?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Linux 4.2.6
[linux/fpc-iii.git] / fs / ext4 / super.c
bloba5e8c744e96260466275108af32ec621b4559d28
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/proc_fs.h>
38 #include <linux/ctype.h>
39 #include <linux/log2.h>
40 #include <linux/crc16.h>
41 #include <linux/cleancache.h>
42 #include <asm/uaccess.h>
43
44 #include <linux/kthread.h>
45 #include <linux/freezer.h>
46
47 #include "ext4.h"
48 #include "ext4_extents.h" /* Needed for trace points definition */
49 #include "ext4_jbd2.h"
50 #include "xattr.h"
51 #include "acl.h"
52 #include "mballoc.h"
53
54 #define CREATE_TRACE_POINTS
55 #include <trace/events/ext4.h>
56
57 static struct proc_dir_entry *ext4_proc_root;
58 static struct kset *ext4_kset;
59 static struct ext4_lazy_init *ext4_li_info;
60 static struct mutex ext4_li_mtx;
61 static struct ext4_features *ext4_feat;
62 static int ext4_mballoc_ready;
63
64 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
65 unsigned long journal_devnum);
66 static int ext4_show_options(struct seq_file *seq, struct dentry *root);
67 static int ext4_commit_super(struct super_block *sb, int sync);
68 static void ext4_mark_recovery_complete(struct super_block *sb,
69 struct ext4_super_block *es);
70 static void ext4_clear_journal_err(struct super_block *sb,
71 struct ext4_super_block *es);
72 static int ext4_sync_fs(struct super_block *sb, int wait);
73 static int ext4_remount(struct super_block *sb, int *flags, char *data);
74 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
75 static int ext4_unfreeze(struct super_block *sb);
76 static int ext4_freeze(struct super_block *sb);
77 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
78 const char *dev_name, void *data);
79 static inline int ext2_feature_set_ok(struct super_block *sb);
80 static inline int ext3_feature_set_ok(struct super_block *sb);
81 static int ext4_feature_set_ok(struct super_block *sb, int readonly);
82 static void ext4_destroy_lazyinit_thread(void);
83 static void ext4_unregister_li_request(struct super_block *sb);
84 static void ext4_clear_request_list(void);
85 static int ext4_reserve_clusters(struct ext4_sb_info *, ext4_fsblk_t);
86
87 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
88 static struct file_system_type ext2_fs_type = {
89 .owner = THIS_MODULE,
90 .name = "ext2",
91 .mount = ext4_mount,
92 .kill_sb = kill_block_super,
93 .fs_flags = FS_REQUIRES_DEV,
94 };
95 MODULE_ALIAS_FS("ext2");
96 MODULE_ALIAS("ext2");
97 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
98 #else
99 #define IS_EXT2_SB(sb) (0)
100 #endif
101
102
103 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
104 static struct file_system_type ext3_fs_type = {
105 .owner = THIS_MODULE,
106 .name = "ext3",
107 .mount = ext4_mount,
108 .kill_sb = kill_block_super,
109 .fs_flags = FS_REQUIRES_DEV,
110 };
111 MODULE_ALIAS_FS("ext3");
112 MODULE_ALIAS("ext3");
113 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
114 #else
115 #define IS_EXT3_SB(sb) (0)
116 #endif
117
118 static int ext4_verify_csum_type(struct super_block *sb,
119 struct ext4_super_block *es)
120 {
121 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
122 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
123 return 1;
124
125 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
126 }
127
128 static __le32 ext4_superblock_csum(struct super_block *sb,
129 struct ext4_super_block *es)
130 {
131 struct ext4_sb_info *sbi = EXT4_SB(sb);
132 int offset = offsetof(struct ext4_super_block, s_checksum);
133 __u32 csum;
134
135 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
136
137 return cpu_to_le32(csum);
138 }
139
140 static int ext4_superblock_csum_verify(struct super_block *sb,
141 struct ext4_super_block *es)
142 {
143 if (!ext4_has_metadata_csum(sb))
144 return 1;
145
146 return es->s_checksum == ext4_superblock_csum(sb, es);
147 }
148
149 void ext4_superblock_csum_set(struct super_block *sb)
150 {
151 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
152
153 if (!ext4_has_metadata_csum(sb))
154 return;
155
156 es->s_checksum = ext4_superblock_csum(sb, es);
157 }
158
159 void *ext4_kvmalloc(size_t size, gfp_t flags)
160 {
161 void *ret;
162
163 ret = kmalloc(size, flags | __GFP_NOWARN);
164 if (!ret)
165 ret = __vmalloc(size, flags, PAGE_KERNEL);
166 return ret;
167 }
168
169 void *ext4_kvzalloc(size_t size, gfp_t flags)
170 {
171 void *ret;
172
173 ret = kzalloc(size, flags | __GFP_NOWARN);
174 if (!ret)
175 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
176 return ret;
177 }
178
179 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
180 struct ext4_group_desc *bg)
181 {
182 return le32_to_cpu(bg->bg_block_bitmap_lo) |
183 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
184 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
185 }
186
187 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
188 struct ext4_group_desc *bg)
189 {
190 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
191 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
192 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
193 }
194
195 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
196 struct ext4_group_desc *bg)
197 {
198 return le32_to_cpu(bg->bg_inode_table_lo) |
199 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
200 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
201 }
202
203 __u32 ext4_free_group_clusters(struct super_block *sb,
204 struct ext4_group_desc *bg)
205 {
206 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
207 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
208 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
209 }
210
211 __u32 ext4_free_inodes_count(struct super_block *sb,
212 struct ext4_group_desc *bg)
213 {
214 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
215 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
216 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
217 }
218
219 __u32 ext4_used_dirs_count(struct super_block *sb,
220 struct ext4_group_desc *bg)
221 {
222 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
223 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
224 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
225 }
226
227 __u32 ext4_itable_unused_count(struct super_block *sb,
228 struct ext4_group_desc *bg)
229 {
230 return le16_to_cpu(bg->bg_itable_unused_lo) |
231 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
232 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
233 }
234
235 void ext4_block_bitmap_set(struct super_block *sb,
236 struct ext4_group_desc *bg, ext4_fsblk_t blk)
237 {
238 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
239 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
240 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
241 }
242
243 void ext4_inode_bitmap_set(struct super_block *sb,
244 struct ext4_group_desc *bg, ext4_fsblk_t blk)
245 {
246 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
247 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
248 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
249 }
250
251 void ext4_inode_table_set(struct super_block *sb,
252 struct ext4_group_desc *bg, ext4_fsblk_t blk)
253 {
254 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
255 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
256 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
257 }
258
259 void ext4_free_group_clusters_set(struct super_block *sb,
260 struct ext4_group_desc *bg, __u32 count)
261 {
262 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
263 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
264 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
265 }
266
267 void ext4_free_inodes_set(struct super_block *sb,
268 struct ext4_group_desc *bg, __u32 count)
269 {
270 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
271 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
272 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
273 }
274
275 void ext4_used_dirs_set(struct super_block *sb,
276 struct ext4_group_desc *bg, __u32 count)
277 {
278 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
279 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
280 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
281 }
282
283 void ext4_itable_unused_set(struct super_block *sb,
284 struct ext4_group_desc *bg, __u32 count)
285 {
286 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
287 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
288 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
289 }
290
291
292 static void __save_error_info(struct super_block *sb, const char *func,
293 unsigned int line)
294 {
295 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
296
297 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
298 if (bdev_read_only(sb->s_bdev))
299 return;
300 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
301 es->s_last_error_time = cpu_to_le32(get_seconds());
302 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
303 es->s_last_error_line = cpu_to_le32(line);
304 if (!es->s_first_error_time) {
305 es->s_first_error_time = es->s_last_error_time;
306 strncpy(es->s_first_error_func, func,
307 sizeof(es->s_first_error_func));
308 es->s_first_error_line = cpu_to_le32(line);
309 es->s_first_error_ino = es->s_last_error_ino;
310 es->s_first_error_block = es->s_last_error_block;
311 }
312 /*
313 * Start the daily error reporting function if it hasn't been
314 * started already
315 */
316 if (!es->s_error_count)
317 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
318 le32_add_cpu(&es->s_error_count, 1);
319 }
320
321 static void save_error_info(struct super_block *sb, const char *func,
322 unsigned int line)
323 {
324 __save_error_info(sb, func, line);
325 ext4_commit_super(sb, 1);
326 }
327
328 /*
329 * The del_gendisk() function uninitializes the disk-specific data
330 * structures, including the bdi structure, without telling anyone
331 * else. Once this happens, any attempt to call mark_buffer_dirty()
332 * (for example, by ext4_commit_super), will cause a kernel OOPS.
333 * This is a kludge to prevent these oops until we can put in a proper
334 * hook in del_gendisk() to inform the VFS and file system layers.
335 */
336 static int block_device_ejected(struct super_block *sb)
337 {
338 struct inode *bd_inode = sb->s_bdev->bd_inode;
339 struct backing_dev_info *bdi = inode_to_bdi(bd_inode);
340
341 return bdi->dev == NULL;
342 }
343
344 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
345 {
346 struct super_block *sb = journal->j_private;
347 struct ext4_sb_info *sbi = EXT4_SB(sb);
348 int error = is_journal_aborted(journal);
349 struct ext4_journal_cb_entry *jce;
350
351 BUG_ON(txn->t_state == T_FINISHED);
352 spin_lock(&sbi->s_md_lock);
353 while (!list_empty(&txn->t_private_list)) {
354 jce = list_entry(txn->t_private_list.next,
355 struct ext4_journal_cb_entry, jce_list);
356 list_del_init(&jce->jce_list);
357 spin_unlock(&sbi->s_md_lock);
358 jce->jce_func(sb, jce, error);
359 spin_lock(&sbi->s_md_lock);
360 }
361 spin_unlock(&sbi->s_md_lock);
362 }
363
364 /* Deal with the reporting of failure conditions on a filesystem such as
365 * inconsistencies detected or read IO failures.
366 *
367 * On ext2, we can store the error state of the filesystem in the
368 * superblock. That is not possible on ext4, because we may have other
369 * write ordering constraints on the superblock which prevent us from
370 * writing it out straight away; and given that the journal is about to
371 * be aborted, we can't rely on the current, or future, transactions to
372 * write out the superblock safely.
373 *
374 * We'll just use the jbd2_journal_abort() error code to record an error in
375 * the journal instead. On recovery, the journal will complain about
376 * that error until we've noted it down and cleared it.
377 */
378
379 static void ext4_handle_error(struct super_block *sb)
380 {
381 if (sb->s_flags & MS_RDONLY)
382 return;
383
384 if (!test_opt(sb, ERRORS_CONT)) {
385 journal_t *journal = EXT4_SB(sb)->s_journal;
386
387 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
388 if (journal)
389 jbd2_journal_abort(journal, -EIO);
390 }
391 if (test_opt(sb, ERRORS_RO)) {
392 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
393 /*
394 * Make sure updated value of ->s_mount_flags will be visible
395 * before ->s_flags update
396 */
397 smp_wmb();
398 sb->s_flags |= MS_RDONLY;
399 }
400 if (test_opt(sb, ERRORS_PANIC))
401 panic("EXT4-fs (device %s): panic forced after error\n",
402 sb->s_id);
403 }
404
405 #define ext4_error_ratelimit(sb) \
406 ___ratelimit(&(EXT4_SB(sb)->s_err_ratelimit_state), \
407 "EXT4-fs error")
408
409 void __ext4_error(struct super_block *sb, const char *function,
410 unsigned int line, const char *fmt, ...)
411 {
412 struct va_format vaf;
413 va_list args;
414
415 if (ext4_error_ratelimit(sb)) {
416 va_start(args, fmt);
417 vaf.fmt = fmt;
418 vaf.va = &args;
419 printk(KERN_CRIT
420 "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
421 sb->s_id, function, line, current->comm, &vaf);
422 va_end(args);
423 }
424 save_error_info(sb, function, line);
425 ext4_handle_error(sb);
426 }
427
428 void __ext4_error_inode(struct inode *inode, const char *function,
429 unsigned int line, ext4_fsblk_t block,
430 const char *fmt, ...)
431 {
432 va_list args;
433 struct va_format vaf;
434 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
435
436 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
437 es->s_last_error_block = cpu_to_le64(block);
438 if (ext4_error_ratelimit(inode->i_sb)) {
439 va_start(args, fmt);
440 vaf.fmt = fmt;
441 vaf.va = &args;
442 if (block)
443 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
444 "inode #%lu: block %llu: comm %s: %pV\n",
445 inode->i_sb->s_id, function, line, inode->i_ino,
446 block, current->comm, &vaf);
447 else
448 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
449 "inode #%lu: comm %s: %pV\n",
450 inode->i_sb->s_id, function, line, inode->i_ino,
451 current->comm, &vaf);
452 va_end(args);
453 }
454 save_error_info(inode->i_sb, function, line);
455 ext4_handle_error(inode->i_sb);
456 }
457
458 void __ext4_error_file(struct file *file, const char *function,
459 unsigned int line, ext4_fsblk_t block,
460 const char *fmt, ...)
461 {
462 va_list args;
463 struct va_format vaf;
464 struct ext4_super_block *es;
465 struct inode *inode = file_inode(file);
466 char pathname[80], *path;
467
468 es = EXT4_SB(inode->i_sb)->s_es;
469 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
470 if (ext4_error_ratelimit(inode->i_sb)) {
471 path = file_path(file, pathname, sizeof(pathname));
472 if (IS_ERR(path))
473 path = "(unknown)";
474 va_start(args, fmt);
475 vaf.fmt = fmt;
476 vaf.va = &args;
477 if (block)
478 printk(KERN_CRIT
479 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
480 "block %llu: comm %s: path %s: %pV\n",
481 inode->i_sb->s_id, function, line, inode->i_ino,
482 block, current->comm, path, &vaf);
483 else
484 printk(KERN_CRIT
485 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
486 "comm %s: path %s: %pV\n",
487 inode->i_sb->s_id, function, line, inode->i_ino,
488 current->comm, path, &vaf);
489 va_end(args);
490 }
491 save_error_info(inode->i_sb, function, line);
492 ext4_handle_error(inode->i_sb);
493 }
494
495 const char *ext4_decode_error(struct super_block *sb, int errno,
496 char nbuf[16])
497 {
498 char *errstr = NULL;
499
500 switch (errno) {
501 case -EIO:
502 errstr = "IO failure";
503 break;
504 case -ENOMEM:
505 errstr = "Out of memory";
506 break;
507 case -EROFS:
508 if (!sb || (EXT4_SB(sb)->s_journal &&
509 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
510 errstr = "Journal has aborted";
511 else
512 errstr = "Readonly filesystem";
513 break;
514 default:
515 /* If the caller passed in an extra buffer for unknown
516 * errors, textualise them now. Else we just return
517 * NULL. */
518 if (nbuf) {
519 /* Check for truncated error codes... */
520 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
521 errstr = nbuf;
522 }
523 break;
524 }
525
526 return errstr;
527 }
528
529 /* __ext4_std_error decodes expected errors from journaling functions
530 * automatically and invokes the appropriate error response. */
531
532 void __ext4_std_error(struct super_block *sb, const char *function,
533 unsigned int line, int errno)
534 {
535 char nbuf[16];
536 const char *errstr;
537
538 /* Special case: if the error is EROFS, and we're not already
539 * inside a transaction, then there's really no point in logging
540 * an error. */
541 if (errno == -EROFS && journal_current_handle() == NULL &&
542 (sb->s_flags & MS_RDONLY))
543 return;
544
545 if (ext4_error_ratelimit(sb)) {
546 errstr = ext4_decode_error(sb, errno, nbuf);
547 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
548 sb->s_id, function, line, errstr);
549 }
550
551 save_error_info(sb, function, line);
552 ext4_handle_error(sb);
553 }
554
555 /*
556 * ext4_abort is a much stronger failure handler than ext4_error. The
557 * abort function may be used to deal with unrecoverable failures such
558 * as journal IO errors or ENOMEM at a critical moment in log management.
559 *
560 * We unconditionally force the filesystem into an ABORT|READONLY state,
561 * unless the error response on the fs has been set to panic in which
562 * case we take the easy way out and panic immediately.
563 */
564
565 void __ext4_abort(struct super_block *sb, const char *function,
566 unsigned int line, const char *fmt, ...)
567 {
568 va_list args;
569
570 save_error_info(sb, function, line);
571 va_start(args, fmt);
572 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
573 function, line);
574 vprintk(fmt, args);
575 printk("\n");
576 va_end(args);
577
578 if ((sb->s_flags & MS_RDONLY) == 0) {
579 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
580 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
581 /*
582 * Make sure updated value of ->s_mount_flags will be visible
583 * before ->s_flags update
584 */
585 smp_wmb();
586 sb->s_flags |= MS_RDONLY;
587 if (EXT4_SB(sb)->s_journal)
588 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
589 save_error_info(sb, function, line);
590 }
591 if (test_opt(sb, ERRORS_PANIC))
592 panic("EXT4-fs panic from previous error\n");
593 }
594
595 void __ext4_msg(struct super_block *sb,
596 const char *prefix, const char *fmt, ...)
597 {
598 struct va_format vaf;
599 va_list args;
600
601 if (!___ratelimit(&(EXT4_SB(sb)->s_msg_ratelimit_state), "EXT4-fs"))
602 return;
603
604 va_start(args, fmt);
605 vaf.fmt = fmt;
606 vaf.va = &args;
607 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
608 va_end(args);
609 }
610
611 #define ext4_warning_ratelimit(sb) \
612 ___ratelimit(&(EXT4_SB(sb)->s_warning_ratelimit_state), \
613 "EXT4-fs warning")
614
615 void __ext4_warning(struct super_block *sb, const char *function,
616 unsigned int line, const char *fmt, ...)
617 {
618 struct va_format vaf;
619 va_list args;
620
621 if (!ext4_warning_ratelimit(sb))
622 return;
623
624 va_start(args, fmt);
625 vaf.fmt = fmt;
626 vaf.va = &args;
627 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
628 sb->s_id, function, line, &vaf);
629 va_end(args);
630 }
631
632 void __ext4_warning_inode(const struct inode *inode, const char *function,
633 unsigned int line, const char *fmt, ...)
634 {
635 struct va_format vaf;
636 va_list args;
637
638 if (!ext4_warning_ratelimit(inode->i_sb))
639 return;
640
641 va_start(args, fmt);
642 vaf.fmt = fmt;
643 vaf.va = &args;
644 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: "
645 "inode #%lu: comm %s: %pV\n", inode->i_sb->s_id,
646 function, line, inode->i_ino, current->comm, &vaf);
647 va_end(args);
648 }
649
650 void __ext4_grp_locked_error(const char *function, unsigned int line,
651 struct super_block *sb, ext4_group_t grp,
652 unsigned long ino, ext4_fsblk_t block,
653 const char *fmt, ...)
654 __releases(bitlock)
655 __acquires(bitlock)
656 {
657 struct va_format vaf;
658 va_list args;
659 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
660
661 es->s_last_error_ino = cpu_to_le32(ino);
662 es->s_last_error_block = cpu_to_le64(block);
663 __save_error_info(sb, function, line);
664
665 if (ext4_error_ratelimit(sb)) {
666 va_start(args, fmt);
667 vaf.fmt = fmt;
668 vaf.va = &args;
669 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
670 sb->s_id, function, line, grp);
671 if (ino)
672 printk(KERN_CONT "inode %lu: ", ino);
673 if (block)
674 printk(KERN_CONT "block %llu:",
675 (unsigned long long) block);
676 printk(KERN_CONT "%pV\n", &vaf);
677 va_end(args);
678 }
679
680 if (test_opt(sb, ERRORS_CONT)) {
681 ext4_commit_super(sb, 0);
682 return;
683 }
684
685 ext4_unlock_group(sb, grp);
686 ext4_handle_error(sb);
687 /*
688 * We only get here in the ERRORS_RO case; relocking the group
689 * may be dangerous, but nothing bad will happen since the
690 * filesystem will have already been marked read/only and the
691 * journal has been aborted. We return 1 as a hint to callers
692 * who might what to use the return value from
693 * ext4_grp_locked_error() to distinguish between the
694 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
695 * aggressively from the ext4 function in question, with a
696 * more appropriate error code.
697 */
698 ext4_lock_group(sb, grp);
699 return;
700 }
701
702 void ext4_update_dynamic_rev(struct super_block *sb)
703 {
704 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
705
706 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
707 return;
708
709 ext4_warning(sb,
710 "updating to rev %d because of new feature flag, "
711 "running e2fsck is recommended",
712 EXT4_DYNAMIC_REV);
713
714 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
715 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
716 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
717 /* leave es->s_feature_*compat flags alone */
718 /* es->s_uuid will be set by e2fsck if empty */
719
720 /*
721 * The rest of the superblock fields should be zero, and if not it
722 * means they are likely already in use, so leave them alone. We
723 * can leave it up to e2fsck to clean up any inconsistencies there.
724 */
725 }
726
727 /*
728 * Open the external journal device
729 */
730 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
731 {
732 struct block_device *bdev;
733 char b[BDEVNAME_SIZE];
734
735 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
736 if (IS_ERR(bdev))
737 goto fail;
738 return bdev;
739
740 fail:
741 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
742 __bdevname(dev, b), PTR_ERR(bdev));
743 return NULL;
744 }
745
746 /*
747 * Release the journal device
748 */
749 static void ext4_blkdev_put(struct block_device *bdev)
750 {
751 blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
752 }
753
754 static void ext4_blkdev_remove(struct ext4_sb_info *sbi)
755 {
756 struct block_device *bdev;
757 bdev = sbi->journal_bdev;
758 if (bdev) {
759 ext4_blkdev_put(bdev);
760 sbi->journal_bdev = NULL;
761 }
762 }
763
764 static inline struct inode *orphan_list_entry(struct list_head *l)
765 {
766 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
767 }
768
769 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
770 {
771 struct list_head *l;
772
773 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
774 le32_to_cpu(sbi->s_es->s_last_orphan));
775
776 printk(KERN_ERR "sb_info orphan list:\n");
777 list_for_each(l, &sbi->s_orphan) {
778 struct inode *inode = orphan_list_entry(l);
779 printk(KERN_ERR " "
780 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
781 inode->i_sb->s_id, inode->i_ino, inode,
782 inode->i_mode, inode->i_nlink,
783 NEXT_ORPHAN(inode));
784 }
785 }
786
787 static void ext4_put_super(struct super_block *sb)
788 {
789 struct ext4_sb_info *sbi = EXT4_SB(sb);
790 struct ext4_super_block *es = sbi->s_es;
791 int i, err;
792
793 ext4_unregister_li_request(sb);
794 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
795
796 flush_workqueue(sbi->rsv_conversion_wq);
797 destroy_workqueue(sbi->rsv_conversion_wq);
798
799 if (sbi->s_journal) {
800 err = jbd2_journal_destroy(sbi->s_journal);
801 sbi->s_journal = NULL;
802 if (err < 0)
803 ext4_abort(sb, "Couldn't clean up the journal");
804 }
805
806 ext4_es_unregister_shrinker(sbi);
807 del_timer_sync(&sbi->s_err_report);
808 ext4_release_system_zone(sb);
809 ext4_mb_release(sb);
810 ext4_ext_release(sb);
811 ext4_xattr_put_super(sb);
812
813 if (!(sb->s_flags & MS_RDONLY)) {
814 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
815 es->s_state = cpu_to_le16(sbi->s_mount_state);
816 }
817 if (!(sb->s_flags & MS_RDONLY))
818 ext4_commit_super(sb, 1);
819
820 if (sbi->s_proc) {
821 remove_proc_entry("options", sbi->s_proc);
822 remove_proc_entry(sb->s_id, ext4_proc_root);
823 }
824 kobject_del(&sbi->s_kobj);
825
826 for (i = 0; i < sbi->s_gdb_count; i++)
827 brelse(sbi->s_group_desc[i]);
828 kvfree(sbi->s_group_desc);
829 kvfree(sbi->s_flex_groups);
830 percpu_counter_destroy(&sbi->s_freeclusters_counter);
831 percpu_counter_destroy(&sbi->s_freeinodes_counter);
832 percpu_counter_destroy(&sbi->s_dirs_counter);
833 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
834 brelse(sbi->s_sbh);
835 #ifdef CONFIG_QUOTA
836 for (i = 0; i < EXT4_MAXQUOTAS; i++)
837 kfree(sbi->s_qf_names[i]);
838 #endif
839
840 /* Debugging code just in case the in-memory inode orphan list
841 * isn't empty. The on-disk one can be non-empty if we've
842 * detected an error and taken the fs readonly, but the
843 * in-memory list had better be clean by this point. */
844 if (!list_empty(&sbi->s_orphan))
845 dump_orphan_list(sb, sbi);
846 J_ASSERT(list_empty(&sbi->s_orphan));
847
848 sync_blockdev(sb->s_bdev);
849 invalidate_bdev(sb->s_bdev);
850 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
851 /*
852 * Invalidate the journal device's buffers. We don't want them
853 * floating about in memory - the physical journal device may
854 * hotswapped, and it breaks the `ro-after' testing code.
855 */
856 sync_blockdev(sbi->journal_bdev);
857 invalidate_bdev(sbi->journal_bdev);
858 ext4_blkdev_remove(sbi);
859 }
860 if (sbi->s_mb_cache) {
861 ext4_xattr_destroy_cache(sbi->s_mb_cache);
862 sbi->s_mb_cache = NULL;
863 }
864 if (sbi->s_mmp_tsk)
865 kthread_stop(sbi->s_mmp_tsk);
866 sb->s_fs_info = NULL;
867 /*
868 * Now that we are completely done shutting down the
869 * superblock, we need to actually destroy the kobject.
870 */
871 kobject_put(&sbi->s_kobj);
872 wait_for_completion(&sbi->s_kobj_unregister);
873 if (sbi->s_chksum_driver)
874 crypto_free_shash(sbi->s_chksum_driver);
875 kfree(sbi->s_blockgroup_lock);
876 kfree(sbi);
877 }
878
879 static struct kmem_cache *ext4_inode_cachep;
880
881 /*
882 * Called inside transaction, so use GFP_NOFS
883 */
884 static struct inode *ext4_alloc_inode(struct super_block *sb)
885 {
886 struct ext4_inode_info *ei;
887
888 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
889 if (!ei)
890 return NULL;
891
892 ei->vfs_inode.i_version = 1;
893 spin_lock_init(&ei->i_raw_lock);
894 INIT_LIST_HEAD(&ei->i_prealloc_list);
895 spin_lock_init(&ei->i_prealloc_lock);
896 ext4_es_init_tree(&ei->i_es_tree);
897 rwlock_init(&ei->i_es_lock);
898 INIT_LIST_HEAD(&ei->i_es_list);
899 ei->i_es_all_nr = 0;
900 ei->i_es_shk_nr = 0;
901 ei->i_es_shrink_lblk = 0;
902 ei->i_reserved_data_blocks = 0;
903 ei->i_reserved_meta_blocks = 0;
904 ei->i_allocated_meta_blocks = 0;
905 ei->i_da_metadata_calc_len = 0;
906 ei->i_da_metadata_calc_last_lblock = 0;
907 spin_lock_init(&(ei->i_block_reservation_lock));
908 #ifdef CONFIG_QUOTA
909 ei->i_reserved_quota = 0;
910 memset(&ei->i_dquot, 0, sizeof(ei->i_dquot));
911 #endif
912 ei->jinode = NULL;
913 INIT_LIST_HEAD(&ei->i_rsv_conversion_list);
914 spin_lock_init(&ei->i_completed_io_lock);
915 ei->i_sync_tid = 0;
916 ei->i_datasync_tid = 0;
917 atomic_set(&ei->i_ioend_count, 0);
918 atomic_set(&ei->i_unwritten, 0);
919 INIT_WORK(&ei->i_rsv_conversion_work, ext4_end_io_rsv_work);
920 #ifdef CONFIG_EXT4_FS_ENCRYPTION
921 ei->i_crypt_info = NULL;
922 #endif
923 return &ei->vfs_inode;
924 }
925
926 static int ext4_drop_inode(struct inode *inode)
927 {
928 int drop = generic_drop_inode(inode);
929
930 trace_ext4_drop_inode(inode, drop);
931 return drop;
932 }
933
934 static void ext4_i_callback(struct rcu_head *head)
935 {
936 struct inode *inode = container_of(head, struct inode, i_rcu);
937 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
938 }
939
940 static void ext4_destroy_inode(struct inode *inode)
941 {
942 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
943 ext4_msg(inode->i_sb, KERN_ERR,
944 "Inode %lu (%p): orphan list check failed!",
945 inode->i_ino, EXT4_I(inode));
946 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
947 EXT4_I(inode), sizeof(struct ext4_inode_info),
948 true);
949 dump_stack();
950 }
951 call_rcu(&inode->i_rcu, ext4_i_callback);
952 }
953
954 static void init_once(void *foo)
955 {
956 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
957
958 INIT_LIST_HEAD(&ei->i_orphan);
959 init_rwsem(&ei->xattr_sem);
960 init_rwsem(&ei->i_data_sem);
961 inode_init_once(&ei->vfs_inode);
962 }
963
964 static int __init init_inodecache(void)
965 {
966 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
967 sizeof(struct ext4_inode_info),
968 0, (SLAB_RECLAIM_ACCOUNT|
969 SLAB_MEM_SPREAD),
970 init_once);
971 if (ext4_inode_cachep == NULL)
972 return -ENOMEM;
973 return 0;
974 }
975
976 static void destroy_inodecache(void)
977 {
978 /*
979 * Make sure all delayed rcu free inodes are flushed before we
980 * destroy cache.
981 */
982 rcu_barrier();
983 kmem_cache_destroy(ext4_inode_cachep);
984 }
985
986 void ext4_clear_inode(struct inode *inode)
987 {
988 invalidate_inode_buffers(inode);
989 clear_inode(inode);
990 dquot_drop(inode);
991 ext4_discard_preallocations(inode);
992 ext4_es_remove_extent(inode, 0, EXT_MAX_BLOCKS);
993 if (EXT4_I(inode)->jinode) {
994 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
995 EXT4_I(inode)->jinode);
996 jbd2_free_inode(EXT4_I(inode)->jinode);
997 EXT4_I(inode)->jinode = NULL;
998 }
999 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1000 if (EXT4_I(inode)->i_crypt_info)
1001 ext4_free_encryption_info(inode, EXT4_I(inode)->i_crypt_info);
1002 #endif
1003 }
1004
1005 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1006 u64 ino, u32 generation)
1007 {
1008 struct inode *inode;
1009
1010 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
1011 return ERR_PTR(-ESTALE);
1012 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
1013 return ERR_PTR(-ESTALE);
1014
1015 /* iget isn't really right if the inode is currently unallocated!!
1016 *
1017 * ext4_read_inode will return a bad_inode if the inode had been
1018 * deleted, so we should be safe.
1019 *
1020 * Currently we don't know the generation for parent directory, so
1021 * a generation of 0 means "accept any"
1022 */
1023 inode = ext4_iget_normal(sb, ino);
1024 if (IS_ERR(inode))
1025 return ERR_CAST(inode);
1026 if (generation && inode->i_generation != generation) {
1027 iput(inode);
1028 return ERR_PTR(-ESTALE);
1029 }
1030
1031 return inode;
1032 }
1033
1034 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1035 int fh_len, int fh_type)
1036 {
1037 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1038 ext4_nfs_get_inode);
1039 }
1040
1041 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1042 int fh_len, int fh_type)
1043 {
1044 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1045 ext4_nfs_get_inode);
1046 }
1047
1048 /*
1049 * Try to release metadata pages (indirect blocks, directories) which are
1050 * mapped via the block device. Since these pages could have journal heads
1051 * which would prevent try_to_free_buffers() from freeing them, we must use
1052 * jbd2 layer's try_to_free_buffers() function to release them.
1053 */
1054 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1055 gfp_t wait)
1056 {
1057 journal_t *journal = EXT4_SB(sb)->s_journal;
1058
1059 WARN_ON(PageChecked(page));
1060 if (!page_has_buffers(page))
1061 return 0;
1062 if (journal)
1063 return jbd2_journal_try_to_free_buffers(journal, page,
1064 wait & ~__GFP_WAIT);
1065 return try_to_free_buffers(page);
1066 }
1067
1068 #ifdef CONFIG_QUOTA
1069 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1070 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1071
1072 static int ext4_write_dquot(struct dquot *dquot);
1073 static int ext4_acquire_dquot(struct dquot *dquot);
1074 static int ext4_release_dquot(struct dquot *dquot);
1075 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1076 static int ext4_write_info(struct super_block *sb, int type);
1077 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1078 struct path *path);
1079 static int ext4_quota_off(struct super_block *sb, int type);
1080 static int ext4_quota_on_mount(struct super_block *sb, int type);
1081 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1082 size_t len, loff_t off);
1083 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1084 const char *data, size_t len, loff_t off);
1085 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1086 unsigned int flags);
1087 static int ext4_enable_quotas(struct super_block *sb);
1088
1089 static struct dquot **ext4_get_dquots(struct inode *inode)
1090 {
1091 return EXT4_I(inode)->i_dquot;
1092 }
1093
1094 static const struct dquot_operations ext4_quota_operations = {
1095 .get_reserved_space = ext4_get_reserved_space,
1096 .write_dquot = ext4_write_dquot,
1097 .acquire_dquot = ext4_acquire_dquot,
1098 .release_dquot = ext4_release_dquot,
1099 .mark_dirty = ext4_mark_dquot_dirty,
1100 .write_info = ext4_write_info,
1101 .alloc_dquot = dquot_alloc,
1102 .destroy_dquot = dquot_destroy,
1103 };
1104
1105 static const struct quotactl_ops ext4_qctl_operations = {
1106 .quota_on = ext4_quota_on,
1107 .quota_off = ext4_quota_off,
1108 .quota_sync = dquot_quota_sync,
1109 .get_state = dquot_get_state,
1110 .set_info = dquot_set_dqinfo,
1111 .get_dqblk = dquot_get_dqblk,
1112 .set_dqblk = dquot_set_dqblk
1113 };
1114 #endif
1115
1116 static const struct super_operations ext4_sops = {
1117 .alloc_inode = ext4_alloc_inode,
1118 .destroy_inode = ext4_destroy_inode,
1119 .write_inode = ext4_write_inode,
1120 .dirty_inode = ext4_dirty_inode,
1121 .drop_inode = ext4_drop_inode,
1122 .evict_inode = ext4_evict_inode,
1123 .put_super = ext4_put_super,
1124 .sync_fs = ext4_sync_fs,
1125 .freeze_fs = ext4_freeze,
1126 .unfreeze_fs = ext4_unfreeze,
1127 .statfs = ext4_statfs,
1128 .remount_fs = ext4_remount,
1129 .show_options = ext4_show_options,
1130 #ifdef CONFIG_QUOTA
1131 .quota_read = ext4_quota_read,
1132 .quota_write = ext4_quota_write,
1133 .get_dquots = ext4_get_dquots,
1134 #endif
1135 .bdev_try_to_free_page = bdev_try_to_free_page,
1136 };
1137
1138 static const struct export_operations ext4_export_ops = {
1139 .fh_to_dentry = ext4_fh_to_dentry,
1140 .fh_to_parent = ext4_fh_to_parent,
1141 .get_parent = ext4_get_parent,
1142 };
1143
1144 enum {
1145 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1146 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1147 Opt_nouid32, Opt_debug, Opt_removed,
1148 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1149 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1150 Opt_commit, Opt_min_batch_time, Opt_max_batch_time, Opt_journal_dev,
1151 Opt_journal_path, Opt_journal_checksum, Opt_journal_async_commit,
1152 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1153 Opt_data_err_abort, Opt_data_err_ignore, Opt_test_dummy_encryption,
1154 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1155 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1156 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1157 Opt_usrquota, Opt_grpquota, Opt_i_version, Opt_dax,
1158 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1159 Opt_lazytime, Opt_nolazytime,
1160 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1161 Opt_inode_readahead_blks, Opt_journal_ioprio,
1162 Opt_dioread_nolock, Opt_dioread_lock,
1163 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1164 Opt_max_dir_size_kb, Opt_nojournal_checksum,
1165 };
1166
1167 static const match_table_t tokens = {
1168 {Opt_bsd_df, "bsddf"},
1169 {Opt_minix_df, "minixdf"},
1170 {Opt_grpid, "grpid"},
1171 {Opt_grpid, "bsdgroups"},
1172 {Opt_nogrpid, "nogrpid"},
1173 {Opt_nogrpid, "sysvgroups"},
1174 {Opt_resgid, "resgid=%u"},
1175 {Opt_resuid, "resuid=%u"},
1176 {Opt_sb, "sb=%u"},
1177 {Opt_err_cont, "errors=continue"},
1178 {Opt_err_panic, "errors=panic"},
1179 {Opt_err_ro, "errors=remount-ro"},
1180 {Opt_nouid32, "nouid32"},
1181 {Opt_debug, "debug"},
1182 {Opt_removed, "oldalloc"},
1183 {Opt_removed, "orlov"},
1184 {Opt_user_xattr, "user_xattr"},
1185 {Opt_nouser_xattr, "nouser_xattr"},
1186 {Opt_acl, "acl"},
1187 {Opt_noacl, "noacl"},
1188 {Opt_noload, "norecovery"},
1189 {Opt_noload, "noload"},
1190 {Opt_removed, "nobh"},
1191 {Opt_removed, "bh"},
1192 {Opt_commit, "commit=%u"},
1193 {Opt_min_batch_time, "min_batch_time=%u"},
1194 {Opt_max_batch_time, "max_batch_time=%u"},
1195 {Opt_journal_dev, "journal_dev=%u"},
1196 {Opt_journal_path, "journal_path=%s"},
1197 {Opt_journal_checksum, "journal_checksum"},
1198 {Opt_nojournal_checksum, "nojournal_checksum"},
1199 {Opt_journal_async_commit, "journal_async_commit"},
1200 {Opt_abort, "abort"},
1201 {Opt_data_journal, "data=journal"},
1202 {Opt_data_ordered, "data=ordered"},
1203 {Opt_data_writeback, "data=writeback"},
1204 {Opt_data_err_abort, "data_err=abort"},
1205 {Opt_data_err_ignore, "data_err=ignore"},
1206 {Opt_offusrjquota, "usrjquota="},
1207 {Opt_usrjquota, "usrjquota=%s"},
1208 {Opt_offgrpjquota, "grpjquota="},
1209 {Opt_grpjquota, "grpjquota=%s"},
1210 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1211 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1212 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1213 {Opt_grpquota, "grpquota"},
1214 {Opt_noquota, "noquota"},
1215 {Opt_quota, "quota"},
1216 {Opt_usrquota, "usrquota"},
1217 {Opt_barrier, "barrier=%u"},
1218 {Opt_barrier, "barrier"},
1219 {Opt_nobarrier, "nobarrier"},
1220 {Opt_i_version, "i_version"},
1221 {Opt_dax, "dax"},
1222 {Opt_stripe, "stripe=%u"},
1223 {Opt_delalloc, "delalloc"},
1224 {Opt_lazytime, "lazytime"},
1225 {Opt_nolazytime, "nolazytime"},
1226 {Opt_nodelalloc, "nodelalloc"},
1227 {Opt_removed, "mblk_io_submit"},
1228 {Opt_removed, "nomblk_io_submit"},
1229 {Opt_block_validity, "block_validity"},
1230 {Opt_noblock_validity, "noblock_validity"},
1231 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1232 {Opt_journal_ioprio, "journal_ioprio=%u"},
1233 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1234 {Opt_auto_da_alloc, "auto_da_alloc"},
1235 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1236 {Opt_dioread_nolock, "dioread_nolock"},
1237 {Opt_dioread_lock, "dioread_lock"},
1238 {Opt_discard, "discard"},
1239 {Opt_nodiscard, "nodiscard"},
1240 {Opt_init_itable, "init_itable=%u"},
1241 {Opt_init_itable, "init_itable"},
1242 {Opt_noinit_itable, "noinit_itable"},
1243 {Opt_max_dir_size_kb, "max_dir_size_kb=%u"},
1244 {Opt_test_dummy_encryption, "test_dummy_encryption"},
1245 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1246 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1247 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1248 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1249 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
1250 {Opt_err, NULL},
1251 };
1252
1253 static ext4_fsblk_t get_sb_block(void **data)
1254 {
1255 ext4_fsblk_t sb_block;
1256 char *options = (char *) *data;
1257
1258 if (!options || strncmp(options, "sb=", 3) != 0)
1259 return 1; /* Default location */
1260
1261 options += 3;
1262 /* TODO: use simple_strtoll with >32bit ext4 */
1263 sb_block = simple_strtoul(options, &options, 0);
1264 if (*options && *options != ',') {
1265 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1266 (char *) *data);
1267 return 1;
1268 }
1269 if (*options == ',')
1270 options++;
1271 *data = (void *) options;
1272
1273 return sb_block;
1274 }
1275
1276 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1277 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1278 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1279
1280 #ifdef CONFIG_QUOTA
1281 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1282 {
1283 struct ext4_sb_info *sbi = EXT4_SB(sb);
1284 char *qname;
1285 int ret = -1;
1286
1287 if (sb_any_quota_loaded(sb) &&
1288 !sbi->s_qf_names[qtype]) {
1289 ext4_msg(sb, KERN_ERR,
1290 "Cannot change journaled "
1291 "quota options when quota turned on");
1292 return -1;
1293 }
1294 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1295 ext4_msg(sb, KERN_ERR, "Cannot set journaled quota options "
1296 "when QUOTA feature is enabled");
1297 return -1;
1298 }
1299 qname = match_strdup(args);
1300 if (!qname) {
1301 ext4_msg(sb, KERN_ERR,
1302 "Not enough memory for storing quotafile name");
1303 return -1;
1304 }
1305 if (sbi->s_qf_names[qtype]) {
1306 if (strcmp(sbi->s_qf_names[qtype], qname) == 0)
1307 ret = 1;
1308 else
1309 ext4_msg(sb, KERN_ERR,
1310 "%s quota file already specified",
1311 QTYPE2NAME(qtype));
1312 goto errout;
1313 }
1314 if (strchr(qname, '/')) {
1315 ext4_msg(sb, KERN_ERR,
1316 "quotafile must be on filesystem root");
1317 goto errout;
1318 }
1319 sbi->s_qf_names[qtype] = qname;
1320 set_opt(sb, QUOTA);
1321 return 1;
1322 errout:
1323 kfree(qname);
1324 return ret;
1325 }
1326
1327 static int clear_qf_name(struct super_block *sb, int qtype)
1328 {
1329
1330 struct ext4_sb_info *sbi = EXT4_SB(sb);
1331
1332 if (sb_any_quota_loaded(sb) &&
1333 sbi->s_qf_names[qtype]) {
1334 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1335 " when quota turned on");
1336 return -1;
1337 }
1338 kfree(sbi->s_qf_names[qtype]);
1339 sbi->s_qf_names[qtype] = NULL;
1340 return 1;
1341 }
1342 #endif
1343
1344 #define MOPT_SET 0x0001
1345 #define MOPT_CLEAR 0x0002
1346 #define MOPT_NOSUPPORT 0x0004
1347 #define MOPT_EXPLICIT 0x0008
1348 #define MOPT_CLEAR_ERR 0x0010
1349 #define MOPT_GTE0 0x0020
1350 #ifdef CONFIG_QUOTA
1351 #define MOPT_Q 0
1352 #define MOPT_QFMT 0x0040
1353 #else
1354 #define MOPT_Q MOPT_NOSUPPORT
1355 #define MOPT_QFMT MOPT_NOSUPPORT
1356 #endif
1357 #define MOPT_DATAJ 0x0080
1358 #define MOPT_NO_EXT2 0x0100
1359 #define MOPT_NO_EXT3 0x0200
1360 #define MOPT_EXT4_ONLY (MOPT_NO_EXT2 | MOPT_NO_EXT3)
1361 #define MOPT_STRING 0x0400
1362
1363 static const struct mount_opts {
1364 int token;
1365 int mount_opt;
1366 int flags;
1367 } ext4_mount_opts[] = {
1368 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1369 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1370 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1371 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1372 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1373 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1374 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK,
1375 MOPT_EXT4_ONLY | MOPT_SET},
1376 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK,
1377 MOPT_EXT4_ONLY | MOPT_CLEAR},
1378 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1379 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1380 {Opt_delalloc, EXT4_MOUNT_DELALLOC,
1381 MOPT_EXT4_ONLY | MOPT_SET | MOPT_EXPLICIT},
1382 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC,
1383 MOPT_EXT4_ONLY | MOPT_CLEAR},
1384 {Opt_nojournal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1385 MOPT_EXT4_ONLY | MOPT_CLEAR},
1386 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM,
1387 MOPT_EXT4_ONLY | MOPT_SET},
1388 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1389 EXT4_MOUNT_JOURNAL_CHECKSUM),
1390 MOPT_EXT4_ONLY | MOPT_SET},
1391 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_NO_EXT2 | MOPT_SET},
1392 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1393 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1394 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1395 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT,
1396 MOPT_NO_EXT2 | MOPT_SET},
1397 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT,
1398 MOPT_NO_EXT2 | MOPT_CLEAR},
1399 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1400 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1401 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1402 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1403 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1404 {Opt_commit, 0, MOPT_GTE0},
1405 {Opt_max_batch_time, 0, MOPT_GTE0},
1406 {Opt_min_batch_time, 0, MOPT_GTE0},
1407 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1408 {Opt_init_itable, 0, MOPT_GTE0},
1409 {Opt_dax, EXT4_MOUNT_DAX, MOPT_SET},
1410 {Opt_stripe, 0, MOPT_GTE0},
1411 {Opt_resuid, 0, MOPT_GTE0},
1412 {Opt_resgid, 0, MOPT_GTE0},
1413 {Opt_journal_dev, 0, MOPT_GTE0},
1414 {Opt_journal_path, 0, MOPT_STRING},
1415 {Opt_journal_ioprio, 0, MOPT_GTE0},
1416 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1417 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_NO_EXT2 | MOPT_DATAJ},
1418 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA,
1419 MOPT_NO_EXT2 | MOPT_DATAJ},
1420 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1421 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1422 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1423 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1424 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1425 #else
1426 {Opt_acl, 0, MOPT_NOSUPPORT},
1427 {Opt_noacl, 0, MOPT_NOSUPPORT},
1428 #endif
1429 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1430 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1431 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1432 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1433 MOPT_SET | MOPT_Q},
1434 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1435 MOPT_SET | MOPT_Q},
1436 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1437 EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1438 {Opt_usrjquota, 0, MOPT_Q},
1439 {Opt_grpjquota, 0, MOPT_Q},
1440 {Opt_offusrjquota, 0, MOPT_Q},
1441 {Opt_offgrpjquota, 0, MOPT_Q},
1442 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1443 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1444 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1445 {Opt_max_dir_size_kb, 0, MOPT_GTE0},
1446 {Opt_test_dummy_encryption, 0, MOPT_GTE0},
1447 {Opt_err, 0, 0}
1448 };
1449
1450 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1451 substring_t *args, unsigned long *journal_devnum,
1452 unsigned int *journal_ioprio, int is_remount)
1453 {
1454 struct ext4_sb_info *sbi = EXT4_SB(sb);
1455 const struct mount_opts *m;
1456 kuid_t uid;
1457 kgid_t gid;
1458 int arg = 0;
1459
1460 #ifdef CONFIG_QUOTA
1461 if (token == Opt_usrjquota)
1462 return set_qf_name(sb, USRQUOTA, &args[0]);
1463 else if (token == Opt_grpjquota)
1464 return set_qf_name(sb, GRPQUOTA, &args[0]);
1465 else if (token == Opt_offusrjquota)
1466 return clear_qf_name(sb, USRQUOTA);
1467 else if (token == Opt_offgrpjquota)
1468 return clear_qf_name(sb, GRPQUOTA);
1469 #endif
1470 switch (token) {
1471 case Opt_noacl:
1472 case Opt_nouser_xattr:
1473 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1474 break;
1475 case Opt_sb:
1476 return 1; /* handled by get_sb_block() */
1477 case Opt_removed:
1478 ext4_msg(sb, KERN_WARNING, "Ignoring removed %s option", opt);
1479 return 1;
1480 case Opt_abort:
1481 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1482 return 1;
1483 case Opt_i_version:
1484 sb->s_flags |= MS_I_VERSION;
1485 return 1;
1486 case Opt_lazytime:
1487 sb->s_flags |= MS_LAZYTIME;
1488 return 1;
1489 case Opt_nolazytime:
1490 sb->s_flags &= ~MS_LAZYTIME;
1491 return 1;
1492 }
1493
1494 for (m = ext4_mount_opts; m->token != Opt_err; m++)
1495 if (token == m->token)
1496 break;
1497
1498 if (m->token == Opt_err) {
1499 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1500 "or missing value", opt);
1501 return -1;
1502 }
1503
1504 if ((m->flags & MOPT_NO_EXT2) && IS_EXT2_SB(sb)) {
1505 ext4_msg(sb, KERN_ERR,
1506 "Mount option \"%s\" incompatible with ext2", opt);
1507 return -1;
1508 }
1509 if ((m->flags & MOPT_NO_EXT3) && IS_EXT3_SB(sb)) {
1510 ext4_msg(sb, KERN_ERR,
1511 "Mount option \"%s\" incompatible with ext3", opt);
1512 return -1;
1513 }
1514
1515 if (args->from && !(m->flags & MOPT_STRING) && match_int(args, &arg))
1516 return -1;
1517 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1518 return -1;
1519 if (m->flags & MOPT_EXPLICIT)
1520 set_opt2(sb, EXPLICIT_DELALLOC);
1521 if (m->flags & MOPT_CLEAR_ERR)
1522 clear_opt(sb, ERRORS_MASK);
1523 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1524 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1525 "options when quota turned on");
1526 return -1;
1527 }
1528
1529 if (m->flags & MOPT_NOSUPPORT) {
1530 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1531 } else if (token == Opt_commit) {
1532 if (arg == 0)
1533 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1534 sbi->s_commit_interval = HZ * arg;
1535 } else if (token == Opt_max_batch_time) {
1536 sbi->s_max_batch_time = arg;
1537 } else if (token == Opt_min_batch_time) {
1538 sbi->s_min_batch_time = arg;
1539 } else if (token == Opt_inode_readahead_blks) {
1540 if (arg && (arg > (1 << 30) || !is_power_of_2(arg))) {
1541 ext4_msg(sb, KERN_ERR,
1542 "EXT4-fs: inode_readahead_blks must be "
1543 "0 or a power of 2 smaller than 2^31");
1544 return -1;
1545 }
1546 sbi->s_inode_readahead_blks = arg;
1547 } else if (token == Opt_init_itable) {
1548 set_opt(sb, INIT_INODE_TABLE);
1549 if (!args->from)
1550 arg = EXT4_DEF_LI_WAIT_MULT;
1551 sbi->s_li_wait_mult = arg;
1552 } else if (token == Opt_max_dir_size_kb) {
1553 sbi->s_max_dir_size_kb = arg;
1554 } else if (token == Opt_stripe) {
1555 sbi->s_stripe = arg;
1556 } else if (token == Opt_resuid) {
1557 uid = make_kuid(current_user_ns(), arg);
1558 if (!uid_valid(uid)) {
1559 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1560 return -1;
1561 }
1562 sbi->s_resuid = uid;
1563 } else if (token == Opt_resgid) {
1564 gid = make_kgid(current_user_ns(), arg);
1565 if (!gid_valid(gid)) {
1566 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1567 return -1;
1568 }
1569 sbi->s_resgid = gid;
1570 } else if (token == Opt_journal_dev) {
1571 if (is_remount) {
1572 ext4_msg(sb, KERN_ERR,
1573 "Cannot specify journal on remount");
1574 return -1;
1575 }
1576 *journal_devnum = arg;
1577 } else if (token == Opt_journal_path) {
1578 char *journal_path;
1579 struct inode *journal_inode;
1580 struct path path;
1581 int error;
1582
1583 if (is_remount) {
1584 ext4_msg(sb, KERN_ERR,
1585 "Cannot specify journal on remount");
1586 return -1;
1587 }
1588 journal_path = match_strdup(&args[0]);
1589 if (!journal_path) {
1590 ext4_msg(sb, KERN_ERR, "error: could not dup "
1591 "journal device string");
1592 return -1;
1593 }
1594
1595 error = kern_path(journal_path, LOOKUP_FOLLOW, &path);
1596 if (error) {
1597 ext4_msg(sb, KERN_ERR, "error: could not find "
1598 "journal device path: error %d", error);
1599 kfree(journal_path);
1600 return -1;
1601 }
1602
1603 journal_inode = d_inode(path.dentry);
1604 if (!S_ISBLK(journal_inode->i_mode)) {
1605 ext4_msg(sb, KERN_ERR, "error: journal path %s "
1606 "is not a block device", journal_path);
1607 path_put(&path);
1608 kfree(journal_path);
1609 return -1;
1610 }
1611
1612 *journal_devnum = new_encode_dev(journal_inode->i_rdev);
1613 path_put(&path);
1614 kfree(journal_path);
1615 } else if (token == Opt_journal_ioprio) {
1616 if (arg > 7) {
1617 ext4_msg(sb, KERN_ERR, "Invalid journal IO priority"
1618 " (must be 0-7)");
1619 return -1;
1620 }
1621 *journal_ioprio =
1622 IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1623 } else if (token == Opt_test_dummy_encryption) {
1624 #ifdef CONFIG_EXT4_FS_ENCRYPTION
1625 sbi->s_mount_flags |= EXT4_MF_TEST_DUMMY_ENCRYPTION;
1626 ext4_msg(sb, KERN_WARNING,
1627 "Test dummy encryption mode enabled");
1628 #else
1629 ext4_msg(sb, KERN_WARNING,
1630 "Test dummy encryption mount option ignored");
1631 #endif
1632 } else if (m->flags & MOPT_DATAJ) {
1633 if (is_remount) {
1634 if (!sbi->s_journal)
1635 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1636 else if (test_opt(sb, DATA_FLAGS) != m->mount_opt) {
1637 ext4_msg(sb, KERN_ERR,
1638 "Cannot change data mode on remount");
1639 return -1;
1640 }
1641 } else {
1642 clear_opt(sb, DATA_FLAGS);
1643 sbi->s_mount_opt |= m->mount_opt;
1644 }
1645 #ifdef CONFIG_QUOTA
1646 } else if (m->flags & MOPT_QFMT) {
1647 if (sb_any_quota_loaded(sb) &&
1648 sbi->s_jquota_fmt != m->mount_opt) {
1649 ext4_msg(sb, KERN_ERR, "Cannot change journaled "
1650 "quota options when quota turned on");
1651 return -1;
1652 }
1653 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
1654 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
1655 ext4_msg(sb, KERN_ERR,
1656 "Cannot set journaled quota options "
1657 "when QUOTA feature is enabled");
1658 return -1;
1659 }
1660 sbi->s_jquota_fmt = m->mount_opt;
1661 #endif
1662 #ifndef CONFIG_FS_DAX
1663 } else if (token == Opt_dax) {
1664 ext4_msg(sb, KERN_INFO, "dax option not supported");
1665 return -1;
1666 #endif
1667 } else {
1668 if (!args->from)
1669 arg = 1;
1670 if (m->flags & MOPT_CLEAR)
1671 arg = !arg;
1672 else if (unlikely(!(m->flags & MOPT_SET))) {
1673 ext4_msg(sb, KERN_WARNING,
1674 "buggy handling of option %s", opt);
1675 WARN_ON(1);
1676 return -1;
1677 }
1678 if (arg != 0)
1679 sbi->s_mount_opt |= m->mount_opt;
1680 else
1681 sbi->s_mount_opt &= ~m->mount_opt;
1682 }
1683 return 1;
1684 }
1685
1686 static int parse_options(char *options, struct super_block *sb,
1687 unsigned long *journal_devnum,
1688 unsigned int *journal_ioprio,
1689 int is_remount)
1690 {
1691 struct ext4_sb_info *sbi = EXT4_SB(sb);
1692 char *p;
1693 substring_t args[MAX_OPT_ARGS];
1694 int token;
1695
1696 if (!options)
1697 return 1;
1698
1699 while ((p = strsep(&options, ",")) != NULL) {
1700 if (!*p)
1701 continue;
1702 /*
1703 * Initialize args struct so we know whether arg was
1704 * found; some options take optional arguments.
1705 */
1706 args[0].to = args[0].from = NULL;
1707 token = match_token(p, tokens, args);
1708 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1709 journal_ioprio, is_remount) < 0)
1710 return 0;
1711 }
1712 #ifdef CONFIG_QUOTA
1713 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
1714 (test_opt(sb, USRQUOTA) || test_opt(sb, GRPQUOTA))) {
1715 ext4_msg(sb, KERN_ERR, "Cannot set quota options when QUOTA "
1716 "feature is enabled");
1717 return 0;
1718 }
1719 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1720 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1721 clear_opt(sb, USRQUOTA);
1722
1723 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1724 clear_opt(sb, GRPQUOTA);
1725
1726 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1727 ext4_msg(sb, KERN_ERR, "old and new quota "
1728 "format mixing");
1729 return 0;
1730 }
1731
1732 if (!sbi->s_jquota_fmt) {
1733 ext4_msg(sb, KERN_ERR, "journaled quota format "
1734 "not specified");
1735 return 0;
1736 }
1737 }
1738 #endif
1739 if (test_opt(sb, DIOREAD_NOLOCK)) {
1740 int blocksize =
1741 BLOCK_SIZE << le32_to_cpu(sbi->s_es->s_log_block_size);
1742
1743 if (blocksize < PAGE_CACHE_SIZE) {
1744 ext4_msg(sb, KERN_ERR, "can't mount with "
1745 "dioread_nolock if block size != PAGE_SIZE");
1746 return 0;
1747 }
1748 }
1749 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA &&
1750 test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
1751 ext4_msg(sb, KERN_ERR, "can't mount with journal_async_commit "
1752 "in data=ordered mode");
1753 return 0;
1754 }
1755 return 1;
1756 }
1757
1758 static inline void ext4_show_quota_options(struct seq_file *seq,
1759 struct super_block *sb)
1760 {
1761 #if defined(CONFIG_QUOTA)
1762 struct ext4_sb_info *sbi = EXT4_SB(sb);
1763
1764 if (sbi->s_jquota_fmt) {
1765 char *fmtname = "";
1766
1767 switch (sbi->s_jquota_fmt) {
1768 case QFMT_VFS_OLD:
1769 fmtname = "vfsold";
1770 break;
1771 case QFMT_VFS_V0:
1772 fmtname = "vfsv0";
1773 break;
1774 case QFMT_VFS_V1:
1775 fmtname = "vfsv1";
1776 break;
1777 }
1778 seq_printf(seq, ",jqfmt=%s", fmtname);
1779 }
1780
1781 if (sbi->s_qf_names[USRQUOTA])
1782 seq_show_option(seq, "usrjquota", sbi->s_qf_names[USRQUOTA]);
1783
1784 if (sbi->s_qf_names[GRPQUOTA])
1785 seq_show_option(seq, "grpjquota", sbi->s_qf_names[GRPQUOTA]);
1786 #endif
1787 }
1788
1789 static const char *token2str(int token)
1790 {
1791 const struct match_token *t;
1792
1793 for (t = tokens; t->token != Opt_err; t++)
1794 if (t->token == token && !strchr(t->pattern, '='))
1795 break;
1796 return t->pattern;
1797 }
1798
1799 /*
1800 * Show an option if
1801 * - it's set to a non-default value OR
1802 * - if the per-sb default is different from the global default
1803 */
1804 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1805 int nodefs)
1806 {
1807 struct ext4_sb_info *sbi = EXT4_SB(sb);
1808 struct ext4_super_block *es = sbi->s_es;
1809 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
1810 const struct mount_opts *m;
1811 char sep = nodefs ? '\n' : ',';
1812
1813 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1814 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
1815
1816 if (sbi->s_sb_block != 1)
1817 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1818
1819 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1820 int want_set = m->flags & MOPT_SET;
1821 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1822 (m->flags & MOPT_CLEAR_ERR))
1823 continue;
1824 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
1825 continue; /* skip if same as the default */
1826 if ((want_set &&
1827 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1828 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1829 continue; /* select Opt_noFoo vs Opt_Foo */
1830 SEQ_OPTS_PRINT("%s", token2str(m->token));
1831 }
1832
1833 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
1834 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
1835 SEQ_OPTS_PRINT("resuid=%u",
1836 from_kuid_munged(&init_user_ns, sbi->s_resuid));
1837 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
1838 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
1839 SEQ_OPTS_PRINT("resgid=%u",
1840 from_kgid_munged(&init_user_ns, sbi->s_resgid));
1841 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
1842 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1843 SEQ_OPTS_PUTS("errors=remount-ro");
1844 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
1845 SEQ_OPTS_PUTS("errors=continue");
1846 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
1847 SEQ_OPTS_PUTS("errors=panic");
1848 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
1849 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
1850 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
1851 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
1852 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
1853 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
1854 if (sb->s_flags & MS_I_VERSION)
1855 SEQ_OPTS_PUTS("i_version");
1856 if (nodefs || sbi->s_stripe)
1857 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
1858 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
1859 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1860 SEQ_OPTS_PUTS("data=journal");
1861 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1862 SEQ_OPTS_PUTS("data=ordered");
1863 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1864 SEQ_OPTS_PUTS("data=writeback");
1865 }
1866 if (nodefs ||
1867 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1868 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1869 sbi->s_inode_readahead_blks);
1870
1871 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1872 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
1873 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
1874 if (nodefs || sbi->s_max_dir_size_kb)
1875 SEQ_OPTS_PRINT("max_dir_size_kb=%u", sbi->s_max_dir_size_kb);
1876
1877 ext4_show_quota_options(seq, sb);
1878 return 0;
1879 }
1880
1881 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1882 {
1883 return _ext4_show_options(seq, root->d_sb, 0);
1884 }
1885
1886 static int options_seq_show(struct seq_file *seq, void *offset)
1887 {
1888 struct super_block *sb = seq->private;
1889 int rc;
1890
1891 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1892 rc = _ext4_show_options(seq, sb, 1);
1893 seq_puts(seq, "\n");
1894 return rc;
1895 }
1896
1897 static int options_open_fs(struct inode *inode, struct file *file)
1898 {
1899 return single_open(file, options_seq_show, PDE_DATA(inode));
1900 }
1901
1902 static const struct file_operations ext4_seq_options_fops = {
1903 .owner = THIS_MODULE,
1904 .open = options_open_fs,
1905 .read = seq_read,
1906 .llseek = seq_lseek,
1907 .release = single_release,
1908 };
1909
1910 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1911 int read_only)
1912 {
1913 struct ext4_sb_info *sbi = EXT4_SB(sb);
1914 int res = 0;
1915
1916 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1917 ext4_msg(sb, KERN_ERR, "revision level too high, "
1918 "forcing read-only mode");
1919 res = MS_RDONLY;
1920 }
1921 if (read_only)
1922 goto done;
1923 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1924 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1925 "running e2fsck is recommended");
1926 else if (sbi->s_mount_state & EXT4_ERROR_FS)
1927 ext4_msg(sb, KERN_WARNING,
1928 "warning: mounting fs with errors, "
1929 "running e2fsck is recommended");
1930 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
1931 le16_to_cpu(es->s_mnt_count) >=
1932 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1933 ext4_msg(sb, KERN_WARNING,
1934 "warning: maximal mount count reached, "
1935 "running e2fsck is recommended");
1936 else if (le32_to_cpu(es->s_checkinterval) &&
1937 (le32_to_cpu(es->s_lastcheck) +
1938 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1939 ext4_msg(sb, KERN_WARNING,
1940 "warning: checktime reached, "
1941 "running e2fsck is recommended");
1942 if (!sbi->s_journal)
1943 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1944 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1945 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1946 le16_add_cpu(&es->s_mnt_count, 1);
1947 es->s_mtime = cpu_to_le32(get_seconds());
1948 ext4_update_dynamic_rev(sb);
1949 if (sbi->s_journal)
1950 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1951
1952 ext4_commit_super(sb, 1);
1953 done:
1954 if (test_opt(sb, DEBUG))
1955 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1956 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
1957 sb->s_blocksize,
1958 sbi->s_groups_count,
1959 EXT4_BLOCKS_PER_GROUP(sb),
1960 EXT4_INODES_PER_GROUP(sb),
1961 sbi->s_mount_opt, sbi->s_mount_opt2);
1962
1963 cleancache_init_fs(sb);
1964 return res;
1965 }
1966
1967 int ext4_alloc_flex_bg_array(struct super_block *sb, ext4_group_t ngroup)
1968 {
1969 struct ext4_sb_info *sbi = EXT4_SB(sb);
1970 struct flex_groups *new_groups;
1971 int size;
1972
1973 if (!sbi->s_log_groups_per_flex)
1974 return 0;
1975
1976 size = ext4_flex_group(sbi, ngroup - 1) + 1;
1977 if (size <= sbi->s_flex_groups_allocated)
1978 return 0;
1979
1980 size = roundup_pow_of_two(size * sizeof(struct flex_groups));
1981 new_groups = ext4_kvzalloc(size, GFP_KERNEL);
1982 if (!new_groups) {
1983 ext4_msg(sb, KERN_ERR, "not enough memory for %d flex groups",
1984 size / (int) sizeof(struct flex_groups));
1985 return -ENOMEM;
1986 }
1987
1988 if (sbi->s_flex_groups) {
1989 memcpy(new_groups, sbi->s_flex_groups,
1990 (sbi->s_flex_groups_allocated *
1991 sizeof(struct flex_groups)));
1992 kvfree(sbi->s_flex_groups);
1993 }
1994 sbi->s_flex_groups = new_groups;
1995 sbi->s_flex_groups_allocated = size / sizeof(struct flex_groups);
1996 return 0;
1997 }
1998
1999 static int ext4_fill_flex_info(struct super_block *sb)
2000 {
2001 struct ext4_sb_info *sbi = EXT4_SB(sb);
2002 struct ext4_group_desc *gdp = NULL;
2003 ext4_group_t flex_group;
2004 int i, err;
2005
2006 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
2007 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
2008 sbi->s_log_groups_per_flex = 0;
2009 return 1;
2010 }
2011
2012 err = ext4_alloc_flex_bg_array(sb, sbi->s_groups_count);
2013 if (err)
2014 goto failed;
2015
2016 for (i = 0; i < sbi->s_groups_count; i++) {
2017 gdp = ext4_get_group_desc(sb, i, NULL);
2018
2019 flex_group = ext4_flex_group(sbi, i);
2020 atomic_add(ext4_free_inodes_count(sb, gdp),
2021 &sbi->s_flex_groups[flex_group].free_inodes);
2022 atomic64_add(ext4_free_group_clusters(sb, gdp),
2023 &sbi->s_flex_groups[flex_group].free_clusters);
2024 atomic_add(ext4_used_dirs_count(sb, gdp),
2025 &sbi->s_flex_groups[flex_group].used_dirs);
2026 }
2027
2028 return 1;
2029 failed:
2030 return 0;
2031 }
2032
2033 static __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
2034 struct ext4_group_desc *gdp)
2035 {
2036 int offset;
2037 __u16 crc = 0;
2038 __le32 le_group = cpu_to_le32(block_group);
2039
2040 if (ext4_has_metadata_csum(sbi->s_sb)) {
2041 /* Use new metadata_csum algorithm */
2042 __le16 save_csum;
2043 __u32 csum32;
2044
2045 save_csum = gdp->bg_checksum;
2046 gdp->bg_checksum = 0;
2047 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
2048 sizeof(le_group));
2049 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp,
2050 sbi->s_desc_size);
2051 gdp->bg_checksum = save_csum;
2052
2053 crc = csum32 & 0xFFFF;
2054 goto out;
2055 }
2056
2057 /* old crc16 code */
2058 if (!(sbi->s_es->s_feature_ro_compat &
2059 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_GDT_CSUM)))
2060 return 0;
2061
2062 offset = offsetof(struct ext4_group_desc, bg_checksum);
2063
2064 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
2065 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
2066 crc = crc16(crc, (__u8 *)gdp, offset);
2067 offset += sizeof(gdp->bg_checksum); /* skip checksum */
2068 /* for checksum of struct ext4_group_desc do the rest...*/
2069 if ((sbi->s_es->s_feature_incompat &
2070 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
2071 offset < le16_to_cpu(sbi->s_es->s_desc_size))
2072 crc = crc16(crc, (__u8 *)gdp + offset,
2073 le16_to_cpu(sbi->s_es->s_desc_size) -
2074 offset);
2075
2076 out:
2077 return cpu_to_le16(crc);
2078 }
2079
2080 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2081 struct ext4_group_desc *gdp)
2082 {
2083 if (ext4_has_group_desc_csum(sb) &&
2084 (gdp->bg_checksum != ext4_group_desc_csum(EXT4_SB(sb),
2085 block_group, gdp)))
2086 return 0;
2087
2088 return 1;
2089 }
2090
2091 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2092 struct ext4_group_desc *gdp)
2093 {
2094 if (!ext4_has_group_desc_csum(sb))
2095 return;
2096 gdp->bg_checksum = ext4_group_desc_csum(EXT4_SB(sb), block_group, gdp);
2097 }
2098
2099 /* Called at mount-time, super-block is locked */
2100 static int ext4_check_descriptors(struct super_block *sb,
2101 ext4_group_t *first_not_zeroed)
2102 {
2103 struct ext4_sb_info *sbi = EXT4_SB(sb);
2104 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2105 ext4_fsblk_t last_block;
2106 ext4_fsblk_t block_bitmap;
2107 ext4_fsblk_t inode_bitmap;
2108 ext4_fsblk_t inode_table;
2109 int flexbg_flag = 0;
2110 ext4_group_t i, grp = sbi->s_groups_count;
2111
2112 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2113 flexbg_flag = 1;
2114
2115 ext4_debug("Checking group descriptors");
2116
2117 for (i = 0; i < sbi->s_groups_count; i++) {
2118 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2119
2120 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2121 last_block = ext4_blocks_count(sbi->s_es) - 1;
2122 else
2123 last_block = first_block +
2124 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2125
2126 if ((grp == sbi->s_groups_count) &&
2127 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2128 grp = i;
2129
2130 block_bitmap = ext4_block_bitmap(sb, gdp);
2131 if (block_bitmap < first_block || block_bitmap > last_block) {
2132 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2133 "Block bitmap for group %u not in group "
2134 "(block %llu)!", i, block_bitmap);
2135 return 0;
2136 }
2137 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2138 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2139 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2140 "Inode bitmap for group %u not in group "
2141 "(block %llu)!", i, inode_bitmap);
2142 return 0;
2143 }
2144 inode_table = ext4_inode_table(sb, gdp);
2145 if (inode_table < first_block ||
2146 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2147 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2148 "Inode table for group %u not in group "
2149 "(block %llu)!", i, inode_table);
2150 return 0;
2151 }
2152 ext4_lock_group(sb, i);
2153 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2154 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2155 "Checksum for group %u failed (%u!=%u)",
2156 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2157 gdp)), le16_to_cpu(gdp->bg_checksum));
2158 if (!(sb->s_flags & MS_RDONLY)) {
2159 ext4_unlock_group(sb, i);
2160 return 0;
2161 }
2162 }
2163 ext4_unlock_group(sb, i);
2164 if (!flexbg_flag)
2165 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2166 }
2167 if (NULL != first_not_zeroed)
2168 *first_not_zeroed = grp;
2169 return 1;
2170 }
2171
2172 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2173 * the superblock) which were deleted from all directories, but held open by
2174 * a process at the time of a crash. We walk the list and try to delete these
2175 * inodes at recovery time (only with a read-write filesystem).
2176 *
2177 * In order to keep the orphan inode chain consistent during traversal (in
2178 * case of crash during recovery), we link each inode into the superblock
2179 * orphan list_head and handle it the same way as an inode deletion during
2180 * normal operation (which journals the operations for us).
2181 *
2182 * We only do an iget() and an iput() on each inode, which is very safe if we
2183 * accidentally point at an in-use or already deleted inode. The worst that
2184 * can happen in this case is that we get a "bit already cleared" message from
2185 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2186 * e2fsck was run on this filesystem, and it must have already done the orphan
2187 * inode cleanup for us, so we can safely abort without any further action.
2188 */
2189 static void ext4_orphan_cleanup(struct super_block *sb,
2190 struct ext4_super_block *es)
2191 {
2192 unsigned int s_flags = sb->s_flags;
2193 int nr_orphans = 0, nr_truncates = 0;
2194 #ifdef CONFIG_QUOTA
2195 int i;
2196 #endif
2197 if (!es->s_last_orphan) {
2198 jbd_debug(4, "no orphan inodes to clean up\n");
2199 return;
2200 }
2201
2202 if (bdev_read_only(sb->s_bdev)) {
2203 ext4_msg(sb, KERN_ERR, "write access "
2204 "unavailable, skipping orphan cleanup");
2205 return;
2206 }
2207
2208 /* Check if feature set would not allow a r/w mount */
2209 if (!ext4_feature_set_ok(sb, 0)) {
2210 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2211 "unknown ROCOMPAT features");
2212 return;
2213 }
2214
2215 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2216 /* don't clear list on RO mount w/ errors */
2217 if (es->s_last_orphan && !(s_flags & MS_RDONLY)) {
2218 ext4_msg(sb, KERN_INFO, "Errors on filesystem, "
2219 "clearing orphan list.\n");
2220 es->s_last_orphan = 0;
2221 }
2222 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2223 return;
2224 }
2225
2226 if (s_flags & MS_RDONLY) {
2227 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2228 sb->s_flags &= ~MS_RDONLY;
2229 }
2230 #ifdef CONFIG_QUOTA
2231 /* Needed for iput() to work correctly and not trash data */
2232 sb->s_flags |= MS_ACTIVE;
2233 /* Turn on quotas so that they are updated correctly */
2234 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2235 if (EXT4_SB(sb)->s_qf_names[i]) {
2236 int ret = ext4_quota_on_mount(sb, i);
2237 if (ret < 0)
2238 ext4_msg(sb, KERN_ERR,
2239 "Cannot turn on journaled "
2240 "quota: error %d", ret);
2241 }
2242 }
2243 #endif
2244
2245 while (es->s_last_orphan) {
2246 struct inode *inode;
2247
2248 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2249 if (IS_ERR(inode)) {
2250 es->s_last_orphan = 0;
2251 break;
2252 }
2253
2254 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2255 dquot_initialize(inode);
2256 if (inode->i_nlink) {
2257 if (test_opt(sb, DEBUG))
2258 ext4_msg(sb, KERN_DEBUG,
2259 "%s: truncating inode %lu to %lld bytes",
2260 __func__, inode->i_ino, inode->i_size);
2261 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2262 inode->i_ino, inode->i_size);
2263 mutex_lock(&inode->i_mutex);
2264 truncate_inode_pages(inode->i_mapping, inode->i_size);
2265 ext4_truncate(inode);
2266 mutex_unlock(&inode->i_mutex);
2267 nr_truncates++;
2268 } else {
2269 if (test_opt(sb, DEBUG))
2270 ext4_msg(sb, KERN_DEBUG,
2271 "%s: deleting unreferenced inode %lu",
2272 __func__, inode->i_ino);
2273 jbd_debug(2, "deleting unreferenced inode %lu\n",
2274 inode->i_ino);
2275 nr_orphans++;
2276 }
2277 iput(inode); /* The delete magic happens here! */
2278 }
2279
2280 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2281
2282 if (nr_orphans)
2283 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2284 PLURAL(nr_orphans));
2285 if (nr_truncates)
2286 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2287 PLURAL(nr_truncates));
2288 #ifdef CONFIG_QUOTA
2289 /* Turn quotas off */
2290 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
2291 if (sb_dqopt(sb)->files[i])
2292 dquot_quota_off(sb, i);
2293 }
2294 #endif
2295 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2296 }
2297
2298 /*
2299 * Maximal extent format file size.
2300 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2301 * extent format containers, within a sector_t, and within i_blocks
2302 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2303 * so that won't be a limiting factor.
2304 *
2305 * However there is other limiting factor. We do store extents in the form
2306 * of starting block and length, hence the resulting length of the extent
2307 * covering maximum file size must fit into on-disk format containers as
2308 * well. Given that length is always by 1 unit bigger than max unit (because
2309 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2310 *
2311 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2312 */
2313 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2314 {
2315 loff_t res;
2316 loff_t upper_limit = MAX_LFS_FILESIZE;
2317
2318 /* small i_blocks in vfs inode? */
2319 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2320 /*
2321 * CONFIG_LBDAF is not enabled implies the inode
2322 * i_block represent total blocks in 512 bytes
2323 * 32 == size of vfs inode i_blocks * 8
2324 */
2325 upper_limit = (1LL << 32) - 1;
2326
2327 /* total blocks in file system block size */
2328 upper_limit >>= (blkbits - 9);
2329 upper_limit <<= blkbits;
2330 }
2331
2332 /*
2333 * 32-bit extent-start container, ee_block. We lower the maxbytes
2334 * by one fs block, so ee_len can cover the extent of maximum file
2335 * size
2336 */
2337 res = (1LL << 32) - 1;
2338 res <<= blkbits;
2339
2340 /* Sanity check against vm- & vfs- imposed limits */
2341 if (res > upper_limit)
2342 res = upper_limit;
2343
2344 return res;
2345 }
2346
2347 /*
2348 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2349 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2350 * We need to be 1 filesystem block less than the 2^48 sector limit.
2351 */
2352 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2353 {
2354 loff_t res = EXT4_NDIR_BLOCKS;
2355 int meta_blocks;
2356 loff_t upper_limit;
2357 /* This is calculated to be the largest file size for a dense, block
2358 * mapped file such that the file's total number of 512-byte sectors,
2359 * including data and all indirect blocks, does not exceed (2^48 - 1).
2360 *
2361 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2362 * number of 512-byte sectors of the file.
2363 */
2364
2365 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2366 /*
2367 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2368 * the inode i_block field represents total file blocks in
2369 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2370 */
2371 upper_limit = (1LL << 32) - 1;
2372
2373 /* total blocks in file system block size */
2374 upper_limit >>= (bits - 9);
2375
2376 } else {
2377 /*
2378 * We use 48 bit ext4_inode i_blocks
2379 * With EXT4_HUGE_FILE_FL set the i_blocks
2380 * represent total number of blocks in
2381 * file system block size
2382 */
2383 upper_limit = (1LL << 48) - 1;
2384
2385 }
2386
2387 /* indirect blocks */
2388 meta_blocks = 1;
2389 /* double indirect blocks */
2390 meta_blocks += 1 + (1LL << (bits-2));
2391 /* tripple indirect blocks */
2392 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2393
2394 upper_limit -= meta_blocks;
2395 upper_limit <<= bits;
2396
2397 res += 1LL << (bits-2);
2398 res += 1LL << (2*(bits-2));
2399 res += 1LL << (3*(bits-2));
2400 res <<= bits;
2401 if (res > upper_limit)
2402 res = upper_limit;
2403
2404 if (res > MAX_LFS_FILESIZE)
2405 res = MAX_LFS_FILESIZE;
2406
2407 return res;
2408 }
2409
2410 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2411 ext4_fsblk_t logical_sb_block, int nr)
2412 {
2413 struct ext4_sb_info *sbi = EXT4_SB(sb);
2414 ext4_group_t bg, first_meta_bg;
2415 int has_super = 0;
2416
2417 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2418
2419 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2420 nr < first_meta_bg)
2421 return logical_sb_block + nr + 1;
2422 bg = sbi->s_desc_per_block * nr;
2423 if (ext4_bg_has_super(sb, bg))
2424 has_super = 1;
2425
2426 /*
2427 * If we have a meta_bg fs with 1k blocks, group 0's GDT is at
2428 * block 2, not 1. If s_first_data_block == 0 (bigalloc is enabled
2429 * on modern mke2fs or blksize > 1k on older mke2fs) then we must
2430 * compensate.
2431 */
2432 if (sb->s_blocksize == 1024 && nr == 0 &&
2433 le32_to_cpu(EXT4_SB(sb)->s_es->s_first_data_block) == 0)
2434 has_super++;
2435
2436 return (has_super + ext4_group_first_block_no(sb, bg));
2437 }
2438
2439 /**
2440 * ext4_get_stripe_size: Get the stripe size.
2441 * @sbi: In memory super block info
2442 *
2443 * If we have specified it via mount option, then
2444 * use the mount option value. If the value specified at mount time is
2445 * greater than the blocks per group use the super block value.
2446 * If the super block value is greater than blocks per group return 0.
2447 * Allocator needs it be less than blocks per group.
2448 *
2449 */
2450 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2451 {
2452 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2453 unsigned long stripe_width =
2454 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2455 int ret;
2456
2457 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2458 ret = sbi->s_stripe;
2459 else if (stripe_width <= sbi->s_blocks_per_group)
2460 ret = stripe_width;
2461 else if (stride <= sbi->s_blocks_per_group)
2462 ret = stride;
2463 else
2464 ret = 0;
2465
2466 /*
2467 * If the stripe width is 1, this makes no sense and
2468 * we set it to 0 to turn off stripe handling code.
2469 */
2470 if (ret <= 1)
2471 ret = 0;
2472
2473 return ret;
2474 }
2475
2476 /* sysfs supprt */
2477
2478 struct ext4_attr {
2479 struct attribute attr;
2480 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2481 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2482 const char *, size_t);
2483 union {
2484 int offset;
2485 int deprecated_val;
2486 } u;
2487 };
2488
2489 static int parse_strtoull(const char *buf,
2490 unsigned long long max, unsigned long long *value)
2491 {
2492 int ret;
2493
2494 ret = kstrtoull(skip_spaces(buf), 0, value);
2495 if (!ret && *value > max)
2496 ret = -EINVAL;
2497 return ret;
2498 }
2499
2500 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2501 struct ext4_sb_info *sbi,
2502 char *buf)
2503 {
2504 return snprintf(buf, PAGE_SIZE, "%llu\n",
2505 (s64) EXT4_C2B(sbi,
2506 percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
2507 }
2508
2509 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2510 struct ext4_sb_info *sbi, char *buf)
2511 {
2512 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2513
2514 if (!sb->s_bdev->bd_part)
2515 return snprintf(buf, PAGE_SIZE, "0\n");
2516 return snprintf(buf, PAGE_SIZE, "%lu\n",
2517 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2518 sbi->s_sectors_written_start) >> 1);
2519 }
2520
2521 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2522 struct ext4_sb_info *sbi, char *buf)
2523 {
2524 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2525
2526 if (!sb->s_bdev->bd_part)
2527 return snprintf(buf, PAGE_SIZE, "0\n");
2528 return snprintf(buf, PAGE_SIZE, "%llu\n",
2529 (unsigned long long)(sbi->s_kbytes_written +
2530 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2531 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
2532 }
2533
2534 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2535 struct ext4_sb_info *sbi,
2536 const char *buf, size_t count)
2537 {
2538 unsigned long t;
2539 int ret;
2540
2541 ret = kstrtoul(skip_spaces(buf), 0, &t);
2542 if (ret)
2543 return ret;
2544
2545 if (t && (!is_power_of_2(t) || t > 0x40000000))
2546 return -EINVAL;
2547
2548 sbi->s_inode_readahead_blks = t;
2549 return count;
2550 }
2551
2552 static ssize_t sbi_ui_show(struct ext4_attr *a,
2553 struct ext4_sb_info *sbi, char *buf)
2554 {
2555 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->u.offset);
2556
2557 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2558 }
2559
2560 static ssize_t sbi_ui_store(struct ext4_attr *a,
2561 struct ext4_sb_info *sbi,
2562 const char *buf, size_t count)
2563 {
2564 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->u.offset);
2565 unsigned long t;
2566 int ret;
2567
2568 ret = kstrtoul(skip_spaces(buf), 0, &t);
2569 if (ret)
2570 return ret;
2571 *ui = t;
2572 return count;
2573 }
2574
2575 static ssize_t es_ui_show(struct ext4_attr *a,
2576 struct ext4_sb_info *sbi, char *buf)
2577 {
2578
2579 unsigned int *ui = (unsigned int *) (((char *) sbi->s_es) +
2580 a->u.offset);
2581
2582 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2583 }
2584
2585 static ssize_t reserved_clusters_show(struct ext4_attr *a,
2586 struct ext4_sb_info *sbi, char *buf)
2587 {
2588 return snprintf(buf, PAGE_SIZE, "%llu\n",
2589 (unsigned long long) atomic64_read(&sbi->s_resv_clusters));
2590 }
2591
2592 static ssize_t reserved_clusters_store(struct ext4_attr *a,
2593 struct ext4_sb_info *sbi,
2594 const char *buf, size_t count)
2595 {
2596 unsigned long long val;
2597 int ret;
2598
2599 if (parse_strtoull(buf, -1ULL, &val))
2600 return -EINVAL;
2601 ret = ext4_reserve_clusters(sbi, val);
2602
2603 return ret ? ret : count;
2604 }
2605
2606 static ssize_t trigger_test_error(struct ext4_attr *a,
2607 struct ext4_sb_info *sbi,
2608 const char *buf, size_t count)
2609 {
2610 int len = count;
2611
2612 if (!capable(CAP_SYS_ADMIN))
2613 return -EPERM;
2614
2615 if (len && buf[len-1] == '\n')
2616 len--;
2617
2618 if (len)
2619 ext4_error(sbi->s_sb, "%.*s", len, buf);
2620 return count;
2621 }
2622
2623 static ssize_t sbi_deprecated_show(struct ext4_attr *a,
2624 struct ext4_sb_info *sbi, char *buf)
2625 {
2626 return snprintf(buf, PAGE_SIZE, "%d\n", a->u.deprecated_val);
2627 }
2628
2629 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2630 static struct ext4_attr ext4_attr_##_name = { \
2631 .attr = {.name = __stringify(_name), .mode = _mode }, \
2632 .show = _show, \
2633 .store = _store, \
2634 .u = { \
2635 .offset = offsetof(struct ext4_sb_info, _elname),\
2636 }, \
2637 }
2638
2639 #define EXT4_ATTR_OFFSET_ES(_name,_mode,_show,_store,_elname) \
2640 static struct ext4_attr ext4_attr_##_name = { \
2641 .attr = {.name = __stringify(_name), .mode = _mode }, \
2642 .show = _show, \
2643 .store = _store, \
2644 .u = { \
2645 .offset = offsetof(struct ext4_super_block, _elname), \
2646 }, \
2647 }
2648
2649 #define EXT4_ATTR(name, mode, show, store) \
2650 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2651
2652 #define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
2653 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2654 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2655
2656 #define EXT4_RO_ATTR_ES_UI(name, elname) \
2657 EXT4_ATTR_OFFSET_ES(name, 0444, es_ui_show, NULL, elname)
2658 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2659 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2660
2661 #define ATTR_LIST(name) &ext4_attr_##name.attr
2662 #define EXT4_DEPRECATED_ATTR(_name, _val) \
2663 static struct ext4_attr ext4_attr_##_name = { \
2664 .attr = {.name = __stringify(_name), .mode = 0444 }, \
2665 .show = sbi_deprecated_show, \
2666 .u = { \
2667 .deprecated_val = _val, \
2668 }, \
2669 }
2670
2671 EXT4_RO_ATTR(delayed_allocation_blocks);
2672 EXT4_RO_ATTR(session_write_kbytes);
2673 EXT4_RO_ATTR(lifetime_write_kbytes);
2674 EXT4_RW_ATTR(reserved_clusters);
2675 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2676 inode_readahead_blks_store, s_inode_readahead_blks);
2677 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2678 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2679 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2680 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2681 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2682 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2683 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2684 EXT4_DEPRECATED_ATTR(max_writeback_mb_bump, 128);
2685 EXT4_RW_ATTR_SBI_UI(extent_max_zeroout_kb, s_extent_max_zeroout_kb);
2686 EXT4_ATTR(trigger_fs_error, 0200, NULL, trigger_test_error);
2687 EXT4_RW_ATTR_SBI_UI(err_ratelimit_interval_ms, s_err_ratelimit_state.interval);
2688 EXT4_RW_ATTR_SBI_UI(err_ratelimit_burst, s_err_ratelimit_state.burst);
2689 EXT4_RW_ATTR_SBI_UI(warning_ratelimit_interval_ms, s_warning_ratelimit_state.interval);
2690 EXT4_RW_ATTR_SBI_UI(warning_ratelimit_burst, s_warning_ratelimit_state.burst);
2691 EXT4_RW_ATTR_SBI_UI(msg_ratelimit_interval_ms, s_msg_ratelimit_state.interval);
2692 EXT4_RW_ATTR_SBI_UI(msg_ratelimit_burst, s_msg_ratelimit_state.burst);
2693 EXT4_RO_ATTR_ES_UI(errors_count, s_error_count);
2694 EXT4_RO_ATTR_ES_UI(first_error_time, s_first_error_time);
2695 EXT4_RO_ATTR_ES_UI(last_error_time, s_last_error_time);
2696
2697 static struct attribute *ext4_attrs[] = {
2698 ATTR_LIST(delayed_allocation_blocks),
2699 ATTR_LIST(session_write_kbytes),
2700 ATTR_LIST(lifetime_write_kbytes),
2701 ATTR_LIST(reserved_clusters),
2702 ATTR_LIST(inode_readahead_blks),
2703 ATTR_LIST(inode_goal),
2704 ATTR_LIST(mb_stats),
2705 ATTR_LIST(mb_max_to_scan),
2706 ATTR_LIST(mb_min_to_scan),
2707 ATTR_LIST(mb_order2_req),
2708 ATTR_LIST(mb_stream_req),
2709 ATTR_LIST(mb_group_prealloc),
2710 ATTR_LIST(max_writeback_mb_bump),
2711 ATTR_LIST(extent_max_zeroout_kb),
2712 ATTR_LIST(trigger_fs_error),
2713 ATTR_LIST(err_ratelimit_interval_ms),
2714 ATTR_LIST(err_ratelimit_burst),
2715 ATTR_LIST(warning_ratelimit_interval_ms),
2716 ATTR_LIST(warning_ratelimit_burst),
2717 ATTR_LIST(msg_ratelimit_interval_ms),
2718 ATTR_LIST(msg_ratelimit_burst),
2719 ATTR_LIST(errors_count),
2720 ATTR_LIST(first_error_time),
2721 ATTR_LIST(last_error_time),
2722 NULL,
2723 };
2724
2725 /* Features this copy of ext4 supports */
2726 EXT4_INFO_ATTR(lazy_itable_init);
2727 EXT4_INFO_ATTR(batched_discard);
2728 EXT4_INFO_ATTR(meta_bg_resize);
2729 EXT4_INFO_ATTR(encryption);
2730
2731 static struct attribute *ext4_feat_attrs[] = {
2732 ATTR_LIST(lazy_itable_init),
2733 ATTR_LIST(batched_discard),
2734 ATTR_LIST(meta_bg_resize),
2735 ATTR_LIST(encryption),
2736 NULL,
2737 };
2738
2739 static ssize_t ext4_attr_show(struct kobject *kobj,
2740 struct attribute *attr, char *buf)
2741 {
2742 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2743 s_kobj);
2744 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2745
2746 return a->show ? a->show(a, sbi, buf) : 0;
2747 }
2748
2749 static ssize_t ext4_attr_store(struct kobject *kobj,
2750 struct attribute *attr,
2751 const char *buf, size_t len)
2752 {
2753 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2754 s_kobj);
2755 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2756
2757 return a->store ? a->store(a, sbi, buf, len) : 0;
2758 }
2759
2760 static void ext4_sb_release(struct kobject *kobj)
2761 {
2762 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2763 s_kobj);
2764 complete(&sbi->s_kobj_unregister);
2765 }
2766
2767 static const struct sysfs_ops ext4_attr_ops = {
2768 .show = ext4_attr_show,
2769 .store = ext4_attr_store,
2770 };
2771
2772 static struct kobj_type ext4_ktype = {
2773 .default_attrs = ext4_attrs,
2774 .sysfs_ops = &ext4_attr_ops,
2775 .release = ext4_sb_release,
2776 };
2777
2778 static void ext4_feat_release(struct kobject *kobj)
2779 {
2780 complete(&ext4_feat->f_kobj_unregister);
2781 }
2782
2783 static ssize_t ext4_feat_show(struct kobject *kobj,
2784 struct attribute *attr, char *buf)
2785 {
2786 return snprintf(buf, PAGE_SIZE, "supported\n");
2787 }
2788
2789 /*
2790 * We can not use ext4_attr_show/store because it relies on the kobject
2791 * being embedded in the ext4_sb_info structure which is definitely not
2792 * true in this case.
2793 */
2794 static const struct sysfs_ops ext4_feat_ops = {
2795 .show = ext4_feat_show,
2796 .store = NULL,
2797 };
2798
2799 static struct kobj_type ext4_feat_ktype = {
2800 .default_attrs = ext4_feat_attrs,
2801 .sysfs_ops = &ext4_feat_ops,
2802 .release = ext4_feat_release,
2803 };
2804
2805 /*
2806 * Check whether this filesystem can be mounted based on
2807 * the features present and the RDONLY/RDWR mount requested.
2808 * Returns 1 if this filesystem can be mounted as requested,
2809 * 0 if it cannot be.
2810 */
2811 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2812 {
2813 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2814 ext4_msg(sb, KERN_ERR,
2815 "Couldn't mount because of "
2816 "unsupported optional features (%x)",
2817 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2818 ~EXT4_FEATURE_INCOMPAT_SUPP));
2819 return 0;
2820 }
2821
2822 if (readonly)
2823 return 1;
2824
2825 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_READONLY)) {
2826 ext4_msg(sb, KERN_INFO, "filesystem is read-only");
2827 sb->s_flags |= MS_RDONLY;
2828 return 1;
2829 }
2830
2831 /* Check that feature set is OK for a read-write mount */
2832 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2833 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2834 "unsupported optional features (%x)",
2835 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2836 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2837 return 0;
2838 }
2839 /*
2840 * Large file size enabled file system can only be mounted
2841 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2842 */
2843 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2844 if (sizeof(blkcnt_t) < sizeof(u64)) {
2845 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2846 "cannot be mounted RDWR without "
2847 "CONFIG_LBDAF");
2848 return 0;
2849 }
2850 }
2851 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2852 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2853 ext4_msg(sb, KERN_ERR,
2854 "Can't support bigalloc feature without "
2855 "extents feature\n");
2856 return 0;
2857 }
2858
2859 #ifndef CONFIG_QUOTA
2860 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
2861 !readonly) {
2862 ext4_msg(sb, KERN_ERR,
2863 "Filesystem with quota feature cannot be mounted RDWR "
2864 "without CONFIG_QUOTA");
2865 return 0;
2866 }
2867 #endif /* CONFIG_QUOTA */
2868 return 1;
2869 }
2870
2871 /*
2872 * This function is called once a day if we have errors logged
2873 * on the file system
2874 */
2875 static void print_daily_error_info(unsigned long arg)
2876 {
2877 struct super_block *sb = (struct super_block *) arg;
2878 struct ext4_sb_info *sbi;
2879 struct ext4_super_block *es;
2880
2881 sbi = EXT4_SB(sb);
2882 es = sbi->s_es;
2883
2884 if (es->s_error_count)
2885 /* fsck newer than v1.41.13 is needed to clean this condition. */
2886 ext4_msg(sb, KERN_NOTICE, "error count since last fsck: %u",
2887 le32_to_cpu(es->s_error_count));
2888 if (es->s_first_error_time) {
2889 printk(KERN_NOTICE "EXT4-fs (%s): initial error at time %u: %.*s:%d",
2890 sb->s_id, le32_to_cpu(es->s_first_error_time),
2891 (int) sizeof(es->s_first_error_func),
2892 es->s_first_error_func,
2893 le32_to_cpu(es->s_first_error_line));
2894 if (es->s_first_error_ino)
2895 printk(": inode %u",
2896 le32_to_cpu(es->s_first_error_ino));
2897 if (es->s_first_error_block)
2898 printk(": block %llu", (unsigned long long)
2899 le64_to_cpu(es->s_first_error_block));
2900 printk("\n");
2901 }
2902 if (es->s_last_error_time) {
2903 printk(KERN_NOTICE "EXT4-fs (%s): last error at time %u: %.*s:%d",
2904 sb->s_id, le32_to_cpu(es->s_last_error_time),
2905 (int) sizeof(es->s_last_error_func),
2906 es->s_last_error_func,
2907 le32_to_cpu(es->s_last_error_line));
2908 if (es->s_last_error_ino)
2909 printk(": inode %u",
2910 le32_to_cpu(es->s_last_error_ino));
2911 if (es->s_last_error_block)
2912 printk(": block %llu", (unsigned long long)
2913 le64_to_cpu(es->s_last_error_block));
2914 printk("\n");
2915 }
2916 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2917 }
2918
2919 /* Find next suitable group and run ext4_init_inode_table */
2920 static int ext4_run_li_request(struct ext4_li_request *elr)
2921 {
2922 struct ext4_group_desc *gdp = NULL;
2923 ext4_group_t group, ngroups;
2924 struct super_block *sb;
2925 unsigned long timeout = 0;
2926 int ret = 0;
2927
2928 sb = elr->lr_super;
2929 ngroups = EXT4_SB(sb)->s_groups_count;
2930
2931 sb_start_write(sb);
2932 for (group = elr->lr_next_group; group < ngroups; group++) {
2933 gdp = ext4_get_group_desc(sb, group, NULL);
2934 if (!gdp) {
2935 ret = 1;
2936 break;
2937 }
2938
2939 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2940 break;
2941 }
2942
2943 if (group >= ngroups)
2944 ret = 1;
2945
2946 if (!ret) {
2947 timeout = jiffies;
2948 ret = ext4_init_inode_table(sb, group,
2949 elr->lr_timeout ? 0 : 1);
2950 if (elr->lr_timeout == 0) {
2951 timeout = (jiffies - timeout) *
2952 elr->lr_sbi->s_li_wait_mult;
2953 elr->lr_timeout = timeout;
2954 }
2955 elr->lr_next_sched = jiffies + elr->lr_timeout;
2956 elr->lr_next_group = group + 1;
2957 }
2958 sb_end_write(sb);
2959
2960 return ret;
2961 }
2962
2963 /*
2964 * Remove lr_request from the list_request and free the
2965 * request structure. Should be called with li_list_mtx held
2966 */
2967 static void ext4_remove_li_request(struct ext4_li_request *elr)
2968 {
2969 struct ext4_sb_info *sbi;
2970
2971 if (!elr)
2972 return;
2973
2974 sbi = elr->lr_sbi;
2975
2976 list_del(&elr->lr_request);
2977 sbi->s_li_request = NULL;
2978 kfree(elr);
2979 }
2980
2981 static void ext4_unregister_li_request(struct super_block *sb)
2982 {
2983 mutex_lock(&ext4_li_mtx);
2984 if (!ext4_li_info) {
2985 mutex_unlock(&ext4_li_mtx);
2986 return;
2987 }
2988
2989 mutex_lock(&ext4_li_info->li_list_mtx);
2990 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2991 mutex_unlock(&ext4_li_info->li_list_mtx);
2992 mutex_unlock(&ext4_li_mtx);
2993 }
2994
2995 static struct task_struct *ext4_lazyinit_task;
2996
2997 /*
2998 * This is the function where ext4lazyinit thread lives. It walks
2999 * through the request list searching for next scheduled filesystem.
3000 * When such a fs is found, run the lazy initialization request
3001 * (ext4_rn_li_request) and keep track of the time spend in this
3002 * function. Based on that time we compute next schedule time of
3003 * the request. When walking through the list is complete, compute
3004 * next waking time and put itself into sleep.
3005 */
3006 static int ext4_lazyinit_thread(void *arg)
3007 {
3008 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
3009 struct list_head *pos, *n;
3010 struct ext4_li_request *elr;
3011 unsigned long next_wakeup, cur;
3012
3013 BUG_ON(NULL == eli);
3014
3015 cont_thread:
3016 while (true) {
3017 next_wakeup = MAX_JIFFY_OFFSET;
3018
3019 mutex_lock(&eli->li_list_mtx);
3020 if (list_empty(&eli->li_request_list)) {
3021 mutex_unlock(&eli->li_list_mtx);
3022 goto exit_thread;
3023 }
3024
3025 list_for_each_safe(pos, n, &eli->li_request_list) {
3026 elr = list_entry(pos, struct ext4_li_request,
3027 lr_request);
3028
3029 if (time_after_eq(jiffies, elr->lr_next_sched)) {
3030 if (ext4_run_li_request(elr) != 0) {
3031 /* error, remove the lazy_init job */
3032 ext4_remove_li_request(elr);
3033 continue;
3034 }
3035 }
3036
3037 if (time_before(elr->lr_next_sched, next_wakeup))
3038 next_wakeup = elr->lr_next_sched;
3039 }
3040 mutex_unlock(&eli->li_list_mtx);
3041
3042 try_to_freeze();
3043
3044 cur = jiffies;
3045 if ((time_after_eq(cur, next_wakeup)) ||
3046 (MAX_JIFFY_OFFSET == next_wakeup)) {
3047 cond_resched();
3048 continue;
3049 }
3050
3051 schedule_timeout_interruptible(next_wakeup - cur);
3052
3053 if (kthread_should_stop()) {
3054 ext4_clear_request_list();
3055 goto exit_thread;
3056 }
3057 }
3058
3059 exit_thread:
3060 /*
3061 * It looks like the request list is empty, but we need
3062 * to check it under the li_list_mtx lock, to prevent any
3063 * additions into it, and of course we should lock ext4_li_mtx
3064 * to atomically free the list and ext4_li_info, because at
3065 * this point another ext4 filesystem could be registering
3066 * new one.
3067 */
3068 mutex_lock(&ext4_li_mtx);
3069 mutex_lock(&eli->li_list_mtx);
3070 if (!list_empty(&eli->li_request_list)) {
3071 mutex_unlock(&eli->li_list_mtx);
3072 mutex_unlock(&ext4_li_mtx);
3073 goto cont_thread;
3074 }
3075 mutex_unlock(&eli->li_list_mtx);
3076 kfree(ext4_li_info);
3077 ext4_li_info = NULL;
3078 mutex_unlock(&ext4_li_mtx);
3079
3080 return 0;
3081 }
3082
3083 static void ext4_clear_request_list(void)
3084 {
3085 struct list_head *pos, *n;
3086 struct ext4_li_request *elr;
3087
3088 mutex_lock(&ext4_li_info->li_list_mtx);
3089 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
3090 elr = list_entry(pos, struct ext4_li_request,
3091 lr_request);
3092 ext4_remove_li_request(elr);
3093 }
3094 mutex_unlock(&ext4_li_info->li_list_mtx);
3095 }
3096
3097 static int ext4_run_lazyinit_thread(void)
3098 {
3099 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
3100 ext4_li_info, "ext4lazyinit");
3101 if (IS_ERR(ext4_lazyinit_task)) {
3102 int err = PTR_ERR(ext4_lazyinit_task);
3103 ext4_clear_request_list();
3104 kfree(ext4_li_info);
3105 ext4_li_info = NULL;
3106 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
3107 "initialization thread\n",
3108 err);
3109 return err;
3110 }
3111 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
3112 return 0;
3113 }
3114
3115 /*
3116 * Check whether it make sense to run itable init. thread or not.
3117 * If there is at least one uninitialized inode table, return
3118 * corresponding group number, else the loop goes through all
3119 * groups and return total number of groups.
3120 */
3121 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
3122 {
3123 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
3124 struct ext4_group_desc *gdp = NULL;
3125
3126 for (group = 0; group < ngroups; group++) {
3127 gdp = ext4_get_group_desc(sb, group, NULL);
3128 if (!gdp)
3129 continue;
3130
3131 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
3132 break;
3133 }
3134
3135 return group;
3136 }
3137
3138 static int ext4_li_info_new(void)
3139 {
3140 struct ext4_lazy_init *eli = NULL;
3141
3142 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
3143 if (!eli)
3144 return -ENOMEM;
3145
3146 INIT_LIST_HEAD(&eli->li_request_list);
3147 mutex_init(&eli->li_list_mtx);
3148
3149 eli->li_state |= EXT4_LAZYINIT_QUIT;
3150
3151 ext4_li_info = eli;
3152
3153 return 0;
3154 }
3155
3156 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
3157 ext4_group_t start)
3158 {
3159 struct ext4_sb_info *sbi = EXT4_SB(sb);
3160 struct ext4_li_request *elr;
3161
3162 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
3163 if (!elr)
3164 return NULL;
3165
3166 elr->lr_super = sb;
3167 elr->lr_sbi = sbi;
3168 elr->lr_next_group = start;
3169
3170 /*
3171 * Randomize first schedule time of the request to
3172 * spread the inode table initialization requests
3173 * better.
3174 */
3175 elr->lr_next_sched = jiffies + (prandom_u32() %
3176 (EXT4_DEF_LI_MAX_START_DELAY * HZ));
3177 return elr;
3178 }
3179
3180 int ext4_register_li_request(struct super_block *sb,
3181 ext4_group_t first_not_zeroed)
3182 {
3183 struct ext4_sb_info *sbi = EXT4_SB(sb);
3184 struct ext4_li_request *elr = NULL;
3185 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
3186 int ret = 0;
3187
3188 mutex_lock(&ext4_li_mtx);
3189 if (sbi->s_li_request != NULL) {
3190 /*
3191 * Reset timeout so it can be computed again, because
3192 * s_li_wait_mult might have changed.
3193 */
3194 sbi->s_li_request->lr_timeout = 0;
3195 goto out;
3196 }
3197
3198 if (first_not_zeroed == ngroups ||
3199 (sb->s_flags & MS_RDONLY) ||
3200 !test_opt(sb, INIT_INODE_TABLE))
3201 goto out;
3202
3203 elr = ext4_li_request_new(sb, first_not_zeroed);
3204 if (!elr) {
3205 ret = -ENOMEM;
3206 goto out;
3207 }
3208
3209 if (NULL == ext4_li_info) {
3210 ret = ext4_li_info_new();
3211 if (ret)
3212 goto out;
3213 }
3214
3215 mutex_lock(&ext4_li_info->li_list_mtx);
3216 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3217 mutex_unlock(&ext4_li_info->li_list_mtx);
3218
3219 sbi->s_li_request = elr;
3220 /*
3221 * set elr to NULL here since it has been inserted to
3222 * the request_list and the removal and free of it is
3223 * handled by ext4_clear_request_list from now on.
3224 */
3225 elr = NULL;
3226
3227 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3228 ret = ext4_run_lazyinit_thread();
3229 if (ret)
3230 goto out;
3231 }
3232 out:
3233 mutex_unlock(&ext4_li_mtx);
3234 if (ret)
3235 kfree(elr);
3236 return ret;
3237 }
3238
3239 /*
3240 * We do not need to lock anything since this is called on
3241 * module unload.
3242 */
3243 static void ext4_destroy_lazyinit_thread(void)
3244 {
3245 /*
3246 * If thread exited earlier
3247 * there's nothing to be done.
3248 */
3249 if (!ext4_li_info || !ext4_lazyinit_task)
3250 return;
3251
3252 kthread_stop(ext4_lazyinit_task);
3253 }
3254
3255 static int set_journal_csum_feature_set(struct super_block *sb)
3256 {
3257 int ret = 1;
3258 int compat, incompat;
3259 struct ext4_sb_info *sbi = EXT4_SB(sb);
3260
3261 if (ext4_has_metadata_csum(sb)) {
3262 /* journal checksum v3 */
3263 compat = 0;
3264 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V3;
3265 } else {
3266 /* journal checksum v1 */
3267 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3268 incompat = 0;
3269 }
3270
3271 jbd2_journal_clear_features(sbi->s_journal,
3272 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3273 JBD2_FEATURE_INCOMPAT_CSUM_V3 |
3274 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3275 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3276 ret = jbd2_journal_set_features(sbi->s_journal,
3277 compat, 0,
3278 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3279 incompat);
3280 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3281 ret = jbd2_journal_set_features(sbi->s_journal,
3282 compat, 0,
3283 incompat);
3284 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3285 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3286 } else {
3287 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3288 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3289 }
3290
3291 return ret;
3292 }
3293
3294 /*
3295 * Note: calculating the overhead so we can be compatible with
3296 * historical BSD practice is quite difficult in the face of
3297 * clusters/bigalloc. This is because multiple metadata blocks from
3298 * different block group can end up in the same allocation cluster.
3299 * Calculating the exact overhead in the face of clustered allocation
3300 * requires either O(all block bitmaps) in memory or O(number of block
3301 * groups**2) in time. We will still calculate the superblock for
3302 * older file systems --- and if we come across with a bigalloc file
3303 * system with zero in s_overhead_clusters the estimate will be close to
3304 * correct especially for very large cluster sizes --- but for newer
3305 * file systems, it's better to calculate this figure once at mkfs
3306 * time, and store it in the superblock. If the superblock value is
3307 * present (even for non-bigalloc file systems), we will use it.
3308 */
3309 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3310 char *buf)
3311 {
3312 struct ext4_sb_info *sbi = EXT4_SB(sb);
3313 struct ext4_group_desc *gdp;
3314 ext4_fsblk_t first_block, last_block, b;
3315 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3316 int s, j, count = 0;
3317
3318 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC))
3319 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3320 sbi->s_itb_per_group + 2);
3321
3322 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3323 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3324 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3325 for (i = 0; i < ngroups; i++) {
3326 gdp = ext4_get_group_desc(sb, i, NULL);
3327 b = ext4_block_bitmap(sb, gdp);
3328 if (b >= first_block && b <= last_block) {
3329 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3330 count++;
3331 }
3332 b = ext4_inode_bitmap(sb, gdp);
3333 if (b >= first_block && b <= last_block) {
3334 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3335 count++;
3336 }
3337 b = ext4_inode_table(sb, gdp);
3338 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3339 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3340 int c = EXT4_B2C(sbi, b - first_block);
3341 ext4_set_bit(c, buf);
3342 count++;
3343 }
3344 if (i != grp)
3345 continue;
3346 s = 0;
3347 if (ext4_bg_has_super(sb, grp)) {
3348 ext4_set_bit(s++, buf);
3349 count++;
3350 }
3351 for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
3352 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3353 count++;
3354 }
3355 }
3356 if (!count)
3357 return 0;
3358 return EXT4_CLUSTERS_PER_GROUP(sb) -
3359 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3360 }
3361
3362 /*
3363 * Compute the overhead and stash it in sbi->s_overhead
3364 */
3365 int ext4_calculate_overhead(struct super_block *sb)
3366 {
3367 struct ext4_sb_info *sbi = EXT4_SB(sb);
3368 struct ext4_super_block *es = sbi->s_es;
3369 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3370 ext4_fsblk_t overhead = 0;
3371 char *buf = (char *) get_zeroed_page(GFP_NOFS);
3372
3373 if (!buf)
3374 return -ENOMEM;
3375
3376 /*
3377 * Compute the overhead (FS structures). This is constant
3378 * for a given filesystem unless the number of block groups
3379 * changes so we cache the previous value until it does.
3380 */
3381
3382 /*
3383 * All of the blocks before first_data_block are overhead
3384 */
3385 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3386
3387 /*
3388 * Add the overhead found in each block group
3389 */
3390 for (i = 0; i < ngroups; i++) {
3391 int blks;
3392
3393 blks = count_overhead(sb, i, buf);
3394 overhead += blks;
3395 if (blks)
3396 memset(buf, 0, PAGE_SIZE);
3397 cond_resched();
3398 }
3399 /* Add the internal journal blocks as well */
3400 if (sbi->s_journal && !sbi->journal_bdev)
3401 overhead += EXT4_NUM_B2C(sbi, sbi->s_journal->j_maxlen);
3402
3403 sbi->s_overhead = overhead;
3404 smp_wmb();
3405 free_page((unsigned long) buf);
3406 return 0;
3407 }
3408
3409
3410 static ext4_fsblk_t ext4_calculate_resv_clusters(struct super_block *sb)
3411 {
3412 ext4_fsblk_t resv_clusters;
3413
3414 /*
3415 * There's no need to reserve anything when we aren't using extents.
3416 * The space estimates are exact, there are no unwritten extents,
3417 * hole punching doesn't need new metadata... This is needed especially
3418 * to keep ext2/3 backward compatibility.
3419 */
3420 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS))
3421 return 0;
3422 /*
3423 * By default we reserve 2% or 4096 clusters, whichever is smaller.
3424 * This should cover the situations where we can not afford to run
3425 * out of space like for example punch hole, or converting
3426 * unwritten extents in delalloc path. In most cases such
3427 * allocation would require 1, or 2 blocks, higher numbers are
3428 * very rare.
3429 */
3430 resv_clusters = ext4_blocks_count(EXT4_SB(sb)->s_es) >>
3431 EXT4_SB(sb)->s_cluster_bits;
3432
3433 do_div(resv_clusters, 50);
3434 resv_clusters = min_t(ext4_fsblk_t, resv_clusters, 4096);
3435
3436 return resv_clusters;
3437 }
3438
3439
3440 static int ext4_reserve_clusters(struct ext4_sb_info *sbi, ext4_fsblk_t count)
3441 {
3442 ext4_fsblk_t clusters = ext4_blocks_count(sbi->s_es) >>
3443 sbi->s_cluster_bits;
3444
3445 if (count >= clusters)
3446 return -EINVAL;
3447
3448 atomic64_set(&sbi->s_resv_clusters, count);
3449 return 0;
3450 }
3451
3452 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3453 {
3454 char *orig_data = kstrdup(data, GFP_KERNEL);
3455 struct buffer_head *bh;
3456 struct ext4_super_block *es = NULL;
3457 struct ext4_sb_info *sbi;
3458 ext4_fsblk_t block;
3459 ext4_fsblk_t sb_block = get_sb_block(&data);
3460 ext4_fsblk_t logical_sb_block;
3461 unsigned long offset = 0;
3462 unsigned long journal_devnum = 0;
3463 unsigned long def_mount_opts;
3464 struct inode *root;
3465 const char *descr;
3466 int ret = -ENOMEM;
3467 int blocksize, clustersize;
3468 unsigned int db_count;
3469 unsigned int i;
3470 int needs_recovery, has_huge_files, has_bigalloc;
3471 __u64 blocks_count;
3472 int err = 0;
3473 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3474 ext4_group_t first_not_zeroed;
3475
3476 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3477 if (!sbi)
3478 goto out_free_orig;
3479
3480 sbi->s_blockgroup_lock =
3481 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3482 if (!sbi->s_blockgroup_lock) {
3483 kfree(sbi);
3484 goto out_free_orig;
3485 }
3486 sb->s_fs_info = sbi;
3487 sbi->s_sb = sb;
3488 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3489 sbi->s_sb_block = sb_block;
3490 if (sb->s_bdev->bd_part)
3491 sbi->s_sectors_written_start =
3492 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3493
3494 /* Cleanup superblock name */
3495 strreplace(sb->s_id, '/', '!');
3496
3497 /* -EINVAL is default */
3498 ret = -EINVAL;
3499 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3500 if (!blocksize) {
3501 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3502 goto out_fail;
3503 }
3504
3505 /*
3506 * The ext4 superblock will not be buffer aligned for other than 1kB
3507 * block sizes. We need to calculate the offset from buffer start.
3508 */
3509 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3510 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3511 offset = do_div(logical_sb_block, blocksize);
3512 } else {
3513 logical_sb_block = sb_block;
3514 }
3515
3516 if (!(bh = sb_bread_unmovable(sb, logical_sb_block))) {
3517 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3518 goto out_fail;
3519 }
3520 /*
3521 * Note: s_es must be initialized as soon as possible because
3522 * some ext4 macro-instructions depend on its value
3523 */
3524 es = (struct ext4_super_block *) (bh->b_data + offset);
3525 sbi->s_es = es;
3526 sb->s_magic = le16_to_cpu(es->s_magic);
3527 if (sb->s_magic != EXT4_SUPER_MAGIC)
3528 goto cantfind_ext4;
3529 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3530
3531 /* Warn if metadata_csum and gdt_csum are both set. */
3532 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3533 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
3534 EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_GDT_CSUM))
3535 ext4_warning(sb, "metadata_csum and uninit_bg are "
3536 "redundant flags; please run fsck.");
3537
3538 /* Check for a known checksum algorithm */
3539 if (!ext4_verify_csum_type(sb, es)) {
3540 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3541 "unknown checksum algorithm.");
3542 silent = 1;
3543 goto cantfind_ext4;
3544 }
3545
3546 /* Load the checksum driver */
3547 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3548 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3549 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3550 if (IS_ERR(sbi->s_chksum_driver)) {
3551 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3552 ret = PTR_ERR(sbi->s_chksum_driver);
3553 sbi->s_chksum_driver = NULL;
3554 goto failed_mount;
3555 }
3556 }
3557
3558 /* Check superblock checksum */
3559 if (!ext4_superblock_csum_verify(sb, es)) {
3560 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3561 "invalid superblock checksum. Run e2fsck?");
3562 silent = 1;
3563 goto cantfind_ext4;
3564 }
3565
3566 /* Precompute checksum seed for all metadata */
3567 if (ext4_has_metadata_csum(sb))
3568 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3569 sizeof(es->s_uuid));
3570
3571 /* Set defaults before we parse the mount options */
3572 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3573 set_opt(sb, INIT_INODE_TABLE);
3574 if (def_mount_opts & EXT4_DEFM_DEBUG)
3575 set_opt(sb, DEBUG);
3576 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3577 set_opt(sb, GRPID);
3578 if (def_mount_opts & EXT4_DEFM_UID16)
3579 set_opt(sb, NO_UID32);
3580 /* xattr user namespace & acls are now defaulted on */
3581 set_opt(sb, XATTR_USER);
3582 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3583 set_opt(sb, POSIX_ACL);
3584 #endif
3585 /* don't forget to enable journal_csum when metadata_csum is enabled. */
3586 if (ext4_has_metadata_csum(sb))
3587 set_opt(sb, JOURNAL_CHECKSUM);
3588
3589 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3590 set_opt(sb, JOURNAL_DATA);
3591 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3592 set_opt(sb, ORDERED_DATA);
3593 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3594 set_opt(sb, WRITEBACK_DATA);
3595
3596 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3597 set_opt(sb, ERRORS_PANIC);
3598 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3599 set_opt(sb, ERRORS_CONT);
3600 else
3601 set_opt(sb, ERRORS_RO);
3602 /* block_validity enabled by default; disable with noblock_validity */
3603 set_opt(sb, BLOCK_VALIDITY);
3604 if (def_mount_opts & EXT4_DEFM_DISCARD)
3605 set_opt(sb, DISCARD);
3606
3607 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3608 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
3609 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3610 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3611 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3612
3613 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3614 set_opt(sb, BARRIER);
3615
3616 /*
3617 * enable delayed allocation by default
3618 * Use -o nodelalloc to turn it off
3619 */
3620 if (!IS_EXT3_SB(sb) && !IS_EXT2_SB(sb) &&
3621 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3622 set_opt(sb, DELALLOC);
3623
3624 /*
3625 * set default s_li_wait_mult for lazyinit, for the case there is
3626 * no mount option specified.
3627 */
3628 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3629
3630 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3631 &journal_devnum, &journal_ioprio, 0)) {
3632 ext4_msg(sb, KERN_WARNING,
3633 "failed to parse options in superblock: %s",
3634 sbi->s_es->s_mount_opts);
3635 }
3636 sbi->s_def_mount_opt = sbi->s_mount_opt;
3637 if (!parse_options((char *) data, sb, &journal_devnum,
3638 &journal_ioprio, 0))
3639 goto failed_mount;
3640
3641 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3642 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3643 "with data=journal disables delayed "
3644 "allocation and O_DIRECT support!\n");
3645 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3646 ext4_msg(sb, KERN_ERR, "can't mount with "
3647 "both data=journal and delalloc");
3648 goto failed_mount;
3649 }
3650 if (test_opt(sb, DIOREAD_NOLOCK)) {
3651 ext4_msg(sb, KERN_ERR, "can't mount with "
3652 "both data=journal and dioread_nolock");
3653 goto failed_mount;
3654 }
3655 if (test_opt(sb, DAX)) {
3656 ext4_msg(sb, KERN_ERR, "can't mount with "
3657 "both data=journal and dax");
3658 goto failed_mount;
3659 }
3660 if (test_opt(sb, DELALLOC))
3661 clear_opt(sb, DELALLOC);
3662 }
3663
3664 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3665 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3666
3667 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3668 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3669 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3670 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
3671 ext4_msg(sb, KERN_WARNING,
3672 "feature flags set on rev 0 fs, "
3673 "running e2fsck is recommended");
3674
3675 if (es->s_creator_os == cpu_to_le32(EXT4_OS_HURD)) {
3676 set_opt2(sb, HURD_COMPAT);
3677 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
3678 EXT4_FEATURE_INCOMPAT_64BIT)) {
3679 ext4_msg(sb, KERN_ERR,
3680 "The Hurd can't support 64-bit file systems");
3681 goto failed_mount;
3682 }
3683 }
3684
3685 if (IS_EXT2_SB(sb)) {
3686 if (ext2_feature_set_ok(sb))
3687 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3688 "using the ext4 subsystem");
3689 else {
3690 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3691 "to feature incompatibilities");
3692 goto failed_mount;
3693 }
3694 }
3695
3696 if (IS_EXT3_SB(sb)) {
3697 if (ext3_feature_set_ok(sb))
3698 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3699 "using the ext4 subsystem");
3700 else {
3701 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3702 "to feature incompatibilities");
3703 goto failed_mount;
3704 }
3705 }
3706
3707 /*
3708 * Check feature flags regardless of the revision level, since we
3709 * previously didn't change the revision level when setting the flags,
3710 * so there is a chance incompat flags are set on a rev 0 filesystem.
3711 */
3712 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3713 goto failed_mount;
3714
3715 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3716 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3717 blocksize > EXT4_MAX_BLOCK_SIZE) {
3718 ext4_msg(sb, KERN_ERR,
3719 "Unsupported filesystem blocksize %d", blocksize);
3720 goto failed_mount;
3721 }
3722
3723 if (sbi->s_mount_opt & EXT4_MOUNT_DAX) {
3724 if (blocksize != PAGE_SIZE) {
3725 ext4_msg(sb, KERN_ERR,
3726 "error: unsupported blocksize for dax");
3727 goto failed_mount;
3728 }
3729 if (!sb->s_bdev->bd_disk->fops->direct_access) {
3730 ext4_msg(sb, KERN_ERR,
3731 "error: device does not support dax");
3732 goto failed_mount;
3733 }
3734 }
3735
3736 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_ENCRYPT) &&
3737 es->s_encryption_level) {
3738 ext4_msg(sb, KERN_ERR, "Unsupported encryption level %d",
3739 es->s_encryption_level);
3740 goto failed_mount;
3741 }
3742
3743 if (sb->s_blocksize != blocksize) {
3744 /* Validate the filesystem blocksize */
3745 if (!sb_set_blocksize(sb, blocksize)) {
3746 ext4_msg(sb, KERN_ERR, "bad block size %d",
3747 blocksize);
3748 goto failed_mount;
3749 }
3750
3751 brelse(bh);
3752 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3753 offset = do_div(logical_sb_block, blocksize);
3754 bh = sb_bread_unmovable(sb, logical_sb_block);
3755 if (!bh) {
3756 ext4_msg(sb, KERN_ERR,
3757 "Can't read superblock on 2nd try");
3758 goto failed_mount;
3759 }
3760 es = (struct ext4_super_block *)(bh->b_data + offset);
3761 sbi->s_es = es;
3762 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3763 ext4_msg(sb, KERN_ERR,
3764 "Magic mismatch, very weird!");
3765 goto failed_mount;
3766 }
3767 }
3768
3769 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3770 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
3771 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3772 has_huge_files);
3773 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3774
3775 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3776 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3777 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3778 } else {
3779 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3780 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3781 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3782 (!is_power_of_2(sbi->s_inode_size)) ||
3783 (sbi->s_inode_size > blocksize)) {
3784 ext4_msg(sb, KERN_ERR,
3785 "unsupported inode size: %d",
3786 sbi->s_inode_size);
3787 goto failed_mount;
3788 }
3789 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3790 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3791 }
3792
3793 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3794 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
3795 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3796 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3797 !is_power_of_2(sbi->s_desc_size)) {
3798 ext4_msg(sb, KERN_ERR,
3799 "unsupported descriptor size %lu",
3800 sbi->s_desc_size);
3801 goto failed_mount;
3802 }
3803 } else
3804 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3805
3806 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3807 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3808 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
3809 goto cantfind_ext4;
3810
3811 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3812 if (sbi->s_inodes_per_block == 0)
3813 goto cantfind_ext4;
3814 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3815 sbi->s_inodes_per_block;
3816 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3817 sbi->s_sbh = bh;
3818 sbi->s_mount_state = le16_to_cpu(es->s_state);
3819 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3820 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3821
3822 for (i = 0; i < 4; i++)
3823 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3824 sbi->s_def_hash_version = es->s_def_hash_version;
3825 if (EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_DIR_INDEX)) {
3826 i = le32_to_cpu(es->s_flags);
3827 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3828 sbi->s_hash_unsigned = 3;
3829 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3830 #ifdef __CHAR_UNSIGNED__
3831 if (!(sb->s_flags & MS_RDONLY))
3832 es->s_flags |=
3833 cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3834 sbi->s_hash_unsigned = 3;
3835 #else
3836 if (!(sb->s_flags & MS_RDONLY))
3837 es->s_flags |=
3838 cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3839 #endif
3840 }
3841 }
3842
3843 /* Handle clustersize */
3844 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3845 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3846 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3847 if (has_bigalloc) {
3848 if (clustersize < blocksize) {
3849 ext4_msg(sb, KERN_ERR,
3850 "cluster size (%d) smaller than "
3851 "block size (%d)", clustersize, blocksize);
3852 goto failed_mount;
3853 }
3854 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3855 le32_to_cpu(es->s_log_block_size);
3856 sbi->s_clusters_per_group =
3857 le32_to_cpu(es->s_clusters_per_group);
3858 if (sbi->s_clusters_per_group > blocksize * 8) {
3859 ext4_msg(sb, KERN_ERR,
3860 "#clusters per group too big: %lu",
3861 sbi->s_clusters_per_group);
3862 goto failed_mount;
3863 }
3864 if (sbi->s_blocks_per_group !=
3865 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3866 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3867 "clusters per group (%lu) inconsistent",
3868 sbi->s_blocks_per_group,
3869 sbi->s_clusters_per_group);
3870 goto failed_mount;
3871 }
3872 } else {
3873 if (clustersize != blocksize) {
3874 ext4_warning(sb, "fragment/cluster size (%d) != "
3875 "block size (%d)", clustersize,
3876 blocksize);
3877 clustersize = blocksize;
3878 }
3879 if (sbi->s_blocks_per_group > blocksize * 8) {
3880 ext4_msg(sb, KERN_ERR,
3881 "#blocks per group too big: %lu",
3882 sbi->s_blocks_per_group);
3883 goto failed_mount;
3884 }
3885 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3886 sbi->s_cluster_bits = 0;
3887 }
3888 sbi->s_cluster_ratio = clustersize / blocksize;
3889
3890 if (sbi->s_inodes_per_group > blocksize * 8) {
3891 ext4_msg(sb, KERN_ERR,
3892 "#inodes per group too big: %lu",
3893 sbi->s_inodes_per_group);
3894 goto failed_mount;
3895 }
3896
3897 /* Do we have standard group size of clustersize * 8 blocks ? */
3898 if (sbi->s_blocks_per_group == clustersize << 3)
3899 set_opt2(sb, STD_GROUP_SIZE);
3900
3901 /*
3902 * Test whether we have more sectors than will fit in sector_t,
3903 * and whether the max offset is addressable by the page cache.
3904 */
3905 err = generic_check_addressable(sb->s_blocksize_bits,
3906 ext4_blocks_count(es));
3907 if (err) {
3908 ext4_msg(sb, KERN_ERR, "filesystem"
3909 " too large to mount safely on this system");
3910 if (sizeof(sector_t) < 8)
3911 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3912 goto failed_mount;
3913 }
3914
3915 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3916 goto cantfind_ext4;
3917
3918 /* check blocks count against device size */
3919 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3920 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3921 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3922 "exceeds size of device (%llu blocks)",
3923 ext4_blocks_count(es), blocks_count);
3924 goto failed_mount;
3925 }
3926
3927 /*
3928 * It makes no sense for the first data block to be beyond the end
3929 * of the filesystem.
3930 */
3931 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3932 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
3933 "block %u is beyond end of filesystem (%llu)",
3934 le32_to_cpu(es->s_first_data_block),
3935 ext4_blocks_count(es));
3936 goto failed_mount;
3937 }
3938 blocks_count = (ext4_blocks_count(es) -
3939 le32_to_cpu(es->s_first_data_block) +
3940 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3941 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3942 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3943 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3944 "(block count %llu, first data block %u, "
3945 "blocks per group %lu)", sbi->s_groups_count,
3946 ext4_blocks_count(es),
3947 le32_to_cpu(es->s_first_data_block),
3948 EXT4_BLOCKS_PER_GROUP(sb));
3949 goto failed_mount;
3950 }
3951 sbi->s_groups_count = blocks_count;
3952 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3953 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3954 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3955 EXT4_DESC_PER_BLOCK(sb);
3956 sbi->s_group_desc = ext4_kvmalloc(db_count *
3957 sizeof(struct buffer_head *),
3958 GFP_KERNEL);
3959 if (sbi->s_group_desc == NULL) {
3960 ext4_msg(sb, KERN_ERR, "not enough memory");
3961 ret = -ENOMEM;
3962 goto failed_mount;
3963 }
3964
3965 if (ext4_proc_root)
3966 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
3967
3968 if (sbi->s_proc)
3969 proc_create_data("options", S_IRUGO, sbi->s_proc,
3970 &ext4_seq_options_fops, sb);
3971
3972 bgl_lock_init(sbi->s_blockgroup_lock);
3973
3974 for (i = 0; i < db_count; i++) {
3975 block = descriptor_loc(sb, logical_sb_block, i);
3976 sbi->s_group_desc[i] = sb_bread_unmovable(sb, block);
3977 if (!sbi->s_group_desc[i]) {
3978 ext4_msg(sb, KERN_ERR,
3979 "can't read group descriptor %d", i);
3980 db_count = i;
3981 goto failed_mount2;
3982 }
3983 }
3984 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
3985 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3986 goto failed_mount2;
3987 }
3988
3989 sbi->s_gdb_count = db_count;
3990 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3991 spin_lock_init(&sbi->s_next_gen_lock);
3992
3993 setup_timer(&sbi->s_err_report, print_daily_error_info,
3994 (unsigned long) sb);
3995
3996 /* Register extent status tree shrinker */
3997 if (ext4_es_register_shrinker(sbi))
3998 goto failed_mount3;
3999
4000 sbi->s_stripe = ext4_get_stripe_size(sbi);
4001 sbi->s_extent_max_zeroout_kb = 32;
4002
4003 /*
4004 * set up enough so that it can read an inode
4005 */
4006 sb->s_op = &ext4_sops;
4007 sb->s_export_op = &ext4_export_ops;
4008 sb->s_xattr = ext4_xattr_handlers;
4009 #ifdef CONFIG_QUOTA
4010 sb->dq_op = &ext4_quota_operations;
4011 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
4012 sb->s_qcop = &dquot_quotactl_sysfile_ops;
4013 else
4014 sb->s_qcop = &ext4_qctl_operations;
4015 sb->s_quota_types = QTYPE_MASK_USR | QTYPE_MASK_GRP;
4016 #endif
4017 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
4018
4019 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
4020 mutex_init(&sbi->s_orphan_lock);
4021
4022 sb->s_root = NULL;
4023
4024 needs_recovery = (es->s_last_orphan != 0 ||
4025 EXT4_HAS_INCOMPAT_FEATURE(sb,
4026 EXT4_FEATURE_INCOMPAT_RECOVER));
4027
4028 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
4029 !(sb->s_flags & MS_RDONLY))
4030 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
4031 goto failed_mount3a;
4032
4033 /*
4034 * The first inode we look at is the journal inode. Don't try
4035 * root first: it may be modified in the journal!
4036 */
4037 if (!test_opt(sb, NOLOAD) &&
4038 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4039 if (ext4_load_journal(sb, es, journal_devnum))
4040 goto failed_mount3a;
4041 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
4042 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
4043 ext4_msg(sb, KERN_ERR, "required journal recovery "
4044 "suppressed and not mounted read-only");
4045 goto failed_mount_wq;
4046 } else {
4047 clear_opt(sb, DATA_FLAGS);
4048 sbi->s_journal = NULL;
4049 needs_recovery = 0;
4050 goto no_journal;
4051 }
4052
4053 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT) &&
4054 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
4055 JBD2_FEATURE_INCOMPAT_64BIT)) {
4056 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
4057 goto failed_mount_wq;
4058 }
4059
4060 if (!set_journal_csum_feature_set(sb)) {
4061 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
4062 "feature set");
4063 goto failed_mount_wq;
4064 }
4065
4066 /* We have now updated the journal if required, so we can
4067 * validate the data journaling mode. */
4068 switch (test_opt(sb, DATA_FLAGS)) {
4069 case 0:
4070 /* No mode set, assume a default based on the journal
4071 * capabilities: ORDERED_DATA if the journal can
4072 * cope, else JOURNAL_DATA
4073 */
4074 if (jbd2_journal_check_available_features
4075 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
4076 set_opt(sb, ORDERED_DATA);
4077 else
4078 set_opt(sb, JOURNAL_DATA);
4079 break;
4080
4081 case EXT4_MOUNT_ORDERED_DATA:
4082 case EXT4_MOUNT_WRITEBACK_DATA:
4083 if (!jbd2_journal_check_available_features
4084 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
4085 ext4_msg(sb, KERN_ERR, "Journal does not support "
4086 "requested data journaling mode");
4087 goto failed_mount_wq;
4088 }
4089 default:
4090 break;
4091 }
4092 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4093
4094 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
4095
4096 no_journal:
4097 if (ext4_mballoc_ready) {
4098 sbi->s_mb_cache = ext4_xattr_create_cache(sb->s_id);
4099 if (!sbi->s_mb_cache) {
4100 ext4_msg(sb, KERN_ERR, "Failed to create an mb_cache");
4101 goto failed_mount_wq;
4102 }
4103 }
4104
4105 if ((DUMMY_ENCRYPTION_ENABLED(sbi) ||
4106 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_ENCRYPT)) &&
4107 (blocksize != PAGE_CACHE_SIZE)) {
4108 ext4_msg(sb, KERN_ERR,
4109 "Unsupported blocksize for fs encryption");
4110 goto failed_mount_wq;
4111 }
4112
4113 if (DUMMY_ENCRYPTION_ENABLED(sbi) &&
4114 !(sb->s_flags & MS_RDONLY) &&
4115 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_ENCRYPT)) {
4116 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_ENCRYPT);
4117 ext4_commit_super(sb, 1);
4118 }
4119
4120 /*
4121 * Get the # of file system overhead blocks from the
4122 * superblock if present.
4123 */
4124 if (es->s_overhead_clusters)
4125 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
4126 else {
4127 err = ext4_calculate_overhead(sb);
4128 if (err)
4129 goto failed_mount_wq;
4130 }
4131
4132 /*
4133 * The maximum number of concurrent works can be high and
4134 * concurrency isn't really necessary. Limit it to 1.
4135 */
4136 EXT4_SB(sb)->rsv_conversion_wq =
4137 alloc_workqueue("ext4-rsv-conversion", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
4138 if (!EXT4_SB(sb)->rsv_conversion_wq) {
4139 printk(KERN_ERR "EXT4-fs: failed to create workqueue\n");
4140 ret = -ENOMEM;
4141 goto failed_mount4;
4142 }
4143
4144 /*
4145 * The jbd2_journal_load will have done any necessary log recovery,
4146 * so we can safely mount the rest of the filesystem now.
4147 */
4148
4149 root = ext4_iget(sb, EXT4_ROOT_INO);
4150 if (IS_ERR(root)) {
4151 ext4_msg(sb, KERN_ERR, "get root inode failed");
4152 ret = PTR_ERR(root);
4153 root = NULL;
4154 goto failed_mount4;
4155 }
4156 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
4157 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
4158 iput(root);
4159 goto failed_mount4;
4160 }
4161 sb->s_root = d_make_root(root);
4162 if (!sb->s_root) {
4163 ext4_msg(sb, KERN_ERR, "get root dentry failed");
4164 ret = -ENOMEM;
4165 goto failed_mount4;
4166 }
4167
4168 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
4169 sb->s_flags |= MS_RDONLY;
4170
4171 /* determine the minimum size of new large inodes, if present */
4172 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
4173 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4174 EXT4_GOOD_OLD_INODE_SIZE;
4175 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4176 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
4177 if (sbi->s_want_extra_isize <
4178 le16_to_cpu(es->s_want_extra_isize))
4179 sbi->s_want_extra_isize =
4180 le16_to_cpu(es->s_want_extra_isize);
4181 if (sbi->s_want_extra_isize <
4182 le16_to_cpu(es->s_min_extra_isize))
4183 sbi->s_want_extra_isize =
4184 le16_to_cpu(es->s_min_extra_isize);
4185 }
4186 }
4187 /* Check if enough inode space is available */
4188 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
4189 sbi->s_inode_size) {
4190 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
4191 EXT4_GOOD_OLD_INODE_SIZE;
4192 ext4_msg(sb, KERN_INFO, "required extra inode space not"
4193 "available");
4194 }
4195
4196 err = ext4_reserve_clusters(sbi, ext4_calculate_resv_clusters(sb));
4197 if (err) {
4198 ext4_msg(sb, KERN_ERR, "failed to reserve %llu clusters for "
4199 "reserved pool", ext4_calculate_resv_clusters(sb));
4200 goto failed_mount4a;
4201 }
4202
4203 err = ext4_setup_system_zone(sb);
4204 if (err) {
4205 ext4_msg(sb, KERN_ERR, "failed to initialize system "
4206 "zone (%d)", err);
4207 goto failed_mount4a;
4208 }
4209
4210 ext4_ext_init(sb);
4211 err = ext4_mb_init(sb);
4212 if (err) {
4213 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
4214 err);
4215 goto failed_mount5;
4216 }
4217
4218 block = ext4_count_free_clusters(sb);
4219 ext4_free_blocks_count_set(sbi->s_es,
4220 EXT4_C2B(sbi, block));
4221 err = percpu_counter_init(&sbi->s_freeclusters_counter, block,
4222 GFP_KERNEL);
4223 if (!err) {
4224 unsigned long freei = ext4_count_free_inodes(sb);
4225 sbi->s_es->s_free_inodes_count = cpu_to_le32(freei);
4226 err = percpu_counter_init(&sbi->s_freeinodes_counter, freei,
4227 GFP_KERNEL);
4228 }
4229 if (!err)
4230 err = percpu_counter_init(&sbi->s_dirs_counter,
4231 ext4_count_dirs(sb), GFP_KERNEL);
4232 if (!err)
4233 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0,
4234 GFP_KERNEL);
4235 if (err) {
4236 ext4_msg(sb, KERN_ERR, "insufficient memory");
4237 goto failed_mount6;
4238 }
4239
4240 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
4241 if (!ext4_fill_flex_info(sb)) {
4242 ext4_msg(sb, KERN_ERR,
4243 "unable to initialize "
4244 "flex_bg meta info!");
4245 goto failed_mount6;
4246 }
4247
4248 err = ext4_register_li_request(sb, first_not_zeroed);
4249 if (err)
4250 goto failed_mount6;
4251
4252 sbi->s_kobj.kset = ext4_kset;
4253 init_completion(&sbi->s_kobj_unregister);
4254 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
4255 "%s", sb->s_id);
4256 if (err)
4257 goto failed_mount7;
4258
4259 #ifdef CONFIG_QUOTA
4260 /* Enable quota usage during mount. */
4261 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
4262 !(sb->s_flags & MS_RDONLY)) {
4263 err = ext4_enable_quotas(sb);
4264 if (err)
4265 goto failed_mount8;
4266 }
4267 #endif /* CONFIG_QUOTA */
4268
4269 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
4270 ext4_orphan_cleanup(sb, es);
4271 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
4272 if (needs_recovery) {
4273 ext4_msg(sb, KERN_INFO, "recovery complete");
4274 ext4_mark_recovery_complete(sb, es);
4275 }
4276 if (EXT4_SB(sb)->s_journal) {
4277 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
4278 descr = " journalled data mode";
4279 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
4280 descr = " ordered data mode";
4281 else
4282 descr = " writeback data mode";
4283 } else
4284 descr = "out journal";
4285
4286 if (test_opt(sb, DISCARD)) {
4287 struct request_queue *q = bdev_get_queue(sb->s_bdev);
4288 if (!blk_queue_discard(q))
4289 ext4_msg(sb, KERN_WARNING,
4290 "mounting with \"discard\" option, but "
4291 "the device does not support discard");
4292 }
4293
4294 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
4295 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
4296 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
4297
4298 if (es->s_error_count)
4299 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
4300
4301 /* Enable message ratelimiting. Default is 10 messages per 5 secs. */
4302 ratelimit_state_init(&sbi->s_err_ratelimit_state, 5 * HZ, 10);
4303 ratelimit_state_init(&sbi->s_warning_ratelimit_state, 5 * HZ, 10);
4304 ratelimit_state_init(&sbi->s_msg_ratelimit_state, 5 * HZ, 10);
4305
4306 kfree(orig_data);
4307 return 0;
4308
4309 cantfind_ext4:
4310 if (!silent)
4311 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
4312 goto failed_mount;
4313
4314 #ifdef CONFIG_QUOTA
4315 failed_mount8:
4316 kobject_del(&sbi->s_kobj);
4317 #endif
4318 failed_mount7:
4319 ext4_unregister_li_request(sb);
4320 failed_mount6:
4321 ext4_mb_release(sb);
4322 if (sbi->s_flex_groups)
4323 kvfree(sbi->s_flex_groups);
4324 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4325 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4326 percpu_counter_destroy(&sbi->s_dirs_counter);
4327 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
4328 failed_mount5:
4329 ext4_ext_release(sb);
4330 ext4_release_system_zone(sb);
4331 failed_mount4a:
4332 dput(sb->s_root);
4333 sb->s_root = NULL;
4334 failed_mount4:
4335 ext4_msg(sb, KERN_ERR, "mount failed");
4336 if (EXT4_SB(sb)->rsv_conversion_wq)
4337 destroy_workqueue(EXT4_SB(sb)->rsv_conversion_wq);
4338 failed_mount_wq:
4339 if (sbi->s_journal) {
4340 jbd2_journal_destroy(sbi->s_journal);
4341 sbi->s_journal = NULL;
4342 }
4343 failed_mount3a:
4344 ext4_es_unregister_shrinker(sbi);
4345 failed_mount3:
4346 del_timer_sync(&sbi->s_err_report);
4347 if (sbi->s_mmp_tsk)
4348 kthread_stop(sbi->s_mmp_tsk);
4349 failed_mount2:
4350 for (i = 0; i < db_count; i++)
4351 brelse(sbi->s_group_desc[i]);
4352 kvfree(sbi->s_group_desc);
4353 failed_mount:
4354 if (sbi->s_chksum_driver)
4355 crypto_free_shash(sbi->s_chksum_driver);
4356 if (sbi->s_proc) {
4357 remove_proc_entry("options", sbi->s_proc);
4358 remove_proc_entry(sb->s_id, ext4_proc_root);
4359 }
4360 #ifdef CONFIG_QUOTA
4361 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4362 kfree(sbi->s_qf_names[i]);
4363 #endif
4364 ext4_blkdev_remove(sbi);
4365 brelse(bh);
4366 out_fail:
4367 sb->s_fs_info = NULL;
4368 kfree(sbi->s_blockgroup_lock);
4369 kfree(sbi);
4370 out_free_orig:
4371 kfree(orig_data);
4372 return err ? err : ret;
4373 }
4374
4375 /*
4376 * Setup any per-fs journal parameters now. We'll do this both on
4377 * initial mount, once the journal has been initialised but before we've
4378 * done any recovery; and again on any subsequent remount.
4379 */
4380 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
4381 {
4382 struct ext4_sb_info *sbi = EXT4_SB(sb);
4383
4384 journal->j_commit_interval = sbi->s_commit_interval;
4385 journal->j_min_batch_time = sbi->s_min_batch_time;
4386 journal->j_max_batch_time = sbi->s_max_batch_time;
4387
4388 write_lock(&journal->j_state_lock);
4389 if (test_opt(sb, BARRIER))
4390 journal->j_flags |= JBD2_BARRIER;
4391 else
4392 journal->j_flags &= ~JBD2_BARRIER;
4393 if (test_opt(sb, DATA_ERR_ABORT))
4394 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4395 else
4396 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
4397 write_unlock(&journal->j_state_lock);
4398 }
4399
4400 static journal_t *ext4_get_journal(struct super_block *sb,
4401 unsigned int journal_inum)
4402 {
4403 struct inode *journal_inode;
4404 journal_t *journal;
4405
4406 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4407
4408 /* First, test for the existence of a valid inode on disk. Bad
4409 * things happen if we iget() an unused inode, as the subsequent
4410 * iput() will try to delete it. */
4411
4412 journal_inode = ext4_iget(sb, journal_inum);
4413 if (IS_ERR(journal_inode)) {
4414 ext4_msg(sb, KERN_ERR, "no journal found");
4415 return NULL;
4416 }
4417 if (!journal_inode->i_nlink) {
4418 make_bad_inode(journal_inode);
4419 iput(journal_inode);
4420 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
4421 return NULL;
4422 }
4423
4424 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4425 journal_inode, journal_inode->i_size);
4426 if (!S_ISREG(journal_inode->i_mode)) {
4427 ext4_msg(sb, KERN_ERR, "invalid journal inode");
4428 iput(journal_inode);
4429 return NULL;
4430 }
4431
4432 journal = jbd2_journal_init_inode(journal_inode);
4433 if (!journal) {
4434 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
4435 iput(journal_inode);
4436 return NULL;
4437 }
4438 journal->j_private = sb;
4439 ext4_init_journal_params(sb, journal);
4440 return journal;
4441 }
4442
4443 static journal_t *ext4_get_dev_journal(struct super_block *sb,
4444 dev_t j_dev)
4445 {
4446 struct buffer_head *bh;
4447 journal_t *journal;
4448 ext4_fsblk_t start;
4449 ext4_fsblk_t len;
4450 int hblock, blocksize;
4451 ext4_fsblk_t sb_block;
4452 unsigned long offset;
4453 struct ext4_super_block *es;
4454 struct block_device *bdev;
4455
4456 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4457
4458 bdev = ext4_blkdev_get(j_dev, sb);
4459 if (bdev == NULL)
4460 return NULL;
4461
4462 blocksize = sb->s_blocksize;
4463 hblock = bdev_logical_block_size(bdev);
4464 if (blocksize < hblock) {
4465 ext4_msg(sb, KERN_ERR,
4466 "blocksize too small for journal device");
4467 goto out_bdev;
4468 }
4469
4470 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4471 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
4472 set_blocksize(bdev, blocksize);
4473 if (!(bh = __bread(bdev, sb_block, blocksize))) {
4474 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4475 "external journal");
4476 goto out_bdev;
4477 }
4478
4479 es = (struct ext4_super_block *) (bh->b_data + offset);
4480 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
4481 !(le32_to_cpu(es->s_feature_incompat) &
4482 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4483 ext4_msg(sb, KERN_ERR, "external journal has "
4484 "bad superblock");
4485 brelse(bh);
4486 goto out_bdev;
4487 }
4488
4489 if ((le32_to_cpu(es->s_feature_ro_compat) &
4490 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
4491 es->s_checksum != ext4_superblock_csum(sb, es)) {
4492 ext4_msg(sb, KERN_ERR, "external journal has "
4493 "corrupt superblock");
4494 brelse(bh);
4495 goto out_bdev;
4496 }
4497
4498 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4499 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4500 brelse(bh);
4501 goto out_bdev;
4502 }
4503
4504 len = ext4_blocks_count(es);
4505 start = sb_block + 1;
4506 brelse(bh); /* we're done with the superblock */
4507
4508 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4509 start, len, blocksize);
4510 if (!journal) {
4511 ext4_msg(sb, KERN_ERR, "failed to create device journal");
4512 goto out_bdev;
4513 }
4514 journal->j_private = sb;
4515 ll_rw_block(READ | REQ_META | REQ_PRIO, 1, &journal->j_sb_buffer);
4516 wait_on_buffer(journal->j_sb_buffer);
4517 if (!buffer_uptodate(journal->j_sb_buffer)) {
4518 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4519 goto out_journal;
4520 }
4521 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4522 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4523 "user (unsupported) - %d",
4524 be32_to_cpu(journal->j_superblock->s_nr_users));
4525 goto out_journal;
4526 }
4527 EXT4_SB(sb)->journal_bdev = bdev;
4528 ext4_init_journal_params(sb, journal);
4529 return journal;
4530
4531 out_journal:
4532 jbd2_journal_destroy(journal);
4533 out_bdev:
4534 ext4_blkdev_put(bdev);
4535 return NULL;
4536 }
4537
4538 static int ext4_load_journal(struct super_block *sb,
4539 struct ext4_super_block *es,
4540 unsigned long journal_devnum)
4541 {
4542 journal_t *journal;
4543 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4544 dev_t journal_dev;
4545 int err = 0;
4546 int really_read_only;
4547
4548 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4549
4550 if (journal_devnum &&
4551 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4552 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4553 "numbers have changed");
4554 journal_dev = new_decode_dev(journal_devnum);
4555 } else
4556 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4557
4558 really_read_only = bdev_read_only(sb->s_bdev);
4559
4560 /*
4561 * Are we loading a blank journal or performing recovery after a
4562 * crash? For recovery, we need to check in advance whether we
4563 * can get read-write access to the device.
4564 */
4565 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
4566 if (sb->s_flags & MS_RDONLY) {
4567 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4568 "required on readonly filesystem");
4569 if (really_read_only) {
4570 ext4_msg(sb, KERN_ERR, "write access "
4571 "unavailable, cannot proceed");
4572 return -EROFS;
4573 }
4574 ext4_msg(sb, KERN_INFO, "write access will "
4575 "be enabled during recovery");
4576 }
4577 }
4578
4579 if (journal_inum && journal_dev) {
4580 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4581 "and inode journals!");
4582 return -EINVAL;
4583 }
4584
4585 if (journal_inum) {
4586 if (!(journal = ext4_get_journal(sb, journal_inum)))
4587 return -EINVAL;
4588 } else {
4589 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4590 return -EINVAL;
4591 }
4592
4593 if (!(journal->j_flags & JBD2_BARRIER))
4594 ext4_msg(sb, KERN_INFO, "barriers disabled");
4595
4596 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
4597 err = jbd2_journal_wipe(journal, !really_read_only);
4598 if (!err) {
4599 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4600 if (save)
4601 memcpy(save, ((char *) es) +
4602 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4603 err = jbd2_journal_load(journal);
4604 if (save)
4605 memcpy(((char *) es) + EXT4_S_ERR_START,
4606 save, EXT4_S_ERR_LEN);
4607 kfree(save);
4608 }
4609
4610 if (err) {
4611 ext4_msg(sb, KERN_ERR, "error loading journal");
4612 jbd2_journal_destroy(journal);
4613 return err;
4614 }
4615
4616 EXT4_SB(sb)->s_journal = journal;
4617 ext4_clear_journal_err(sb, es);
4618
4619 if (!really_read_only && journal_devnum &&
4620 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4621 es->s_journal_dev = cpu_to_le32(journal_devnum);
4622
4623 /* Make sure we flush the recovery flag to disk. */
4624 ext4_commit_super(sb, 1);
4625 }
4626
4627 return 0;
4628 }
4629
4630 static int ext4_commit_super(struct super_block *sb, int sync)
4631 {
4632 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4633 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4634 int error = 0;
4635
4636 if (!sbh || block_device_ejected(sb))
4637 return error;
4638 if (buffer_write_io_error(sbh)) {
4639 /*
4640 * Oh, dear. A previous attempt to write the
4641 * superblock failed. This could happen because the
4642 * USB device was yanked out. Or it could happen to
4643 * be a transient write error and maybe the block will
4644 * be remapped. Nothing we can do but to retry the
4645 * write and hope for the best.
4646 */
4647 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4648 "superblock detected");
4649 clear_buffer_write_io_error(sbh);
4650 set_buffer_uptodate(sbh);
4651 }
4652 /*
4653 * If the file system is mounted read-only, don't update the
4654 * superblock write time. This avoids updating the superblock
4655 * write time when we are mounting the root file system
4656 * read/only but we need to replay the journal; at that point,
4657 * for people who are east of GMT and who make their clock
4658 * tick in localtime for Windows bug-for-bug compatibility,
4659 * the clock is set in the future, and this will cause e2fsck
4660 * to complain and force a full file system check.
4661 */
4662 if (!(sb->s_flags & MS_RDONLY))
4663 es->s_wtime = cpu_to_le32(get_seconds());
4664 if (sb->s_bdev->bd_part)
4665 es->s_kbytes_written =
4666 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4667 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4668 EXT4_SB(sb)->s_sectors_written_start) >> 1));
4669 else
4670 es->s_kbytes_written =
4671 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4672 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeclusters_counter))
4673 ext4_free_blocks_count_set(es,
4674 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4675 &EXT4_SB(sb)->s_freeclusters_counter)));
4676 if (percpu_counter_initialized(&EXT4_SB(sb)->s_freeinodes_counter))
4677 es->s_free_inodes_count =
4678 cpu_to_le32(percpu_counter_sum_positive(
4679 &EXT4_SB(sb)->s_freeinodes_counter));
4680 BUFFER_TRACE(sbh, "marking dirty");
4681 ext4_superblock_csum_set(sb);
4682 mark_buffer_dirty(sbh);
4683 if (sync) {
4684 error = sync_dirty_buffer(sbh);
4685 if (error)
4686 return error;
4687
4688 error = buffer_write_io_error(sbh);
4689 if (error) {
4690 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4691 "superblock");
4692 clear_buffer_write_io_error(sbh);
4693 set_buffer_uptodate(sbh);
4694 }
4695 }
4696 return error;
4697 }
4698
4699 /*
4700 * Have we just finished recovery? If so, and if we are mounting (or
4701 * remounting) the filesystem readonly, then we will end up with a
4702 * consistent fs on disk. Record that fact.
4703 */
4704 static void ext4_mark_recovery_complete(struct super_block *sb,
4705 struct ext4_super_block *es)
4706 {
4707 journal_t *journal = EXT4_SB(sb)->s_journal;
4708
4709 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4710 BUG_ON(journal != NULL);
4711 return;
4712 }
4713 jbd2_journal_lock_updates(journal);
4714 if (jbd2_journal_flush(journal) < 0)
4715 goto out;
4716
4717 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
4718 sb->s_flags & MS_RDONLY) {
4719 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4720 ext4_commit_super(sb, 1);
4721 }
4722
4723 out:
4724 jbd2_journal_unlock_updates(journal);
4725 }
4726
4727 /*
4728 * If we are mounting (or read-write remounting) a filesystem whose journal
4729 * has recorded an error from a previous lifetime, move that error to the
4730 * main filesystem now.
4731 */
4732 static void ext4_clear_journal_err(struct super_block *sb,
4733 struct ext4_super_block *es)
4734 {
4735 journal_t *journal;
4736 int j_errno;
4737 const char *errstr;
4738
4739 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4740
4741 journal = EXT4_SB(sb)->s_journal;
4742
4743 /*
4744 * Now check for any error status which may have been recorded in the
4745 * journal by a prior ext4_error() or ext4_abort()
4746 */
4747
4748 j_errno = jbd2_journal_errno(journal);
4749 if (j_errno) {
4750 char nbuf[16];
4751
4752 errstr = ext4_decode_error(sb, j_errno, nbuf);
4753 ext4_warning(sb, "Filesystem error recorded "
4754 "from previous mount: %s", errstr);
4755 ext4_warning(sb, "Marking fs in need of filesystem check.");
4756
4757 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4758 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4759 ext4_commit_super(sb, 1);
4760
4761 jbd2_journal_clear_err(journal);
4762 jbd2_journal_update_sb_errno(journal);
4763 }
4764 }
4765
4766 /*
4767 * Force the running and committing transactions to commit,
4768 * and wait on the commit.
4769 */
4770 int ext4_force_commit(struct super_block *sb)
4771 {
4772 journal_t *journal;
4773
4774 if (sb->s_flags & MS_RDONLY)
4775 return 0;
4776
4777 journal = EXT4_SB(sb)->s_journal;
4778 return ext4_journal_force_commit(journal);
4779 }
4780
4781 static int ext4_sync_fs(struct super_block *sb, int wait)
4782 {
4783 int ret = 0;
4784 tid_t target;
4785 bool needs_barrier = false;
4786 struct ext4_sb_info *sbi = EXT4_SB(sb);
4787
4788 trace_ext4_sync_fs(sb, wait);
4789 flush_workqueue(sbi->rsv_conversion_wq);
4790 /*
4791 * Writeback quota in non-journalled quota case - journalled quota has
4792 * no dirty dquots
4793 */
4794 dquot_writeback_dquots(sb, -1);
4795 /*
4796 * Data writeback is possible w/o journal transaction, so barrier must
4797 * being sent at the end of the function. But we can skip it if
4798 * transaction_commit will do it for us.
4799 */
4800 if (sbi->s_journal) {
4801 target = jbd2_get_latest_transaction(sbi->s_journal);
4802 if (wait && sbi->s_journal->j_flags & JBD2_BARRIER &&
4803 !jbd2_trans_will_send_data_barrier(sbi->s_journal, target))
4804 needs_barrier = true;
4805
4806 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4807 if (wait)
4808 ret = jbd2_log_wait_commit(sbi->s_journal,
4809 target);
4810 }
4811 } else if (wait && test_opt(sb, BARRIER))
4812 needs_barrier = true;
4813 if (needs_barrier) {
4814 int err;
4815 err = blkdev_issue_flush(sb->s_bdev, GFP_KERNEL, NULL);
4816 if (!ret)
4817 ret = err;
4818 }
4819
4820 return ret;
4821 }
4822
4823 /*
4824 * LVM calls this function before a (read-only) snapshot is created. This
4825 * gives us a chance to flush the journal completely and mark the fs clean.
4826 *
4827 * Note that only this function cannot bring a filesystem to be in a clean
4828 * state independently. It relies on upper layer to stop all data & metadata
4829 * modifications.
4830 */
4831 static int ext4_freeze(struct super_block *sb)
4832 {
4833 int error = 0;
4834 journal_t *journal;
4835
4836 if (sb->s_flags & MS_RDONLY)
4837 return 0;
4838
4839 journal = EXT4_SB(sb)->s_journal;
4840
4841 if (journal) {
4842 /* Now we set up the journal barrier. */
4843 jbd2_journal_lock_updates(journal);
4844
4845 /*
4846 * Don't clear the needs_recovery flag if we failed to
4847 * flush the journal.
4848 */
4849 error = jbd2_journal_flush(journal);
4850 if (error < 0)
4851 goto out;
4852
4853 /* Journal blocked and flushed, clear needs_recovery flag. */
4854 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4855 }
4856
4857 error = ext4_commit_super(sb, 1);
4858 out:
4859 if (journal)
4860 /* we rely on upper layer to stop further updates */
4861 jbd2_journal_unlock_updates(journal);
4862 return error;
4863 }
4864
4865 /*
4866 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4867 * flag here, even though the filesystem is not technically dirty yet.
4868 */
4869 static int ext4_unfreeze(struct super_block *sb)
4870 {
4871 if (sb->s_flags & MS_RDONLY)
4872 return 0;
4873
4874 if (EXT4_SB(sb)->s_journal) {
4875 /* Reset the needs_recovery flag before the fs is unlocked. */
4876 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4877 }
4878
4879 ext4_commit_super(sb, 1);
4880 return 0;
4881 }
4882
4883 /*
4884 * Structure to save mount options for ext4_remount's benefit
4885 */
4886 struct ext4_mount_options {
4887 unsigned long s_mount_opt;
4888 unsigned long s_mount_opt2;
4889 kuid_t s_resuid;
4890 kgid_t s_resgid;
4891 unsigned long s_commit_interval;
4892 u32 s_min_batch_time, s_max_batch_time;
4893 #ifdef CONFIG_QUOTA
4894 int s_jquota_fmt;
4895 char *s_qf_names[EXT4_MAXQUOTAS];
4896 #endif
4897 };
4898
4899 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4900 {
4901 struct ext4_super_block *es;
4902 struct ext4_sb_info *sbi = EXT4_SB(sb);
4903 unsigned long old_sb_flags;
4904 struct ext4_mount_options old_opts;
4905 int enable_quota = 0;
4906 ext4_group_t g;
4907 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4908 int err = 0;
4909 #ifdef CONFIG_QUOTA
4910 int i, j;
4911 #endif
4912 char *orig_data = kstrdup(data, GFP_KERNEL);
4913
4914 /* Store the original options */
4915 old_sb_flags = sb->s_flags;
4916 old_opts.s_mount_opt = sbi->s_mount_opt;
4917 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4918 old_opts.s_resuid = sbi->s_resuid;
4919 old_opts.s_resgid = sbi->s_resgid;
4920 old_opts.s_commit_interval = sbi->s_commit_interval;
4921 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4922 old_opts.s_max_batch_time = sbi->s_max_batch_time;
4923 #ifdef CONFIG_QUOTA
4924 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4925 for (i = 0; i < EXT4_MAXQUOTAS; i++)
4926 if (sbi->s_qf_names[i]) {
4927 old_opts.s_qf_names[i] = kstrdup(sbi->s_qf_names[i],
4928 GFP_KERNEL);
4929 if (!old_opts.s_qf_names[i]) {
4930 for (j = 0; j < i; j++)
4931 kfree(old_opts.s_qf_names[j]);
4932 kfree(orig_data);
4933 return -ENOMEM;
4934 }
4935 } else
4936 old_opts.s_qf_names[i] = NULL;
4937 #endif
4938 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4939 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4940
4941 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
4942 err = -EINVAL;
4943 goto restore_opts;
4944 }
4945
4946 if ((old_opts.s_mount_opt & EXT4_MOUNT_JOURNAL_CHECKSUM) ^
4947 test_opt(sb, JOURNAL_CHECKSUM)) {
4948 ext4_msg(sb, KERN_ERR, "changing journal_checksum "
4949 "during remount not supported; ignoring");
4950 sbi->s_mount_opt ^= EXT4_MOUNT_JOURNAL_CHECKSUM;
4951 }
4952
4953 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
4954 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
4955 ext4_msg(sb, KERN_ERR, "can't mount with "
4956 "both data=journal and delalloc");
4957 err = -EINVAL;
4958 goto restore_opts;
4959 }
4960 if (test_opt(sb, DIOREAD_NOLOCK)) {
4961 ext4_msg(sb, KERN_ERR, "can't mount with "
4962 "both data=journal and dioread_nolock");
4963 err = -EINVAL;
4964 goto restore_opts;
4965 }
4966 if (test_opt(sb, DAX)) {
4967 ext4_msg(sb, KERN_ERR, "can't mount with "
4968 "both data=journal and dax");
4969 err = -EINVAL;
4970 goto restore_opts;
4971 }
4972 }
4973
4974 if ((sbi->s_mount_opt ^ old_opts.s_mount_opt) & EXT4_MOUNT_DAX) {
4975 ext4_msg(sb, KERN_WARNING, "warning: refusing change of "
4976 "dax flag with busy inodes while remounting");
4977 sbi->s_mount_opt ^= EXT4_MOUNT_DAX;
4978 }
4979
4980 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4981 ext4_abort(sb, "Abort forced by user");
4982
4983 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
4984 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
4985
4986 es = sbi->s_es;
4987
4988 if (sbi->s_journal) {
4989 ext4_init_journal_params(sb, sbi->s_journal);
4990 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4991 }
4992
4993 if (*flags & MS_LAZYTIME)
4994 sb->s_flags |= MS_LAZYTIME;
4995
4996 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4997 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
4998 err = -EROFS;
4999 goto restore_opts;
5000 }
5001
5002 if (*flags & MS_RDONLY) {
5003 err = sync_filesystem(sb);
5004 if (err < 0)
5005 goto restore_opts;
5006 err = dquot_suspend(sb, -1);
5007 if (err < 0)
5008 goto restore_opts;
5009
5010 /*
5011 * First of all, the unconditional stuff we have to do
5012 * to disable replay of the journal when we next remount
5013 */
5014 sb->s_flags |= MS_RDONLY;
5015
5016 /*
5017 * OK, test if we are remounting a valid rw partition
5018 * readonly, and if so set the rdonly flag and then
5019 * mark the partition as valid again.
5020 */
5021 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
5022 (sbi->s_mount_state & EXT4_VALID_FS))
5023 es->s_state = cpu_to_le16(sbi->s_mount_state);
5024
5025 if (sbi->s_journal)
5026 ext4_mark_recovery_complete(sb, es);
5027 } else {
5028 /* Make sure we can mount this feature set readwrite */
5029 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
5030 EXT4_FEATURE_RO_COMPAT_READONLY) ||
5031 !ext4_feature_set_ok(sb, 0)) {
5032 err = -EROFS;
5033 goto restore_opts;
5034 }
5035 /*
5036 * Make sure the group descriptor checksums
5037 * are sane. If they aren't, refuse to remount r/w.
5038 */
5039 for (g = 0; g < sbi->s_groups_count; g++) {
5040 struct ext4_group_desc *gdp =
5041 ext4_get_group_desc(sb, g, NULL);
5042
5043 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
5044 ext4_msg(sb, KERN_ERR,
5045 "ext4_remount: Checksum for group %u failed (%u!=%u)",
5046 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
5047 le16_to_cpu(gdp->bg_checksum));
5048 err = -EINVAL;
5049 goto restore_opts;
5050 }
5051 }
5052
5053 /*
5054 * If we have an unprocessed orphan list hanging
5055 * around from a previously readonly bdev mount,
5056 * require a full umount/remount for now.
5057 */
5058 if (es->s_last_orphan) {
5059 ext4_msg(sb, KERN_WARNING, "Couldn't "
5060 "remount RDWR because of unprocessed "
5061 "orphan inode list. Please "
5062 "umount/remount instead");
5063 err = -EINVAL;
5064 goto restore_opts;
5065 }
5066
5067 /*
5068 * Mounting a RDONLY partition read-write, so reread
5069 * and store the current valid flag. (It may have
5070 * been changed by e2fsck since we originally mounted
5071 * the partition.)
5072 */
5073 if (sbi->s_journal)
5074 ext4_clear_journal_err(sb, es);
5075 sbi->s_mount_state = le16_to_cpu(es->s_state);
5076 if (!ext4_setup_super(sb, es, 0))
5077 sb->s_flags &= ~MS_RDONLY;
5078 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
5079 EXT4_FEATURE_INCOMPAT_MMP))
5080 if (ext4_multi_mount_protect(sb,
5081 le64_to_cpu(es->s_mmp_block))) {
5082 err = -EROFS;
5083 goto restore_opts;
5084 }
5085 enable_quota = 1;
5086 }
5087 }
5088
5089 /*
5090 * Reinitialize lazy itable initialization thread based on
5091 * current settings
5092 */
5093 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
5094 ext4_unregister_li_request(sb);
5095 else {
5096 ext4_group_t first_not_zeroed;
5097 first_not_zeroed = ext4_has_uninit_itable(sb);
5098 ext4_register_li_request(sb, first_not_zeroed);
5099 }
5100
5101 ext4_setup_system_zone(sb);
5102 if (sbi->s_journal == NULL && !(old_sb_flags & MS_RDONLY))
5103 ext4_commit_super(sb, 1);
5104
5105 #ifdef CONFIG_QUOTA
5106 /* Release old quota file names */
5107 for (i = 0; i < EXT4_MAXQUOTAS; i++)
5108 kfree(old_opts.s_qf_names[i]);
5109 if (enable_quota) {
5110 if (sb_any_quota_suspended(sb))
5111 dquot_resume(sb, -1);
5112 else if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
5113 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
5114 err = ext4_enable_quotas(sb);
5115 if (err)
5116 goto restore_opts;
5117 }
5118 }
5119 #endif
5120
5121 *flags = (*flags & ~MS_LAZYTIME) | (sb->s_flags & MS_LAZYTIME);
5122 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
5123 kfree(orig_data);
5124 return 0;
5125
5126 restore_opts:
5127 sb->s_flags = old_sb_flags;
5128 sbi->s_mount_opt = old_opts.s_mount_opt;
5129 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
5130 sbi->s_resuid = old_opts.s_resuid;
5131 sbi->s_resgid = old_opts.s_resgid;
5132 sbi->s_commit_interval = old_opts.s_commit_interval;
5133 sbi->s_min_batch_time = old_opts.s_min_batch_time;
5134 sbi->s_max_batch_time = old_opts.s_max_batch_time;
5135 #ifdef CONFIG_QUOTA
5136 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
5137 for (i = 0; i < EXT4_MAXQUOTAS; i++) {
5138 kfree(sbi->s_qf_names[i]);
5139 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
5140 }
5141 #endif
5142 kfree(orig_data);
5143 return err;
5144 }
5145
5146 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
5147 {
5148 struct super_block *sb = dentry->d_sb;
5149 struct ext4_sb_info *sbi = EXT4_SB(sb);
5150 struct ext4_super_block *es = sbi->s_es;
5151 ext4_fsblk_t overhead = 0, resv_blocks;
5152 u64 fsid;
5153 s64 bfree;
5154 resv_blocks = EXT4_C2B(sbi, atomic64_read(&sbi->s_resv_clusters));
5155
5156 if (!test_opt(sb, MINIX_DF))
5157 overhead = sbi->s_overhead;
5158
5159 buf->f_type = EXT4_SUPER_MAGIC;
5160 buf->f_bsize = sb->s_blocksize;
5161 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, overhead);
5162 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
5163 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
5164 /* prevent underflow in case that few free space is available */
5165 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
5166 buf->f_bavail = buf->f_bfree -
5167 (ext4_r_blocks_count(es) + resv_blocks);
5168 if (buf->f_bfree < (ext4_r_blocks_count(es) + resv_blocks))
5169 buf->f_bavail = 0;
5170 buf->f_files = le32_to_cpu(es->s_inodes_count);
5171 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
5172 buf->f_namelen = EXT4_NAME_LEN;
5173 fsid = le64_to_cpup((void *)es->s_uuid) ^
5174 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
5175 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
5176 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
5177
5178 return 0;
5179 }
5180
5181 /* Helper function for writing quotas on sync - we need to start transaction
5182 * before quota file is locked for write. Otherwise the are possible deadlocks:
5183 * Process 1 Process 2
5184 * ext4_create() quota_sync()
5185 * jbd2_journal_start() write_dquot()
5186 * dquot_initialize() down(dqio_mutex)
5187 * down(dqio_mutex) jbd2_journal_start()
5188 *
5189 */
5190
5191 #ifdef CONFIG_QUOTA
5192
5193 static inline struct inode *dquot_to_inode(struct dquot *dquot)
5194 {
5195 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_id.type];
5196 }
5197
5198 static int ext4_write_dquot(struct dquot *dquot)
5199 {
5200 int ret, err;
5201 handle_t *handle;
5202 struct inode *inode;
5203
5204 inode = dquot_to_inode(dquot);
5205 handle = ext4_journal_start(inode, EXT4_HT_QUOTA,
5206 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
5207 if (IS_ERR(handle))
5208 return PTR_ERR(handle);
5209 ret = dquot_commit(dquot);
5210 err = ext4_journal_stop(handle);
5211 if (!ret)
5212 ret = err;
5213 return ret;
5214 }
5215
5216 static int ext4_acquire_dquot(struct dquot *dquot)
5217 {
5218 int ret, err;
5219 handle_t *handle;
5220
5221 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5222 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
5223 if (IS_ERR(handle))
5224 return PTR_ERR(handle);
5225 ret = dquot_acquire(dquot);
5226 err = ext4_journal_stop(handle);
5227 if (!ret)
5228 ret = err;
5229 return ret;
5230 }
5231
5232 static int ext4_release_dquot(struct dquot *dquot)
5233 {
5234 int ret, err;
5235 handle_t *handle;
5236
5237 handle = ext4_journal_start(dquot_to_inode(dquot), EXT4_HT_QUOTA,
5238 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
5239 if (IS_ERR(handle)) {
5240 /* Release dquot anyway to avoid endless cycle in dqput() */
5241 dquot_release(dquot);
5242 return PTR_ERR(handle);
5243 }
5244 ret = dquot_release(dquot);
5245 err = ext4_journal_stop(handle);
5246 if (!ret)
5247 ret = err;
5248 return ret;
5249 }
5250
5251 static int ext4_mark_dquot_dirty(struct dquot *dquot)
5252 {
5253 struct super_block *sb = dquot->dq_sb;
5254 struct ext4_sb_info *sbi = EXT4_SB(sb);
5255
5256 /* Are we journaling quotas? */
5257 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) ||
5258 sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
5259 dquot_mark_dquot_dirty(dquot);
5260 return ext4_write_dquot(dquot);
5261 } else {
5262 return dquot_mark_dquot_dirty(dquot);
5263 }
5264 }
5265
5266 static int ext4_write_info(struct super_block *sb, int type)
5267 {
5268 int ret, err;
5269 handle_t *handle;
5270
5271 /* Data block + inode block */
5272 handle = ext4_journal_start(d_inode(sb->s_root), EXT4_HT_QUOTA, 2);
5273 if (IS_ERR(handle))
5274 return PTR_ERR(handle);
5275 ret = dquot_commit_info(sb, type);
5276 err = ext4_journal_stop(handle);
5277 if (!ret)
5278 ret = err;
5279 return ret;
5280 }
5281
5282 /*
5283 * Turn on quotas during mount time - we need to find
5284 * the quota file and such...
5285 */
5286 static int ext4_quota_on_mount(struct super_block *sb, int type)
5287 {
5288 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
5289 EXT4_SB(sb)->s_jquota_fmt, type);
5290 }
5291
5292 /*
5293 * Standard function to be called on quota_on
5294 */
5295 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
5296 struct path *path)
5297 {
5298 int err;
5299
5300 if (!test_opt(sb, QUOTA))
5301 return -EINVAL;
5302
5303 /* Quotafile not on the same filesystem? */
5304 if (path->dentry->d_sb != sb)
5305 return -EXDEV;
5306 /* Journaling quota? */
5307 if (EXT4_SB(sb)->s_qf_names[type]) {
5308 /* Quotafile not in fs root? */
5309 if (path->dentry->d_parent != sb->s_root)
5310 ext4_msg(sb, KERN_WARNING,
5311 "Quota file not on filesystem root. "
5312 "Journaled quota will not work");
5313 }
5314
5315 /*
5316 * When we journal data on quota file, we have to flush journal to see
5317 * all updates to the file when we bypass pagecache...
5318 */
5319 if (EXT4_SB(sb)->s_journal &&
5320 ext4_should_journal_data(d_inode(path->dentry))) {
5321 /*
5322 * We don't need to lock updates but journal_flush() could
5323 * otherwise be livelocked...
5324 */
5325 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
5326 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
5327 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
5328 if (err)
5329 return err;
5330 }
5331
5332 return dquot_quota_on(sb, type, format_id, path);
5333 }
5334
5335 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
5336 unsigned int flags)
5337 {
5338 int err;
5339 struct inode *qf_inode;
5340 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5341 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5342 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5343 };
5344
5345 BUG_ON(!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA));
5346
5347 if (!qf_inums[type])
5348 return -EPERM;
5349
5350 qf_inode = ext4_iget(sb, qf_inums[type]);
5351 if (IS_ERR(qf_inode)) {
5352 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
5353 return PTR_ERR(qf_inode);
5354 }
5355
5356 /* Don't account quota for quota files to avoid recursion */
5357 qf_inode->i_flags |= S_NOQUOTA;
5358 err = dquot_enable(qf_inode, type, format_id, flags);
5359 iput(qf_inode);
5360
5361 return err;
5362 }
5363
5364 /* Enable usage tracking for all quota types. */
5365 static int ext4_enable_quotas(struct super_block *sb)
5366 {
5367 int type, err = 0;
5368 unsigned long qf_inums[EXT4_MAXQUOTAS] = {
5369 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
5370 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
5371 };
5372
5373 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
5374 for (type = 0; type < EXT4_MAXQUOTAS; type++) {
5375 if (qf_inums[type]) {
5376 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
5377 DQUOT_USAGE_ENABLED);
5378 if (err) {
5379 ext4_warning(sb,
5380 "Failed to enable quota tracking "
5381 "(type=%d, err=%d). Please run "
5382 "e2fsck to fix.", type, err);
5383 return err;
5384 }
5385 }
5386 }
5387 return 0;
5388 }
5389
5390 static int ext4_quota_off(struct super_block *sb, int type)
5391 {
5392 struct inode *inode = sb_dqopt(sb)->files[type];
5393 handle_t *handle;
5394
5395 /* Force all delayed allocation blocks to be allocated.
5396 * Caller already holds s_umount sem */
5397 if (test_opt(sb, DELALLOC))
5398 sync_filesystem(sb);
5399
5400 if (!inode)
5401 goto out;
5402
5403 /* Update modification times of quota files when userspace can
5404 * start looking at them */
5405 handle = ext4_journal_start(inode, EXT4_HT_QUOTA, 1);
5406 if (IS_ERR(handle))
5407 goto out;
5408 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
5409 ext4_mark_inode_dirty(handle, inode);
5410 ext4_journal_stop(handle);
5411
5412 out:
5413 return dquot_quota_off(sb, type);
5414 }
5415
5416 /* Read data from quotafile - avoid pagecache and such because we cannot afford
5417 * acquiring the locks... As quota files are never truncated and quota code
5418 * itself serializes the operations (and no one else should touch the files)
5419 * we don't have to be afraid of races */
5420 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
5421 size_t len, loff_t off)
5422 {
5423 struct inode *inode = sb_dqopt(sb)->files[type];
5424 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5425 int offset = off & (sb->s_blocksize - 1);
5426 int tocopy;
5427 size_t toread;
5428 struct buffer_head *bh;
5429 loff_t i_size = i_size_read(inode);
5430
5431 if (off > i_size)
5432 return 0;
5433 if (off+len > i_size)
5434 len = i_size-off;
5435 toread = len;
5436 while (toread > 0) {
5437 tocopy = sb->s_blocksize - offset < toread ?
5438 sb->s_blocksize - offset : toread;
5439 bh = ext4_bread(NULL, inode, blk, 0);
5440 if (IS_ERR(bh))
5441 return PTR_ERR(bh);
5442 if (!bh) /* A hole? */
5443 memset(data, 0, tocopy);
5444 else
5445 memcpy(data, bh->b_data+offset, tocopy);
5446 brelse(bh);
5447 offset = 0;
5448 toread -= tocopy;
5449 data += tocopy;
5450 blk++;
5451 }
5452 return len;
5453 }
5454
5455 /* Write to quotafile (we know the transaction is already started and has
5456 * enough credits) */
5457 static ssize_t ext4_quota_write(struct super_block *sb, int type,
5458 const char *data, size_t len, loff_t off)
5459 {
5460 struct inode *inode = sb_dqopt(sb)->files[type];
5461 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5462 int err, offset = off & (sb->s_blocksize - 1);
5463 int retries = 0;
5464 struct buffer_head *bh;
5465 handle_t *handle = journal_current_handle();
5466
5467 if (EXT4_SB(sb)->s_journal && !handle) {
5468 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5469 " cancelled because transaction is not started",
5470 (unsigned long long)off, (unsigned long long)len);
5471 return -EIO;
5472 }
5473 /*
5474 * Since we account only one data block in transaction credits,
5475 * then it is impossible to cross a block boundary.
5476 */
5477 if (sb->s_blocksize - offset < len) {
5478 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5479 " cancelled because not block aligned",
5480 (unsigned long long)off, (unsigned long long)len);
5481 return -EIO;
5482 }
5483
5484 do {
5485 bh = ext4_bread(handle, inode, blk,
5486 EXT4_GET_BLOCKS_CREATE |
5487 EXT4_GET_BLOCKS_METADATA_NOFAIL);
5488 } while (IS_ERR(bh) && (PTR_ERR(bh) == -ENOSPC) &&
5489 ext4_should_retry_alloc(inode->i_sb, &retries));
5490 if (IS_ERR(bh))
5491 return PTR_ERR(bh);
5492 if (!bh)
5493 goto out;
5494 BUFFER_TRACE(bh, "get write access");
5495 err = ext4_journal_get_write_access(handle, bh);
5496 if (err) {
5497 brelse(bh);
5498 return err;
5499 }
5500 lock_buffer(bh);
5501 memcpy(bh->b_data+offset, data, len);
5502 flush_dcache_page(bh->b_page);
5503 unlock_buffer(bh);
5504 err = ext4_handle_dirty_metadata(handle, NULL, bh);
5505 brelse(bh);
5506 out:
5507 if (inode->i_size < off + len) {
5508 i_size_write(inode, off + len);
5509 EXT4_I(inode)->i_disksize = inode->i_size;
5510 ext4_mark_inode_dirty(handle, inode);
5511 }
5512 return len;
5513 }
5514
5515 #endif
5516
5517 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5518 const char *dev_name, void *data)
5519 {
5520 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
5521 }
5522
5523 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5524 static inline void register_as_ext2(void)
5525 {
5526 int err = register_filesystem(&ext2_fs_type);
5527 if (err)
5528 printk(KERN_WARNING
5529 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5530 }
5531
5532 static inline void unregister_as_ext2(void)
5533 {
5534 unregister_filesystem(&ext2_fs_type);
5535 }
5536
5537 static inline int ext2_feature_set_ok(struct super_block *sb)
5538 {
5539 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
5540 return 0;
5541 if (sb->s_flags & MS_RDONLY)
5542 return 1;
5543 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
5544 return 0;
5545 return 1;
5546 }
5547 #else
5548 static inline void register_as_ext2(void) { }
5549 static inline void unregister_as_ext2(void) { }
5550 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
5551 #endif
5552
5553 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5554 static inline void register_as_ext3(void)
5555 {
5556 int err = register_filesystem(&ext3_fs_type);
5557 if (err)
5558 printk(KERN_WARNING
5559 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5560 }
5561
5562 static inline void unregister_as_ext3(void)
5563 {
5564 unregister_filesystem(&ext3_fs_type);
5565 }
5566
5567 static inline int ext3_feature_set_ok(struct super_block *sb)
5568 {
5569 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
5570 return 0;
5571 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
5572 return 0;
5573 if (sb->s_flags & MS_RDONLY)
5574 return 1;
5575 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
5576 return 0;
5577 return 1;
5578 }
5579 #else
5580 static inline void register_as_ext3(void) { }
5581 static inline void unregister_as_ext3(void) { }
5582 static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
5583 #endif
5584
5585 static struct file_system_type ext4_fs_type = {
5586 .owner = THIS_MODULE,
5587 .name = "ext4",
5588 .mount = ext4_mount,
5589 .kill_sb = kill_block_super,
5590 .fs_flags = FS_REQUIRES_DEV,
5591 };
5592 MODULE_ALIAS_FS("ext4");
5593
5594 static int __init ext4_init_feat_adverts(void)
5595 {
5596 struct ext4_features *ef;
5597 int ret = -ENOMEM;
5598
5599 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
5600 if (!ef)
5601 goto out;
5602
5603 ef->f_kobj.kset = ext4_kset;
5604 init_completion(&ef->f_kobj_unregister);
5605 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
5606 "features");
5607 if (ret) {
5608 kfree(ef);
5609 goto out;
5610 }
5611
5612 ext4_feat = ef;
5613 ret = 0;
5614 out:
5615 return ret;
5616 }
5617
5618 static void ext4_exit_feat_adverts(void)
5619 {
5620 kobject_put(&ext4_feat->f_kobj);
5621 wait_for_completion(&ext4_feat->f_kobj_unregister);
5622 kfree(ext4_feat);
5623 }
5624
5625 /* Shared across all ext4 file systems */
5626 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5627 struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5628
5629 static int __init ext4_init_fs(void)
5630 {
5631 int i, err;
5632
5633 ext4_li_info = NULL;
5634 mutex_init(&ext4_li_mtx);
5635
5636 /* Build-time check for flags consistency */
5637 ext4_check_flag_values();
5638
5639 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5640 mutex_init(&ext4__aio_mutex[i]);
5641 init_waitqueue_head(&ext4__ioend_wq[i]);
5642 }
5643
5644 err = ext4_init_es();
5645 if (err)
5646 return err;
5647
5648 err = ext4_init_pageio();
5649 if (err)
5650 goto out7;
5651
5652 err = ext4_init_system_zone();
5653 if (err)
5654 goto out6;
5655 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
5656 if (!ext4_kset) {
5657 err = -ENOMEM;
5658 goto out5;
5659 }
5660 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
5661
5662 err = ext4_init_feat_adverts();
5663 if (err)
5664 goto out4;
5665
5666 err = ext4_init_mballoc();
5667 if (err)
5668 goto out2;
5669 else
5670 ext4_mballoc_ready = 1;
5671 err = init_inodecache();
5672 if (err)
5673 goto out1;
5674 register_as_ext3();
5675 register_as_ext2();
5676 err = register_filesystem(&ext4_fs_type);
5677 if (err)
5678 goto out;
5679
5680 return 0;
5681 out:
5682 unregister_as_ext2();
5683 unregister_as_ext3();
5684 destroy_inodecache();
5685 out1:
5686 ext4_mballoc_ready = 0;
5687 ext4_exit_mballoc();
5688 out2:
5689 ext4_exit_feat_adverts();
5690 out4:
5691 if (ext4_proc_root)
5692 remove_proc_entry("fs/ext4", NULL);
5693 kset_unregister(ext4_kset);
5694 out5:
5695 ext4_exit_system_zone();
5696 out6:
5697 ext4_exit_pageio();
5698 out7:
5699 ext4_exit_es();
5700
5701 return err;
5702 }
5703
5704 static void __exit ext4_exit_fs(void)
5705 {
5706 ext4_exit_crypto();
5707 ext4_destroy_lazyinit_thread();
5708 unregister_as_ext2();
5709 unregister_as_ext3();
5710 unregister_filesystem(&ext4_fs_type);
5711 destroy_inodecache();
5712 ext4_exit_mballoc();
5713 ext4_exit_feat_adverts();
5714 remove_proc_entry("fs/ext4", NULL);
5715 kset_unregister(ext4_kset);
5716 ext4_exit_system_zone();
5717 ext4_exit_pageio();
5718 ext4_exit_es();
5719 }
5720
5721 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5722 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5723 MODULE_LICENSE("GPL");
5724 module_init(ext4_init_fs)
5725 module_exit(ext4_exit_fs)
Linux with added FPC-III support