
865
866 va_start(args, fmt);
867 vaf.fmt = fmt;
868 vaf.va = &args;
869 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
870 va_end(args);
871 }
872
873 static int ext4_warning_ratelimit(struct super_block *sb)
874 {
875 atomic_inc(&EXT4_SB(sb)->s_warning_count);
876 return ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state),
877 "EXT4-fs warning");
878 }
879
880 void __ext4_warning(struct super_block *sb, const char *function,
881 unsigned int line, const char *fmt, ...)
882 {
883 struct va_format vaf;
884 va_list args;
885
886 if (!ext4_warning_ratelimit(sb))
887 return;
888
889 va_start(args, fmt);
890 vaf.fmt = fmt;
891 vaf.va = &args;
892 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
893 sb->s_id, function, line, &vaf);
894 va_end(args);
895 }
896
897 void __ext4_warning_inode(const struct inode *inode, const char *function,
898 unsigned int line, const char *fmt, ...)
899 {
900 struct va_format vaf;
901 va_list args;
902
903 if (!ext4_warning_ratelimit(inode->i_sb))
904 return;
905
906 va_start(args, fmt);
907 vaf.fmt = fmt;
908 vaf.va = &args;
909 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
910 "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
911 function, line, inode->i_ino, current->comm, &vaf);
912 va_end(args);
913 }
914
915 void __ext4_grp_locked_error(const char *function, unsigned int line,
916 struct super_block *sb, ext4_group_t grp,
917 unsigned long ino, ext4_fsblk_t block,
918 const char *fmt, ...)
919 __releases(bitlock)
920 __acquires(bitlock)
921 {
922 struct va_format vaf;
923 va_list args;
924
925 if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
926 return;
927
928 trace_ext4_error(sb, function, line);
929 if (ext4_error_ratelimit(sb)) {
930 va_start(args, fmt);
931 vaf.fmt = fmt;
932 vaf.va = &args;
933 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
934 sb->s_id, function, line, grp);
935 if (ino)
936 printk(KERN_CONT "inode %lu: ", ino);
937 if (block)
938 printk(KERN_CONT "block %llu:",
939 (unsigned long long) block);
940 printk(KERN_CONT "%pV\n", &vaf);
941 va_end(args);
942 }
943
944 if (test_opt(sb, ERRORS_CONT)) {
945 if (test_opt(sb, WARN_ON_ERROR))
946 WARN_ON_ONCE(1);
947 __save_error_info(sb, EFSCORRUPTED, ino, block, function, line);
948 schedule_work(&EXT4_SB(sb)->s_error_work);
949 return;
950 }
951 ext4_unlock_group(sb, grp);
952 save_error_info(sb, EFSCORRUPTED, ino, block, function, line);
953 ext4_handle_error(sb, false);
954 /*
955 * We only get here in the ERRORS_RO case; relocking the group
956 * may be dangerous, but nothing bad will happen since the
957 * filesystem will have already been marked read/only and the
958 * journal has been aborted. We return 1 as a hint to callers
959 * who might what to use the return value from
960 * ext4_grp_locked_error() to distinguish between the
961 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
962 * aggressively from the ext4 function in question, with a
963 * more appropriate error code.
964 */
965 ext4_lock_group(sb, grp);
966 return;
967 }
968
969 void ext4_mark_group_bitmap_corrupted(struct super_block *sb,
970 ext4_group_t group,
971 unsigned int flags)
972 {
973 struct ext4_sb_info *sbi = EXT4_SB(sb);
974 struct ext4_group_info *grp = ext4_get_group_info(sb, group);
975 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, group, NULL);
976 int ret;
977
978 if (flags & EXT4_GROUP_INFO_BBITMAP_CORRUPT) {
979 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT,
980 &grp->bb_state);
981 if (!ret)
982 percpu_counter_sub(&sbi->s_freeclusters_counter,
983 grp->bb_free);
984 }
985
986 if (flags & EXT4_GROUP_INFO_IBITMAP_CORRUPT) {
987 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT,
988 &grp->bb_state);
989 if (!ret && gdp) {
990 int count;
991
992 count = ext4_free_inodes_count(sb, gdp);
993 percpu_counter_sub(&sbi->s_freeinodes_counter,
994 count);
995 }
996 }
997 }
998
999 void ext4_update_dynamic_rev(struct super_block *sb)
1000 {
1001 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
1002
1003 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
1004 return;
1005
1006 ext4_warning(sb,
1007 "updating to rev %d because of new feature flag, "
1008 "running e2fsck is recommended",
1009 EXT4_DYNAMIC_REV);
1010
1011 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
1012 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
1013 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
1014 /* leave es->s_feature_*compat flags alone */
1015 /* es->s_uuid will be set by e2fsck if empty */
1016
1017 /*
1018 * The rest of the superblock fields should be zero, and if not it
1019 * means they are likely already in use, so leave them alone. We
1020 * can leave it up to e2fsck to clean up any inconsistencies there.
1021 */
1022 }
1023
1024 /*
1025 * Open the external journal device
1026 */
1027 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
1028 {
1029 struct block_device *bdev;
1030
1031 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
1032 if (IS_ERR(bdev))
1033 goto fail;
1034 return bdev;
1035
1036 fail:
1037 ext4_msg(sb, KERN_ERR,
1038 "failed to open journal device unknown-block(%u,%u) %ld",
1039 MAJOR(dev), MINOR(dev), PTR_ERR(bdev));
1040 return NULL;
1041 }
1042
1043 /*
1044 * Release the journal device
1045 */
1046 static void ext4_blkdev_put(struct block_device *bdev)
1047 {
1048 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
1049 }
1050
1051 static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
1052 {
1053 struct block_device *bdev;
1054 bdev = sbi->s_journal_bdev;
1055 if (bdev) {
1056 ext4_blkdev_put(bdev);
1057 sbi->s_journal_bdev = NULL;
1058 }
1059 }
1060
1061 static inline struct inode *orphan_list_entry(struct list_head *l)
1062 {
1063 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
1064 }
1065
1066 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
1067 {
1068 struct list_head *l;
1069
1070 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
1071 le32_to_cpu(sbi->s_es->s_last_orphan));
1072
1073 printk(KERN_ERR "sb_info orphan list:\n");
1074 list_for_each(l, &sbi->s_orphan) {
1075 struct inode *inode = orphan_list_entry(l);
1076 printk(KERN_ERR " "
1077 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
1078 inode->i_sb->s_id, inode->i_ino, inode,
1079 inode->i_mode, inode->i_nlink,
1080 NEXT_ORPHAN(inode));
1081 }
1082 }
1083
1084 #ifdef CONFIG_QUOTA
1085 static int ext4_quota_off(struct super_block *sb, int type);
1086
1087 static inline void ext4_quota_off_umount(struct super_block *sb)
1088 {
1089 int type;
1090
1091 /* Use our quota_off function to clear inode flags etc. */
1092 for (type = 0; type < EXT4_MAXQUOTAS; type++)
1093 ext4_quota_off(sb, type);
1094 }
1095
1096 /*
1097 * This is a helper function which is used in the mount/remount
1098 * codepaths (which holds s_umount) to fetch the quota file name.
1099 */
1100 static inline char *get_qf_name(struct super_block *sb,
1101 struct ext4_sb_info *sbi,
1102 int type)
1103 {
1104 return rcu_dereference_protected(sbi->s_qf_names[type],
1105 lockdep_is_held(&sb->s_umount));
1106 }
1107 #else
1108 static inline void ext4_quota_off_umount(struct super_block *sb)
1109 {
1110 }
1111 #endif
1112
1113 static void ext4_put_super(struct super_block *sb)
1114 {
1115 struct ext4_sb_info *sbi = EXT4_SB(sb);
1116 struct ext4_super_block *es = sbi->s_es;
1117 struct buffer_head **group_desc;
1118 struct flex_groups **flex_groups;
1119 int aborted = 0;
1120 int i, err;
1121
1122 ext4_unregister_li_request(sb);
1123 ext4_quota_off_umount(sb);
1124
1125 flush_work(&sbi->s_error_work);
1126 destroy_workqueue(sbi->rsv_conversion_wq);
1127
1128 /*
1129 * Unregister sysfs before destroying jbd2 journal.
1130 * Since we could still access attr_journal_task attribute via sysfs
1131 * path which could have sbi->s_journal->j_task as NULL
1132 */
1133 ext4_unregister_sysfs(sb);
1134
1135 if (sbi->s_journal) {
1136 aborted = is_journal_aborted(sbi->s_journal);
1137 err = jbd2_journal_destroy(sbi->s_journal);
1138 sbi->s_journal = NULL;
1139 if ((err < 0) && !aborted) {
1140 ext4_abort(sb, -err, "Couldn't clean up the journal");
1141 }
1142 }
1143
1144 ext4_es_unregister_shrinker(sbi);
1145 del_timer_sync(&sbi->s_err_report);
1146 ext4_release_system_zone(sb);
1147 ext4_mb_release(sb);
1148 ext4_ext_release(sb);
1149
1150 if (!sb_rdonly(sb) && !aborted) {
1151 ext4_clear_feature_journal_needs_recovery(sb);
1152 es->s_state = cpu_to_le16(sbi->s_mount_state);
1153 }
1154 if (!sb_rdonly(sb))
1155 ext4_commit_super(sb, 1);
1156
1157 rcu_read_lock();
1158 group_desc = rcu_dereference(sbi->s_group_desc);
1159 for (i = 0; i < sbi->s_gdb_count; i++)
1160 brelse(group_desc[i]);
1161 kvfree(group_desc);
1162 flex_groups = rcu_dereference(sbi->s_flex_groups);
1163 if (flex_groups) {
1164 for (i = 0; i < sbi->s_flex_groups_allocated; i++)
1165 kvfree(flex_groups[i]);
1166 kvfree(flex_groups);
1167 }
1168 rcu_read_unlock();
1169 percpu_counter_destroy(&sbi->s_freeclusters_counter);
1170 percpu_counter_destroy(&sbi->s_freeinodes_counter);
1171 percpu_counter_destroy(&sbi->s_dirs_counter);
1172 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
1173 percpu_free_rwsem(&sbi->s_writepages_rwsem);
1174 #ifdef CONFIG_QUOTA
1175 for (i = 0; i < EXT4_MAXQUOTAS; i++)
1176 kfree(get_qf_name(sb, sbi, i));
1177 #endif
1178
1179 /* Debugging code just in case the in-memory inode orphan list
1180 * isn't empty. The on-disk one can be non-empty if we've
1181 * detected an error and taken the fs readonly, but the
1182 * in-memory list had better be clean by this point. */
1183 if (!list_empty(&sbi->s_orphan))
1184 dump_orphan_list(sb, sbi);
1185 ASSERT(list_empty(&sbi->s_orphan));
1186
1187 sync_blockdev(sb->s_bdev);
1188 invalidate_bdev(sb->s_bdev);
1189 if (sbi->s_journal_bdev && sbi->s_journal_bdev != sb->s_bdev) {
1190 /*
1191 * Invalidate the journal device's buffers. We don't want them
1192 * floating about in memory - the physical journal device may
1193 * hotswapped, and it breaks the `ro-after' testing code.
1194 */
1195 sync_blockdev(sbi->s_journal_bdev);
1196 invalidate_bdev(sbi->s_journal_bdev);
1197 ext4_blkdev_remove(sbi);
1198 }
1199
1200 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
1201 sbi->s_ea_inode_cache = NULL;
1202
1203 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
1204 sbi->s_ea_block_cache = NULL;
1205
1206 if (sbi->s_mmp_tsk)
1207 kthread_stop(sbi->s_mmp_tsk);
1208 brelse(sbi->s_sbh);
1209 sb->s_fs_info = NULL;
1210 /*
1211 * Now that we are completely done shutting down the
1212 * superblock, we need to actually destroy the kobject.
1213 */
1214 kobject_put(&sbi->s_kobj);
1215 wait_for_completion(&sbi->s_kobj_unregister);
1216 if (sbi->s_chksum_driver)
1217 crypto_free_shash(sbi->s_chksum_driver);
1218 kfree(sbi->s_blockgroup_lock);
1219 fs_put_dax(sbi->s_daxdev);
1220 fscrypt_free_dummy_policy(&sbi->s_dummy_enc_policy);
1221 #ifdef CONFIG_UNICODE
1222 utf8_unload(sb->s_encoding);
1223 #endif
1224 kfree(sbi);
1225 }
1226
1227 static struct kmem_cache *ext4_inode_cachep;
1228
1229 /*
1230 * Called inside transaction, so use GFP_NOFS
1231 */
1232 static struct inode *ext4_alloc_inode(struct super_block *sb)
1233 {
1234 struct ext4_inode_info *ei;
1235
1236 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
1237 if (!ei)
1238 return NULL;
1239
1240 inode_set_iversion(&ei->vfs_inode, 1);
1241 spin_lock_init(&ei->i_raw_lock);
1242 INIT_LIST_HEAD(&ei->i_prealloc_list);
1243 atomic_set(&ei->i_prealloc_active, 0);
1244 spin_lock_init(&ei->i_prealloc_lock);
1245 ext4_es_init_tree(&ei->i_es_tree);
1246 rwlock_init(&ei->i_es_lock);
1247 INIT_LIST_HEAD(&ei->i_es_list);
1248 ei->i_es_all_nr = 0;
1249 ei->i_es_shk_nr = 0;
1250 ei->i_es_shrink_lblk = 0;
1251 ei->i_reserved_data_blocks = 0;
1252 spin_lock_init(&(ei->i_block_reservation_lock));
1253 ext4_init_pending_tree(&ei->i_pending_tree);
1254 #ifdef CONFIG_QUOTA
1255 ei->i_reserved_quota = 0;
1256 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
1257 #endif
1258 ei->jinode = NULL;
1259 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
1260 spin_lock_init(&ei->i_completed_io_lock);
1261 ei->i_sync_tid = 0;
1262 ei->i_datasync_tid = 0;
1263 atomic_set(&ei->i_unwritten, 0);
1264 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
1265 ext4_fc_init_inode(&ei->vfs_inode);
1266 mutex_init(&ei->i_fc_lock);
1267 return &ei->vfs_inode;
1268 }
1269
1270 static int ext4_drop_inode(struct inode *inode)
1271 {
1272 int drop = generic_drop_inode(inode);
1273
1274 if (!drop)
1275 drop = fscrypt_drop_inode(inode);
1276
1277 trace_ext4_drop_inode(inode, drop);
1278 return drop;
1279 }
1280
1281 static void ext4_free_in_core_inode(struct inode *inode)
1282 {
1283 fscrypt_free_inode(inode);
1284 if (!list_empty(&(EXT4_I(inode)->i_fc_list))) {
1285 pr_warn("%s: inode %ld still in fc list",
1286 __func__, inode->i_ino);
1287 }
1288 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
1289 }
1290
1291 static void ext4_destroy_inode(struct inode *inode)
1292 {
1293 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
1294 ext4_msg(inode->i_sb, KERN_ERR,
1295 "Inode %lu (%p): orphan list check failed!",
1296 inode->i_ino, EXT4_I(inode));
1297 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
1298 EXT4_I(inode), sizeof(struct ext4_inode_info),
1299 true);
1300 dump_stack();
1301 }
1302 }
1303
1304 static void init_once(void *foo)
1305 {
1306 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
1307
1308 INIT_LIST_HEAD(&ei->i_orphan);
1309 init_rwsem(&ei->xattr_sem);
1310 init_rwsem(&ei->i_data_sem);
1311 init_rwsem(&ei->i_mmap_sem);
1312 inode_init_once(&ei->vfs_inode);
1313 ext4_fc_init_inode(&ei->vfs_inode);
1314 }
1315
1316 static int __init init_inodecache(void)
1317 {
1318 ext4_inode_cachep = kmem_cache_create_usercopy("ext4_inode_cache",
1319 sizeof(struct ext4_inode_info), 0,
1320 (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD|
1321 SLAB_ACCOUNT),
1322 offsetof(struct ext4_inode_info, i_data),
1323 sizeof_field(struct ext4_inode_info, i_data),
1324 init_once);
1325 if (ext4_inode_cachep == NULL)
1326 return -ENOMEM;
1327 return 0;
1328 }
1329
1330 static void destroy_inodecache(void)
1331 {
1332 /*
1333 * Make sure all delayed rcu free inodes are flushed before we
1334 * destroy cache.
1335 */
1336 rcu_barrier();
1337 kmem_cache_destroy(ext4_inode_cachep);
1338 }
1339
1340 void ext4_clear_inode(struct inode *inode)
1341 {
1342 ext4_fc_del(inode);
1343 invalidate_inode_buffers(inode);
1344 clear_inode(inode);
1345 ext4_discard_preallocations(inode, 0);
1346 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
1347 dquot_drop(inode);
1348 if (EXT4_I(inode)->jinode) {
1349 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1350 EXT4_I(inode)->jinode);
1351 jbd2_free_inode(EXT4_I(inode)->jinode);
1352 EXT4_I(inode)->jinode = NULL;
1353 }
1354 fscrypt_put_encryption_info(inode);
1355 fsverity_cleanup_inode(inode);
1356 }
1357
1358 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1359 u64 ino, u32 generation)
1360 {
1361 struct inode *inode;
1362
1363 /*
1364 * Currently we don't know the generation for parent directory, so
1365 * a generation of 0 means "accept any"
1366 */
1367 inode = ext4_iget(sb, ino, EXT4_IGET_HANDLE);
1368 if (IS_ERR(inode))
1369 return ERR_CAST(inode);
1370 if (generation && inode->i_generation != generation) {
1371 iput(inode);
1372 return ERR_PTR(-ESTALE);
1373 }
1374
1375 return inode;
1376 }
1377
1378 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1379 int fh_len, int fh_type)
1380 {
1381 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1382 ext4_nfs_get_inode);
1383 }
1384
1385 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1386 int fh_len, int fh_type)
1387 {
1388 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1389 ext4_nfs_get_inode);
1390 }
1391
1392 static int ext4_nfs_commit_metadata(struct inode *inode)
1393 {
1394 struct writeback_control wbc = {
1395 .sync_mode = WB_SYNC_ALL
1396 };
1397
1398 trace_ext4_nfs_commit_metadata(inode);
1399 return ext4_write_inode(inode, &wbc);
1400 }
1401
1402 /*
1403 * Try to release metadata pages (indirect blocks, directories) which are
1404 * mapped via the block device. Since these pages could have journal heads
1405 * which would prevent try_to_free_buffers() from freeing them, we must use
1406 * jbd2 layer's try_to_free_buffers() function to release them.
1407 */
1408 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1409 gfp_t wait)
1410 {
1411 journal_t *journal = EXT4_SB(sb)->s_journal;
1412
1413 WARN_ON(PageChecked(page));
1414 if (!page_has_buffers(page))
1415 return 0;
1416 if (journal)
1417 return jbd2_journal_try_to_free_buffers(journal, page);
1418
1419 return try_to_free_buffers(page);
1420 }
1421
1422 #ifdef CONFIG_FS_ENCRYPTION
1423 static int ext4_get_context(struct inode *inode, void *ctx, size_t len)
1424 {
1425 return ext4_xattr_get(inode, EXT4_XATTR_INDEX_ENCRYPTION,
1426 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx, len);
1427 }
1428
1429 static int ext4_set_context(struct inode *inode, const void *ctx, size_t len,
1430 void *fs_data)
1431 {
1432 handle_t *handle = fs_data;
1433 int res, res2, credits, retries = 0;
1434
1435 /*
1436 * Encrypting the root directory is not allowed because e2fsck expects
1437 * lost+found to exist and be unencrypted, and encrypting the root
1438 * directory would imply encrypting the lost+found directory as well as
1439 * the filename "lost+found" itself.
1440 */
1441 if (inode->i_ino == EXT4_ROOT_INO)
1442 return -EPERM;
1443
1444 if (WARN_ON_ONCE(IS_DAX(inode) && i_size_read(inode)))
1445 return -EINVAL;
1446
1447 if (ext4_test_inode_flag(inode, EXT4_INODE_DAX))
1448 return -EOPNOTSUPP;
1449
1450 res = ext4_convert_inline_data(inode);
1451 if (res)
1452 return res;
1453
1454 /*
1455 * If a journal handle was specified, then the encryption context is
1456 * being set on a new inode via inheritance and is part of a larger
1457 * transaction to create the inode. Otherwise the encryption context is
1458 * being set on an existing inode in its own transaction. Only in the
1459 * latter case should the "retry on ENOSPC" logic be used.
1460 */
1461
1462 if (handle) {
1463 res = ext4_xattr_set_handle(handle, inode,
1464 EXT4_XATTR_INDEX_ENCRYPTION,
1465 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1466 ctx, len, 0);
1467 if (!res) {
1468 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1469 ext4_clear_inode_state(inode,
1470 EXT4_STATE_MAY_INLINE_DATA);
1471 /*
1472 * Update inode->i_flags - S_ENCRYPTED will be enabled,
1473 * S_DAX may be disabled
1474 */
1475 ext4_set_inode_flags(inode, false);
1476 }
1477 return res;
1478 }
1479
1480 res = dquot_initialize(inode);
1481 if (res)
1482 return res;
1483 retry:
1484 res = ext4_xattr_set_credits(inode, len, false /* is_create */,
1485 &credits);
1486 if (res)
1487 return res;
1488
1489 handle = ext4_journal_start(inode, EXT4_HT_MISC, credits);
1490 if (IS_ERR(handle))
1491 return PTR_ERR(handle);
1492
1493 res = ext4_xattr_set_handle(handle, inode, EXT4_XATTR_INDEX_ENCRYPTION,
1494 EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1495 ctx, len, 0);
1496 if (!res) {
1497 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1498 /*
1499 * Update inode->i_flags - S_ENCRYPTED will be enabled,
1500 * S_DAX may be disabled
1501 */
1502 ext4_set_inode_flags(inode, false);
1503 res = ext4_mark_inode_dirty(handle, inode);
1504 if (res)
1505 EXT4_ERROR_INODE(inode, "Failed to mark inode dirty");
1506 }
1507 res2 = ext4_journal_stop(handle);
1508
1509 if (res == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries))
1510 goto retry;
1511 if (!res)
1512 res = res2;
1513 return res;
1514 }
1515
1516 static const union fscrypt_policy *ext4_get_dummy_policy(struct super_block *sb)
1517 {
1518 return EXT4_SB(sb)->s_dummy_enc_policy.policy;
1519 }
1520
1521 static bool ext4_has_stable_inodes(struct super_block *sb)
1522 {
1523 return ext4_has_feature_stable_inodes(sb);
1524 }
1525
1526 static void ext4_get_ino_and_lblk_bits(struct super_block *sb,
1527 int *ino_bits_ret, int *lblk_bits_ret)
1528 {
1529 *ino_bits_ret = 8 * sizeof(EXT4_SB(sb)->s_es->s_inodes_count);
1530 *lblk_bits_ret = 8 * sizeof(ext4_lblk_t);
1531 }
1532
1533 static const struct fscrypt_operations ext4_cryptops = {
1534 .key_prefix = "ext4:",
1535 .get_context = ext4_get_context,
1536 .set_context = ext4_set_context,
1537 .get_dummy_policy = ext4_get_dummy_policy,
1538 .empty_dir = ext4_empty_dir,
1539 .max_namelen = EXT4_NAME_LEN,
1540 .has_stable_inodes = ext4_has_stable_inodes,
1541 .get_ino_and_lblk_bits = ext4_get_ino_and_lblk_bits,
1542 };
1543 #endif
1544
1545 #ifdef CONFIG_QUOTA
1546 static const char * const quotatypes[] = INITQFNAMES;
1547 #define QTYPE2NAME(t) (quotatypes[t])
1548
1549 static int ext4_write_dquot(struct dquot *dquot);
1550 static int ext4_acquire_dquot(struct dquot *dquot);
1551 static int ext4_release_dquot(struct dquot *dquot);
1552 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1553 static int ext4_write_info(struct super_block *sb, int type);
1554 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1555 const struct path *path);
1556 static int ext4_quota_on_mount(struct super_block *sb, int type);
1557 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1558 size_t len, loff_t off);
1559 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1560 const char *data, size_t len, loff_t off);
1561 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1562 unsigned int flags);
1563 static int ext4_enable_quotas(struct super_block *sb);
1564
1565 static struct dquot **ext4_get_dquots(struct inode *inode)
1566 {
1567 return EXT4_I(inode)->i_dquot;
1568 }
1569
1570 static const struct dquot_operations ext4_quota_operations = {
1571 .get_reserved_space = ext4_get_reserved_space,
1572 .write_dquot = ext4_write_dquot,
1573 .acquire_dquot = ext4_acquire_dquot,
1574 .release_dquot = ext4_release_dquot,
1575 .mark_dirty = ext4_mark_dquot_dirty,
1576 .write_info = ext4_write_info,
1577 .alloc_dquot = dquot_alloc,
1578 .destroy_dquot = dquot_destroy,
1579 .get_projid = ext4_get_projid,
1580 .get_inode_usage = ext4_get_inode_usage,
1581 .get_next_id = dquot_get_next_id,
1582 };
1583
1584 static const struct quotactl_ops ext4_qctl_operations = {
1585 .quota_on = ext4_quota_on,
1586 .quota_off = ext4_quota_off,
1587 .quota_sync = dquot_quota_sync,
1588 .get_state = dquot_get_state,
1589 .set_info = dquot_set_dqinfo,
1590 .get_dqblk = dquot_get_dqblk,
1591 .set_dqblk = dquot_set_dqblk,
1592 .get_nextdqblk = dquot_get_next_dqblk,
1593 };
1594 #endif
1595
1596 static const struct super_operations ext4_sops = {
1597 .alloc_inode = ext4_alloc_inode,
1598 .free_inode = ext4_free_in_core_inode,
1599 .destroy_inode = ext4_destroy_inode,
1600 .write_inode = ext4_write_inode,
1601 .dirty_inode = ext4_dirty_inode,
1602 .drop_inode = ext4_drop_inode,
1603 .evict_inode = ext4_evict_inode,
1604 .put_super = ext4_put_super,
1605 .sync_fs = ext4_sync_fs,
1606 .freeze_fs = ext4_freeze,
1607 .unfreeze_fs = ext4_unfreeze,
1608 .statfs = ext4_statfs,
1609 .remount_fs = ext4_remount,
1610 .show_options = ext4_show_options,
1611 #ifdef CONFIG_QUOTA
1612 .quota_read = ext4_quota_read,
1613 .quota_write = ext4_quota_write,
1614 .get_dquots = ext4_get_dquots,
1615 #endif
1616 .bdev_try_to_free_page = bdev_try_to_free_page,
1617 };
1618
1619 static const struct export_operations ext4_export_ops = {
1620 .fh_to_dentry = ext4_fh_to_dentry,
1621 .fh_to_parent = ext4_fh_to_parent,
1622 .get_parent = ext4_get_parent,
1623 .commit_metadata = ext4_nfs_commit_metadata,
1624 };
1625
1626 enum {
1627 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1628 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1629 Opt_nouid32, Opt_debug, Opt_removed,
1630 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1631 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1632 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1633 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1634 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1635 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
1636 Opt_inlinecrypt,
1637 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1638 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1639 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1640 Opt_usrquota, Opt_grpquota, Opt_prjquota, Opt_i_version,
1641 Opt_dax, Opt_dax_always, Opt_dax_inode, Opt_dax_never,
1642 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_warn_on_error,
1643 Opt_nowarn_on_error, Opt_mblk_io_submit,
1644 Opt_lazytime, Opt_nolazytime, Opt_debug_want_extra_isize,
1645 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1646 Opt_inode_readahead_blks, Opt_journal_ioprio,
1647 Opt_dioread_nolock, Opt_dioread_lock,
1648 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1649 Opt_max_dir_size_kb, Opt_nojournal_checksum, Opt_nombcache,
1650 Opt_prefetch_block_bitmaps,
1651 #ifdef CONFIG_EXT4_DEBUG
1652 Opt_fc_debug_max_replay, Opt_fc_debug_force
1653 #endif
1654 };
1655
1656 static const match_table_t tokens = {
1657 {Opt_bsd_df, "bsddf"},
1658 {Opt_minix_df, "minixdf"},
1659 {Opt_grpid, "grpid"},
1660 {Opt_grpid, "bsdgroups"},
1661 {Opt_nogrpid, "nogrpid"},
1662 {Opt_nogrpid, "sysvgroups"},
1663 {Opt_resgid, "resgid=%u"},
1664 {Opt_resuid, "resuid=%u"},
1665 {Opt_sb, "sb=%u"},
1666 {Opt_err_cont, "errors=continue"},
1667 {Opt_err_panic, "errors=panic"},
1668 {Opt_err_ro, "errors=remount-ro"},
1669 {Opt_nouid32, "nouid32"},
1670 {Opt_debug, "debug"},
1671 {Opt_removed, "oldalloc"},
1672 {Opt_removed, "orlov"},
1673 {Opt_user_xattr, "user_xattr"},
1674 {Opt_nouser_xattr, "nouser_xattr"},
1675 {Opt_acl, "acl"},
1676 {Opt_noacl, "noacl"},
1677 {Opt_noload, "norecovery"},
1678 {Opt_noload, "noload"},
1679 {Opt_removed, "nobh"},
1680 {Opt_removed, "bh"},
1681 {Opt_commit, "commit=%u"},
1682 {Opt_min_batch_time, "min_batch_time=%u"},
1683 {Opt_max_batch_time, "max_batch_time=%u"},
1684 {Opt_journal_dev, "journal_dev=%u"},
1685 {Opt_journal_path, "journal_path=%s"},
1686 {Opt_journal_checksum, "journal_checksum"},
1687 {Opt_nojournal_checksum, "nojournal_checksum"},
1688 {Opt_journal_async_commit, "journal_async_commit"},
1689 {Opt_abort, "abort"},
1690 {Opt_data_journal, "data=journal"},
1691 {Opt_data_ordered, "data=ordered"},
1692 {Opt_data_writeback, "data=writeback"},
1693 {Opt_data_err_abort, "data_err=abort"},
1694 {Opt_data_err_ignore, "data_err=ignore"},
1695 {Opt_offusrjquota, "usrjquota="},
1696 {Opt_usrjquota, "usrjquota=%s"},
1697 {Opt_offgrpjquota, "grpjquota="},
1698 {Opt_grpjquota, "grpjquota=%s"},
1699 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1700 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1701 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1702 {Opt_grpquota, "grpquota"},
1703 {Opt_noquota, "noquota"},
1704 {Opt_quota, "quota"},
1705 {Opt_usrquota, "usrquota"},
1706 {Opt_prjquota, "prjquota"},
1707 {Opt_barrier, "barrier=%u"},
1708 {Opt_barrier, "barrier"},
1709 {Opt_nobarrier, "nobarrier"},
1710 {Opt_i_version, "i_version"},
1711 {Opt_dax, "dax"},
1712 {Opt_dax_always, "dax=always"},
1713 {Opt_dax_inode, "dax=inode"},
1714 {Opt_dax_never, "dax=never"},
1715 {Opt_stripe, "stripe=%u"},
1716 {Opt_delalloc, "delalloc"},
1717 {Opt_warn_on_error, "warn_on_error"},
1718 {Opt_nowarn_on_error, "nowarn_on_error"},
1719 {Opt_lazytime, "lazytime"},
1720 {Opt_nolazytime, "nolazytime"},
1721 {Opt_debug_want_extra_isize, "debug_want_extra_isize=%u"},
1722 {Opt_nodelalloc, "nodelalloc"},
1723 {Opt_removed, "mblk_io_submit"},
1724 {Opt_removed, "nomblk_io_submit"},
1725 {Opt_block_validity, "block_validity"},
1726 {Opt_noblock_validity, "noblock_validity"},
1727 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1728 {Opt_journal_ioprio, "journal_ioprio=%u"},
1729 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1730 {Opt_auto_da_alloc, "auto_da_alloc"},
1731 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1732 {Opt_dioread_nolock, "dioread_nolock"},
1733 {Opt_dioread_lock, "nodioread_nolock"},
1734 {Opt_dioread_lock, "dioread_lock"},
1735 {Opt_discard, "discard"},
1736 {Opt_nodiscard, "nodiscard"},
1737 {Opt_init_itable, "init_itable=%u"},
1738 {Opt_init_itable, "init_itable"},
1739 {Opt_noinit_itable, "noinit_itable"},
1740 #ifdef CONFIG_EXT4_DEBUG
1741 {Opt_fc_debug_force, "fc_debug_force"},
1742 {Opt_fc_debug_max_replay, "fc_debug_max_replay=%u"},
1743 #endif
1744 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1745 {Opt_test_dummy_encryption, "test_dummy_encryption=%s"},
1746 {Opt_test_dummy_encryption, "test_dummy_encryption"},
1747 {Opt_inlinecrypt, "inlinecrypt"},
1748 {Opt_nombcache, "nombcache"},
1749 {Opt_nombcache, "no_mbcache"}, /* for backward compatibility */
1750 {Opt_prefetch_block_bitmaps, "prefetch_block_bitmaps"},
1751 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1752 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1753 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1754 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1755 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
1756 {Opt_err, NULL},
1757 };
1758
1759 static ext4_fsblk_t get_sb_block(void **data)
1760 {
1761 ext4_fsblk_t sb_block;
1762 char *options = (char *) *data;
1763
1764 if (!options || strncmp(options, "sb=", 3) != 0)
1765 return 1; /* Default location */
1766
1767 options += 3;
1768 /* TODO: use simple_strtoll with >32bit ext4 */
1769 sb_block = simple_strtoul(options, &options, 0);
1770 if (*options && *options != ',') {
1771 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1772 (char *) *data);
1773 return 1;
1774 }
1775 if (*options == ',')
1776 options++;
1777 *data = (void *) options;
1778
1779 return sb_block;
1780 }
1781
1782 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1783 static const char deprecated_msg[] =
1784 "Mount option \"%s\" will be removed by %s\n"
1785 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1786
1787 #ifdef CONFIG_QUOTA
1788 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1789 {
1790 struct ext4_sb_info *sbi = EXT4_SB(sb);
1791 char *qname, *old_qname = get_qf_name(sb, sbi, qtype);
1792 int ret = -1;
1793
1794 if (sb_any_quota_loaded(sb) && !old_qname) {
1795 ext4_msg(sb, KERN_ERR,
1796 "Cannot change journaled "
1797 "quota options when quota turned on");
1798 return -1;
1799 }
1800 if (ext4_has_feature_quota(sb)) {
1801 ext4_msg(sb, KERN_INFO, "Journaled quota options "
1802 "ignored when QUOTA feature is enabled");
1803 return 1;
1804 }
1805 qname = match_strdup(args);
1806 if (!qname) {
1807 ext4_msg(sb, KERN_ERR,
1808 "Not enough memory for storing quotafile name");
1809 return -1;
1810 }
1811 if (old_qname) {
1812 if (strcmp(old_qname, qname) == 0)
1813 ret = 1;
1814 else
1815 ext4_msg(sb, KERN_ERR,
1816 "%s quota file already specified",
1817 QTYPE2NAME(qtype));
1818 goto errout;
1819 }
1820 if (strchr(qname, '/')) {
1821 ext4_msg(sb, KERN_ERR,
1822 "quotafile must be on filesystem root");
1823 goto errout;
1824 }
1825 rcu_assign_pointer(sbi->s_qf_names[qtype], qname);
1826 set_opt(sb, QUOTA);
1827 return 1;
1828 errout:
1829 kfree(qname);
1830 return ret;
1831 }
1832
1833 static int clear_qf_name(struct super_block *sb, int qtype)
1834 {
1835
1836 struct ext4_sb_info *sbi = EXT4_SB(sb);
1837 char *old_qname = get_qf_name(sb, sbi, qtype);
1838
1839 if (sb_any_quota_loaded(sb) && old_qname) {
1840 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1841 " when quota turned on");
1842 return -1;
1843 }
1844 rcu_assign_pointer(sbi->s_qf_names[qtype], NULL);
1845 synchronize_rcu();
1846 kfree(old_qname);
1847 return 1;
1848 }
1849 #endif
1850
1851 #define MOPT_SET 0x0001
1852 #define MOPT_CLEAR 0x0002
1853 #define MOPT_NOSUPPORT 0x0004
1854 #define MOPT_EXPLICIT 0x0008
1855 #define MOPT_CLEAR_ERR 0x0010
1856 #define MOPT_GTE0 0x0020
1857 #ifdef CONFIG_QUOTA
1858 #define MOPT_Q 0
1859 #define MOPT_QFMT 0x0040
1860 #else
1861 #define MOPT_Q MOPT_NOSUPPORT
1862 #define MOPT_QFMT MOPT_NOSUPPORT
1863 #endif
1864 #define MOPT_DATAJ 0x0080
1865 #define MOPT_NO_EXT2 0x0100
1866 #define MOPT_NO_EXT3 0x0200
1867 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1868 #define MOPT_STRING 0x0400
1869 #define MOPT_SKIP 0x0800
1870 #define MOPT_2 0x1000
1871
1872 static const struct mount_opts {
1873 int token;
1874 int mount_opt;
1875 int flags;
1876 } ext4_mount_opts[] = {
1877 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1878 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1879 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1880 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1881 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1882 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1883 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1884 MOPT_EXT4_ONLY | MOPT_SET},
1885 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1886 MOPT_EXT4_ONLY | MOPT_CLEAR},
1887 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1888 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1889 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1890 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1891 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1892 MOPT_EXT4_ONLY | MOPT_CLEAR},
1893 {Opt_warn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_SET},
1894 {Opt_nowarn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_CLEAR},
1895 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1896 MOPT_EXT4_ONLY | MOPT_CLEAR},
1897 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1898 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1899 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1900 EXT4_MOUNT_JOURNAL_CHECKSUM),
1901 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1902 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1903 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1904 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1905 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1906 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1907 MOPT_NO_EXT2},
1908 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1909 MOPT_NO_EXT2},
1910 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1911 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1912 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1913 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1914 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1915 {Opt_commit, 0, MOPT_GTE0},
1916 {Opt_max_batch_time, 0, MOPT_GTE0},
1917 {Opt_min_batch_time, 0, MOPT_GTE0},
1918 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1919 {Opt_init_itable, 0, MOPT_GTE0},
1920 {Opt_dax, EXT4_MOUNT_DAX_ALWAYS, MOPT_SET | MOPT_SKIP},
1921 {Opt_dax_always, EXT4_MOUNT_DAX_ALWAYS,
1922 MOPT_EXT4_ONLY | MOPT_SET | MOPT_SKIP},
1923 {Opt_dax_inode, EXT4_MOUNT2_DAX_INODE,
1924 MOPT_EXT4_ONLY | MOPT_SET | MOPT_SKIP},
1925 {Opt_dax_never, EXT4_MOUNT2_DAX_NEVER,
1926 MOPT_EXT4_ONLY | MOPT_SET | MOPT_SKIP},
1927 {Opt_stripe, 0, MOPT_GTE0},
1928 {Opt_resuid, 0, MOPT_GTE0},
1929 {Opt_resgid, 0, MOPT_GTE0},
1930 {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1931 {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING},
1932 {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1933 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1934 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1935 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1936 MOPT_NO_EXT2 | MOPT_DATAJ},
1937 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1938 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1939 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1940 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1941 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1942 #else
1943 {Opt_acl, 0, MOPT_NOSUPPORT},
1944 {Opt_noacl, 0, MOPT_NOSUPPORT},
1945 #endif
1946 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1947 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1948 {Opt_debug_want_extra_isize, 0, MOPT_GTE0},
1949 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1950 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1951 MOPT_SET | MOPT_Q},
1952 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1953 MOPT_SET | MOPT_Q},
1954 {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA,
1955 MOPT_SET | MOPT_Q},
1956 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1957 EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA),
1958 MOPT_CLEAR | MOPT_Q},
1959 {Opt_usrjquota, 0, MOPT_Q | MOPT_STRING},
1960 {Opt_grpjquota, 0, MOPT_Q | MOPT_STRING},
1961 {Opt_offusrjquota, 0, MOPT_Q},
1962 {Opt_offgrpjquota, 0, MOPT_Q},
1963 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1964 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1965 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1966 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
1967 {Opt_test_dummy_encryption, 0, MOPT_STRING},
1968 {Opt_nombcache, EXT4_MOUNT_NO_MBCACHE, MOPT_SET},
1969 {Opt_prefetch_block_bitmaps, EXT4_MOUNT_PREFETCH_BLOCK_BITMAPS,
1970 MOPT_SET},
1971 #ifdef CONFIG_EXT4_DEBUG
1972 {Opt_fc_debug_force, EXT4_MOUNT2_JOURNAL_FAST_COMMIT,
1973 MOPT_SET | MOPT_2 | MOPT_EXT4_ONLY},
1974 {Opt_fc_debug_max_replay, 0, MOPT_GTE0},
1975 #endif
1976 {Opt_err, 0, 0}
1977 };
1978
1979 #ifdef CONFIG_UNICODE
1980 static const struct ext4_sb_encodings {
1981 __u16 magic;
1982 char *name;
1983 char *version;
1984 } ext4_sb_encoding_map[] = {
1985 {EXT4_ENC_UTF8_12_1, "utf8", "12.1.0"},
1986 };
1987
1988 static int ext4_sb_read_encoding(const struct ext4_super_block *es,
1989 const struct ext4_sb_encodings **encoding,
1990 __u16 *flags)
1991 {
1992 __u16 magic = le16_to_cpu(es->s_encoding);
1993 int i;
1994
1995 for (i = 0; i < ARRAY_SIZE(ext4_sb_encoding_map); i++)
1996 if (magic == ext4_sb_encoding_map[i].magic)
1997 break;
1998
1999 if (i >= ARRAY_SIZE(ext4_sb_encoding_map))
2000 return -EINVAL;
2001
2002 *encoding = &ext4_sb_encoding_map[i];
2003 *flags = le16_to_cpu(es->s_encoding_flags);
2004
2005 return 0;
2006 }
2007 #endif
2008
2009 static int ext4_set_test_dummy_encryption(struct super_block *sb,
2010 const char *opt,
2011 const substring_t *arg,
2012 bool is_remount)
2013 {
2014 #ifdef CONFIG_FS_ENCRYPTION
2015 struct ext4_sb_info *sbi = EXT4_SB(sb);
2016 int err;
2017
2018 /*
2019 * This mount option is just for testing, and it's not worthwhile to
2020 * implement the extra complexity (e.g. RCU protection) that would be
2021 * needed to allow it to be set or changed during remount. We do allow
2022 * it to be specified during remount, but only if there is no change.
2023 */
2024 if (is_remount && !sbi->s_dummy_enc_policy.policy) {
2025 ext4_msg(sb, KERN_WARNING,
2026 "Can't set test_dummy_encryption on remount");
2027 return -1;
2028 }
2029 err = fscrypt_set_test_dummy_encryption(sb, arg->from,
2030 &sbi->s_dummy_enc_policy);
2031 if (err) {
2032 if (err == -EEXIST)
2033 ext4_msg(sb, KERN_WARNING,
2034 "Can't change test_dummy_encryption on remount");
2035 else if (err == -EINVAL)
2036 ext4_msg(sb, KERN_WARNING,
2037 "Value of option \"%s\" is unrecognized", opt);
2038 else
2039 ext4_msg(sb, KERN_WARNING,
2040 "Error processing option \"%s\" [%d]",
2041 opt, err);
2042 return -1;
2043 }
2044 ext4_msg(sb, KERN_WARNING, "Test dummy encryption mode enabled");
2045 #else
2046 ext4_msg(sb, KERN_WARNING,
2047 "Test dummy encryption mount option ignored");
2048 #endif
2049 return 1;
2050 }
2051
2052 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
2053 substring_t *args, unsigned long *journal_devnum,
2054 unsigned int *journal_ioprio, int is_remount)
2055 {
2056 struct ext4_sb_info *sbi = EXT4_SB(sb);
2057 const struct mount_opts *m;
2058 kuid_t uid;
2059 kgid_t gid;
2060 int arg = 0;
2061
2062 #ifdef CONFIG_QUOTA
2063 if (token == Opt_usrjquota)
2064 return set_qf_name(sb, USRQUOTA, &args[0]);
2065 else if (token == Opt_grpjquota)
2066 return set_qf_name(sb, GRPQUOTA, &args[0]);
2067 else if (token == Opt_offusrjquota)
2068 return clear_qf_name(sb, USRQUOTA);
2069 else if (token == Opt_offgrpjquota)
2070 return clear_qf_name(sb, GRPQUOTA);
2071 #endif
2072 switch (token) {
2073 case Opt_noacl:
2074 case Opt_nouser_xattr:
2075 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
2076 break;
2077 case Opt_sb:
2078 return 1; /* handled by get_sb_block() */
2079 case Opt_removed:
2080 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
2081 return 1;
2082 case Opt_abort:
2083 ext4_set_mount_flag(sb, EXT4_MF_FS_ABORTED);
2084 return 1;
2085 case Opt_i_version:
2086 sb->s_flags |= SB_I_VERSION;
2087 return 1;
2088 case Opt_lazytime:
2089 sb->s_flags |= SB_LAZYTIME;
2090 return 1;
2091 case Opt_nolazytime:
2092 sb->s_flags &= ~SB_LAZYTIME;
2093 return 1;
2094 case Opt_inlinecrypt:
2095 #ifdef CONFIG_FS_ENCRYPTION_INLINE_CRYPT
2096 sb->s_flags |= SB_INLINECRYPT;
2097 #else
2098 ext4_msg(sb, KERN_ERR, "inline encryption not supported");
2099 #endif
2100 return 1;
2101 }
2102
2103 for (m = ext4_mount_opts; m->token != Opt_err; m++)
2104 if (token == m->token)
2105 break;
2106
2107 if (m->token == Opt_err) {
2108 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
2109 "or missing value", opt);
2110 return -1;
2111 }
2112
2113 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
2114 ext4_msg(sb, KERN_ERR,
2115 "Mount option \"%s\" incompatible with ext2", opt);
2116 return -1;
2117 }
2118 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
2119 ext4_msg(sb, KERN_ERR,
2120 "Mount option \"%s\" incompatible with ext3", opt);
2121 return -1;
2122 }
2123
2124 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
2125 return -1;
2126 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
2127 return -1;
2128 if (m->flags & MOPT_EXPLICIT) {
2129 if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
2130 set_opt2(sb, EXPLICIT_DELALLOC);
2131 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
2132 set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM);
2133 } else
2134 return -1;
2135 }
2136 if (m->flags & MOPT_CLEAR_ERR)
2137 clear_opt(sb, ERRORS_MASK);
2138 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
2139 ext4_msg(sb, KERN_ERR, "Cannot change quota "
2140 "options when quota turned on");
2141 return -1;
2142 }
2143
2144 if (m->flags & MOPT_NOSUPPORT) {
2145 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
2146 } else if (token == Opt_commit) {
2147 if (arg == 0)
2148 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
2149 else if (arg > INT_MAX / HZ) {
2150 ext4_msg(sb, KERN_ERR,
2151 "Invalid commit interval %d, "
2152 "must be smaller than %d",
2153 arg, INT_MAX / HZ);
2154 return -1;
2155 }
2156 sbi->s_commit_interval = HZ * arg;
2157 } else if (token == Opt_debug_want_extra_isize) {
2158 if ((arg & 1) ||
2159 (arg < 4) ||
2160 (arg > (sbi->s_inode_size - EXT4_GOOD_OLD_INODE_SIZE))) {
2161 ext4_msg(sb, KERN_ERR,
2162 "Invalid want_extra_isize %d", arg);
2163 return -1;
2164 }
2165 sbi->s_want_extra_isize = arg;
2166 } else if (token == Opt_max_batch_time) {
2167 sbi->s_max_batch_time = arg;
2168 } else if (token == Opt_min_batch_time) {
2169 sbi->s_min_batch_time = arg;
2170 } else if (token == Opt_inode_readahead_blks) {
2171 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
2172 ext4_msg(sb, KERN_ERR,
2173 "EXT4-fs: inode_readahead_blks must be "
2174 "0 or a power of 2 smaller than 2^31");
2175 return -1;
2176 }
2177 sbi->s_inode_readahead_blks = arg;
2178 } else if (token == Opt_init_itable) {
2179 set_opt(sb, INIT_INODE_TABLE);
2180 if (!args->from)
2181 arg = EXT4_DEF_LI_WAIT_MULT;
2182 sbi->s_li_wait_mult = arg;
2183 } else if (token == Opt_max_dir_size_kb) {
2184 sbi->s_max_dir_size_kb = arg;
2185 #ifdef CONFIG_EXT4_DEBUG
2186 } else if (token == Opt_fc_debug_max_replay) {
2187 sbi->s_fc_debug_max_replay = arg;
2188 #endif
2189 } else if (token == Opt_stripe) {
2190 sbi->s_stripe = arg;
2191 } else if (token == Opt_resuid) {
2192 uid = make_kuid(current_user_ns(), arg);
2193 if (!uid_valid(uid)) {
2194 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
2195 return -1;
2196 }
2197 sbi->s_resuid = uid;
2198 } else if (token == Opt_resgid) {
2199 gid = make_kgid(current_user_ns(), arg);
2200 if (!gid_valid(gid)) {
2201 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
2202 return -1;
2203 }
2204 sbi->s_resgid = gid;
2205 } else if (token == Opt_journal_dev) {
2206 if (is_remount) {
2207 ext4_msg(sb, KERN_ERR,
2208 "Cannot specify journal on remount");
2209 return -1;
2210 }
2211 *journal_devnum = arg;
2212 } else if (token == Opt_journal_path) {
2213 char *journal_path;
2214 struct inode *journal_inode;
2215 struct path path;
2216 int error;
2217
2218 if (is_remount) {
2219 ext4_msg(sb, KERN_ERR,
2220 "Cannot specify journal on remount");
2221 return -1;
2222 }
2223 journal_path = match_strdup(&args[0]);
2224 if (!journal_path) {
2225 ext4_msg(sb, KERN_ERR, "error: could not dup "
2226 "journal device string");
2227 return -1;
2228 }
2229
2230 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
2231 if (error) {
2232 ext4_msg(sb, KERN_ERR, "error: could not find "
2233 "journal device path: error %d", error);
2234 kfree(journal_path);
2235 return -1;
2236 }
2237
2238 journal_inode = d_inode(path.dentry);
2239 if (!S_ISBLK(journal_inode->i_mode)) {
2240 ext4_msg(sb, KERN_ERR, "error: journal path %s "
2241 "is not a block device", journal_path);
2242 path_put(&path);
2243 kfree(journal_path);
2244 return -1;
2245 }
2246
2247 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
2248 path_put(&path);
2249 kfree(journal_path);
2250 } else if (token == Opt_journal_ioprio) {
2251 if (arg > 7) {
2252 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
2253 " (must be 0-7)");
2254 return -1;
2255 }
2256 *journal_ioprio =
2257 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
2258 } else if (token == Opt_test_dummy_encryption) {
2259 return ext4_set_test_dummy_encryption(sb, opt, &args[0],
2260 is_remount);
2261 } else if (m->flags & MOPT_DATAJ) {
2262 if (is_remount) {
2263 if (!sbi->s_journal)
2264 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
2265 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
2266 ext4_msg(sb, KERN_ERR,
2267 "Cannot change data mode on remount");
2268 return -1;
2269 }
2270 } else {
2271 clear_opt(sb, DATA_FLAGS);
2272 sbi->s_mount_opt |= m->mount_opt;
2273 }
2274 #ifdef CONFIG_QUOTA
2275 } else if (m->flags & MOPT_QFMT) {
2276 if (sb_any_quota_loaded(sb) &&
2277 sbi->s_jquota_fmt != m->mount_opt) {
2278 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
2279 "quota options when quota turned on");
2280 return -1;
2281 }
2282 if (ext4_has_feature_quota(sb)) {
2283 ext4_msg(sb, KERN_INFO,
2284 "Quota format mount options ignored "
2285 "when QUOTA feature is enabled");
2286 return 1;
2287 }
2288 sbi->s_jquota_fmt = m->mount_opt;
2289 #endif
2290 } else if (token == Opt_dax || token == Opt_dax_always ||
2291 token == Opt_dax_inode || token == Opt_dax_never) {
2292 #ifdef CONFIG_FS_DAX
2293 switch (token) {
2294 case Opt_dax:
2295 case Opt_dax_always:
2296 if (is_remount &&
2297 (!(sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) ||
2298 (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER))) {
2299 fail_dax_change_remount:
2300 ext4_msg(sb, KERN_ERR, "can't change "
2301 "dax mount option while remounting");
2302 return -1;
2303 }
2304 if (is_remount &&
2305 (test_opt(sb, DATA_FLAGS) ==
2306 EXT4_MOUNT_JOURNAL_DATA)) {
2307 ext4_msg(sb, KERN_ERR, "can't mount with "
2308 "both data=journal and dax");
2309 return -1;
2310 }
2311 ext4_msg(sb, KERN_WARNING,
2312 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
2313 sbi->s_mount_opt |= EXT4_MOUNT_DAX_ALWAYS;
2314 sbi->s_mount_opt2 &= ~EXT4_MOUNT2_DAX_NEVER;
2315 break;
2316 case Opt_dax_never:
2317 if (is_remount &&
2318 (!(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) ||
2319 (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS)))
2320 goto fail_dax_change_remount;
2321 sbi->s_mount_opt2 |= EXT4_MOUNT2_DAX_NEVER;
2322 sbi->s_mount_opt &= ~EXT4_MOUNT_DAX_ALWAYS;
2323 break;
2324 case Opt_dax_inode:
2325 if (is_remount &&
2326 ((sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) ||
2327 (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) ||
2328 !(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_INODE)))
2329 goto fail_dax_change_remount;
2330 sbi->s_mount_opt &= ~EXT4_MOUNT_DAX_ALWAYS;
2331 sbi->s_mount_opt2 &= ~EXT4_MOUNT2_DAX_NEVER;
2332 /* Strictly for printing options */
2333 sbi->s_mount_opt2 |= EXT4_MOUNT2_DAX_INODE;
2334 break;
2335 }
2336 #else
2337 ext4_msg(sb, KERN_INFO, "dax option not supported");
2338 sbi->s_mount_opt2 |= EXT4_MOUNT2_DAX_NEVER;
2339 sbi->s_mount_opt &= ~EXT4_MOUNT_DAX_ALWAYS;
2340 return -1;
2341 #endif
2342 } else if (token == Opt_data_err_abort) {
2343 sbi->s_mount_opt |= m->mount_opt;
2344 } else if (token == Opt_data_err_ignore) {
2345 sbi->s_mount_opt &= ~m->mount_opt;
2346 } else {
2347 if (!args->from)
2348 arg = 1;
2349 if (m->flags & MOPT_CLEAR)
2350 arg = !arg;
2351 else if (unlikely(!(m->flags & MOPT_SET))) {
2352 ext4_msg(sb, KERN_WARNING,
2353 "buggy handling of option %s", opt);
2354 WARN_ON(1);
2355 return -1;
2356 }
2357 if (m->flags & MOPT_2) {
2358 if (arg != 0)
2359 sbi->s_mount_opt2 |= m->mount_opt;
2360 else
2361 sbi->s_mount_opt2 &= ~m->mount_opt;
2362 } else {
2363 if (arg != 0)
2364 sbi->s_mount_opt |= m->mount_opt;
2365 else
2366 sbi->s_mount_opt &= ~m->mount_opt;
2367 }
2368 }
2369 return 1;
2370 }
2371
2372 static int parse_options(char *options, struct super_block *sb,
2373 unsigned long *journal_devnum,
2374 unsigned int *journal_ioprio,
2375 int is_remount)
2376 {
2377 struct ext4_sb_info __maybe_unused *sbi = EXT4_SB(sb);
2378 char *p, __maybe_unused *usr_qf_name, __maybe_unused *grp_qf_name;
2379 substring_t args[MAX_OPT_ARGS];
2380 int token;
2381
2382 if (!options)
2383 return 1;
2384
2385 while ((p = strsep(&options, ",")) != NULL) {
2386 if (!*p)
2387 continue;
2388 /*
2389 * Initialize args struct so we know whether arg was
2390 * found; some options take optional arguments.
2391 */
2392 args[0].to = args[0].from = NULL;
2393 token = match_token(p, tokens, args);
2394 if (handle_mount_opt(sb, p, token, args, journal_devnum,
2395 journal_ioprio, is_remount) < 0)
2396 return 0;
2397 }
2398 #ifdef CONFIG_QUOTA
2399 /*
2400 * We do the test below only for project quotas. 'usrquota' and
2401 * 'grpquota' mount options are allowed even without quota feature
2402 * to support legacy quotas in quota files.
2403 */
2404 if (test_opt(sb, PRJQUOTA) && !ext4_has_feature_project(sb)) {
2405 ext4_msg(sb, KERN_ERR, "Project quota feature not enabled. "
2406 "Cannot enable project quota enforcement.");
2407 return 0;
2408 }
2409 usr_qf_name = get_qf_name(sb, sbi, USRQUOTA);
2410 grp_qf_name = get_qf_name(sb, sbi, GRPQUOTA);
2411 if (usr_qf_name || grp_qf_name) {
2412 if (test_opt(sb, USRQUOTA) && usr_qf_name)
2413 clear_opt(sb, USRQUOTA);
2414
2415 if (test_opt(sb, GRPQUOTA) && grp_qf_name)
2416 clear_opt(sb, GRPQUOTA);
2417
2418 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
2419 ext4_msg(sb, KERN_ERR, "old and new quota "
2420 "format mixing");
2421 return 0;
2422 }
2423
2424 if (!sbi->s_jquota_fmt) {
2425 ext4_msg(sb, KERN_ERR, "journaled quota format "
2426 "not specified");
2427 return 0;
2428 }
2429 }
2430 #endif
2431 if (test_opt(sb, DIOREAD_NOLOCK)) {
2432 int blocksize =
2433 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
2434 if (blocksize < PAGE_SIZE)
2435 ext4_msg(sb, KERN_WARNING, "Warning: mounting with an "
2436 "experimental mount option 'dioread_nolock' "
2437 "for blocksize < PAGE_SIZE");
2438 }
2439 return 1;
2440 }
2441
2442 static inline void ext4_show_quota_options(struct seq_file *seq,
2443 struct super_block *sb)
2444 {
2445 #if defined(CONFIG_QUOTA)
2446 struct ext4_sb_info *sbi = EXT4_SB(sb);
2447 char *usr_qf_name, *grp_qf_name;
2448
2449 if (sbi->s_jquota_fmt) {
2450 char *fmtname = "";
2451
2452 switch (sbi->s_jquota_fmt) {
2453 case QFMT_VFS_OLD:
2454 fmtname = "vfsold";
2455 break;
2456 case QFMT_VFS_V0:
2457 fmtname = "vfsv0";
2458 break;
2459 case QFMT_VFS_V1:
2460 fmtname = "vfsv1";
2461 break;
2462 }
2463 seq_printf(seq, ",jqfmt=%s", fmtname);
2464 }
2465
2466 rcu_read_lock();
2467 usr_qf_name = rcu_dereference(sbi->s_qf_names[USRQUOTA]);
2468 grp_qf_name = rcu_dereference(sbi->s_qf_names[GRPQUOTA]);
2469 if (usr_qf_name)
2470 seq_show_option(seq, "usrjquota", usr_qf_name);
2471 if (grp_qf_name)
2472 seq_show_option(seq, "grpjquota", grp_qf_name);
2473 rcu_read_unlock();
2474 #endif
2475 }
2476
2477 static const char *token2str(int token)
2478 {
2479 const struct match_token *t;
2480
2481 for (t = tokens; t->token != Opt_err; t++)
2482 if (t->token == token && !strchr(t->pattern, '='))
2483 break;
2484 return t->pattern;
2485 }
2486
2487 /*
2488 * Show an option if
2489 * - it's set to a non-default value OR
2490 * - if the per-sb default is different from the global default
2491 */
2492 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
2493 int nodefs)
2494 {
2495 struct ext4_sb_info *sbi = EXT4_SB(sb);
2496 struct ext4_super_block *es = sbi->s_es;
2497 int def_errors, def_mount_opt = sbi->s_def_mount_opt;
2498 const struct mount_opts *m;
2499 char sep = nodefs ? '\n' : ',';
2500
2501 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
2502 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2503
2504 if (sbi->s_sb_block != 1)
2505 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
2506
2507 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
2508 int want_set = m->flags & MOPT_SET;
2509 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
2510 (m->flags & MOPT_CLEAR_ERR) || m->flags & MOPT_SKIP)
2511 continue;
2512 if (!nodefs && !(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
2513 continue; /* skip if same as the default */
2514 if ((want_set &&
2515 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
2516 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
2517 continue; /* select Opt_noFoo vs Opt_Foo */
2518 SEQ_OPTS_PRINT("%s", token2str(m->token));
2519 }
2520
2521 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
2522 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
2523 SEQ_OPTS_PRINT("resuid=%u",
2524 from_kuid_munged(&init_user_ns, sbi->s_resuid));
2525 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
2526 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
2527 SEQ_OPTS_PRINT("resgid=%u",
2528 from_kgid_munged(&init_user_ns, sbi->s_resgid));
2529 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
2530 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
2531 SEQ_OPTS_PUTS("errors=remount-ro");
2532 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
2533 SEQ_OPTS_PUTS("errors=continue");
2534 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
2535 SEQ_OPTS_PUTS("errors=panic");
2536 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
2537 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
2538 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
2539 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
2540 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
2541 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2542 if (sb->s_flags & SB_I_VERSION)
2543 SEQ_OPTS_PUTS("i_version");
2544 if (nodefs || sbi->s_stripe)
2545 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
2546 if (nodefs || EXT4_MOUNT_DATA_FLAGS &
2547 (sbi->s_mount_opt ^ def_mount_opt)) {
2548 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2549 SEQ_OPTS_PUTS("data=journal");
2550 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2551 SEQ_OPTS_PUTS("data=ordered");
2552 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
2553 SEQ_OPTS_PUTS("data=writeback");
2554 }
2555 if (nodefs ||
2556 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
2557 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
2558 sbi->s_inode_readahead_blks);
2559
2560 if (test_opt(sb, INIT_INODE_TABLE) && (nodefs ||
2561 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
2562 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
2563 if (nodefs || sbi->s_max_dir_size_kb)
2564 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
2565 if (test_opt(sb, DATA_ERR_ABORT))
2566 SEQ_OPTS_PUTS("data_err=abort");
2567
2568 fscrypt_show_test_dummy_encryption(seq, sep, sb);
2569
2570 if (sb->s_flags & SB_INLINECRYPT)
2571 SEQ_OPTS_PUTS("inlinecrypt");
2572
2573 if (test_opt(sb, DAX_ALWAYS)) {
2574 if (IS_EXT2_SB(sb))
2575 SEQ_OPTS_PUTS("dax");
2576 else
2577 SEQ_OPTS_PUTS("dax=always");
2578 } else if (test_opt2(sb, DAX_NEVER)) {
2579 SEQ_OPTS_PUTS("dax=never");
2580 } else if (test_opt2(sb, DAX_INODE)) {
2581 SEQ_OPTS_PUTS("dax=inode");
2582 }
2583 ext4_show_quota_options(seq, sb);
2584 return 0;
2585 }
2586
2587 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
2588 {
2589 return _ext4_show_options(seq, root->d_sb, 0);
2590 }
2591
2592 int ext4_seq_options_show(struct seq_file *seq, void *offset)
2593 {
2594 struct super_block *sb = seq->private;
2595 int rc;
2596
2597 seq_puts(seq, sb_rdonly(sb) ? "ro" : "rw");
2598 rc = _ext4_show_options(seq, sb, 1);
2599 seq_puts(seq, "\n");
2600 return rc;
2601 }
2602
2603 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
2604 int read_only)
2605 {
2606 struct ext4_sb_info *sbi = EXT4_SB(sb);
2607 int err = 0;
2608
2609 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
2610 ext4_msg(sb, KERN_ERR, "revision level too high, "
2611 "forcing read-only mode");
2612 err = -EROFS;
2613 goto done;
2614 }
2615 if (read_only)
2616 goto done;
2617 if (!(sbi->s_mount_state & EXT4_VALID_FS))
2618 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
2619 "running e2fsck is recommended");
2620 else if (sbi->s_mount_state & EXT4_ERROR_FS)
2621 ext4_msg(sb, KERN_WARNING,
2622 "warning: mounting fs with errors, "
2623 "running e2fsck is recommended");
2624 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
2625 le16_to_cpu(es->s_mnt_count) >=
2626 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
2627 ext4_msg(sb, KERN_WARNING,
2628 "warning: maximal mount count reached, "
2629 "running e2fsck is recommended");
2630 else if (le32_to_cpu(es->s_checkinterval) &&
2631 (ext4_get_tstamp(es, s_lastcheck) +
2632 le32_to_cpu(es->s_checkinterval) <= ktime_get_real_seconds()))
2633 ext4_msg(sb, KERN_WARNING,
2634 "warning: checktime reached, "
2635 "running e2fsck is recommended");
2636 if (!sbi->s_journal)
2637 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
2638 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
2639 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
2640 le16_add_cpu(&es->s_mnt_count, 1);
2641 ext4_update_tstamp(es, s_mtime);
2642 if (sbi->s_journal)
2643 ext4_set_feature_journal_needs_recovery(sb);
2644
2645 err = ext4_commit_super(sb, 1);
2646 done:
2647 if (test_opt(sb, DEBUG))
2648 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
2649 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
2650 sb->s_blocksize,
2651 sbi->s_groups_count,
2652 EXT4_BLOCKS_PER_GROUP(sb),
2653 EXT4_INODES_PER_GROUP(sb),
2654 sbi->s_mount_opt, sbi->s_mount_opt2);
2655
2656 cleancache_init_fs(sb);
2657 return err;
2658 }
2659
2660 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
2661 {
2662 struct ext4_sb_info *sbi = EXT4_SB(sb);
2663 struct flex_groups **old_groups, **new_groups;
2664 int size, i, j;
2665
2666 if (!sbi->s_log_groups_per_flex)
2667 return 0;
2668
2669 size = ext4_flex_group(sbi, ngroup - 1) + 1;
2670 if (size <= sbi->s_flex_groups_allocated)
2671 return 0;
2672
2673 new_groups = kvzalloc(roundup_pow_of_two(size *
2674 sizeof(*sbi->s_flex_groups)), GFP_KERNEL);
2675 if (!new_groups) {
2676 ext4_msg(sb, KERN_ERR,
2677 "not enough memory for %d flex group pointers", size);
2678 return -ENOMEM;
2679 }
2680 for (i = sbi->s_flex_groups_allocated; i < size; i++) {
2681 new_groups[i] = kvzalloc(roundup_pow_of_two(
2682 sizeof(struct flex_groups)),
2683 GFP_KERNEL);
2684 if (!new_groups[i]) {
2685 for (j = sbi->s_flex_groups_allocated; j < i; j++)
2686 kvfree(new_groups[j]);
2687 kvfree(new_groups);
2688 ext4_msg(sb, KERN_ERR,
2689 "not enough memory for %d flex groups", size);
2690 return -ENOMEM;
2691 }
2692 }
2693 rcu_read_lock();
2694 old_groups = rcu_dereference(sbi->s_flex_groups);
2695 if (old_groups)
2696 memcpy(new_groups, old_groups,
2697 (sbi->s_flex_groups_allocated *
2698 sizeof(struct flex_groups *)));
2699 rcu_read_unlock();
2700 rcu_assign_pointer(sbi->s_flex_groups, new_groups);
2701 sbi->s_flex_groups_allocated = size;
2702 if (old_groups)
2703 ext4_kvfree_array_rcu(old_groups);
2704 return 0;
2705 }
2706
2707 static int ext4_fill_flex_info(struct super_block *sb)
2708 {
2709 struct ext4_sb_info *sbi = EXT4_SB(sb);
2710 struct ext4_group_desc *gdp = NULL;
2711 struct flex_groups *fg;
2712 ext4_group_t flex_group;
2713 int i, err;
2714
2715 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
2716 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
2717 sbi->s_log_groups_per_flex = 0;
2718 return 1;
2719 }
2720
2721 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
2722 if (err)
2723 goto failed;
2724
2725 for (i = 0; i < sbi->s_groups_count; i++) {
2726 gdp = ext4_get_group_desc(sb, i, NULL);
2727
2728 flex_group = ext4_flex_group(sbi, i);
2729 fg = sbi_array_rcu_deref(sbi, s_flex_groups, flex_group);
2730 atomic_add(ext4_free_inodes_count(sb, gdp), &fg->free_inodes);
2731 atomic64_add(ext4_free_group_clusters(sb, gdp),
2732 &fg->free_clusters);
2733 atomic_add(ext4_used_dirs_count(sb, gdp), &fg->used_dirs);
2734 }
2735
2736 return 1;
2737 failed:
2738 return 0;
2739 }
2740
2741 static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
2742 struct ext4_group_desc *gdp)
2743 {
2744 int offset = offsetof(struct ext4_group_desc, bg_checksum);
2745 __u16 crc = 0;
2746 __le32 le_group = cpu_to_le32(block_group);
2747 struct ext4_sb_info *sbi = EXT4_SB(sb);
2748
2749 if (ext4_has_metadata_csum(sbi->s_sb)) {
2750 /* Use new metadata_csum algorithm */
2751 __u32 csum32;
2752 __u16 dummy_csum = 0;
2753
2754 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2755 sizeof(le_group));
2756 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset);
2757 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum,
2758 sizeof(dummy_csum));
2759 offset += sizeof(dummy_csum);
2760 if (offset < sbi->s_desc_size)
2761 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset,
2762 sbi->s_desc_size - offset);
2763
2764 crc = csum32 & 0xFFFF;
2765 goto out;
2766 }
2767
2768 /* old crc16 code */
2769 if (!ext4_has_feature_gdt_csum(sb))
2770 return 0;
2771
2772 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2773 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2774 crc = crc16(crc, (__u8 *)gdp, offset);
2775 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2776 /* for checksum of struct ext4_group_desc do the rest...*/
2777 if (ext4_has_feature_64bit(sb) &&
2778 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2779 crc = crc16(crc, (__u8 *)gdp + offset,
2780 le16_to_cpu(sbi->s_es->s_desc_size) -
2781 offset);
2782
2783 out:
2784 return cpu_to_le16(crc);
2785 }
2786
2787 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2788 struct ext4_group_desc *gdp)
2789 {
2790 if (ext4_has_group_desc_csum(sb) &&
2791 (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
2792 return 0;
2793
2794 return 1;
2795 }
2796
2797 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2798 struct ext4_group_desc *gdp)
2799 {
2800 if (!ext4_has_group_desc_csum(sb))
2801 return;
2802 gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
2803 }
2804
2805 /* Called at mount-time, super-block is locked */
2806 static int ext4_check_descriptors(struct super_block *sb,
2807 ext4_fsblk_t sb_block,
2808 ext4_group_t *first_not_zeroed)
2809 {
2810 struct ext4_sb_info *sbi = EXT4_SB(sb);
2811 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2812 ext4_fsblk_t last_block;
2813 ext4_fsblk_t last_bg_block = sb_block + ext4_bg_num_gdb(sb, 0);
2814 ext4_fsblk_t block_bitmap;
2815 ext4_fsblk_t inode_bitmap;
2816 ext4_fsblk_t inode_table;
2817 int flexbg_flag = 0;
2818 ext4_group_t i, grp = sbi->s_groups_count;
2819
2820 if (ext4_has_feature_flex_bg(sb))
2821 flexbg_flag = 1;
2822
2823 ext4_debug("Checking group descriptors");
2824
2825 for (i = 0; i < sbi->s_groups_count; i++) {
2826 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2827
2828 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2829 last_block = ext4_blocks_count(sbi->s_es) - 1;
2830 else
2831 last_block = first_block +
2832 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2833
2834 if ((grp == sbi->s_groups_count) &&
2835 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2836 grp = i;
2837
2838 block_bitmap = ext4_block_bitmap(sb, gdp);
2839 if (block_bitmap == sb_block) {
2840 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2841 "Block bitmap for group %u overlaps "
2842 "superblock", i);
2843 if (!sb_rdonly(sb))
2844 return 0;
2845 }
2846 if (block_bitmap >= sb_block + 1 &&
2847 block_bitmap <= last_bg_block) {
2848 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2849 "Block bitmap for group %u overlaps "
2850 "block group descriptors", i);
2851 if (!sb_rdonly(sb))
2852 return 0;
2853 }
2854 if (block_bitmap < first_block || block_bitmap > last_block) {
2855 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2856 "Block bitmap for group %u not in group "
2857 "(block %llu)!", i, block_bitmap);
2858 return 0;
2859 }
2860 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2861 if (inode_bitmap == sb_block) {
2862 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2863 "Inode bitmap for group %u overlaps "
2864 "superblock", i);
2865 if (!sb_rdonly(sb))
2866 return 0;
2867 }
2868 if (inode_bitmap >= sb_block + 1 &&
2869 inode_bitmap <= last_bg_block) {
2870 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2871 "Inode bitmap for group %u overlaps "
2872 "block group descriptors", i);
2873 if (!sb_rdonly(sb))
2874 return 0;
2875 }
2876 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2877 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2878 "Inode bitmap for group %u not in group "
2879 "(block %llu)!", i, inode_bitmap);
2880 return 0;
2881 }
2882 inode_table = ext4_inode_table(sb, gdp);
2883 if (inode_table == sb_block) {
2884 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2885 "Inode table for group %u overlaps "
2886 "superblock", i);
2887 if (!sb_rdonly(sb))
2888 return 0;
2889 }
2890 if (inode_table >= sb_block + 1 &&
2891 inode_table <= last_bg_block) {
2892 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2893 "Inode table for group %u overlaps "
2894 "block group descriptors", i);
2895 if (!sb_rdonly(sb))
2896 return 0;
2897 }
2898 if (inode_table < first_block ||
2899 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2900 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2901 "Inode table for group %u not in group "
2902 "(block %llu)!", i, inode_table);
2903 return 0;
2904 }
2905 ext4_lock_group(sb, i);
2906 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2907 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2908 "Checksum for group %u failed (%u!=%u)",
2909 i, le16_to_cpu(ext4_group_desc_csum(sb, i,
2910 gdp)), le16_to_cpu(gdp->bg_checksum));
2911 if (!sb_rdonly(sb)) {
2912 ext4_unlock_group(sb, i);
2913 return 0;
2914 }
2915 }
2916 ext4_unlock_group(sb, i);
2917 if (!flexbg_flag)
2918 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2919 }
2920 if (NULL != first_not_zeroed)
2921 *first_not_zeroed = grp;
2922 return 1;
2923 }
2924
2925 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2926 * the superblock) which were deleted from all directories, but held open by
2927 * a process at the time of a crash. We walk the list and try to delete these
2928 * inodes at recovery time (only with a read-write filesystem).
2929 *
2930 * In order to keep the orphan inode chain consistent during traversal (in
2931 * case of crash during recovery), we link each inode into the superblock
2932 * orphan list_head and handle it the same way as an inode deletion during
2933 * normal operation (which journals the operations for us).
2934 *
2935 * We only do an iget() and an iput() on each inode, which is very safe if we
2936 * accidentally point at an in-use or already deleted inode. The worst that
2937 * can happen in this case is that we get a "bit already cleared" message from
2938 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2939 * e2fsck was run on this filesystem, and it must have already done the orphan
2940 * inode cleanup for us, so we can safely abort without any further action.
2941 */
2942 static void ext4_orphan_cleanup(struct super_block *sb,
2943 struct ext4_super_block *es)
2944 {
2945 unsigned int s_flags = sb->s_flags;
2946 int ret, nr_orphans = 0, nr_truncates = 0;
2947 #ifdef CONFIG_QUOTA
2948 int quota_update = 0;
2949 int i;
2950 #endif
2951 if (!es->s_last_orphan) {
2952 jbd_debug(4, "no orphan inodes to clean up\n");
2953 return;
2954 }
2955
2956 if (bdev_read_only(sb->s_bdev)) {
2957 ext4_msg(sb, KERN_ERR, "write access "
2958 "unavailable, skipping orphan cleanup");
2959 return;
2960 }
2961
2962 /* Check if feature set would not allow a r/w mount */
2963 if (!ext4_feature_set_ok(sb, 0)) {
2964 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2965 "unknown ROCOMPAT features");
2966 return;
2967 }
2968
2969 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2970 /* don't clear list on RO mount w/ errors */
2971 if (es->s_last_orphan && !(s_flags & SB_RDONLY)) {
2972 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
2973 "clearing orphan list.\n");
2974 es->s_last_orphan = 0;
2975 }
2976 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2977 return;
2978 }
2979
2980 if (s_flags & SB_RDONLY) {
2981 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2982 sb->s_flags &= ~SB_RDONLY;
2983 }
2984 #ifdef CONFIG_QUOTA
2985 /* Needed for iput() to work correctly and not trash data */
2986 sb->s_flags |= SB_ACTIVE;
2987
2988 /*
2989 * Turn on quotas which were not enabled for read-only mounts if
2990 * filesystem has quota feature, so that they are updated correctly.
2991 */
2992 if (ext4_has_feature_quota(sb) && (s_flags & SB_RDONLY)) {
2993 int ret = ext4_enable_quotas(sb);
2994
2995 if (!ret)
2996 quota_update = 1;
2997 else
2998 ext4_msg(sb, KERN_ERR,
2999 "Cannot turn on quotas: error %d", ret);
3000 }
3001
3002 /* Turn on journaled quotas used for old sytle */
3003 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
3004 if (EXT4_SB(sb)->s_qf_names[i]) {
3005 int ret = ext4_quota_on_mount(sb, i);
3006
3007 if (!ret)
3008 quota_update = 1;
3009 else
3010 ext4_msg(sb, KERN_ERR,
3011 "Cannot turn on journaled "
3012 "quota: type %d: error %d", i, ret);
3013 }
3014 }
3015 #endif
3016
3017 while (es->s_last_orphan) {
3018 struct inode *inode;
3019
3020 /*
3021 * We may have encountered an error during cleanup; if
3022 * so, skip the rest.
3023 */
3024 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
3025 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
3026 es->s_last_orphan = 0;
3027 break;
3028 }
3029
3030 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
3031 if (IS_ERR(inode)) {
3032 es->s_last_orphan = 0;
3033 break;
3034 }
3035
3036 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
3037 dquot_initialize(inode);
3038 if (inode->i_nlink) {
3039 if (test_opt(sb, DEBUG))
3040 ext4_msg(sb, KERN_DEBUG,
3041 "%s: truncating inode %lu to %lld bytes",
3042 __func__, inode->i_ino, inode->i_size);
3043 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
3044 inode->i_ino, inode->i_size);
3045 inode_lock(inode);
3046 truncate_inode_pages(inode->i_mapping, inode->i_size);
3047 ret = ext4_truncate(inode);
3048 if (ret)
3049 ext4_std_error(inode->i_sb, ret);
3050 inode_unlock(inode);
3051 nr_truncates++;
3052 } else {
3053 if (test_opt(sb, DEBUG))
3054 ext4_msg(sb, KERN_DEBUG,
3055 "%s: deleting unreferenced inode %lu",
3056 __func__, inode->i_ino);
3057 jbd_debug(2, "deleting unreferenced inode %lu\n",
3058 inode->i_ino);
3059 nr_orphans++;
3060 }
3061 iput(inode); /* The delete magic happens here! */
3062 }
3063
3064 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
3065
3066 if (nr_orphans)
3067 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
3068 PLURAL(nr_orphans));
3069 if (nr_truncates)
3070 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
3071 PLURAL(nr_truncates));
3072 #ifdef CONFIG_QUOTA
3073 /* Turn off quotas if they were enabled for orphan cleanup */
3074 if (quota_update) {
3075 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
3076 if (sb_dqopt(sb)->files[i])
3077 dquot_quota_off(sb, i);
3078 }
3079 }
3080 #endif
3081 sb->s_flags = s_flags; /* Restore SB_RDONLY status */
3082 }
3083
3084 /*
3085 * Maximal extent format file size.
3086 * Resulting logical blkno at s_maxbytes must fit in our on-disk
3087 * extent format containers, within a sector_t, and within i_blocks
3088 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
3089 * so that won't be a limiting factor.
3090 *
3091 * However there is other limiting factor. We do store extents in the form
3092 * of starting block and length, hence the resulting length of the extent
3093 * covering maximum file size must fit into on-disk format containers as
3094 * well. Given that length is always by 1 unit bigger than max unit (because
3095 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
3096 *
3097 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
3098 */
3099 static loff_t ext4_max_size(int blkbits, int has_huge_files)
3100 {
3101 loff_t res;
3102 loff_t upper_limit = MAX_LFS_FILESIZE;
3103
3104 BUILD_BUG_ON(sizeof(blkcnt_t) < sizeof(u64));
3105
3106 if (!has_huge_files) {
3107 upper_limit = (1LL << 32) - 1;
3108
3109 /* total blocks in file system block size */
3110 upper_limit >>= (blkbits - 9);
3111 upper_limit <<= blkbits;
3112 }
3113
3114 /*
3115 * 32-bit extent-start container, ee_block. We lower the maxbytes
3116 * by one fs block, so ee_len can cover the extent of maximum file
3117 * size
3118 */
3119 res = (1LL << 32) - 1;
3120 res <<= blkbits;
3121
3122 /* Sanity check against vm- & vfs- imposed limits */
3123 if (res > upper_limit)
3124 res = upper_limit;
3125
3126 return res;
3127 }
3128
3129 /*
3130 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
3131 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
3132 * We need to be 1 filesystem block less than the 2^48 sector limit.
3133 */
3134 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
3135 {
3136 loff_t res = EXT4_NDIR_BLOCKS;
3137 int meta_blocks;
3138 loff_t upper_limit;
3139 /* This is calculated to be the largest file size for a dense, block
3140 * mapped file such that the file's total number of 512-byte sectors,
3141 * including data and all indirect blocks, does not exceed (2^48 - 1).
3142 *
3143 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
3144 * number of 512-byte sectors of the file.
3145 */
3146
3147 if (!has_huge_files) {
3148 /*
3149 * !has_huge_files or implies that the inode i_block field
3150 * represents total file blocks in 2^32 512-byte sectors ==
3151 * size of vfs inode i_blocks * 8
3152 */
3153 upper_limit = (1LL << 32) - 1;
3154
3155 /* total blocks in file system block size */
3156 upper_limit >>= (bits - 9);
3157
3158 } else {
3159 /*
3160 * We use 48 bit ext4_inode i_blocks
3161 * With EXT4_HUGE_FILE_FL set the i_blocks
3162 * represent total number of blocks in
3163 * file system block size
3164 */
3165 upper_limit = (1LL << 48) - 1;
3166
3167 }
3168
3169 /* indirect blocks */
3170 meta_blocks = 1;
3171 /* double indirect blocks */
3172 meta_blocks += 1 + (1LL << (bits-2));
3173 /* tripple indirect blocks */
3174 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
3175
3176 upper_limit -= meta_blocks;
3177 upper_limit <<= bits;
3178
3179 res += 1LL << (bits-2);
3180 res += 1LL << (2*(bits-2));
3181 res += 1LL << (3*(bits-2));
3182 res <<= bits;
3183 if (res > upper_limit)
3184 res = upper_limit;
3185
3186 if (res > MAX_LFS_FILESIZE)
3187 res = MAX_LFS_FILESIZE;
3188
3189 return res;
3190 }
3191
3192 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
3193 ext4_fsblk_t logical_sb_block, int nr)
3194 {
3195 struct ext4_sb_info *sbi = EXT4_SB(sb);
3196 ext4_group_t bg, first_meta_bg;
3197 int has_super = 0;
3198
3199 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
3200
3201 if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
3202 return logical_sb_block + nr + 1;
3203 bg = sbi->s_desc_per_block * nr;
3204 if (ext4_bg_has_super(sb, bg))
3205 has_super = 1;
3206
3207 /*
3208 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
3209 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
3210 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
3211 * compensate.
3212 */
3213 if (sb->s_blocksize == 1024 && nr == 0 &&
3214 le32_to_cpu(sbi->s_es->s_first_data_block) == 0)
3215 has_super++;
3216
3217 return (has_super + ext4_group_first_block_no(sb, bg));
3218 }
3219
3220 /**
3221 * ext4_get_stripe_size: Get the stripe size.
3222 * @sbi: In memory super block info
3223 *
3224 * If we have specified it via mount option, then
3225 * use the mount option value. If the value specified at mount time is
3226 * greater than the blocks per group use the super block value.
3227 * If the super block value is greater than blocks per group return 0.
3228 * Allocator needs it be less than blocks per group.
3229 *
3230 */
3231 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
3232 {
3233 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
3234 unsigned long stripe_width =
3235 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3236 int ret;
3237
3238 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3239 ret = sbi->s_stripe;
3240 else if (stripe_width && stripe_width <= sbi->s_blocks_per_group)
3241 ret = stripe_width;
3242 else if (stride && stride <= sbi->s_blocks_per_group)
3243 ret = stride;
3244 else
3245 ret = 0;
3246
3247 /*
3248 * If the stripe width is 1, this makes no sense and
3249 * we set it to 0 to turn off stripe handling code.
3250 */
3251 if (ret <= 1)
3252 ret = 0;
3253
3254 return ret;
3255 }
3256
3257 /*
3258 * Check whether this filesystem can be mounted based on
3259 * the features present and the RDONLY/RDWR mount requested.
3260 * Returns 1 if this filesystem can be mounted as requested,
3261 * 0 if it cannot be.
3262 */
3263 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
3264 {
3265 if (ext4_has_unknown_ext4_incompat_features(sb)) {
3266 ext4_msg(sb, KERN_ERR,
3267 "Couldn't mount because of "
3268 "unsupported optional features (%x)",
3269 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
3270 ~EXT4_FEATURE_INCOMPAT_SUPP));
3271 return 0;
3272 }
3273
3274 #ifndef CONFIG_UNICODE
3275 if (ext4_has_feature_casefold(sb)) {
3276 ext4_msg(sb, KERN_ERR,
3277 "Filesystem with casefold feature cannot be "
3278 "mounted without CONFIG_UNICODE");
3279 return 0;
3280 }
3281 #endif
3282
3283 if (readonly)
3284 return 1;
3285
3286 if (ext4_has_feature_readonly(sb)) {
3287 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
3288 sb->s_flags |= SB_RDONLY;
3289 return 1;
3290 }
3291
3292 /* Check that feature set is OK for a read-write mount */
3293 if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
3294 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
3295 "unsupported optional features (%x)",
3296 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
3297 ~EXT4_FEATURE_RO_COMPAT_SUPP));
3298 return 0;
3299 }
3300 if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
3301 ext4_msg(sb, KERN_ERR,
3302 "Can't support bigalloc feature without "
3303 "extents feature\n");
3304 return 0;
3305 }
3306
3307 #if !IS_ENABLED(CONFIG_QUOTA) || !IS_ENABLED(CONFIG_QFMT_V2)
3308 if (!readonly && (ext4_has_feature_quota(sb) ||
3309 ext4_has_feature_project(sb))) {
3310 ext4_msg(sb, KERN_ERR,
3311 "The kernel was not built with CONFIG_QUOTA and CONFIG_QFMT_V2");
3312 return 0;
3313 }
3314 #endif /* CONFIG_QUOTA */
3315 return 1;
3316 }
3317
3318 /*
3319 * This function is called once a day if we have errors logged
3320 * on the file system
3321 */
3322 static void print_daily_error_info(struct timer_list *t)
3323 {
3324 struct ext4_sb_info *sbi = from_timer(sbi, t, s_err_report);
3325 struct super_block *sb = sbi->s_sb;
3326 struct ext4_super_block *es = sbi->s_es;
3327
3328 if (es->s_error_count)
3329 /* fsck newer than v1.41.13 is needed to clean this condition. */
3330 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
3331 le32_to_cpu(es->s_error_count));
3332 if (es->s_first_error_time) {
3333 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %llu: %.*s:%d",
3334 sb->s_id,
3335 ext4_get_tstamp(es, s_first_error_time),
3336 (int) sizeof(es->s_first_error_func),
3337 es->s_first_error_func,
3338 le32_to_cpu(es->s_first_error_line));
3339 if (es->s_first_error_ino)
3340 printk(KERN_CONT ": inode %u",
3341 le32_to_cpu(es->s_first_error_ino));
3342 if (es->s_first_error_block)
3343 printk(KERN_CONT ": block %llu", (unsigned long long)
3344 le64_to_cpu(es->s_first_error_block));
3345 printk(KERN_CONT "\n");
3346 }
3347 if (es->s_last_error_time) {
3348 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %llu: %.*s:%d",
3349 sb->s_id,
3350 ext4_get_tstamp(es, s_last_error_time),
3351 (int) sizeof(es->s_last_error_func),
3352 es->s_last_error_func,
3353 le32_to_cpu(es->s_last_error_line));
3354 if (es->s_last_error_ino)
3355 printk(KERN_CONT ": inode %u",
3356 le32_to_cpu(es->s_last_error_ino));
3357 if (es->s_last_error_block)
3358 printk(KERN_CONT ": block %llu", (unsigned long long)
3359 le64_to_cpu(es->s_last_error_block));
3360 printk(KERN_CONT "\n");
3361 }
3362 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
3363 }
3364
3365 /* Find next suitable group and run ext4_init_inode_table */
3366 static int ext4_run_li_request(struct ext4_li_request *elr)
3367 {
3368 struct ext4_group_desc *gdp = NULL;
3369 struct super_block *sb = elr->lr_super;
3370 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3371 ext4_group_t group = elr->lr_next_group;
3372 unsigned long timeout = 0;
3373 unsigned int prefetch_ios = 0;
3374 int ret = 0;
3375
3376 if (elr->lr_mode == EXT4_LI_MODE_PREFETCH_BBITMAP) {
3377 elr->lr_next_group = ext4_mb_prefetch(sb, group,
3378 EXT4_SB(sb)->s_mb_prefetch, &prefetch_ios);
3379 if (prefetch_ios)
3380 ext4_mb_prefetch_fini(sb, elr->lr_next_group,
3381 prefetch_ios);
3382 trace_ext4_prefetch_bitmaps(sb, group, elr->lr_next_group,
3383 prefetch_ios);
3384 if (group >= elr->lr_next_group) {
3385 ret = 1;
3386 if (elr->lr_first_not_zeroed != ngroups &&
3387 !sb_rdonly(sb) && test_opt(sb, INIT_INODE_TABLE)) {
3388 elr->lr_next_group = elr->lr_first_not_zeroed;
3389 elr->lr_mode = EXT4_LI_MODE_ITABLE;
3390 ret = 0;
3391 }
3392 }
3393 return ret;
3394 }
3395
3396 for (; group < ngroups; group++) {
3397 gdp = ext4_get_group_desc(sb, group, NULL);
3398 if (!gdp) {
3399 ret = 1;
3400 break;
3401 }
3402
3403 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3404 break;
3405 }
3406
3407 if (group >= ngroups)
3408 ret = 1;
3409
3410 if (!ret) {
3411 timeout = jiffies;
3412 ret = ext4_init_inode_table(sb, group,
3413 elr->lr_timeout ? 0 : 1);
3414 trace_ext4_lazy_itable_init(sb, group);
3415 if (elr->lr_timeout == 0) {
3416 timeout = (jiffies - timeout) *
3417 EXT4_SB(elr->lr_super)->s_li_wait_mult;
3418 elr->lr_timeout = timeout;
3419 }
3420 elr->lr_next_sched = jiffies + elr->lr_timeout;
3421 elr->lr_next_group = group + 1;
3422 }
3423 return ret;
3424 }
3425
3426 /*
3427 * Remove lr_request from the list_request and free the
3428 * request structure. Should be called with li_list_mtx held
3429 */
3430 static void ext4_remove_li_request(struct ext4_li_request *elr)
3431 {
3432 if (!elr)
3433 return;
3434
3435 list_del(&elr->lr_request);
3436 EXT4_SB(elr->lr_super)->s_li_request = NULL;
3437 kfree(elr);
3438 }
3439
3440 static void ext4_unregister_li_request(struct super_block *sb)
3441 {
3442 mutex_lock(&ext4_li_mtx);
3443 if (!ext4_li_info) {
3444 mutex_unlock(&ext4_li_mtx);
3445 return;
3446 }
3447
3448 mutex_lock(&ext4_li_info->li_list_mtx);
3449 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
3450 mutex_unlock(&ext4_li_info->li_list_mtx);
3451 mutex_unlock(&ext4_li_mtx);
3452 }
3453
3454 static struct task_struct *ext4_lazyinit_task;
3455
3456 /*
3457 * This is the function where ext4lazyinit thread lives. It walks
3458 * through the request list searching for next scheduled filesystem.
3459 * When such a fs is found, run the lazy initialization request
3460 * (ext4_rn_li_request) and keep track of the time spend in this
3461 * function. Based on that time we compute next schedule time of
3462 * the request. When walking through the list is complete, compute
3463 * next waking time and put itself into sleep.
3464 */
3465 static int ext4_lazyinit_thread(void *arg)
3466 {
3467 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
3468 struct list_head *pos, *n;
3469 struct ext4_li_request *elr;
3470 unsigned long next_wakeup, cur;
3471
3472 BUG_ON(NULL == eli);
3473
3474 cont_thread:
3475 while (true) {
3476 next_wakeup = MAX_JIFFY_OFFSET;
3477
3478 mutex_lock(&eli->li_list_mtx);
3479 if (list_empty(&eli->li_request_list)) {
3480 mutex_unlock(&eli->li_list_mtx);
3481 goto exit_thread;
3482 }
3483 list_for_each_safe(pos, n, &eli->li_request_list) {
3484 int err = 0;
3485 int progress = 0;
3486 elr = list_entry(pos, struct ext4_li_request,
3487 lr_request);
3488
3489 if (time_before(jiffies, elr->lr_next_sched)) {
3490 if (time_before(elr->lr_next_sched, next_wakeup))
3491 next_wakeup = elr->lr_next_sched;
3492 continue;
3493 }
3494 if (down_read_trylock(&elr->lr_super->s_umount)) {
3495 if (sb_start_write_trylock(elr->lr_super)) {
3496 progress = 1;
3497 /*
3498 * We hold sb->s_umount, sb can not
3499 * be removed from the list, it is
3500 * now safe to drop li_list_mtx
3501 */
3502 mutex_unlock(&eli->li_list_mtx);
3503 err = ext4_run_li_request(elr);
3504 sb_end_write(elr->lr_super);
3505 mutex_lock(&eli->li_list_mtx);
3506 n = pos->next;
3507 }
3508 up_read((&elr->lr_super->s_umount));
3509 }
3510 /* error, remove the lazy_init job */
3511 if (err) {
3512 ext4_remove_li_request(elr);
3513 continue;
3514 }
3515 if (!progress) {
3516 elr->lr_next_sched = jiffies +
3517 (prandom_u32()
3518 % (EXT4_DEF_LI_MAX_START_DELAY * HZ));
3519 }
3520 if (time_before(elr->lr_next_sched, next_wakeup))
3521 next_wakeup = elr->lr_next_sched;
3522 }
3523 mutex_unlock(&eli->li_list_mtx);
3524
3525 try_to_freeze();
3526
3527 cur = jiffies;
3528 if ((time_after_eq(cur, next_wakeup)) ||
3529 (MAX_JIFFY_OFFSET == next_wakeup)) {
3530 cond_resched();
3531 continue;
3532 }
3533
3534 schedule_timeout_interruptible(next_wakeup - cur);
3535
3536 if (kthread_should_stop()) {
3537 ext4_clear_request_list();
3538 goto exit_thread;
3539 }
3540 }
3541
3542 exit_thread:
3543 /*
3544 * It looks like the request list is empty, but we need
3545 * to check it under the li_list_mtx lock, to prevent any
3546 * additions into it, and of course we should lock ext4_li_mtx
3547 * to atomically free the list and ext4_li_info, because at
3548 * this point another ext4 filesystem could be registering
3549 * new one.
3550 */
3551 mutex_lock(&ext4_li_mtx);
3552 mutex_lock(&eli->li_list_mtx);
3553 if (!list_empty(&eli->li_request_list)) {
3554 mutex_unlock(&eli->li_list_mtx);
3555 mutex_unlock(&ext4_li_mtx);
3556 goto cont_thread;
3557 }
3558 mutex_unlock(&eli->li_list_mtx);
3559 kfree(ext4_li_info);
3560 ext4_li_info = NULL;
3561 mutex_unlock(&ext4_li_mtx);
3562
3563 return 0;
3564 }
3565
3566 static void ext4_clear_request_list(void)
3567 {
3568 struct list_head *pos, *n;
3569 struct ext4_li_request *elr;
3570
3571 mutex_lock(&ext4_li_info->li_list_mtx);
3572 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3573 elr = list_entry(pos, struct ext4_li_request,
3574 lr_request);
3575 ext4_remove_li_request(elr);
3576 }
3577 mutex_unlock(&ext4_li_info->li_list_mtx);
3578 }
3579
3580 static int ext4_run_lazyinit_thread(void)
3581 {
3582 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3583 ext4_li_info, "ext4lazyinit");
3584 if (IS_ERR(ext4_lazyinit_task)) {
3585 int err = PTR_ERR(ext4_lazyinit_task);
3586 ext4_clear_request_list();
3587 kfree(ext4_li_info);
3588 ext4_li_info = NULL;
3589 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
3590 "initialization thread\n",
3591 err);
3592 return err;
3593 }
3594 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
3595 return 0;
3596 }
3597
3598 /*
3599 * Check whether it make sense to run itable init. thread or not.
3600 * If there is at least one uninitialized inode table, return
3601 * corresponding group number, else the loop goes through all
3602 * groups and return total number of groups.
3603 */
3604 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3605 {
3606 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3607 struct ext4_group_desc *gdp = NULL;
3608
3609 if (!ext4_has_group_desc_csum(sb))
3610 return ngroups;
3611
3612 for (group = 0; group < ngroups; group++) {
3613 gdp = ext4_get_group_desc(sb, group, NULL);
3614 if (!gdp)
3615 continue;
3616
3617 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3618 break;
3619 }
3620
3621 return group;
3622 }
3623
3624 static int ext4_li_info_new(void)
3625 {
3626 struct ext4_lazy_init *eli = NULL;
3627
3628 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3629 if (!eli)
3630 return -ENOMEM;
3631
3632 INIT_LIST_HEAD(&eli->li_request_list);
3633 mutex_init(&eli->li_list_mtx);
3634
3635 eli->li_state |= EXT4_LAZYINIT_QUIT;
3636
3637 ext4_li_info = eli;
3638
3639 return 0;
3640 }
3641
3642 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3643 ext4_group_t start)
3644 {
3645 struct ext4_li_request *elr;
3646
3647 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3648 if (!elr)
3649 return NULL;
3650
3651 elr->lr_super = sb;
3652 elr->lr_first_not_zeroed = start;
3653 if (test_opt(sb, PREFETCH_BLOCK_BITMAPS))
3654 elr->lr_mode = EXT4_LI_MODE_PREFETCH_BBITMAP;
3655 else {
3656 elr->lr_mode = EXT4_LI_MODE_ITABLE;
3657 elr->lr_next_group = start;
3658 }
3659
3660 /*
3661 * Randomize first schedule time of the request to
3662 * spread the inode table initialization requests
3663 * better.
3664 */
3665 elr->lr_next_sched = jiffies + (prandom_u32() %
3666 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
3667 return elr;
3668 }
3669
3670 int ext4_register_li_request(struct super_block *sb,
3671 ext4_group_t first_not_zeroed)
3672 {
3673 struct ext4_sb_info *sbi = EXT4_SB(sb);
3674 struct ext4_li_request *elr = NULL;
3675 ext4_group_t ngroups = sbi->s_groups_count;
3676 int ret = 0;
3677
3678 mutex_lock(&ext4_li_mtx);
3679 if (sbi->s_li_request != NULL) {
3680 /*
3681 * Reset timeout so it can be computed again, because
3682 * s_li_wait_mult might have changed.
3683 */
3684 sbi->s_li_request->lr_timeout = 0;
3685 goto out;
3686 }
3687
3688 if (!test_opt(sb, PREFETCH_BLOCK_BITMAPS) &&
3689 (first_not_zeroed == ngroups || sb_rdonly(sb) ||
3690 !test_opt(sb, INIT_INODE_TABLE)))
3691 goto out;
3692
3693 elr = ext4_li_request_new(sb, first_not_zeroed);
3694 if (!elr) {
3695 ret = -ENOMEM;
3696 goto out;
3697 }
3698
3699 if (NULL == ext4_li_info) {
3700 ret = ext4_li_info_new();
3701 if (ret)
3702 goto out;
3703 }
3704
3705 mutex_lock(&ext4_li_info->li_list_mtx);
3706 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3707 mutex_unlock(&ext4_li_info->li_list_mtx);
3708
3709 sbi->s_li_request = elr;
3710 /*
3711 * set elr to NULL here since it has been inserted to
3712 * the request_list and the removal and free of it is
3713 * handled by ext4_clear_request_list from now on.
3714 */
3715 elr = NULL;
3716
3717 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3718 ret = ext4_run_lazyinit_thread();
3719 if (ret)
3720 goto out;
3721 }
3722 out:
3723 mutex_unlock(&ext4_li_mtx);
3724 if (ret)
3725 kfree(elr);
3726 return ret;
3727 }
3728
3729 /*
3730 * We do not need to lock anything since this is called on
3731 * module unload.
3732 */
3733 static void ext4_destroy_lazyinit_thread(void)
3734 {
3735 /*
3736 * If thread exited earlier
3737 * there's nothing to be done.
3738 */
3739 if (!ext4_li_info || !ext4_lazyinit_task)
3740 return;
3741
3742 kthread_stop(ext4_lazyinit_task);
3743 }
3744
3745 static int set_journal_csum_feature_set(struct super_block *sb)
3746 {
3747 int ret = 1;
3748 int compat, incompat;
3749 struct ext4_sb_info *sbi = EXT4_SB(sb);
3750
3751 if (ext4_has_metadata_csum(sb)) {
3752 /* journal checksum v3 */
3753 compat = 0;
3754 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
3755 } else {
3756 /* journal checksum v1 */
3757 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3758 incompat = 0;
3759 }
3760
3761 jbd2_journal_clear_features(sbi->s_journal,
3762 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3763 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3764 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3765 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3766 ret = jbd2_journal_set_features(sbi->s_journal,
3767 compat, 0,
3768 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3769 incompat);
3770 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3771 ret = jbd2_journal_set_features(sbi->s_journal,
3772 compat, 0,
3773 incompat);
3774 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3775 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3776 } else {
3777 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3778 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3779 }
3780
3781 return ret;
3782 }
3783
3784 /*
3785 * Note: calculating the overhead so we can be compatible with
3786 * historical BSD practice is quite difficult in the face of
3787 * clusters/bigalloc. This is because multiple metadata blocks from
3788 * different block group can end up in the same allocation cluster.
3789 * Calculating the exact overhead in the face of clustered allocation
3790 * requires either O(all block bitmaps) in memory or O(number of block
3791 * groups**2) in time. We will still calculate the superblock for
3792 * older file systems --- and if we come across with a bigalloc file
3793 * system with zero in s_overhead_clusters the estimate will be close to
3794 * correct especially for very large cluster sizes --- but for newer
3795 * file systems, it's better to calculate this figure once at mkfs
3796 * time, and store it in the superblock. If the superblock value is
3797 * present (even for non-bigalloc file systems), we will use it.
3798 */
3799 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3800 char *buf)
3801 {
3802 struct ext4_sb_info *sbi = EXT4_SB(sb);
3803 struct ext4_group_desc *gdp;
3804 ext4_fsblk_t first_block, last_block, b;
3805 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3806 int s, j, count = 0;
3807
3808 if (!ext4_has_feature_bigalloc(sb))
3809 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3810 sbi->s_itb_per_group + 2);
3811
3812 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3813 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3814 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3815 for (i = 0; i < ngroups; i++) {
3816 gdp = ext4_get_group_desc(sb, i, NULL);
3817 b = ext4_block_bitmap(sb, gdp);
3818 if (b >= first_block && b <= last_block) {
3819 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3820 count++;
3821 }
3822 b = ext4_inode_bitmap(sb, gdp);
3823 if (b >= first_block && b <= last_block) {
3824 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3825 count++;
3826 }
3827 b = ext4_inode_table(sb, gdp);
3828 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3829 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3830 int c = EXT4_B2C(sbi, b - first_block);
3831 ext4_set_bit(c, buf);
3832 count++;
3833 }
3834 if (i != grp)
3835 continue;
3836 s = 0;
3837 if (ext4_bg_has_super(sb, grp)) {
3838 ext4_set_bit(s++, buf);
3839 count++;
3840 }
3841 j = ext4_bg_num_gdb(sb, grp);
3842 if (s + j > EXT4_BLOCKS_PER_GROUP(sb)) {
3843 ext4_error(sb, "Invalid number of block group "
3844 "descriptor blocks: %d", j);
3845 j = EXT4_BLOCKS_PER_GROUP(sb) - s;
3846 }
3847 count += j;
3848 for (; j > 0; j--)
3849 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3850 }
3851 if (!count)
3852 return 0;
3853 return EXT4_CLUSTERS_PER_GROUP(sb) -
3854 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3855 }
3856
3857 /*
3858 * Compute the overhead and stash it in sbi->s_overhead
3859 */
3860 int ext4_calculate_overhead(struct super_block *sb)
3861 {
3862 struct ext4_sb_info *sbi = EXT4_SB(sb);
3863 struct ext4_super_block *es = sbi->s_es;
3864 struct inode *j_inode;
3865 unsigned int j_blocks, j_inum = le32_to_cpu(es->s_journal_inum);
3866 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3867 ext4_fsblk_t overhead = 0;
3868 char *buf = (char *) get_zeroed_page(GFP_NOFS);
3869
3870 if (!buf)
3871 return -ENOMEM;
3872
3873 /*
3874 * Compute the overhead (FS structures). This is constant
3875 * for a given filesystem unless the number of block groups
3876 * changes so we cache the previous value until it does.
3877 */
3878
3879 /*
3880 * All of the blocks before first_data_block are overhead
3881 */
3882 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3883
3884 /*
3885 * Add the overhead found in each block group
3886 */
3887 for (i = 0; i < ngroups; i++) {
3888 int blks;
3889
3890 blks = count_overhead(sb, i, buf);
3891 overhead += blks;
3892 if (blks)
3893 memset(buf, 0, PAGE_SIZE);
3894 cond_resched();
3895 }
3896
3897 /*
3898 * Add the internal journal blocks whether the journal has been
3899 * loaded or not
3900 */
3901 if (sbi->s_journal && !sbi->s_journal_bdev)
3902 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_total_len);
3903 else if (ext4_has_feature_journal(sb) && !sbi->s_journal && j_inum) {
3904 /* j_inum for internal journal is non-zero */
3905 j_inode = ext4_get_journal_inode(sb, j_inum);
3906 if (j_inode) {
3907 j_blocks = j_inode->i_size >> sb->s_blocksize_bits;
3908 overhead += EXT4_NUM_B2C(sbi, j_blocks);
3909 iput(j_inode);
3910 } else {
3911 ext4_msg(sb, KERN_ERR, "can't get journal size");
3912 }
3913 }
3914 sbi->s_overhead = overhead;
3915 smp_wmb();
3916 free_page((unsigned long) buf);
3917 return 0;
3918 }
3919
3920 static void ext4_set_resv_clusters(struct super_block *sb)
3921 {
3922 ext4_fsblk_t resv_clusters;
3923 struct ext4_sb_info *sbi = EXT4_SB(sb);
3924
3925 /*
3926 * There's no need to reserve anything when we aren't using extents.
3927 * The space estimates are exact, there are no unwritten extents,
3928 * hole punching doesn't need new metadata... This is needed especially
3929 * to keep ext2/3 backward compatibility.
3930 */
3931 if (!ext4_has_feature_extents(sb))
3932 return;
3933 /*
3934 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3935 * This should cover the situations where we can not afford to run
3936 * out of space like for example punch hole, or converting
3937 * unwritten extents in delalloc path. In most cases such
3938 * allocation would require 1, or 2 blocks, higher numbers are
3939 * very rare.
3940 */
3941 resv_clusters = (ext4_blocks_count(sbi->s_es) >>
3942 sbi->s_cluster_bits);
3943
3944 do_div(resv_clusters, 50);
3945 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3946
3947 atomic64_set(&sbi->s_resv_clusters, resv_clusters);
3948 }
3949
3950 static const char *ext4_quota_mode(struct super_block *sb)
3951 {
3952 #ifdef CONFIG_QUOTA
3953 if (!ext4_quota_capable(sb))
3954 return "none";
3955
3956 if (EXT4_SB(sb)->s_journal && ext4_is_quota_journalled(sb))
3957 return "journalled";
3958 else
3959 return "writeback";
3960 #else
3961 return "disabled";
3962 #endif
3963 }
3964
3965 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3966 {
3967 struct dax_device *dax_dev = fs_dax_get_by_bdev(sb->s_bdev);
3968 char *orig_data = kstrdup(data, GFP_KERNEL);
3969 struct buffer_head *bh, **group_desc;
3970 struct ext4_super_block *es = NULL;
3971 struct ext4_sb_info *sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3972 struct flex_groups **flex_groups;
3973 ext4_fsblk_t block;
3974 ext4_fsblk_t sb_block = get_sb_block(&data);
3975 ext4_fsblk_t logical_sb_block;
3976 unsigned long offset = 0;
3977 unsigned long journal_devnum = 0;
3978 unsigned long def_mount_opts;
3979 struct inode *root;
3980 const char *descr;
3981 int ret = -ENOMEM;
3982 int blocksize, clustersize;
3983 unsigned int db_count;
3984 unsigned int i;
3985 int needs_recovery, has_huge_files;
3986 __u64 blocks_count;
3987 int err = 0;
3988 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3989 ext4_group_t first_not_zeroed;
3990
3991 if ((data && !orig_data) || !sbi)
3992 goto out_free_base;
3993
3994 sbi->s_daxdev = dax_dev;
3995 sbi->s_blockgroup_lock =
3996 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3997 if (!sbi->s_blockgroup_lock)
3998 goto out_free_base;
3999
4000 sb->s_fs_info = sbi;
4001 sbi->s_sb = sb;
4002 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
4003 sbi->s_sb_block = sb_block;
4004 sbi->s_sectors_written_start =
4005 part_stat_read(sb->s_bdev, sectors[STAT_WRITE]);
4006
4007 /* Cleanup superblock name */
4008 strreplace(sb->s_id, '/', '!');
4009
4010 /* -EINVAL is default */
4011 ret = -EINVAL;
4012 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
4013 if (!blocksize) {
4014 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
4015 goto out_fail;
4016 }
4017
4018 /*
4019 * The ext4 superblock will not be buffer aligned for other than 1kB
4020 * block sizes. We need to calculate the offset from buffer start.
4021 */
4022 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
4023 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
4024 offset = do_div(logical_sb_block, blocksize);
4025 } else {
4026 logical_sb_block = sb_block;
4027 }
4028
4029 bh = ext4_sb_bread_unmovable(sb, logical_sb_block);
4030 if (IS_ERR(bh)) {
4031 ext4_msg(sb, KERN_ERR, "unable to read superblock");
4032 ret = PTR_ERR(bh);
4033 goto out_fail;
4034 }
4035 /*
4036 * Note: s_es must be initialized as soon as possible because
4037 * some ext4 macro-instructions depend on its value
4038 */
4039 es = (struct ext4_super_block *) (bh->b_data + offset);
4040 sbi->s_es = es;
4041 sb->s_magic = le16_to_cpu(es->s_magic);
4042 if (sb->s_magic != EXT4_SUPER_MAGIC)
4043 goto cantfind_ext4;
4044 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
4045
4046 /* Warn if metadata_csum and gdt_csum are both set. */
4047 if (ext4_has_feature_metadata_csum(sb) &&
4048 ext4_has_feature_gdt_csum(sb))
4049 ext4_warning(sb, "metadata_csum and uninit_bg are "
4050 "redundant flags; please run fsck.");
4051
4052 /* Check for a known checksum algorithm */
4053 if (!ext4_verify_csum_type(sb, es)) {
4054 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
4055 "unknown checksum algorithm.");
4056 silent = 1;
4057 goto cantfind_ext4;
4058 }
4059
4060 /* Load the checksum driver */
4061 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
4062 if (IS_ERR(sbi->s_chksum_driver)) {
4063 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
4064 ret = PTR_ERR(sbi->s_chksum_driver);
4065 sbi->s_chksum_driver = NULL;
4066 goto failed_mount;
4067 }
4068
4069 /* Check superblock checksum */
4070 if (!ext4_superblock_csum_verify(sb, es)) {
4071 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
4072 "invalid superblock checksum. Run e2fsck?");
4073 silent = 1;
4074 ret = -EFSBADCRC;
4075 goto cantfind_ext4;
4076 }
4077
4078 /* Precompute checksum seed for all metadata */
4079 if (ext4_has_feature_csum_seed(sb))
4080 sbi->s_csum_seed = le32_to_cpu(es->s_checksum_seed);
4081 else if (ext4_has_metadata_csum(sb) || ext4_has_feature_ea_inode(sb))
4082 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
4083 sizeof(es->s_uuid));
4084
4085 /* Set defaults before we parse the mount options */
4086 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
4087 set_opt(sb, INIT_INODE_TABLE);
4088 if (def_mount_opts & EXT4_DEFM_DEBUG)
4089 set_opt(sb, DEBUG);
4090 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
4091 set_opt(sb, GRPID);
4092 if (def_mount_opts & EXT4_DEFM_UID16)
4093 set_opt(sb, NO_UID32);
4094 /* xattr user namespace & acls are now defaulted on */
4095 set_opt(sb, XATTR_USER);
4096 #ifdef CONFIG_EXT4_FS_POSIX_ACL
4097 set_opt(sb, POSIX_ACL);
4098 #endif
4099 if (ext4_has_feature_fast_commit(sb))
4100 set_opt2(sb, JOURNAL_FAST_COMMIT);
4101 /* don't forget to enable journal_csum when metadata_csum is enabled. */
4102 if (ext4_has_metadata_csum(sb))
4103 set_opt(sb, JOURNAL_CHECKSUM);
4104
4105 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
4106 set_opt(sb, JOURNAL_DATA);
4107 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
4108 set_opt(sb, ORDERED_DATA);
4109 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
4110 set_opt(sb, WRITEBACK_DATA);
4111
4112 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
4113 set_opt(sb, ERRORS_PANIC);
4114 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
4115 set_opt(sb, ERRORS_CONT);
4116 else
4117 set_opt(sb, ERRORS_RO);
4118 /* block_validity enabled by default; disable with noblock_validity */
4119 set_opt(sb, BLOCK_VALIDITY);
4120 if (def_mount_opts & EXT4_DEFM_DISCARD)
4121 set_opt(sb, DISCARD);
4122
4123 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
4124 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
4125 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
4126 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
4127 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
4128
4129 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
4130 set_opt(sb, BARRIER);
4131
4132 /*
4133 * enable delayed allocation by default
4134 * Use -o nodelalloc to turn it off
4135 */
4136 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
4137 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
4138 set_opt(sb, DELALLOC);
4139
4140 /*
4141 * set default s_li_wait_mult for lazyinit, for the case there is
4142 * no mount option specified.
4143 */
4144 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
4145
4146 if (le32_to_cpu(es->s_log_block_size) >
4147 (EXT4_MAX_BLOCK_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
4148 ext4_msg(sb, KERN_ERR,
4149 "Invalid log block size: %u",
4150 le32_to_cpu(es->s_log_block_size));
4151 goto failed_mount;
4152 }
4153 if (le32_to_cpu(es->s_log_cluster_size) >
4154 (EXT4_MAX_CLUSTER_LOG_SIZE - EXT4_MIN_BLOCK_LOG_SIZE)) {
4155 ext4_msg(sb, KERN_ERR,
4156 "Invalid log cluster size: %u",
4157 le32_to_cpu(es->s_log_cluster_size));
4158 goto failed_mount;
4159 }
4160
4161 blocksize = EXT4_MIN_BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
4162
4163 if (blocksize == PAGE_SIZE)
4164 set_opt(sb, DIOREAD_NOLOCK);
4165
4166 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
4167 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
4168 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
4169 } else {
4170 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
4171 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
4172 if (sbi->s_first_ino < EXT4_GOOD_OLD_FIRST_INO) {
4173 ext4_msg(sb, KERN_ERR, "invalid first ino: %u",
4174 sbi->s_first_ino);
4175 goto failed_mount;
4176 }
4177 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
4178 (!is_power_of_2(sbi->s_inode_size)) ||
4179 (sbi->s_inode_size > blocksize)) {
4180 ext4_msg(sb, KERN_ERR,
4181 "unsupported inode size: %d",
4182 sbi->s_inode_size);
4183 ext4_msg(sb, KERN_ERR, "blocksize: %d", blocksize);
4184 goto failed_mount;
4185 }
4186 /*
4187 * i_atime_extra is the last extra field available for
4188 * [acm]times in struct ext4_inode. Checking for that
4189 * field should suffice to ensure we have extra space
4190 * for all three.
4191 */
4192 if (sbi->s_inode_size >= offsetof(struct ext4_inode, i_atime_extra) +
4193 sizeof(((struct ext4_inode *)0)->i_atime_extra)) {
4194 sb->s_time_gran = 1;
4195 sb->s_time_max = EXT4_EXTRA_TIMESTAMP_MAX;
4196 } else {
4197 sb->s_time_gran = NSEC_PER_SEC;
4198 sb->s_time_max = EXT4_NON_EXTRA_TIMESTAMP_MAX;
4199 }
4200 sb->s_time_min = EXT4_TIMESTAMP_MIN;
4201 }
4202 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
4203 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4204 EXT4_GOOD_OLD_INODE_SIZE;
4205 if (ext4_has_feature_extra_isize(sb)) {
4206 unsigned v, max = (sbi->s_inode_size -
4207 EXT4_GOOD_OLD_INODE_SIZE);
4208
4209 v = le16_to_cpu(es->s_want_extra_isize);
4210 if (v > max) {
4211 ext4_msg(sb, KERN_ERR,
4212 "bad s_want_extra_isize: %d", v);
4213 goto failed_mount;
4214 }
4215 if (sbi->s_want_extra_isize < v)
4216 sbi->s_want_extra_isize = v;
4217
4218 v = le16_to_cpu(es->s_min_extra_isize);
4219 if (v > max) {
4220 ext4_msg(sb, KERN_ERR,
4221 "bad s_min_extra_isize: %d", v);
4222 goto failed_mount;
4223 }
4224 if (sbi->s_want_extra_isize < v)
4225 sbi->s_want_extra_isize = v;
4226 }
4227 }
4228
4229 if (sbi->s_es->s_mount_opts[0]) {
4230 char *s_mount_opts = kstrndup(sbi->s_es->s_mount_opts,
4231 sizeof(sbi->s_es->s_mount_opts),
4232 GFP_KERNEL);
4233 if (!s_mount_opts)
4234 goto failed_mount;
4235 if (!parse_options(s_mount_opts, sb, &journal_devnum,
4236 &journal_ioprio, 0)) {
4237 ext4_msg(sb, KERN_WARNING,
4238 "failed to parse options in superblock: %s",
4239 s_mount_opts);
4240 }
4241 kfree(s_mount_opts);
4242 }
4243 sbi->s_def_mount_opt = sbi->s_mount_opt;
4244 if (!parse_options((char *) data, sb, &journal_devnum,
4245 &journal_ioprio, 0))
4246 goto failed_mount;
4247
4248 #ifdef CONFIG_UNICODE
4249 if (ext4_has_feature_casefold(sb) && !sb->s_encoding) {
4250 const struct ext4_sb_encodings *encoding_info;
4251 struct unicode_map *encoding;
4252 __u16 encoding_flags;
4253
4254 if (ext4_has_feature_encrypt(sb)) {
4255 ext4_msg(sb, KERN_ERR,
4256 "Can't mount with encoding and encryption");
4257 goto failed_mount;
4258 }
4259
4260 if (ext4_sb_read_encoding(es, &encoding_info,
4261 &encoding_flags)) {
4262 ext4_msg(sb, KERN_ERR,
4263 "Encoding requested by superblock is unknown");
4264 goto failed_mount;
4265 }
4266
4267 encoding = utf8_load(encoding_info->version);
4268 if (IS_ERR(encoding)) {
4269 ext4_msg(sb, KERN_ERR,
4270 "can't mount with superblock charset: %s-%s "
4271 "not supported by the kernel. flags: 0x%x.",
4272 encoding_info->name, encoding_info->version,
4273 encoding_flags);
4274 goto failed_mount;
4275 }
4276 ext4_msg(sb, KERN_INFO,"Using encoding defined by superblock: "
4277 "%s-%s with flags 0x%hx", encoding_info->name,
4278 encoding_info->version?:"\b", encoding_flags);
4279
4280 sb->s_encoding = encoding;
4281 sb->s_encoding_flags = encoding_flags;
4282 }
4283 #endif
4284
4285 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4286 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting with data=journal disables delayed allocation, dioread_nolock, O_DIRECT and fast_commit support!\n");
4287 /* can't mount with both data=journal and dioread_nolock. */
4288 clear_opt(sb, DIOREAD_NOLOCK);
4289 clear_opt2(sb, JOURNAL_FAST_COMMIT);
4290 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4291 ext4_msg(sb, KERN_ERR, "can't mount with "
4292 "both data=journal and delalloc");
4293 goto failed_mount;
4294 }
4295 if (test_opt(sb, DAX_ALWAYS)) {
4296 ext4_msg(sb, KERN_ERR, "can't mount with "
4297 "both data=journal and dax");
4298 goto failed_mount;
4299 }
4300 if (ext4_has_feature_encrypt(sb)) {
4301 ext4_msg(sb, KERN_WARNING,
4302 "encrypted files will use data=ordered "
4303 "instead of data journaling mode");
4304 }
4305 if (test_opt(sb, DELALLOC))
4306 clear_opt(sb, DELALLOC);
4307 } else {
4308 sb->s_iflags |= SB_I_CGROUPWB;
4309 }
4310
4311 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
4312 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
4313
4314 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
4315 (ext4_has_compat_features(sb) ||
4316 ext4_has_ro_compat_features(sb) ||
4317 ext4_has_incompat_features(sb)))
4318 ext4_msg(sb, KERN_WARNING,
4319 "feature flags set on rev 0 fs, "
4320 "running e2fsck is recommended");
4321
4322 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
4323 set_opt2(sb, HURD_COMPAT);
4324 if (ext4_has_feature_64bit(sb)) {
4325 ext4_msg(sb, KERN_ERR,
4326 "The Hurd can't support 64-bit file systems");
4327 goto failed_mount;
4328 }
4329
4330 /*
4331 * ea_inode feature uses l_i_version field which is not
4332 * available in HURD_COMPAT mode.
4333 */
4334 if (ext4_has_feature_ea_inode(sb)) {
4335 ext4_msg(sb, KERN_ERR,
4336 "ea_inode feature is not supported for Hurd");
4337 goto failed_mount;
4338 }
4339 }
4340
4341 if (IS_EXT2_SB(sb)) {
4342 if (ext2_feature_set_ok(sb))
4343 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
4344 "using the ext4 subsystem");
4345 else {
4346 /*
4347 * If we're probing be silent, if this looks like
4348 * it's actually an ext[34] filesystem.
4349 */
4350 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
4351 goto failed_mount;
4352 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
4353 "to feature incompatibilities");
4354 goto failed_mount;
4355 }
4356 }
4357
4358 if (IS_EXT3_SB(sb)) {
4359 if (ext3_feature_set_ok(sb))
4360 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
4361 "using the ext4 subsystem");
4362 else {
4363 /*
4364 * If we're probing be silent, if this looks like
4365 * it's actually an ext4 filesystem.
4366 */
4367 if (silent && ext4_feature_set_ok(sb, sb_rdonly(sb)))
4368 goto failed_mount;
4369 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
4370 "to feature incompatibilities");
4371 goto failed_mount;
4372 }
4373 }
4374
4375 /*
4376 * Check feature flags regardless of the revision level, since we
4377 * previously didn't change the revision level when setting the flags,
4378 * so there is a chance incompat flags are set on a rev 0 filesystem.
4379 */
4380 if (!ext4_feature_set_ok(sb, (sb_rdonly(sb))))
4381 goto failed_mount;
4382
4383 if (le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) > (blocksize / 4)) {
4384 ext4_msg(sb, KERN_ERR,
4385 "Number of reserved GDT blocks insanely large: %d",
4386 le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks));
4387 goto failed_mount;
4388 }
4389
4390 if (bdev_dax_supported(sb->s_bdev, blocksize))
4391 set_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags);
4392
4393 if (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) {
4394 if (ext4_has_feature_inline_data(sb)) {
4395 ext4_msg(sb, KERN_ERR, "Cannot use DAX on a filesystem"
4396 " that may contain inline data");
4397 goto failed_mount;
4398 }
4399 if (!test_bit(EXT4_FLAGS_BDEV_IS_DAX, &sbi->s_ext4_flags)) {
4400 ext4_msg(sb, KERN_ERR,
4401 "DAX unsupported by block device.");
4402 goto failed_mount;
4403 }
4404 }
4405
4406 if (ext4_has_feature_encrypt(sb) && es->s_encryption_level) {
4407 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
4408 es->s_encryption_level);
4409 goto failed_mount;
4410 }
4411
4412 if (sb->s_blocksize != blocksize) {
4413 /* Validate the filesystem blocksize */
4414 if (!sb_set_blocksize(sb, blocksize)) {
4415 ext4_msg(sb, KERN_ERR, "bad block size %d",
4416 blocksize);
4417 goto failed_mount;
4418 }
4419
4420 brelse(bh);
4421 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
4422 offset = do_div(logical_sb_block, blocksize);
4423 bh = ext4_sb_bread_unmovable(sb, logical_sb_block);
4424 if (IS_ERR(bh)) {
4425 ext4_msg(sb, KERN_ERR,
4426 "Can't read superblock on 2nd try");
4427 ret = PTR_ERR(bh);
4428 bh = NULL;
4429 goto failed_mount;
4430 }
4431 es = (struct ext4_super_block *)(bh->b_data + offset);
4432 sbi->s_es = es;
4433 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
4434 ext4_msg(sb, KERN_ERR,
4435 "Magic mismatch, very weird!");
4436 goto failed_mount;
4437 }
4438 }
4439
4440 has_huge_files = ext4_has_feature_huge_file(sb);
4441 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
4442 has_huge_files);
4443 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
4444
4445 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
4446 if (ext4_has_feature_64bit(sb)) {
4447 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
4448 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
4449 !is_power_of_2(sbi->s_desc_size)) {
4450 ext4_msg(sb, KERN_ERR,
4451 "unsupported descriptor size %lu",
4452 sbi->s_desc_size);
4453 goto failed_mount;
4454 }
4455 } else
4456 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
4457
4458 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
4459 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
4460
4461 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
4462 if (sbi->s_inodes_per_block == 0)
4463 goto cantfind_ext4;
4464 if (sbi->s_inodes_per_group < sbi->s_inodes_per_block ||
4465 sbi->s_inodes_per_group > blocksize * 8) {
4466 ext4_msg(sb, KERN_ERR, "invalid inodes per group: %lu\n",
4467 sbi->s_inodes_per_group);
4468 goto failed_mount;
4469 }
4470 sbi->s_itb_per_group = sbi->s_inodes_per_group /
4471 sbi->s_inodes_per_block;
4472 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
4473 sbi->s_sbh = bh;
4474 sbi->s_mount_state = le16_to_cpu(es->s_state);
4475 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
4476 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
4477
4478 for (i = 0; i < 4; i++)
4479 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
4480 sbi->s_def_hash_version = es->s_def_hash_version;
4481 if (ext4_has_feature_dir_index(sb)) {
4482 i = le32_to_cpu(es->s_flags);
4483 if (i & EXT2_FLAGS_UNSIGNED_HASH)
4484 sbi->s_hash_unsigned = 3;
4485 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
4486 #ifdef __CHAR_UNSIGNED__
4487 if (!sb_rdonly(sb))
4488 es->s_flags |=
4489 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
4490 sbi->s_hash_unsigned = 3;
4491 #else
4492 if (!sb_rdonly(sb))
4493 es->s_flags |=
4494 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
4495 #endif
4496 }
4497 }
4498
4499 /* Handle clustersize */
4500 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
4501 if (ext4_has_feature_bigalloc(sb)) {
4502 if (clustersize < blocksize) {
4503 ext4_msg(sb, KERN_ERR,
4504 "cluster size (%d) smaller than "
4505 "block size (%d)", clustersize, blocksize);
4506 goto failed_mount;
4507 }
4508 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
4509 le32_to_cpu(es->s_log_block_size);
4510 sbi->s_clusters_per_group =
4511 le32_to_cpu(es->s_clusters_per_group);
4512 if (sbi->s_clusters_per_group > blocksize * 8) {
4513 ext4_msg(sb, KERN_ERR,
4514 "#clusters per group too big: %lu",
4515 sbi->s_clusters_per_group);
4516 goto failed_mount;
4517 }
4518 if (sbi->s_blocks_per_group !=
4519 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
4520 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
4521 "clusters per group (%lu) inconsistent",
4522 sbi->s_blocks_per_group,
4523 sbi->s_clusters_per_group);
4524 goto failed_mount;
4525 }
4526 } else {
4527 if (clustersize != blocksize) {
4528 ext4_msg(sb, KERN_ERR,
4529 "fragment/cluster size (%d) != "
4530 "block size (%d)", clustersize, blocksize);
4531 goto failed_mount;
4532 }
4533 if (sbi->s_blocks_per_group > blocksize * 8) {
4534 ext4_msg(sb, KERN_ERR,
4535 "#blocks per group too big: %lu",
4536 sbi->s_blocks_per_group);
4537 goto failed_mount;
4538 }
4539 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
4540 sbi->s_cluster_bits = 0;
4541 }
4542 sbi->s_cluster_ratio = clustersize / blocksize;
4543
4544 /* Do we have standard group size of clustersize * 8 blocks ? */
4545 if (sbi->s_blocks_per_group == clustersize << 3)
4546 set_opt2(sb, STD_GROUP_SIZE);
4547
4548 /*
4549 * Test whether we have more sectors than will fit in sector_t,
4550 * and whether the max offset is addressable by the page cache.
4551 */
4552 err = generic_check_addressable(sb->s_blocksize_bits,
4553 ext4_blocks_count(es));
4554 if (err) {
4555 ext4_msg(sb, KERN_ERR, "filesystem"
4556 " too large to mount safely on this system");
4557 goto failed_mount;
4558 }
4559
4560 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
4561 goto cantfind_ext4;
4562
4563 /* check blocks count against device size */
4564 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
4565 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
4566 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
4567 "exceeds size of device (%llu blocks)",
4568 ext4_blocks_count(es), blocks_count);
4569 goto failed_mount;
4570 }
4571
4572 /*
4573 * It makes no sense for the first data block to be beyond the end
4574 * of the filesystem.
4575 */
4576 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
4577 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
4578 "block %u is beyond end of filesystem (%llu)",
4579 le32_to_cpu(es->s_first_data_block),
4580 ext4_blocks_count(es));
4581 goto failed_mount;
4582 }
4583 if ((es->s_first_data_block == 0) && (es->s_log_block_size == 0) &&
4584 (sbi->s_cluster_ratio == 1)) {
4585 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
4586 "block is 0 with a 1k block and cluster size");
4587 goto failed_mount;
4588 }
4589
4590 blocks_count = (ext4_blocks_count(es) -
4591 le32_to_cpu(es->s_first_data_block) +
4592 EXT4_BLOCKS_PER_GROUP(sb) - 1);
4593 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
4594 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
4595 ext4_msg(sb, KERN_WARNING, "groups count too large: %llu "
4596 "(block count %llu, first data block %u, "
4597 "blocks per group %lu)", blocks_count,
4598 ext4_blocks_count(es),
4599 le32_to_cpu(es->s_first_data_block),
4600 EXT4_BLOCKS_PER_GROUP(sb));
4601 goto failed_mount;
4602 }
4603 sbi->s_groups_count = blocks_count;
4604 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
4605 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
4606 if (((u64)sbi->s_groups_count * sbi->s_inodes_per_group) !=
4607 le32_to_cpu(es->s_inodes_count)) {
4608 ext4_msg(sb, KERN_ERR, "inodes count not valid: %u vs %llu",
4609 le32_to_cpu(es->s_inodes_count),
4610 ((u64)sbi->s_groups_count * sbi->s_inodes_per_group));
4611 ret = -EINVAL;
4612 goto failed_mount;
4613 }
4614 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
4615 EXT4_DESC_PER_BLOCK(sb);
4616 if (ext4_has_feature_meta_bg(sb)) {
4617 if (le32_to_cpu(es->s_first_meta_bg) > db_count) {
4618 ext4_msg(sb, KERN_WARNING,
4619 "first meta block group too large: %u "
4620 "(group descriptor block count %u)",
4621 le32_to_cpu(es->s_first_meta_bg), db_count);
4622 goto failed_mount;
4623 }
4624 }
4625 rcu_assign_pointer(sbi->s_group_desc,
4626 kvmalloc_array(db_count,
4627 sizeof(struct buffer_head *),
4628 GFP_KERNEL));
4629 if (sbi->s_group_desc == NULL) {
4630 ext4_msg(sb, KERN_ERR, "not enough memory");
4631 ret = -ENOMEM;
4632 goto failed_mount;
4633 }
4634
4635 bgl_lock_init(sbi->s_blockgroup_lock);
4636
4637 /* Pre-read the descriptors into the buffer cache */
4638 for (i = 0; i < db_count; i++) {
4639 block = descriptor_loc(sb, logical_sb_block, i);
4640 ext4_sb_breadahead_unmovable(sb, block);
4641 }
4642
4643 for (i = 0; i < db_count; i++) {
4644 struct buffer_head *bh;
4645
4646 block = descriptor_loc(sb, logical_sb_block, i);
4647 bh = ext4_sb_bread_unmovable(sb, block);
4648 if (IS_ERR(bh)) {
4649 ext4_msg(sb, KERN_ERR,
4650 "can't read group descriptor %d", i);
4651 db_count = i;
4652 ret = PTR_ERR(bh);
4653 goto failed_mount2;
4654 }
4655 rcu_read_lock();
4656 rcu_dereference(sbi->s_group_desc)[i] = bh;
4657 rcu_read_unlock();
4658 }
4659 sbi->s_gdb_count = db_count;
4660 if (!ext4_check_descriptors(sb, logical_sb_block, &first_not_zeroed)) {
4661 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
4662 ret = -EFSCORRUPTED;
4663 goto failed_mount2;
4664 }
4665
4666 timer_setup(&sbi->s_err_report, print_daily_error_info, 0);
4667 spin_lock_init(&sbi->s_error_lock);
4668 INIT_WORK(&sbi->s_error_work, flush_stashed_error_work);
4669
4670 /* Register extent status tree shrinker */
4671 if (ext4_es_register_shrinker(sbi))
4672 goto failed_mount3;
4673
4674 sbi->s_stripe = ext4_get_stripe_size(sbi);
4675 sbi->s_extent_max_zeroout_kb = 32;
4676
4677 /*
4678 * set up enough so that it can read an inode
4679 */
4680 sb->s_op = &ext4_sops;
4681 sb->s_export_op = &ext4_export_ops;
4682 sb->s_xattr = ext4_xattr_handlers;
4683 #ifdef CONFIG_FS_ENCRYPTION
4684 sb->s_cop = &ext4_cryptops;
4685 #endif
4686 #ifdef CONFIG_FS_VERITY
4687 sb->s_vop = &ext4_verityops;
4688 #endif
4689 #ifdef CONFIG_QUOTA
4690 sb->dq_op = &ext4_quota_operations;
4691 if (ext4_has_feature_quota(sb))
4692 sb->s_qcop = &dquot_quotactl_sysfile_ops;
4693 else
4694 sb->s_qcop = &ext4_qctl_operations;
4695 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP | QTYPE_MASK_PRJ;
4696 #endif
4697 memcpy(&sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
4698
4699 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
4700 mutex_init(&sbi->s_orphan_lock);
4701
4702 /* Initialize fast commit stuff */
4703 atomic_set(&sbi->s_fc_subtid, 0);
4704 atomic_set(&sbi->s_fc_ineligible_updates, 0);
4705 INIT_LIST_HEAD(&sbi->s_fc_q[FC_Q_MAIN]);
4706 INIT_LIST_HEAD(&sbi->s_fc_q[FC_Q_STAGING]);
4707 INIT_LIST_HEAD(&sbi->s_fc_dentry_q[FC_Q_MAIN]);
4708 INIT_LIST_HEAD(&sbi->s_fc_dentry_q[FC_Q_STAGING]);
4709 sbi->s_fc_bytes = 0;
4710 ext4_clear_mount_flag(sb, EXT4_MF_FC_INELIGIBLE);
4711 ext4_clear_mount_flag(sb, EXT4_MF_FC_COMMITTING);
4712 spin_lock_init(&sbi->s_fc_lock);
4713 memset(&sbi->s_fc_stats, 0, sizeof(sbi->s_fc_stats));
4714 sbi->s_fc_replay_state.fc_regions = NULL;
4715 sbi->s_fc_replay_state.fc_regions_size = 0;
4716 sbi->s_fc_replay_state.fc_regions_used = 0;
4717 sbi->s_fc_replay_state.fc_regions_valid = 0;
4718 sbi->s_fc_replay_state.fc_modified_inodes = NULL;
4719 sbi->s_fc_replay_state.fc_modified_inodes_size = 0;
4720 sbi->s_fc_replay_state.fc_modified_inodes_used = 0;
4721
4722 sb->s_root = NULL;
4723
4724 needs_recovery = (es->s_last_orphan != 0 ||
4725 ext4_has_feature_journal_needs_recovery(sb));
4726
4727 if (ext4_has_feature_mmp(sb) && !sb_rdonly(sb))
4728 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
4729 goto failed_mount3a;
4730
4731 /*
4732 * The first inode we look at is the journal inode. Don't try
4733 * root first: it may be modified in the journal!
4734 */
4735 if (!test_opt(sb, NOLOAD) && ext4_has_feature_journal(sb)) {
4736 err = ext4_load_journal(sb, es, journal_devnum);
4737 if (err)
4738 goto failed_mount3a;
4739 } else if (test_opt(sb, NOLOAD) && !sb_rdonly(sb) &&
4740 ext4_has_feature_journal_needs_recovery(sb)) {
4741 ext4_msg(sb, KERN_ERR, "required journal recovery "
4742 "suppressed and not mounted read-only");
4743 goto failed_mount_wq;
4744 } else {
4745 /* Nojournal mode, all journal mount options are illegal */
4746 if (test_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM)) {
4747 ext4_msg(sb, KERN_ERR, "can't mount with "
4748 "journal_checksum, fs mounted w/o journal");
4749 goto failed_mount_wq;
4750 }
4751 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4752 ext4_msg(sb, KERN_ERR, "can't mount with "
4753 "journal_async_commit, fs mounted w/o journal");
4754 goto failed_mount_wq;
4755 }
4756 if (sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ) {
4757 ext4_msg(sb, KERN_ERR, "can't mount with "
4758 "commit=%lu, fs mounted w/o journal",
4759 sbi->s_commit_interval / HZ);
4760 goto failed_mount_wq;
4761 }
4762 if (EXT4_MOUNT_DATA_FLAGS &
4763 (sbi->s_mount_opt ^ sbi->s_def_mount_opt)) {
4764 ext4_msg(sb, KERN_ERR, "can't mount with "
4765 "data=, fs mounted w/o journal");
4766 goto failed_mount_wq;
4767 }
4768 sbi->s_def_mount_opt &= ~EXT4_MOUNT_JOURNAL_CHECKSUM;
4769 clear_opt(sb, JOURNAL_CHECKSUM);
4770 clear_opt(sb, DATA_FLAGS);
4771 clear_opt2(sb, JOURNAL_FAST_COMMIT);
4772 sbi->s_journal = NULL;
4773 needs_recovery = 0;
4774 goto no_journal;
4775 }
4776
4777 if (ext4_has_feature_64bit(sb) &&
4778 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4779 JBD2_FEATURE_INCOMPAT_64BIT)) {
4780 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
4781 goto failed_mount_wq;
4782 }
4783
4784 if (!set_journal_csum_feature_set(sb)) {
4785 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
4786 "feature set");
4787 goto failed_mount_wq;
4788 }
4789
4790 if (test_opt2(sb, JOURNAL_FAST_COMMIT) &&
4791 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4792 JBD2_FEATURE_INCOMPAT_FAST_COMMIT)) {
4793 ext4_msg(sb, KERN_ERR,
4794 "Failed to set fast commit journal feature");
4795 goto failed_mount_wq;
4796 }
4797
4798 /* We have now updated the journal if required, so we can
4799 * validate the data journaling mode. */
4800 switch (test_opt(sb, DATA_FLAGS)) {
4801 case 0:
4802 /* No mode set, assume a default based on the journal
4803 * capabilities: ORDERED_DATA if the journal can
4804 * cope, else JOURNAL_DATA
4805 */
4806 if (jbd2_journal_check_available_features
4807 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4808 set_opt(sb, ORDERED_DATA);
4809 sbi->s_def_mount_opt |= EXT4_MOUNT_ORDERED_DATA;
4810 } else {
4811 set_opt(sb, JOURNAL_DATA);
4812 sbi->s_def_mount_opt |= EXT4_MOUNT_JOURNAL_DATA;
4813 }
4814 break;
4815
4816 case EXT4_MOUNT_ORDERED_DATA:
4817 case EXT4_MOUNT_WRITEBACK_DATA:
4818 if (!jbd2_journal_check_available_features
4819 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4820 ext4_msg(sb, KERN_ERR, "Journal does not support "
4821 "requested data journaling mode");
4822 goto failed_mount_wq;
4823 }
4824 break;
4825 default:
4826 break;
4827 }
4828
4829 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
4830 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
4831 ext4_msg(sb, KERN_ERR, "can't mount with "
4832 "journal_async_commit in data=ordered mode");
4833 goto failed_mount_wq;
4834 }
4835
4836 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4837
4838 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
4839 sbi->s_journal->j_submit_inode_data_buffers =
4840 ext4_journal_submit_inode_data_buffers;
4841 sbi->s_journal->j_finish_inode_data_buffers =
4842 ext4_journal_finish_inode_data_buffers;
4843
4844 no_journal:
4845 if (!test_opt(sb, NO_MBCACHE)) {
4846 sbi->s_ea_block_cache = ext4_xattr_create_cache();
4847 if (!sbi->s_ea_block_cache) {
4848 ext4_msg(sb, KERN_ERR,
4849 "Failed to create ea_block_cache");
4850 goto failed_mount_wq;
4851 }
4852
4853 if (ext4_has_feature_ea_inode(sb)) {
4854 sbi->s_ea_inode_cache = ext4_xattr_create_cache();
4855 if (!sbi->s_ea_inode_cache) {
4856 ext4_msg(sb, KERN_ERR,
4857 "Failed to create ea_inode_cache");
4858 goto failed_mount_wq;
4859 }
4860 }
4861 }
4862
4863 if (ext4_has_feature_verity(sb) && blocksize != PAGE_SIZE) {
4864 ext4_msg(sb, KERN_ERR, "Unsupported blocksize for fs-verity");
4865 goto failed_mount_wq;
4866 }
4867
4868 if (DUMMY_ENCRYPTION_ENABLED(sbi) && !sb_rdonly(sb) &&
4869 !ext4_has_feature_encrypt(sb)) {
4870 ext4_set_feature_encrypt(sb);
4871 ext4_commit_super(sb, 1);
4872 }
4873
4874 /*
4875 * Get the # of file system overhead blocks from the
4876 * superblock if present.
4877 */
4878 if (es->s_overhead_clusters)
4879 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4880 else {
4881 err = ext4_calculate_overhead(sb);
4882 if (err)
4883 goto failed_mount_wq;
4884 }
4885
4886 /*
4887 * The maximum number of concurrent works can be high and
4888 * concurrency isn't really necessary. Limit it to 1.
4889 */
4890 EXT4_SB(sb)->rsv_conversion_wq =
4891 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4892 if (!EXT4_SB(sb)->rsv_conversion_wq) {
4893 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
4894 ret = -ENOMEM;
4895 goto failed_mount4;
4896 }
4897
4898 /*
4899 * The jbd2_journal_load will have done any necessary log recovery,
4900 * so we can safely mount the rest of the filesystem now.
4901 */
4902
4903 root = ext4_iget(sb, EXT4_ROOT_INO, EXT4_IGET_SPECIAL);
4904 if (IS_ERR(root)) {
4905 ext4_msg(sb, KERN_ERR, "get root inode failed");
4906 ret = PTR_ERR(root);
4907 root = NULL;
4908 goto failed_mount4;
4909 }
4910 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
4911 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
4912 iput(root);
4913 goto failed_mount4;
4914 }
4915
4916 sb->s_root = d_make_root(root);
4917 if (!sb->s_root) {
4918 ext4_msg(sb, KERN_ERR, "get root dentry failed");
4919 ret = -ENOMEM;
4920 goto failed_mount4;
4921 }
4922
4923 ret = ext4_setup_super(sb, es, sb_rdonly(sb));
4924 if (ret == -EROFS) {
4925 sb->s_flags |= SB_RDONLY;
4926 ret = 0;
4927 } else if (ret)
4928 goto failed_mount4a;
4929
4930 ext4_set_resv_clusters(sb);
4931
4932 if (test_opt(sb, BLOCK_VALIDITY)) {
4933 err = ext4_setup_system_zone(sb);
4934 if (err) {
4935 ext4_msg(sb, KERN_ERR, "failed to initialize system "
4936 "zone (%d)", err);
4937 goto failed_mount4a;
4938 }
4939 }
4940 ext4_fc_replay_cleanup(sb);
4941
4942 ext4_ext_init(sb);
4943 err = ext4_mb_init(sb);
4944 if (err) {
4945 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4946 err);
4947 goto failed_mount5;
4948 }
4949
4950 block = ext4_count_free_clusters(sb);
4951 ext4_free_blocks_count_set(sbi->s_es,
4952 EXT4_C2B(sbi, block));
4953 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
4954 GFP_KERNEL);
4955 if (!err) {
4956 unsigned long freei = ext4_count_free_inodes(sb);
4957 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
4958 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
4959 GFP_KERNEL);
4960 }
4961 if (!err)
4962 err = percpu_counter_init(&sbi->s_dirs_counter,
4963 ext4_count_dirs(sb), GFP_KERNEL);
4964 if (!err)
4965 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
4966 GFP_KERNEL);
4967 if (!err)
4968 err = percpu_init_rwsem(&sbi->s_writepages_rwsem);
4969
4970 if (err) {
4971 ext4_msg(sb, KERN_ERR, "insufficient memory");
4972 goto failed_mount6;
4973 }
4974
4975 if (ext4_has_feature_flex_bg(sb))
4976 if (!ext4_fill_flex_info(sb)) {
4977 ext4_msg(sb, KERN_ERR,
4978 "unable to initialize "
4979 "flex_bg meta info!");
4980 goto failed_mount6;
4981 }
4982
4983 err = ext4_register_li_request(sb, first_not_zeroed);
4984 if (err)
4985 goto failed_mount6;
4986
4987 err = ext4_register_sysfs(sb);
4988 if (err)
4989 goto failed_mount7;
4990
4991 #ifdef CONFIG_QUOTA
4992 /* Enable quota usage during mount. */
4993 if (ext4_has_feature_quota(sb) && !sb_rdonly(sb)) {
4994 err = ext4_enable_quotas(sb);
4995 if (err)
4996 goto failed_mount8;
4997 }
4998 #endif /* CONFIG_QUOTA */
4999
5000 /*
5001 * Save the original bdev mapping's wb_err value which could be
5002 * used to detect the metadata async write error.
5003 */
5004 spin_lock_init(&sbi->s_bdev_wb_lock);
5005 errseq_check_and_advance(&sb->s_bdev->bd_inode->i_mapping->wb_err,
5006 &sbi->s_bdev_wb_err);
5007 sb->s_bdev->bd_super = sb;
5008 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
5009 ext4_orphan_cleanup(sb, es);
5010 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
5011 if (needs_recovery) {
5012 ext4_msg(sb, KERN_INFO, "recovery complete");
5013 err = ext4_mark_recovery_complete(sb, es);
5014 if (err)
5015 goto failed_mount8;
5016 }
5017 if (EXT4_SB(sb)->s_journal) {
5018 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
5019 descr = " journalled data mode";
5020 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
5021 descr = " ordered data mode";
5022 else
5023 descr = " writeback data mode";
5024 } else
5025 descr = "out journal";
5026
5027 if (test_opt(sb, DISCARD)) {
5028 struct request_queue *q = bdev_get_queue(sb->s_bdev);
5029 if (!blk_queue_discard(q))
5030 ext4_msg(sb, KERN_WARNING,
5031 "mounting with \"discard\" option, but "
5032 "the device does not support discard");
5033 }
5034
5035 if (___ratelimit(&ext4_mount_msg_ratelimit, "EXT4-fs mount"))
5036 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
5037 "Opts: %.*s%s%s. Quota mode: %s.", descr,
5038 (int) sizeof(sbi->s_es->s_mount_opts),
5039 sbi->s_es->s_mount_opts,
5040 *sbi->s_es->s_mount_opts ? "; " : "", orig_data,
5041 ext4_quota_mode(sb));
5042
5043 if (es->s_error_count)
5044 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
5045
5046 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
5047 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
5048 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
5049 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
5050 atomic_set(&sbi->s_warning_count, 0);
5051 atomic_set(&sbi->s_msg_count, 0);
5052
5053 kfree(orig_data);
5054 return 0;
5055
5056 cantfind_ext4:
5057 if (!silent)
5058 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
5059 goto failed_mount;
5060
5061 failed_mount8:
5062 ext4_unregister_sysfs(sb);
5063 kobject_put(&sbi->s_kobj);
5064 failed_mount7:
5065 ext4_unregister_li_request(sb);
5066 failed_mount6:
5067 ext4_mb_release(sb);
5068 rcu_read_lock();
5069 flex_groups = rcu_dereference(sbi->s_flex_groups);
5070 if (flex_groups) {
5071 for (i = 0; i < sbi->s_flex_groups_allocated; i++)
5072 kvfree(flex_groups[i]);
5073 kvfree(flex_groups);
5074 }
5075 rcu_read_unlock();
5076 percpu_counter_destroy(&sbi->s_freeclusters_counter);
5077 percpu_counter_destroy(&sbi->s_freeinodes_counter);
5078 percpu_counter_destroy(&sbi->s_dirs_counter);
5079 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
5080 percpu_free_rwsem(&sbi->s_writepages_rwsem);
5081 failed_mount5:
5082 ext4_ext_release(sb);
5083 ext4_release_system_zone(sb);
5084 failed_mount4a:
5085 dput(sb->s_root);
5086 sb->s_root = NULL;
5087 failed_mount4:
5088 ext4_msg(sb, KERN_ERR, "mount failed");
5089 if (EXT4_SB(sb)->rsv_conversion_wq)
5090 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
5091 failed_mount_wq:
5092 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
5093 sbi->s_ea_inode_cache = NULL;
5094
5095 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
5096 sbi->s_ea_block_cache = NULL;
5097
5098 if (sbi->s_journal) {
5099 jbd2_journal_destroy(sbi->s_journal);
5100 sbi->s_journal = NULL;
5101 }
5102 failed_mount3a:
5103 ext4_es_unregister_shrinker(sbi);
5104 failed_mount3:
5105 del_timer_sync(&sbi->s_err_report);
5106 flush_work(&sbi->s_error_work);
5107 if (sbi->s_mmp_tsk)
5108 kthread_stop(sbi->s_mmp_tsk);
5109 failed_mount2:
5110 rcu_read_lock();
5111 group_desc = rcu_dereference(sbi->s_group_desc);
5112 for (i = 0; i < db_count; i++)
5113 brelse(group_desc[i]);
5114 kvfree(group_desc);
5115 rcu_read_unlock();
5116 failed_mount:
5117 if (sbi->s_chksum_driver)
5118 crypto_free_shash(sbi->s_chksum_driver);
5119
5120 #ifdef CONFIG_UNICODE
5121 utf8_unload(sb->s_encoding);
5122 #endif
5123
5124 #ifdef CONFIG_QUOTA
5125 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5126 kfree(get_qf_name(sb, sbi, i));
5127 #endif
5128 fscrypt_free_dummy_policy(&sbi->s_dummy_enc_policy);
5129 ext4_blkdev_remove(sbi);
5130 brelse(bh);
5131 out_fail:
5132 sb->s_fs_info = NULL;
5133 kfree(sbi->s_blockgroup_lock);
5134 out_free_base:
5135 kfree(sbi);
5136 kfree(orig_data);
5137 fs_put_dax(dax_dev);
5138 return err ? err : ret;
5139 }
5140
5141 /*
5142 * Setup any per-fs journal parameters now. We'll do this both on
5143 * initial mount, once the journal has been initialised but before we've
5144 * done any recovery; and again on any subsequent remount.
5145 */
5146 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
5147 {
5148 struct ext4_sb_info *sbi = EXT4_SB(sb);
5149
5150 journal->j_commit_interval = sbi->s_commit_interval;
5151 journal->j_min_batch_time = sbi->s_min_batch_time;
5152 journal->j_max_batch_time = sbi->s_max_batch_time;
5153 ext4_fc_init(sb, journal);
5154
5155 write_lock(&journal->j_state_lock);
5156 if (test_opt(sb, BARRIER))
5157 journal->j_flags |= JBD2_BARRIER;
5158 else
5159 journal->j_flags &= ~JBD2_BARRIER;
5160 if (test_opt(sb, DATA_ERR_ABORT))
5161 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
5162 else
5163 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
5164 write_unlock(&journal->j_state_lock);
5165 }
5166
5167 static struct inode *ext4_get_journal_inode(struct super_block *sb,
5168 unsigned int journal_inum)
5169 {
5170 struct inode *journal_inode;
5171
5172 /*
5173 * Test for the existence of a valid inode on disk. Bad things
5174 * happen if we iget() an unused inode, as the subsequent iput()
5175 * will try to delete it.
5176 */
5177 journal_inode = ext4_iget(sb, journal_inum, EXT4_IGET_SPECIAL);
5178 if (IS_ERR(journal_inode)) {
5179 ext4_msg(sb, KERN_ERR, "no journal found");
5180 return NULL;
5181 }
5182 if (!journal_inode->i_nlink) {
5183 make_bad_inode(journal_inode);
5184 iput(journal_inode);
5185 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
5186 return NULL;
5187 }
5188
5189 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
5190 journal_inode, journal_inode->i_size);
5191 if (!S_ISREG(journal_inode->i_mode)) {
5192 ext4_msg(sb, KERN_ERR, "invalid journal inode");
5193 iput(journal_inode);
5194 return NULL;
5195 }
5196 return journal_inode;
5197 }
5198
5199 static journal_t *ext4_get_journal(struct super_block *sb,
5200 unsigned int journal_inum)
5201 {
5202 struct inode *journal_inode;
5203 journal_t *journal;
5204
5205 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5206 return NULL;
5207
5208 journal_inode = ext4_get_journal_inode(sb, journal_inum);
5209 if (!journal_inode)
5210 return NULL;
5211
5212 journal = jbd2_journal_init_inode(journal_inode);
5213 if (!journal) {
5214 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
5215 iput(journal_inode);
5216 return NULL;
5217 }
5218 journal->j_private = sb;
5219 ext4_init_journal_params(sb, journal);
5220 return journal;
5221 }
5222
5223 static journal_t *ext4_get_dev_journal(struct super_block *sb,
5224 dev_t j_dev)
5225 {
5226 struct buffer_head *bh;
5227 journal_t *journal;
5228 ext4_fsblk_t start;
5229 ext4_fsblk_t len;
5230 int hblock, blocksize;
5231 ext4_fsblk_t sb_block;
5232 unsigned long offset;
5233 struct ext4_super_block *es;
5234 struct block_device *bdev;
5235
5236 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5237 return NULL;
5238
5239 bdev = ext4_blkdev_get(j_dev, sb);
5240 if (bdev == NULL)
5241 return NULL;
5242
5243 blocksize = sb->s_blocksize;
5244 hblock = bdev_logical_block_size(bdev);
5245 if (blocksize < hblock) {
5246 ext4_msg(sb, KERN_ERR,
5247 "blocksize too small for journal device");
5248 goto out_bdev;
5249 }
5250
5251 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
5252 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
5253 set_blocksize(bdev, blocksize);
5254 if (!(bh = __bread(bdev, sb_block, blocksize))) {
5255 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
5256 "external journal");
5257 goto out_bdev;
5258 }
5259
5260 es = (struct ext4_super_block *) (bh->b_data + offset);
5261 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
5262 !(le32_to_cpu(es->s_feature_incompat) &
5263 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
5264 ext4_msg(sb, KERN_ERR, "external journal has "
5265 "bad superblock");
5266 brelse(bh);
5267 goto out_bdev;
5268 }
5269
5270 if ((le32_to_cpu(es->s_feature_ro_compat) &
5271 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
5272 es->s_checksum != ext4_superblock_csum(sb, es)) {
5273 ext4_msg(sb, KERN_ERR, "external journal has "
5274 "corrupt superblock");
5275 brelse(bh);
5276 goto out_bdev;
5277 }
5278
5279 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
5280 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
5281 brelse(bh);
5282 goto out_bdev;
5283 }
5284
5285 len = ext4_blocks_count(es);
5286 start = sb_block + 1;
5287 brelse(bh); /* we're done with the superblock */
5288
5289 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
5290 start, len, blocksize);
5291 if (!journal) {
5292 ext4_msg(sb, KERN_ERR, "failed to create device journal");
5293 goto out_bdev;
5294 }
5295 journal->j_private = sb;
5296 if (ext4_read_bh_lock(journal->j_sb_buffer, REQ_META | REQ_PRIO, true)) {
5297 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
5298 goto out_journal;
5299 }
5300 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
5301 ext4_msg(sb, KERN_ERR, "External journal has more than one "
5302 "user (unsupported) - %d",
5303 be32_to_cpu(journal->j_superblock->s_nr_users));
5304 goto out_journal;
5305 }
5306 EXT4_SB(sb)->s_journal_bdev = bdev;
5307 ext4_init_journal_params(sb, journal);
5308 return journal;
5309
5310 out_journal:
5311 jbd2_journal_destroy(journal);
5312 out_bdev:
5313 ext4_blkdev_put(bdev);
5314 return NULL;
5315 }
5316
5317 static int ext4_load_journal(struct super_block *sb,
5318 struct ext4_super_block *es,
5319 unsigned long journal_devnum)
5320 {
5321 journal_t *journal;
5322 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
5323 dev_t journal_dev;
5324 int err = 0;
5325 int really_read_only;
5326 int journal_dev_ro;
5327
5328 if (WARN_ON_ONCE(!ext4_has_feature_journal(sb)))
5329 return -EFSCORRUPTED;
5330
5331 if (journal_devnum &&
5332 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
5333 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
5334 "numbers have changed");
5335 journal_dev = new_decode_dev(journal_devnum);
5336 } else
5337 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
5338
5339 if (journal_inum && journal_dev) {
5340 ext4_msg(sb, KERN_ERR,
5341 "filesystem has both journal inode and journal device!");
5342 return -EINVAL;
5343 }
5344
5345 if (journal_inum) {
5346 journal = ext4_get_journal(sb, journal_inum);
5347 if (!journal)
5348 return -EINVAL;
5349 } else {
5350 journal = ext4_get_dev_journal(sb, journal_dev);
5351 if (!journal)
5352 return -EINVAL;
5353 }
5354
5355 journal_dev_ro = bdev_read_only(journal->j_dev);
5356 really_read_only = bdev_read_only(sb->s_bdev) | journal_dev_ro;
5357
5358 if (journal_dev_ro && !sb_rdonly(sb)) {
5359 ext4_msg(sb, KERN_ERR,
5360 "journal device read-only, try mounting with '-o ro'");
5361 err = -EROFS;
5362 goto err_out;
5363 }
5364
5365 /*
5366 * Are we loading a blank journal or performing recovery after a
5367 * crash? For recovery, we need to check in advance whether we
5368 * can get read-write access to the device.
5369 */
5370 if (ext4_has_feature_journal_needs_recovery(sb)) {
5371 if (sb_rdonly(sb)) {
5372 ext4_msg(sb, KERN_INFO, "INFO: recovery "
5373 "required on readonly filesystem");
5374 if (really_read_only) {
5375 ext4_msg(sb, KERN_ERR, "write access "
5376 "unavailable, cannot proceed "
5377 "(try mounting with noload)");
5378 err = -EROFS;
5379 goto err_out;
5380 }
5381 ext4_msg(sb, KERN_INFO, "write access will "
5382 "be enabled during recovery");
5383 }
5384 }
5385
5386 if (!(journal->j_flags & JBD2_BARRIER))
5387 ext4_msg(sb, KERN_INFO, "barriers disabled");
5388
5389 if (!ext4_has_feature_journal_needs_recovery(sb))
5390 err = jbd2_journal_wipe(journal, !really_read_only);
5391 if (!err) {
5392 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
5393 if (save)
5394 memcpy(save, ((char *) es) +
5395 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
5396 err = jbd2_journal_load(journal);
5397 if (save)
5398 memcpy(((char *) es) + EXT4_S_ERR_START,
5399 save, EXT4_S_ERR_LEN);
5400 kfree(save);
5401 }
5402
5403 if (err) {
5404 ext4_msg(sb, KERN_ERR, "error loading journal");
5405 goto err_out;
5406 }
5407
5408 EXT4_SB(sb)->s_journal = journal;
5409 err = ext4_clear_journal_err(sb, es);
5410 if (err) {
5411 EXT4_SB(sb)->s_journal = NULL;
5412 jbd2_journal_destroy(journal);
5413 return err;
5414 }
5415
5416 if (!really_read_only && journal_devnum &&
5417 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
5418 es->s_journal_dev = cpu_to_le32(journal_devnum);
5419
5420 /* Make sure we flush the recovery flag to disk. */
5421 ext4_commit_super(sb, 1);
5422 }
5423
5424 return 0;
5425
5426 err_out:
5427 jbd2_journal_destroy(journal);
5428 return err;
5429 }
5430
5431 static int ext4_commit_super(struct super_block *sb, int sync)
5432 {
5433 struct ext4_sb_info *sbi = EXT4_SB(sb);
5434 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
5435 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
5436 int error = 0;
5437
5438 if (!sbh || block_device_ejected(sb))
5439 return error;
5440
5441 /*
5442 * If the file system is mounted read-only, don't update the
5443 * superblock write time. This avoids updating the superblock
5444 * write time when we are mounting the root file system
5445 * read/only but we need to replay the journal; at that point,
5446 * for people who are east of GMT and who make their clock
5447 * tick in localtime for Windows bug-for-bug compatibility,
5448 * the clock is set in the future, and this will cause e2fsck
5449 * to complain and force a full file system check.
5450 */
5451 if (!(sb->s_flags & SB_RDONLY))
5452 ext4_update_tstamp(es, s_wtime);
5453 es->s_kbytes_written =
5454 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
5455 ((part_stat_read(sb->s_bdev, sectors[STAT_WRITE]) -
5456 EXT4_SB(sb)->s_sectors_written_start) >> 1));
5457 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
5458 ext4_free_blocks_count_set(es,
5459 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
5460 &EXT4_SB(sb)->s_freeclusters_counter)));
5461 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
5462 es->s_free_inodes_count =
5463 cpu_to_le32(percpu_counter_sum_positive(
5464 &EXT4_SB(sb)->s_freeinodes_counter));
5465 /* Copy error information to the on-disk superblock */
5466 spin_lock(&sbi->s_error_lock);
5467 if (sbi->s_add_error_count > 0) {
5468 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
5469 if (!es->s_first_error_time && !es->s_first_error_time_hi) {
5470 __ext4_update_tstamp(&es->s_first_error_time,
5471 &es->s_first_error_time_hi,
5472 sbi->s_first_error_time);
5473 strncpy(es->s_first_error_func, sbi->s_first_error_func,
5474 sizeof(es->s_first_error_func));
5475 es->s_first_error_line =
5476 cpu_to_le32(sbi->s_first_error_line);
5477 es->s_first_error_ino =
5478 cpu_to_le32(sbi->s_first_error_ino);
5479 es->s_first_error_block =
5480 cpu_to_le64(sbi->s_first_error_block);
5481 es->s_first_error_errcode =
5482 ext4_errno_to_code(sbi->s_first_error_code);
5483 }
5484 __ext4_update_tstamp(&es->s_last_error_time,
5485 &es->s_last_error_time_hi,
5486 sbi->s_last_error_time);
5487 strncpy(es->s_last_error_func, sbi->s_last_error_func,
5488 sizeof(es->s_last_error_func));
5489 es->s_last_error_line = cpu_to_le32(sbi->s_last_error_line);
5490 es->s_last_error_ino = cpu_to_le32(sbi->s_last_error_ino);
5491 es->s_last_error_block = cpu_to_le64(sbi->s_last_error_block);
5492 es->s_last_error_errcode =
5493 ext4_errno_to_code(sbi->s_last_error_code);
5494 /*
5495 * Start the daily error reporting function if it hasn't been
5496 * started already
5497 */
5498 if (!es->s_error_count)
5499 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ);
5500 le32_add_cpu(&es->s_error_count, sbi->s_add_error_count);
5501 sbi->s_add_error_count = 0;
5502 }
5503 spin_unlock(&sbi->s_error_lock);
5504
5505 BUFFER_TRACE(sbh, "marking dirty");
5506 ext4_superblock_csum_set(sb);
5507 if (sync)
5508 lock_buffer(sbh);
5509 if (buffer_write_io_error(sbh) || !buffer_uptodate(sbh)) {
5510 /*
5511 * Oh, dear. A previous attempt to write the
5512 * superblock failed. This could happen because the
5513 * USB device was yanked out. Or it could happen to
5514 * be a transient write error and maybe the block will
5515 * be remapped. Nothing we can do but to retry the
5516 * write and hope for the best.
5517 */
5518 ext4_msg(sb, KERN_ERR, "previous I/O error to "
5519 "superblock detected");
5520 clear_buffer_write_io_error(sbh);
5521 set_buffer_uptodate(sbh);
5522 }
5523 mark_buffer_dirty(sbh);
5524 if (sync) {
5525 unlock_buffer(sbh);
5526 error = __sync_dirty_buffer(sbh,
5527 REQ_SYNC | (test_opt(sb, BARRIER) ? REQ_FUA : 0));
5528 if (buffer_write_io_error(sbh)) {
5529 ext4_msg(sb, KERN_ERR, "I/O error while writing "
5530 "superblock");
5531 clear_buffer_write_io_error(sbh);
5532 set_buffer_uptodate(sbh);
5533 }
5534 }
5535 return error;
5536 }
5537
5538 /*
5539 * Have we just finished recovery? If so, and if we are mounting (or
5540 * remounting) the filesystem readonly, then we will end up with a
5541 * consistent fs on disk. Record that fact.
5542 */
5543 static int ext4_mark_recovery_complete(struct super_block *sb,
5544 struct ext4_super_block *es)
5545 {
5546 int err;
5547 journal_t *journal = EXT4_SB(sb)->s_journal;
5548
5549 if (!ext4_has_feature_journal(sb)) {
5550 if (journal != NULL) {
5551 ext4_error(sb, "Journal got removed while the fs was "
5552 "mounted!");
5553 return -EFSCORRUPTED;
5554 }
5555 return 0;
5556 }
5557 jbd2_journal_lock_updates(journal);
5558 err = jbd2_journal_flush(journal);
5559 if (err < 0)
5560 goto out;
5561
5562 if (ext4_has_feature_journal_needs_recovery(sb) && sb_rdonly(sb)) {
5563 ext4_clear_feature_journal_needs_recovery(sb);
5564 ext4_commit_super(sb, 1);
5565 }
5566 out:
5567 jbd2_journal_unlock_updates(journal);
5568 return err;
5569 }
5570
5571 /*
5572 * If we are mounting (or read-write remounting) a filesystem whose journal
5573 * has recorded an error from a previous lifetime, move that error to the
5574 * main filesystem now.
5575 */
5576 static int ext4_clear_journal_err(struct super_block *sb,
5577 struct ext4_super_block *es)
5578 {
5579 journal_t *journal;
5580 int j_errno;
5581 const char *errstr;
5582
5583 if (!ext4_has_feature_journal(sb)) {
5584 ext4_error(sb, "Journal got removed while the fs was mounted!");
5585 return -EFSCORRUPTED;
5586 }
5587
5588 journal = EXT4_SB(sb)->s_journal;
5589
5590 /*
5591 * Now check for any error status which may have been recorded in the
5592 * journal by a prior ext4_error() or ext4_abort()
5593 */
5594
5595 j_errno = jbd2_journal_errno(journal);
5596 if (j_errno) {
5597 char nbuf[16];
5598
5599 errstr = ext4_decode_error(sb, j_errno, nbuf);
5600 ext4_warning(sb, "Filesystem error recorded "
5601 "from previous mount: %s", errstr);
5602 ext4_warning(sb, "Marking fs in need of filesystem check.");
5603
5604 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
5605 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
5606 ext4_commit_super(sb, 1);
5607
5608 jbd2_journal_clear_err(journal);
5609 jbd2_journal_update_sb_errno(journal);
5610 }
5611 return 0;
5612 }
5613
5614 /*
5615 * Force the running and committing transactions to commit,
5616 * and wait on the commit.
5617 */
5618 int ext4_force_commit(struct super_block *sb)
5619 {
5620 journal_t *journal;
5621
5622 if (sb_rdonly(sb))
5623 return 0;
5624
5625 journal = EXT4_SB(sb)->s_journal;
5626 return ext4_journal_force_commit(journal);
5627 }
5628
5629 static int ext4_sync_fs(struct super_block *sb, int wait)
5630 {
5631 int ret = 0;
5632 tid_t target;
5633 bool needs_barrier = false;
5634 struct ext4_sb_info *sbi = EXT4_SB(sb);
5635
5636 if (unlikely(ext4_forced_shutdown(sbi)))
5637 return 0;
5638
5639 trace_ext4_sync_fs(sb, wait);
5640 flush_workqueue(sbi->rsv_conversion_wq);
5641 /*
5642 * Writeback quota in non-journalled quota case - journalled quota has
5643 * no dirty dquots
5644 */
5645 dquot_writeback_dquots(sb, -1);
5646 /*
5647 * Data writeback is possible w/o journal transaction, so barrier must
5648 * being sent at the end of the function. But we can skip it if
5649 * transaction_commit will do it for us.
5650 */
5651 if (sbi->s_journal) {
5652 target = jbd2_get_latest_transaction(sbi->s_journal);
5653 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
5654 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
5655 needs_barrier = true;
5656
5657 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
5658 if (wait)
5659 ret = jbd2_log_wait_commit(sbi->s_journal,
5660 target);
5661 }
5662 } else if (wait && test_opt(sb, BARRIER))
5663 needs_barrier = true;
5664 if (needs_barrier) {
5665 int err;
5666 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL);
5667 if (!ret)
5668 ret = err;
5669 }
5670
5671 return ret;
5672 }
5673
5674 /*
5675 * LVM calls this function before a (read-only) snapshot is created. This
5676 * gives us a chance to flush the journal completely and mark the fs clean.
5677 *
5678 * Note that only this function cannot bring a filesystem to be in a clean
5679 * state independently. It relies on upper layer to stop all data & metadata
5680 * modifications.
5681 */
5682 static int ext4_freeze(struct super_block *sb)
5683 {
5684 int error = 0;
5685 journal_t *journal;
5686
5687 if (sb_rdonly(sb))
5688 return 0;
5689
5690 journal = EXT4_SB(sb)->s_journal;
5691
5692 if (journal) {
5693 /* Now we set up the journal barrier. */
5694 jbd2_journal_lock_updates(journal);
5695
5696 /*
5697 * Don't clear the needs_recovery flag if we failed to
5698 * flush the journal.
5699 */
5700 error = jbd2_journal_flush(journal);
5701 if (error < 0)
5702 goto out;
5703
5704 /* Journal blocked and flushed, clear needs_recovery flag. */
5705 ext4_clear_feature_journal_needs_recovery(sb);
5706 }
5707
5708 error = ext4_commit_super(sb, 1);
5709 out:
5710 if (journal)
5711 /* we rely on upper layer to stop further updates */
5712 jbd2_journal_unlock_updates(journal);
5713 return error;
5714 }
5715
5716 /*
5717 * Called by LVM after the snapshot is done. We need to reset the RECOVER
5718 * flag here, even though the filesystem is not technically dirty yet.
5719 */
5720 static int ext4_unfreeze(struct super_block *sb)
5721 {
5722 if (sb_rdonly(sb) || ext4_forced_shutdown(EXT4_SB(sb)))
5723 return 0;
5724
5725 if (EXT4_SB(sb)->s_journal) {
5726 /* Reset the needs_recovery flag before the fs is unlocked. */
5727 ext4_set_feature_journal_needs_recovery(sb);
5728 }
5729
5730 ext4_commit_super(sb, 1);
5731 return 0;
5732 }
5733
5734 /*
5735 * Structure to save mount options for ext4_remount's benefit
5736 */
5737 struct ext4_mount_options {
5738 unsigned long s_mount_opt;
5739 unsigned long s_mount_opt2;
5740 kuid_t s_resuid;
5741 kgid_t s_resgid;
5742 unsigned long s_commit_interval;
5743 u32 s_min_batch_time, s_max_batch_time;
5744 #ifdef CONFIG_QUOTA
5745 int s_jquota_fmt;
5746 char *s_qf_names[EXT4_MAXQUOTAS];
5747 #endif
5748 };
5749
5750 static int ext4_remount(struct super_block *sb, int *flags, char *data)
5751 {
5752 struct ext4_super_block *es;
5753 struct ext4_sb_info *sbi = EXT4_SB(sb);
5754 unsigned long old_sb_flags, vfs_flags;
5755 struct ext4_mount_options old_opts;
5756 int enable_quota = 0;
5757 ext4_group_t g;
5758 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
5759 int err = 0;
5760 #ifdef CONFIG_QUOTA
5761 int i, j;
5762 char *to_free[EXT4_MAXQUOTAS];
5763 #endif
5764 char *orig_data = kstrdup(data, GFP_KERNEL);
5765
5766 if (data && !orig_data)
5767 return -ENOMEM;
5768
5769 /* Store the original options */
5770 old_sb_flags = sb->s_flags;
5771 old_opts.s_mount_opt = sbi->s_mount_opt;
5772 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
5773 old_opts.s_resuid = sbi->s_resuid;
5774 old_opts.s_resgid = sbi->s_resgid;
5775 old_opts.s_commit_interval = sbi->s_commit_interval;
5776 old_opts.s_min_batch_time = sbi->s_min_batch_time;
5777 old_opts.s_max_batch_time = sbi->s_max_batch_time;
5778 #ifdef CONFIG_QUOTA
5779 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
5780 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5781 if (sbi->s_qf_names[i]) {
5782 char *qf_name = get_qf_name(sb, sbi, i);
5783
5784 old_opts.s_qf_names[i] = kstrdup(qf_name, GFP_KERNEL);
5785 if (!old_opts.s_qf_names[i]) {
5786 for (j = 0; j < i; j++)
5787 kfree(old_opts.s_qf_names[j]);
5788 kfree(orig_data);
5789 return -ENOMEM;
5790 }
5791 } else
5792 old_opts.s_qf_names[i] = NULL;
5793 #endif
5794 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
5795 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
5796
5797 /*
5798 * Some options can be enabled by ext4 and/or by VFS mount flag
5799 * either way we need to make sure it matches in both *flags and
5800 * s_flags. Copy those selected flags from *flags to s_flags
5801 */
5802 vfs_flags = SB_LAZYTIME | SB_I_VERSION;
5803 sb->s_flags = (sb->s_flags & ~vfs_flags) | (*flags & vfs_flags);
5804
5805 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
5806 err = -EINVAL;
5807 goto restore_opts;
5808 }
5809
5810 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
5811 test_opt(sb, JOURNAL_CHECKSUM)) {
5812 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
5813 "during remount not supported; ignoring");
5814 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
5815 }
5816
5817 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
5818 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
5819 ext4_msg(sb, KERN_ERR, "can't mount with "
5820 "both data=journal and delalloc");
5821 err = -EINVAL;
5822 goto restore_opts;
5823 }
5824 if (test_opt(sb, DIOREAD_NOLOCK)) {
5825 ext4_msg(sb, KERN_ERR, "can't mount with "
5826 "both data=journal and dioread_nolock");
5827 err = -EINVAL;
5828 goto restore_opts;
5829 }
5830 } else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA) {
5831 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
5832 ext4_msg(sb, KERN_ERR, "can't mount with "
5833 "journal_async_commit in data=ordered mode");
5834 err = -EINVAL;
5835 goto restore_opts;
5836 }
5837 }
5838
5839 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_NO_MBCACHE) {
5840 ext4_msg(sb, KERN_ERR, "can't enable nombcache during remount");
5841 err = -EINVAL;
5842 goto restore_opts;
5843 }
5844
5845 if (ext4_test_mount_flag(sb, EXT4_MF_FS_ABORTED))
5846 ext4_abort(sb, EXT4_ERR_ESHUTDOWN, "Abort forced by user");
5847
5848 sb->s_flags = (sb->s_flags & ~SB_POSIXACL) |
5849 (test_opt(sb, POSIX_ACL) ? SB_POSIXACL : 0);
5850
5851 es = sbi->s_es;
5852
5853 if (sbi->s_journal) {
5854 ext4_init_journal_params(sb, sbi->s_journal);
5855 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
5856 }
5857
5858 /* Flush outstanding errors before changing fs state */
5859 flush_work(&sbi->s_error_work);
5860
5861 if ((bool)(*flags & SB_RDONLY) != sb_rdonly(sb)) {
5862 if (ext4_test_mount_flag(sb, EXT4_MF_FS_ABORTED)) {
5863 err = -EROFS;
5864 goto restore_opts;
5865 }
5866
5867 if (*flags & SB_RDONLY) {
5868 err = sync_filesystem(sb);
5869 if (err < 0)
5870 goto restore_opts;
5871 err = dquot_suspend(sb, -1);
5872 if (err < 0)
5873 goto restore_opts;
5874
5875 /*
5876 * First of all, the unconditional stuff we have to do
5877 * to disable replay of the journal when we next remount
5878 */
5879 sb->s_flags |= SB_RDONLY;
5880
5881 /*
5882 * OK, test if we are remounting a valid rw partition
5883 * readonly, and if so set the rdonly flag and then
5884 * mark the partition as valid again.
5885 */
5886 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
5887 (sbi->s_mount_state & EXT4_VALID_FS))
5888 es->s_state = cpu_to_le16(sbi->s_mount_state);
5889
5890 if (sbi->s_journal) {
5891 /*
5892 * We let remount-ro finish even if marking fs
5893 * as clean failed...
5894 */
5895 ext4_mark_recovery_complete(sb, es);
5896 }
5897 if (sbi->s_mmp_tsk)
5898 kthread_stop(sbi->s_mmp_tsk);
5899 } else {
5900 /* Make sure we can mount this feature set readwrite */
5901 if (ext4_has_feature_readonly(sb) ||
5902 !ext4_feature_set_ok(sb, 0)) {
5903 err = -EROFS;
5904 goto restore_opts;
5905 }
5906 /*
5907 * Make sure the group descriptor checksums
5908 * are sane. If they aren't, refuse to remount r/w.
5909 */
5910 for (g = 0; g < sbi->s_groups_count; g++) {
5911 struct ext4_group_desc *gdp =
5912 ext4_get_group_desc(sb, g, NULL);
5913
5914 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
5915 ext4_msg(sb, KERN_ERR,
5916 "ext4_remount: Checksum for group %u failed (%u!=%u)",
5917 g, le16_to_cpu(ext4_group_desc_csum(sb, g, gdp)),
5918 le16_to_cpu(gdp->bg_checksum));
5919 err = -EFSBADCRC;
5920 goto restore_opts;
5921 }
5922 }
5923
5924 /*
5925 * If we have an unprocessed orphan list hanging
5926 * around from a previously readonly bdev mount,
5927 * require a full umount/remount for now.
5928 */
5929 if (es->s_last_orphan) {
5930 ext4_msg(sb, KERN_WARNING, "Couldn't "
5931 "remount RDWR because of unprocessed "
5932 "orphan inode list. Please "
5933 "umount/remount instead");
5934 err = -EINVAL;
5935 goto restore_opts;
5936 }
5937
5938 /*
5939 * Mounting a RDONLY partition read-write, so reread
5940 * and store the current valid flag. (It may have
5941 * been changed by e2fsck since we originally mounted
5942 * the partition.)
5943 */
5944 if (sbi->s_journal) {
5945 err = ext4_clear_journal_err(sb, es);
5946 if (err)
5947 goto restore_opts;
5948 }
5949 sbi->s_mount_state = le16_to_cpu(es->s_state);
5950
5951 err = ext4_setup_super(sb, es, 0);
5952 if (err)
5953 goto restore_opts;
5954
5955 sb->s_flags &= ~SB_RDONLY;
5956 if (ext4_has_feature_mmp(sb))
5957 if (ext4_multi_mount_protect(sb,
5958 le64_to_cpu(es->s_mmp_block))) {
5959 err = -EROFS;
5960 goto restore_opts;
5961 }
5962 enable_quota = 1;
5963 }
5964 }
5965
5966 /*
5967 * Reinitialize lazy itable initialization thread based on
5968 * current settings
5969 */
5970 if (sb_rdonly(sb) || !test_opt(sb, INIT_INODE_TABLE))
5971 ext4_unregister_li_request(sb);
5972 else {
5973 ext4_group_t first_not_zeroed;
5974 first_not_zeroed = ext4_has_uninit_itable(sb);
5975 ext4_register_li_request(sb, first_not_zeroed);
5976 }
5977
5978 /*
5979 * Handle creation of system zone data early because it can fail.
5980 * Releasing of existing data is done when we are sure remount will
5981 * succeed.
5982 */
5983 if (test_opt(sb, BLOCK_VALIDITY) && !sbi->s_system_blks) {
5984 err = ext4_setup_system_zone(sb);
5985 if (err)
5986 goto restore_opts;
5987 }
5988
5989 if (sbi->s_journal == NULL && !(old_sb_flags & SB_RDONLY)) {
5990 err = ext4_commit_super(sb, 1);
5991 if (err)
5992 goto restore_opts;
5993 }
5994
5995 #ifdef CONFIG_QUOTA
5996 /* Release old quota file names */
5997 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5998 kfree(old_opts.s_qf_names[i]);
5999 if (enable_quota) {
6000 if (sb_any_quota_suspended(sb))
6001 dquot_resume(sb, -1);
6002 else if (ext4_has_feature_quota(sb)) {
6003 err = ext4_enable_quotas(sb);
6004 if (err)
6005 goto restore_opts;
6006 }
6007 }
6008 #endif
6009 if (!test_opt(sb, BLOCK_VALIDITY) && sbi->s_system_blks)
6010 ext4_release_system_zone(sb);
6011
6012 /*
6013 * Some options can be enabled by ext4 and/or by VFS mount flag
6014 * either way we need to make sure it matches in both *flags and
6015 * s_flags. Copy those selected flags from s_flags to *flags
6016 */
6017 *flags = (*flags & ~vfs_flags) | (sb->s_flags & vfs_flags);
6018
6019 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s. Quota mode: %s.",
6020 orig_data, ext4_quota_mode(sb));
6021 kfree(orig_data);
6022 return 0;
6023
6024 restore_opts:
6025 sb->s_flags = old_sb_flags;
6026 sbi->s_mount_opt = old_opts.s_mount_opt;
6027 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
6028 sbi->s_resuid = old_opts.s_resuid;
6029 sbi->s_resgid = old_opts.s_resgid;
6030 sbi->s_commit_interval = old_opts.s_commit_interval;
6031 sbi->s_min_batch_time = old_opts.s_min_batch_time;
6032 sbi->s_max_batch_time = old_opts.s_max_batch_time;
6033 if (!test_opt(sb, BLOCK_VALIDITY) && sbi->s_system_blks)
6034 ext4_release_system_zone(sb);
6035 #ifdef CONFIG_QUOTA
6036 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
6037 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
6038 to_free[i] = get_qf_name(sb, sbi, i);
6039 rcu_assign_pointer(sbi->s_qf_names[i], old_opts.s_qf_names[i]);
6040 }
6041 synchronize_rcu();
6042 for (i = 0; i < EXT4_MAXQUOTAS; i++)
6043 kfree(to_free[i]);
6044 #endif
6045 kfree(orig_data);
6046 return err;
6047 }
6048
6049 #ifdef CONFIG_QUOTA
6050 static int ext4_statfs_project(struct super_block *sb,
6051 kprojid_t projid, struct kstatfs *buf)
6052 {
6053 struct kqid qid;
6054 struct dquot *dquot;
6055 u64 limit;
6056 u64 curblock;
6057
6058 qid = make_kqid_projid(projid);
6059 dquot = dqget(sb, qid);
6060 if (IS_ERR(dquot))
6061 return PTR_ERR(dquot);
6062 spin_lock(&dquot->dq_dqb_lock);
6063
6064 limit = min_not_zero(dquot->dq_dqb.dqb_bsoftlimit,
6065 dquot->dq_dqb.dqb_bhardlimit);
6066 limit >>= sb->s_blocksize_bits;
6067
6068 if (limit && buf->f_blocks > limit) {
6069 curblock = (dquot->dq_dqb.dqb_curspace +
6070 dquot->dq_dqb.dqb_rsvspace) >> sb->s_blocksize_bits;
6071 buf->f_blocks = limit;
6072 buf->f_bfree = buf->f_bavail =
6073 (buf->f_blocks > curblock) ?
6074 (buf->f_blocks - curblock) : 0;
6075 }
6076
6077 limit = min_not_zero(dquot->dq_dqb.dqb_isoftlimit,
6078 dquot->dq_dqb.dqb_ihardlimit);
6079 if (limit && buf->f_files > limit) {
6080 buf->f_files = limit;
6081 buf->f_ffree =
6082 (buf->f_files > dquot->dq_dqb.dqb_curinodes) ?
6083 (buf->f_files - dquot->dq_dqb.dqb_curinodes) : 0;
6084 }
6085
6086 spin_unlock(&dquot->dq_dqb_lock);
6087 dqput(dquot);
6088 return 0;
6089 }
6090 #endif
6091
6092 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
6093 {
6094 struct super_block *sb = dentry->d_sb;
6095 struct ext4_sb_info *sbi = EXT4_SB(sb);
6096 struct ext4_super_block *es = sbi->s_es;
6097 ext4_fsblk_t overhead = 0, resv_blocks;
6098 u64 fsid;
6099 s64 bfree;
6100 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
6101
6102 if (!test_opt(sb, MINIX_DF))
6103 overhead = sbi->s_overhead;
6104
6105 buf->f_type = EXT4_SUPER_MAGIC;
6106 buf->f_bsize = sb->s_blocksize;
6107 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
6108 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
6109 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
6110 /* prevent underflow in case that few free space is available */
6111 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
6112 buf->f_bavail = buf->f_bfree -
6113 (ext4_r_blocks_count(es) + resv_blocks);
6114 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
6115 buf->f_bavail = 0;
6116 buf->f_files = le32_to_cpu(es->s_inodes_count);
6117 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
6118 buf->f_namelen = EXT4_NAME_LEN;
6119 fsid = le64_to_cpup((void *)es->s_uuid) ^
6120 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
6121 buf->f_fsid = u64_to_fsid(fsid);
6122
6123 #ifdef CONFIG_QUOTA
6124 if (ext4_test_inode_flag(dentry->d_inode, EXT4_INODE_PROJINHERIT) &&
6125 sb_has_quota_limits_enabled(sb, PRJQUOTA))
6126 ext4_statfs_project(sb, EXT4_I(dentry->d_inode)->i_projid, buf);
6127 #endif
6128 return 0;
6129 }
6130
6131
6132 #ifdef CONFIG_QUOTA
6133
6134 /*
6135 * Helper functions so that transaction is started before we acquire dqio_sem
6136 * to keep correct lock ordering of transaction > dqio_sem
6137 */
6138 static inline struct inode *dquot_to_inode(struct dquot *dquot)
6139 {
6140 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
6141 }
6142
6143 static int ext4_write_dquot(struct dquot *dquot)
6144 {
6145 int ret, err;
6146 handle_t *handle;
6147 struct inode *inode;
6148
6149 inode = dquot_to_inode(dquot);
6150 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
6151 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
6152 if (IS_ERR(handle))
6153 return PTR_ERR(handle);
6154 ret = dquot_commit(dquot);
6155 err = ext4_journal_stop(handle);
6156 if (!ret)
6157 ret = err;
6158 return ret;
6159 }
6160
6161 static int ext4_acquire_dquot(struct dquot *dquot)
6162 {
6163 int ret, err;
6164 handle_t *handle;
6165
6166 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
6167 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
6168 if (IS_ERR(handle))
6169 return PTR_ERR(handle);
6170 ret = dquot_acquire(dquot);
6171 err = ext4_journal_stop(handle);
6172 if (!ret)
6173 ret = err;
6174 return ret;
6175 }
6176
6177 static int ext4_release_dquot(struct dquot *dquot)
6178 {
6179 int ret, err;
6180 handle_t *handle;
6181
6182 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
6183 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
6184 if (IS_ERR(handle)) {
6185 /* Release dquot anyway to avoid endless cycle in dqput() */
6186 dquot_release(dquot);
6187 return PTR_ERR(handle);
6188 }
6189 ret = dquot_release(dquot);
6190 err = ext4_journal_stop(handle);
6191 if (!ret)
6192 ret = err;
6193 return ret;
6194 }
6195
6196 static int ext4_mark_dquot_dirty(struct dquot *dquot)
6197 {
6198 struct super_block *sb = dquot->dq_sb;
6199
6200 if (ext4_is_quota_journalled(sb)) {
6201 dquot_mark_dquot_dirty(dquot);
6202 return ext4_write_dquot(dquot);
6203 } else {
6204 return dquot_mark_dquot_dirty(dquot);
6205 }
6206 }
6207
6208 static int ext4_write_info(struct super_block *sb, int type)
6209 {
6210 int ret, err;
6211 handle_t *handle;
6212
6213 /* Data block + inode block */
6214 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
6215 if (IS_ERR(handle))
6216 return PTR_ERR(handle);
6217 ret = dquot_commit_info(sb, type);
6218 err = ext4_journal_stop(handle);
6219 if (!ret)
6220 ret = err;
6221 return ret;
6222 }
6223
6224 /*
6225 * Turn on quotas during mount time - we need to find
6226 * the quota file and such...
6227 */
6228 static int ext4_quota_on_mount(struct super_block *sb, int type)
6229 {
6230 return dquot_quota_on_mount(sb, get_qf_name(sb, EXT4_SB(sb), type),
6231 EXT4_SB(sb)->s_jquota_fmt, type);
6232 }
6233
6234 static void lockdep_set_quota_inode(struct inode *inode, int subclass)
6235 {
6236 struct ext4_inode_info *ei = EXT4_I(inode);
6237
6238 /* The first argument of lockdep_set_subclass has to be
6239 * *exactly* the same as the argument to init_rwsem() --- in
6240 * this case, in init_once() --- or lockdep gets unhappy
6241 * because the name of the lock is set using the
6242 * stringification of the argument to init_rwsem().
6243 */
6244 (void) ei; /* shut up clang warning if !CONFIG_LOCKDEP */
6245 lockdep_set_subclass(&ei->i_data_sem, subclass);
6246 }
6247
6248 /*
6249 * Standard function to be called on quota_on
6250 */
6251 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
6252 const struct path *path)
6253 {
6254 int err;
6255
6256 if (!test_opt(sb, QUOTA))
6257 return -EINVAL;
6258
6259 /* Quotafile not on the same filesystem? */
6260 if (path->dentry->d_sb != sb)
6261 return -EXDEV;
6262
6263 /* Quota already enabled for this file? */
6264 if (IS_NOQUOTA(d_inode(path->dentry)))
6265 return -EBUSY;
6266
6267 /* Journaling quota? */
6268 if (EXT4_SB(sb)->s_qf_names[type]) {
6269 /* Quotafile not in fs root? */
6270 if (path->dentry->d_parent != sb->s_root)
6271 ext4_msg(sb, KERN_WARNING,
6272 "Quota file not on filesystem root. "
6273 "Journaled quota will not work");
6274 sb_dqopt(sb)->flags |= DQUOT_NOLIST_DIRTY;
6275 } else {
6276 /*
6277 * Clear the flag just in case mount options changed since
6278 * last time.
6279 */
6280 sb_dqopt(sb)->flags &= ~DQUOT_NOLIST_DIRTY;
6281 }
6282
6283 /*
6284 * When we journal data on quota file, we have to flush journal to see
6285 * all updates to the file when we bypass pagecache...
6286 */
6287 if (EXT4_SB(sb)->s_journal &&
6288 ext4_should_journal_data(d_inode(path->dentry))) {
6289 /*
6290 * We don't need to lock updates but journal_flush() could
6291 * otherwise be livelocked...
6292 */
6293 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
6294 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
6295 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
6296 if (err)
6297 return err;
6298 }
6299
6300 lockdep_set_quota_inode(path->dentry->d_inode, I_DATA_SEM_QUOTA);
6301 err = dquot_quota_on(sb, type, format_id, path);
6302 if (err) {
6303 lockdep_set_quota_inode(path->dentry->d_inode,
6304 I_DATA_SEM_NORMAL);
6305 } else {
6306 struct inode *inode = d_inode(path->dentry);
6307 handle_t *handle;
6308
6309 /*
6310 * Set inode flags to prevent userspace from messing with quota
6311 * files. If this fails, we return success anyway since quotas
6312 * are already enabled and this is not a hard failure.
6313 */
6314 inode_lock(inode);
6315 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
6316 if (IS_ERR(handle))
6317 goto unlock_inode;
6318 EXT4_I(inode)->i_flags |= EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL;
6319 inode_set_flags(inode, S_NOATIME | S_IMMUTABLE,
6320 S_NOATIME | S_IMMUTABLE);
6321 err = ext4_mark_inode_dirty(handle, inode);
6322 ext4_journal_stop(handle);
6323 unlock_inode:
6324 inode_unlock(inode);
6325 }
6326 return err;
6327 }
6328
6329 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
6330 unsigned int flags)
6331 {
6332 int err;
6333 struct inode *qf_inode;
6334 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
6335 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
6336 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
6337 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
6338 };
6339
6340 BUG_ON(!ext4_has_feature_quota(sb));
6341
6342 if (!qf_inums[type])
6343 return -EPERM;
6344
6345 qf_inode = ext4_iget(sb, qf_inums[type], EXT4_IGET_SPECIAL);
6346 if (IS_ERR(qf_inode)) {
6347 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
6348 return PTR_ERR(qf_inode);
6349 }
6350
6351 /* Don't account quota for quota files to avoid recursion */
6352 qf_inode->i_flags |= S_NOQUOTA;
6353 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_QUOTA);
6354 err = dquot_load_quota_inode(qf_inode, type, format_id, flags);
6355 if (err)
6356 lockdep_set_quota_inode(qf_inode, I_DATA_SEM_NORMAL);
6357 iput(qf_inode);
6358
6359 return err;
6360 }
6361
6362 /* Enable usage tracking for all quota types. */
6363 static int ext4_enable_quotas(struct super_block *sb)
6364 {
6365 int type, err = 0;
6366 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
6367 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
6368 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum),
6369 le32_to_cpu(EXT4_SB(sb)->s_es->s_prj_quota_inum)
6370 };
6371 bool quota_mopt[EXT4_MAXQUOTAS] = {
6372 test_opt(sb, USRQUOTA),
6373 test_opt(sb, GRPQUOTA),
6374 test_opt(sb, PRJQUOTA),
6375 };
6376
6377 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE | DQUOT_NOLIST_DIRTY;
6378 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
6379 if (qf_inums[type]) {
6380 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
6381 DQUOT_USAGE_ENABLED |
6382 (quota_mopt[type] ? DQUOT_LIMITS_ENABLED : 0));
6383 if (err) {
6384 ext4_warning(sb,
6385 "Failed to enable quota tracking "
6386 "(type=%d, err=%d). Please run "
6387 "e2fsck to fix.", type, err);
6388 for (type--; type >= 0; type--)
6389 dquot_quota_off(sb, type);
6390
6391 return err;
6392 }
6393 }
6394 }
6395 return 0;
6396 }
6397
6398 static int ext4_quota_off(struct super_block *sb, int type)
6399 {
6400 struct inode *inode = sb_dqopt(sb)->files[type];
6401 handle_t *handle;
6402 int err;
6403
6404 /* Force all delayed allocation blocks to be allocated.
6405 * Caller already holds s_umount sem */
6406 if (test_opt(sb, DELALLOC))
6407 sync_filesystem(sb);
6408
6409 if (!inode || !igrab(inode))
6410 goto out;
6411
6412 err = dquot_quota_off(sb, type);
6413 if (err || ext4_has_feature_quota(sb))
6414 goto out_put;
6415
6416 inode_lock(inode);
6417 /*
6418 * Update modification times of quota files when userspace can
6419 * start looking at them. If we fail, we return success anyway since
6420 * this is not a hard failure and quotas are already disabled.
6421 */
6422 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
6423 if (IS_ERR(handle)) {
6424 err = PTR_ERR(handle);
6425 goto out_unlock;
6426 }
6427 EXT4_I(inode)->i_flags &= ~(EXT4_NOATIME_FL | EXT4_IMMUTABLE_FL);
6428 inode_set_flags(inode, 0, S_NOATIME | S_IMMUTABLE);
6429 inode->i_mtime = inode->i_ctime = current_time(inode);
6430 err = ext4_mark_inode_dirty(handle, inode);
6431 ext4_journal_stop(handle);
6432 out_unlock:
6433 inode_unlock(inode);
6434 out_put:
6435 lockdep_set_quota_inode(inode, I_DATA_SEM_NORMAL);
6436 iput(inode);
6437 return err;
6438 out:
6439 return dquot_quota_off(sb, type);
6440 }
6441
6442 /* Read data from quotafile - avoid pagecache and such because we cannot afford
6443 * acquiring the locks... As quota files are never truncated and quota code
6444 * itself serializes the operations (and no one else should touch the files)
6445 * we don't have to be afraid of races */
6446 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
6447 size_t len, loff_t off)
6448 {
6449 struct inode *inode = sb_dqopt(sb)->files[type];
6450 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
6451 int offset = off & (sb->s_blocksize - 1);
6452 int tocopy;
6453 size_t toread;
6454 struct buffer_head *bh;
6455 loff_t i_size = i_size_read(inode);
6456
6457 if (off > i_size)
6458 return 0;
6459 if (off+len > i_size)
6460 len = i_size-off;
6461 toread = len;
6462 while (toread > 0) {
6463 tocopy = sb->s_blocksize - offset < toread ?
6464 sb->s_blocksize - offset : toread;
6465 bh = ext4_bread(NULL, inode, blk, 0);
6466 if (IS_ERR(bh))
6467 return PTR_ERR(bh);
6468 if (!bh) /* A hole? */
6469 memset(data, 0, tocopy);
6470 else
6471 memcpy(data, bh->b_data+offset, tocopy);
6472 brelse(bh);
6473 offset = 0;
6474 toread -= tocopy;
6475 data += tocopy;
6476 blk++;
6477 }
6478 return len;
6479 }
6480
6481 /* Write to quotafile (we know the transaction is already started and has
6482 * enough credits) */
6483 static ssize_t ext4_quota_write(struct super_block *sb, int type,
6484 const char *data, size_t len, loff_t off)
6485 {
6486 struct inode *inode = sb_dqopt(sb)->files[type];
6487 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
6488 int err = 0, err2 = 0, offset = off & (sb->s_blocksize - 1);
6489 int retries = 0;
6490 struct buffer_head *bh;
6491 handle_t *handle = journal_current_handle();
6492
6493 if (EXT4_SB(sb)->s_journal && !handle) {
6494 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
6495 " cancelled because transaction is not started",
6496 (unsigned long long)off, (unsigned long long)len);
6497 return -EIO;
6498 }
6499 /*
6500 * Since we account only one data block in transaction credits,
6501 * then it is impossible to cross a block boundary.
6502 */
6503 if (sb->s_blocksize - offset < len) {
6504 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
6505 " cancelled because not block aligned",
6506 (unsigned long long)off, (unsigned long long)len);
6507 return -EIO;
6508 }
6509
6510 do {
6511 bh = ext4_bread(handle, inode, blk,
6512 EXT4_GET_BLOCKS_CREATE |
6513 EXT4_GET_BLOCKS_METADATA_NOFAIL);
6514 } while (PTR_ERR(bh) == -ENOSPC &&
6515 ext4_should_retry_alloc(inode->i_sb, &retries));
6516 if (IS_ERR(bh))
6517 return PTR_ERR(bh);
6518 if (!bh)
6519 goto out;
6520 BUFFER_TRACE(bh, "get write access");
6521 err = ext4_journal_get_write_access(handle, bh);
6522 if (err) {
6523 brelse(bh);
6524 return err;
6525 }
6526 lock_buffer(bh);
6527 memcpy(bh->b_data+offset, data, len);
6528 flush_dcache_page(bh->b_page);
6529 unlock_buffer(bh);
6530 err = ext4_handle_dirty_metadata(handle, NULL, bh);
6531 brelse(bh);
6532 out:
6533 if (inode->i_size < off + len) {
6534 i_size_write(inode, off + len);
6535 EXT4_I(inode)->i_disksize = inode->i_size;
6536 err2 = ext4_mark_inode_dirty(handle, inode);
6537 if (unlikely(err2 && !err))
6538 err = err2;
6539 }
6540 return err ? err : len;
6541 }
6542 #endif
6543
6544 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
6545 const char *dev_name, void *data)
6546 {
6547 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
6548 }
6549
6550 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
6551 static inline void register_as_ext2(void)
6552 {
6553 int err = register_filesystem(&ext2_fs_type);
6554 if (err)
6555 printk(KERN_WARNING
6556 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
6557 }
6558
6559 static inline void unregister_as_ext2(void)
6560 {
6561 unregister_filesystem(&ext2_fs_type);
6562 }
6563
6564 static inline int ext2_feature_set_ok(struct super_block *sb)
6565 {
6566 if (ext4_has_unknown_ext2_incompat_features(sb))
6567 return 0;
6568 if (sb_rdonly(sb))
6569 return 1;
6570 if (ext4_has_unknown_ext2_ro_compat_features(sb))
6571 return 0;
6572 return 1;
6573 }
6574 #else
6575 static inline void register_as_ext2(void) { }
6576 static inline void unregister_as_ext2(void) { }
6577 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
6578 #endif
6579
6580 static inline void register_as_ext3(void)
6581 {
6582 int err = register_filesystem(&ext3_fs_type);
6583 if (err)
6584 printk(KERN_WARNING
6585 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
6586 }
6587
6588 static inline void unregister_as_ext3(void)
6589 {
6590 unregister_filesystem(&ext3_fs_type);
6591 }
6592
6593 static inline int ext3_feature_set_ok(struct super_block *sb)
6594 {
6595 if (ext4_has_unknown_ext3_incompat_features(sb))
6596 return 0;
6597 if (!ext4_has_feature_journal(sb))
6598 return 0;
6599 if (sb_rdonly(sb))
6600 return 1;
6601 if (ext4_has_unknown_ext3_ro_compat_features(sb))
6602 return 0;
6603 return 1;
6604 }
6605
6606 static struct file_system_type ext4_fs_type = {
6607 .owner = THIS_MODULE,
6608 .name = "ext4",
6609 .mount = ext4_mount,
6610 .kill_sb = kill_block_super,
6611 .fs_flags = FS_REQUIRES_DEV,
6612 };
6613 MODULE_ALIAS_FS("ext4");
6614
6615 /* Shared across all ext4 file systems */
6616 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
6617
6618 static int __init ext4_init_fs(void)
6619 {
6620 int i, err;
6621
6622 ratelimit_state_init(&ext4_mount_msg_ratelimit, 30 * HZ, 64);
6623 ext4_li_info = NULL;
6624 mutex_init(&ext4_li_mtx);
6625
6626 /* Build-time check for flags consistency */
6627 ext4_check_flag_values();
6628
6629 for (i = 0; i < EXT4_WQ_HASH_SZ; i++)
6630 init_waitqueue_head(&ext4__ioend_wq[i]);
6631
6632 err = ext4_init_es();
6633 if (err)
6634 return err;
6635
6636 err = ext4_init_pending();
6637 if (err)
6638 goto out7;
6639
6640 err = ext4_init_post_read_processing();
6641 if (err)
6642 goto out6;
6643
6644 err = ext4_init_pageio();
6645 if (err)
6646 goto out5;
6647
6648 err = ext4_init_system_zone();
6649 if (err)
6650 goto out4;
6651
6652 err = ext4_init_sysfs();
6653 if (err)
6654 goto out3;
6655
6656 err = ext4_init_mballoc();
6657 if (err)
6658 goto out2;
6659 err = init_inodecache();
6660 if (err)
6661 goto out1;
6662
6663 err = ext4_fc_init_dentry_cache();
6664 if (err)
6665 goto out05;
6666
6667 register_as_ext3();
6668 register_as_ext2();
6669 err = register_filesystem(&ext4_fs_type);
6670 if (err)
6671 goto out;
6672
6673 return 0;
6674 out:
6675 unregister_as_ext2();
6676 unregister_as_ext3();
6677 out05:
6678 destroy_inodecache();
6679 out1:
6680 ext4_exit_mballoc();
6681 out2:
6682 ext4_exit_sysfs();
6683 out3:
6684 ext4_exit_system_zone();
6685 out4:
6686 ext4_exit_pageio();
6687 out5:
6688 ext4_exit_post_read_processing();
6689 out6:
6690 ext4_exit_pending();
6691 out7:
6692 ext4_exit_es();
6693
6694 return err;
6695 }
6696
6697 static void __exit ext4_exit_fs(void)
6698 {
6699 ext4_destroy_lazyinit_thread();
6700 unregister_as_ext2();
6701 unregister_as_ext3();
6702 unregister_filesystem(&ext4_fs_type);
6703 destroy_inodecache();
6704 ext4_exit_mballoc();
6705 ext4_exit_sysfs();
6706 ext4_exit_system_zone();
6707 ext4_exit_pageio();
6708 ext4_exit_post_read_processing();
6709 ext4_exit_es();
6710 ext4_exit_pending();
6711 }
6712
6713 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
6714 MODULE_DESCRIPTION("Fourth Extended Filesystem");
6715 MODULE_LICENSE("GPL");
6716 MODULE_SOFTDEP("pre: crc32c");
6717 module_init(ext4_init_fs)
6718 module_exit(ext4_exit_fs)
Linux with added FPC-III support