?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Linux 3.6.10
[linux/fpc-iii.git] / fs / ext4 / super.c
blob2b5fb603b2b6a1da2fc4dd0029891ff520fb42f3
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/jbd2.h>
25 #include <linux/slab.h>
26 #include <linux/init.h>
27 #include <linux/blkdev.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"
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
63 static int ext4_load_journal(struct super_block *, struct ext4_super_block *,
64 unsigned long journal_devnum);
65 static int ext4_show_options(struct seq_file *seq, struct dentry *root);
66 static int ext4_commit_super(struct super_block *sb, int sync);
67 static void ext4_mark_recovery_complete(struct super_block *sb,
68 struct ext4_super_block *es);
69 static void ext4_clear_journal_err(struct super_block *sb,
70 struct ext4_super_block *es);
71 static int ext4_sync_fs(struct super_block *sb, int wait);
72 static const char *ext4_decode_error(struct super_block *sb, int errno,
73 char nbuf[16]);
74 static int ext4_remount(struct super_block *sb, int *flags, char *data);
75 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf);
76 static int ext4_unfreeze(struct super_block *sb);
77 static int ext4_freeze(struct super_block *sb);
78 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
79 const char *dev_name, void *data);
80 static inline int ext2_feature_set_ok(struct super_block *sb);
81 static inline int ext3_feature_set_ok(struct super_block *sb);
82 static int ext4_feature_set_ok(struct super_block *sb, int readonly);
83 static void ext4_destroy_lazyinit_thread(void);
84 static void ext4_unregister_li_request(struct super_block *sb);
85 static void ext4_clear_request_list(void);
86
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 #define IS_EXT2_SB(sb) ((sb)->s_bdev->bd_holder == &ext2_fs_type)
96 #else
97 #define IS_EXT2_SB(sb) (0)
98 #endif
99
100
101 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
102 static struct file_system_type ext3_fs_type = {
103 .owner = THIS_MODULE,
104 .name = "ext3",
105 .mount = ext4_mount,
106 .kill_sb = kill_block_super,
107 .fs_flags = FS_REQUIRES_DEV,
108 };
109 #define IS_EXT3_SB(sb) ((sb)->s_bdev->bd_holder == &ext3_fs_type)
110 #else
111 #define IS_EXT3_SB(sb) (0)
112 #endif
113
114 static int ext4_verify_csum_type(struct super_block *sb,
115 struct ext4_super_block *es)
116 {
117 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
118 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
119 return 1;
120
121 return es->s_checksum_type == EXT4_CRC32C_CHKSUM;
122 }
123
124 static __le32 ext4_superblock_csum(struct super_block *sb,
125 struct ext4_super_block *es)
126 {
127 struct ext4_sb_info *sbi = EXT4_SB(sb);
128 int offset = offsetof(struct ext4_super_block, s_checksum);
129 __u32 csum;
130
131 csum = ext4_chksum(sbi, ~0, (char *)es, offset);
132
133 return cpu_to_le32(csum);
134 }
135
136 int ext4_superblock_csum_verify(struct super_block *sb,
137 struct ext4_super_block *es)
138 {
139 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
140 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
141 return 1;
142
143 return es->s_checksum == ext4_superblock_csum(sb, es);
144 }
145
146 void ext4_superblock_csum_set(struct super_block *sb)
147 {
148 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
149
150 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb,
151 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
152 return;
153
154 es->s_checksum = ext4_superblock_csum(sb, es);
155 }
156
157 void *ext4_kvmalloc(size_t size, gfp_t flags)
158 {
159 void *ret;
160
161 ret = kmalloc(size, flags);
162 if (!ret)
163 ret = __vmalloc(size, flags, PAGE_KERNEL);
164 return ret;
165 }
166
167 void *ext4_kvzalloc(size_t size, gfp_t flags)
168 {
169 void *ret;
170
171 ret = kzalloc(size, flags);
172 if (!ret)
173 ret = __vmalloc(size, flags | __GFP_ZERO, PAGE_KERNEL);
174 return ret;
175 }
176
177 void ext4_kvfree(void *ptr)
178 {
179 if (is_vmalloc_addr(ptr))
180 vfree(ptr);
181 else
182 kfree(ptr);
183
184 }
185
186 ext4_fsblk_t ext4_block_bitmap(struct super_block *sb,
187 struct ext4_group_desc *bg)
188 {
189 return le32_to_cpu(bg->bg_block_bitmap_lo) |
190 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
191 (ext4_fsblk_t)le32_to_cpu(bg->bg_block_bitmap_hi) << 32 : 0);
192 }
193
194 ext4_fsblk_t ext4_inode_bitmap(struct super_block *sb,
195 struct ext4_group_desc *bg)
196 {
197 return le32_to_cpu(bg->bg_inode_bitmap_lo) |
198 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
199 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_bitmap_hi) << 32 : 0);
200 }
201
202 ext4_fsblk_t ext4_inode_table(struct super_block *sb,
203 struct ext4_group_desc *bg)
204 {
205 return le32_to_cpu(bg->bg_inode_table_lo) |
206 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
207 (ext4_fsblk_t)le32_to_cpu(bg->bg_inode_table_hi) << 32 : 0);
208 }
209
210 __u32 ext4_free_group_clusters(struct super_block *sb,
211 struct ext4_group_desc *bg)
212 {
213 return le16_to_cpu(bg->bg_free_blocks_count_lo) |
214 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
215 (__u32)le16_to_cpu(bg->bg_free_blocks_count_hi) << 16 : 0);
216 }
217
218 __u32 ext4_free_inodes_count(struct super_block *sb,
219 struct ext4_group_desc *bg)
220 {
221 return le16_to_cpu(bg->bg_free_inodes_count_lo) |
222 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
223 (__u32)le16_to_cpu(bg->bg_free_inodes_count_hi) << 16 : 0);
224 }
225
226 __u32 ext4_used_dirs_count(struct super_block *sb,
227 struct ext4_group_desc *bg)
228 {
229 return le16_to_cpu(bg->bg_used_dirs_count_lo) |
230 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
231 (__u32)le16_to_cpu(bg->bg_used_dirs_count_hi) << 16 : 0);
232 }
233
234 __u32 ext4_itable_unused_count(struct super_block *sb,
235 struct ext4_group_desc *bg)
236 {
237 return le16_to_cpu(bg->bg_itable_unused_lo) |
238 (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT ?
239 (__u32)le16_to_cpu(bg->bg_itable_unused_hi) << 16 : 0);
240 }
241
242 void ext4_block_bitmap_set(struct super_block *sb,
243 struct ext4_group_desc *bg, ext4_fsblk_t blk)
244 {
245 bg->bg_block_bitmap_lo = cpu_to_le32((u32)blk);
246 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
247 bg->bg_block_bitmap_hi = cpu_to_le32(blk >> 32);
248 }
249
250 void ext4_inode_bitmap_set(struct super_block *sb,
251 struct ext4_group_desc *bg, ext4_fsblk_t blk)
252 {
253 bg->bg_inode_bitmap_lo = cpu_to_le32((u32)blk);
254 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
255 bg->bg_inode_bitmap_hi = cpu_to_le32(blk >> 32);
256 }
257
258 void ext4_inode_table_set(struct super_block *sb,
259 struct ext4_group_desc *bg, ext4_fsblk_t blk)
260 {
261 bg->bg_inode_table_lo = cpu_to_le32((u32)blk);
262 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
263 bg->bg_inode_table_hi = cpu_to_le32(blk >> 32);
264 }
265
266 void ext4_free_group_clusters_set(struct super_block *sb,
267 struct ext4_group_desc *bg, __u32 count)
268 {
269 bg->bg_free_blocks_count_lo = cpu_to_le16((__u16)count);
270 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
271 bg->bg_free_blocks_count_hi = cpu_to_le16(count >> 16);
272 }
273
274 void ext4_free_inodes_set(struct super_block *sb,
275 struct ext4_group_desc *bg, __u32 count)
276 {
277 bg->bg_free_inodes_count_lo = cpu_to_le16((__u16)count);
278 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
279 bg->bg_free_inodes_count_hi = cpu_to_le16(count >> 16);
280 }
281
282 void ext4_used_dirs_set(struct super_block *sb,
283 struct ext4_group_desc *bg, __u32 count)
284 {
285 bg->bg_used_dirs_count_lo = cpu_to_le16((__u16)count);
286 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
287 bg->bg_used_dirs_count_hi = cpu_to_le16(count >> 16);
288 }
289
290 void ext4_itable_unused_set(struct super_block *sb,
291 struct ext4_group_desc *bg, __u32 count)
292 {
293 bg->bg_itable_unused_lo = cpu_to_le16((__u16)count);
294 if (EXT4_DESC_SIZE(sb) >= EXT4_MIN_DESC_SIZE_64BIT)
295 bg->bg_itable_unused_hi = cpu_to_le16(count >> 16);
296 }
297
298
299 /* Just increment the non-pointer handle value */
300 static handle_t *ext4_get_nojournal(void)
301 {
302 handle_t *handle = current->journal_info;
303 unsigned long ref_cnt = (unsigned long)handle;
304
305 BUG_ON(ref_cnt >= EXT4_NOJOURNAL_MAX_REF_COUNT);
306
307 ref_cnt++;
308 handle = (handle_t *)ref_cnt;
309
310 current->journal_info = handle;
311 return handle;
312 }
313
314
315 /* Decrement the non-pointer handle value */
316 static void ext4_put_nojournal(handle_t *handle)
317 {
318 unsigned long ref_cnt = (unsigned long)handle;
319
320 BUG_ON(ref_cnt == 0);
321
322 ref_cnt--;
323 handle = (handle_t *)ref_cnt;
324
325 current->journal_info = handle;
326 }
327
328 /*
329 * Wrappers for jbd2_journal_start/end.
330 */
331 handle_t *ext4_journal_start_sb(struct super_block *sb, int nblocks)
332 {
333 journal_t *journal;
334
335 trace_ext4_journal_start(sb, nblocks, _RET_IP_);
336 if (sb->s_flags & MS_RDONLY)
337 return ERR_PTR(-EROFS);
338
339 WARN_ON(sb->s_writers.frozen == SB_FREEZE_COMPLETE);
340 journal = EXT4_SB(sb)->s_journal;
341 if (!journal)
342 return ext4_get_nojournal();
343 /*
344 * Special case here: if the journal has aborted behind our
345 * backs (eg. EIO in the commit thread), then we still need to
346 * take the FS itself readonly cleanly.
347 */
348 if (is_journal_aborted(journal)) {
349 ext4_abort(sb, "Detected aborted journal");
350 return ERR_PTR(-EROFS);
351 }
352 return jbd2_journal_start(journal, nblocks);
353 }
354
355 int __ext4_journal_stop(const char *where, unsigned int line, handle_t *handle)
356 {
357 struct super_block *sb;
358 int err;
359 int rc;
360
361 if (!ext4_handle_valid(handle)) {
362 ext4_put_nojournal(handle);
363 return 0;
364 }
365 sb = handle->h_transaction->t_journal->j_private;
366 err = handle->h_err;
367 rc = jbd2_journal_stop(handle);
368
369 if (!err)
370 err = rc;
371 if (err)
372 __ext4_std_error(sb, where, line, err);
373 return err;
374 }
375
376 void ext4_journal_abort_handle(const char *caller, unsigned int line,
377 const char *err_fn, struct buffer_head *bh,
378 handle_t *handle, int err)
379 {
380 char nbuf[16];
381 const char *errstr = ext4_decode_error(NULL, err, nbuf);
382
383 BUG_ON(!ext4_handle_valid(handle));
384
385 if (bh)
386 BUFFER_TRACE(bh, "abort");
387
388 if (!handle->h_err)
389 handle->h_err = err;
390
391 if (is_handle_aborted(handle))
392 return;
393
394 printk(KERN_ERR "EXT4-fs: %s:%d: aborting transaction: %s in %s\n",
395 caller, line, errstr, err_fn);
396
397 jbd2_journal_abort_handle(handle);
398 }
399
400 static void __save_error_info(struct super_block *sb, const char *func,
401 unsigned int line)
402 {
403 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
404
405 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
406 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
407 es->s_last_error_time = cpu_to_le32(get_seconds());
408 strncpy(es->s_last_error_func, func, sizeof(es->s_last_error_func));
409 es->s_last_error_line = cpu_to_le32(line);
410 if (!es->s_first_error_time) {
411 es->s_first_error_time = es->s_last_error_time;
412 strncpy(es->s_first_error_func, func,
413 sizeof(es->s_first_error_func));
414 es->s_first_error_line = cpu_to_le32(line);
415 es->s_first_error_ino = es->s_last_error_ino;
416 es->s_first_error_block = es->s_last_error_block;
417 }
418 /*
419 * Start the daily error reporting function if it hasn't been
420 * started already
421 */
422 if (!es->s_error_count)
423 mod_timer(&EXT4_SB(sb)->s_err_report, jiffies + 24*60*60*HZ);
424 es->s_error_count = cpu_to_le32(le32_to_cpu(es->s_error_count) + 1);
425 }
426
427 static void save_error_info(struct super_block *sb, const char *func,
428 unsigned int line)
429 {
430 __save_error_info(sb, func, line);
431 ext4_commit_super(sb, 1);
432 }
433
434 /*
435 * The del_gendisk() function uninitializes the disk-specific data
436 * structures, including the bdi structure, without telling anyone
437 * else. Once this happens, any attempt to call mark_buffer_dirty()
438 * (for example, by ext4_commit_super), will cause a kernel OOPS.
439 * This is a kludge to prevent these oops until we can put in a proper
440 * hook in del_gendisk() to inform the VFS and file system layers.
441 */
442 static int block_device_ejected(struct super_block *sb)
443 {
444 struct inode *bd_inode = sb->s_bdev->bd_inode;
445 struct backing_dev_info *bdi = bd_inode->i_mapping->backing_dev_info;
446
447 return bdi->dev == NULL;
448 }
449
450 static void ext4_journal_commit_callback(journal_t *journal, transaction_t *txn)
451 {
452 struct super_block *sb = journal->j_private;
453 struct ext4_sb_info *sbi = EXT4_SB(sb);
454 int error = is_journal_aborted(journal);
455 struct ext4_journal_cb_entry *jce, *tmp;
456
457 spin_lock(&sbi->s_md_lock);
458 list_for_each_entry_safe(jce, tmp, &txn->t_private_list, jce_list) {
459 list_del_init(&jce->jce_list);
460 spin_unlock(&sbi->s_md_lock);
461 jce->jce_func(sb, jce, error);
462 spin_lock(&sbi->s_md_lock);
463 }
464 spin_unlock(&sbi->s_md_lock);
465 }
466
467 /* Deal with the reporting of failure conditions on a filesystem such as
468 * inconsistencies detected or read IO failures.
469 *
470 * On ext2, we can store the error state of the filesystem in the
471 * superblock. That is not possible on ext4, because we may have other
472 * write ordering constraints on the superblock which prevent us from
473 * writing it out straight away; and given that the journal is about to
474 * be aborted, we can't rely on the current, or future, transactions to
475 * write out the superblock safely.
476 *
477 * We'll just use the jbd2_journal_abort() error code to record an error in
478 * the journal instead. On recovery, the journal will complain about
479 * that error until we've noted it down and cleared it.
480 */
481
482 static void ext4_handle_error(struct super_block *sb)
483 {
484 if (sb->s_flags & MS_RDONLY)
485 return;
486
487 if (!test_opt(sb, ERRORS_CONT)) {
488 journal_t *journal = EXT4_SB(sb)->s_journal;
489
490 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
491 if (journal)
492 jbd2_journal_abort(journal, -EIO);
493 }
494 if (test_opt(sb, ERRORS_RO)) {
495 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
496 sb->s_flags |= MS_RDONLY;
497 }
498 if (test_opt(sb, ERRORS_PANIC))
499 panic("EXT4-fs (device %s): panic forced after error\n",
500 sb->s_id);
501 }
502
503 void __ext4_error(struct super_block *sb, const char *function,
504 unsigned int line, const char *fmt, ...)
505 {
506 struct va_format vaf;
507 va_list args;
508
509 va_start(args, fmt);
510 vaf.fmt = fmt;
511 vaf.va = &args;
512 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: comm %s: %pV\n",
513 sb->s_id, function, line, current->comm, &vaf);
514 va_end(args);
515 save_error_info(sb, function, line);
516
517 ext4_handle_error(sb);
518 }
519
520 void ext4_error_inode(struct inode *inode, const char *function,
521 unsigned int line, ext4_fsblk_t block,
522 const char *fmt, ...)
523 {
524 va_list args;
525 struct va_format vaf;
526 struct ext4_super_block *es = EXT4_SB(inode->i_sb)->s_es;
527
528 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
529 es->s_last_error_block = cpu_to_le64(block);
530 save_error_info(inode->i_sb, function, line);
531 va_start(args, fmt);
532 vaf.fmt = fmt;
533 vaf.va = &args;
534 if (block)
535 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
536 "inode #%lu: block %llu: comm %s: %pV\n",
537 inode->i_sb->s_id, function, line, inode->i_ino,
538 block, current->comm, &vaf);
539 else
540 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: "
541 "inode #%lu: comm %s: %pV\n",
542 inode->i_sb->s_id, function, line, inode->i_ino,
543 current->comm, &vaf);
544 va_end(args);
545
546 ext4_handle_error(inode->i_sb);
547 }
548
549 void ext4_error_file(struct file *file, const char *function,
550 unsigned int line, ext4_fsblk_t block,
551 const char *fmt, ...)
552 {
553 va_list args;
554 struct va_format vaf;
555 struct ext4_super_block *es;
556 struct inode *inode = file->f_dentry->d_inode;
557 char pathname[80], *path;
558
559 es = EXT4_SB(inode->i_sb)->s_es;
560 es->s_last_error_ino = cpu_to_le32(inode->i_ino);
561 save_error_info(inode->i_sb, function, line);
562 path = d_path(&(file->f_path), pathname, sizeof(pathname));
563 if (IS_ERR(path))
564 path = "(unknown)";
565 va_start(args, fmt);
566 vaf.fmt = fmt;
567 vaf.va = &args;
568 if (block)
569 printk(KERN_CRIT
570 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
571 "block %llu: comm %s: path %s: %pV\n",
572 inode->i_sb->s_id, function, line, inode->i_ino,
573 block, current->comm, path, &vaf);
574 else
575 printk(KERN_CRIT
576 "EXT4-fs error (device %s): %s:%d: inode #%lu: "
577 "comm %s: path %s: %pV\n",
578 inode->i_sb->s_id, function, line, inode->i_ino,
579 current->comm, path, &vaf);
580 va_end(args);
581
582 ext4_handle_error(inode->i_sb);
583 }
584
585 static const char *ext4_decode_error(struct super_block *sb, int errno,
586 char nbuf[16])
587 {
588 char *errstr = NULL;
589
590 switch (errno) {
591 case -EIO:
592 errstr = "IO failure";
593 break;
594 case -ENOMEM:
595 errstr = "Out of memory";
596 break;
597 case -EROFS:
598 if (!sb || (EXT4_SB(sb)->s_journal &&
599 EXT4_SB(sb)->s_journal->j_flags & JBD2_ABORT))
600 errstr = "Journal has aborted";
601 else
602 errstr = "Readonly filesystem";
603 break;
604 default:
605 /* If the caller passed in an extra buffer for unknown
606 * errors, textualise them now. Else we just return
607 * NULL. */
608 if (nbuf) {
609 /* Check for truncated error codes... */
610 if (snprintf(nbuf, 16, "error %d", -errno) >= 0)
611 errstr = nbuf;
612 }
613 break;
614 }
615
616 return errstr;
617 }
618
619 /* __ext4_std_error decodes expected errors from journaling functions
620 * automatically and invokes the appropriate error response. */
621
622 void __ext4_std_error(struct super_block *sb, const char *function,
623 unsigned int line, int errno)
624 {
625 char nbuf[16];
626 const char *errstr;
627
628 /* Special case: if the error is EROFS, and we're not already
629 * inside a transaction, then there's really no point in logging
630 * an error. */
631 if (errno == -EROFS && journal_current_handle() == NULL &&
632 (sb->s_flags & MS_RDONLY))
633 return;
634
635 errstr = ext4_decode_error(sb, errno, nbuf);
636 printk(KERN_CRIT "EXT4-fs error (device %s) in %s:%d: %s\n",
637 sb->s_id, function, line, errstr);
638 save_error_info(sb, function, line);
639
640 ext4_handle_error(sb);
641 }
642
643 /*
644 * ext4_abort is a much stronger failure handler than ext4_error. The
645 * abort function may be used to deal with unrecoverable failures such
646 * as journal IO errors or ENOMEM at a critical moment in log management.
647 *
648 * We unconditionally force the filesystem into an ABORT|READONLY state,
649 * unless the error response on the fs has been set to panic in which
650 * case we take the easy way out and panic immediately.
651 */
652
653 void __ext4_abort(struct super_block *sb, const char *function,
654 unsigned int line, const char *fmt, ...)
655 {
656 va_list args;
657
658 save_error_info(sb, function, line);
659 va_start(args, fmt);
660 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: ", sb->s_id,
661 function, line);
662 vprintk(fmt, args);
663 printk("\n");
664 va_end(args);
665
666 if ((sb->s_flags & MS_RDONLY) == 0) {
667 ext4_msg(sb, KERN_CRIT, "Remounting filesystem read-only");
668 sb->s_flags |= MS_RDONLY;
669 EXT4_SB(sb)->s_mount_flags |= EXT4_MF_FS_ABORTED;
670 if (EXT4_SB(sb)->s_journal)
671 jbd2_journal_abort(EXT4_SB(sb)->s_journal, -EIO);
672 save_error_info(sb, function, line);
673 }
674 if (test_opt(sb, ERRORS_PANIC))
675 panic("EXT4-fs panic from previous error\n");
676 }
677
678 void ext4_msg(struct super_block *sb, const char *prefix, const char *fmt, ...)
679 {
680 struct va_format vaf;
681 va_list args;
682
683 va_start(args, fmt);
684 vaf.fmt = fmt;
685 vaf.va = &args;
686 printk("%sEXT4-fs (%s): %pV\n", prefix, sb->s_id, &vaf);
687 va_end(args);
688 }
689
690 void __ext4_warning(struct super_block *sb, const char *function,
691 unsigned int line, const char *fmt, ...)
692 {
693 struct va_format vaf;
694 va_list args;
695
696 va_start(args, fmt);
697 vaf.fmt = fmt;
698 vaf.va = &args;
699 printk(KERN_WARNING "EXT4-fs warning (device %s): %s:%d: %pV\n",
700 sb->s_id, function, line, &vaf);
701 va_end(args);
702 }
703
704 void __ext4_grp_locked_error(const char *function, unsigned int line,
705 struct super_block *sb, ext4_group_t grp,
706 unsigned long ino, ext4_fsblk_t block,
707 const char *fmt, ...)
708 __releases(bitlock)
709 __acquires(bitlock)
710 {
711 struct va_format vaf;
712 va_list args;
713 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
714
715 es->s_last_error_ino = cpu_to_le32(ino);
716 es->s_last_error_block = cpu_to_le64(block);
717 __save_error_info(sb, function, line);
718
719 va_start(args, fmt);
720
721 vaf.fmt = fmt;
722 vaf.va = &args;
723 printk(KERN_CRIT "EXT4-fs error (device %s): %s:%d: group %u, ",
724 sb->s_id, function, line, grp);
725 if (ino)
726 printk(KERN_CONT "inode %lu: ", ino);
727 if (block)
728 printk(KERN_CONT "block %llu:", (unsigned long long) block);
729 printk(KERN_CONT "%pV\n", &vaf);
730 va_end(args);
731
732 if (test_opt(sb, ERRORS_CONT)) {
733 ext4_commit_super(sb, 0);
734 return;
735 }
736
737 ext4_unlock_group(sb, grp);
738 ext4_handle_error(sb);
739 /*
740 * We only get here in the ERRORS_RO case; relocking the group
741 * may be dangerous, but nothing bad will happen since the
742 * filesystem will have already been marked read/only and the
743 * journal has been aborted. We return 1 as a hint to callers
744 * who might what to use the return value from
745 * ext4_grp_locked_error() to distinguish between the
746 * ERRORS_CONT and ERRORS_RO case, and perhaps return more
747 * aggressively from the ext4 function in question, with a
748 * more appropriate error code.
749 */
750 ext4_lock_group(sb, grp);
751 return;
752 }
753
754 void ext4_update_dynamic_rev(struct super_block *sb)
755 {
756 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
757
758 if (le32_to_cpu(es->s_rev_level) > EXT4_GOOD_OLD_REV)
759 return;
760
761 ext4_warning(sb,
762 "updating to rev %d because of new feature flag, "
763 "running e2fsck is recommended",
764 EXT4_DYNAMIC_REV);
765
766 es->s_first_ino = cpu_to_le32(EXT4_GOOD_OLD_FIRST_INO);
767 es->s_inode_size = cpu_to_le16(EXT4_GOOD_OLD_INODE_SIZE);
768 es->s_rev_level = cpu_to_le32(EXT4_DYNAMIC_REV);
769 /* leave es->s_feature_*compat flags alone */
770 /* es->s_uuid will be set by e2fsck if empty */
771
772 /*
773 * The rest of the superblock fields should be zero, and if not it
774 * means they are likely already in use, so leave them alone. We
775 * can leave it up to e2fsck to clean up any inconsistencies there.
776 */
777 }
778
779 /*
780 * Open the external journal device
781 */
782 static struct block_device *ext4_blkdev_get(dev_t dev, struct super_block *sb)
783 {
784 struct block_device *bdev;
785 char b[BDEVNAME_SIZE];
786
787 bdev = blkdev_get_by_dev(dev, FMODE_READ|FMODE_WRITE|FMODE_EXCL, sb);
788 if (IS_ERR(bdev))
789 goto fail;
790 return bdev;
791
792 fail:
793 ext4_msg(sb, KERN_ERR, "failed to open journal device %s: %ld",
794 __bdevname(dev, b), PTR_ERR(bdev));
795 return NULL;
796 }
797
798 /*
799 * Release the journal device
800 */
801 static int ext4_blkdev_put(struct block_device *bdev)
802 {
803 return blkdev_put(bdev, FMODE_READ|FMODE_WRITE|FMODE_EXCL);
804 }
805
806 static int ext4_blkdev_remove(struct ext4_sb_info *sbi)
807 {
808 struct block_device *bdev;
809 int ret = -ENODEV;
810
811 bdev = sbi->journal_bdev;
812 if (bdev) {
813 ret = ext4_blkdev_put(bdev);
814 sbi->journal_bdev = NULL;
815 }
816 return ret;
817 }
818
819 static inline struct inode *orphan_list_entry(struct list_head *l)
820 {
821 return &list_entry(l, struct ext4_inode_info, i_orphan)->vfs_inode;
822 }
823
824 static void dump_orphan_list(struct super_block *sb, struct ext4_sb_info *sbi)
825 {
826 struct list_head *l;
827
828 ext4_msg(sb, KERN_ERR, "sb orphan head is %d",
829 le32_to_cpu(sbi->s_es->s_last_orphan));
830
831 printk(KERN_ERR "sb_info orphan list:\n");
832 list_for_each(l, &sbi->s_orphan) {
833 struct inode *inode = orphan_list_entry(l);
834 printk(KERN_ERR " "
835 "inode %s:%lu at %p: mode %o, nlink %d, next %d\n",
836 inode->i_sb->s_id, inode->i_ino, inode,
837 inode->i_mode, inode->i_nlink,
838 NEXT_ORPHAN(inode));
839 }
840 }
841
842 static void ext4_put_super(struct super_block *sb)
843 {
844 struct ext4_sb_info *sbi = EXT4_SB(sb);
845 struct ext4_super_block *es = sbi->s_es;
846 int i, err;
847
848 ext4_unregister_li_request(sb);
849 dquot_disable(sb, -1, DQUOT_USAGE_ENABLED | DQUOT_LIMITS_ENABLED);
850
851 flush_workqueue(sbi->dio_unwritten_wq);
852 destroy_workqueue(sbi->dio_unwritten_wq);
853
854 lock_super(sb);
855 if (sbi->s_journal) {
856 err = jbd2_journal_destroy(sbi->s_journal);
857 sbi->s_journal = NULL;
858 if (err < 0)
859 ext4_abort(sb, "Couldn't clean up the journal");
860 }
861
862 del_timer(&sbi->s_err_report);
863 ext4_release_system_zone(sb);
864 ext4_mb_release(sb);
865 ext4_ext_release(sb);
866 ext4_xattr_put_super(sb);
867
868 if (!(sb->s_flags & MS_RDONLY)) {
869 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
870 es->s_state = cpu_to_le16(sbi->s_mount_state);
871 }
872 if (!(sb->s_flags & MS_RDONLY))
873 ext4_commit_super(sb, 1);
874
875 if (sbi->s_proc) {
876 remove_proc_entry("options", sbi->s_proc);
877 remove_proc_entry(sb->s_id, ext4_proc_root);
878 }
879 kobject_del(&sbi->s_kobj);
880
881 for (i = 0; i < sbi->s_gdb_count; i++)
882 brelse(sbi->s_group_desc[i]);
883 ext4_kvfree(sbi->s_group_desc);
884 ext4_kvfree(sbi->s_flex_groups);
885 percpu_counter_destroy(&sbi->s_freeclusters_counter);
886 percpu_counter_destroy(&sbi->s_freeinodes_counter);
887 percpu_counter_destroy(&sbi->s_dirs_counter);
888 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
889 brelse(sbi->s_sbh);
890 #ifdef CONFIG_QUOTA
891 for (i = 0; i < MAXQUOTAS; i++)
892 kfree(sbi->s_qf_names[i]);
893 #endif
894
895 /* Debugging code just in case the in-memory inode orphan list
896 * isn't empty. The on-disk one can be non-empty if we've
897 * detected an error and taken the fs readonly, but the
898 * in-memory list had better be clean by this point. */
899 if (!list_empty(&sbi->s_orphan))
900 dump_orphan_list(sb, sbi);
901 J_ASSERT(list_empty(&sbi->s_orphan));
902
903 invalidate_bdev(sb->s_bdev);
904 if (sbi->journal_bdev && sbi->journal_bdev != sb->s_bdev) {
905 /*
906 * Invalidate the journal device's buffers. We don't want them
907 * floating about in memory - the physical journal device may
908 * hotswapped, and it breaks the `ro-after' testing code.
909 */
910 sync_blockdev(sbi->journal_bdev);
911 invalidate_bdev(sbi->journal_bdev);
912 ext4_blkdev_remove(sbi);
913 }
914 if (sbi->s_mmp_tsk)
915 kthread_stop(sbi->s_mmp_tsk);
916 sb->s_fs_info = NULL;
917 /*
918 * Now that we are completely done shutting down the
919 * superblock, we need to actually destroy the kobject.
920 */
921 unlock_super(sb);
922 kobject_put(&sbi->s_kobj);
923 wait_for_completion(&sbi->s_kobj_unregister);
924 if (sbi->s_chksum_driver)
925 crypto_free_shash(sbi->s_chksum_driver);
926 kfree(sbi->s_blockgroup_lock);
927 kfree(sbi);
928 }
929
930 static struct kmem_cache *ext4_inode_cachep;
931
932 /*
933 * Called inside transaction, so use GFP_NOFS
934 */
935 static struct inode *ext4_alloc_inode(struct super_block *sb)
936 {
937 struct ext4_inode_info *ei;
938
939 ei = kmem_cache_alloc(ext4_inode_cachep, GFP_NOFS);
940 if (!ei)
941 return NULL;
942
943 ei->vfs_inode.i_version = 1;
944 ei->vfs_inode.i_data.writeback_index = 0;
945 memset(&ei->i_cached_extent, 0, sizeof(struct ext4_ext_cache));
946 INIT_LIST_HEAD(&ei->i_prealloc_list);
947 spin_lock_init(&ei->i_prealloc_lock);
948 ei->i_reserved_data_blocks = 0;
949 ei->i_reserved_meta_blocks = 0;
950 ei->i_allocated_meta_blocks = 0;
951 ei->i_da_metadata_calc_len = 0;
952 ei->i_da_metadata_calc_last_lblock = 0;
953 spin_lock_init(&(ei->i_block_reservation_lock));
954 #ifdef CONFIG_QUOTA
955 ei->i_reserved_quota = 0;
956 #endif
957 ei->jinode = NULL;
958 INIT_LIST_HEAD(&ei->i_completed_io_list);
959 spin_lock_init(&ei->i_completed_io_lock);
960 ei->cur_aio_dio = NULL;
961 ei->i_sync_tid = 0;
962 ei->i_datasync_tid = 0;
963 atomic_set(&ei->i_ioend_count, 0);
964 atomic_set(&ei->i_aiodio_unwritten, 0);
965
966 return &ei->vfs_inode;
967 }
968
969 static int ext4_drop_inode(struct inode *inode)
970 {
971 int drop = generic_drop_inode(inode);
972
973 trace_ext4_drop_inode(inode, drop);
974 return drop;
975 }
976
977 static void ext4_i_callback(struct rcu_head *head)
978 {
979 struct inode *inode = container_of(head, struct inode, i_rcu);
980 kmem_cache_free(ext4_inode_cachep, EXT4_I(inode));
981 }
982
983 static void ext4_destroy_inode(struct inode *inode)
984 {
985 if (!list_empty(&(EXT4_I(inode)->i_orphan))) {
986 ext4_msg(inode->i_sb, KERN_ERR,
987 "Inode %lu (%p): orphan list check failed!",
988 inode->i_ino, EXT4_I(inode));
989 print_hex_dump(KERN_INFO, "", DUMP_PREFIX_ADDRESS, 16, 4,
990 EXT4_I(inode), sizeof(struct ext4_inode_info),
991 true);
992 dump_stack();
993 }
994 call_rcu(&inode->i_rcu, ext4_i_callback);
995 }
996
997 static void init_once(void *foo)
998 {
999 struct ext4_inode_info *ei = (struct ext4_inode_info *) foo;
1000
1001 INIT_LIST_HEAD(&ei->i_orphan);
1002 #ifdef CONFIG_EXT4_FS_XATTR
1003 init_rwsem(&ei->xattr_sem);
1004 #endif
1005 init_rwsem(&ei->i_data_sem);
1006 inode_init_once(&ei->vfs_inode);
1007 }
1008
1009 static int init_inodecache(void)
1010 {
1011 ext4_inode_cachep = kmem_cache_create("ext4_inode_cache",
1012 sizeof(struct ext4_inode_info),
1013 0, (SLAB_RECLAIM_ACCOUNT|
1014 SLAB_MEM_SPREAD),
1015 init_once);
1016 if (ext4_inode_cachep == NULL)
1017 return -ENOMEM;
1018 return 0;
1019 }
1020
1021 static void destroy_inodecache(void)
1022 {
1023 kmem_cache_destroy(ext4_inode_cachep);
1024 }
1025
1026 void ext4_clear_inode(struct inode *inode)
1027 {
1028 invalidate_inode_buffers(inode);
1029 clear_inode(inode);
1030 dquot_drop(inode);
1031 ext4_discard_preallocations(inode);
1032 if (EXT4_I(inode)->jinode) {
1033 jbd2_journal_release_jbd_inode(EXT4_JOURNAL(inode),
1034 EXT4_I(inode)->jinode);
1035 jbd2_free_inode(EXT4_I(inode)->jinode);
1036 EXT4_I(inode)->jinode = NULL;
1037 }
1038 }
1039
1040 static struct inode *ext4_nfs_get_inode(struct super_block *sb,
1041 u64 ino, u32 generation)
1042 {
1043 struct inode *inode;
1044
1045 if (ino < EXT4_FIRST_INO(sb) && ino != EXT4_ROOT_INO)
1046 return ERR_PTR(-ESTALE);
1047 if (ino > le32_to_cpu(EXT4_SB(sb)->s_es->s_inodes_count))
1048 return ERR_PTR(-ESTALE);
1049
1050 /* iget isn't really right if the inode is currently unallocated!!
1051 *
1052 * ext4_read_inode will return a bad_inode if the inode had been
1053 * deleted, so we should be safe.
1054 *
1055 * Currently we don't know the generation for parent directory, so
1056 * a generation of 0 means "accept any"
1057 */
1058 inode = ext4_iget(sb, ino);
1059 if (IS_ERR(inode))
1060 return ERR_CAST(inode);
1061 if (generation && inode->i_generation != generation) {
1062 iput(inode);
1063 return ERR_PTR(-ESTALE);
1064 }
1065
1066 return inode;
1067 }
1068
1069 static struct dentry *ext4_fh_to_dentry(struct super_block *sb, struct fid *fid,
1070 int fh_len, int fh_type)
1071 {
1072 return generic_fh_to_dentry(sb, fid, fh_len, fh_type,
1073 ext4_nfs_get_inode);
1074 }
1075
1076 static struct dentry *ext4_fh_to_parent(struct super_block *sb, struct fid *fid,
1077 int fh_len, int fh_type)
1078 {
1079 return generic_fh_to_parent(sb, fid, fh_len, fh_type,
1080 ext4_nfs_get_inode);
1081 }
1082
1083 /*
1084 * Try to release metadata pages (indirect blocks, directories) which are
1085 * mapped via the block device. Since these pages could have journal heads
1086 * which would prevent try_to_free_buffers() from freeing them, we must use
1087 * jbd2 layer's try_to_free_buffers() function to release them.
1088 */
1089 static int bdev_try_to_free_page(struct super_block *sb, struct page *page,
1090 gfp_t wait)
1091 {
1092 journal_t *journal = EXT4_SB(sb)->s_journal;
1093
1094 WARN_ON(PageChecked(page));
1095 if (!page_has_buffers(page))
1096 return 0;
1097 if (journal)
1098 return jbd2_journal_try_to_free_buffers(journal, page,
1099 wait & ~__GFP_WAIT);
1100 return try_to_free_buffers(page);
1101 }
1102
1103 #ifdef CONFIG_QUOTA
1104 #define QTYPE2NAME(t) ((t) == USRQUOTA ? "user" : "group")
1105 #define QTYPE2MOPT(on, t) ((t) == USRQUOTA?((on)##USRJQUOTA):((on)##GRPJQUOTA))
1106
1107 static int ext4_write_dquot(struct dquot *dquot);
1108 static int ext4_acquire_dquot(struct dquot *dquot);
1109 static int ext4_release_dquot(struct dquot *dquot);
1110 static int ext4_mark_dquot_dirty(struct dquot *dquot);
1111 static int ext4_write_info(struct super_block *sb, int type);
1112 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
1113 struct path *path);
1114 static int ext4_quota_on_sysfile(struct super_block *sb, int type,
1115 int format_id);
1116 static int ext4_quota_off(struct super_block *sb, int type);
1117 static int ext4_quota_off_sysfile(struct super_block *sb, int type);
1118 static int ext4_quota_on_mount(struct super_block *sb, int type);
1119 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
1120 size_t len, loff_t off);
1121 static ssize_t ext4_quota_write(struct super_block *sb, int type,
1122 const char *data, size_t len, loff_t off);
1123 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
1124 unsigned int flags);
1125 static int ext4_enable_quotas(struct super_block *sb);
1126
1127 static const struct dquot_operations ext4_quota_operations = {
1128 .get_reserved_space = ext4_get_reserved_space,
1129 .write_dquot = ext4_write_dquot,
1130 .acquire_dquot = ext4_acquire_dquot,
1131 .release_dquot = ext4_release_dquot,
1132 .mark_dirty = ext4_mark_dquot_dirty,
1133 .write_info = ext4_write_info,
1134 .alloc_dquot = dquot_alloc,
1135 .destroy_dquot = dquot_destroy,
1136 };
1137
1138 static const struct quotactl_ops ext4_qctl_operations = {
1139 .quota_on = ext4_quota_on,
1140 .quota_off = ext4_quota_off,
1141 .quota_sync = dquot_quota_sync,
1142 .get_info = dquot_get_dqinfo,
1143 .set_info = dquot_set_dqinfo,
1144 .get_dqblk = dquot_get_dqblk,
1145 .set_dqblk = dquot_set_dqblk
1146 };
1147
1148 static const struct quotactl_ops ext4_qctl_sysfile_operations = {
1149 .quota_on_meta = ext4_quota_on_sysfile,
1150 .quota_off = ext4_quota_off_sysfile,
1151 .quota_sync = dquot_quota_sync,
1152 .get_info = dquot_get_dqinfo,
1153 .set_info = dquot_set_dqinfo,
1154 .get_dqblk = dquot_get_dqblk,
1155 .set_dqblk = dquot_set_dqblk
1156 };
1157 #endif
1158
1159 static const struct super_operations ext4_sops = {
1160 .alloc_inode = ext4_alloc_inode,
1161 .destroy_inode = ext4_destroy_inode,
1162 .write_inode = ext4_write_inode,
1163 .dirty_inode = ext4_dirty_inode,
1164 .drop_inode = ext4_drop_inode,
1165 .evict_inode = ext4_evict_inode,
1166 .put_super = ext4_put_super,
1167 .sync_fs = ext4_sync_fs,
1168 .freeze_fs = ext4_freeze,
1169 .unfreeze_fs = ext4_unfreeze,
1170 .statfs = ext4_statfs,
1171 .remount_fs = ext4_remount,
1172 .show_options = ext4_show_options,
1173 #ifdef CONFIG_QUOTA
1174 .quota_read = ext4_quota_read,
1175 .quota_write = ext4_quota_write,
1176 #endif
1177 .bdev_try_to_free_page = bdev_try_to_free_page,
1178 };
1179
1180 static const struct super_operations ext4_nojournal_sops = {
1181 .alloc_inode = ext4_alloc_inode,
1182 .destroy_inode = ext4_destroy_inode,
1183 .write_inode = ext4_write_inode,
1184 .dirty_inode = ext4_dirty_inode,
1185 .drop_inode = ext4_drop_inode,
1186 .evict_inode = ext4_evict_inode,
1187 .put_super = ext4_put_super,
1188 .statfs = ext4_statfs,
1189 .remount_fs = ext4_remount,
1190 .show_options = ext4_show_options,
1191 #ifdef CONFIG_QUOTA
1192 .quota_read = ext4_quota_read,
1193 .quota_write = ext4_quota_write,
1194 #endif
1195 .bdev_try_to_free_page = bdev_try_to_free_page,
1196 };
1197
1198 static const struct export_operations ext4_export_ops = {
1199 .fh_to_dentry = ext4_fh_to_dentry,
1200 .fh_to_parent = ext4_fh_to_parent,
1201 .get_parent = ext4_get_parent,
1202 };
1203
1204 enum {
1205 Opt_bsd_df, Opt_minix_df, Opt_grpid, Opt_nogrpid,
1206 Opt_resgid, Opt_resuid, Opt_sb, Opt_err_cont, Opt_err_panic, Opt_err_ro,
1207 Opt_nouid32, Opt_debug, Opt_removed,
1208 Opt_user_xattr, Opt_nouser_xattr, Opt_acl, Opt_noacl,
1209 Opt_auto_da_alloc, Opt_noauto_da_alloc, Opt_noload,
1210 Opt_commit, Opt_min_batch_time, Opt_max_batch_time,
1211 Opt_journal_dev, Opt_journal_checksum, Opt_journal_async_commit,
1212 Opt_abort, Opt_data_journal, Opt_data_ordered, Opt_data_writeback,
1213 Opt_data_err_abort, Opt_data_err_ignore,
1214 Opt_usrjquota, Opt_grpjquota, Opt_offusrjquota, Opt_offgrpjquota,
1215 Opt_jqfmt_vfsold, Opt_jqfmt_vfsv0, Opt_jqfmt_vfsv1, Opt_quota,
1216 Opt_noquota, Opt_barrier, Opt_nobarrier, Opt_err,
1217 Opt_usrquota, Opt_grpquota, Opt_i_version,
1218 Opt_stripe, Opt_delalloc, Opt_nodelalloc, Opt_mblk_io_submit,
1219 Opt_nomblk_io_submit, Opt_block_validity, Opt_noblock_validity,
1220 Opt_inode_readahead_blks, Opt_journal_ioprio,
1221 Opt_dioread_nolock, Opt_dioread_lock,
1222 Opt_discard, Opt_nodiscard, Opt_init_itable, Opt_noinit_itable,
1223 };
1224
1225 static const match_table_t tokens = {
1226 {Opt_bsd_df, "bsddf"},
1227 {Opt_minix_df, "minixdf"},
1228 {Opt_grpid, "grpid"},
1229 {Opt_grpid, "bsdgroups"},
1230 {Opt_nogrpid, "nogrpid"},
1231 {Opt_nogrpid, "sysvgroups"},
1232 {Opt_resgid, "resgid=%u"},
1233 {Opt_resuid, "resuid=%u"},
1234 {Opt_sb, "sb=%u"},
1235 {Opt_err_cont, "errors=continue"},
1236 {Opt_err_panic, "errors=panic"},
1237 {Opt_err_ro, "errors=remount-ro"},
1238 {Opt_nouid32, "nouid32"},
1239 {Opt_debug, "debug"},
1240 {Opt_removed, "oldalloc"},
1241 {Opt_removed, "orlov"},
1242 {Opt_user_xattr, "user_xattr"},
1243 {Opt_nouser_xattr, "nouser_xattr"},
1244 {Opt_acl, "acl"},
1245 {Opt_noacl, "noacl"},
1246 {Opt_noload, "norecovery"},
1247 {Opt_noload, "noload"},
1248 {Opt_removed, "nobh"},
1249 {Opt_removed, "bh"},
1250 {Opt_commit, "commit=%u"},
1251 {Opt_min_batch_time, "min_batch_time=%u"},
1252 {Opt_max_batch_time, "max_batch_time=%u"},
1253 {Opt_journal_dev, "journal_dev=%u"},
1254 {Opt_journal_checksum, "journal_checksum"},
1255 {Opt_journal_async_commit, "journal_async_commit"},
1256 {Opt_abort, "abort"},
1257 {Opt_data_journal, "data=journal"},
1258 {Opt_data_ordered, "data=ordered"},
1259 {Opt_data_writeback, "data=writeback"},
1260 {Opt_data_err_abort, "data_err=abort"},
1261 {Opt_data_err_ignore, "data_err=ignore"},
1262 {Opt_offusrjquota, "usrjquota="},
1263 {Opt_usrjquota, "usrjquota=%s"},
1264 {Opt_offgrpjquota, "grpjquota="},
1265 {Opt_grpjquota, "grpjquota=%s"},
1266 {Opt_jqfmt_vfsold, "jqfmt=vfsold"},
1267 {Opt_jqfmt_vfsv0, "jqfmt=vfsv0"},
1268 {Opt_jqfmt_vfsv1, "jqfmt=vfsv1"},
1269 {Opt_grpquota, "grpquota"},
1270 {Opt_noquota, "noquota"},
1271 {Opt_quota, "quota"},
1272 {Opt_usrquota, "usrquota"},
1273 {Opt_barrier, "barrier=%u"},
1274 {Opt_barrier, "barrier"},
1275 {Opt_nobarrier, "nobarrier"},
1276 {Opt_i_version, "i_version"},
1277 {Opt_stripe, "stripe=%u"},
1278 {Opt_delalloc, "delalloc"},
1279 {Opt_nodelalloc, "nodelalloc"},
1280 {Opt_mblk_io_submit, "mblk_io_submit"},
1281 {Opt_nomblk_io_submit, "nomblk_io_submit"},
1282 {Opt_block_validity, "block_validity"},
1283 {Opt_noblock_validity, "noblock_validity"},
1284 {Opt_inode_readahead_blks, "inode_readahead_blks=%u"},
1285 {Opt_journal_ioprio, "journal_ioprio=%u"},
1286 {Opt_auto_da_alloc, "auto_da_alloc=%u"},
1287 {Opt_auto_da_alloc, "auto_da_alloc"},
1288 {Opt_noauto_da_alloc, "noauto_da_alloc"},
1289 {Opt_dioread_nolock, "dioread_nolock"},
1290 {Opt_dioread_lock, "dioread_lock"},
1291 {Opt_discard, "discard"},
1292 {Opt_nodiscard, "nodiscard"},
1293 {Opt_init_itable, "init_itable=%u"},
1294 {Opt_init_itable, "init_itable"},
1295 {Opt_noinit_itable, "noinit_itable"},
1296 {Opt_removed, "check=none"}, /* mount option from ext2/3 */
1297 {Opt_removed, "nocheck"}, /* mount option from ext2/3 */
1298 {Opt_removed, "reservation"}, /* mount option from ext2/3 */
1299 {Opt_removed, "noreservation"}, /* mount option from ext2/3 */
1300 {Opt_removed, "journal=%u"}, /* mount option from ext2/3 */
1301 {Opt_err, NULL},
1302 };
1303
1304 static ext4_fsblk_t get_sb_block(void **data)
1305 {
1306 ext4_fsblk_t sb_block;
1307 char *options = (char *) *data;
1308
1309 if (!options || strncmp(options, "sb=", 3) != 0)
1310 return 1; /* Default location */
1311
1312 options += 3;
1313 /* TODO: use simple_strtoll with >32bit ext4 */
1314 sb_block = simple_strtoul(options, &options, 0);
1315 if (*options && *options != ',') {
1316 printk(KERN_ERR "EXT4-fs: Invalid sb specification: %s\n",
1317 (char *) *data);
1318 return 1;
1319 }
1320 if (*options == ',')
1321 options++;
1322 *data = (void *) options;
1323
1324 return sb_block;
1325 }
1326
1327 #define DEFAULT_JOURNAL_IOPRIO (IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, 3))
1328 static char deprecated_msg[] = "Mount option \"%s\" will be removed by %s\n"
1329 "Contact linux-ext4@vger.kernel.org if you think we should keep it.\n";
1330
1331 #ifdef CONFIG_QUOTA
1332 static int set_qf_name(struct super_block *sb, int qtype, substring_t *args)
1333 {
1334 struct ext4_sb_info *sbi = EXT4_SB(sb);
1335 char *qname;
1336
1337 if (sb_any_quota_loaded(sb) &&
1338 !sbi->s_qf_names[qtype]) {
1339 ext4_msg(sb, KERN_ERR,
1340 "Cannot change journaled "
1341 "quota options when quota turned on");
1342 return -1;
1343 }
1344 qname = match_strdup(args);
1345 if (!qname) {
1346 ext4_msg(sb, KERN_ERR,
1347 "Not enough memory for storing quotafile name");
1348 return -1;
1349 }
1350 if (sbi->s_qf_names[qtype] &&
1351 strcmp(sbi->s_qf_names[qtype], qname)) {
1352 ext4_msg(sb, KERN_ERR,
1353 "%s quota file already specified", QTYPE2NAME(qtype));
1354 kfree(qname);
1355 return -1;
1356 }
1357 sbi->s_qf_names[qtype] = qname;
1358 if (strchr(sbi->s_qf_names[qtype], '/')) {
1359 ext4_msg(sb, KERN_ERR,
1360 "quotafile must be on filesystem root");
1361 kfree(sbi->s_qf_names[qtype]);
1362 sbi->s_qf_names[qtype] = NULL;
1363 return -1;
1364 }
1365 set_opt(sb, QUOTA);
1366 return 1;
1367 }
1368
1369 static int clear_qf_name(struct super_block *sb, int qtype)
1370 {
1371
1372 struct ext4_sb_info *sbi = EXT4_SB(sb);
1373
1374 if (sb_any_quota_loaded(sb) &&
1375 sbi->s_qf_names[qtype]) {
1376 ext4_msg(sb, KERN_ERR, "Cannot change journaled quota options"
1377 " when quota turned on");
1378 return -1;
1379 }
1380 /*
1381 * The space will be released later when all options are confirmed
1382 * to be correct
1383 */
1384 sbi->s_qf_names[qtype] = NULL;
1385 return 1;
1386 }
1387 #endif
1388
1389 #define MOPT_SET 0x0001
1390 #define MOPT_CLEAR 0x0002
1391 #define MOPT_NOSUPPORT 0x0004
1392 #define MOPT_EXPLICIT 0x0008
1393 #define MOPT_CLEAR_ERR 0x0010
1394 #define MOPT_GTE0 0x0020
1395 #ifdef CONFIG_QUOTA
1396 #define MOPT_Q 0
1397 #define MOPT_QFMT 0x0040
1398 #else
1399 #define MOPT_Q MOPT_NOSUPPORT
1400 #define MOPT_QFMT MOPT_NOSUPPORT
1401 #endif
1402 #define MOPT_DATAJ 0x0080
1403
1404 static const struct mount_opts {
1405 int token;
1406 int mount_opt;
1407 int flags;
1408 } ext4_mount_opts[] = {
1409 {Opt_minix_df, EXT4_MOUNT_MINIX_DF, MOPT_SET},
1410 {Opt_bsd_df, EXT4_MOUNT_MINIX_DF, MOPT_CLEAR},
1411 {Opt_grpid, EXT4_MOUNT_GRPID, MOPT_SET},
1412 {Opt_nogrpid, EXT4_MOUNT_GRPID, MOPT_CLEAR},
1413 {Opt_mblk_io_submit, EXT4_MOUNT_MBLK_IO_SUBMIT, MOPT_SET},
1414 {Opt_nomblk_io_submit, EXT4_MOUNT_MBLK_IO_SUBMIT, MOPT_CLEAR},
1415 {Opt_block_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_SET},
1416 {Opt_noblock_validity, EXT4_MOUNT_BLOCK_VALIDITY, MOPT_CLEAR},
1417 {Opt_dioread_nolock, EXT4_MOUNT_DIOREAD_NOLOCK, MOPT_SET},
1418 {Opt_dioread_lock, EXT4_MOUNT_DIOREAD_NOLOCK, MOPT_CLEAR},
1419 {Opt_discard, EXT4_MOUNT_DISCARD, MOPT_SET},
1420 {Opt_nodiscard, EXT4_MOUNT_DISCARD, MOPT_CLEAR},
1421 {Opt_delalloc, EXT4_MOUNT_DELALLOC, MOPT_SET | MOPT_EXPLICIT},
1422 {Opt_nodelalloc, EXT4_MOUNT_DELALLOC, MOPT_CLEAR | MOPT_EXPLICIT},
1423 {Opt_journal_checksum, EXT4_MOUNT_JOURNAL_CHECKSUM, MOPT_SET},
1424 {Opt_journal_async_commit, (EXT4_MOUNT_JOURNAL_ASYNC_COMMIT |
1425 EXT4_MOUNT_JOURNAL_CHECKSUM), MOPT_SET},
1426 {Opt_noload, EXT4_MOUNT_NOLOAD, MOPT_SET},
1427 {Opt_err_panic, EXT4_MOUNT_ERRORS_PANIC, MOPT_SET | MOPT_CLEAR_ERR},
1428 {Opt_err_ro, EXT4_MOUNT_ERRORS_RO, MOPT_SET | MOPT_CLEAR_ERR},
1429 {Opt_err_cont, EXT4_MOUNT_ERRORS_CONT, MOPT_SET | MOPT_CLEAR_ERR},
1430 {Opt_data_err_abort, EXT4_MOUNT_DATA_ERR_ABORT, MOPT_SET},
1431 {Opt_data_err_ignore, EXT4_MOUNT_DATA_ERR_ABORT, MOPT_CLEAR},
1432 {Opt_barrier, EXT4_MOUNT_BARRIER, MOPT_SET},
1433 {Opt_nobarrier, EXT4_MOUNT_BARRIER, MOPT_CLEAR},
1434 {Opt_noauto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_SET},
1435 {Opt_auto_da_alloc, EXT4_MOUNT_NO_AUTO_DA_ALLOC, MOPT_CLEAR},
1436 {Opt_noinit_itable, EXT4_MOUNT_INIT_INODE_TABLE, MOPT_CLEAR},
1437 {Opt_commit, 0, MOPT_GTE0},
1438 {Opt_max_batch_time, 0, MOPT_GTE0},
1439 {Opt_min_batch_time, 0, MOPT_GTE0},
1440 {Opt_inode_readahead_blks, 0, MOPT_GTE0},
1441 {Opt_init_itable, 0, MOPT_GTE0},
1442 {Opt_stripe, 0, MOPT_GTE0},
1443 {Opt_data_journal, EXT4_MOUNT_JOURNAL_DATA, MOPT_DATAJ},
1444 {Opt_data_ordered, EXT4_MOUNT_ORDERED_DATA, MOPT_DATAJ},
1445 {Opt_data_writeback, EXT4_MOUNT_WRITEBACK_DATA, MOPT_DATAJ},
1446 #ifdef CONFIG_EXT4_FS_XATTR
1447 {Opt_user_xattr, EXT4_MOUNT_XATTR_USER, MOPT_SET},
1448 {Opt_nouser_xattr, EXT4_MOUNT_XATTR_USER, MOPT_CLEAR},
1449 #else
1450 {Opt_user_xattr, 0, MOPT_NOSUPPORT},
1451 {Opt_nouser_xattr, 0, MOPT_NOSUPPORT},
1452 #endif
1453 #ifdef CONFIG_EXT4_FS_POSIX_ACL
1454 {Opt_acl, EXT4_MOUNT_POSIX_ACL, MOPT_SET},
1455 {Opt_noacl, EXT4_MOUNT_POSIX_ACL, MOPT_CLEAR},
1456 #else
1457 {Opt_acl, 0, MOPT_NOSUPPORT},
1458 {Opt_noacl, 0, MOPT_NOSUPPORT},
1459 #endif
1460 {Opt_nouid32, EXT4_MOUNT_NO_UID32, MOPT_SET},
1461 {Opt_debug, EXT4_MOUNT_DEBUG, MOPT_SET},
1462 {Opt_quota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA, MOPT_SET | MOPT_Q},
1463 {Opt_usrquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA,
1464 MOPT_SET | MOPT_Q},
1465 {Opt_grpquota, EXT4_MOUNT_QUOTA | EXT4_MOUNT_GRPQUOTA,
1466 MOPT_SET | MOPT_Q},
1467 {Opt_noquota, (EXT4_MOUNT_QUOTA | EXT4_MOUNT_USRQUOTA |
1468 EXT4_MOUNT_GRPQUOTA), MOPT_CLEAR | MOPT_Q},
1469 {Opt_usrjquota, 0, MOPT_Q},
1470 {Opt_grpjquota, 0, MOPT_Q},
1471 {Opt_offusrjquota, 0, MOPT_Q},
1472 {Opt_offgrpjquota, 0, MOPT_Q},
1473 {Opt_jqfmt_vfsold, QFMT_VFS_OLD, MOPT_QFMT},
1474 {Opt_jqfmt_vfsv0, QFMT_VFS_V0, MOPT_QFMT},
1475 {Opt_jqfmt_vfsv1, QFMT_VFS_V1, MOPT_QFMT},
1476 {Opt_err, 0, 0}
1477 };
1478
1479 static int handle_mount_opt(struct super_block *sb, char *opt, int token,
1480 substring_t *args, unsigned long *journal_devnum,
1481 unsigned int *journal_ioprio, int is_remount)
1482 {
1483 struct ext4_sb_info *sbi = EXT4_SB(sb);
1484 const struct mount_opts *m;
1485 kuid_t uid;
1486 kgid_t gid;
1487 int arg = 0;
1488
1489 #ifdef CONFIG_QUOTA
1490 if (token == Opt_usrjquota)
1491 return set_qf_name(sb, USRQUOTA, &args[0]);
1492 else if (token == Opt_grpjquota)
1493 return set_qf_name(sb, GRPQUOTA, &args[0]);
1494 else if (token == Opt_offusrjquota)
1495 return clear_qf_name(sb, USRQUOTA);
1496 else if (token == Opt_offgrpjquota)
1497 return clear_qf_name(sb, GRPQUOTA);
1498 #endif
1499 if (args->from && match_int(args, &arg))
1500 return -1;
1501 switch (token) {
1502 case Opt_noacl:
1503 case Opt_nouser_xattr:
1504 ext4_msg(sb, KERN_WARNING, deprecated_msg, opt, "3.5");
1505 break;
1506 case Opt_sb:
1507 return 1; /* handled by get_sb_block() */
1508 case Opt_removed:
1509 ext4_msg(sb, KERN_WARNING,
1510 "Ignoring removed %s option", opt);
1511 return 1;
1512 case Opt_resuid:
1513 uid = make_kuid(current_user_ns(), arg);
1514 if (!uid_valid(uid)) {
1515 ext4_msg(sb, KERN_ERR, "Invalid uid value %d", arg);
1516 return -1;
1517 }
1518 sbi->s_resuid = uid;
1519 return 1;
1520 case Opt_resgid:
1521 gid = make_kgid(current_user_ns(), arg);
1522 if (!gid_valid(gid)) {
1523 ext4_msg(sb, KERN_ERR, "Invalid gid value %d", arg);
1524 return -1;
1525 }
1526 sbi->s_resgid = gid;
1527 return 1;
1528 case Opt_abort:
1529 sbi->s_mount_flags |= EXT4_MF_FS_ABORTED;
1530 return 1;
1531 case Opt_i_version:
1532 sb->s_flags |= MS_I_VERSION;
1533 return 1;
1534 case Opt_journal_dev:
1535 if (is_remount) {
1536 ext4_msg(sb, KERN_ERR,
1537 "Cannot specify journal on remount");
1538 return -1;
1539 }
1540 *journal_devnum = arg;
1541 return 1;
1542 case Opt_journal_ioprio:
1543 if (arg < 0 || arg > 7)
1544 return -1;
1545 *journal_ioprio = IOPRIO_PRIO_VALUE(IOPRIO_CLASS_BE, arg);
1546 return 1;
1547 }
1548
1549 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1550 if (token != m->token)
1551 continue;
1552 if (args->from && (m->flags & MOPT_GTE0) && (arg < 0))
1553 return -1;
1554 if (m->flags & MOPT_EXPLICIT)
1555 set_opt2(sb, EXPLICIT_DELALLOC);
1556 if (m->flags & MOPT_CLEAR_ERR)
1557 clear_opt(sb, ERRORS_MASK);
1558 if (token == Opt_noquota && sb_any_quota_loaded(sb)) {
1559 ext4_msg(sb, KERN_ERR, "Cannot change quota "
1560 "options when quota turned on");
1561 return -1;
1562 }
1563
1564 if (m->flags & MOPT_NOSUPPORT) {
1565 ext4_msg(sb, KERN_ERR, "%s option not supported", opt);
1566 } else if (token == Opt_commit) {
1567 if (arg == 0)
1568 arg = JBD2_DEFAULT_MAX_COMMIT_AGE;
1569 sbi->s_commit_interval = HZ * arg;
1570 } else if (token == Opt_max_batch_time) {
1571 if (arg == 0)
1572 arg = EXT4_DEF_MAX_BATCH_TIME;
1573 sbi->s_max_batch_time = arg;
1574 } else if (token == Opt_min_batch_time) {
1575 sbi->s_min_batch_time = arg;
1576 } else if (token == Opt_inode_readahead_blks) {
1577 if (arg > (1 << 30))
1578 return -1;
1579 if (arg && !is_power_of_2(arg)) {
1580 ext4_msg(sb, KERN_ERR,
1581 "EXT4-fs: inode_readahead_blks"
1582 " must be a power of 2");
1583 return -1;
1584 }
1585 sbi->s_inode_readahead_blks = arg;
1586 } else if (token == Opt_init_itable) {
1587 set_opt(sb, INIT_INODE_TABLE);
1588 if (!args->from)
1589 arg = EXT4_DEF_LI_WAIT_MULT;
1590 sbi->s_li_wait_mult = arg;
1591 } else if (token == Opt_stripe) {
1592 sbi->s_stripe = arg;
1593 } else if (m->flags & MOPT_DATAJ) {
1594 if (is_remount) {
1595 if (!sbi->s_journal)
1596 ext4_msg(sb, KERN_WARNING, "Remounting file system with no journal so ignoring journalled data option");
1597 else if (test_opt(sb, DATA_FLAGS) !=
1598 m->mount_opt) {
1599 ext4_msg(sb, KERN_ERR,
1600 "Cannot change data mode on remount");
1601 return -1;
1602 }
1603 } else {
1604 clear_opt(sb, DATA_FLAGS);
1605 sbi->s_mount_opt |= m->mount_opt;
1606 }
1607 #ifdef CONFIG_QUOTA
1608 } else if (m->flags & MOPT_QFMT) {
1609 if (sb_any_quota_loaded(sb) &&
1610 sbi->s_jquota_fmt != m->mount_opt) {
1611 ext4_msg(sb, KERN_ERR, "Cannot "
1612 "change journaled quota options "
1613 "when quota turned on");
1614 return -1;
1615 }
1616 sbi->s_jquota_fmt = m->mount_opt;
1617 #endif
1618 } else {
1619 if (!args->from)
1620 arg = 1;
1621 if (m->flags & MOPT_CLEAR)
1622 arg = !arg;
1623 else if (unlikely(!(m->flags & MOPT_SET))) {
1624 ext4_msg(sb, KERN_WARNING,
1625 "buggy handling of option %s", opt);
1626 WARN_ON(1);
1627 return -1;
1628 }
1629 if (arg != 0)
1630 sbi->s_mount_opt |= m->mount_opt;
1631 else
1632 sbi->s_mount_opt &= ~m->mount_opt;
1633 }
1634 return 1;
1635 }
1636 ext4_msg(sb, KERN_ERR, "Unrecognized mount option \"%s\" "
1637 "or missing value", opt);
1638 return -1;
1639 }
1640
1641 static int parse_options(char *options, struct super_block *sb,
1642 unsigned long *journal_devnum,
1643 unsigned int *journal_ioprio,
1644 int is_remount)
1645 {
1646 #ifdef CONFIG_QUOTA
1647 struct ext4_sb_info *sbi = EXT4_SB(sb);
1648 #endif
1649 char *p;
1650 substring_t args[MAX_OPT_ARGS];
1651 int token;
1652
1653 if (!options)
1654 return 1;
1655
1656 while ((p = strsep(&options, ",")) != NULL) {
1657 if (!*p)
1658 continue;
1659 /*
1660 * Initialize args struct so we know whether arg was
1661 * found; some options take optional arguments.
1662 */
1663 args[0].to = args[0].from = 0;
1664 token = match_token(p, tokens, args);
1665 if (handle_mount_opt(sb, p, token, args, journal_devnum,
1666 journal_ioprio, is_remount) < 0)
1667 return 0;
1668 }
1669 #ifdef CONFIG_QUOTA
1670 if (sbi->s_qf_names[USRQUOTA] || sbi->s_qf_names[GRPQUOTA]) {
1671 if (test_opt(sb, USRQUOTA) && sbi->s_qf_names[USRQUOTA])
1672 clear_opt(sb, USRQUOTA);
1673
1674 if (test_opt(sb, GRPQUOTA) && sbi->s_qf_names[GRPQUOTA])
1675 clear_opt(sb, GRPQUOTA);
1676
1677 if (test_opt(sb, GRPQUOTA) || test_opt(sb, USRQUOTA)) {
1678 ext4_msg(sb, KERN_ERR, "old and new quota "
1679 "format mixing");
1680 return 0;
1681 }
1682
1683 if (!sbi->s_jquota_fmt) {
1684 ext4_msg(sb, KERN_ERR, "journaled quota format "
1685 "not specified");
1686 return 0;
1687 }
1688 } else {
1689 if (sbi->s_jquota_fmt) {
1690 ext4_msg(sb, KERN_ERR, "journaled quota format "
1691 "specified with no journaling "
1692 "enabled");
1693 return 0;
1694 }
1695 }
1696 #endif
1697 return 1;
1698 }
1699
1700 static inline void ext4_show_quota_options(struct seq_file *seq,
1701 struct super_block *sb)
1702 {
1703 #if defined(CONFIG_QUOTA)
1704 struct ext4_sb_info *sbi = EXT4_SB(sb);
1705
1706 if (sbi->s_jquota_fmt) {
1707 char *fmtname = "";
1708
1709 switch (sbi->s_jquota_fmt) {
1710 case QFMT_VFS_OLD:
1711 fmtname = "vfsold";
1712 break;
1713 case QFMT_VFS_V0:
1714 fmtname = "vfsv0";
1715 break;
1716 case QFMT_VFS_V1:
1717 fmtname = "vfsv1";
1718 break;
1719 }
1720 seq_printf(seq, ",jqfmt=%s", fmtname);
1721 }
1722
1723 if (sbi->s_qf_names[USRQUOTA])
1724 seq_printf(seq, ",usrjquota=%s", sbi->s_qf_names[USRQUOTA]);
1725
1726 if (sbi->s_qf_names[GRPQUOTA])
1727 seq_printf(seq, ",grpjquota=%s", sbi->s_qf_names[GRPQUOTA]);
1728
1729 if (test_opt(sb, USRQUOTA))
1730 seq_puts(seq, ",usrquota");
1731
1732 if (test_opt(sb, GRPQUOTA))
1733 seq_puts(seq, ",grpquota");
1734 #endif
1735 }
1736
1737 static const char *token2str(int token)
1738 {
1739 const struct match_token *t;
1740
1741 for (t = tokens; t->token != Opt_err; t++)
1742 if (t->token == token && !strchr(t->pattern, '='))
1743 break;
1744 return t->pattern;
1745 }
1746
1747 /*
1748 * Show an option if
1749 * - it's set to a non-default value OR
1750 * - if the per-sb default is different from the global default
1751 */
1752 static int _ext4_show_options(struct seq_file *seq, struct super_block *sb,
1753 int nodefs)
1754 {
1755 struct ext4_sb_info *sbi = EXT4_SB(sb);
1756 struct ext4_super_block *es = sbi->s_es;
1757 int def_errors, def_mount_opt = nodefs ? 0 : sbi->s_def_mount_opt;
1758 const struct mount_opts *m;
1759 char sep = nodefs ? '\n' : ',';
1760
1761 #define SEQ_OPTS_PUTS(str) seq_printf(seq, "%c" str, sep)
1762 #define SEQ_OPTS_PRINT(str, arg) seq_printf(seq, "%c" str, sep, arg)
1763
1764 if (sbi->s_sb_block != 1)
1765 SEQ_OPTS_PRINT("sb=%llu", sbi->s_sb_block);
1766
1767 for (m = ext4_mount_opts; m->token != Opt_err; m++) {
1768 int want_set = m->flags & MOPT_SET;
1769 if (((m->flags & (MOPT_SET|MOPT_CLEAR)) == 0) ||
1770 (m->flags & MOPT_CLEAR_ERR))
1771 continue;
1772 if (!(m->mount_opt & (sbi->s_mount_opt ^ def_mount_opt)))
1773 continue; /* skip if same as the default */
1774 if ((want_set &&
1775 (sbi->s_mount_opt & m->mount_opt) != m->mount_opt) ||
1776 (!want_set && (sbi->s_mount_opt & m->mount_opt)))
1777 continue; /* select Opt_noFoo vs Opt_Foo */
1778 SEQ_OPTS_PRINT("%s", token2str(m->token));
1779 }
1780
1781 if (nodefs || !uid_eq(sbi->s_resuid, make_kuid(&init_user_ns, EXT4_DEF_RESUID)) ||
1782 le16_to_cpu(es->s_def_resuid) != EXT4_DEF_RESUID)
1783 SEQ_OPTS_PRINT("resuid=%u",
1784 from_kuid_munged(&init_user_ns, sbi->s_resuid));
1785 if (nodefs || !gid_eq(sbi->s_resgid, make_kgid(&init_user_ns, EXT4_DEF_RESGID)) ||
1786 le16_to_cpu(es->s_def_resgid) != EXT4_DEF_RESGID)
1787 SEQ_OPTS_PRINT("resgid=%u",
1788 from_kgid_munged(&init_user_ns, sbi->s_resgid));
1789 def_errors = nodefs ? -1 : le16_to_cpu(es->s_errors);
1790 if (test_opt(sb, ERRORS_RO) && def_errors != EXT4_ERRORS_RO)
1791 SEQ_OPTS_PUTS("errors=remount-ro");
1792 if (test_opt(sb, ERRORS_CONT) && def_errors != EXT4_ERRORS_CONTINUE)
1793 SEQ_OPTS_PUTS("errors=continue");
1794 if (test_opt(sb, ERRORS_PANIC) && def_errors != EXT4_ERRORS_PANIC)
1795 SEQ_OPTS_PUTS("errors=panic");
1796 if (nodefs || sbi->s_commit_interval != JBD2_DEFAULT_MAX_COMMIT_AGE*HZ)
1797 SEQ_OPTS_PRINT("commit=%lu", sbi->s_commit_interval / HZ);
1798 if (nodefs || sbi->s_min_batch_time != EXT4_DEF_MIN_BATCH_TIME)
1799 SEQ_OPTS_PRINT("min_batch_time=%u", sbi->s_min_batch_time);
1800 if (nodefs || sbi->s_max_batch_time != EXT4_DEF_MAX_BATCH_TIME)
1801 SEQ_OPTS_PRINT("max_batch_time=%u", sbi->s_max_batch_time);
1802 if (sb->s_flags & MS_I_VERSION)
1803 SEQ_OPTS_PUTS("i_version");
1804 if (nodefs || sbi->s_stripe)
1805 SEQ_OPTS_PRINT("stripe=%lu", sbi->s_stripe);
1806 if (EXT4_MOUNT_DATA_FLAGS & (sbi->s_mount_opt ^ def_mount_opt)) {
1807 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
1808 SEQ_OPTS_PUTS("data=journal");
1809 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
1810 SEQ_OPTS_PUTS("data=ordered");
1811 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_WRITEBACK_DATA)
1812 SEQ_OPTS_PUTS("data=writeback");
1813 }
1814 if (nodefs ||
1815 sbi->s_inode_readahead_blks != EXT4_DEF_INODE_READAHEAD_BLKS)
1816 SEQ_OPTS_PRINT("inode_readahead_blks=%u",
1817 sbi->s_inode_readahead_blks);
1818
1819 if (nodefs || (test_opt(sb, INIT_INODE_TABLE) &&
1820 (sbi->s_li_wait_mult != EXT4_DEF_LI_WAIT_MULT)))
1821 SEQ_OPTS_PRINT("init_itable=%u", sbi->s_li_wait_mult);
1822
1823 ext4_show_quota_options(seq, sb);
1824 return 0;
1825 }
1826
1827 static int ext4_show_options(struct seq_file *seq, struct dentry *root)
1828 {
1829 return _ext4_show_options(seq, root->d_sb, 0);
1830 }
1831
1832 static int options_seq_show(struct seq_file *seq, void *offset)
1833 {
1834 struct super_block *sb = seq->private;
1835 int rc;
1836
1837 seq_puts(seq, (sb->s_flags & MS_RDONLY) ? "ro" : "rw");
1838 rc = _ext4_show_options(seq, sb, 1);
1839 seq_puts(seq, "\n");
1840 return rc;
1841 }
1842
1843 static int options_open_fs(struct inode *inode, struct file *file)
1844 {
1845 return single_open(file, options_seq_show, PDE(inode)->data);
1846 }
1847
1848 static const struct file_operations ext4_seq_options_fops = {
1849 .owner = THIS_MODULE,
1850 .open = options_open_fs,
1851 .read = seq_read,
1852 .llseek = seq_lseek,
1853 .release = single_release,
1854 };
1855
1856 static int ext4_setup_super(struct super_block *sb, struct ext4_super_block *es,
1857 int read_only)
1858 {
1859 struct ext4_sb_info *sbi = EXT4_SB(sb);
1860 int res = 0;
1861
1862 if (le32_to_cpu(es->s_rev_level) > EXT4_MAX_SUPP_REV) {
1863 ext4_msg(sb, KERN_ERR, "revision level too high, "
1864 "forcing read-only mode");
1865 res = MS_RDONLY;
1866 }
1867 if (read_only)
1868 goto done;
1869 if (!(sbi->s_mount_state & EXT4_VALID_FS))
1870 ext4_msg(sb, KERN_WARNING, "warning: mounting unchecked fs, "
1871 "running e2fsck is recommended");
1872 else if ((sbi->s_mount_state & EXT4_ERROR_FS))
1873 ext4_msg(sb, KERN_WARNING,
1874 "warning: mounting fs with errors, "
1875 "running e2fsck is recommended");
1876 else if ((__s16) le16_to_cpu(es->s_max_mnt_count) > 0 &&
1877 le16_to_cpu(es->s_mnt_count) >=
1878 (unsigned short) (__s16) le16_to_cpu(es->s_max_mnt_count))
1879 ext4_msg(sb, KERN_WARNING,
1880 "warning: maximal mount count reached, "
1881 "running e2fsck is recommended");
1882 else if (le32_to_cpu(es->s_checkinterval) &&
1883 (le32_to_cpu(es->s_lastcheck) +
1884 le32_to_cpu(es->s_checkinterval) <= get_seconds()))
1885 ext4_msg(sb, KERN_WARNING,
1886 "warning: checktime reached, "
1887 "running e2fsck is recommended");
1888 if (!sbi->s_journal)
1889 es->s_state &= cpu_to_le16(~EXT4_VALID_FS);
1890 if (!(__s16) le16_to_cpu(es->s_max_mnt_count))
1891 es->s_max_mnt_count = cpu_to_le16(EXT4_DFL_MAX_MNT_COUNT);
1892 le16_add_cpu(&es->s_mnt_count, 1);
1893 es->s_mtime = cpu_to_le32(get_seconds());
1894 ext4_update_dynamic_rev(sb);
1895 if (sbi->s_journal)
1896 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
1897
1898 ext4_commit_super(sb, 1);
1899 done:
1900 if (test_opt(sb, DEBUG))
1901 printk(KERN_INFO "[EXT4 FS bs=%lu, gc=%u, "
1902 "bpg=%lu, ipg=%lu, mo=%04x, mo2=%04x]\n",
1903 sb->s_blocksize,
1904 sbi->s_groups_count,
1905 EXT4_BLOCKS_PER_GROUP(sb),
1906 EXT4_INODES_PER_GROUP(sb),
1907 sbi->s_mount_opt, sbi->s_mount_opt2);
1908
1909 cleancache_init_fs(sb);
1910 return res;
1911 }
1912
1913 static int ext4_fill_flex_info(struct super_block *sb)
1914 {
1915 struct ext4_sb_info *sbi = EXT4_SB(sb);
1916 struct ext4_group_desc *gdp = NULL;
1917 ext4_group_t flex_group_count;
1918 ext4_group_t flex_group;
1919 unsigned int groups_per_flex = 0;
1920 size_t size;
1921 int i;
1922
1923 sbi->s_log_groups_per_flex = sbi->s_es->s_log_groups_per_flex;
1924 if (sbi->s_log_groups_per_flex < 1 || sbi->s_log_groups_per_flex > 31) {
1925 sbi->s_log_groups_per_flex = 0;
1926 return 1;
1927 }
1928 groups_per_flex = 1 << sbi->s_log_groups_per_flex;
1929
1930 /* We allocate both existing and potentially added groups */
1931 flex_group_count = ((sbi->s_groups_count + groups_per_flex - 1) +
1932 ((le16_to_cpu(sbi->s_es->s_reserved_gdt_blocks) + 1) <<
1933 EXT4_DESC_PER_BLOCK_BITS(sb))) / groups_per_flex;
1934 size = flex_group_count * sizeof(struct flex_groups);
1935 sbi->s_flex_groups = ext4_kvzalloc(size, GFP_KERNEL);
1936 if (sbi->s_flex_groups == NULL) {
1937 ext4_msg(sb, KERN_ERR, "not enough memory for %u flex groups",
1938 flex_group_count);
1939 goto failed;
1940 }
1941
1942 for (i = 0; i < sbi->s_groups_count; i++) {
1943 gdp = ext4_get_group_desc(sb, i, NULL);
1944
1945 flex_group = ext4_flex_group(sbi, i);
1946 atomic_add(ext4_free_inodes_count(sb, gdp),
1947 &sbi->s_flex_groups[flex_group].free_inodes);
1948 atomic_add(ext4_free_group_clusters(sb, gdp),
1949 &sbi->s_flex_groups[flex_group].free_clusters);
1950 atomic_add(ext4_used_dirs_count(sb, gdp),
1951 &sbi->s_flex_groups[flex_group].used_dirs);
1952 }
1953
1954 return 1;
1955 failed:
1956 return 0;
1957 }
1958
1959 static __le16 ext4_group_desc_csum(struct ext4_sb_info *sbi, __u32 block_group,
1960 struct ext4_group_desc *gdp)
1961 {
1962 int offset;
1963 __u16 crc = 0;
1964 __le32 le_group = cpu_to_le32(block_group);
1965
1966 if ((sbi->s_es->s_feature_ro_compat &
1967 cpu_to_le32(EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))) {
1968 /* Use new metadata_csum algorithm */
1969 __u16 old_csum;
1970 __u32 csum32;
1971
1972 old_csum = gdp->bg_checksum;
1973 gdp->bg_checksum = 0;
1974 csum32 = ext4_chksum(sbi, sbi->s_csum_seed, (__u8 *)&le_group,
1975 sizeof(le_group));
1976 csum32 = ext4_chksum(sbi, csum32, (__u8 *)gdp,
1977 sbi->s_desc_size);
1978 gdp->bg_checksum = old_csum;
1979
1980 crc = csum32 & 0xFFFF;
1981 goto out;
1982 }
1983
1984 /* old crc16 code */
1985 offset = offsetof(struct ext4_group_desc, bg_checksum);
1986
1987 crc = crc16(~0, sbi->s_es->s_uuid, sizeof(sbi->s_es->s_uuid));
1988 crc = crc16(crc, (__u8 *)&le_group, sizeof(le_group));
1989 crc = crc16(crc, (__u8 *)gdp, offset);
1990 offset += sizeof(gdp->bg_checksum); /* skip checksum */
1991 /* for checksum of struct ext4_group_desc do the rest...*/
1992 if ((sbi->s_es->s_feature_incompat &
1993 cpu_to_le32(EXT4_FEATURE_INCOMPAT_64BIT)) &&
1994 offset < le16_to_cpu(sbi->s_es->s_desc_size))
1995 crc = crc16(crc, (__u8 *)gdp + offset,
1996 le16_to_cpu(sbi->s_es->s_desc_size) -
1997 offset);
1998
1999 out:
2000 return cpu_to_le16(crc);
2001 }
2002
2003 int ext4_group_desc_csum_verify(struct super_block *sb, __u32 block_group,
2004 struct ext4_group_desc *gdp)
2005 {
2006 if (ext4_has_group_desc_csum(sb) &&
2007 (gdp->bg_checksum != ext4_group_desc_csum(EXT4_SB(sb),
2008 block_group, gdp)))
2009 return 0;
2010
2011 return 1;
2012 }
2013
2014 void ext4_group_desc_csum_set(struct super_block *sb, __u32 block_group,
2015 struct ext4_group_desc *gdp)
2016 {
2017 if (!ext4_has_group_desc_csum(sb))
2018 return;
2019 gdp->bg_checksum = ext4_group_desc_csum(EXT4_SB(sb), block_group, gdp);
2020 }
2021
2022 /* Called at mount-time, super-block is locked */
2023 static int ext4_check_descriptors(struct super_block *sb,
2024 ext4_group_t *first_not_zeroed)
2025 {
2026 struct ext4_sb_info *sbi = EXT4_SB(sb);
2027 ext4_fsblk_t first_block = le32_to_cpu(sbi->s_es->s_first_data_block);
2028 ext4_fsblk_t last_block;
2029 ext4_fsblk_t block_bitmap;
2030 ext4_fsblk_t inode_bitmap;
2031 ext4_fsblk_t inode_table;
2032 int flexbg_flag = 0;
2033 ext4_group_t i, grp = sbi->s_groups_count;
2034
2035 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
2036 flexbg_flag = 1;
2037
2038 ext4_debug("Checking group descriptors");
2039
2040 for (i = 0; i < sbi->s_groups_count; i++) {
2041 struct ext4_group_desc *gdp = ext4_get_group_desc(sb, i, NULL);
2042
2043 if (i == sbi->s_groups_count - 1 || flexbg_flag)
2044 last_block = ext4_blocks_count(sbi->s_es) - 1;
2045 else
2046 last_block = first_block +
2047 (EXT4_BLOCKS_PER_GROUP(sb) - 1);
2048
2049 if ((grp == sbi->s_groups_count) &&
2050 !(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2051 grp = i;
2052
2053 block_bitmap = ext4_block_bitmap(sb, gdp);
2054 if (block_bitmap < first_block || block_bitmap > last_block) {
2055 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2056 "Block bitmap for group %u not in group "
2057 "(block %llu)!", i, block_bitmap);
2058 return 0;
2059 }
2060 inode_bitmap = ext4_inode_bitmap(sb, gdp);
2061 if (inode_bitmap < first_block || inode_bitmap > last_block) {
2062 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2063 "Inode bitmap for group %u not in group "
2064 "(block %llu)!", i, inode_bitmap);
2065 return 0;
2066 }
2067 inode_table = ext4_inode_table(sb, gdp);
2068 if (inode_table < first_block ||
2069 inode_table + sbi->s_itb_per_group - 1 > last_block) {
2070 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2071 "Inode table for group %u not in group "
2072 "(block %llu)!", i, inode_table);
2073 return 0;
2074 }
2075 ext4_lock_group(sb, i);
2076 if (!ext4_group_desc_csum_verify(sb, i, gdp)) {
2077 ext4_msg(sb, KERN_ERR, "ext4_check_descriptors: "
2078 "Checksum for group %u failed (%u!=%u)",
2079 i, le16_to_cpu(ext4_group_desc_csum(sbi, i,
2080 gdp)), le16_to_cpu(gdp->bg_checksum));
2081 if (!(sb->s_flags & MS_RDONLY)) {
2082 ext4_unlock_group(sb, i);
2083 return 0;
2084 }
2085 }
2086 ext4_unlock_group(sb, i);
2087 if (!flexbg_flag)
2088 first_block += EXT4_BLOCKS_PER_GROUP(sb);
2089 }
2090 if (NULL != first_not_zeroed)
2091 *first_not_zeroed = grp;
2092
2093 ext4_free_blocks_count_set(sbi->s_es,
2094 EXT4_C2B(sbi, ext4_count_free_clusters(sb)));
2095 sbi->s_es->s_free_inodes_count =cpu_to_le32(ext4_count_free_inodes(sb));
2096 return 1;
2097 }
2098
2099 /* ext4_orphan_cleanup() walks a singly-linked list of inodes (starting at
2100 * the superblock) which were deleted from all directories, but held open by
2101 * a process at the time of a crash. We walk the list and try to delete these
2102 * inodes at recovery time (only with a read-write filesystem).
2103 *
2104 * In order to keep the orphan inode chain consistent during traversal (in
2105 * case of crash during recovery), we link each inode into the superblock
2106 * orphan list_head and handle it the same way as an inode deletion during
2107 * normal operation (which journals the operations for us).
2108 *
2109 * We only do an iget() and an iput() on each inode, which is very safe if we
2110 * accidentally point at an in-use or already deleted inode. The worst that
2111 * can happen in this case is that we get a "bit already cleared" message from
2112 * ext4_free_inode(). The only reason we would point at a wrong inode is if
2113 * e2fsck was run on this filesystem, and it must have already done the orphan
2114 * inode cleanup for us, so we can safely abort without any further action.
2115 */
2116 static void ext4_orphan_cleanup(struct super_block *sb,
2117 struct ext4_super_block *es)
2118 {
2119 unsigned int s_flags = sb->s_flags;
2120 int nr_orphans = 0, nr_truncates = 0;
2121 #ifdef CONFIG_QUOTA
2122 int i;
2123 #endif
2124 if (!es->s_last_orphan) {
2125 jbd_debug(4, "no orphan inodes to clean up\n");
2126 return;
2127 }
2128
2129 if (bdev_read_only(sb->s_bdev)) {
2130 ext4_msg(sb, KERN_ERR, "write access "
2131 "unavailable, skipping orphan cleanup");
2132 return;
2133 }
2134
2135 /* Check if feature set would not allow a r/w mount */
2136 if (!ext4_feature_set_ok(sb, 0)) {
2137 ext4_msg(sb, KERN_INFO, "Skipping orphan cleanup due to "
2138 "unknown ROCOMPAT features");
2139 return;
2140 }
2141
2142 if (EXT4_SB(sb)->s_mount_state & EXT4_ERROR_FS) {
2143 if (es->s_last_orphan)
2144 jbd_debug(1, "Errors on filesystem, "
2145 "clearing orphan list.\n");
2146 es->s_last_orphan = 0;
2147 jbd_debug(1, "Skipping orphan recovery on fs with errors.\n");
2148 return;
2149 }
2150
2151 if (s_flags & MS_RDONLY) {
2152 ext4_msg(sb, KERN_INFO, "orphan cleanup on readonly fs");
2153 sb->s_flags &= ~MS_RDONLY;
2154 }
2155 #ifdef CONFIG_QUOTA
2156 /* Needed for iput() to work correctly and not trash data */
2157 sb->s_flags |= MS_ACTIVE;
2158 /* Turn on quotas so that they are updated correctly */
2159 for (i = 0; i < MAXQUOTAS; i++) {
2160 if (EXT4_SB(sb)->s_qf_names[i]) {
2161 int ret = ext4_quota_on_mount(sb, i);
2162 if (ret < 0)
2163 ext4_msg(sb, KERN_ERR,
2164 "Cannot turn on journaled "
2165 "quota: error %d", ret);
2166 }
2167 }
2168 #endif
2169
2170 while (es->s_last_orphan) {
2171 struct inode *inode;
2172
2173 inode = ext4_orphan_get(sb, le32_to_cpu(es->s_last_orphan));
2174 if (IS_ERR(inode)) {
2175 es->s_last_orphan = 0;
2176 break;
2177 }
2178
2179 list_add(&EXT4_I(inode)->i_orphan, &EXT4_SB(sb)->s_orphan);
2180 dquot_initialize(inode);
2181 if (inode->i_nlink) {
2182 ext4_msg(sb, KERN_DEBUG,
2183 "%s: truncating inode %lu to %lld bytes",
2184 __func__, inode->i_ino, inode->i_size);
2185 jbd_debug(2, "truncating inode %lu to %lld bytes\n",
2186 inode->i_ino, inode->i_size);
2187 ext4_truncate(inode);
2188 nr_truncates++;
2189 } else {
2190 ext4_msg(sb, KERN_DEBUG,
2191 "%s: deleting unreferenced inode %lu",
2192 __func__, inode->i_ino);
2193 jbd_debug(2, "deleting unreferenced inode %lu\n",
2194 inode->i_ino);
2195 nr_orphans++;
2196 }
2197 iput(inode); /* The delete magic happens here! */
2198 }
2199
2200 #define PLURAL(x) (x), ((x) == 1) ? "" : "s"
2201
2202 if (nr_orphans)
2203 ext4_msg(sb, KERN_INFO, "%d orphan inode%s deleted",
2204 PLURAL(nr_orphans));
2205 if (nr_truncates)
2206 ext4_msg(sb, KERN_INFO, "%d truncate%s cleaned up",
2207 PLURAL(nr_truncates));
2208 #ifdef CONFIG_QUOTA
2209 /* Turn quotas off */
2210 for (i = 0; i < MAXQUOTAS; i++) {
2211 if (sb_dqopt(sb)->files[i])
2212 dquot_quota_off(sb, i);
2213 }
2214 #endif
2215 sb->s_flags = s_flags; /* Restore MS_RDONLY status */
2216 }
2217
2218 /*
2219 * Maximal extent format file size.
2220 * Resulting logical blkno at s_maxbytes must fit in our on-disk
2221 * extent format containers, within a sector_t, and within i_blocks
2222 * in the vfs. ext4 inode has 48 bits of i_block in fsblock units,
2223 * so that won't be a limiting factor.
2224 *
2225 * However there is other limiting factor. We do store extents in the form
2226 * of starting block and length, hence the resulting length of the extent
2227 * covering maximum file size must fit into on-disk format containers as
2228 * well. Given that length is always by 1 unit bigger than max unit (because
2229 * we count 0 as well) we have to lower the s_maxbytes by one fs block.
2230 *
2231 * Note, this does *not* consider any metadata overhead for vfs i_blocks.
2232 */
2233 static loff_t ext4_max_size(int blkbits, int has_huge_files)
2234 {
2235 loff_t res;
2236 loff_t upper_limit = MAX_LFS_FILESIZE;
2237
2238 /* small i_blocks in vfs inode? */
2239 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2240 /*
2241 * CONFIG_LBDAF is not enabled implies the inode
2242 * i_block represent total blocks in 512 bytes
2243 * 32 == size of vfs inode i_blocks * 8
2244 */
2245 upper_limit = (1LL << 32) - 1;
2246
2247 /* total blocks in file system block size */
2248 upper_limit >>= (blkbits - 9);
2249 upper_limit <<= blkbits;
2250 }
2251
2252 /*
2253 * 32-bit extent-start container, ee_block. We lower the maxbytes
2254 * by one fs block, so ee_len can cover the extent of maximum file
2255 * size
2256 */
2257 res = (1LL << 32) - 1;
2258 res <<= blkbits;
2259
2260 /* Sanity check against vm- & vfs- imposed limits */
2261 if (res > upper_limit)
2262 res = upper_limit;
2263
2264 return res;
2265 }
2266
2267 /*
2268 * Maximal bitmap file size. There is a direct, and {,double-,triple-}indirect
2269 * block limit, and also a limit of (2^48 - 1) 512-byte sectors in i_blocks.
2270 * We need to be 1 filesystem block less than the 2^48 sector limit.
2271 */
2272 static loff_t ext4_max_bitmap_size(int bits, int has_huge_files)
2273 {
2274 loff_t res = EXT4_NDIR_BLOCKS;
2275 int meta_blocks;
2276 loff_t upper_limit;
2277 /* This is calculated to be the largest file size for a dense, block
2278 * mapped file such that the file's total number of 512-byte sectors,
2279 * including data and all indirect blocks, does not exceed (2^48 - 1).
2280 *
2281 * __u32 i_blocks_lo and _u16 i_blocks_high represent the total
2282 * number of 512-byte sectors of the file.
2283 */
2284
2285 if (!has_huge_files || sizeof(blkcnt_t) < sizeof(u64)) {
2286 /*
2287 * !has_huge_files or CONFIG_LBDAF not enabled implies that
2288 * the inode i_block field represents total file blocks in
2289 * 2^32 512-byte sectors == size of vfs inode i_blocks * 8
2290 */
2291 upper_limit = (1LL << 32) - 1;
2292
2293 /* total blocks in file system block size */
2294 upper_limit >>= (bits - 9);
2295
2296 } else {
2297 /*
2298 * We use 48 bit ext4_inode i_blocks
2299 * With EXT4_HUGE_FILE_FL set the i_blocks
2300 * represent total number of blocks in
2301 * file system block size
2302 */
2303 upper_limit = (1LL << 48) - 1;
2304
2305 }
2306
2307 /* indirect blocks */
2308 meta_blocks = 1;
2309 /* double indirect blocks */
2310 meta_blocks += 1 + (1LL << (bits-2));
2311 /* tripple indirect blocks */
2312 meta_blocks += 1 + (1LL << (bits-2)) + (1LL << (2*(bits-2)));
2313
2314 upper_limit -= meta_blocks;
2315 upper_limit <<= bits;
2316
2317 res += 1LL << (bits-2);
2318 res += 1LL << (2*(bits-2));
2319 res += 1LL << (3*(bits-2));
2320 res <<= bits;
2321 if (res > upper_limit)
2322 res = upper_limit;
2323
2324 if (res > MAX_LFS_FILESIZE)
2325 res = MAX_LFS_FILESIZE;
2326
2327 return res;
2328 }
2329
2330 static ext4_fsblk_t descriptor_loc(struct super_block *sb,
2331 ext4_fsblk_t logical_sb_block, int nr)
2332 {
2333 struct ext4_sb_info *sbi = EXT4_SB(sb);
2334 ext4_group_t bg, first_meta_bg;
2335 int has_super = 0;
2336
2337 first_meta_bg = le32_to_cpu(sbi->s_es->s_first_meta_bg);
2338
2339 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_META_BG) ||
2340 nr < first_meta_bg)
2341 return logical_sb_block + nr + 1;
2342 bg = sbi->s_desc_per_block * nr;
2343 if (ext4_bg_has_super(sb, bg))
2344 has_super = 1;
2345
2346 return (has_super + ext4_group_first_block_no(sb, bg));
2347 }
2348
2349 /**
2350 * ext4_get_stripe_size: Get the stripe size.
2351 * @sbi: In memory super block info
2352 *
2353 * If we have specified it via mount option, then
2354 * use the mount option value. If the value specified at mount time is
2355 * greater than the blocks per group use the super block value.
2356 * If the super block value is greater than blocks per group return 0.
2357 * Allocator needs it be less than blocks per group.
2358 *
2359 */
2360 static unsigned long ext4_get_stripe_size(struct ext4_sb_info *sbi)
2361 {
2362 unsigned long stride = le16_to_cpu(sbi->s_es->s_raid_stride);
2363 unsigned long stripe_width =
2364 le32_to_cpu(sbi->s_es->s_raid_stripe_width);
2365 int ret;
2366
2367 if (sbi->s_stripe && sbi->s_stripe <= sbi->s_blocks_per_group)
2368 ret = sbi->s_stripe;
2369 else if (stripe_width <= sbi->s_blocks_per_group)
2370 ret = stripe_width;
2371 else if (stride <= sbi->s_blocks_per_group)
2372 ret = stride;
2373 else
2374 ret = 0;
2375
2376 /*
2377 * If the stripe width is 1, this makes no sense and
2378 * we set it to 0 to turn off stripe handling code.
2379 */
2380 if (ret <= 1)
2381 ret = 0;
2382
2383 return ret;
2384 }
2385
2386 /* sysfs supprt */
2387
2388 struct ext4_attr {
2389 struct attribute attr;
2390 ssize_t (*show)(struct ext4_attr *, struct ext4_sb_info *, char *);
2391 ssize_t (*store)(struct ext4_attr *, struct ext4_sb_info *,
2392 const char *, size_t);
2393 int offset;
2394 };
2395
2396 static int parse_strtoul(const char *buf,
2397 unsigned long max, unsigned long *value)
2398 {
2399 char *endp;
2400
2401 *value = simple_strtoul(skip_spaces(buf), &endp, 0);
2402 endp = skip_spaces(endp);
2403 if (*endp || *value > max)
2404 return -EINVAL;
2405
2406 return 0;
2407 }
2408
2409 static ssize_t delayed_allocation_blocks_show(struct ext4_attr *a,
2410 struct ext4_sb_info *sbi,
2411 char *buf)
2412 {
2413 return snprintf(buf, PAGE_SIZE, "%llu\n",
2414 (s64) EXT4_C2B(sbi,
2415 percpu_counter_sum(&sbi->s_dirtyclusters_counter)));
2416 }
2417
2418 static ssize_t session_write_kbytes_show(struct ext4_attr *a,
2419 struct ext4_sb_info *sbi, char *buf)
2420 {
2421 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2422
2423 if (!sb->s_bdev->bd_part)
2424 return snprintf(buf, PAGE_SIZE, "0\n");
2425 return snprintf(buf, PAGE_SIZE, "%lu\n",
2426 (part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2427 sbi->s_sectors_written_start) >> 1);
2428 }
2429
2430 static ssize_t lifetime_write_kbytes_show(struct ext4_attr *a,
2431 struct ext4_sb_info *sbi, char *buf)
2432 {
2433 struct super_block *sb = sbi->s_buddy_cache->i_sb;
2434
2435 if (!sb->s_bdev->bd_part)
2436 return snprintf(buf, PAGE_SIZE, "0\n");
2437 return snprintf(buf, PAGE_SIZE, "%llu\n",
2438 (unsigned long long)(sbi->s_kbytes_written +
2439 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
2440 EXT4_SB(sb)->s_sectors_written_start) >> 1)));
2441 }
2442
2443 static ssize_t inode_readahead_blks_store(struct ext4_attr *a,
2444 struct ext4_sb_info *sbi,
2445 const char *buf, size_t count)
2446 {
2447 unsigned long t;
2448
2449 if (parse_strtoul(buf, 0x40000000, &t))
2450 return -EINVAL;
2451
2452 if (t && !is_power_of_2(t))
2453 return -EINVAL;
2454
2455 sbi->s_inode_readahead_blks = t;
2456 return count;
2457 }
2458
2459 static ssize_t sbi_ui_show(struct ext4_attr *a,
2460 struct ext4_sb_info *sbi, char *buf)
2461 {
2462 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2463
2464 return snprintf(buf, PAGE_SIZE, "%u\n", *ui);
2465 }
2466
2467 static ssize_t sbi_ui_store(struct ext4_attr *a,
2468 struct ext4_sb_info *sbi,
2469 const char *buf, size_t count)
2470 {
2471 unsigned int *ui = (unsigned int *) (((char *) sbi) + a->offset);
2472 unsigned long t;
2473
2474 if (parse_strtoul(buf, 0xffffffff, &t))
2475 return -EINVAL;
2476 *ui = t;
2477 return count;
2478 }
2479
2480 static ssize_t trigger_test_error(struct ext4_attr *a,
2481 struct ext4_sb_info *sbi,
2482 const char *buf, size_t count)
2483 {
2484 int len = count;
2485
2486 if (!capable(CAP_SYS_ADMIN))
2487 return -EPERM;
2488
2489 if (len && buf[len-1] == '\n')
2490 len--;
2491
2492 if (len)
2493 ext4_error(sbi->s_sb, "%.*s", len, buf);
2494 return count;
2495 }
2496
2497 #define EXT4_ATTR_OFFSET(_name,_mode,_show,_store,_elname) \
2498 static struct ext4_attr ext4_attr_##_name = { \
2499 .attr = {.name = __stringify(_name), .mode = _mode }, \
2500 .show = _show, \
2501 .store = _store, \
2502 .offset = offsetof(struct ext4_sb_info, _elname), \
2503 }
2504 #define EXT4_ATTR(name, mode, show, store) \
2505 static struct ext4_attr ext4_attr_##name = __ATTR(name, mode, show, store)
2506
2507 #define EXT4_INFO_ATTR(name) EXT4_ATTR(name, 0444, NULL, NULL)
2508 #define EXT4_RO_ATTR(name) EXT4_ATTR(name, 0444, name##_show, NULL)
2509 #define EXT4_RW_ATTR(name) EXT4_ATTR(name, 0644, name##_show, name##_store)
2510 #define EXT4_RW_ATTR_SBI_UI(name, elname) \
2511 EXT4_ATTR_OFFSET(name, 0644, sbi_ui_show, sbi_ui_store, elname)
2512 #define ATTR_LIST(name) &ext4_attr_##name.attr
2513
2514 EXT4_RO_ATTR(delayed_allocation_blocks);
2515 EXT4_RO_ATTR(session_write_kbytes);
2516 EXT4_RO_ATTR(lifetime_write_kbytes);
2517 EXT4_ATTR_OFFSET(inode_readahead_blks, 0644, sbi_ui_show,
2518 inode_readahead_blks_store, s_inode_readahead_blks);
2519 EXT4_RW_ATTR_SBI_UI(inode_goal, s_inode_goal);
2520 EXT4_RW_ATTR_SBI_UI(mb_stats, s_mb_stats);
2521 EXT4_RW_ATTR_SBI_UI(mb_max_to_scan, s_mb_max_to_scan);
2522 EXT4_RW_ATTR_SBI_UI(mb_min_to_scan, s_mb_min_to_scan);
2523 EXT4_RW_ATTR_SBI_UI(mb_order2_req, s_mb_order2_reqs);
2524 EXT4_RW_ATTR_SBI_UI(mb_stream_req, s_mb_stream_request);
2525 EXT4_RW_ATTR_SBI_UI(mb_group_prealloc, s_mb_group_prealloc);
2526 EXT4_RW_ATTR_SBI_UI(max_writeback_mb_bump, s_max_writeback_mb_bump);
2527 EXT4_ATTR(trigger_fs_error, 0200, NULL, trigger_test_error);
2528
2529 static struct attribute *ext4_attrs[] = {
2530 ATTR_LIST(delayed_allocation_blocks),
2531 ATTR_LIST(session_write_kbytes),
2532 ATTR_LIST(lifetime_write_kbytes),
2533 ATTR_LIST(inode_readahead_blks),
2534 ATTR_LIST(inode_goal),
2535 ATTR_LIST(mb_stats),
2536 ATTR_LIST(mb_max_to_scan),
2537 ATTR_LIST(mb_min_to_scan),
2538 ATTR_LIST(mb_order2_req),
2539 ATTR_LIST(mb_stream_req),
2540 ATTR_LIST(mb_group_prealloc),
2541 ATTR_LIST(max_writeback_mb_bump),
2542 ATTR_LIST(trigger_fs_error),
2543 NULL,
2544 };
2545
2546 /* Features this copy of ext4 supports */
2547 EXT4_INFO_ATTR(lazy_itable_init);
2548 EXT4_INFO_ATTR(batched_discard);
2549
2550 static struct attribute *ext4_feat_attrs[] = {
2551 ATTR_LIST(lazy_itable_init),
2552 ATTR_LIST(batched_discard),
2553 NULL,
2554 };
2555
2556 static ssize_t ext4_attr_show(struct kobject *kobj,
2557 struct attribute *attr, char *buf)
2558 {
2559 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2560 s_kobj);
2561 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2562
2563 return a->show ? a->show(a, sbi, buf) : 0;
2564 }
2565
2566 static ssize_t ext4_attr_store(struct kobject *kobj,
2567 struct attribute *attr,
2568 const char *buf, size_t len)
2569 {
2570 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2571 s_kobj);
2572 struct ext4_attr *a = container_of(attr, struct ext4_attr, attr);
2573
2574 return a->store ? a->store(a, sbi, buf, len) : 0;
2575 }
2576
2577 static void ext4_sb_release(struct kobject *kobj)
2578 {
2579 struct ext4_sb_info *sbi = container_of(kobj, struct ext4_sb_info,
2580 s_kobj);
2581 complete(&sbi->s_kobj_unregister);
2582 }
2583
2584 static const struct sysfs_ops ext4_attr_ops = {
2585 .show = ext4_attr_show,
2586 .store = ext4_attr_store,
2587 };
2588
2589 static struct kobj_type ext4_ktype = {
2590 .default_attrs = ext4_attrs,
2591 .sysfs_ops = &ext4_attr_ops,
2592 .release = ext4_sb_release,
2593 };
2594
2595 static void ext4_feat_release(struct kobject *kobj)
2596 {
2597 complete(&ext4_feat->f_kobj_unregister);
2598 }
2599
2600 static struct kobj_type ext4_feat_ktype = {
2601 .default_attrs = ext4_feat_attrs,
2602 .sysfs_ops = &ext4_attr_ops,
2603 .release = ext4_feat_release,
2604 };
2605
2606 /*
2607 * Check whether this filesystem can be mounted based on
2608 * the features present and the RDONLY/RDWR mount requested.
2609 * Returns 1 if this filesystem can be mounted as requested,
2610 * 0 if it cannot be.
2611 */
2612 static int ext4_feature_set_ok(struct super_block *sb, int readonly)
2613 {
2614 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT4_FEATURE_INCOMPAT_SUPP)) {
2615 ext4_msg(sb, KERN_ERR,
2616 "Couldn't mount because of "
2617 "unsupported optional features (%x)",
2618 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_incompat) &
2619 ~EXT4_FEATURE_INCOMPAT_SUPP));
2620 return 0;
2621 }
2622
2623 if (readonly)
2624 return 1;
2625
2626 /* Check that feature set is OK for a read-write mount */
2627 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT4_FEATURE_RO_COMPAT_SUPP)) {
2628 ext4_msg(sb, KERN_ERR, "couldn't mount RDWR because of "
2629 "unsupported optional features (%x)",
2630 (le32_to_cpu(EXT4_SB(sb)->s_es->s_feature_ro_compat) &
2631 ~EXT4_FEATURE_RO_COMPAT_SUPP));
2632 return 0;
2633 }
2634 /*
2635 * Large file size enabled file system can only be mounted
2636 * read-write on 32-bit systems if kernel is built with CONFIG_LBDAF
2637 */
2638 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_HUGE_FILE)) {
2639 if (sizeof(blkcnt_t) < sizeof(u64)) {
2640 ext4_msg(sb, KERN_ERR, "Filesystem with huge files "
2641 "cannot be mounted RDWR without "
2642 "CONFIG_LBDAF");
2643 return 0;
2644 }
2645 }
2646 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC) &&
2647 !EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_EXTENTS)) {
2648 ext4_msg(sb, KERN_ERR,
2649 "Can't support bigalloc feature without "
2650 "extents feature\n");
2651 return 0;
2652 }
2653
2654 #ifndef CONFIG_QUOTA
2655 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
2656 !readonly) {
2657 ext4_msg(sb, KERN_ERR,
2658 "Filesystem with quota feature cannot be mounted RDWR "
2659 "without CONFIG_QUOTA");
2660 return 0;
2661 }
2662 #endif /* CONFIG_QUOTA */
2663 return 1;
2664 }
2665
2666 /*
2667 * This function is called once a day if we have errors logged
2668 * on the file system
2669 */
2670 static void print_daily_error_info(unsigned long arg)
2671 {
2672 struct super_block *sb = (struct super_block *) arg;
2673 struct ext4_sb_info *sbi;
2674 struct ext4_super_block *es;
2675
2676 sbi = EXT4_SB(sb);
2677 es = sbi->s_es;
2678
2679 if (es->s_error_count)
2680 ext4_msg(sb, KERN_NOTICE, "error count: %u",
2681 le32_to_cpu(es->s_error_count));
2682 if (es->s_first_error_time) {
2683 printk(KERN_NOTICE "EXT4-fs (%s): initial error at %u: %.*s:%d",
2684 sb->s_id, le32_to_cpu(es->s_first_error_time),
2685 (int) sizeof(es->s_first_error_func),
2686 es->s_first_error_func,
2687 le32_to_cpu(es->s_first_error_line));
2688 if (es->s_first_error_ino)
2689 printk(": inode %u",
2690 le32_to_cpu(es->s_first_error_ino));
2691 if (es->s_first_error_block)
2692 printk(": block %llu", (unsigned long long)
2693 le64_to_cpu(es->s_first_error_block));
2694 printk("\n");
2695 }
2696 if (es->s_last_error_time) {
2697 printk(KERN_NOTICE "EXT4-fs (%s): last error at %u: %.*s:%d",
2698 sb->s_id, le32_to_cpu(es->s_last_error_time),
2699 (int) sizeof(es->s_last_error_func),
2700 es->s_last_error_func,
2701 le32_to_cpu(es->s_last_error_line));
2702 if (es->s_last_error_ino)
2703 printk(": inode %u",
2704 le32_to_cpu(es->s_last_error_ino));
2705 if (es->s_last_error_block)
2706 printk(": block %llu", (unsigned long long)
2707 le64_to_cpu(es->s_last_error_block));
2708 printk("\n");
2709 }
2710 mod_timer(&sbi->s_err_report, jiffies + 24*60*60*HZ); /* Once a day */
2711 }
2712
2713 /* Find next suitable group and run ext4_init_inode_table */
2714 static int ext4_run_li_request(struct ext4_li_request *elr)
2715 {
2716 struct ext4_group_desc *gdp = NULL;
2717 ext4_group_t group, ngroups;
2718 struct super_block *sb;
2719 unsigned long timeout = 0;
2720 int ret = 0;
2721
2722 sb = elr->lr_super;
2723 ngroups = EXT4_SB(sb)->s_groups_count;
2724
2725 sb_start_write(sb);
2726 for (group = elr->lr_next_group; group < ngroups; group++) {
2727 gdp = ext4_get_group_desc(sb, group, NULL);
2728 if (!gdp) {
2729 ret = 1;
2730 break;
2731 }
2732
2733 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2734 break;
2735 }
2736
2737 if (group == ngroups)
2738 ret = 1;
2739
2740 if (!ret) {
2741 timeout = jiffies;
2742 ret = ext4_init_inode_table(sb, group,
2743 elr->lr_timeout ? 0 : 1);
2744 if (elr->lr_timeout == 0) {
2745 timeout = (jiffies - timeout) *
2746 elr->lr_sbi->s_li_wait_mult;
2747 elr->lr_timeout = timeout;
2748 }
2749 elr->lr_next_sched = jiffies + elr->lr_timeout;
2750 elr->lr_next_group = group + 1;
2751 }
2752 sb_end_write(sb);
2753
2754 return ret;
2755 }
2756
2757 /*
2758 * Remove lr_request from the list_request and free the
2759 * request structure. Should be called with li_list_mtx held
2760 */
2761 static void ext4_remove_li_request(struct ext4_li_request *elr)
2762 {
2763 struct ext4_sb_info *sbi;
2764
2765 if (!elr)
2766 return;
2767
2768 sbi = elr->lr_sbi;
2769
2770 list_del(&elr->lr_request);
2771 sbi->s_li_request = NULL;
2772 kfree(elr);
2773 }
2774
2775 static void ext4_unregister_li_request(struct super_block *sb)
2776 {
2777 mutex_lock(&ext4_li_mtx);
2778 if (!ext4_li_info) {
2779 mutex_unlock(&ext4_li_mtx);
2780 return;
2781 }
2782
2783 mutex_lock(&ext4_li_info->li_list_mtx);
2784 ext4_remove_li_request(EXT4_SB(sb)->s_li_request);
2785 mutex_unlock(&ext4_li_info->li_list_mtx);
2786 mutex_unlock(&ext4_li_mtx);
2787 }
2788
2789 static struct task_struct *ext4_lazyinit_task;
2790
2791 /*
2792 * This is the function where ext4lazyinit thread lives. It walks
2793 * through the request list searching for next scheduled filesystem.
2794 * When such a fs is found, run the lazy initialization request
2795 * (ext4_rn_li_request) and keep track of the time spend in this
2796 * function. Based on that time we compute next schedule time of
2797 * the request. When walking through the list is complete, compute
2798 * next waking time and put itself into sleep.
2799 */
2800 static int ext4_lazyinit_thread(void *arg)
2801 {
2802 struct ext4_lazy_init *eli = (struct ext4_lazy_init *)arg;
2803 struct list_head *pos, *n;
2804 struct ext4_li_request *elr;
2805 unsigned long next_wakeup, cur;
2806
2807 BUG_ON(NULL == eli);
2808
2809 cont_thread:
2810 while (true) {
2811 next_wakeup = MAX_JIFFY_OFFSET;
2812
2813 mutex_lock(&eli->li_list_mtx);
2814 if (list_empty(&eli->li_request_list)) {
2815 mutex_unlock(&eli->li_list_mtx);
2816 goto exit_thread;
2817 }
2818
2819 list_for_each_safe(pos, n, &eli->li_request_list) {
2820 elr = list_entry(pos, struct ext4_li_request,
2821 lr_request);
2822
2823 if (time_after_eq(jiffies, elr->lr_next_sched)) {
2824 if (ext4_run_li_request(elr) != 0) {
2825 /* error, remove the lazy_init job */
2826 ext4_remove_li_request(elr);
2827 continue;
2828 }
2829 }
2830
2831 if (time_before(elr->lr_next_sched, next_wakeup))
2832 next_wakeup = elr->lr_next_sched;
2833 }
2834 mutex_unlock(&eli->li_list_mtx);
2835
2836 try_to_freeze();
2837
2838 cur = jiffies;
2839 if ((time_after_eq(cur, next_wakeup)) ||
2840 (MAX_JIFFY_OFFSET == next_wakeup)) {
2841 cond_resched();
2842 continue;
2843 }
2844
2845 schedule_timeout_interruptible(next_wakeup - cur);
2846
2847 if (kthread_should_stop()) {
2848 ext4_clear_request_list();
2849 goto exit_thread;
2850 }
2851 }
2852
2853 exit_thread:
2854 /*
2855 * It looks like the request list is empty, but we need
2856 * to check it under the li_list_mtx lock, to prevent any
2857 * additions into it, and of course we should lock ext4_li_mtx
2858 * to atomically free the list and ext4_li_info, because at
2859 * this point another ext4 filesystem could be registering
2860 * new one.
2861 */
2862 mutex_lock(&ext4_li_mtx);
2863 mutex_lock(&eli->li_list_mtx);
2864 if (!list_empty(&eli->li_request_list)) {
2865 mutex_unlock(&eli->li_list_mtx);
2866 mutex_unlock(&ext4_li_mtx);
2867 goto cont_thread;
2868 }
2869 mutex_unlock(&eli->li_list_mtx);
2870 kfree(ext4_li_info);
2871 ext4_li_info = NULL;
2872 mutex_unlock(&ext4_li_mtx);
2873
2874 return 0;
2875 }
2876
2877 static void ext4_clear_request_list(void)
2878 {
2879 struct list_head *pos, *n;
2880 struct ext4_li_request *elr;
2881
2882 mutex_lock(&ext4_li_info->li_list_mtx);
2883 list_for_each_safe(pos, n, &ext4_li_info->li_request_list) {
2884 elr = list_entry(pos, struct ext4_li_request,
2885 lr_request);
2886 ext4_remove_li_request(elr);
2887 }
2888 mutex_unlock(&ext4_li_info->li_list_mtx);
2889 }
2890
2891 static int ext4_run_lazyinit_thread(void)
2892 {
2893 ext4_lazyinit_task = kthread_run(ext4_lazyinit_thread,
2894 ext4_li_info, "ext4lazyinit");
2895 if (IS_ERR(ext4_lazyinit_task)) {
2896 int err = PTR_ERR(ext4_lazyinit_task);
2897 ext4_clear_request_list();
2898 kfree(ext4_li_info);
2899 ext4_li_info = NULL;
2900 printk(KERN_CRIT "EXT4-fs: error %d creating inode table "
2901 "initialization thread\n",
2902 err);
2903 return err;
2904 }
2905 ext4_li_info->li_state |= EXT4_LAZYINIT_RUNNING;
2906 return 0;
2907 }
2908
2909 /*
2910 * Check whether it make sense to run itable init. thread or not.
2911 * If there is at least one uninitialized inode table, return
2912 * corresponding group number, else the loop goes through all
2913 * groups and return total number of groups.
2914 */
2915 static ext4_group_t ext4_has_uninit_itable(struct super_block *sb)
2916 {
2917 ext4_group_t group, ngroups = EXT4_SB(sb)->s_groups_count;
2918 struct ext4_group_desc *gdp = NULL;
2919
2920 for (group = 0; group < ngroups; group++) {
2921 gdp = ext4_get_group_desc(sb, group, NULL);
2922 if (!gdp)
2923 continue;
2924
2925 if (!(gdp->bg_flags & cpu_to_le16(EXT4_BG_INODE_ZEROED)))
2926 break;
2927 }
2928
2929 return group;
2930 }
2931
2932 static int ext4_li_info_new(void)
2933 {
2934 struct ext4_lazy_init *eli = NULL;
2935
2936 eli = kzalloc(sizeof(*eli), GFP_KERNEL);
2937 if (!eli)
2938 return -ENOMEM;
2939
2940 INIT_LIST_HEAD(&eli->li_request_list);
2941 mutex_init(&eli->li_list_mtx);
2942
2943 eli->li_state |= EXT4_LAZYINIT_QUIT;
2944
2945 ext4_li_info = eli;
2946
2947 return 0;
2948 }
2949
2950 static struct ext4_li_request *ext4_li_request_new(struct super_block *sb,
2951 ext4_group_t start)
2952 {
2953 struct ext4_sb_info *sbi = EXT4_SB(sb);
2954 struct ext4_li_request *elr;
2955 unsigned long rnd;
2956
2957 elr = kzalloc(sizeof(*elr), GFP_KERNEL);
2958 if (!elr)
2959 return NULL;
2960
2961 elr->lr_super = sb;
2962 elr->lr_sbi = sbi;
2963 elr->lr_next_group = start;
2964
2965 /*
2966 * Randomize first schedule time of the request to
2967 * spread the inode table initialization requests
2968 * better.
2969 */
2970 get_random_bytes(&rnd, sizeof(rnd));
2971 elr->lr_next_sched = jiffies + (unsigned long)rnd %
2972 (EXT4_DEF_LI_MAX_START_DELAY * HZ);
2973
2974 return elr;
2975 }
2976
2977 static int ext4_register_li_request(struct super_block *sb,
2978 ext4_group_t first_not_zeroed)
2979 {
2980 struct ext4_sb_info *sbi = EXT4_SB(sb);
2981 struct ext4_li_request *elr;
2982 ext4_group_t ngroups = EXT4_SB(sb)->s_groups_count;
2983 int ret = 0;
2984
2985 if (sbi->s_li_request != NULL) {
2986 /*
2987 * Reset timeout so it can be computed again, because
2988 * s_li_wait_mult might have changed.
2989 */
2990 sbi->s_li_request->lr_timeout = 0;
2991 return 0;
2992 }
2993
2994 if (first_not_zeroed == ngroups ||
2995 (sb->s_flags & MS_RDONLY) ||
2996 !test_opt(sb, INIT_INODE_TABLE))
2997 return 0;
2998
2999 elr = ext4_li_request_new(sb, first_not_zeroed);
3000 if (!elr)
3001 return -ENOMEM;
3002
3003 mutex_lock(&ext4_li_mtx);
3004
3005 if (NULL == ext4_li_info) {
3006 ret = ext4_li_info_new();
3007 if (ret)
3008 goto out;
3009 }
3010
3011 mutex_lock(&ext4_li_info->li_list_mtx);
3012 list_add(&elr->lr_request, &ext4_li_info->li_request_list);
3013 mutex_unlock(&ext4_li_info->li_list_mtx);
3014
3015 sbi->s_li_request = elr;
3016 /*
3017 * set elr to NULL here since it has been inserted to
3018 * the request_list and the removal and free of it is
3019 * handled by ext4_clear_request_list from now on.
3020 */
3021 elr = NULL;
3022
3023 if (!(ext4_li_info->li_state & EXT4_LAZYINIT_RUNNING)) {
3024 ret = ext4_run_lazyinit_thread();
3025 if (ret)
3026 goto out;
3027 }
3028 out:
3029 mutex_unlock(&ext4_li_mtx);
3030 if (ret)
3031 kfree(elr);
3032 return ret;
3033 }
3034
3035 /*
3036 * We do not need to lock anything since this is called on
3037 * module unload.
3038 */
3039 static void ext4_destroy_lazyinit_thread(void)
3040 {
3041 /*
3042 * If thread exited earlier
3043 * there's nothing to be done.
3044 */
3045 if (!ext4_li_info || !ext4_lazyinit_task)
3046 return;
3047
3048 kthread_stop(ext4_lazyinit_task);
3049 }
3050
3051 static int set_journal_csum_feature_set(struct super_block *sb)
3052 {
3053 int ret = 1;
3054 int compat, incompat;
3055 struct ext4_sb_info *sbi = EXT4_SB(sb);
3056
3057 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3058 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3059 /* journal checksum v2 */
3060 compat = 0;
3061 incompat = JBD2_FEATURE_INCOMPAT_CSUM_V2;
3062 } else {
3063 /* journal checksum v1 */
3064 compat = JBD2_FEATURE_COMPAT_CHECKSUM;
3065 incompat = 0;
3066 }
3067
3068 if (test_opt(sb, JOURNAL_ASYNC_COMMIT)) {
3069 ret = jbd2_journal_set_features(sbi->s_journal,
3070 compat, 0,
3071 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3072 incompat);
3073 } else if (test_opt(sb, JOURNAL_CHECKSUM)) {
3074 ret = jbd2_journal_set_features(sbi->s_journal,
3075 compat, 0,
3076 incompat);
3077 jbd2_journal_clear_features(sbi->s_journal, 0, 0,
3078 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT);
3079 } else {
3080 jbd2_journal_clear_features(sbi->s_journal,
3081 JBD2_FEATURE_COMPAT_CHECKSUM, 0,
3082 JBD2_FEATURE_INCOMPAT_ASYNC_COMMIT |
3083 JBD2_FEATURE_INCOMPAT_CSUM_V2);
3084 }
3085
3086 return ret;
3087 }
3088
3089 /*
3090 * Note: calculating the overhead so we can be compatible with
3091 * historical BSD practice is quite difficult in the face of
3092 * clusters/bigalloc. This is because multiple metadata blocks from
3093 * different block group can end up in the same allocation cluster.
3094 * Calculating the exact overhead in the face of clustered allocation
3095 * requires either O(all block bitmaps) in memory or O(number of block
3096 * groups**2) in time. We will still calculate the superblock for
3097 * older file systems --- and if we come across with a bigalloc file
3098 * system with zero in s_overhead_clusters the estimate will be close to
3099 * correct especially for very large cluster sizes --- but for newer
3100 * file systems, it's better to calculate this figure once at mkfs
3101 * time, and store it in the superblock. If the superblock value is
3102 * present (even for non-bigalloc file systems), we will use it.
3103 */
3104 static int count_overhead(struct super_block *sb, ext4_group_t grp,
3105 char *buf)
3106 {
3107 struct ext4_sb_info *sbi = EXT4_SB(sb);
3108 struct ext4_group_desc *gdp;
3109 ext4_fsblk_t first_block, last_block, b;
3110 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3111 int s, j, count = 0;
3112
3113 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_BIGALLOC))
3114 return (ext4_bg_has_super(sb, grp) + ext4_bg_num_gdb(sb, grp) +
3115 sbi->s_itb_per_group + 2);
3116
3117 first_block = le32_to_cpu(sbi->s_es->s_first_data_block) +
3118 (grp * EXT4_BLOCKS_PER_GROUP(sb));
3119 last_block = first_block + EXT4_BLOCKS_PER_GROUP(sb) - 1;
3120 for (i = 0; i < ngroups; i++) {
3121 gdp = ext4_get_group_desc(sb, i, NULL);
3122 b = ext4_block_bitmap(sb, gdp);
3123 if (b >= first_block && b <= last_block) {
3124 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3125 count++;
3126 }
3127 b = ext4_inode_bitmap(sb, gdp);
3128 if (b >= first_block && b <= last_block) {
3129 ext4_set_bit(EXT4_B2C(sbi, b - first_block), buf);
3130 count++;
3131 }
3132 b = ext4_inode_table(sb, gdp);
3133 if (b >= first_block && b + sbi->s_itb_per_group <= last_block)
3134 for (j = 0; j < sbi->s_itb_per_group; j++, b++) {
3135 int c = EXT4_B2C(sbi, b - first_block);
3136 ext4_set_bit(c, buf);
3137 count++;
3138 }
3139 if (i != grp)
3140 continue;
3141 s = 0;
3142 if (ext4_bg_has_super(sb, grp)) {
3143 ext4_set_bit(s++, buf);
3144 count++;
3145 }
3146 for (j = ext4_bg_num_gdb(sb, grp); j > 0; j--) {
3147 ext4_set_bit(EXT4_B2C(sbi, s++), buf);
3148 count++;
3149 }
3150 }
3151 if (!count)
3152 return 0;
3153 return EXT4_CLUSTERS_PER_GROUP(sb) -
3154 ext4_count_free(buf, EXT4_CLUSTERS_PER_GROUP(sb) / 8);
3155 }
3156
3157 /*
3158 * Compute the overhead and stash it in sbi->s_overhead
3159 */
3160 int ext4_calculate_overhead(struct super_block *sb)
3161 {
3162 struct ext4_sb_info *sbi = EXT4_SB(sb);
3163 struct ext4_super_block *es = sbi->s_es;
3164 ext4_group_t i, ngroups = ext4_get_groups_count(sb);
3165 ext4_fsblk_t overhead = 0;
3166 char *buf = (char *) get_zeroed_page(GFP_KERNEL);
3167
3168 memset(buf, 0, PAGE_SIZE);
3169 if (!buf)
3170 return -ENOMEM;
3171
3172 /*
3173 * Compute the overhead (FS structures). This is constant
3174 * for a given filesystem unless the number of block groups
3175 * changes so we cache the previous value until it does.
3176 */
3177
3178 /*
3179 * All of the blocks before first_data_block are overhead
3180 */
3181 overhead = EXT4_B2C(sbi, le32_to_cpu(es->s_first_data_block));
3182
3183 /*
3184 * Add the overhead found in each block group
3185 */
3186 for (i = 0; i < ngroups; i++) {
3187 int blks;
3188
3189 blks = count_overhead(sb, i, buf);
3190 overhead += blks;
3191 if (blks)
3192 memset(buf, 0, PAGE_SIZE);
3193 cond_resched();
3194 }
3195 sbi->s_overhead = overhead;
3196 smp_wmb();
3197 free_page((unsigned long) buf);
3198 return 0;
3199 }
3200
3201 static int ext4_fill_super(struct super_block *sb, void *data, int silent)
3202 {
3203 char *orig_data = kstrdup(data, GFP_KERNEL);
3204 struct buffer_head *bh;
3205 struct ext4_super_block *es = NULL;
3206 struct ext4_sb_info *sbi;
3207 ext4_fsblk_t block;
3208 ext4_fsblk_t sb_block = get_sb_block(&data);
3209 ext4_fsblk_t logical_sb_block;
3210 unsigned long offset = 0;
3211 unsigned long journal_devnum = 0;
3212 unsigned long def_mount_opts;
3213 struct inode *root;
3214 char *cp;
3215 const char *descr;
3216 int ret = -ENOMEM;
3217 int blocksize, clustersize;
3218 unsigned int db_count;
3219 unsigned int i;
3220 int needs_recovery, has_huge_files, has_bigalloc;
3221 __u64 blocks_count;
3222 int err;
3223 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
3224 ext4_group_t first_not_zeroed;
3225
3226 sbi = kzalloc(sizeof(*sbi), GFP_KERNEL);
3227 if (!sbi)
3228 goto out_free_orig;
3229
3230 sbi->s_blockgroup_lock =
3231 kzalloc(sizeof(struct blockgroup_lock), GFP_KERNEL);
3232 if (!sbi->s_blockgroup_lock) {
3233 kfree(sbi);
3234 goto out_free_orig;
3235 }
3236 sb->s_fs_info = sbi;
3237 sbi->s_sb = sb;
3238 sbi->s_mount_opt = 0;
3239 sbi->s_resuid = make_kuid(&init_user_ns, EXT4_DEF_RESUID);
3240 sbi->s_resgid = make_kgid(&init_user_ns, EXT4_DEF_RESGID);
3241 sbi->s_inode_readahead_blks = EXT4_DEF_INODE_READAHEAD_BLKS;
3242 sbi->s_sb_block = sb_block;
3243 if (sb->s_bdev->bd_part)
3244 sbi->s_sectors_written_start =
3245 part_stat_read(sb->s_bdev->bd_part, sectors[1]);
3246
3247 /* Cleanup superblock name */
3248 for (cp = sb->s_id; (cp = strchr(cp, '/'));)
3249 *cp = '!';
3250
3251 ret = -EINVAL;
3252 blocksize = sb_min_blocksize(sb, EXT4_MIN_BLOCK_SIZE);
3253 if (!blocksize) {
3254 ext4_msg(sb, KERN_ERR, "unable to set blocksize");
3255 goto out_fail;
3256 }
3257
3258 /*
3259 * The ext4 superblock will not be buffer aligned for other than 1kB
3260 * block sizes. We need to calculate the offset from buffer start.
3261 */
3262 if (blocksize != EXT4_MIN_BLOCK_SIZE) {
3263 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3264 offset = do_div(logical_sb_block, blocksize);
3265 } else {
3266 logical_sb_block = sb_block;
3267 }
3268
3269 if (!(bh = sb_bread(sb, logical_sb_block))) {
3270 ext4_msg(sb, KERN_ERR, "unable to read superblock");
3271 goto out_fail;
3272 }
3273 /*
3274 * Note: s_es must be initialized as soon as possible because
3275 * some ext4 macro-instructions depend on its value
3276 */
3277 es = (struct ext4_super_block *) (bh->b_data + offset);
3278 sbi->s_es = es;
3279 sb->s_magic = le16_to_cpu(es->s_magic);
3280 if (sb->s_magic != EXT4_SUPER_MAGIC)
3281 goto cantfind_ext4;
3282 sbi->s_kbytes_written = le64_to_cpu(es->s_kbytes_written);
3283
3284 /* Warn if metadata_csum and gdt_csum are both set. */
3285 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3286 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM) &&
3287 EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_GDT_CSUM))
3288 ext4_warning(sb, KERN_INFO "metadata_csum and uninit_bg are "
3289 "redundant flags; please run fsck.");
3290
3291 /* Check for a known checksum algorithm */
3292 if (!ext4_verify_csum_type(sb, es)) {
3293 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3294 "unknown checksum algorithm.");
3295 silent = 1;
3296 goto cantfind_ext4;
3297 }
3298
3299 /* Load the checksum driver */
3300 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3301 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM)) {
3302 sbi->s_chksum_driver = crypto_alloc_shash("crc32c", 0, 0);
3303 if (IS_ERR(sbi->s_chksum_driver)) {
3304 ext4_msg(sb, KERN_ERR, "Cannot load crc32c driver.");
3305 ret = PTR_ERR(sbi->s_chksum_driver);
3306 sbi->s_chksum_driver = NULL;
3307 goto failed_mount;
3308 }
3309 }
3310
3311 /* Check superblock checksum */
3312 if (!ext4_superblock_csum_verify(sb, es)) {
3313 ext4_msg(sb, KERN_ERR, "VFS: Found ext4 filesystem with "
3314 "invalid superblock checksum. Run e2fsck?");
3315 silent = 1;
3316 goto cantfind_ext4;
3317 }
3318
3319 /* Precompute checksum seed for all metadata */
3320 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3321 EXT4_FEATURE_RO_COMPAT_METADATA_CSUM))
3322 sbi->s_csum_seed = ext4_chksum(sbi, ~0, es->s_uuid,
3323 sizeof(es->s_uuid));
3324
3325 /* Set defaults before we parse the mount options */
3326 def_mount_opts = le32_to_cpu(es->s_default_mount_opts);
3327 set_opt(sb, INIT_INODE_TABLE);
3328 if (def_mount_opts & EXT4_DEFM_DEBUG)
3329 set_opt(sb, DEBUG);
3330 if (def_mount_opts & EXT4_DEFM_BSDGROUPS)
3331 set_opt(sb, GRPID);
3332 if (def_mount_opts & EXT4_DEFM_UID16)
3333 set_opt(sb, NO_UID32);
3334 /* xattr user namespace & acls are now defaulted on */
3335 #ifdef CONFIG_EXT4_FS_XATTR
3336 set_opt(sb, XATTR_USER);
3337 #endif
3338 #ifdef CONFIG_EXT4_FS_POSIX_ACL
3339 set_opt(sb, POSIX_ACL);
3340 #endif
3341 set_opt(sb, MBLK_IO_SUBMIT);
3342 if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_DATA)
3343 set_opt(sb, JOURNAL_DATA);
3344 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_ORDERED)
3345 set_opt(sb, ORDERED_DATA);
3346 else if ((def_mount_opts & EXT4_DEFM_JMODE) == EXT4_DEFM_JMODE_WBACK)
3347 set_opt(sb, WRITEBACK_DATA);
3348
3349 if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_PANIC)
3350 set_opt(sb, ERRORS_PANIC);
3351 else if (le16_to_cpu(sbi->s_es->s_errors) == EXT4_ERRORS_CONTINUE)
3352 set_opt(sb, ERRORS_CONT);
3353 else
3354 set_opt(sb, ERRORS_RO);
3355 if (def_mount_opts & EXT4_DEFM_BLOCK_VALIDITY)
3356 set_opt(sb, BLOCK_VALIDITY);
3357 if (def_mount_opts & EXT4_DEFM_DISCARD)
3358 set_opt(sb, DISCARD);
3359
3360 sbi->s_resuid = make_kuid(&init_user_ns, le16_to_cpu(es->s_def_resuid));
3361 sbi->s_resgid = make_kgid(&init_user_ns, le16_to_cpu(es->s_def_resgid));
3362 sbi->s_commit_interval = JBD2_DEFAULT_MAX_COMMIT_AGE * HZ;
3363 sbi->s_min_batch_time = EXT4_DEF_MIN_BATCH_TIME;
3364 sbi->s_max_batch_time = EXT4_DEF_MAX_BATCH_TIME;
3365
3366 if ((def_mount_opts & EXT4_DEFM_NOBARRIER) == 0)
3367 set_opt(sb, BARRIER);
3368
3369 /*
3370 * enable delayed allocation by default
3371 * Use -o nodelalloc to turn it off
3372 */
3373 if (!IS_EXT3_SB(sb) &&
3374 ((def_mount_opts & EXT4_DEFM_NODELALLOC) == 0))
3375 set_opt(sb, DELALLOC);
3376
3377 /*
3378 * set default s_li_wait_mult for lazyinit, for the case there is
3379 * no mount option specified.
3380 */
3381 sbi->s_li_wait_mult = EXT4_DEF_LI_WAIT_MULT;
3382
3383 if (!parse_options((char *) sbi->s_es->s_mount_opts, sb,
3384 &journal_devnum, &journal_ioprio, 0)) {
3385 ext4_msg(sb, KERN_WARNING,
3386 "failed to parse options in superblock: %s",
3387 sbi->s_es->s_mount_opts);
3388 }
3389 sbi->s_def_mount_opt = sbi->s_mount_opt;
3390 if (!parse_options((char *) data, sb, &journal_devnum,
3391 &journal_ioprio, 0))
3392 goto failed_mount;
3393
3394 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA) {
3395 printk_once(KERN_WARNING "EXT4-fs: Warning: mounting "
3396 "with data=journal disables delayed "
3397 "allocation and O_DIRECT support!\n");
3398 if (test_opt2(sb, EXPLICIT_DELALLOC)) {
3399 ext4_msg(sb, KERN_ERR, "can't mount with "
3400 "both data=journal and delalloc");
3401 goto failed_mount;
3402 }
3403 if (test_opt(sb, DIOREAD_NOLOCK)) {
3404 ext4_msg(sb, KERN_ERR, "can't mount with "
3405 "both data=journal and delalloc");
3406 goto failed_mount;
3407 }
3408 if (test_opt(sb, DELALLOC))
3409 clear_opt(sb, DELALLOC);
3410 }
3411
3412 blocksize = BLOCK_SIZE << le32_to_cpu(es->s_log_block_size);
3413 if (test_opt(sb, DIOREAD_NOLOCK)) {
3414 if (blocksize < PAGE_SIZE) {
3415 ext4_msg(sb, KERN_ERR, "can't mount with "
3416 "dioread_nolock if block size != PAGE_SIZE");
3417 goto failed_mount;
3418 }
3419 }
3420
3421 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
3422 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
3423
3424 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV &&
3425 (EXT4_HAS_COMPAT_FEATURE(sb, ~0U) ||
3426 EXT4_HAS_RO_COMPAT_FEATURE(sb, ~0U) ||
3427 EXT4_HAS_INCOMPAT_FEATURE(sb, ~0U)))
3428 ext4_msg(sb, KERN_WARNING,
3429 "feature flags set on rev 0 fs, "
3430 "running e2fsck is recommended");
3431
3432 if (IS_EXT2_SB(sb)) {
3433 if (ext2_feature_set_ok(sb))
3434 ext4_msg(sb, KERN_INFO, "mounting ext2 file system "
3435 "using the ext4 subsystem");
3436 else {
3437 ext4_msg(sb, KERN_ERR, "couldn't mount as ext2 due "
3438 "to feature incompatibilities");
3439 goto failed_mount;
3440 }
3441 }
3442
3443 if (IS_EXT3_SB(sb)) {
3444 if (ext3_feature_set_ok(sb))
3445 ext4_msg(sb, KERN_INFO, "mounting ext3 file system "
3446 "using the ext4 subsystem");
3447 else {
3448 ext4_msg(sb, KERN_ERR, "couldn't mount as ext3 due "
3449 "to feature incompatibilities");
3450 goto failed_mount;
3451 }
3452 }
3453
3454 /*
3455 * Check feature flags regardless of the revision level, since we
3456 * previously didn't change the revision level when setting the flags,
3457 * so there is a chance incompat flags are set on a rev 0 filesystem.
3458 */
3459 if (!ext4_feature_set_ok(sb, (sb->s_flags & MS_RDONLY)))
3460 goto failed_mount;
3461
3462 if (blocksize < EXT4_MIN_BLOCK_SIZE ||
3463 blocksize > EXT4_MAX_BLOCK_SIZE) {
3464 ext4_msg(sb, KERN_ERR,
3465 "Unsupported filesystem blocksize %d", blocksize);
3466 goto failed_mount;
3467 }
3468
3469 if (sb->s_blocksize != blocksize) {
3470 /* Validate the filesystem blocksize */
3471 if (!sb_set_blocksize(sb, blocksize)) {
3472 ext4_msg(sb, KERN_ERR, "bad block size %d",
3473 blocksize);
3474 goto failed_mount;
3475 }
3476
3477 brelse(bh);
3478 logical_sb_block = sb_block * EXT4_MIN_BLOCK_SIZE;
3479 offset = do_div(logical_sb_block, blocksize);
3480 bh = sb_bread(sb, logical_sb_block);
3481 if (!bh) {
3482 ext4_msg(sb, KERN_ERR,
3483 "Can't read superblock on 2nd try");
3484 goto failed_mount;
3485 }
3486 es = (struct ext4_super_block *)(bh->b_data + offset);
3487 sbi->s_es = es;
3488 if (es->s_magic != cpu_to_le16(EXT4_SUPER_MAGIC)) {
3489 ext4_msg(sb, KERN_ERR,
3490 "Magic mismatch, very weird!");
3491 goto failed_mount;
3492 }
3493 }
3494
3495 has_huge_files = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3496 EXT4_FEATURE_RO_COMPAT_HUGE_FILE);
3497 sbi->s_bitmap_maxbytes = ext4_max_bitmap_size(sb->s_blocksize_bits,
3498 has_huge_files);
3499 sb->s_maxbytes = ext4_max_size(sb->s_blocksize_bits, has_huge_files);
3500
3501 if (le32_to_cpu(es->s_rev_level) == EXT4_GOOD_OLD_REV) {
3502 sbi->s_inode_size = EXT4_GOOD_OLD_INODE_SIZE;
3503 sbi->s_first_ino = EXT4_GOOD_OLD_FIRST_INO;
3504 } else {
3505 sbi->s_inode_size = le16_to_cpu(es->s_inode_size);
3506 sbi->s_first_ino = le32_to_cpu(es->s_first_ino);
3507 if ((sbi->s_inode_size < EXT4_GOOD_OLD_INODE_SIZE) ||
3508 (!is_power_of_2(sbi->s_inode_size)) ||
3509 (sbi->s_inode_size > blocksize)) {
3510 ext4_msg(sb, KERN_ERR,
3511 "unsupported inode size: %d",
3512 sbi->s_inode_size);
3513 goto failed_mount;
3514 }
3515 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE)
3516 sb->s_time_gran = 1 << (EXT4_EPOCH_BITS - 2);
3517 }
3518
3519 sbi->s_desc_size = le16_to_cpu(es->s_desc_size);
3520 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT)) {
3521 if (sbi->s_desc_size < EXT4_MIN_DESC_SIZE_64BIT ||
3522 sbi->s_desc_size > EXT4_MAX_DESC_SIZE ||
3523 !is_power_of_2(sbi->s_desc_size)) {
3524 ext4_msg(sb, KERN_ERR,
3525 "unsupported descriptor size %lu",
3526 sbi->s_desc_size);
3527 goto failed_mount;
3528 }
3529 } else
3530 sbi->s_desc_size = EXT4_MIN_DESC_SIZE;
3531
3532 sbi->s_blocks_per_group = le32_to_cpu(es->s_blocks_per_group);
3533 sbi->s_inodes_per_group = le32_to_cpu(es->s_inodes_per_group);
3534 if (EXT4_INODE_SIZE(sb) == 0 || EXT4_INODES_PER_GROUP(sb) == 0)
3535 goto cantfind_ext4;
3536
3537 sbi->s_inodes_per_block = blocksize / EXT4_INODE_SIZE(sb);
3538 if (sbi->s_inodes_per_block == 0)
3539 goto cantfind_ext4;
3540 sbi->s_itb_per_group = sbi->s_inodes_per_group /
3541 sbi->s_inodes_per_block;
3542 sbi->s_desc_per_block = blocksize / EXT4_DESC_SIZE(sb);
3543 sbi->s_sbh = bh;
3544 sbi->s_mount_state = le16_to_cpu(es->s_state);
3545 sbi->s_addr_per_block_bits = ilog2(EXT4_ADDR_PER_BLOCK(sb));
3546 sbi->s_desc_per_block_bits = ilog2(EXT4_DESC_PER_BLOCK(sb));
3547
3548 for (i = 0; i < 4; i++)
3549 sbi->s_hash_seed[i] = le32_to_cpu(es->s_hash_seed[i]);
3550 sbi->s_def_hash_version = es->s_def_hash_version;
3551 i = le32_to_cpu(es->s_flags);
3552 if (i & EXT2_FLAGS_UNSIGNED_HASH)
3553 sbi->s_hash_unsigned = 3;
3554 else if ((i & EXT2_FLAGS_SIGNED_HASH) == 0) {
3555 #ifdef __CHAR_UNSIGNED__
3556 es->s_flags |= cpu_to_le32(EXT2_FLAGS_UNSIGNED_HASH);
3557 sbi->s_hash_unsigned = 3;
3558 #else
3559 es->s_flags |= cpu_to_le32(EXT2_FLAGS_SIGNED_HASH);
3560 #endif
3561 }
3562
3563 /* Handle clustersize */
3564 clustersize = BLOCK_SIZE << le32_to_cpu(es->s_log_cluster_size);
3565 has_bigalloc = EXT4_HAS_RO_COMPAT_FEATURE(sb,
3566 EXT4_FEATURE_RO_COMPAT_BIGALLOC);
3567 if (has_bigalloc) {
3568 if (clustersize < blocksize) {
3569 ext4_msg(sb, KERN_ERR,
3570 "cluster size (%d) smaller than "
3571 "block size (%d)", clustersize, blocksize);
3572 goto failed_mount;
3573 }
3574 sbi->s_cluster_bits = le32_to_cpu(es->s_log_cluster_size) -
3575 le32_to_cpu(es->s_log_block_size);
3576 sbi->s_clusters_per_group =
3577 le32_to_cpu(es->s_clusters_per_group);
3578 if (sbi->s_clusters_per_group > blocksize * 8) {
3579 ext4_msg(sb, KERN_ERR,
3580 "#clusters per group too big: %lu",
3581 sbi->s_clusters_per_group);
3582 goto failed_mount;
3583 }
3584 if (sbi->s_blocks_per_group !=
3585 (sbi->s_clusters_per_group * (clustersize / blocksize))) {
3586 ext4_msg(sb, KERN_ERR, "blocks per group (%lu) and "
3587 "clusters per group (%lu) inconsistent",
3588 sbi->s_blocks_per_group,
3589 sbi->s_clusters_per_group);
3590 goto failed_mount;
3591 }
3592 } else {
3593 if (clustersize != blocksize) {
3594 ext4_warning(sb, "fragment/cluster size (%d) != "
3595 "block size (%d)", clustersize,
3596 blocksize);
3597 clustersize = blocksize;
3598 }
3599 if (sbi->s_blocks_per_group > blocksize * 8) {
3600 ext4_msg(sb, KERN_ERR,
3601 "#blocks per group too big: %lu",
3602 sbi->s_blocks_per_group);
3603 goto failed_mount;
3604 }
3605 sbi->s_clusters_per_group = sbi->s_blocks_per_group;
3606 sbi->s_cluster_bits = 0;
3607 }
3608 sbi->s_cluster_ratio = clustersize / blocksize;
3609
3610 if (sbi->s_inodes_per_group > blocksize * 8) {
3611 ext4_msg(sb, KERN_ERR,
3612 "#inodes per group too big: %lu",
3613 sbi->s_inodes_per_group);
3614 goto failed_mount;
3615 }
3616
3617 /*
3618 * Test whether we have more sectors than will fit in sector_t,
3619 * and whether the max offset is addressable by the page cache.
3620 */
3621 err = generic_check_addressable(sb->s_blocksize_bits,
3622 ext4_blocks_count(es));
3623 if (err) {
3624 ext4_msg(sb, KERN_ERR, "filesystem"
3625 " too large to mount safely on this system");
3626 if (sizeof(sector_t) < 8)
3627 ext4_msg(sb, KERN_WARNING, "CONFIG_LBDAF not enabled");
3628 ret = err;
3629 goto failed_mount;
3630 }
3631
3632 if (EXT4_BLOCKS_PER_GROUP(sb) == 0)
3633 goto cantfind_ext4;
3634
3635 /* check blocks count against device size */
3636 blocks_count = sb->s_bdev->bd_inode->i_size >> sb->s_blocksize_bits;
3637 if (blocks_count && ext4_blocks_count(es) > blocks_count) {
3638 ext4_msg(sb, KERN_WARNING, "bad geometry: block count %llu "
3639 "exceeds size of device (%llu blocks)",
3640 ext4_blocks_count(es), blocks_count);
3641 goto failed_mount;
3642 }
3643
3644 /*
3645 * It makes no sense for the first data block to be beyond the end
3646 * of the filesystem.
3647 */
3648 if (le32_to_cpu(es->s_first_data_block) >= ext4_blocks_count(es)) {
3649 ext4_msg(sb, KERN_WARNING, "bad geometry: first data "
3650 "block %u is beyond end of filesystem (%llu)",
3651 le32_to_cpu(es->s_first_data_block),
3652 ext4_blocks_count(es));
3653 goto failed_mount;
3654 }
3655 blocks_count = (ext4_blocks_count(es) -
3656 le32_to_cpu(es->s_first_data_block) +
3657 EXT4_BLOCKS_PER_GROUP(sb) - 1);
3658 do_div(blocks_count, EXT4_BLOCKS_PER_GROUP(sb));
3659 if (blocks_count > ((uint64_t)1<<32) - EXT4_DESC_PER_BLOCK(sb)) {
3660 ext4_msg(sb, KERN_WARNING, "groups count too large: %u "
3661 "(block count %llu, first data block %u, "
3662 "blocks per group %lu)", sbi->s_groups_count,
3663 ext4_blocks_count(es),
3664 le32_to_cpu(es->s_first_data_block),
3665 EXT4_BLOCKS_PER_GROUP(sb));
3666 goto failed_mount;
3667 }
3668 sbi->s_groups_count = blocks_count;
3669 sbi->s_blockfile_groups = min_t(ext4_group_t, sbi->s_groups_count,
3670 (EXT4_MAX_BLOCK_FILE_PHYS / EXT4_BLOCKS_PER_GROUP(sb)));
3671 db_count = (sbi->s_groups_count + EXT4_DESC_PER_BLOCK(sb) - 1) /
3672 EXT4_DESC_PER_BLOCK(sb);
3673 sbi->s_group_desc = ext4_kvmalloc(db_count *
3674 sizeof(struct buffer_head *),
3675 GFP_KERNEL);
3676 if (sbi->s_group_desc == NULL) {
3677 ext4_msg(sb, KERN_ERR, "not enough memory");
3678 ret = -ENOMEM;
3679 goto failed_mount;
3680 }
3681
3682 if (ext4_proc_root)
3683 sbi->s_proc = proc_mkdir(sb->s_id, ext4_proc_root);
3684
3685 if (sbi->s_proc)
3686 proc_create_data("options", S_IRUGO, sbi->s_proc,
3687 &ext4_seq_options_fops, sb);
3688
3689 bgl_lock_init(sbi->s_blockgroup_lock);
3690
3691 for (i = 0; i < db_count; i++) {
3692 block = descriptor_loc(sb, logical_sb_block, i);
3693 sbi->s_group_desc[i] = sb_bread(sb, block);
3694 if (!sbi->s_group_desc[i]) {
3695 ext4_msg(sb, KERN_ERR,
3696 "can't read group descriptor %d", i);
3697 db_count = i;
3698 goto failed_mount2;
3699 }
3700 }
3701 if (!ext4_check_descriptors(sb, &first_not_zeroed)) {
3702 ext4_msg(sb, KERN_ERR, "group descriptors corrupted!");
3703 goto failed_mount2;
3704 }
3705 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_FLEX_BG))
3706 if (!ext4_fill_flex_info(sb)) {
3707 ext4_msg(sb, KERN_ERR,
3708 "unable to initialize "
3709 "flex_bg meta info!");
3710 goto failed_mount2;
3711 }
3712
3713 sbi->s_gdb_count = db_count;
3714 get_random_bytes(&sbi->s_next_generation, sizeof(u32));
3715 spin_lock_init(&sbi->s_next_gen_lock);
3716
3717 init_timer(&sbi->s_err_report);
3718 sbi->s_err_report.function = print_daily_error_info;
3719 sbi->s_err_report.data = (unsigned long) sb;
3720
3721 err = percpu_counter_init(&sbi->s_freeclusters_counter,
3722 ext4_count_free_clusters(sb));
3723 if (!err) {
3724 err = percpu_counter_init(&sbi->s_freeinodes_counter,
3725 ext4_count_free_inodes(sb));
3726 }
3727 if (!err) {
3728 err = percpu_counter_init(&sbi->s_dirs_counter,
3729 ext4_count_dirs(sb));
3730 }
3731 if (!err) {
3732 err = percpu_counter_init(&sbi->s_dirtyclusters_counter, 0);
3733 }
3734 if (err) {
3735 ext4_msg(sb, KERN_ERR, "insufficient memory");
3736 ret = err;
3737 goto failed_mount3;
3738 }
3739
3740 sbi->s_stripe = ext4_get_stripe_size(sbi);
3741 sbi->s_max_writeback_mb_bump = 128;
3742
3743 /*
3744 * set up enough so that it can read an inode
3745 */
3746 if (!test_opt(sb, NOLOAD) &&
3747 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
3748 sb->s_op = &ext4_sops;
3749 else
3750 sb->s_op = &ext4_nojournal_sops;
3751 sb->s_export_op = &ext4_export_ops;
3752 sb->s_xattr = ext4_xattr_handlers;
3753 #ifdef CONFIG_QUOTA
3754 sb->s_qcop = &ext4_qctl_operations;
3755 sb->dq_op = &ext4_quota_operations;
3756
3757 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA)) {
3758 /* Use qctl operations for hidden quota files. */
3759 sb->s_qcop = &ext4_qctl_sysfile_operations;
3760 }
3761 #endif
3762 memcpy(sb->s_uuid, es->s_uuid, sizeof(es->s_uuid));
3763
3764 INIT_LIST_HEAD(&sbi->s_orphan); /* unlinked but open files */
3765 mutex_init(&sbi->s_orphan_lock);
3766 sbi->s_resize_flags = 0;
3767
3768 sb->s_root = NULL;
3769
3770 needs_recovery = (es->s_last_orphan != 0 ||
3771 EXT4_HAS_INCOMPAT_FEATURE(sb,
3772 EXT4_FEATURE_INCOMPAT_RECOVER));
3773
3774 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_MMP) &&
3775 !(sb->s_flags & MS_RDONLY))
3776 if (ext4_multi_mount_protect(sb, le64_to_cpu(es->s_mmp_block)))
3777 goto failed_mount3;
3778
3779 /*
3780 * The first inode we look at is the journal inode. Don't try
3781 * root first: it may be modified in the journal!
3782 */
3783 if (!test_opt(sb, NOLOAD) &&
3784 EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
3785 if (ext4_load_journal(sb, es, journal_devnum))
3786 goto failed_mount3;
3787 } else if (test_opt(sb, NOLOAD) && !(sb->s_flags & MS_RDONLY) &&
3788 EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
3789 ext4_msg(sb, KERN_ERR, "required journal recovery "
3790 "suppressed and not mounted read-only");
3791 goto failed_mount_wq;
3792 } else {
3793 clear_opt(sb, DATA_FLAGS);
3794 sbi->s_journal = NULL;
3795 needs_recovery = 0;
3796 goto no_journal;
3797 }
3798
3799 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_64BIT) &&
3800 !jbd2_journal_set_features(EXT4_SB(sb)->s_journal, 0, 0,
3801 JBD2_FEATURE_INCOMPAT_64BIT)) {
3802 ext4_msg(sb, KERN_ERR, "Failed to set 64-bit journal feature");
3803 goto failed_mount_wq;
3804 }
3805
3806 if (!set_journal_csum_feature_set(sb)) {
3807 ext4_msg(sb, KERN_ERR, "Failed to set journal checksum "
3808 "feature set");
3809 goto failed_mount_wq;
3810 }
3811
3812 /* We have now updated the journal if required, so we can
3813 * validate the data journaling mode. */
3814 switch (test_opt(sb, DATA_FLAGS)) {
3815 case 0:
3816 /* No mode set, assume a default based on the journal
3817 * capabilities: ORDERED_DATA if the journal can
3818 * cope, else JOURNAL_DATA
3819 */
3820 if (jbd2_journal_check_available_features
3821 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE))
3822 set_opt(sb, ORDERED_DATA);
3823 else
3824 set_opt(sb, JOURNAL_DATA);
3825 break;
3826
3827 case EXT4_MOUNT_ORDERED_DATA:
3828 case EXT4_MOUNT_WRITEBACK_DATA:
3829 if (!jbd2_journal_check_available_features
3830 (sbi->s_journal, 0, 0, JBD2_FEATURE_INCOMPAT_REVOKE)) {
3831 ext4_msg(sb, KERN_ERR, "Journal does not support "
3832 "requested data journaling mode");
3833 goto failed_mount_wq;
3834 }
3835 default:
3836 break;
3837 }
3838 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
3839
3840 sbi->s_journal->j_commit_callback = ext4_journal_commit_callback;
3841
3842 /*
3843 * The journal may have updated the bg summary counts, so we
3844 * need to update the global counters.
3845 */
3846 percpu_counter_set(&sbi->s_freeclusters_counter,
3847 ext4_count_free_clusters(sb));
3848 percpu_counter_set(&sbi->s_freeinodes_counter,
3849 ext4_count_free_inodes(sb));
3850 percpu_counter_set(&sbi->s_dirs_counter,
3851 ext4_count_dirs(sb));
3852 percpu_counter_set(&sbi->s_dirtyclusters_counter, 0);
3853
3854 no_journal:
3855 /*
3856 * Get the # of file system overhead blocks from the
3857 * superblock if present.
3858 */
3859 if (es->s_overhead_clusters)
3860 sbi->s_overhead = le32_to_cpu(es->s_overhead_clusters);
3861 else {
3862 ret = ext4_calculate_overhead(sb);
3863 if (ret)
3864 goto failed_mount_wq;
3865 }
3866
3867 /*
3868 * The maximum number of concurrent works can be high and
3869 * concurrency isn't really necessary. Limit it to 1.
3870 */
3871 EXT4_SB(sb)->dio_unwritten_wq =
3872 alloc_workqueue("ext4-dio-unwritten", WQ_MEM_RECLAIM | WQ_UNBOUND, 1);
3873 if (!EXT4_SB(sb)->dio_unwritten_wq) {
3874 printk(KERN_ERR "EXT4-fs: failed to create DIO workqueue\n");
3875 goto failed_mount_wq;
3876 }
3877
3878 /*
3879 * The jbd2_journal_load will have done any necessary log recovery,
3880 * so we can safely mount the rest of the filesystem now.
3881 */
3882
3883 root = ext4_iget(sb, EXT4_ROOT_INO);
3884 if (IS_ERR(root)) {
3885 ext4_msg(sb, KERN_ERR, "get root inode failed");
3886 ret = PTR_ERR(root);
3887 root = NULL;
3888 goto failed_mount4;
3889 }
3890 if (!S_ISDIR(root->i_mode) || !root->i_blocks || !root->i_size) {
3891 ext4_msg(sb, KERN_ERR, "corrupt root inode, run e2fsck");
3892 iput(root);
3893 goto failed_mount4;
3894 }
3895 sb->s_root = d_make_root(root);
3896 if (!sb->s_root) {
3897 ext4_msg(sb, KERN_ERR, "get root dentry failed");
3898 ret = -ENOMEM;
3899 goto failed_mount4;
3900 }
3901
3902 if (ext4_setup_super(sb, es, sb->s_flags & MS_RDONLY))
3903 sb->s_flags |= MS_RDONLY;
3904
3905 /* determine the minimum size of new large inodes, if present */
3906 if (sbi->s_inode_size > EXT4_GOOD_OLD_INODE_SIZE) {
3907 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3908 EXT4_GOOD_OLD_INODE_SIZE;
3909 if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
3910 EXT4_FEATURE_RO_COMPAT_EXTRA_ISIZE)) {
3911 if (sbi->s_want_extra_isize <
3912 le16_to_cpu(es->s_want_extra_isize))
3913 sbi->s_want_extra_isize =
3914 le16_to_cpu(es->s_want_extra_isize);
3915 if (sbi->s_want_extra_isize <
3916 le16_to_cpu(es->s_min_extra_isize))
3917 sbi->s_want_extra_isize =
3918 le16_to_cpu(es->s_min_extra_isize);
3919 }
3920 }
3921 /* Check if enough inode space is available */
3922 if (EXT4_GOOD_OLD_INODE_SIZE + sbi->s_want_extra_isize >
3923 sbi->s_inode_size) {
3924 sbi->s_want_extra_isize = sizeof(struct ext4_inode) -
3925 EXT4_GOOD_OLD_INODE_SIZE;
3926 ext4_msg(sb, KERN_INFO, "required extra inode space not"
3927 "available");
3928 }
3929
3930 err = ext4_setup_system_zone(sb);
3931 if (err) {
3932 ext4_msg(sb, KERN_ERR, "failed to initialize system "
3933 "zone (%d)", err);
3934 goto failed_mount4a;
3935 }
3936
3937 ext4_ext_init(sb);
3938 err = ext4_mb_init(sb);
3939 if (err) {
3940 ext4_msg(sb, KERN_ERR, "failed to initialize mballoc (%d)",
3941 err);
3942 goto failed_mount5;
3943 }
3944
3945 err = ext4_register_li_request(sb, first_not_zeroed);
3946 if (err)
3947 goto failed_mount6;
3948
3949 sbi->s_kobj.kset = ext4_kset;
3950 init_completion(&sbi->s_kobj_unregister);
3951 err = kobject_init_and_add(&sbi->s_kobj, &ext4_ktype, NULL,
3952 "%s", sb->s_id);
3953 if (err)
3954 goto failed_mount7;
3955
3956 EXT4_SB(sb)->s_mount_state |= EXT4_ORPHAN_FS;
3957 ext4_orphan_cleanup(sb, es);
3958 EXT4_SB(sb)->s_mount_state &= ~EXT4_ORPHAN_FS;
3959 if (needs_recovery) {
3960 ext4_msg(sb, KERN_INFO, "recovery complete");
3961 ext4_mark_recovery_complete(sb, es);
3962 }
3963 if (EXT4_SB(sb)->s_journal) {
3964 if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_JOURNAL_DATA)
3965 descr = " journalled data mode";
3966 else if (test_opt(sb, DATA_FLAGS) == EXT4_MOUNT_ORDERED_DATA)
3967 descr = " ordered data mode";
3968 else
3969 descr = " writeback data mode";
3970 } else
3971 descr = "out journal";
3972
3973 #ifdef CONFIG_QUOTA
3974 /* Enable quota usage during mount. */
3975 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA) &&
3976 !(sb->s_flags & MS_RDONLY)) {
3977 ret = ext4_enable_quotas(sb);
3978 if (ret)
3979 goto failed_mount7;
3980 }
3981 #endif /* CONFIG_QUOTA */
3982
3983 ext4_msg(sb, KERN_INFO, "mounted filesystem with%s. "
3984 "Opts: %s%s%s", descr, sbi->s_es->s_mount_opts,
3985 *sbi->s_es->s_mount_opts ? "; " : "", orig_data);
3986
3987 if (es->s_error_count)
3988 mod_timer(&sbi->s_err_report, jiffies + 300*HZ); /* 5 minutes */
3989
3990 kfree(orig_data);
3991 return 0;
3992
3993 cantfind_ext4:
3994 if (!silent)
3995 ext4_msg(sb, KERN_ERR, "VFS: Can't find ext4 filesystem");
3996 goto failed_mount;
3997
3998 failed_mount7:
3999 ext4_unregister_li_request(sb);
4000 failed_mount6:
4001 ext4_mb_release(sb);
4002 failed_mount5:
4003 ext4_ext_release(sb);
4004 ext4_release_system_zone(sb);
4005 failed_mount4a:
4006 dput(sb->s_root);
4007 sb->s_root = NULL;
4008 failed_mount4:
4009 ext4_msg(sb, KERN_ERR, "mount failed");
4010 destroy_workqueue(EXT4_SB(sb)->dio_unwritten_wq);
4011 failed_mount_wq:
4012 if (sbi->s_journal) {
4013 jbd2_journal_destroy(sbi->s_journal);
4014 sbi->s_journal = NULL;
4015 }
4016 failed_mount3:
4017 del_timer(&sbi->s_err_report);
4018 if (sbi->s_flex_groups)
4019 ext4_kvfree(sbi->s_flex_groups);
4020 percpu_counter_destroy(&sbi->s_freeclusters_counter);
4021 percpu_counter_destroy(&sbi->s_freeinodes_counter);
4022 percpu_counter_destroy(&sbi->s_dirs_counter);
4023 percpu_counter_destroy(&sbi->s_dirtyclusters_counter);
4024 if (sbi->s_mmp_tsk)
4025 kthread_stop(sbi->s_mmp_tsk);
4026 failed_mount2:
4027 for (i = 0; i < db_count; i++)
4028 brelse(sbi->s_group_desc[i]);
4029 ext4_kvfree(sbi->s_group_desc);
4030 failed_mount:
4031 if (sbi->s_chksum_driver)
4032 crypto_free_shash(sbi->s_chksum_driver);
4033 if (sbi->s_proc) {
4034 remove_proc_entry("options", sbi->s_proc);
4035 remove_proc_entry(sb->s_id, ext4_proc_root);
4036 }
4037 #ifdef CONFIG_QUOTA
4038 for (i = 0; i < MAXQUOTAS; i++)
4039 kfree(sbi->s_qf_names[i]);
4040 #endif
4041 ext4_blkdev_remove(sbi);
4042 brelse(bh);
4043 out_fail:
4044 sb->s_fs_info = NULL;
4045 kfree(sbi->s_blockgroup_lock);
4046 kfree(sbi);
4047 out_free_orig:
4048 kfree(orig_data);
4049 return ret;
4050 }
4051
4052 /*
4053 * Setup any per-fs journal parameters now. We'll do this both on
4054 * initial mount, once the journal has been initialised but before we've
4055 * done any recovery; and again on any subsequent remount.
4056 */
4057 static void ext4_init_journal_params(struct super_block *sb, journal_t *journal)
4058 {
4059 struct ext4_sb_info *sbi = EXT4_SB(sb);
4060
4061 journal->j_commit_interval = sbi->s_commit_interval;
4062 journal->j_min_batch_time = sbi->s_min_batch_time;
4063 journal->j_max_batch_time = sbi->s_max_batch_time;
4064
4065 write_lock(&journal->j_state_lock);
4066 if (test_opt(sb, BARRIER))
4067 journal->j_flags |= JBD2_BARRIER;
4068 else
4069 journal->j_flags &= ~JBD2_BARRIER;
4070 if (test_opt(sb, DATA_ERR_ABORT))
4071 journal->j_flags |= JBD2_ABORT_ON_SYNCDATA_ERR;
4072 else
4073 journal->j_flags &= ~JBD2_ABORT_ON_SYNCDATA_ERR;
4074 write_unlock(&journal->j_state_lock);
4075 }
4076
4077 static journal_t *ext4_get_journal(struct super_block *sb,
4078 unsigned int journal_inum)
4079 {
4080 struct inode *journal_inode;
4081 journal_t *journal;
4082
4083 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4084
4085 /* First, test for the existence of a valid inode on disk. Bad
4086 * things happen if we iget() an unused inode, as the subsequent
4087 * iput() will try to delete it. */
4088
4089 journal_inode = ext4_iget(sb, journal_inum);
4090 if (IS_ERR(journal_inode)) {
4091 ext4_msg(sb, KERN_ERR, "no journal found");
4092 return NULL;
4093 }
4094 if (!journal_inode->i_nlink) {
4095 make_bad_inode(journal_inode);
4096 iput(journal_inode);
4097 ext4_msg(sb, KERN_ERR, "journal inode is deleted");
4098 return NULL;
4099 }
4100
4101 jbd_debug(2, "Journal inode found at %p: %lld bytes\n",
4102 journal_inode, journal_inode->i_size);
4103 if (!S_ISREG(journal_inode->i_mode)) {
4104 ext4_msg(sb, KERN_ERR, "invalid journal inode");
4105 iput(journal_inode);
4106 return NULL;
4107 }
4108
4109 journal = jbd2_journal_init_inode(journal_inode);
4110 if (!journal) {
4111 ext4_msg(sb, KERN_ERR, "Could not load journal inode");
4112 iput(journal_inode);
4113 return NULL;
4114 }
4115 journal->j_private = sb;
4116 ext4_init_journal_params(sb, journal);
4117 return journal;
4118 }
4119
4120 static journal_t *ext4_get_dev_journal(struct super_block *sb,
4121 dev_t j_dev)
4122 {
4123 struct buffer_head *bh;
4124 journal_t *journal;
4125 ext4_fsblk_t start;
4126 ext4_fsblk_t len;
4127 int hblock, blocksize;
4128 ext4_fsblk_t sb_block;
4129 unsigned long offset;
4130 struct ext4_super_block *es;
4131 struct block_device *bdev;
4132
4133 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4134
4135 bdev = ext4_blkdev_get(j_dev, sb);
4136 if (bdev == NULL)
4137 return NULL;
4138
4139 blocksize = sb->s_blocksize;
4140 hblock = bdev_logical_block_size(bdev);
4141 if (blocksize < hblock) {
4142 ext4_msg(sb, KERN_ERR,
4143 "blocksize too small for journal device");
4144 goto out_bdev;
4145 }
4146
4147 sb_block = EXT4_MIN_BLOCK_SIZE / blocksize;
4148 offset = EXT4_MIN_BLOCK_SIZE % blocksize;
4149 set_blocksize(bdev, blocksize);
4150 if (!(bh = __bread(bdev, sb_block, blocksize))) {
4151 ext4_msg(sb, KERN_ERR, "couldn't read superblock of "
4152 "external journal");
4153 goto out_bdev;
4154 }
4155
4156 es = (struct ext4_super_block *) (bh->b_data + offset);
4157 if ((le16_to_cpu(es->s_magic) != EXT4_SUPER_MAGIC) ||
4158 !(le32_to_cpu(es->s_feature_incompat) &
4159 EXT4_FEATURE_INCOMPAT_JOURNAL_DEV)) {
4160 ext4_msg(sb, KERN_ERR, "external journal has "
4161 "bad superblock");
4162 brelse(bh);
4163 goto out_bdev;
4164 }
4165
4166 if (memcmp(EXT4_SB(sb)->s_es->s_journal_uuid, es->s_uuid, 16)) {
4167 ext4_msg(sb, KERN_ERR, "journal UUID does not match");
4168 brelse(bh);
4169 goto out_bdev;
4170 }
4171
4172 len = ext4_blocks_count(es);
4173 start = sb_block + 1;
4174 brelse(bh); /* we're done with the superblock */
4175
4176 journal = jbd2_journal_init_dev(bdev, sb->s_bdev,
4177 start, len, blocksize);
4178 if (!journal) {
4179 ext4_msg(sb, KERN_ERR, "failed to create device journal");
4180 goto out_bdev;
4181 }
4182 journal->j_private = sb;
4183 ll_rw_block(READ, 1, &journal->j_sb_buffer);
4184 wait_on_buffer(journal->j_sb_buffer);
4185 if (!buffer_uptodate(journal->j_sb_buffer)) {
4186 ext4_msg(sb, KERN_ERR, "I/O error on journal device");
4187 goto out_journal;
4188 }
4189 if (be32_to_cpu(journal->j_superblock->s_nr_users) != 1) {
4190 ext4_msg(sb, KERN_ERR, "External journal has more than one "
4191 "user (unsupported) - %d",
4192 be32_to_cpu(journal->j_superblock->s_nr_users));
4193 goto out_journal;
4194 }
4195 EXT4_SB(sb)->journal_bdev = bdev;
4196 ext4_init_journal_params(sb, journal);
4197 return journal;
4198
4199 out_journal:
4200 jbd2_journal_destroy(journal);
4201 out_bdev:
4202 ext4_blkdev_put(bdev);
4203 return NULL;
4204 }
4205
4206 static int ext4_load_journal(struct super_block *sb,
4207 struct ext4_super_block *es,
4208 unsigned long journal_devnum)
4209 {
4210 journal_t *journal;
4211 unsigned int journal_inum = le32_to_cpu(es->s_journal_inum);
4212 dev_t journal_dev;
4213 int err = 0;
4214 int really_read_only;
4215
4216 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4217
4218 if (journal_devnum &&
4219 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4220 ext4_msg(sb, KERN_INFO, "external journal device major/minor "
4221 "numbers have changed");
4222 journal_dev = new_decode_dev(journal_devnum);
4223 } else
4224 journal_dev = new_decode_dev(le32_to_cpu(es->s_journal_dev));
4225
4226 really_read_only = bdev_read_only(sb->s_bdev);
4227
4228 /*
4229 * Are we loading a blank journal or performing recovery after a
4230 * crash? For recovery, we need to check in advance whether we
4231 * can get read-write access to the device.
4232 */
4233 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER)) {
4234 if (sb->s_flags & MS_RDONLY) {
4235 ext4_msg(sb, KERN_INFO, "INFO: recovery "
4236 "required on readonly filesystem");
4237 if (really_read_only) {
4238 ext4_msg(sb, KERN_ERR, "write access "
4239 "unavailable, cannot proceed");
4240 return -EROFS;
4241 }
4242 ext4_msg(sb, KERN_INFO, "write access will "
4243 "be enabled during recovery");
4244 }
4245 }
4246
4247 if (journal_inum && journal_dev) {
4248 ext4_msg(sb, KERN_ERR, "filesystem has both journal "
4249 "and inode journals!");
4250 return -EINVAL;
4251 }
4252
4253 if (journal_inum) {
4254 if (!(journal = ext4_get_journal(sb, journal_inum)))
4255 return -EINVAL;
4256 } else {
4257 if (!(journal = ext4_get_dev_journal(sb, journal_dev)))
4258 return -EINVAL;
4259 }
4260
4261 if (!(journal->j_flags & JBD2_BARRIER))
4262 ext4_msg(sb, KERN_INFO, "barriers disabled");
4263
4264 if (!EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER))
4265 err = jbd2_journal_wipe(journal, !really_read_only);
4266 if (!err) {
4267 char *save = kmalloc(EXT4_S_ERR_LEN, GFP_KERNEL);
4268 if (save)
4269 memcpy(save, ((char *) es) +
4270 EXT4_S_ERR_START, EXT4_S_ERR_LEN);
4271 err = jbd2_journal_load(journal);
4272 if (save)
4273 memcpy(((char *) es) + EXT4_S_ERR_START,
4274 save, EXT4_S_ERR_LEN);
4275 kfree(save);
4276 }
4277
4278 if (err) {
4279 ext4_msg(sb, KERN_ERR, "error loading journal");
4280 jbd2_journal_destroy(journal);
4281 return err;
4282 }
4283
4284 EXT4_SB(sb)->s_journal = journal;
4285 ext4_clear_journal_err(sb, es);
4286
4287 if (!really_read_only && journal_devnum &&
4288 journal_devnum != le32_to_cpu(es->s_journal_dev)) {
4289 es->s_journal_dev = cpu_to_le32(journal_devnum);
4290
4291 /* Make sure we flush the recovery flag to disk. */
4292 ext4_commit_super(sb, 1);
4293 }
4294
4295 return 0;
4296 }
4297
4298 static int ext4_commit_super(struct super_block *sb, int sync)
4299 {
4300 struct ext4_super_block *es = EXT4_SB(sb)->s_es;
4301 struct buffer_head *sbh = EXT4_SB(sb)->s_sbh;
4302 int error = 0;
4303
4304 if (!sbh || block_device_ejected(sb))
4305 return error;
4306 if (buffer_write_io_error(sbh)) {
4307 /*
4308 * Oh, dear. A previous attempt to write the
4309 * superblock failed. This could happen because the
4310 * USB device was yanked out. Or it could happen to
4311 * be a transient write error and maybe the block will
4312 * be remapped. Nothing we can do but to retry the
4313 * write and hope for the best.
4314 */
4315 ext4_msg(sb, KERN_ERR, "previous I/O error to "
4316 "superblock detected");
4317 clear_buffer_write_io_error(sbh);
4318 set_buffer_uptodate(sbh);
4319 }
4320 /*
4321 * If the file system is mounted read-only, don't update the
4322 * superblock write time. This avoids updating the superblock
4323 * write time when we are mounting the root file system
4324 * read/only but we need to replay the journal; at that point,
4325 * for people who are east of GMT and who make their clock
4326 * tick in localtime for Windows bug-for-bug compatibility,
4327 * the clock is set in the future, and this will cause e2fsck
4328 * to complain and force a full file system check.
4329 */
4330 if (!(sb->s_flags & MS_RDONLY))
4331 es->s_wtime = cpu_to_le32(get_seconds());
4332 if (sb->s_bdev->bd_part)
4333 es->s_kbytes_written =
4334 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written +
4335 ((part_stat_read(sb->s_bdev->bd_part, sectors[1]) -
4336 EXT4_SB(sb)->s_sectors_written_start) >> 1));
4337 else
4338 es->s_kbytes_written =
4339 cpu_to_le64(EXT4_SB(sb)->s_kbytes_written);
4340 ext4_free_blocks_count_set(es,
4341 EXT4_C2B(EXT4_SB(sb), percpu_counter_sum_positive(
4342 &EXT4_SB(sb)->s_freeclusters_counter)));
4343 es->s_free_inodes_count =
4344 cpu_to_le32(percpu_counter_sum_positive(
4345 &EXT4_SB(sb)->s_freeinodes_counter));
4346 BUFFER_TRACE(sbh, "marking dirty");
4347 ext4_superblock_csum_set(sb);
4348 mark_buffer_dirty(sbh);
4349 if (sync) {
4350 error = sync_dirty_buffer(sbh);
4351 if (error)
4352 return error;
4353
4354 error = buffer_write_io_error(sbh);
4355 if (error) {
4356 ext4_msg(sb, KERN_ERR, "I/O error while writing "
4357 "superblock");
4358 clear_buffer_write_io_error(sbh);
4359 set_buffer_uptodate(sbh);
4360 }
4361 }
4362 return error;
4363 }
4364
4365 /*
4366 * Have we just finished recovery? If so, and if we are mounting (or
4367 * remounting) the filesystem readonly, then we will end up with a
4368 * consistent fs on disk. Record that fact.
4369 */
4370 static void ext4_mark_recovery_complete(struct super_block *sb,
4371 struct ext4_super_block *es)
4372 {
4373 journal_t *journal = EXT4_SB(sb)->s_journal;
4374
4375 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL)) {
4376 BUG_ON(journal != NULL);
4377 return;
4378 }
4379 jbd2_journal_lock_updates(journal);
4380 if (jbd2_journal_flush(journal) < 0)
4381 goto out;
4382
4383 if (EXT4_HAS_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER) &&
4384 sb->s_flags & MS_RDONLY) {
4385 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4386 ext4_commit_super(sb, 1);
4387 }
4388
4389 out:
4390 jbd2_journal_unlock_updates(journal);
4391 }
4392
4393 /*
4394 * If we are mounting (or read-write remounting) a filesystem whose journal
4395 * has recorded an error from a previous lifetime, move that error to the
4396 * main filesystem now.
4397 */
4398 static void ext4_clear_journal_err(struct super_block *sb,
4399 struct ext4_super_block *es)
4400 {
4401 journal_t *journal;
4402 int j_errno;
4403 const char *errstr;
4404
4405 BUG_ON(!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL));
4406
4407 journal = EXT4_SB(sb)->s_journal;
4408
4409 /*
4410 * Now check for any error status which may have been recorded in the
4411 * journal by a prior ext4_error() or ext4_abort()
4412 */
4413
4414 j_errno = jbd2_journal_errno(journal);
4415 if (j_errno) {
4416 char nbuf[16];
4417
4418 errstr = ext4_decode_error(sb, j_errno, nbuf);
4419 ext4_warning(sb, "Filesystem error recorded "
4420 "from previous mount: %s", errstr);
4421 ext4_warning(sb, "Marking fs in need of filesystem check.");
4422
4423 EXT4_SB(sb)->s_mount_state |= EXT4_ERROR_FS;
4424 es->s_state |= cpu_to_le16(EXT4_ERROR_FS);
4425 ext4_commit_super(sb, 1);
4426
4427 jbd2_journal_clear_err(journal);
4428 jbd2_journal_update_sb_errno(journal);
4429 }
4430 }
4431
4432 /*
4433 * Force the running and committing transactions to commit,
4434 * and wait on the commit.
4435 */
4436 int ext4_force_commit(struct super_block *sb)
4437 {
4438 journal_t *journal;
4439 int ret = 0;
4440
4441 if (sb->s_flags & MS_RDONLY)
4442 return 0;
4443
4444 journal = EXT4_SB(sb)->s_journal;
4445 if (journal)
4446 ret = ext4_journal_force_commit(journal);
4447
4448 return ret;
4449 }
4450
4451 static int ext4_sync_fs(struct super_block *sb, int wait)
4452 {
4453 int ret = 0;
4454 tid_t target;
4455 struct ext4_sb_info *sbi = EXT4_SB(sb);
4456
4457 trace_ext4_sync_fs(sb, wait);
4458 flush_workqueue(sbi->dio_unwritten_wq);
4459 /*
4460 * Writeback quota in non-journalled quota case - journalled quota has
4461 * no dirty dquots
4462 */
4463 dquot_writeback_dquots(sb, -1);
4464 if (jbd2_journal_start_commit(sbi->s_journal, &target)) {
4465 if (wait)
4466 jbd2_log_wait_commit(sbi->s_journal, target);
4467 }
4468 return ret;
4469 }
4470
4471 /*
4472 * LVM calls this function before a (read-only) snapshot is created. This
4473 * gives us a chance to flush the journal completely and mark the fs clean.
4474 *
4475 * Note that only this function cannot bring a filesystem to be in a clean
4476 * state independently. It relies on upper layer to stop all data & metadata
4477 * modifications.
4478 */
4479 static int ext4_freeze(struct super_block *sb)
4480 {
4481 int error = 0;
4482 journal_t *journal;
4483
4484 if (sb->s_flags & MS_RDONLY)
4485 return 0;
4486
4487 journal = EXT4_SB(sb)->s_journal;
4488
4489 /* Now we set up the journal barrier. */
4490 jbd2_journal_lock_updates(journal);
4491
4492 /*
4493 * Don't clear the needs_recovery flag if we failed to flush
4494 * the journal.
4495 */
4496 error = jbd2_journal_flush(journal);
4497 if (error < 0)
4498 goto out;
4499
4500 /* Journal blocked and flushed, clear needs_recovery flag. */
4501 EXT4_CLEAR_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4502 error = ext4_commit_super(sb, 1);
4503 out:
4504 /* we rely on upper layer to stop further updates */
4505 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4506 return error;
4507 }
4508
4509 /*
4510 * Called by LVM after the snapshot is done. We need to reset the RECOVER
4511 * flag here, even though the filesystem is not technically dirty yet.
4512 */
4513 static int ext4_unfreeze(struct super_block *sb)
4514 {
4515 if (sb->s_flags & MS_RDONLY)
4516 return 0;
4517
4518 lock_super(sb);
4519 /* Reset the needs_recovery flag before the fs is unlocked. */
4520 EXT4_SET_INCOMPAT_FEATURE(sb, EXT4_FEATURE_INCOMPAT_RECOVER);
4521 ext4_commit_super(sb, 1);
4522 unlock_super(sb);
4523 return 0;
4524 }
4525
4526 /*
4527 * Structure to save mount options for ext4_remount's benefit
4528 */
4529 struct ext4_mount_options {
4530 unsigned long s_mount_opt;
4531 unsigned long s_mount_opt2;
4532 kuid_t s_resuid;
4533 kgid_t s_resgid;
4534 unsigned long s_commit_interval;
4535 u32 s_min_batch_time, s_max_batch_time;
4536 #ifdef CONFIG_QUOTA
4537 int s_jquota_fmt;
4538 char *s_qf_names[MAXQUOTAS];
4539 #endif
4540 };
4541
4542 static int ext4_remount(struct super_block *sb, int *flags, char *data)
4543 {
4544 struct ext4_super_block *es;
4545 struct ext4_sb_info *sbi = EXT4_SB(sb);
4546 unsigned long old_sb_flags;
4547 struct ext4_mount_options old_opts;
4548 int enable_quota = 0;
4549 ext4_group_t g;
4550 unsigned int journal_ioprio = DEFAULT_JOURNAL_IOPRIO;
4551 int err = 0;
4552 #ifdef CONFIG_QUOTA
4553 int i;
4554 #endif
4555 char *orig_data = kstrdup(data, GFP_KERNEL);
4556
4557 /* Store the original options */
4558 lock_super(sb);
4559 old_sb_flags = sb->s_flags;
4560 old_opts.s_mount_opt = sbi->s_mount_opt;
4561 old_opts.s_mount_opt2 = sbi->s_mount_opt2;
4562 old_opts.s_resuid = sbi->s_resuid;
4563 old_opts.s_resgid = sbi->s_resgid;
4564 old_opts.s_commit_interval = sbi->s_commit_interval;
4565 old_opts.s_min_batch_time = sbi->s_min_batch_time;
4566 old_opts.s_max_batch_time = sbi->s_max_batch_time;
4567 #ifdef CONFIG_QUOTA
4568 old_opts.s_jquota_fmt = sbi->s_jquota_fmt;
4569 for (i = 0; i < MAXQUOTAS; i++)
4570 old_opts.s_qf_names[i] = sbi->s_qf_names[i];
4571 #endif
4572 if (sbi->s_journal && sbi->s_journal->j_task->io_context)
4573 journal_ioprio = sbi->s_journal->j_task->io_context->ioprio;
4574
4575 /*
4576 * Allow the "check" option to be passed as a remount option.
4577 */
4578 if (!parse_options(data, sb, NULL, &journal_ioprio, 1)) {
4579 err = -EINVAL;
4580 goto restore_opts;
4581 }
4582
4583 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED)
4584 ext4_abort(sb, "Abort forced by user");
4585
4586 sb->s_flags = (sb->s_flags & ~MS_POSIXACL) |
4587 (test_opt(sb, POSIX_ACL) ? MS_POSIXACL : 0);
4588
4589 es = sbi->s_es;
4590
4591 if (sbi->s_journal) {
4592 ext4_init_journal_params(sb, sbi->s_journal);
4593 set_task_ioprio(sbi->s_journal->j_task, journal_ioprio);
4594 }
4595
4596 if ((*flags & MS_RDONLY) != (sb->s_flags & MS_RDONLY)) {
4597 if (sbi->s_mount_flags & EXT4_MF_FS_ABORTED) {
4598 err = -EROFS;
4599 goto restore_opts;
4600 }
4601
4602 if (*flags & MS_RDONLY) {
4603 err = dquot_suspend(sb, -1);
4604 if (err < 0)
4605 goto restore_opts;
4606
4607 /*
4608 * First of all, the unconditional stuff we have to do
4609 * to disable replay of the journal when we next remount
4610 */
4611 sb->s_flags |= MS_RDONLY;
4612
4613 /*
4614 * OK, test if we are remounting a valid rw partition
4615 * readonly, and if so set the rdonly flag and then
4616 * mark the partition as valid again.
4617 */
4618 if (!(es->s_state & cpu_to_le16(EXT4_VALID_FS)) &&
4619 (sbi->s_mount_state & EXT4_VALID_FS))
4620 es->s_state = cpu_to_le16(sbi->s_mount_state);
4621
4622 if (sbi->s_journal)
4623 ext4_mark_recovery_complete(sb, es);
4624 } else {
4625 /* Make sure we can mount this feature set readwrite */
4626 if (!ext4_feature_set_ok(sb, 0)) {
4627 err = -EROFS;
4628 goto restore_opts;
4629 }
4630 /*
4631 * Make sure the group descriptor checksums
4632 * are sane. If they aren't, refuse to remount r/w.
4633 */
4634 for (g = 0; g < sbi->s_groups_count; g++) {
4635 struct ext4_group_desc *gdp =
4636 ext4_get_group_desc(sb, g, NULL);
4637
4638 if (!ext4_group_desc_csum_verify(sb, g, gdp)) {
4639 ext4_msg(sb, KERN_ERR,
4640 "ext4_remount: Checksum for group %u failed (%u!=%u)",
4641 g, le16_to_cpu(ext4_group_desc_csum(sbi, g, gdp)),
4642 le16_to_cpu(gdp->bg_checksum));
4643 err = -EINVAL;
4644 goto restore_opts;
4645 }
4646 }
4647
4648 /*
4649 * If we have an unprocessed orphan list hanging
4650 * around from a previously readonly bdev mount,
4651 * require a full umount/remount for now.
4652 */
4653 if (es->s_last_orphan) {
4654 ext4_msg(sb, KERN_WARNING, "Couldn't "
4655 "remount RDWR because of unprocessed "
4656 "orphan inode list. Please "
4657 "umount/remount instead");
4658 err = -EINVAL;
4659 goto restore_opts;
4660 }
4661
4662 /*
4663 * Mounting a RDONLY partition read-write, so reread
4664 * and store the current valid flag. (It may have
4665 * been changed by e2fsck since we originally mounted
4666 * the partition.)
4667 */
4668 if (sbi->s_journal)
4669 ext4_clear_journal_err(sb, es);
4670 sbi->s_mount_state = le16_to_cpu(es->s_state);
4671 if (!ext4_setup_super(sb, es, 0))
4672 sb->s_flags &= ~MS_RDONLY;
4673 if (EXT4_HAS_INCOMPAT_FEATURE(sb,
4674 EXT4_FEATURE_INCOMPAT_MMP))
4675 if (ext4_multi_mount_protect(sb,
4676 le64_to_cpu(es->s_mmp_block))) {
4677 err = -EROFS;
4678 goto restore_opts;
4679 }
4680 enable_quota = 1;
4681 }
4682 }
4683
4684 /*
4685 * Reinitialize lazy itable initialization thread based on
4686 * current settings
4687 */
4688 if ((sb->s_flags & MS_RDONLY) || !test_opt(sb, INIT_INODE_TABLE))
4689 ext4_unregister_li_request(sb);
4690 else {
4691 ext4_group_t first_not_zeroed;
4692 first_not_zeroed = ext4_has_uninit_itable(sb);
4693 ext4_register_li_request(sb, first_not_zeroed);
4694 }
4695
4696 ext4_setup_system_zone(sb);
4697 if (sbi->s_journal == NULL)
4698 ext4_commit_super(sb, 1);
4699
4700 unlock_super(sb);
4701 #ifdef CONFIG_QUOTA
4702 /* Release old quota file names */
4703 for (i = 0; i < MAXQUOTAS; i++)
4704 if (old_opts.s_qf_names[i] &&
4705 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4706 kfree(old_opts.s_qf_names[i]);
4707 if (enable_quota) {
4708 if (sb_any_quota_suspended(sb))
4709 dquot_resume(sb, -1);
4710 else if (EXT4_HAS_RO_COMPAT_FEATURE(sb,
4711 EXT4_FEATURE_RO_COMPAT_QUOTA)) {
4712 err = ext4_enable_quotas(sb);
4713 if (err) {
4714 lock_super(sb);
4715 goto restore_opts;
4716 }
4717 }
4718 }
4719 #endif
4720
4721 ext4_msg(sb, KERN_INFO, "re-mounted. Opts: %s", orig_data);
4722 kfree(orig_data);
4723 return 0;
4724
4725 restore_opts:
4726 sb->s_flags = old_sb_flags;
4727 sbi->s_mount_opt = old_opts.s_mount_opt;
4728 sbi->s_mount_opt2 = old_opts.s_mount_opt2;
4729 sbi->s_resuid = old_opts.s_resuid;
4730 sbi->s_resgid = old_opts.s_resgid;
4731 sbi->s_commit_interval = old_opts.s_commit_interval;
4732 sbi->s_min_batch_time = old_opts.s_min_batch_time;
4733 sbi->s_max_batch_time = old_opts.s_max_batch_time;
4734 #ifdef CONFIG_QUOTA
4735 sbi->s_jquota_fmt = old_opts.s_jquota_fmt;
4736 for (i = 0; i < MAXQUOTAS; i++) {
4737 if (sbi->s_qf_names[i] &&
4738 old_opts.s_qf_names[i] != sbi->s_qf_names[i])
4739 kfree(sbi->s_qf_names[i]);
4740 sbi->s_qf_names[i] = old_opts.s_qf_names[i];
4741 }
4742 #endif
4743 unlock_super(sb);
4744 kfree(orig_data);
4745 return err;
4746 }
4747
4748 static int ext4_statfs(struct dentry *dentry, struct kstatfs *buf)
4749 {
4750 struct super_block *sb = dentry->d_sb;
4751 struct ext4_sb_info *sbi = EXT4_SB(sb);
4752 struct ext4_super_block *es = sbi->s_es;
4753 ext4_fsblk_t overhead = 0;
4754 u64 fsid;
4755 s64 bfree;
4756
4757 if (!test_opt(sb, MINIX_DF))
4758 overhead = sbi->s_overhead;
4759
4760 buf->f_type = EXT4_SUPER_MAGIC;
4761 buf->f_bsize = sb->s_blocksize;
4762 buf->f_blocks = ext4_blocks_count(es) - EXT4_C2B(sbi, sbi->s_overhead);
4763 bfree = percpu_counter_sum_positive(&sbi->s_freeclusters_counter) -
4764 percpu_counter_sum_positive(&sbi->s_dirtyclusters_counter);
4765 /* prevent underflow in case that few free space is available */
4766 buf->f_bfree = EXT4_C2B(sbi, max_t(s64, bfree, 0));
4767 buf->f_bavail = buf->f_bfree - ext4_r_blocks_count(es);
4768 if (buf->f_bfree < ext4_r_blocks_count(es))
4769 buf->f_bavail = 0;
4770 buf->f_files = le32_to_cpu(es->s_inodes_count);
4771 buf->f_ffree = percpu_counter_sum_positive(&sbi->s_freeinodes_counter);
4772 buf->f_namelen = EXT4_NAME_LEN;
4773 fsid = le64_to_cpup((void *)es->s_uuid) ^
4774 le64_to_cpup((void *)es->s_uuid + sizeof(u64));
4775 buf->f_fsid.val[0] = fsid & 0xFFFFFFFFUL;
4776 buf->f_fsid.val[1] = (fsid >> 32) & 0xFFFFFFFFUL;
4777
4778 return 0;
4779 }
4780
4781 /* Helper function for writing quotas on sync - we need to start transaction
4782 * before quota file is locked for write. Otherwise the are possible deadlocks:
4783 * Process 1 Process 2
4784 * ext4_create() quota_sync()
4785 * jbd2_journal_start() write_dquot()
4786 * dquot_initialize() down(dqio_mutex)
4787 * down(dqio_mutex) jbd2_journal_start()
4788 *
4789 */
4790
4791 #ifdef CONFIG_QUOTA
4792
4793 static inline struct inode *dquot_to_inode(struct dquot *dquot)
4794 {
4795 return sb_dqopt(dquot->dq_sb)->files[dquot->dq_type];
4796 }
4797
4798 static int ext4_write_dquot(struct dquot *dquot)
4799 {
4800 int ret, err;
4801 handle_t *handle;
4802 struct inode *inode;
4803
4804 inode = dquot_to_inode(dquot);
4805 handle = ext4_journal_start(inode,
4806 EXT4_QUOTA_TRANS_BLOCKS(dquot->dq_sb));
4807 if (IS_ERR(handle))
4808 return PTR_ERR(handle);
4809 ret = dquot_commit(dquot);
4810 err = ext4_journal_stop(handle);
4811 if (!ret)
4812 ret = err;
4813 return ret;
4814 }
4815
4816 static int ext4_acquire_dquot(struct dquot *dquot)
4817 {
4818 int ret, err;
4819 handle_t *handle;
4820
4821 handle = ext4_journal_start(dquot_to_inode(dquot),
4822 EXT4_QUOTA_INIT_BLOCKS(dquot->dq_sb));
4823 if (IS_ERR(handle))
4824 return PTR_ERR(handle);
4825 ret = dquot_acquire(dquot);
4826 err = ext4_journal_stop(handle);
4827 if (!ret)
4828 ret = err;
4829 return ret;
4830 }
4831
4832 static int ext4_release_dquot(struct dquot *dquot)
4833 {
4834 int ret, err;
4835 handle_t *handle;
4836
4837 handle = ext4_journal_start(dquot_to_inode(dquot),
4838 EXT4_QUOTA_DEL_BLOCKS(dquot->dq_sb));
4839 if (IS_ERR(handle)) {
4840 /* Release dquot anyway to avoid endless cycle in dqput() */
4841 dquot_release(dquot);
4842 return PTR_ERR(handle);
4843 }
4844 ret = dquot_release(dquot);
4845 err = ext4_journal_stop(handle);
4846 if (!ret)
4847 ret = err;
4848 return ret;
4849 }
4850
4851 static int ext4_mark_dquot_dirty(struct dquot *dquot)
4852 {
4853 /* Are we journaling quotas? */
4854 if (EXT4_SB(dquot->dq_sb)->s_qf_names[USRQUOTA] ||
4855 EXT4_SB(dquot->dq_sb)->s_qf_names[GRPQUOTA]) {
4856 dquot_mark_dquot_dirty(dquot);
4857 return ext4_write_dquot(dquot);
4858 } else {
4859 return dquot_mark_dquot_dirty(dquot);
4860 }
4861 }
4862
4863 static int ext4_write_info(struct super_block *sb, int type)
4864 {
4865 int ret, err;
4866 handle_t *handle;
4867
4868 /* Data block + inode block */
4869 handle = ext4_journal_start(sb->s_root->d_inode, 2);
4870 if (IS_ERR(handle))
4871 return PTR_ERR(handle);
4872 ret = dquot_commit_info(sb, type);
4873 err = ext4_journal_stop(handle);
4874 if (!ret)
4875 ret = err;
4876 return ret;
4877 }
4878
4879 /*
4880 * Turn on quotas during mount time - we need to find
4881 * the quota file and such...
4882 */
4883 static int ext4_quota_on_mount(struct super_block *sb, int type)
4884 {
4885 return dquot_quota_on_mount(sb, EXT4_SB(sb)->s_qf_names[type],
4886 EXT4_SB(sb)->s_jquota_fmt, type);
4887 }
4888
4889 /*
4890 * Standard function to be called on quota_on
4891 */
4892 static int ext4_quota_on(struct super_block *sb, int type, int format_id,
4893 struct path *path)
4894 {
4895 int err;
4896
4897 if (!test_opt(sb, QUOTA))
4898 return -EINVAL;
4899
4900 /* Quotafile not on the same filesystem? */
4901 if (path->dentry->d_sb != sb)
4902 return -EXDEV;
4903 /* Journaling quota? */
4904 if (EXT4_SB(sb)->s_qf_names[type]) {
4905 /* Quotafile not in fs root? */
4906 if (path->dentry->d_parent != sb->s_root)
4907 ext4_msg(sb, KERN_WARNING,
4908 "Quota file not on filesystem root. "
4909 "Journaled quota will not work");
4910 }
4911
4912 /*
4913 * When we journal data on quota file, we have to flush journal to see
4914 * all updates to the file when we bypass pagecache...
4915 */
4916 if (EXT4_SB(sb)->s_journal &&
4917 ext4_should_journal_data(path->dentry->d_inode)) {
4918 /*
4919 * We don't need to lock updates but journal_flush() could
4920 * otherwise be livelocked...
4921 */
4922 jbd2_journal_lock_updates(EXT4_SB(sb)->s_journal);
4923 err = jbd2_journal_flush(EXT4_SB(sb)->s_journal);
4924 jbd2_journal_unlock_updates(EXT4_SB(sb)->s_journal);
4925 if (err)
4926 return err;
4927 }
4928
4929 return dquot_quota_on(sb, type, format_id, path);
4930 }
4931
4932 static int ext4_quota_enable(struct super_block *sb, int type, int format_id,
4933 unsigned int flags)
4934 {
4935 int err;
4936 struct inode *qf_inode;
4937 unsigned long qf_inums[MAXQUOTAS] = {
4938 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
4939 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
4940 };
4941
4942 BUG_ON(!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA));
4943
4944 if (!qf_inums[type])
4945 return -EPERM;
4946
4947 qf_inode = ext4_iget(sb, qf_inums[type]);
4948 if (IS_ERR(qf_inode)) {
4949 ext4_error(sb, "Bad quota inode # %lu", qf_inums[type]);
4950 return PTR_ERR(qf_inode);
4951 }
4952
4953 err = dquot_enable(qf_inode, type, format_id, flags);
4954 iput(qf_inode);
4955
4956 return err;
4957 }
4958
4959 /* Enable usage tracking for all quota types. */
4960 static int ext4_enable_quotas(struct super_block *sb)
4961 {
4962 int type, err = 0;
4963 unsigned long qf_inums[MAXQUOTAS] = {
4964 le32_to_cpu(EXT4_SB(sb)->s_es->s_usr_quota_inum),
4965 le32_to_cpu(EXT4_SB(sb)->s_es->s_grp_quota_inum)
4966 };
4967
4968 sb_dqopt(sb)->flags |= DQUOT_QUOTA_SYS_FILE;
4969 for (type = 0; type < MAXQUOTAS; type++) {
4970 if (qf_inums[type]) {
4971 err = ext4_quota_enable(sb, type, QFMT_VFS_V1,
4972 DQUOT_USAGE_ENABLED);
4973 if (err) {
4974 ext4_warning(sb,
4975 "Failed to enable quota (type=%d) "
4976 "tracking. Please run e2fsck to fix.",
4977 type);
4978 return err;
4979 }
4980 }
4981 }
4982 return 0;
4983 }
4984
4985 /*
4986 * quota_on function that is used when QUOTA feature is set.
4987 */
4988 static int ext4_quota_on_sysfile(struct super_block *sb, int type,
4989 int format_id)
4990 {
4991 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
4992 return -EINVAL;
4993
4994 /*
4995 * USAGE was enabled at mount time. Only need to enable LIMITS now.
4996 */
4997 return ext4_quota_enable(sb, type, format_id, DQUOT_LIMITS_ENABLED);
4998 }
4999
5000 static int ext4_quota_off(struct super_block *sb, int type)
5001 {
5002 struct inode *inode = sb_dqopt(sb)->files[type];
5003 handle_t *handle;
5004
5005 /* Force all delayed allocation blocks to be allocated.
5006 * Caller already holds s_umount sem */
5007 if (test_opt(sb, DELALLOC))
5008 sync_filesystem(sb);
5009
5010 if (!inode)
5011 goto out;
5012
5013 /* Update modification times of quota files when userspace can
5014 * start looking at them */
5015 handle = ext4_journal_start(inode, 1);
5016 if (IS_ERR(handle))
5017 goto out;
5018 inode->i_mtime = inode->i_ctime = CURRENT_TIME;
5019 ext4_mark_inode_dirty(handle, inode);
5020 ext4_journal_stop(handle);
5021
5022 out:
5023 return dquot_quota_off(sb, type);
5024 }
5025
5026 /*
5027 * quota_off function that is used when QUOTA feature is set.
5028 */
5029 static int ext4_quota_off_sysfile(struct super_block *sb, int type)
5030 {
5031 if (!EXT4_HAS_RO_COMPAT_FEATURE(sb, EXT4_FEATURE_RO_COMPAT_QUOTA))
5032 return -EINVAL;
5033
5034 /* Disable only the limits. */
5035 return dquot_disable(sb, type, DQUOT_LIMITS_ENABLED);
5036 }
5037
5038 /* Read data from quotafile - avoid pagecache and such because we cannot afford
5039 * acquiring the locks... As quota files are never truncated and quota code
5040 * itself serializes the operations (and no one else should touch the files)
5041 * we don't have to be afraid of races */
5042 static ssize_t ext4_quota_read(struct super_block *sb, int type, char *data,
5043 size_t len, loff_t off)
5044 {
5045 struct inode *inode = sb_dqopt(sb)->files[type];
5046 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5047 int err = 0;
5048 int offset = off & (sb->s_blocksize - 1);
5049 int tocopy;
5050 size_t toread;
5051 struct buffer_head *bh;
5052 loff_t i_size = i_size_read(inode);
5053
5054 if (off > i_size)
5055 return 0;
5056 if (off+len > i_size)
5057 len = i_size-off;
5058 toread = len;
5059 while (toread > 0) {
5060 tocopy = sb->s_blocksize - offset < toread ?
5061 sb->s_blocksize - offset : toread;
5062 bh = ext4_bread(NULL, inode, blk, 0, &err);
5063 if (err)
5064 return err;
5065 if (!bh) /* A hole? */
5066 memset(data, 0, tocopy);
5067 else
5068 memcpy(data, bh->b_data+offset, tocopy);
5069 brelse(bh);
5070 offset = 0;
5071 toread -= tocopy;
5072 data += tocopy;
5073 blk++;
5074 }
5075 return len;
5076 }
5077
5078 /* Write to quotafile (we know the transaction is already started and has
5079 * enough credits) */
5080 static ssize_t ext4_quota_write(struct super_block *sb, int type,
5081 const char *data, size_t len, loff_t off)
5082 {
5083 struct inode *inode = sb_dqopt(sb)->files[type];
5084 ext4_lblk_t blk = off >> EXT4_BLOCK_SIZE_BITS(sb);
5085 int err = 0;
5086 int offset = off & (sb->s_blocksize - 1);
5087 struct buffer_head *bh;
5088 handle_t *handle = journal_current_handle();
5089
5090 if (EXT4_SB(sb)->s_journal && !handle) {
5091 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5092 " cancelled because transaction is not started",
5093 (unsigned long long)off, (unsigned long long)len);
5094 return -EIO;
5095 }
5096 /*
5097 * Since we account only one data block in transaction credits,
5098 * then it is impossible to cross a block boundary.
5099 */
5100 if (sb->s_blocksize - offset < len) {
5101 ext4_msg(sb, KERN_WARNING, "Quota write (off=%llu, len=%llu)"
5102 " cancelled because not block aligned",
5103 (unsigned long long)off, (unsigned long long)len);
5104 return -EIO;
5105 }
5106
5107 bh = ext4_bread(handle, inode, blk, 1, &err);
5108 if (!bh)
5109 goto out;
5110 err = ext4_journal_get_write_access(handle, bh);
5111 if (err) {
5112 brelse(bh);
5113 goto out;
5114 }
5115 lock_buffer(bh);
5116 memcpy(bh->b_data+offset, data, len);
5117 flush_dcache_page(bh->b_page);
5118 unlock_buffer(bh);
5119 err = ext4_handle_dirty_metadata(handle, NULL, bh);
5120 brelse(bh);
5121 out:
5122 if (err)
5123 return err;
5124 if (inode->i_size < off + len) {
5125 i_size_write(inode, off + len);
5126 EXT4_I(inode)->i_disksize = inode->i_size;
5127 ext4_mark_inode_dirty(handle, inode);
5128 }
5129 return len;
5130 }
5131
5132 #endif
5133
5134 static struct dentry *ext4_mount(struct file_system_type *fs_type, int flags,
5135 const char *dev_name, void *data)
5136 {
5137 return mount_bdev(fs_type, flags, dev_name, data, ext4_fill_super);
5138 }
5139
5140 #if !defined(CONFIG_EXT2_FS) && !defined(CONFIG_EXT2_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5141 static inline void register_as_ext2(void)
5142 {
5143 int err = register_filesystem(&ext2_fs_type);
5144 if (err)
5145 printk(KERN_WARNING
5146 "EXT4-fs: Unable to register as ext2 (%d)\n", err);
5147 }
5148
5149 static inline void unregister_as_ext2(void)
5150 {
5151 unregister_filesystem(&ext2_fs_type);
5152 }
5153
5154 static inline int ext2_feature_set_ok(struct super_block *sb)
5155 {
5156 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT2_FEATURE_INCOMPAT_SUPP))
5157 return 0;
5158 if (sb->s_flags & MS_RDONLY)
5159 return 1;
5160 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT2_FEATURE_RO_COMPAT_SUPP))
5161 return 0;
5162 return 1;
5163 }
5164 MODULE_ALIAS("ext2");
5165 #else
5166 static inline void register_as_ext2(void) { }
5167 static inline void unregister_as_ext2(void) { }
5168 static inline int ext2_feature_set_ok(struct super_block *sb) { return 0; }
5169 #endif
5170
5171 #if !defined(CONFIG_EXT3_FS) && !defined(CONFIG_EXT3_FS_MODULE) && defined(CONFIG_EXT4_USE_FOR_EXT23)
5172 static inline void register_as_ext3(void)
5173 {
5174 int err = register_filesystem(&ext3_fs_type);
5175 if (err)
5176 printk(KERN_WARNING
5177 "EXT4-fs: Unable to register as ext3 (%d)\n", err);
5178 }
5179
5180 static inline void unregister_as_ext3(void)
5181 {
5182 unregister_filesystem(&ext3_fs_type);
5183 }
5184
5185 static inline int ext3_feature_set_ok(struct super_block *sb)
5186 {
5187 if (EXT4_HAS_INCOMPAT_FEATURE(sb, ~EXT3_FEATURE_INCOMPAT_SUPP))
5188 return 0;
5189 if (!EXT4_HAS_COMPAT_FEATURE(sb, EXT4_FEATURE_COMPAT_HAS_JOURNAL))
5190 return 0;
5191 if (sb->s_flags & MS_RDONLY)
5192 return 1;
5193 if (EXT4_HAS_RO_COMPAT_FEATURE(sb, ~EXT3_FEATURE_RO_COMPAT_SUPP))
5194 return 0;
5195 return 1;
5196 }
5197 MODULE_ALIAS("ext3");
5198 #else
5199 static inline void register_as_ext3(void) { }
5200 static inline void unregister_as_ext3(void) { }
5201 static inline int ext3_feature_set_ok(struct super_block *sb) { return 0; }
5202 #endif
5203
5204 static struct file_system_type ext4_fs_type = {
5205 .owner = THIS_MODULE,
5206 .name = "ext4",
5207 .mount = ext4_mount,
5208 .kill_sb = kill_block_super,
5209 .fs_flags = FS_REQUIRES_DEV,
5210 };
5211
5212 static int __init ext4_init_feat_adverts(void)
5213 {
5214 struct ext4_features *ef;
5215 int ret = -ENOMEM;
5216
5217 ef = kzalloc(sizeof(struct ext4_features), GFP_KERNEL);
5218 if (!ef)
5219 goto out;
5220
5221 ef->f_kobj.kset = ext4_kset;
5222 init_completion(&ef->f_kobj_unregister);
5223 ret = kobject_init_and_add(&ef->f_kobj, &ext4_feat_ktype, NULL,
5224 "features");
5225 if (ret) {
5226 kfree(ef);
5227 goto out;
5228 }
5229
5230 ext4_feat = ef;
5231 ret = 0;
5232 out:
5233 return ret;
5234 }
5235
5236 static void ext4_exit_feat_adverts(void)
5237 {
5238 kobject_put(&ext4_feat->f_kobj);
5239 wait_for_completion(&ext4_feat->f_kobj_unregister);
5240 kfree(ext4_feat);
5241 }
5242
5243 /* Shared across all ext4 file systems */
5244 wait_queue_head_t ext4__ioend_wq[EXT4_WQ_HASH_SZ];
5245 struct mutex ext4__aio_mutex[EXT4_WQ_HASH_SZ];
5246
5247 static int __init ext4_init_fs(void)
5248 {
5249 int i, err;
5250
5251 ext4_li_info = NULL;
5252 mutex_init(&ext4_li_mtx);
5253
5254 ext4_check_flag_values();
5255
5256 for (i = 0; i < EXT4_WQ_HASH_SZ; i++) {
5257 mutex_init(&ext4__aio_mutex[i]);
5258 init_waitqueue_head(&ext4__ioend_wq[i]);
5259 }
5260
5261 err = ext4_init_pageio();
5262 if (err)
5263 return err;
5264 err = ext4_init_system_zone();
5265 if (err)
5266 goto out6;
5267 ext4_kset = kset_create_and_add("ext4", NULL, fs_kobj);
5268 if (!ext4_kset)
5269 goto out5;
5270 ext4_proc_root = proc_mkdir("fs/ext4", NULL);
5271
5272 err = ext4_init_feat_adverts();
5273 if (err)
5274 goto out4;
5275
5276 err = ext4_init_mballoc();
5277 if (err)
5278 goto out3;
5279
5280 err = ext4_init_xattr();
5281 if (err)
5282 goto out2;
5283 err = init_inodecache();
5284 if (err)
5285 goto out1;
5286 register_as_ext3();
5287 register_as_ext2();
5288 err = register_filesystem(&ext4_fs_type);
5289 if (err)
5290 goto out;
5291
5292 return 0;
5293 out:
5294 unregister_as_ext2();
5295 unregister_as_ext3();
5296 destroy_inodecache();
5297 out1:
5298 ext4_exit_xattr();
5299 out2:
5300 ext4_exit_mballoc();
5301 out3:
5302 ext4_exit_feat_adverts();
5303 out4:
5304 if (ext4_proc_root)
5305 remove_proc_entry("fs/ext4", NULL);
5306 kset_unregister(ext4_kset);
5307 out5:
5308 ext4_exit_system_zone();
5309 out6:
5310 ext4_exit_pageio();
5311 return err;
5312 }
5313
5314 static void __exit ext4_exit_fs(void)
5315 {
5316 ext4_destroy_lazyinit_thread();
5317 unregister_as_ext2();
5318 unregister_as_ext3();
5319 unregister_filesystem(&ext4_fs_type);
5320 destroy_inodecache();
5321 ext4_exit_xattr();
5322 ext4_exit_mballoc();
5323 ext4_exit_feat_adverts();
5324 remove_proc_entry("fs/ext4", NULL);
5325 kset_unregister(ext4_kset);
5326 ext4_exit_system_zone();
5327 ext4_exit_pageio();
5328 }
5329
5330 MODULE_AUTHOR("Remy Card, Stephen Tweedie, Andrew Morton, Andreas Dilger, Theodore Ts'o and others");
5331 MODULE_DESCRIPTION("Fourth Extended Filesystem");
5332 MODULE_LICENSE("GPL");
5333 module_init(ext4_init_fs)
5334 module_exit(ext4_exit_fs)
Linux with added FPC-III support