ext4: add MODULE_SOFTDEP to ensure crc32c is included in the initramfs
[sfrench/cifs-2.6.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
46 #include <linux/kthread.h>
47 #include <linux/freezer.h>
48
49 #include "ext4.h"
50 #include "ext4_extents.h"       /* Needed for trace points definition */
51 #include "ext4_jbd2.h"
52 #include "xattr.h"
53 #include "acl.h"
54 #include "mballoc.h"
55 #include "fsmap.h"
56
57 #define CREATE_TRACE_POINTS
58 #include <trace/events/ext4.h>
59
60 static struct ext4_lazy_init *ext4_li_info;
61 static struct mutex ext4_li_mtx;
62 static struct ratelimit_state ext4_mount_msg_ratelimit;
63
64 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
65                              unsigned long journal_devnum);
66 static int ext4_show_options(struct seq_file *seq, struct dentry *root);
67 static int ext4_commit_super(struct super_block *sb, int sync);
68 static void ext4_mark_recovery_complete(struct super_block *sb,
69                                         struct ext4_super_block *es);
70 static void ext4_clear_journal_err(struct super_block *sb,
71                                    struct ext4_super_block *es);
72 static int ext4_sync_fs(struct super_block *sb, int wait);
73 static int ext4_remount(struct super_block *sb, int *flags, char *data);
74 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
75 static int ext4_unfreeze(struct super_block *sb);
76 static int ext4_freeze(struct super_block *sb);
77 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
78                        const char *dev_name, void *data);
79 static inline int ext2_feature_set_ok(struct super_block *sb);
80 static inline int ext3_feature_set_ok(struct super_block *sb);
81 static int ext4_feature_set_ok(struct super_block *sb, int readonly);
82 static void ext4_destroy_lazyinit_thread(void);
83 static void ext4_unregister_li_request(struct super_block *sb);
84 static void ext4_clear_request_list(void);
85 static struct inode *ext4_get_journal_inode(struct super_block *sb,
86                                             unsigned int journal_inum);
87
88 /*
89  * Lock ordering
90  *
91  * Note the difference between i_mmap_sem (EXT4_I(inode)->i_mmap_sem) and
92  * i_mmap_rwsem (inode->i_mmap_rwsem)!
93  *
94  * page fault path:
95  * mmap_sem -> sb_start_pagefault -> i_mmap_sem (r) -> transaction start ->
96  *   page lock -> i_data_sem (rw)
97  *
98  * buffered write path:
99  * sb_start_write -> i_mutex -> mmap_sem
100  * sb_start_write -> i_mutex -> transaction start -> page lock ->
101  *   i_data_sem (rw)
102  *
103  * truncate:
104  * sb_start_write -> i_mutex -> i_mmap_sem (w) -> i_mmap_rwsem (w) -> page lock
105  * sb_start_write -> i_mutex -> i_mmap_sem (w) -> transaction start ->
106  *   i_data_sem (rw)
107  *
108  * direct IO:
109  * sb_start_write -> i_mutex -> mmap_sem
110  * sb_start_write -> i_mutex -> transaction start -> i_data_sem (rw)
111  *
112  * writepages:
113  * transaction start -> page lock(s) -> i_data_sem (rw)
114  */
115
116 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
117 static struct file_system_type ext2_fs_type = {
118         .owner          = THIS_MODULE,
119         .name           = "ext2",
120         .mount          = ext4_mount,
121         .kill_sb        = kill_block_super,
122         .fs_flags       = FS_REQUIRES_DEV,
123 };
124 MODULE_ALIAS_FS("ext2");
125 MODULE_ALIAS("ext2");
126 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
127 #else
128 #define IS_EXT2_SB(sb) (0)
129 #endif
130
131
132 static struct file_system_type ext3_fs_type = {
133         .owner          = THIS_MODULE,
134         .name           = "ext3",
135         .mount          = ext4_mount,
136         .kill_sb        = kill_block_super,
137         .fs_flags       = FS_REQUIRES_DEV,
138 };
139 MODULE_ALIAS_FS("ext3");
140 MODULE_ALIAS("ext3");
141 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
142
143 static int ext4_verify_csum_type(struct super_block *sb,
144                                  struct ext4_super_block *es)
145 {
146         if (!ext4_has_feature_metadata_csum(sb))
147                 return 1;
148
149         return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
150 }
151
152 static __le32 ext4_superblock_csum(struct super_block *sb,
153                                    struct ext4_super_block *es)
154 {
155         struct ext4_sb_info *sbi = EXT4_SB(sb);
156         int offset = offsetof(struct ext4_super_block, s_checksum);
157         __u32 csum;
158
159         csum = ext4_chksum(sbi, ~0, (char *)es, offset);
160
161         return cpu_to_le32(csum);
162 }
163
164 static int ext4_superblock_csum_verify(struct super_block *sb,
165                                        struct ext4_super_block *es)
166 {
167         if (!ext4_has_metadata_csum(sb))
168                 return 1;
169
170         return es->s_checksum == ext4_superblock_csum(sb, es);
171 }
172
173 void ext4_superblock_csum_set(struct super_block *sb)
174 {
175         struct ext4_super_block *es = EXT4_SB(sb)->s_es;
176
177         if (!ext4_has_metadata_csum(sb))
178                 return;
179
180         es->s_checksum = ext4_superblock_csum(sb, es);
181 }
182
183 void *ext4_kvmalloc(size_t size, gfp_t flags)
184 {
185         void *ret;
186
187         ret = kmalloc(size, flags | __GFP_NOWARN);
188         if (!ret)
189                 ret = __vmalloc(size, flags, PAGE_KERNEL);
190         return ret;
191 }
192
193 void *ext4_kvzalloc(size_t size, gfp_t flags)
194 {
195         void *ret;
196
197         ret = kzalloc(size, flags | __GFP_NOWARN);
198         if (!ret)
199                 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
200         return ret;
201 }
202
203 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
204                                struct ext4_group_desc *bg)
205 {
206         return le32_to_cpu(bg->bg_block_bitmap_lo) |
207                 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
208                  (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
209 }
210
211 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
212                                struct ext4_group_desc *bg)
213 {
214         return le32_to_cpu(bg->bg_inode_bitmap_lo) |
215                 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
216                  (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
217 }
218
219 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
220                               struct ext4_group_desc *bg)
221 {
222         return le32_to_cpu(bg->bg_inode_table_lo) |
223                 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
224                  (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
225 }
226
227 __u32 ext4_free_group_clusters(struct super_block *sb,
228                                struct ext4_group_desc *bg)
229 {
230         return le16_to_cpu(bg->bg_free_blocks_count_lo) |
231                 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
232                  (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
233 }
234
235 __u32 ext4_free_inodes_count(struct super_block *sb,
236                               struct ext4_group_desc *bg)
237 {
238         return le16_to_cpu(bg->bg_free_inodes_count_lo) |
239                 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
240                  (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
241 }
242
243 __u32 ext4_used_dirs_count(struct super_block *sb,
244                               struct ext4_group_desc *bg)
245 {
246         return le16_to_cpu(bg->bg_used_dirs_count_lo) |
247                 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
248                  (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
249 }
250
251 __u32 ext4_itable_unused_count(struct super_block *sb,
252                               struct ext4_group_desc *bg)
253 {
254         return le16_to_cpu(bg->bg_itable_unused_lo) |
255                 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
256                  (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
257 }
258
259 void ext4_block_bitmap_set(struct super_block *sb,
260                            struct ext4_group_desc *bg, ext4_fsblk_t blk)
261 {
262         bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
263         if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
264                 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
265 }
266
267 void ext4_inode_bitmap_set(struct super_block *sb,
268                            struct ext4_group_desc *bg, ext4_fsblk_t blk)
269 {
270         bg->bg_inode_bitmap_lo  = cpu_to_le32((u32)blk);
271         if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
272                 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
273 }
274
275 void ext4_inode_table_set(struct super_block *sb,
276                           struct ext4_group_desc *bg, ext4_fsblk_t blk)
277 {
278         bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
279         if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
280                 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
281 }
282
283 void ext4_free_group_clusters_set(struct super_block *sb,
284                                   struct ext4_group_desc *bg, __u32 count)
285 {
286         bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
287         if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
288                 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
289 }
290
291 void ext4_free_inodes_set(struct super_block *sb,
292                           struct ext4_group_desc *bg, __u32 count)
293 {
294         bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
295         if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
296                 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
297 }
298
299 void ext4_used_dirs_set(struct super_block *sb,
300                           struct ext4_group_desc *bg, __u32 count)
301 {
302         bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
303         if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
304                 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
305 }
306
307 void ext4_itable_unused_set(struct super_block *sb,
308                           struct ext4_group_desc *bg, __u32 count)
309 {
310         bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
311         if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
312                 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
313 }
314
315
316 static void __save_error_info(struct super_block *sb, const char *func,
317                             unsigned int line)
318 {
319         struct ext4_super_block *es = EXT4_SB(sb)->s_es;
320
321         EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
322         if (bdev_read_only(sb->s_bdev))
323                 return;
324         es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
325         es->s_last_error_time = cpu_to_le32(get_seconds());
326         strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
327         es->s_last_error_line = cpu_to_le32(line);
328         if (!es->s_first_error_time) {
329                 es->s_first_error_time = es->s_last_error_time;
330                 strncpy(es->s_first_error_func, func,
331                         sizeof(es->s_first_error_func));
332                 es->s_first_error_line = cpu_to_le32(line);
333                 es->s_first_error_ino = es->s_last_error_ino;
334                 es->s_first_error_block = es->s_last_error_block;
335         }
336         /*
337          * Start the daily error reporting function if it hasn't been
338          * started already
339          */
340         if (!es->s_error_count)
341                 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
342         le32_add_cpu(&es->s_error_count, 1);
343 }
344
345 static void save_error_info(struct super_block *sb, const char *func,
346                             unsigned int line)
347 {
348         __save_error_info(sb, func, line);
349         ext4_commit_super(sb, 1);
350 }
351
352 /*
353  * The del_gendisk() function uninitializes the disk-specific data
354  * structures, including the bdi structure, without telling anyone
355  * else.  Once this happens, any attempt to call mark_buffer_dirty()
356  * (for example, by ext4_commit_super), will cause a kernel OOPS.
357  * This is a kludge to prevent these oops until we can put in a proper
358  * hook in del_gendisk() to inform the VFS and file system layers.
359  */
360 static int block_device_ejected(struct super_block *sb)
361 {
362         struct inode *bd_inode = sb->s_bdev->bd_inode;
363         struct backing_dev_info *bdi = inode_to_bdi(bd_inode);
364
365         return bdi->dev == NULL;
366 }
367
368 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
369 {
370         struct super_block              *sb = journal->j_private;
371         struct ext4_sb_info             *sbi = EXT4_SB(sb);
372         int                             error = is_journal_aborted(journal);
373         struct ext4_journal_cb_entry    *jce;
374
375         BUG_ON(txn->t_state == T_FINISHED);
376
377         ext4_process_freed_data(sb, txn->t_tid);
378
379         spin_lock(&sbi->s_md_lock);
380         while (!list_empty(&txn->t_private_list)) {
381                 jce = list_entry(txn->t_private_list.next,
382                                  struct ext4_journal_cb_entry, jce_list);
383                 list_del_init(&jce->jce_list);
384                 spin_unlock(&sbi->s_md_lock);
385                 jce->jce_func(sb, jce, error);
386                 spin_lock(&sbi->s_md_lock);
387         }
388         spin_unlock(&sbi->s_md_lock);
389 }
390
391 /* Deal with the reporting of failure conditions on a filesystem such as
392  * inconsistencies detected or read IO failures.
393  *
394  * On ext2, we can store the error state of the filesystem in the
395  * superblock.  That is not possible on ext4, because we may have other
396  * write ordering constraints on the superblock which prevent us from
397  * writing it out straight away; and given that the journal is about to
398  * be aborted, we can't rely on the current, or future, transactions to
399  * write out the superblock safely.
400  *
401  * We'll just use the jbd2_journal_abort() error code to record an error in
402  * the journal instead.  On recovery, the journal will complain about
403  * that error until we've noted it down and cleared it.
404  */
405
406 static void ext4_handle_error(struct super_block *sb)
407 {
408         if (sb_rdonly(sb))
409                 return;
410
411         if (!test_opt(sb, ERRORS_CONT)) {
412                 journal_t *journal = EXT4_SB(sb)->s_journal;
413
414                 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
415                 if (journal)
416                         jbd2_journal_abort(journal, -EIO);
417         }
418         if (test_opt(sb, ERRORS_RO)) {
419                 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
420                 /*
421                  * Make sure updated value of ->s_mount_flags will be visible
422                  * before ->s_flags update
423                  */
424                 smp_wmb();
425                 sb->s_flags |= SB_RDONLY;
426         }
427         if (test_opt(sb, ERRORS_PANIC)) {
428                 if (EXT4_SB(sb)->s_journal &&
429                   !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
430                         return;
431                 panic("EXT4-fs (device %s): panic forced after error\n",
432                         sb->s_id);
433         }
434 }
435
436 #define ext4_error_ratelimit(sb)                                        \
437                 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state),     \
438                              "EXT4-fs error")
439
440 void __ext4_error(struct super_block *sb, const char *function,
441                   unsigned int line, const char *fmt, ...)
442 {
443         struct va_format vaf;
444         va_list args;
445
446         if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
447                 return;
448
449         trace_ext4_error(sb, function, line);
450         if (ext4_error_ratelimit(sb)) {
451                 va_start(args, fmt);
452                 vaf.fmt = fmt;
453                 vaf.va = &args;
454                 printk(KERN_CRIT
455                        "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
456                        sb->s_id, function, line, current->comm, &vaf);
457                 va_end(args);
458         }
459         save_error_info(sb, function, line);
460         ext4_handle_error(sb);
461 }
462
463 void __ext4_error_inode(struct inode *inode, const char *function,
464                         unsigned int line, ext4_fsblk_t block,
465                         const char *fmt, ...)
466 {
467         va_list args;
468         struct va_format vaf;
469         struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
470
471         if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
472                 return;
473
474         trace_ext4_error(inode->i_sb, function, line);
475         es->s_last_error_ino = cpu_to_le32(inode->i_ino);
476         es->s_last_error_block = cpu_to_le64(block);
477         if (ext4_error_ratelimit(inode->i_sb)) {
478                 va_start(args, fmt);
479                 vaf.fmt = fmt;
480                 vaf.va = &args;
481                 if (block)
482                         printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
483                                "inode #%lu: block %llu: comm %s: %pV\n",
484                                inode->i_sb->s_id, function, line, inode->i_ino,
485                                block, current->comm, &vaf);
486                 else
487                         printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
488                                "inode #%lu: comm %s: %pV\n",
489                                inode->i_sb->s_id, function, line, inode->i_ino,
490                                current->comm, &vaf);
491                 va_end(args);
492         }
493         save_error_info(inode->i_sb, function, line);
494         ext4_handle_error(inode->i_sb);
495 }
496
497 void __ext4_error_file(struct file *file, const char *function,
498                        unsigned int line, ext4_fsblk_t block,
499                        const char *fmt, ...)
500 {
501         va_list args;
502         struct va_format vaf;
503         struct ext4_super_block *es;
504         struct inode *inode = file_inode(file);
505         char pathname[80], *path;
506
507         if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
508                 return;
509
510         trace_ext4_error(inode->i_sb, function, line);
511         es = EXT4_SB(inode->i_sb)->s_es;
512         es->s_last_error_ino = cpu_to_le32(inode->i_ino);
513         if (ext4_error_ratelimit(inode->i_sb)) {
514                 path = file_path(file, pathname, sizeof(pathname));
515                 if (IS_ERR(path))
516                         path = "(unknown)";
517                 va_start(args, fmt);
518                 vaf.fmt = fmt;
519                 vaf.va = &args;
520                 if (block)
521                         printk(KERN_CRIT
522                                "EXT4-fs error (device %s): %s:%d: inode #%lu: "
523                                "block %llu: comm %s: path %s: %pV\n",
524                                inode->i_sb->s_id, function, line, inode->i_ino,
525                                block, current->comm, path, &vaf);
526                 else
527                         printk(KERN_CRIT
528                                "EXT4-fs error (device %s): %s:%d: inode #%lu: "
529                                "comm %s: path %s: %pV\n",
530                                inode->i_sb->s_id, function, line, inode->i_ino,
531                                current->comm, path, &vaf);
532                 va_end(args);
533         }
534         save_error_info(inode->i_sb, function, line);
535         ext4_handle_error(inode->i_sb);
536 }
537
538 const char *ext4_decode_error(struct super_block *sb, int errno,
539                               char nbuf[16])
540 {
541         char *errstr = NULL;
542
543         switch (errno) {
544         case -EFSCORRUPTED:
545                 errstr = "Corrupt filesystem";
546                 break;
547         case -EFSBADCRC:
548                 errstr = "Filesystem failed CRC";
549                 break;
550         case -EIO:
551                 errstr = "IO failure";
552                 break;
553         case -ENOMEM:
554                 errstr = "Out of memory";
555                 break;
556         case -EROFS:
557                 if (!sb || (EXT4_SB(sb)->s_journal &&
558                             EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
559                         errstr = "Journal has aborted";
560                 else
561                         errstr = "Readonly filesystem";
562                 break;
563         default:
564                 /* If the caller passed in an extra buffer for unknown
565                  * errors, textualise them now.  Else we just return
566                  * NULL. */
567                 if (nbuf) {
568                         /* Check for truncated error codes... */
569                         if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
570                                 errstr = nbuf;
571                 }
572                 break;
573         }
574
575         return errstr;
576 }
577
578 /* __ext4_std_error decodes expected errors from journaling functions
579  * automatically and invokes the appropriate error response.  */
580
581 void __ext4_std_error(struct super_block *sb, const char *function,
582                       unsigned int line, int errno)
583 {
584         char nbuf[16];
585         const char *errstr;
586
587         if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
588                 return;
589
590         /* Special case: if the error is EROFS, and we're not already
591          * inside a transaction, then there's really no point in logging
592          * an error. */
593         if (errno == -EROFS && journal_current_handle() == NULL && sb_rdonly(sb))
594                 return;
595
596         if (ext4_error_ratelimit(sb)) {
597                 errstr = ext4_decode_error(sb, errno, nbuf);
598                 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
599                        sb->s_id, function, line, errstr);
600         }
601
602         save_error_info(sb, function, line);
603         ext4_handle_error(sb);
604 }
605
606 /*
607  * ext4_abort is a much stronger failure handler than ext4_error.  The
608  * abort function may be used to deal with unrecoverable failures such
609  * as journal IO errors or ENOMEM at a critical moment in log management.
610  *
611  * We unconditionally force the filesystem into an ABORT|READONLY state,
612  * unless the error response on the fs has been set to panic in which
613  * case we take the easy way out and panic immediately.
614  */
615
616 void __ext4_abort(struct super_block *sb, const char *function,
617                 unsigned int line, const char *fmt, ...)
618 {
619         struct va_format vaf;
620         va_list args;
621
622         if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
623                 return;
624
625         save_error_info(sb, function, line);
626         va_start(args, fmt);
627         vaf.fmt = fmt;
628         vaf.va = &args;
629         printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: %pV\n",
630                sb->s_id, function, line, &vaf);
631         va_end(args);
632
633         if (sb_rdonly(sb) == 0) {
634                 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
635                 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
636                 /*
637                  * Make sure updated value of ->s_mount_flags will be visible
638                  * before ->s_flags update
639                  */
640                 smp_wmb();
641                 sb->s_flags |= SB_RDONLY;
642                 if (EXT4_SB(sb)->s_journal)
643                         jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
644                 save_error_info(sb, function, line);
645         }
646         if (test_opt(sb, ERRORS_PANIC)) {
647                 if (EXT4_SB(sb)->s_journal &&
648                   !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
649                         return;
650                 panic("EXT4-fs panic from previous error\n");
651         }
652 }
653
654 void __ext4_msg(struct super_block *sb,
655                 const char *prefix, const char *fmt, ...)
656 {
657         struct va_format vaf;
658         va_list args;
659
660         if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
661                 return;
662
663         va_start(args, fmt);
664         vaf.fmt = fmt;
665         vaf.va = &args;
666         printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
667         va_end(args);
668 }
669
670 #define ext4_warning_ratelimit(sb)                                      \
671                 ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state), \
672                              "EXT4-fs warning")
673
674 void __ext4_warning(struct super_block *sb, const char *function,
675                     unsigned int line, const char *fmt, ...)
676 {
677         struct va_format vaf;
678         va_list args;
679
680         if (!ext4_warning_ratelimit(sb))
681                 return;
682
683         va_start(args, fmt);
684         vaf.fmt = fmt;
685         vaf.va = &args;
686         printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
687                sb->s_id, function, line, &vaf);
688         va_end(args);
689 }
690
691 void __ext4_warning_inode(const struct inode *inode, const char *function,
692                           unsigned int line, const char *fmt, ...)
693 {
694         struct va_format vaf;
695         va_list args;
696
697         if (!ext4_warning_ratelimit(inode->i_sb))
698                 return;
699
700         va_start(args, fmt);
701         vaf.fmt = fmt;
702         vaf.va = &args;
703         printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
704                "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
705                function, line, inode->i_ino, current->comm, &vaf);
706         va_end(args);
707 }
708
709 void __ext4_grp_locked_error(const char *function, unsigned int line,
710                              struct super_block *sb, ext4_group_t grp,
711                              unsigned long ino, ext4_fsblk_t block,
712                              const char *fmt, ...)
713 __releases(bitlock)
714 __acquires(bitlock)
715 {
716         struct va_format vaf;
717         va_list args;
718         struct ext4_super_block *es = EXT4_SB(sb)->s_es;
719
720         if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
721                 return;
722
723         trace_ext4_error(sb, function, line);
724         es->s_last_error_ino = cpu_to_le32(ino);
725         es->s_last_error_block = cpu_to_le64(block);
726         __save_error_info(sb, function, line);
727
728         if (ext4_error_ratelimit(sb)) {
729                 va_start(args, fmt);
730                 vaf.fmt = fmt;
731                 vaf.va = &args;
732                 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
733                        sb->s_id, function, line, grp);
734                 if (ino)
735                         printk(KERN_CONT "inode %lu: ", ino);
736                 if (block)
737                         printk(KERN_CONT "block %llu:",
738                                (unsigned long long) block);
739                 printk(KERN_CONT "%pV\n", &vaf);
740                 va_end(args);
741         }
742
743         if (test_opt(sb, ERRORS_CONT)) {
744                 ext4_commit_super(sb, 0);
745                 return;
746         }
747
748         ext4_unlock_group(sb, grp);
749         ext4_commit_super(sb, 1);
750         ext4_handle_error(sb);
751         /*
752          * We only get here in the ERRORS_RO case; relocking the group
753          * may be dangerous, but nothing bad will happen since the
754          * filesystem will have already been marked read/only and the
755          * journal has been aborted.  We return 1 as a hint to callers
756          * who might what to use the return value from
757          * ext4_grp_locked_error() to distinguish between the
758          * ERRORS_CONT and ERRORS_RO case, and perhaps return more
759          * aggressively from the ext4 function in question, with a
760          * more appropriate error code.
761          */
762         ext4_lock_group(sb, grp);
763         return;
764 }
765
766 void ext4_update_dynamic_rev(struct super_block *sb)
767 {
768         struct ext4_super_block *es = EXT4_SB(sb)->s_es;
769
770         if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
771                 return;
772
773         ext4_warning(sb,
774                      "updating to rev %d because of new feature flag, "
775                      "running e2fsck is recommended",
776                      EXT4_DYNAMIC_REV);
777
778         es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
779         es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
780         es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
781         /* leave es->s_feature_*compat flags alone */
782         /* es->s_uuid will be set by e2fsck if empty */
783
784         /*
785          * The rest of the superblock fields should be zero, and if not it
786          * means they are likely already in use, so leave them alone.  We
787          * can leave it up to e2fsck to clean up any inconsistencies there.
788          */
789 }
790
791 /*
792  * Open the external journal device
793  */
794 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
795 {
796         struct block_device *bdev;
797         char b[BDEVNAME_SIZE];
798
799         bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
800         if (IS_ERR(bdev))
801                 goto fail;
802         return bdev;
803
804 fail:
805         ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
806                         __bdevname(dev, b), PTR_ERR(bdev));
807         return NULL;
808 }
809
810 /*
811  * Release the journal device
812  */
813 static void ext4_blkdev_put(struct block_device *bdev)
814 {
815         blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
816 }
817
818 static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
819 {
820         struct block_device *bdev;
821         bdev = sbi->journal_bdev;
822         if (bdev) {
823                 ext4_blkdev_put(bdev);
824                 sbi->journal_bdev = NULL;
825         }
826 }
827
828 static inline struct inode *orphan_list_entry(struct list_head *l)
829 {
830         return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
831 }
832
833 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
834 {
835         struct list_head *l;
836
837         ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
838                  le32_to_cpu(sbi->s_es->s_last_orphan));
839
840         printk(KERN_ERR "sb_info orphan list:\n");
841         list_for_each(l, &sbi->s_orphan) {
842                 struct inode *inode = orphan_list_entry(l);
843                 printk(KERN_ERR "  "
844                        "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
845                        inode->i_sb->s_id, inode->i_ino, inode,
846                        inode->i_mode, inode->i_nlink,
847                        NEXT_ORPHAN(inode));
848         }
849 }
850
851 #ifdef CONFIG_QUOTA
852 static int ext4_quota_off(struct super_block *sb, int type);
853
854 static inline void ext4_quota_off_umount(struct super_block *sb)
855 {
856         int type;
857
858         /* Use our quota_off function to clear inode flags etc. */
859         for (type = 0; type < EXT4_MAXQUOTAS; type++)
860                 ext4_quota_off(sb, type);
861 }
862 #else
863 static inline void ext4_quota_off_umount(struct super_block *sb)
864 {
865 }
866 #endif
867
868 static void ext4_put_super(struct super_block *sb)
869 {
870         struct ext4_sb_info *sbi = EXT4_SB(sb);
871         struct ext4_super_block *es = sbi->s_es;
872         int aborted = 0;
873         int i, err;
874
875         ext4_unregister_li_request(sb);
876         ext4_quota_off_umount(sb);
877
878         destroy_workqueue(sbi->rsv_conversion_wq);
879
880         if (sbi->s_journal) {
881                 aborted = is_journal_aborted(sbi->s_journal);
882                 err = jbd2_journal_destroy(sbi->s_journal);
883                 sbi->s_journal = NULL;
884                 if ((err < 0) && !aborted)
885                         ext4_abort(sb, "Couldn't clean up the journal");
886         }
887
888         ext4_unregister_sysfs(sb);
889         ext4_es_unregister_shrinker(sbi);
890         del_timer_sync(&sbi->s_err_report);
891         ext4_release_system_zone(sb);
892         ext4_mb_release(sb);
893         ext4_ext_release(sb);
894
895         if (!sb_rdonly(sb) && !aborted) {
896                 ext4_clear_feature_journal_needs_recovery(sb);
897                 es->s_state = cpu_to_le16(sbi->s_mount_state);
898         }
899         if (!sb_rdonly(sb))
900                 ext4_commit_super(sb, 1);
901
902         for (i = 0; i < sbi->s_gdb_count; i++)
903                 brelse(sbi->s_group_desc[i]);
904         kvfree(sbi->s_group_desc);
905         kvfree(sbi->s_flex_groups);
906         percpu_counter_destroy(&sbi->s_freeclusters_counter);
907         percpu_counter_destroy(&sbi->s_freeinodes_counter);
908         percpu_counter_destroy(&sbi->s_dirs_counter);
909         percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
910         percpu_free_rwsem(&sbi->s_journal_flag_rwsem);
911 #ifdef CONFIG_QUOTA
912         for (i = 0; i < EXT4_MAXQUOTAS; i++)
913                 kfree(sbi->s_qf_names[i]);
914 #endif
915
916         /* Debugging code just in case the in-memory inode orphan list
917          * isn't empty.  The on-disk one can be non-empty if we've
918          * detected an error and taken the fs readonly, but the
919          * in-memory list had better be clean by this point. */
920         if (!list_empty(&sbi->s_orphan))
921                 dump_orphan_list(sb, sbi);
922         J_ASSERT(list_empty(&sbi->s_orphan));
923
924         sync_blockdev(sb->s_bdev);
925         invalidate_bdev(sb->s_bdev);
926         if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
927                 /*
928                  * Invalidate the journal device's buffers.  We don't want them
929                  * floating about in memory - the physical journal device may
930                  * hotswapped, and it breaks the `ro-after' testing code.
931                  */
932                 sync_blockdev(sbi->journal_bdev);
933                 invalidate_bdev(sbi->journal_bdev);
934                 ext4_blkdev_remove(sbi);
935         }
936         if (sbi->s_ea_inode_cache) {
937                 ext4_xattr_destroy_cache(sbi->s_ea_inode_cache);
938                 sbi->s_ea_inode_cache = NULL;
939         }
940         if (sbi->s_ea_block_cache) {
941                 ext4_xattr_destroy_cache(sbi->s_ea_block_cache);
942                 sbi->s_ea_block_cache = NULL;
943         }
944         if (sbi->s_mmp_tsk)
945                 kthread_stop(sbi->s_mmp_tsk);
946         brelse(sbi->s_sbh);
947         sb->s_fs_info = NULL;
948         /*
949          * Now that we are completely done shutting down the
950          * superblock, we need to actually destroy the kobject.
951          */
952         kobject_put(&sbi->s_kobj);
953         wait_for_completion(&sbi->s_kobj_unregister);
954         if (sbi->s_chksum_driver)
955                 crypto_free_shash(sbi->s_chksum_driver);
956         kfree(sbi->s_blockgroup_lock);
957         fs_put_dax(sbi->s_daxdev);
958         kfree(sbi);
959 }
960
961 static struct kmem_cache *ext4_inode_cachep;
962
963 /*
964  * Called inside transaction, so use GFP_NOFS
965  */
966 static struct inode *ext4_alloc_inode(struct super_block *sb)
967 {
968         struct ext4_inode_info *ei;
969
970         ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
971         if (!ei)
972                 return NULL;
973
974         inode_set_iversion(&ei->vfs_inode, 1);
975         spin_lock_init(&ei->i_raw_lock);
976         INIT_LIST_HEAD(&ei->i_prealloc_list);
977         spin_lock_init(&ei->i_prealloc_lock);
978         ext4_es_init_tree(&ei->i_es_tree);
979         rwlock_init(&ei->i_es_lock);
980         INIT_LIST_HEAD(&ei->i_es_list);
981         ei->i_es_all_nr = 0;
982         ei->i_es_shk_nr = 0;
983         ei->i_es_shrink_lblk = 0;
984         ei->i_reserved_data_blocks = 0;
985         ei->i_da_metadata_calc_len = 0;
986         ei->i_da_metadata_calc_last_lblock = 0;
987         spin_lock_init(&(ei->i_block_reservation_lock));
988 #ifdef CONFIG_QUOTA
989         ei->i_reserved_quota = 0;
990         memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
991 #endif
992         ei->jinode = NULL;
993         INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
994         spin_lock_init(&ei->i_completed_io_lock);
995         ei->i_sync_tid = 0;
996         ei->i_datasync_tid = 0;
997         atomic_set(&ei->i_unwritten, 0);
998         INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
999         return &ei->vfs_inode;
1000 }
1001
1002 static int ext4_drop_inode(struct inode *inode)
1003 {
1004         int drop = generic_drop_inode(inode);
1005
1006         trace_ext4_drop_inode(inode, drop);
1007         return drop;
1008 }
1009
1010 static void ext4_i_callback(struct rcu_head *head)
1011 {
1012         struct inode *inode = container_of(head, struct inode, i_rcu);
1013         kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
1014 }
1015
1016 static void ext4_destroy_inode(struct inode *inode)
1017 {
1018         if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
1019                 ext4_msg(inode->i_sb, KERN_ERR,
1020                          "Inode %lu (%p): orphan list check failed!",
1021                          inode->i_ino, EXT4_I(inode));
1022                 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
1023                                 EXT4_I(inode), sizeof(struct ext4_inode_info),
1024                                 true);
1025                 dump_stack();
1026         }
1027         call_rcu(&inode->i_rcu, ext4_i_callback);
1028 }
1029
1030 static void init_once(void *foo)
1031 {
1032         struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
1033
1034         INIT_LIST_HEAD(&ei->i_orphan);
1035         init_rwsem(&ei->xattr_sem);
1036         init_rwsem(&ei->i_data_sem);
1037         init_rwsem(&ei->i_mmap_sem);
1038         inode_init_once(&ei->vfs_inode);
1039 }
1040
1041 static int __init init_inodecache(void)
1042 {
1043         ext4_inode_cachep = kmem_cache_create_usercopy("ext4_inode_cache",
1044                                 sizeof(struct ext4_inode_info), 0,
1045                                 (SLAB_RECLAIM_ACCOUNT|SLAB_MEM_SPREAD|
1046                                         SLAB_ACCOUNT),
1047                                 offsetof(struct ext4_inode_info, i_data),
1048                                 sizeof_field(struct ext4_inode_info, i_data),
1049                                 init_once);
1050         if (ext4_inode_cachep == NULL)
1051                 return -ENOMEM;
1052         return 0;
1053 }
1054
1055 static void destroy_inodecache(void)
1056 {
1057         /*
1058          * Make sure all delayed rcu free inodes are flushed before we
1059          * destroy cache.
1060          */
1061         rcu_barrier();
1062         kmem_cache_destroy(ext4_inode_cachep);
1063 }
1064
1065 void ext4_clear_inode(struct inode *inode)
1066 {
1067         invalidate_inode_buffers(inode);
1068         clear_inode(inode);
1069         dquot_drop(inode);
1070         ext4_discard_preallocations(inode);
1071         ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
1072         if (EXT4_I(inode)->jinode) {
1073                 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1074                                                EXT4_I(inode)->jinode);
1075                 jbd2_free_inode(EXT4_I(inode)->jinode);
1076                 EXT4_I(inode)->jinode = NULL;
1077         }
1078         fscrypt_put_encryption_info(inode);
1079 }
1080
1081 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1082                                         u64 ino, u32 generation)
1083 {
1084         struct inode *inode;
1085
1086         if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
1087                 return ERR_PTR(-ESTALE);
1088         if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
1089                 return ERR_PTR(-ESTALE);
1090
1091         /* iget isn't really right if the inode is currently unallocated!!
1092          *
1093          * ext4_read_inode will return a bad_inode if the inode had been
1094          * deleted, so we should be safe.
1095          *
1096          * Currently we don't know the generation for parent directory, so
1097          * a generation of 0 means "accept any"
1098          */
1099         inode = ext4_iget_normal(sb, ino);
1100         if (IS_ERR(inode))
1101                 return ERR_CAST(inode);
1102         if (generation && inode->i_generation != generation) {
1103                 iput(inode);
1104                 return ERR_PTR(-ESTALE);
1105         }
1106
1107         return inode;
1108 }
1109
1110 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1111                                         int fh_len, int fh_type)
1112 {
1113         return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1114                                     ext4_nfs_get_inode);
1115 }
1116
1117 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1118                                         int fh_len, int fh_type)
1119 {
1120         return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1121                                     ext4_nfs_get_inode);
1122 }
1123
1124 /*
1125  * Try to release metadata pages (indirect blocks, directories) which are
1126  * mapped via the block device.  Since these pages could have journal heads
1127  * which would prevent try_to_free_buffers() from freeing them, we must use
1128  * jbd2 layer's try_to_free_buffers() function to release them.
1129  */
1130 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1131                                  gfp_t wait)
1132 {
1133         journal_t *journal = EXT4_SB(sb)->s_journal;
1134
1135         WARN_ON(PageChecked(page));
1136         if (!page_has_buffers(page))
1137                 return 0;
1138         if (journal)
1139                 return jbd2_journal_try_to_free_buffers(journal, page,
1140                                                 wait & ~__GFP_DIRECT_RECLAIM);
1141         return try_to_free_buffers(page);
1142 }
1143
1144 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1145 static int ext4_get_context(struct inode *inode, void *ctx, size_t len)
1146 {
1147         return ext4_xattr_get(inode, EXT4_XATTR_INDEX_ENCRYPTION,
1148                                  EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx, len);
1149 }
1150
1151 static int ext4_set_context(struct inode *inode, const void *ctx, size_t len,
1152                                                         void *fs_data)
1153 {
1154         handle_t *handle = fs_data;
1155         int res, res2, credits, retries = 0;
1156
1157         /*
1158          * Encrypting the root directory is not allowed because e2fsck expects
1159          * lost+found to exist and be unencrypted, and encrypting the root
1160          * directory would imply encrypting the lost+found directory as well as
1161          * the filename "lost+found" itself.
1162          */
1163         if (inode->i_ino == EXT4_ROOT_INO)
1164                 return -EPERM;
1165
1166         if (WARN_ON_ONCE(IS_DAX(inode) && i_size_read(inode)))
1167                 return -EINVAL;
1168
1169         res = ext4_convert_inline_data(inode);
1170         if (res)
1171                 return res;
1172
1173         /*
1174          * If a journal handle was specified, then the encryption context is
1175          * being set on a new inode via inheritance and is part of a larger
1176          * transaction to create the inode.  Otherwise the encryption context is
1177          * being set on an existing inode in its own transaction.  Only in the
1178          * latter case should the "retry on ENOSPC" logic be used.
1179          */
1180
1181         if (handle) {
1182                 res = ext4_xattr_set_handle(handle, inode,
1183                                             EXT4_XATTR_INDEX_ENCRYPTION,
1184                                             EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1185                                             ctx, len, 0);
1186                 if (!res) {
1187                         ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1188                         ext4_clear_inode_state(inode,
1189                                         EXT4_STATE_MAY_INLINE_DATA);
1190                         /*
1191                          * Update inode->i_flags - S_ENCRYPTED will be enabled,
1192                          * S_DAX may be disabled
1193                          */
1194                         ext4_set_inode_flags(inode);
1195                 }
1196                 return res;
1197         }
1198
1199         res = dquot_initialize(inode);
1200         if (res)
1201                 return res;
1202 retry:
1203         res = ext4_xattr_set_credits(inode, len, false /* is_create */,
1204                                      &credits);
1205         if (res)
1206                 return res;
1207
1208         handle = ext4_journal_start(inode, EXT4_HT_MISC, credits);
1209         if (IS_ERR(handle))
1210                 return PTR_ERR(handle);
1211
1212         res = ext4_xattr_set_handle(handle, inode, EXT4_XATTR_INDEX_ENCRYPTION,
1213                                     EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1214                                     ctx, len, 0);
1215         if (!res) {
1216                 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1217                 /*
1218                  * Update inode->i_flags - S_ENCRYPTED will be enabled,
1219                  * S_DAX may be disabled
1220                  */
1221                 ext4_set_inode_flags(inode);
1222                 res = ext4_mark_inode_dirty(handle, inode);
1223                 if (res)
1224                         EXT4_ERROR_INODE(inode, "Failed to mark inode dirty");
1225         }
1226         res2 = ext4_journal_stop(handle);
1227
1228         if (res == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries))
1229                 goto retry;
1230         if (!res)
1231                 res = res2;
1232         return res;
1233 }
1234
1235 static bool ext4_dummy_context(struct inode *inode)
1236 {
1237         return DUMMY_ENCRYPTION_ENABLED(EXT4_SB(inode->i_sb));
1238 }
1239
1240 static unsigned ext4_max_namelen(struct inode *inode)
1241 {
1242         return S_ISLNK(inode->i_mode) ? inode->i_sb->s_blocksize :
1243                 EXT4_NAME_LEN;
1244 }
1245
1246 static const struct fscrypt_operations ext4_cryptops = {
1247         .key_prefix             = "ext4:",
1248         .get_context            = ext4_get_context,
1249         .set_context            = ext4_set_context,
1250         .dummy_context          = ext4_dummy_context,
1251         .empty_dir              = ext4_empty_dir,
1252         .max_namelen            = ext4_max_namelen,
1253 };
1254 #endif
1255
1256 #ifdef CONFIG_QUOTA
1257 static const char * const quotatypes[] = INITQFNAMES;
1258 #define QTYPE2NAME(t) (quotatypes[t])
1259
1260 static int ext4_write_dquot(struct dquot *dquot);
1261 static int ext4_acquire_dquot(struct dquot *dquot);
1262 static int ext4_release_dquot(struct dquot *dquot);
1263 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1264 static int ext4_write_info(struct super_block *sb, int type);
1265 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1266                          const struct path *path);
1267 static int ext4_quota_on_mount(struct super_block *sb, int type);
1268 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1269                                size_t len, loff_t off);
1270 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1271                                 const char *data, size_t len, loff_t off);
1272 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1273                              unsigned int flags);
1274 static int ext4_enable_quotas(struct super_block *sb);
1275 static int ext4_get_next_id(struct super_block *sb, struct kqid *qid);
1276
1277 static struct dquot **ext4_get_dquots(struct inode *inode)
1278 {
1279         return EXT4_I(inode)->i_dquot;
1280 }
1281
1282 static const struct dquot_operations ext4_quota_operations = {
1283         .get_reserved_space     = ext4_get_reserved_space,
1284         .write_dquot            = ext4_write_dquot,
1285         .acquire_dquot          = ext4_acquire_dquot,
1286         .release_dquot          = ext4_release_dquot,
1287         .mark_dirty             = ext4_mark_dquot_dirty,
1288         .write_info             = ext4_write_info,
1289         .alloc_dquot            = dquot_alloc,
1290         .destroy_dquot          = dquot_destroy,
1291         .get_projid             = ext4_get_projid,
1292         .get_inode_usage        = ext4_get_inode_usage,
1293         .get_next_id            = ext4_get_next_id,
1294 };
1295
1296 static const struct quotactl_ops ext4_qctl_operations = {
1297         .quota_on       = ext4_quota_on,
1298         .quota_off      = ext4_quota_off,
1299         .quota_sync     = dquot_quota_sync,
1300         .get_state      = dquot_get_state,
1301         .set_info       = dquot_set_dqinfo,
1302         .get_dqblk      = dquot_get_dqblk,
1303         .set_dqblk      = dquot_set_dqblk,
1304         .get_nextdqblk  = dquot_get_next_dqblk,
1305 };
1306 #endif
1307
1308 static const struct super_operations ext4_sops = {
1309         .alloc_inode    = ext4_alloc_inode,
1310         .destroy_inode  = ext4_destroy_inode,
1311         .write_inode    = ext4_write_inode,
1312         .dirty_inode    = ext4_dirty_inode,
1313         .drop_inode     = ext4_drop_inode,
1314         .evict_inode    = ext4_evict_inode,
1315         .put_super      = ext4_put_super,
1316         .sync_fs        = ext4_sync_fs,
1317         .freeze_fs      = ext4_freeze,
1318         .unfreeze_fs    = ext4_unfreeze,
1319         .statfs         = ext4_statfs,
1320         .remount_fs     = ext4_remount,
1321         .show_options   = ext4_show_options,
1322 #ifdef CONFIG_QUOTA
1323         .quota_read     = ext4_quota_read,
1324         .quota_write    = ext4_quota_write,
1325         .get_dquots     = ext4_get_dquots,
1326 #endif
1327         .bdev_try_to_free_page = bdev_try_to_free_page,
1328 };
1329
1330 static const struct export_operations ext4_export_ops = {
1331         .fh_to_dentry = ext4_fh_to_dentry,
1332         .fh_to_parent = ext4_fh_to_parent,
1333         .get_parent = ext4_get_parent,
1334 };
1335
1336 enum {
1337         Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1338         Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1339         Opt_nouid32, Opt_debug, Opt_removed,
1340         Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1341         Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1342         Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1343         Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1344         Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1345         Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
1346         Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1347         Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1348         Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1349         Opt_usrquota, Opt_grpquota, Opt_prjquota, Opt_i_version, Opt_dax,
1350         Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1351         Opt_lazytime, Opt_nolazytime, Opt_debug_want_extra_isize,
1352         Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1353         Opt_inode_readahead_blks, Opt_journal_ioprio,
1354         Opt_dioread_nolock, Opt_dioread_lock,
1355         Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1356         Opt_max_dir_size_kb, Opt_nojournal_checksum, Opt_nombcache,
1357 };
1358
1359 static const match_table_t tokens = {
1360         {Opt_bsd_df, "bsddf"},
1361         {Opt_minix_df, "minixdf"},
1362         {Opt_grpid, "grpid"},
1363         {Opt_grpid, "bsdgroups"},
1364         {Opt_nogrpid, "nogrpid"},
1365         {Opt_nogrpid, "sysvgroups"},
1366         {Opt_resgid, "resgid=%u"},
1367         {Opt_resuid, "resuid=%u"},
1368         {Opt_sb, "sb=%u"},
1369         {Opt_err_cont, "errors=continue"},
1370         {Opt_err_panic, "errors=panic"},
1371         {Opt_err_ro, "errors=remount-ro"},
1372         {Opt_nouid32, "nouid32"},
1373         {Opt_debug, "debug"},
1374         {Opt_removed, "oldalloc"},
1375         {Opt_removed, "orlov"},
1376         {Opt_user_xattr, "user_xattr"},
1377         {Opt_nouser_xattr, "nouser_xattr"},
1378         {Opt_acl, "acl"},
1379         {Opt_noacl, "noacl"},
1380         {Opt_noload, "norecovery"},
1381         {Opt_noload, "noload"},
1382         {Opt_removed, "nobh"},
1383         {Opt_removed, "bh"},
1384         {Opt_commit, "commit=%u"},
1385         {Opt_min_batch_time, "min_batch_time=%u"},
1386         {Opt_max_batch_time, "max_batch_time=%u"},
1387         {Opt_journal_dev, "journal_dev=%u"},
1388         {Opt_journal_path, "journal_path=%s"},
1389         {Opt_journal_checksum, "journal_checksum"},
1390         {Opt_nojournal_checksum, "nojournal_checksum"},
1391         {Opt_journal_async_commit, "journal_async_commit"},
1392         {Opt_abort, "abort"},
1393         {Opt_data_journal, "data=journal"},
1394         {Opt_data_ordered, "data=ordered"},
1395         {Opt_data_writeback, "data=writeback"},
1396         {Opt_data_err_abort, "data_err=abort"},
1397         {Opt_data_err_ignore, "data_err=ignore"},
1398         {Opt_offusrjquota, "usrjquota="},
1399         {Opt_usrjquota, "usrjquota=%s"},
1400         {Opt_offgrpjquota, "grpjquota="},
1401         {Opt_grpjquota, "grpjquota=%s"},
1402         {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1403         {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1404         {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1405         {Opt_grpquota, "grpquota"},
1406         {Opt_noquota, "noquota"},
1407         {Opt_quota, "quota"},
1408         {Opt_usrquota, "usrquota"},
1409         {Opt_prjquota, "prjquota"},
1410         {Opt_barrier, "barrier=%u"},
1411         {Opt_barrier, "barrier"},
1412         {Opt_nobarrier, "nobarrier"},
1413         {Opt_i_version, "i_version"},
1414         {Opt_dax, "dax"},
1415         {Opt_stripe, "stripe=%u"},
1416         {Opt_delalloc, "delalloc"},
1417         {Opt_lazytime, "lazytime"},
1418         {Opt_nolazytime, "nolazytime"},
1419         {Opt_debug_want_extra_isize, "debug_want_extra_isize=%u"},
1420         {Opt_nodelalloc, "nodelalloc"},
1421         {Opt_removed, "mblk_io_submit"},
1422         {Opt_removed, "nomblk_io_submit"},
1423         {Opt_block_validity, "block_validity"},
1424         {Opt_noblock_validity, "noblock_validity"},
1425         {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1426         {Opt_journal_ioprio, "journal_ioprio=%u"},
1427         {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1428         {Opt_auto_da_alloc, "auto_da_alloc"},
1429         {Opt_noauto_da_alloc, "noauto_da_alloc"},
1430         {Opt_dioread_nolock, "dioread_nolock"},
1431         {Opt_dioread_lock, "dioread_lock"},
1432         {Opt_discard, "discard"},
1433         {Opt_nodiscard, "nodiscard"},
1434         {Opt_init_itable, "init_itable=%u"},
1435         {Opt_init_itable, "init_itable"},
1436         {Opt_noinit_itable, "noinit_itable"},
1437         {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1438         {Opt_test_dummy_encryption, "test_dummy_encryption"},
1439         {Opt_nombcache, "nombcache"},
1440         {Opt_nombcache, "no_mbcache"},  /* for backward compatibility */
1441         {Opt_removed, "check=none"},    /* mount option from ext2/3 */
1442         {Opt_removed, "nocheck"},       /* mount option from ext2/3 */
1443         {Opt_removed, "reservation"},   /* mount option from ext2/3 */
1444         {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1445         {Opt_removed, "journal=%u"},    /* mount option from ext2/3 */
1446         {Opt_err, NULL},
1447 };
1448
1449 static ext4_fsblk_t get_sb_block(void **data)
1450 {
1451         ext4_fsblk_t    sb_block;
1452         char            *options = (char *) *data;
1453
1454         if (!options || strncmp(options, "sb=", 3) != 0)
1455                 return 1;       /* Default location */
1456
1457         options += 3;
1458         /* TODO: use simple_strtoll with >32bit ext4 */
1459         sb_block = simple_strtoul(options, &options, 0);
1460         if (*options && *options != ',') {
1461                 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1462                        (char *) *data);
1463                 return 1;
1464         }
1465         if (*options == ',')
1466                 options++;
1467         *data = (void *) options;
1468
1469         return sb_block;
1470 }
1471
1472 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1473 static const char deprecated_msg[] =
1474         "Mount option \"%s\" will be removed by %s\n"
1475         "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1476
1477 #ifdef CONFIG_QUOTA
1478 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1479 {
1480         struct ext4_sb_info *sbi = EXT4_SB(sb);
1481         char *qname;
1482         int ret = -1;
1483
1484         if (sb_any_quota_loaded(sb) &&
1485                 !sbi->s_qf_names[qtype]) {
1486                 ext4_msg(sb, KERN_ERR,
1487                         "Cannot change journaled "
1488                         "quota options when quota turned on");
1489                 return -1;
1490         }
1491         if (ext4_has_feature_quota(sb)) {
1492                 ext4_msg(sb, KERN_INFO, "Journaled quota options "
1493                          "ignored when QUOTA feature is enabled");
1494                 return 1;
1495         }
1496         qname = match_strdup(args);
1497         if (!qname) {
1498                 ext4_msg(sb, KERN_ERR,
1499                         "Not enough memory for storing quotafile name");
1500                 return -1;
1501         }
1502         if (sbi->s_qf_names[qtype]) {
1503                 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1504                         ret = 1;
1505                 else
1506                         ext4_msg(sb, KERN_ERR,
1507                                  "%s quota file already specified",
1508                                  QTYPE2NAME(qtype));
1509                 goto errout;
1510         }
1511         if (strchr(qname, '/')) {
1512                 ext4_msg(sb, KERN_ERR,
1513                         "quotafile must be on filesystem root");
1514                 goto errout;
1515         }
1516         sbi->s_qf_names[qtype] = qname;
1517         set_opt(sb, QUOTA);
1518         return 1;
1519 errout:
1520         kfree(qname);
1521         return ret;
1522 }
1523
1524 static int clear_qf_name(struct super_block *sb, int qtype)
1525 {
1526
1527         struct ext4_sb_info *sbi = EXT4_SB(sb);
1528
1529         if (sb_any_quota_loaded(sb) &&
1530                 sbi->s_qf_names[qtype]) {
1531                 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1532                         " when quota turned on");
1533                 return -1;
1534         }
1535         kfree(sbi->s_qf_names[qtype]);
1536         sbi->s_qf_names[qtype] = NULL;
1537         return 1;
1538 }
1539 #endif
1540
1541 #define MOPT_SET        0x0001
1542 #define MOPT_CLEAR      0x0002
1543 #define MOPT_NOSUPPORT  0x0004
1544 #define MOPT_EXPLICIT   0x0008
1545 #define MOPT_CLEAR_ERR  0x0010
1546 #define MOPT_GTE0       0x0020
1547 #ifdef CONFIG_QUOTA
1548 #define MOPT_Q          0
1549 #define MOPT_QFMT       0x0040
1550 #else
1551 #define MOPT_Q          MOPT_NOSUPPORT
1552 #define MOPT_QFMT       MOPT_NOSUPPORT
1553 #endif
1554 #define MOPT_DATAJ      0x0080
1555 #define MOPT_NO_EXT2    0x0100
1556 #define MOPT_NO_EXT3    0x0200
1557 #define MOPT_EXT4_ONLY  (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1558 #define MOPT_STRING     0x0400
1559
1560 static const struct mount_opts {
1561         int     token;
1562         int     mount_opt;
1563         int     flags;
1564 } ext4_mount_opts[] = {
1565         {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1566         {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1567         {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1568         {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1569         {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1570         {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1571         {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1572          MOPT_EXT4_ONLY | MOPT_SET},
1573         {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1574          MOPT_EXT4_ONLY | MOPT_CLEAR},
1575         {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1576         {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1577         {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1578          MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1579         {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1580          MOPT_EXT4_ONLY | MOPT_CLEAR},
1581         {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1582          MOPT_EXT4_ONLY | MOPT_CLEAR},
1583         {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1584          MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1585         {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1586                                     EXT4_MOUNT_JOURNAL_CHECKSUM),
1587          MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1588         {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1589         {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1590         {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1591         {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1592         {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1593          MOPT_NO_EXT2},
1594         {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1595          MOPT_NO_EXT2},
1596         {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1597         {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1598         {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1599         {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1600         {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1601         {Opt_commit, 0, MOPT_GTE0},
1602         {Opt_max_batch_time, 0, MOPT_GTE0},
1603         {Opt_min_batch_time, 0, MOPT_GTE0},
1604         {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1605         {Opt_init_itable, 0, MOPT_GTE0},
1606         {Opt_dax, EXT4_MOUNT_DAX, MOPT_SET},
1607         {Opt_stripe, 0, MOPT_GTE0},
1608         {Opt_resuid, 0, MOPT_GTE0},
1609         {Opt_resgid, 0, MOPT_GTE0},
1610         {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1611         {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING},
1612         {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1613         {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1614         {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1615         {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1616          MOPT_NO_EXT2 | MOPT_DATAJ},
1617         {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1618         {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1619 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1620         {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1621         {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1622 #else
1623         {Opt_acl, 0, MOPT_NOSUPPORT},
1624         {Opt_noacl, 0, MOPT_NOSUPPORT},
1625 #endif
1626         {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1627         {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1628         {Opt_debug_want_extra_isize, 0, MOPT_GTE0},
1629         {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1630         {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1631                                                         MOPT_SET | MOPT_Q},
1632         {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1633                                                         MOPT_SET | MOPT_Q},
1634         {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA,
1635                                                         MOPT_SET | MOPT_Q},
1636         {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1637                        EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA),
1638                                                         MOPT_CLEAR | MOPT_Q},
1639         {Opt_usrjquota, 0, MOPT_Q},
1640         {Opt_grpjquota, 0, MOPT_Q},
1641         {Opt_offusrjquota, 0, MOPT_Q},
1642         {Opt_offgrpjquota, 0, MOPT_Q},
1643         {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1644         {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1645         {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1646         {Opt_max_dir_size_kb, 0, MOPT_GTE0},
1647         {Opt_test_dummy_encryption, 0, MOPT_GTE0},
1648         {Opt_nombcache, EXT4_MOUNT_NO_MBCACHE, MOPT_SET},
1649         {Opt_err, 0, 0}
1650 };
1651
1652 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1653                             substring_t *args, unsigned long *journal_devnum,
1654                             unsigned int *journal_ioprio, int is_remount)
1655 {
1656         struct ext4_sb_info *sbi = EXT4_SB(sb);
1657         const struct mount_opts *m;
1658         kuid_t uid;
1659         kgid_t gid;
1660         int arg = 0;
1661
1662 #ifdef CONFIG_QUOTA
1663         if (token == Opt_usrjquota)
1664                 return set_qf_name(sb, USRQUOTA, &args[0]);
1665         else if (token == Opt_grpjquota)
1666                 return set_qf_name(sb, GRPQUOTA, &args[0]);
1667         else if (token == Opt_offusrjquota)
1668                 return clear_qf_name(sb, USRQUOTA);
1669         else if (token == Opt_offgrpjquota)
1670                 return clear_qf_name(sb, GRPQUOTA);
1671 #endif
1672         switch (token) {
1673         case Opt_noacl:
1674         case Opt_nouser_xattr:
1675                 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1676                 break;
1677         case Opt_sb:
1678                 return 1;       /* handled by get_sb_block() */
1679         case Opt_removed:
1680                 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
1681                 return 1;
1682         case Opt_abort:
1683                 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1684                 return 1;
1685         case Opt_i_version:
1686                 sb->s_flags |= SB_I_VERSION;
1687                 return 1;
1688         case Opt_lazytime:
1689                 sb->s_flags |= SB_LAZYTIME;
1690                 return 1;
1691         case Opt_nolazytime:
1692                 sb->s_flags &= ~SB_LAZYTIME;
1693                 return 1;
1694         }
1695
1696         for (m = ext4_mount_opts; m->token != Opt_err; m++)
1697                 if (token == m->token)
1698                         break;
1699
1700         if (m->token == Opt_err) {
1701                 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1702                          "or missing value", opt);
1703                 return -1;
1704         }
1705
1706         if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1707                 ext4_msg(sb, KERN_ERR,
1708                          "Mount option \"%s\" incompatible with ext2", opt);
1709                 return -1;
1710         }
1711         if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1712                 ext4_msg(sb, KERN_ERR,
1713                          "Mount option \"%s\" incompatible with ext3", opt);
1714                 return -1;
1715         }
1716
1717         if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
1718                 return -1;
1719         if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1720                 return -1;
1721         if (m->flags & MOPT_EXPLICIT) {
1722                 if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
1723                         set_opt2(sb, EXPLICIT_DELALLOC);
1724                 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
1725                         set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM);
1726                 } else
1727                         return -1;
1728         }
1729         if (m->flags & MOPT_CLEAR_ERR)
1730                 clear_opt(sb, ERRORS_MASK);
1731         if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1732                 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1733                          "options when quota turned on");
1734                 return -1;
1735         }
1736
1737         if (m->flags & MOPT_NOSUPPORT) {
1738                 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1739         } else if (token == Opt_commit) {
1740                 if (arg == 0)
1741                         arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1742                 sbi->s_commit_interval = HZ * arg;
1743         } else if (token == Opt_debug_want_extra_isize) {
1744                 sbi->s_want_extra_isize = arg;
1745         } else if (token == Opt_max_batch_time) {
1746                 sbi->s_max_batch_time = arg;
1747         } else if (token == Opt_min_batch_time) {
1748                 sbi->s_min_batch_time = arg;
1749         } else if (token == Opt_inode_readahead_blks) {
1750                 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1751                         ext4_msg(sb, KERN_ERR,
1752                                  "EXT4-fs: inode_readahead_blks must be "
1753                                  "0 or a power of 2 smaller than 2^31");
1754                         return -1;
1755                 }
1756                 sbi->s_inode_readahead_blks = arg;
1757         } else if (token == Opt_init_itable) {
1758                 set_opt(sb, INIT_INODE_TABLE);
1759                 if (!args->from)
1760                         arg = EXT4_DEF_LI_WAIT_MULT;
1761                 sbi->s_li_wait_mult = arg;
1762         } else if (token == Opt_max_dir_size_kb) {
1763                 sbi->s_max_dir_size_kb = arg;
1764         } else if (token == Opt_stripe) {
1765                 sbi->s_stripe = arg;
1766         } else if (token == Opt_resuid) {
1767                 uid = make_kuid(current_user_ns(), arg);
1768                 if (!uid_valid(uid)) {
1769                         ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1770                         return -1;
1771                 }
1772                 sbi->s_resuid = uid;
1773         } else if (token == Opt_resgid) {
1774                 gid = make_kgid(current_user_ns(), arg);
1775                 if (!gid_valid(gid)) {
1776                         ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1777                         return -1;
1778                 }
1779                 sbi->s_resgid = gid;
1780         } else if (token == Opt_journal_dev) {
1781                 if (is_remount) {
1782                         ext4_msg(sb, KERN_ERR,
1783                                  "Cannot specify journal on remount");
1784                         return -1;
1785                 }
1786                 *journal_devnum = arg;
1787         } else if (token == Opt_journal_path) {
1788                 char *journal_path;
1789                 struct inode *journal_inode;
1790                 struct path path;
1791                 int error;
1792
1793                 if (is_remount) {
1794                         ext4_msg(sb, KERN_ERR,
1795                                  "Cannot specify journal on remount");
1796                         return -1;
1797                 }
1798                 journal_path = match_strdup(&args[0]);
1799                 if (!journal_path) {
1800                         ext4_msg(sb, KERN_ERR, "error: could not dup "
1801                                 "journal device string");
1802                         return -1;
1803                 }
1804
1805                 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1806                 if (error) {
1807                         ext4_msg(sb, KERN_ERR, "error: could not find "
1808                                 "journal device path: error %d", error);
1809                         kfree(journal_path);
1810                         return -1;
1811                 }
1812
1813                 journal_inode = d_inode(path.dentry);
1814                 if (!S_ISBLK(journal_inode->i_mode)) {
1815                         ext4_msg(sb, KERN_ERR, "error: journal path %s "
1816                                 "is not a block device", journal_path);
1817                         path_put(&path);
1818                         kfree(journal_path);
1819                         return -1;
1820                 }
1821
1822                 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
1823                 path_put(&path);
1824                 kfree(journal_path);
1825         } else if (token == Opt_journal_ioprio) {
1826                 if (arg > 7) {
1827                         ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1828                                  " (must be 0-7)");
1829                         return -1;
1830                 }
1831                 *journal_ioprio =
1832                         IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1833         } else if (token == Opt_test_dummy_encryption) {
1834 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1835                 sbi->s_mount_flags |= EXT4_MF_TEST_DUMMY_ENCRYPTION;
1836                 ext4_msg(sb, KERN_WARNING,
1837                          "Test dummy encryption mode enabled");
1838 #else
1839                 ext4_msg(sb, KERN_WARNING,
1840                          "Test dummy encryption mount option ignored");
1841 #endif
1842         } else if (m->flags & MOPT_DATAJ) {
1843                 if (is_remount) {
1844                         if (!sbi->s_journal)
1845                                 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1846                         else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
1847                                 ext4_msg(sb, KERN_ERR,
1848                                          "Cannot change data mode on remount");
1849                                 return -1;
1850                         }
1851                 } else {
1852                         clear_opt(sb, DATA_FLAGS);
1853                         sbi->s_mount_opt |= m->mount_opt;
1854                 }
1855 #ifdef CONFIG_QUOTA
1856         } else if (m->flags & MOPT_QFMT) {
1857                 if (sb_any_quota_loaded(sb) &&
1858                     sbi->s_jquota_fmt != m->mount_opt) {
1859                         ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1860                                  "quota options when quota turned on");
1861                         return -1;
1862                 }
1863                 if (ext4_has_feature_quota(sb)) {
1864                         ext4_msg(sb, KERN_INFO,
1865                                  "Quota format mount options ignored "
1866                                  "when QUOTA feature is enabled");
1867                         return 1;
1868                 }
1869                 sbi->s_jquota_fmt = m->mount_opt;
1870 #endif
1871         } else if (token == Opt_dax) {
1872 #ifdef CONFIG_FS_DAX
1873                 ext4_msg(sb, KERN_WARNING,
1874                 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
1875                         sbi->s_mount_opt |= m->mount_opt;
1876 #else
1877                 ext4_msg(sb, KERN_INFO, "dax option not supported");
1878                 return -1;
1879 #endif
1880         } else if (token == Opt_data_err_abort) {
1881                 sbi->s_mount_opt |= m->mount_opt;
1882         } else if (token == Opt_data_err_ignore) {
1883                 sbi->s_mount_opt &= ~m->mount_opt;
1884         } else {
1885                 if (!args->from)
1886                         arg = 1;
1887                 if (m->flags & MOPT_CLEAR)
1888                         arg = !arg;
1889                 else if (unlikely(!(m->flags & MOPT_SET))) {
1890                         ext4_msg(sb, KERN_WARNING,
1891                                  "buggy handling of option %s", opt);
1892                         WARN_ON(1);
1893                         return -1;
1894                 }
1895                 if (arg != 0)
1896                         sbi->s_mount_opt |= m->mount_opt;
1897                 else
1898                         sbi->s_mount_opt &= ~m->mount_opt;
1899         }
1900         return 1;
1901 }
1902
1903 static int parse_options(char *options, struct super_block *sb,
1904                          unsigned long *journal_devnum,
1905                          unsigned int *journal_ioprio,
1906                          int is_remount)
1907 {
1908         struct ext4_sb_info *sbi = EXT4_SB(sb);
1909         char *p;
1910         substring_t args[MAX_OPT_ARGS];
1911         int token;
1912
1913         if (!options)
1914                 return 1;
1915
1916         while ((p = strsep(&options, ",")) != NULL) {
1917                 if (!*p)
1918                         continue;
1919                 /*
1920                  * Initialize args struct so we know whether arg was
1921                  * found; some options take optional arguments.
1922                  */
1923                 args[0].to = args[0].from = NULL;
1924                 token = match_token(p, tokens, args);
1925                 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1926                                      journal_ioprio, is_remount) < 0)
1927                         return 0;
1928         }
1929 #ifdef CONFIG_QUOTA
1930         /*
1931          * We do the test below only for project quotas. 'usrquota' and
1932          * 'grpquota' mount options are allowed even without quota feature
1933          * to support legacy quotas in quota files.
1934          */
1935         if (test_opt(sb, PRJQUOTA) && !ext4_has_feature_project(sb)) {
1936                 ext4_msg(sb, KERN_ERR, "Project quota feature not enabled. "
1937                          "Cannot enable project quota enforcement.");
1938                 return 0;
1939         }
1940         if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1941                 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1942                         clear_opt(sb, USRQUOTA);
1943
1944                 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1945                         clear_opt(sb, GRPQUOTA);
1946
1947                 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1948                         ext4_msg(sb, KERN_ERR, "old and new quota "
1949                                         "format mixing");
1950                         return 0;
1951                 }
1952
1953                 if (!sbi->s_jquota_fmt) {
1954                         ext4_msg(sb, KERN_ERR, "journaled quota format "
1955                                         "not specified");
1956                         return 0;
1957                 }
1958         }
1959 #endif
1960         if (test_opt(sb, DIOREAD_NOLOCK)) {
1961                 int blocksize =
1962                         BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1963
1964                 if (blocksize < PAGE_SIZE) {
1965                         ext4_msg(sb, KERN_ERR, "can't mount with "
1966                                  "dioread_nolock if block size != PAGE_SIZE");
1967                         return 0;
1968                 }
1969         }
1970         return 1;
1971 }
1972
1973 static inline void ext4_show_quota_options(struct seq_file *seq,
1974                                            struct super_block *sb)
1975 {
1976 #if defined(CONFIG_QUOTA)
1977         struct ext4_sb_info *sbi = EXT4_SB(sb);
1978
1979         if (sbi->s_jquota_fmt) {
1980                 char *fmtname = "";
1981
1982                 switch (sbi->s_jquota_fmt) {
1983                 case QFMT_VFS_OLD:
1984                         fmtname = "vfsold";
1985                         break;
1986                 case QFMT_VFS_V0:
1987                         fmtname = "vfsv0";
1988                         break;
1989                 case QFMT_VFS_V1:
1990                         fmtname = "vfsv1";
1991                         break;
1992                 }
1993                 seq_printf(seq, ",jqfmt=%s", fmtname);
1994         }
1995
1996         if (sbi->s_qf_names[USRQUOTA])
1997                 seq_show_option(seq, "usrjquota", sbi->s_qf_names[USRQUOTA]);
1998
1999         if (sbi->s_qf_names[GRPQUOTA])
2000                 seq_show_option(seq, "grpjquota", sbi->s_qf_names[GRPQUOTA]);
2001 #endif
2002 }
2003
2004 static const char *token2str(int token)
2005 {
2006         const struct match_token *t;
2007
2008         for (t = tokens; t->token != Opt_err; t++)
2009                 if (t->token == token && !strchr(t->pattern, '='))
2010                         break;
2011         return t->pattern;
2012 }
2013
2014 /*
2015  * Show an option if
2016  *  - it's set to a non-default value OR
2017  *  - if the per-sb default is different from the global default
2018  */
2019 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
2020                               int nodefs)
2021 {
2022         struct ext4_sb_info *sbi = EXT4_SB(sb);
2023         struct ext4_super_block *es = sbi->s_es;
2024         int def_errors, def_mount_opt = sbi->s_def_mount_opt;
2025         const struct mount_opts *m;
2026         char sep = nodefs ? '\n' : ',';
2027
2028 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
2029 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2030
2031         if (sbi->s_sb_block != 1)
2032                 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
2033
2034         for (m = ext4_mount_opts; m->token != Opt_err; m++) {
2035                 int want_set = m->flags & MOPT_SET;
2036                 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
2037                     (m->flags & MOPT_CLEAR_ERR))
2038                         continue;
2039                 if (!nodefs && !(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
2040                         continue; /* skip if same as the default */
2041                 if ((want_set &&
2042                      (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
2043                     (!want_set && (sbi->s_mount_opt & m->mount_opt)))
2044                         continue; /* select Opt_noFoo vs Opt_Foo */
2045                 SEQ_OPTS_PRINT("%s", token2str(m->token));
2046         }
2047
2048         if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
2049             le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
2050                 SEQ_OPTS_PRINT("resuid=%u",
2051                                 from_kuid_munged(&init_user_ns, sbi->s_resuid));
2052         if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
2053             le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
2054                 SEQ_OPTS_PRINT("resgid=%u",
2055                                 from_kgid_munged(&init_user_ns, sbi->s_resgid));
2056         def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
2057         if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
2058                 SEQ_OPTS_PUTS("errors=remount-ro");
2059         if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
2060                 SEQ_OPTS_PUTS("errors=continue");
2061         if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
2062                 SEQ_OPTS_PUTS("errors=panic");
2063         if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
2064                 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
2065         if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
2066                 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
2067         if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
2068                 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2069         if (sb->s_flags & SB_I_VERSION)
2070                 SEQ_OPTS_PUTS("i_version");
2071         if (nodefs || sbi->s_stripe)
2072                 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
2073         if (nodefs || EXT4_MOUNT_DATA_FLAGS &
2074                         (sbi->s_mount_opt ^ def_mount_opt)) {
2075                 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2076                         SEQ_OPTS_PUTS("data=journal");
2077                 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2078                         SEQ_OPTS_PUTS("data=ordered");
2079                 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
2080                         SEQ_OPTS_PUTS("data=writeback");
2081         }
2082         if (nodefs ||
2083             sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
2084                 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
2085                                sbi->s_inode_readahead_blks);
2086
2087         if (test_opt(sb, INIT_INODE_TABLE) && (nodefs ||
2088                        (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
2089                 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
2090         if (nodefs || sbi->s_max_dir_size_kb)
2091                 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
2092         if (test_opt(sb, DATA_ERR_ABORT))
2093                 SEQ_OPTS_PUTS("data_err=abort");
2094
2095         ext4_show_quota_options(seq, sb);
2096         return 0;
2097 }
2098
2099 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
2100 {
2101         return _ext4_show_options(seq, root->d_sb, 0);
2102 }
2103
2104 int ext4_seq_options_show(struct seq_file *seq, void *offset)
2105 {
2106         struct super_block *sb = seq->private;
2107         int rc;
2108
2109         seq_puts(seq, sb_rdonly(sb) ? "ro" : "rw");
2110         rc = _ext4_show_options(seq, sb, 1);
2111         seq_puts(seq, "\n");
2112         return rc;
2113 }
2114
2115 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
2116                             int read_only)
2117 {
2118         struct ext4_sb_info *sbi = EXT4_SB(sb);
2119         int res = 0;
2120
2121         if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
2122                 ext4_msg(sb, KERN_ERR, "revision level too high, "
2123                          "forcing read-only mode");
2124                 res = SB_RDONLY;
2125         }
2126         if (read_only)
2127                 goto done;
2128         if (!(sbi->s_mount_state & EXT4_VALID_FS))
2129                 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
2130                          "running e2fsck is recommended");
2131         else if (sbi->s_mount_state & EXT4_ERROR_FS)
2132                 ext4_msg(sb, KERN_WARNING,
2133                          "warning: mounting fs with errors, "
2134                          "running e2fsck is recommended");
2135         else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
2136                  le16_to_cpu(es->s_mnt_count) >=
2137                  (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
2138                 ext4_msg(sb, KERN_WARNING,
2139                          "warning: maximal mount count reached, "
2140                          "running e2fsck is recommended");
2141         else if (le32_to_cpu(es->s_checkinterval) &&
2142                 (le32_to_cpu(es->s_lastcheck) +
2143                         le32_to_cpu(es->s_checkinterval) <= get_seconds()))
2144                 ext4_msg(sb, KERN_WARNING,
2145                          "warning: checktime reached, "
2146                          "running e2fsck is recommended");
2147         if (!sbi->s_journal)
2148                 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
2149         if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
2150                 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
2151         le16_add_cpu(&es->s_mnt_count, 1);
2152         es->s_mtime = cpu_to_le32(get_seconds());
2153         ext4_update_dynamic_rev(sb);
2154         if (sbi->s_journal)
2155                 ext4_set_feature_journal_needs_recovery(sb);
2156
2157         ext4_commit_super(sb, 1);
2158 done:
2159         if (test_opt(sb, DEBUG))
2160                 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
2161                                 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
2162                         sb->s_blocksize,
2163                         sbi->s_groups_count,
2164                         EXT4_BLOCKS_PER_GROUP(sb),
2165                         EXT4_INODES_PER_GROUP(sb),
2166                         sbi->s_mount_opt, sbi->s_mount_opt2);
2167
2168         cleancache_init_fs(sb);
2169         return res;
2170 }
2171
2172 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
2173 {
2174         struct ext4_sb_info *sbi = EXT4_SB(sb);
2175         struct flex_groups *new_groups;
2176         int size;
2177
2178         if (!sbi->s_log_groups_per_flex)
2179                 return 0;
2180
2181         size = ext4_flex_group(sbi, ngroup - 1) + 1;
2182         if (size <= sbi->s_flex_groups_allocated)
2183                 return 0;
2184
2185         size = roundup_pow_of_two(size * sizeof(struct flex_groups));
2186         new_groups = kvzalloc(size, GFP_KERNEL);
2187         if (!new_groups) {
2188                 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
2189                          size / (int) sizeof(struct flex_groups));
2190                 return -ENOMEM;
2191         }
2192
2193         if (sbi->s_flex_groups) {
2194                 memcpy(new_groups, sbi->s_flex_groups,
2195                        (sbi->s_flex_groups_allocated *
2196                         sizeof(struct flex_groups)));
2197                 kvfree(sbi->s_flex_groups);
2198         }
2199         sbi->s_flex_groups = new_groups;
2200         sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
2201         return 0;
2202 }
2203
2204 static int ext4_fill_flex_info(struct super_block *sb)
2205 {
2206         struct ext4_sb_info *sbi = EXT4_SB(sb);
2207         struct ext4_group_desc *gdp = NULL;
2208         ext4_group_t flex_group;
2209         int i, err;
2210
2211         sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
2212         if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
2213                 sbi->s_log_groups_per_flex = 0;
2214                 return 1;
2215         }
2216
2217         err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
2218         if (err)
2219                 goto failed;
2220
2221         for (i = 0; i < sbi->s_groups_count; i++) {
2222                 gdp = ext4_get_group_desc(sb, i, NULL);
2223
2224                 flex_group = ext4_flex_group(sbi, i);
2225                 atomic_add(ext4_free_inodes_count(sb, gdp),
2226                            &sbi->s_flex_groups[flex_group].free_inodes);
2227                 atomic64_add(ext4_free_group_clusters(sb, gdp),
2228                              &sbi->s_flex_groups[flex_group].free_clusters);
2229                 atomic_add(ext4_used_dirs_count(sb, gdp),
2230                            &sbi->s_flex_groups[flex_group].used_dirs);
2231         }
2232
2233         return 1;
2234 failed:
2235         return 0;
2236 }
2237
2238 static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
2239                                    struct ext4_group_desc *gdp)
2240 {
2241         int offset = offsetof(struct ext4_group_desc, bg_checksum);
2242         __u16 crc = 0;
2243         __le32 le_group = cpu_to_le32(block_group);
2244         struct ext4_sb_info *sbi = EXT4_SB(sb);
2245
2246         if (ext4_has_metadata_csum(sbi->s_sb)) {
2247                 /* Use new metadata_csum algorithm */
2248                 __u32 csum32;
2249                 __u16 dummy_csum = 0;
2250
2251                 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2252                                      sizeof(le_group));
2253                 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset);
2254                 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum,
2255                                      sizeof(dummy_csum));
2256                 offset += sizeof(dummy_csum);
2257                 if (offset < sbi->s_desc_size)
2258                         csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset,
2259                                              sbi->s_desc_size - offset);
2260
2261                 crc = csum32 & 0xFFFF;
2262                 goto out;
2263         }
2264
2265         /* old crc16 code */
2266         if (!ext4_has_feature_gdt_csum(sb))
2267                 return 0;
2268
2269         crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2270         crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2271         crc = crc16(crc, (__u8 *)gdp, offset);
2272         offset += sizeof(gdp->bg_checksum); /* skip checksum */
2273         /* for checksum of struct ext4_group_desc do the rest...*/
2274         if (ext4_has_feature_64bit(sb) &&
2275             offset < le16_to_cpu(sbi->s_es->s_desc_size))
2276                 crc = crc16(crc, (__u8 *)gdp + offset,
2277                             le16_to_cpu(sbi->s_es->s_desc_size) -
2278                                 offset);
2279
2280 out:
2281         return cpu_to_le16(crc);
2282 }
2283
2284 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2285                                 struct ext4_group_desc *gdp)
2286 {
2287         if (ext4_has_group_desc_csum(sb) &&
2288             (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
2289                 return 0;
2290
2291         return 1;
2292 }
2293
2294 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2295                               struct ext4_group_desc *gdp)
2296 {
2297         if (!ext4_has_group_desc_csum(sb))
2298                 return;
2299         gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
2300 }
2301
2302 /* Called at mount-time, super-block is locked */
2303 static int ext4_check_descriptors(struct super_block *sb,
2304                                   ext4_fsblk_t sb_block,
2305                                   ext4_group_t *first_not_zeroed)
2306 {
2307         struct ext4_sb_info *sbi = EXT4_SB(sb);
2308         ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2309         ext4_fsblk_t last_block;
2310         ext4_fsblk_t block_bitmap;
2311         ext4_fsblk_t inode_bitmap;
2312         ext4_fsblk_t inode_table;
2313         int flexbg_flag = 0;
2314         ext4_group_t i, grp = sbi->s_groups_count;
2315
2316         if (ext4_has_feature_flex_bg(sb))
2317                 flexbg_flag = 1;
2318
2319         ext4_debug("Checking group descriptors");
2320
2321         for (i = 0; i < sbi->s_groups_count; i++) {
2322                 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2323
2324                 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2325                         last_block = ext4_blocks_count(sbi->s_es) - 1;
2326                 else
2327                         last_block = first_block +
2328                                 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2329
2330                 if ((grp == sbi->s_groups_count) &&
2331                    !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2332                         grp = i;
2333
2334                 block_bitmap = ext4_block_bitmap(sb, gdp);
2335                 if (block_bitmap == sb_block) {
2336                         ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2337                                  "Block bitmap for group %u overlaps "
2338                                  "superblock", i);
2339                         if (!sb_rdonly(sb))
2340                                 return 0;
2341                 }
2342                 if (block_bitmap < first_block || block_bitmap > last_block) {
2343                         ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2344                                "Block bitmap for group %u not in group "
2345                                "(block %llu)!", i, block_bitmap);
2346                         return 0;
2347                 }
2348                 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2349                 if (inode_bitmap == sb_block) {
2350                         ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2351                                  "Inode bitmap for group %u overlaps "
2352                                  "superblock", i);
2353                         if (!sb_rdonly(sb))
2354                                 return 0;
2355                 }
2356                 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2357                         ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2358                                "Inode bitmap for group %u not in group "
2359                                "(block %llu)!", i, inode_bitmap);
2360                         return 0;
2361                 }
2362                 inode_table = ext4_inode_table(sb, gdp);
2363                 if (inode_table == sb_block) {
2364                         ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2365                                  "Inode table for group %u overlaps "
2366                                  "superblock", i);
2367                         if (!sb_rdonly(sb))
2368                                 return 0;
2369                 }
2370                 if (inode_table < first_block ||
2371                     inode_table + sbi->s_itb_per_group - 1 > last_block) {
2372                         ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2373                                "Inode table for group %u not in group "
2374                                "(block %llu)!", i, inode_table);
2375                         return 0;
2376                 }
2377                 ext4_lock_group(sb, i);
2378                 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2379                         ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2380                                  "Checksum for group %u failed (%u!=%u)",
2381                                  i, le16_to_cpu(ext4_group_desc_csum(sb, i,
2382                                      gdp)), le16_to_cpu(gdp->bg_checksum));
2383                         if (!sb_rdonly(sb)) {
2384                                 ext4_unlock_group(sb, i);
2385                                 return 0;
2386                         }
2387                 }
2388                 ext4_unlock_group(sb, i);
2389                 if (!flexbg_flag)
2390                         first_block += EXT4_BLOCKS_PER_GROUP(sb);
2391         }
2392         if (NULL != first_not_zeroed)
2393                 *first_not_zeroed = grp;
2394         return 1;
2395 }
2396
2397 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2398  * the superblock) which were deleted from all directories, but held open by
2399  * a process at the time of a crash.  We walk the list and try to delete these
2400  * inodes at recovery time (only with a read-write filesystem).
2401  *
2402  * In order to keep the orphan inode chain consistent during traversal (in
2403  * case of crash during recovery), we link each inode into the superblock
2404  * orphan list_head and handle it the same way as an inode deletion during
2405  * normal operation (which journals the operations for us).
2406  *
2407  * We only do an iget() and an iput() on each inode, which is very safe if we
2408  * accidentally point at an in-use or already deleted inode.  The worst that
2409  * can happen in this case is that we get a "bit already cleared" message from
2410  * ext4_free_inode().  The only reason we would point at a wrong inode is if
2411  * e2fsck was run on this filesystem, and it must have already done the orphan
2412  * inode cleanup for us, so we can safely abort without any further action.
2413  */
2414 static void ext4_orphan_cleanup(struct super_block *sb,
2415                                 struct ext4_super_block *es)
2416 {
2417         unsigned int s_flags = sb->s_flags;
2418         int ret, nr_orphans = 0, nr_truncates = 0;
2419 #ifdef CONFIG_QUOTA
2420         int quota_update = 0;
2421         int i;
2422 #endif
2423         if (!es->s_last_orphan) {
2424                 jbd_debug(4, "no orphan inodes to clean up\n");
2425                 return;
2426         }
2427
2428         if (bdev_read_only(sb->s_bdev)) {
2429                 ext4_msg(sb, KERN_ERR, "write access "
2430                         "unavailable, skipping orphan cleanup");
2431                 return;
2432         }
2433
2434         /* Check if feature set would not allow a r/w mount */
2435         if (!ext4_feature_set_ok(sb, 0)) {
2436                 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2437                          "unknown ROCOMPAT features");
2438                 return;
2439         }
2440
2441         if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2442                 /* don't clear list on RO mount w/ errors */
2443                 if (es->s_last_orphan && !(s_flags & SB_RDONLY)) {
2444                         ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
2445                                   "clearing orphan list.\n");
2446                         es->s_last_orphan = 0;
2447                 }
2448                 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2449                 return;
2450         }
2451
2452         if (s_flags & SB_RDONLY) {
2453                 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2454                 sb->s_flags &= ~SB_RDONLY;
2455         }
2456 #ifdef CONFIG_QUOTA
2457         /* Needed for iput() to work correctly and not trash data */
2458         sb->s_flags |= SB_ACTIVE;
2459
2460         /*
2461          * Turn on quotas which were not enabled for read-only mounts if
2462          * filesystem has quota feature, so that they are updated correctly.
2463          */
2464         if (ext4_has_feature_quota(sb) && (s_flags & SB_RDONLY)) {
2465                 int ret = ext4_enable_quotas(sb);
2466
2467                 if (!ret)
2468                         quota_update = 1;
2469                 else
2470                         ext4_msg(sb, KERN_ERR,
2471                                 "Cannot turn on quotas: error %d", ret);
2472         }
2473
2474         /* Turn on journaled quotas used for old sytle */
2475         for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2476                 if (EXT4_SB(sb)->s_qf_names[i]) {
2477                         int ret = ext4_quota_on_mount(sb, i);
2478
2479                         if (!ret)
2480                                 quota_update = 1;
2481                         else
2482                                 ext4_msg(sb, KERN_ERR,
2483                                         "Cannot turn on journaled "
2484                                         "quota: type %d: error %d", i, ret);
2485                 }
2486         }
2487 #endif
2488
2489         while (es->s_last_orphan) {
2490                 struct inode *inode;
2491
2492                 /*
2493                  * We may have encountered an error during cleanup; if
2494                  * so, skip the rest.
2495                  */
2496                 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2497                         jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2498                         es->s_last_orphan = 0;
2499                         break;
2500                 }
2501
2502                 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2503                 if (IS_ERR(inode)) {
2504                         es->s_last_orphan = 0;
2505                         break;
2506                 }
2507
2508                 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2509                 dquot_initialize(inode);
2510                 if (inode->i_nlink) {
2511                         if (test_opt(sb, DEBUG))
2512                                 ext4_msg(sb, KERN_DEBUG,
2513                                         "%s: truncating inode %lu to %lld bytes",
2514                                         __func__, inode->i_ino, inode->i_size);
2515                         jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2516                                   inode->i_ino, inode->i_size);
2517                         inode_lock(inode);
2518                         truncate_inode_pages(inode->i_mapping, inode->i_size);
2519                         ret = ext4_truncate(inode);
2520                         if (ret)
2521                                 ext4_std_error(inode->i_sb, ret);
2522                         inode_unlock(inode);
2523                         nr_truncates++;
2524                 } else {
2525                         if (test_opt(sb, DEBUG))
2526                                 ext4_msg(sb, KERN_DEBUG,
2527                                         "%s: deleting unreferenced inode %lu",
2528                                         __func__, inode->i_ino);
2529                         jbd_debug(2, "deleting unreferenced inode %lu\n",
2530                                   inode->i_ino);
2531                         nr_orphans++;
2532                 }
2533                 iput(inode);  /* The delete magic happens here! */
2534         }
2535
2536 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2537
2538         if (nr_orphans)
2539                 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2540                        PLURAL(nr_orphans));
2541         if (nr_truncates)
2542                 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2543                        PLURAL(nr_truncates));
2544 #ifdef CONFIG_QUOTA
2545         /* Turn off quotas if they were enabled for orphan cleanup */
2546         if (quota_update) {
2547                 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2548                         if (sb_dqopt(sb)->files[i])
2549                                 dquot_quota_off(sb, i);
2550                 }
2551         }
2552 #endif
2553         sb->s_flags = s_flags; /* Restore SB_RDONLY status */
2554 }
2555
2556 /*
2557  * Maximal extent format file size.
2558  * Resulting logical blkno at s_maxbytes must fit in our on-disk
2559  * extent format containers, within a sector_t, and within i_blocks
2560  * in the vfs.  ext4 inode has 48 bits of i_block in fsblock units,
2561  * so that won't be a limiting factor.
2562  *
2563  * However there is other limiting factor. We do store extents in the form
2564  * of starting block and length, hence the resulting length of the extent
2565  * covering maximum file size must fit into on-disk format containers as
2566  * well. Given that length is always by 1 unit bigger than max unit (because
2567  * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2568  *
2569  * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2570  */
2571 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2572 {
2573         loff_t res;
2574         loff_t upper_limit = MAX_LFS_FILESIZE;
2575
2576         /* small i_blocks in vfs inode? */
2577         if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2578                 /*
2579                  * CONFIG_LBDAF is not enabled implies the inode
2580                  * i_block represent total blocks in 512 bytes
2581                  * 32 == size of vfs inode i_blocks * 8
2582                  */
2583                 upper_limit = (1LL << 32) - 1;
2584
2585                 /* total blocks in file system block size */
2586                 upper_limit >>= (blkbits - 9);
2587                 upper_limit <<= blkbits;
2588         }
2589
2590         /*
2591          * 32-bit extent-start container, ee_block. We lower the maxbytes
2592          * by one fs block, so ee_len can cover the extent of maximum file
2593          * size
2594          */
2595         res = (1LL << 32) - 1;
2596         res <<= blkbits;
2597
2598         /* Sanity check against vm- & vfs- imposed limits */
2599         if (res > upper_limit)
2600                 res = upper_limit;
2601
2602         return res;
2603 }
2604
2605 /*
2606  * Maximal bitmap file size.  There is a direct, and {,double-,triple-}indirect
2607  * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2608  * We need to be 1 filesystem block less than the 2^48 sector limit.
2609  */
2610 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2611 {
2612         loff_t res = EXT4_NDIR_BLOCKS;
2613         int meta_blocks;
2614         loff_t upper_limit;
2615         /* This is calculated to be the largest file size for a dense, block
2616          * mapped file such that the file's total number of 512-byte sectors,
2617          * including data and all indirect blocks, does not exceed (2^48 - 1).
2618          *
2619          * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2620          * number of 512-byte sectors of the file.
2621          */
2622
2623         if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2624                 /*
2625                  * !has_huge_files or CONFIG_LBDAF not enabled implies that
2626                  * the inode i_block field represents total file blocks in
2627                  * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2628                  */
2629                 upper_limit = (1LL << 32) - 1;
2630
2631                 /* total blocks in file system block size */
2632                 upper_limit >>= (bits - 9);
2633
2634         } else {
2635                 /*
2636                  * We use 48 bit ext4_inode i_blocks
2637                  * With EXT4_HUGE_FILE_FL set the i_blocks
2638                  * represent total number of blocks in
2639                  * file system block size
2640                  */
2641                 upper_limit = (1LL << 48) - 1;
2642
2643         }
2644
2645         /* indirect blocks */
2646         meta_blocks = 1;
2647         /* double indirect blocks */
2648         meta_blocks += 1 + (1LL << (bits-2));
2649         /* tripple indirect blocks */
2650         meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2651
2652         upper_limit -= meta_blocks;
2653         upper_limit <<= bits;
2654
2655         res += 1LL << (bits-2);
2656         res += 1LL << (2*(bits-2));
2657         res += 1LL << (3*(bits-2));
2658         res <<= bits;
2659         if (res > upper_limit)
2660                 res = upper_limit;
2661
2662         if (res > MAX_LFS_FILESIZE)
2663                 res = MAX_LFS_FILESIZE;
2664
2665         return res;
2666 }
2667
2668 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2669                                    ext4_fsblk_t logical_sb_block, int nr)
2670 {
2671         struct ext4_sb_info *sbi = EXT4_SB(sb);
2672         ext4_group_t bg, first_meta_bg;
2673         int has_super = 0;
2674
2675         first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2676
2677         if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
2678                 return logical_sb_block + nr + 1;
2679         bg = sbi->s_desc_per_block * nr;
2680         if (ext4_bg_has_super(sb, bg))
2681                 has_super = 1;
2682
2683         /*
2684          * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2685          * block 2, not 1.  If s_first_data_block == 0 (bigalloc is enabled
2686          * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2687          * compensate.
2688          */
2689         if (sb->s_blocksize == 1024 && nr == 0 &&
2690             le32_to_cpu(sbi->s_es->s_first_data_block) == 0)
2691                 has_super++;
2692
2693         return (has_super + ext4_group_first_block_no(sb, bg));
2694 }
2695
2696 /**
2697  * ext4_get_stripe_size: Get the stripe size.
2698  * @sbi: In memory super block info
2699  *
2700  * If we have specified it via mount option, then
2701  * use the mount option value. If the value specified at mount time is
2702  * greater than the blocks per group use the super block value.
2703  * If the super block value is greater than blocks per group return 0.
2704  * Allocator needs it be less than blocks per group.
2705  *
2706  */
2707 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2708 {
2709         unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2710         unsigned long stripe_width =
2711                         le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2712         int ret;
2713
2714         if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2715                 ret = sbi->s_stripe;
2716         else if (stripe_width && stripe_width <= sbi->s_blocks_per_group)
2717                 ret = stripe_width;
2718         else if (stride && stride <= sbi->s_blocks_per_group)
2719                 ret = stride;
2720         else
2721                 ret = 0;
2722
2723         /*
2724          * If the stripe width is 1, this makes no sense and
2725          * we set it to 0 to turn off stripe handling code.
2726          */
2727         if (ret <= 1)
2728                 ret = 0;
2729
2730         return ret;
2731 }
2732
2733 /*
2734  * Check whether this filesystem can be mounted based on
2735  * the features present and the RDONLY/RDWR mount requested.
2736  * Returns 1 if this filesystem can be mounted as requested,
2737  * 0 if it cannot be.
2738  */
2739 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2740 {
2741         if (ext4_has_unknown_ext4_incompat_features(sb)) {
2742                 ext4_msg(sb, KERN_ERR,
2743                         "Couldn't mount because of "
2744                         "unsupported optional features (%x)",
2745                         (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2746                         ~EXT4_FEATURE_INCOMPAT_SUPP));
2747                 return 0;
2748         }
2749
2750         if (readonly)
2751                 return 1;
2752
2753         if (ext4_has_feature_readonly(sb)) {
2754                 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
2755                 sb->s_flags |= SB_RDONLY;
2756                 return 1;
2757         }
2758
2759         /* Check that feature set is OK for a read-write mount */
2760         if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
2761                 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2762                          "unsupported optional features (%x)",
2763                          (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2764                                 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2765                 return 0;
2766         }
2767         /*
2768          * Large file size enabled file system can only be mounted
2769          * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2770          */
2771         if (ext4_has_feature_huge_file(sb)) {
2772                 if (sizeof(blkcnt_t) < sizeof(u64)) {
2773                         ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2774                                  "cannot be mounted RDWR without "
2775                                  "CONFIG_LBDAF");
2776                         return 0;
2777                 }
2778         }
2779         if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
2780                 ext4_msg(sb, KERN_ERR,
2781                          "Can't support bigalloc feature without "
2782                          "extents feature\n");
2783                 return 0;
2784         }
2785
2786 #ifndef CONFIG_QUOTA
2787         if (ext4_has_feature_quota(sb) && !readonly) {
2788                 ext4_msg(sb, KERN_ERR,
2789                          "Filesystem with quota feature cannot be mounted RDWR "
2790                          "without CONFIG_QUOTA");
2791                 return 0;
2792         }
2793         if (ext4_has_feature_project(sb) && !readonly) {
2794                 ext4_msg(sb, KERN_ERR,
2795                          "Filesystem with project quota feature cannot be mounted RDWR "
2796                          "without CONFIG_QUOTA");
2797                 return 0;
2798         }
2799 #endif  /* CONFIG_QUOTA */
2800         return 1;
2801 }
2802
2803 /*
2804  * This function is called once a day if we have errors logged
2805  * on the file system
2806  */
2807 static void print_daily_error_info(struct timer_list *t)
2808 {
2809         struct ext4_sb_info *sbi = from_timer(sbi, t, s_err_report);
2810         struct super_block *sb = sbi->s_sb;
2811         struct ext4_super_block *es = sbi->s_es;
2812
2813         if (es->s_error_count)
2814                 /* fsck newer than v1.41.13 is needed to clean this condition. */
2815                 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
2816                          le32_to_cpu(es->s_error_count));
2817         if (es->s_first_error_time) {
2818                 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d",
2819                        sb->s_id, le32_to_cpu(es->s_first_error_time),
2820                        (int) sizeof(es->s_first_error_func),
2821                        es->s_first_error_func,
2822                        le32_to_cpu(es->s_first_error_line));
2823                 if (es->s_first_error_ino)
2824                         printk(KERN_CONT ": inode %u",
2825                                le32_to_cpu(es->s_first_error_ino));
2826                 if (es->s_first_error_block)
2827                         printk(KERN_CONT ": block %llu", (unsigned long long)
2828                                le64_to_cpu(es->s_first_error_block));
2829                 printk(KERN_CONT "\n");
2830         }
2831         if (es->s_last_error_time) {
2832                 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d",
2833                        sb->s_id, le32_to_cpu(es->s_last_error_time),
2834                        (int) sizeof(es->s_last_error_func),
2835                        es->s_last_error_func,
2836                        le32_to_cpu(es->s_last_error_line));
2837                 if (es->s_last_error_ino)
2838                         printk(KERN_CONT ": inode %u",
2839                                le32_to_cpu(es->s_last_error_ino));
2840                 if (es->s_last_error_block)
2841                         printk(KERN_CONT ": block %llu", (unsigned long long)
2842                                le64_to_cpu(es->s_last_error_block));
2843                 printk(KERN_CONT "\n");
2844         }
2845         mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ);  /* Once a day */
2846 }
2847
2848 /* Find next suitable group and run ext4_init_inode_table */
2849 static int ext4_run_li_request(struct ext4_li_request *elr)
2850 {
2851         struct ext4_group_desc *gdp = NULL;
2852         ext4_group_t group, ngroups;
2853         struct super_block *sb;
2854         unsigned long timeout = 0;
2855         int ret = 0;
2856
2857         sb = elr->lr_super;
2858         ngroups = EXT4_SB(sb)->s_groups_count;
2859
2860         for (group = elr->lr_next_group; group < ngroups; group++) {
2861                 gdp = ext4_get_group_desc(sb, group, NULL);
2862                 if (!gdp) {
2863                         ret = 1;
2864                         break;
2865                 }
2866
2867                 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2868                         break;
2869         }
2870
2871         if (group >= ngroups)
2872                 ret = 1;
2873
2874         if (!ret) {
2875                 timeout = jiffies;
2876                 ret = ext4_init_inode_table(sb, group,
2877                                             elr->lr_timeout ? 0 : 1);
2878                 if (elr->lr_timeout == 0) {
2879                         timeout = (jiffies - timeout) *
2880                                   elr->lr_sbi->s_li_wait_mult;
2881                         elr->lr_timeout = timeout;
2882                 }
2883                 elr->lr_next_sched = jiffies + elr->lr_timeout;
2884                 elr->lr_next_group = group + 1;
2885         }
2886         return ret;
2887 }
2888
2889 /*
2890  * Remove lr_request from the list_request and free the
2891  * request structure. Should be called with li_list_mtx held
2892  */
2893 static void ext4_remove_li_request(struct ext4_li_request *elr)
2894 {
2895         struct ext4_sb_info *sbi;
2896
2897         if (!elr)
2898                 return;
2899
2900         sbi = elr->lr_sbi;
2901
2902         list_del(&elr->lr_request);
2903         sbi->s_li_request = NULL;
2904         kfree(elr);
2905 }
2906
2907 static void ext4_unregister_li_request(struct super_block *sb)
2908 {
2909         mutex_lock(&ext4_li_mtx);
2910         if (!ext4_li_info) {
2911                 mutex_unlock(&ext4_li_mtx);
2912                 return;
2913         }
2914
2915         mutex_lock(&ext4_li_info->li_list_mtx);
2916         ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2917         mutex_unlock(&ext4_li_info->li_list_mtx);
2918         mutex_unlock(&ext4_li_mtx);
2919 }
2920
2921 static struct task_struct *ext4_lazyinit_task;
2922
2923 /*
2924  * This is the function where ext4lazyinit thread lives. It walks
2925  * through the request list searching for next scheduled filesystem.
2926  * When such a fs is found, run the lazy initialization request
2927  * (ext4_rn_li_request) and keep track of the time spend in this
2928  * function. Based on that time we compute next schedule time of
2929  * the request. When walking through the list is complete, compute
2930  * next waking time and put itself into sleep.
2931  */
2932 static int ext4_lazyinit_thread(void *arg)
2933 {
2934         struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2935         struct list_head *pos, *n;
2936         struct ext4_li_request *elr;
2937         unsigned long next_wakeup, cur;
2938
2939         BUG_ON(NULL == eli);
2940
2941 cont_thread:
2942         while (true) {
2943                 next_wakeup = MAX_JIFFY_OFFSET;
2944
2945                 mutex_lock(&eli->li_list_mtx);
2946                 if (list_empty(&eli->li_request_list)) {
2947                         mutex_unlock(&eli->li_list_mtx);
2948                         goto exit_thread;
2949                 }
2950                 list_for_each_safe(pos, n, &eli->li_request_list) {
2951                         int err = 0;
2952                         int progress = 0;
2953                         elr = list_entry(pos, struct ext4_li_request,
2954                                          lr_request);
2955
2956                         if (time_before(jiffies, elr->lr_next_sched)) {
2957                                 if (time_before(elr->lr_next_sched, next_wakeup))
2958                                         next_wakeup = elr->lr_next_sched;
2959                                 continue;
2960                         }
2961                         if (down_read_trylock(&elr->lr_super->s_umount)) {
2962                                 if (sb_start_write_trylock(elr->lr_super)) {
2963                                         progress = 1;
2964                                         /*
2965                                          * We hold sb->s_umount, sb can not
2966                                          * be removed from the list, it is
2967                                          * now safe to drop li_list_mtx
2968                                          */
2969                                         mutex_unlock(&eli->li_list_mtx);
2970                                         err = ext4_run_li_request(elr);
2971                                         sb_end_write(elr->lr_super);
2972                                         mutex_lock(&eli->li_list_mtx);
2973                                         n = pos->next;
2974                                 }
2975                                 up_read((&elr->lr_super->s_umount));
2976                         }
2977                         /* error, remove the lazy_init job */
2978                         if (err) {
2979                                 ext4_remove_li_request(elr);
2980                                 continue;
2981                         }
2982                         if (!progress) {
2983                                 elr->lr_next_sched = jiffies +
2984                                         (prandom_u32()
2985                                          % (EXT4_DEF_LI_MAX_START_DELAY * HZ));
2986                         }
2987                         if (time_before(elr->lr_next_sched, next_wakeup))
2988                                 next_wakeup = elr->lr_next_sched;
2989                 }
2990                 mutex_unlock(&eli->li_list_mtx);
2991
2992                 try_to_freeze();
2993
2994                 cur = jiffies;
2995                 if ((time_after_eq(cur, next_wakeup)) ||
2996                     (MAX_JIFFY_OFFSET == next_wakeup)) {
2997                         cond_resched();
2998                         continue;
2999                 }
3000
3001                 schedule_timeout_interruptible(next_wakeup - cur);
3002
3003                 if (kthread_should_stop()) {
3004                         ext4_clear_request_list();
3005                         goto exit_thread;
3006                 }
3007         }
3008
3009 exit_thread:
3010         /*
3011          * It looks like the request list is empty, but we need
3012          * to check it under the li_list_mtx lock, to prevent any
3013          * additions into it, and of course we should lock ext4_li_mtx
3014          * to atomically free the list and ext4_li_info, because at
3015          * this point another ext4 filesystem could be registering
3016          * new one.
3017          */
3018         mutex_lock(&ext4_li_mtx);
3019         mutex_lock(&eli->li_list_mtx);
3020         if (!list_empty(&eli->li_request_list)) {
3021                 mutex_unlock(&eli->li_list_mtx);
3022                 mutex_unlock(&ext4_li_mtx);
3023                 goto cont_thread;
3024         }
3025         mutex_unlock(&eli->li_list_mtx);
3026         kfree(ext4_li_info);
3027         ext4_li_info = NULL;
3028         mutex_unlock(&ext4_li_mtx);
3029
3030         return 0;
3031 }
3032
3033 static void ext4_clear_request_list(void)
3034 {
3035         struct list_head *pos, *n;
3036         struct ext4_li_request *elr;
3037
3038         mutex_lock(&ext4_li_info->li_list_mtx);
3039         list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3040                 elr = list_entry(pos, struct ext4_li_request,
3041                                  lr_request);
3042                 ext4_remove_li_request(elr);
3043         }
3044         mutex_unlock(&ext4_li_info->li_list_mtx);
3045 }
3046
3047 static int ext4_run_lazyinit_thread(void)
3048 {
3049         ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3050                                          ext4_li_info, "ext4lazyinit");
3051         if (IS_ERR(ext4_lazyinit_task)) {
3052                 int err = PTR_ERR(ext4_lazyinit_task);
3053                 ext4_clear_request_list();
3054                 kfree(ext4_li_info);
3055                 ext4_li_info = NULL;
3056                 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
3057                                  "initialization thread\n",
3058                                  err);
3059                 return err;
3060         }
3061         ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
3062         return 0;
3063 }
3064
3065 /*
3066  * Check whether it make sense to run itable init. thread or not.
3067  * If there is at least one uninitialized inode table, return
3068  * corresponding group number, else the loop goes through all
3069  * groups and return total number of groups.
3070  */
3071 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3072 {
3073         ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3074         struct ext4_group_desc *gdp = NULL;
3075
3076         for (group = 0; group < ngroups; group++) {
3077                 gdp = ext4_get_group_desc(sb, group, NULL);
3078                 if (!gdp)
3079                         continue;
3080
3081                 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3082                         break;
3083         }
3084
3085         return group;
3086 }
3087
3088 static int ext4_li_info_new(void)
3089 {
3090         struct ext4_lazy_init *eli = NULL;
3091
3092         eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3093         if (!eli)
3094                 return -ENOMEM;
3095
3096         INIT_LIST_HEAD(&eli->li_request_list);
3097         mutex_init(&eli->li_list_mtx);
3098
3099         eli->li_state |= EXT4_LAZYINIT_QUIT;
3100
3101         ext4_li_info = eli;
3102
3103         return 0;
3104 }
3105
3106 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3107                                             ext4_group_t start)
3108 {
3109         struct ext4_sb_info *sbi = EXT4_SB(sb);
3110         struct ext4_li_request *elr;
3111
3112         elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3113         if (!elr)
3114                 return NULL;
3115
3116         elr->lr_super = sb;
3117         elr->lr_sbi = sbi;
3118         elr->lr_next_group = start;
3119
3120         /*
3121          * Randomize first schedule time of the request to
3122          * spread the inode table initialization requests
3123          * better.
3124          */
3125         elr->lr_next_sched = jiffies + (prandom_u32() %
3126                                 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
3127         return elr;
3128 }
3129
3130 int ext4_register_li_request(struct super_block *sb,
3131                              ext4_group_t first_not_zeroed)
3132 {
3133         struct ext4_sb_info *sbi = EXT4_SB(sb);
3134         struct ext4_li_request *elr = NULL;
3135         ext4_group_t ngroups = sbi->s_groups_count;
3136         int ret = 0;
3137
3138         mutex_lock(&ext4_li_mtx);
3139         if (sbi->s_li_request != NULL) {
3140                 /*
3141                  * Reset timeout so it can be computed again, because
3142                  * s_li_wait_mult might have changed.
3143                  */
3144                 sbi->s_li_request->lr_timeout = 0;
3145                 goto out;
3146         }
3147
3148         if (first_not_zeroed == ngroups || sb_rdonly(sb) ||
3149             !test_opt(sb, INIT_INODE_TABLE))
3150                 goto out;
3151
3152         elr = ext4_li_request_new(sb, first_not_zeroed);
3153         if (!elr) {
3154                 ret = -ENOMEM;
3155                 goto out;
3156         }
3157
3158         if (NULL == ext4_li_info) {
3159                 ret = ext4_li_info_new();
3160                 if (ret)
3161                         goto out;
3162         }
3163
3164         mutex_lock(&ext4_li_info->li_list_mtx);
3165         list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3166         mutex_unlock(&ext4_li_info->li_list_mtx);
3167
3168         sbi->s_li_request = elr;
3169         /*
3170          * set elr to NULL here since it has been inserted to
3171          * the request_list and the removal and free of it is
3172          * handled by ext4_clear_request_list from now on.
3173          */
3174         elr = NULL;
3175
3176         if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3177                 ret = ext4_run_lazyinit_thread();
3178                 if (ret)
3179                         goto out;
3180         }
3181 out:
3182         mutex_unlock(&ext4_li_mtx);
3183         if (ret)
3184                 kfree(elr);
3185         return ret;
3186 }
3187
3188 /*
3189  * We do not need to lock anything since this is called on
3190  * module unload.
3191  */
3192 static void ext4_destroy_lazyinit_thread(void)
3193 {
3194         /*
3195          * If thread exited earlier
3196          * there's nothing to be done.
3197          */
3198         if (!ext4_li_info || !ext4_lazyinit_task)
3199                 return;
3200
3201         kthread_stop(ext4_lazyinit_task);
3202 }
3203
3204 static int set_journal_csum_feature_set(struct super_block *sb)
3205 {
3206         int ret = 1;
3207         int compat, incompat;
3208         struct ext4_sb_info *sbi = EXT4_SB(sb);
3209
3210         if (ext4_has_metadata_csum(sb)) {
3211                 /* journal checksum v3 */
3212                 compat = 0;
3213                 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
3214         } else {
3215                 /* journal checksum v1 */
3216                 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3217                 incompat = 0;
3218         }
3219
3220         jbd2_journal_clear_features(sbi->s_journal,
3221                         JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3222                         JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3223                         JBD2_FEATURE_INCOMPAT_CSUM_V2);
3224         if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3225                 ret = jbd2_journal_set_features(sbi->s_journal,
3226                                 compat, 0,
3227                                 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |