ext4: Set flags on quota files directly
[sfrench/cifs-2.6.git] / fs / ext4 / super.c
1 /*
2  *  linux/fs/ext4/super.c
3  *
4  * Copyright (C) 1992, 1993, 1994, 1995
5  * Remy Card (card@masi.ibp.fr)
6  * Laboratoire MASI - Institut Blaise Pascal
7  * Universite Pierre et Marie Curie (Paris VI)
8  *
9  *  from
10  *
11  *  linux/fs/minix/inode.c
12  *
13  *  Copyright (C) 1991, 1992  Linus Torvalds
14  *
15  *  Big-endian to little-endian byte-swapping/bitmaps by
16  *        David S. Miller (davem@caip.rutgers.edu), 1995
17  */
18
19 #include <linux/module.h>
20 #include <linux/string.h>
21 #include <linux/fs.h>
22 #include <linux/time.h>
23 #include <linux/vmalloc.h>
24 #include <linux/slab.h>
25 #include <linux/init.h>
26 #include <linux/blkdev.h>
27 #include <linux/backing-dev.h>
28 #include <linux/parser.h>
29 #include <linux/buffer_head.h>
30 #include <linux/exportfs.h>
31 #include <linux/vfs.h>
32 #include <linux/random.h>
33 #include <linux/mount.h>
34 #include <linux/namei.h>
35 #include <linux/quotaops.h>
36 #include <linux/seq_file.h>
37 #include <linux/ctype.h>
38 #include <linux/log2.h>
39 #include <linux/crc16.h>
40 #include <linux/cleancache.h>
41 #include <linux/uaccess.h>
42
43 #include <linux/kthread.h>
44 #include <linux/freezer.h>
45
46 #include "ext4.h"
47 #include "ext4_extents.h"       /* Needed for trace points definition */
48 #include "ext4_jbd2.h"
49 #include "xattr.h"
50 #include "acl.h"
51 #include "mballoc.h"
52
53 #define CREATE_TRACE_POINTS
54 #include <trace/events/ext4.h>
55
56 static struct ext4_lazy_init *ext4_li_info;
57 static struct mutex ext4_li_mtx;
58 static struct ratelimit_state ext4_mount_msg_ratelimit;
59
60 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
61                              unsigned long journal_devnum);
62 static int ext4_show_options(struct seq_file *seq, struct dentry *root);
63 static int ext4_commit_super(struct super_block *sb, int sync);
64 static void ext4_mark_recovery_complete(struct super_block *sb,
65                                         struct ext4_super_block *es);
66 static void ext4_clear_journal_err(struct super_block *sb,
67                                    struct ext4_super_block *es);
68 static int ext4_sync_fs(struct super_block *sb, int wait);
69 static int ext4_remount(struct super_block *sb, int *flags, char *data);
70 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
71 static int ext4_unfreeze(struct super_block *sb);
72 static int ext4_freeze(struct super_block *sb);
73 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
74                        const char *dev_name, void *data);
75 static inline int ext2_feature_set_ok(struct super_block *sb);
76 static inline int ext3_feature_set_ok(struct super_block *sb);
77 static int ext4_feature_set_ok(struct super_block *sb, int readonly);
78 static void ext4_destroy_lazyinit_thread(void);
79 static void ext4_unregister_li_request(struct super_block *sb);
80 static void ext4_clear_request_list(void);
81 static struct inode *ext4_get_journal_inode(struct super_block *sb,
82                                             unsigned int journal_inum);
83
84 /*
85  * Lock ordering
86  *
87  * Note the difference between i_mmap_sem (EXT4_I(inode)->i_mmap_sem) and
88  * i_mmap_rwsem (inode->i_mmap_rwsem)!
89  *
90  * page fault path:
91  * mmap_sem -> sb_start_pagefault -> i_mmap_sem (r) -> transaction start ->
92  *   page lock -> i_data_sem (rw)
93  *
94  * buffered write path:
95  * sb_start_write -> i_mutex -> mmap_sem
96  * sb_start_write -> i_mutex -> transaction start -> page lock ->
97  *   i_data_sem (rw)
98  *
99  * truncate:
100  * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (w) -> i_mmap_sem (w) ->
101  *   i_mmap_rwsem (w) -> page lock
102  * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (w) -> i_mmap_sem (w) ->
103  *   transaction start -> i_data_sem (rw)
104  *
105  * direct IO:
106  * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (r) -> mmap_sem
107  * sb_start_write -> i_mutex -> EXT4_STATE_DIOREAD_LOCK (r) ->
108  *   transaction start -> i_data_sem (rw)
109  *
110  * writepages:
111  * transaction start -> page lock(s) -> i_data_sem (rw)
112  */
113
114 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT2)
115 static struct file_system_type ext2_fs_type = {
116         .owner          = THIS_MODULE,
117         .name           = "ext2",
118         .mount          = ext4_mount,
119         .kill_sb        = kill_block_super,
120         .fs_flags       = FS_REQUIRES_DEV,
121 };
122 MODULE_ALIAS_FS("ext2");
123 MODULE_ALIAS("ext2");
124 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
125 #else
126 #define IS_EXT2_SB(sb) (0)
127 #endif
128
129
130 static struct file_system_type ext3_fs_type = {
131         .owner          = THIS_MODULE,
132         .name           = "ext3",
133         .mount          = ext4_mount,
134         .kill_sb        = kill_block_super,
135         .fs_flags       = FS_REQUIRES_DEV,
136 };
137 MODULE_ALIAS_FS("ext3");
138 MODULE_ALIAS("ext3");
139 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
140
141 static int ext4_verify_csum_type(struct super_block *sb,
142                                  struct ext4_super_block *es)
143 {
144         if (!ext4_has_feature_metadata_csum(sb))
145                 return 1;
146
147         return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
148 }
149
150 static __le32 ext4_superblock_csum(struct super_block *sb,
151                                    struct ext4_super_block *es)
152 {
153         struct ext4_sb_info *sbi = EXT4_SB(sb);
154         int offset = offsetof(struct ext4_super_block, s_checksum);
155         __u32 csum;
156
157         csum = ext4_chksum(sbi, ~0, (char *)es, offset);
158
159         return cpu_to_le32(csum);
160 }
161
162 static int ext4_superblock_csum_verify(struct super_block *sb,
163                                        struct ext4_super_block *es)
164 {
165         if (!ext4_has_metadata_csum(sb))
166                 return 1;
167
168         return es->s_checksum == ext4_superblock_csum(sb, es);
169 }
170
171 void ext4_superblock_csum_set(struct super_block *sb)
172 {
173         struct ext4_super_block *es = EXT4_SB(sb)->s_es;
174
175         if (!ext4_has_metadata_csum(sb))
176                 return;
177
178         es->s_checksum = ext4_superblock_csum(sb, es);
179 }
180
181 void *ext4_kvmalloc(size_t size, gfp_t flags)
182 {
183         void *ret;
184
185         ret = kmalloc(size, flags | __GFP_NOWARN);
186         if (!ret)
187                 ret = __vmalloc(size, flags, PAGE_KERNEL);
188         return ret;
189 }
190
191 void *ext4_kvzalloc(size_t size, gfp_t flags)
192 {
193         void *ret;
194
195         ret = kzalloc(size, flags | __GFP_NOWARN);
196         if (!ret)
197                 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
198         return ret;
199 }
200
201 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
202                                struct ext4_group_desc *bg)
203 {
204         return le32_to_cpu(bg->bg_block_bitmap_lo) |
205                 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
206                  (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
207 }
208
209 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
210                                struct ext4_group_desc *bg)
211 {
212         return le32_to_cpu(bg->bg_inode_bitmap_lo) |
213                 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
214                  (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
215 }
216
217 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
218                               struct ext4_group_desc *bg)
219 {
220         return le32_to_cpu(bg->bg_inode_table_lo) |
221                 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
222                  (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
223 }
224
225 __u32 ext4_free_group_clusters(struct super_block *sb,
226                                struct ext4_group_desc *bg)
227 {
228         return le16_to_cpu(bg->bg_free_blocks_count_lo) |
229                 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
230                  (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
231 }
232
233 __u32 ext4_free_inodes_count(struct super_block *sb,
234                               struct ext4_group_desc *bg)
235 {
236         return le16_to_cpu(bg->bg_free_inodes_count_lo) |
237                 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
238                  (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
239 }
240
241 __u32 ext4_used_dirs_count(struct super_block *sb,
242                               struct ext4_group_desc *bg)
243 {
244         return le16_to_cpu(bg->bg_used_dirs_count_lo) |
245                 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
246                  (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
247 }
248
249 __u32 ext4_itable_unused_count(struct super_block *sb,
250                               struct ext4_group_desc *bg)
251 {
252         return le16_to_cpu(bg->bg_itable_unused_lo) |
253                 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
254                  (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
255 }
256
257 void ext4_block_bitmap_set(struct super_block *sb,
258                            struct ext4_group_desc *bg, ext4_fsblk_t blk)
259 {
260         bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
261         if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
262                 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
263 }
264
265 void ext4_inode_bitmap_set(struct super_block *sb,
266                            struct ext4_group_desc *bg, ext4_fsblk_t blk)
267 {
268         bg->bg_inode_bitmap_lo  = cpu_to_le32((u32)blk);
269         if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
270                 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
271 }
272
273 void ext4_inode_table_set(struct super_block *sb,
274                           struct ext4_group_desc *bg, ext4_fsblk_t blk)
275 {
276         bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
277         if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
278                 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
279 }
280
281 void ext4_free_group_clusters_set(struct super_block *sb,
282                                   struct ext4_group_desc *bg, __u32 count)
283 {
284         bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
285         if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
286                 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
287 }
288
289 void ext4_free_inodes_set(struct super_block *sb,
290                           struct ext4_group_desc *bg, __u32 count)
291 {
292         bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
293         if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
294                 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
295 }
296
297 void ext4_used_dirs_set(struct super_block *sb,
298                           struct ext4_group_desc *bg, __u32 count)
299 {
300         bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
301         if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
302                 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
303 }
304
305 void ext4_itable_unused_set(struct super_block *sb,
306                           struct ext4_group_desc *bg, __u32 count)
307 {
308         bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
309         if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
310                 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
311 }
312
313
314 static void __save_error_info(struct super_block *sb, const char *func,
315                             unsigned int line)
316 {
317         struct ext4_super_block *es = EXT4_SB(sb)->s_es;
318
319         EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
320         if (bdev_read_only(sb->s_bdev))
321                 return;
322         es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
323         es->s_last_error_time = cpu_to_le32(get_seconds());
324         strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
325         es->s_last_error_line = cpu_to_le32(line);
326         if (!es->s_first_error_time) {
327                 es->s_first_error_time = es->s_last_error_time;
328                 strncpy(es->s_first_error_func, func,
329                         sizeof(es->s_first_error_func));
330                 es->s_first_error_line = cpu_to_le32(line);
331                 es->s_first_error_ino = es->s_last_error_ino;
332                 es->s_first_error_block = es->s_last_error_block;
333         }
334         /*
335          * Start the daily error reporting function if it hasn't been
336          * started already
337          */
338         if (!es->s_error_count)
339                 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
340         le32_add_cpu(&es->s_error_count, 1);
341 }
342
343 static void save_error_info(struct super_block *sb, const char *func,
344                             unsigned int line)
345 {
346         __save_error_info(sb, func, line);
347         ext4_commit_super(sb, 1);
348 }
349
350 /*
351  * The del_gendisk() function uninitializes the disk-specific data
352  * structures, including the bdi structure, without telling anyone
353  * else.  Once this happens, any attempt to call mark_buffer_dirty()
354  * (for example, by ext4_commit_super), will cause a kernel OOPS.
355  * This is a kludge to prevent these oops until we can put in a proper
356  * hook in del_gendisk() to inform the VFS and file system layers.
357  */
358 static int block_device_ejected(struct super_block *sb)
359 {
360         struct inode *bd_inode = sb->s_bdev->bd_inode;
361         struct backing_dev_info *bdi = inode_to_bdi(bd_inode);
362
363         return bdi->dev == NULL;
364 }
365
366 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
367 {
368         struct super_block              *sb = journal->j_private;
369         struct ext4_sb_info             *sbi = EXT4_SB(sb);
370         int                             error = is_journal_aborted(journal);
371         struct ext4_journal_cb_entry    *jce;
372
373         BUG_ON(txn->t_state == T_FINISHED);
374         spin_lock(&sbi->s_md_lock);
375         while (!list_empty(&txn->t_private_list)) {
376                 jce = list_entry(txn->t_private_list.next,
377                                  struct ext4_journal_cb_entry, jce_list);
378                 list_del_init(&jce->jce_list);
379                 spin_unlock(&sbi->s_md_lock);
380                 jce->jce_func(sb, jce, error);
381                 spin_lock(&sbi->s_md_lock);
382         }
383         spin_unlock(&sbi->s_md_lock);
384 }
385
386 /* Deal with the reporting of failure conditions on a filesystem such as
387  * inconsistencies detected or read IO failures.
388  *
389  * On ext2, we can store the error state of the filesystem in the
390  * superblock.  That is not possible on ext4, because we may have other
391  * write ordering constraints on the superblock which prevent us from
392  * writing it out straight away; and given that the journal is about to
393  * be aborted, we can't rely on the current, or future, transactions to
394  * write out the superblock safely.
395  *
396  * We'll just use the jbd2_journal_abort() error code to record an error in
397  * the journal instead.  On recovery, the journal will complain about
398  * that error until we've noted it down and cleared it.
399  */
400
401 static void ext4_handle_error(struct super_block *sb)
402 {
403         if (sb->s_flags & MS_RDONLY)
404                 return;
405
406         if (!test_opt(sb, ERRORS_CONT)) {
407                 journal_t *journal = EXT4_SB(sb)->s_journal;
408
409                 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
410                 if (journal)
411                         jbd2_journal_abort(journal, -EIO);
412         }
413         if (test_opt(sb, ERRORS_RO)) {
414                 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
415                 /*
416                  * Make sure updated value of ->s_mount_flags will be visible
417                  * before ->s_flags update
418                  */
419                 smp_wmb();
420                 sb->s_flags |= MS_RDONLY;
421         }
422         if (test_opt(sb, ERRORS_PANIC)) {
423                 if (EXT4_SB(sb)->s_journal &&
424                   !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
425                         return;
426                 panic("EXT4-fs (device %s): panic forced after error\n",
427                         sb->s_id);
428         }
429 }
430
431 #define ext4_error_ratelimit(sb)                                        \
432                 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state),     \
433                              "EXT4-fs error")
434
435 void __ext4_error(struct super_block *sb, const char *function,
436                   unsigned int line, const char *fmt, ...)
437 {
438         struct va_format vaf;
439         va_list args;
440
441         if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
442                 return;
443
444         if (ext4_error_ratelimit(sb)) {
445                 va_start(args, fmt);
446                 vaf.fmt = fmt;
447                 vaf.va = &args;
448                 printk(KERN_CRIT
449                        "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
450                        sb->s_id, function, line, current->comm, &vaf);
451                 va_end(args);
452         }
453         save_error_info(sb, function, line);
454         ext4_handle_error(sb);
455 }
456
457 void __ext4_error_inode(struct inode *inode, const char *function,
458                         unsigned int line, ext4_fsblk_t block,
459                         const char *fmt, ...)
460 {
461         va_list args;
462         struct va_format vaf;
463         struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
464
465         if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
466                 return;
467
468         es->s_last_error_ino = cpu_to_le32(inode->i_ino);
469         es->s_last_error_block = cpu_to_le64(block);
470         if (ext4_error_ratelimit(inode->i_sb)) {
471                 va_start(args, fmt);
472                 vaf.fmt = fmt;
473                 vaf.va = &args;
474                 if (block)
475                         printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
476                                "inode #%lu: block %llu: comm %s: %pV\n",
477                                inode->i_sb->s_id, function, line, inode->i_ino,
478                                block, current->comm, &vaf);
479                 else
480                         printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
481                                "inode #%lu: comm %s: %pV\n",
482                                inode->i_sb->s_id, function, line, inode->i_ino,
483                                current->comm, &vaf);
484                 va_end(args);
485         }
486         save_error_info(inode->i_sb, function, line);
487         ext4_handle_error(inode->i_sb);
488 }
489
490 void __ext4_error_file(struct file *file, const char *function,
491                        unsigned int line, ext4_fsblk_t block,
492                        const char *fmt, ...)
493 {
494         va_list args;
495         struct va_format vaf;
496         struct ext4_super_block *es;
497         struct inode *inode = file_inode(file);
498         char pathname[80], *path;
499
500         if (unlikely(ext4_forced_shutdown(EXT4_SB(inode->i_sb))))
501                 return;
502
503         es = EXT4_SB(inode->i_sb)->s_es;
504         es->s_last_error_ino = cpu_to_le32(inode->i_ino);
505         if (ext4_error_ratelimit(inode->i_sb)) {
506                 path = file_path(file, pathname, sizeof(pathname));
507                 if (IS_ERR(path))
508                         path = "(unknown)";
509                 va_start(args, fmt);
510                 vaf.fmt = fmt;
511                 vaf.va = &args;
512                 if (block)
513                         printk(KERN_CRIT
514                                "EXT4-fs error (device %s): %s:%d: inode #%lu: "
515                                "block %llu: comm %s: path %s: %pV\n",
516                                inode->i_sb->s_id, function, line, inode->i_ino,
517                                block, current->comm, path, &vaf);
518                 else
519                         printk(KERN_CRIT
520                                "EXT4-fs error (device %s): %s:%d: inode #%lu: "
521                                "comm %s: path %s: %pV\n",
522                                inode->i_sb->s_id, function, line, inode->i_ino,
523                                current->comm, path, &vaf);
524                 va_end(args);
525         }
526         save_error_info(inode->i_sb, function, line);
527         ext4_handle_error(inode->i_sb);
528 }
529
530 const char *ext4_decode_error(struct super_block *sb, int errno,
531                               char nbuf[16])
532 {
533         char *errstr = NULL;
534
535         switch (errno) {
536         case -EFSCORRUPTED:
537                 errstr = "Corrupt filesystem";
538                 break;
539         case -EFSBADCRC:
540                 errstr = "Filesystem failed CRC";
541                 break;
542         case -EIO:
543                 errstr = "IO failure";
544                 break;
545         case -ENOMEM:
546                 errstr = "Out of memory";
547                 break;
548         case -EROFS:
549                 if (!sb || (EXT4_SB(sb)->s_journal &&
550                             EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
551                         errstr = "Journal has aborted";
552                 else
553                         errstr = "Readonly filesystem";
554                 break;
555         default:
556                 /* If the caller passed in an extra buffer for unknown
557                  * errors, textualise them now.  Else we just return
558                  * NULL. */
559                 if (nbuf) {
560                         /* Check for truncated error codes... */
561                         if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
562                                 errstr = nbuf;
563                 }
564                 break;
565         }
566
567         return errstr;
568 }
569
570 /* __ext4_std_error decodes expected errors from journaling functions
571  * automatically and invokes the appropriate error response.  */
572
573 void __ext4_std_error(struct super_block *sb, const char *function,
574                       unsigned int line, int errno)
575 {
576         char nbuf[16];
577         const char *errstr;
578
579         if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
580                 return;
581
582         /* Special case: if the error is EROFS, and we're not already
583          * inside a transaction, then there's really no point in logging
584          * an error. */
585         if (errno == -EROFS && journal_current_handle() == NULL &&
586             (sb->s_flags & MS_RDONLY))
587                 return;
588
589         if (ext4_error_ratelimit(sb)) {
590                 errstr = ext4_decode_error(sb, errno, nbuf);
591                 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
592                        sb->s_id, function, line, errstr);
593         }
594
595         save_error_info(sb, function, line);
596         ext4_handle_error(sb);
597 }
598
599 /*
600  * ext4_abort is a much stronger failure handler than ext4_error.  The
601  * abort function may be used to deal with unrecoverable failures such
602  * as journal IO errors or ENOMEM at a critical moment in log management.
603  *
604  * We unconditionally force the filesystem into an ABORT|READONLY state,
605  * unless the error response on the fs has been set to panic in which
606  * case we take the easy way out and panic immediately.
607  */
608
609 void __ext4_abort(struct super_block *sb, const char *function,
610                 unsigned int line, const char *fmt, ...)
611 {
612         struct va_format vaf;
613         va_list args;
614
615         if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
616                 return;
617
618         save_error_info(sb, function, line);
619         va_start(args, fmt);
620         vaf.fmt = fmt;
621         vaf.va = &args;
622         printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: %pV\n",
623                sb->s_id, function, line, &vaf);
624         va_end(args);
625
626         if ((sb->s_flags & MS_RDONLY) == 0) {
627                 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
628                 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
629                 /*
630                  * Make sure updated value of ->s_mount_flags will be visible
631                  * before ->s_flags update
632                  */
633                 smp_wmb();
634                 sb->s_flags |= MS_RDONLY;
635                 if (EXT4_SB(sb)->s_journal)
636                         jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
637                 save_error_info(sb, function, line);
638         }
639         if (test_opt(sb, ERRORS_PANIC)) {
640                 if (EXT4_SB(sb)->s_journal &&
641                   !(EXT4_SB(sb)->s_journal->j_flags & JBD2_REC_ERR))
642                         return;
643                 panic("EXT4-fs panic from previous error\n");
644         }
645 }
646
647 void __ext4_msg(struct super_block *sb,
648                 const char *prefix, const char *fmt, ...)
649 {
650         struct va_format vaf;
651         va_list args;
652
653         if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
654                 return;
655
656         va_start(args, fmt);
657         vaf.fmt = fmt;
658         vaf.va = &args;
659         printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
660         va_end(args);
661 }
662
663 #define ext4_warning_ratelimit(sb)                                      \
664                 ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state), \
665                              "EXT4-fs warning")
666
667 void __ext4_warning(struct super_block *sb, const char *function,
668                     unsigned int line, const char *fmt, ...)
669 {
670         struct va_format vaf;
671         va_list args;
672
673         if (!ext4_warning_ratelimit(sb))
674                 return;
675
676         va_start(args, fmt);
677         vaf.fmt = fmt;
678         vaf.va = &args;
679         printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
680                sb->s_id, function, line, &vaf);
681         va_end(args);
682 }
683
684 void __ext4_warning_inode(const struct inode *inode, const char *function,
685                           unsigned int line, const char *fmt, ...)
686 {
687         struct va_format vaf;
688         va_list args;
689
690         if (!ext4_warning_ratelimit(inode->i_sb))
691                 return;
692
693         va_start(args, fmt);
694         vaf.fmt = fmt;
695         vaf.va = &args;
696         printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
697                "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
698                function, line, inode->i_ino, current->comm, &vaf);
699         va_end(args);
700 }
701
702 void __ext4_grp_locked_error(const char *function, unsigned int line,
703                              struct super_block *sb, ext4_group_t grp,
704                              unsigned long ino, ext4_fsblk_t block,
705                              const char *fmt, ...)
706 __releases(bitlock)
707 __acquires(bitlock)
708 {
709         struct va_format vaf;
710         va_list args;
711         struct ext4_super_block *es = EXT4_SB(sb)->s_es;
712
713         if (unlikely(ext4_forced_shutdown(EXT4_SB(sb))))
714                 return;
715
716         es->s_last_error_ino = cpu_to_le32(ino);
717         es->s_last_error_block = cpu_to_le64(block);
718         __save_error_info(sb, function, line);
719
720         if (ext4_error_ratelimit(sb)) {
721                 va_start(args, fmt);
722                 vaf.fmt = fmt;
723                 vaf.va = &args;
724                 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
725                        sb->s_id, function, line, grp);
726                 if (ino)
727                         printk(KERN_CONT "inode %lu: ", ino);
728                 if (block)
729                         printk(KERN_CONT "block %llu:",
730                                (unsigned long long) block);
731                 printk(KERN_CONT "%pV\n", &vaf);
732                 va_end(args);
733         }
734
735         if (test_opt(sb, ERRORS_CONT)) {
736                 ext4_commit_super(sb, 0);
737                 return;
738         }
739
740         ext4_unlock_group(sb, grp);
741         ext4_handle_error(sb);
742         /*
743          * We only get here in the ERRORS_RO case; relocking the group
744          * may be dangerous, but nothing bad will happen since the
745          * filesystem will have already been marked read/only and the
746          * journal has been aborted.  We return 1 as a hint to callers
747          * who might what to use the return value from
748          * ext4_grp_locked_error() to distinguish between the
749          * ERRORS_CONT and ERRORS_RO case, and perhaps return more
750          * aggressively from the ext4 function in question, with a
751          * more appropriate error code.
752          */
753         ext4_lock_group(sb, grp);
754         return;
755 }
756
757 void ext4_update_dynamic_rev(struct super_block *sb)
758 {
759         struct ext4_super_block *es = EXT4_SB(sb)->s_es;
760
761         if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
762                 return;
763
764         ext4_warning(sb,
765                      "updating to rev %d because of new feature flag, "
766                      "running e2fsck is recommended",
767                      EXT4_DYNAMIC_REV);
768
769         es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
770         es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
771         es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
772         /* leave es->s_feature_*compat flags alone */
773         /* es->s_uuid will be set by e2fsck if empty */
774
775         /*
776          * The rest of the superblock fields should be zero, and if not it
777          * means they are likely already in use, so leave them alone.  We
778          * can leave it up to e2fsck to clean up any inconsistencies there.
779          */
780 }
781
782 /*
783  * Open the external journal device
784  */
785 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
786 {
787         struct block_device *bdev;
788         char b[BDEVNAME_SIZE];
789
790         bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
791         if (IS_ERR(bdev))
792                 goto fail;
793         return bdev;
794
795 fail:
796         ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
797                         __bdevname(dev, b), PTR_ERR(bdev));
798         return NULL;
799 }
800
801 /*
802  * Release the journal device
803  */
804 static void ext4_blkdev_put(struct block_device *bdev)
805 {
806         blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
807 }
808
809 static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
810 {
811         struct block_device *bdev;
812         bdev = sbi->journal_bdev;
813         if (bdev) {
814                 ext4_blkdev_put(bdev);
815                 sbi->journal_bdev = NULL;
816         }
817 }
818
819 static inline struct inode *orphan_list_entry(struct list_head *l)
820 {
821         return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
822 }
823
824 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
825 {
826         struct list_head *l;
827
828         ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
829                  le32_to_cpu(sbi->s_es->s_last_orphan));
830
831         printk(KERN_ERR "sb_info orphan list:\n");
832         list_for_each(l, &sbi->s_orphan) {
833                 struct inode *inode = orphan_list_entry(l);
834                 printk(KERN_ERR "  "
835                        "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
836                        inode->i_sb->s_id, inode->i_ino, inode,
837                        inode->i_mode, inode->i_nlink,
838                        NEXT_ORPHAN(inode));
839         }
840 }
841
842 #ifdef CONFIG_QUOTA
843 static int ext4_quota_off(struct super_block *sb, int type);
844
845 static inline void ext4_quota_off_umount(struct super_block *sb)
846 {
847         int type;
848
849         if (ext4_has_feature_quota(sb)) {
850                 dquot_disable(sb, -1,
851                               DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
852         } else {
853                 /* Use our quota_off function to clear inode flags etc. */
854                 for (type = 0; type < EXT4_MAXQUOTAS; type++)
855                         ext4_quota_off(sb, type);
856         }
857 }
858 #else
859 static inline void ext4_quota_off_umount(struct super_block *sb)
860 {
861 }
862 #endif
863
864 static void ext4_put_super(struct super_block *sb)
865 {
866         struct ext4_sb_info *sbi = EXT4_SB(sb);
867         struct ext4_super_block *es = sbi->s_es;
868         int aborted = 0;
869         int i, err;
870
871         ext4_unregister_li_request(sb);
872         ext4_quota_off_umount(sb);
873
874         flush_workqueue(sbi->rsv_conversion_wq);
875         destroy_workqueue(sbi->rsv_conversion_wq);
876
877         if (sbi->s_journal) {
878                 aborted = is_journal_aborted(sbi->s_journal);
879                 err = jbd2_journal_destroy(sbi->s_journal);
880                 sbi->s_journal = NULL;
881                 if ((err < 0) && !aborted)
882                         ext4_abort(sb, "Couldn't clean up the journal");
883         }
884
885         ext4_unregister_sysfs(sb);
886         ext4_es_unregister_shrinker(sbi);
887         del_timer_sync(&sbi->s_err_report);
888         ext4_release_system_zone(sb);
889         ext4_mb_release(sb);
890         ext4_ext_release(sb);
891
892         if (!(sb->s_flags & MS_RDONLY) && !aborted) {
893                 ext4_clear_feature_journal_needs_recovery(sb);
894                 es->s_state = cpu_to_le16(sbi->s_mount_state);
895         }
896         if (!(sb->s_flags & MS_RDONLY))
897                 ext4_commit_super(sb, 1);
898
899         for (i = 0; i < sbi->s_gdb_count; i++)
900                 brelse(sbi->s_group_desc[i]);
901         kvfree(sbi->s_group_desc);
902         kvfree(sbi->s_flex_groups);
903         percpu_counter_destroy(&sbi->s_freeclusters_counter);
904         percpu_counter_destroy(&sbi->s_freeinodes_counter);
905         percpu_counter_destroy(&sbi->s_dirs_counter);
906         percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
907         percpu_free_rwsem(&sbi->s_journal_flag_rwsem);
908 #ifdef CONFIG_QUOTA
909         for (i = 0; i < EXT4_MAXQUOTAS; i++)
910                 kfree(sbi->s_qf_names[i]);
911 #endif
912
913         /* Debugging code just in case the in-memory inode orphan list
914          * isn't empty.  The on-disk one can be non-empty if we've
915          * detected an error and taken the fs readonly, but the
916          * in-memory list had better be clean by this point. */
917         if (!list_empty(&sbi->s_orphan))
918                 dump_orphan_list(sb, sbi);
919         J_ASSERT(list_empty(&sbi->s_orphan));
920
921         sync_blockdev(sb->s_bdev);
922         invalidate_bdev(sb->s_bdev);
923         if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
924                 /*
925                  * Invalidate the journal device's buffers.  We don't want them
926                  * floating about in memory - the physical journal device may
927                  * hotswapped, and it breaks the `ro-after' testing code.
928                  */
929                 sync_blockdev(sbi->journal_bdev);
930                 invalidate_bdev(sbi->journal_bdev);
931                 ext4_blkdev_remove(sbi);
932         }
933         if (sbi->s_mb_cache) {
934                 ext4_xattr_destroy_cache(sbi->s_mb_cache);
935                 sbi->s_mb_cache = NULL;
936         }
937         if (sbi->s_mmp_tsk)
938                 kthread_stop(sbi->s_mmp_tsk);
939         brelse(sbi->s_sbh);
940         sb->s_fs_info = NULL;
941         /*
942          * Now that we are completely done shutting down the
943          * superblock, we need to actually destroy the kobject.
944          */
945         kobject_put(&sbi->s_kobj);
946         wait_for_completion(&sbi->s_kobj_unregister);
947         if (sbi->s_chksum_driver)
948                 crypto_free_shash(sbi->s_chksum_driver);
949         kfree(sbi->s_blockgroup_lock);
950         kfree(sbi);
951 }
952
953 static struct kmem_cache *ext4_inode_cachep;
954
955 /*
956  * Called inside transaction, so use GFP_NOFS
957  */
958 static struct inode *ext4_alloc_inode(struct super_block *sb)
959 {
960         struct ext4_inode_info *ei;
961
962         ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
963         if (!ei)
964                 return NULL;
965
966         ei->vfs_inode.i_version = 1;
967         spin_lock_init(&ei->i_raw_lock);
968         INIT_LIST_HEAD(&ei->i_prealloc_list);
969         spin_lock_init(&ei->i_prealloc_lock);
970         ext4_es_init_tree(&ei->i_es_tree);
971         rwlock_init(&ei->i_es_lock);
972         INIT_LIST_HEAD(&ei->i_es_list);
973         ei->i_es_all_nr = 0;
974         ei->i_es_shk_nr = 0;
975         ei->i_es_shrink_lblk = 0;
976         ei->i_reserved_data_blocks = 0;
977         ei->i_reserved_meta_blocks = 0;
978         ei->i_allocated_meta_blocks = 0;
979         ei->i_da_metadata_calc_len = 0;
980         ei->i_da_metadata_calc_last_lblock = 0;
981         spin_lock_init(&(ei->i_block_reservation_lock));
982 #ifdef CONFIG_QUOTA
983         ei->i_reserved_quota = 0;
984         memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
985 #endif
986         ei->jinode = NULL;
987         INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
988         spin_lock_init(&ei->i_completed_io_lock);
989         ei->i_sync_tid = 0;
990         ei->i_datasync_tid = 0;
991         atomic_set(&ei->i_unwritten, 0);
992         INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
993         return &ei->vfs_inode;
994 }
995
996 static int ext4_drop_inode(struct inode *inode)
997 {
998         int drop = generic_drop_inode(inode);
999
1000         trace_ext4_drop_inode(inode, drop);
1001         return drop;
1002 }
1003
1004 static void ext4_i_callback(struct rcu_head *head)
1005 {
1006         struct inode *inode = container_of(head, struct inode, i_rcu);
1007         kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
1008 }
1009
1010 static void ext4_destroy_inode(struct inode *inode)
1011 {
1012         if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
1013                 ext4_msg(inode->i_sb, KERN_ERR,
1014                          "Inode %lu (%p): orphan list check failed!",
1015                          inode->i_ino, EXT4_I(inode));
1016                 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
1017                                 EXT4_I(inode), sizeof(struct ext4_inode_info),
1018                                 true);
1019                 dump_stack();
1020         }
1021         call_rcu(&inode->i_rcu, ext4_i_callback);
1022 }
1023
1024 static void init_once(void *foo)
1025 {
1026         struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
1027
1028         INIT_LIST_HEAD(&ei->i_orphan);
1029         init_rwsem(&ei->xattr_sem);
1030         init_rwsem(&ei->i_data_sem);
1031         init_rwsem(&ei->i_mmap_sem);
1032         inode_init_once(&ei->vfs_inode);
1033 }
1034
1035 static int __init init_inodecache(void)
1036 {
1037         ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
1038                                              sizeof(struct ext4_inode_info),
1039                                              0, (SLAB_RECLAIM_ACCOUNT|
1040                                                 SLAB_MEM_SPREAD|SLAB_ACCOUNT),
1041                                              init_once);
1042         if (ext4_inode_cachep == NULL)
1043                 return -ENOMEM;
1044         return 0;
1045 }
1046
1047 static void destroy_inodecache(void)
1048 {
1049         /*
1050          * Make sure all delayed rcu free inodes are flushed before we
1051          * destroy cache.
1052          */
1053         rcu_barrier();
1054         kmem_cache_destroy(ext4_inode_cachep);
1055 }
1056
1057 void ext4_clear_inode(struct inode *inode)
1058 {
1059         invalidate_inode_buffers(inode);
1060         clear_inode(inode);
1061         dquot_drop(inode);
1062         ext4_discard_preallocations(inode);
1063         ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
1064         if (EXT4_I(inode)->jinode) {
1065                 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1066                                                EXT4_I(inode)->jinode);
1067                 jbd2_free_inode(EXT4_I(inode)->jinode);
1068                 EXT4_I(inode)->jinode = NULL;
1069         }
1070 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1071         fscrypt_put_encryption_info(inode, NULL);
1072 #endif
1073 }
1074
1075 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1076                                         u64 ino, u32 generation)
1077 {
1078         struct inode *inode;
1079
1080         if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
1081                 return ERR_PTR(-ESTALE);
1082         if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
1083                 return ERR_PTR(-ESTALE);
1084
1085         /* iget isn't really right if the inode is currently unallocated!!
1086          *
1087          * ext4_read_inode will return a bad_inode if the inode had been
1088          * deleted, so we should be safe.
1089          *
1090          * Currently we don't know the generation for parent directory, so
1091          * a generation of 0 means "accept any"
1092          */
1093         inode = ext4_iget_normal(sb, ino);
1094         if (IS_ERR(inode))
1095                 return ERR_CAST(inode);
1096         if (generation && inode->i_generation != generation) {
1097                 iput(inode);
1098                 return ERR_PTR(-ESTALE);
1099         }
1100
1101         return inode;
1102 }
1103
1104 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1105                                         int fh_len, int fh_type)
1106 {
1107         return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1108                                     ext4_nfs_get_inode);
1109 }
1110
1111 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1112                                         int fh_len, int fh_type)
1113 {
1114         return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1115                                     ext4_nfs_get_inode);
1116 }
1117
1118 /*
1119  * Try to release metadata pages (indirect blocks, directories) which are
1120  * mapped via the block device.  Since these pages could have journal heads
1121  * which would prevent try_to_free_buffers() from freeing them, we must use
1122  * jbd2 layer's try_to_free_buffers() function to release them.
1123  */
1124 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1125                                  gfp_t wait)
1126 {
1127         journal_t *journal = EXT4_SB(sb)->s_journal;
1128
1129         WARN_ON(PageChecked(page));
1130         if (!page_has_buffers(page))
1131                 return 0;
1132         if (journal)
1133                 return jbd2_journal_try_to_free_buffers(journal, page,
1134                                                 wait & ~__GFP_DIRECT_RECLAIM);
1135         return try_to_free_buffers(page);
1136 }
1137
1138 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1139 static int ext4_get_context(struct inode *inode, void *ctx, size_t len)
1140 {
1141         return ext4_xattr_get(inode, EXT4_XATTR_INDEX_ENCRYPTION,
1142                                  EXT4_XATTR_NAME_ENCRYPTION_CONTEXT, ctx, len);
1143 }
1144
1145 static int ext4_prepare_context(struct inode *inode)
1146 {
1147         return ext4_convert_inline_data(inode);
1148 }
1149
1150 static int ext4_set_context(struct inode *inode, const void *ctx, size_t len,
1151                                                         void *fs_data)
1152 {
1153         handle_t *handle = fs_data;
1154         int res, res2, retries = 0;
1155
1156         /*
1157          * If a journal handle was specified, then the encryption context is
1158          * being set on a new inode via inheritance and is part of a larger
1159          * transaction to create the inode.  Otherwise the encryption context is
1160          * being set on an existing inode in its own transaction.  Only in the
1161          * latter case should the "retry on ENOSPC" logic be used.
1162          */
1163
1164         if (handle) {
1165                 res = ext4_xattr_set_handle(handle, inode,
1166                                             EXT4_XATTR_INDEX_ENCRYPTION,
1167                                             EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1168                                             ctx, len, 0);
1169                 if (!res) {
1170                         ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1171                         ext4_clear_inode_state(inode,
1172                                         EXT4_STATE_MAY_INLINE_DATA);
1173                         /*
1174                          * Update inode->i_flags - e.g. S_DAX may get disabled
1175                          */
1176                         ext4_set_inode_flags(inode);
1177                 }
1178                 return res;
1179         }
1180
1181 retry:
1182         handle = ext4_journal_start(inode, EXT4_HT_MISC,
1183                         ext4_jbd2_credits_xattr(inode));
1184         if (IS_ERR(handle))
1185                 return PTR_ERR(handle);
1186
1187         res = ext4_xattr_set_handle(handle, inode, EXT4_XATTR_INDEX_ENCRYPTION,
1188                                     EXT4_XATTR_NAME_ENCRYPTION_CONTEXT,
1189                                     ctx, len, 0);
1190         if (!res) {
1191                 ext4_set_inode_flag(inode, EXT4_INODE_ENCRYPT);
1192                 /* Update inode->i_flags - e.g. S_DAX may get disabled */
1193                 ext4_set_inode_flags(inode);
1194                 res = ext4_mark_inode_dirty(handle, inode);
1195                 if (res)
1196                         EXT4_ERROR_INODE(inode, "Failed to mark inode dirty");
1197         }
1198         res2 = ext4_journal_stop(handle);
1199
1200         if (res == -ENOSPC && ext4_should_retry_alloc(inode->i_sb, &retries))
1201                 goto retry;
1202         if (!res)
1203                 res = res2;
1204         return res;
1205 }
1206
1207 static int ext4_dummy_context(struct inode *inode)
1208 {
1209         return DUMMY_ENCRYPTION_ENABLED(EXT4_SB(inode->i_sb));
1210 }
1211
1212 static unsigned ext4_max_namelen(struct inode *inode)
1213 {
1214         return S_ISLNK(inode->i_mode) ? inode->i_sb->s_blocksize :
1215                 EXT4_NAME_LEN;
1216 }
1217
1218 static const struct fscrypt_operations ext4_cryptops = {
1219         .key_prefix             = "ext4:",
1220         .get_context            = ext4_get_context,
1221         .prepare_context        = ext4_prepare_context,
1222         .set_context            = ext4_set_context,
1223         .dummy_context          = ext4_dummy_context,
1224         .is_encrypted           = ext4_encrypted_inode,
1225         .empty_dir              = ext4_empty_dir,
1226         .max_namelen            = ext4_max_namelen,
1227 };
1228 #else
1229 static const struct fscrypt_operations ext4_cryptops = {
1230         .is_encrypted           = ext4_encrypted_inode,
1231 };
1232 #endif
1233
1234 #ifdef CONFIG_QUOTA
1235 static char *quotatypes[] = INITQFNAMES;
1236 #define QTYPE2NAME(t) (quotatypes[t])
1237
1238 static int ext4_write_dquot(struct dquot *dquot);
1239 static int ext4_acquire_dquot(struct dquot *dquot);
1240 static int ext4_release_dquot(struct dquot *dquot);
1241 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1242 static int ext4_write_info(struct super_block *sb, int type);
1243 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1244                          const struct path *path);
1245 static int ext4_quota_on_mount(struct super_block *sb, int type);
1246 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1247                                size_t len, loff_t off);
1248 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1249                                 const char *data, size_t len, loff_t off);
1250 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1251                              unsigned int flags);
1252 static int ext4_enable_quotas(struct super_block *sb);
1253 static int ext4_get_next_id(struct super_block *sb, struct kqid *qid);
1254
1255 static struct dquot **ext4_get_dquots(struct inode *inode)
1256 {
1257         return EXT4_I(inode)->i_dquot;
1258 }
1259
1260 static const struct dquot_operations ext4_quota_operations = {
1261         .get_reserved_space = ext4_get_reserved_space,
1262         .write_dquot    = ext4_write_dquot,
1263         .acquire_dquot  = ext4_acquire_dquot,
1264         .release_dquot  = ext4_release_dquot,
1265         .mark_dirty     = ext4_mark_dquot_dirty,
1266         .write_info     = ext4_write_info,
1267         .alloc_dquot    = dquot_alloc,
1268         .destroy_dquot  = dquot_destroy,
1269         .get_projid     = ext4_get_projid,
1270         .get_next_id    = ext4_get_next_id,
1271 };
1272
1273 static const struct quotactl_ops ext4_qctl_operations = {
1274         .quota_on       = ext4_quota_on,
1275         .quota_off      = ext4_quota_off,
1276         .quota_sync     = dquot_quota_sync,
1277         .get_state      = dquot_get_state,
1278         .set_info       = dquot_set_dqinfo,
1279         .get_dqblk      = dquot_get_dqblk,
1280         .set_dqblk      = dquot_set_dqblk,
1281         .get_nextdqblk  = dquot_get_next_dqblk,
1282 };
1283 #endif
1284
1285 static const struct super_operations ext4_sops = {
1286         .alloc_inode    = ext4_alloc_inode,
1287         .destroy_inode  = ext4_destroy_inode,
1288         .write_inode    = ext4_write_inode,
1289         .dirty_inode    = ext4_dirty_inode,
1290         .drop_inode     = ext4_drop_inode,
1291         .evict_inode    = ext4_evict_inode,
1292         .put_super      = ext4_put_super,
1293         .sync_fs        = ext4_sync_fs,
1294         .freeze_fs      = ext4_freeze,
1295         .unfreeze_fs    = ext4_unfreeze,
1296         .statfs         = ext4_statfs,
1297         .remount_fs     = ext4_remount,
1298         .show_options   = ext4_show_options,
1299 #ifdef CONFIG_QUOTA
1300         .quota_read     = ext4_quota_read,
1301         .quota_write    = ext4_quota_write,
1302         .get_dquots     = ext4_get_dquots,
1303 #endif
1304         .bdev_try_to_free_page = bdev_try_to_free_page,
1305 };
1306
1307 static const struct export_operations ext4_export_ops = {
1308         .fh_to_dentry = ext4_fh_to_dentry,
1309         .fh_to_parent = ext4_fh_to_parent,
1310         .get_parent = ext4_get_parent,
1311 };
1312
1313 enum {
1314         Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1315         Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1316         Opt_nouid32, Opt_debug, Opt_removed,
1317         Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1318         Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1319         Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1320         Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1321         Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1322         Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
1323         Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1324         Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1325         Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1326         Opt_usrquota, Opt_grpquota, Opt_prjquota, Opt_i_version, Opt_dax,
1327         Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1328         Opt_lazytime, Opt_nolazytime, Opt_debug_want_extra_isize,
1329         Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1330         Opt_inode_readahead_blks, Opt_journal_ioprio,
1331         Opt_dioread_nolock, Opt_dioread_lock,
1332         Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1333         Opt_max_dir_size_kb, Opt_nojournal_checksum,
1334 };
1335
1336 static const match_table_t tokens = {
1337         {Opt_bsd_df, "bsddf"},
1338         {Opt_minix_df, "minixdf"},
1339         {Opt_grpid, "grpid"},
1340         {Opt_grpid, "bsdgroups"},
1341         {Opt_nogrpid, "nogrpid"},
1342         {Opt_nogrpid, "sysvgroups"},
1343         {Opt_resgid, "resgid=%u"},
1344         {Opt_resuid, "resuid=%u"},
1345         {Opt_sb, "sb=%u"},
1346         {Opt_err_cont, "errors=continue"},
1347         {Opt_err_panic, "errors=panic"},
1348         {Opt_err_ro, "errors=remount-ro"},
1349         {Opt_nouid32, "nouid32"},
1350         {Opt_debug, "debug"},
1351         {Opt_removed, "oldalloc"},
1352         {Opt_removed, "orlov"},
1353         {Opt_user_xattr, "user_xattr"},
1354         {Opt_nouser_xattr, "nouser_xattr"},
1355         {Opt_acl, "acl"},
1356         {Opt_noacl, "noacl"},
1357         {Opt_noload, "norecovery"},
1358         {Opt_noload, "noload"},
1359         {Opt_removed, "nobh"},
1360         {Opt_removed, "bh"},
1361         {Opt_commit, "commit=%u"},
1362         {Opt_min_batch_time, "min_batch_time=%u"},
1363         {Opt_max_batch_time, "max_batch_time=%u"},
1364         {Opt_journal_dev, "journal_dev=%u"},
1365         {Opt_journal_path, "journal_path=%s"},
1366         {Opt_journal_checksum, "journal_checksum"},
1367         {Opt_nojournal_checksum, "nojournal_checksum"},
1368         {Opt_journal_async_commit, "journal_async_commit"},
1369         {Opt_abort, "abort"},
1370         {Opt_data_journal, "data=journal"},
1371         {Opt_data_ordered, "data=ordered"},
1372         {Opt_data_writeback, "data=writeback"},
1373         {Opt_data_err_abort, "data_err=abort"},
1374         {Opt_data_err_ignore, "data_err=ignore"},
1375         {Opt_offusrjquota, "usrjquota="},
1376         {Opt_usrjquota, "usrjquota=%s"},
1377         {Opt_offgrpjquota, "grpjquota="},
1378         {Opt_grpjquota, "grpjquota=%s"},
1379         {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1380         {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1381         {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1382         {Opt_grpquota, "grpquota"},
1383         {Opt_noquota, "noquota"},
1384         {Opt_quota, "quota"},
1385         {Opt_usrquota, "usrquota"},
1386         {Opt_prjquota, "prjquota"},
1387         {Opt_barrier, "barrier=%u"},
1388         {Opt_barrier, "barrier"},
1389         {Opt_nobarrier, "nobarrier"},
1390         {Opt_i_version, "i_version"},
1391         {Opt_dax, "dax"},
1392         {Opt_stripe, "stripe=%u"},
1393         {Opt_delalloc, "delalloc"},
1394         {Opt_lazytime, "lazytime"},
1395         {Opt_nolazytime, "nolazytime"},
1396         {Opt_debug_want_extra_isize, "debug_want_extra_isize=%u"},
1397         {Opt_nodelalloc, "nodelalloc"},
1398         {Opt_removed, "mblk_io_submit"},
1399         {Opt_removed, "nomblk_io_submit"},
1400         {Opt_block_validity, "block_validity"},
1401         {Opt_noblock_validity, "noblock_validity"},
1402         {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1403         {Opt_journal_ioprio, "journal_ioprio=%u"},
1404         {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1405         {Opt_auto_da_alloc, "auto_da_alloc"},
1406         {Opt_noauto_da_alloc, "noauto_da_alloc"},
1407         {Opt_dioread_nolock, "dioread_nolock"},
1408         {Opt_dioread_lock, "dioread_lock"},
1409         {Opt_discard, "discard"},
1410         {Opt_nodiscard, "nodiscard"},
1411         {Opt_init_itable, "init_itable=%u"},
1412         {Opt_init_itable, "init_itable"},
1413         {Opt_noinit_itable, "noinit_itable"},
1414         {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1415         {Opt_test_dummy_encryption, "test_dummy_encryption"},
1416         {Opt_removed, "check=none"},    /* mount option from ext2/3 */
1417         {Opt_removed, "nocheck"},       /* mount option from ext2/3 */
1418         {Opt_removed, "reservation"},   /* mount option from ext2/3 */
1419         {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1420         {Opt_removed, "journal=%u"},    /* mount option from ext2/3 */
1421         {Opt_err, NULL},
1422 };
1423
1424 static ext4_fsblk_t get_sb_block(void **data)
1425 {
1426         ext4_fsblk_t    sb_block;
1427         char            *options = (char *) *data;
1428
1429         if (!options || strncmp(options, "sb=", 3) != 0)
1430                 return 1;       /* Default location */
1431
1432         options += 3;
1433         /* TODO: use simple_strtoll with >32bit ext4 */
1434         sb_block = simple_strtoul(options, &options, 0);
1435         if (*options && *options != ',') {
1436                 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1437                        (char *) *data);
1438                 return 1;
1439         }
1440         if (*options == ',')
1441                 options++;
1442         *data = (void *) options;
1443
1444         return sb_block;
1445 }
1446
1447 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1448 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1449         "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1450
1451 #ifdef CONFIG_QUOTA
1452 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1453 {
1454         struct ext4_sb_info *sbi = EXT4_SB(sb);
1455         char *qname;
1456         int ret = -1;
1457
1458         if (sb_any_quota_loaded(sb) &&
1459                 !sbi->s_qf_names[qtype]) {
1460                 ext4_msg(sb, KERN_ERR,
1461                         "Cannot change journaled "
1462                         "quota options when quota turned on");
1463                 return -1;
1464         }
1465         if (ext4_has_feature_quota(sb)) {
1466                 ext4_msg(sb, KERN_INFO, "Journaled quota options "
1467                          "ignored when QUOTA feature is enabled");
1468                 return 1;
1469         }
1470         qname = match_strdup(args);
1471         if (!qname) {
1472                 ext4_msg(sb, KERN_ERR,
1473                         "Not enough memory for storing quotafile name");
1474                 return -1;
1475         }
1476         if (sbi->s_qf_names[qtype]) {
1477                 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1478                         ret = 1;
1479                 else
1480                         ext4_msg(sb, KERN_ERR,
1481                                  "%s quota file already specified",
1482                                  QTYPE2NAME(qtype));
1483                 goto errout;
1484         }
1485         if (strchr(qname, '/')) {
1486                 ext4_msg(sb, KERN_ERR,
1487                         "quotafile must be on filesystem root");
1488                 goto errout;
1489         }
1490         sbi->s_qf_names[qtype] = qname;
1491         set_opt(sb, QUOTA);
1492         return 1;
1493 errout:
1494         kfree(qname);
1495         return ret;
1496 }
1497
1498 static int clear_qf_name(struct super_block *sb, int qtype)
1499 {
1500
1501         struct ext4_sb_info *sbi = EXT4_SB(sb);
1502
1503         if (sb_any_quota_loaded(sb) &&
1504                 sbi->s_qf_names[qtype]) {
1505                 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1506                         " when quota turned on");
1507                 return -1;
1508         }
1509         kfree(sbi->s_qf_names[qtype]);
1510         sbi->s_qf_names[qtype] = NULL;
1511         return 1;
1512 }
1513 #endif
1514
1515 #define MOPT_SET        0x0001
1516 #define MOPT_CLEAR      0x0002
1517 #define MOPT_NOSUPPORT  0x0004
1518 #define MOPT_EXPLICIT   0x0008
1519 #define MOPT_CLEAR_ERR  0x0010
1520 #define MOPT_GTE0       0x0020
1521 #ifdef CONFIG_QUOTA
1522 #define MOPT_Q          0
1523 #define MOPT_QFMT       0x0040
1524 #else
1525 #define MOPT_Q          MOPT_NOSUPPORT
1526 #define MOPT_QFMT       MOPT_NOSUPPORT
1527 #endif
1528 #define MOPT_DATAJ      0x0080
1529 #define MOPT_NO_EXT2    0x0100
1530 #define MOPT_NO_EXT3    0x0200
1531 #define MOPT_EXT4_ONLY  (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1532 #define MOPT_STRING     0x0400
1533
1534 static const struct mount_opts {
1535         int     token;
1536         int     mount_opt;
1537         int     flags;
1538 } ext4_mount_opts[] = {
1539         {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1540         {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1541         {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1542         {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1543         {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1544         {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1545         {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1546          MOPT_EXT4_ONLY | MOPT_SET},
1547         {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1548          MOPT_EXT4_ONLY | MOPT_CLEAR},
1549         {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1550         {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1551         {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1552          MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1553         {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1554          MOPT_EXT4_ONLY | MOPT_CLEAR},
1555         {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1556          MOPT_EXT4_ONLY | MOPT_CLEAR},
1557         {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1558          MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1559         {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1560                                     EXT4_MOUNT_JOURNAL_CHECKSUM),
1561          MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1562         {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1563         {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1564         {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1565         {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1566         {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1567          MOPT_NO_EXT2},
1568         {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1569          MOPT_NO_EXT2},
1570         {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1571         {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1572         {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1573         {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1574         {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1575         {Opt_commit, 0, MOPT_GTE0},
1576         {Opt_max_batch_time, 0, MOPT_GTE0},
1577         {Opt_min_batch_time, 0, MOPT_GTE0},
1578         {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1579         {Opt_init_itable, 0, MOPT_GTE0},
1580         {Opt_dax, EXT4_MOUNT_DAX, MOPT_SET},
1581         {Opt_stripe, 0, MOPT_GTE0},
1582         {Opt_resuid, 0, MOPT_GTE0},
1583         {Opt_resgid, 0, MOPT_GTE0},
1584         {Opt_journal_dev, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1585         {Opt_journal_path, 0, MOPT_NO_EXT2 | MOPT_STRING},
1586         {Opt_journal_ioprio, 0, MOPT_NO_EXT2 | MOPT_GTE0},
1587         {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1588         {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1589         {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1590          MOPT_NO_EXT2 | MOPT_DATAJ},
1591         {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1592         {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1593 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1594         {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1595         {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1596 #else
1597         {Opt_acl, 0, MOPT_NOSUPPORT},
1598         {Opt_noacl, 0, MOPT_NOSUPPORT},
1599 #endif
1600         {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1601         {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1602         {Opt_debug_want_extra_isize, 0, MOPT_GTE0},
1603         {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1604         {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1605                                                         MOPT_SET | MOPT_Q},
1606         {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1607                                                         MOPT_SET | MOPT_Q},
1608         {Opt_prjquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_PRJQUOTA,
1609                                                         MOPT_SET | MOPT_Q},
1610         {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1611                        EXT4_MOUNT_GRPQUOTA | EXT4_MOUNT_PRJQUOTA),
1612                                                         MOPT_CLEAR | MOPT_Q},
1613         {Opt_usrjquota, 0, MOPT_Q},
1614         {Opt_grpjquota, 0, MOPT_Q},
1615         {Opt_offusrjquota, 0, MOPT_Q},
1616         {Opt_offgrpjquota, 0, MOPT_Q},
1617         {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1618         {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1619         {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1620         {Opt_max_dir_size_kb, 0, MOPT_GTE0},
1621         {Opt_test_dummy_encryption, 0, MOPT_GTE0},
1622         {Opt_err, 0, 0}
1623 };
1624
1625 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1626                             substring_t *args, unsigned long *journal_devnum,
1627                             unsigned int *journal_ioprio, int is_remount)
1628 {
1629         struct ext4_sb_info *sbi = EXT4_SB(sb);
1630         const struct mount_opts *m;
1631         kuid_t uid;
1632         kgid_t gid;
1633         int arg = 0;
1634
1635 #ifdef CONFIG_QUOTA
1636         if (token == Opt_usrjquota)
1637                 return set_qf_name(sb, USRQUOTA, &args[0]);
1638         else if (token == Opt_grpjquota)
1639                 return set_qf_name(sb, GRPQUOTA, &args[0]);
1640         else if (token == Opt_offusrjquota)
1641                 return clear_qf_name(sb, USRQUOTA);
1642         else if (token == Opt_offgrpjquota)
1643                 return clear_qf_name(sb, GRPQUOTA);
1644 #endif
1645         switch (token) {
1646         case Opt_noacl:
1647         case Opt_nouser_xattr:
1648                 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1649                 break;
1650         case Opt_sb:
1651                 return 1;       /* handled by get_sb_block() */
1652         case Opt_removed:
1653                 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
1654                 return 1;
1655         case Opt_abort:
1656                 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1657                 return 1;
1658         case Opt_i_version:
1659                 sb->s_flags |= MS_I_VERSION;
1660                 return 1;
1661         case Opt_lazytime:
1662                 sb->s_flags |= MS_LAZYTIME;
1663                 return 1;
1664         case Opt_nolazytime:
1665                 sb->s_flags &= ~MS_LAZYTIME;
1666                 return 1;
1667         }
1668
1669         for (m = ext4_mount_opts; m->token != Opt_err; m++)
1670                 if (token == m->token)
1671                         break;
1672
1673         if (m->token == Opt_err) {
1674                 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1675                          "or missing value", opt);
1676                 return -1;
1677         }
1678
1679         if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1680                 ext4_msg(sb, KERN_ERR,
1681                          "Mount option \"%s\" incompatible with ext2", opt);
1682                 return -1;
1683         }
1684         if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1685                 ext4_msg(sb, KERN_ERR,
1686                          "Mount option \"%s\" incompatible with ext3", opt);
1687                 return -1;
1688         }
1689
1690         if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
1691                 return -1;
1692         if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1693                 return -1;
1694         if (m->flags & MOPT_EXPLICIT) {
1695                 if (m->mount_opt & EXT4_MOUNT_DELALLOC) {
1696                         set_opt2(sb, EXPLICIT_DELALLOC);
1697                 } else if (m->mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) {
1698                         set_opt2(sb, EXPLICIT_JOURNAL_CHECKSUM);
1699                 } else
1700                         return -1;
1701         }
1702         if (m->flags & MOPT_CLEAR_ERR)
1703                 clear_opt(sb, ERRORS_MASK);
1704         if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1705                 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1706                          "options when quota turned on");
1707                 return -1;
1708         }
1709
1710         if (m->flags & MOPT_NOSUPPORT) {
1711                 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1712         } else if (token == Opt_commit) {
1713                 if (arg == 0)
1714                         arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1715                 sbi->s_commit_interval = HZ * arg;
1716         } else if (token == Opt_debug_want_extra_isize) {
1717                 sbi->s_want_extra_isize = arg;
1718         } else if (token == Opt_max_batch_time) {
1719                 sbi->s_max_batch_time = arg;
1720         } else if (token == Opt_min_batch_time) {
1721                 sbi->s_min_batch_time = arg;
1722         } else if (token == Opt_inode_readahead_blks) {
1723                 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1724                         ext4_msg(sb, KERN_ERR,
1725                                  "EXT4-fs: inode_readahead_blks must be "
1726                                  "0 or a power of 2 smaller than 2^31");
1727                         return -1;
1728                 }
1729                 sbi->s_inode_readahead_blks = arg;
1730         } else if (token == Opt_init_itable) {
1731                 set_opt(sb, INIT_INODE_TABLE);
1732                 if (!args->from)
1733                         arg = EXT4_DEF_LI_WAIT_MULT;
1734                 sbi->s_li_wait_mult = arg;
1735         } else if (token == Opt_max_dir_size_kb) {
1736                 sbi->s_max_dir_size_kb = arg;
1737         } else if (token == Opt_stripe) {
1738                 sbi->s_stripe = arg;
1739         } else if (token == Opt_resuid) {
1740                 uid = make_kuid(current_user_ns(), arg);
1741                 if (!uid_valid(uid)) {
1742                         ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1743                         return -1;
1744                 }
1745                 sbi->s_resuid = uid;
1746         } else if (token == Opt_resgid) {
1747                 gid = make_kgid(current_user_ns(), arg);
1748                 if (!gid_valid(gid)) {
1749                         ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1750                         return -1;
1751                 }
1752                 sbi->s_resgid = gid;
1753         } else if (token == Opt_journal_dev) {
1754                 if (is_remount) {
1755                         ext4_msg(sb, KERN_ERR,
1756                                  "Cannot specify journal on remount");
1757                         return -1;
1758                 }
1759                 *journal_devnum = arg;
1760         } else if (token == Opt_journal_path) {
1761                 char *journal_path;
1762                 struct inode *journal_inode;
1763                 struct path path;
1764                 int error;
1765
1766                 if (is_remount) {
1767                         ext4_msg(sb, KERN_ERR,
1768                                  "Cannot specify journal on remount");
1769                         return -1;
1770                 }
1771                 journal_path = match_strdup(&args[0]);
1772                 if (!journal_path) {
1773                         ext4_msg(sb, KERN_ERR, "error: could not dup "
1774                                 "journal device string");
1775                         return -1;
1776                 }
1777
1778                 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1779                 if (error) {
1780                         ext4_msg(sb, KERN_ERR, "error: could not find "
1781                                 "journal device path: error %d", error);
1782                         kfree(journal_path);
1783                         return -1;
1784                 }
1785
1786                 journal_inode = d_inode(path.dentry);
1787                 if (!S_ISBLK(journal_inode->i_mode)) {
1788                         ext4_msg(sb, KERN_ERR, "error: journal path %s "
1789                                 "is not a block device", journal_path);
1790                         path_put(&path);
1791                         kfree(journal_path);
1792                         return -1;
1793                 }
1794
1795                 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
1796                 path_put(&path);
1797                 kfree(journal_path);
1798         } else if (token == Opt_journal_ioprio) {
1799                 if (arg > 7) {
1800                         ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1801                                  " (must be 0-7)");
1802                         return -1;
1803                 }
1804                 *journal_ioprio =
1805                         IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1806         } else if (token == Opt_test_dummy_encryption) {
1807 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1808                 sbi->s_mount_flags |= EXT4_MF_TEST_DUMMY_ENCRYPTION;
1809                 ext4_msg(sb, KERN_WARNING,
1810                          "Test dummy encryption mode enabled");
1811 #else
1812                 ext4_msg(sb, KERN_WARNING,
1813                          "Test dummy encryption mount option ignored");
1814 #endif
1815         } else if (m->flags & MOPT_DATAJ) {
1816                 if (is_remount) {
1817                         if (!sbi->s_journal)
1818                                 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1819                         else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
1820                                 ext4_msg(sb, KERN_ERR,
1821                                          "Cannot change data mode on remount");
1822                                 return -1;
1823                         }
1824                 } else {
1825                         clear_opt(sb, DATA_FLAGS);
1826                         sbi->s_mount_opt |= m->mount_opt;
1827                 }
1828 #ifdef CONFIG_QUOTA
1829         } else if (m->flags & MOPT_QFMT) {
1830                 if (sb_any_quota_loaded(sb) &&
1831                     sbi->s_jquota_fmt != m->mount_opt) {
1832                         ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1833                                  "quota options when quota turned on");
1834                         return -1;
1835                 }
1836                 if (ext4_has_feature_quota(sb)) {
1837                         ext4_msg(sb, KERN_INFO,
1838                                  "Quota format mount options ignored "
1839                                  "when QUOTA feature is enabled");
1840                         return 1;
1841                 }
1842                 sbi->s_jquota_fmt = m->mount_opt;
1843 #endif
1844         } else if (token == Opt_dax) {
1845 #ifdef CONFIG_FS_DAX
1846                 ext4_msg(sb, KERN_WARNING,
1847                 "DAX enabled. Warning: EXPERIMENTAL, use at your own risk");
1848                         sbi->s_mount_opt |= m->mount_opt;
1849 #else
1850                 ext4_msg(sb, KERN_INFO, "dax option not supported");
1851                 return -1;
1852 #endif
1853         } else if (token == Opt_data_err_abort) {
1854                 sbi->s_mount_opt |= m->mount_opt;
1855         } else if (token == Opt_data_err_ignore) {
1856                 sbi->s_mount_opt &= ~m->mount_opt;
1857         } else {
1858                 if (!args->from)
1859                         arg = 1;
1860                 if (m->flags & MOPT_CLEAR)
1861                         arg = !arg;
1862                 else if (unlikely(!(m->flags & MOPT_SET))) {
1863                         ext4_msg(sb, KERN_WARNING,
1864                                  "buggy handling of option %s", opt);
1865                         WARN_ON(1);
1866                         return -1;
1867                 }
1868                 if (arg != 0)
1869                         sbi->s_mount_opt |= m->mount_opt;
1870                 else
1871                         sbi->s_mount_opt &= ~m->mount_opt;
1872         }
1873         return 1;
1874 }
1875
1876 static int parse_options(char *options, struct super_block *sb,
1877                          unsigned long *journal_devnum,
1878                          unsigned int *journal_ioprio,
1879                          int is_remount)
1880 {
1881         struct ext4_sb_info *sbi = EXT4_SB(sb);
1882         char *p;
1883         substring_t args[MAX_OPT_ARGS];
1884         int token;
1885
1886         if (!options)
1887                 return 1;
1888
1889         while ((p = strsep(&options, ",")) != NULL) {
1890                 if (!*p)
1891                         continue;
1892                 /*
1893                  * Initialize args struct so we know whether arg was
1894                  * found; some options take optional arguments.
1895                  */
1896                 args[0].to = args[0].from = NULL;
1897                 token = match_token(p, tokens, args);
1898                 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1899                                      journal_ioprio, is_remount) < 0)
1900                         return 0;
1901         }
1902 #ifdef CONFIG_QUOTA
1903         /*
1904          * We do the test below only for project quotas. 'usrquota' and
1905          * 'grpquota' mount options are allowed even without quota feature
1906          * to support legacy quotas in quota files.
1907          */
1908         if (test_opt(sb, PRJQUOTA) && !ext4_has_feature_project(sb)) {
1909                 ext4_msg(sb, KERN_ERR, "Project quota feature not enabled. "
1910                          "Cannot enable project quota enforcement.");
1911                 return 0;
1912         }
1913         if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1914                 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1915                         clear_opt(sb, USRQUOTA);
1916
1917                 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1918                         clear_opt(sb, GRPQUOTA);
1919
1920                 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1921                         ext4_msg(sb, KERN_ERR, "old and new quota "
1922                                         "format mixing");
1923                         return 0;
1924                 }
1925
1926                 if (!sbi->s_jquota_fmt) {
1927                         ext4_msg(sb, KERN_ERR, "journaled quota format "
1928                                         "not specified");
1929                         return 0;
1930                 }
1931         }
1932 #endif
1933         if (test_opt(sb, DIOREAD_NOLOCK)) {
1934                 int blocksize =
1935                         BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1936
1937                 if (blocksize < PAGE_SIZE) {
1938                         ext4_msg(sb, KERN_ERR, "can't mount with "
1939                                  "dioread_nolock if block size != PAGE_SIZE");
1940                         return 0;
1941                 }
1942         }
1943         return 1;
1944 }
1945
1946 static inline void ext4_show_quota_options(struct seq_file *seq,
1947                                            struct super_block *sb)
1948 {
1949 #if defined(CONFIG_QUOTA)
1950         struct ext4_sb_info *sbi = EXT4_SB(sb);
1951
1952         if (sbi->s_jquota_fmt) {
1953                 char *fmtname = "";
1954
1955                 switch (sbi->s_jquota_fmt) {
1956                 case QFMT_VFS_OLD:
1957                         fmtname = "vfsold";
1958                         break;
1959                 case QFMT_VFS_V0:
1960                         fmtname = "vfsv0";
1961                         break;
1962                 case QFMT_VFS_V1:
1963                         fmtname = "vfsv1";
1964                         break;
1965                 }
1966                 seq_printf(seq, ",jqfmt=%s", fmtname);
1967         }
1968
1969         if (sbi->s_qf_names[USRQUOTA])
1970                 seq_show_option(seq, "usrjquota", sbi->s_qf_names[USRQUOTA]);
1971
1972         if (sbi->s_qf_names[GRPQUOTA])
1973                 seq_show_option(seq, "grpjquota", sbi->s_qf_names[GRPQUOTA]);
1974 #endif
1975 }
1976
1977 static const char *token2str(int token)
1978 {
1979         const struct match_token *t;
1980
1981         for (t = tokens; t->token != Opt_err; t++)
1982                 if (t->token == token && !strchr(t->pattern, '='))
1983                         break;
1984         return t->pattern;
1985 }
1986
1987 /*
1988  * Show an option if
1989  *  - it's set to a non-default value OR
1990  *  - if the per-sb default is different from the global default
1991  */
1992 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1993                               int nodefs)
1994 {
1995         struct ext4_sb_info *sbi = EXT4_SB(sb);
1996         struct ext4_super_block *es = sbi->s_es;
1997         int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
1998         const struct mount_opts *m;
1999         char sep = nodefs ? '\n' : ',';
2000
2001 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
2002 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
2003
2004         if (sbi->s_sb_block != 1)
2005                 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
2006
2007         for (m = ext4_mount_opts; m->token != Opt_err; m++) {
2008                 int want_set = m->flags & MOPT_SET;
2009                 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
2010                     (m->flags & MOPT_CLEAR_ERR))
2011                         continue;
2012                 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
2013                         continue; /* skip if same as the default */
2014                 if ((want_set &&
2015                      (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
2016                     (!want_set && (sbi->s_mount_opt & m->mount_opt)))
2017                         continue; /* select Opt_noFoo vs Opt_Foo */
2018                 SEQ_OPTS_PRINT("%s", token2str(m->token));
2019         }
2020
2021         if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
2022             le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
2023                 SEQ_OPTS_PRINT("resuid=%u",
2024                                 from_kuid_munged(&init_user_ns, sbi->s_resuid));
2025         if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
2026             le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
2027                 SEQ_OPTS_PRINT("resgid=%u",
2028                                 from_kgid_munged(&init_user_ns, sbi->s_resgid));
2029         def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
2030         if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
2031                 SEQ_OPTS_PUTS("errors=remount-ro");
2032         if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
2033                 SEQ_OPTS_PUTS("errors=continue");
2034         if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
2035                 SEQ_OPTS_PUTS("errors=panic");
2036         if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
2037                 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
2038         if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
2039                 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
2040         if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
2041                 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
2042         if (sb->s_flags & MS_I_VERSION)
2043                 SEQ_OPTS_PUTS("i_version");
2044         if (nodefs || sbi->s_stripe)
2045                 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
2046         if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
2047                 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
2048                         SEQ_OPTS_PUTS("data=journal");
2049                 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
2050                         SEQ_OPTS_PUTS("data=ordered");
2051                 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
2052                         SEQ_OPTS_PUTS("data=writeback");
2053         }
2054         if (nodefs ||
2055             sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
2056                 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
2057                                sbi->s_inode_readahead_blks);
2058
2059         if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
2060                        (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
2061                 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
2062         if (nodefs || sbi->s_max_dir_size_kb)
2063                 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
2064         if (test_opt(sb, DATA_ERR_ABORT))
2065                 SEQ_OPTS_PUTS("data_err=abort");
2066
2067         ext4_show_quota_options(seq, sb);
2068         return 0;
2069 }
2070
2071 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
2072 {
2073         return _ext4_show_options(seq, root->d_sb, 0);
2074 }
2075
2076 int ext4_seq_options_show(struct seq_file *seq, void *offset)
2077 {
2078         struct super_block *sb = seq->private;
2079         int rc;
2080
2081         seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
2082         rc = _ext4_show_options(seq, sb, 1);
2083         seq_puts(seq, "\n");
2084         return rc;
2085 }
2086
2087 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
2088                             int read_only)
2089 {
2090         struct ext4_sb_info *sbi = EXT4_SB(sb);
2091         int res = 0;
2092
2093         if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
2094                 ext4_msg(sb, KERN_ERR, "revision level too high, "
2095                          "forcing read-only mode");
2096                 res = MS_RDONLY;
2097         }
2098         if (read_only)
2099                 goto done;
2100         if (!(sbi->s_mount_state & EXT4_VALID_FS))
2101                 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
2102                          "running e2fsck is recommended");
2103         else if (sbi->s_mount_state & EXT4_ERROR_FS)
2104                 ext4_msg(sb, KERN_WARNING,
2105                          "warning: mounting fs with errors, "
2106                          "running e2fsck is recommended");
2107         else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
2108                  le16_to_cpu(es->s_mnt_count) >=
2109                  (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
2110                 ext4_msg(sb, KERN_WARNING,
2111                          "warning: maximal mount count reached, "
2112                          "running e2fsck is recommended");
2113         else if (le32_to_cpu(es->s_checkinterval) &&
2114                 (le32_to_cpu(es->s_lastcheck) +
2115                         le32_to_cpu(es->s_checkinterval) <= get_seconds()))
2116                 ext4_msg(sb, KERN_WARNING,
2117                          "warning: checktime reached, "
2118                          "running e2fsck is recommended");
2119         if (!sbi->s_journal)
2120                 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
2121         if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
2122                 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
2123         le16_add_cpu(&es->s_mnt_count, 1);
2124         es->s_mtime = cpu_to_le32(get_seconds());
2125         ext4_update_dynamic_rev(sb);
2126         if (sbi->s_journal)
2127                 ext4_set_feature_journal_needs_recovery(sb);
2128
2129         ext4_commit_super(sb, 1);
2130 done:
2131         if (test_opt(sb, DEBUG))
2132                 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
2133                                 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
2134                         sb->s_blocksize,
2135                         sbi->s_groups_count,
2136                         EXT4_BLOCKS_PER_GROUP(sb),
2137                         EXT4_INODES_PER_GROUP(sb),
2138                         sbi->s_mount_opt, sbi->s_mount_opt2);
2139
2140         cleancache_init_fs(sb);
2141         return res;
2142 }
2143
2144 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
2145 {
2146         struct ext4_sb_info *sbi = EXT4_SB(sb);
2147         struct flex_groups *new_groups;
2148         int size;
2149
2150         if (!sbi->s_log_groups_per_flex)
2151                 return 0;
2152
2153         size = ext4_flex_group(sbi, ngroup - 1) + 1;
2154         if (size <= sbi->s_flex_groups_allocated)
2155                 return 0;
2156
2157         size = roundup_pow_of_two(size * sizeof(struct flex_groups));
2158         new_groups = ext4_kvzalloc(size, GFP_KERNEL);
2159         if (!new_groups) {
2160                 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
2161                          size / (int) sizeof(struct flex_groups));
2162                 return -ENOMEM;
2163         }
2164
2165         if (sbi->s_flex_groups) {
2166                 memcpy(new_groups, sbi->s_flex_groups,
2167                        (sbi->s_flex_groups_allocated *
2168                         sizeof(struct flex_groups)));
2169                 kvfree(sbi->s_flex_groups);
2170         }
2171         sbi->s_flex_groups = new_groups;
2172         sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
2173         return 0;
2174 }
2175
2176 static int ext4_fill_flex_info(struct super_block *sb)
2177 {
2178         struct ext4_sb_info *sbi = EXT4_SB(sb);
2179         struct ext4_group_desc *gdp = NULL;
2180         ext4_group_t flex_group;
2181         int i, err;
2182
2183         sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
2184         if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
2185                 sbi->s_log_groups_per_flex = 0;
2186                 return 1;
2187         }
2188
2189         err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
2190         if (err)
2191                 goto failed;
2192
2193         for (i = 0; i < sbi->s_groups_count; i++) {
2194                 gdp = ext4_get_group_desc(sb, i, NULL);
2195
2196                 flex_group = ext4_flex_group(sbi, i);
2197                 atomic_add(ext4_free_inodes_count(sb, gdp),
2198                            &sbi->s_flex_groups[flex_group].free_inodes);
2199                 atomic64_add(ext4_free_group_clusters(sb, gdp),
2200                              &sbi->s_flex_groups[flex_group].free_clusters);
2201                 atomic_add(ext4_used_dirs_count(sb, gdp),
2202                            &sbi->s_flex_groups[flex_group].used_dirs);
2203         }
2204
2205         return 1;
2206 failed:
2207         return 0;
2208 }
2209
2210 static __le16 ext4_group_desc_csum(struct super_block *sb, __u32 block_group,
2211                                    struct ext4_group_desc *gdp)
2212 {
2213         int offset = offsetof(struct ext4_group_desc, bg_checksum);
2214         __u16 crc = 0;
2215         __le32 le_group = cpu_to_le32(block_group);
2216         struct ext4_sb_info *sbi = EXT4_SB(sb);
2217
2218         if (ext4_has_metadata_csum(sbi->s_sb)) {
2219                 /* Use new metadata_csum algorithm */
2220                 __u32 csum32;
2221                 __u16 dummy_csum = 0;
2222
2223                 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2224                                      sizeof(le_group));
2225                 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp, offset);
2226                 csum32 = ext4_chksum(sbi, csum32, (__u8 *)&dummy_csum,
2227                                      sizeof(dummy_csum));
2228                 offset += sizeof(dummy_csum);
2229                 if (offset < sbi->s_desc_size)
2230                         csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp + offset,
2231                                              sbi->s_desc_size - offset);
2232
2233                 crc = csum32 & 0xFFFF;
2234                 goto out;
2235         }
2236
2237         /* old crc16 code */
2238         if (!ext4_has_feature_gdt_csum(sb))
2239                 return 0;
2240
2241         crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2242         crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2243         crc = crc16(crc, (__u8 *)gdp, offset);
2244         offset += sizeof(gdp->bg_checksum); /* skip checksum */
2245         /* for checksum of struct ext4_group_desc do the rest...*/
2246         if (ext4_has_feature_64bit(sb) &&
2247             offset < le16_to_cpu(sbi->s_es->s_desc_size))
2248                 crc = crc16(crc, (__u8 *)gdp + offset,
2249                             le16_to_cpu(sbi->s_es->s_desc_size) -
2250                                 offset);
2251
2252 out:
2253         return cpu_to_le16(crc);
2254 }
2255
2256 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2257                                 struct ext4_group_desc *gdp)
2258 {
2259         if (ext4_has_group_desc_csum(sb) &&
2260             (gdp->bg_checksum != ext4_group_desc_csum(sb, block_group, gdp)))
2261                 return 0;
2262
2263         return 1;
2264 }
2265
2266 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2267                               struct ext4_group_desc *gdp)
2268 {
2269         if (!ext4_has_group_desc_csum(sb))
2270                 return;
2271         gdp->bg_checksum = ext4_group_desc_csum(sb, block_group, gdp);
2272 }
2273
2274 /* Called at mount-time, super-block is locked */
2275 static int ext4_check_descriptors(struct super_block *sb,
2276                                   ext4_fsblk_t sb_block,
2277                                   ext4_group_t *first_not_zeroed)
2278 {
2279         struct ext4_sb_info *sbi = EXT4_SB(sb);
2280         ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2281         ext4_fsblk_t last_block;
2282         ext4_fsblk_t block_bitmap;
2283         ext4_fsblk_t inode_bitmap;
2284         ext4_fsblk_t inode_table;
2285         int flexbg_flag = 0;
2286         ext4_group_t i, grp = sbi->s_groups_count;
2287
2288         if (ext4_has_feature_flex_bg(sb))
2289                 flexbg_flag = 1;
2290
2291         ext4_debug("Checking group descriptors");
2292
2293         for (i = 0; i < sbi->s_groups_count; i++) {
2294                 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2295
2296                 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2297                         last_block = ext4_blocks_count(sbi->s_es) - 1;
2298                 else
2299                         last_block = first_block +
2300                                 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2301
2302                 if ((grp == sbi->s_groups_count) &&
2303                    !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2304                         grp = i;
2305
2306                 block_bitmap = ext4_block_bitmap(sb, gdp);
2307                 if (block_bitmap == sb_block) {
2308                         ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2309                                  "Block bitmap for group %u overlaps "
2310                                  "superblock", i);
2311                 }
2312                 if (block_bitmap < first_block || block_bitmap > last_block) {
2313                         ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2314                                "Block bitmap for group %u not in group "
2315                                "(block %llu)!", i, block_bitmap);
2316                         return 0;
2317                 }
2318                 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2319                 if (inode_bitmap == sb_block) {
2320                         ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2321                                  "Inode bitmap for group %u overlaps "
2322                                  "superblock", i);
2323                 }
2324                 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2325                         ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2326                                "Inode bitmap for group %u not in group "
2327                                "(block %llu)!", i, inode_bitmap);
2328                         return 0;
2329                 }
2330                 inode_table = ext4_inode_table(sb, gdp);
2331                 if (inode_table == sb_block) {
2332                         ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2333                                  "Inode table for group %u overlaps "
2334                                  "superblock", i);
2335                 }
2336                 if (inode_table < first_block ||
2337                     inode_table + sbi->s_itb_per_group - 1 > last_block) {
2338                         ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2339                                "Inode table for group %u not in group "
2340                                "(block %llu)!", i, inode_table);
2341                         return 0;
2342                 }
2343                 ext4_lock_group(sb, i);
2344                 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2345                         ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2346                                  "Checksum for group %u failed (%u!=%u)",
2347                                  i, le16_to_cpu(ext4_group_desc_csum(sb, i,
2348                                      gdp)), le16_to_cpu(gdp->bg_checksum));
2349                         if (!(sb->s_flags & MS_RDONLY)) {
2350                                 ext4_unlock_group(sb, i);
2351                                 return 0;
2352                         }
2353                 }
2354                 ext4_unlock_group(sb, i);
2355                 if (!flexbg_flag)
2356                         first_block += EXT4_BLOCKS_PER_GROUP(sb);
2357         }
2358         if (NULL != first_not_zeroed)
2359                 *first_not_zeroed = grp;
2360         return 1;
2361 }
2362
2363 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2364  * the superblock) which were deleted from all directories, but held open by
2365  * a process at the time of a crash.  We walk the list and try to delete these
2366  * inodes at recovery time (only with a read-write filesystem).
2367  *
2368  * In order to keep the orphan inode chain consistent during traversal (in
2369  * case of crash during recovery), we link each inode into the superblock
2370  * orphan list_head and handle it the same way as an inode deletion during
2371  * normal operation (which journals the operations for us).
2372  *
2373  * We only do an iget() and an iput() on each inode, which is very safe if we
2374  * accidentally point at an in-use or already deleted inode.  The worst that
2375  * can happen in this case is that we get a "bit already cleared" message from
2376  * ext4_free_inode().  The only reason we would point at a wrong inode is if
2377  * e2fsck was run on this filesystem, and it must have already done the orphan
2378  * inode cleanup for us, so we can safely abort without any further action.
2379  */
2380 static void ext4_orphan_cleanup(struct super_block *sb,
2381                                 struct ext4_super_block *es)
2382 {
2383         unsigned int s_flags = sb->s_flags;
2384         int ret, nr_orphans = 0, nr_truncates = 0;
2385 #ifdef CONFIG_QUOTA
2386         int i;
2387 #endif
2388         if (!es->s_last_orphan) {
2389                 jbd_debug(4, "no orphan inodes to clean up\n");
2390                 return;
2391         }
2392
2393         if (bdev_read_only(sb->s_bdev)) {
2394                 ext4_msg(sb, KERN_ERR, "write access "
2395                         "unavailable, skipping orphan cleanup");
2396                 return;
2397         }
2398
2399         /* Check if feature set would not allow a r/w mount */
2400         if (!ext4_feature_set_ok(sb, 0)) {
2401                 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2402                          "unknown ROCOMPAT features");
2403                 return;
2404         }
2405
2406         if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2407                 /* don't clear list on RO mount w/ errors */
2408                 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
2409                         ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
2410                                   "clearing orphan list.\n");
2411                         es->s_last_orphan = 0;
2412                 }
2413                 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2414                 return;
2415         }
2416
2417         if (s_flags & MS_RDONLY) {
2418                 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2419                 sb->s_flags &= ~MS_RDONLY;
2420         }
2421 #ifdef CONFIG_QUOTA
2422         /* Needed for iput() to work correctly and not trash data */
2423         sb->s_flags |= MS_ACTIVE;
2424         /* Turn on quotas so that they are updated correctly */
2425         for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2426                 if (EXT4_SB(sb)->s_qf_names[i]) {
2427                         int ret = ext4_quota_on_mount(sb, i);
2428                         if (ret < 0)
2429                                 ext4_msg(sb, KERN_ERR,
2430                                         "Cannot turn on journaled "
2431                                         "quota: error %d", ret);
2432                 }
2433         }
2434 #endif
2435
2436         while (es->s_last_orphan) {
2437                 struct inode *inode;
2438
2439                 /*
2440                  * We may have encountered an error during cleanup; if
2441                  * so, skip the rest.
2442                  */
2443                 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2444                         jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2445                         es->s_last_orphan = 0;
2446                         break;
2447                 }
2448
2449                 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2450                 if (IS_ERR(inode)) {
2451                         es->s_last_orphan = 0;
2452                         break;
2453                 }
2454
2455                 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2456                 dquot_initialize(inode);
2457                 if (inode->i_nlink) {
2458                         if (test_opt(sb, DEBUG))
2459                                 ext4_msg(sb, KERN_DEBUG,
2460                                         "%s: truncating inode %lu to %lld bytes",
2461                                         __func__, inode->i_ino, inode->i_size);
2462                         jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2463                                   inode->i_ino, inode->i_size);
2464                         inode_lock(inode);
2465                         truncate_inode_pages(inode->i_mapping, inode->i_size);
2466                         ret = ext4_truncate(inode);
2467                         if (ret)
2468                                 ext4_std_error(inode->i_sb, ret);
2469                         inode_unlock(inode);
2470                         nr_truncates++;
2471                 } else {
2472                         if (test_opt(sb, DEBUG))
2473                                 ext4_msg(sb, KERN_DEBUG,
2474                                         "%s: deleting unreferenced inode %lu",
2475                                         __func__, inode->i_ino);
2476                         jbd_debug(2, "deleting unreferenced inode %lu\n",
2477                                   inode->i_ino);
2478                         nr_orphans++;
2479                 }
2480                 iput(inode);  /* The delete magic happens here! */
2481         }
2482
2483 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2484
2485         if (nr_orphans)
2486                 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2487                        PLURAL(nr_orphans));
2488         if (nr_truncates)
2489                 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2490                        PLURAL(nr_truncates));
2491 #ifdef CONFIG_QUOTA
2492         /* Turn quotas off */
2493         for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2494                 if (sb_dqopt(sb)->files[i])
2495                         dquot_quota_off(sb, i);
2496         }
2497 #endif
2498         sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2499 }
2500
2501 /*
2502  * Maximal extent format file size.
2503  * Resulting logical blkno at s_maxbytes must fit in our on-disk
2504  * extent format containers, within a sector_t, and within i_blocks
2505  * in the vfs.  ext4 inode has 48 bits of i_block in fsblock units,
2506  * so that won't be a limiting factor.
2507  *
2508  * However there is other limiting factor. We do store extents in the form
2509  * of starting block and length, hence the resulting length of the extent
2510  * covering maximum file size must fit into on-disk format containers as
2511  * well. Given that length is always by 1 unit bigger than max unit (because
2512  * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2513  *
2514  * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2515  */
2516 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2517 {
2518         loff_t res;
2519         loff_t upper_limit = MAX_LFS_FILESIZE;
2520
2521         /* small i_blocks in vfs inode? */
2522         if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2523                 /*
2524                  * CONFIG_LBDAF is not enabled implies the inode
2525                  * i_block represent total blocks in 512 bytes
2526                  * 32 == size of vfs inode i_blocks * 8
2527                  */
2528                 upper_limit = (1LL << 32) - 1;
2529
2530                 /* total blocks in file system block size */
2531                 upper_limit >>= (blkbits - 9);
2532                 upper_limit <<= blkbits;
2533         }
2534
2535         /*
2536          * 32-bit extent-start container, ee_block. We lower the maxbytes
2537          * by one fs block, so ee_len can cover the extent of maximum file
2538          * size
2539          */
2540         res = (1LL << 32) - 1;
2541         res <<= blkbits;
2542
2543         /* Sanity check against vm- & vfs- imposed limits */
2544         if (res > upper_limit)
2545                 res = upper_limit;
2546
2547         return res;
2548 }
2549
2550 /*
2551  * Maximal bitmap file size.  There is a direct, and {,double-,triple-}indirect
2552  * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2553  * We need to be 1 filesystem block less than the 2^48 sector limit.
2554  */
2555 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2556 {
2557         loff_t res = EXT4_NDIR_BLOCKS;
2558         int meta_blocks;
2559         loff_t upper_limit;
2560         /* This is calculated to be the largest file size for a dense, block
2561          * mapped file such that the file's total number of 512-byte sectors,
2562          * including data and all indirect blocks, does not exceed (2^48 - 1).
2563          *
2564          * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2565          * number of 512-byte sectors of the file.
2566          */
2567
2568         if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2569                 /*
2570                  * !has_huge_files or CONFIG_LBDAF not enabled implies that
2571                  * the inode i_block field represents total file blocks in
2572                  * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2573                  */
2574                 upper_limit = (1LL << 32) - 1;
2575
2576                 /* total blocks in file system block size */
2577                 upper_limit >>= (bits - 9);
2578
2579         } else {
2580                 /*
2581                  * We use 48 bit ext4_inode i_blocks
2582                  * With EXT4_HUGE_FILE_FL set the i_blocks
2583                  * represent total number of blocks in
2584                  * file system block size
2585                  */
2586                 upper_limit = (1LL << 48) - 1;
2587
2588         }
2589
2590         /* indirect blocks */
2591         meta_blocks = 1;
2592         /* double indirect blocks */
2593         meta_blocks += 1 + (1LL << (bits-2));
2594         /* tripple indirect blocks */
2595         meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2596
2597         upper_limit -= meta_blocks;
2598         upper_limit <<= bits;
2599
2600         res += 1LL << (bits-2);
2601         res += 1LL << (2*(bits-2));
2602         res += 1LL << (3*(bits-2));
2603         res <<= bits;
2604         if (res > upper_limit)
2605                 res = upper_limit;
2606
2607         if (res > MAX_LFS_FILESIZE)
2608                 res = MAX_LFS_FILESIZE;
2609
2610         return res;
2611 }
2612
2613 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2614                                    ext4_fsblk_t logical_sb_block, int nr)
2615 {
2616         struct ext4_sb_info *sbi = EXT4_SB(sb);
2617         ext4_group_t bg, first_meta_bg;
2618         int has_super = 0;
2619
2620         first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2621
2622         if (!ext4_has_feature_meta_bg(sb) || nr < first_meta_bg)
2623                 return logical_sb_block + nr + 1;
2624         bg = sbi->s_desc_per_block * nr;
2625         if (ext4_bg_has_super(sb, bg))
2626                 has_super = 1;
2627
2628         /*
2629          * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2630          * block 2, not 1.  If s_first_data_block == 0 (bigalloc is enabled
2631          * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2632          * compensate.
2633          */
2634         if (sb->s_blocksize == 1024 && nr == 0 &&
2635             le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block) == 0)
2636                 has_super++;
2637
2638         return (has_super + ext4_group_first_block_no(sb, bg));
2639 }
2640
2641 /**
2642  * ext4_get_stripe_size: Get the stripe size.
2643  * @sbi: In memory super block info
2644  *
2645  * If we have specified it via mount option, then
2646  * use the mount option value. If the value specified at mount time is
2647  * greater than the blocks per group use the super block value.
2648  * If the super block value is greater than blocks per group return 0.
2649  * Allocator needs it be less than blocks per group.
2650  *
2651  */
2652 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2653 {
2654         unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2655         unsigned long stripe_width =
2656                         le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2657         int ret;
2658
2659         if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2660                 ret = sbi->s_stripe;
2661         else if (stripe_width && stripe_width <= sbi->s_blocks_per_group)
2662                 ret = stripe_width;
2663         else if (stride && stride <= sbi->s_blocks_per_group)
2664                 ret = stride;
2665         else
2666                 ret = 0;
2667
2668         /*
2669          * If the stripe width is 1, this makes no sense and
2670          * we set it to 0 to turn off stripe handling code.
2671          */
2672         if (ret <= 1)
2673                 ret = 0;
2674
2675         return ret;
2676 }
2677
2678 /*
2679  * Check whether this filesystem can be mounted based on
2680  * the features present and the RDONLY/RDWR mount requested.
2681  * Returns 1 if this filesystem can be mounted as requested,
2682  * 0 if it cannot be.
2683  */
2684 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2685 {
2686         if (ext4_has_unknown_ext4_incompat_features(sb)) {
2687                 ext4_msg(sb, KERN_ERR,
2688                         "Couldn't mount because of "
2689                         "unsupported optional features (%x)",
2690                         (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2691                         ~EXT4_FEATURE_INCOMPAT_SUPP));
2692                 return 0;
2693         }
2694
2695         if (readonly)
2696                 return 1;
2697
2698         if (ext4_has_feature_readonly(sb)) {
2699                 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
2700                 sb->s_flags |= MS_RDONLY;
2701                 return 1;
2702         }
2703
2704         /* Check that feature set is OK for a read-write mount */
2705         if (ext4_has_unknown_ext4_ro_compat_features(sb)) {
2706                 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2707                          "unsupported optional features (%x)",
2708                          (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2709                                 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2710                 return 0;
2711         }
2712         /*
2713          * Large file size enabled file system can only be mounted
2714          * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2715          */
2716         if (ext4_has_feature_huge_file(sb)) {
2717                 if (sizeof(blkcnt_t) < sizeof(u64)) {
2718                         ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2719                                  "cannot be mounted RDWR without "
2720                                  "CONFIG_LBDAF");
2721                         return 0;
2722                 }
2723         }
2724         if (ext4_has_feature_bigalloc(sb) && !ext4_has_feature_extents(sb)) {
2725                 ext4_msg(sb, KERN_ERR,
2726                          "Can't support bigalloc feature without "
2727                          "extents feature\n");
2728                 return 0;
2729         }
2730
2731 #ifndef CONFIG_QUOTA
2732         if (ext4_has_feature_quota(sb) && !readonly) {
2733                 ext4_msg(sb, KERN_ERR,
2734                          "Filesystem with quota feature cannot be mounted RDWR "
2735                          "without CONFIG_QUOTA");
2736                 return 0;
2737         }
2738         if (ext4_has_feature_project(sb) && !readonly) {
2739                 ext4_msg(sb, KERN_ERR,
2740                          "Filesystem with project quota feature cannot be mounted RDWR "
2741                          "without CONFIG_QUOTA");
2742                 return 0;
2743         }
2744 #endif  /* CONFIG_QUOTA */
2745         return 1;
2746 }
2747
2748 /*
2749  * This function is called once a day if we have errors logged
2750  * on the file system
2751  */
2752 static void print_daily_error_info(unsigned long arg)
2753 {
2754         struct super_block *sb = (struct super_block *) arg;
2755         struct ext4_sb_info *sbi;
2756         struct ext4_super_block *es;
2757
2758         sbi = EXT4_SB(sb);
2759         es = sbi->s_es;
2760
2761         if (es->s_error_count)
2762                 /* fsck newer than v1.41.13 is needed to clean this condition. */
2763                 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
2764                          le32_to_cpu(es->s_error_count));
2765         if (es->s_first_error_time) {
2766                 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d",
2767                        sb->s_id, le32_to_cpu(es->s_first_error_time),
2768                        (int) sizeof(es->s_first_error_func),
2769                        es->s_first_error_func,
2770                        le32_to_cpu(es->s_first_error_line));
2771                 if (es->s_first_error_ino)
2772                         printk(KERN_CONT ": inode %u",
2773                                le32_to_cpu(es->s_first_error_ino));
2774                 if (es->s_first_error_block)
2775                         printk(KERN_CONT ": block %llu", (unsigned long long)
2776                                le64_to_cpu(es->s_first_error_block));
2777                 printk(KERN_CONT "\n");
2778         }
2779         if (es->s_last_error_time) {
2780                 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d",
2781                        sb->s_id, le32_to_cpu(es->s_last_error_time),
2782                        (int) sizeof(es->s_last_error_func),
2783                        es->s_last_error_func,
2784                        le32_to_cpu(es->s_last_error_line));
2785                 if (es->s_last_error_ino)
2786                         printk(KERN_CONT ": inode %u",
2787                                le32_to_cpu(es->s_last_error_ino));
2788                 if (es->s_last_error_block)
2789                         printk(KERN_CONT ": block %llu", (unsigned long long)
2790                                le64_to_cpu(es->s_last_error_block));
2791                 printk(KERN_CONT "\n");
2792         }
2793         mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ);  /* Once a day */
2794 }
2795
2796 /* Find next suitable group and run ext4_init_inode_table */
2797 static int ext4_run_li_request(struct ext4_li_request *elr)
2798 {
2799         struct ext4_group_desc *gdp = NULL;
2800         ext4_group_t group, ngroups;
2801         struct super_block *sb;
2802         unsigned long timeout = 0;
2803         int ret = 0;
2804
2805         sb = elr->lr_super;
2806         ngroups = EXT4_SB(sb)->s_groups_count;
2807
2808         for (group = elr->lr_next_group; group < ngroups; group++) {
2809                 gdp = ext4_get_group_desc(sb, group, NULL);
2810                 if (!gdp) {
2811                         ret = 1;
2812                         break;
2813                 }
2814
2815                 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2816                         break;
2817         }
2818
2819         if (group >= ngroups)
2820                 ret = 1;
2821
2822         if (!ret) {
2823                 timeout = jiffies;
2824                 ret = ext4_init_inode_table(sb, group,
2825                                             elr->lr_timeout ? 0 : 1);
2826                 if (elr->lr_timeout == 0) {
2827                         timeout = (jiffies - timeout) *
2828                                   elr->lr_sbi->s_li_wait_mult;
2829                         elr->lr_timeout = timeout;
2830                 }
2831                 elr->lr_next_sched = jiffies + elr->lr_timeout;
2832                 elr->lr_next_group = group + 1;
2833         }
2834         return ret;
2835 }
2836
2837 /*
2838  * Remove lr_request from the list_request and free the
2839  * request structure. Should be called with li_list_mtx held
2840  */
2841 static void ext4_remove_li_request(struct ext4_li_request *elr)
2842 {
2843         struct ext4_sb_info *sbi;
2844
2845         if (!elr)
2846                 return;
2847
2848         sbi = elr->lr_sbi;
2849
2850         list_del(&elr->lr_request);
2851         sbi->s_li_request = NULL;
2852         kfree(elr);
2853 }
2854
2855 static void ext4_unregister_li_request(struct super_block *sb)
2856 {
2857         mutex_lock(&ext4_li_mtx);
2858         if (!ext4_li_info) {
2859                 mutex_unlock(&ext4_li_mtx);
2860                 return;
2861         }
2862
2863         mutex_lock(&ext4_li_info->li_list_mtx);
2864         ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2865         mutex_unlock(&ext4_li_info->li_list_mtx);
2866         mutex_unlock(&ext4_li_mtx);
2867 }
2868
2869 static struct task_struct *ext4_lazyinit_task;
2870
2871 /*
2872  * This is the function where ext4lazyinit thread lives. It walks
2873  * through the request list searching for next scheduled filesystem.
2874  * When such a fs is found, run the lazy initialization request
2875  * (ext4_rn_li_request) and keep track of the time spend in this
2876  * function. Based on that time we compute next schedule time of
2877  * the request. When walking through the list is complete, compute
2878  * next waking time and put itself into sleep.
2879  */
2880 static int ext4_lazyinit_thread(void *arg)
2881 {
2882         struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2883         struct list_head *pos, *n;
2884         struct ext4_li_request *elr;
2885         unsigned long next_wakeup, cur;
2886
2887         BUG_ON(NULL == eli);
2888
2889 cont_thread:
2890         while (true) {
2891                 next_wakeup = MAX_JIFFY_OFFSET;
2892
2893                 mutex_lock(&eli->li_list_mtx);
2894                 if (list_empty(&eli->li_request_list)) {
2895                         mutex_unlock(&eli->li_list_mtx);
2896                         goto exit_thread;
2897                 }
2898                 list_for_each_safe(pos, n, &eli->li_request_list) {
2899                         int err = 0;
2900                         int progress = 0;
2901                         elr = list_entry(pos, struct ext4_li_request,
2902                                          lr_request);
2903
2904                         if (time_before(jiffies, elr->lr_next_sched)) {
2905                                 if (time_before(elr->lr_next_sched, next_wakeup))
2906                                         next_wakeup = elr->lr_next_sched;
2907                                 continue;
2908                         }
2909                         if (down_read_trylock(&elr->lr_super->s_umount)) {
2910                                 if (sb_start_write_trylock(elr->lr_super)) {
2911                                         progress = 1;
2912                                         /*
2913                                          * We hold sb->s_umount, sb can not
2914                                          * be removed from the list, it is
2915                                          * now safe to drop li_list_mtx
2916                                          */
2917                                         mutex_unlock(&eli->li_list_mtx);
2918                                         err = ext4_run_li_request(elr);
2919                                         sb_end_write(elr->lr_super);
2920                                         mutex_lock(&eli->li_list_mtx);
2921                                         n = pos->next;
2922                                 }
2923                                 up_read((&elr->lr_super->s_umount));
2924                         }
2925                         /* error, remove the lazy_init job */
2926                         if (err) {
2927                                 ext4_remove_li_request(elr);
2928                                 continue;
2929                         }
2930                         if (!progress) {
2931                                 elr->lr_next_sched = jiffies +
2932                                         (prandom_u32()
2933                                          % (EXT4_DEF_LI_MAX_START_DELAY * HZ));
2934                         }
2935                         if (time_before(elr->lr_next_sched, next_wakeup))
2936                                 next_wakeup = elr->lr_next_sched;
2937                 }
2938                 mutex_unlock(&eli->li_list_mtx);
2939
2940                 try_to_freeze();
2941
2942                 cur = jiffies;
2943                 if ((time_after_eq(cur, next_wakeup)) ||
2944                     (MAX_JIFFY_OFFSET == next_wakeup)) {
2945                         cond_resched();
2946                         continue;
2947                 }
2948
2949                 schedule_timeout_interruptible(next_wakeup - cur);
2950
2951                 if (kthread_should_stop()) {
2952                         ext4_clear_request_list();
2953                         goto exit_thread;
2954                 }
2955         }
2956
2957 exit_thread:
2958         /*
2959          * It looks like the request list is empty, but we need
2960          * to check it under the li_list_mtx lock, to prevent any
2961          * additions into it, and of course we should lock ext4_li_mtx
2962          * to atomically free the list and ext4_li_info, because at
2963          * this point another ext4 filesystem could be registering
2964          * new one.
2965          */
2966         mutex_lock(&ext4_li_mtx);
2967         mutex_lock(&eli->li_list_mtx);
2968         if (!list_empty(&eli->li_request_list)) {
2969                 mutex_unlock(&eli->li_list_mtx);
2970                 mutex_unlock(&ext4_li_mtx);
2971                 goto cont_thread;
2972         }
2973         mutex_unlock(&eli->li_list_mtx);
2974         kfree(ext4_li_info);
2975         ext4_li_info = NULL;
2976         mutex_unlock(&ext4_li_mtx);
2977
2978         return 0;
2979 }
2980
2981 static void ext4_clear_request_list(void)
2982 {
2983         struct list_head *pos, *n;
2984         struct ext4_li_request *elr;
2985
2986         mutex_lock(&ext4_li_info->li_list_mtx);
2987         list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2988                 elr = list_entry(pos, struct ext4_li_request,
2989                                  lr_request);
2990                 ext4_remove_li_request(elr);
2991         }
2992         mutex_unlock(&ext4_li_info->li_list_mtx);
2993 }
2994
2995 static int ext4_run_lazyinit_thread(void)
2996 {
2997         ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2998                                          ext4_li_info, "ext4lazyinit");
2999         if (IS_ERR(ext4_lazyinit_task)) {
3000                 int err = PTR_ERR(ext4_lazyinit_task);
3001                 ext4_clear_request_list();
3002                 kfree(ext4_li_info);
3003                 ext4_li_info = NULL;
3004                 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
3005                                  "initialization thread\n",
3006                                  err);
3007                 return err;
3008         }
3009         ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
3010         return 0;
3011 }
3012
3013 /*
3014  * Check whether it make sense to run itable init. thread or not.
3015  * If there is at least one uninitialized inode table, return
3016  * corresponding group number, else the loop goes through all
3017  * groups and return total number of groups.
3018  */
3019 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3020 {
3021         ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3022         struct ext4_group_desc *gdp = NULL;
3023
3024         for (group = 0; group < ngroups; group++) {
3025                 gdp = ext4_get_group_desc(sb, group, NULL);
3026                 if (!gdp)
3027                         continue;
3028
3029                 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3030                         break;
3031         }
3032
3033         return group;
3034 }
3035
3036 static int ext4_li_info_new(void)
3037 {
3038         struct ext4_lazy_init *eli = NULL;
3039
3040         eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3041         if (!eli)
3042                 return -ENOMEM;
3043
3044         INIT_LIST_HEAD(&eli->li_request_list);
3045         mutex_init(&eli->li_list_mtx);
3046
3047         eli->li_state |= EXT4_LAZYINIT_QUIT;
3048
3049         ext4_li_info = eli;
3050
3051         return 0;
3052 }
3053
3054 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3055                                             ext4_group_t start)
3056 {
3057         struct ext4_sb_info *sbi = EXT4_SB(sb);
3058         struct ext4_li_request *elr;
3059
3060         elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3061         if (!elr)
3062                 return NULL;
3063
3064         elr->lr_super = sb;
3065         elr->lr_sbi = sbi;
3066         elr->lr_next_group = start;
3067
3068         /*
3069          * Randomize first schedule time of the request to
3070          * spread the inode table initialization requests
3071          * better.
3072          */
3073         elr->lr_next_sched = jiffies + (prandom_u32() %
3074                                 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
3075         return elr;
3076 }
3077
3078 int ext4_register_li_request(struct super_block *sb,
3079                              ext4_group_t first_not_zeroed)
3080 {
3081         struct ext4_sb_info *sbi = EXT4_SB(sb);
3082         struct ext4_li_request *elr = NULL;
3083         ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3084         int ret = 0;
3085
3086         mutex_lock(&ext4_li_mtx);
3087         if (sbi->s_li_request != NULL) {
3088                 /*
3089                  * Reset timeout so it can be computed again, because
3090                  * s_li_wait_mult might have changed.
3091                  */
3092                 sbi->s_li_request->lr_timeout = 0;
3093                 goto out;
3094         }
3095
3096         if (first_not_zeroed == ngroups ||
3097             (sb->s_flags & MS_RDONLY) ||
3098             !test_opt(sb, INIT_INODE_TABLE))
3099                 goto out;
3100
3101         elr = ext4_li_request_new(sb, first_not_zeroed);
3102         if (!elr) {
3103                 ret = -ENOMEM;
3104                 goto out;
3105         }
3106
3107         if (NULL == ext4_li_info) {
3108                 ret = ext4_li_info_new();
3109                 if (ret)
3110                         goto out;
3111         }
3112
3113         mutex_lock(&ext4_li_info->li_list_mtx);
3114         list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3115         mutex_unlock(&ext4_li_info->li_list_mtx);
3116
3117         sbi->s_li_request = elr;
3118         /*
3119          * set elr to NULL here since it has been inserted to
3120          * the request_list and the removal and free of it is
3121          * handled by ext4_clear_request_list from now on.
3122          */
3123         elr = NULL;
3124
3125         if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3126                 ret = ext4_run_lazyinit_thread();
3127                 if (ret)
3128                         goto out;
3129         }
3130 out:
3131         mutex_unlock(&ext4_li_mtx);
3132         if (ret)
3133                 kfree(elr);
3134         return ret;
3135 }
3136
3137 /*
3138  * We do not need to lock anything since this is called on
3139  * module unload.
3140  */
3141 static void ext4_destroy_lazyinit_thread(void)
3142 {
3143         /*
3144          * If thread exited earlier
3145          * there's nothing to be done.
3146          */
3147         if (!ext4_li_info || !ext4_lazyinit_task)
3148                 return;
3149
3150         kthread_stop(ext4_lazyinit_task);
3151 }
3152
3153 static int set_journal_csum_feature_set(struct super_block *sb)
3154 {
3155         int ret = 1;
3156         int compat, incompat;
3157         struct ext4_sb_info *sbi = EXT4_SB(sb);
3158
3159         if (ext4_has_metadata_csum(sb)) {
3160                 /* journal checksum v3 */
3161                 compat = 0;
3162                 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
3163         } else {
3164                 /* journal checksum v1 */
3165                 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3166                 incompat = 0;
3167         }
3168
3169         jbd2_journal_clear_features(sbi->s_journal,
3170                         JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3171                         JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3172                         JBD2_FEATURE_INCOMPAT_CSUM_V2);
3173         if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3174                 ret = jbd2_journal_set_features(sbi->s_journal,
3175                                 compat, 0,
3176                                 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3177                                 incompat);
3178         } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3179                 ret = jbd2_journal_set_features(sbi->s_journal,
3180                                 compat, 0,
3181                                 incompat);
3182                 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3183                                 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3184         } else {
3185                 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3186                                 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3187         }
3188
3189         return ret;
3190 }
3191
3192 /*
3193  * Note: calculating the overhead so we can be compatible with
3194  * historical BSD practice is quite difficult in the face of
3195  * clusters/bigalloc.  This is because multiple metadata blocks from
3196  * different block group can end up in the same allocation cluster.
3197  * Calculating the exact overhead in the face of clustered allocation
3198  * requires either O(all block bitmaps) in memory or O(number of block
3199  * groups**2) in time.  We will still calculate the superblock for
3200  * older file systems --- and if we come across with a bigalloc file
3201  * system with zero in s_overhead_clusters the estimate will be close to
3202  * correct especially for very large cluster sizes --- but for newer
3203  * file systems, it's better to calculate this figure once at mkfs
3204  * time, and store it in the superblock.  If the superblock value is
3205  * present (even for non-bigalloc file systems), we will use it.
3206  */
3207 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3208                           char *buf)
3209 {
3210         struct ext4_sb_info     *sbi = EXT4_SB(sb);
3211         struct ext4_group_desc  *gdp;
3212         ext4_fsblk_t            first_block, last_block, b;
3213         ext4_group_t            i, ngroups = ext4_get_groups_count(sb);
3214         int                     s, j, count = 0;
3215
3216         if (!ext4_has_feature_bigalloc(sb))
3217                 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3218                         sbi->s_itb_per_group + 2);
3219
3220         first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3221                 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3222         last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3223         for (i = 0; i < ngroups; i++) {
3224                 gdp = ext4_get_group_desc(sb, i, NULL);
3225                 b = ext4_block_bitmap(sb, gdp);
3226                 if (b >= first_block && b <= last_block) {
3227                         ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3228                         count++;