136940af00b888c17f37e1875d2b5af7dbbe1be7
[linux-2.6-microblaze.git] / fs / ext4 / super.c
1 // SPDX-License-Identifier: GPL-2.0
2 /*
3  *  linux/fs/ext4/super.c
4  *
5  * Copyright (C) 1992, 1993, 1994, 1995
6  * Remy Card (card@masi.ibp.fr)
7  * Laboratoire MASI - Institut Blaise Pascal
8  * Universite Pierre et Marie Curie (Paris VI)
9  *
10  *  from
11  *
12  *  linux/fs/minix/inode.c
13  *
14  *  Copyright (C) 1991, 1992  Linus Torvalds
15  *
16  *  Big-endian to little-endian byte-swapping/bitmaps by
17  *        David S. Miller (davem@caip.rutgers.edu), 1995
18  */
19
20 #include <linux/module.h>
21 #include <linux/string.h>
22 #include <linux/fs.h>
23 #include <linux/time.h>
24 #include <linux/vmalloc.h>
25 #include <linux/slab.h>
26 #include <linux/init.h>
27 #include <linux/blkdev.h>
28 #include <linux/backing-dev.h>
29 #include <linux/parser.h>
30 #include <linux/buffer_head.h>
31 #include <linux/exportfs.h>
32 #include <linux/vfs.h>
33 #include <linux/random.h>
34 #include <linux/mount.h>
35 #include <linux/namei.h>
36 #include <linux/quotaops.h>
37 #include <linux/seq_file.h>
38 #include <linux/ctype.h>
39 #include <linux/log2.h>
40 #include <linux/crc16.h>
41 #include <linux/dax.h>
42 #include <linux/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 DEFINE_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 void ext4_update_super(struct super_block *sb);
69 static int ext4_commit_super(struct super_block *sb);
70 static int ext4_mark_recovery_complete(struct super_block *sb,
71                                         struct ext4_super_block *es);
72 static int ext4_clear_journal_err(struct super_block *sb,
73                                   struct ext4_super_block *es);
74 static int ext4_sync_fs(struct super_block *sb, int wait);
75 static int ext4_remount(struct super_block *sb, int *flags, char *data);
76 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
77 static int ext4_unfreeze(struct super_block *sb);
78 static int ext4_freeze(struct super_block *sb);
79 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
80                        const char *dev_name, void *data);
81 static inline int ext2_feature_set_ok(struct super_block *sb);
82 static inline int ext3_feature_set_ok(struct super_block *sb);
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  * page fault path:
93  * mmap_lock -> sb_start_pagefault -> invalidate_lock (r) -> transaction start
94  *   -> page lock -> i_data_sem (rw)
95  *
96  * buffered write path:
97  * sb_start_write -> i_mutex -> mmap_lock
98  * sb_start_write -> i_mutex -> transaction start -> page lock ->
99  *   i_data_sem (rw)
100  *
101  * truncate:
102  * sb_start_write -> i_mutex -> invalidate_lock (w) -> i_mmap_rwsem (w) ->
103  *   page lock
104  * sb_start_write -> i_mutex -> invalidate_lock (w) -> transaction start ->
105  *   i_data_sem (rw)
106  *
107  * direct IO:
108  * sb_start_write -> i_mutex -> mmap_lock
109  * sb_start_write -> i_mutex -> transaction start -> i_data_sem (rw)
110  *
111  * writepages:
112  * transaction start -> page lock(s) -> i_data_sem (rw)
113  */
114
115 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
116 static struct file_system_type ext2_fs_type = {
117         .owner          = THIS_MODULE,
118         .name           = "ext2",
119         .mount          = ext4_mount,
120         .kill_sb        = kill_block_super,
121         .fs_flags       = FS_REQUIRES_DEV,
122 };
123 MODULE_ALIAS_FS("ext2");
124 MODULE_ALIAS("ext2");
125 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
126 #else
127 #define IS_EXT2_SB(sb) (0)
128 #endif
129
130
131 static struct file_system_type ext3_fs_type = {
132         .owner          = THIS_MODULE,
133         .name           = "ext3",
134         .mount          = ext4_mount,
135         .kill_sb        = kill_block_super,
136         .fs_flags       = FS_REQUIRES_DEV,
137 };
138 MODULE_ALIAS_FS("ext3");
139 MODULE_ALIAS("ext3");
140 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
141
142
143 static inline void __ext4_read_bh(struct buffer_head *bh, int op_flags,
144                                   bh_end_io_t *end_io)
145 {
146         /*
147          * buffer's verified bit is no longer valid after reading from
148          * disk again due to write out error, clear it to make sure we
149          * recheck the buffer contents.
150          */
151         clear_buffer_verified(bh);
152
153         bh->b_end_io = end_io ? end_io : end_buffer_read_sync;
154         get_bh(bh);
155         submit_bh(REQ_OP_READ, op_flags, bh);
156 }
157
158 void ext4_read_bh_nowait(struct buffer_head *bh, int op_flags,
159                          bh_end_io_t *end_io)
160 {
161         BUG_ON(!buffer_locked(bh));
162
163         if (ext4_buffer_uptodate(bh)) {
164                 unlock_buffer(bh);
165                 return;
166         }
167         __ext4_read_bh(bh, op_flags, end_io);
168 }
169
170 int ext4_read_bh(struct buffer_head *bh, int op_flags, bh_end_io_t *end_io)
171 {
172         BUG_ON(!buffer_locked(bh));
173
174         if (ext4_buffer_uptodate(bh)) {
175                 unlock_buffer(bh);
176                 return 0;
177         }
178
179         __ext4_read_bh(bh, op_flags, end_io);
180
181         wait_on_buffer(bh);
182         if (buffer_uptodate(bh))
183                 return 0;
184         return -EIO;
185 }
186
187 int ext4_read_bh_lock(struct buffer_head *bh, int op_flags, bool wait)
188 {
189         if (trylock_buffer(bh)) {
190                 if (wait)
191                         return ext4_read_bh(bh, op_flags, NULL);
192                 ext4_read_bh_nowait(bh, op_flags, NULL);
193                 return 0;
194         }
195         if (wait) {
196                 wait_on_buffer(bh);
197                 if (buffer_uptodate(bh))
198                         return 0;
199                 return -EIO;
200         }
201         return 0;
202 }
203
204 /*
205  * This works like __bread_gfp() except it uses ERR_PTR for error
206  * returns.  Currently with sb_bread it's impossible to distinguish
207  * between ENOMEM and EIO situations (since both result in a NULL
208  * return.
209  */
210 static struct buffer_head *__ext4_sb_bread_gfp(struct super_block *sb,
211                                                sector_t block, int op_flags,
212                                                gfp_t gfp)
213 {
214         struct buffer_head *bh;
215         int ret;
216
217         bh = sb_getblk_gfp(sb, block, gfp);
218         if (bh == NULL)
219                 return ERR_PTR(-ENOMEM);
220         if (ext4_buffer_uptodate(bh))
221                 return bh;
222
223         ret = ext4_read_bh_lock(bh, REQ_META | op_flags, true);
224         if (ret) {
225                 put_bh(bh);
226                 return ERR_PTR(ret);
227         }
228         return bh;
229 }
230
231 struct buffer_head *ext4_sb_bread(struct super_block *sb, sector_t block,
232                                    int op_flags)
233 {
234         return __ext4_sb_bread_gfp(sb, block, op_flags, __GFP_MOVABLE);
235 }
236
237 struct buffer_head *ext4_sb_bread_unmovable(struct super_block *sb,
238                                             sector_t block)
239 {
240         return __ext4_sb_bread_gfp(sb, block, 0, 0);
241 }
242
243 void ext4_sb_breadahead_unmovable(struct super_block *sb, sector_t block)
244 {
245         struct buffer_head *bh = sb_getblk_gfp(sb, block, 0);
246
247         if (likely(bh)) {
248                 ext4_read_bh_lock(bh, REQ_RAHEAD, false);
249                 brelse(bh);
250         }
251 }
252
253 static int ext4_verify_csum_type(struct super_block *sb,
254                                  struct ext4_super_block *es)
255 {
256         if (!ext4_has_feature_metadata_csum(sb))
257                 return 1;
258
259         return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
260 }
261
262 static __le32 ext4_superblock_csum(struct super_block *sb,
263                                    struct ext4_super_block *es)
264 {
265         struct ext4_sb_info *sbi = EXT4_SB(sb);
266         int offset = offsetof(struct ext4_super_block, s_checksum);
267         __u32 csum;
268
269         csum = ext4_chksum(sbi, ~0, (char *)es, offset);
270
271         return cpu_to_le32(csum);
272 }
273
274 static int ext4_superblock_csum_verify(struct super_block *sb,
275                                        struct ext4_super_block *es)
276 {
277         if (!ext4_has_metadata_csum(sb))
278                 return 1;
279
280         return es->s_checksum == ext4_superblock_csum(sb, es);
281 }
282
283 void ext4_superblock_csum_set(struct super_block *sb)
284 {
285         struct ext4_super_block *es = EXT4_SB(sb)->s_es;
286
287         if (!ext4_has_metadata_csum(sb))
288                 return;
289
290         es->s_checksum = ext4_superblock_csum(sb, es);
291 }
292
293 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
294                                struct ext4_group_desc *bg)
295 {
296         return le32_to_cpu(bg->bg_block_bitmap_lo) |
297                 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
298                  (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
299 }
300
301 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
302                                struct ext4_group_desc *bg)
303 {
304         return le32_to_cpu(bg->bg_inode_bitmap_lo) |
305                 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
306                  (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
307 }
308
309 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
310                               struct ext4_group_desc *bg)
311 {
312         return le32_to_cpu(bg->bg_inode_table_lo) |
313                 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
314                  (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
315 }
316
317 __u32 ext4_free_group_clusters(struct super_block *sb,
318                                struct ext4_group_desc *bg)
319 {
320         return le16_to_cpu(bg->bg_free_blocks_count_lo) |
321                 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
322                  (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
323 }
324
325 __u32 ext4_free_inodes_count(struct super_block *sb,
326                               struct ext4_group_desc *bg)
327 {
328         return le16_to_cpu(bg->bg_free_inodes_count_lo) |
329                 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
330                  (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
331 }
332
333 __u32 ext4_used_dirs_count(struct super_block *sb,
334                               struct ext4_group_desc *bg)
335 {
336         return le16_to_cpu(bg->bg_used_dirs_count_lo) |
337                 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
338                  (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
339 }
340
341 __u32 ext4_itable_unused_count(struct super_block *sb,
342                               struct ext4_group_desc *bg)
343 {
344         return le16_to_cpu(bg->bg_itable_unused_lo) |
345                 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
346                  (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
347 }
348
349 void ext4_block_bitmap_set(struct super_block *sb,
350                            struct ext4_group_desc *bg, ext4_fsblk_t blk)
351 {
352         bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
353         if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
354                 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
355 }
356
357 void ext4_inode_bitmap_set(struct super_block *sb,
358                            struct ext4_group_desc *bg, ext4_fsblk_t blk)
359 {
360         bg->bg_inode_bitmap_lo  = cpu_to_le32((u32)blk);
361         if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
362                 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
363 }
364
365 void ext4_inode_table_set(struct super_block *sb,
366                           struct ext4_group_desc *bg, ext4_fsblk_t blk)
367 {
368         bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
369         if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
370                 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
371 }
372
373 void ext4_free_group_clusters_set(struct super_block *sb,
374                                   struct ext4_group_desc *bg, __u32 count)
375 {
376         bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
377         if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
378                 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
379 }
380
381 void ext4_free_inodes_set(struct super_block *sb,
382                           struct ext4_group_desc *bg, __u32 count)
383 {
384         bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
385         if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
386                 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
387 }
388
389 void ext4_used_dirs_set(struct super_block *sb,
390                           struct ext4_group_desc *bg, __u32 count)
391 {
392         bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
393         if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
394                 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
395 }
396
397 void ext4_itable_unused_set(struct super_block *sb,
398                           struct ext4_group_desc *bg, __u32 count)
399 {
400         bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
401         if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
402                 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
403 }
404
405 static void __ext4_update_tstamp(__le32 *lo, __u8 *hi, time64_t now)
406 {
407         now = clamp_val(now, 0, (1ull << 40) - 1);
408
409         *lo = cpu_to_le32(lower_32_bits(now));
410         *hi = upper_32_bits(now);
411 }
412
413 static time64_t __ext4_get_tstamp(__le32 *lo, __u8 *hi)
414 {
415         return ((time64_t)(*hi) << 32) + le32_to_cpu(*lo);
416 }
417 #define ext4_update_tstamp(es, tstamp) \
418         __ext4_update_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi, \
419                              ktime_get_real_seconds())
420 #define ext4_get_tstamp(es, tstamp) \
421         __ext4_get_tstamp(&(es)->tstamp, &(es)->tstamp ## _hi)
422
423 /*
424  * The del_gendisk() function uninitializes the disk-specific data
425  * structures, including the bdi structure, without telling anyone
426  * else.  Once this happens, any attempt to call mark_buffer_dirty()
427  * (for example, by ext4_commit_super), will cause a kernel OOPS.
428  * This is a kludge to prevent these oops until we can put in a proper
429  * hook in del_gendisk() to inform the VFS and file system layers.
430  */
431 static int block_device_ejected(struct super_block *sb)
432 {
433         struct inode *bd_inode = sb->s_bdev->bd_inode;
434         struct backing_dev_info *bdi = inode_to_bdi(bd_inode);
435
436         return bdi->dev == NULL;
437 }
438
439 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
440 {
441         struct super_block              *sb = journal->j_private;
442         struct ext4_sb_info             *sbi = EXT4_SB(sb);
443         int                             error = is_journal_aborted(journal);
444         struct ext4_journal_cb_entry    *jce;
445
446         BUG_ON(txn->t_state == T_FINISHED);
447
448         ext4_process_freed_data(sb, txn->t_tid);
449
450         spin_lock(&sbi->s_md_lock);
451         while (!list_empty(&txn->t_private_list)) {
452                 jce = list_entry(txn->t_private_list.next,
453                                  struct ext4_journal_cb_entry, jce_list);
454                 list_del_init(&jce->jce_list);
455                 spin_unlock(&sbi->s_md_lock);
456                 jce->jce_func(sb, jce, error);
457                 spin_lock(&sbi->s_md_lock);
458         }
459         spin_unlock(&sbi->s_md_lock);
460 }
461
462 /*
463  * This writepage callback for write_cache_pages()
464  * takes care of a few cases after page cleaning.
465  *
466  * write_cache_pages() already checks for dirty pages
467  * and calls clear_page_dirty_for_io(), which we want,
468  * to write protect the pages.
469  *
470  * However, we may have to redirty a page (see below.)
471  */
472 static int ext4_journalled_writepage_callback(struct page *page,
473                                               struct writeback_control *wbc,
474                                               void *data)
475 {
476         transaction_t *transaction = (transaction_t *) data;
477         struct buffer_head *bh, *head;
478         struct journal_head *jh;
479
480         bh = head = page_buffers(page);
481         do {
482                 /*
483                  * We have to redirty a page in these cases:
484                  * 1) If buffer is dirty, it means the page was dirty because it
485                  * contains a buffer that needs checkpointing. So the dirty bit
486                  * needs to be preserved so that checkpointing writes the buffer
487                  * properly.
488                  * 2) If buffer is not part of the committing transaction
489                  * (we may have just accidentally come across this buffer because
490                  * inode range tracking is not exact) or if the currently running
491                  * transaction already contains this buffer as well, dirty bit
492                  * needs to be preserved so that the buffer gets writeprotected
493                  * properly on running transaction's commit.
494                  */
495                 jh = bh2jh(bh);
496                 if (buffer_dirty(bh) ||
497                     (jh && (jh->b_transaction != transaction ||
498                             jh->b_next_transaction))) {
499                         redirty_page_for_writepage(wbc, page);
500                         goto out;
501                 }
502         } while ((bh = bh->b_this_page) != head);
503
504 out:
505         return AOP_WRITEPAGE_ACTIVATE;
506 }
507
508 static int ext4_journalled_submit_inode_data_buffers(struct jbd2_inode *jinode)
509 {
510         struct address_space *mapping = jinode->i_vfs_inode->i_mapping;
511         struct writeback_control wbc = {
512                 .sync_mode =  WB_SYNC_ALL,
513                 .nr_to_write = LONG_MAX,
514                 .range_start = jinode->i_dirty_start,
515                 .range_end = jinode->i_dirty_end,
516         };
517
518         return write_cache_pages(mapping, &wbc,
519                                  ext4_journalled_writepage_callback,
520                                  jinode->i_transaction);
521 }
522
523 static int ext4_journal_submit_inode_data_buffers(struct jbd2_inode *jinode)
524 {
525         int ret;
526
527         if (ext4_should_journal_data(jinode->i_vfs_inode))
528                 ret = ext4_journalled_submit_inode_data_buffers(jinode);
529         else
530                 ret = jbd2_journal_submit_inode_data_buffers(jinode);
531
532         return ret;
533 }
534
535 static int ext4_journal_finish_inode_data_buffers(struct jbd2_inode *jinode)
536 {
537         int ret = 0;
538
539         if (!ext4_should_journal_data(jinode->i_vfs_inode))
540                 ret = jbd2_journal_finish_inode_data_buffers(jinode);
541
542         return ret;
543 }
544
545 static bool system_going_down(void)
546 {
547         return system_state == SYSTEM_HALT || system_state == SYSTEM_POWER_OFF
548                 || system_state == SYSTEM_RESTART;
549 }
550
551 struct ext4_err_translation {
552         int code;
553         int errno;
554 };
555
556 #define EXT4_ERR_TRANSLATE(err) { .code = EXT4_ERR_##err, .errno = err }
557
558 static struct ext4_err_translation err_translation[] = {
559         EXT4_ERR_TRANSLATE(EIO),
560         EXT4_ERR_TRANSLATE(ENOMEM),
561         EXT4_ERR_TRANSLATE(EFSBADCRC),
562         EXT4_ERR_TRANSLATE(EFSCORRUPTED),
563         EXT4_ERR_TRANSLATE(ENOSPC),
564         EXT4_ERR_TRANSLATE(ENOKEY),
565         EXT4_ERR_TRANSLATE(EROFS),
566         EXT4_ERR_TRANSLATE(EFBIG),
567         EXT4_ERR_TRANSLATE(EEXIST),
568         EXT4_ERR_TRANSLATE(ERANGE),
569         EXT4_ERR_TRANSLATE(EOVERFLOW),
570         EXT4_ERR_TRANSLATE(EBUSY),
571         EXT4_ERR_TRANSLATE(ENOTDIR),
572         EXT4_ERR_TRANSLATE(ENOTEMPTY),
573         EXT4_ERR_TRANSLATE(ESHUTDOWN),
574         EXT4_ERR_TRANSLATE(EFAULT),
575 };
576
577 static int ext4_errno_to_code(int errno)
578 {
579         int i;
580
581         for (i = 0; i < ARRAY_SIZE(err_translation); i++)
582                 if (err_translation[i].errno == errno)
583                         return err_translation[i].code;
584         return EXT4_ERR_UNKNOWN;
585 }
586
587 static void save_error_info(struct super_block *sb, int error,
588                             __u32 ino, __u64 block,
589                             const char *func, unsigned int line)
590 {
591         struct ext4_sb_info *sbi = EXT4_SB(sb);
592
593         /* We default to EFSCORRUPTED error... */
594         if (error == 0)
595                 error = EFSCORRUPTED;
596
597         spin_lock(&sbi->s_error_lock);
598         sbi->s_add_error_count++;
599         sbi->s_last_error_code = error;
600         sbi->s_last_error_line = line;
601         sbi->s_last_error_ino = ino;
602         sbi->s_last_error_block = block;
603         sbi->s_last_error_func = func;
604         sbi->s_last_error_time = ktime_get_real_seconds();
605         if (!sbi->s_first_error_time) {
606                 sbi->s_first_error_code = error;
607                 sbi->s_first_error_line = line;
608                 sbi->s_first_error_ino = ino;
609                 sbi->s_first_error_block = block;
610                 sbi->s_first_error_func = func;
611                 sbi->s_first_error_time = sbi->s_last_error_time;
612         }
613         spin_unlock(&sbi->s_error_lock);
614 }
615
616 /* Deal with the reporting of failure conditions on a filesystem such as
617  * inconsistencies detected or read IO failures.
618  *
619  * On ext2, we can store the error state of the filesystem in the
620  * superblock.  That is not possible on ext4, because we may have other
621  * write ordering constraints on the superblock which prevent us from
622  * writing it out straight away; and given that the journal is about to
623  * be aborted, we can't rely on the current, or future, transactions to
624  * write out the superblock safely.
625  *
626  * We'll just use the jbd2_journal_abort() error code to record an error in
627  * the journal instead.  On recovery, the journal will complain about
628  * that error until we've noted it down and cleared it.
629  *
630  * If force_ro is set, we unconditionally force the filesystem into an
631  * ABORT|READONLY state, unless the error response on the fs has been set to
632  * panic in which case we take the easy way out and panic immediately. This is
633  * used to deal with unrecoverable failures such as journal IO errors or ENOMEM
634  * at a critical moment in log management.
635  */
636 static void ext4_handle_error(struct super_block *sb, bool force_ro, int error,
637                               __u32 ino, __u64 block,
638                               const char *func, unsigned int line)
639 {
640         journal_t *journal = EXT4_SB(sb)->s_journal;
641         bool continue_fs = !force_ro && test_opt(sb, ERRORS_CONT);
642
643         EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
644         if (test_opt(sb, WARN_ON_ERROR))
645                 WARN_ON_ONCE(1);
646
647         if (!continue_fs && !sb_rdonly(sb)) {
648                 ext4_set_mount_flag(sb, EXT4_MF_FS_ABORTED);
649                 if (journal)
650                         jbd2_journal_abort(journal, -EIO);
651         }
652
653         if (!bdev_read_only(sb->s_bdev)) {
654                 save_error_info(sb, error, ino, block, func, line);
655                 /*
656                  * In case the fs should keep running, we need to writeout
657                  * superblock through the journal. Due to lock ordering
658                  * constraints, it may not be safe to do it right here so we
659                  * defer superblock flushing to a workqueue.
660                  */
661                 if (continue_fs)
662                         schedule_work(&EXT4_SB(sb)->s_error_work);
663                 else
664                         ext4_commit_super(sb);
665         }
666
667         /*
668          * We force ERRORS_RO behavior when system is rebooting. Otherwise we
669          * could panic during 'reboot -f' as the underlying device got already
670          * disabled.
671          */
672         if (test_opt(sb, ERRORS_PANIC) && !system_going_down()) {
673                 panic("EXT4-fs (device %s): panic forced after error\n",
674                         sb->s_id);
675         }
676
677         if (sb_rdonly(sb) || continue_fs)
678                 return;
679
680         ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
681         /*
682          * Make sure updated value of ->s_mount_flags will be visible before
683          * ->s_flags update
684          */
685         smp_wmb();
686         sb->s_flags |= SB_RDONLY;
687 }
688
689 static void flush_stashed_error_work(struct work_struct *work)
690 {
691         struct ext4_sb_info *sbi = container_of(work, struct ext4_sb_info,
692                                                 s_error_work);
693         journal_t *journal = sbi->s_journal;
694         handle_t *handle;
695
696         /*
697          * If the journal is still running, we have to write out superblock
698          * through the journal to avoid collisions of other journalled sb
699          * updates.
700          *
701          * We use directly jbd2 functions here to avoid recursing back into
702          * ext4 error handling code during handling of previous errors.
703          */
704         if (!sb_rdonly(sbi->s_sb) && journal) {
705                 struct buffer_head *sbh = sbi->s_sbh;
706                 handle = jbd2_journal_start(journal, 1);
707                 if (IS_ERR(handle))
708                         goto write_directly;
709                 if (jbd2_journal_get_write_access(handle, sbh)) {
710                         jbd2_journal_stop(handle);
711                         goto write_directly;
712                 }
713                 ext4_update_super(sbi->s_sb);
714                 if (buffer_write_io_error(sbh) || !buffer_uptodate(sbh)) {
715                         ext4_msg(sbi->s_sb, KERN_ERR, "previous I/O error to "
716                                  "superblock detected");
717                         clear_buffer_write_io_error(sbh);
718                         set_buffer_uptodate(sbh);
719                 }
720
721                 if (jbd2_journal_dirty_metadata(handle, sbh)) {
722                         jbd2_journal_stop(handle);
723                         goto write_directly;
724                 }
725                 jbd2_journal_stop(handle);
726                 ext4_notify_error_sysfs(sbi);
727                 return;
728         }
729 write_directly:
730         /*
731          * Write through journal failed. Write sb directly to get error info
732          * out and hope for the best.
733          */
734         ext4_commit_super(sbi->s_sb);
735         ext4_notify_error_sysfs(sbi);
736 }
737
738 #define ext4_error_ratelimit(sb)                                        \
739                 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state),     \
740                              "EXT4-fs error")
741
742 void __ext4_error(struct super_block *sb, const char *function,
743                   unsigned int line, bool force_ro, int error, __u64 block,
744                   const char *fmt, ...)
745 {
746         struct va_format vaf;
747         va_list args;
748
749         if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
750                 return;
751
752         trace_ext4_error(sb, function, line);
753         if (ext4_error_ratelimit(sb)) {
754                 va_start(args, fmt);
755                 vaf.fmt = fmt;
756                 vaf.va = &args;
757                 printk(KERN_CRIT
758                        "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
759                        sb->s_id, function, line, current->comm, &vaf);
760                 va_end(args);
761         }
762         ext4_handle_error(sb, force_ro, error, 0, block, function, line);
763 }
764
765 void __ext4_error_inode(struct inode *inode, const char *function,
766                         unsigned int line, ext4_fsblk_t block, int error,
767                         const char *fmt, ...)
768 {
769         va_list args;
770         struct va_format vaf;
771
772         if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
773                 return;
774
775         trace_ext4_error(inode->i_sb, function, line);
776         if (ext4_error_ratelimit(inode->i_sb)) {
777                 va_start(args, fmt);
778                 vaf.fmt = fmt;
779                 vaf.va = &args;
780                 if (block)
781                         printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
782                                "inode #%lu: block %llu: comm %s: %pV\n",
783                                inode->i_sb->s_id, function, line, inode->i_ino,
784                                block, current->comm, &vaf);
785                 else
786                         printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
787                                "inode #%lu: comm %s: %pV\n",
788                                inode->i_sb->s_id, function, line, inode->i_ino,
789                                current->comm, &vaf);
790                 va_end(args);
791         }
792         ext4_handle_error(inode->i_sb, false, error, inode->i_ino, block,
793                           function, line);
794 }
795
796 void __ext4_error_file(struct file *file, const char *function,
797                        unsigned int line, ext4_fsblk_t block,
798                        const char *fmt, ...)
799 {
800         va_list args;
801         struct va_format vaf;
802         struct inode *inode = file_inode(file);
803         char pathname[80], *path;
804
805         if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
806                 return;
807
808         trace_ext4_error(inode->i_sb, function, line);
809         if (ext4_error_ratelimit(inode->i_sb)) {
810                 path = file_path(file, pathname, sizeof(pathname));
811                 if (IS_ERR(path))
812                         path = "(unknown)";
813                 va_start(args, fmt);
814                 vaf.fmt = fmt;
815                 vaf.va = &args;
816                 if (block)
817                         printk(KERN_CRIT
818                                "EXT4-fs error (device %s): %s:%d: inode #%lu: "
819                                "block %llu: comm %s: path %s: %pV\n",
820                                inode->i_sb->s_id, function, line, inode->i_ino,
821                                block, current->comm, path, &vaf);
822                 else
823                         printk(KERN_CRIT
824                                "EXT4-fs error (device %s): %s:%d: inode #%lu: "
825                                "comm %s: path %s: %pV\n",
826                                inode->i_sb->s_id, function, line, inode->i_ino,
827                                current->comm, path, &vaf);
828                 va_end(args);
829         }
830         ext4_handle_error(inode->i_sb, false, EFSCORRUPTED, inode->i_ino, block,
831                           function, line);
832 }
833
834 const char *ext4_decode_error(struct super_block *sb, int errno,
835                               char nbuf[16])
836 {
837         char *errstr = NULL;
838
839         switch (errno) {
840         case -EFSCORRUPTED:
841                 errstr = "Corrupt filesystem";
842                 break;
843         case -EFSBADCRC:
844                 errstr = "Filesystem failed CRC";
845                 break;
846         case -EIO:
847                 errstr = "IO failure";
848                 break;
849         case -ENOMEM:
850                 errstr = "Out of memory";
851                 break;
852         case -EROFS:
853                 if (!sb || (EXT4_SB(sb)->s_journal &&
854                             EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
855                         errstr = "Journal has aborted";
856                 else
857                         errstr = "Readonly filesystem";
858                 break;
859         default:
860                 /* If the caller passed in an extra buffer for unknown
861                  * errors, textualise them now.  Else we just return
862                  * NULL. */
863                 if (nbuf) {
864                         /* Check for truncated error codes... */
865                         if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
866                                 errstr = nbuf;
867                 }
868                 break;
869         }
870
871         return errstr;
872 }
873
874 /* __ext4_std_error decodes expected errors from journaling functions
875  * automatically and invokes the appropriate error response.  */
876
877 void __ext4_std_error(struct super_block *sb, const char *function,
878                       unsigned int line, int errno)
879 {
880         char nbuf[16];
881         const char *errstr;
882
883         if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
884                 return;
885
886         /* Special case: if the error is EROFS, and we're not already
887          * inside a transaction, then there's really no point in logging
888          * an error. */
889         if (errno == -EROFS && journal_current_handle() == NULL && sb_rdonly(sb))
890                 return;
891
892         if (ext4_error_ratelimit(sb)) {
893                 errstr = ext4_decode_error(sb, errno, nbuf);
894                 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
895                        sb->s_id, function, line, errstr);
896         }
897
898         ext4_handle_error(sb, false, -errno, 0, 0, function, line);
899 }
900
901 void __ext4_msg(struct super_block *sb,
902                 const char *prefix, const char *fmt, ...)
903 {
904         struct va_format vaf;
905         va_list args;
906
907         atomic_inc(&EXT4_SB(sb)->s_msg_count);
908         if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
909                 return;
910
911         va_start(args, fmt);
912         vaf.fmt = fmt;
913         vaf.va = &args;
914         printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
915         va_end(args);
916 }
917
918 static int ext4_warning_ratelimit(struct super_block *sb)
919 {
920         atomic_inc(&EXT4_SB(sb)->s_warning_count);
921         return ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state),
922                             "EXT4-fs warning");
923 }
924
925 void __ext4_warning(struct super_block *sb, const char *function,
926                     unsigned int line, const char *fmt, ...)
927 {
928         struct va_format vaf;
929         va_list args;
930
931         if (!ext4_warning_ratelimit(sb))
932                 return;
933
934         va_start(args, fmt);
935         vaf.fmt = fmt;
936         vaf.va = &args;
937         printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
938                sb->s_id, function, line, &vaf);
939         va_end(args);
940 }
941
942 void __ext4_warning_inode(const struct inode *inode, const char *function,
943                           unsigned int line, const char *fmt, ...)
944 {
945         struct va_format vaf;
946         va_list args;
947
948         if (!ext4_warning_ratelimit(inode->i_sb))
949                 return;
950
951         va_start(args, fmt);
952         vaf.fmt = fmt;
953         vaf.va = &args;
954         printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
955                "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
956                function, line, inode->i_ino, current->comm, &vaf);
957         va_end(args);
958 }
959
960 void __ext4_grp_locked_error(const char *function, unsigned int line,
961                              struct super_block *sb, ext4_group_t grp,
962                              unsigned long ino, ext4_fsblk_t block,
963                              const char *fmt, ...)
964 __releases(bitlock)
965 __acquires(bitlock)
966 {
967         struct va_format vaf;
968         va_list args;
969
970         if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
971                 return;
972
973         trace_ext4_error(sb, function, line);
974         if (ext4_error_ratelimit(sb)) {
975                 va_start(args, fmt);
976                 vaf.fmt = fmt;
977                 vaf.va = &args;
978                 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
979                        sb->s_id, function, line, grp);
980                 if (ino)
981                         printk(KERN_CONT "inode %lu: ", ino);
982                 if (block)
983                         printk(KERN_CONT "block %llu:",
984                                (unsigned long long) block);
985                 printk(KERN_CONT "%pV\n", &vaf);
986                 va_end(args);
987         }
988
989         if (test_opt(sb, ERRORS_CONT)) {
990                 if (test_opt(sb, WARN_ON_ERROR))
991                         WARN_ON_ONCE(1);
992                 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
993                 if (!bdev_read_only(sb->s_bdev)) {
994                         save_error_info(sb, EFSCORRUPTED, ino, block, function,
995                                         line);
996                         schedule_work(&EXT4_SB(sb)->s_error_work);
997                 }
998                 return;
999         }
1000         ext4_unlock_group(sb, grp);
1001         ext4_handle_error(sb, false, EFSCORRUPTED, ino, block, function, line);
1002         /*
1003          * We only get here in the ERRORS_RO case; relocking the group
1004          * may be dangerous, but nothing bad will happen since the
1005          * filesystem will have already been marked read/only and the
1006          * journal has been aborted.  We return 1 as a hint to callers
1007          * who might what to use the return value from
1008          * ext4_grp_locked_error() to distinguish between the
1009          * ERRORS_CONT and ERRORS_RO case, and perhaps return more
1010          * aggressively from the ext4 function in question, with a
1011          * more appropriate error code.
1012          */
1013         ext4_lock_group(sb, grp);
1014         return;
1015 }
1016
1017 void ext4_mark_group_bitmap_corrupted(struct super_block *sb,
1018                                      ext4_group_t group,
1019                                      unsigned int flags)
1020 {
1021         struct ext4_sb_info *sbi = EXT4_SB(sb);
1022         struct ext4_group_info *grp = ext4_get_group_info(sb, group);
1023         struct ext4_group_desc *gdp = ext4_get_group_desc(sb, group, NULL);
1024         int ret;
1025
1026         if (flags & EXT4_GROUP_INFO_BBITMAP_CORRUPT) {
1027                 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_BBITMAP_CORRUPT_BIT,
1028                                             &grp->bb_state);
1029                 if (!ret)
1030                         percpu_counter_sub(&sbi->s_freeclusters_counter,
1031                                            grp->bb_free);
1032         }
1033
1034         if (flags & EXT4_GROUP_INFO_IBITMAP_CORRUPT) {
1035                 ret = ext4_test_and_set_bit(EXT4_GROUP_INFO_IBITMAP_CORRUPT_BIT,
1036                                             &grp->bb_state);
1037                 if (!ret && gdp) {
1038                         int count;
1039
1040                         count = ext4_free_inodes_count(sb, gdp);
1041                         percpu_counter_sub(&sbi->s_freeinodes_counter,
1042                                            count);
1043                 }
1044         }
1045 }
1046
1047 void ext4_update_dynamic_rev(struct super_block *sb)
1048 {
1049         struct ext4_super_block *es = EXT4_SB(sb)->s_es;
1050
1051         if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
1052                 return;
1053
1054         ext4_warning(sb,
1055                      "updating to rev %d because of new feature flag, "
1056                      "running e2fsck is recommended",
1057                      EXT4_DYNAMIC_REV);
1058
1059         es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
1060         es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
1061         es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
1062         /* leave es->s_feature_*compat flags alone */
1063         /* es->s_uuid will be set by e2fsck if empty */
1064
1065         /*
1066          * The rest of the superblock fields should be zero, and if not it
1067          * means they are likely already in use, so leave them alone.  We
1068          * can leave it up to e2fsck to clean up any inconsistencies there.
1069          */
1070 }
1071
1072 /*
1073  * Open the external journal device
1074  */
1075 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
1076 {
1077         struct block_device *bdev;
1078
1079         bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
1080         if (IS_ERR(bdev))
1081                 goto fail;
1082         return bdev;
1083
1084 fail:
1085         ext4_msg(sb, KERN_ERR,
1086                  "failed to open journal device unknown-block(%u,%u) %ld",
1087                  MAJOR(dev), MINOR(dev), PTR_ERR(bdev));
1088         return NULL;
1089 }
1090
1091 /*
1092  * Release the journal device
1093  */
1094 static void ext4_blkdev_put(struct block_device *bdev)
1095 {
1096         blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
1097 }
1098
1099 static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
1100 {
1101         struct block_device *bdev;
1102         bdev = sbi->s_journal_bdev;
1103         if (bdev) {
1104                 ext4_blkdev_put(bdev);
1105                 sbi->s_journal_bdev = NULL;
1106         }
1107 }
1108
1109 static inline struct inode *orphan_list_entry(struct list_head *l)
1110 {
1111         return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
1112 }
1113
1114 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
1115 {
1116         struct list_head *l;
1117
1118         ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
1119                  le32_to_cpu(sbi->s_es->s_last_orphan));
1120
1121         printk(KERN_ERR "sb_info orphan list:\n");
1122         list_for_each(l, &sbi->s_orphan) {
1123                 struct inode *inode = orphan_list_entry(l);
1124                 printk(KERN_ERR "  "
1125                        "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
1126                        inode->i_sb->s_id, inode->i_ino, inode,
1127                        inode->i_mode, inode->i_nlink,
1128                        NEXT_ORPHAN(inode));
1129         }
1130 }
1131
1132 #ifdef CONFIG_QUOTA
1133 static int ext4_quota_off(struct super_block *sb, int type);
1134
1135 static inline void ext4_quota_off_umount(struct super_block *sb)
1136 {
1137         int type;
1138
1139         /* Use our quota_off function to clear inode flags etc. */
1140         for (type = 0; type < EXT4_MAXQUOTAS; type++)
1141                 ext4_quota_off(sb, type);
1142 }
1143
1144 /*
1145  * This is a helper function which is used in the mount/remount
1146  * codepaths (which holds s_umount) to fetch the quota file name.
1147  */
1148 static inline char *get_qf_name(struct super_block *sb,
1149                                 struct ext4_sb_info *sbi,
1150                                 int type)
1151 {
1152         return rcu_dereference_protected(sbi->s_qf_names[type],
1153                                          lockdep_is_held(&sb->s_umount));
1154 }
1155 #else
1156 static inline void ext4_quota_off_umount(struct super_block *sb)
1157 {
1158 }
1159 #endif
1160
1161 static void ext4_put_super(struct super_block *sb)
1162 {
1163         struct ext4_sb_info *sbi = EXT4_SB(sb);
1164         struct ext4_super_block *es = sbi->s_es;
1165         struct buffer_head **group_desc;
1166         struct flex_groups **flex_groups;
1167         int aborted = 0;
1168         int i, err;
1169
1170         ext4_unregister_li_request(sb);
1171         ext4_quota_off_umount(sb);
1172
1173         flush_work(&sbi->s_error_work);
1174         destroy_workqueue(sbi->rsv_conversion_wq);
1175         ext4_release_orphan_info(sb);
1176
1177         /*
1178          * Unregister sysfs before destroying jbd2 journal.
1179          * Since we could still access attr_journal_task attribute via sysfs
1180          * path which could have sbi->s_journal->j_task as NULL
1181          */
1182         ext4_unregister_sysfs(sb);
1183
1184         if (sbi->s_journal) {
1185                 aborted = is_journal_aborted(sbi->s_journal);
1186                 err = jbd2_journal_destroy(sbi->s_journal);
1187                 sbi->s_journal = NULL;
1188                 if ((err < 0) && !aborted) {
1189                         ext4_abort(sb, -err, "Couldn't clean up the journal");
1190                 }
1191         }
1192
1193         ext4_es_unregister_shrinker(sbi);
1194         del_timer_sync(&sbi->s_err_report);
1195         ext4_release_system_zone(sb);
1196         ext4_mb_release(sb);
1197         ext4_ext_release(sb);
1198
1199         if (!sb_rdonly(sb) && !aborted) {
1200                 ext4_clear_feature_journal_needs_recovery(sb);
1201                 ext4_clear_feature_orphan_present(sb);
1202                 es->s_state = cpu_to_le16(sbi->s_mount_state);
1203         }
1204         if (!sb_rdonly(sb))
1205                 ext4_commit_super(sb);
1206
1207         rcu_read_lock();
1208         group_desc = rcu_dereference(sbi->s_group_desc);
1209         for (i = 0; i < sbi->s_gdb_count; i++)
1210                 brelse(group_desc[i]);
1211         kvfree(group_desc);
1212         flex_groups = rcu_dereference(sbi->s_flex_groups);
1213         if (flex_groups) {
1214                 for (i = 0; i < sbi->s_flex_groups_allocated; i++)
1215                         kvfree(flex_groups[i]);
1216                 kvfree(flex_groups);
1217         }
1218         rcu_read_unlock();
1219         percpu_counter_destroy(&sbi->s_freeclusters_counter);
1220         percpu_counter_destroy(&sbi->s_freeinodes_counter);
1221         percpu_counter_destroy(&sbi->s_dirs_counter);
1222         percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
1223         percpu_counter_destroy(&sbi->s_sra_exceeded_retry_limit);
1224         percpu_free_rwsem(&sbi->s_writepages_rwsem);
1225 #ifdef CONFIG_QUOTA
1226         for (i = 0; i < EXT4_MAXQUOTAS; i++)
1227                 kfree(get_qf_name(sb, sbi, i));
1228 #endif
1229
1230         /* Debugging code just in case the in-memory inode orphan list
1231          * isn't empty.  The on-disk one can be non-empty if we've
1232          * detected an error and taken the fs readonly, but the
1233          * in-memory list had better be clean by this point. */
1234         if (!list_empty(&sbi->s_orphan))
1235                 dump_orphan_list(sb, sbi);
1236         ASSERT(list_empty(&sbi->s_orphan));
1237
1238         sync_blockdev(sb->s_bdev);
1239         invalidate_bdev(sb->s_bdev);
1240         if (sbi->s_journal_bdev && sbi->s_journal_bdev != sb->s_bdev) {
1241                 /*
1242                  * Invalidate the journal device's buffers.  We don't want them
1243                  * floating about in memory - the physical journal device may
1244                  * hotswapped, and it breaks the `ro-after' testing code.
1245                  */
1246                 sync_blockdev(sbi->s_journal_bdev);
1247                 invalidate_bdev(sbi->s_journal_bdev);
1248                 ext4_blkdev_remove(sbi);
1249         }
1250
1251         ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
1252         sbi->s_ea_inode_cache = NULL;
1253
1254         ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
1255         sbi->s_ea_block_cache = NULL;
1256
1257         ext4_stop_mmpd(sbi);
1258
1259         brelse(sbi->s_sbh);
1260         sb->s_fs_info = NULL;
1261         /*
1262          * Now that we are completely done shutting down the
1263          * superblock, we need to actually destroy the kobject.
1264          */
1265         kobject_put(&sbi->s_kobj);
1266         wait_for_completion(&sbi->s_kobj_unregister);
1267         if (sbi->s_chksum_driver)
1268                 crypto_free_shash(sbi->s_chksum_driver);
1269         kfree(sbi->s_blockgroup_lock);
1270         fs_put_dax(sbi->s_daxdev);
1271         fscrypt_free_dummy_policy(&sbi->s_dummy_enc_policy);
1272 #ifdef CONFIG_UNICODE
1273         utf8_unload(sb->s_encoding);
1274 #endif
1275         kfree(sbi);
1276 }
1277
1278 static struct kmem_cache *ext4_inode_cachep;
1279
1280 /*
1281  * Called inside transaction, so use GFP_NOFS
1282  */
1283 static struct inode *ext4_alloc_inode(struct super_block *sb)
1284 {
1285         struct ext4_inode_info *ei;
1286
1287         ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
1288         if (!ei)
1289                 return NULL;
1290
1291         inode_set_iversion(&ei->vfs_inode, 1);
1292         spin_lock_init(&ei->i_raw_lock);
1293         INIT_LIST_HEAD(&ei->i_prealloc_list);
1294         atomic_set(&ei->i_prealloc_active, 0);
1295         spin_lock_init(&ei->i_prealloc_lock);
1296         ext4_es_init_tree(&ei->i_es_tree);
1297         rwlock_init(&ei->i_es_lock);
1298         INIT_LIST_HEAD(&ei->i_es_list);
1299         ei->i_es_all_nr = 0;
1300         ei->i_es_shk_nr = 0;
1301         ei->i_es_shrink_lblk = 0;
1302         ei->i_reserved_data_blocks = 0;
1303         spin_lock_init(&(ei->i_block_reservation_lock));
1304         ext4_init_pending_tree(&ei->i_pending_tree);
1305 #ifdef CONFIG_QUOTA
1306         ei->i_reserved_quota = 0;
1307         memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
1308 #endif
1309         ei->jinode = NULL;
1310         INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
1311         spin_lock_init(&ei->i_completed_io_lock);
1312         ei->i_sync_tid = 0;
1313         ei->i_datasync_tid = 0;
1314         atomic_set(&ei->i_unwritten, 0);
1315         INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
1316         ext4_fc_init_inode(&ei->vfs_inode);
1317         mutex_init(&ei->i_fc_lock);
1318         return &ei->vfs_inode;
1319 }
1320
1321 static int ext4_drop_inode(struct inode *inode)
1322 {
1323         int drop = generic_drop_inode(inode);
1324
1325         if (!drop)
1326                 drop = fscrypt_drop_inode(inode);
1327
1328         trace_ext4_drop_inode(inode, drop);
1329         return drop;
1330 }
1331
1332 static void ext4_free_in_core_inode(struct inode *inode)
1333 {
1334         fscrypt_free_inode(inode);
1335         if (!list_empty(&(EXT4_I(inode)->i_fc_list))) {
1336                 pr_warn("%s: inode %ld still in fc list",
1337                         __func__, inode->i_ino);
1338         }
1339         kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
1340 }
1341
1342 static void ext4_destroy_inode(struct inode *inode)
1343 {
1344         if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
1345                 ext4_msg(inode->i_sb, KERN_ERR,
1346                          "Inode %lu (%p): orphan list check failed!",
1347                          inode->i_ino, EXT4_I(inode));
1348                 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
1349                                 EXT4_I(inode), sizeof(struct ext4_inode_info),
1350                                 true);
1351                 dump_stack();
1352         }
1353 }
1354
1355 static void init_once(void *foo)
1356 {
1357         struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
1358
1359         INIT_LIST_HEAD(&ei->i_orphan);
1360         init_rwsem(&ei->xattr_sem);
1361         init_rwsem(&ei->i_data_sem);
1362         inode_init_once(&ei->vfs_inode);
1363         ext4_fc_init_inode(&ei->vfs_inode);
1364 }
1365
1366 static int __init init_inodecache(void)
1367 {
1368         ext4_inode_cachep = kmem_cache_create_usercopy("ext4_inode_cache",
1369                                 sizeof(struct ext4_inode_info), 0,
1370                                 (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD|
1371                                         SLAB_ACCOUNT),
1372                                 offsetof(struct ext4_inode_info, i_data),
1373                                 sizeof_field(struct ext4_inode_info, i_data),
1374                                 init_once);
1375         if (ext4_inode_cachep == NULL)
1376                 return -ENOMEM;
1377         return 0;
1378 }
1379
1380 static void destroy_inodecache(void)
1381 {
1382         /*
1383          * Make sure all delayed rcu free inodes are flushed before we
1384          * destroy cache.
1385          */
1386         rcu_barrier();
1387         kmem_cache_destroy(ext4_inode_cachep);
1388 }
1389
1390 void ext4_clear_inode(struct inode *inode)
1391 {
1392         ext4_fc_del(inode);
1393         invalidate_inode_buffers(inode);
1394         clear_inode(inode);
1395         ext4_discard_preallocations(inode, 0);
1396         ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
1397         dquot_drop(inode);
1398         if (EXT4_I(inode)->jinode) {
1399                 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1400                                                EXT4_I(inode)->jinode);
1401                 jbd2_free_inode(EXT4_I(inode)->jinode);
1402                 EXT4_I(inode)->jinode = NULL;
1403         }
1404         fscrypt_put_encryption_info(inode);
1405         fsverity_cleanup_inode(inode);
1406 }
1407
1408 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1409                                         u64 ino, u32 generation)
1410 {
1411         struct inode *inode;
1412
1413         /*
1414          * Currently we don't know the generation for parent directory, so
1415          * a generation of 0 means "accept any"
1416          */
1417         inode = ext4_iget(sb, ino, EXT4_IGET_HANDLE);
1418         if (IS_ERR(inode))
1419                 return ERR_CAST(inode);
1420         if (generation && inode->i_generation != generation) {
1421                 iput(inode);
1422                 return ERR_PTR(-ESTALE);
1423         }
1424
1425         return inode;
1426 }
1427
1428 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1429                                         int fh_len, int fh_type)
1430 {
1431         return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1432                                     ext4_nfs_get_inode);
1433 }
1434
1435 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1436                                         int fh_len, int fh_type)
1437 {
1438         return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1439                                     ext4_nfs_get_inode);
1440 }
1441
1442 static int ext4_nfs_commit_metadata(struct inode *inode)
1443 {
1444         struct writeback_control wbc = {
1445                 .sync_mode = WB_SYNC_ALL
1446         };
1447
1448         trace_ext4_nfs_commit_metadata(inode);
1449         return ext4_write_inode(inode, &wbc);
1450 }
1451
1452 #ifdef CONFIG_FS_ENCRYPTION
1453 static int ext4_get_context(struct inode *inode, void *ctx, size_t len)
1454 {
1455         return ext4_xattr_get(inode, EXT4_XATTR_INDEX_ENCRYPTION,
1456                                  EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx, len);
1457 }
1458
1459 static int ext4_set_context(struct inode *inode, const void *ctx, size_t len,
1460                                                         void *fs_data)
1461 {
1462         handle_t *handle = fs_data;
1463         int res, res2, credits, retries = 0;
1464
1465         /*
1466          * Encrypting the root directory is not allowed because e2fsck expects
1467          * lost+found to exist and be unencrypted, and encrypting the root
1468          * directory would imply encrypting the lost+found directory as well as
1469          * the filename "lost+found" itself.
1470          */
1471         if (inode->i_ino == EXT4_ROOT_INO)
1472                 return -EPERM;
1473
1474         if (WARN_ON_ONCE(IS_DAX(inode) && i_size_read(inode)))
1475                 return -EINVAL;
1476
1477         if (ext4_test_inode_flag(inode, EXT4_INODE_DAX))
1478                 return -EOPNOTSUPP;
1479
1480         res = ext4_convert_inline_data(inode);
1481         if (res)
1482                 return res;
1483
1484         /*
1485          * If a journal handle was specified, then the encryption context is
1486          * being set on a new inode via inheritance and is part of a larger
1487          * transaction to create the inode.  Otherwise the encryption context is
1488          * being set on an existing inode in its own transaction.  Only in the
1489          * latter case should the "retry on ENOSPC" logic be used.
1490          */
1491
1492         if (handle) {
1493                 res = ext4_xattr_set_handle(handle, inode,
1494                                             EXT4_XATTR_INDEX_ENCRYPTION,
1495                                             EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1496                                             ctx, len, 0);
1497                 if (!res) {
1498                         ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1499                         ext4_clear_inode_state(inode,
1500                                         EXT4_STATE_MAY_INLINE_DATA);
1501                         /*
1502                          * Update inode->i_flags - S_ENCRYPTED will be enabled,
1503                          * S_DAX may be disabled
1504                          */
1505                         ext4_set_inode_flags(inode, false);
1506                 }
1507                 return res;
1508         }
1509
1510         res = dquot_initialize(inode);
1511         if (res)
1512                 return res;
1513 retry:
1514         res = ext4_xattr_set_credits(inode, len, false /* is_create */,
1515                                      &credits);
1516         if (res)
1517                 return res;
1518
1519         handle = ext4_journal_start(inode, EXT4_HT_MISC, credits);
1520         if (IS_ERR(handle))
1521                 return PTR_ERR(handle);
1522
1523         res = ext4_xattr_set_handle(handle, inode, EXT4_XATTR_INDEX_ENCRYPTION,
1524                                     EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1525                                     ctx, len, 0);
1526         if (!res) {
1527                 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1528                 /*
1529                  * Update inode->i_flags - S_ENCRYPTED will be enabled,
1530                  * S_DAX may be disabled
1531                  */
1532                 ext4_set_inode_flags(inode, false);
1533                 res = ext4_mark_inode_dirty(handle, inode);
1534                 if (res)
1535                         EXT4_ERROR_INODE(inode, "Failed to mark inode dirty");
1536         }
1537         res2 = ext4_journal_stop(handle);
1538
1539         if (res == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries))
1540                 goto retry;
1541         if (!res)
1542                 res = res2;
1543         return res;
1544 }
1545
1546 static const union fscrypt_policy *ext4_get_dummy_policy(struct super_block *sb)
1547 {
1548         return EXT4_SB(sb)->s_dummy_enc_policy.policy;
1549 }
1550
1551 static bool ext4_has_stable_inodes(struct super_block *sb)
1552 {
1553         return ext4_has_feature_stable_inodes(sb);
1554 }
1555
1556 static void ext4_get_ino_and_lblk_bits(struct super_block *sb,
1557                                        int *ino_bits_ret, int *lblk_bits_ret)
1558 {
1559         *ino_bits_ret = 8 * sizeof(EXT4_SB(sb)->s_es->s_inodes_count);
1560         *lblk_bits_ret = 8 * sizeof(ext4_lblk_t);
1561 }
1562
1563 static const struct fscrypt_operations ext4_cryptops = {
1564         .key_prefix             = "ext4:",
1565         .get_context            = ext4_get_context,
1566         .set_context            = ext4_set_context,
1567         .get_dummy_policy       = ext4_get_dummy_policy,
1568         .empty_dir              = ext4_empty_dir,
1569         .max_namelen            = EXT4_NAME_LEN,
1570         .has_stable_inodes      = ext4_has_stable_inodes,
1571         .get_ino_and_lblk_bits  = ext4_get_ino_and_lblk_bits,
1572 };
1573 #endif
1574
1575 #ifdef CONFIG_QUOTA
1576 static const char * const quotatypes[] = INITQFNAMES;
1577 #define QTYPE2NAME(t) (quotatypes[t])
1578
1579 static int ext4_write_dquot(struct dquot *dquot);
1580 static int ext4_acquire_dquot(struct dquot *dquot);
1581 static int ext4_release_dquot(struct dquot *dquot);
1582 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1583 static int ext4_write_info(struct super_block *sb, int type);
1584 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1585                          const struct path *path);
1586 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1587                                size_t len, loff_t off);
1588 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1589                                 const char *data, size_t len, loff_t off);
1590 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1591                              unsigned int flags);
1592
1593 static struct dquot **ext4_get_dquots(struct inode *inode)
1594 {
1595         return EXT4_I(inode)->i_dquot;
1596 }
1597
1598 static const struct dquot_operations ext4_quota_operations = {
1599         .get_reserved_space     = ext4_get_reserved_space,
1600         .write_dquot            = ext4_write_dquot,
1601         .acquire_dquot          = ext4_acquire_dquot,
1602         .release_dquot          = ext4_release_dquot,
1603         .mark_dirty             = ext4_mark_dquot_dirty,
1604         .write_info             = ext4_write_info,
1605         .alloc_dquot            = dquot_alloc,
1606         .destroy_dquot          = dquot_destroy,
1607         .get_projid             = ext4_get_projid,
1608         .get_inode_usage        = ext4_get_inode_usage,
1609         .get_next_id            = dquot_get_next_id,
1610 };
1611
1612 static const struct quotactl_ops ext4_qctl_operations = {
1613         .quota_on       = ext4_quota_on,
1614         .quota_off      = ext4_quota_off,
1615         .quota_sync     = dquot_quota_sync,
1616         .get_state      = dquot_get_state,
1617         .set_info       = dquot_set_dqinfo,
1618         .get_dqblk      = dquot_get_dqblk,
1619         .set_dqblk      = dquot_set_dqblk,
1620         .get_nextdqblk  = dquot_get_next_dqblk,
1621 };
1622 #endif
1623
1624 static const struct super_operations ext4_sops = {
1625         .alloc_inode    = ext4_alloc_inode,
1626         .free_inode     = ext4_free_in_core_inode,
1627         .destroy_inode  = ext4_destroy_inode,
1628         .write_inode    = ext4_write_inode,
1629         .dirty_inode    = ext4_dirty_inode,
1630         .drop_inode     = ext4_drop_inode,
1631         .evict_inode    = ext4_evict_inode,
1632         .put_super      = ext4_put_super,
1633         .sync_fs        = ext4_sync_fs,
1634         .freeze_fs      = ext4_freeze,
1635         .unfreeze_fs    = ext4_unfreeze,
1636         .statfs         = ext4_statfs,
1637         .remount_fs     = ext4_remount,
1638         .show_options   = ext4_show_options,
1639 #ifdef CONFIG_QUOTA
1640         .quota_read     = ext4_quota_read,
1641         .quota_write    = ext4_quota_write,
1642         .get_dquots     = ext4_get_dquots,
1643 #endif
1644 };
1645
1646 static const struct export_operations ext4_export_ops = {
1647         .fh_to_dentry = ext4_fh_to_dentry,
1648         .fh_to_parent = ext4_fh_to_parent,
1649         .get_parent = ext4_get_parent,
1650         .commit_metadata = ext4_nfs_commit_metadata,
1651 };
1652
1653 enum {
1654         Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1655         Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1656         Opt_nouid32, Opt_debug, Opt_removed,
1657         Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1658         Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1659         Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1660         Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1661         Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1662         Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
1663         Opt_inlinecrypt,
1664         Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1665         Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1666         Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1667         Opt_usrquota, Opt_grpquota, Opt_prjquota, Opt_i_version,
1668         Opt_dax, Opt_dax_always, Opt_dax_inode, Opt_dax_never,
1669         Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_warn_on_error,
1670         Opt_nowarn_on_error, Opt_mblk_io_submit,
1671         Opt_lazytime, Opt_nolazytime, Opt_debug_want_extra_isize,
1672         Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1673         Opt_inode_readahead_blks, Opt_journal_ioprio,
1674         Opt_dioread_nolock, Opt_dioread_lock,
1675         Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1676         Opt_max_dir_size_kb, Opt_nojournal_checksum, Opt_nombcache,
1677         Opt_no_prefetch_block_bitmaps, Opt_mb_optimize_scan,
1678 #ifdef CONFIG_EXT4_DEBUG
1679         Opt_fc_debug_max_replay, Opt_fc_debug_force
1680 #endif
1681 };
1682
1683 static const match_table_t tokens = {
1684         {Opt_bsd_df, "bsddf"},
1685         {Opt_minix_df, "minixdf"},
1686         {Opt_grpid, "grpid"},
1687         {Opt_grpid, "bsdgroups"},
1688         {Opt_nogrpid, "nogrpid"},
1689         {Opt_nogrpid, "sysvgroups"},
1690         {Opt_resgid, "resgid=%u"},
1691         {Opt_resuid, "resuid=%u"},
1692         {Opt_sb, "sb=%u"},
1693         {Opt_err_cont, "errors=continue"},
1694         {Opt_err_panic, "errors=panic"},
1695         {Opt_err_ro, "errors=remount-ro"},
1696         {Opt_nouid32, "nouid32"},
1697         {Opt_debug, "debug"},
1698         {Opt_removed, "oldalloc"},
1699         {Opt_removed, "orlov"},
1700         {Opt_user_xattr, "user_xattr"},
1701         {Opt_nouser_xattr, "nouser_xattr"},
1702         {Opt_acl, "acl"},
1703         {Opt_noacl, "noacl"},
1704         {Opt_noload, "norecovery"},
1705         {Opt_noload, "noload"},
1706         {Opt_removed, "nobh"},
1707         {Opt_removed, "bh"},
1708         {Opt_commit, "commit=%u"},
1709         {Opt_min_batch_time, "min_batch_time=%u"},
1710         {Opt_max_batch_time, "max_batch_time=%u"},
1711         {Opt_journal_dev, "journal_dev=%u"},
1712         {Opt_journal_path, "journal_path=%s"},
1713         {Opt_journal_checksum, "journal_checksum"},
1714         {Opt_nojournal_checksum, "nojournal_checksum"},
1715         {Opt_journal_async_commit, "journal_async_commit"},
1716         {Opt_abort, "abort"},
1717         {Opt_data_journal, "data=journal"},
1718         {Opt_data_ordered, "data=ordered"},
1719         {Opt_data_writeback, "data=writeback"},
1720         {Opt_data_err_abort, "data_err=abort"},
1721         {Opt_data_err_ignore, "data_err=ignore"},
1722         {Opt_offusrjquota, "usrjquota="},
1723         {Opt_usrjquota, "usrjquota=%s"},
1724         {Opt_offgrpjquota, "grpjquota="},
1725         {Opt_grpjquota, "grpjquota=%s"},
1726         {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1727         {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1728         {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1729         {Opt_grpquota, "grpquota"},
1730         {Opt_noquota, "noquota"},
1731         {Opt_quota, "quota"},
1732         {Opt_usrquota, "usrquota"},
1733         {Opt_prjquota, "prjquota"},
1734         {Opt_barrier, "barrier=%u"},
1735         {Opt_barrier, "barrier"},
1736         {Opt_nobarrier, "nobarrier"},
1737         {Opt_i_version, "i_version"},
1738         {Opt_dax, "dax"},
1739         {Opt_dax_always, "dax=always"},
1740         {Opt_dax_inode, "dax=inode"},
1741         {Opt_dax_never, "dax=never"},
1742         {Opt_stripe, "stripe=%u"},
1743         {Opt_delalloc, "delalloc"},
1744         {Opt_warn_on_error, "warn_on_error"},
1745         {Opt_nowarn_on_error, "nowarn_on_error"},
1746         {Opt_lazytime, "lazytime"},
1747         {Opt_nolazytime, "nolazytime"},
1748         {Opt_debug_want_extra_isize, "debug_want_extra_isize=%u"},
1749         {Opt_nodelalloc, "nodelalloc"},
1750         {Opt_removed, "mblk_io_submit"},
1751         {Opt_removed, "nomblk_io_submit"},
1752         {Opt_block_validity, "block_validity"},
1753         {Opt_noblock_validity, "noblock_validity"},
1754         {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1755         {Opt_journal_ioprio, "journal_ioprio=%u"},
1756         {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1757         {Opt_auto_da_alloc, "auto_da_alloc"},
1758         {Opt_noauto_da_alloc, "noauto_da_alloc"},
1759         {Opt_dioread_nolock, "dioread_nolock"},
1760         {Opt_dioread_lock, "nodioread_nolock"},
1761         {Opt_dioread_lock, "dioread_lock"},
1762         {Opt_discard, "discard"},
1763         {Opt_nodiscard, "nodiscard"},
1764         {Opt_init_itable, "init_itable=%u"},
1765         {Opt_init_itable, "init_itable"},
1766         {Opt_noinit_itable, "noinit_itable"},
1767 #ifdef CONFIG_EXT4_DEBUG
1768         {Opt_fc_debug_force, "fc_debug_force"},
1769         {Opt_fc_debug_max_replay, "fc_debug_max_replay=%u"},
1770 #endif
1771         {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1772         {Opt_test_dummy_encryption, "test_dummy_encryption=%s"},
1773         {Opt_test_dummy_encryption, "test_dummy_encryption"},
1774         {Opt_inlinecrypt, "inlinecrypt"},
1775         {Opt_nombcache, "nombcache"},
1776         {Opt_nombcache, "no_mbcache"},  /* for backward compatibility */
1777         {Opt_removed, "prefetch_block_bitmaps"},
1778         {Opt_no_prefetch_block_bitmaps, "no_prefetch_block_bitmaps"},
1779         {Opt_mb_optimize_scan, "mb_optimize_scan=%d"},
1780         {Opt_removed, "check=none"},    /* mount option from ext2/3 */
1781         {Opt_removed, "nocheck"},       /* mount option from ext2/3 */
1782         {Opt_removed, "reservation"},   /* mount option from ext2/3 */
1783         {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1784         {Opt_removed, "journal=%u"},    /* mount option from ext2/3 */
1785         {Opt_err, NULL},
1786 };
1787
1788 static ext4_fsblk_t get_sb_block(void **data)
1789 {
1790         ext4_fsblk_t    sb_block;
1791         char            *options = (char *) *data;
1792
1793         if (!options || strncmp(options, "sb=", 3) != 0)
1794                 return 1;       /* Default location */
1795
1796         options += 3;
1797         /* TODO: use simple_strtoll with >32bit ext4 */
1798         sb_block = simple_strtoul(options, &options, 0);
1799         if (*options && *options != ',') {
1800                 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1801                        (char *) *data);
1802                 return 1;
1803         }
1804         if (*options == ',')
1805                 options++;
1806         *data = (void *) options;
1807
1808         return sb_block;
1809 }
1810
1811 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1812 #define DEFAULT_MB_OPTIMIZE_SCAN        (-1)
1813
1814 static const char deprecated_msg[] =
1815         "Mount option \"%s\" will be removed by %s\n"
1816         "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1817
1818 #ifdef CONFIG_QUOTA
1819 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1820 {
1821         struct ext4_sb_info *sbi = EXT4_SB(sb);
1822         char *qname, *old_qname = get_qf_name(sb, sbi, qtype);
1823         int ret = -1;
1824
1825         if (sb_any_quota_loaded(sb) && !old_qname) {
1826                 ext4_msg(sb, KERN_ERR,
1827                         "Cannot change journaled "
1828                         "quota options when quota turned on");
1829                 return -1;
1830         }
1831         if (ext4_has_feature_quota(sb)) {
1832                 ext4_msg(sb, KERN_INFO, "Journaled quota options "
1833                          "ignored when QUOTA feature is enabled");
1834                 return 1;
1835         }
1836         qname = match_strdup(args);
1837         if (!qname) {
1838                 ext4_msg(sb, KERN_ERR,
1839                         "Not enough memory for storing quotafile name");
1840                 return -1;
1841         }
1842         if (old_qname) {
1843                 if (strcmp(old_qname, qname) == 0)
1844                         ret = 1;
1845                 else
1846                         ext4_msg(sb, KERN_ERR,
1847                                  "%s quota file already specified",
1848                                  QTYPE2NAME(qtype));
1849                 goto errout;
1850         }
1851         if (strchr(qname, '/')) {
1852                 ext4_msg(sb, KERN_ERR,
1853                         "quotafile must be on filesystem root");
1854                 goto errout;
1855         }
1856         rcu_assign_pointer(sbi->s_qf_names[qtype], qname);
1857         set_opt(sb, QUOTA);
1858         return 1;
1859 errout:
1860         kfree(qname);
1861         return ret;
1862 }
1863
1864 static int clear_qf_name(struct super_block *sb, int qtype)
1865 {
1866
1867         struct ext4_sb_info *sbi = EXT4_SB(sb);
1868         char *old_qname = get_qf_name(sb, sbi, qtype);
1869
1870         if (sb_any_quota_loaded(sb) && old_qname) {
1871                 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1872                         " when quota turned on");
1873                 return -1;
1874         }
1875         rcu_assign_pointer(sbi->s_qf_names[qtype], NULL);
1876         synchronize_rcu();
1877         kfree(old_qname);
1878         return 1;
1879 }
1880 #endif
1881
1882 #define MOPT_SET        0x0001
1883 #define MOPT_CLEAR      0x0002
1884 #define MOPT_NOSUPPORT  0x0004
1885 #define MOPT_EXPLICIT   0x0008
1886 #define MOPT_CLEAR_ERR  0x0010
1887 #define MOPT_GTE0       0x0020
1888 #ifdef CONFIG_QUOTA
1889 #define MOPT_Q          0
1890 #define MOPT_QFMT       0x0040
1891 #else
1892 #define MOPT_Q          MOPT_NOSUPPORT
1893 #define MOPT_QFMT       MOPT_NOSUPPORT
1894 #endif
1895 #define MOPT_DATAJ      0x0080
1896 #define MOPT_NO_EXT2    0x0100
1897 #define MOPT_NO_EXT3    0x0200
1898 #define MOPT_EXT4_ONLY  (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1899 #define MOPT_STRING     0x0400
1900 #define MOPT_SKIP       0x0800
1901 #define MOPT_2          0x1000
1902
1903 static const struct mount_opts {
1904         int     token;
1905         int     mount_opt;
1906         int     flags;
1907 } ext4_mount_opts[] = {
1908         {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1909         {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1910         {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1911         {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1912         {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1913         {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1914         {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1915          MOPT_EXT4_ONLY | MOPT_SET},
1916         {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1917          MOPT_EXT4_ONLY | MOPT_CLEAR},
1918         {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1919         {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1920         {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1921          MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1922         {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1923          MOPT_EXT4_ONLY | MOPT_CLEAR},
1924         {Opt_warn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_SET},
1925         {Opt_nowarn_on_error, EXT4_MOUNT_WARN_ON_ERROR, MOPT_CLEAR},
1926         {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1927          MOPT_EXT4_ONLY | MOPT_CLEAR},
1928         {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1929          MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1930         {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1931                                     EXT4_MOUNT_JOURNAL_CHECKSUM),
1932          MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1933         {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1934         {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1935         {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1936         {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1937         {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1938          MOPT_NO_EXT2},
1939         {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1940          MOPT_NO_EXT2},
1941         {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1942         {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1943         {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1944         {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1945         {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1946         {Opt_commit, 0, MOPT_GTE0},
1947         {Opt_max_batch_time, 0, MOPT_GTE0},
1948         {Opt_min_batch_time, 0, MOPT_GTE0},
1949         {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1950         {Opt_init_itable, 0, MOPT_GTE0},
1951         {Opt_dax, EXT4_MOUNT_DAX_ALWAYS, MOPT_SET | MOPT_SKIP},
1952         {Opt_dax_always, EXT4_MOUNT_DAX_ALWAYS,
1953                 MOPT_EXT4_ONLY | MOPT_SET | MOPT_SKIP},
1954         {Opt_dax_inode, EXT4_MOUNT2_DAX_INODE,
1955                 MOPT_EXT4_ONLY | MOPT_SET | MOPT_SKIP},
1956         {Opt_dax_never, EXT4_MOUNT2_DAX_NEVER,
1957                 MOPT_EXT4_ONLY | MOPT_SET | MOPT_SKIP},
1958         {Opt_stripe, 0, MOPT_GTE0},
1959         {Opt_resuid, 0, MOPT_GTE0},
1960         {Opt_resgid, 0, MOPT_GTE0},
1961         {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1962         {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING},
1963         {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1964         {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1965         {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1966         {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1967          MOPT_NO_EXT2 | MOPT_DATAJ},
1968         {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1969         {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1970 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1971         {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1972         {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1973 #else
1974         {Opt_acl, 0, MOPT_NOSUPPORT},
1975         {Opt_noacl, 0, MOPT_NOSUPPORT},
1976 #endif
1977         {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1978         {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1979         {Opt_debug_want_extra_isize, 0, MOPT_GTE0},
1980         {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1981         {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1982                                                         MOPT_SET | MOPT_Q},
1983         {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1984                                                         MOPT_SET | MOPT_Q},
1985         {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA,
1986                                                         MOPT_SET | MOPT_Q},
1987         {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1988                        EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA),
1989                                                         MOPT_CLEAR | MOPT_Q},
1990         {Opt_usrjquota, 0, MOPT_Q | MOPT_STRING},
1991         {Opt_grpjquota, 0, MOPT_Q | MOPT_STRING},
1992         {Opt_offusrjquota, 0, MOPT_Q},
1993         {Opt_offgrpjquota, 0, MOPT_Q},
1994         {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1995         {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1996         {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1997         {Opt_max_dir_size_kb, 0, MOPT_GTE0},
1998         {Opt_test_dummy_encryption, 0, MOPT_STRING},
1999         {Opt_nombcache, EXT4_MOUNT_NO_MBCACHE, MOPT_SET},
2000         {Opt_no_prefetch_block_bitmaps, EXT4_MOUNT_NO_PREFETCH_BLOCK_BITMAPS,
2001          MOPT_SET},
2002         {Opt_mb_optimize_scan, EXT4_MOUNT2_MB_OPTIMIZE_SCAN, MOPT_GTE0},
2003 #ifdef CONFIG_EXT4_DEBUG
2004         {Opt_fc_debug_force, EXT4_MOUNT2_JOURNAL_FAST_COMMIT,
2005          MOPT_SET | MOPT_2 | MOPT_EXT4_ONLY},
2006         {Opt_fc_debug_max_replay, 0, MOPT_GTE0},
2007 #endif
2008         {Opt_err, 0, 0}
2009 };
2010
2011 #ifdef CONFIG_UNICODE
2012 static const struct ext4_sb_encodings {
2013         __u16 magic;
2014         char *name;
2015         char *version;
2016 } ext4_sb_encoding_map[] = {
2017         {EXT4_ENC_UTF8_12_1, "utf8", "12.1.0"},
2018 };
2019
2020 static int ext4_sb_read_encoding(const struct ext4_super_block *es,
2021                                  const struct ext4_sb_encodings **encoding,
2022                                  __u16 *flags)
2023 {
2024         __u16 magic = le16_to_cpu(es->s_encoding);
2025         int i;
2026
2027         for (i = 0; i < ARRAY_SIZE(ext4_sb_encoding_map); i++)
2028                 if (magic == ext4_sb_encoding_map[i].magic)
2029                         break;
2030
2031         if (i >= ARRAY_SIZE(ext4_sb_encoding_map))
2032                 return -EINVAL;
2033
2034         *encoding = &ext4_sb_encoding_map[i];
2035         *flags = le16_to_cpu(es->s_encoding_flags);
2036
2037         return 0;
2038 }
2039 #endif
2040
2041 static int ext4_set_test_dummy_encryption(struct super_block *sb,
2042                                           const char *opt,
2043                                           const substring_t *arg,
2044                                           bool is_remount)
2045 {
2046 #ifdef CONFIG_FS_ENCRYPTION
2047         struct ext4_sb_info *sbi = EXT4_SB(sb);
2048         int err;
2049
2050         /*
2051          * This mount option is just for testing, and it's not worthwhile to
2052          * implement the extra complexity (e.g. RCU protection) that would be
2053          * needed to allow it to be set or changed during remount.  We do allow
2054          * it to be specified during remount, but only if there is no change.
2055          */
2056         if (is_remount && !sbi->s_dummy_enc_policy.policy) {
2057                 ext4_msg(sb, KERN_WARNING,
2058                          "Can't set test_dummy_encryption on remount");
2059                 return -1;
2060         }
2061         err = fscrypt_set_test_dummy_encryption(sb, arg->from,
2062                                                 &sbi->s_dummy_enc_policy);
2063         if (err) {
2064                 if (err == -EEXIST)
2065                         ext4_msg(sb, KERN_WARNING,
2066                                  "Can't change test_dummy_encryption on remount");
2067                 else if (err == -EINVAL)
2068                         ext4_msg(sb, KERN_WARNING,
2069                                  "Value of option \"%s\" is unrecognized", opt);
2070                 else
2071                         ext4_msg(sb, KERN_WARNING,
2072                                  "Error processing option \"%s\" [%d]",
2073                                  opt, err);
2074                 return -1;
2075         }
2076         ext4_msg(sb, KERN_WARNING, "Test dummy encryption mode enabled");
2077 #else
2078         ext4_msg(sb, KERN_WARNING,
2079                  "Test dummy encryption mount option ignored");
2080 #endif
2081         return 1;
2082 }
2083
2084 struct ext4_parsed_options {
2085         unsigned long journal_devnum;
2086         unsigned int journal_ioprio;
2087         int mb_optimize_scan;
2088 };
2089
2090 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
2091                             substring_t *args, struct ext4_parsed_options *parsed_opts,
2092                             int is_remount)
2093 {
2094         struct ext4_sb_info *sbi = EXT4_SB(sb);
2095         const struct mount_opts *m;
2096         kuid_t uid;
2097         kgid_t gid;
2098         int arg = 0;
2099
2100 #ifdef CONFIG_QUOTA
2101         if (token == Opt_usrjquota)
2102                 return set_qf_name(sb, USRQUOTA, &args[0]);
2103         else if (token == Opt_grpjquota)
2104                 return set_qf_name(sb, GRPQUOTA, &args[0]);
2105         else if (token == Opt_offusrjquota)
2106                 return clear_qf_name(sb, USRQUOTA);
2107         else if (token == Opt_offgrpjquota)
2108                 return clear_qf_name(sb, GRPQUOTA);
2109 #endif
2110         switch (token) {
2111         case Opt_noacl:
2112         case Opt_nouser_xattr:
2113                 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
2114                 break;
2115         case Opt_sb:
2116                 return 1;       /* handled by get_sb_block() */
2117         case Opt_removed:
2118                 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
2119                 return 1;
2120         case Opt_abort:
2121                 ext4_set_mount_flag(sb, EXT4_MF_FS_ABORTED);
2122                 return 1;
2123         case Opt_i_version:
2124                 sb->s_flags |= SB_I_VERSION;
2125                 return 1;
2126         case Opt_lazytime:
2127                 sb->s_flags |= SB_LAZYTIME;
2128                 return 1;
2129         case Opt_nolazytime:
2130                 sb->s_flags &= ~SB_LAZYTIME;
2131                 return 1;
2132         case Opt_inlinecrypt:
2133 #ifdef CONFIG_FS_ENCRYPTION_INLINE_CRYPT
2134                 sb->s_flags |= SB_INLINECRYPT;
2135 #else
2136                 ext4_msg(sb, KERN_ERR, "inline encryption not supported");
2137 #endif
2138                 return 1;
2139         }
2140
2141         for (m = ext4_mount_opts; m->token != Opt_err; m++)
2142                 if (token == m->token)
2143                         break;
2144
2145         if (m->token == Opt_err) {
2146                 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
2147                          "or missing value", opt);
2148                 return -1;
2149         }
2150
2151         if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
2152                 ext4_msg(sb, KERN_ERR,
2153                          "Mount option \"%s\" incompatible with ext2", opt);
2154                 return -1;
2155         }
2156         if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
2157                 ext4_msg(sb, KERN_ERR,
2158                          "Mount option \"%s\" incompatible with ext3", opt);
2159                 return -1;
2160         }
2161
2162         if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
2163                 return -1;
2164         if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
2165                 return -1;
2166         if (m->flags & MOPT_EXPLICIT) {
2167                 if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
2168                         set_opt2(sb, EXPLICIT_DELALLOC);
2169                 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
2170                         set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM);
2171                 } else
2172                         return -1;
2173         }
2174         if (m->flags & MOPT_CLEAR_ERR)
2175                 clear_opt(sb, ERRORS_MASK);
2176         if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
2177                 ext4_msg(sb, KERN_ERR, "Cannot change quota "
2178                          "options when quota turned on");
2179                 return -1;
2180         }
2181
2182         if (m->flags & MOPT_NOSUPPORT) {
2183                 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
2184         } else if (token == Opt_commit) {
2185                 if (arg == 0)
2186                         arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
2187                 else if (arg > INT_MAX / HZ) {
2188                         ext4_msg(sb, KERN_ERR,
2189                                  "Invalid commit interval %d, "
2190                                  "must be smaller than %d",
2191                                  arg, INT_MAX / HZ);
2192                         return -1;
2193                 }
2194                 sbi->s_commit_interval = HZ * arg;
2195         } else if (token == Opt_debug_want_extra_isize) {
2196                 if ((arg & 1) ||
2197                     (arg < 4) ||
2198                     (arg > (sbi->s_inode_size - EXT4_GOOD_OLD_INODE_SIZE))) {
2199                         ext4_msg(sb, KERN_ERR,
2200                                  "Invalid want_extra_isize %d", arg);
2201                         return -1;
2202                 }
2203                 sbi->s_want_extra_isize = arg;
2204         } else if (token == Opt_max_batch_time) {
2205                 sbi->s_max_batch_time = arg;
2206         } else if (token == Opt_min_batch_time) {
2207                 sbi->s_min_batch_time = arg;
2208         } else if (token == Opt_inode_readahead_blks) {
2209                 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
2210                         ext4_msg(sb, KERN_ERR,
2211                                  "EXT4-fs: inode_readahead_blks must be "
2212                                  "0 or a power of 2 smaller than 2^31");
2213                         return -1;
2214                 }
2215                 sbi->s_inode_readahead_blks = arg;
2216         } else if (token == Opt_init_itable) {
2217                 set_opt(sb, INIT_INODE_TABLE);
2218                 if (!args->from)
2219                         arg = EXT4_DEF_LI_WAIT_MULT;
2220                 sbi->s_li_wait_mult = arg;
2221         } else if (token == Opt_max_dir_size_kb) {
2222                 sbi->s_max_dir_size_kb = arg;
2223 #ifdef CONFIG_EXT4_DEBUG
2224         } else if (token == Opt_fc_debug_max_replay) {
2225                 sbi->s_fc_debug_max_replay = arg;
2226 #endif
2227         } else if (token == Opt_stripe) {
2228                 sbi->s_stripe = arg;
2229         } else if (token == Opt_resuid) {
2230                 uid = make_kuid(current_user_ns(), arg);
2231                 if (!uid_valid(uid)) {
2232                         ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
2233                         return -1;
2234                 }
2235                 sbi->s_resuid = uid;
2236         } else if (token == Opt_resgid) {
2237                 gid = make_kgid(current_user_ns(), arg);
2238                 if (!gid_valid(gid)) {
2239                         ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
2240                         return -1;
2241                 }
2242                 sbi->s_resgid = gid;
2243         } else if (token == Opt_journal_dev) {
2244                 if (is_remount) {
2245                         ext4_msg(sb, KERN_ERR,
2246                                  "Cannot specify journal on remount");
2247                         return -1;
2248                 }
2249                 parsed_opts->journal_devnum = arg;
2250         } else if (token == Opt_journal_path) {
2251                 char *journal_path;
2252                 struct inode *journal_inode;
2253                 struct path path;
2254                 int error;
2255
2256                 if (is_remount) {
2257                         ext4_msg(sb, KERN_ERR,
2258                                  "Cannot specify journal on remount");
2259                         return -1;
2260                 }
2261                 journal_path = match_strdup(&args[0]);
2262                 if (!journal_path) {
2263                         ext4_msg(sb, KERN_ERR, "error: could not dup "
2264                                 "journal device string");
2265                         return -1;
2266                 }
2267
2268                 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
2269                 if (error) {
2270                         ext4_msg(sb, KERN_ERR, "error: could not find "
2271                                 "journal device path: error %d", error);
2272                         kfree(journal_path);
2273                         return -1;
2274                 }
2275
2276                 journal_inode = d_inode(path.dentry);
2277                 if (!S_ISBLK(journal_inode->i_mode)) {
2278                         ext4_msg(sb, KERN_ERR, "error: journal path %s "
2279                                 "is not a block device", journal_path);
2280                         path_put(&path);
2281                         kfree(journal_path);
2282                         return -1;
2283                 }
2284
2285                 parsed_opts->journal_devnum = new_encode_dev(journal_inode->i_rdev);
2286                 path_put(&path);
2287                 kfree(journal_path);
2288         } else if (token == Opt_journal_ioprio) {
2289                 if (arg > 7) {
2290                         ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
2291                                  " (must be 0-7)");
2292                         return -1;
2293                 }
2294                 parsed_opts->journal_ioprio =
2295                         IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
2296         } else if (token == Opt_test_dummy_encryption) {
2297                 return ext4_set_test_dummy_encryption(sb, opt, &args[0],
2298                                                       is_remount);
2299         } else if (m->flags & MOPT_DATAJ) {
2300                 if (is_remount) {
2301                         if (!sbi->s_journal)
2302                                 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
2303                         else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
2304                                 ext4_msg(sb, KERN_ERR,
2305                                          "Cannot change data mode on remount");
2306                                 return -1;
2307                         }
2308                 } else {
2309                         clear_opt(sb, DATA_FLAGS);
2310                         sbi->s_mount_opt |= m->mount_opt;
2311                 }
2312 #ifdef CONFIG_QUOTA
2313         } else if (m->flags & MOPT_QFMT) {
2314                 if (sb_any_quota_loaded(sb) &&
2315                     sbi->s_jquota_fmt != m->mount_opt) {
2316                         ext4_msg(sb, KERN_ERR, "Cannot change journaled "
2317                                  "quota options when quota turned on");
2318                         return -1;
2319                 }
2320                 if (ext4_has_feature_quota(sb)) {
2321                         ext4_msg(sb, KERN_INFO,
2322                                  "Quota format mount options ignored "
2323                                  "when QUOTA feature is enabled");
2324                         return 1;
2325                 }
2326                 sbi->s_jquota_fmt = m->mount_opt;
2327 #endif
2328         } else if (token == Opt_dax || token == Opt_dax_always ||
2329                    token == Opt_dax_inode || token == Opt_dax_never) {
2330 #ifdef CONFIG_FS_DAX
2331                 switch (token) {
2332                 case Opt_dax:
2333                 case Opt_dax_always:
2334                         if (is_remount &&
2335                             (!(sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) ||
2336                              (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER))) {
2337                         fail_dax_change_remount:
2338                                 ext4_msg(sb, KERN_ERR, "can't change "
2339                                          "dax mount option while remounting");
2340                                 return -1;
2341                         }
2342                         if (is_remount &&
2343                             (test_opt(sb, DATA_FLAGS) ==
2344                              EXT4_MOUNT_JOURNAL_DATA)) {
2345                                     ext4_msg(sb, KERN_ERR, "can't mount with "
2346                                              "both data=journal and dax");
2347                                     return -1;
2348                         }
2349                         ext4_msg(sb, KERN_WARNING,
2350                                 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
2351                         sbi->s_mount_opt |= EXT4_MOUNT_DAX_ALWAYS;
2352                         sbi->s_mount_opt2 &= ~EXT4_MOUNT2_DAX_NEVER;
2353                         break;
2354                 case Opt_dax_never:
2355                         if (is_remount &&
2356                             (!(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) ||
2357                              (sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS)))
2358                                 goto fail_dax_change_remount;
2359                         sbi->s_mount_opt2 |= EXT4_MOUNT2_DAX_NEVER;
2360                         sbi->s_mount_opt &= ~EXT4_MOUNT_DAX_ALWAYS;
2361                         break;
2362                 case Opt_dax_inode:
2363                         if (is_remount &&
2364                             ((sbi->s_mount_opt & EXT4_MOUNT_DAX_ALWAYS) ||
2365                              (sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_NEVER) ||
2366                              !(sbi->s_mount_opt2 & EXT4_MOUNT2_DAX_INODE)))
2367                                 goto fail_dax_change_remount;
2368                         sbi->s_mount_opt &= ~EXT4_MOUNT_DAX_ALWAYS;
2369                         sbi->s_mount_opt2 &= ~EXT4_MOUNT2_DAX_NEVER;
2370                         /* Strictly for printing options */
2371                         sbi->s_mount_opt2 |= EXT4_MOUNT2_DAX_INODE;
2372                         break;
2373                 }
2374 #else
2375                 ext4_msg(sb, KERN_INFO, "dax option not supported");
2376                 sbi->s_mount_opt2 |= EXT4_MOUNT2_DAX_NEVER;
2377                 sbi->s_mount_opt &= ~EXT4_MOUNT_DAX_ALWAYS;
2378                 return -1;
2379 #endif
2380         } else if (token == Opt_data_err_abort) {
2381                 sbi->s_mount_opt |= m->mount_opt;
2382         } else if (token == Opt_data_err_ignore) {
2383                 sbi->s_mount_opt &= ~m->mount_opt;
2384         } else if (token == Opt_mb_optimize_scan) {
2385                 if (arg != 0 && arg != 1) {
2386                         ext4_msg(sb, KERN_WARNING,
2387                                  "mb_optimize_scan should be set to 0 or 1.");
2388                         return -1;
2389                 }
2390                 parsed_opts->mb_optimize_scan = arg;
2391         } else {
2392                 if (!args->from)
2393                         arg = 1;
2394                 if (m->flags & MOPT_CLEAR)
2395                         arg = !arg;
2396                 else if (unlikely(!(m->flags & MOPT_SET))) {
2397                         ext4_msg(sb, KERN_WARNING,
2398                                  "buggy handling of option %s", opt);
2399                         WARN_ON(1);
2400                         return -1;
2401                 }
2402                 if (m->flags & MOPT_2) {
2403                         if (arg != 0)
2404                                 sbi->s_mount_opt2 |= m->mount_opt;
2405                         else
2406                                 sbi->s_mount_opt2 &= ~m->mount_opt;
2407                 } else {
2408                         if (arg != 0)
2409                                 sbi->s_mount_opt |= m->mount_opt;
2410                         else
2411                                 sbi->s_mount_opt &= ~m->mount_opt;
2412                 }
2413         }
2414         return 1;
2415 }
2416
2417 static int parse_options(char *options, struct super_block *sb,
2418                          struct ext4_parsed_options *ret_opts,
2419                          int is_remount)
2420 {
2421         struct ext4_sb_info __maybe_unused *sbi = EXT4_SB(sb);
2422         char *p, __maybe_unused *usr_qf_name, __maybe_unused *grp_qf_name;
2423         substring_t args[MAX_OPT_ARGS];
2424         int token;
2425
2426         if (!options)
2427                 return 1;
2428
2429         while ((p = strsep(&options, ",")) != NULL) {
2430                 if (!*p)
2431                         continue;
2432                 /*
2433                  * Initialize args struct so we know whether arg was
2434                  * found; some options take optional arguments.
2435                  */
2436                 args[0].to = args[0].from = NULL;
2437                 token = match_token(p, tokens, args);
2438                 if (handle_mount_opt(sb, p, token, args, ret_opts,
2439                                      is_remount) < 0)
2440                         return 0;
2441         }
2442 #ifdef CONFIG_QUOTA
2443         /*
2444          * We do the test below only for project quotas. 'usrquota' and
2445          * 'grpquota' mount options are allowed even without quota feature
2446          * to support legacy quotas in quota files.
2447          */
2448         if (test_opt(sb, PRJQUOTA) && !ext4_has_feature_project(sb)) {
2449                 ext4_msg(sb, KERN_ERR, "Project quota feature not enabled. "
2450                          "Cannot enable project quota enforcement.");
2451                 return 0;
2452         }
2453         usr_qf_name = get_qf_name(sb, sbi, USRQUOTA);
2454         grp_qf_name = get_qf_name(sb, sbi, GRPQUOTA);
2455         if (usr_qf_name || grp_qf_name) {
2456                 if (test_opt(sb, USRQUOTA) && usr_qf_name)
2457                         clear_opt(sb, USRQUOTA);
2458
2459                 if (test_opt(sb, GRPQUOTA) && grp_qf_name)
2460                         clear_opt(sb, GRPQUOTA);
2461
2462                 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
2463                         ext4_msg(sb, KERN_ERR, "old and new quota "
2464                                         "format mixing");
2465                         return 0;
2466                 }
2467
2468                 if (!sbi->s_jquota_fmt) {
2469                         ext4_msg(sb, KERN_ERR, "journaled quota format "
2470                                         "not specified");
2471                         return 0;
2472                 }
2473         }
2474 #endif
2475         if (test_opt(sb, DIOREAD_NOLOCK)) {
2476                 int blocksize =
2477                         BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
2478                 if (blocksize < PAGE_SIZE)
2479                         ext4_msg(sb, KERN_WARNING, "Warning: mounting with an "
2480                                  "experimental mount option 'dioread_nolock' "
2481                                  "for blocksize < PAGE_SIZE");
2482         }
2483         return 1;
2484 }
2485
2486 static inline void ext4_show_quota_options(struct seq_file *seq,
2487                                            struct super_block *sb)
2488 {
2489 #if defined(CONFIG_QUOTA)
2490         struct ext4_sb_info *sbi = EXT4_SB(sb);
2491         char *usr_qf_name, *grp_qf_name;
2492
2493         if (sbi->s_jquota_fmt) {
2494                 char *fmtname = "";
2495
2496                 switch (sbi->s_jquota_fmt) {
2497                 case QFMT_VFS_OLD:
2498                         fmtname = "vfsold";
2499                         break;
2500                 case QFMT_VFS_V0:
2501                         fmtname = "vfsv0";
2502                         break;
2503                 case QFMT_VFS_V1:
2504                         fmtname = "vfsv1";
2505                         break;
2506                 }
2507                 seq_printf(seq, ",jqfmt=%s", fmtname);
2508         }
2509
2510         rcu_read_lock();
2511         usr_qf_name = rcu_dereference(sbi->s_qf_names[USRQUOTA]);
2512         grp_qf_name = rcu_dereference(sbi->s_qf_names[GRPQUOTA]);
2513         if (usr_qf_name)
2514                 seq_show_option(seq, "usrjquota", usr_qf_name);
2515         if (grp_qf_name)
2516                 seq_show_option(seq, "grpjquota", grp_qf_name);
2517         rcu_read_unlock();
2518 #endif
2519 }
2520
2521 static const char *token2str(int token)
2522 {
2523         const struct match_token *t;
2524
2525         for (t = tokens; t->token != Opt_err; t++)
2526                 if (t->token == token && !strchr(t->pattern, '='))
2527                         break;
2528         return t->pattern;
2529 }
2530
2531 /*
2532  * Show an option if
2533  *  - it's set to a non-default value OR
2534  *  - if the per-sb default is different from the global default
2535  */
2536 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
2537                               int nodefs)
2538 {
2539         struct ext4_sb_info *sbi = EXT4_SB(sb);
2540         struct ext4_super_block *es = sbi->s_es;
2541         int def_errors, def_mount_opt = sbi->s_def_mount_opt;
2542         const struct mount_opts *m;
2543         char sep = nodefs ? '\n' : ',';
2544
2545 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
2546 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2547
2548         if (sbi->s_sb_block != 1)
2549                 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
2550
2551         for (m = ext4_mount_opts; m->token != Opt_err; m++) {
2552                 int want_set = m->flags & MOPT_SET;
2553                 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
2554                     (m->flags & MOPT_CLEAR_ERR) || m->flags & MOPT_SKIP)
2555                         continue;
2556                 if (!nodefs && !(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
2557                         continue; /* skip if same as the default */
2558                 if ((want_set &&
2559                      (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
2560                     (!want_set && (sbi->s_mount_opt & m->mount_opt)))
2561                         continue; /* select Opt_noFoo vs Opt_Foo */
2562                 SEQ_OPTS_PRINT("%s", token2str(m->token));
2563         }
2564
2565         if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
2566             le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
2567                 SEQ_OPTS_PRINT("resuid=%u",
2568                                 from_kuid_munged(&init_user_ns, sbi->s_resuid));
2569         if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
2570             le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
2571                 SEQ_OPTS_PRINT("resgid=%u",
2572                                 from_kgid_munged(&init_user_ns, sbi->s_resgid));
2573         def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
2574         if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
2575                 SEQ_OPTS_PUTS("errors=remount-ro");
2576         if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
2577                 SEQ_OPTS_PUTS("errors=continue");
2578         if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
2579                 SEQ_OPTS_PUTS("errors=panic");
2580         if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
2581                 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
2582         if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
2583                 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
2584         if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
2585                 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2586         if (sb->s_flags & SB_I_VERSION)
2587                 SEQ_OPTS_PUTS("i_version");
2588         if (nodefs || sbi->s_stripe)
2589                 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
2590         if (nodefs || EXT4_MOUNT_DATA_FLAGS &
2591                         (sbi->s_mount_opt ^ def_mount_opt)) {
2592                 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2593                         SEQ_OPTS_PUTS("data=journal");
2594                 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2595                         SEQ_OPTS_PUTS("data=ordered");
2596                 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
2597                         SEQ_OPTS_PUTS("data=writeback");
2598         }
2599         if (nodefs ||
2600             sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
2601                 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
2602                                sbi->s_inode_readahead_blks);
2603
2604         if (test_opt(sb, INIT_INODE_TABLE) && (nodefs ||
2605                        (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
2606                 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
2607         if (nodefs || sbi->s_max_dir_size_kb)
2608                 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
2609         if (test_opt(sb, DATA_ERR_ABORT))
2610                 SEQ_OPTS_PUTS("data_err=abort");
2611
2612         fscrypt_show_test_dummy_encryption(seq, sep, sb);
2613
2614         if (sb->s_flags & SB_INLINECRYPT)
2615                 SEQ_OPTS_PUTS("inlinecrypt");
2616
2617         if (test_opt(sb, DAX_ALWAYS)) {
2618                 if (IS_EXT2_SB(sb))
2619                         SEQ_OPTS_PUTS("dax");
2620                 else
2621                         SEQ_OPTS_PUTS("dax=always");
2622         } else if (test_opt2(sb, DAX_NEVER)) {
2623                 SEQ_OPTS_PUTS("dax=never");
2624         } else if (test_opt2(sb, DAX_INODE)) {
2625                 SEQ_OPTS_PUTS("dax=inode");
2626         }
2627         ext4_show_quota_options(seq, sb);
2628         return 0;
2629 }
2630
2631 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
2632 {
2633         return _ext4_show_options(seq, root->d_sb, 0);
2634 }
2635
2636 int ext4_seq_options_show(struct seq_file *seq, void *offset)
2637 {
2638         struct super_block *sb = seq->private;
2639         int rc;
2640
2641         seq_puts(seq, sb_rdonly(sb) ? "ro" : "rw");
2642         rc = _ext4_show_options(seq, sb, 1);
2643         seq_puts(seq, "\n");
2644         return rc;
2645 }
2646
2647 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
2648                             int read_only)
2649 {
2650         struct ext4_sb_info *sbi = EXT4_SB(sb);
2651         int err = 0;
2652
2653         if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
2654                 ext4_msg(sb, KERN_ERR, "revision level too high, "
2655                          "forcing read-only mode");
2656                 err = -EROFS;
2657                 goto done;
2658         }
2659         if (read_only)
2660                 goto done;
2661         if (!(sbi->s_mount_state & EXT4_VALID_FS))
2662                 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
2663                          "running e2fsck is recommended");
2664         else if (sbi->s_mount_state & EXT4_ERROR_FS)
2665                 ext4_msg(sb, KERN_WARNING,
2666                          "warning: mounting fs with errors, "
2667                          "running e2fsck is recommended");
2668         else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
2669                  le16_to_cpu(es->s_mnt_count) >=
2670                  (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
2671                 ext4_msg(sb, KERN_WARNING,
2672                          "warning: maximal mount count reached, "
2673                          "running e2fsck is recommended");
2674         else if (le32_to_cpu(es->s_checkinterval) &&
2675                  (ext4_get_tstamp(es, s_lastcheck) +
2676                   le32_to_cpu(es->s_checkinterval) <= ktime_get_real_seconds()))
2677                 ext4_msg(sb, KERN_WARNING,
2678                          "warning: checktime reached, "
2679                          "running e2fsck is recommended");
2680         if (!sbi->s_journal)
2681                 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
2682         if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
2683                 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
2684         le16_add_cpu(&es->s_mnt_count, 1);
2685         ext4_update_tstamp(es, s_mtime);
2686         if (sbi->s_journal) {
2687                 ext4_set_feature_journal_needs_recovery(sb);
2688                 if (ext4_has_feature_orphan_file(sb))
2689                         ext4_set_feature_orphan_present(sb);
2690         }
2691
2692         err = ext4_commit_super(sb);
2693 done:
2694         if (test_opt(sb, DEBUG))
2695                 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
2696                                 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
2697                         sb->s_blocksize,
2698                         sbi->s_groups_count,
2699                         EXT4_BLOCKS_PER_GROUP(sb),
2700                         EXT4_INODES_PER_GROUP(sb),
2701                         sbi->s_mount_opt, sbi->s_mount_opt2);
2702
2703         cleancache_init_fs(sb);
2704         return err;
2705 }
2706
2707 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
2708 {
2709         struct ext4_sb_info *sbi = EXT4_SB(sb);
2710         struct flex_groups **old_groups, **new_groups;
2711         int size, i, j;
2712
2713         if (!sbi->s_log_groups_per_flex)
2714                 return 0;
2715
2716         size = ext4_flex_group(sbi, ngroup - 1) + 1;
2717         if (size <= sbi->s_flex_groups_allocated)
2718                 return 0;
2719
2720         new_groups = kvzalloc(roundup_pow_of_two(size *
2721                               sizeof(*sbi->s_flex_groups)), GFP_KERNEL);
2722         if (!new_groups) {
2723                 ext4_msg(sb, KERN_ERR,
2724                          "not enough memory for %d flex group pointers", size);
2725                 return -ENOMEM;
2726         }
2727         for (i = sbi->s_flex_groups_allocated; i < size; i++) {
2728                 new_groups[i] = kvzalloc(roundup_pow_of_two(
2729                                          sizeof(struct flex_groups)),
2730                                          GFP_KERNEL);
2731                 if (!new_groups[i]) {
2732                         for (j = sbi->s_flex_groups_allocated; j < i; j++)
2733                                 kvfree(new_groups[j]);
2734                         kvfree(new_groups);
2735                         ext4_msg(sb, KERN_ERR,
2736                                  "not enough memory for %d flex groups", size);
2737                         return -ENOMEM;
2738                 }
2739         }
2740         rcu_read_lock();
2741         old_groups = rcu_dereference(sbi->s_flex_groups);
2742         if (old_groups)
2743                 memcpy(new_groups, old_groups,
2744                        (sbi->s_flex_groups_allocated *
2745                         sizeof(struct flex_groups *)));
2746         rcu_read_unlock();
2747         rcu_assign_pointer(sbi->s_flex_groups, new_groups);
2748         sbi->s_flex_groups_allocated = size;
2749         if (old_groups)
2750                 ext4_kvfree_array_rcu(old_groups);
2751         return 0;
2752 }
2753
2754 static int ext4_fill_flex_info(struct super_block *sb)
2755 {
2756         struct ext4_sb_info *sbi = EXT4_SB(sb);
2757         struct ext4_group_desc *gdp = NULL;
2758         struct flex_groups *fg;
2759         ext4_group_t flex_group;
2760         int i, err;
2761
2762         sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
2763         if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
2764                 sbi->s_log_groups_per_flex = 0;
2765                 return 1;
2766         }
2767
2768         err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
2769         if (err)
2770                 goto failed;
2771
2772         for (i = 0; i < sbi->s_groups_count; i++) {
2773                 gdp = ext4_get_group_desc(sb, i, NULL);
2774
2775                 flex_group = ext4_flex_group(sbi, i);
2776                 fg = sbi_array_rcu_deref(sbi, s_flex_groups, flex_group);
2777                 atomic_add(ext4_free_inodes_count(sb, gdp), &fg->free_inodes);
2778                 atomic64_add(ext4_free_group_clusters(sb, gdp),
2779                              &fg->free_clusters);
2780                 atomic_add(ext4_used_dirs_count(sb, gdp), &fg->used_dirs);
2781         }
2782
2783         return 1;
2784 failed:
2785         return 0;
2786 }
2787
2788 static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
2789                                    struct ext4_group_desc *gdp)
2790 {
2791         int offset = offsetof(struct ext4_group_desc, bg_checksum);
2792         __u16 crc = 0;
2793         __le32 le_group = cpu_to_le32(block_group);
2794         struct ext4_sb_info *sbi = EXT4_SB(sb);
2795
2796         if (ext4_has_metadata_csum(sbi->s_sb)) {
2797                 /* Use new metadata_csum algorithm */
2798                 __u32 csum32;
2799                 __u16 dummy_csum = 0;
2800
2801                 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2802                                      sizeof(le_group));
2803                 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset);
2804                 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum,
2805                                      sizeof(dummy_csum));
2806                 offset += sizeof(dummy_csum);
2807                 if (offset < sbi->s_desc_size)
2808                         csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset,
2809                                              sbi->s_desc_size - offset);
2810
2811                 crc = csum32 & 0xFFFF;
2812                 goto out;
2813         }
2814
2815         /* old crc16 code */
2816         if (!ext4_has_feature_gdt_csum(sb))
2817                 return 0;
2818
2819         crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2820         crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2821         crc = crc16(crc, (__u8 *)gdp, offset);
2822         offset += sizeof(gdp->bg_checksum); /* skip checksum */
2823         /* for checksum of struct ext4_group_desc do the rest...*/
2824         if (ext4_has_feature_64bit(sb) &&
2825             offset < le16_to_cpu(sbi->s_es->s_desc_size))
2826                 crc = crc16(crc, (__u8 *)gdp + offset,
2827                             le16_to_cpu(sbi->s_es->s_desc_size) -
2828                                 offset);
2829
2830 out:
2831         return cpu_to_le16(crc);
2832 }
2833
2834 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2835                                 struct ext4_group_desc *gdp)
2836 {
2837         if (ext4_has_group_desc_csum(sb) &&
2838             (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
2839                 return 0;
2840
2841         return 1;
2842 }
2843
2844 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2845                               struct ext4_group_desc *gdp)
2846 {
2847         if (!ext4_has_group_desc_csum(sb))
2848                 return;
2849         gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
2850 }
2851
2852 /* Called at mount-time, super-block is locked */
2853 static int ext4_check_descriptors(struct super_block *sb,
2854                                   ext4_fsblk_t sb_block,
2855                                   ext4_group_t *first_not_zeroed)
2856 {
2857         struct ext4_sb_info *sbi = EXT4_SB(sb);
2858         ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2859         ext4_fsblk_t last_block;
2860         ext4_fsblk_t last_bg_block = sb_block + ext4_bg_num_gdb(sb, 0);
2861         ext4_fsblk_t block_bitmap;
2862         ext4_fsblk_t inode_bitmap;
2863         ext4_fsblk_t inode_table;
2864         int flexbg_flag = 0;
2865         ext4_group_t i, grp = sbi->s_groups_count;
2866
2867         if (ext4_has_feature_flex_bg(sb))
2868                 flexbg_flag = 1;
2869
2870         ext4_debug("Checking group descriptors");
2871
2872         for (i = 0; i < sbi->s_groups_count; i++) {
2873                 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2874
2875                 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2876                         last_block = ext4_blocks_count(sbi->s_es) - 1;
2877                 else
2878                         last_block = first_block +
2879                                 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2880
2881                 if ((grp == sbi->s_groups_count) &&
2882                    !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2883                         grp = i;
2884
2885                 block_bitmap = ext4_block_bitmap(sb, gdp);
2886                 if (block_bitmap == sb_block) {
2887                         ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2888                                  "Block bitmap for group %u overlaps "
2889                                  "superblock", i);
2890                         if (!sb_rdonly(sb))
2891                                 return 0;
2892                 }
2893                 if (block_bitmap >= sb_block + 1 &&
2894                     block_bitmap <= last_bg_block) {
2895                         ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2896                                  "Block bitmap for group %u overlaps "
2897                                  "block group descriptors", i);
2898                         if (!sb_rdonly(sb))
2899                                 return 0;
2900                 }
2901                 if (block_bitmap < first_block || block_bitmap > last_block) {
2902                         ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2903                                "Block bitmap for group %u not in group "
2904                                "(block %llu)!", i, block_bitmap);
2905                         return 0;
2906                 }
2907                 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2908                 if (inode_bitmap == sb_block) {
2909                         ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2910                                  "Inode bitmap for group %u overlaps "
2911                                  "superblock", i);
2912                         if (!sb_rdonly(sb))
2913                                 return 0;
2914                 }
2915                 if (inode_bitmap >= sb_block + 1 &&
2916                     inode_bitmap <= last_bg_block) {
2917                         ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2918                                  "Inode bitmap for group %u overlaps "
2919                                  "block group descriptors", i);
2920                         if (!sb_rdonly(sb))
2921                                 return 0;
2922                 }
2923                 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2924                         ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2925                                "Inode bitmap for group %u not in group "
2926                                "(block %llu)!", i, inode_bitmap);
2927                         return 0;
2928                 }
2929                 inode_table = ext4_inode_table(sb, gdp);
2930                 if (inode_table == sb_block) {
2931                         ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2932                                  "Inode table for group %u overlaps "
2933                                  "superblock", i);
2934                         if (!sb_rdonly(sb))
2935                                 return 0;
2936                 }
2937                 if (inode_table >= sb_block + 1 &&
2938                     inode_table <= last_bg_block) {
2939                         ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2940                                  "Inode table for group %u overlaps "
2941                                  "block group descriptors", i);
2942                         if (!sb_rdonly(sb))
2943                                 return 0;
2944                 }
2945                 if (inode_table < first_block ||
2946                     inode_table + sbi->s_itb_per_group - 1 > last_block) {
2947                         ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2948                                "Inode table for group %u not in group "
2949                                "(block %llu)!", i, inode_table);
2950                         return 0;
2951                 }
2952                 ext4_lock_group(sb, i);
2953                 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2954                         ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2955                                  "Checksum for group %u failed (%u!=%u)",
2956                                  i, le16_to_cpu(ext4_group_desc_csum(sb, i,
2957                                      gdp)), le16_to_cpu(gdp->bg_checksum));
2958                         if (!sb_rdonly(sb)) {
2959                                 ext4_unlock_group(sb, i);
2960                                 return 0;
2961                         }
2962                 }
2963                 ext4_unlock_group(sb, i);
2964                 if (!flexbg_flag)
2965                         first_block += EXT4_BLOCKS_PER_GROUP(sb);
2966         }
2967         if (NULL != first_not_zeroed)
2968                 *first_not_zeroed = grp;
2969         return 1;
2970 }
2971
2972 /*
2973  * Maximal extent format file size.
2974  * Resulting logical blkno at s_maxbytes must fit in our on-disk
2975  * extent format containers, within a sector_t, and within i_blocks
2976  * in the vfs.  ext4 inode has 48 bits of i_block in fsblock units,
2977  * so that won't be a limiting factor.
2978  *
2979  * However there is other limiting factor. We do store extents in the form
2980  * of starting block and length, hence the resulting length of the extent
2981  * covering maximum file size must fit into on-disk format containers as
2982  * well. Given that length is always by 1 unit bigger than max unit (because
2983  * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2984  *
2985  * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2986  */
2987 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2988 {
2989         loff_t res;
2990         loff_t upper_limit = MAX_LFS_FILESIZE;
2991
2992         BUILD_BUG_ON(sizeof(blkcnt_t) < sizeof(u64));
2993
2994         if (!has_huge_files) {
2995                 upper_limit = (1LL << 32) - 1;
2996
2997                 /* total blocks in file system block size */
2998                 upper_limit >>= (blkbits - 9);
2999                 upper_limit <<= blkbits;
3000         }
3001
3002         /*
3003          * 32-bit extent-start container, ee_block. We lower the maxbytes
3004          * by one fs block, so ee_len can cover the extent of maximum file
3005          * size
3006          */
3007         res = (1LL << 32) - 1;
3008         res <<= blkbits;
3009
3010         /* Sanity check against vm- & vfs- imposed limits */
3011         if (res > upper_limit)
3012                 res = upper_limit;
3013
3014         return res;
3015 }
3016
3017 /*
3018  * Maximal bitmap file size.  There is a direct, and {,double-,triple-}indirect
3019  * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
3020  * We need to be 1 filesystem block less than the 2^48 sector limit.
3021  */
3022 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
3023 {
3024         loff_t res = EXT4_NDIR_BLOCKS;
3025         int meta_blocks;
3026         loff_t upper_limit;
3027         /* This is calculated to be the largest file size for a dense, block
3028          * mapped file such that the file's total number of 512-byte sectors,
3029          * including data and all indirect blocks, does not exceed (2^48 - 1).
3030          *
3031          * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
3032          * number of 512-byte sectors of the file.
3033          */
3034
3035         if (!has_huge_files) {
3036                 /*
3037                  * !has_huge_files or implies that the inode i_block field
3038                  * represents total file blocks in 2^32 512-byte sectors ==
3039                  * size of vfs inode i_blocks * 8
3040                  */
3041                 upper_limit = (1LL << 32) - 1;
3042
3043                 /* total blocks in file system block size */
3044                 upper_limit >>= (bits - 9);
3045
3046         } else {
3047                 /*
3048                  * We use 48 bit ext4_inode i_blocks
3049                  * With EXT4_HUGE_FILE_FL set the i_blocks
3050                  * represent total number of blocks in
3051                  * file system block size
3052                  */
3053                 upper_limit = (1LL << 48) - 1;
3054
3055         }
3056
3057         /* indirect blocks */
3058         meta_blocks = 1;
3059         /* double indirect blocks */
3060         meta_blocks += 1 + (1LL << (bits-2));
3061         /* tripple indirect blocks */
3062         meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
3063
3064         upper_limit -= meta_blocks;
3065         upper_limit <<= bits;
3066
3067         res += 1LL << (bits-2);
3068         res += 1LL << (2*(bits-2));
3069         res += 1LL << (3*(bits-2));
3070         res <<= bits;
3071         if (res > upper_limit)
3072                 res = upper_limit;
3073
3074         if (res > MAX_LFS_FILESIZE)
3075                 res = MAX_LFS_FILESIZE;
3076
3077         return res;
3078 }
3079
3080 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
3081                                    ext4_fsblk_t logical_sb_block, int nr)
3082 {
3083         struct ext4_sb_info *sbi = EXT4_SB(sb);
3084         ext4_group_t bg, first_meta_bg;
3085         int has_super = 0;
3086
3087         first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
3088
3089         if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
3090                 return logical_sb_block + nr + 1;
3091         bg = sbi->s_desc_per_block * nr;
3092         if (ext4_bg_has_super(sb, bg))
3093                 has_super = 1;
3094
3095         /*
3096          * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
3097          * block 2, not 1.  If s_first_data_block == 0 (bigalloc is enabled
3098          * on modern mke2fs or blksize > 1k on older mke2fs) then we must
3099          * compensate.
3100          */
3101         if (sb->s_blocksize == 1024 && nr == 0 &&
3102             le32_to_cpu(sbi->s_es->s_first_data_block) == 0)
3103                 has_super++;
3104
3105         return (has_super + ext4_group_first_block_no(sb, bg));
3106 }
3107
3108 /**
3109  * ext4_get_stripe_size: Get the stripe size.
3110  * @sbi: In memory super block info
3111  *
3112  * If we have specified it via mount option, then
3113  * use the mount option value. If the value specified at mount time is
3114  * greater than the blocks per group use the super block value.
3115  * If the super block value is greater than blocks per group return 0.
3116  * Allocator needs it be less than blocks per group.
3117  *
3118  */
3119 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
3120 {
3121         unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
3122         unsigned long stripe_width =
3123                         le32_to_cpu(sbi->s_es->s_raid_stripe_width);
3124         int ret;
3125
3126         if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
3127                 ret = sbi->s_stripe;
3128         else if (stripe_width && stripe_width <= sbi->s_blocks_per_group)
3129                 ret = stripe_width;
3130         else if (stride && stride <= sbi->s_blocks_per_group)
3131                 ret = stride;
3132         else
3133                 ret = 0;
3134
3135         /*
3136          * If the stripe width is 1, this makes no sense and
3137          * we set it to 0 to turn off stripe handling code.
3138          */
3139         if (ret <= 1)
3140                 ret = 0;
3141
3142         return ret;
3143 }
3144
3145 /*
3146  * Check whether this filesystem can be mounted based on
3147  * the features present and the RDONLY/RDWR mount requested.
3148  * Returns 1 if this filesystem can be mounted as requested,
3149  * 0 if it cannot be.
3150  */
3151 int ext4_feature_set_ok(struct super_block *sb, int readonly)
3152 {
3153         if (ext4_has_unknown_ext4_incompat_features(sb)) {
3154                 ext4_msg(sb, KERN_ERR,
3155                         "Couldn't mount because of "
3156                         "unsupported optional features (%x)",
3157                         (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
3158                         ~EXT4_FEATURE_INCOMPAT_SUPP));
3159                 return 0;
3160         }
3161
3162 #ifndef CONFIG_UNICODE
3163         if (ext4_has_feature_casefold(sb)) {
3164                 ext4_msg(sb, KERN_ERR,
3165                          "Filesystem with casefold feature cannot be "
3166                          "mounted without CONFIG_UNICODE");
3167                 return 0;
3168         }
3169 #endif
3170
3171         if (readonly)
3172                 return 1;
3173
3174         if (ext4_has_feature_readonly(sb)) {
3175                 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
3176                 sb->s_flags |= SB_RDONLY;
3177                 return 1;
3178         }
3179
3180         /* Check that feature set is OK for a read-write mount */
3181         if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
3182                 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
3183                          "unsupported optional features (%x)",
3184                          (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
3185                                 ~EXT4_FEATURE_RO_COMPAT_SUPP));
3186                 return 0;
3187         }
3188         if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
3189                 ext4_msg(sb, KERN_ERR,
3190                          "Can't support bigalloc feature without "
3191                          "extents feature\n");
3192                 return 0;
3193         }
3194
3195 #if !IS_ENABLED(CONFIG_QUOTA) || !IS_ENABLED(CONFIG_QFMT_V2)
3196         if (!readonly && (ext4_has_feature_quota(sb) ||
3197                           ext4_has_feature_project(sb))) {
3198                 ext4_msg(sb, KERN_ERR,
3199                          "The kernel was not built with CONFIG_QUOTA and CONFIG_QFMT_V2");
3200                 return 0;
3201         }
3202 #endif  /* CONFIG_QUOTA */
3203         return 1;
3204 }
3205
3206 /*
3207  * This function is called once a day if we have errors logged
3208  * on the file system
3209  */
3210 static void print_daily_error_info(struct timer_list *t)
3211 {
3212         struct ext4_sb_info *sbi = from_timer(sbi, t, s_err_report);
3213         struct super_block *sb = sbi->s_sb;
3214         struct ext4_super_block *es = sbi->s_es;
3215
3216         if (es->s_error_count)
3217                 /* fsck newer than v1.41.13 is needed to clean this condition. */
3218                 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
3219                          le32_to_cpu(es->s_error_count));
3220         if (es->s_first_error_time) {
3221                 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %llu: %.*s:%d",
3222                        sb->s_id,
3223                        ext4_get_tstamp(es, s_first_error_time),
3224                        (int) sizeof(es->s_first_error_func),
3225                        es->s_first_error_func,
3226                        le32_to_cpu(es->s_first_error_line));
3227                 if (es->s_first_error_ino)
3228                         printk(KERN_CONT ": inode %u",
3229                                le32_to_cpu(es->s_first_error_ino));
3230                 if (es->s_first_error_block)
3231                         printk(KERN_CONT ": block %llu", (unsigned long long)
3232                                le64_to_cpu(es->s_first_error_block));
3233                 printk(KERN_CONT "\n");
3234         }
3235         if (es->s_last_error_time) {
3236                 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %llu: %.*s:%d",
3237                        sb->s_id,
3238                        ext4_get_tstamp(es, s_last_error_time),
3239                        (int) sizeof(es->s_last_error_func),
3240                        es->s_last_error_func,
3241                        le32_to_cpu(es->s_last_error_line));
3242                 if (es->s_last_error_ino)
3243                         printk(KERN_CONT ": inode %u",
3244                                le32_to_cpu(es->s_last_error_ino));
3245                 if (es->s_last_error_block)
3246                         printk(KERN_CONT ": block %llu", (unsigned long long)
3247                                le64_to_cpu(es->s_last_error_block));
3248                 printk(KERN_CONT "\n");
3249         }
3250         mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ);  /* Once a day */
3251 }
3252
3253 /* Find next suitable group and run ext4_init_inode_table */
3254 static int ext4_run_li_request(struct ext4_li_request *elr)
3255 {
3256         struct ext4_group_desc *gdp = NULL;
3257         struct super_block *sb = elr->lr_super;
3258         ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3259         ext4_group_t group = elr->lr_next_group;
3260         unsigned long timeout = 0;
3261         unsigned int prefetch_ios = 0;
3262         int ret = 0;
3263
3264         if (elr->lr_mode == EXT4_LI_MODE_PREFETCH_BBITMAP) {
3265                 elr->lr_next_group = ext4_mb_prefetch(sb, group,
3266                                 EXT4_SB(sb)->s_mb_prefetch, &prefetch_ios);
3267                 if (prefetch_ios)
3268                         ext4_mb_prefetch_fini(sb, elr->lr_next_group,
3269                                               prefetch_ios);
3270                 trace_ext4_prefetch_bitmaps(sb, group, elr->lr_next_group,
3271                                             prefetch_ios);
3272                 if (group >= elr->lr_next_group) {
3273                         ret = 1;
3274                         if (elr->lr_first_not_zeroed != ngroups &&
3275                             !sb_rdonly(sb) && test_opt(sb, INIT_INODE_TABLE)) {
3276                                 elr->lr_next_group = elr->lr_first_not_zeroed;
3277                                 elr->lr_mode = EXT4_LI_MODE_ITABLE;
3278                                 ret = 0;
3279                         }